NBER WORKING PAPER SERIES

THE ECONOMIC IMPACTS OF THE US-CHINA TRADE WAR

Pablo Fajgelbaum Amit Khandelwal

Working Paper 29315 http://www.nber.org/papers/w29315

NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 September 2021 , Revised December 2021

Maximilian Schwarz provided excellent research assistance. We thank the Editor, Gene Grossman, Jonathan Vogel, and Stephen Redding for helpful comments and suggestions. E-mail: pfajgelb@princeton.edu, ak2796@columbia.edu. When citing this paper, please use the following: Fajgelbaum PD, Khandelwal AK. 2021. The Economic Impacts of the Trade War. Annu. Rev. Econ. 3: Submitted. DOI: 10.1146/annurev-economics-051420-110410. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been peerreviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications.

• © 2021 by Pablo Fajgelbaum and Amit Khandelwal. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source.

The Economic Impacts of the US-China Trade War Pablo Fajgelbaum and Amit Khandelwal NBER Working Paper No. 29315 September 2021, Revised December 2021 JEL No. F0

ABSTRACT

In 2018, the US launched a trade war with China, an abrupt departure from its historical leadership in integrating global markets. By late 2019, the US had imposed tariffs on roughly $350 billion of Chinese imports, and China had retaliated on $100 billion US exports. Economists have used a diversity of data and methods to assess the impacts of the trade war on the US, China and other countries. This article reviews what we have learned to date from this work.

Pablo Fajgelbaum Princeton University Department of Economics International Economics Section Room 294 Julis Romo Rabinowitz Bldg Princeton, NJ 08544 and NBER pfajgelb@princeton.edu

Amit Khandelwal Graduate School of Business Columbia University Uris Hall 606, 3022 Broadway New York, NY 10027 and NBER ak2796@columbia.edu

1 Introduction

In early 2018, the US raised tariffs on a few large import items-washing machines, solar panels, steel and aluminum. While these tariffs did not discriminate by origin, it soon became apparent that US trade policies were targeting China. The US subsequently increased tariffs on thousands of products from China between 2018-19, ultimately targeting roughly $350 billion of imports from that country. China retaliated over several tariff waves, targeting about $100 billion of US exports. The two parties signed an agreement to halt further tariff escalations in January 2020, but the existing tariffs remained in place as of 2021.

The trade war stands out as among the largest and most abrupt change in US trade policy history, particularly when juxtaposed against the leading role historically played by the US in driving tariff reductions around the globe. As the trade war unfolded, economists attempted to assess its economic impacts. This article reviews these efforts and synthesizes what we have learned.

The research has largely explored the two central questions in international economics: What are the aggregate welfare consequences of trade barriers? How is this aggregate change distributed within a country?

To assess the aggregate impacts of the trade war, standard trade models reveal that a crucial component is the pass-through of tariffs to import prices. From the previous literature, and given the presumption that the economies of the US and China are large enough to affect prices, it would have been natural to expect an incomplete pass-through of tariffs (i.e., that import prices before applying tariffs would fall with tariffs). In contrast, empirical work has found complete passthrough of tariffs to tariff-inclusive import prices (i.e., tariff-inclusive import prices rise one-for-one with the tariff changes) when looking across exporting countries or products differentially exposed to tariffs. We discuss potential explanations for this finding. The aggregate impacts also depend on producer effects that materialize through export prices, and on changes in tariff revenue. We review approaches that have estimated these components.

The main takeaways from this research is that US consumers of imported goods have borne the brunt of the tariffs through higher prices, and that the trade war has lowered aggregate real income in both the US and China, although not by large magnitudes relative to GDP.

We also review papers that have explored the distributional consequences of the trade war through consumers and producers, employment across sectors, and spatial impacts on income and consumption across the US. We conclude with thoughts on the open questions that would be important to address in future work.

2 Background

The US-China trade war unfolded over a series of tariff waves between 2018-2019. In January 2020, the two countries signed the Phase One agreement to deescalate trade tensions in January 2020, yet the tariffs remained in place as of late 2021. We summarize the key events of the trade war, but refer readers to the summary in Bown and Kolb (2021) and the comprehensive discussion

in Bown (2021) for more details. 1

In February 2018, following a Section 201 investigation of solar panels and washing machines, the US International Trade Commission determined that imports of these products had injured domestic producers, and President Trump imposed safeguard tariffs. These first tariff waves targeted specific products from many countries. Shortly thereafter, additional tariffs on steel and aluminum were imposed based on Section 232 investigations by the Commerce Department. 2 These tariffs also targeted several countries including China, with some large economies (e.g., the European Union and Canada) initially being exempted. China and other trade partners imposed retaliatory tariffs in response.

Subsequent stages of the trade war were largely conducted between the US and China. In August 2017, the US initiated a Section 301 investigation against China's trade practices, and on March 22, 2018 the Office of the US Trade Representative accused China of unfair trade practices ranging from the forced transfer of technology to Chinese firms and intellectual property theft. On these grounds, the US ultimately implemented five tariff rounds on Chinese exports-July 2018, August 2018, September 2018, June 2019, and September 2019-with China retaliating at each stage. The US and China canceled a sixth tariff wave in December 2019 in anticipation of the Phase One agreement. Once the deal was signed in January 2020, both sides agreed to reduce the tariffs from the September 2019 wave by half, but the tariffs remain in place as of September 2021.

In terms of magnitudes, the US imposed tariffs (including other trade partners) on 17.6% of its 2017 imports. Imports as a share of 2016 GDP was about 15%, so the US raised tariffs on import transactions corresponding to about 2.6% of GDP, with average tariffs increasing from 3.7% to 25.8%. On the export side, trade partners retaliated on 8.7% of 2017 exports. 3 Exports as a share of 2016 GDP was about 12%, so trade partners imposed retaliations on exports corresponding to about 1% of US GDP, with average tariffs rising from 7.7% to 20.8%. So, the US and Chinese tariffs targeted imports and exports amounting to 3.6% of US GDP. China raised tariffs on about 11% of imports, and about 18% of their exports were targeted by the US (Chang et al. 2021). With import and export shares of 2017 GDP of 17.9% and 19.7%, respectively, the trade war affected transactions equivalent to about 5.5% of China's GDP. 4 China further hampered US access to its

2 Section 201 of the Trade Act of 1974 allows the President to impose safeguard tariffs in response to injury to a domestic industry. Section 232 of the Trade Expansion Act of 1962 allows the President to protectionist measures on imports posing a national security threat.

market by reducing its Most Favored Nation (MFN) tariffs on about 10% of its imports.

In comparison, the 1930 Smoot-Hawley legislation raised average tariffs from 34.6% to 42.5% on dutiable imports that were equivalent to 1.4% of GDP, and several foreign trade partners retaliated (Canada, which accounted for 20% of US exports, raised duties on a third of US exports to Canada); see Irwin (1998) and Irwin (2017). Thus, by the metric of GDP targeted by tariffs, the USChina trade war appears more substantial than Smoot-Hawley tariffs. 5 Another way to gauge the magnitude of the trade war is to consider the fraction of products targeted by tariffs. Roughly twothirds of ten-digit imported and exported products were targeted with tariff increases (Fajgelbaum et al., 2021a), while Smoot-Hawley raised tariffs on 27% of dutiable products (Irwin, 2017).

Of course, the context of the US-China trade war differs substantially from the Smoot-Hawley tariffs. Those tariffs were launched on the eve of the Great Depression, while global real growth in 2017 was 3.7% (IMF, 2018). 6 The nature of globalization today is also quite different, most notably through the volume of trade and its composition, with two-thirds of global trade now in intermediate goods (Johnson and Noguera, 2012; Antràs and Chor, 2021). A general tariff increase today may affect not only the prices that final consumers pay, but also the costs for firms that use those goods as inputs for production, a force that the papers discussed below try to account for. The US imposed tariffs on 67% of imported intermediate inputs and capital goods from China (representing 62% of the total Chinese imports targeted), and China also imposed tariffs on 67% of intermediate and capital goods from US (representing 81% of US imports targeted); see Bown (2021).

While we focus on the economic consequences, understanding the political motivations is also important. This episode stands alongside recent backlashes against globalization: it was launched two years after the 2016 UK Brexit referendum, a year after the US withdrew from the Trans Pacific Partnership, and a year before the US blocked the appointment of judges to the WTO Appellate Body. 7 These episodes materialized against a background of growing inequality related, among other forces, to China's ascent in the world economy (Autor et al., 2016, 2019; Pierce and Schott, 2016, 2020) and labor-saving or skill-biased technological change (Goldin and Katz, 2010; Acemoglu and Restrepo, 2019). In this context, trade policy can become a powerful electoral tool if deployed to benefit some groups. 8 Electoral motivations were indeed visible as part of the 2016 Trump Presidential campaign, which ran on a anti-globalization platform of tariffs on China. Fajgelbaum et al. (2020) provide suggestive evidence of electoral motivation; consistent with a median-voter view of politically motivated tariffs (Dixit and Londregan, 1995; Mayer, 1984). They find that counties with approximately balanced Republican and Democrat constituencies in the 2016 Presi-

dential election received more import protection from the 2018 tariff waves than heavily Democratic and Republican counties. 9 However, they also find that China's retaliatory tariffs coincided with Republican-leaning counties, largely because these counties tend to be rural and the Chinese tariffs were large in agriculture. Blanchard et al. (2019) argue that the political consequences of the trade war did not pay off for the Republican party in 2018 Congressional election, as counties more exposed to the retaliatory tariffs reduced support for Republican candidates.

3 Welfare in The Standard Trade Model

We organize the discussion around a well-known formula dating to at least Dixit and Norman (1980). In neoclassical models, the aggregate equivalent variation, EV , corresponding to a change in import tariffs-the sum of money that would suffice, if properly distributed across the agents in the economy, to leave them indifferent with a tariff change-can be expressed, to a first-order approximation, as follows:

EV = -m ' ∆( p ∗ m (1 + τ )) ︸ ︷︷ ︸ EV m + x ' ∆ p ∗ x ︸ ︷︷ ︸ EV x +∆ R, (1)

where m and x are vectors with quantities of imported and exported commodities before tariffs change, p ∗ m and p ∗ x are the import and export prices, τ are ad-valorem import tariffs, R is tariff revenue, and ∆ denotes the difference between the post- and pre-trade war outcomes. 10

The import prices p ∗ m are what the importing country faces when buying from the world, i.e., the price at the port. But buyers inside the country pay the tariff-inclusive price p ∗ m (1 + τ ). So the 'importer cost,' EV m encompasses all buyers of imported goods, both firms and final consumers. The export prices ∆ p ∗ x are the prices faced by domestic producers selling abroad, and EV x is the change in export value that they perceive. The last term, ∆ R , is the tariff revenue received by the government.

The part m ' ∆( p ∗ m τ ) of the consumer cost EV m is transferred to the government as tariff revenue and washes out from the previous expression, which simplifies to:

EV = ( p ∗ m m ) ' ( -∆ ln p ∗ m + τ ∆ ln m ) + ( p ∗ x x ) ' ∆ln p ∗ x . (2)

This formula holds in neoclassical environments regardless of how complex they may be. It holds for any input-output structure, or with heterogeneous consumers. Therefore, if one is content with the assumptions of neoclassical trade theory and with first-order approximations, to measure EV it suffices to use information on the trade flows and import tariffs before the trade war, and on the changes in international prices and in import quantities caused by the trade war.

Before going on, it is worth pointing out that this expression holds in models with perfect competition. Virtually all the estimates we review in this article were implemented under this

assumption. The welfare impacts of tariffs can be different in contexts of imperfect competition if profit-shifting considerations are strong (see Brander (1995) for a review of strategic trade policy theory).

The formula says that a country is made better off by terms-of-trade increases (lower import prices p ∗ m or higher export prices p ∗ x ) and worse off by distortions ( τ ∆ln m < 0). Conveniently, and perhaps counter-intuitively, it depends on gross trade, which is readily observable, rather than on value-added trade, which would be much harder to measure. 11 The other components, the changes in international prices and import quantities caused by the trade war, present a challenge. The empirical and quantitative approaches reviewed below impose structure to measure how ( p ∗ m , p ∗ x , m ) responded to the trade war tariffs.

Before discussing the estimates, it is worth reviewing how tariffs affect each term in (2). In principle every price and trade flow could respond to every tariff, but the theory guides the signs and welfare implications of some of these relationships. Consider first import prices. The mechanical effect of a tariff is to raise the tariff-inclusive price of the commodity being taxed, lowering the demand for imports of that commodity. Through market clearing, and assuming away well known 'paradoxes' such as in Lerner (1936) and Metzler (1949), this reduction in demand lowers the before-tariff import price but typically not enough to fully offset the tariff increase, so that the tariff-inclusive price increases. The increase in the tariff-inclusive import price relative to the tariff increase is the tariff pass-through rate ; the smaller it is (i.e., the larger the decline in p ∗ m ), the greater the gains extracted by the importer from trade partners through terms-of-trade improvements. 12 Second, the reductions in import quantities due to tariffs matter through their impact on tariff revenue. Turning to export prices, the logic we have just discussed implies that prices of exported commodities should fall with retaliatory tariffs. In addition, the flip-side of before-tariff import prices falling with tariffs is rising export prices, a manifestation of Lerner's symmetry: as tariffs reallocate resources away from export-oriented activities and demand away from imports, US producer prices increase and so do export prices. 13 Finally, import and exports prices may be indirectly affected through cost increases via input-output linkages if imported intermediates are used in production of exports.

In the next two sections, we review papers that have computed import and export prices necessary to construct the welfare consequences of the war. As noted, doing so requires measuring the full distribution of price changes triggered by the tariffs. In practice, identifying this distribution is

12 As noted by Johnson (1953), the terms-of-trade externality arising from a trade war generally gives rise to an inefficient world equilibrium. Governments may cooperate to avoid this bad outcome, as in the theory of trade agreements developed by Bagwell and Staiger (1999). Evidence that government policy is guided by terms-of-trade manipulation is provided by Broda et al. (2008), Bagwell and Staiger (2011), Bown and Crowley (2013), Ludema and Mayda (2013), and Nicita et al. (2018). Beshkar et al. (2015) argue that flexibility concerns may overturn the relationship between import market power and applied tariffs.

challenging. Tariffs affect the prices of highly disaggregated imported transactions, with a change in one tariff potentially affecting the import prices of several goods as demand is reallocated and general-equilibrium adjustments take place. Since tariff changes are not randomly assigned, identifying their impacts typically requires controlling for potential confounders at the product, sector, and/or country level; the cost of including these controls is that price effects on goods other than that being taxed cannot be inferred. For example, these controls may absorb the impacts that the tariff changes-particularly large ones-may have across products, sectors, or countries through wage adjustments required for market clearing at those levels.

The papers that we review in Section 4 typically identify the changes in US import prices from China relative to other origins within a product, or across imported products within a sector. In doing so, these empirical specifications are only able to measure relative price changes within products or within sectors, but do not estimate the full distribution of price changes. They do not identify the product-, sector-, or country-level components of the price changes potentially caused by the war. This gap between the partial distribution of price changes estimated in the regressions and the entire distribution of price changes needed for (2) can be filled in through additional assumptions in general equilibrium models, and we review those in Section 6.

4 Importers

4.1 Tariff Pass-Through

A group of papers studies the response of US import prices to US tariffs (Amiti et al., 2019; Fajgelbaum et al., 2020; Cavallo et al., 2021; Flaaen et al., 2020). This measurement allows to compute EV m . These analyses define a commodity as a product-by-origin dyad in international trade data. The presumption is that, within a product category, countries sell differentiated varieties. The typical regression examines export prices (i.e., the prices of goods at the dock before tariffs are levied):

∆ln p ∗ igt = controls -β ∆ln(1 + τ igt ) + /epsilon1 igt , (3)

where g is a product code, t is a month, i is a country exporting to the US, and the controls (that we discuss below) include fixed effects to account for trends at a broader level. 14 These fixed effects determine the source of variation that identifies pass-through and are important to interpret the results, as we discuss below. Dealing with endogeneity of tariffs is, of course, a concern in analyses of the effects of trade policy (Goldberg and Pavcnik, 2016). In the case of the trade war, this research has demonstrated that, conditional on the fixed effects that enter in the controls, the tariff changes were uncorrelated with previous price and import trends across products (e.g., Cavallo et al. 2021, Fajgelbaum et al. 2020, Amiti et al. 2020b).

The pass-through rate from tariffs to tariff-inclusive prices we have discussed in the previous section is 1 -β . Incomplete pass-through means a β between zero and one. In the case of small importing countries whose demand is unlikely to impact international prices, the pass-through should be close to complete (a β of zero). The pass-through should be incomplete (a β greater than zero) and larger for importers carrying enough weight in world demand to influence prices. For example, if an importer commands a very large share of an exporter's total sales of a product, the exporter may not easily reallocate to other markets, making supply more inelastic at least over the short run. A more inelastic supply, in turn, implies lower pass-through. So, when prices are expressed before tariffs as in (3), complete pass-through is revealed by β = 0 and incomplete pass-though by β > 0. 15

Various strands of the literature prior to the trade war provided support for incomplete passthrough; i.e., countries seem typically able to affect the terms of trade in their favor using tariffs.

First, the most closely related strand directly examined regressions similar to (3). Despite the centrality of tariff pass-through for the welfare effects of trade policy, the list of papers directly estimating tariff pass-through that were written before the trade war is small. This list includes: Kreinin (1961), Feenstra (1989), Winkelmann and Winkelmann (1998), Hummels and Skiba (2004), Mallick and Marques (2008), Ural Marchand (2012), Irwin (2014), Ludema and Yu (2016), and De Loecker et al. (2016). 16 These papers mostly find incomplete pass-through in applications across different countries and time periods (see Table 1). For example, Feenstra (1989) estimates a β of 0 . 43 from US tariff increases on Japanese trucks in the 1980s and Irwin (2014) finds a β of 0.24 from US import tariffs on sugars between 1891 and 1914. Outside this range, Winkelmann and Winkelmann (1998) find complete pass-through in the context of the trade liberalization of a small importer (New Zealand in the mid-1980's).

Second, several studies have estimated import demand and export supply elasticities that can be used to inform tariff pass-through. 17 Romalis (2007) uses NAFTA tariff cuts and finds evidence of incomplete pass-through, with an implied β of 0.79. 18 Broda and Weinstein (2006), Broda et al. (2008), Soderbery (2015), and Irwin and Soderbery (2021) do not use tariff variation, but estimate import demand and export supply elasticities following the approach from Feenstra (1994). The estimates from these papers also imply incomplete pass-through. For example, Broda et al. (2008) estimate elasticities at the product level for 15 non-WTO countries from the late 90's to



Notes: Table reports direct estimates of tariff pass-through from papers published from previous trade policy episodes. It reports the article, period and country of study, and trade policy episode. The following indicates the source of the passthrough estimate for each article (if multiple columns are reported, the estimate is an average): Row 1, Table 1 columns 1 and3; Row 2, Table 1 column 2 and 3-4; Row 3, Table 2 column 1; Row 4, Table 2 column 2; Row 5, Table 3 column 1; Row 6, Table 2 column A1-A2; Row 7, Table 2 column 4; Row 8, Table 2 column 2; Row 9, Table 8 columns 1-2; Row 10, Figure 9. The final column standardizes the estimate according to how β is defined in the text. The paper in row 6 reports pass-through to domestic prices.

early 00's. Their estimates imply a median β across all products within each country that ranges between 0 . 79 and 0 . 95 across countries. That is, their elasticities imply that every country in their sample-despite consisting of small importers except for China and Russia-has potentially strong power to manipulate international prices with tariffs. Furthermore, they find positive correlations between tariffs and inverse supply elasticities, suggesting that countries engage in some terms of trade manipulation.

Finally, a large literature has examined the related concept of exchange-rate pass-through. From the perspective of an exporter to the US, a depreciation of the exporter's currency is a positive demand shifter as much as a tariff reduction. Hence, appreciations of the exporter's currency and import tariffs should have similar effects on prices. The typical finding has been incomplete exchange-rate pass-through: a depreciation of the exporter's currency typically increases the foreign-currency prices faced by the importer. Goldberg and Knetter (1997) review earlier results that imply a pass-through rate of about β = 0 . 4; see also Burstein and Gopinath (2014). In Japanese auto imports, Feenstra (1989) finds support for the symmetry hypothesis.

Against these priors, it is surprising that several papers found pass-through to be virtually complete during this trade war (a β = 0). Moreover, the US and China account for a reasonably large fraction of each others' trade across many of the product categories affected by tariffs, so one would expect tariffs to affect bilateral prices that clear the US-China markets. Across six-digit

products in 2017, China accounted for an average 23% of US imports, and the US accounted for an average 12% of Chinese imports. The pass-through evidence is robust and it was identified by several research teams independently using different data sources and at various time horizons.

Fajgelbaum et al. (2020) match the tariff changes to publicly-available US import and export data at the HS10 product level, 19 and implement the specification in (3) for both the US and China as importers at a monthly frequency. As controls, their benchmark specification includes fixed effects by product-time, exporter-time, and exporter-sector. Therefore, the identification comes from differential variation in tariff changes across exporters to the US within a product. That is, they ask if China lowered prices relative to other exporters within products in response to tariffs. Fajgelbaum et al. (2020) estimate a β = 0 . 00 (se 0 . 08). Using the same data and a similar regression but 12-month differences instead of monthly changes, Amiti et al. (2019) estimate β = -0 . 012 (se 0 . 023). In follow-up analysis that includes the September 2019 waves, Amiti et al. (2020b) and Fajgelbaum et al. (2021a) find that the pass-through remains complete over a one-year horizon.

Those studies used publicly available trade data. While the HS10 product definitions used in these analyses are narrow, they consist of aggregates of potentially many transactions across individual goods and sellers. Cavallo et al. (2021) estimate tariff pass-through using confidential micro data on the import prices that underlie the import price indices constructed by the BLS. They estimate (3) at the monthly level with a lag structure and with the bilateral exchange rate as an additional independent variable, relying on both variation across exporters to the US and across products. 20 They find a cumulative one-year value of β = 0 . 018 (se 0 . 030). In contrast to the near complete tariff pass-through, they estimate incomplete exchange-rate pass through of 25-35%, in line with previous work. 21

The finding of complete pass-through implies that the importer cost ( EV m ) from (1) can be easily calculated as the increase in the cost of the pre-trade war import basket keeping import prices constant. This cost is the product of three terms: the import share of GDP (15%), the fraction of US imports targeted by tariff increases (17.6%), and the average import price increase among targeted varieties (which equals the average tariff increase of 22.1% due to complete pass-through). This calculation implies that imports buyers lost in aggregate 0.58% of GDP. This number means that importers have been 0.58% worse off in real terms relative to the pre-trade war scenario for the entire time that the tariffs have been in place. 22

So far we discussed the result for the US, as this has been the focus of most research. But Chang et al. (2021) and Ma et al. (2021) implement the specifications in Fajgelbaum et al. (2020) and Amiti et al. (2019), respectively, from the perspective of China, and both find complete passthrough. The complete pass-through in China reinforces the surprising nature of the results: in standard general equilibrium, if an importer faces an elastic export supply from an exporter, then that exporter would typically be able to manipulate the terms of trade using trade policy. But during the trade war, neither country seemed able to manipulate the terms of trade in its favor. 23

As we have mentioned, these pass-through calculations come with an important caveat. The estimates compare price changes across exporters within a product, or across products. So impacts of tariffs that are common across exporter or products are not reflected. For example, the results do not rule out that US tariffs lowered the overall wage level in China relative to the US, and therefore lowered the price level of all Chinese goods exported to the US. Assessing these effects require a general-equilibrium model, which we discuss in Section 6.

4.2 What Might Explain Complete Pass-Through?

We speculate on possible explanations for the complete pass-through result, although we stress that more research into the mechanism is required. For this, it is worth our reminding that complete pass-through reflects inelastic demand or elastic supply-in either case, the incidence of the tax falls on the importer. More explicitly, regression (3) can be thought of as the reduced-form of an import demand and foreign export supply system. Let

ln m = A -σ ln ( p ∗ (1 + τ )) (4)

be the US import demand of a variety (a commodity from an exporting country), where σ is the domestic import demand elasticity, p ∗ is the import price, τ is an ad-valorem tariff, and A is a demand shifter accounting for expenditures and preference shocks; 24 and let

ln p ∗ = Z + ω ln m (5)

be the inverse foreign supply from that exporter to the importing country, where ω is the inverse foreign export supply elasticity and Z is a cost shifter accounting for wages and productivity. Combining these two equations gives (3), with a β = 1 / ( 1 + ( ωσ ) -1 ) . So β = 0 occurs with

24 This foreign-export supply is a reduced-form formula that represents the difference between a foreign exporter supply and demand at market prices under assumptions of perfect competition. An important caveat is that this expression is not derived from a general-equilibrium in the foreign market, and that it assumes price-taking behavior of exporters.

infinitely inelastic domestic import demand ( σ → 0) or infinitely elastic foreign export supply ( ω → 0).

Therefore, in this simple framework, to understand complete pass-through one needs a story of inelastic demand or elastic supply. 25 Short of these alternatives, the explanation for the results could lie on the behavior of the demand and supply shifters, A and Z above, which may be also responding to the tariffs. We review the plausibility of these and other possible explanations next.

Inelastic Demand?

An explanation of complete pass-through based on inelastic demand can be plausibly ruled out by results in Fajgelbaum et al. (2020). They use the tariff variation as instruments to recover ( σ, ω ) from (4) and (5), and find σ = 2 . 53 (se 0 . 26). 26 Hence, their findings are consistent with an infinitely elastic foreign export supply but with a finite and relatively low demand elasticity across origins.

Elastic Supply?

Another possibility is that, at the variety level, supply happens to be very elastic: this would mean that China can easily reallocate exports from the US to other destinations when demand falls. Indeed, Fajgelbaum et al. (2020) estimate ω = -0 . 002 (se 0 . 05) for Chinese exporters. This hypothesis would imply that, at the product level, the value of exports to the world for the Chinese products taxed by the US should not have fallen much. Fajgelbaum et al. (2021b) estimate that, across products, US tariffs lead China to reduce exports to the US but also to increase exports to the rest of the world, and cannot reject that Chinese (product-level) exports to the world remained constant. Hence, US tariffs did not seem to impact product-level Chinese exports, suggesting seamless reallocation of Chinese exports away from US into other markets within products. Jiao et al. (2020) uses firm-level data on the universe of firms from one Chinese prefecture, and their estimates suggest no declines in firm-level sales to the world despite a decline to the US, which is again consistent with relatively easy reallocation across destinations of a product, at least for Chinese producers.

Demand Shifters?

Demand-side explanations could mask a decline in import prices before tariffs if the demand shifter A increases in products with higher tariffs. This could happen due to dynamic decisions of importers of durable goods: if current tariffs are an indication of even bigger future restrictions, buyers who can carry inventory would increase current demand in anticipation of even further

tariff increases. Alessandria and Mix (2021) provide evidence of anticipatory effects in previous tariff events, but this hypothesis has not been linked to pass-through. Another possibility related to demand shifters is that the tariffs triggered simultaneous improvements in average product quality that offset (quality-unadjusted) price declines. An extensive literature documents qualitybased selection, such that firms selling higher-quality and therefore more expensive products enter tougher destinations (see Hummels and Skiba (2004) and Manova and Zhang (2012), among others). Ludema and Yu (2016) note that quality-based selection matters for the how tariffs affect qualityunadjusted import prices: as tariffs increase, firms selling more expensive products survive, pushing import prices up. However, this explanation for complete pass-through seems unlikely for continuing products because Cavallo et al. (2021) and Flaaen et al. (2020) also document complete pass-through in more disaggregated price data. It is possible, however, for quality upgrading to operate on an extensive margin of product entry and exit, an hypothesis that has not yet been assessed.

Supply Shifters?

Various mechanisms related to the supply shifter Z could mask an upward sloping supply (and hence a finding of incomplete pass-through if this supply shifter was properly controlled for). First, firms' unit costs vary with imported intermediates, 27 and particularly so for Chinese exporters (e.g., Brandt et al. (2017) and Amiti et al. (2020a)). If firms import and export the same products, and taking into account that there is overlap in the product categories taxed by the US and China (64 percent of six-digit HS codes are tariffed by both countries), then a specification like (3) could yield very small or complete pass-through because the tariffs would raise the costs of Chinese exporters, pushing up exporter prices and potentially offsetting the reduction in prices due to lower demand. Relatedly, if supply chains involve two-way trade within narrow product categories, then US import tariffs may raise costs along the chain, also increasing the costs of Chinese exporters. Second, there is considerable evidence that trade liberalizations drive improvements in productivity. 28 So, it could be that decreased access to US markets reduced firm productivity, pushing costs up, but this seems unlikely at this time horizon. A third explanation concerns unobserved policies. It is possible that the Chinese and American governments provided subsidies to firms, such as farm subsidies in the US (Blanchard et al., 2019). If these subsidies were systematically chosen to offset foreign tariffs, Z and τ would be correlated in way to leave export prices constant.

Contracts with Sticky Prices?

The tariffs were imposed without much warning to firms, so the duties were levied on top of prices that had already been contracted. This could explain complete pass-through early on. However, complete pass-through persists for up to two years, hence it must have applied to new orders after tariffs were enacted. Jiao et al. (2020) surveyed 600 Chinese firms on how the tariffs affected their export prices, with 21% indicating inflexibility to adjust prices due to contractual agreements.

Thus, price stickiness could explain some of the pass-through finding. Still, as documented by Fajgelbaum et al. (2020) and Amiti et al. (2019), imports did fall sharply with the tariffs, suggesting that pre-existing commitments were unlikely to be binding beyond the period in which the trade war started. This means that stories based on contract inflexibility must allow for sticky prices with lower quantities. A related idea from the exchange-rate literature is that pass-through would be complete if import prices are sticky and denominated in dollars (Gopinath et al., 2010). This idea has the extra appeal of reconciling the complete pass-through to tariffs in the trade war with incomplete exchange rate pass-through. As discussed earlier, such stickiness would need to operate at horizons beyond a year to explain the data.

Market Structure?

The prediction that import prices fall with tariffs is borne out of standard neoclassical assumptions of perfect competition and increasing marginal costs. However, tariffs may increase import prices if there is foreign market power, and this force may offset the declining prices due to increasing marginal costs. This could be the case under specific conditions on the curvature of the import demand faced by a foreign monopolist (Brander, 1995), or in models with import bargaining where the terms-of-trade impacts of trade policy depends on bargaining power. 29

Level of Aggregation?

The result that β = 0 in the estimation of (3) with exporter-time fixed effects is consistent with standard general-equilibrium quantitative trade models such as the Armington model in Anderson and Van Wincoop (2003): in this model, the terms of trade move with relative wages, responding to tariffs in order to preserve market clearing. Supply is infinitely elastic to each destination, conditional on the wage. 30 So the result of complete pass-through is not surprising in the context of that model. Still, standard trade models would predict incomplete pass-through at some level of aggregation for a sufficiently large importer, or for a small importer trading differentiated products. So, even if pass-through is complete when comparing imported varieties of a product, it may not be when comparing, for example, imports of differentially exposed products. However, some of the previous papers continue to find evidence of complete pass-through across products: Fajgelbaum et al. (2020) estimate a version of (3) at the product-level (i.e., examining the change in a productlevel price index instead of ∆ ln p ∗ gct ), obtaining a product-level β = -. 09 (se 0 . 20), and Cavallo et al. (2021) rely on product-level variation for their analysis in which the find β close to zero. 31

Nevertheless, market clearing conditions necessarily lead to incomplete pass-through at some level of aggregation, such as at the sector and country level. Indeed, Cavallo et al. (2021) find that, without any controls or fixed-effects to absorb country-level trends, their regression yields some incomplete pass-through, β = 0 . 079 (0 . 026). While this result is suggestive that pass-through is present at some higher levels of aggregation, the pass-through at these higher levels is still difficult to estimate due to possible correlation with country-level trends. As discussed below, simulations of quantitative models can reveal the magnitude of these effects.

4.3 Consumer Prices

The pass-through estimations discussed above look at prices at the port . These are the prices faced by the country as a whole, and that therefore matter for aggregate welfare. As we discuss below, the complete pass-through result discussed above illustrated one important dimension of redistribution: there were considerable losses for US imports buyers, matched almost completely by gains to US producers. Of course, many buyers of imports are firms rather than final consumers.

The most direct way to assess consumer effects through higher prices is by looking at consumer prices. Flaaen et al. (2020) analyzes the retail price of one product of the trade war-washing machines-which was among the earlier products to be tariffed. As the washing-machines tax was non-discriminatory, the authors compare prices of washing machines versus other appliances. Using oven ranges as a control product, they find no evidence of incomplete tariff pass-through; if anything, their result implies more than complete pass-through with consumer washing machine prices increasing 125 percent.

However, Cavallo et al. (2021) suggest that complete pass-through to retail prices may not hold broadly. They use online prices from two large US retailers and are able to identify goods that are subject to the tariffs. In contrast to the complete pass-through at the port, they find that final consumer prices are barely affected. So, for retail goods such as a household goods and electronics, the cost of the tariffs fell on the retailers, not final consumers. However, they also document that retailers increased purchases ahead of tariff announcements, and were therefore not necessarily exposed to the tariffs. This would translate the incidence of the tariff on importers of non-retail goods (which included a large fraction of all tariffed goods), or on agents that were unable to anticipate the tariffs or stock up. It is also possible that, as the tariffs remain in place and the stocks that retailers accumulated before tariffs dwindle, the prices are eventually passed on to final consumers. In either case, an important path for future work is to assess the incidence of the tariffs across final consumers, retailers, wholesalers, and other agents within the supply chain.

5 Producers

5.1 Export Prices

The welfare formula in (2) says that the producer impact hinges on how export prices respond to tariffs. As discussed above, tariffs could affect export prices through three channels: foreign tariffs could dampen foreign demand, while domestic tariffs could increase imported input costs or reallocate expenditures to domestic goods.

Consider first the retaliatory tariff channel. US export or producer prices may fall in response to Chinese retaliation, leading to lower welfare through EV x in (1) by lowering p ∗ x . Fajgelbaum et al. (2020) and Amiti et al. (2020b) show that US variety-level export prices to China relative to other destinations did not fall in response to retaliatory tariffs, suggesting that US producers may adjust flexibly across destinations. Similar to our previous caveat that import price responses estimated across origins within a product do not capture product-level import price changes, these regressions do not capture that US producer prices could fall in products or sectors facing higher Chinese tariffs, regardless of where these products are shipped. Indeed, using BLS micro-data on exported goods, Cavallo et al. (2021) find relative price reductions in US products targeted by China. At more aggregate levels, Fajgelbaum et al. (2020) present evidence that US sector-level export price indices fell with retaliatory tariffs in the same sector. Neither them nor Flaaen and Pierce (2019) find evidence of sector-level producer prices falling with retaliation, but capturing sectoral price impacts from retaliations may be difficult since exports are typically a small share of sectors' total sales.

Turning to the second channel, US export prices could also increase through more costly access to imported inputs. In the aggregate welfare calculation of (1), these export price increases would offset part of the greater buyer cost from EV m , but on net the cost increases would be welfarereducing. Benguria and Saffie (2019), Flaaen and Pierce (2019), and Handley et al. (2020) estimate this impact of US tariffs, and generally find that indeed export prices raise with US tariffs through rising cost for imported inputs. For example, Flaaen and Pierce (2019) find that an interquartile shift in exposure to rising input costs is associated with 4.5 percent relative increase in factory-gate prices.

Finally, export prices may also increase with import tariffs, leading to higher welfare through EV x in (1). As domestic demand is reallocated away from imports into domestically produced goods, export supply is restricted and domestic goods become more scarce internationally, and therefore more expensive. Fajgelbaum et al. (2020) and Amiti et al. (2020b) find evidence suggesting that the US PPI increases with the import tariff.

5.2 Reallocations

The trade and employment reallocations from the trade war are related to changes in relative incomes and expenditures, and thus also give a sense of winners and losers. The estimated trade reallocations between US and China have the expected signs: within products, US and Chinese

imports and exports are reallocated away from each other into other origins and destinations, respectively. For example, Amiti et al. (2019) and Fajgelbaum et al. (2020) use specifications similar to (3) but with import quantity as dependent variable, finding that imports decline with tariffs at an elasticity of β = 1 . 31 (0 . 09) and β = 1 . 47 (0 . 24), respectively. Their event studies suggest that this response is persistent. On the export side, several studies also document sharp decreases in US exports to China relative to other destination in response to the lower Chinese demand from retaliatory tariffs (see Fajgelbaum et al. (2020), Amiti et al. (2020b), and Benguria and Saffie (2019)).

Fajgelbaum et al. (2021b) shift the focus away from US and China to study how other countries not directly exposed to trade-war tariffs reallocated trade. Their main specification considers, for each country, the change in product-level exports to the US, China, and the rest of the world as a function of the US and China tariffs on each other. They find that, on average, countries reallocated exports into the US and away from China in response to the US-China tariffs, and strongly increased their exports to the rest of the world. Their findings suggest downward sloping supply curves due to scale economies or other forces for a subset of countries that increased exports to both US or China and the rest of the world. Using pre-war export shares to weight the predicted export changes across products, they also find that the growth in total exports induced by the trade war was highly heterogeneous across countries, but this heterogeneity was due to country-specific supply curves or demand substitution elasticities with US and China, rather than by product-level specialization patterns.

The previous papers study trade reallocations across origins or products. Flaaen and Pierce (2019) instead examine employment reallocation. They match the tariffs to monthly industry-level employment from the Current Employment Statistics program of the Bureau of Labor Statistics. Their interest is in understanding both the impacts of tariffs on employment across industries and the role of the three channels we have already mentioned: import protection; rising input cost; and foreign retaliation. Their results indicate that the first channel modestly raises employment, but these gains are more than offset by the negative consequences of the other two channels. Overall, moving an industry from the 25th percentile to the 75th percentile of exposure along these three measures of tariff exposure reduces manufacturing employment by 2.3 percent. Thus, this study suggests that trade policy in a world with supply chains and the prospects of retaliation can undo positive direct impacts of tariffs on employment. 32

Waugh (2019) examines the impact of the Chinese retaliatory tariffs on county-level employment data from the BLS Quarterly Census of Employment and Wages. His estimates suggest that total employment responds negatively to the retaliatory tariffs: a county at the upper quartile of the tariff distribution experienced 0.75 percentage point decline in employment growth relative to the lower quartile. These magnitudes are about twice as large for private-sector employment in the goods sector, which was directly affected by the trade war tariffs.

Thus, the early evidence from these papers suggests that the trade war has not raised manufacturing employment in the US.

6 Aggregate and Distributional Effects

6.1 Adding up Consumer and Producer Impacts

Combining the impacts on prices and trade reallocations, we can compute the aggregate effects in (2). As we have discussed, a possible reading of the results on import tariff pass-through and export price effects is that neither import nor export prices moved much. Starting from a situation close to free trade, and assuming small changes in imports for products that did not face import tariffs, the absence of price changes readily implies that a first-order approximation to EV in (2) was approximately to zero. Under the same neoclassical assumptions, a closed-form expression for the second-order approximation is: EV = 1 / 2 (∆ m ) ' ∆ τ . The tariff change ∆ τ is observed, whereas ∆ m is estimated by Amiti et al. (2019) and Fajgelbaum et al. (2020), as we have discussed. Applying this formula and excluding 2019 tariffs, Amiti et al. (2019) finds a loss equivalent to 0 . 044% of GDP and Fajgelbaum et al. (2020) estimate a loss of 0 . 059%. Further including 2019 tariffs, Fajgelbaum et al. (2021a) find a loss of 0 . 17%.

These approximations are computed assuming complete tariff pass-through. However, as discussed in the last conjecture of Section 4.2, the empirical analyses at the variety level do not rule out terms-of-trade effects through changes in prices at the sector or country level. Also, these backof-envelope calculations do not consider the possible impacts of retaliatory tariffs at these higher levels of aggregation. These numbers are difficult to estimate empirically, but existing research has simulated these effects in general equilibrium models.

Fajgelbaum et al. (2020) combine the demand-side parameters estimated from a nested CES import demand system in the style of Broda et al. (2008), which accounts for substitution across products and across origins within products, with a supply side of the US economy that incorporates the three channels through which export prices respond to tariffs discussed in (5). For the supply side, they assume a static general equilibrium model where fixed factors give rise to upward sloping supply at the sector level, perfect competition, flexible prices, and an input-output structure with unitary elasticities, calibrated to match US input-output tables. Their simulations imply that US export prices p ∗ x overall rose 0 . 7% due to tariffs in 2018. When multiplied by a 7 . 4% non-service export share of GDP, this yields an increase in EV x in (1) of 0.05% of GDP. Further including the 2019 tariffs, this component increases to 0.13%.

Including all the tariffs, further adding up a simulated tariff revenue gain of 0.34% of GDP 33 and the previous result that the consumer cost EV m was 0 . 58%, the aggregate welfare loss in (1) is 0 . 10% of GDP (Fajgelbaum et al. (2020) estimate losses of 0 . 04% of GDP from only the 2018

tariffs). Chang et al. (2021) replicate their methodology on 2018-19 tariffs and find an aggregate welfare loss in China of 0.29%.

The simulations in Fajgelbaum et al. (2020) keep wages and aggregate demands in foreign countries constant. Several analyses, including Balistreri et al. (2018), Caliendo and Parro (2021), Charbonneau and Landry (2018), and IMF (2018), simulate general equilibrium impacts using multi-country environments in the style of Eaton and Kortum (2002) that further account for the equilibrium of the world economy. 34 Despite the many methodological differences with Fajgelbaum et al. (2020), the aggregate effects are similar and consistently found to be small relative to GDP and negative for both the US and China. For example, Caliendo and Parro (2021) obtain that the trade war tariffs lower US and Chinese welfare by 0 . 01% and 0 . 09% , respectively. 35 This magnitude and similarity is not surprising, given the observed trade to GDP ratios. As a benchmark, Costinot and Rodríguez-Clare (2014) show that, in a standard parametrization of these frameworks, a 100% uniform tariff imposed by the US reduces welfare by approximately 0 . 3%, whereas Baqaee and Farhi (2021) compute that a 10% universal tariff shock would increase US and Chinese welfare by 0.09% and 0.16%, respectively. 36,37

These welfare effects appear small relative to GDP, but this does not mean that the distortions due to tariffs are small. Finkelstein and Hendren (2020) calculate that the US-China tariffs have a marginal value of public funds (MVPF), defined as the ratio of real income costs of a policy to its revenue benefit, of (minus) 1.2-1.5. This implies that the tariffs are particularly costly relative to many other public policies. 38

35 The differences include the solution method (first-order in Fajgelbaum et al. (2020) versus exact solution in Caliendo and Parro (2021)), the level of aggregation for tariffs (product versus sector-level), the demand structure, and the general equilibrium analysis (fixed wages everywhere but in the US vs full wage adjustment across the world). The somewhat smaller welfare loss for the US in Caliendo and Parro (2021) could be due to the positive terms-of-trade effects for the US that they incorporate. Baqaee and Farhi (2021) show the accuracy of first-order approximations in solving quantitative trade frameworks.

36 These models have in common that they only consider static distortions from tariffs. Reyes-Heroles et al. (2020) simulate the long-run impacts of the tariffs by making steady-state comparisons in a multi-country model with capital accumulation. They find declines in GDP in the US and China of around 1% and relatively larger benefits for emerging markets than for other advanced economies due to their factor endowments and imported inputs intensity.

37 Many studies simulated the effects of the trade war using static computable general equilibrium (CGE) models with a rich sectoral structure such as GTAP (e.g., see Freund et al. (2020), Bollen and Rojas-Romagosa (2018), Bellora and Fontagné (2019), Ciuriak and Xiao (2018), Li and Whalley (2021), Gentile et al. (2020), and Itakura, 2020) or macro DSGE models with dynamic forces (e.g., capital accumulation, adjustment costs, sticky prices, or wage rigidities) such as the GIMF developed by the IMF (e.g., see Berthou et al. (2018), Bundesbank (2020), ECB (2019), Georgiadis et al. (2021), IMF (2018), and OECD, 2019). The first group reports effects on US (Chinese) output in the range between -0.17% and -0.41% (-0.30% and -1.20%). The DSGE studies estimate effects on US (Chinese) output in the range of -0.30% and -0.50% (-0.30% and -0.55%). Hunt et al. (2020) discusses how the features of these models affect these estimates.

38 In Section 2 we noted that Smoot-Hawley tariff increases were somewhat smaller but comparable to the trade war tariffs. In simulations of a general equilibrium model, Irwin (1998) finds that the tariffs lowered US welfare by 0.1%-0.4%, and by up to 1.1% when allowing the (specific) tariffs to increase because of deflation. To our knowledge,

6.2 Stock Market and Uncertainty

As the trade war unfolded, the announcements of impending tariffs through social media and official government proclamations grabbed the headlines. These announcements coincided with sharp and typically negative movements in equity markets. Some analyses have relied on these movements to identify the impact of the trade war tariffs on the valuation of exposed firms, and then add up these responses through different approaches to compute the aggregate impacts of the war. 39

Huang et al. (2020) run a three-day event study centered around March 22, 2018, when the Trump administration announced the pending imposition of 25% tariffs on the first wave of Chinese imports. Using a sample of US listed non-financial firms with sales in China or trade with that country, they find that the tariff announcements resulted in 4.3% market decline losses. They argue that about a quarter of the market losses is driven by firms' direct import and export exposure to China, while the rest is driven by changes in macro variables or indirect exposure via supply chains. Amiti et al. (2021) use 11 tariff announcements and implementations between 2018 and 2019. The market dropped a cumulative 12.9% over a three-day window around those events, and their approach attributes the vast majority of this decline to the trade war.

These numbers stand in contrast with the results based on price changes and terms of trade in static trade models, e.g. they are an order of magnitude or two larger than the approximately 0.01% to 0.1% decline from the models discussed in the last subsection. The reasons behind the striking difference in these numbers is an open question. As argued by Amiti et al. (2021), one possibility is that static models do not incorporate potentially important mechanisms, such dynamic losses or growing uncertainty. 40 The counter view is that very short-run stock market response at the time of tariff announcements may not reflect the actual impact of tariffs on fundamentals. For example, market participants may not have experience pricing an uncommon policy shock and may require a longer window in order to assess the potential impacts on the real economy. Methodologically, the event windows used in stock market analyses are much shorter and temporary than the before vs. after comparisons of tariff implementations used in the event studies we discussed in Section (4.1).

Efforts to reconcile the stock market responses with welfare impacts from trade models is a promising area of work.

6.3 Local Labor Markets and Distributional Consequences

Trade shocks including tariffs have different impacts across regions within countries, particularly when labor is imperfectly mobile (Topalova, 2010; Kovak, 2013; Autor et al., 2016). Fajgelbaum et al. (2020) merge the product-level US and China tariffs with (pre-war) counties' sectoral employment wage bills. and use their model to simulate real wage changes across US counties. Their results indicate large dispersion: on average, real wages in the tradeable sector decline by 1% but with a large standard deviation of 0.5% across counties. This dispersion reflects specialization patterns: the Great Lakes region of the Midwest and the industrial areas of the Northeast received higher tariff protection, while the Midwestern plains and Mountain West faced higher tariff retaliation, as about 27% of China's retaliations targeted US agricultural goods (Bown, 2021). Caliendo and Parro (2021) simulate state-level effects in a framework that additionally allows for internal trade costs, so that differences in real income also arise from differential price-index effects, with losses ranging from about 0 . 1% to 0 . 2% across states.

Lack of data availability has so far prevented an analysis of actual wage changes across regions to the trade war, and recent studies considered other less standard outcomes. Waugh (2019) studies welfare effects by examining consumption patterns across counties. 41 The paper exploits data on monthly-level new automobile registration, which is available at the county level in nearly real time, as a proxy for consumer expenditures. The data captures consumption of only one (albeit important) durable good and records counts as opposed to values. Assuming that county price changes are uncorrelated with tariff exposure, differences in auto expenditures are indicative of income changes. The paper finds that, relative to a county in the lower quartile of the retaliatory tariff distribution, a county in the upper quartile experiences a 3.8 percentage point decline in auto sales. In China, Chor and Li (2021) consider the spatial impact of the trade war by examining changes in nightlight intensity against regional exposure to tariffs, and their analysis also reveals large heterogeneity in impacts across locations.

7 Open Questions

The US-China trade war is the most important trade policy shift in recent decades, and it provides researchers with an unusual opportunity to study the mechanics of global trade. Throughout this article we mentioned several questions that remain under-explored, including the various conjectures for complete pass-through that we have discussed. We conclude with a few additional questions that deserve further scrutiny.

First, a natural follow-up is to examine the longer-run aggregate effects of the trade war. This includes re-examining pass-through at different time horizons and levels of aggregation, but also examining systematically how pass-through varies along the supply chain from the dock to retail

prices. Second, what are the impacts of trade barriers imposed by US and China in terms of production relocation, risk versus efficiency trade-offs, and national security implications? Third, as noted, US and Chinese firms could request exemptions from import duties, which raises natural questions: How did expectations of potential tariff rebates affect importer behavior, and what political-economy channels influenced the decisions of firms to seek exemptions and the government decisions to grant them? Fourth, what are the dynamic implications from potential reductions in investment, changes in capacity, and access to imported inputs?

Turning to the distributional consequences of the trade war, the studies to date analyze proxies for consumption, short-run employment effects, and model-based implications for wages. As time passes, what will survey and administrative data reveal about the distributional consequences through consumption and income channels? What will be the long-run implications for the spatial distribution of economic activity; e.g., will the trade war 'undo' the labor market impacts of the China shock (Autor et al., 2016)? (As discussed, the evidence thus far suggests 'no.')

Finally, outside the scope of this review are questions about the geopolitical implications of the US-China trade war. Answering such questions often do not lend themselves easily to clean econometric studies, but insights may be learned from theoretical frameworks and work from other fields. Such questions include: How much global 'soft power' did the US gain or lose by initiating the trade war? How will the trade war affect the political relationship between US and China? And, how will the world trading system evolve in response; for example, Staiger (2021) is a recent contribution to this debate.

References

Acemoglu, D. and P. Restrepo (2019): 'Automation and New Tasks: How Technology Displaces and Reinstates Labor,' Journal of Economic Perspectives , 33, 3-30.

Alessandria, G. A. and C. B. Mix (2021): 'Trade Policy is Real News: Theory and Evidence,' Tech. rep., National Bureau of Economic Research.

Amiti, M., M. Dai, R. C. Feenstra, and J. Romalis (2020a): 'How did China's WTO entry affect US prices?' Journal of International Economics , 126, 103339.

Amiti, M., S. H. Kong, and D. Weinstein (2021): 'Trade Protection, Stock Market Returns, and Welfare,' Tech. rep., National Bureau of Economic Research.

Amiti, M. and J. Konings (2007): 'Trade liberalization, intermediate inputs, and productivity: Evidence from Indonesia,' American Economic Review , 97, 1611-1638.

Amiti, M., S. J. Redding, and D. E. Weinstein (2019): 'The impact of the 2018 tariffs on prices and welfare,' Journal of Economic Perspectives , 33, 187-210.

---(2020b): 'Who's paying for the US tariffs? A longer-term perspective,' in AEA Papers and Proceedings , vol. 110, 541-46.







Johnson, H. G. (1953): 'Optimum tariffs and retaliation,' The Review of Economic Studies , 21, 142-153.

Johnson, R. C. and G. Noguera (2012): 'Accounting for intermediates: Production sharing and trade in value added,' Journal of international Economics , 86, 224-236.

Kang, Z. Y. and S. Vasserman (2021): 'Robust Bounds for Welfare Analysis,' Available at SSRN 3926399 .

Kovak, B. K. (2013): 'Regional effects of trade reform: What is the correct measure of liberalization?' American Economic Review , 103, 1960-76.

Kreinin, M. E. (1961): 'Effect of tariff changes on the prices and volume of imports,' The American Economic Review , 51, 310-324.

Lerner, A. P. (1936): 'The symmetry between import and export taxes,' Economica , 3, 306-313.

Li, C. and J. Whalley (2021): 'Trade protectionism and US manufacturing employment,' Economic Modelling , 96, 353-361.

Ludema, R. D. and A. M. Mayda (2013): 'Do terms-of-trade effects matter for trade agreements? Theory and evidence from WTO countries,' The Quarterly Journal of Economics , 128, 1837-1893.

Ludema, R. D. and Z. Yu (2016): 'Tariff pass-through, firm heterogeneity and product quality,' Journal of International Economics , 103, 234-249.

Ma, H., J. Ning, and M. J. Xu (2021): 'An eye for an eye? The trade and price effects of China's retaliatory tariffs on US exports,' China Economic Review , 101685.

Mallick, S. and H. Marques (2008): 'Passthrough of exchange rate and tariffs into import prices of India: currency depreciation versus import liberalization,' Review of International Economics , 16, 765-782.

Manova, K. and Z. Zhang (2012): 'Export prices across firms and destinations,' The Quarterly Journal of Economics , 127, 379-436.

Mayer, W. (1984): 'Endogenous tariff formation,' The American Economic Review , 74, 970-985.

McLaren, J. (1997): 'Size, sunk costs, and Judge Bowker's objection to free trade,' The American Economic Review , 400-420.

Metzler, L. A. (1949): 'Tariffs, the terms of trade, and the distribution of national income,' Journal of Political Economy , 57, 1-29.

Nicita, A., M. Olarreaga, and P. Silva (2018): 'Cooperation in WTO tariff waters?' Journal of Political Economy , 126, 1302-1338.

OECD (2019): 'OECD Economic Outlook,' .

Pavcnik, N. (2002): 'Trade liberalization, exit, and productivity improvements: Evidence from Chilean plants,' The Review of economic studies , 69, 245-276.

Pierce, J. R. and P. K. Schott (2016): 'The Surprisingly Swift Decline of US Manufacturing Employment,' American Economic Review , 106, 1632-62.