The Dollar and Emerging Market Economies: Financial Vulnerabilities Meet the International Trade System

K.7 The Dollar and Emerging Market Economies: Financial Vulnerabilities Meet the International Trade System Shousha, Samer Please cite paper as: Shousha, Samer (2019). The Dollar and Emerging Market Economies: Financial Vulnerabilities Meet the International Trade System. International Finance Discussion Papers 1258. https://doi.org/10.17016/IFDP.2019.1258 International Finance Discussion Papers Board of Governors of the Federal Reserve System Number 1258 October 2019

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 1258 October 2019 The Dollar and Emerging Market Economies: Financial Vulnerabilities Meet the International Trade System Samer Shousha NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References to International Finance Discussion Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors. 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.

The Dollar and Emerging Market Economies: Financial ∗ Vulnerabilities Meet the International Trade System Samer Shousha Federal Reserve Board† October 2019 Abstract This paper shows that dollar appreciations lead to declines in GDP, investment, and credit to the private sector in emerging market economies (EMEs). These results imply that the transmission of dollar movements to EMEs occurs mainly through financial conditions rather than net exports, contrary to what would be expected from the conventional Mundell-Fleming model. Moreover, the central role of the U.S. dollar in global trade invoicing and financing - the dominant currency paradigm - and the increased integration of EMEs into international supply chains weaken the traditional trade channel. Finally, as expected if financial vulnerabilities are prominent, EMEs with higher exposure to credit denominated in dollars and lower monetary policy credibility experience greater contractions during dollar appreciations. JEL classification: F31, F34, F36, F41, F44 Keywords: Dollar, balance sheet mismatch, dominant currency paradigm, global value chain, monetary policy credibility ∗Theviewsinthispaperaresolelytheresponsibilityoftheauthorandshouldnotbeinterpreted asreflectingtheviewsoftheBoardofGovernorsoftheFederalReserveSystemoranyotherperson associated with the Federal Reserve System. I thank Shaghil Ahmed for extensive comments and suggestionsonanearlierdraftofthispaper.Iwouldalsoliketothank,forveryusefulcommentsand suggestions,DaniloCascaldi-Garcia,StephanieCurcuru,ChristopherErceg,ThiagoFerreira,Felipe Iachan, Matteo Iacoviello, Joao Victor Issler, Steve Kamin, Friederike Niepmann, Andre Silva, and seminar participants at FGV-EPGE and the International Finance Workshop of the Federal Reserve Board. †DivisionofInternationalFinance,BoardofGovernorsoftheFederalReserveSystem,Washington,D.C.20551U.S.A.Email:samer.f.shousha@frb.gov 1

1 Introduction The predominance of the U.S. dollar (dollar henceforth) in the international trade andfinancialsystemhaveledtoincreasingfocusontheimportanceofdollarmovements for the global economy, especially for emerging market economies (EMEs). Figure1showstherelationshipbetweenthebroadrealdollarandaveragedetrended EME GDP (on the left) and EME investment (on the right).1 Based on these simple correlations, a stronger dollar (and, therefore, weaker EME currencies) is related to weakerGDPandinvestmentinEMEs,contrarytowhatyouwouldexpectfromtraditional trade channels.2 This paper focuses on the question: Do these relationships only represent simple correlations, or are there causal channels through which a strongerdollarleadstoacontractionofactivityinEMEs? FIGURE 1: DollarandEmergingMarketEconomiesBusinessCycles Source:NationalsourcesandFRED. I answer this question using a structural empirical model to evaluate quantitatively the effects of movements in the broad real dollar on business cycles fluctuationsinEMEsandthedifferenttransmissionchannelsoftheseeffects.Morespecifically, my analysis proceeds in two stages. First, I estimate a panel VAR system with 1The broad real dollar is a weighted average of the foreign exchange value of the U.S. dollar againstthecurrenciesofabroadgroupofmajorU.S.tradingpartnersinrealterms(adjustedusing consumerpriceindexes). 2The conventional textbook trade channel states that a depreciation of the local currency makes exportsrelativelycheaperandimportsmoreexpensive.Thisresultsinhighernetexportsandashift ofconsumersawayfromimportedgoodstowarddomesticones. 2

thirteen EMEs.3 I show that dollar shocks are important sources of business cycle fluctuations in EMEs, and positive dollar shocks (dollar appreciations) lead to declines in GDP, investment, and real credit to the private sector and to an increase in EMEsovereignspreads. Inthesecondpartofmyanalysis,Ievaluatetheimportanceofthedifferentchannelsthroughwhichdollarmovementscanbetransmittedtotherealeconomy.More specifically, I evaluate the importance of four previously studied mechanisms. The firsttwo,dominantcurrencypricingandglobalvaluechainintegration,weakenthe effects of the traditional trade channel, while the other two, balance sheet vulnerabilitiesandlackofmonetarypolicycredibility,canrevertitseffects (i) Dominant currency pricing: The prevalence of dollar invoicing in trade leads to trade prices being sticky in dollars rather than in local currency, mitigating the effects of dollar movements on exports while amplifying their effects on imports; (ii) Global value chain integration: Increased integration into international supply chains leads trade flows to be less sensitive to exchange rate movements, as the higher import content of exports causes imports and exports to tend to movetogether. (iii) Balancesheetvulnerabilities:Ifaneconomyhassubstantialdollar-denominated debt,dollarappreciationsexpandthedomestic-currencyvalueofliabilitiesrelativetoassets,weakeningbalancesheetsandtighteningfinancialconditions; (iv) Lackofmonetarypolicycredibility:Countrieswithlowmonetarypolicycredibility have higher exchange rate pass-through to domestic inflation, which mightrequireadditionalmonetarypolicytighteningafterdollarappreciations. After constructing indices to quantitatively rank the countries in terms of the importance to them of each mechanism, I split them in two groups, one with rela- 3Iprefertousethepaneldatamethodologybecauseitincreasestheefficiencyandpowerofthe analysis,asindividualcountries’VARswouldhavetoomanyparameterscomparedtothetimeseries length. 3

tively high exposure and other with relatively low exposure to each of the indices, and reestimate the panel VAR for each of the two groups. The analysis shows that, regardless of how we divide the countries, dollar appreciations lead to GDP, investment, and credit contractions in EMEs and have negligible effects on exports. Moreover,countrieswithrelativelyhigherintegrationinglobalvaluechainshavea smallercontractionofimports,asyouwouldexpectwithahigherimportcontentof exports, while countries with higher dollar invoicing have a greater contraction on imports, also as expected. Finally, countries with higher balance sheet vulnerabilities and lower monetary policy credibility experience greater contractions in GDP, investment,andimports,suggestingthatbothbalancesheetvulnerabilitiesandlack ofmonetarypolicycredibilityareimportantingeneratingthecontractionaryeffects ofdollarappreciationsinEMEs. Itisimportanttonotethatthispaperdoesnotultimatelyidentifythefundamentalsourcesoffluctuationsinthedollar,whichisanassetpriceandconsequentlycan be affected by a myriad of factors. Indeed, it is difficult to disentangle the sources of exchange rates movements more generally. For example, there is a vast literature about the so-called ”exchange rate disconnect puzzle”, namely that macroeconomic fundamentals are largely disconnected from exchange rates, especially over short horizons (Meese and Rogoff (1983), Obstfeld and Rogoff (2000), Itskhoki and Mukhin (2019)).4 The paper’s main objective is to understand how EMEs respond to movements in the broad real dollar, irrespective of their source, and which characteristics are more relevant to the transmission of dollar fluctuations to EMEs’ real variables.. This paper is related to a growing body of the literature that studies the role of thedollarastheworld’sdominantcurrency.GoldbergandTille(2009)andGopinath (2016) document the ample use of the dollar in trade invoicing. Additionally, Boz etal.(2018)showthatthepresenceofa‘dominantcurrency’affectsthetransmission of exchange rate fluctuations to terms-of-trade, export and import quantities, and 4To alleviate concerns about the endogeneity of the dollar, I control for other variables usually relatedtotheglobalfinancialcyclesuchastheVIX,ameasureofglobaluncertainty,theU.S.2-year realinterestrate,andU.S.realGDP. 4

the exchange rate pass-through into import prices. Indeed, Bruno, Kim, and Shin (2018) show that the combination of the ‘dominant currency paradigm’ with the fact that most of trade credit is denominated in dollars - what they call the working capital channel of trade fluctuations - implies a negligible effect or even a contraction on exports after a local currency depreciation. I contribute to this literature by showing that exports barely move in EMEs after dollar appreciations, in line with theworksofBozetal.(2018)andBruno,Kim,andShin(2018)inthe‘dominantcurrency paradigm’ and the working capital channel of trade fluctuations. I also show thatimportscontractmoreinEMEswithahighershareofdollartradeinvoicingafterdollarappreciations,againinlinewiththepredictionsofthe’dominantcurrency paradigm’. Thispaperalsosupportsthefindingsoftheliteraturethatstudiestheincreasing relevance of international supply chains, as shown, for example, by Johnson and Noguera (2017). On the theoretical side, Amiti, Itskhoki, and Konings (2014) develop a theoretical framework that predicts that firms with high import shares and high market shares have low exchange rate pass-through into their export prices, and they confirm the model predictions using Belgian firm-product-level data on imports and exports. On the empirical side, Johnson (2014) shows that using valueadded exports in place of gross exports has three implications: (i) countries appear less exposed to foreign expenditure changes; (ii) at the sectoral-level, the manufacturing sector looks substantially less exposed, and non-manufacturing sectors look substantially more exposed to foreign shocks; and (iii) the importance of shocks originating in particular export destinations differs. Moreover, he shows that integrationininternationalsupplychainschangeshowmovementsintheexchangerate are transmitted to export prices, depending also on the import content of exports. I contribute to this literature by showing that indeed higher integration in global value chains weakens the traditional trade channel of exchange rate movements, consistentwithexportsandimportsmovingmoreintandem. The paper is also related to the literature that challenges the conventional textbooknotionthatadepreciationofthecurrencyisexpansionarythroughitseffecton 5

net exports by arguing that financial channels can go the other way and more than offset the trade channel. For example, Krugman (1999) and Cespedes, Chang, and Velasco (2004) show in theoretical models that if an economy has a large share of debtdenominatedinforeigncurrency,aweakercurrencycanleadtoadeterioration in the balance sheets of domestic banks and firms and end up being contractionary. Kalemli-Ozcan, Kamil, and Villegas-Sanchez (2016) provide empirical evidence on this balance sheet channel, showing that in the presence of currency mismatches large depreciation events can lead to a persistent decline in output and investment. More recently, Bruno and Shin (2015), using a model where regional banks borrow indollarsfromglobalbankstolendtolocalcorporates,showthatinthepresenceof currency mismatches there is a tight link between local currency appreciation and loosening of financial conditions. Caballero, Fernandez, and Park (2019) also document a considerable increase in foreign financing through bond issuance in EMEs since the early 2000s, which unveiled an additional transmission channel of dollar movements to economic activity in these economies. Finally, Avdjiev et al. (2019) provide empirical evidence of a “triangular” relationship between dollar movements, cross-border bank flows, and real investment in EMEs. Using both macro (country-level) and micro (firm-level) data, they show that a dollar appreciation leadstoafallincross-borderbanklendingtoEMEsandlowerrealinvestment.Icontributetothisliteraturebyevaluatingempiricallytheeffectsofdollarmovementsin EMEs and showing that GDP, investment, and credit to the private sector contract afterdollarappreciations,andmoresoincountrieswithhigherdollar-denominated debt. Finally, this paper is related to the literature that links monetary policy credibility, exchange rate pass-through, and the effects of currency depreciations in real activity.Severalauthorshavefoundevidenceofalowerexchangeratepass-through toconsumerpriceswhencountrieshavegreatermonetarypolicycredibility(see,for example Carriere-Swallow et al. (2016) and Ha, Stocker, and Yilmazkuday (2019)). I contributetothisliteraturebyshowingthatthemonetarypolicycredibilitychannel is also important to explain the contractionary effect of dollar appreciations on real 6

activityinEMEs,especiallyincountrieswithalonghistoryofhighinflation. The remainder of the paper is organized as follows. Section 2 discusses in more detail the different transmission channels of exchange rate movements. Section 3 highlightsthemainstylizedfacts,describesthepanelVARspecification,andpresents andinterpretsthebaselineresults.Section4evaluatestheimportanceofeachofthe different transmission channels. Section 5 presents several robustness checks. Section6concludes. 2 How Are Dollar Movements Transmitted to EMEs? The literature that studies the transmission of exchange rate movements to real activity typically emphasizes the traditional trade channel (embodied in the seminal workofMundell(1963)andFleming(1962))asthekeydeterminantoftheresponse ofeconomiestocurrencymovements.Inthisframework,alocalcurrencydepreciationisexpansionarythroughitseffectsonnetexports.Assumingpricesarestickyin the currency of the producing country, a local currency depreciation leads to a declineinthepriceofexportsandariseinthepriceofimports,expandingthequantity exported and contracting the quantity imported, and thus leading to higher net exports.Moreover,thedepreciationincreasesthedemandforhomegoodsthroughthe change in relative prices, shifting consumers away from imported goods towards domesticonesand,consequently,leadingtoanoutputexpansion. However, there are some features in the current international trade system that weakentheeffectofthetraditionaltradechannel.First,thereisampleevidencethat most trade, especially in EMEs, is invoiced in dollars. This so-called dominant currencyparadigmstatesthattheprevalenceofdollarinvoicingintradeleadstoexport prices that are essentially denominated in dollars and, consequently, exchange rate movements against the dollar should have negligible effects on exports. Moreover, the dollar also affects credit conditions for working capital, as 80 percent of bank trade credit is denominated in dollars. These facts together could even lead to a contractioninexportsafteradollarappreciation,stronglyweakeningthetraditional 7

tradechannel.Finally,asimportpricesarealsopredominantlydenominatedindollars, dollar appreciations lead to greater contractions in import quantities. Figure 2 providesevidenceontheprominenceofdollarinvoicinginglobaltrade. FIGURE 2: DollarDominanceinTradeInvoicing Source: National sources, Kamps (2006), Goldberg and Tille (2009), Gopinath (2016), Castellares (2017),Labbe(2018),andGiulianoandLuttini(2019). Second, a higher integration into global value chains - higher import content of exports - dampens trade volume responses to exchange rate movements, as export and import volumes tend to move together. Moreover, in the presence of dominant currency pricing, this effect is amplified, as export prices and marginal costs also tend to move together. Thus, the correlation of export and import volumes should behigherthegreatertheimportcontentofexports. Moving to financial channels that can revert the effects of the traditional trade channel, let us first consider the balance channel channel. The literature on this channel argues that local currency depreciations in the presence of currency mismatches of assets and liabilities weaken balance sheets of domestic bank and firms, tightening financial conditions and contracting lending and real activity.5 Figure 3 5Baskayaetal.(2017)provideevidencethatEMEs’domesticbanks,especiallythosewithhigher 8

illustratesthemechanism.Whenthelocalcurrencydepreciates,thenetworthofthe domestic bank or firm decreases from n to n . This reduction happens because t t+1 the foreign-currency value of local assets contracts (from S ∗d to S ∗d , with d t t t+1 t being the local currency value of assets and S being the exchange rate) relative to liabilities (d ∗B in dollars). This decrease in net worth leads to a contraction in credit t extensionfromforeignlendersandtheneedforafurthercontractioninassets. FIGURE 3: BalanceSheetChannel This channel operates through both the demand and supply in the dollar global credit market. On the demand side, borrowers that are heavily indebted in dollars seeadeteriorationoftheirbalancesheetsanddecreasetheirpropensitytoconsume and invest. Onthe supply side, local banks that draw on cross-borderbank lending tolendtolocalborrowersareaffectedbylocalcurrencydepreciations,asitincreases the effective credit risk faced by banks, given the presence of currency mismatches. Consequently, episodes of appreciation of the dollar are associated with deleveragnon-core financing, extend more credit during periods of high capital inflows, using as a funding sourcetheinternationalcapitalmarket.Additionally,Kalemli-Ozcan,Liu,andShim(2018)showthat firms with a higher volume of foreign exchange debt before the exchange rate appreciates increase theirleverage(andconsequentlyrisktaking)relativelymoreaftertheappreciation,supportingthe balancesheetchannel. 9

ingofglobalbanksandanoveralltighteningofglobalfinancialconditions.6 AsFigure 4 shows, the broad real dollar is inversely related to dollar-denominated crossborder bank lending to non-U.S. residents and international dollar bonds issued by EMEs non-financial corporations, consistent with this mechanism.7 This feedback loop reinforces the contractionary effects of dollar appreciations, leading to a contraction in GDP, investment, and credit and an increase in risk spreads. This contractionshouldbestrongerthehighertheexposureofprivatesectorliabilitiestothe dollar, ie, the higher the share of dollar credit in the total credit to the private sector inthecountry. FIGURE 4: BroadDollarandDollarLoansandBonds Source:BISlocationalbankingstatisticsandFRED. Finally, monetary policy credibility and the anchoring of inflation expectations affect the extent to which relative price shocks have secondary effects. Countries with less anchored inflation expectations could have a higher exchange rate passthroughtoconsumerpricesandconsequentlyneedtotightenmonetarypolicyafter 6Bruno and Shin (2015) formulate a model with this interaction between global and local banks andthetightlinkbetweendollarmovementsandglobalfinancialconditions.Theyalsoshowusing apanelstudyof46countriesthatthepredictionsofthemodelaresupportedbythedata. 7Shin(2018)providessimilarevidence. 10

a local currency depreciation. Indeed, Ha, Stocker, and Yilmazkuday (2019) show that, although the exchange rate pass-through can vary considerably depending on the nature of the shocks, exchange rate pass-through is significantly lower in countriesthatadoptinflationtargetingandhavehighercentralbankindependence.The tighteninginmonetarypolicycouldmorethanoffsettheexpansionaryeffectsstemmingfromthetraditionaltradechannelandalsoleadtocontractionsinrealactivity afterlocalcurrencydepreciations,especiallyforEMEswithverylowcredibilityand higherratesofinflation. 3 Empirical Model and Results IfirstestimateastructuralpanelVARforthirteenemergingmarketeconomies(EMEs) toevaluatetheeffectsofdollarshocks. 3.1 Data and Panel VAR Specification Myempiricalmodeltakestheformofafirst-orderVAR: p ∑ Ay = η + B y +(cid:101) (1) i,t i k i,t−k i,t k=1 where η isacountryfixedeffect,idenotescountries,tdenotestimeperiod,and i y = [yf ,yh ] i,t i,t i,t yf = [gdpUS,rUS,vix ,reerUS], yh = [gdp ,inv ,exp ,imp ,crt ,r ,reer ] i,t t t t t i,t i,t i,t i,t i,t i,t i,t i,t gdpUSdenotestheU.S.GDP,rUSdenotestheU.S.2-yearrealinterestrate,vix det t t notestheindexoftheimpliedvolatilityinS&P500stockindexoptionpricesfromthe ChicagoBoardOptionsExchange(CBOE),reerUS denotesthebroadtrade-weighted t U.S.realexchangerate,gdp denotesrealgrossdomesticproduct,inv denotesreal i,t i,t grossfixedcapitalformation,exp denotesrealexports,imp denotesrealimports, i,t i,t crt denotes real credit volume to the non-financial private sector, r denotes the i,t i,t country-specificinterestrate,andreer denotestherealexchangerate.Allvariables i,t 11

are log deviations from a log-linear and a log-quadratic trend with the exception of U.S.GDP,VIX,interestrates,andexchangerateswhicharedeviationsfromalinear trend.Ialsoremovethesamplemeanafterdetrendingforeachvariableseparately. IestimatethepanelVARforthirteenEMEs-Argentina,Brazil,Chile,Colombia, Indonesia, Korea, Malaysia, Mexico, Peru, Philippines, South Africa, Thailand, and Turkey - using quarterly data from 1996 to 2018. The countries selected have welldevelopedfinancialmarketsandatleast15yearsofdata.Thedatasourcesarelisted intheAppendix. Table1showsbusinesscyclestatisticsforsamplecountries,averagingovercountryspecific moments. As mentioned earlier, Figure 1 shows that real activity variables (GDP and investment) are strongly negatively correlated with the dollar. More surprisingly, even exports are slightly negatively correlated with the dollar - the correlation coefficients with the broad real dollar are shown in the third column of Table 1. Also, consistent with previous work, the country interest rate is countercyclical in EMEs (column 2 in Table 1). Finally, the country interest rate has a positive comovementwiththedollarinEMEs. TABLE I SUMMARY STATISTICS FOR DETRENDED VARIABLES σ ρ(X ,BroadUSD ) ρ(X ,Y) X t t t t Y 0.03 -0.80 1.00 I 0.11 -0.81 0.98 Exp 0.04 -0.36 0.62 Imp 0.09 -0.77 0.95 Crt 0.07 -0.36 0.46 R 0.02 0.80 -0.81 REER 0.06 -0.74 0.83 Note: The data are the simple average of the indicators for Argentina, Brazil, Chile, Colombia, Indonesia, Korea, Malaysia, Mexico, Peru, Philippines, South Africa, Thailand, and Turkey. The data sources are listed in the Appendix. The data are sampled quarterly from 1996:Q1 to 2018:Q4. The columnslabeledY,I,Exp,Imp,Crt,R,andREERrefer,respectively,todetrendedGDP,investment, realexports,realimports,realcredit,countryrealinterestrate,andrealeffectiveexchangerate. σ X represents the standard deviation of each variable, ρ(X ,BroadUSD ) represents the correlation of t t eachvariablewiththebroadrealdollar,andρ(X ,Y)representsthecorrelationofeachvariablewith t t detrendedoutput. 12

IidentifythepanelVARbyasimplerecursivestructure,imposingthatthematrix A is lower triangular in equation (1), with the variables ordered in the same order presented in y . This means that the dollar can have contemporaneous and lagged i,t effects on EME variables, but affects other U.S. variables only with a lag. The idea is to isolate dollar shocks that are independent of contemporaneous movements in U.S.GDP,U.S.2-yearrealinterestrate,andtheVIX.8 I use the least square dummy variable (LSDV) estimator to estimate the panel VAR for each group. As T >>> N, the LSDV strategy is preferred to Generalized MethodsofMoments(GMM)estimatorsasithasbetterfinitesamplepropertiesand efficiency, especially if the degree of cross-section to time series variation is large. Also,withalargeT,Nickel(1981)critiqueregardingthebiasoftheLSDVestimator islessimportant.IusetheAkaikeInformationCriteria(AIC)toselectthelaglength andgetp=2asoptimal.Icalculatetheerrorbandsusingbootstrapmethods. 3.2 Baseline Results Figure5showstheimpulseresponsefunctionsfora10%positiveshocktothedollar on the variables in the model. Dollar appreciations lead to contractions in GDP, investment, and real credit as well as, an increase in sovereign risk in EMEs, contrary totheconventionaltextbooknotionthatadepreciationofthecurrencyisexpansionarythroughitseffectonnetexports.Theresultsaregenerallystatisticallysignificant andthecontractionininvestmentandrealcredittotheprivatesectorisparticularly strong,consistentwiththeresultsobtainedbyAvdjievetal.(2019).Moreover,wesee a negligible effect and even an initial small contraction in exports, consistent with the interaction between the ‘dominant currency paradigm’ and the working capital channeloftradefluctuations.Finally,importsexperienceastrongcontraction. Tounderstandthecontributionofeachshockfordifferentvariables,Iperforma variance decomposition of the forecast errors. Table 2 shows the results. According to my estimates, innovations in the dollar are responsible for about 21% of move- 8Iusethe2-yearrealinterestratetocontrolalsoforchangesinexpectationsaboutfuturemonetary policy,notonlyitscurrentstance. 13

FIGURE 5: Impulseresponsetoa10%dollarshock. Note: Marked black lines show point estimates of impulse responses respectively for the baseline panelVAR;and68%and95%confidencebandsaredepictedwithdark-grayandlight-grayshaded areas,respectively.Bootstrapconfidencebandsarebasedon100,000repetitions. mentsinGDPandinvestmentinEMEs.Forcredittothenon-financialprivatesector, dollar innovations are responsible for around 29% of fluctuations in EMEs. Moreover, the share of the variance of forecast errors explained by dollar innovations for imports is around 20%, while for exports it is much smaller, around 8%, again consistent with the dominant currency paradigm, which predicts that dollar movementsshouldhavenegligibleeffectsonexports.Finally,forthecountryinterestrate the share of the variance of forecast errors explained by dollar innovations is small, around10%. 4 Evaluating the Importance of the Channels To examine the importance of the different channels, I first construct indices that measure the exposure of countries to each channel. Then, I split the countries into twogroups,onewithrelativelyhighandotherwithrelativelylowexposuretoeach ofthechannels. 14

TABLE II VARIANCE DECOMPOSITION OF DETRENDED VARIABLES: PERCENT OF VARIATION EXPLAINED BY BROAD DOLLAR SHOCKS Forecasthorizonh Y I Exp Imp Crt R REER 1 4% 2% 1% 3% 0% 1% 1% 4 12% 10% 5% 11% 1% 3% 3% 8 18% 17% 5% 17% 6% 6% 6% 12 20% 20% 5% 19% 13% 7% 8% 24 21% 21% 7% 20% 27% 8% 11% 60 21% 21% 8% 20% 29% 10% 11% Note:ThecolumnslabeledY,I,Exp,Imp,Crt,RandREERrefer,respectively,todetrendedGDP,investment,realexports,realimports,realcredit,countryrealinterestrateandrealeffectiveexchange rate. 4.1 Construction of the Exposure Indices I follow Iacoviello and Navarro (2019) and construct each exposure index in three steps.First,Istandardizetheexposurevariablebysubtractingitsmeananddividing by its variance to make all exposure variables comparable. Denote the standardize variable by vk, where k refers to one of the four mechanisms in question. Then, I it construct a logistic transformation of the standardized measure getting lk = exp(v i k t ) it 1+vk it tomaptheexposurevariablesintheunitintervaltomakethemcomparableamong eachother. Theexposurevariablesforeachchannelare: (i) Globalvaluechainintegration:importcontentofexports; (ii) Dominantcurrencychannel:shareofexportsinvoicedindollars; (iii) Balance sheet channel: share of credit to the non-financial private sector denominatedindollars; (iv) Monetarypolicycredibility:averageinflationrate. Figure 6 presents the indices constructed. The detailed data sources are listed in theAppendix. 15

FIGURE 6: ExposureIndices Source: Author’s calculations based on national sources, BIS, Kamps (2006), Goldberg and Tille (2009),Gopinath(2016),Castellares(2017),Labbe(2018),andGiulianoandLuttini(2019). 4.2 Subgroup Analysis I evaluate the effects of each channel by doing a subgroup analysis, using the split thesamplein”relativelyhigh”and”relativelylow”exposuregroupsconstructedin section4.1. Figure 7 shows the results for the dominant currency channel. Countries with higher dollar trade invoicing have a greater contraction on imports, as expected. Moreover, both export responses are muted and even initially negative, supporting the interaction between the dominant currency paradigm and the working capital channeloftradefluctuationsproposedbyBruno,Kim,andShin(2018).9 Figure 8 shows the results for the global value chain integration intensity. The 9The low discrepancy of export responses between the two groups might be related to the fact that apart from South Africa and Turkey, all the remaining countries in my sample have a share of over80%oftradeinvoicedindollars,whichmakesithardertodifferentiatebetweenbothgroups. 16

FIGURE 7: DominantCurrencyChannel-Impulseresponsetoa10%dollarshock. Note: Solid blue and dashed red lines show point estimates of impulse responses respectively for thegroupwithrelativelyhighexposureandthatwithrelativelylowexposure;light-blueandlightred shaded areas represent 68% (1 standard deviation) confidence bands for each group. Bootstrap confidencebandsarebasedon100,000repetitions. effects on imports are significantly smaller for countries that have a higher global value chain integration, i.e., a higher import content on exports. These results are consistentwithahigherco-movementofexportsandimportsincountrieswithhigh integrationintointernationalsupplychains,whichmitigatestheexpansionaryeffect ofthetraditionalchannel. For all subgroups shown thus far, the effects of dollar appreciations on EME GDP, investment, and real credit are negative, with a more pronounced contraction in countries with lower global value chain integration and higher dollar trade invoicing.10 Thus,Imovenowtotheothertwochannels,balancesheetvulnerabilities andlackofmonetarypolicycredibility,whichcanreverttheeffectspredictedbythe traditionaltradechannel. Figure 9 shows the results for the balance sheet channel. Again, both groups 10Itisimportanttonotethatcountrieswithlowerglobalvaluechainintegrationandhigherdollar tradeinvoicingarealsomostlymajorcommodityexporters,whichusuallysufferadditionalnegative effectsfromdollarappreciationsduetothenegativeeffectfromsuchanappreciationoncommodity prices (see Cheng and Xiong (2014) for more details on the financialization of commodity markets andtherelationshipbetweencommoditypricesandthedollar). 17

FIGURE 8: GlobalValueChainIntegrationChannel-Impulseresponsetoa10%dollarshock. Note: Solid blue and dashed red lines show point estimates of impulse responses respectively for thegroupwithrelativelyhighexposureandthatwithrelativelylowexposure;light-blueandlightred shaded areas represent 68% (1 standard deviation) confidence bands for each group. Bootstrap confidencebandsarebasedon100,000repetitions. have a contraction in GDP, investment, and real credit and an increase in sovereign spreads. Moreover, countries with a higher share of dollar credit in total credit to the private sector experience greater contractions in GDP, investment, imports, and credit than those with a lower share.11 These results are consistent with adverse balance sheet effects being important in the transmission of dollar movements to EMEs. Finally, Figure 10 shows the results for the monetary policy credibility channel. Countries with persistently higher inflation also experience greater contractions in GDP, investment, imports, and real credit and a much more pronounced increase in country spreads. These results indicate that lack of monetary policy credibility is also a relevant distinguishing factor in the transmission of dollar movements to EMEs. 11Thegreatercontractiononimportscouldberelatedtothehighshareofimportedcapitalgoods onEMEinvestment,asshownbyEatonandKortum(2001). 18

FIGURE 9: BalanceSheetChannel-Impulseresponsetoa10%dollarshock. Note: Solid blue and dashed red lines show point estimates of impulse responses respectively for thegroupwithrelativelyhighexposureandthatwithrelativelylowexposure;light-blueandlightred shaded areas represent 68% (1 standard deviation) confidence bands for each group. Bootstrap confidencebandsarebasedon100,000repetitions. FIGURE 10: MonetaryPolicyCredibilityChannel-Impulseresponsetoa10%dollarshock. Note: Solid blue and dashed red lines show point estimates of impulse responses respectively for thegroupwithrelativelyhighexposureandthatwithrelativelylowexposure;light-blueandlightred shaded areas represent 68% (1 standard deviation) confidence bands for each group. Bootstrap confidencebandsarebasedon100,000repetitions. 19

Altogether, the subgroup analysis indicates that all the channels emphasized in therecentliteraturethatweakenorreversethetraditionaltradechannel-highdollar trade invoicing, integration into international supply chains, balance sheet vulnerabilities, and the lack of monetary policy credibility - are important for the transmission of dollar movements to EMEs. However, it is important to note that the subgroupanalysisonlyseparatesonecharacteristicatatimeandthuscannotspeak totherelativeimportanceofthedifferentchannelsinanestedmodel. 5 Robustness Analysis Robustness analysis confirm that the results hold under a host of alternative specificationswhichinclude(i)allowingcontemporaryeffectsofEMEs’realvariableson thedollarandonlylaggedeffectsofthedollaronthemandcheckingalsotheeffects of EME output shocks on the dollar to evaluate the possibility of reverse causality; (ii)estimatingthemodelonlyforthepreGlobalFinancialCrisisperiod;(iii)including commodity prices in the panel VAR, having both contemporaneous and lagged effects or only lagged effects on the dollar; and (iv) estimating individual VARs for eachEME.Idescribebeloweachrobustnessexercise,whilethefiguresareshownin theappendix. 5.1 EME Feedback Effects on the Dollar I reestimate the panel VAR allowing now for contemporary effects of EMEs’ real variables - output and investment - on the dollar to check any potential feedback effectsstemmingfromEMEstothedollar.First,Icheckiftheresultsofbroaddollar shocks are different if EME variables are put before the broad dollar in the causal ordering. The impulse responses for all variables are almost identical, with just a smaller effect in the short run on EME GDP and investment as the initial impact is zero by construction (Figure B.1). However, both impulse responses end up mostly at the same level of the baseline estimation. Second, I also check if there is any sig- 20

nificanteffectofEMEs’outputshocksonthedollar.Confirmingtheresultsobtained forthedollarshock,EMEs’GDPshockshaveanegligibleeffectonthedollar(Figure B.2). 5.2 Pre-Global Financial Crisis (GFC) Analysis Lilley et al. (2019) document a correlation between changes in U.S. foreign bond holdingsandthedollarthatemergessincetheGFC.Also,severalproxiesforglobal risk factors start to co-move strongly with the dollar and changes in U.S. foreign bond holdings around 2007, suggesting that risk plays a key role in this finding.12 I thus check the robustness of my findings, estimating the model only up to the pre-GFC period, when this co-movement is not present.13 Although the results for the U.S. GDP and U.S. real interest rate reverse, the results for EME variables all gointhesamedirectionasinthebaselinescenario,althoughtheyarequantitatively smallerforEMEGDP,investment,imports,andrealcredit(FigureB.3).Theseresults also indicate that even when the sources of dollar movements are different, dollar appreciationsarecontractionaryforEMEs. 5.3 Commodity Price Effects on the Dollar The subgroup analysis indicated that commodity exporters experience greater contractionsafterdollarappreciations,hintingthattheconnectionbetweencommodity pricesandthedollarcouldbeimportanttoexplainthecontractionaryeffectsofdollar appreciations, due to the negative relation between commodity prices and the dollar. I check that by including commodity prices in the panel VAR. The impulse responsesareverysimilartothebaselineestimation,irrespectiveofwhetherweput thedollarbeforecommoditypricesinthecausalordering(FigureB.4)orthereverse (FigureB.5). 12I thank Stephanie Curcuru and Friederike Niepmann for pointing out this fact and suggesting thisrobustnessexercise. 13IexcludeIndonesiafromthesampleforthisanalysisbecausethecountry’ssamplewouldbetoo short. 21

5.4 Country-specific VARs I estimate country-specific VARs to explore the cross-section of countries and evaluate the relation between the long-run effects on real activity and trade variables andtheexposureindices.Broadly,theresultsareverysimilartothesubgroupanalysis.14 For dollar invoicing, we see, again, a stronger contraction in imports with higher dollar invoicing. For global value chain integration, a higher integration means again a smaller contraction of imports. Moreover, countries with a higher share of dollar credit and lower monetary policy credibility experience greater contractions in output, investment, and imports. Finally, exports are, again, broadly insensitivetodollarmovements,regardlessofcountrycharacteristics. 6 Conclusion Thispaperevaluatestheeffectsofdollarmovementsonemergingmarketeconomies. First,IestimateapanelVARandshowthatdollarappreciationsleadtocontractions inGDP,investmentandcredittotheprivatesectorandanincreaseinsovereignrisk inEMEs.Moreover,asubgroupanalysisshowsthattheseresultsaretrueregardless of how we divide the countries, although the magnitudes vary across groups with different characteristics. These results suggest that the channels emphasized in the recent literature that weaken or reverse the traditional trade channel (high dollar invoicing, integration of supply chains, adverse balance sheet effects, and lack of monetarypolicycredibility)shouldbegivengreaterattention. The quantitatively small role of the trade channel and predominance of financial channels is coherent with previous works such as Bruno and Shin (2015). More recently, Hofmann, Shim, and Shin (2017) and Avdjiev et al. (2019) show that currency appreciations against the dollar lead to easier financial conditions, compressing sovereign bond spreads, and increasing dollar-denominated cross-border bank flows, which then lead to higher real investment. My results are consistent with these findings and show that there is also an interaction between financial vulner- 14Resultsforeachindividualcountryareavailableuponrequest. 22

abilities and the international trade system.15 A more complete modeling of the interactionsbetweenthesetwochannelsisapromisingavenueofwork. It is important to note that the results in this paper do not suggest that EMEs are better off with fixed or heavily managed currencies. Fixing or managing exchange rates have other negative effects, such as the loss of monetary autonomy andahigherriskofabalanceofpaymentcrisisduetovolatilecapitalflows.Infact, Kalemli-Ozcan (2019) shows that free-floating EMEs are much more insulated from international risk spillovers than EMEs with managed floats. Thus, although I find fewerbenefitsfromflexibleexchangerates,theseresultsdonotchangethefactthat EMEs are still better off with flexible rather than fixed or managed exchange rates. A more clear policy implication stemming from this work would be to emphasize the need for improved institutions to lower countries’ financial vulnerabilities such as better monetary policy implementation and possibly the use of macroprudential measurestolimit”excessive”borrowinginforeigncurrency. One important caveat is that this empirical model is agnostic about the exact sourcesofdollarmovements.AlthoughwelikelycanagreethatEMEdevelopments themselves are not the major factor behind movements in the dollar, dollar movements related, say, to changing expectations about U.S. growth prospects, or fiscal policy changes, or monetary policy shifts may have different effects on EMEs, as emphasizedinotherworks.16 References Amiti, Mary, Oleg Itskhoki, and Jozef Konings. 2014. “Importers, Exporters, and ExchangeRateDisconnect.” AmericanEconomicReview104(7):1942–1978. 15Gopinath and Stein (2018) provide a unified explanation for the emergence of a dominant currencyinbothtradeandfinancialdimension,emphasizingthestrategiccomplementaritybetweenthe roleofthedollarasunitofaccountandsafestoreofvalueandreinforcingthecontractionaryeffects ofdollarappreciationsinEMEs. 16For the different spillover effects of U.S. monetary policy in emerging markets and advanced economiesanditsrelationwithglobalriskperceptionsandEMEsvulnerabilities,see,forexample, Kalemli-Ozcan(2019)andIacovielloandNavarro(2019). 23

Avdjiev, Stefan, Valentina Bruno, Catherine Koch, and Hyun Song Shin. 2019. “The Dollar Exchange Rate as a Global Risk Factor: Evidence from Investment.” IMF EconomicReview67(1):151–173. Baskaya, Yusuf Soner, Julian di Giovanni, Sebnem Kalemli-Ozcan, Jose-Luis Peydro, and Mehmet Fatih Ulu. 2017. “Capital Flows and the International Credit Channel.” JournalofInternationalEconomics108(S1):S15–S22. Boz, Emine, Camila Casas, Federico J. Diez, Gita Gopinath, Pierre-Olivier Gourinchas,and MikkelPlagborg-Moller. 2018. “DominantCurrencyParadigm.” AmericanEconomicReviewForthcoming. Bruno, Valentina, Se-Jik Kim, and Hyun Song Shin. 2018. “Exchange Rates and the Working Capital Channel of Trade Fluctuations.” AEA Papers and Proceedings 108:531–536. Bruno, Valentina and Hyun Song Shin. 2015. “Cross-Border Banking and Global Liquidity.” ReviewofEconomicStudies82(2):535–564. Caballero, Julian, Andres Fernandez, and Jongho Park. 2019. “On Corporate Borrowing, Credit Spreads and Economic Activity in Emerging Economies: An EmpiricalInvestigation.” JournalofInternationalEconomics118(C):160–178. Carriere-Swallow, Yan, Bertrand Gruss, Nicolas E. Magud, and Fabian Valencia. 2016. “MonetaryPolicyCredibilityandExchangeRatePass-Through.” IMFWorkingPaperWP/16/240. Castellares, Renzo. 2017. “Condiciones de Mercado y Calidad como Determinantes delTraspasodelTipodeCambio.” RevistaEstudiosEconomicos (33):29–41. Cespedes, Luis Felipe, Roberto Chang, and Andres Velasco. 2004. “Balance Sheets andExchangeRatePolicy.” AmericanEconomicReview94(4):1183–1193. Cheng, Ing-Haw and Wei Xiong. 2014. “Financialization of Commodity Markets.” AnnualReviewofFinancialEconomics6:419–441. 24

Eaton, Jonathan and Samuel Kortum. 2001. “Trade in Capital Goods.” European EconomicReview45(7):1195–1235. Fleming, J. Marcus. 1962. “Domestic Financial Policies Under Fixed and Floating ExchangeRates.” IMFStaffPapers9:369–379. Giuliano, Fernando and Emiliano Luttini. 2019. “Import Prices and Invoice Currency:EvidencefromChile.” BISWorkingPapers784. Goldberg, Linda and Cedric Tille. 2009. “Macroeconomic Interdependence and the InternationalRoleoftheDollar.” JournalofMonetaryEconomics56(7):990–1003. Gopinath, Gita. 2016. “The International Price System.” Jackson Hole Symposium Proceedings. Gopinath, Gita and Jeremy C. Stein. 2018. “Banking, Trade, and the Making of a DominantCurrency.” Unpublished. Ha, Jongrim, M. Marc Stocker, and Hakan Yilmazkuday. 2019. “Inflation and ExchangeRatePass-Through.” PolicyResearchWorkingPaperWPS8780. Hofmann, Boris, Ilhyock Shim, and Hyun Song Shin. 2017. “Sovereign Yields and theRisk-takingChannelofCurrencyAppreciation.” BISWorkingPapers538. Iacoviello, Matteo and Gaston Navarro. 2019. “Foreign Effects of Higher U.S. InterestRates.” JournalofInternationalMoneyandFinance95:232–250. Itskhoki, Oleg and Dmitry Mukhin. 2019. “Exchange Rate Disconnect in General Equilibrium.” Unpublished. Johnson, Robert C. 2014. “Five Facts about Value-Added Exports and Implications for Macroeconomics and Trade Research.” Journal of Economic Perspectives 28(2):119–142. Johnson, Robert C. and Guillermo Noguera. 2012. “Accounting for Intermediates: ProductionSharingandTradeinValue Added.” JournalofInternationalEconomics 86(2):224–236. 25

———. 2017. “A Portrait of Trade in Value-Added over Four Decades.” The Review ofEconomicsandStatistics99(5):896–911. Kalemli-Ozcan, Sebnem. 2019. “U.S. Monetary Policy and International Risk Spillovers.” JacksonHoleSymposiumProceedings. Kalemli-Ozcan, Sebnem, Herman Kamil, and Carolina Villegas-Sanchez. 2016. “What Hinders Investment in the Aftermath of Financial Crises: Insolvent Firms orIlliquidBanks?” ReviewofEconomicsandStatistics98(4):756–769. Kalemli-Ozcan, Sebnem, Xiaoxi Liu, and Ilhyock Shim. 2018. “Exchange Rate AppreciationsandCorporateRiskTaking.” BISWorkingPapers710. Kamps,Anette.2006. “TheEuroasInvoicingCurrencyinInternationalTrade.” ECB WorkingPaper665. Krugman,Paul.1999. “BalanceSheets,theTransferProblem,andFinancialCrises.” InternationalTaxandPublicFinance6(4):459–472. Labbe, Felipe. 2018. “Multi-Destination Exporters, Market Penetration, and the InvoicingCurrencyofTrade.” Unpublished. Lilley, Andrew, Matteo Maggiori, Brent Neiman, and Jesse Schreger. 2019. “ExchangeRateReconnect.” NBERWorkingPaper26046. Meese, Richard A and Kenneth Rogoff. 1983. “Empirical Exchange Rate Models of the Seventies: Do They Fit Out of Sample?” Journal of International Economics 14(1-2):3–24. Mundell, Robert A. 1963. “Capital Mobility and Stabilization Policy Under Fixed and Flexible Exchange Rates.” The Canadian Journal of Economics and Political Science4(29):475–485. Nickel, Stephen J. 1981. “Biases in Dynamic Models with Fixed Effects.” Econometrica49(6):1417–1426. 26

Obstfeld, Maurice and Kenneth Rogoff. 2000. “The Six Major Puzzles in International Macroeconomics: Is There a Common Cause?” In NBER Macroeconomics Annual2000,editedbyBenBernankeandKennethRogoff.MITPress,339–390. Shin, Hyun Song. 2018. “Exchange Rates and Monetary Policy Frameworks in Emerging Market Economies.” ECB Conference on Monetary Policy: Bridging ScienceandPractice. 27

A Data ThedatasetincludesquarterlydataforArgentina,Brazil,Chile,Colombia,Indonesia,Korea,Malaysia,Mexico,Peru,Philippines,SouthAfrica,Thailand,Turkey,and the United States. The sample periods vary across countries. They are: Argentina, Brazil,Mexico,Philippines,SouthAfricaandUnitedStates,1996:Q1-2018:Q4;Turkey, 1996:Q2-2018:Q3;Malaysia,1996:Q4-2018:Q3;ColombiaandPeru,1997:Q1-2018:Q3; KoreaandThailand,1998:Q1-2018:Q3;Chile,1999:Q2-2018:Q3;andIndonesia,2004Q2- 2018Q3. U.S.RealGDP,U.S.RealInterestRate,VIX,andBroadU.S.Dollar:allthedata are from FRED. The U.S. real interest rate is measured by the interest rate on the 2-yearU.S.TreasurybondminusameasureofU.S.expectedinflation. Real GDP, Real Investment, Real Exports, and Real Imports: all the data are from national accounts, deflated by each own deflator and seasonally adjusted usingARIMAX-12. Real Credit: obtained by dividing nominal credit to the non-financial private sectorbytheCPIandseasonallyadjustedusingARIMAX-12.ForArgentina,Brazil, Chile,Indonesia,Korea,Malaysia,Mexico,SouthAfrica,Thailand,andTurkey,nominal credit to the non-financial sector is obtained from the BIS at http://www.bis. org/statistics/totcredit.htm.ForColombia,Peru,andPhilippines,nominalcredit to the non-financial sector is obtained from each country’s central bank. CPI is obtainedfromnationalstatisticalagencies. RealInterestRate:thecountryspecificinterestrateintheinternationalfinancial markets, R, is measured as the sum of J. P. Morgan’s EMBI+ sovereign spread and theU.S.realinterestrate.EMBI+isacompositeindexofdifferentdollar-denominated bonds on four markets: Brady bonds, Eurobonds, U.S. dollar local markets, and 28

loans. The spreads are computed as an arithmetic, market-capitalization-weighted averageofbondspreadsovertheU.S.Treasurybondsofcomparableduration. RealExchangeRates:obtainedfromtheBISeffectiveexchangerateindicesdatabase, particularly the quarterly average of the broad indices. Real EERs are calculated as geometric weighted averages of bilateral exchange rates adjusted by relative consumer prices. The weighting pattern is time-varying, and the most recent weights are based on trade in the 2008-10 period (see broad and narrow weights in http: //www.bis.org/statistics/eer.htm). An increase in the index indicates an appreciation. Import Content of Exports: I use the VAX ratio, which represents the valueadded ratio in international trade, defined as the domestic value-added in gross exports divided by total gross exports, obtained from Johnson and Noguera (2012) andJohnsonandNoguera(2017). Share of Dollar in Trade Invoicing: for Brazil, Indonesia, Korea, Thailand, and Turkey,Iobtaindatafromnationalsources.ForMalaysiaandSouthAfrica,Iobtain the data from Kamps (2006). For Argentina and Colombia, I obtain the data from Gopinath (2016). For Chile, I obtain the data from Labbe (2018) and Giuliano and Luttini (2019). For Peru, I obtain the data from Castellares (2017). For Mexico, the shares are proxied by data provided by the Bank of Thailand for trade flows betweenMexicoandThailand. Share of Credit to Non-Financial Private Sector Denominated in Dollars: obtained by dividing nominal credit to non-financial private sector denominated in dollars divided by total nominal credit to the non-financial private sector. For the credit denominated in dollars, I obtain the data for Argentina, Brazil, Chile, Indonesia, Korea, Malaysia, Mexico, South Africa, Thailand, and Turkey from the BIS GlobalLiquidity Indicators.ForColombia, Peru,andPhilippines, nominalcreditto 29

thenon-financialsectordenominatedindollarsisobtainedfromeachcountry’scentral bank. Total nominal credit to the non-financial private sector is obtained from thesamesourcesusedintheconstructionofrealcredit. Inflation:allthedataarefromnationalstatisticalagencies. B Robustness Analysis B.1 EMEs Feedback Effects on the Dollar FIGURE B.1: Impulseresponsetoa10%dollarshock-AllowingFeedbackEffectsfromEMEs. Note:Markedblackandsolidredlinesshowpointestimatesofimpulseresponses,respectively,for thebaselinepanelVARandthepanelVARallowingforcontemporaryeffectsofEMErealvariables - output and investment - on the dollar; 68% and 95% confidence bands are depicted with darkgray and light-gray shaded areas, respectively. Bootstrap confidence bands are based on 100,000 repetitions. 30

FIGURE B.2: Impulseresponsetoa1%EMEGDPshock. Note:MarkedblacklinesshowpointestimatesofimpulseresponsesforthepanelVARallowingfor contemporary effects of EME real variables - output and investment - on the dollar; 68% and 95% confidence bands are depicted with dark-gray and light-gray shaded areas, respectively. Bootstrap confidencebandsarebasedon100,000repetitions. B.2 Pre-Global Financial Crisis Analysis FIGURE B.3: Impulseresponsetoa10%dollarshock-Pre-GFCperiod. Note:Markedblackandsolidredlinesshowpointestimatesofimpulseresponses,respectively,for thebaselinepanelVARandthepanelVARestimatedonlyforthepre-globalfinancialcrisisperiod; 68% and 95% confidence bands are depicted with dark-gray and light-gray shaded areas, respectively.Bootstrapconfidencebandsarebasedon100,000repetitions. 31

B.3 Commodity Price Effects on the Dollar FIGURE B.4: Impulseresponsetoa10%dollarshock-CommodityPricesAftertheDollar. Note: Marked black and solid red lines show point estimates of impulse responses, respectively, for the baseline panel VAR and the panel VAR including commodity prices with the dollar before commodity prices in the causal ordering; 68% and 95% confidence bands are depicted with darkgray and light-gray shaded areas, respectively. Bootstrap confidence bands are based on 100,000 repetitions. FIGURE B.5: Impulseresponsetoa10%dollarshock-CommodityPricesBeforetheDollar. Note: Marked black and solid red lines show point estimates of impulse responses, respectively, forthebaselinepanelVARandthepanelVARincludingcommoditypriceswithcommodityprices before the dollar in the causal ordering; 68% and 95% confidence bands are depicted with darkgray and light-gray shaded areas, respectively. Bootstrap confidence bands are based on 100,000 repetitions. 32