1、 The Impact of U.S.Monetary Policy on Foreign Firms Julian di Giovanni|John Rogers NO.1039 NOVEMBER 2022 The Impact of U.S.Monetary Policy on Foreign Firms Julian di Giovanni and John Rogers Federal Reserve Bank of New York Staff Reports,no.1039 November 2022 JEL classification:E52,F40 Abstract This

2、 paper uses cross-country firm-level data to explore the impact of U.S.monetary policy shocks on firms sales,investment,and employment.We estimate a sizable impact of U.S.monetary policy on the average foreign firm,while controlling for other macroeconomic and financial variables like the VIX and ex

3、change rate fluctuations that accompany U.S.monetary policy changes.We then quantify the role of international trade exposure and financial constraints in transmitting monetary policy shocks to firms,allowing for a better identification of the importance of external demand effects and the interest r

4、ate channel.We first exploit cross-country sector-level data on intermediate and final goods trade to show that greater global production linkages amplify the impact of U.S.monetary policy at the firm level.We then show that the impact varies along the firm-level distribution of proxies for firms fi

5、nancial constraints(for example,size and net worth),with the impact being significantly attenuated for less constrained firms.Key words:U.S.monetary policy spillovers,foreign firms,international production linkages,financial constraints _ Giovanni:Federal Reserve Bank of New York(email:julian.digiov

6、anniny.frb.org).Rogers:Fudan University(email:).The authors thank Wenchuan Dong,Neel Lahiri,and Yijing Ren for providing excellent research assistance.This paper presents preliminary findings and is being distributed to economists and other interested readers solely to stimulate discussion and elici

7、t comments.The views expressed in this paper are those of the author(s)and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System.Any errors or omissions are the responsibility of the author(s).To view the authors disclosure statements,visit htt

8、ps:/www.newyorkfed.org/research/staff_reports/sr1039.html.1IntroductionThe impact of U.S.monetary policy on the real economy is a long-studied topic,and one that is ofprimary importance to understand today as the Fed and other central banks have entered a globaltightening cycle.These policy actions

9、are not taken in a vacuum,and some economists,such asObstfeld(2022)and Wei(2022),have argued that there is risk of central banks dampening aggregatedemand excessively.Indeed,spillovers of U.S.monetary policy may impact foreign economies viaseveral channels independently of domestic policy actions.Th

10、is paper merges firm,sectoral and macroeconomic data for a large cross-section of countriesto quantify how international trade exposure and the financial channel of interest rate changesaffect transmission of U.S.monetary policy shocks to foreign firm activity.We study these twochannels given that t

11、he recent confluence of escalating protectionism,Covid-19,disrupted supplychains,Brexit,OFAC sanctions,corporate delistings,and geopolitical tensions has raised questionsabout whether the decades-long trend toward globalization in trade and financial markets,as wellas the rise of“megafirms”(Autor,Do

12、rn,Katz,Patterson and van Reenen,2020),is reversing.Such“end-of-globalization”considerations are important for global welfare depending on the degree towhich and channels through which shocks such as monetary policy tightening are fundamentallytransmitted.Focusing on the firm level is particularly s

13、alient given the role of“granular”firms indriving aggregate fluctuations(Gabaix,2011).Firm heterogeneity further interacts with exposureto the world economy,particularly via international trade,to play a large role in aggregate inter-national business cycle co-movement(di Giovanni,Levchenko and Meje

14、an,2014,2018,2022;Weiand Xie,2020).We begin by estimating the effect of U.S.monetary policy shocks on the change in the averageforeign firms investment-to-capital share,sales-to-capital share,and employment growth in a givencountry.Our methodology utilizes a panel regression model,which allows us to

15、 control for time-varying firm-level and macroeconomic variables,as well as a rich set of non-time varying fixed effects.The main results imply that the tightening of U.S.monetary policy has a statistically significantcontractionary effect on the change of a firms investment and sales ratios,while e

16、mployment growthalso falls,but not sufficiently to detect a statistically significant effect.Results are also economicallymeaningful.For example,a one percentage point contraction in U.S.monetary policy translates to afall in the investment ratio equivalent to sixty-five percent of the median change

17、 in the investmentratio across over the sample period.We then ask how this spillover effect varies along multiplecountry dimensions.For example,we document significant differences between emerging marketeconomies(EMEs)and advanced economies(Kalemli-Ozcan,2019).We also examine how financialaccount an

18、d trade openness at the aggregate levels affect the magnitude of U.S.monetary policytransmission to foreign firms.1We next turn to a more in-depth analysis of the impact of a firms trade exposure,both to theworld economy as well as with respect to the United States alone.To do so,we construct fourex

19、port-oriented measures of trade using cross-country sector-level data on intermediate and finalgoods trade as well as sectoral output sourced from the World Input-Output Database(WIOD)from Timmer,Dietzenbacher,Los,Stehrer and de Vries(2015).Specifically,we construct a country-sectors(i)total exports

20、-to-output ratio,(ii)final goods exports-to-output ratio,(iii)intermediateexports-to-output ratio,and(iv)export-based weighted outdegree.The latter measure captureshow important a sector is as supplier of intermediates in the production of one unit of its country-sector export partners output.We int

21、eract these variables with the monetary policy shock in thenext set of regressions,focusing on the impact on firm-level investment.The approach allows usto identify how the variation in trade exposure impacts shock spillover to the average firm withina country-sector.Given that we exploit variation

22、at the countrysectoryear level,we are able tocontrol for time-varying fixed effects at the country and/or sector level.We document that total export exposure plays a significant role in the transmission of U.S.monetary policy shocks to firm investment.Movements along the distribution of country-sect

23、orexport openness from low(bottom decile)to high(top decile)amplifies the impact of the shockby forty percent relative to the impact on the average firm.Interestingly,decomposing the totalexport-to-output ratio,we find that it is intermediate goods and services trade that drives theoverall export ex

24、posure findings,both for trade with the whole world and bilaterally with the UnitedStates.Finally,the estimated coefficient on the weighted outdegree measure is also economically andstatistically significant,indicating that its not just the importance of overall intermediate exports indriving the tr

25、ansmission of U.S.monetary policy shocks to foreign firms,but also the amplificationof demand shocks via global production linkages.The results of this external demand channel viainternational trade and production linkages is in line with recent findings in the literature usingmore aggregated data,s

26、uch as Br auning and Sheremirov(2021)and di Giovanni and Hale(2022).To provide evidence on the role of differential financial constraints on the interest rate channel,we run panel regressions interacting proxies of financial constraints(size or net worth)with theU.S.monetary policy shock.This allows

27、 us to exploit time-varying firm-level variation in theinteraction variable to identify this transmission mechanism,and thereby include an exhaustive setof time-varying fixed effects at the countrysectoryear level along with non-time varying firmfixed effects.Results show that foreign firms with low

28、er financial constraints are able to attenuatethe impact of monetary policy shocks on their investment,consistent with recent micro studies ofdomestic firms by Cloyne,Ferreira,Froemel and Surico(2020)and Ottonello and Winberry(2020).The magnitude of this effect is large.For example,moving over the i

29、nterquartile range of thefirm-net worth distribution implies that less financially constrained firms are able to attenuate theimpact of U.S.monetary policy shocks by roughly one-quarter of the impact on the mean firm.2The final set of heterogeneity regressions combines the trade exposure measures wi

30、th the finan-cial constraint proxies in order to estimate the joint impact of these channels.These regressionsyield some interesting results.First,the magnitude and significance of the trade and financial inter-action coefficients do not change dramatically when included together.Second,our quantifi

31、cationexercises imply that the dampening effect of looser financial constraints of larger firms dominatesthe amplification effect of greater trade exposure.While the trade measures are at the country-sector level,large firms tend to dominate the export market(Melitz,2003;Freund and Pierola,2015),so

32、our overall results point to these“granular”foreign firms being impacted less on net byU.S.monetary policy shocks given the channels identified in our regressions.Related LiteratureThe empirical literature on cross-border spillovers of monetary policy shocks is voluminous.Mostof this research,includ

33、ing early papers on the Global Financial Cycle,relied on aggregate data.Pioneering research on the GFC includes Rey(2013),Rey(2016),Kalemli-Ozcan(2019),Han andWei(2018),and Miranda-Agrippino and Rey(2020).Early work on spillovers from U.S.monetarypolicy shocks includes Eichenbaum and Evans(1995),Rog

34、ers(1999),Kim and Roubini(2000),Faust and Rogers(2003),and Faust,Rogers,Swanson and Wright(2003),who focused on foreigninterest rates and exchange rates in VARs.Rogers,Scotti and Wright(2014)examine the effectsof unconventional monetary policy by the Fed,BOE,ECB,and BOJ on cross-border bond yieldsan

35、d stock prices,as well as exchange rates.1Br auning and Sheremirov(2021)document that tradeplays a key role in explaining cross-country heterogeneity in the effects of U.S.monetary shocks onaggregate output,interest rates,and trade flows in a large panel of countries.Degasperi,Hong andRicco(2021)fin

36、d that a U.S.monetary policy tightening has large contractionary effects on bothadvanced and emerging economies,with financial channels dominating over demand and exchangerate channels in the transmission to real variables.On the micro side,Br auning and Ivashina(2020)examine the role of U.S.monetar

37、y policy inaffecting credit conditions of EME firms.They show that the spillover is stronger in higher-yieldingand more financially open markets and for firms with a higher reliance on foreign bank credit.Morais,Peydr o,Rold an-Pe na and Ruiz-Ortega(2019)analyze the universe of corporate loans inMex

38、ico,matched with firm and bank balance-sheet data,to identify the spillover effects of advancedeconomy monetary policy shocks.They find that a tightening of foreign monetary policy increases1See also Georgiadis(2016),who finds that the magnitude of U.S.monetary policy spillovers depends on a host of

39、receiving country characteristics,including trade and financial integration,exchange rate regime,and participation inglobal value chains;Dedola,Rivolta and Stracca(2017),who find that a surprise U.S.monetary policy tightening leadsto a dollar appreciation,decline in foreign industrial production,rea

40、l GDP,and inflation,and rise in unemploymentin a panel of advanced and emerging economies;and Kearns,Schrimpf and Xia(2019),who measure monetary policyshocks for seven advanced economy central banks and spillovers to 47 advanced and emerging market economies.Theyfind no evidence that spillovers are

41、related to real linkages such as trade flows,but some importance for exchangerate regimes,with the key country characteristic being financial openness.3the supply of credit of foreign banks to Mexican firms and that this occurs via their respective(countrys)banks.Di Giovanni and Hale(2022)examine sp

42、illovers of U.S.monetary policy shocksto sectoral stock returns.They derive a model in which firms in all countries are affected by amonetary shock,by an amount that is proportional to a firms global production linkages,and findthat the global production network plays the key role in transmitting U.

43、S.monetary policy shocksto cross-border stock returns,even conditioning on financial channel variables.In addition to our paper being related to the large literature on international spillovers of U.S.monetary policy,it is closely related to Claessens,Tong and Wei(2012)and Dao,Minoiu andOstry(2021).

44、Although neither of these papers examines U.S.monetary policy,Claessens et al.(2012)examine how the global financial crisis affected firms profits,sales,and investment,thefocus of our paper.They find that the crisis had a bigger negative effect on firms with greatersensitivity to business cycle and

45、trade developments,particularly in countries more open to trade.Dao et al.(2021)examine the relationship between real exchange rate fluctuations and firm-levelinvestment and growth.They show that real depreciation boost profits,investment,and assetgrowth of tradable sector firms that have higher lab

46、or shares and are relatively more financiallyconstrained,interpreting this finding as evidence for an“internal financing channel.”2Our paper is also related to work on trade and transport costs in international trade and macro(Obstfeld and Rogoff,2000;Anderson and van Wincoop,2004).These authors pre

47、sent evidence thattotal trade costs in rich countries are large,with an estimated ad valorem tax equivalent of about170 percent,while poor countries face even higher trade costs.More recent estimates indicate thatglobal trade costs have declined by 15 per cent between 2000 and 2018(http:/tradecosts.

48、wto.org/).Clearly,there is a lot of variation across time,countries,and goods,features of the data that weexploit.Our forensic investigation of the linkages from U.S.monetary policy shocks to cross-borderfirms investment,sales,and employment uncovers a key role for trade networks,consistent withBr a

49、uning and Sheremirov(2021)and di Giovanni and Hale(2022).Finally,our paper is also related to the literature on the investment channel of monetary policytransmission in closed economies.This literature emphasizes the importance of firm heterogeneityfor the transmission of monetary policy,with much a

50、ttention paid to“balance sheet effects.”Thebalance sheet channel broadly refers to feedback effects between the health of borrowers balancesheets,as measured for example by net worth,and investment(e.g.or output,asset prices,etc.).In this framework,financially constrained firms borrow in order to un

51、dertake productive long-termprojects.Firms finance projects by issuing claims to lenders.The cash flows associated withfirms projects are exposed to an aggregate shock that may generate fluctuations in net worth,2While putting the finishing touches on the ARC version of this paper,we became aware of

52、 contemporaneous workin progress by Arbatli-Saxegaard,Firat,Furceri and Verrier(2022).These authors also examine the cross-bordereffects of U.S.monetary policy shocks in a large panel of firm-level data.Although we do much more analysis oftrade network channels and use different measures of Fed mone

53、tary policy shocks and investment,the two papershave a similar focus.4which could in turn trigger liquidations of capital and affect investment,the price of capital,andaggregate output.Monetary policy shocks,for example,would give rise to such effects.Seminalpapers include Bernanke and Gertler(1989)

54、and Kiyotaki and Moore(1997)and more recent workby Cloyne et al.(2020),Ottonello and Winberry(2020),Caglio,Darst and Kalemli-Ozcan(2021).The rest of the paper proceeds as follows.Section 2 describes our empirical methodology.Section 3 describes the data and presents summary statistics.Section 4 prov

55、ides regression resultsthat focus on the role of country-level characteristics,while our analysis of firm-level trade andfinancial constraint heterogeneity is in Section 5.Section 6 concludes.2MethodologyWe first estimate the unconditional impact of U.S.monetary policy shocks on foreign firms annual

56、change in investment,sales,and employment.Our regression analysis follows the approaches usedin a closed-economy setting by running panel regressions,where we allow for the possibility oftracing out the dynamic impulse of endogeneous variables using local projections(Jord a,2005).The baseline regres

57、sion that estimates the average effect of monetary policy shocks on all firms is:Yfsc,t+h Yfsc,t1=+MPUSt1+Zfsc,t1+Xc,t1+fsc,t+h,(1)where f denotes a firm,s the sector and c the country.Yfsc,t+his the firm-level outcome measuredin year t+h,h=0,1,2,.,T.The firm-level outcomes are either(i)the investme

58、nt-to-lagged fixedcapital ratio(It/Kt1),(ii)the sales-to-lagged fixed capital ratio(St/Kt1),or(iii)log employment(lnEt).Given the use of annual data,our baseline is to estimate the model for h=0 only.In thiscase,the left-hand side of(1)measures either the annual change in the investment or sales sha

59、res,or annual employment growth.MPUSt1is the U.S.monetary policy shock variable from Bu,Rogersand Wu(2021)(BRW)at t 1,thus accounting for the lagged impact of monetary policy on thereal economy.As described below,the BRW shock is a measure of monetary surprises centeredon each of the eight FOMC meet

60、ings per year.To match our annual firm-level real variables,weaggregate the eight shock observations throughout the calendar year,which is customary in theliterature.This timing issue further motivates the use of a lagged monetary policy shock variableas opposed to a contemporaneous one.3If a moneta

61、ry policy tightening(loosening),MPUS 0(MPUS 0),depresses(stimulates)firms activity,we would expect that 0.We further control for other standard firm-level controls,Z,which we lag one period.Thesevariables include firm size(measured as the log of total assets),net worth,and change in the cashflow-to-

62、asset ratio.4We also include the lag of macroeconomic controls,X,which may vary at3We experimented with additional lags,but this did not yield any additional insights.4We also experiment by including firm age,Tobins Q,and other measures of firms financial health such as changesin its leverage ratio.

63、Including these regressors did not impact our results but cut the sample size in several cases5the country or global levels.These include domestic real GDP growth,change in the log nominalexchange rate against the U.S.dollar,the change in short-term domestic interest rates,and logVIX.Given the panel

64、 setup,we are also able to include a set of non-time-varying fixed effects,(e.g.,at the country,sector,or firm-level).Finally,is the error term.Given that the monetarypolicy shock is repeated across all firms in a given year,we cluster standard errors at the annuallevel and further cluster at the fi

65、rm level to control for potential autocorrelation in the errors.Equation(1)is a useful baseline specification to estimate the impact of U.S.monetary policyon the average firm in a given country.We can then“unpack”the potential heterogeneous impactsof monetary policy by allowing for to vary across mu

66、ltiple dimensions.To begin,we examinehow the impact of U.S.monetary policy on foreign firms varies across countries via simple samplesplits and interactions with country characteristics.For example,we examine whether differs be-tween emerging market economies(EMEs)and industrial countries.We also ex

67、amine how financialaccount and trade openness at the aggregate levels impact the estimates of.Role of Trade LinkagesChanges in U.S.monetary policy may impact foreign firms activity directly given the resultingcontraction/expansion of U.S.demand.This channel might be expected to have an out-sized imp

68、acton firms or sectors depending on how involved they are in international trade.Further,given theexpansion of global production networks over time,firms that are more integrated in global valuechains may be even more impacted given spillovers across countries arising from the change inU.S.monetary

69、policy.Given data limitations,we are forced to exploit trade heterogeneity at thecountry-sector level rather than firm level in our estimation.5Thus,our first extended regression specification exploits heterogeneity across country-sectorswithin a year:Yfsc,t+h Yfsc,t1=+1MPUSt1+2(Tradecs,t1 MPUSt1)+T

70、radecs,t1+Zfsc,t1+Xc,t1+fsc,t+h,(2)where Tradecsis a measure of a country-sectors trade exposure.We construct several measures ofa country-sectors exposure to demand shocks by exploiting heterogeneity in a country-sectors linksto either the world or U.S.only via exports along four possible dimension

71、s:a country-sectors(i)total exports-to-output ratio,(ii)final goods exports-to-output ratio,(iii)intermediate exports-to-output ratio,and(iv)export-based weighted outdegree.As we describe in Section 3,the weightedoutdegree measure captures how important a country-sectors output is for all other coun

72、try-sectors(for example,Italian firms do not report the age variable in our dataset).Therefore,in order to maximize samplesize we constrain the inclusion of firm-level controls in the final analysis.5See di Giovanni et al.(2022)for evidence that sales growth of firms more exposed to trade are more s

73、ensitive tochanges in world GDP.6production and thus captures the importance of a country-sector in the global value chain.6Notethat we also explored related import-based measures,consistent with the idea that the generalequilibrium impact of U.S.monetary policy shocks may also feed through to firms

74、/sectors costsvia imports,but results based on import measures were never significant so we omit for brevity.Conditional on U.S.monetary policy having a greater impact on firms in sectors that have largertrade exposure measures,we would expect that 2 0.The most stringent set of fixed effects may now

75、 vary at more granular levels along the timedimensions specifically at the countrysectoryear level since identification of the interactionterms is exploiting variation at the firmyear level.Therefore,we are able to identify differentialimpacts of financial constraints within a year along the firm di

76、stribution while controlling for time-varying countrysector characteristics or shocks.Given the literature that studies the balance sheet effect of external shocks(e.g.,exchangerate changes),we extend the estimation of(3)along several dimensions.For example,we interactother macro variables,such as c

77、hanges in the exchange rate or VIX,with monetary policy shocks,thus estimating several interaction coefficients.Further,we allow 2to vary across different cross-sections of the data,such as the country level.We consider such further“unpacking”of the interestrate channel(and trade channel)in robustne

78、ss analysis.Firm-Level and Country-Sector Trade HeterogeneityOur final specification combines the insights from regressions(2)and(3)in order to estimate therelative importance of the interest rate and trade channels.Specifically,we estimate the following:Yfsc,t+h Yfsc,t1=+1MPUSt1+2(Zfcs,t1 MPUSt1)+3

79、(Tradecs,t1 MPUSt1)+Tradecs,t1+Zfsc,t1+Xc,t1+fsc,t+h,(4)where variables are defined as above.Importantly,relative to the firm heterogeneity regressionsof(3),we cannot exploit time-varying fixed effects at the countrysectoryear level given theinclusion of the trade variables.This regression specifica

80、tion allows us to quantify the relative importance of financial constraintsand trade channels across the distribution of firms and country-sectors in our sample.We detailthis quantification exercise when presenting results below.We further experimented with moregranular specifications by interacting

81、 the firm-level Z and country-sector Trade variables with themonetary policy shocks.Besides being difficult to interpret,the triple-interaction coefficients werestatistically insignificant for the majority of specifications.83Data3.1Monetary Policy ShocksAs our baseline,we use the Bu et al.(2021)mon

82、etary policy shock series,which is plotted at theannual frequency in Figure 1.This series is derived from a two-step,partial-least squares estimationusing daily interest rate data across a wide spectrum of maturities.The general idea behindconstruction of the measure is to use Fama and MacBeth(1973)

83、two-step regressions to estimatethe unobservable monetary policy shock.The method works initially through the sensitivity ofoutcome variables to FOMC announcements.In the first step,time-series regressions are run toestimate the sensitivity of interest rates at different maturities to FOMC announcem

84、ents.This isequivalent to the asset beta in the original Fama-MacBeth method.In the second step,all outcomevariables are regressed onto the corresponding estimated sensitivity index from step one,for eachtime t.In this way,the monetary policy shock is derived as the series of estimated coefficients

85、fromthe Fama-MacBeth style second step regressions.Bu et al.(2021)scale the shock series such thatit has a one-to-one contemporaneous effect on the 2-year Treasury Bill rate.9The Bu et al.(2021)shock measure has three appealing features,which together distinguishit from other shock series in the lit

86、erature.First,by using the full maturity spectrum of interestrates,this series stably bridges periods of conventional and unconventional monetary policy.Second,the shock is largely devoid of the central bank information effect,the notion that monetary policyannouncements,in addition to providing a p

87、ure monetary surprise,also reveal information regardingthe central banks future macroeconomic outlook(Nakamura and Steinsson,2018).And third,theBu et al.(2021)shock series is largely unpredictable from available information,including BlueChip forecasts,“big data”measures of economic activity,news re

88、leases and consumer sentiment.10For robustness,we also examine two alternative U.S.monetary policy shock series.The first isthe policy news shock of Nakamura and Steinsson(2018),which we depict in Figure A1.The authorsconstruct their measure using changes in five interest rate futures:the Fed Funds

89、future for currentmonth and the month of the next FOMC meeting,and the 3-month Eurodollar futures at horizons oftwo,three,and four quarters.The policy news shock is the first principle component of the changein these five interest rate futures over a 30-minute window around scheduled FOMC announceme

90、nts.Our second robustness check uses Swanson(2021)s forward guidance shock,depicted in Figure A2.9To provide further meaning,Bu et al.(2021)regress contemporaneous changes in interest rates of differentmaturities on the shock.The response coefficient reaches its maximum at the 2-year interest rate(n

91、ormalized tobe 1.0).The response of the 5-year interest rate is of comparable magnitude,also large and significant.Responsecoefficients for all other maturities(3-mo.,6-mo.,1-yr,10-yr and 30-yr)are significant but smaller.Thus,both theshort and long ends of the yield curve respond to the BRW shock b

92、y less than do the 2-and 5-yr rates.This is similarto the experiment in G urkayanak,Sack and Swanson(2005),who show that the long rate responds relatively moreto their estimated“path factor”while the short rate responds relatively more to the“target factor.”10See,for example,Ramey(2016),Miranda-Agri

93、ppino(2016),and Bauer and Swanson(2020)for critiques of earliermonetary policy shock series that exhibited predictability.9Figure 1.U.S.Monetary Policy ShocksNotes:This figure plots the annual aggregate of the pure monetary policy shock constructed by Bu et al.(2021)(updated March 4,2021).A noticeab

94、le difference in both of these series relative to BRW is the large negative values in 2001,almost all of which occurred after the 9/11 terrorist attack.11Finally,we also examine the shockBu et al.(2021)constructed for the ECB(Figure A3)to examine robustness to the precise sourceof the monetary polic

95、y impulse.Given that we run regressions using annual firm-level data,we must aggregate the monetarypolicy to the annual level as well.This aggregation has the potential of netting out positive andnegative monetary policy innovations within a year and thus may bias the estimated impact ofmonetary pol

96、icy on investment towards zero.Therefore,for identification we will rely on thepersistent nature of monetary policy action within a year as well as the lagged effect of monetarypolicy on the real economy.11Note that the scales of the policy news shock and the forward guidance shock are also arbitrar

97、y.Nakamura andSteinsson(2018)rescale their series such that its effect on the 1-year nominal Treasury yield is equal to one.Swanson(2021)offers one natural way to interpret his forward guidance shock:a 25bp change in the expected federal fundsrate one year ahead,which would be very large by historic

98、al standards,about 4.4 standard deviations.Applying thatto his estimates suggests that a forward guidance surprise of this magnitude would raise the 2-Yr Treasury bill rateby around 20bp.Concerning values in 2001,Cochrane and Piasezzi(2002)argue that it is problematic to interpretmovements in intere

99、st rates around September 11,2001 as a shock versus an expected movement.Their measure,like ours,does not exhibit this feature.103.2Firm DataWe source firm-level data from Worldscope for a large cross-section of countries and sectors spanningthe time period 1995-2019 at the annual level.These data a

100、re reported for publicly listed firms,soare skewed towards covering medium-size to large firms.This firm coverage is similar to the onein studies of the impact of monetary policy on firm outcomes in the United States that rely onCompustat data,and studies in an international setting such as Claessen

101、s et al.(2012).Our cleanedsample covers twenty countries,which we choose based on the availability of a sufficient numberof firms over the whole time period(at least 5,000 firm-year observations per country)and anapproximately equal split between emerging market economies(EMEs)and industrial countri

102、es.12We further constrain the final regression sample to firms with at least five years of data.Table A1 presents summary statistics for the firm-level outcome variables,explanatory variables,and controls we experimented with and that are commonly used in the literature.The three outcomevariables ar

103、e:(i)the investment-to-(lagged)fixed capital ratios,where we follow Cloyne et al.(2020)and define fixed capital by net property,plant and equipment,sales-to-(lagged)fixed capital ratios,and employment growth.We winsorize the data at the 1%level to clean outliers.13The summary statistics indicate sub

104、stantial cross-sectional heterogeneity in the three outcomevariables,with the medians approximately centered around zero.Turning to the firm-level ex-planatory variables,there is also a good deal of cross-sectional heterogeneity.We focus on two keyfirm-level variables both because they proxy for fin

105、ancial constraints and offer maximal coverage:size and net worth.Size is defined as the logarithm of total assets while net worth is the log ofthe difference between total assets and total liabilities.In looking at Table A1,we see that thesevariables are quite skewed,which is not surprising given th

106、e granular nature of many firm-levelcharacteristics,such as the size distribution(Gabaix,2011).This also holds true for other possibleproxies for size such as employment and the age distribution.Furthermore,the absolute size offirms along the distribution differs across country size,such that there

107、is a positive correlationbetween the largest firms within a country and country size(di Giovanni and Levchenko,2012).We take this cross-country difference in distributions into account before running regressions bynormalizing both firm size and net worth.Specifically,for each country-year we normali

108、ze eachvariable around its mean,so that the distribution is centered at zero.This normalization ensuresthat we do not confound estimates that vary across the firm distribution for country differences inour regressions below.1412The country sample includes Australia,Brazil,Canada,China,France,Germany

109、,India,Indonesia,Italy,Japan,South Korea,Malaysia,Poland,Russia,Sweden,Taiwan,Thailand,Turkey,United Kingdom,and Vietnam.13One exception is the sales ratio,which we winsorize at the 5%level.14The inclusion of country or country-sector fixed effects would also help assuage this concern.However,given

110、thatwe run interaction regressions with firm characteristics and the monetary policy shock,it is best to first demean thefirm variables.113.3Trade DataWe use the 2013 edition of the World Input-Output Database(WIOD)from Timmer et al.(2015)toconstruct trade exposure measures at the country-sector lev

111、el.This database contains informationon bilateral trade flows in final and intermediate goods and services for 40 countries and the rest ofthe world as well as 35 sectors.15The database also contains country-sector value added and grossoutput measures.The database begins in 1995 and ends in 2011.We

112、opt for this database ratherthan the more recent version(which covers 2000-2014)given that there is interesting monetarypolicy variation in the late 1990s that we would like to include.The downside to this approach isthat we are forced to fill in trade data for 2011 onward in order to exploit the ad

113、ditional eight yearsof monetary policy shocks and firm-level data we have.However,given that world trade has beenstagnant since the Great Financial Crisis(Antr as,2021)and the relative stability of the the worldI-O matrix,we are not overly concerned about potential bias this extrapolation might crea

114、te.We construct four measures of trade exposure at the country-sector level.These are meantto capture exposure to demand shocks resulting from U.S.monetary policy shocks.The first is acountry-sectors total exports-to-output ratio.We next break this measure into(i)the final goodsexports-to-output rat

115、io,and(ii)the intermediate goods-to-output ratio.16Our final measure is anexport-based weighted outdegree.This variable captures how important is a country-sectors outputthat it exports for all of its customers(foreign country-sectors)production.More specifically forthis fourth measure,letmi,nj=Sale

116、sminjOutputnjbe country-sector mi sales to country-sector nj deflated by njs output.Then the export-basedweighted outdegree for country-sector pair mi is defined as:WtOutdegmi=NXn=mJXj=1mi,nj.Note that the weighted-outdegree measure only captures the first-order importance of a country-sector as a s

117、upplier in global production given that it does not measure the importance of country-sector mis customers in supplying their intermediate goods further downstream in global produc-tion process.However,given the relative high level aggregation of the WIOD and the sparsityin international linkages,th

118、e cross-sectional heterogeneity of the first-order linkages are sufficientto capture the relative importance of a country-sector in the global production network.Indeed,the distribution of these weighted-outdegree measures is quite skewed and follows a power law(seedi Giovanni and Hale,2022,for exam

119、ple).15We use the rest-of-the-world(ROW)variables for three countries that are missing data:Malaysia,Thailand andVietnam.Given the sparse data for Asia,the ROW data cover many of the smaller Asian economies,so we viewthis approximation to be reasonable.If anything,this assumption will bias against o

120、ur regressions finding any tradeeffects as we are killing some cross-sectional heterogeneity by imputing the same numbers for several country-sectors.16Note that when we write“goods”these might be services depending on the export sector.12We construct all the trade measures with respect to world tra

121、de and bilateral U.S.trade only.With the trade flows measure,for example,the outdegree for each country m would only besummed over sectors in country n=United States.We consider both sets of measures in orderto help tease out both direct and indirect trade channels that would impact foreign firms gi

122、venboth a U.S.response in demand but also U.S.monetary policy shocks that directly impact othercountries demand and which spill over to their import demand.17Tables A3 and A4 presentsummary statistics of these measures for the year 2000,where we calculate statistics across ourcountry sample in a giv

123、en sector,for world trade and U.S.-trade,respectively.There is considerableheterogeneity both across sectors(comparing the Mean columns)and countries within a sector(comparing the St.Dev.columns)according to all trade exposure measures.3.4Other Macro ControlsTable A6 presents summary statistics acro

124、ss countries and over time for the annual macroeconomicdata we use:(i)the log of the CBOE Volatility Index(VIX),(ii)real GDP growth in domesticcurrency,(iii)the percentage change of the local currency-to-U.S.dollar nominal exchange rate,(iv)the change in the domestic short-term rate,(v)one minus the

125、 Fern andez,Klein,Rebucci,Schindlerand Uribe(2016)index of financial account repression(Fin.Openness),and(vi)the(exports plusimports)-to-GDP ratio(Trade/GDP).All financial series are calculated using the annual averageof the underlying variable,while macroeconomic and trade data are end-of-year seri

126、es.4Baseline Results:The Role of Aggregate FactorsWe begin with a set of baseline regressions that provide an interesting first look at the data andpoint to potential channels through which U.S.monetary policy may have differing effects on foreignfirms.We also show that results are robust to several

127、 checks including the split between emergingand developed economies and choice of monetary policy shock.In order to better identify potentialchannels and quantify their relative importance,we then move on to exploiting cross country-sectorand/or firm-level heterogeneity in the following section.4.1B

128、aseline SpecificationWe begin by estimating regression specification(1)for h=0.Table 1 presents our baseline resultsfor investment,sales,and employment.For each variable,regressions include either countrysectoror firm fixed effects.The negative coefficient on the MPUSshock variable indicates that a

129、surprisemonetary policy tightening(MPUS 0)is associated with fall in investment,as seen in columns(1)-(2),or sales(columns(3)-(4)in the following year.These results are robust across both sets of17See di Giovanni and Hale(2022)for a structural econometric analysis of this problem.13fixed effects and

130、 statistically significant at the one-percent level for the more stringent set of firmfixed effects.Turning to the employment growth regressions in columns(5)-(6),the coefficient onMPUSis also negative,but insignificant at standard confidence levels.Quantitatively,the impact of a U.S.monetary policy

131、 shock is sizeable for both foreign firmsinvestment and sales.For example,in the regressions with firm fixed effects a one-percentage pointtightening(which would be very large by historical standards)implies that the investment ratiofalls by 0.13 percentage points in the following year.This is large

132、 relative to the median changein the investment ratio across all firms over the sample period,which is 0.2 percentage points(seeTable A1).A similar calculation holds for sales,with the sales ratio falling by 1.1 percentage pointsfollowing a one hundred basis point tightening.This is almost four time

133、s as large as the mediansales ratio change across firms in the sample(0.3 percentage points).The estimated coefficients on firm-level controls are consistent with those reported in the in-vestment literature.As seen in Table 1,both cash flow and net worth enter positively,while sizeis negative.Dao e

134、t al.(2021)also find negative and significant effects of firm size(measured byemployment)on investment in a panel of firms similar to ours.18Turning to the macro controls,the VIX is negatively correlated with firm activity,as are changes in the domestic interest rate(though not robustly).Domestic re

135、al GDP growth tends to be negatively correlated with nextperiods firm investment and sales changes when including firm-level fixed effects,but is positivelycorrelated with employment growth.Meanwhile,changes in the nominal exchange rate are typicallynot statistically significant,only weakly so for t

136、he investment regression in column(2).4.2Cross-Country Heterogeneity and RobustnessEffects for Industrial versus Emerging Market EconomiesTable 2 presents estimates of the baseline regression with firm fixed effects separately for industrialand EME country samples.Examining the coefficients on MPUS,

137、we see that the results are verysimilar to the baseline regressions.Interestingly,and perhaps not surprisingly,the monetary policyshock coefficients(the only ones we report,to save space)are larger in absolute value for theemerging market economies.The coefficient differences across country samples

138、are not statisticallydistinguishable given their overlapping confidence intervals,however.Leave-One-Out AnalysisThe large,expansionary U.S.monetary policy shock in 2009(Figure 1)and resurgence of globalinvestment coming out of the Great Recession motivates a sensitivity check of the baseline results

139、18The authors further control for size,leverage,Tobins Q,and sales growth.We have also explored including thesevariables.While doing so cuts sample size substantially,our baseline result does not change.Gulen and Ion(2016),who examine political uncertainty and investment,control for Tobins Q,cash fl

140、ow,and sales growth in regressions forU.S.firm-level investment and find all of these controls to be positive and significant,consistent with our regressions.14Table 1.Effect of U.S.Monetary Policy Shocks on Firms Investment,Sales,and Employment:Baseline Estimates(Investmentt/FixAssetst1)(Salest/Fix

141、Assetst1)Employment Growtht(1)(2)(3)(4)(5)(6)MPUSt1-0.134b-0.161a-1.119a-1.302a-0.020-0.030(0.051)(0.054)(0.391)(0.402)(0.030)(0.027)(CF/TA)t10.0005b0.001a0.0012E-05-3E-05-8E-05(0.0002)(0.0002)(0.001)(0.001)(0.0001)(0.0001)Sizet1-0.007a-0.081a-0.008-0.362a-0.027a-0.103a(0.002)(0.010)(0.021)(0.060)(0

142、.002)(0.008)Net Wortht10.0030.002-0.062a-0.212a0.022a0.047a(0.002)(0.004)(0.018)(0.032)(0.002)(0.003)ln(RGDPD)t1-0.0002-0.004b-0.012-0.039a0.003b0.002c(0.001)(0.002)(0.010)(0.013)(0.001)(0.001)ln(VIXt1)-0.076a-0.104a-0.622a-0.819a-0.026-0.024(0.024)(0.018)(0.203)(0.177)(0.017)(0.017)ln(NXR)t1-0.051-

143、0.099c-0.478-0.794-0.029-0.044(0.042)(0.051)(0.455)(0.493)(0.028)(0.031)IntRateDt1-0.375b-0.201-3.456c-2.295-0.082-0.042(0.179)(0.216)(1.785)(2.041)(0.119)(0.132)Observations374,864374,360374,687374,179256,108254,414R20.0050.0570.0090.1060.0220.176Countrysector FEYesNoYesNoYesNoFirm FENoYesNoYesNoYe

144、sNotes:This table presents firm-level panel regression results based on the estimation of regression(1)for the changein the investment-to-fixed capital ratio(columns 1 and 2),the change in the sales-to-fixed capital ratio(columns 3and 4),and employment growth(columns 5 and 6).The sample uses firms w

145、ith at least five years of observations over1995-2019.All regressors are lagged one period,where MPUSis the monetary policy shock from Bu et al.(2021),CF/TA is a firms cash flow-to-total assets ratio,Size is the within country-year measure of a firms size basedon the log of total assets,Net worth is

146、 the within country-year measure of a firms net worth based on the log ofnet worth(assets minus liabilities),RGDPDis a countrys real GDP,NXR is a countrys nominal exchange rateagainst the U.S.dollar,VIX is the CBOE Volatility Index,and IntRateD is a countrys short-term interest rate(annual average).

147、We include fixed effects at various levels of disaggregation.Standard errors are double clusteredat the firm and year level,whereaindicates significance at the 1%level,bat the 5%level,andcat the 10%level.for possible outliers.In Figure A4,we display the estimated obtained by running regression(1)mul

148、tiple times while omitting one years observations at a time.As we see,every estimate isnegative and significant.Leaving out financial crisis years(2009-11,which implies leaving out the2008-10 shocks)weakens the negative effect of U.S.monetary policy on global investment,but noneof the coefficients i

149、s significantly different from any of the others throughout the sample.Alternative Measures of Monetary Policy ShocksTable A7 shows how the baseline results are affected by using three alternative measures of15Table 2.Effect of U.S.Monetary Policy Shocks on Firms Investment,Sales,and Employment:Base

150、line Estimates for EMEs and Industrial Countries(Investmentt/FixAssetst1)(Salest/FixAssetst1)Employment GrowthtIndustrialEmergingIndustrialEmergingIndustrialEmerging(1)(2)(3)(4)(5)(6)MPUSt1-0.143b-0.168a-0.978c-1.515a-0.007-0.053(0.064)(0.052)(0.484)(0.459)(0.029)(0.035)Observations207,263167,097207

151、,155167,024152,789101,625R20.0610.0530.1010.1140.1990.151Macro controlsYesYesYesYesYesYesFirm controlsYesYesYesYesYesYesFirm FEYesYesYesYesYesYesNotes:This table presents firm-level panel regression results based on the estimation of regression(1)for the samplesplit between emerging market economies

152、 and industrial countries for the change in the investment-to-fixed capitalratio(columns 1 and 2),the change in the sales-to-fixed capital ratio(columns 3 and 4),and employment growth(columns 5 and 6).The sample uses firms with at least five years of observations over 1995-2019.All regressors arelag

153、ged one period,where MPUSis the monetary policy shock from Bu et al.(2021).We include lagged firm andmacroeconomic variables as in the baseline estimation in Table 1,and firm-level fixed effects.Standard errors aredouble clustered at the firm and year level,whereaindicates significance at the 1%leve

154、l,bat the 5%level,andcatthe 10%level.monetary policy shocks:the Nakamura and Steinsson(2018)and Swanson(2021)measures for theFed and the Bu et al.(2021)shock for the ECB.19We also include lagged changes in either the2-yr.or 5-yr.U.S.Treasury bill rate to control for the more general effects of U.S.i

155、nterest ratechanges on foreign firm investment.The first two columns indicate that the baseline results usingthe BRW shock are robust,with the coefficients on MPUSrising in all cases and even becomingstatistically significant in the employment regressions.In columns(3)-(8)we replace the BRWshock wit

156、h one of the alternatives.Results using the forward guidance shock are similar to thebaseline findings:U.S.monetary policy tightenings significantly reduce foreign firm investment andsales growth.With the policy news shock,however,the coefficient estimates are insignificantlydifferent from zero,like

157、ly reflecting the“central bank information effect”which is the subject ofNakamura and Steinsson(2018).The final two columns indicate that the ECB monetary policyshock is insignificantly different from zero.Notice that in all regressions the coefficients on laggedchanges in U.S.T-bill rates are posit

158、ive.This is consistent with higher U.S.aggregate demand,andthus interest rates,spilling over to increase investment,sales,and employment by foreign firms.DynamicsAlthough our primary objective is to exploit the rich cross-section of firms,sectors,and countries19These were computed using the same met

159、hod described above for the Fed.16in our annual data set,we also estimate dynamic effects of U.S.monetary policy shocks using Jord a(2005)s local projections regressions.We re-estimate equation(1)for h=0,.,3 and display thecumulative impulse responses of the investment share,sales share,and log empl

160、oyment in thethree panels of Figure A5.The results indicate that U.S.monetary policy tightenings have fairlypersistent negative effects on the levels of these variables,but that the initial response(h=0)thatwe estimate in our static regression captures the largest impact.The results for employment a

161、renot statistically significant,however.Country-Level Trade and Financial OpennessBefore moving on to more micro identification,we run a set of regressions to examine how tradeand financial openness at the country level affect the transmission of U.S.monetary policy shocksat the firm level.We estima

162、te these regressions by interacting measures of a countrys total trade toGDP and its financial openness,as described in Section 3,with the monetary policy shock variable.As Table A8 shows,the coefficient on the U.S.monetary policy shock is largely unaffected relativeto the baseline estimation of reg

163、ression(1).Focusing on the firm-level fixed effect regressions incolumns(2),(4),and(6)a common theme emerges:the impact of U.S.monetary policy shocksis greater for countries that are more open to trade,20while being attenuated for more financiallyopen countries.However,the coefficients on the intera

164、ctions with the trade and financial opennessvariables are insignificant for investment and sales,but significant for employment.As we showbelow,the influence of openness is manifest not so much at the country level but by sector.5Firm Heterogeneity ResultsTo gauge the importance of the external dema

165、nd and interest rate channels of U.S.monetary policytransmission abroad,we next focus on heterogeneity at a more granular level,with a particular focuson international trade exposure and proxies for firms financial constraints.We begin by extendingthe baseline specification to allow for heterogeneou

166、s effects of international trade linkages at thecountry-sector level,and report results for different specifications of regression(2).We then utilizeproxies for firm-level financial constraints and report results for different specifications of regression(3).Finally,we combine the country-sector and

167、 firm-level data to examine the impact of trade andthe interest rate channel jointly by reporting results for different specifications of regression(4).Forthe sake of brevity,we present tables for the investment regressions in the main text and relegatethe sales and employment regressions to Appendi

168、x B.20Note that results are similar if we use the exports-to-GDP ratio rather than the total trade-to-GDP ratio.175.1Trade ExposureTable 3 reports OLS estimates for regression(2)for the change in the investment share.We leaveout time-varying fixed effects in order to retain the main coefficient on M

169、PUS,but do include firmfixed effects in all specifications.Columns(1)-(4)use the trade measures based on global trade,while columns(5)-(8)use only trade flows with the United States.The coefficient on the non-interacted U.S.monetary policy shock variable remains negative andstrongly significant in a

170、ll specifications.Turning to the coefficient on the total exports-to-outputratio(TotExp/Output),we see that country-sectors that are more dependent on trade with boththe world or the U.S.alone are relatively more affected by U.S.monetary policy shocks.We dissectthis result further by examining wheth

171、er the type of trade matters and find that it does.First,while the coefficients on the final goods exports-to-output ratio(FinExp/Output)are negative,they are tiny and statistically insignificant.In contrast,when we turn to the intermediate goodsexports-to-output ratio(IntExp/Output)regressions,the

172、coefficients are negative and significantfor both global and U.S.-only trade.This indicates the key role of intermediate goods trade intransmitting monetary policy shocks to firm investment.Finally,the coefficient interaction withthe export weighted outdegree(WtOutdeg),which captures the importance

173、of a country-sectoras a supplier to other country-sectors production,is also negative and significant,both for globaltrade and U.S.bilateral trade only.Table 4 extends the regressions to include time-varying fixed effects(thus eliminating the maineffect of MPUS)by including countryyear fixed effects

174、.The advantage of including these fixedeffects is that we are able to control for time-varying country-level characteristics and shocks,such as overall trade openness or unobserved aggregate shocks,which may be correlated with U.S.monetary policy shocks.Looking across columns(1)-(8),we see that the

175、coefficients on the tradevariables are similar to those reported in Table 3.If anything,the coefficients on the interactionterms are larger(in absolute terms)and tend to be more statistically significant.21Before quantifying the importance of trade in transmitting monetary policy shocks to firms tha

176、tare exposed differently,it is worth commenting on the regression results for sales and employment.Tables A9 and A10 present the results for the regressions without and with time-varying fixedeffects,respectively.The coefficients on the trade interaction terms are generally insignificant inregressio

177、ns for both variables,whether or not we include time-varying country fixed effects.Quantifying the Trade ChannelWe exploit the country-sector distribution of the(normalized)trade measures in order to quantify21We also experimented with including sectoryear fixed effects and obtained similar results

178、as our baseline OLSregressions.Regressions including both countryyear and sectoryear fixed effects yield similar coefficients as ourmain regressions,though the majority of the coefficients are no longer significant.This finding is not surprising giventhat the inclusion of both country and sector tim

179、e-varying fixed effects greatly reduces degrees of freedom.18Table 3.Effect of U.S.Monetary Policy Shocks on Firms Investment:The Importance of TradeIntegration,Non-Time-Varying FE Estimates(Investmentt/FixAssetst1)Global TradeU.S.Trade(1)(2)(3)(4)(5)(6)(7)(8)MPUSt1-0.160a-0.161a-0.160a-0.161a-0.161

180、a-0.161a-0.161a-0.161a(0.054)(0.054)(0.054)(0.054)(0.054)(0.054)(0.054)(0.054)MPUSt1?TotExpOutput?t1-0.089b-0.291c(0.038)(0.156)MPUSt1?FinExpOutput?t1-0.008-0.012(0.054)(0.148)MPUSt1?IntExpOutput?t1-0.149b-0.462c(0.061)(0.259)MPUSt1 WtOutdegt1-0.025b-1.127b(0.012)(0.449)Observations374,360374,360374

181、,360374,360374,360374,360374,360374,360R20.0580.0570.0580.0580.0580.0570.0580.058Macro ControlsYesYesYesYesYesYesYesYesFirm ControlsYesYesYesYesYesYesYesYesFirm FEYesYesYesYesYesYesYesYesNotes:This table presents firm-level panel regression results based on the estimation of regression(2)for the cha

182、ngein the investment-to-fixed capital ratio,where we interact different measures of country-sectors trade integrationwith the monetary policy shock.Columns(1)-(4)use trade measures based on country-sector exports with the world,while columns(5)-(8)use U.S.-only exports data.The country-sectors trade

183、 measure include(i)total trade-to-output ratio(TotExp/Output),(ii)final goods trade-to-output ratio(FinExp/Output),(iii)intermediate goodstrade-to-output ratio(IntExp/Output),and(iv)the weighted outdegree(WtOutdeg).The sample uses firms withat least five years of observations over 1995-2019.All regr

184、essors are lagged one period,where MPUSis the monetarypolicy shock from Bu et al.(2021),Size is the within country-year measure of a firms size based on the log of totalassets,and Net worth is the within country-year measure of a firms net worth based on the log of net worth(assetsminus liabilities)

185、.We include lagged firm and macroeconomic variables as in the baseline estimation in Table 1,andfixed effects at various levels of disaggregation.Standard errors are double clustered at the firm and year level,whereaindicates significance at the 1%level,bat the 5%level,andcat the 10%level.their impo

186、rtance in transmitting U.S.monetary policy shocks to firm investment in Tables 3 and4.First note that the normalized versions of these variables are constructed around a mean ofzero in a given country and year.This implies that the distribution we exploit for the regressionsis centered around zero(s

187、ee Table A5).Therefore,the mean-firms trade variables are equal tozero and the impact of the U.S.monetary policy shock on firm investment is simply equal to thenon-interacted coefficient on MPUS.Indeed,this is confirmed by comparing the coefficients in thefirst row of Table 3 to those of the firm-le

188、vel fixed effects in column(2)of Table 1.We take two approaches to examining the relative importance of trade exposure on monetarypolicy transmission across firms.The first is to compute the impact of MPUSon firms across the19Table 4.Effect of U.S.Monetary Policy Shocks on Firms Investment:The Impor

189、tance of TradeIntegration,Time-Varying FE Estimates(Investmentt/FixAssetst1)Global TradeU.S.Trade(1)(2)(3)(4)(5)(6)(7)(8)MPUSt1?TotExpOutput?t1-0.120a-0.387b(0.041)(0.162)MPUSt1?FinExpOutput?t1-0.011-0.013(0.054)(0.150)MPUSt1?IntExpOutput?t1-0.208a-0.820b(0.073)(0.348)MPUSt1 WtOutdegt1-0.031b-1.614a

190、(0.012)(0.570)Observations374,359374,359374,359374,359374,359374,359374,359374,359R20.0690.0690.0690.0690.0690.0690.0690.069Countryyear FEYesYesYesYesYesYesYesYesFirm ControlsYesYesYesYesYesYesYesYesFirm FEYesYesYesYesYesYesYesYesNotes:This table presents firm-level panel regression results based on

191、 the estimation of regression(2),with time-varying fixed effects,for the change in the investment-to-fixed capital ratio,where we interact different measuresof country-sectors trade integration with the monetary policy shock.Columns(1)-(4)use trade measures basedon country-sector exports with the wo

192、rld,while columns(5)-(8)use U.S.-only exports data.The country-sectorstrade measure include(i)total trade-to-output ratio(TotExp/Output),(ii)final goods trade-to-output ratio(Fin-Exp/Output),(iii)intermediate goods trade-to-output ratio(IntExp/Output),and(iv)the weighted outdegree(WtOutdeg).The samp

193、le uses firms with at least five years of observations over 1995-2019.All regressors arelagged one period,where MPUSis the monetary policy shock from Bu et al.(2021),Size is the within country-yearmeasure of a firms size based on the log of total assets,and Net worth is the within country-year measu

194、re of a firmsnet worth based on the log of net worth(assets minus liabilities).We include lagged firm variables as in the baselineestimation in Table 1,and fixed effects at various levels of disaggregation.Standard errors are double clustered atthe firm and year level,whereaindicates significance at

195、 the 1%level,bat the 5%level,andcat the 10%level.interquartile range(IQR)of the country-sectors trade exposure measures.22Second,given thatthe trade exposure measures are skewed,we also look at the differential impact between the topand bottom deciles of the distribution.To be clear,as we exploit di

196、fferences across country-sectorpairs,it is only possible to interpret the following exercises for a representative firm in a givencountry-sector,irrespective of its trading behavior or other firm-level characteristics.Our calibration results in Table 5 are based on the coefficients in Table 4 in ord

197、er to controlfor the more conservative set of fixed effects.Moving from the bottom quartile to the top quartilecountry-sector in the world total export-to-output ratio distribution shows that greater exportexposure amplifies the spillover effects of U.S.monetary policy shocks.Specifically,the moving

198、along the IQR implies that a one percentage point surprise contraction in U.S.monetary policy22This is akin to looking at a standard deviation of the distribution,but given that the normalized variables arestill somewhat skewed,we opt for the IQR.20Table 5.Quantification Exercise of the Heterogeneou

199、s Impacts on Investment of Trade Exposureto U.S.Monetary Policy ShocksGlobal TradeU.S.TradeCoef.IQRP90-P10Coef.IQRP90-P10(1)(2)(3)(4)(5)(6)MPUSt1?TotExpOutput?t1-0.120-0.037-0.066-0.387-0.018-0.0340.2300.4080.1110.211MPUSt1?FinExpOutput?t1-0.011-0.001-0.002-0.0130.000-0.0010.0060.0150.0010.004MPUSt1

200、?IntExpOutput?t1-0.208-0.038-0.093-0.82-0.017-0.0380.2340.5810.1080.237MPUSt1 WtOutdegt1-0.031-0.016-0.035-1.614-0.013-0.0450.0980.2180.0820.279Notes:This table presents quantification results based on firm-level panel regression results from the estimation ofregression(2),with time-varying fixed ef

201、fects as reported in Table 4 combined with information from Table A5.TheCoef.column reports the coefficients on the interacted variable,IQR/P90-P10 measure the coefficients impliedimpact of a U.S.monetary policy shock when moving from the lower quartile/decile to top quartile/decile of thegiven trad

202、e exposure variable.Numbers in square brackets represent the share(in absolute values)of these impactsrelative to the impact of a monetary policy shock on a mean firm.MPUSis the monetary policy shock from Buet al.(2021),and the trade exposure variables are(i)total trade-to-output ratio(TotExp/Output

203、),(ii)final goodstrade-to-output ratio(FinExp/Output),(iii)intermediate goods trade-to-output ratio(IntExp/Output),and(iv)the weighted outdegree(WtOutdeg).intensifies the decrease in the investment ratio by an additional 0.037 percentage points.This isequal to about one quarter of the average effect

204、 of the monetary policy shock,per the entries insquare brackets.Performing a similar calculation using the difference between the top and bottomdeciles implies that the same U.S.monetary policy contraction lowers the investment ratio by 0.066percentage points when considering the world trade ratio,r

205、oughly equal to forty percent of theaverage effect of the shock.The calculations using the U.S.-only trade ratio yields about half ofthe effect relative to exposure to world trade(square brackets,top row).Turning to intermediate exports,we also find an amplifying effect of trade exposure.Movingfrom

206、the bottom quartile to the top quartile country-sector in the world intermediate trade-to-output ratio distribution implies that a one percentage point shock to U.S.monetary policy willhave almost identical effects as moving over the IQR of the total exports ratio.However,a similarcalculation using

207、the difference between the top and bottom deciles of intermediate trade impliesthat the same U.S.monetary policy contraction will lower the investment ratio by an additional0.093 percentage points when considering the world trade ratio,which is around sixty percent ofthe average effect of the moneta

208、ry policy shock.The calculations using the U.S.-only intermediatetrade ratio again yield about half of the effect relative to exposure to world trade.21Finally,we consider the network measure of international trade,the weighted outdegree.Per-forming the interquartile quantification implies that movi

209、ng from the bottom quartile to the topquartile country-sector in the world weighted outdegree distribution implies that a one percentagepoint contraction in U.S.monetary policy leads to firm investment falling by 0.016 percentagepoints,or about ten percent of the average effect of the monetary polic

210、y shock.Considering thedifference between deciles roughly doubles the effect relative to the IQR calculation.Interestingly,comparing the IQR of the U.S.-only weighted-outdegree distribution yields similar results as theworld distribution,while moving between the deciles for the U.S.-only weighted ou

211、tdegree impliesa larger impact than moving along the world distribution.These facts capture the importance ofthe U.S.as customer country for our country-sector sample of suppliers,as well as the skewness ofthe weighted-outdegree distribution.Overall,we show that there are important heterogeneous eff

212、ects on firms conditional on theirsectors exposure to demand shocks being transmitted via exporting behavior.The magnitudeof the amplifying effect arising from the interaction between U.S.monetary policy shocks andintermediate good trade and global production linkages on firm-level investment is lar

213、ge.5.2Financial ConstraintsWe next examine the importance of financial constraints at the firm-level,conditioning on standardfirm-level measures as in regression(3).Here we allow for transmission to vary by firm characteristicZ.The two characteristics we use to proxy for firm financial constraints a

214、re size and net worth.23As noted in Section 2,in this specification the set of fixed effects()may now vary over time,allowing us to control for unobserved time-varying country-and/or sector-level characteristics(e.g.,how a countrys trade openness varies over time).Results are reported in Table 6.We

215、display results for investment only,with sales and employ-ment results in Appendix B.Moving from left to right,we begin by omitting time-varying fixedeffects,then include countryyear fixed effects,and finally include the most stringent set of fixedeffects of countrysectoryear.Looking at columns(1)an

216、d(4),which omit time-varying fixedeffects and control for size and net worth respectively,we see that a contractionary U.S.mone-tary policy shock has a slightly larger negative effect on investment growth than in our baselineestimation.As indicated in rows two and three,where the coefficient on the

217、interaction of eithersize or net worth and MPUSis positive,the contractionary effect is smaller for firms that are lessfinancially constrained.This finding holds irrespective of the proxy for financial constraints andthe set of fixed effects.Our interaction results echo those of Cloyne et al.(2020)a

218、nd Ottonello andWinberry(2020),who analyze U.S.firm investment and also find a smaller impact of monetary23Results are qualitatively similar if we instead use age or measure size by employment using a smaller subset offirms for which these data exist.22Table 6.Effect of U.S.Monetary Policy Shocks on

219、 Firms Investment:Firm-Level heterogeneity(Investmentt/FixAssetst1)SizeNet Worth(1)(2)(3)(4)(5)(6)MPUSt1-0.165a-0.164a(0.055)(0.054)MPUSt1Sizet10.018a0.020a0.021a(0.005)(0.006)(0.006)MPUSt1Net Wortht10.017a0.018a0.018a(0.005)(0.005)(0.005)Observations374,360374,359373,241374,360374,359373,241R20.058

220、0.0690.0960.0580.0690.096Countryyear FENoYesNoNoYesNoCountrysectoryear FENoNoYesNoNoYesMacro ControlsYesNoNoYesNoNoFirm ControlsYesYesYesYesYesYesFirm FEYesYesYesYesYesYesNotes:This table presents firm-level panel regression results based on the estimation of regression(3)for the changein the invest

221、ment-to-fixed capital ratio,where we interact firm characteristics with the monetary policy shock.Thesample uses firms with at least five years of observations over 1995-2019.All regressors are lagged one period,whereMPUSis the monetary policy shock from Bu et al.(2021),Size is the within country-ye

222、ar measure of a firms sizebased on the log of total assets,and Net worth is the within country-year measure of a firms net worth based onthe log of net worth(assets minus liabilities).We include lagged firm and macroeconomic variables as in the baselineestimation in Table 1,and fixed effects at vari

223、ous levels of disaggregation.Standard errors are double clustered atthe firm and year level,whereaindicates significance at the 1%level,bat the 5%level,andcat the 10%level.policy shocks on the investment of firms with less binding financial constraints.24Table A11 presents the size and net worth int

224、eraction results for sales and employments.Theresults for sales are qualitatively similar to those using investment shares in Panel A of the table.Turning to the employment regressions in Panel B,we see that there is no significant effect of MPUSon employment growth,as in the baseline regressions.Ho

225、wever,the coefficients on the interactionswith size or net worth are negative and significant when not including the most stringent set oftime-varying fixed effects,indicating that larger/high net worth firms contract employment morethan smaller/low net worth firms during periods of monetary tighten

226、ing.Quantifying the Financial Constraints ChannelWe next utilize the firm-level distribution of firm characteristics to quantify the heterogeneous24The firm size results also matches how small(U.S.)firms cut investment by more than large firms following amonetary contraction,the key result in early

227、work by Gertler and Gilchrist(1994).23impact of firms financial constraints on their investment reaction to monetary policy shocks.Sim-ilar to the trade exposure quantification exercise above,we examine the differential impact acrossthe firm-size distribution,in this case focusing on size and net wo

228、rth where each variable is nor-malized around mean zero(see Table A2).Notably,in contrast to trade exposure,the impact ofU.S.monetary policy shocks on firms that are in the upper tail of the distribution is attenuatedrather than amplified relative to those firms in the lower tail of the distribution

229、.Given the similarity in point estimates across the set of fixed effects in Table 6,we providenumbers based on the countrysectoryear specifications of columns(3)and(6)in Table 7.Het-erogeneity in the impact of monetary policy shocks across the firm distribution is large.First,moving across the IQR o

230、f the size distribution from smaller to larger firms implies an attenuationof the impact of U.S.monetary policy shocks of 0.052 percentage points,approximately one-thirdthe impact on the mean firm(based on column(1)of Table 6:0.165 p.p.).Moving from the lowerto upper decile of the firm-size distribu

231、tion implies a large attenuation arising from the looseningof financial constraints:0.108 percentage points,or two-thirds the impact on the average firm.Second,the net worth measure of financial constraints yields similar results to what we find forsize.Moving across the IQR of the net worth distrib

232、ution from more financially constrained to lessfinancially constrained firms implies an attenuation of 0.042 percentage points,which is approxi-mately one quarter of the impact of the shock on the mean firm(based on column(4)of Table 6:0.164 p.p.).Moving from the lower to upper decile of the firm-ne

233、t worth distribution implies alarge attenuation arising from the loosening of financial constraints:0.086 percentage points,orover one half the impact on the average firm.5.3Trade Exposure and Financial ConstraintsOur final set of core estimation results examines the heterogeneous impact of monetary

234、 policyshocks on foreign firms conditional on their trade exposure and financial constraints jointly.Table 8presents results for the investment regressions using the size interaction,while we relegate the networth regression to Table A14 since results are qualitatively similar.25All regressions are

235、run withcountryyear fixed effects.Looking across coefficients for the size and trade variables in Table 8and contrasting them with Tables 4 and 6(countryyear specifications),we see that the coefficientsare remarkably similar even when controlling for trade exposure and financial constraint proxiesjo

236、intly.A similar story holds for the net worth regressions as well as the employment and salesregressions presented in Appendix B.We next move to quantification exercises.Although comparing the impact of heterogeneityin the trade exposure and financial constraint proxies distributions is not perfect

237、given that thetrade variables are based on sector-level data,it is useful to remember that the largest firms in a25For completeness,we also present the sales and employment regressions in Tables A15 and A16.24Table 7.Quantification Exercise of the Heterogeneous Impacts on Investment of Financial Con

238、-straints to U.S.Monetary Policy ShocksCoef.IQRP90-P10(1)(2)(3)MPUSt1Sizet10.0210.0520.1080.3140.658MPUSt1Net Wortht10.0180.0420.0860.2560.527Notes:This table presents quantification results based on firm-level panel regression results from the estimation ofregression(3),with time-varying fixed effe

239、cts as reported in Table 6 combined with information from Table A2.TheCoef.column reports the coefficients on the interacted variable,IQR/P90-P10 measure the coefficients impliedimpact of a U.S.monetary policy shock when moving from the lower quartile/decile to top quartile/decile of thegiven firm c

240、onstraint variable.Numbers in square brackets represent the share(in absolute values)of these impactsrelative to the impact of a monetary policy shock on a mean firm.MPUSis the monetary policy shock from Bu etal.(2021),Size is the within country-year measure of a firms size based on the log of total

241、 assets,and Net worthis the within country-year measure of a firms net worth based on the log of net worth(assets minus liabilities).given sector also dominate exports(Melitz,2003;Freund and Pierola,2015).Therefore,contrastingimpacts of the trade and interest rate channels when looking at firms alon

242、g the size distributionacross sectors may indeed be a good approximation to having firm-level trade data to exploit.We begin by asking how small firms compare to large ones when moving from low to hightrade exposed sectors in Table 9.Focusing on intermediate goods trade exposure in the first tworows

243、,we utilize coefficients from either columns(3)or(7)of Table 8.First,looking at the IQRfor the size variable,the differential impact between a less financially constrained(larger)and amore constrained(smaller)firm from a one percentage point monetary policy tightening is 0.044p.p.,an attenuation of

244、roughly one quarter relative to the total impact on the mean firm(0.161p.p.contraction in investment).However,once we include the impact difference in the IQR of theintermediate world trade exposure and consider a movement from a less open to more open sector,this attenuation falls to 0.012 percenta

245、ge points(i.e.,0.044 0.032=0.012).Assuming that thedistribution of intermediate trade openness within a sector is similar to that across sectors(e.g.,the power law distributions of both trade exposures have the same slope),then this quantitativeexperiment would imply that,on net,the impact of large

246、firms being less financially constrainedwhile also being more exposed to world demand shocks via trade produces a slight attenuation ofthe effect of U.S.monetary policy shocks relative to the average firm.Put concretely,this indicatesthat the exacerbation of the impact of U.S.monetary policy shocks

247、due to increasedtrade exposure is dominated by the attenuation associated with being less financiallyconstrained.Further calculations yield the same qualitative results for firms exposures to U.S.intermediate goods trade,as well as their exposure to world production networks as measured by25Table 8.

248、Effect of U.S.Monetary Policy Shocks on Firms Investment:The Importance of Sizeand Trade Integration(Investmentt/FixAssetst1)Global TradeU.S.Trade(1)(2)(3)(4)(5)(6)(7)(8)MPUSt1Sizet10.019a0.020a0.018a0.019a0.019a0.020a0.018a0.019a(0.006)(0.006)(0.005)(0.006)(0.006)(0.006)(0.006)(0.006)MPUSt1?TotExpO

249、utput?t1-0.095b-0.284c(0.038)(0.155)MPUSt1?FinExpOutput?t10.0260.083(0.059)(0.150)MPUSt1?IntExpOutput?t1-0.178b-0.683b(0.069)(0.329)MPUSt1 WtOutdegt1-0.027b-1.426b(0.012)(0.536)Observations374,359374,359374,359374,359374,359374,359374,359374,359R20.0690.0690.0690.0690.0690.0690.0690.069Countryyear F

250、EYesYesYesYesYesYesYesYesFirm ControlsYesYesYesYesYesYesYesYesFirm FEYesYesYesYesYesYesYesYesNotes:This table presents firm-level panel regression results based on the estimation of regression(4)for the changein the investment-to-fixed capital ratio,where we interact firm size in addition to differe

251、nt measures country-sectorstrade integration with the monetary policy shock.Columns(1)-(4)use trade measures based on country-sector exportswith the world,while columns(5)-(8)use U.S.-only exports data.The country-sectors trade measure include(i)totaltrade-to-output ratio(TotExp/Output),(ii)final go

252、ods trade-to-output ratio(FinExp/Output),(iii)intermediategoods trade-to-output ratio(IntExp/Output),and(iv)the weighted outdegree(WtOutdeg).The sample usesfirms with at least five years of observations over 1995-2019.All regressors are lagged one period,where MPUSis themonetary policy shock from Bu

253、 et al.(2021),Size is the within country-year measure of a firms size based on thelog of total assets,and Net worth is the within country-year measure of a firms net worth based on the log of networth(assets minus liabilities).We include lagged firm and macroeconomic variables as in the baseline est

254、imationin Table 1,and fixed effects at various levels of disaggregation.Standard errors are double clustered at the firm andyear level,whereaindicates significance at the 1%level,bat the 5%level,andcat the 10%level.weighted outdegree.There,however,the dampening effects of less binding financial cons

255、traints oflarger firms are somewhat larger.5.4Heterogeneous Effects across CountriesWe exploit the cross-country dimension of our dataset in order to ask whether there is any het-erogeneity in the relative impact of either the trade exposure or financial constraint variables byestimating regressions

256、(2)and(3)allowing for the coefficients on the trade exposure or financialconstraint interaction terms(the 2s)to vary across countries.26Figure A6 plots the cross-countrydistribution of twenty different estimated coefficients on the interaction of the monetary policy26We also allow for heterogeneity

257、in the non-interacted coefficients to avoid omitted variable bias.26Table 9.Quantification Exercise of the Heterogeneous Impacts on Investment of Trade Exposureand Financial Constraints to U.S.Monetary Policy ShocksGlobal TradeU.S.TradeCoef.IQRP90-P10Coef.IQRP90-P10(1)(2)(3)(4)(5)(6)MPUSt1Sizet10.01

258、80.0440.0930.0180.0440.0930.2760.5750.2760.575MPUSt1?IntExpOutput?t1-0.178-0.032-0.080-0.683-0.015-0.0320.2000.4970.0900.197Total0.0120.0130.0300.0610.0760.0780.1860.377MPUSt1Sizet10.0190.0470.0980.0190.0470.0980.2910.6070.2910.607MPUSt1 WtOutdegt1-0.027-0.014-0.031-1.426-0.012-0.0400.0860.1900.0730

259、.247Total0.0330.0670.0350.0580.2060.4170.2190.360Notes:This table presents quantification results based on firm-level panel regression results from the estimationof regression(4)as reported in Table 8 combined with information from Tables A2 and A5.The Coef.columnreports the coefficients on the inte

260、racted variable,IQR/P90-P10 measure the coefficients implied impact of a U.S.monetary policy shock when moving from the lower quartile/decile to top quartile/decile of the given firm constraintvariable.Numbers in square brackets represent the share(in absolute values)of these impacts relative to the

261、 impactof a monetary policy shock on a mean firm.MPUSis the monetary policy shock from Bu et al.(2021),Size is thewithin country-year measure of a firms size based on the log of total assets,and Net worth is the within country-yearmeasure of a firms net worth based on the log of net worth(assets min

262、us liabilities).shock with the four world trade exposure measures.The estimates are based on regressions withcountryyear fixed effects and we include 95%confidence intervals in the figures.We reject ho-mogeneity across the three coefficients that appear significant in Table 4 in panels(a),(c),and(d)

263、total exports,intermediate exports,and weighted outdegree,respectively.It clear from thefigures that there is heterogeneity in the estimates,with some coefficients being positive rather thannegative and others insignificant.However,given the unbalanced nature of the panel along withusing country-sec

264、tor variables rather than firm ones,it is hard to draw any concrete conclusions.We repeat this for the financial constraint interactions in Figure A7,which plots coefficients forthe size and net worth interactions in panels(a)and(b),respectively.We reject homogeneity ofcoefficients,but all coefficie

265、nts are positive and many statistically significant.Further Robustness ChecksWe conduct additional robustness checks for the interaction re-gressions.In particular,we first replace both the country-sector trade and firm-level financial27constraint variables with beginning-of-period values rather tha

266、n using time-varying values.Over-all,results are robust and the coefficients on the interaction terms do not change dramatically,either quantitatively or in terms of statistical significance.Second,rather than using beginning-of-period values we use the interaction variables averaged over time.Again

267、,our main findings arerobust to this change of spsecification.6ConclusionThis paper documents two broad results.First,there are significant effects of Fed monetary policyshocks on foreign firms investment,sales,and employment.This spillover effect varies betweenemerging market economies(EMEs)and adv

268、anced economies,but not according to country-levelvariation in measures such as the degree of financial account and trade openness.Second,drillingdown to more granular levels of heterogeneity across sectors and firms,we find interesting patternsin the data that suggest potential channels for the amp

269、lification or attenuation in the spilloversof U.S.monetary policy shocks.Namely,greater exposure to intermediate goods trade and globalproduction linkages contribute to amplifying the cross-country transmission of U.S.monetary policyshocks to firms.However,these effects are attenuated for larger fir

270、ms and firms with greater networth given less binding financial constraints,which dampen the interest rate channel of monetarypolicy.These findings highlight the importance of both external demand channel and interest ratechannels for monetary policy spillovers to foreign activity.28ReferencesAnders

271、on,James E.and Eric van Wincoop,“Trade Costs,”Journal of Economic Literature,2004,42(3),691751.Antr as,Pol,“De-Globalisation?Global Value Chains in the Post-COVID-19 Age,”in“2021 ECBForum:Central Banks in a Shifting World Conference Proceedings”2021.Arbatli-Saxegaard,Elif,Melih Firat,Davide Furceri,

272、and Jeanne Verrier,“U.S.Monetary PolicyShock Spillovers:Evidence from Firm-Level Data,”2022.IMF Working Paper 22/191.Autor,David,David Dorn,Lawrence F.Katz,Christina Patterson,and John van Reenen,“TheFall of the Labor Share and the Rise of Superstar Firms,”The Quarterly Journal of Economics,2020,135

273、(2),645709.Bauer,Michael D.and Eric T.Swanson,“An Alternative Explanation for the Fed InformationEffect,”2020.NBER Working Paper 27013.Bernanke,Ben S.and Mark Gertler,“Agency Costs,Net Worth,and Business Fluctuations,”American Economic Review,1989,79(1),1431.Br auning,Falk and Viacheslav Sheremirov,

274、“The Transmission Mechanisms of International Busi-ness Cycles:Output Spillovers through Trade and Financial Linkages,”2021.Federal ReserveBank of Boston Research Department Working Papers No.21-13.and Victoria Ivashina,“U.S.Monetary Policy and Emerging Market Credit Cycles,”Journal ofMonetary Econo

275、mics,2020,112,5776.Bu,Chunya,John H.Rogers,and Wenbin Wu,“A unified measure of Fed monetary policy shocks,”Journal of Monetary Economics,2021,118,331349.Caglio,Cecilia R,R.Matthew Darst,and S ebnem Kalemli-Ozcan,“Risk-Taking and MonetaryPolicy Transmission:Evidence from Loans to SMEs and Large Firms

276、,”April 2021.NBERWorking Paper 28685.Carvalho,Vasco M.,“From Micro to Macro via Production Networks,”Journal of Economic Per-spectives,Fall 2014,28(4),2348.Claessens,Stijn,Hui Tong,and Shang-Jin Wei,“From the Financial Crisis to the Real Economy:Using Firm-level Data to Identify Transmission Channel

277、s,”Journal of International Economics,2012,88.Cloyne,James,Clodomiro Ferreira,Maren Froemel,and Paolo Surico,“Monetary Policy,CorporateFinance and Investment,”2020.Forthcoming,Journal of the European Economic Association.Cochrane,John and Monika Piasezzi,“The Fed and Interest Rates:A High-Frequency

278、Identifica-tion,”American Economic Review P&P,2002,92(2),9095.Dao,Mai Chi,Camelia Minoiu,and Jonathan D.Ostry,“Corporate Investment and the RealExchange Rate,”Journal of International Economics,2021,131.Dedola,Luca,Giulia Rivolta,and Livio Stracca,“If the Fed sneezes,who catches a cold?,”Journalof I

279、nternational Economics,2017,108,S23S41.Degasperi,Riccardo,Seokki Simon Hong,and Giovanni Ricco,“The Global Transmission of USMonetary Policy,”2021.Mimeo,University of Warwick.di Giovanni,Julian and Andrei A.Levchenko,“Country Size,International Trade and AggregateFluctuations in Granular Economies,”

280、Journal of Political Economy,2012,120(6),10831132.and Galina Hale,“Stock Market Spillovers via the Global Production Network:Transmission ofU.S.Monetary Policy,”2022.Forthcoming,Journal of Finance.29,Andrei A.Levchenko,and Isabelle Mejean,“Firms,Destinations,and Aggregate Fluctuations,”Econometrica,

281、2014,82(4),13031340.,and,“The Micro Origins of International Business Cycle Comovement,”American Eco-nomic Review,2018,108(1),82108.,and,“Foreign Shocks as Granular Fluctuations,”2022.NBER Working Paper 28123.,Sebnem Kalemli-Ozcan,Mehmet F.Ulu,and Yusef S.Baskaya,“International Spillovers andLocal C

282、redit Cycles,”Review of Economic Studies,2021,89(2),733773.Eichenbaum,Martin and Charles Evans,“Some Empirical Evidence on the Effects of Shocks toMonetary Policy on Exchange Rates,”The Quarterly Journal of Economics,1995,110(4),9751009.Fama,Eugene F.and James D.MacBeth,“Risk,Return,and Equilibrium:

283、Empirical Tests,”Journal of Political Economy,1973,81(3),607636.Faust,Jon and John H.Rogers,“Monetary Policys Role in Exchange Rate Behavior,”Journal ofMonetary Economics,2003,50(7),14031424.,Eric T.Swanson,and Jonathan H.Wright,“Identifying the Effects of Monetary PolicyShocks on Exchange Rates Usi

284、ng High Frequency Data,”Journal of the European EconomicAssociation,2003,1(5),10311057.Fern andez,Andr es,Michael W.Klein,Alessandro Rebucci,Martin Schindler,and Martin Uribe,“Capital control measures:A new dataset,”IMF Economic Review,2016,64(3),548574.Freund,Caroline and Martha Denisse Pierola,“Ex

285、port Superstars,”Review of Economics andStatistics,2015,97(5),10231032.Gabaix,Xavier,“The Granular Origins of Aggregate Fluctuations,”Econometrica,2011,79(3),733772.Georgiadis,Georgios,“Determinants of Global Spillovers from US Monetary Policy,”Journal ofInternational Money and Finance,2016,67,4161.

286、Gertler,Mark and Simon Gilchrist,“Monetary Policy,Business Cycles,and the Behavior of SmallManufacturing Firms,”Quarterly Journal of Economics,1994,109(2),309340.Gopinath,Gita,Sebnem Kalemli-Ozcan,Loukas Karabarbounis,and Carolina Villegas-Sanchez,“Capital Allocation and Productivity in South Europe

287、,”Quarterly Journal of Economics,2017,132,19151967.Gulen,Huseyin and Mihai Ion,“Policy Uncertainty and Corporate Investment,”Review of Finan-cial Studies,2016,29(3),523564.G urkayanak,Refet S.,Brian Sack,and Eric T.Swanson,“Do Actions Speak Louder Than Words?The Response of Asset Prices to Monetary

288、Policy Actions and Statements,”International Journalof Central Banking,2005,1(1),5593.Han,Xuehui and Shang-Jin Wei,“International Transmissions of Monetary Shocks:Between aTrilemma and a Dilemma,”Journal of International Economics,2018,110,205219.Kalemli-Ozcan,Sebnem,“U.S.Monetary Policy and Interna

289、tional Risk Spillovers,”2019.FederalReserve Bank of Kansas City Jackson Hole Economic Policy Symposium.Kearns,Jonathan,Andreas Schrimpf,and Fan Dora Xia,“Explaining Monetary Spillovers:TheMatrix Reloaded,”2019.Mimeo,Reserve Bank of Australia.Kim,Soyoung and Nouriel Roubini,“Exchange rate anomalies i

290、n the industrial countries:A solu-tion with a structural VAR approach,”Journal of Monetary Economics,2000,45(3),561586.30Kiyotaki,Nobu and John Moore,“Credit Cycles,”Journal of Political Economy,1997,105(2).Melitz,Marc J.,“The Impact of Trade on Intra-Industry Reallocations and Aggregate IndustryPro

291、ductivity,”Econometrica,2003,71(6),16951725.Miranda-Agrippino,Silvia,“Unsurprising shocks:information,premia,and the monetary trans-mission,”2016.Bank of England Working Paper.and H el ene Rey,“US monetary policy and the global financial cycle,”Review of EconomicStudies,2020,87(6),27542776.Morais,Be

292、rnardo,Jos e-Luis Peydr o,Jessica Rold an-Pe na,and Claudia Ruiz-Ortega,“The Interna-tional Bank Lending Channel of Monetary Policy Rates and QE:Credit Supply,Reach-for-Yield,and Real Effects,”Journal of Finance,2019,74(1),5590.Nakamura,Emi and J on Steinsson,“High-frequency identification of moneta

293、ry non-neutrality:theinformation effect,”Quarterly Journal of Economics,2018,133(3),12831330.Obstfeld,Maurice,“Uncoordinated Monetary Policies Risk a Historic Global Slowdown,”2022.Peterson Institute for International Economics,September 12,2022.and Kenneth Rogoff,“The Six Major Puzzles in Internati

294、onal Macroeconomics:Is There aCommon Cause?,”NBER Macroeconomics Annual,2000,15,339390.Oscar Jord a,“Estimation and inference of impulse responses by local projections,”American Eco-nomic Review,2005,95(1),161182.Ottonello,Pablo and Thomas Winberry,“Financial Heterogeneity and the Investment Channel

295、 ofMonetary Policy,”Econometrica,2020,88(6),24732502.Ramey,Valerie A.,“Macroeconomic shocks and their propagation,”in“Handbook of Macroeco-nomics,”Vol.2,Elsevier,2016,pp.71162.Rey,H el ene,“Dilemma not trilemma:the global financial cycle and monetary policy independence,”2013.Federal Reserve Bank of

296、 Kansas City Jackson Hole Economic Policy Symposium.,“International channels of transmission of monetary policy and the Mundellian trilemma,”IMFEconomic Review,2016,64(1),635.Rogers,John H.,“Monetary Shocks and Real Exchange Rates,”Journal of International Eco-nomics,1999,48,26988.,Chiara Scotti,and

297、 Jonathan H.Wright,“Evaluating asset-market effects of unconventionalmonetary policy:a multi-country review,”Economic Policy,2014,29,749799.Swanson,Eric T.,“Measuring the effects of federal reserve forward guidance and asset purchaseson financial markets,”Journal of Monetary Economics,2021,118,3253.

298、Timmer,Marcel P.,Erik Dietzenbacher,Bart Los,Robert Stehrer,and Gaaitzen J.de Vries,“AnIllustrated User Guide to the World InputOutput Database:the Case of Global AutomotiveProduction,”Review of International Economics,August 2015,23(3),575605.Wei,Shang-Jin,“The Risk of Competitive Interest Rate Hik

299、es,”2022.Project Syndicate,September8,2022.and Yinxi Xie,“Monetary policy in an era of global supply chains,”Journal of InternationalEconomics,2020,124(2),283296.31Appendix AAdditional FiguresFigure A1.Alternative U.S.Monetary Policy Shocks:Nakamura and SteinssonNotes:This figure plots the annual ag

300、gregate of the policy news shock constructed by Nakamura and Steinsson(2018)(updated).Figure A2.Alternative U.S.Monetary Policy Shocks:Swansons Forward GuidanceNotes:This figure plots the annual aggregate of the Forward Guidance factor estimated by Swanson(2021).32Figure A3.European Monetary Policy

301、ShocksNotes:This figure plots the annual aggregate of the pure European monetary policy shock constructed by Bu et al.(2021).Figure A4.Estimated Coefficient on U.S.Monetary Policy Shock Leaving Out One YearNotes:This figure plots the estimated obtained from estimating Equation(1)multiple times leavi

302、ng out one yearsworth of observations at a time.The left-out year is indicated on the horizontal axis.33Figure A5.Cumulative Impulse Responses for Investment,Sales,and Employment of a onePercentage Point Contraction in U.S.Monetary Policy(a)Investment/Fixed Assets(b)Sales/Fixed Assets(c)ln(Employmen

303、t)Notes:This figure plots the cumulative impulse response function of a one percentage point contraction in U.S.monetary policy(Bu et al.,2021)for(a)investment ratio,(b)sales ratio,and(c)log employment(in millions).Estimation is based on local projection method(Jord a,2005)of the baseline regression

304、(1)with h=0,.,3,controlling for firm-level fixed effects.90%confidence intervals are plotted in dashed lines,and regressions areclustering at the firm and year levels.34Figure A6.Heterogeneous Impact of Trade Exposure on the Transmission of U.S.MonetaryPolicy shocks across Countries(a)TotExp/Output(

305、b)FinExp/Output(c)IntExp/Output(d)WtOutdegNotes:This figure plots coefficients for the financial constraint interaction with the monetary policy shock fromregression(2)(2)where we allow the coefficient to vary across countries.Panel(a)plots the coefficients for theTotExp/Out variable interaction,pan

306、el(b)for the FinExp/Out variable interaction,panel(c)for the IntExp/Outvariable interaction,and panel(d)for the WtOutdeg variable interaction.All regressions were run with firm controlsand countryyear fixed effects,clustering at the firm and year levels.The blue standard error bounds are for the95%l

307、evel.35Figure A7.Heterogeneous Impact of Financial Constraints on the Transmission of U.S.MonetaryPolicy shocks across Countries(a)Size(b)Net WorthNotes:This figure plots coefficients for the financial constraint interaction with the monetary policy shock fromregression(3)(2)where we allow the coeff

308、icient to vary across countries.Panel(a)plots the coefficients for theSize variable interaction,while panel(b)plots the coefficients for the Net Worth variable interaction.All regressionswere run with firm controls and countrysectoryear fixed effects,clustering at the firm and year levels.The bluest

309、andard error bounds are for the 95%level.36Appendix BAdditional TablesTable A1.Firm-level Summary Statistics for Country Sample,1995-2019Obs.MeanMedianSt.Dev.MinMax(Investment/Assets)438,300-0.024-0.0020.533-22(Sales/Assets)438,0390.0640.0033.774-1010Employment growth297,1520.0740.0160.328-12log(Cas

310、h flow)332,13219.4219.383.1110.9326.45Sales growth423,5670.130.060.44-12log(Assets)480,72921.5421.593.4810.1128.95Age387,64928.512123.890211log(Sales)463,35321.1721.303.649.9028.33log(EBITDA)375,45319.7719.713.0211.9426.67Tobins Q143,7792.081.284.190.4280.80Liquidity ratio467,6780.010.040.30-4.230.4

311、0Leverage477,2630.240.190.3003.49log(Debt)411,66720.0620.243.709.7427.84log(Int.pay)425,45916.7416.893.636.9124.78log(Collateral)458,24420.7720.953.599.5728.11log(Dividends)272,38518.2618.332.8110.0424.86log(Equity)439,23921.1321.403.2612.9328.09Notes:This table presents firm-level summary statistic

312、s for all firms in with at least five years of data and that arein our baseline regression sample over 1995-2019.Summary statistics are based on the pooled sample of firms,whereall variables have been winsorized at the 1%level,except for the change in sales-to-asset ratio which is winsorized atthe 5

313、%level.All measures are in nominal terms and in USD.Table A2.Summary Statistics for Normalized Firm-Level Financial Constraint Proxy Measuresacross FirmsObs.MeanSt.Dev.p10p25p50p75p90Size438,3000.0002.120-2.382-1.338-0.2081.1322.761Net Worth438,3000.0001.965-2.258-1.237-0.1481.1062.540Notes:This tab

314、le presents firm-level summary statistics on the normalized size and net worth variables.Eachvariable is normalized across firms within a country-year.Summary statistics presented across all years.37Table A3.Summary Statistics Sector-Level Trade Measures for World Trade across Country Sample,2000Tot

315、Exp/OutputFinExp/OutputIntExp/OutputWtOutdegSectorCodeMeanSt.Dev.MeanSt.Dev.MeanSt.Dev.MeanSt.Dev.AgricultureAtB0.0920.0770.0240.0180.0690.0680.2520.081Air Transport620.3250.1900.1040.0620.2210.1280.1010.212Automotive500.0140.0290.0050.0130.0090.0170.0030.035Carbon/Nuclear Fuels230.1950.1430.0630.05

316、40.1320.0910.3910.126Chemicals240.3620.2160.0910.0930.2710.1451.0750.087ConstructionF0.0080.0180.0030.0090.0050.0100.0260.228EducationM0.0040.0060.0010.0020.0030.0040.0070.151Electrical Equipment30t330.5490.2630.2560.1360.2930.1360.8650.161Financial IntermediationJ0.0450.0470.0070.0060.0370.0410.126

317、0.197Food15t160.1250.0740.0950.0580.0300.0230.1470.198General Machinery290.3980.2380.2500.1400.1480.1020.4070.125General Manufacturing36t370.3960.2170.2870.1610.1090.0680.1110.183Health/Social WorkN0.0030.0040.0020.0020.0020.0020.0050.110Hotels and RestaurantsH0.0650.0730.0230.0250.0420.0500.0900.15

318、3Inland Transport600.0510.0430.0100.0100.0410.0350.0730.116Leather190.5160.2280.3690.1890.1470.0620.2500.124Metals27t280.2960.1290.0220.0180.2730.1220.8680.037MiningC0.2850.2414E-040.0280.2840.2251.6060.011Non-Metallic Minerals260.1710.0950.0220.0160.1490.0910.1260.040Other Business Activities71t740

319、.0980.0810.0130.0150.0860.0700.4990.072Other ServicesO0.0440.0360.0120.0110.0320.0250.1200.031Other Transport630.0890.0740.0130.0140.0760.0630.1670.093Paper21t220.1790.1600.0210.0140.1570.1550.2980.133Post and Telecommunications640.0410.0270.0080.0080.0330.0230.0620.260Public AdministrationL0.0030.0

320、020.0020.0010.0020.0010.0040.198Real Estate700.0070.0100.0020.0030.0050.0070.0210.181Retail Trade520.0290.0430.0200.0370.0090.0140.0200.081Rubber and Plastics250.2760.1770.0560.0460.2200.1390.3260.123Textiles17t180.4750.2190.3020.1650.1720.1280.3890.008Transport Equipment34t350.3430.2410.1960.1620.1

321、470.0980.4660.090UtilitiesE0.0210.0290.0050.0060.0160.0230.0440.007Water Transport610.4150.3670.0980.1010.3170.2830.0820.014Wholesale Trade510.0480.0550.0110.0120.0370.0460.1200.006Wood200.2150.1760.0130.0140.2020.1740.1830.054Notes:This table presents sector-level summary statistics on the(i)total

322、exports-to-output ratio(TotExp/Output),(ii)final goods exports-to-output ratio(FinExp/Output),(iii)intermediate goods exports-to-output ratio(IntExp/Output),and(iv)export weighted outdegree(WtOutdeg)at the sector level for tradewith the world.The Mean variable is the average value of the ratio acros

323、s countries within a sector,while St.Dev.is the standard deviation of the ratio acrosscountries within a sector.We calculate both the mean and standard deviation of these ratios across countries for the year 2000.38Table A4.Summary Statistics for Sector-Level Trade Measures for U.S.Trade across Coun

324、try Sample,2000TotExp/OutputFinExp/OutputIntExp/OutputWtOutdegSectorCodeMeanSt.Dev.MeanSt.Dev.MeanSt.Dev.MeanSt.Dev.AgricultureAtB0.0150.0220.0040.0050.0110.0160.0040.007Air Transport620.1040.0760.0330.0250.0710.0520.0020.001Automotive500.0020.0040.0010.0020.0010.0021E-051E-05Carbon/Nuclear Fuels230

325、.0370.0480.0120.0140.0250.0340.0040.006Chemicals240.0720.0930.0200.0220.0520.0780.0110.014ConstructionF0.0010.0017E-051E-045E-040.0018E-051E-04EducationM0.0010.0012E-044E-045E-040.0013E-056E-05Electrical Equipment30t330.1480.1680.0870.1006E-020.0710.0140.021Financial IntermediationJ0.0040.0060.0010.

326、0010.0030.0050.0010.001Food15t160.0200.0320.0170.0270.0030.0060.0010.001General Machinery290.0930.0990.0570.0550.0370.0460.0030.003General Manufacturing36t370.1790.1680.1350.1210.0440.0510.0030.005Health/Social WorkN0.0010.0013E-044E-044E-040.0014E-055E-05Hotels and RestaurantsH0.0020.0060.0010.0053

327、E-040.0013E-059E-05Inland Transport600.0020.0044E-040.0010.0020.0040.0010.002Leather190.1760.0850.1290.0630.0480.0220.0090.011Metals27t280.0590.0840.0080.0150.0510.0690.0090.010MiningC0.0390.0700.0010.0030.0380.0680.0360.085Non-Metallic Minerals260.0480.0700.0070.0070.0410.0710.0020.002Other Busines

328、s Activities71t740.0420.0460.0060.0100.0360.0400.0080.012Other ServicesO0.0060.0080.0020.0030.0050.0060.0010.001Other Transport630.0040.0120.0010.0040.0030.0086E-051E-04Paper21t220.0380.0750.0080.0080.0300.0700.0030.007Post and Telecommunications640.0010.0022E-042E-040.0010.0011E-043E-04Public Admin

329、istrationL4E-056E-052E-053E-052E-053E-051E-061E-06Real Estate704E-040.0029E-053E-043E-040.0014E-051E-04Retail Trade520.0030.0050.0020.0030.0010.0032E-044E-04Rubber and Plastics250.0580.1310.0190.0450.0390.0860.0020.004Textiles17t180.1220.1350.1100.1220.0120.0230.0020.002Transport Equipment34t350.095

330、0.1700.0640.1310.0310.0420.0060.010UtilitiesE0.0060.0150.0010.0030.0040.0115E-040.001Water Transport610.0040.0050.0010.0010.0030.0045E-057E-05Wholesale Trade510.0020.0040.0010.0010.0020.0030.0010.003Wood200.0500.1090.0040.0040.0460.1090.0040.010Notes:This table presents sector-level summary statisti

331、cs on the(i)total exports-to-output ratio(TotExp/Output),(ii)final goods exports-to-output ratio(FinExp/Output),(iii)intermediate goods exports-to-output ratio(IntExp/Output),and(iv)export weighted outdegree(WtOutdeg)at the sector level for tradewith the U.S.only.The Mean variable is the average val

332、ue of the ratio across countries within a sector,while St.Dev.is the standard deviation of the ratioacross countries within a sector.We calculate both the mean and standard deviation of these ratios across countries for the year 2000.39Table A5.Summary Statistics for Normalized Sector-Level Trade Me

333、asures across FirmsObs.MeanSt.Dev.p10p25p50p75p90TotExp/Output438,3000.0000.223-0.239-0.131-0.0550.1770.309FinExp/Output438,3000.0000.102-0.086-0.056-0.0340.0320.136IntExp/Output438,3000.0000.169-0.182-0.086-0.0340.0940.267WtOutdeg438,3000.0000.720-0.678-0.299-0.0960.2110.455TotExp/Output,U.S.438,3000.0000.060-0.035-0.022-0.0090.0240.053FinExp/Output,U.S.438,3000.0000.036-0.021-0.014-0.0080.0030.0