I. Introduction

After five decades of negotiations that culminated with the creation of the WTO in 1995, low tariffs in advanced economies became one of the pillars of the international trading system, facilitating trade, prompting subsequent liberalization in emerging markets, and underpinning the development of global supply chains. While there was, and still is, room for further tariff reductions in certain sectors of advanced economies, average tariffs had reached such low levels that the negotiating agendas of advanced economies eventually evolved to address non-tariff barriers ranging from regulatory cooperation to digital trade. This landscape changed abruptly in 2018 when the U.S. administration imposed, or announced they might impose, higher tariffs targeting specific sectors or specific trading partners. From the U.S. perspective, the tariffs respond to concerns about various longstanding market-distorting government interventions and policies by trading partners, including on technology transfer and subsidies to state-owned enterprises, and in some cases concern regarding U.S. national security interests. With the prospect of higher tariffs taking center stage again, their possible adverse effects on different economies is a key concern among policy makers around the world. In this paper, we use a multi-sector trade model to systematically estimate the effects of three potential trade policy actions by the U.S. that might have systemic importance.

While trade policy agendas slowly expanded to other issues, the theory of international trade continued to make progress in developing models to understand how tariff barriers might affect global production and consumption patterns. We take advantage of this progress by simulating the effect of tariff changes using one such state-of-the-art computable general equilibrium model, developed by Caliendo, Feenstra, Romalis and Taylor (2017; CFRT henceforth), we simulate the effect of potential tariff changes using input-output data covering 165 countries and 17 sectors. We use detailed tariff data, including tariff data predating the current period of heightened trade tensions.

Two fronts of ongoing U.S.-related trade disputes give rise to the three scenarios we study. The first one relates to the potential introduction of tariffs on imported autos and auto-parts, in the context of the U.S. Commerce Department’s Section 232 (national security) investigation of the sector. The second set of scenarios corresponds to U.S.-China trade relations, which for the U.S. at least partly falls under the umbrella of a Section 301 (technology transfer) investigation. In particular, we model two possible outcomes of this dispute. The first one involves what we call a ‘transactional deal’ between the two countries, which resembles a managed-trade outcome. In our framework, such a deal could potentially involve an increase in Chinese purchases of U.S. goods and/or voluntary export restraints by China that limit its sales of certain goods to the U.S. The scenarios are calibrated such that each side closes half of the bilateral trade deficit between the two countries. In the absence of such a deal, an escalation of bilateral tariffs becomes more likely. Such an escalation is the focus of our third and final scenario.

It is important to note upfront that, as discussed in more detail subsequently, the implementation of these scenarios involves certain modeling choices. This is undoubtedly a reflection of the uncertainty surrounding the possible outcomes of ongoing trade negotiations. For instance, retaliation by trading partners following the introduction of tariffs by the U.S. could take different forms; a transactional deal between the U.S. and China could involve other sectors or include reforms that are pursued on a most-favored-nation basis; and an escalation of tensions between these two countries may go hand-in-hand with liberalization efforts vis-à-vis other partners. Given this intrinsic uncertainty, this paper aims to contribute more broadly to the understanding of the main drivers and channels through which some potential increases in trade barriers can affect production chains, trade, and consumption patterns around the world.

The uncertainty around ongoing negotiations is compounded in some cases with the fact that no perfect model exists to fully assess the effects of certain policy changes. For example, existing models may not be well suited to analyze changes in areas such as investment regimes and intellectual property, or―more fundamentally―the risks posed by a rejection of the rules-based global trading system (see e.g. Mattoo and Staiger, 2019). While the focus of this study is modeling the effect of changes in certain trade barriers, it is important to bear in mind that some policy changes not modeled explicitly within this framework can have first-order effects on economic activity.

Despite their varied nature, some common features emerge across all three scenarios we model. First, that while aggregate GDP effects are relatively small, they are negative for most countries. This includes the U.S., which is estimated to lose in most scenarios as ‘protected’ sectors do not expand sufficiently to make up for the loss of competitiveness in the rest of the economy. This is largely a result of increased protection applying to sectors for which foreign production is difficult to substitute domestically. Second, spillovers to third countries are generally a function of direct exposures to (i.e. gross trade flows with) the markets affected by the tariffs, although in some cases indirect exposures through supply chains are of first-order importance. Third, there is a wide variation at the sectoral level, with the results suggesting that there would likely be very high adjustment costs in the process of attaining a new equilibrium. This variation at the sectoral level also translates into heterogeneous impacts across U.S. geographic regions and across the income distribution. A transactional deal with China would, in particular, benefit U.S. states with a high share of primary sector (mining, hydrocarbons, and agriculture) activity in total value added, such as Arkansas, North Dakota, Oklahoma, and Wyoming, whereas these states would suffer most under the auto tariffs scenario. Importantly, richer states seem to benefit from certain trade scenarios, whereas poorer ones seldom benefit from any of these. Finally, we show how the estimated impacts are a function of various assumptions, including – crucially – the extent to which the U.S. is able to re-shore production in the affected sectors.

To see how the results of this paper contribute to our understanding of the potential effects of trade tensions, it is instructive to contrast the tool we use with other modeling approaches.² In particular, sectoral trade models (like the CFRT framework we use) can provide a useful complement to macroeconomic models, such as the IMF’s Global Integrated Monetary and Fiscal (GIMF) model.³ While the GIMF model can yield valuable information on shorter-term demand effects and the role for rigidities and longer-term macro distortions, trade models can shed light on potential long-term supply-chain disruptions. The trade model and the data that we use are particularly well-suited to study the longer-term allocative-inefficiency of tariffs that are targeted to specific productive sectors (17 in our calibration), and their spillovers to specific countries (165 in our calibration). Of course, no model can capture all aspects of reality, and this one is no exception. Most important to bear in mind is that the model we use does not have any dynamics, and as such it describes the steady state implications and ignores any effect trade tensions may have on uncertainty. As has been widely documented in recent research, the uncertainty related to trade policy can have detrimental effects on investment and economic activity.⁴

Our overarching finding that the trade scenarios we model tend to produce negative outcomes for the parties involved does not imply that the status quo is optimal. Rather, the U.S. and its trading partners could work constructively, not only to better address distortions in the trading system and in individual countries’ trade regimes, but also to tap trade policy areas that have high growth potential (IMF-WB-WTO, 2018). Our results both underscore the value of adhering to the existing levels of liberalization, and highlight the risks associated with a fragmentation or even a complete breakdown of the trading system where new and longstanding issues remain unresolved.

The rest of the paper is organized as follows. Section II presents the model as well as the main databases used to estimate the impact of tariff changes. This section also discusses the estimation of elasticities of substitution, which are key for the model results. Section III describes the analyzed scenarios and presents the corresponding results. Section IV discusses the sensitivity of the results to the calibration of key parameters, including the assumed elasticities of substitution. Section V concludes.

II. Model, Data and Calibration

B. Construction of Main Database

The use of input-output (I-O) data is key to quantify and account for the bilateral interaction matrix not only across countries, but also across sectors. In this paper we rely on the Eora global supply chain database, which consists of a multi-region input-output table (MRIO), containing data for many countries in the world.

The original Eora database represents a balanced global MRIO table linking 26 sectors across 190 countries, with a complete annual time-series over the period 1990–2015. See Lenzen et al (2013) for a detailed description of the Eora database. Our trade-flow data correspond to the latest (2015) Eora table.

For the analysis presented in this paper, the Eora database is combined with 2015 applied-tariff data from UNCTAD’s Trade Analysis Information System accessed through WITS. The intersection of both databases results in a combined I-O database, with 17 sectors and 165 countries. Table 1 lists the 17 sectors included in our analysis.

Table 1:

Economic sectors included in the analysis

Source: IMF staff calculations.

Table 1:

Economic sectors included in the analysis

Goods		Services
	Agriculture	Construction
	Extractive industries	Wholesale & retail
	Food	Hotels & restaurants
	Textile	Transport & communication
	Wood & paper	Financial services
	Chemicals & oil derivatives	Other services
	Metals
	Electronics
	Transport goods
	Other manufacturing
	Electricity

Source: IMF staff calculations.

View Table

Table 1:

Economic sectors included in the analysis

Goods		Services
	Agriculture	Construction
	Extractive industries	Wholesale & retail
	Food	Hotels & restaurants
	Textile	Transport & communication
	Wood & paper	Financial services
	Chemicals & oil derivatives	Other services
	Metals
	Electronics
	Transport goods
	Other manufacturing
	Electricity

Source: IMF staff calculations.

This database thus enables the analysis of all bilateral exposures―among the 165 countries―at the sectoral level. As an illustration of the data, Figure 1 presents the details of the sectoral gross exports (imports) of the U.S. to (from) some of its main trading partners, including Canada, China, Germany and Mexico, among others.

View Full Size

U.S. Sectoral Gross Trade Flows with Selected Countries, 2015

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: Eora database; and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

C. Calibration of Elasticities of Substitution

In CFRT, there are two sets of sector-level elasticities that play a key role in determining the effect of new tariffs: (i) elasticities of substitution across foreign or domestic intermediate inputs (os, where s indexes sectors); and (ii) elasticities of substitution across composite foreign and domestic inputs (cos). Elasticities in (i) are assumed larger than those in (ii). Such a determination is motivated by technical considerations, as explained in sub-section 2.A, as well as empirical considerations, since in practice switching inputs between two foreign suppliers is easier than switching from a foreign to a domestic supplier.

Estimating elasticities across intermediate inputs (σ_s)

The first set of elasticities – those across foreign or domestic intermediate inputs – are newly estimated in this paper. We follow the steps described just after Table 1 in CFRT. We start with estimating Γ^s in equation (1) following the triangular trade procedure described in Section 4 of Caliendo and Parro (CP 2015) for the goods-producing sectors in Eora. Data on tariffs for each sector, s, between any two countries, j and k, $\left({\tau }_{kj}^s\right)$ and corresponding trade flows $\left(X_{kj}^s\right)$ for the year 2016 are from UNCTAD Trade Analysis Information System and sourced through WITS.

$\begin{array}{cc}\ln \left(\frac{X_{ni}^s{X}_{ih}^s{X}_{hn}^s}{X_{in}^s{X}_{hi}^s{X}_{nh}^s}\right)=-{\mathrm{\Gamma }}^s\ln \left(\frac{{\tau }_{ni}^s{\tau }_{ih}^s{\tau }_{hn}^s}{{\tau }_{in}^s{\tau }_{hi}^s{\tau }_{nh}^s}\right)+{\tilde{\epsilon }}^s& \left(1\right)\end{array}$

Equation (1) is estimated over three samples, one with all bilateral trade data, another excluding 5 percent of the smallest flows and another excluding 10 percent of the smallest flows, the latter being our preferred sample as it excludes small outliers that unduly tilt the estimation.

Table 2 shows our preferred calibration of Γ^s and compares it to the estimates in CP (2015).¹⁰ Note that for sectors 1 and 10 we simply use the values from CP, as our estimates are not robust across samples. Estimates for sector 9 were also not robust as had been the case for CP, and – following CP – we take the mean elasticity across all other goods-producing sectors.

Table 2.

Sector-level trade elasticities

Sources: Caliendo and Parro (2015); and IMF staff calculations. Notes: ^1/ Simple-averaged elasticities of more disaggregated sectors. ^2/ Using the same as CP. ^3/ Simple-average of elasticities across goods producing sectors.

Table 2.

Sector-level trade elasticities

Sector #	Simplified sector name	CP (2015)	Preferred
1	Agriculture, Hunting, Forestry and Fishing	9.1	9.12/
2	Extractive Industries	13.5	19.0
3	Food, Beverages and Tobacco	2.6	2.5
4	Textiles, Textile Products, leather and footwear	8.1	3.0
5	Wood, Paper, Printing and Publishing	14.01/	5.4
6	Petroleum, Chemical, Non-Metallic Mineral Products	18.11/	12.1
7	Metal Products	7.0	13.9
8	Electrical equipment and Machinery	4.81/	2.6
9	Transport equipment	8.2	8.03/
10	Manufacturing, Nec; Recycling	4.0	4.02/

Sources: Caliendo and Parro (2015); and IMF staff calculations. Notes: ^1/ Simple-averaged elasticities of more disaggregated sectors. ^2/ Using the same as CP. ^3/ Simple-average of elasticities across goods producing sectors.

View Table

Table 2.

Sector-level trade elasticities

Sector #	Simplified sector name	CP (2015)	Preferred
1	Agriculture, Hunting, Forestry and Fishing	9.1	9.12/
2	Extractive Industries	13.5	19.0
3	Food, Beverages and Tobacco	2.6	2.5
4	Textiles, Textile Products, leather and footwear	8.1	3.0
5	Wood, Paper, Printing and Publishing	14.01/	5.4
6	Petroleum, Chemical, Non-Metallic Mineral Products	18.11/	12.1
7	Metal Products	7.0	13.9
8	Electrical equipment and Machinery	4.81/	2.6
9	Transport equipment	8.2	8.03/
10	Manufacturing, Nec; Recycling	4.0	4.02/

Sources: Caliendo and Parro (2015); and IMF staff calculations. Notes: ^1/ Simple-averaged elasticities of more disaggregated sectors. ^2/ Using the same as CP. ^3/ Simple-average of elasticities across goods producing sectors.

The estimates in Table 2 for Γ^s correspond to the first set of numbers in Table 1 of CFRT. We follow the procedure in CFRT and calibrate os based on the relation between the sectoral elasticities of substitution and the Pareto shape parameter estimated by Eaton, Kortum and Kramarz (2011). In particular, if Γ^s denotes our estimated elasticities from Table 2, then σ_s=(Γ^s + 1)/1.5.

Calibrating elasticities across final domestic and foreign goods (ω_s)

To calibrate the foreign-domestic elasticities of substitution we rely on the estimates by Feenstra et al (2018; FLOR henceforth).

The codes provided in the published version of FLOR were used to estimate os and cos, under their two-stage least squares procedure consistent with the estimates presented in their table 3. Then we computed the ratio os/ros for each sector s and matched their sectors to the Eora-based sectors we use, to finally compute ros using our own estimates of os. For sectors that could not be matched, we used the simple average of the ratio, which is 1.76. The resulting sectoral elasticities of substitution are presented in Figure 2.

View Full Size

Estimated Sectoral Elasticities of Substitution

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

III. Scenarios and Results

This section is split in three sub-sections. The first defines and describes the scenarios analyzed. The second presents the main results for each of these scenarios using the multi-sector model on the U.S. and China. Finally, the third sub-section discusses spillovers to ‘third countries’.

B. Effects on the U.S.

An overarching result from our simulations is that U.S. GDP falls under all three scenarios ―by around 0.2–0.3 percent of GDP―following the introduction of tariffs (Figure 3).¹⁷ In other words, U.S. GDP falls under a wide range of trade ‘tensions’ and ‘deal’ scenarios.

View Full Size

Real GDP impact on the U.S. and selected trading partners

(percent change)

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Significant sectoral disruptions take place in the background (Figure 4). Moreover, and contrary to the effect on GDP, which was similar across scenarios, the changes in sectoral value added are fairly heterogenous and highly dependent on the scenario under consideration. For instance, the U.S. agricultural sector benefits noticeably from larger purchases of U.S. goods under a “transactional deal”, as China treats such goods highly preferentially. However, the agricultural sector contracts under the auto tariffs scenario as several U.S. trading partners retaliate by increasing tariffs on that particular sector in order to match the intensity of the U.S. tariffs.

View Full Size

Impact on U.S. Sectors

percent change

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Sources: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

That the U.S. might lose out as a result of higher tariffs on a specific sector and across all partners (Scenario 1) or due to higher tariffs on a specific partner across all sectors (Scenario 3) is relatively unsurprising. Increased U.S. tariffs imply higher distortions for U.S. firms and consumers, with the new equilibrium associated with less-efficient and more-expensive suppliers. This effect is compounded by the income losses firms must grapple with as a result of U.S. partners’ retaliation.¹⁸

Perhaps more surprising is that – on net – U.S. overall GDP appears to fall under the transactional-deal scenario with China (Scenario 2). This is due to the opposing effects of the two parts of the deal that were implemented: on one hand, additional purchases by China clearly benefit some sectors in the U.S. (see corresponding blue bars in Figure 3). However, the reduction in Chinese exports of selected manufacturing goods (“electronics” and “other manufacturing” sectors) to the U.S. does not lead to a meaningful expansion of the protected sectors in the U.S., which themselves partly rely on Chinese inputs to produce. This inability of U.S. protected sectors to expand also implies that all U.S. sectors now face higher input costs in those sectors, as foreign inputs are not easily substitutable with domestic production.¹⁹

Another important issue, that is somewhat masked by the effect on overall GDP, is that the heterogeneity in the response at the sectoral level would lead to a diverse outcome at the regional level within the U.S. This is due to the relative geographical specialization of the different U.S. states across economic sectors (Box 1). In particular a “transactional deal” where China buys more from the U.S. would benefit states with a high share of the primary sector (mining, hydrocarbons, and agriculture) in total value added―for instance, Arkansas, North Dakota, Oklahoma, and Wyoming―whereas these states would suffer most under the “auto tariffs” scenario. The impact of an intensification of ongoing trade tensions between the U.S. and China or a “transaction deal” where China restraint some of its U.S.-bound exports would have a more homogeneous effect, with most states being negatively affected by these sector-wide tariff increases. Moreover, these results have important implications across the income distribution―as richer states (with higher median incomes and a higher proportion of the rich) would benefit relative to poorer ones in the auto tariff scenario as well an intensification of the ongoing trade war with China, while poorer states would only benefit when China purchases more primary products from the U.S.

Box 1.

Trade Policy and its Potential Effects on U.S. States

Our simulations show that the introduction of higher tariffs to imports of certain goods or coming from certain countries would have, in the long run, negative effects on U.S. overall output. The impact at the sectoral level, however, is in some cases particularly large. In light of this heterogeneity, and given the diverse specialization patterns across U.S. states, this box explores the regional implications of these potential policy actions. More precisely, U.S. sectoral impacts obtained in our simulation results shown above are mapped (outside of the model) to changes in output at the state level based on the share of each productive sector in U.S. states’ output.²⁰

In a scenario where the U.S. introduces a 25 percent tariff on imports of autos and auto-parts across the board, and trading partners retaliate with the same tariff rate on U.S. auto goods and a uniform tariff rate across all other U.S. goods, primary good sectors (i.e. agriculture, mining, and hydrocarbon extraction) tend to suffer large falls in value-added (Figure 4). This in turn translates into larger falls in GDP in states where these sectors represent a relatively large share of output―such as Arkansas, North Dakota, Oklahoma, West Virginia, and Wyoming―which, in addition, do not reap the benefits from the increase in output in the U.S. auto industry under this scenario (see Box Figure 1 and Box Figure 3, Panel A). Certainly, the magnitude of these effects would directly depend on the specifics of the potential retaliation by trading partners. As a sensitivity analysis, we replace the uniform tariff retaliation by trading partners by a more selective retaliation, mainly targeting some ‘emblematic’ U.S. products (e.g. Harley Davidson motorcycles, blue jeans, bourbon whiskey, among other).²¹ In this case, although these states are still among the ones with the higher GDP falls, their losses are largely reduced as their exposure to the “targeted sectors” is lower Box Figure 3, Panel B).

View Full Size

U.S. States Exposure to Selected Sectors

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: Eora Database (2015); and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Impacts Tend to Harm Poorer Regions and Lower-Income Households

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Notes: green check denotes “positive effect”; red cross denotes “negative effect”; empty cell denotes that the impact is not statistically significant at the 10 percent level. See Box Table 1 for details.Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

View Full Size

View Full Size

Impact of Trade Scenarios on GDP at the State-Level

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download Figure
• Download Figure
• Download figure as PowerPoint slide

In a scenario where there is an intensification of trade tensions between the U.S. and China, with both countries introducing a uniform 25 percent tariff rate across all products, and thus with most sectors experiencing a fall in output, most U.S. states exhibit a negative change in GDP. Moreover, the effects appear to be relatively less heterogeneous across states than under most other trade scenarios (Box Figure 3, Panel C).

Finally, in a scenario where a “trade deal” is reached between the U.S. and China, and where China agrees to buy more selected U.S. goods―including in particular agriculture and hydrocarbon products―the states which are relatively exposed to these sectors and are overall less diversified, would tend to benefit from such a deal (Box Figure 3, Panel E). Nevertheless, a deal restraining some of China’s exports to the U.S. would lead to losses in all states (Box Figure 3, Panel D).

These differences regarding the impact across different geographic areas and under certain trade scenarios would likely lead to varying effects on other socio-economic indicators such as income, poverty and inequality. This is due to the fact that regions are themselves heterogeneous in terms of income and income distribution.

In fact, positive changes in GDP at the state level under the auto tariff scenario or a scenario with higher bilateral tariffs between the U.S. and China are positively associated with states having higher median income levels as well as a higher proportion of rich individuals (and lower proportion of the poor). In other words, based on simple correlations these specific trade scenarios tend to favor those richer states, and which have a higher proportion of wealthy people and lower poverty rates (see Box Figure 2 and Box Table 1).

Box Table 1:

Link between Gains/Losses and Income Measures at the State-Level

Correlation of estimated state-level GDP changes with state-level income measures

Notes: *, **, and *** denote statistical significance at the 10, 5, and 1 percent confidence level. “low income households” and “high-income households” denote those households with an annual income of less than $15,000 and more than $150,000, respectively. Source: IMF staff calculations.

Box Table 1:

Link between Gains/Losses and Income Measures at the State-Level

Correlation of estimated state-level GDP changes with state-level income measures

	States with higher median income	States with higher mean income	Higher proportion of low income household	Higher proportion of high-income households
Auto tariffs, uniform retaliation	0.3 **	0.4 ***	-0.1	0.4 ***
Auto tariffs, selected retaliation	0.5 ***	0.6 ***	-0.2	0.6 ***
Intensification of trade war with China	0.5 ***	0.6 ***	-0.3 *	0.6 ***
Trade deal: restraint on China’s exports to U.S.	0.2	0.3 **	-0.1	0.3 **
Trade deal: China purchases of U.S. goods	-0.1	-0.2 *	0.0	-0.3 *

Notes: *, **, and *** denote statistical significance at the 10, 5, and 1 percent confidence level. “low income households” and “high-income households” denote those households with an annual income of less than $15,000 and more than $150,000, respectively. Source: IMF staff calculations.

View Table

Box Table 1:

Link between Gains/Losses and Income Measures at the State-Level

Correlation of estimated state-level GDP changes with state-level income measures

	States with higher median income	States with higher mean income	Higher proportion of low income household	Higher proportion of high-income households
Auto tariffs, uniform retaliation	0.3 **	0.4 ***	-0.1	0.4 ***
Auto tariffs, selected retaliation	0.5 ***	0.6 ***	-0.2	0.6 ***
Intensification of trade war with China	0.5 ***	0.6 ***	-0.3 *	0.6 ***
Trade deal: restraint on China’s exports to U.S.	0.2	0.3 **	-0.1	0.3 **
Trade deal: China purchases of U.S. goods	-0.1	-0.2 *	0.0	-0.3 *

Notes: *, **, and *** denote statistical significance at the 10, 5, and 1 percent confidence level. “low income households” and “high-income households” denote those households with an annual income of less than $15,000 and more than $150,000, respectively. Source: IMF staff calculations.

By contrast, a “trade deal” scenario leading to additional purchases of certain U.S. goods by China, which tends to favor those states with a relatively higher share of primary good sectors (mining and agriculture), would seem to somewhat benefit states with lower average income, a higher proportion of lower earners, and a lower share of the wealthy.

C. Impact and Spillovers to Other Countries

Impact on main trading partners

Figure 3 also shows the effect on overall GDP of some of the U.S. main trading partners―namely, Canada, China, Germany, Japan, and Mexico. Together, these five economies represent more than half of all U.S. imports. Contrary to the impact on the U.S., which is broadly similar across scenarios, the impact on these countries is rather heterogenous and highly dependent on the scenario under consideration.

In particular, when the U.S. and China impose bilateral tariffs on each other, both countries tend to suffer from this. Other countries―particularly Canada and Mexico, which are highly integrated with the U.S.―would tend to benefit in the long-run from less-intense competition with China in the U.S. market. Nevertheless, in Scenario 1 (tariffs on vehicles and auto-parts), these two countries would suffer the most given their significant exposure to the U.S. (Figure 5). This is also the case, to a lesser extent, for Germany and Japan.

View Full Size

Trade Exposure to the U.S. in Selected Countries direct exposures

(i.e. gross trade flows)

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: Eora Database (2015); and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

The heterogeneous response across countries under the different scenarios is even more striking across the different sectors. To illustrate this, Figure 6 shows the sectoral impacts for Canada, China, and Mexico. The transportation equipment sector―which includes autos and auto-parts ―in both Canada and Mexico exhibit significant losses under a scenario where the U.S. imposes 25 percent tariffs on vehicles and auto-parts. In the case of China, some manufacturing sectors (such as electronics and other manufacturing) tend to experience significant losses, both when the U.S. increases tariffs on all Chinese goods (Scenario 3) and also when China agrees to reduce its exports of these goods to the U.S. under a bilateral “transactional deal” (Scenario 2).

View Full Size

Sectoral Impact in Selected Countries

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

USMCA exemptions

Side-letters to the United States-Mexico-Canada (USMCA) trade agreement stipulate certain exemptions for Canada and Mexico if the U.S. were to impose Section 232 tariffs on automobile imports.²² If ratified, new Section 232 tariffs may be waived for Canada and Mexico, up to certain quantity and value quotas that vary by product and partner. Canada would be allowed to ship up to 2.6 million cars to the U.S. under USMCA preferences, well above the 1.8 million it sent in 2018, and $32.4 billion worth of auto parts without getting hit by tariffs. Mexico’s side-letter is similar, except that the quota for auto parts is higher, at $108 billion.

To assess whether these quotas might become binding, we re-ran Scenario 1 but assuming that Canada and Mexico are fully exempt. We then applied the simulated increase in real transport equipment sales from USMCA partners to the U.S. to the corresponding observed 2018 exports of passenger vehicles, thus obtaining counterfactual exports of passenger vehicles for both countries. Since the quotas for auto-parts are set in terms of values rather than quantities, our counterfactual exports of auto-parts were obtained in the same way but using the simulated increase in nominal transport equipment sales. Our results, shown in Figure 7, indicate that exports of auto-parts would likely will remain within the USMCA quotas. In the case of passenger vehicles, however, the combination of more-stringent quotas with a significant simulated expansion in exports ―as they gain market share relative to all other countries to which the auto and auto-parts tariffs still apply―implies that the quotas established by the USMCA would become binding.²³ In other words, tariffs would indeed kick-in on U.S. vehicle imports from Canada and Mexico.²⁴

View Full Size

Potential Effect of the USMCA on Exports of Auto and Auto-parts

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Impact on other countries

In a world that is highly integrated, changes in trade barriers―either at the bilateral or multilateral level―can have a significant effect on other countries, even those not directly exposed to changing trade barriers. In this context, it is useful that our simulations include results for 165 countries. In order to summarize the information concisely, Figures 10 to 13 in the Appendix present color-coded maps representing the countries that would benefit or lose (in terms of overall GDP and real export changes) from these different trade scenarios.

View Full Size

Scenario 2 – A transactional U.S.-China deal

Ex-ante exposures (left) and simulated percent change in exports (right)

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: Eora Database (2015); and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Sensitivity of GDP Gain/Losses to Changes in Elasticities of Substitution

real GDP effects under baseline (horizontal axes) and alternative (vertical axes) calibrations

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Impact of “Auto-Tariffs” on All Countries

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Sources: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Impact of “Trade War” Between the U.S. and China on All Countries

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Impact of “Trade Deal: Restraint on China’s Exports to the U.S.” on All Countries

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Impact of “Trade Deal: Additional Purchases of U.S. Goods by China” on All Countries

Citation: IMF Working Papers 2019, 143; 10.5089/9781498318440.001.A001

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

What determines who wins and who loses? A key determinant of the outcome for each country is the ex-ante sectoral exposure of the different countries to the U.S., and to any other country directly involved in the imposition of tariffs (e.g. China under both Scenarios 2 and 3). To see this, it will be instructive to zoom in on one of the scenarios, bearing in mind that the observations made here regarding economic channels hold more generally across scenarios. For the purposes of illustration, let us then focus on the transactional deal between the U.S. and China (Scenario 2). On the one hand, any country selling e.g. agricultural goods to China would be negatively affected by the preferential treatment granted to U.S. producers. This is measured on the horizontal axis of the left-hand-side chart of Figure 8. On the other hand, countries that sell e.g. electronics to the U.S. stand to benefit from the decreased competition with China; this is measured on the vertical axis of the chart. As can be clearly seen from the Figure, some countries, like Vietnam, would ex-ante be negatively exposed to this scenario, whereas others such as Mexico lie on the opposite end of the spectrum.

This prior is largely borne out by our simulations (Figure 8, right panel). As expected, Vietnam experiences a large negative effect due to the preferential tariff reduction on U.S. goods coming into China. Similarly, Mexico benefits exclusively because of the lower competition with Chinese producers in the U.S. market.

Not all effects follow as closely the pattern of direct exposures, underscoring the benefits of the model we are using to gauge the effects of tariffs. For example, consider the contrast between Korea and Malaysia. The panel on the left-hand-side of Figure 8 suggests that their ex-ante direct exposures to Chinese competition in the U.S. market is very similar. To this first-order approximation using direct exposures, therefore, one might expect both countries to experience similar benefits from decreased Chinese competition. However, while Malaysia sees large long-term gains, Korea barely benefits. The reason is that Chinese exports of electronics and other manufacturing going to the U.S. have a large Korean value-added component. Indirect exposures through global supply chains can be of first-order importance in cases such as this one.

Another crucial factor driving the quantitative gain/losses among different countries relates to the relative ease with which countries, particularly by the U.S, are able to replace foreign imported goods by domestically produced ones. This is analyzed in the following section.

IV. The Role of the Elasticity of Substitution Between Domestic and Foreign Goods

In addition to initial trade exposures, elasticities of substitution play a key role in driving the magnitudes and signs of the effects of these trade scenarios. Thus, this section analyzes the impact of varying the elasticities between final domestic and foreign-produced goods.

In particular, we consider the sensitivity of our baseline results to increasing the elasticity of substitution between foreign and domestic goods, which in the current context can be read as making it easier for any country to re-shore production. We implement this robustness check by raising sectoral foreign-domestic elasticities to be equal to foreign-foreign ones. In practice, it is as if every dot in Figure 2 above were to be raised upwards to the 45-degree line.²⁵

We find that, in general, if re-shoring production were to be easier for the U.S., then potential losses would be larger for most other countries. This is shown by the fact that in Figure 9 most countries other than the U.S. lie below the 45-degree line.²⁶ Intuitively, in our baseline when the U.S. raises tariffs on its imports from any specific country (and/or in particular sectors), the goods imported from that country are likely to be substituted by imports of the same goods from other countries. This trade-diverting effect is driven by our baseline calibration based on the work by Feenstra et al. (2018): upon an increase in barriers with a specific partner, it is relatively easier to find another foreign supplier than it is to find a domestic one (see discussion in Section II.C). However, if it was equally easy for the U.S. to substitute those goods with domestic production, then trade diversion would be more muted and adverse spillovers would be much larger.

In the case of the scenarios primarily involving the U.S. and China, Figure 9 clearly shows that countries heavily involved in either the Asian or American value chain―such as Canada, Korea, and Mexico―would be hurt if re-shoring was easier. In the case of Korea, gains may even turn into losses.

V. Concluding Remarks

Following decades of growing trade flows and cooperation, the process of global trade integration faces an evident risk of stalling or going in reverse. In this paper, we use a multi-sector general equilibrium calibrated to 165 countries to quantify the potential effects of three specific scenarios that could, in principle, have systemic consequences for how global production and consumption are organized. We analyze three relevant and plausible scenarios in detail: (i) a scenario in which the U.S. imposes uniform tariffs on all imports of vehicles and auto-parts from all trading partners, who then retaliate by imposing equivalent tariffs on the U.S.; (ii) a scenario in which a transactional trade deal is reached between the U.S. and China to eliminate the bilateral deficit, with the latter purchasing more goods (especially primary products and other selected goods such as aircrafts) from the U.S., and limiting exports to the U.S. of selected manufacturing goods; and, in the absence of such deal, (iii) a possible escalation of the trade tensions between the U.S. and China, in which both economies introduce a 25 percent tariff on all their bilateral imports.

We find that GDP in the U.S. falls under most scenarios. While possibly expected in scenarios that include retaliation, we find that U.S. output losses also arise in the case where China voluntarily limits its exports of selected goods to the U.S. This is because those foreign goods (electronics and other manufacturing goods) are particularly difficult to substitute with domestic production, and thus tend to be replaced by additional imports from other countries.

Some production sectors and geographic regions in the U.S. exhibit substantial changes. A “transactional deal” with China would, in particular, benefit U.S. states with a high share of primary sector (mining, hydrocarbons, and agriculture) activity in total value added, whereas these states would suffer most under the “auto tariffs” scenario. While our model features no adjustment costs in the move toward the new long-run equilibrium, these results are indicative of the potentially disruptive nature of the tariff changes considered. Which sectors gain and which lose depends on the specific scenario under consideration. Moreover, these sectoral changes can have important socio-economic consequences across different U.S. regions, as some areas tend to be specialized in a small number of sectors that can greatly benefit―or suffer―from the introduction of tariffs.

Regarding the effect on U.S. trading partners and other economies, the potential gains/losses as well as sector-level disruptions are also scenario-dependent. The initial sectoral exposures to the U.S. and other key players appear to be an important determinant of spillovers, although in some cases indirect exposures through global supply chains can be of first-order importance.

Besides input-output data, the other key inputs used to obtain these results are the sectoral elasticities of substitution. Following recent advances in the field of international trade, we calibrate separately the elasticity of substitution between foreign suppliers, and the elasticity of substitution between foreign and domestic suppliers. Key in this calibration is that, in every sector, it is easier to find another foreign supplier than it is to find a domestic one in response to a price change of a foreign good. In robustness checks we find that, if it were easier to re-shore production than it is in our baseline calibration, U.S. real GDP losses would be lower (though they generally remain negative) and most third countries would be worse off as they do not gain through trade diversion. In other words, spillovers would likely be harsher if re-shoring production to the U.S. ends up being easier than in our baseline calibration.

Our overarching finding that the trade scenarios we model tend to produce negative outcomes for the parties involved does not imply that the status quo is optimal. Rather, the U.S. and its trading partners could work constructively, not only to better address distortions in the trading system and in individual countries’ trade regimes, but also to tap trade policy areas that have high growth potential (IMF-WB-WTO, 2018). Our results both underscore the value of adhering to the existing levels of liberalization, and highlight the risks associated with a fragmentation or even a complete breakdown of the trading system where new and longstanding issues remain unresolved.