Trade Costs and Inflation Dynamics

Board of Governors of the Federal Reserve System International Finance Discussion Papers ISSN 1073-2500 (Print) ISSN 2767-4509 (Online) Number 1411 July 2025 Trade Costs and Inflation Dynamics Pablo Cuba-Borda, Albert Queralto, Ricardo Reyes-Heroles, and Mika¨el Scaramucci Please cite this paper as: Cuba-Borda, Pablo, Albert Queralto, Ricardo Reyes-Heroles, and Mika¨el Scaramucci (2025). “Trade Costs and Inflation Dynamics,” International Finance Discussion Papers 1411. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/IFDP.2025.1411. NOTE: International Finance Discussion Papers (IFDPs) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the International Finance Discussion Papers Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from the Social Science Research Network electronic library at www.ssrn.com.

∗ Trade Costs and Inflation Dynamics Pablo Cuba-Borda Albert Queralto Federal Reserve Board Federal Reserve Board Ricardo Reyes-Heroles Mika¨el Scaramucci Federal Reserve Bank of Dallas UCLA May 2025 Abstract Westudyhowtradecostshocksinfluenceinflation. Usingbilateraltradeflowsfromdetailed globalinput-outputdataandagravityframework,weestimatetradecostshocksandtheireffects on CPI inflation. Higher trade costs for final goods cause large but short-lived inflation spikes, while increased costs for intermediate inputs trigger more persistent inflation. A multi-country model of inflation with trade in final goods and intermediate inputs replicates these patterns. We show that trade cost shocks and tariffs on imported inputs transmit through global value chainsandworsenmonetarypolicytrade-offs. Weusethemodeltoquantifytheeffectsoftrade costsduringthe2018–2019U.S.-Chinatradewarandtoestimatethecontributionoftradecosts during the post-pandemic inflation surge. Novel data on U.S. domestic sourcing shares allow us to estimate trade cost shocks for the U.S. using Bayesian methods. JEL Codes: E1, E3, E5, F4, F6. Keywords: Inflation, international trade, trade costs, New Keynesian model, post-pandemic inflation, monetary policy, gravity equations. ∗Firstversion: February2024. Theviewsinthispaperaresolelytheresponsibilityoftheauthorsandshouldnot beinterpretedasreflectingtheviewsoftheBoardofGovernorsoftheFederalReserveSystemorofanyotherperson associated with the Federal Reserve System. We thank seminar participants at the University of Washington, the Federal Reserve Board, the Dallas Fed, the LAJCB Conference, CEBRA 2024 Annual Meetings, 2025 ASSA/AEA Annual Meetings, Brandeis University, West Coast Workshop in International Finance, the Mid-Atlantic Trade Workshop, and SCIEA Meeting at the NY Fed. We are particularly grateful to our discussants Matteo Cacciatore, Thuy Lan Nguyen, and Julian Di Giovanni, and to Francesco Bianchi, Daniel Xu, Carter Mix, Martin Uribe and Stephanie Schmitt-Groh´e for comments and suggestions. Pablo Cuba-Borda: pablo.a.cubaborda@frb.gov. Albert Queralto: albert.queralto@frb.gov Ricardo Reyes-Heroles: ricardo.reyes-heroles@dal.frb.org Mika¨el Scaramucci: mscaramucci@g.ucla.edu

1 Introduction The global economy is deeply interconnected. Over the past half century, economic integration by low- and middle-income countries and the deepening of global production networks—also referred to as global value chains (GVCs)—fueled world trade flows, particularly those of goods used in intermediate stages of production.1 In this integrated global economy, shocks to the costs of transaction and transport associated with the exchange of goods across borders—that is, broadlydefined trade costs—can have important macroeconomic consequences. Recent events, such as shifts in U.S. trade policy since 2017, disruptions in GVCs following the COVID-19 pandemic, or trade fragmentation triggered by geopolitical events such as the Russian invasion of Ukraine in 2022, highlight the potential consequences of trade cost shocks for real economic activity and inflation. Yet, existing work has focused on the real effects of trade costs, with scant attention paid to how these shocks impact inflation.2 This gap reflects the emphasis in the literature on the role of productivity and demand shocks for open economies and the divide between international trade models and the workhorse framework for the analysis of inflation dynamics and monetary policy.3 This paper studies how trade cost shocks shape inflation dynamics in a global economy. We followtwocomplementaryapproaches. Empirically, weexploitglobalinput-outputdatatomeasure trade costs in both final and intermediate goods across countries and time and estimate the causal effects of shocks to these costs on inflation. On the theory front, we develop a dynamic general equilibrium model of trade and inflation to quantify and elucidate the mechanisms through which trade cost shocks propagate to inflation. Thus, our study contributes to bridging the divide in the literature between international trade and international monetary economics. Our analysis proceeds in three steps. First, we estimate bilateral trade costs for final goods and intermediate inputs by combining detailed sectoral data from global input-output tables with gravity equations derived from a static Armington model of trade in final and intermediate goods.4 1World exports rose from 16 percent of world GDP in the 1970s to 29 percent by the late 2010s. Hanson (2012) and Reyes-Heroles et al. (2020) document the rise of low- and middle-income countries in global trade. Antra`s and Chor(2022)surveytheliteratureonGVCs. Worldimportsofintermediategoodsaccountedforatleasthalfofworld trade every year from 1996 to 2022 (Cuba-Borda et al., 2025). 2Recent exceptions that consider the effects on inflation of trade cost shocks are Comin and Johnson (2020), Barattieri et al. (2021), Di Giovanni et al. (2023), Comin et al. (2024), and Kalemli-O¨zcan et al. (2025). Furceri et al. (2018) studies empirically the macroeconomic implications of tariff shocks. 3For the role of productivity shocks see Backus et al. (1992) and Heathcote and Perri (2002). Stockman and Tesar (1995) and Bai and R´ıos-Rull (2015) study the implications of demand shocks. A relatively small number of works study the implications of trade costs for outcomes other than inflation (Fitzgerald, 2012; Eaton et al., 2016b; Reyes-Heroles, 2017; Alessandria and Choi, 2021). See Gal´ı (2015) for an introduction to the workhorse framework to study inflation, economic fluctuations, and monetary policy. 4See Head and Mayer (2014) for an overview of gravity models in international trade. 1

Using bilateral trade flows and aggregate spending data together with an estimate of the trade elasticity, we apply the ratio-type estimator of Head and Ries (2001) to infer bilateral trade costs across 44 countries from 1995 to 2020. This approach captures the cross-country variation in bilateraltradeflowsthatcannotbeexplainedbyeitherimporter-orexporter-specificcharacteristics and recovers bilateral trade costs as the factor that explains the remaining variation as implied by a structural gravity equation—that is, we recover bilateral residuals. Our estimated bilateral trade costs vary across time and space. The evolution of trade costs provides evidence of a process of increasing globalization, which has stalled since 2010. Across space, bilateral trade costs are sizable and heterogeneous, particularly for low- and middle-income countriesrelativelyisolatedfrominternationaltrademarketsinthe1990s. Ourmeasureofbilateral trade costs also captures changes in trade policies, such as the tariff increases between the U.S. and China between 2018 and 2019. We then study the causal relationship between changes in trade costs and inflation using panel local projection methods, as in Jord`a (2005). To do so, we construct country-wide import costs by aggregating bilateral trade costs.5 Thus, our measure of import costs also varies across time and space. We show that our measure of import costs is correlated with effective tariff rates across countries, providing reassurance that changes in trade policies are reflected in our measure of costs. We exploit the variation in import costs to identify the causal effect of trade costs on inflation. Our local projections specification controls for time- and country-fixed effects and additional macroeconomic covariates. Our empirical strategy effectively controls for importer- and exporter-specific characteristics across countries year-by-year, giving us confidence that we can identify a causal relationship between trade costs and inflation. Our findings provide novel evidence that higher import costs translate into higher inflation. We also find that the resulting inflation dynamics depend on the nature of the trade cost shocks. While higher trade costs for final goods lead to large but short-lived increases in inflation, higher trade costs for intermediate inputs generate more modest but longer-lasting inflation. Quantitatively, a ten percentage point rise in relative import costs of final goods leads to an average relative increase in inflation of 0.7 percentage points within one year that dissipates quickly thereafter. The same increase in the import cost of intermediate inputs increases inflation by 0.6 percentage points in the first year, but in this case, inflation remains elevated by about 0.2 percentage points for up to five 5OurapproachresemblesthoseofFrankelandRomer(1999)andFrankelandRose(2002),whoaggregatepredicted bilateral trade flows based on geographic characteristics in gravity models. 2

years. We also explore the effects of higher trade costs on GDP and show that increases in import costs trigger declines in output. Thus, our analysis provides evidence that the macroeconomic effects of higher trade costs—which can be triggered by non-policy and policy-related factors like tariffs—mirror those of adverse supply shocks. In the second step of our analysis, we develop a dynamic multi-country general equilibrium model of trade and inflation that features international trade in both final and intermediate goods andstaggeredpriceandwageadjustments. Thegoalofdevelopingthisframeworkistwofold. First, weaimtostudythetransmissionmechanismsthroughwhichtradecostshockspropagatetoinflation and the macroeconomy more generally. Second, we use our model to provide complementary estimates of the macroeconomic effects of higher trade costs on inflation, given that our empirical estimates only identify relative effects—that is, our empirical analysis is subject to the “missing intercept” problem that arises from including time-fixed effects in our panel local projections. Our general equilibrium model embeds the static Armington model of trade into a dynamic framework. Each country produces a unique homogeneous good that can be traded across borders, subject to good-specific iceberg-type trade costs and ad valorem tariffs. Goods can be used as final goods or intermediate inputs in producing nontradable differentiated goods in each country. The model implies static gravity equations for final and intermediate goods that are in line with those that we exploit to measure trade costs empirically. Firms producing nontradable differentiated goods adjust prices infrequently, and households face nominal wage rigidities in labor markets. International trade in financial assets is restricted to non-state-contingent bonds denominated in U.S.dollars. WeassumethatmonetarypoliciesfollowstandardTaylor-typerules. Wecalibratethe model to observed trade flows across five regions: the U.S., China, advanced non-U.S. economies, Asian emerging market economies (EMEs) excluding China, and other EMEs. To elucidate the mechanisms common across countries that shape the effects of trade costs shocks on inflation, we keep heterogeneity across countries limited in our calibration and only allow countries to differ in their sizes and trade flows with their trading partners. Our model reproduces the inflation dynamics caused by the higher trade costs we estimate in the data. In the model, increases in trade costs on intermediate inputs trigger more persistent inflation because these increases lead to persistently higher marginal costs for domestic producers, which rise as higher input prices propagate through global value chains. These higher marginal costs are passed through slowly to consumer prices as domestic producers gradually adjust prices to reflect higher marginal costs. In contrast, the model generates large but short-lived increases in 3

inflation in response to higher trade costs for final goods, as these higher costs immediately pass through to final consumer prices via higher prices of imported final goods—leading to dynamics that resemble a one-time rise in the price level. We also use our model to analyze the macroeconomic dynamics in response to trade shocks under an alternative monetary policy rule that targets inflation of domestically produced goods only, rather than all consumption goods, including imported final goods. We show that in the presence of shocks to trade costs of intermediate inputs, this alternative rule does not have the stabilization properties suggested in previous literature (Corsetti et al., 2010), nor does it “see through” temporary increases in inflation driven by the imposition of tariffs—an issue that has featured prominently in past monetary policymaking discussions.6 In the third and final step of our analysis, we use the model to estimate the global economic impact of the 2018-19 U.S.-China trade war and the role of trade costs in driving the U.S. inflation surge post-2020. The fact that our model accommodates multiple countries and that it is amenable to estimation makes it a useful tool to study these events. Moreover, this analysis highlights how our model can be used to understand episodes related to tariff changes and those involving broader tradecosts. Inourfirstexperiment,theincreaseinU.S.-imposedtariffsandtheretaliationobserved in 2018-2019 led to a rise in U.S. inflation of up to 0.3 percentage points per year and a decline in the level of U.S. GDP of around 0.4 percent. Because the inflation response is persistent, the price level rises by up to 30 percent more than what is implied by rules of thumb based on estimates of the pass-through of tariffs to import prices (Amiti et al., 2019; Fajgelbaum and Khandelwal, 2022). Hence, our model analysis underscores the role of general equilibrium effects of higher import costs of intermediate inputs on prices. In the second experiment, we construct novel quarterly data that helps identify variation in U.S. trade costs and use Bayesian methods to estimate the role of trade costs in the post-pandemic surge in inflation. We find that shocks to trade costs contributed to preventing deflation in the early phases of the pandemic and added about one percentage point to inflation in 2022 and 2023, thus helping explain the persistence of the inflation surge experienced after the COVID-19 pandemic. Relation to the Literature. This paper is most closely related to recent work that incorporates trade in intermediate goods and production networks to study the effects of trade and trade costs on inflation.7 Comin and Johnson (2020) study how trade integration in final 6See, for example, the alternative scenarios in the Federal Reserve’s September 2018 Tealbook: https://www. federalreserve.gov/monetarypolicy/files/FOMC20180131tealbooka20180119.pdf. 7A body of existing work—partially motivated by Obstfeld and Rogoff (2000)—studies the real macroeconomic 4

and intermediate goods has shaped the long-term trend of U.S. inflation. Barattieri et al. (2021) studytheshort-runmacroeconomiceffects—includinginflation—ofprotectionistpoliciesinCanada. Di Giovanni et al. (2023) study the drivers of inflation during the COVID-19 pandemic through the lens of a multi-country and sector New Keynesian model. Kalemli-O¨zcan et al. (2025) develop a multi-country New-Keynesian model with multiple sectors and input-output linkages to study the interaction between monetary policy and trade.8 Relative to this body of work, our contribution is threefold. First, we document how broadlydefined trade costs, nesting non-policy and policy-related trade barriers, vary across time and space. Second, we exploit panel data for a large number of countries to document the effects of tradecostshocksoninflationandprovidenovelevidenceaboutthemagnitudeandpersistenceofthe inflationary effects of these shocks. We show that trade cost shocks act as negative supply shocks, leading to higher inflation and lower output in the short run. Third, we develop and estimate a multi-country general equilibrium New Keynesian model that elucidates the mechanisms that explain our empirical results and that enables us to quantify the effects of increases in tariffs and other types of trade disruptions.9 Ourpaperalsocontributestotherecentliteratureonthemacroeconomicconsequencesoftariffs and their implications for monetary policy. Furceri et al. (2018) and Caldara et al. (2020) provide empirical evidence of the macroeconomic effects of tariffs and uncertainty induced by them. Erceg et al. (2023) analyze the interaction between trade and fiscal policies in open economies. Auclert et al. (2025) study the short-run macroeconomic effects of tariffs. Some of this work has focused primarily on the monetary policy implications of tariffs in open economy New Keynesian models (Bergin and Corsetti, 2023; Bianchi and Coulibaly, 2025; Werning et al., 2025). Our paper differs from the above literature along key dimensions. First, we study the effects of shocks to broadly defined trade barriers rather than focusing on tariff shocks. Our measure of trade cost shocks nests tariff shocks, is consistent with gravity models of international trade (Head and Mayer, 2014), and avoids issues related to missing tariff data (Teti, 2025). Second, we provide empirical estimates of the effects of these shocks on the macroeconomy using panel data. Third, we effects of trade costs in the presence of trade in intermediates and production networks (Fitzgerald, 2012; Reyes- Heroles, 2017; Eaton et al., 2016b; Alessandria et al., 2023). 8Additional related work considers the effects of import constraints (Comin et al., 2024) and import competition (Amiti et al., 2024) on inflation; the effects of globalization on international inflation co-movement (Ho et al., 2022) and inflation dynamics (Bianchi and Civelli, 2015; Hottman and Reyes-Heroles, 2023). 9Our focus on the transmission of trade cost shocks through value chains in intermediate inputs aligns with evidence in Flaaen and Pierce (2019), which shows that the 2018-19 U.S.-China trade war had significant effects on U.S. manufacturing prices and employment through this channel. 5

emphasize the multi-country feature of our framework, which allows us to take into account trade diversion following trade cost shocks. Fourth, we estimate our model using Bayesian techniques to quantify the role of trade disruptions in shaping the recent surge in U.S. inflation. Lastly, this paper is also related to the literature in international trade using gravity models to estimate trade costs. Within this literature, a small number of works have exploited the evolution of estimates of trade costs over time to study their effects on the macroeconomy (Jacks et al., 2008, 2011; Fitzgerald, 2012; Eaton et al., 2016b,a; Reyes-Heroles, 2017). We contribute to this line of work by establishing a causal link between trade cost shocks and inflation dynamics.10 Roadmap. The rest of the paper is organized as follows. Sections 2 and 3 describe our procedure to identify trade costs and estimate the effects of trade cost shocks on inflation. Section 4 presents the model and its calibration. Sections 5 and 6 analyze the predictions of the model and quantify the effects of tariffs and trade shocks during the 2018-2019 U.S.-China trade war post-pandemic inflation surge. Section 7 concludes. 2 Trade Costs Across Time and Space 2.1 Measuring Trade Costs Trade costs are the centerpiece of our analysis. Observing or directly measuring the total cost of shipping goods across borders is impossible (Anderson and van Wincoop, 2004). Therefore, to measure these costs, we follow the literature in international trade that infers trade costs based on observed trade flows and gravity equations implied by a type of static trade model usually referred to as Gravity model.11 While this approach requires imposing a lot of structure on the elements of a model that dictate bilateral trade flows across countries, it allows us to back out trade costs as model residuals in a very transparent fashion. In addition, our approach allows us to circumvent any issues related to obtaining reliable direct measures of trade costs and to compare trade costs across space and over time. In the following section, we present the trade bloc of the open economy dynamic model that we will develop to study the mechanisms through which trade cost shocks shape inflation dynamics. Section 4 will make clear how we embed the static bloc into our dynamic framework. The static 10OurapproachisalsorelatedtoFrankelandRomer(1999)andFrankelandRose(2002),whoestimatetheeffects of openness and currency unions, respectively, on output relying on the gravity model of trade and cross-sectional data. 11Gravity models constitute the workhorse framework in international trade to estimate bilateral trade flows and their determinants (Head and Mayer, 2014). 6

blocofthemodeldeliversgravityequationsthatallowustorecovertradecostsfromobservedtrade flows. Morespecifically,weconsiderastandardArmingtonmodeloftradeforourstaticbloc.12 The equilibrium of this bloc delivers predictions for bilateral trade flows for both final and intermediate goods in any given period t that depend only on prices of goods and aggregate spending in period t. Hence, the model implies static gravity equations. However, in the dynamic model that we present in Section 4, prices and aggregate expenditures are determined endogenously and in line with optimal dynamic decisions. 2.2 Model I: Intratemporal Trade and Gravity In any given period t = 1,2,..., the world is comprised of multiple countries indexed by i,h ∈ I = {1,...,N}. Each country produces a unique tradable good that is available to all countries— that is, there is national product differentiation. There are two types of economic agents in each country: householdsandfirms. Goodsproducedinanycountrycanbeboughteitherbyhouseholds or firms in all countries around the world. Households buy goods for final use and firms buy them to use as intermediate inputs. In country i, households aggregate goods across sources into a single nontradable composite consumption good, C . Similarly, firms aggregate goods to obtain a i,t composite intermediate input to be used in production, M . This aggregation is done according i,t to a constant elasticity of substitution (CES) aggregator given by (cid:32) (cid:88) N ηQ−1 (cid:33) ηQ ηQ −1 Q i,t = (Q ih,t ) ηQ , (1) h=1 where Q ∈ {C,M} and ηQ > 1. In (1), Q denotes the use by country i of goods of type ih,t Q ∈ {C,M} produced in h at time t, where C and M stand for final and intermediate goods, respectively. Let P denote the price of the goods produced and sold in country i expressed in local currency i,t units. If E denotes the nominal bilateral exchange rate between countries i and h expressed in ih,t terms of country i’s currency units per unit of country h’s currency, then the price of a good produced and sold in country h in terms of country i’s currency is defined by P ≡ E P . ih,t ih,t h,t Trade across countries is subject to iceberg-type trade costs given by τQ ≥ 1, implying that ih,t for one unit of good of type Q ∈ {C,M} produced in h to be delivered to i, τQ units have to be ih,t 12Our model is isomorphic to one in which trade arises from Ricardian comparative advantages as in Eaton and Kortum (2002). 7

shipped at time t. That is, τQ −1 units of the good disappear when it is shipped internationally ih,t from country h to country i.13 We normalize domestic trade costs such that τQ = 1 for every i ii,t and Q ∈ {C,M}. We assume that international goods markets are perfectly competitive, which implies that the Law of One Price (LOP) holds across countries. Hence, the price in local currency units that country i has to pay to acquire one unit of the good produced in country h, PQ , is given by ih,t PQ = τQ P , (2) ih,t ih,t ih,t for goods of type Q ∈ {C,M}.14 Householdsandfirmsincountryiseektominimizeexpenditureonfinalandintermediategoods, respectively, when choosing {Q } for Q ∈ {C,M}. The solution to this minimization problem ih,t h delivers conditional demand functions for goods of type Q ∈ {C,M} given by (cid:32) τQ P (cid:33)−ηQ ih,t ih,t Q = Q , (3) ih,t PQ i,t i,t where PQ ≡ (cid:32) (cid:88) N (cid:16) τQ P (cid:17)1−ηQ (cid:33) 1− 1 ηQ (4) i,t ih,t ih,t h=1 denotes the ideal price index for composite goods Q. Let XQ ≡ PQ Q denote expenditure by ih,t ih,t ih,t country i on goods of type Q produced in country h—that is, bilateral imports by i from h. Then, XQ = (cid:80)N XQ = PQQ denotes total expenditure by country i on Q-type goods. Equation i,t h=1 ih,t i,t ih,t (3) then implies that (cid:32) τQ P (cid:33)−(ηQ−1) XQ = ih,t ih,t XQ, (5) ih,t PQ i,t i,t which determines bilateral trade flows given current trade costs, prices, and aggregate expenditure. Let ωQ denote the share of expenditure by country i on goods of type Q produced in country ih,t 13In section 4, we introduce into the model tariffs that differ from non-tariff barriers. 14Note that our assumption that (2) holds implies that bilateral trade costs, τQ , in our model also capture any ih,t bilateral factors driving deviations from the LOP including destination-specific markups by exporters. 8

h, ωQ ≡ X i Q h,t. Equation (5) implies that these shares are given by ih,t XQ i,t (cid:32) τQ P (cid:33)−(ηQ−1) ωQ = ih,t ih,t , (6) ih,t PQ i,t where the trade elasticity in this model is given by ηQ − 1. Note that (6) is a gravity-type equation, implying that bilateral trade flows across countries can be expressed in terms of importer i characteristics, exporter h characteristics, and a measure of bilateral trade frictions inclusive of the bilateral nominal exchange rate, τQ E , summarizing all frictions that impede trade across ih,t ih,t any two countries. We proceed now to describe how we can exploit (6) to infer bilateral trade costs relying only on data for bilateral trade flows. Note from (6) that we can use a country’s domestic sourcing share, given by ω , to control for ii,t the price of the goods produced in the exporting country (denominated in local currency units) by dividing ωQ by ωQ . More specifically, given (6) for importer i and exporter h, we can express ih,t hh,t bilateral trade costs between these countries as a function of importer i’s bilateral trade share, the exporter h’s domestic sourcing share, and prices as follows: (cid:32) ωQ (cid:33)− ηQ 1 −1 PQ τQ = ih,t i,t . (7) ih,t ωQ E PQ hh,t ih,t h,t Hence, the equilibrium of our model implies that, given data on bilateral trade shares, domestic sourcing shares, and relative prices across countries for each type of good Q ∈ {C,M}, we can recover bilateral trade costs in any given period t conditional on a value of the parameter ηQ > 1.15 Relative prices across countries are difficult to measure, and time series for these prices are scarce. These issues imply that it is difficult to gather reliable time series data for these prices.16 However, we can circumvent these issues by further manipulating equation (7) to obtain a measure of bilateral trade frictions that only depends on bilateral trade flows and not prices. By switching the roles of the importing and exporting countries in (7) to control for relative price differences, we obtain a measure of trade frictions between individual country pairs given by TQ ≡ (τQ τQ )2 1 = (cid:18) ω ih,t ω hi,t (cid:19)− 2(ηQ 1 −1) , (8) ih,t ih,t hi,t ω ω hh,t ii,t 15See Reyes-Heroles (2017) for an application of this procedure. 16See Jacks et al. (2011) and Jacks et al. (2008) for other works exploiting the availability of historical data to construct HR indices for the period 1870-2000. 9

whichdefineswhattheliteraturereferstoastheHead-Ries(HR)indexforcountrypair(i,h)(Head and Ries, 2001; Eaton et al., 2016b). The HR index measures bilateral trade frictions in period t by considering the geometric mean of bilateral trade cost τj for any pair of countries. Note that ih,t TQ = 1, which is consistent with the notion that trade with oneself is costless and that, under ii,t the assumption of symmetric trade costs, the index becomes the actual bilateral trade cost. While this measure has multiple appealing features, it cannot account for asymmetries in bilateral trade costs. However, for a given importer i, changes in these bilateral measures will reflect changes in import costs against all of i’s trading partners. We exploit this feature of this bilateral measure of frictions in Section 2.5 to construct a measure of import costs at the country level. From now on, whenever we refer to bilateral trade costs, we will be referring to TQ . ih,t Ourproceduretoconstructbilateraltradefrictionsseparatesthefractionofvariationinbilateral trade flows across countries that is not driven by either importer- or exporter-wide characteristics and assigns this fraction of variation to bilateral trade costs—that is, a bilateral residual. Hence, variation across space in our measured frictions is explained by bilateral factors—these could be geographical characteristics typically considered in gravity model such as distance between two countries—and controls for country-wide factors, such as differences in country size or inflation, which could also explain part of trade flows. In that sense, our bilateral trade costs capture factors that could be exogenous to aggregate supply or demand shocks.17 In Section 3 we describe the strategy we follow to identify the causal effect of these shocks on inflation. Note that there are two important requirements for constructing bilateral trade costs. First, we require data on countries’ “domestic” trade flows, XQ —or equivalently domestic sourcing shares, ii,t ωQ —forfinalandintermediategoods. Dataforbilateraltradeflowsatthegoodandsectorallevels ii,t are provided by multiple datasets, but this is not the case for domestic trade. Second, we require measures of the elasticities, ηQ for Q ∈ {C,M}, a natural requirement given that our procedure is similar to inferring prices from quantities. In the following section, we describe the data that allow us to compute domestic trade and discuss our choice of values for elasticities. Given values for elasticities, our data will enable us to construct bilateral trade costs for many countries over three decades. 17See Frankel and Romer (1999) for a similar approach to identify exogenous variation in trade. 10

2.3 World Input-Output Data and Trade Elasticities Weusedataonworldinput-outputtables. Thesedataallowus(i)tocomputeinternationalbilateral trade flows, (ii) to compute domestic sourcing flows that are consistent with international trade flows and production data, and (iii) to distinguish these flows between goods used either as final goods or as intermediates in production. OurmaindatasourceistheInter-CountryInput-OutputTables(ICIO)publishedbytheOECD. The ICIO provides global input-output tables, which map flows of production and expenditure within countries and flows of international trade between countries, broken down by economic activityandcountry,globally.18 TheICIOconsiders76countriesandtherestoftheworldaggregate (ROW) from 1995 through 2020. We aggregate the ICIO data to 41 countries—40 individual countries and one ROW aggregate. This aggregation allows us to compare our measures of bilateral trade costs with those obtained usingtheWorldInput-OutputDatabase(WIOD)(Timmeretal.,2015),adatasetwidelyusedinthe existing literature.19 We exclude non-tradable goods from our analysis—basically, service sectors— and consider tradable sectors, which we aggregate to obtain flows for final goods—consumption or investment—separately from those used as intermediate inputs. Based on these data, we construct domestic and bilateral expenditure share—ωQ for all i and h—for final and intermediate goods, ih,t Q ∈ {C,M}, which we then use to compute bilateral trade costs according to (8) for a given value of the elasticity ηQ.20 For our baseline trade elasticities (see also section 4.5), we assume that elasticities do not vary across types of goods, η ≡ ηC = ηM, and that η = 5 implying a trade elasticity equal to four, η − 1 = 4. This value is in line with the estimates in the trade literature that are obtained by relying on static Gravity models (Head and Ries, 2001; Simonovska and Waugh, 2014; Caliendo and Parro, 2014), and closer to the higher-end estimates of Boehm et al. (2023) for the long-run trade elasticity. We consider this a reasonable value for the elasticity because it is consistent with valuesintheliteratureestimatedusingvariationintradeflowsacrosscountries—thesamevariation that we use to identify bilateral trade costs. Appendix C.3 explores the robustness of our main empirical results established in Section 3 to alternative values of η. In addition, we show in Section 5.1 that for our baseline value of η, the dynamic model that we propose replicates our empirical 18See details here: https://www.oecd.org/en/data/datasets/inter-country-input-output-tables.html. 19We choose the ICIO as our primary data source because it covers a longer time period than the WIOD, which only provides data until 2014. 20See Appendix A for further details on data cleaning and manipulation. 11

results remarkably well. 2.4 Stylized Facts We now proceed to describe the trends and features of our measure of bilateral trade costs. We establishthreefactsthathelpustoelucidatehowthesetrendsandfeaturesreflectchangesinpolicy and non-policy related trade barriers over time, across space, and across types of countries. We will rely precisely on variation in bilateral trade costs across time and space to identify the effects of trade cost shocks on inflation and other macroeconomic variables in Section 3. We start by exploring how these costs evolved over time. Fact 1: Bilateral trade costs declined significantly between 1995 and 2008 and stabilized thereafter, remaining unchanged until 2020. Figure 1 plots the evolution of the distribution of bilateral trade costs. The solid lines in Figure 1a plot the median costs for final (red line) and intermediate (blue line) goods in the cross-section of countries for any given year. These costs are expressed as a percentage of the sales price of the good—intermsofourdefinitionin(8), thefigureshows(TQ −1)×100forQ ∈ {C,M}. According ih,t to our estimates shown in panel (a), median bilateral trade costs across countries for both final and intermediate goods fell approximately 80 percentage points from 1995 to 2008. These changes are also reflected by the black lines in Figure 2a, which show changes in trade costs relative to their 1995 levels. This is a significant decline given that in 1995 these costs were approximately 380 and 420 percent for final and intermediate goods, respectively.21 The solid lines in Figure 1a and the black lines in Figure 2a also show that, since 2008, median trade costs have remained pretty much unchanged. Hence, the long-run evolution of global trade costs is consistent with a long period of globalization that stalled after 2010.22 The evolution of the distribution of bilateral trade costs constructed using WIOD data is very similar to the one shown in Figure 1a (Appendix B, Figure A.1). Oursecondfactisabouthowvariationinthesecostsacrossspacehasevolvedovertime. Hence, we focus on the evolution of the cross-sectional dispersion in bilateral trade costs. 21Intermsofmagnitudes,ourestimatesofbilateraltradecostsareinlinewithpreviousliteratureshowingthatthese are large (Anderson and van Wincoop, 2004). It is worth emphasizing that these magnitudes are not unreasonable given that our measure of trade costs includes all frictions that impede trade across any two countries. 22A steady process of globalization can be traced back to 1965 using historical WIOD data (Appendix B). 12

Figure 1: Evolution of Global Bilateral Trade Costs (a) Evolution of distribution (all 41 countries) (b) Evolution of medians by country group Note: Trade costs are expressed as a percentage of the sales price of the good. That is, the figures show the evolution of (TQ −1)×100 for Q ∈ {C,M}. All C denotes all countries, AE-AE refers to trade between ih,t two Advanced Economies, EE-EE refers to trade between two Emerging Economies, and AE-EE refers to trade between two economies of different types. Fact 2: The dispersion in bilateral trade costs across country pairs remained stable from 1995 to 2020. Figure 1a plots the evolution of the dispersion of trade costs over time. The charts plot the 20- 80 percentile bands for final consumption (red shaded area) and intermediate inputs (blue shaded area). Variation in bilateral trade costs is substantial at any given point in time and it remains sizable over the years. For instance, the ratio of 80th-percentile trade costs to those in the 20th percentilewasabout2.5forbothfinalandintermediategoodsin1995. Thisratiobarelydeclinedto 2.3 by 2019 before experiencing a small increase in 2020. Hence, there’s substantial and persistent variation in trade costs across space. We can split our sample of countries into Advanced Economies (AE) and Emerging Market Economies (EE) to explore how bilateral trade costs across different country-types have evolved over time. We classify trade flows between any two countries as either between AEs, between EEs, or between different types of countries (AE/EE of EE/AE). The next fact establishes the trends in trade costs for these three types of trade flows. 13

Fact 3: From 1995 to 2010, bilateral trade costs for trade flows between pairs of EEs declined considerably more than those for trade between pairs of AEs or AE-EE/EE-AE pairs. Since 2010, bilateral trade costs for any type of country pair remained relatively unchanged. Moreover, trade costs across AE-EE/EE-AE pairs for intermediate goods decreased by more than those for final goods. However, this has not been the case for the other types of country pairs. Figures 1b plots the evolution of median bilateral trade costs for specific types of country pairs. Trade between EEs faces the highest costs. However, in terms of changes since 1995, shown in Figure 2a, median costs for this type of country pair also experienced the largest decline over the 1995-2020 period. These trends align with the fact that trade between EEs has grown the most over the last 25 years (Reyes-Heroles et al., 2020). Interestingly, trade costs across AEs remained relatively stable, reflecting the fact that these countries were already substantially integrated prior to 1995. Figure 2a also shows that integration between EEs has occurred more rapidly for final than intermediate goods, but this has not been the case for the other types of country pairs. Figure 2: Changes in Trade Costs (relative to 1995) (a) Changes in median trade costs (b) Changes in bilateral trade costs for by country group selected U.S. trading partners Note: Trade costs are expressed as a percentage of the sales price of the good. That is, the figure shows the evolution of (TQ −1)×100 for Q ∈ {C,M}. All C denotes all countries, AE-AE refers to trade between ih,t two Advanced Economies, EE-EE refers to trade between two Advanced Economies, and AE-EE refers to trade between two economies of different types. Ourmeasureoftradecostsdoesnotonlycapturelong-runtrendsinglobaltradeintegration—as established by Facts 1 through 3—but it can also capture changes in trade policies. Let us consider 14

the case of bilateral trade costs for the U.S. shown in Figure 2b. The figure plots the changes since 1995 in U.S. bilateral trade costs with its three main trading partners—Canada (blue lines), Mexico (green lines), and China (red lines). Focusing first on the case of Canada and Mexico, note that bilateral trade costs with the U.S. declined substantially between 1995 and 2000, in line with the integration process sparked by the North American Free Trade Agreement (NAFTA) signed in 1994. Turning to the case of China, note that these costs experienced large declines from 1995 until around 2007. These declines are in line with the literature on the emergence of China in international trade markets, particularly as an exporter to the U.S. (Autor et al., 2013). During the 2018–19 period, U.S.–China trade costs for intermediate and final goods increased by 20 and 11 percentage points, respectively. Our estimated changes in bilateral trade costs are in line with the 16 percentage point increase in the weighted average tariff imposed by the U.S. on China and China’s partial retaliation, as well as the fact that these increases were tilted toward intermediate goods (Bown, 2021). Hence, bilateral trade costs also capture recent changes in trade policies. In summary, bilateral trade costs capture changes over time in policy and non-policy related trade barriers that vary across countries. Moreover, a sizable share of the variation in these costs is not driven by global factors—such as shipping technologies—nor by time-invariant countrypair-specific factors. A regression of bilateral trade costs on country-pair and time-fixed effects, TQ = γQ+γQ +εQ , implies that around 20 percent of variation in trade costs is not explained ih,t t ih ih,t by the fixed effects γQ or γQ. Our aim is to exploit this variation in trade costs across time and t ih space to identify their causal effect on inflation. As it will be made clear in Section 3, our estimates will control for these kinds of factors. 2.5 Import Costs The methodology we follow to estimate the effects of changes in trade costs on inflation is based on panel data local projections (Jord`a, 2005). Hence, we aggregate bilateral trade costs into countryspecific import costs for final and intermediate goods. These import costs will inherit the variation across time and space from a country’s bilateral trade costs that we documented in Section 2.4. More specifically, we construct country-specific import costs τQ for Q ∈ {C,M} by aggregating i,t 15

bilateral trade costs using import weights.23 More specifically, our import costs are given by N (cid:32) XQ (cid:33) τQ = (cid:88) ih,t TQ , (9) i,t (cid:80) XQ ih,t h=1 k̸=i ik,t for Q ∈ {C,M}. A key question is if our measure of import costs can systematically capture variation in directly measurable trade policies such as tariffs. This question is particularly relevant given the observed increases in tariffs around the globe since 2017. Hence, we compare our measure of import costs with data for effective tariffs. We collect data for tariff rates from 1996 to 2018 at the sectoral level provided by Reyes-Heroles et al. (2020) and aggregate these data into a single country-specific effective tariff rate using sectoral import shares. For the case of the U.S., we also compute an alternativemeasureofitseffectivetariffratefortheperiod1995-2020usingtheBureauofEconomic Analysis (BEA) data on customs duties and imports of goods. Figure 3 plots our measure of total U.S. import costs against the two measures of U.S. effective tariff rates. To obtain total import costs, we simply consider an import-weighted average of import costs for final and intermediate goods. The figure shows there is a clear and strong positive correlation between import costs and both measures of effective tariff rates. To explore this correlation for our sample of countries, we consider the following pooled regression logτ = γ +β ×log(1+tariff ), (10) i,t i i,t where τ denotes country i’s import costs at time t and tariff stands for its corresponding tariff. i,t i,t We estimate a value of β = 1.1 that is statistically significant at the 1 percent level and we cannot reject β being different from one. Hence, we conclude that our measure of import costs reflects variation in effective tariff rates. In the following section, we now turn to the causal effect of higher trade costs on inflation and other macroeconomic variables. In estimating these effects, our aim is to control for other forces that could be driving differences in inflation across countries. 23Wecouldconsideralternativeweightingschemesincludingtime-invariantuniformweightssimilartotheuniform weights in Frankel and Romer (1999). 16

Figure 3: Trade Costs and Effective Tariffs in the U.S. Note: U.S. trade costs in years 1995-2020 (y-axis) against effective U.S. import tariff rate from NIPA (blue dots and line) and effective tariff rates from Reyes-Heroles et al. (2020) aggregated using sectoral import weights for the period 1996-2018. 3 Estimating the Effect of Trade Costs on Inflation 3.1 Estimation Strategy To estimate the causal effects of changes in import costs on inflation, we rely on panel data local projections (Jord`a, 2005) and estimate the following panel specification: y = δQ +δQ +βQ·∆τQ +γQ·∆τ̸Q +Γ′QZ +εQ for h ≥ 1, (11) i,t+h i,h t,h h i,t h i,t h i,t−1 i,t+h where y is the dependent variable of interest for country i in period t+h. Our main coefficient i,t+h of interest in equation (11) is βQ, which captures the average h-period ahead response of y h i,t+h following an initial one-period change in trade costs relative to other countries. We focus on the dynamic effects of trade costs on CPI inflation, π , in which case the dependent variable is t y = π . We also study the impact of higher trade costs on the level of GDP, in which case i,t+h i,t+h we define y = log(GDP ) − log(GDP ). Appendix C presents the empirical impulsei,t+h i,t+h i,t−1 responsefunctionsforothermacroeconomicoutcomesincludingexports,imports,therealexchange rate, and the trade balance. Our estimation strategy allows us to identify the relative effects of an increase in country i’s import costs relative to another country’s costs. In that sense, our estimates only capture the macroeconomic effects of increases in import costs under particular assumptions because time-fixed 17

effects in (11) absorb the global effects of trade costs shocks. However, the general equilibrium model that we develop and calibrate in Section 4 allows us to circumvent this issue and provide model-based estimates for the macroeconomic effects of increases in trade costs. Moreover, the multi-country nature of the model permits us to test how the predictions of the model compare to our estimates of βQ obtained from (11). h To isolate the dynamic impact of trade costs, we control for unobserved sources of variation using country- (δ ) and time-fixed effects (δ ). Given that our measure of trade costs is expressed i t as a percent of the final sale price, the coefficient βQ measures the effect of a one percentage h point increase in trade costs. We scale the response coefficients such that the total import costs of final and intermediate goods increase by 10 percentage points. We report estimates for h = 1,...,5 years.24 Wealsocontrolforotherfactorsthatmaysimultaneouslyaffectinflation, likethepotential correlation between trade costs across good types. To account for this correlation, we include ∆τ̸Q i,t in the regression. For example, when Q = C and y = π , our estimate of βC recovers the i,t i,t h effects on inflation of a one-time increase in trade costs on final goods, holding changes in trade costs for intermediate inputs, (cid:19)Q(cid:19) = M, constant. The vector Z i,t−1 controls for other observable characteristics of country i, including lagged values of trade cost changes, ∆τQ ,∆τ̸Q , the i,t−1 i,t−1 lagged value of the dependent variable, y , lagged unemployment rate, and lagged GDP growth. i,t−1 To account for episodes that may have led to macroeconomic turmoil, like inflation surge or a GDP collapse, but are unrelated to changes in trade costs, we include country-year dummy observations from the Global Crises Database.25 3.2 The Dynamic Effect of Trade Cost Shocks Inflationary Response. Figure 4 shows that an increase in trade costs in final goods or intermediate inputs leads to a contemporaneous rise in CPI inflation. Panel (a) shows that a 10 percentage point increase in a country’s trade costs of final goods relative to the trade costs of its trading partners leads to a 0.65 percentage point increase in CPI inflation within the first year. Panel (b) shows that an equally sized rise in trade costs of intermediate inputs leads to a 0.58 percentage point increase in CPI inflation on impact. The shaded areas represent the 70 percent confidence intervals and show that the contemporaneous effects are statistically different from zero. 24We estimate Equation 11 following Correia (2016) with heteroskedasticity-robust standard errors. 25We control for years in which yearly inflation was above 50 percent, years explicitly recorded as an inflation crises and periods classified as a currency crises. See https://www.hbs.edu/ behavioral-finance-and-financial-stability/data/Pages/global.aspx 18

We highlight two features of the inflation response to higher trade costs. First, the persistence of the inflationary effects differs depending on the type of trade cost shock. Higher trade costs of final goods—say, due to tariffs imposed on goods like washing machines—lead to short-lived effects on CPI inflation, which dissipate after one year. In contrast, higher trade costs of intermediate inputs—say, due to a shortage in semiconductors or tariffs imposed on imported Chinese battery cells—have more persistent effects on CPI inflation that take several years to peter out. This novel empirical result has important implications for policymakers in charge of monetary policy decisions as we explore in section 5.3.26 Second, the impact response of CPI inflation varies depending on the type of trade cost shock. Anincreaseinimportcostsforfinalgoodsislargerrelativetotheinflationaryeffectofanincreasein the costs of intermediate inputs. This difference is neither economically nor statistically significant for a horizon of h = 1 as shown in Figure 4. However, the estimates for a horizon of h = 0 shown in columns (2) of the first panel of Table A.2 in Appendix C show that this difference is clearly significant in economic terms, with βC = 1.28 and βM = 0.75.27 This difference is in line with 0 0 the intuition that higher costs of imported consumption goods affect the CPI directly. Meanwhile, higher import costs of intermediate inputs indirectly affect the CPI through production costs. GDP Response. Figure 5 shows our estimates for the effects of higher trade costs on GDP. In this case, the coefficient βQ measures the cumulative percent response of GDP in period t+h h following an increase in trade costs in period t. Panel (a) shows that higher trade costs of final goods lead to a modest GDP decline of less than 0.5 percent in years 1-3, but this response is not statistically different from zero. Panel (b) shows that the hit to GDP is more pronounced following an increase in the trade cost of intermediate inputs. Given our main empirical results, we develop a dynamic model in section 4 to compare its predictions to our empirical estimates. Our calibrated model replicates the empirical estimates in this section.28 We use the model to highlight the transmission mechanism of trade cost shocks to the macroeconomy. We also show that the response of firms’ marginal costs plays a crucial role in shaping the inflationary response. 26The difference in persistence is not driven by differences in the persistence of the shocks as shown in Figure A.5 of Appendix C. 27We do not show the estimates for h = 0 in Figure 4 because of time-aggregation issues that arise with annual data and that make it difficult to interpret the estimates for h=0. 28Model-based impulse-response functions are also in line with our empirical impulse-responses presented in Appendix C for exports, imports, the real exchange rate, and the trade balance. 19

Figure 4: Response of Inflation to a 10 Percentage Point Increase in Trade Costs (a) ResponsetoTradeCostsinFinalGoods (b) ResponsetoTradeCostsinIntermediateGoods Note: The figure shows the consumer price index (CPI) response to a 10 percentage point increase in trade costs. Solid lines show estimated βQ coefficient. Shaded areas show 70 percent confidence intervals. h Figure 5: Cumulative Response of GDP to a 10 Percentage Point Increase in Trade Costs (a) ResponsetoTradeCostsinFinalGoods (b) ResponsetoTradeCostsinIntermediateGoods Note: The figure shows the cumulative response of log(GDP) to a 10 percentage point increase in trade costs. Solid lines show estimated βQ coefficient. Shaded areas show 70 percent confidence intervals. h 4 The Model We develop a dynamic model to understand key mechanisms through which changes in trade costs can affect inflation and macroeconomic outcomes. We build on the New Keynesian literature and extend a multi-country New Keynesian model 20

with nominal price and wage rigidities to also feature trade in final consumption goods and intermediate inputs. Our New Keynesian bloc is similar to canonical open economy models (see Corsetti et al. (2010) for a review). The trade bloc of our model—as described in section 2.2— deviates from standard open economy New Keynesian models by allowing for trade in different types of goods. We will consider not only trade in final goods but also trade in intermediates. A central aspect of this bloc of the model is that it delivers gravity-type bilateral trade equations, which align with how we measure trade costs in section 2. We embed the static trade model described in Section 2.2 into a dynamic framework. To recap, time is discrete and indexed by t = 1,2,..., and the world is comprised of N countries indexed by i,h ∈ I = {1,...,N}. We assume that each of these countries has population ξ , for i = 1,...,N, i and we normalize world populationto unity inevery period. We assume country1 to be the United States. In addition to trading final consumption goods and intermediate inputs—as described in section2.2—countriesalsotradeinfinancialassetsunderincompleteinternationalfinancialmarkets. Moreprecisely,countriescanonlytradearisk-freeinternationalbonddenominatedin(real)dollars, country 1’s currency, across borders. Aside from the fact that country 1’s currency is the one used in international financial markets, countries are otherwise symmetric. We proceed now to describe the structure of a generic country i. 4.1 Households There is a continuum of households indexed by ℓ in each country. Within a country, households engage in monopolistic competition when supplying differentiated labor services to the production sector, as in Erceg et al. (2000). That is, goods-producing firms regard each household’s labor as an imperfect substitute for the labor services of other households. Hence, the objective function of a household ℓ in country i is given by ∞ (cid:88) (cid:104) (cid:105) E βt U (C )−V (Lℓ ) , (12) 0 i i,t i i,t t=0 where C is as specified in (1) for Q = C.29 As a reminder, C is a CES aggregate of C — i,t i,t ih,t country i household’s consumption of the good produced in country h—across source countries h = 1,...,N. In (12), Lℓ denotes labor services (hours) provided by household ℓ in country i. i,t 29As is standard in this class of models, complete financial markets within country i ensure that all households ℓ consume the same amount, so we omit the ℓ index in C . i,t 21

Household ℓ in country i seeks to maximize (12) subject to the budget constraint N (cid:88) τC P C +B + B i1,t ≤ Wℓ Lℓ +R B +R Ψ B i1,t−1 +T (13) ih,t ih,t ih,t ii,t E i,t i,t i,t−1 ii,t−1 1,t−1 i,t−1 E i,t 1i,t 1i,t h=1 for all t, where B denotes holdings of domestically-traded bonds for country i, B denotes ii,t i1,t holdingsofcountry1’sbonddenominatedinU.S.dollars,E denotescountryi’snominalexchange ih,t rate against country h—as defined in section 2.2—and T are transfers to households in country i,t i.30 We allow risk premia to vary across countries through the term Ψ in (13). These risk 1,t−1 premium terms are such that Ψ = 1 and Ψ ≥ 1 for i ̸= 1. More specifically, we assume that for 1,t i,t i ̸= 1, Ψ is given by i,t b Ψ ≡ (1−ψ i1,t )εψ , (14) i,t Q Y i,t 1i,t i,t where b ≡ Bi1,t denotes country i’s borrowing in units of country 1’s good, Y denotes total i1,t PC i,t 1,t tradable output in country i, εψ is an exogenous shock to the risk premium of country i, and Q i,t ih,t denotes the real exchange rate between country i and country h defined as E PC ih,t h,t Q ≡ , (15) ih,t PC i,t where PC is as defined in (4). i,t Note that in (13), we express prices paid for final consumption goods i, PC , explicitly in terms ih,t of the trade costs for final consumption goods, τC —as defined in section 2.2—and the bilateral ih,t prices that exclude the costs of shipping goods across borders, P . We do so to emphasize how ih,t changesinthesetradecostsdirectlyaffectthepricespaidbyfinalconsumersand,therefore,directly impact their behavior and welfare. We now impose additional structure on these trade costs and assume that they are comprised of exogenous iceberg trade costs, dC ≥ 1, and exogenous ad ih,t valorem tariffs, κC ≥ 0, such that total trade costs are given by τC = dC (1+κC ). ih,t ih,t ih,t ih,t Note that, conditional on C , equation (3) for Q = C determines demand for consumption i,t goods across country sources, where PC is given by (4) for Q = C. Hence, conditional on aggregate i,t expenditure and prices, all bilateral trade flows are fully determined by equation (6). Aggregate spending, in turn, is determined by the household’s intertemporal optimality condition for bonds 30Transfers to households include those from firms and the government. The specific transfers are provided in section 4.4. 22

denominated in domestic currency. The Euler equation characterizes this condition: (cid:20) U′(C ) (cid:21) U′(C ) = βR E i i,t+1 , (16) i i,t i,t t π i,t+1 PC where π ≡ i,t denotes CPI inflation. i,t PC i,t−1 For countries that trade bonds denominated in a currency other than their domestic currency, that is, for countries other than country 1, a household’s optimal portfolio choice is characterized by an uncovered interest parity (UIP) condition, which in real terms is given by   (cid:20) U′(C )/U′(C ) (cid:21) U′(C )/U′(C ) R i,t E t i i,t π + i 1 ,t+1 i i,t = R 1,t Ψ i,t E t i π 1 i , , t t + + 1 1 Q Q 1i 1 , i t i , + t 1 i,t . (17) Hence, given our assumption about Ψ in (14), we have that the uncovered interest rate parity i,t condition between countries i and 1 will not hold if bi1,t deviates from 1. Q1i,tYi,t ψ 4.1.1 Wage Setting Following Erceg et al. (2000), we assume that there is an ‘employment agency/union’ in each country that combines households’ labor services (hours) into an aggregate homogeneous labor input supplied to final producers, which we denote by L . The agency combines labor services i,t across households according to (cid:18)(cid:90) 1 ϵw−1 (cid:19) ϵw ϵw −1 L i,t = Lℓ i,t ϵw dℓ , (18) 0 where ϵ > 1. Given a profile of wages across household types, {Wℓ } , the agency seeks to w i,t ℓ minimize its hiring costs by demanding labor services of type ℓ household according to (cid:32) Wℓ (cid:33)−ϵw Lℓ = i,t L , (19) i,t W i,t i,t where W = (cid:18)(cid:90) 1 Wℓ 1−ϵwdℓ (cid:19) 1− 1 ϵw , (20) i,t i,t 0 and W can be interpreted as the aggregate wage index in country i. i,t 23

Householdsineachcountrycompeteinamonopolisticfashioninthelabormarketandtherefore can set wages to maximize their utility. However, we assume that in any given period t, household ℓ can only reset its nominal wage, Wℓ , with probability 1 − θ , and that with probability θ , i,t w w household ℓ’s nominal wage has to be the same as in the previous period. Consider a household that can reset its nominal wage in period t. This household will choose its optimal reset nominal wage, W , to maximize i,t ∞ (cid:88) (cid:16) (cid:17) E βkθk U (C )−V (Lℓ ) , (21) t w i i,t i i,t+k|t k=0 where (cid:18) W (cid:19)−ϵw Lℓ = i,t L (22) i,t+k|t W i,t+k i,t+k denotes labor demand in period t+k for a wage setter that last reset its wage in period t. The optimal reset nominal wage has to be such that the following first-order condition holds: E (cid:88) ∞ βkθkLℓ U′(C ) (cid:32) W i,t − ϵ w V i ′(Lℓ i,t+k|t ) (cid:33) = 0. (23) t w i,t+k|t i i,t+k PC ϵ −1 U′(C ) k=0 i,t+k w i i,t+k Notethatinthecaseinwhichworkerscanadjustwagesineveryperiod, θk = 0, thenweobtainthe w usualoptimalityconditionthatequatestherealwage,W /PC,tothemarginalrateofsubstitution, i,t i,t V′(Lℓ )/U′(C ), adjusted by the monopolistic distortion associated with a positive markup, i i,t+k|t i i,t+k ϵ /(ϵ −1). Since a measure θ of firms keep their price unchanged and 1−θ reset it optimally, w w w the aggregate wage index, W i,t , is such that W i 1 ,t −ϵw = θ w (W i,t−1 )1−ϵw +(1−θ w )(W i,t )1−ϵw. 4.2 Firms There are two types of firms in each country. For the first type, a unit continuum of firms indexed by v ∈ [0,1] produce differentiated goods that cannot be traded across borders. In country i, these goodsareproducedusingthehomogeneouslaborsuppliedincountryiandabundleofintermediate inputs. Thesecondclassoffirmsconsistsofperfectlycompetitiveidenticalretailfirmsthatproduce a final homogeneous good that can be traded internationally, subject to trade costs. These firms produce tradable goods by aggregating the differentiated goods produced by domestic firms. We first describe the technology and problem faced by the first type of firms and then proceed to 24

describe these issues for the second type. 4.2.1 Differentiated Firms: Nontradables Firm v ∈ [0,1] in country i produces nontradable goods according to the production function (cid:20) (cid:21) εy Y i v ,t = A i,t (1−ν)ε 1 yLv i,t εy ε − y 1 +νε 1 yM i v ,t εy ε − y 1 εy−1 , (24) where A denotes exogenous productivity that does not vary across firms, Lv is labor input, and i,t i,t Mv is the amount of an intermediate input bundle used in production.31 The intermediate input, i,t inturn, consistsofanaggregateofgoodsproducedinallcountriesaccordingtotheCESaggregator specified in (1) for Q = M. Note that, as long as ηM < ∞, intermediate inputs are not perfectly substitutable across countries. Given prices of country-specific tradable goods in global markets, {P }N , retail firms in i,t i=1 country i will seek to minimize the total cost of intermediate inputs, N (cid:88) τM P M , (25) ih,t ih,t ih,t h=1 subject to (1) for Q = M and for a given level of M . As for the case of the trade costs faced i,t by households, we assume that the trade costs faced by firms are comprised of exogenous iceberg trade costs, dM ≥ 1, and exogenous ad valorem tariffs, κM ≥ 0, such that total trade costs are ih,t ih,t τM = dM (1+κM ). The solution to the minimization problem yields demands for goods from ih,t ih,t ih,t different countries, M , according to (3), with an ideal price index for the intermediate input ih,t bundle, PM, as specified in (4) for Q = M. i,t Note that, even though households and firms import exactly the same good from any given country,theypaydifferentpricestotheextentthattradecostsdifferbetweenfinalandintermediate goods. Differentiated firms choose the amount of labor and the intermediate input bundle to minimize 31We will restrict attention to a first-order approximation of the model and ignore second-order price dispersion terms. Hence, in this section, we can treat the aggregate production function as being analogous to the individualproducer production function (the difference between the two arises from price dispersion and is therefore of second order). 25

total production costs given by W Lv +PMMv , (26) i,t i,t i,t i,t subject to (24). Given nominal wages and the price of the intermediate input bundle in country i, PM, the solution to the cost minimization problem delivers the marginal cost faced by retail firms, i,t which is the same across firms and given by MC = 1 (cid:104) (1−ν)W 1−εy +ν(PM) 1−εy (cid:105) 1− 1 εy . (27) i,t A i,t i,t i,t Note that changes in trade costs of intermediate inputs affect the nominal marginal cost of the firms directly through their effects on the price of the intermediate input bundle, PM. Therefore, i,t an increase in trade costs leading to a higher PM will increase firms’ marginal costs and decrease i,t production efficiency. Before analyzing the price-setting behavior by firms producing differentiated goods, we describe the technology and problem of the representative retailer producing the homogeneous tradable good. 4.2.2 Retail Firms: Tradables To produce tradable goods, the representative retail firm in country i aggregates differentiated goods available in i according to Y i,t = (cid:18)(cid:90) 1 Y i v ,t ϵ− ϵ 1 dv (cid:19) ϵ− ϵ 1 , (28) 0 where ϵ > 0. The output of the homogeneous good in country i can then be used for final consumption or as an intermediate input in the production of differentiated goods, either domestically or abroad. Given the prices of different varieties, final good producers maximize profits subject to (28). The solution to this maximization problem delivers the demand for variety v in country i: (cid:18)Pv (cid:19)−ϵ Yv = i,t Y , (29) i,t P i,t i,t 26

where (cid:20)(cid:90) 1 (cid:21) 1− 1 ϵ P = Pv 1−ϵdv (30) i,t i,t 0 defines the nominal price of a unit of the homogeneous good produced in country i in terms of its own currency, and Pv denotes the nominal price charged by firm v in country i, also in terms of i,t local currency units. As stated in section 2.2, excluding trade costs, the price of country i’s imports from any country h is given by the price country h producers set domestically, adjusted for the exchange rate between the two countries. Accordingly, P = E P . (31) ih,t ih,t h,t 4.2.3 Price Setting by Differentiated Firms Differentiated firms, in country i, set prices to sell their goods to retailers.32 However, in line with our assumption of staggered wage adjustments, we assume that firm v can only reset its price in period t with probability 1−θ, and with probability θ it must keep its price unchanged relative to last year’s. Let P denote the optimal price for a firm that is able to reset its price in period t. Such a i,t firm in country i will set this price to maximize E (cid:88) ∞ βkθk U i ′ ,t (C i,t ) (cid:0) P −MC (cid:1) (cid:18) P i,t (cid:19)−ϵ Y , (32) t PC i,t i,t+k P i,t+k k=0 i,t+k i,t+k where the marginal cost is specified in (27). The optimal price set by those firms able to adjust prices must be such that the following optimality condition holds: E (cid:88) ∞ βkθk U i ′ ,t (C i,t ) Pϵ Y (cid:20) P − ϵ MC (cid:21) = 0. (33) t PC i,t+k i,t+k i,t ϵ−1 i,t+k k=0 i,t+j Note that under flexible prices (θ = 0), this condition reduces to the usual pricing condition, setting prices equal to a markup over marginal cost. Since a measure θ of firms keep their price unchanged and 1−θ reset prices optimally, from (30) we obtain that P satisfies the law of motion i,t 32Note that in this model, firms engaging in international trade are perfectly competitive. Therefore, firms do not set prices in international markets. However, our assumptions are equivalent to allowing for price setting in internationalmarketsunderproducercurrencypricing(PCP).Thatis,firmssetpricesinthecurrencyofthecountry in which they produce and let their prices in the foreign currency adjust with the exchange rate. 27

P1−ϵ = θP1−ϵ +(1−θ)P 1−ϵ . i,t i,t−1 i,t 4.3 Monetary policy We assume that central banks in all countries follow conventional Taylor-type monetary policy rules. More specifically, the central bank in country i sets the nominal interest rate according to the inertial policy rule given by  (cid:32) (cid:33)ϕy 1−ϕr 1 GDP R i,t = (R i,t−1 )ϕr  β (π i,t )ϕπ GDPf i l , e t x εr i,t , (34) i,t whereGDP denotesrealvalueaddedincountryi,andGDPflex correspondstothesameobjectin i,t i,t the potential economy with flexible prices.33 In (34), ϕr > 0 determines the inertia in the monetary policy response, ϕπ > 0 and ϕy > 0 parameterize the elasticities of the policy rate with respect to changes in inflation and GDP deviations from the central banks zero inflation target and the natural level of GDP, respectively; εr is an exogenous shock to the monetary policy rule. i,t 4.4 Market clearing and balance of payments Tradable goods produced in country i are sold either domestically or abroad to country i’s trading partners. Domesticorforeignbuyersofthesegoodsthenconsumethemorusethemasintermediate inputs. Hence, the market-clearing conditions for these goods are given by N (cid:88) ξ Y = ξ (dC C +dM M ) (35) i i,t h hi,t hi,t hi,t hi,t h=1 for i = 1,...N, where the population terms ξ reflect the fact that all variables are expressed in i per-capita terms. Note that (35) accounts for the goods that are lost when traded across countries because of the iceberg-type trade costs dQ ≥ 1 for Q ∈ {C,M}. hi,t We can derive the balance of payments condition for every country i, other than country 1 (i ̸= 1), that determines the evolution of their holdings of dollar-denominated bonds. To do so, we 33Real value added in country i is given by GDP = Pi,tY − Pi M ,tM . i,t PC i,t PC i,t i,t i,t 28

aggregate domestic budget constraints and obtain N (cid:88) τC P C + B i1,t = W L +R Ψ B i1,t−1 +T , (36) ih,t ih,t ih,t E i,t i,t 1,t−1 i,t−1 E i,t 1i,t 1i,t h=1 where transfers to households in country i include tariff revenues that are rebated lump-sum to households, N (cid:32) τC P C τM P M (cid:33) K = (cid:88) κC ih,t ih,t ih,t +κM ih,t ih,t ih,t , (37) i,t ih,t 1+κC ih,t 1+κM h=1 ih,t ih,t and firms profits, Π , such that T = K +Π . Given that profits are given by Π = P Y − i,t i,t i,t i,t i,t i,t i,t W L − (cid:80)N τM P M , (36) can be rewritten as i,t i,t h=1 ih,t ih,t ih,t N N (cid:88) dC P C + B i1,t = P Y − (cid:88) dM P M +R Ψ B i1,t−1 , (38) ih,t ih,t ih,t E i,t i,t ih,t ih,t ih,t 1,t−1 i,t−1 E 1i,t 1i,t h=1 h=1 which does not depend on tariffs because we have imposed that all tariff revenues are rebated lump-sum to households, which is equivalent to the government running a balanced budget. Hence, we obtain that the evolution of holdings of U.S. bonds for country i—prescribed by the balance-ofpayments condition—is given by B −B = E (EX −IM )+(R Ψ −1)B , (39) i1,t i1,t−1 1i,t i,t i,t 1,t−1 i,t−1 i1,t−1 where EX = P Y −P C −P M and IM = (cid:80) dC P C + (cid:80) dM P M , i,t i,t i,t i,t ii,t ii,t ii,t i,t h̸=i ih,t ih,t ih,t h̸=i ih,t ih,t ih,t denote exports and imports. Condition (39) simply states that country i’s current account is equal to its trade balance plus its net foreign investment income. 4.5 Calibration Our baseline calibration considers five country blocks, N = 5, with countries 1 through 5 representing the United States, China, the advanced non-U.S. economies, the Asian emerging market economies, and the rest of the emerging market economies, respectively. Our calibration strategyrestrictsheterogeneityacrosscountriesinsteadystatetotwosetsofparameters: countries’ sizes, ξ , and degrees of openness determined by countries’ levels of bilateral trade costs in steady i state as discussed in this section. All other values of model parameters will be the same across 29

countries. By following this strategy, we aim to elucidate the common mechanisms across countries that shape the effects of trade costs shocks on inflation and other macroeconomic outcomes. For households’ preferences, we allow for habit formation in consumption. Thus, we consider a more general case than in the model presented in section 4, and replace U (C ) with i i,t (C −hC )1−σ −1 i,t i,t−1 U (C ,C ) = , i i,t i,t−1 1−σ where h ≥ 0 modulates the degree of habit formation and σ > 0 is the inverse of the intertemporal elasticity of substitution (IES). For the disutility of labor, we choose the functional form Lℓ 1+φ V (Lℓ ) = i,t , i i,t 1+φ where φ > 0 is the inverse of the labor supply elasticity. The population parameters, ξ , are set to replicate the weights of these five regions in world i GDP. For our numerical experiments, we consider a log-linear approximation of the model around its steady state under balanced trade. Hence, we assume that trade costs can be expressed as (cid:16) (cid:17)1−ηQ τQ = τ¯QεQ , where τ¯Q are time-invariant parameters such that (cid:80)N τ¯Q = 1 for all i, ih,t ih ih,t ih h=1 ih and εQ are stationary shocks to trade costs for Q ∈ {C,M}.34 Note that, by equation (6), we ih,t have that in steady state (τ¯Q)1−ηQ = ωQ, which implies we can recover the steady state level of ih ih the trade costs from observed trade shares.35 Given our assumption that trade is balanced in the steady state, we can only use half of the expenditure shares from the data to pin down the steady state trade costs, with the rest being determined by the restriction that trade must be balanced.36 We choose the model’s parameters by relyingonpreviousliteratureortargetinglong-runmomentsinthedata. Table1liststheparameter values for our calibration. Starting with our choice for preference parameters, our values for the 34Notethatthenormalization (cid:80)N (cid:16) τ¯Q (cid:17)1−ηQ =1foreachcountryidiffersfromthenormalizationτQ =1that h=1 ih ii,t weassumedwhenconstructingbilateraltradecostsfortheempiricalanalysis. Thenormalizationchoiceisinnocuous for the predictions of our model. 35Intermsoftheequilibriumconditionsofthemodel,thisassumptionisequivalenttoassumingthatequation(1) is given by (cid:32) (cid:33) ηQ Q i,t = (cid:88) (ω i Q h )η 1 Q(Q ih,t ) ηQ ηQ −1 ηQ−1 , h and τQ =εQ for Q∈{C,M}. ih,t ih,t 36The ω’s can be chosen recursively such that a steady state with balanced trade and unity prices exists. 30

Table 1: Calibrated Parameters Parameter Description Value Preferences β Discount factor 0.99 σ Inverse IES 0.5 h Habit 0.75 φ Inverse labor supply elasticity 2 ηC Elasticity of substitution for final consumption 5 ϵ Elasticity of substitution across retailers 6 ϵ Elasticity of substitution across labor varieties 6 w Technology ν Share of intermediates in production 0.4 ε Elasticity of substitution labor-intermediates 0.5 y ηm Elasticity of substitution for intermediate inputs 5 Prices and wages θ Price rigidity 0.8 θ Wage rigidity 0.8 w Monetary policy ϕ Taylor rule inflation coefficient 1.5 π ϕ Taylor rule output coefficient 0.2 y ϕ Taylor rule inertia 0.75 r ψ Risk premium elasticity to NFA 0.001 Trade costs ρ Trade cost shock autocorrelation 0.95 τ Global Value Chain [ωC,ωC,ωC,ωC] Consumption trade weights, country 1 [.94,.012,.004,.021] 11 12 13 14 [ωM,ωM,ωM,ωM] Intermediates trade weights, country 1 [.88,.025,.007,.04] 11 12 13 14 [ωC,ωC,ωC] Consumption trade weights, country 2 [.95,.009,.02] 22 23 24 [ωM,ωM,ωM] Intermediates trade weights, country 2 [.94,.01,.014] 22 23 24 [ωC,ωC] Consumption trade weights, country 3 [.94,.014] 33 34 [ωM,ωM] Intermediates trade weights, country 3 [.81,.045] 33 34 ωC Consumption trade weights, country 4 .94 44 ωM Intermediates trade weights, country 4 .89 44 Country Sizes [ξ ,ξ ,ξ ,ξ ,ξ ] Region populations [.20,.19,.19,.27,.14] 1 2 3 4 5 discount factor, β, the inverse of the IES, σ, habit formation, h, and the inverse of the labor supply elasticity, φ, are all standard in the literature on open economy macro models (Bodenstein et al., 2023). For the trade elasticity for final goods, ηC, and for the trade elasticity for intermediate, ηM inputs, we choose a value equal to 5, in line with the long-run trade elasticity we use to recover trade costs in Section 2.37 37Model predictions are little changed with the trade elasticity in the range of estimates in Boehm et al. (2023) 31

For the technological parameters, our choices are also standard and similar to those in the literature. Regarding the New Keynesian bloc of our model, the values for the elasticities of substitutionacrossretailersandlaborvarieties(ϵ, ϵ ), andforrigiditiesofpricesandwages(θ, θ ), w w come from (Bodenstein et al., 2023). To calibrate the technologies of differentiated firms, we follow Comin and Johnson (2020) and measure the share of value added to consumption from the U.S. National Income and Product Accounts to set the steady-state value of ν. We assume that labor and intermediate inputs are complementary (ϵ < 1), and the trade elasticity for intermediates is y equal to that of final goods. For the policy parameters, we assume standard values for the response of the policy rate to deviations of inflation from the central bank’s target ϕ > 1, the deviations π of output from its natural level ϕ < 1, and the persistence parameter 0 < ϕ < 1. We allow for a y r small amount of risk premia in steady-state ψ > 0, which implies that exogenous deviations from UIP have negligible macroeconomic implications. For the parameters determining the trade shares, we set four parameters for the U.S., three for China, and so on, and let the trade balance condition in steady state determine the rest. We choose the values of these parameters based on data on global input-output linkages from the ICIO database. These values are shown at the bottom of Table 1. 5 The Transmission of Trade Cost Shocks To highlight the model’s predictions about the effects of trade disruptions, we first examine the impact of increasing trade costs in the model, mimicking the empirical analysis in Section 3. We explore the transmission mechanism of trade costs focusing on the U.S. and discuss the role of the monetary policy response in shaping the response to tariffs and iceberg trade costs. 5.1 Model vs Data In Section 3, we showed that higher trade costs lead to increased inflation and lower GDP. Here we compare our model to the data. Because our empirical specification is informative about the relative effects of higher trade costs across countries, it is not enough to compute a generic impulse response from the model. Instead, we exploit the multi-country nature of our model to construct relative impulse responses across regions to mimic Equation 11. We use the model to simulate the average country in our dataset in terms of trade openness. We use region 1 in our model as a reference country and simulate a 10 percentage point increase in 32

bilateral trade costs with all its trading partners. Relative to the calibration of Table 1, we adjust the trade shares to capture greater openness of the average country in the data relative to the U.S.38 We collect the simulated path for the variables of interest in deviations from steady-state, yˆ , forj = 1,...,5 corresponding tothe five regions inthe model andh = 1,...,20, because our j,t+h (cid:80)N (yˆ −yˆ ) model’s frequency is quarterly. The estimated effect of interest is given by j=2 1,t+h j,t+h . Our N−1 calibration of trade shocks implies that trade-weighted U.S. import costs rise by twelve percentage points, while import costs of the other regions rise by about two percentage points on average. We repeat this simulation separately for trade costs affecting final goods and intermediate inputs.39 This procedure ensures we construct a model estimate comparable to our empirical results. Figure 6 shows the dynamic response of inflation (top panels) and GDP (bottom panels) in the model and in the data. The green-circled lines correspond to our model’s simulations. The vertical whiskersdepicttheempiricalestimates. Themodel’sresponsesliewithinthe70%confidencebands from our empirical results for nearly all horizons. One exception is that the inflation response following an increase in intermediate trade costs is somewhat less persistent in the model than in the data. This is partly due to substantial heterogeneity in inflation dynamics worldwide, which our model doesn’t capture. Also, our model’s reference country has a relatively low share of intermediates in production (ν = 0.4) under our calibration. The share of intermediate inputs in production varies widely across countries and sectors (Reyes-Heroles et al., 2020). For instance, in manufacturing-intensive countries like China, the share of intermediates of production ν = 0.6 amplifiestheinflationaryresponsetotradeshocksonintermediateinputs. Inflationpersistencealso differssubstantiallyaroundtheworld. IntheU.S.,pre-pandemicinflationdynamicswereconsistent with the view that households and firms put negligible weight on past inflation developments, whereas in other advanced economies inflation tends to be more persistent, perhaps because of greater price and wage indexation (de Michelis et al., 2024). We discuss the robustness of the model fit to various parameters in the Appendix D. 5.2 Unpacking the Mechanism We now explore the transmission mechanism of higher trade costs and use the model to provide quantitative statements about the macroeconomic effects on a particular country. Hence, this 38Wesetthedomesticsourcingshareofcountry1toω =0.91,andωM =0.73,andadjustthetradeshareswith 11 11 other blocks accordingly to obtain an import-to-GDP ratio of 34%, while in the data the average import-to-GDP ratio across countries is 38.5% over the period 1995-2020. 39Alternatively we could estimate Equation 11 using model simulated data. However, our calibration imposes limited heterogeneity across countries for this approach to generate sufficient variation. 33

Figure 6: Model vs Data 1.5 1.5 1 1 0.5 0 0.5 -0.5 0 -1 -0.25 0 4 8 12 16 20 0 4 8 12 16 20 1.2 1.2 0 0 -1 -1 -2 -2 0 4 8 12 16 20 0 4 8 12 16 20 Note: Empirical and model-based estimates of the effects of a 10 percentage point increase in bilateral trade costs of the average country with all its trading partners. Trade shares in the simulation capture the average country’s openness to trade in final goods and intermediate inputs. Vertical whiskers correspond to empirical estimates reported in Section 3. experiment incorporates the global effects that were absorbed by the time-fixed effects when we estimated the panel local projections in (11) of Section 2. We focus on the United States under the calibration in Table 1 and consider a bilateral 10 percentage point increase in trade costs against all trading partners. We assume a symmetric increase in the trade costs foreigners pay for U.S. imports, in line with how we construct bilateral costs in our empirical analysis. Figure 7 shows the dynamic effects of trade cost shocks to final consumption goods (blue circled line) and trade cost shocks to intermediate inputs (red solid line). The key observation is that when trade costs increase for consumption goods, inflation rises by about 0.5 percentage points in the first year, which is close to our empirical estimates for relative effects. However, the effects are short-lived, with 4-quarter inflation falling slightly below the steady state after four quarters. Thus, the effects are akin to a one-time increase in the price level that materializes upon impact of the shock and slowly reverts as the shock itself is unwound. In contrast, when the trade cost shock affectsintermediateinputs,inflationinitiallyrises0.3percentagepoints,buttheeffectismuchmore persistent. While the higher persistence is in line with our empirical estimates for relative effects, our macroeconomic estimate of the effects on CPI inflation after one year (0.3) is approximately half of our relative estimate (0.58), implying that the time-fixed effects δM absorbed a significant t,h part of the macroeconomic effects of an increase in import costs of intermediate inputs. The more persistent inflation response arises because higher import costs raise domestic 34

Figure 7: Effects on the U.S. of a 10 p.p. increase in trade costs 0.8 0 0.75 0.6 0.5 -0.5 0.4 0.25 0.2 -1 0 0 -0.2 -1.5 -0.25 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 0.7 0.4 0.5 0.3 0.2 0 0 0 -0.5 appreciation -0.3 -0.7 -0.2 -1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 0 0 0.4 -10 -10 0.2 0 -20 -20 -0.2 -30 -30 -0.4 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 Note: Effectsofa10percentagepointincreaseintheU.S.’stradecostsfromalltradingpartnersonfinalconsumption goods (blue circled line) and intermediate inputs (red solid line). producers’ marginal costs—not only directly, through pricier imported inputs, but also indirectly via “second-round” effects in the input-output network. With higher input costs, more expensive domestic goods propagate through the global value chain. In each market, the effect mirrors an exogenous decline in aggregate total factor productivity or cost-push shock (Auclert et al., 2025; Werning et al., 2025). Although firms can substitute imported intermediate inputs with domestic alternatives, such as labor, these are imperfect substitutes, generating inefficiencies. As a result, real marginal cost increases. The persistent rise in real marginal cost leads to sustained inflation, which is amplified by sluggish price adjustments. In contrast, increased trade costs for consumption goods lead to lower real wages and reduced marginal cost. GDP falls in both scenarios due to tighter monetary policy and weaker external demand, but the contraction in GDP is more pronounced in response to trade costs affecting 35

intermediate inputs. Imports and exports contract sharply in both cases, with the trade balance deteriorating slightly in the initial periods, partly due to a mild exchange rate appreciation.40 To summarize, and in line with our empirical results, increases in trade costs lead to a contraction in GDP. Moreover, higher trade costs of intermediate inputs generate more persistent inflation relative to the inflationary effects of trade costs on final goods. This result has notable policy implications, as shocks leading to more persistent inflation may worsen the policy trade-off and create larger risks of partially de-anchoring longer-run inflation expectations. 5.3 Trade Costs and Monetary Rules In this section, we compare the effects of higher trade costs under our baseline monetary policy rule—which responds to CPI inflation, that is, inflation in the consumer price index inclusive of importedfinalconsumptiongoods—toanalternativerulethatrespondstoinflationofdomesticallyproduced goods only, which we label PPI rule (for producer price index). Such a rule is attractive because it is easy to implement and has been shown to have desirable stabilization properties in open economies.41 In the context of monetary policy, an often-cited appeal of this type of rule is that it “sees through” temporary increases in inflation driven by the imposition of tariffs.42 We showherethatthepresenceoftradecostsonintermediateinputscallsthatreasoningintoquestion. ThetoprowofFigure8comparesoutcomesunderourbaselineCPIrule(bluecircledlines)with those under the PPI rule (green dotted lines), when trade costs apply only to final consumption goods.43 Under the CPI rule, the monetary policy rate rises significantly, while under the PPI rule the policy rate barely moves. Thus, monetary policy effectively “sees through” the inflation increase due to trade costs, resulting in a noticeably smaller GDP decline in the short run. The second row performs the same experiment, now with trade costs on intermediate inputs. In stark contrasttothepreviouscase,theoutcomesarenowvirtuallyidenticalunderbothrules,withhigher trade costs leading to persistently higher inflation and persistently lower GDP. The reason is that higher trade costs on intermediate inputs lead to higher marginal costs for all domestic producers, which are eventually passed through as higher prices of domestic goods. Thus, in the presence of trade costs on intermediate inputs, a PPI rule will not effectively see through the increase in trade 40This appreciation is in line with our empirical estimates in Appendix C. 41See Corsetti et al. (2010) 42See, for example, the Risks and Uncertainty section of the September 2018 Tealbook published by the Federal Reserve Board. 43We assume here that the increase in trade costs is permanent, which resembles the imposition of a permanent tariff by the home country. See Appendix E for the case when trade costs are transitory. 36

Figure 8: CPI vs. PPI Targeting with Permanent Increase in Trade Costs 0.8 0.5 0.75 0.6 0 0.5 0.4 -0.5 0.2 0.25 -1 0 0 -0.2 -1.5 -0.1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 0.8 0.5 0.75 0.6 0 0.5 0.4 -0.5 0.2 0.25 -1 0 0 -0.2 -1.5 -0.1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 Note: Effects of a permanent 10 percentage point increase in the U.S.’s trade costs from all trading partners under alternative monetary policy rules. costs. Our analysis of the monetary policy response to trade costs relates to recent work focusing on the effect of tariffs (Bergin and Corsetti, 2023; Bianchi and Coulibaly, 2025; Werning et al., 2025). Tariffs can be seen as policy-induced trade costs that, in addition to generating inefficiencies, raise government revenue. Hence, as discussed in the previous section, higher tariffs should lower GDP net of tariff revenues and increase inflation. One additional difference is that, in contrast to trade costs, tariffsdonotgenerateresourcecosts, asshowninEquation39. Nonetheless, theperformance ofthePPItargetingruleinresponsetohighertariffsisqualitativelysimilartotheresultspresented forhighertradecosts. InAppendixF,werepeattheexperimentinthissectionandshowthathigher tariffs on intermediate inputs also worsen monetary policy trade-offs. 6 Quantifying the Effect of Trade Shocks The model developed and calibrated in Section 4 is particularly well suited to analyze the effects of changes in trade costs on inflation and GDP during two salient episodes: the U.S.-China Trade war of 2018-2019 and the post-pandemic inflation surge of 2020-2023. In this section, we use our model to quantify the inflationary effects of trade costs during these two recent events and emphasize the features of our model that make it optimal to carry out these analyses. 37

6.1 U.S.-China Trade War: 2018-2019 In 2018 and 2019, the U.S. and China imposed tariffs and other trade barriers on each other, and reached a partial agreement in 2020. We use our multi-country model to gauge the implications of increased trade barriers between these countries. Two key features of our model make it ideal to analyze these implications. First, relative to a two-country model, our multi-country model can incorporate the trade diversion patterns generated by the bilateral nature of the U.S.-China trade war. Second, our model allows for differences in changes in tariffs between final and intermediate goods—a feature of the 2018-19 U.S.-China trade war (Bown, 2021). Measured changes in tariffs, together with changes in our measure of U.S.-China bilateral trade costs between 2018 and 2019, allow us to construct shocks that are tailor-made to be fed into our model. We calibrate the size of the trade shocks based on the tariffs imposed by the U.S. on China and China’sretaliatoryresponsetotheU.S.(Fajgelbaumetal.,2019;Bown,2021). Table2summarizes our calibration. Throughout the various waves of the trade dispute, the U.S. increased tariffs on Chinese imports by 20 percentage points. In response, China raised tariffs on U.S. exports by 10 percentage points. We target these average tariff changes and assume that in the model, the U.S. imposes a 13 percent tariff on final goods and a 25 percent tariff on intermediate inputs imported from China. This choice reflects the actual nature of the trade war, with approximately two-thirds of imported goods affected by U.S. tariffs being intermediate inputs (Fajgelbaum and Khandelwal, 2022). In terms of retaliation, we assume that China imposes a tariff of 10 percent on final goods and of 15 percent on intermediate inputs. This assumption reflects that China’s response was, on average,halfoftheU.S.actionandwastargetedtowardsU.S.intermediateexports. Ourcalibration implies that the average bilateral tariff increase for each good-type—more precisely, the geometric average ofbilateral tariffs, whichis inline withourmeasure ofbilateral tradecosts in(8)—matches the observed increase in the bilateral trade costs we recovered from the data in Section 2. In the simulation, the tariff increases are transitory, with an expected duration of five years, as in our baseline calibration. In our simulation, U.S. inflation rises, and U.S. GDP growth slows (Figure 9). The effect on inflation is significant: The increase in trade costs drives U.S. inflation up by 0.3 percentage points above the baseline and causes it to remain persistently elevated. The contribution of trade costs in final goods (the blue bars) is short-lived and vanishes after a year.44 By contrast, the contribution 44Thefigureshowsfour-quarterinflationrates. Therefore,aone-timeriseinthepriceleveloccurringintheinitial quarter shows up as an increase in four-quarter inflation for the following four quarters. 38

Table 2: Calibration of U.S.-China Trade War Simulation Tariffs Trade Costs U.S. → China China → U.S. Average tariff U.S. ↔ China 1 (a) (b) [(1+a)(1+b)]2 −1 (2018-2019) Final Goods 13 10 11 10.9 Intermediate Inputs 25 15 20 19.7 Trade weighted average 21 10 11 10.9 of higher trade costs in intermediates (the red bars) induces a persistently elevated inflation rate. As the costs of importing inputs from China rise, U.S. firms react by making greater use of inputs sourcedfromotherregions,includingtheU.S.itself. Thesedifferentinputs,however,arenotperfect substitutes for inputs imported from China, leading to lower production efficiency for U.S. firms. Consequently, U.S. marginal costs increase persistently, resulting in higher inflation for longer. The associated higher policy rates contribute to a persistent drag on GDP relative to the baseline (right panel). Figure 9: Increase in U.S.-China trade costs, effects on U.S. 0.6 0.6 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 0 4 8 12 16 0 4 8 12 16 Note: The figure shows the effects of an 13 and 25 percentage point increase in U.S. tariffs on Chinese imports of finalandintermediategoods,respectively,togetherwithincreasesof10and15percentagepointsinChina’simports fromtheU.S.offinalandintermediategoods,respectively. Thebluebarsshowthecontributionsoffinalgoodstrade costs, and the red bars show the contributions of intermediate goods trade costs. Figure 10 shows the effects on inflation and GDP in China and the rest of the non-U.S. regions. China experienced a larger hit to GDP than the U.S. did and a smaller increase in inflation. These facts reflect that China’s retaliation was only partial, and that exports constitute a larger 39

Figure 10: Increase in U.S.-China trade costs, effects on non-U.S. regions 0.3 0.2 0 0.2 0.1 -0.4 0 -0.1 -0.8 0 4 8 12 16 0 4 8 12 16 Note: The figure shows the effects of a 13 and 25 percentage point increase in U.S. tariffs on Chinese imports of finalandintermediategoods,respectively,togetherwithincreasesof10and15percentagepointsinChina’simports from the U.S. of final and intermediate goods, respectively. The impulse responses show the response of inflation and GDP for China in black, Asian economies excluding China in green, advanced economies excluding the U.S. in purple, and the rest of the world in yellow. share of China’s GDP compared to the U.S. The GDP in regions other than the U.S. and China experienced a boost, as firms and households in the U.S. and China partially diverted trade flows toward imports from these countries. The non-U.S. advanced economies, which have the U.S. as an important trading partner, as well as the rest of the world, experienced larger activity increases compared to the non-China Asian economies. In turn, inflation rises in the foreign economies— reflecting both higher input costs and some depreciation of their currencies against the dollar—but the increase is very modest. 6.2 Post-Pandemic Trade Costs and Inflation In this section, we explore the contribution of trade cost shocks during the 2021-2022 surge in U.S. inflation, in the aftermath of the COVID-19 pandemic. This period provides a natural laboratory to explore the inflationary effects of disruptions in trade flows resulting from factors such as supply chain disruptions, bottlenecks, and higher shipping costs among others. All these factors are captured by iceberg trade cost shocks in our model, which we separately identify from other supply and demand forces at play during this period. To quantify the contributions of trade cost shocks, domestic shocks, and foreign shocks in driving U.S. inflation and GDP growth during 2021-2022, we rely on a two-country version of our 40

model. Restricting ourselves to two countries allows us to estimate the model using quarterly data for the U.S. and a ROW aggregate. We construct quarterly series starting in 1999 for U.S. domestic sourcing shares of final and intermediate goods, which, together with other more common macroeconomic series, permit us to quantify the role of trade cost shocks. Data. Relative to our analysis in Section 2, we assemble novel quarterly data on the evolution of U.S. domestic sourcing shares in final goods and intermediate inputs. Constructing quarterly sourcing shares is challenging because gross output measures at such a frequency are generally unavailable. We follow a procedure similar to Eaton et al. (2016b) to overcome this challenge.45 Figure 11 shows the evolution of our quarterly measure of the domestic sourcing shares. The blue line is the domestic sourcing share in final goods, and the red line is the sourcing share in intermediate inputs. Domestic sourcing shares declined through 2008, consistent with the continuing expansion of global trade. Thereafter, the sourcing share for final goods remained steady, whereas the domestic sourcing share in intermediate inputs trended upward, in line with the slowdown in globalization. Relative to the related analysis by Comin and Johnson (2020) using annual data for the U.S., our quarterly measure has the advantage of capturing higher frequency movements in sourcing shares, which are key to teasing out the different drivers of inflation.46 Inadditiontothenewseriesondomesticsourcingshares, weestimatethemodelusingstandard macroeconomic variables. For both blocs, we collect quarterly real GDP growth and CPI inflation and a measure of nominal interest rates. We measure the rest of the world as a trade-weighted average of the major U.S. trading partners as in Bodenstein et al. (2023). Appendix G provides additional details and data sources.47 Shock inference and identification. We maintain the model calibration reported in Table 1. In addition to trade cost shocks, we incorporate three additional shocks: technology, domestic demand, and monetary policy shocks. These shocks follow an autoregressive process with normally distributedinnovations. Weuseatwo-stepBayesianinferenceproceduretoestimatetheparameters governing these shocks and the realized unobserved disturbances. First, we estimate the shock processparametersusingdatafrom1999:Q1to2019:Q4, avoidingoutliersinducedbythepandemic lockdown in 2020. Second, we extend our sample through 2023:Q4 to recover the full time series 45AsinSection2,wefocusonmanufacturingindustriesandinterpolateannualfiguresfromtheBEA’sinput-output tables using industrial production. See Appendix G.2. 46Because manufacturing industries are more tradable than services, the sourcing shares are lower relative to the whole U.S. economy. We link the demeaned sourcing shares in the data to the model. See Appendix G for details. 47Unless otherwise noted, all quarterly data was collected from Haver Analytics (2024). For the U.S. policy rate, we use the Wu-Xia shadow federal funds rate to account for the tightness of monetary policy during the period of zero interest rates. For countries in RoW, we use a short-term nominal interest rate or policy rate when available. 41

Figure 11: U.S. Quarterly Domestic Sourcing Shares 0.86 0.84 0.82 0.8 0.78 0.76 0.74 0.72 2000 2005 2010 2015 2020 Notes: U.S. sourcing shares interpolated from BEA input-output tables. The blue line corresponds to the domestic sourcing share of final goods. The red line depicts the domestic sourcing share for intermediate inputs. Sourcing shares corresponds to tradable sectors in accordance with the standard NICS classification. See Appendix G for details. of smoothed shocks. See Appendix G for details and parameter estimates. Figure 12 shows that data on domestic sourcing shares are central to identifying trade cost shocks and differentiating their contribution from other sources of aggregate supply fluctuations. The figure shows impulse responses to a total factor productivity shock (blue line), a trade cost shock for final goods (red line), and a trade cost shock for intermediate inputs (yellow line). The main source of identification comes from the correlation between inflation and GDP with domesticsourcingshares. AsshowninthetoppanelsofFigure12, adversetotalfactorproductivity shocks and adverse trade costs shocks produce a negative correlation between inflation and GDP, butthemagnitudeoftheeffectsisquitedifferent. Totalfactorproductivityshockshavestrongerand more persistent effects on these variables relative to the impact of trade cost shocks. Importantly, as shown by the bottom panels, total factor productivity shocks lower the domestic sourcing share through the expenditure switching channel induced by the appreciation of the exchange rate. In contrast, trade cost shocks have the opposite effect on the domestic sourcing shares, with adverse tradecostshocksgeneratinganincreaseinthedomesticsourcingsharesinducedbythereallocation of demand for final goods and intermediate inputs produced domestically. In addition, trade costs shocks generate large movements in sourcing shares and relatively modest changes in GDP and inflation, while the converse is true for TFP shocks. These observations help understand how the 42

model identifies the historical contributions of these shocks to inflation, discussed next.48 Shock decomposition. Exploiting the identification from the domestic sourcing shares we analyze the drivers of inflation in the post-pandemic period. Figure 13 shows the shock decomposition of inflation and GDP growth for the period 2018:Q1-2023:Q4. For ease of interpretation, we classify shocks into three groups: trade costs shocks, combining trade costs to final goods and intermediate inputs; domestic shocks, collecting shocks to total factor productivity, consumption demand, and monetary policy; foreign shocks, which encompass all disturbances originating in the foreign economy, including shocks to the UIP condition. Our estimates reveal that trade costs significantly influenced inflation dynamics during the pandemic and its aftermath. In the early stages, higher trade costs helped counteract deflationary pressures from collapsing demand. As the economy reopened, trade costs partially retraced, alleviating inflationary pressures. Most U.S. inflation in the later part of 2021 and the first half of 2022 stemmed from domestic factors, while trade cost shocks had a more limited effect during 48Appendix A.13 shows impulse responses to demand-side shocks in the estimated model. Figure 12: Identification of Trade Cost Shocks 2.5 0 2 -1 1.5 -2 1 -3 0.5 -4 0 -5 -0.5 -6 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 Notes: Impulse response to a one standard deviation to total factor productivity shock (blue), trade cost shock for final goods (red), trade cost shock for intermediate inputs (yellow). Model calibrated at the estimated posterior mean parameters in Table A.3 43

Figure 13: Decomposition of U.S. CPI and U.S. GDP Growth U.S. CPI Inflation U.S. GDP growth Notes: The black solid lines correspond to four quarter changes in the U.S. Consumer Price Index (left panel) and on U.S. real GDP (right panel) in deviations from the sample average over 1999:Q1-2019:Q4. The bars represent the contribution of trade costs (blue), U.S. shocks (red), and foreign shocks (yellow). U.S. shocks correspond to TFP, consumption preferences, and monetary policy shocks. Foreign shocks are the counterpart but also include UIP shocks. All shocks are estimated with the Kalman smoother and with the model calibrated with the estimated posterior mean parameters in Table A.3. this stage of the inflation surge.49 Starting in the second half of 2022 and afterwards—in line with signs of trade fragmentation following the Russian invasion of Ukraine and tensions in the Middle East—increases in trade costs hindered the disinflation process, contributing approximately one percentage point to additional inflation in 2022 and driving much of the inflation persistence in 2023. Trade shocks also had a notable contribution to economic activity. The right panel of Figure 13 shows the shock decomposition for U.S. GDP growth. Higher trade costs lowered U.S. GDP growth at the onset of the pandemic in 2020; their reversal boosted activity in 2021 and 2022, and higher trade costs became a drag for growth in 2023. 7 Conclusions We provide evidence that the macroeconomic effects of higher import costs are akin to negative supply shocks, leading to higher CPI inflation and lower GDP growth. On impact, a 10 percentage pointincreaseinacountry’simportcostsrelativetoanothercountryleadstoa1.2percentagepoint 49We do not aim to decompose supply vs demand factors, Bianchi et al. (2023), Barro and Bianchi (2024), and GiannoneandPrimiceri(2024)emphasizetheroleoffiscalandotherdemandshocks,whileBlanchardandBernanke (2023) and Gagliardone and Gertler (2023) underline the role of supply shocks. 44

(≈ 0.65+0.58) relative increase in CPI inflation. Interestingly, we also document that these effects depend on the type of good—either final or intermediate—affected by the increase in import costs, with higher trade costs for intermediate inputs leading to significantly more persistent increases in inflationrelativetothoseforfinalgoods. Weareabletodocumentthesefactsbyconsideringshocks to a broad measure of trade costs that we construct by exploiting static Gravity and data from global input-output tables. We estimate bilateral trade costs that can be compared consistently across countries and time, and then aggregate them into country-specific import costs that capture variationinbothpolicy—suchastariffs—andnon-policy—suchasshippingcosts—inducedchanges in trade costs. We develop a dynamic trade model of inflation by embedding intertemporal trade in final and intermediate goods into a multi-country open economy model of inflation. Our approach to incorporate inflation dynamics relies on the off-the-shelf elements of models with staggered price and wage adjustments, as in the New Keynesian literature. Even under relatively stringent assumptions about heterogeneity across countries in our calibration, the model can replicate the empirical impulse responses we estimate in relative terms. Most importantly, our model allows us to provide an estimate for the macroeconomic effects of higher import costs. For the U.S., we find that a 10 percentage point increase in import costs—for example, in the form of blanket tariffs with all its trading partners—leads to a contemporaneous rise in CPI inflation of approximately 0.8 percentage points and a decline in GDP of approximately 1.8 percent. More importantly, the inflationary effects are persistent, resulting from higher trade costs of intermediate goods. We also use our model to analyze the inflationary effect of the U.S.-China 2018-19 trade war and to quantify the role of trade costs during the post-pandemic inflation episode. Although nontrade-related factors affecting aggregate supply and demand explain the buildup of inflation in the post-pandemic recovery, higher trade costs contributed to sustaining inflation above the Federal Reserve’s target, particularly in late 2022 and 2023. 45

References Alessandria, George and Horag Choi, “The dynamics of the U.S. trade balance and real exchange rate: The J curve and trade costs?,” Journal of International Economics, 2021, 132, DOI: 10.1016/j.jinteco.2021.103511. , Shafaat Yar Khan, Armen Khederlarian, Carter Mix, and Kim J. Ruhl, “The aggregate effects of global and local supply chain disruptions: 2020–2022,” Journal of International Economics, 2023, 146, DOI: 10.1016/j.jinteco.2023.103788. NBER International Seminar on Macroeconomics 2022. Amiti, Mary, Sebastian Heise, Fatih Karahan, and Ay¸segu¨l S¸ahin, “Inflation strikes back: The role of import competition and the labor market,” NBER Macroeconomics Annual, 2024, 38 (1), DOI: 10.1086/726645. , Stephen J. Redding, and David E. Weinstein, “The Impact of the 2018 Tariffs on Prices and Welfare,” Journal of Economic Perspectives, 2019, 33 DOI: 10.1257/jep.33.4.187. An, Sungbae and Frank Schorfheide, “Bayesian analysis of DSGE models,” Econometric reviews, 2007, 26 (2-4), DOI: 10.1080/07474930701220071. Anderson, James E. and Eric van Wincoop, “Trade Costs,” Journal of Economic Literature, September 2004, 42 (3), DOI: 10.1257/0022051042177649. Antr`as, Pol and Davin Chor, “Global value chains,” in Gita Gopinath, Elhanan Helpman, and Kenneth Rogoff, eds., Handbook of International Economics, Vol. 5 of Handbook of International Economics, Elsevier, pp. 297–376 2022, DOI: https://doi.org/10.1016/bs.hesint.2022.02.005. Auclert, Adrien, Matthew Rognlie, and Ludwig Straub, “The Macroeconomics of Tariff Shocks,” National Bureau of Economic Research, April 2025. Autor, David H., David Dorn, and Gordon H. Hanson, “The China Syndrome: Local Labor Market Effects of Import Competition in the United States,” American Economic Review, October 2013, 103 (6), DOI: 10.1257/aer.103.6.2121. Backus, David K., Patrick J. Kehoe, and Finn E. Kydland, “International Real Business Cycles,” Journal of Political Economy, 1992, 100 (4), DOI: 10.1086/261838. 46

Bai, Yan and Jos´e-V´ıctor R´ıos-Rull,“DemandShocksandOpenEconomyPuzzles,”American Economic Review, May 2015, 105 (5), DOI: 10.1257/aer.p20151121. Barattieri, Alessandro, Matteo Cacciatore, and Fabio Ghironi, “Protectionism and the business cycle,” Journal of International Economics, 2021, 129, DOI: 10.1016/j.jinteco.2020.103417. Barro, Robert and Francesco Bianchi, “Fiscal Influences on Inflation in OECD Countries, 2020-2023,” NBER Working Paper, 2024. Bergin, Paul R. and Giancarlo Corsetti, “The macroeconomic stabilization of tariff shocks: What is the optimal monetary response?,” Journal of International Economics, 2023, DOI: https://doi.org/10.1016/j.jinteco.2023.103758. Bianchi, Francesco and Andrea Civelli, “Globalization and inflation: Evidence from a timevarying VAR,” Review of Economic Dynamics, 2015, 18 (2), , Renato Faccini, and Leonardo Melosi, “A fiscal theory of persistent inflation,” The Quarterly Journal of Economics, 2023, 138 (4), DOI: 10.1093/qje/qjad027. Bianchi, Javier and Louphou Coulibaly, “TheOptimalMonetaryPolicyResponsetoTariffs,” National Bureau of Economic Research, 2025. Blanchard, Olivier J and Ben S Bernanke, “What caused the US pandemic-era inflation?,” National Bureau of Economic Research, 2023. Bodenstein, Martin, Pablo Cuba-Borda, Nils Gornemann, Ignacio Presno, Andrea Prestipino, Albert Queralto, and Andrea Raffo, “GlobalFlighttoSafety, BusinessCycles, and the Dollar,” International Finance Discussion Paper, 2023, DOI: 10.17016/IFDP.2023.1381. Boehm, Christoph E., Andrei A. Levchenko, and Nitya Pandalai-Nayar, “The Long and Short (Run) of Trade Elasticities,” American Economic Review, April 2023, 113 (4), DOI: 10.1257/aer.20210225. Bown, Chad P, “The US–China trade war and Phase One agreement,” Journal of Policy Modeling, 2021, 43 (4), DOI: 10.1016/j.jpolmod.2021.04.003. 47

Caldara, Dario, Matteo Iacoviello, Patrick Molligo, Andrea Prestipino, and Andrea Raffo,“Theeconomiceffectsoftradepolicyuncertainty,”Journal of Monetary Economics,2020, 109, DOI: 10.1016/j.jmoneco.2019.11.002. Caliendo, Lorenzo and Fernando Parro, “Estimates of the Trade and Welfare Effects of NAFTA,” The Review of Economic Studies, 11 2014, 82 (1), DOI: 10.1093/restud/rdu035. Comin, Diego and Robert Johnson, “Offshoring and Inflation,” C.E.P.R. Discussion Papers, October 2020. , Robert C. Johnson, and Callum Jones, “Import Constraints,” AEA Papers and Proceedings, May 2024, 114, DOI: 10.1257/pandp.20241100. Correia, Sergio, “A feasible estimator for linear models with multi-way fixed effects,” 2016. Corsetti, Giancarlo, Luca Dedola, and Sylvain Leduc, “Optimal Monetary Policy in Open Economies,” in Benjamin M. Friedman and Michael Woodford, eds., Handbook of Monetary Economics, Vol. 3 of Handbook of Monetary Economics, Elsevier, pp. 861–933 2010, DOI: 10.1016/B978-0-444-53454-5.00004-9. Cuba-Borda, Pablo, Albert Queralto, Ricardo Reyes-Heroles, and Mik¨ael Scaramucci, “How Do Trade Disruptions Affect Inflation?,” February 2025. FEDS Notes, Board of Governors of the Federal Reserve System. de Michelis, Andrea, Grace Lofstrom, Mike McHenry, Musa Orak, Albert Queralto, and Mika¨el Scaramucci, “Has the Inflation Process Become More Persistent? Evidence from the Major Advanced Economies,” July 2024. FEDS Notes, Board of Governors of the Federal Reserve System. Eaton, Jonathan and Samuel Kortum, “Technology, Geography, and Trade,” Econometrica, 2002, 70 (5), DOI: 10.1111/1468-0262.00352. , , and Brent Neiman, “Obstfeld and Rogoff’s international macro puzzles: a quantitative assessment,” Journal of Economic Dynamics and Control, 2016, 72, DOI: 10.1016/j.jedc.2016.06.002. , , , and John Romalis, “Trade and the global recession,” American Economic Review, 2016, 106 (11), DOI: 10.1257/aer.20101557. 48

Erceg, Christopher, Andrea Prestipino, and Andrea Raffo, “Trade Policies and Fiscal Devaluations,” American Economic Journal: Macroeconomics, October 2023, 15 (4), DOI: 10.1257/mac.20210163. Erceg, Christopher J, Dale W Henderson, and Andrew T Levin, “Optimal monetary policy with staggered wage and price contracts,” Journal of monetary Economics, 2000, 46 (2), DOI: 10.1016/S0304-3932(00)00028-3. Fajgelbaum, Pablo D. and Amit K. Khandelwal, “The Economic Impacts of the US–China Trade War,” Annual Review of Economics, 2022, DOI: https://doi.org/10.1146/annureveconomics-051420-110410. , Pinelopi K Goldberg, Patrick J Kennedy, and Amit K Khandelwal, “The Return to Protectionism*,” The Quarterly Journal of Economics, 2019, 135 DOI: 10.1093/qje/qjz036. Fitzgerald, Doireann, “Trade Costs, Asset Market Frictions, and Risk Sharing,” American Economic Review, May 2012, 102 (6), DOI: 10.1257/aer.102.6.2700. Flaaen, Aaron and Justin R Pierce, “Disentangling the Effects of the 2018-2019 Tariffs on a Globally Connected US Manufacturing Sector,” Board of Governors of the Federal Reserve System (US), 2019. Frankel, Jeffrey A and Andrew K Rose, “Anestimateoftheeffectofcurrencyunionsontrade andoutput,”Quarterly Journal of Economics,2002,117(2),DOI:10.1162/003355302753650292. Frankel, Jeffrey A. and David H. Romer, “Does Trade Cause Growth?,” American Economic Review, June 1999, 89 (3), DOI: 10.1257/aer.89.3.379. Furceri, Davide, Swarnali A Hannan, Jonathan D Ostry, and Andrew K Rose, “Macroeconomic consequences of tariffs,” National Bureau of Economic Research, 2018. Gagliardone, Luca and Mark Gertler, “Oil prices, monetary policy and inflation surges,” National Bureau of Economic Research, 2023. Gal´ı, Jordi, Monetary Policy, Inflation, and the Business Cycle: An Introduction to the New Keynesian Framework and Its Applications Second editionnumber 10495.In ‘Economics Books.’, Princeton University Press, 2015. 49

Giannone, Domenico and Giorgio Primiceri, “The drivers of post-pandemic inflation,” National Bureau of Economic Research, 2024. Giovanni, Julian Di, S¸ebnem Kalemli-O¨zcan, Alvaro Silva, and Muhammed A Yildirim, “Pandemic-era inflation drivers and global spillovers,” National Bureau of Economic Research, 2023. Hanson, Gordon H., “The Rise of Middle Kingdoms: Emerging Economies in Global Trade,” Journal of Economic Perspectives, May 2012, 26 (2), DOI: 10.1257/jep.26.2.41. Haver Analytics, “Quarterly macroeconomic data retrieved from Haver Analytics,” 2024. Accessed on November 2024. Head, Keith and John Ries, “Increasing Returns versus National Product Differentiation as an Explanation for the Pattern of U.S.-Canada Trade,” American Economic Review, September 2001, 91 (4), DOI: 10.1257/aer.91.4.858. and Thierry Mayer, “Gravity Equations: Workhorse,Toolkit, and Cookbook,” in “Handbook of International Economics,” Vol. 4, Elsevier, chapter 3, pp. 131–195 2014, DOI: 10.1016/B978- 0-444-54314-1.00003-3. Heathcote, Jonathan and Fabrizio Perri, “Financial autarky and international business cycles,” Journal of Monetary Economics, 2002, 49 (3), DOI: 10.1016/S0304-3932(02)00103-4. Ho, Paul, Pierre-Daniel G Sarte, and Felipe F Schwartzman, “Multilateral Comovement in a New Keynesian World: A Little Trade Goes a Long Way,” DOI: 10.21144/wp22-14. Hottman, Colin and Ricardo Reyes-Heroles, “Globalization, Inflation Dynamics, and the Slope of the Phillips Curve,” 2023. Jacks, David S., Christopher M. Meissner, and Dennis Novy, “Trade Costs, 1870-2000,” American Economic Review, May 2008, 98 (2), DOI: 10.1257/aer.98.2.529. , , and , “Trade booms, trade busts, and trade costs,” Journal of International Economics, 2011, 83 (2), DOI: https://doi.org/10.1016/j.jinteco.2010.10.008. Kalemli-O¨zcan, Sebnem, Can Soylu, and Muhammed A Yildirim, “Global Networks, Monetary Policy and Trade,” National Bureau of Economic Research, April 2025. 50

Obstfeld, Maurice and Kenneth Rogoff, “The six major puzzles in international macroeconomics: is there a common cause?,” NBER macroeconomics annual, 2000, 15, DOI: 10.1086/654423. Reyes-Heroles, Ricardo, “The Role of Trade Costs in the Surge of Trade Imbalances,” 2017. , Sharon Traiberman, and Eva Van Leemput, “Emerging markets and the new geography of trade: The effects of rising trade barriers,” IMF Economic Review, 2020, 68, DOI: 10.1057/s41308-020-00115-3. Simonovska, Ina and Michael E. Waugh, “The elasticity of trade: Estimates and evidence,” Journal of International Economics, 2014, 92 (1), DOI: 10.1016/j.jinteco.2013.10.001. Stockman, Alan C. and Linda L. Tesar, “Tastes and Technology in a Two-Country Model of the Business Cycle: Explaining International Comovements,” The American Economic Review, 1995, 85 (1), Teti, Feodora, “Missing Tariffs,” Centre for Economic Policy Research, 2025. Timmer, M. P., E. Dietzenbacher, B. Los, R. Stehrer, and G. J. de Vries, “AnIllustrated User Guide to the World Input–Output Database: the Case of Global Automotive Production,” Review of International Economics, 2015, 23, DOI: 10.1111/roie.12178. Werning, Iv´an, Guido Lorenzoni, and Veronica Guerrieri, “Tariffs as Cost-Push Shocks: Implications for Optimal Monetary Policy,” National Bureau of Economic Research, May 2025. O`scar Jord`a, “Estimation and Inference of Impulse Responses by Local Projections,” American Economic Review, March 2005, 95 (1), DOI: 10.1257/0002828053828518. 51

Appendix A World Input-Output Data Our baseline empirical analysis relies on data from the OECD Inter-Country Input-Output (ICIO) data. This data set provides global input-output tables, which map flows of production and expenditure within countries and flows of international trade between countries, broken down by economic activity and country, globally. The ICIO considers 77 countries, 76 individual countries and a rest of the world aggregate (ROW) from 1995 through 2020. We aggregate the ICIO data to cover the same countries as WIOD data (described below) and use rest of world to carry the adjustment. The ICIO considers 45 sectors or areas of activity. We define sectors 1-22 (codes A-C) as tradable and the remaining sectors as non-tradable. We aggregate sectors into tradable and non-tradable. We then construct total final expenditure (aggregating final demand categories) on tradable sectors across geographic origins and obtain {XC } in every t. Similarly, we consider ih,t i,h total expenditure by tradable sectors on tradable sectors across geographic origins and obtain {XM } in every t. We use these flows to construct bilateral trade costs. We have considered ih,t i,h alternative aggregation procedures and our main empirial results remian robust. We also consider data from the World Input-Output Database (WIOD).We combine the 2013 and 2016 releases of WIOD data. The WIOD 2013 release considers input-output data for 41 countries (including ROW) and 35 sectors for the period 1995-2011. The WIOD 2016 release considers 44 countries (including ROW) and 56 sectors covering the period 2000-2014. We combine both releases to obtain a balanced panel of with 41 countries for the period 2005-2014. Table A.1 provides our industry mapping between the WIOD and ICIO databases. We also consider WIOD’s long-run database which has input-output table estimates for 25 countries and 23 sectors covering the period 1965-2000. We follow the same procedure as with the ICIO data and construct bilateral trade costs relying on tradable sectors only. A.1

Appendix Table A.1: Manufacturing Sector Classification in WIOT and ICIO Sector WIOD2016 ICIO2023 1 Cropandanimalproduction,huntingandrelatedserviceactivities Agriculture,hunting,forestry 2 Forestryandlogging 3 Fishingandaquaculture Fishingandaquaculture 4 Miningandquarrying Miningandquarrying,non-energyproducingproducts Miningsupportserviceactivities 5 Manufactureoffoodproducts,beveragesandtobaccoproducts Foodproducts,beveragesandtobacco 6 Manufactureoftextiles,wearingapparelandleatherproducts 7 Manufactureofwoodandofproductsofwoodandcork, Textiles,textileproducts,leatherandfootwear exceptfurniture;manufactureofarticlesofstraw andplaitingmaterials 8 Manufactureofpaperandpaperproducts Woodandproductsofwoodandcork 9 Printingandreproductionofrecordedmedia Paperproductsandprinting 10 Manufactureofcokeandrefinedpetroleumproducts Cokeandrefinedpetroleumproducts 11 Manufactureofchemicalsandchemicalproducts Chemicalandchemicalproducts 12 Manufactureofbasicpharmaceuticalproducts Pharmaceuticals,medicinalchemical andpharmaceuticalpreparations andbotanicalproducts 13 Manufactureofrubberandplasticproducts Rubberandplasticsproducts 14 Manufactureofothernon-metallicmineralproducts Othernon-metallicmineralproducts 15 Manufactureofbasicmetals Basicmetals 16 Manufactureoffabricatedmetalproducts, Fabricatedmetalproducts exceptmachineryandequipment 17 Manufactureofcomputer,electronicandopticalproducts Computer,electronicandopticalequipment 18 Manufactureofelectricalequipment Electricalequipment 19 Manufactureofmachineryandequipmentn.e.c. Machineryandequipment,nec 20 Manufactureofmotorvehicles,trailersandsemi-trailers Motorvehicles,trailersandsemi-trailers 21 Manufactureofothertransportequipment Othertransportequipment 22 Manufactureoffurniture;othermanufacturing Manufacturingnec;repair andinstallationofmachineryandequipment 23 Repairandinstallationofmachineryandequipment A.2

Appendix B Evolution of Bilateral Trade Costs This Appendix considers alternative data sources to construct bilateral trade costs. Figure A.1 plots the evolution of the distribution of bilateral trade costs constructed using the WIOD data for thesamesetofcountriesthatweconsiderusingtheICIOdata. FigureA.2plotstheevolutionofthe distribution of bilateral trade costs constructed using historical WIOD data. For the overlapping period 1995-2000, trade costs are lower in the historical WIOD data because of the difference in the sample of countries. The historical WIOD does not consider almost any low-income countries. Figure A.1: Evolution of distribution of trade costs (all 41 countries): WIOD Note: Trade costs are expressed as a percentage of the sales price of the good. That is, the figure shows the evolution of (TQ −1) for Q∈{C,M}. ih,t Figure A.2: Historical Evolution of Global Trade Costs (a) Median Trade Cost: Final Goods (C) (b) Median Trade Cost: Intermediate Goods (M) Note: Trade costs are expressed as a percentage of the sales price of the good, divided by 100. That is, the figure shows the evolution of TQ −1 for Q ∈ {C,M}. Shaded areas are bounded by the 20th and 80th ih,t percentiles. A.3

Appendix Figures A.3 and A.4 combine data from the historical WIOD and the 2013 and 2016 WIOD releases to construct and plot the historical evolution of the U.S. domestic sourcing share and U.S. import costs, respectively. These costs are constructed with ηC = ηQ = 5 and weighted using contemporaneous import weights. Figure A.3: Evolution of United States domestic sourcing share Note: data comes from WIOD database. The 2011-2014 numbers have been taken from the WIOD 2016 databaseandstitchedtotheWIOD2013numbers. The1965-1999comefromthehistoricalWIODdatabase Figure A.4: Evolution of trade costs in the United States Note: Trade costs are expressed as a percentage of the sales price of the good, divided by 100. That is, the figure shows the evolution of τQ −1 for Q ∈ {C,M}. Data come from WIOD database. The 2011-2014 i,t numbers have been taken from the WIOD 2016 database and stitched to the WIOD 2013 numbers. The 1965-1999 come from the historical WIOD database A.4

Appendix C Additional Empirical Results C.1 Response of Trade Costs Figure A.5: Local projection of trade cost on itself (a) Final trade cost (b) Intermediate trade cost Note: Country fixed effects and year error clustering are included. The figure considers a 10 percentage point increase in import costs. C.2 Additional Local Projection Estimates Figures A.6 andA.7 plots the estimates of the local projection in Equation 11 for real exports, real imports, the real exchange rate and the trade-balance-to-GDP ratio. A.5

Appendix Figure A.6: Response to Final Trade Costs (a) Real Exports (b) Real Imports (c) Real Exchange Rate (d) TB/GDP A.6

Appendix Figure A.7: Response to Intermediate Trade Costs (a) Real Exports (b) Real Imports (c) Real Exchange Rate (d) TB/GDP A.7

Appendix C.3 The Trade Elasticity Our baseline estimate assumes a parameter of η = η = 5. However, there is a range of estimates M forthisparameterinthetheliterature(Boehmetal.,2023). Wealsoconstructbilateraltradecosts as in (8) using different values of the trade elasticity, η−1 = {2,4,6,8}, and re-estimate our local projections in (11). Table A.2 shows how our estimates of βQ for h = 0,1 depend the value of the h trade elasticity. Given observed bilateral trade flows, a lower trade elasticity is consistent smaller bilateral trade costs and more muted variation in these costs across countries. Hence, for lower values of the trade elasticity, a given change in trade costs generates small changes in bilateral trade flows, and therefore small effects on both CPI inflation and GDP growth. The first two rows of each panel, h = 0 and h = 1, of Table A.2 show how our estimates of βQ for Q ∈ {C,M} become smaller h for lower values of η. An alternative way to present our estimates is to normalize βQ to reflect h a change in trade flows consistent with a 1 percentage point increase in an importer’s domestic sourcing share, ωQ . This can be done by running the local projections specifications in (11) with ii,t countryi’sdomesticsourcingshareasthedependentvariable. Inresultsnotshowninthisappendix (available upon request), we show that normalized estimates are almost identical across different values of the trade elasticity. A.8

Appendix Table A.2: Inflation and GDP growth regressions for different trade elasticities (θ = η−1) (a) h=0 YearonyearInflationRate(h=0) YearonyearGDPgrowth(h=0) (1) (2) (3) (4) (5) (6) (7) (8) η−1=2 η−1=4 η−1=6 η−1=8 η−1=2 η−1=4 η−1=6 η−1=8 %∆Finaltradecost 0.124 1.275 2.915 4.725 0.0131 0.0427 0.0798 0.125 (0.0809) (0.472) (1.019) (1.627) (0.0458) (0.310) (0.677) (1.086) %∆Interm. tradecost 0.159 0.747 1.463 2.234 -0.105 -0.719 -1.546 -2.454 (0.114) (0.680) (1.387) (2.146) (0.0504) (0.365) (0.796) (1.269) CPIrate%(-1) -0.207 -0.197 -0.195 -0.194 (0.223) (0.227) (0.228) (0.229) GDPgrowth%(-1) 2.491 2.565 2.548 2.537 (1.530) (1.540) (1.550) (1.556) Unemployment%(-1) -0.0834 -0.0915 -0.0958 -0.0983 0.0924 0.0983 0.0997 0.100 (0.0612) (0.0625) (0.0624) (0.0624) (0.0393) (0.0401) (0.0401) (0.0401) R-squared 0.857 0.852 0.851 0.850 0.652 0.646 0.645 0.644 Num. ind. 27 27 27 27 27 27 27 27 Num. obs. 542 542 542 542 542 542 542 542 (b) h=1 YearaheadInflationRate(h=1) YearaheadGDPgrowth(h=1) (1) (2) (3) (4) (5) (6) (7) (8) η−1=2 η−1=4 η−1=6 η−1=8 η−1=2 η−1=4 η−1=6 η−1=8 %∆Finaltradecost -0.0741 0.652 1.721 2.906 -0.0453 -0.379 -0.866 -1.406 (0.125) (0.392) (0.895) (1.464) (0.0730) (0.500) (1.098) (1.766) %∆Interm. tradecost 0.317 0.575 0.717 0.833 -0.119 -0.806 -1.688 -2.642 (0.198) (0.543) (0.985) (1.477) (0.0652) (0.486) (1.067) (1.711) CPIrate%(-1) 0.0210 0.0458 0.0495 0.0508 (0.122) (0.126) (0.127) (0.127) GDPgrowth%(-1) 2.098 2.261 2.272 2.273 (2.090) (2.112) (2.134) (2.144) Unemployment%(-1) -0.0850 -0.0730 -0.0724 -0.0726 0.285 0.295 0.297 0.298 (0.0668) (0.0677) (0.0681) (0.0682) (0.0771) (0.0782) (0.0784) (0.0785) R-squared 0.634 0.609 0.606 0.605 0.668 0.661 0.660 0.659 Num. ind. 27 27 27 27 27 27 27 27 Num. obs. 542 542 542 542 542 542 542 542 Note: Country fixed effects and year error clustering are included. Both sourcing share and CPI inflation tables respond to a 10% increase in trade costs. A.9

Appendix D Robustness: Model vs Data D.1 Technological Parameters. We explore the role of two key technological parameters shaping the transmission of trade cost shocks: the share of intermediate inputs in firms’ production (ν), and the elasticity of substitution between intermediate inputs and labor (ε ). Figure A.8 contrasts the effects in our baseline y calibration with ν = 0.4 and ε = 0.5, with alternatives assuming higher and lower values for y these parameters. A higher share of intermediate inputs implies a larger increase in inflation and a bigger decline in GDP. Similarly, a lower elasticity of substitution between intermediates and labor is also associated with amplified GDP and inflation effects of trade shocks. However, the impact of different values of the elasticity parameter is smaller relative to the effects induced by different shares of intermediate inputs. Figure A.8: Model vs Data: Robustness Technological Parameters 1.5 1.5 1 1 0.5 0 0.5 -0.5 0 -1 -0.25 0 4 8 12 16 20 0 4 8 12 16 20 1.2 1.2 0 0 -1 -1 -2 -2 -3 -3 0 4 8 12 16 20 0 4 8 12 16 20 Note: Effectsofapermanent10percentagepointincreaseintheU.S.’stradecostsfromalltradingpartnersunder alternative model calibrations. D.2 Wage and Price Indexation Because of the New Keynesian structure inflation dynamics in the model are influenced by the frequency of price and wage adjustment as well as the degree of price and wage indexation. So if prices and wages are more flexible we should expect to more inflation on impact. If price and wage setting are backward-looking due to indexation we should expect to see more persistence in the inflationary response. We study the transmission of trade shocks under two alternative A.10

Appendix configurations. One with less stickiness in prices and wages for which we set θ = θ = 0.6 w which halves the duration of price and wage contracts relative to our baseline calibration. For the second experiment we extend the model to incorporate price and wage indexation. We assume that country’s i differentiated firms unable to adjust prices follow the indexation rule P = P πι π1−ι, with the parameter ι determining the degree of indexation with respect to i,t t−1 i,t−1 past inflation π . Similarly, country-i’s households unable to reset wage follow the indexation rule: i,t W i,t = W i,t−1 π i ι , w t−1 π1−ιw. Where ι w determines the degree of wage indexation. FigureA.9contraststheeffectsinourbaselinecalibrationwhereθ = θ = 0.8andnoindexation, w with one in which calibration in which prices and wages are more flexible θ = θ = 0.6 and another w calibration in which prices and wages backward-looking with ι = ι = 0.3. w Figure A.9: Model vs Data: Robustness to Price and Wage Setting Parameters 1.5 1.5 1 1 0.5 0 0.5 -0.5 0 -1 -0.25 0 4 8 12 16 20 0 4 8 12 16 20 1.2 1.2 0 0 -1 -1 -2 -2 0 4 8 12 16 20 0 4 8 12 16 20 Note: Effects of a permanent 10 percentage point increase in the U.S.’s trade costs from all trading partners under alternative model calibrations. A.11

Appendix E CPI vs PPI Rule: Transitory Increase in Trade Costs Studyingone-timepermanentincreasesintradecostsisausefulbenchmark. However,thechanging nature of trade policy suggests that changes in trade costs are not permanent. The transitory nature of trade policies can have implications for the desirability of monetary stabilization. Figure A.10 illustrates this point by re-assessing the benefits of PPI targeting following a transitory 10 percentage point increase in bilateral trade costs between the U.S. and all trading partners.50 In thiscase,PPItargetingdoeslittletoamelioratetheGDP,eveninthecaseoffinalgoodstradecosts. As the price of imported goods falls due to the reversal of trade costs, the PPI-based rule generates persistent CPI deflation, increasing the real interest rate and reducing aggregate consumption. Figure A.10: CPI vs. PPI Targeting with Transitory Increase in Trade Costs 0.8 0.5 0.75 0.6 0 0.5 0.4 -0.5 0.2 0.25 -1 0 0 -0.2 -1.5 -0.1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 0.8 0.5 0.75 0.6 0 0.5 0.4 -0.5 0.2 0.25 -1 0 0 -0.2 -1.5 -0.1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 Note: Effects of a transitory 10 percentage point increase in the U.S.’s trade costs from all trading partners under alternative monetary policy rules. 50For this experiment we assume tariffs follow the auto-regressive process in our baseline calibration. A.12

Appendix F Monetary Policy Response to Tariffs We extend our analysis of alternative policy rules for the case of higher tariffs. In the model we use the shock κ to simulate the effect of tariffs imposed by country-i on imports from country-h. ih,t We use the shock κ to simulate the extent of retaliatory tariffs. hi,t Figure A.11 analyzes the case in which the U.S. imposes a transitory 10 percent tariff on the imports from all its trading partners. We assume the persistence of new tariffs equals that of trade cost shocks under our baseline calibration in Table 1. Figure A.12 shows the same experiment for the case of a permanent increase in U.S. tariffs on imports from all its trading partners. In both experiments we assume that U.S.’s trading partners impose retaliatory tariffs of the same magnitude. The top row shows the response to higher tariffs on final goods. The bottom row shows the response to higher tariffs on intermediate inputs. The results are very similar to those discussed in the main text. Importantly, when monetary policy follows a PPI rule, it can provide some relief to the contractionary effect of higher tariffs on final goods but this alternative policy rule does not improve the outcomes following an increase in tariffs for intermediate inputs. Figure A.11: CPI vs. PPI Targeting with Transitory Increase in U.S. Tariffs 0.8 0.5 0.75 0.6 0 0.5 0.4 -0.5 0.2 0.25 -1 0 0 -0.2 -1.5 -0.1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 0.8 0.5 0.75 0.6 0 0.5 0.4 -0.5 0.2 0.25 -1 0 0 -0.2 -1.5 -0.1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 Note: Effects of a transitory 10 percentage point increase in the U.S. tariffs on all trading partners under alternative monetary policy rules. GDP measure is net of tariff revenues. A.13

Appendix Figure A.12: CPI vs. PPI Targeting with Permanent Increase in U.S. Tariffs 0.8 0.5 0.75 0.6 0 0.5 0.4 -0.5 0.2 0.25 -1 0 0 -0.2 -1.5 -0.1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 0.8 0.5 0.75 0.6 0 0.5 0.4 -0.5 0.2 0.25 -1 0 0 -0.2 -1.5 -0.1 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 Note: Effects of a permanent 10 percentage point increase in the U.S. import tariffs on all trading partners under alternative monetary policy rules. GDP measure is net of tariff revenues. G Post-Pandemic Inflation Analysis G.1 Data Summary Unless otherwise noted, all series are at quarterly frequency and seasonally adjusted by the corresponding agency. All relevant series are in per capita terms to be consistent with the model definition. All series are obtained through Haver Analytics (2024) unless otherwise specified. United States. We collect the following data series for the U.S. for the period 1999:Q1- 2023:Q4. • Gross Domestic Product: we collect quarterly real GDP from the Bureau of Economic Analysis (BEA). We take the quarter-on-quarter log difference as our final measure. • Consumer Price Inflation: we take the consumer price inflation index, which we then transform by taking the quarter-on-quarter log difference. • Policy rate: we use the Wu-Xia shadow federal funds rate to measure the interest rate, to prevent from being stuck at the ZLB. The data is assembled by the Federal Reserve Bank of Atlanta. • Real Exchange Rate: our measure of the real exchange rate is the ”Real Broad Effective A.14

Appendix Exchange Rate” for United States obtained from FRED. • U.S. Domestic Sourcing Shares for Final Goods: See Appendix G.2 • U.S. Domestic Sourcing Shares for Intermediate Inputs: See Appendix G.2 Rest-of-World Aggregate For the rest-of-world aggregate we combine country-level time series for country/blocs: Argentina, Australia, Brazil, Bulgaria, Canada, Colombia, Chile, China, Croatia, Czech Republic, Denmark, Euro Area, Hong Kong, Hungary, India, Indonesia, Israel, Japan, Malaysia, Mexico, New Zealand, Philippines, Poland, Romania, Russian Federation, Saudi Arabia,Singapore,SouthAfrica,SouthKorea,Sweden,Taiwan,Thailand,Turkey,UnitedKingdom and the United States. Our sample of countries represents about 85% of PPP-adjusted world GDP in 2019. Unless otherwise note, all data is seasonally adjusted. • Foreign GDP: trade-weighted average of real GDP growth measured as quarter-on-quarter log difference for each country. • Foreign inflation: trade-weighted average of consumer price inflation, measured as quarteron-quarter log difference of the CPI index of each country. • Foreign policy rate: we proxy the foreign policy rates using the money market interest rate whereavailable, otherwiseweusethedepositrate. Theforeignpolicyrateisaggregatedusing U.S. trade weights. G.2 Quarterly Domestic Sourcing Shares From the BEA input-output tables, we can construct the annual intermediate and final domestic sourcing shares as follows. We first filter the sectors of the “Use of Commodities by Industries” tables by selecting only tradable sectors.51 We then aggregate all rows, and the industry columns to compose the intermediate goods, while aggregating the final end-use columns (both denoted as E , or expenditure, where j ∈ {intermediate,final}). We also collect the nominal dollar export j expenditure (which cannot be split into intermediates and final goods due to being a single series) from the same table, denoted as X. We then collect imports for intermediates and final goods using the “Use of imported commodities by industry” tables in a similar aggregation process to 51Tradable sectors are defined as sectors 1-5 and 8-26, in accordance with NICS classifications. A.15

Appendix our earlier tables (denoted as M where j ∈ {intermediate,final}). We use the import series to j construct the domestic sourcing share, as: M j λ = 1− j E j Where j represents the final and intermediate sourcing share. Our data runs annually from 1997 to 2023. We then use the BEA International Trade in Goods, to interpolate our annual sourcing share series. We first collect total intermediate and final exports from this table, which we use to derive intermediate and final “shares”. We then multiply our previously collected nominal exports series, X, into intermediate and final exports using these shares (resulting in X , where X = X + j final X ). We then obtain annual intermediate and final output using the expenditure series intermediate as: O = E −M +X (G.1) j j j j Where once again j ∈ {intermediate,final}. Therefore, we have annual output, imports, exports and expenditure split by intermediate and final goods. We then use quarterly industrial production from FRED (manufacturing and consumer final goods IP) to interpolate our new annual output series for intermediate and final goods respectively (we use PPI to deflate the series before interpolation, and reflate after interpolation). Then, we use the quarterly exports and imports by intermediate and final goods (from our International Trade in Goods database) to interpolate our annual numbers X and M . j j All quantities are nominal, so no deflation is necessary. Thus, we obtain quarterly OQ,EQ,MQ,XQ (by reversing equation (G.1) to obtain EQ), which j j j j j enables us to obtain quarterly domestic sourcing share as λQ = 1 − M j Q . We also compose j EQ j “Annualized Domestic Sourcing Share” λQAnn, still a quarterly series, by taking 4-quarter rolling j sumsofMQ,EQ andrecomputingλ aspreviously. Wenowhavequarterlydomesticsourcingshare j j j ready for our Bayesian estimation exercise. A.16

Appendix G.3 Mapping the Model to the Data In each country-bloc, we observe the following variables: quarterly annualized output growth (∆y˜o ), quarterly annualized inflation measured (π˜o ), and quarterly annualized nominal interest i,t i,t rates (R˜o ), for i = {U.S.,RoW}. The real exchange rate index between the U.S. and RoW, (qˆo), is i,t t measured in deviations from its long-run value of 100. For the U.S. only, we measure the domestic sourcingsharesinfinalgoodsandintermediateinputsλ˜C,oandλ˜M,o,respectively. Variablesdenoted t t with a tilde have been demeaned using their sample averages. We map the observed data series to the model counterparts through the following system of measurement equations: ∆y˜o = 100×log(y /y ), i ∈ {U.S.,RoW} i,t i,t i,t−1 π˜o = 400×logπ , i ∈ {U.S.,RoW} i,t i,t R˜o = 400×logR , i ∈ {U.S.,RoW} i,t i,t qˆo = 100×log(q /q ), t 12,t 12 λˆC,o = 100× (cid:0) λC −ωC(cid:1) , U.S.,t 11,t 11 λˆM,o = 100× (cid:0) λM −ωM(cid:1) , U.S.,t 11,t 11 G.4 Priors, Posterior Sampler, and Estimation Results Columns 2-4 in Table A.3 list prior distributions, along with prior means and standard deviations used to estimate the two-country model of Section 6.2. We assume the statistical independence of estimated parameters under the prior distribution, so we compute the joint prior density from the product of the marginal distributions. Using standard perturbation techniques, we approximate the model solution around its nonstochastic steady state and evaluate the likelihood function using the Kalman filter. We use the standard random walk Metropolis algorithm (RWM) described in An and Schorfheide (2007) to generate draws from the posterior distribution. The covariance matrix of the proposal distribution in the RWM algorithm to obtain an acceptance rate between 30% and 40%. We simulate 50,000 draws from the simulated posterior distribution and retain only the last 25,000 draws for posterior inference. Columns 5 and 6 in Table A.3 show key moments of the posterior distribution of the estimated parameters. A.17

Appendix Table A.3: Estimated Parameters: Two-Country Model Parameter Description Prior Posterior Mean HPD Interval (1) (2) (3) (4) (5) σA Std dev. U.S. TFP shock IG(0.01,0.05) 0.04 [0.04 0.05] 1 σr Std dev. U.S. Monetary policy shock IG(0.01,0.05) 0.01 [0.01 0.01] 1 σD Std dev. U.S. Demand shock IG(0.01,0.05) 0.02 [0.02 0.03] 1 σA Std dev. RoW TFP shock IG(0.01,0.05) 0.01 [0.01 0.02] 2 σr Std dev. RoW Monetary policy shock IG(0.01,0.05) 0.00 [0.00 0.00] 2 σD Std dev. RoW Demand shock IG(0.01,0.05) 0.01 [0.01 0.01] 2 σψ Std dev. RoW UIP shock IG(0.01,0.05) 0.01 [0.00 0.01] 2 σ Std dev. Trade cost shock - C IG(0.01,0.05) 0.04 [0.03 0.04] τC σ Std dev. Trade cost shock - M IG(0.01,0.05) 0.08 [0.08 0.10] τM ρA Persistence U.S. TFP shock B(0.6,0.125) 0.98 [0.96 0.99] 1 ρr Persistence U.S. Monetary policy response B(0.6,0.125) 0.94 [0.92 0.96] 1 ρD Persistence U.S. Demand shock B(0.6,0.125) 0.56 [0.47 0.66] 1 ρA Persistence RoW TFP shock B(0.6,0.125) 0.89 [0.82 0.95] 2 ρr Persistence RoW Monetary policy response B(0.6,0.125) 0.77 [0.67 0.85] 2 ρD Persistence RoW Demand shock B(0.6,0.125) 0.79 [0.72 0.87] 2 ρψ Persistence RoW UIP shock B(0.6,0.125) 0.83 [0.77 0.89] 2 ρτC Persistence Trade cost shock - C B(0.6,0.125) 0.93 [0.90 0.97] ρτM Persistence Trade cost shock - M B(0.6,0.125) 0.88 [0.83 0.92] ρ(ϵD,ϵD) Correlation U.S. and Row demand shock U(0,0.5774) 0.02 [-0.16 0.19] 1 2 ρ(ϵA,ϵA) Correlation U.S. and Row TFP shock U(0,0.5774) 0.15 [-0.00 0.30] 1 2 Notes: The estimation sample is 1991:Q1 - 2019:Q4. Column (3) reports the prior distributions. B is Beta distribution. IG is Inverse Gamma distribution. U is Uniform distribution. The numbers in parentheses denote the prior mean and standard deviation of each distribution. Column (4) report posterior meanss. Column (5) reports the Highest Probabality Density Interval in square brackets. All posterior statistics are based on the last 25,000 draws from a RWMH algorithm, after discarding the first 25,000 draws. A.18

Appendix G.5 Additional Impulse Responses Figure A.13: Identification of Demand and Supply Shocks 2.5 3 3 2 2 2 1 1 1.5 0 -1 0 1 -2 -1 0.5 -3 -2 -4 0 -3 -5 -0.5 -6 -4 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0.8 0.8 0.05 0 0.6 0.6 -0.05 0.4 0.4 -0.1 0.2 0.2 -0.15 -0.2 0 0 -0.25 -0.2 -0.2 -0.3 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 Notes: Impulse response to a one standard deviation to total factor productivity shock (blue), trade cost shock for final goods (red), trade cost shock for intermediate inputs (yellow). Consumption demand shock (purple). Monetary policy shock (green). Model calibrated at the estimated posterior mean parameters in Table A.3 A.19