FirmsinInternationalTrade

FedericoJ.Díez

FederalReserveBankofBoston

JesseMora

OccidentalCollege

AlanC.Spearot

UniversityofCalifornia,SantaCruz

Abstract

Firms play a critical role in the global economy. In this chapter, we survey the

behavioroffirmsintheinternationaleconomy,bothintheoryandinthedata.We

first summarize the key empirical facts thatmotivate the study of firms in trade.

Then,wedetail recent theoreticaldevelopmentson themicro-foundationsof firm

behavior in an international context, focusing on how firms select into exporting,

andhow firms respond to international shocks.Finally,we turn toa “realworld,”

empiricallyfocusedviewofexporting,beginningwiththegrowthdynamicsoffirms

expanding to globalmarkets, and then addressing the critical financing decisions

firmsmakewhenengagingininternationalcommerce.Weconcludewithdirections

forfutureresearch.

1

Introduction

Thelast20yearshavewitnessedasubstantialexpansioninresearchstudyingfirms

in the global economy. This literature on “firms in trade” studies how firms, in

particular the most productive and typically very large firms, self-select into

exporting and shape the landscape of international trade and investment. These

largefirmsininternationaltradeareresponsiblefordecisionsthatcreatelinkages

between major sectors of the world economy and affect the lives of billions of

workersworldwide.

Oneneedstolooknofurtherthansomeofthemostwell-knownbusinesses

tounderstand the importanceof firms inshapingworld trade flows.Forexample,

Boeing, the largest exporter in theUnited States, exported $29billion in2009, or

approximately1.8percentoftotalexports fromtheUnitedStates.1Thisnumber is

moreimpressivewhenweconsiderthatthereareapproximately300,000exporters

in the United States (U.S. Census 2013). Another firm, Fontera—a large dairy

producerandNewZealand’slargestcompany—isresponsiblefor20percentofNew

Zealand merchandise exports. 2 There are many more such examples. In fact,

Bernardetal.(2007)documentthatwhileexportingfirmsintheUnitedStatesarea

minority ofmanufacturing firms, not all exporters are the same and a very small

share of exporting firms (11 percent) account for a very large share of exporting

revenues(92percent).

Clearly, firmsplayan important role inexports from theUnitedStatesand

otherdevelopedeconomiesinawiderangeofindustries.And,aswediscussbelow,

theseanecdotesarejustthetipoftheiceberginhowfirms,inparticularlargefirms,

haveaprofoundeffectonthe internationaleconomy.Thedecisionsofthemodern

exporter are complex and have far-reaching implications for the firm’s financial

well-being, the workers who are employed by the firm, and the consumers who

enjoyitsproducts.

1http://www.slate.com/articles/business/exports/2010/11/the_boeing_co.html2http://www.nzdairycareers.co.nz/?page=Dairy_Industry&subpage=Dairy_Facts

2

Perhaps surprisingly, in spite of these anecdotes, the focus on firms is a

relatively new development. The classic literature on trade focused mainly on

primitives over country and industry characteristics to predict the pattern and

implications of trade. In itsmost basic and general form, the classic trademodel

predicts that countries tend to export goods in which they have a comparative

advantage,which,throughthelensofmostmodels,correspondstoproductswitha

lowwithin-countryautarkypricewhencomparedwiththewithin-countryautarky

priceofothercountries.Thesepricedifferentials,inautarky,canbedrivenbyfactor

endowments, technology differences, or a combination of the two. That is, a

relativelylowautarkypriceisobservedinproductswithhigherproductivity,andin

goodsthatmakeintensiveuseofarelativelyabundantfactorofproduction.

Evenbeforethefocusontheroleoffirms,testingtheclassictrademodels—

perhapstosupportpolicyevaluation—wasoffirst-orderimportance.Indeed,before

thefirms-in-traderevolution,manyroundsoftheWorldTradeOrganization(WTO)

occurred,andotheragreementsweresignedthatwerespecifictocertainindustries

(for example, the U.S.-Canada Automotive Pact). Unfortunately, testing the classic

trademodelshasprovenchallenging.Generalstatementsofcomparativeadvantage

exist(Deardorff1984),althoughsincethecounterfactualofautarkyisarare,ifnota

non-existent, occurrence, themodel is virtually impossible to test.3Further,when

focusing on endowment differences, one rarely observes a precise accounting for

the compositionof trade that is related to factorsof production, especially across

manycountries.Thus,startingwithLeontief(1953)andVanek(1968),researchers

have adopted assumptions on technology to link the classic, endowment-based

results to the imputed “factor content of trade” (for example, the automobile

industry tends toexportXamountofdomesticcapitalpervehicle).This literature

3Oneexceptionexiststothisdataissue,whichisBernhofenandBrown(2004).Here,theauthorsuserefined historical data from 19th century Japan, when autarky was observed, to test the generaltheoryof comparativeadvantage.They find robust support for the theory,whereuponopening totrade,Japanexportedgoodsforwhichtheautarkypricewasrelativelylow.

3

has provided a long list of competing papers, providing mixed support for the

endowment-basedmodels.4

Beyond the empirical challenges, it also became clear that the classic

literature on trade was missing the mark in a more important area—the model

itself. To seewhy, consider awell-known company such asGeneralMotors (GM),

which,duringthe1950sand1960s,hadanapproximately50percentshareofthe

domesticmarket.Yet,theclassictrademodelsassumedthatfirmsweresmallprice

takers, exercising no strategy overwhat to produce,where to produce, orhow to

produce. This is particularly important when considering the international

dimension of these models. In the classic models, every firm used the same

technology, neverbought imported inputs, andmade simpleproductiondecisions

based on domestic factor prices and the homogenous domestic price. Further,

companieslikeGMoftenproducedmanyproductvarieties,andthesevarietieswere

traded between countries, usually in both directions, in stark contrast with the

classical prediction that countries export a homogeneous product in one industry

and import in another. Simply put, themodels did notmatch reality, and theory

needed to catch up to provide a realistic and quantifiable theory to evaluate the

costsandbenefitsoftrade.

In this chapter, we survey the progression of firms in the trade literature,

fromthegroundbreakingworkofPaulKrugman(1979,1980),whichchangedthe

way we model international trade, to the most recent theoretical and empirical

research,whichevaluateshow firmsevolve in theglobal economyandhow firms

financetheirexportoperations.Webegininsectiononewithsomebasicfactsabout

firmsintradethathaveguidedtheliteratureandstandincontrastwiththeclassic,

aggregatetreatmentofworldtrade.Inthesecondsection,weoutlinethecanonical

frameworksformodelingtheexportingdecisionsofheterogeneousfirmsandfocus

onhowtradecostsandtradepolicyaffectthesedecisions. Inthethirdsection,we

4Earlier tests of the Heckscher-Ohlin theorems include Leontief (1953), Leamer (1980), Trefler(1993,1995),DavisandWeinstein(2001).VeryrecentworkinMorrowandTrefler(2016)providesthemostrigoroussupportoftheHOresultstodate.

4

take an empirical view of how exporting evolves over time and of how thismay

influence future firm performance. In the fourth section, we focus on how firms

financetheirexportingoperationsandhowcreditconstraintscanalterthepattern

oftrade.Finally, inthelastsection,weconcludewithideasforfuturedirectionsin

theliterature.

1 EmpiricsofFirmsinTrade

Asmentioned in the introduction, one of themain issueswith the classical trade

model is its inability to explain the presence and impact of firms in the global

economy.Withtheincreasedavailabilityof firm-leveldatainthe1990s, itbecame

clearthattheproblemwasnotonlythemodel’sinabilitytoexplainthebehaviorofa

few firms, but also its inability to explain systematic trade patterns acrossmany

countriesandindustries.Inthissection,wefocusonthefollowingfourstylizedfacts

thataredifficult toreconcilewith thestandardtrademodelandthathave formed

thebasisforthefirms-in-traderevolution:(1)fewfirmsexport,eveninacountry’s

“exporting”sectors;(2)exportingfirmsaredifferentfromnon-exportingfirmsinall

sectors;(3)afew,very-successfulexportingfirmsaccountformostexports;and(4)

exportingfirmsstartsmallandmostfail,butthosethatsurvivetendtoexpand.In

subsequent sections, we outline the leading framework developed to understand

theseempirical findingsanddetailhowthis frameworkallows international trade

economiststoanswernew,policy-relevant,questions.

1.1 FewFirmsExport

Most classic theories of trade, as well as the most basic version of Krugman’s

pioneeringworkonfirmsintrade,assumethatallfirmshavethesamelikelihoodof

exporting. However, this is in stark contrast with data for the United States and

other countries. Simply put, very few firms export, even in export-oriented

industries.

5

Tobegin,consider first thehighly influentialworkofBernardetal. (2007),

whichoutlinesmanystylizedfactsforU.S.exportersandimporters.Incontrastwith

thepredictionsof theclassicmodels, theauthors find thatonly18percentofU.S.

firmsexportandthosefirmsthatdo,exportonly14percentoftheirtotalshipments,

on average. Evenwithin the largest U.S. manufacturing sectors—fabricatedmetal

products,printingandrelatedsupport,andfoodmanufacturing—only14percent,5

percent, and 12 percent, respectively, of firms in those sectors export. These few

exportersalsotendtoexportasmallshareoftheirtotalshipments;12percent,14

percent,and15percent,respectively.

Table1.1:ExportingbyIndustryinColombia

IndustryPercentofFirms

PercentofExporters

MeanRatioofExportstoTotal

RevenueFoodManufacturing 16.9 23.6 27.8Beverages 2.1 20.0 20.9TabacoProducts 0.1 91.2 16.0Textiles 4.4 52.5 16.9Apparel 10.2 58.7 30.7LeatherProducts(ExcludingFootwear) 1.5 77.9 44.7FootwearandRelatedProducts 1.8 55.4 18.6WoodProductsManufacturing 1.6 32.9 26.7Paper,CardboardAndDerivatives 2.2 60.4 22.6PrintingandRelatedSupport 7.9 46.2 18.9ChemicalManufacturing 12.3 56.3 17.8RubberProducts 1.3 59.4 20.4PlasticProducts 8.5 54.6 14.8GlassAndGlassProducts 0.7 73.9 30.0NonmetallicMineralProducts 2.9 40.1 27.2CementandOtherProducts 1.7 26.1 18.9PrimaryMetalManufacturing 2.2 49.9 31.8FabricatedMetalManufacturing 8.3 51.4 20.3AutomotiveVehiclesandParts 3.8 64.4 21.3OtherTransportationVehiclesandParts 0.6 49.9 24.7OtherManufacturingIndustries 9.0 50.1 20.8Note:Authors’calculationsbasedondatafromColombia’scustoms(DIAN)andsupervising(SIREM)agenciesfrom1995to2011.Column2istheaveragepercentageofmanufacturingfirmsinthegivenindustry; Column 3 is the average percentage of firmswithin the given industry that export; andColumn4 is the average ratio of exports to total shipmentswithin a given industry, for firms thatexport.

6

WhiletheseresultscouldbeexplainedbythefactthattheUnitedStateshasa

largedomesticmarket,itisalsothecasethatexportingisnotubiquitousforfirmsin

other, smaller and developing markets. Indeed, using data from Colombia to

corroboratethefindingsinBernardetal.(2007),wefindasimilarpatterninTable

1.1,whereexportingtendstobeasmallshareoftotalsales(22percentonaverage).

InColombia’s largestmanufacturing industries—foodmanufacturing, apparel, and

chemicalmanufacturing—23.6percent, 58.7percent, and56.3percent of firms in

each of those sectors export, respectively. Further, for these same industries, the

mean ratio of exports to total revenue for exporting firms is 27.8 percent, 30.7

percent, and 17.8 percent, respectively.5Clearly, many firms in Colombia do not

export, even in the largest sectors, and those that do, tend to earnmost of their

revenuesfromthedomesticmarket.

1.2 ExportersAreDifferentfromNon-Exporters

Whileourfirststylizedfactmakesitclearthatfewfirmsexport,itisnotclearwhy

this is so. Indeed, it may be that firms are simply reaching different consumers,

someofwhomhappentobeabroad,andthattherearenofundamentaldifferences

between exporters and non-exporters. Alternatively, it could be the case that

exporters are just different from non-exporters in other characteristics, and that

these characteristics induce selection into exporting. The data clearly support the

latter hypothesis, whereby exporters are fundamentally different from non-

exporters,eveninexportingsectors.Itisthisself-selectionthatformsthebasisfor

mostofthetheoreticalworkthatweoutlineinthesectionsthatfollow.

Again drawing on the work of Bernard et al. (2007), U.S. firms display

numerous export “premia”: exporters have higher employment, shipments, value-

5UnlikethedatausedinBernardetal.(2007),ourdataareavailableonlyforthelargestproducers,and the findingshereexclude the firms thatare less likely toparticipate in theexportmarket.Weexpect that thesenumbersareoverstatedandwilldecreasewhensmallermanufacturing firmsareincluded. For further evidence, using a larger (but older) data sample of Colombian firms, Brooks(2006)findsthatonly10–20percentofColombianmanufacturingfirmsexportedinthe1980sandthattheaverageexportsharewasaround20percent.

7

added per worker, productivity, wages, capital per worker, and skill per worker.

Casas, Díez, and González (2016) find a similar exporter premium for Colombian

firms,whichwehavereplicatedinTable1.2.Forexample,theyfindthatexporters

havehigherwages,value-addedperworker,incomeperworker,capitalperworker,

productivity, and investment per worker.6The key finding in these and other

studies is that firmproductivity iscorrelatedwithexporting; that is, inallsectors,

themost-productivefirmsaremorelikelytoexport.7

Table1.2:DifferencesbetweenExportingandNon-ExportingFirms

Wage Value-

AddedIncome Capital Investment Productivity

ExporterPremia

0.299 0.408 0.350 0.347 0.446 0.223(0.0156) (0.0218) (0.0239) (0.0324) (0.0346) (0.0250)

Obs. 25,979 26,042 26,130 26,130 25,091 26,130Notes:Standarderrors(inparentheses)clusteredbyfirm;allestimateshavestatisticalsignificanceat the 1 percent level. Productivity is calculated using themethod found in Gandhi, Navarro, andRivers (2016) and for the other estimates the dependent variables are measured in billions ofColombian pesos of 2005 perworker. All specifications include controls of year and sector. TabletakenfromCasas,Díez,andGonzález(2016).

1.3 ExportingIsVeryConcentrated

Whilethedatashowthatexportingfirmsaredifferentfromnon-exportingfirms,the

data also show that there is heterogeneity even within exporting firms.

Surprisingly,datashowinghowrareitisforfirmstoexportmayactuallyoverstate

howmuchexportingactuallytakesplace.Forexample,againdrawingonBernardet

al. (2007),exportersthatsellonlyoneproducttooneexportmarketrepresent40

percentofexporting firms in theUnitedStates,butonly0.2percentof theexport

6 For other studies identifying differences between non-exporters and exporters, and evendifferenceswithinexportingfirms,seeEaton,Kortum,and Kramarz (2011),Mayer,andOttaviano(2008),Casas,Díez,andGonzález(2015),andClerides,Lach,andTybout(1998).7Whilemostpapers,usingvariousmeasurementsof firmproductivity, findevidence thatonly themost-productivefirmsexport,themostcommonmeasurementofproductivity,TFP,iscalculatedbyestimating the production function. Gandhi, Navarro, and Rivers (2016) argue that estimating theproduction functionsuffers frombiasand find thatcorrecting for thebias, in thecaseofColombiaandChile,resultsinsmallerproductivitydifferencesinmostcasesandnodifferencesinothers.

8

value. Incontrast, firmsexporting fiveormoreproducts to fiveormorecountries

accountfor13.7percentofallexportingfirmsand92.9percentofexportvalue.8A

similarpatternappearsindatafromColombia.AsshowninTable1.3,theshareof

single-product, single-destination exporters in Colombia is 16 percent, but these

firmsaccountforonly0.3percentoftheexportvalue.Ontheotherhand,theshare

offirmsexportingmorethan10productstomorethan10countriesis11.5percent

offirms,butthesefirmsaccountforalmost60percentoftheexportvalue.Clearly,

exportingisconcentratedanddominatedbyafew,very-successfulfirmsthatexport

many products tomany destinations, andmodels of firms in trademust consider

bothselectionintodifferentdestinationsaswellasanexpandedproductset.9

8Another way to analyze the importance of large firms is to focus only on the top exporters(disregardingthenumberofdestinationsorproducts).Forexample,FreundandPierola(2016)focuson the top five exporters in 40 developing countries, and find that these “superstars” on averageaccountforonethirdofexportvalueandoverhalfofexportgrowthoverafive-yearperiod.9Theseissuesarealsogermanetothestudyofmultinationals(seeChapter9of),theorganizationofthefirmintooffshoringandoutsourcing(seeChapter10),andtheroleoftradeintermediaries(seeChapter11).

9

Table1.3:Multiproduct/CountryFirms

(a)Theshareoffirms

Countries

1 2 3 4 5 6-10 >10

Product

1 16.0 3.9 1.4 0.5 0.3 0.5 0.62 4.8 3.6 2.7 1.3 0.7 1.1 0.83 2.3 2.1 1.7 1.3 0.5 1.2 0.64 1.4 1.1 0.9 1.4 0.3 1.5 0.95 0.2 1.0 0.5 0.7 0.2 2.2 0.56-10 2.1 1.9 1.5 1.9 1.3 5.3 2.5>10 1.2 1.1 1.3 1.2 1.0 5.8 11.5

(b)Theshareofexportvalue

Countries

1 2 3 4 5 6-10 >10

Product

1 0.3 0.7 0.1 0.1 0.0 0.3 5.32 0.4 0.2 0.5 0.2 0.2 0.8 3.63 0.1 0.2 0.1 0.1 0.1 0.7 0.44 0.1 0.2 0.1 0.1 0.2 0.6 1.55 0.0 0.1 0.0 0.1 0.1 0.7 4.96-10 0.1 0.2 0.1 0.4 0.3 2.5 5.3>10 0.1 0.2 0.3 0.3 0.4 6.9 59.9

Note: Calculations based on data from SIREM andDIAN/DANE for 2013. Table taken from Casas, Díez, andGonzález(2016).

1.4 ExportingFirmsStartSmallandMostFail,butThoseThatSurviveExpandThefinalstylizedfactswediscussarethesurvivalrateofexportersandthegrowth

ofthosethatdosurvive.Giventhelong-runnatureoftheclassicliterature,dynamics

were rarely considered (along with the firms themselves). However, in many

countries,newfirmsattempttoexportandothersexitineveryyear.InTable1.4we

summarize these statistics for Colombia, where 34 percent of exporters in 2000

were entrants (new exporters), 23 percentwere continuing exporters (firms that

continuouslyexportedbetween1994and2000),and43percentwereexperienced

exporters(firmsthatexportedinsomebutnotallyearsbetween1994and2000).

Thesestatistics,however,understatetheimportanceofcontinuingexporters,which

accounted for 70.5 percent of the export share in 2000. Further, continuing

10

exportersexportanaverageof5.6millionUSDperfirm(comparedwith1.2million

perfirmforexperiencedexportersand0.1millionperfirmforentrants).

Table1.4:ExportsbyFirmType

FirmType NumberofFirms

ShareofFirms(%)

Exports(mn,USD)

ShareofExports(%)

ExportsperFirms(mn,USD)

ContinuingExporters 1,630 23.2 9,098 70.5 5.6ExperiencedExporters 3,016 43.0 3,622 28.1 1.2Entrants 2,372 33.8 179 1.4 0.1Note: Continuing Exporters are firms that exported every year from 1994 to 2000; ExperiencedExportersare firms that exported both in 2000 and in one, but not every, year before 2000; andEntrantsarefirmsthatexportedin2000forthefirsttime.

While the importance of continuing exporters is clear, new exporters are

importantinotherdimensions.Althoughentrantstendtostartsmall,andmostwill

notexportbeyondoneyear,thosefirmsthatdosucceedtendtogrowandbecome

important contributors to export growth.As shown inTable1.5, exportsper firm

increaseforallfirmtypes,butexportsperfirmincreaseatafasterrateforentrants.

Indeed, after five years, exports per firm for entrants increase by a factor of 22

(comparedwith a doubling for continuous exporters and tripling for experienced

exporters),andafter10yearstheyincreasebyafactorof51(comparedwithfour

forcontinuingexportersandsevenforexperiencedexporters).Additionally,aftera

few years of exporting, the hazard rate (the share of firms that exit the export

market at time t, conditional on having survived the previous year) is almost the

same forall firmtypes.Thekeydifference is thehazardrate in2001,which is60

percentforentrants(theirinitialyear).Incontrast,thehazardrateis7.4percentfor

continuing exporters and 27.1 percent for experienced exporters. Clearly, new

exporters have a high likelihood of failure, but, conditional on success, they grow

quickly and their international operations stabilize in line with experienced and

continuingexporters.

11

Table1.5:ExportDynamicsbyFirmType

FirmType 2000 2001 2005 2010

ExportsperFirm(mn,USD)

ContinuingExporters 5.6 5.3 9.7 21.5ExperiencedExporters 1.2 1.5 3.8 8.0Entrants 0.1 0.5 1.7 3.9

HazardRateContinuingExporters - 7.4 3.9 4.6ExperiencedExporters - 27.1 14.6 8.1Entrants - 59.2 17.1 5.4

Note:ContinuingExportersarefirmsthatexportedeveryyearfrom1994to2000;ExperiencedExportersare firms that exportedboth in2000and inone,butnotevery, yearbefore2000;andEntrantsarefirmsthatexportedin2000forthefirsttime.Hazardrate isthefailureratefor exporters (thepercentage of firms that exit the exportmarket) at time t, conditional onhavingsurvivedatt-1.

Newexportersmaystartsmallandmaynotseemimportant,buttheygrow

quickly if they survive and, in aggregate, are the key to future export growth.

Indeed,Eatonetal.(2007)find,forexample,thatnewexportingfirmswillaccount

for almost half of total export growth within a decade. Overall, the dynamics

discussed here have spurred a new literature trying to explain why first-time

exportersstartsmall,whysomanyfirmsstopexportinggiventhehighsunkcoststo

export, why conditional on export survival firms tend to expand, and even why

export failure may have lasting negative consequences for firms. Understanding

these firms and answering the questions posed here are fundamental to

understandingexportdynamicsandexportgrowth.

1.5 MotivatingNewModels

To summarize the key empirical facts from this section, it is clear that there is

substantialheterogeneityininternationaltradethatoriginatesatthefirmlevel.Not

all firmsexport,and those thatdo tend tobemoresuccessful thannon-exporters.

Even within the set of active exporters, there is substantial heterogeneity. The

biggestexporterstendtobereallybig.Theyaccountforamassiveshareofexported

value.Further,thereissubstantialvariationinwhetherexporterssucceed.Somefail

andexitquickly.Otherssucceedandexpandovertime.

12

Onalmosteverydimensionof internationalcommerce, there isselectionat

the firm level. Indeed, this selection has led economists to develop trademodels

with heterogeneous firms to explain this selection, and raise new questions that

cannotbeexplainedbyclassicmodels.Forexample,aretheexportpremiatheresult

of exporters “learning from exporting” or is there something about the exporting

firmsthatallowsthemtobecomemoresuccessfulandabletocovertheadditional

costs of exporting? Or can reallocation of resources from non-exporting firms to

exporters raise aggregate productivity and thus also add a new channel for gains

fromtrade?Wenowoutlinethebasicmodelsusedtoanswerthesequestions.

2 ModelingFirmsinTrade

Withsomebasicstylizedfactsinhand,wenowdevelopthetheorybehindthefirms-

in-traderevolution.Formodeling firms in international trade, the firstquestionto

consideriswhetherfirmsare“big”or“small,”whichdeterminespreciselywhether

thesefirmstakeintoaccounttheeffectsoftheiractionson“economicaggregates.”

Do firms have enough purchasing power to affect the wage? Do firms take into

accountthedirecteffectoftheiroutputchoicesontheincentivesofotherfirms(àla

Cournot)?Ordofirmsactsmall,eithertakingpriceasgivenoronlyhavingcontrol

overasmall“monopoly”marketfortheirownvariety?

Despite the analytical complications from doing so, an early literature on

firms in trade focusedonoligopolyas themain framework formodeling strategic

decisions. For example, Brander and Krugman (1983) and Brander and Spencer

(1985) each evaluate the effects of trade costs (the former) and subsidies (the

latter) on the strategic behavior of firms in an oligopoly framework. Eaton and

Grossman (1986) determine the optimal tariff in a conjectural variations model,

focusing on the how the nature of competition determines this optimal policy.10

Whileenlighteningforpolicydiscussionsovertariffsandsubsidies,theliteratureon

strategictradesuffersfromananalyticalcomplexitythatseverelylimitsitsabilityto10Conjecturalvariationsisamodelingstrategywherefirmsassume,or“conjecture,”theresponseofotherfirmstotheirownstrategicdecisions.

13

appropriatelyrepresent firms in international trade.That is, firm-heterogeneity in

productivity,oneofthehallmarksofrecentempiricalworkintrade,isverydifficult

tomodel in an environmentwith large firms that havemarket powerwithin and

acrossvarieties.11Thisalsomakesitdifficulttoanalyzeself-selectionintoexporting

byproductivity,which is at the center of positive andnormative questions at the

intersectionoftradepolicyandthetheoryofthefirm.

Recognizing the difficulty with oligopolistic competition, a significant

majority of empirically and theoretically focused work on international trade,

starting with the seminal work of Krugman (1979, 1980), begins with the

assumption of monopolistic competition. Monopolistic competition assumes that

firms havemarket power in their own varieties but do not take into account the

effectsoftheirdecisionsoneconomicaggregates(priceindex,wage,marginalutility

ofincome,etc.).Tostudywherethistakesus(andmoreimportantly,whereitdoes

not), we begin our discussion of firms and trade with a simplified version of

Krugman (1979, 1980), which, in many ways, has formed the basis of our

understandingoffirmsintheinternationaleconomy.

2.1 TheKrugmanFramework

The Krugman framework delivers a parsimonious, general equilibrium model in

whichconsumerslovevarietyandfirmscanexploiteconomiesofscale.Itishardto

overstate its simple genius; consequently, the foundations of this model have

formedthebasisforthevastmajorityofworkonfirmsintradesinceitspublication.

The starting point for the Krugman framework is a set of consumerswho

have preferences over a variety of goods. Specifically, consumers have constant

elasticityofsubstitution(CES)preferencesofthefollowingform:

11Recentresearchallowsforamoregeneraltreatmentofmarketpowerwithfirmheterogeneitybyassuming that themost productive firms enter themarket first. SeeAtkeson andBurstein (2008),Eaton,Kortum,andSotelo(2015),andGaubertandItskhoki(2015).

14

𝑈 = 𝑞!!!!!!

!!!

!!!!

.

Here,thereisnooutsidegood,andconsumersearnutilityonlyinthedifferentiated

sector,with𝜎representingtheelasticityofsubstitutionacrossNvarieties,indexed

byj.ThereareNavailablevarieties,witheachvarietyproducedbyonefirm.Inthe

closed-economymodelthatfollows,allvarietiesmustbeproduceddomestically.

Firms in themodelareverysimple.Toenter themarket, firmsmusthireF

workers tosetup the firm,where theseworkersundertakeupfrontcosts, suchas

R&D andmarketing. Eachworker is paid a wagew, and all firms have the same

technology.Afterentering,firmsproducevarietiesusingadditionallabor,withone

unit of labor required for each unit of production.When pricing their individual

variety, firms take aggregates as given, but behave as monopolists in their own

variety.Firmsenteruntiltotalprofitsareequaltozero.

Initsmostbasicform,therearethreeequilibriumconditionsintheKrugman

framework. The first is pricing.Whilewe generalize thepricingdecisionby firms

rigorously in later sections, for now,we simply assert that givenCESpreferences

and the assumption of monopolistic competition, firms charge a price that is a

constantmark-upabovemarginalcost:

𝑝 = !!!!

𝑤.

As𝜎rises and varieties become more substitutable, the mark-up above marginal

costfalls.Ofcourse,aswagesrise,pricesdosoproportionally.12

Thesecondequilibriumconditionisfreeentry;firmsenteruntiltotalprofits

equal zero. This condition yields the following “break-even” production level for

eachfirm:

𝑞 = 𝜎 − 1 𝐹.

Beforeintroducingthethirdcondition,noticethatoptimalpricesandquantitiesin

theCESKrugmanmodelarethesameforallfirms.Thisresultsfromtheassumption

that all firms have identical technology. Given the statistics on firm-level trade

12ThisisanartifactofCESpreferences,apointwereturntolater.

15

presentedinsectionone,thisisakeydownsideoftheKrugmanframeworkthatwill

beenhancedinlatersections.

Finally, the last condition is labormarket clearing.Here, a labormarket of

sizeLmustsupportalllabordemandbyN firms,whetheritbefixedupfrontlabor

demandorvariablelaborrequirementsforproduction.Bysettingdemandequalto

supply,wehave:

𝑁 =𝐿𝜎𝐹.

Here, the number of varieties that consumers value due to CES preferences

increasesinthesizeofthemarket.Thiscompletesthecharacterizationoftheclosed

economy.

Krugman’smodelcanbeextended“internationally” inanynumberofways

(for example, many countries, asymmetric labor endowments, tariffs), but the

easiestway tostudy the impactof tradeon theequilibrium is simply todoubleL.

This represents an experiment of onemarket integratingwith anothermarket of

equal size.With this experiment, and the threeequilibriumconditionsabove, it is

clearthattheonlychangeintheequilibriumisthenumberofvarietiesavailableto

the consumer in eachmarket. Pricing does not change relative to thewage, since

pricingunderCES isaconstantmark-upovermarginalcost.Outputper firmdoes

notchangebecausethemark-updoesnotchangewithmarketsizeandfirmsenter

untilzeroprofitsarereached.Thenumberofvarietiespercountrydoesnotchange,

butsincetradeisfree,thenumberofvarietiesavailabletoeachconsumerhasrisen

bya factorof two.Thus, consumersgain from trade throughanexpandedvariety

set.

2.1.1 MovingBeyondKrugman

TheKrugmanmodelwasnevermeanttopreciselymatchthecharacteristicsoffirms

intheglobaleconomy.Rather, itsmainobjectivewastopresentamodel inwhich

economiesofscale lead to tradebetweenotherwise-identicalcountries. Insteadof

16

tradingendowmentsasembodiedingoods,countriesaretradingvarietiesofsimilar

products.

It does not take an expert to realize that the Krugman framework—in

particular,theversionpresentedabove—ismissingsomeotherkeyfeaturesofthe

datathatmaybeimportant.Again,asintheclassicliterature,allfirmsarethesame

size, operating with the same technology, and presented with the same

opportunities to export. In reality, firms typically havedifferent technologies, and

thosefirmswithbettertechnologyaretypicallyabletoexportmorereadily.Tothis

end,agreatmajorityofresearchstudyingfirmsintradehasusedthefundamental

insight of Krugman within expanded models with heterogeneous technology and

other firm attributes. While maintaining the assumption of monopolistic

competition, researchers then appeal to probability distributions that define firm

attributes to describe the mix of firms and elegantly evaluate how policy and

geographyaffect themixof firms servinga givenmarket.Wenowoutline abasic

frameworkoffirmsandtradethatsummarizesthisliterature,andtheroleoftrade

costsandtradepolicyinguidingthelandscapeofinternationalcommerce.

2.2 ExportingandMonopolisticCompetition

In servinganexportmarket,weassume that firmschoosequantities tomaximize

profits. To reach the destinationmarket, firmsmust first build their product at a

cost c per unit, and may pay a host of costs related to trade and trade policy.

Specifically,therearefourcostsoftradethatmayaffectthedecisiontosellabroad.

The first is themelting iceberg trade cost,𝜏 > 1,which is precisely thenumberof

units of a good that a firm must produce for one unit to reach the destination

market.Thereareanumberofwaystointerpretthistradecost.Explicitly,thereisa

possibility that goods are stolen or damaged en route, and therefore to fill a

customer’sorderthereisanexpectationthatmoregoodsaresentthanisrequired.

In reality, many firms do not send extra goods, but instead insure against this

probability;hence,𝜏mayalsobeinterpretedinthisway.Anotherinterpretationof

the icebergcost is in thecompoundedcostsof financing thatoccurover timeasa

17

good is shipped from one location to the next, or the compounded risk from

shipping(Schmidt-Eisenlohr(2013),seesectionfour).

Exportingfirmsmayalsobesubjecttoaper-unitorspecificcost,s.Thiscost

is also related to distance, and can be easily interpreted as the physical cost of

movingthegood,whetherduetoweightorweightoveradistance.13Further,some

trade policy measures are assessed per unit (as opposed to as a percentage of

value). In recentwork, usingNorwegian firm-level data, Irarrazabal,Moxnes, and

Opromolla (2015) estimate that specific trade costs are14percent of themedian

priceoftradedgoods.

Therearealsofixedcostsofexport,Fx,whichdonotdependonoutput.These

costs tend to be interpreted as organizational, logistical, or other administrative

costs involved in shipping the good from the plant to the customer in another

market.Often,thesecostsarereferredtoas“redtape,”andmanyoccurattheport

ofentry.Indeed,inarecentsurveybytheOECD-WTO(2015),countrieswerealmost

twice as likely to report trying to improve general border procedures, and 30

percentmorelikelytoreporttryingtoimprovetransportinfrastructureinaneffort

tofacilitatemoretradethantheyweretoreducingexplicitpolicycostsandfees.14

A final cost related to international trade is the import tariff, which is

typically advaloremandassessedon thevalueof thegoodwhen it arrivesat the

destination.Tomodel the tariff, it is usually assumed that if the consumerpays a

pricep, the firm receives a pricep/t,where t is the tariff factor (one plus the ad

valorem tariff). This particular valuationmethod is called “Cost of Insurance and

Freight,” or CIF, and it will form the basis of our treatment of trade costs in the

modelthatfollows.15

Havingdefinedthecostsofexporting,wenowmovetodefiningtheobjective

functionofthefirmservingaforeignmarket.Above,wedefinedproductsbyj,and,13Forexample,fuelcostsaretechnicallythecostofenergyrequiredtomoveaspecificmassoveradistance.14SeeFigure2.17inhttps://www.wto.org/english/res_e/booksp_e/aid4trade15_e.pdf15Analternatemethod,“FreeOnBoard”orFOB,isusedbyahandfulofcountries,andleviesthedutyonthevalueofthegoodnetofthecostofinsuranceandfreight.Johnson(1966)discussesthewayinwhichCIFvaluationnaturallydiscriminatesagainstexportersfromdistantcountriescomparedwithFOBvaluation.

18

below,weassumethatfirmseachproduceoneproduct.Thus,theprofitfunctionof

anarbitraryfirmisellingtoanexportmarketiswrittenas:

(1) Π! =!!𝑝(𝑞!)𝑞! − 𝜏𝑐!𝑞! − 𝑠𝑞! − 𝐹! .

In the profit function, we note that the only heterogeneity across firms is in the

marginalcostof firm i, 𝑐! ,and in the firm’soutputchoices(whichare,ofcourse,a

functionof𝑐! in equilibrium).Thismarginal cost is inverselyproportional to firm-

level productivity and, in extendedmodels, can be a function of factor and input

prices.16In the literature, it is common to assume that transport costs and fixed

costs are homogeneous across all firms, although in reality firms may face

heterogeneous trade costs exogenously or endogenously (section four discusses

waysinwhichcostsmaybeendogenousthroughtradefinance).

As mentioned above, firms maximize profits by choosing quantities.

Suppressingi’sfortheremainderofthemanuscript—understandingthatcdefinesa

given firm producing a particular product—the maximization problem yields

optimalpricingfortheexportmarket,

(2) 𝑝 𝑞 = !(!)! ! !!

𝜏𝑐 + 𝑠 𝑡,

where𝜀 𝑞 = − !"!"

!!istheelasticityofdemandinabsoluteterms.Dependingonthe

assumption over the utility function, this elasticity may be constant (CES, as in

Krugman 1980 and Melitz 2003), variable and falling in output (Melitz and

Ottaviano2008),ormayhavesomeotherproperties(seeMrázováandNeary2015,

forageneraltreatmentoftheconvexityofdemandandselectionintoexporting).

Ofnote,thepricederivedin(2)istheconsumerprice,andthispricecanbe

brokenup into three terms: themark-up !(!)! ! !!

,marginal costs (𝜏𝑐 + 𝑠), and the

tariffadjustmentfactor(𝑡).Note,however,thatthetariff factordoesnotaffectthe

pricethattheproducerreceives,! !!,unlessitaffectsthemark-up.Thiswillhavean

importanteffectontheresponseoffirmstoagivensetofinternationalshocks,since

16For thismanuscript,we abstract fromwage and other input prices.However, all results in thismanuscriptfollowwhenassumingthatcostsareinlaborunitsandthatthecostsincludeascalarforthewage.

19

the effect of the shockswill depend on the pricing conduct andmarket power of

eachfirm,asembodiedinthemark-up.

Implicitly, using the particular functional form of𝑝 𝑞 , equation (2) pins

downthevalueofqasafunctionofmarginalcostsandtariffs.Thefirmthensolves

fortheprofitsofservingtheexportmarketanddecideswhetherornottoenter.We

nowmovetothisbasicentrydecision,focusingontheroleoftradecostsandtrade

policyinthedecisiontoexport.

2.3 SelectionintoExporting

Thefirstdecisionafirmmustmakeiswhethertoenteranexportmarketorremain

domestic.Asspecifiedin(1),thefirmbalancesvariableprofitsagainstafixedcostof

exporting whenmaking this decision.17Below, we describe this selection process

usingavarietyofassumptionsovertherevenuefunction,and,later,werefertothis

marginasthe“extensivemarginoftrade”

CESDemand

Like the Krugman framework derived in section 2.1, the canonical model of

exportingwithheterogeneousfirms,asdevelopedinMelitz(2003),usesaconstant-

elasticitydemandfunctiontomodelconsumers.Inthisframework,theelasticityof

demandisconstantandequalto𝜎,and,hence,using(2)consumerpricesare:

(3) 𝑝 𝑞 = !!!!

𝜏𝑐 + 𝑠 𝑡.

Pluggingintotherevenuefunction,wefindthatprofitsare:

(4) Π = 𝐴𝑡!! 𝜏𝑐 + 𝑠 !!! − 𝐹! ,

17A separate literature has developed that allows for variable profits in multiple markets to belinkedthroughthecostfunction,therebymakingtheentrydecisionmorecomplicated.SeeAhnandMcQuoid(2015);Spearot(2012);Blum,Claro,andHorstmann,(2013);McQuoidandRubini(2014);Soderbery, (2014), and Spearot (2013b). Mora (2016) allows formarkets to be linked through afinancialconstraint.

20

where𝐴 is function of aggregate variables and parameters, and, below, will

represent a demand shifter in the export market. The firm chooses to enter the

marketind iftheprofitsfromdoingsoaregreaterthanzero.Inthiscase,variable

profits are always positive due to the CES assumption (discussed below), so the

questionofselectionboilsdowntowhetherthesevariableprofitsaregreaterthan

the fixed cost of exporting, 𝐹! . Intuitively, this occurs if production costs are

sufficientlylow:

(5) 𝑐 < !!

!!!!!

!!!! − !

!.

Notethatin(5),differentlyfromMelitz(2003),wehaveincludedadvaloremtariffs

aswellasper-unittransportcosts.Indeed,withtheadditionofperunitcosts, it is

clear that if per-unit costs are sufficiently high, there exists no positive range of

productioncosts,c,suchthattradeoccurs.Thatis,thereare“tradezeros”through

endogenousfirmselection.Althoughdifferentfromthetreatmentoftradezeroesin

Helpman, Melitz, and Rubinstein (2008), where trade zeros occur because of a

bounded distribution of productivity (firms can only be so good), this result

highlightshownaturalforcesofgeographyandtradecostscanlimittrade,notonly

byfirms,butalsobetweencountries.

Forclarity,wenowsimplify themodel to that in theoriginalMelitz (2003)

framework,where therearenotariffs, t,andnoper-unit transportcosts,s. In this

case,exportingoccursif:

(6) 𝑐 < !!

!!!

!!!!.

In(6),wefindanumberofintuitiverelationshipsbetweentradecostsandselection

into exporting. Both the fixed costs of exporting and the iceberg transport cost,

when larger, reduce the rangeofmarginal costs so that exportingoccurs. That is,

withhighertradecosts,tosuccessfullyexportafirmmustbeevenmoreproductive.

Importantly,thisequationalsomakesitclearthatdestination-marketfactors,such

asthedemandshifter𝐴,affectthe incentivestoexportandtheroleof tradecosts.

Indeed,amoredistantmarketcansupportmoreexportingifthemarketsizeislarge

enough to compensate for the additional costs of trade. Empirically, this is an

21

important issue to consider, since it suggests a necessary empirical strategy that

requiresdestination-marketeffects,potentiallybyindustry.

Non-CESDemand

CES demand holds special properties that, while making analysis tractable, are

empirically implausible. In particular, under the assumption of monopolistic

competition,demandisalwayspositive,regardlessoftheprice,anddemandvalues

areasymptotictoinfinitevalueswhenpricesarelowenough.Ironically,bakedinto

theCESframework—whichusuallyrequiresthatfirmsbesmall—isthepossibility

that a firm is extremely large as it gets very productive. Tomove away from the

possibilityof“superfirms”withinamodelofmonopolisticcompetition,researchers

haveexploredtheuseofotherutilityfunctionsthatmaybelesstractablebutmatch

thedatabetterinthisandotherdimensions.Thetwomostcommonlyusednon-CES

preferencesareTranslogandContinuumQuadratic.18

For the following section, we focus on the continuum quadratic, and, in

particular,anextendedversionoftheMelitzandOttaviano(2008)utilityfunction.

Precisely, continuum quadratic preferences yield a (linear) inverse demand

function,representedby𝑝(𝑞) = 𝐴(𝑄)− 𝑏𝑞,whereA(Q)isanaggregatetermthatis

decreasing inquantity sold to consumers in thatmarketbyall firms.However, as

motivated above, to keep firms small,we assume firms takeA as given.With this

linear demand function and the assumption of small firms, the profitmaximizing

producerpricefrom(2)issimplifiedas:

(7) ! !!= !! !"!! !

!!".

Here,wefindanumberof interestingdifferencesbetweenlinearandCESdemand

functions in termsof theeffectsof tradecostsand tradepolicy.Most importantly,

wefindthatproducerpricesarenotproportionaltomarginalcosts,eitherthecosts

18 Feenstra (2014) outlines a firm-heterogeneity model using a general form of homotheticpreferenceswithvariableelasticities.Bertoletti,Etro,andSimonovska(2016)discussthegainsfromtrade in a utility function with indirect additivity. Rodriguez-Lopez (2011) develops a firm-heterogeneitymodelusingatranslogexpenditurefunction.

22

relatedtofirm-levelproductionorthoserelatedtotransportation.Thisresultsfrom

the variable elasticity of demand under the assumption of linear demand, and, in

particular, the assumption that absolute elasticities fall with lower costs (higher

output).

Plugging (8) into the revenue function, profits under linear demand with

fixedexportingcostsarewrittenas:

(8) Π = !! !"!! ! !

!!"− 𝐹! ,

where it is clear why linear demand becomes less tractable with fixed exporting

costs. While it is true that lower costs increase the probability of exporting, the

functional formof this cutoff is far less tractable than thatofCES.Precisely, firms

exportifcostsmeetthefollowingcondition:

(9) 𝑐 < !!"𝐴 − 2 𝑏𝑡𝐹! − !

!.

AswithCES,iftheper-unitcostsoftransportationsaretoohigh,nofirmwillbeof

sufficient productivity to export. However, unlike with CES, even when per-unit

costs of transportation are zero, it is still possible that no firm will export.

Specifically,settingsequaltozeroin(9),wehave:

(10) 𝑐 < !!"𝐴 − 2 𝑏𝑡𝐹! .

Intuitively, since monopolists never produce beyond the point of unit elasticity,

revenuesareboundedunderlineardemand.Thus,ifthefixedcostsofexportaretoo

high,nofirmwithapositivemarginalcostcanentertheexportmarket.

Given theseproperties, the literaturerarelyuses fixedexportingcostswith

lineardemand.Instead,theliteraturefocusesonselectionthroughtheicebergcost

(Melitz and Ottaviano 2008), or the tariff (Spearot, 2013a), where the selection

condition is simply𝑐 < !!". Additionally, selection in the linear model is also

proportionaltodemandlevelsandtradecosts,asinCES.So,whilethelinearmodel

isslightlymorerestrictiveintermsofthetypesoftransportcoststhatareused,the

modelissimilarintheuseofadvaloremtradecosts,anditultimatelyfacilitatesan

elegantanalysisoftradeinwhichmark-upsmayvary.

23

2.4 Firm-LevelTrade

Theroleoftradecostsdoesnotendwiththedecisiontoexportornot.Inparticular,

therearea varietyof variable trade costs that canaffect the intensitywithwhich

firms trade; we refer to increases in average exports by firms as the intensive

marginof trade in section three.Next,webriefly examine these incentiveswithin

theCESandlineardemandmodels.

CESDemand

Returning to theCESmodel, firm-level trade revenues, conditionalon trading, are

easilyderivedas:

(11) 𝑣 = 𝜎𝐴𝑡!! 𝜏𝑐 + 𝑠 !!! .

To evaluate the role of market and trade costs shocks on this value, we log-

differentiate(11)withrespecttoA,andalltradecostandtradepolicyparameters.

(12) 𝑣 = 𝐴 − 𝜎𝑡 − 𝜎 − 1 !"!"!!

𝜏 − 𝜎 − 1 !!"!!

𝑠.

Here, “hats” refer to percentage changes in the variable. Equation (12) makes it

clearthat,undertheassumptionofCESdemand, firm-level traderevenueschange

proportionally with demand shifters and tariffs, and proportionally with other

changesintradecostsscaledbytheircostshare.Unfortunately,icebergandper-unit

tradecostsaretypicallynotmeasuredseparately(ifatall),sodeterminingthecost

sharesofdifferenttypesoftradecostswouldproveadifficultendeavor.However,

as described above, one could argue that the former are related to insurance and

valueandthelattertoweight,whichwouldgiveasenseoftheircostshares.

Typically, the literature uses the iceberg cost interchangeably as both a

“distance” and “trade policy” parameter. Equation (12) makes it clear that this

interpretation is not correct for two reasons. First, even if there are no per-unit

transportcosts,theelasticityoffirm-leveltradewithrespecttotariffsislargerthan

theelasticitywithrespecttotradecosts.Thisisbecausethetariffreducesrevenues

24

directly, and also through the optimal choice of quantity.19For iceberg costs, the

onlyeffectonrevenueisthroughtheoptimalchoiceofquantity.Second,withnon-

zero per-unit trade costs, it is clear that any elasticitywith respect to distance is

contaminatediftheshareofadvaloremcostsisnotmeasuredappropriately.

Non-CESDemand

Theanalysisofdemandandtradeshocksundertheassumptionoflineardemandis

morecomplicated.First,notethatfirm-leveltraderevenuesarederivedas:

(13) 𝑣 = !!!!! !"!! !.!!"

Log differentiating (13) with respect to A and all trade cost and trade policy

parameters yields a reasonably complicated function related to parameters, trade

costshocks,andthelatter’sshareofmarginalcosts.However,asshowninSpearot

(2013a),thiseffectoftradeshockscanbesimplifiedasfollows:

(14) 𝑣 = !!

!!"#𝐴 − 𝑡 − !"

!"!!𝜏 − !

!"!!𝑠 + 𝑡.

Finally, noting that thehighest value of revenues that a firm can earnwith linear

demandsis𝑣!"# =!!

!!",(14)canbefurthersimplifiedto:

(15) 𝑣 = 2 !!"#!

𝐴 − 𝑡 − !!!"!!

𝜏 − !!"!!

𝑠 + 𝑡.

Equation(15)makescleartheeffectoftradecosts,tradepolicy,anddemandshocks

inanenvironmentwithdemandthatismoreinelasticforhigherquantities.Within

the parentheses in equation (15), the effects of trade cost and policy shocks are

similartothosewithCES.However,alltradeshocksmustbeweightedbyfirm-level

export revenues,𝑣, and it is predicted that a given set of shocks will be less

pronouncedinpercentagetermsonfirm-leveltradewhenthefirmis larger(more

productive).

19Toseethisin(11),notethattheeffectthroughtheoptimalchoiceofquantityissimplythroughtherevenuefunction− 𝜎 − 1 .Withtheextracostoftariffsforthefirm,wereducethisexponentbyoneto−𝜎.

25

2.5 AggregatingFirm-levelTrade

A crucial component of firms in trade, and a major innovation in the analysis of

world trade flows, is aggregating the decisions of firms into country-level trade

flows. Starting with Helpman, Melitz, and Yeaple (2004) and Chaney (2008), the

most common method of aggregating firms is to use the Pareto distribution to

represent heterogeneity in the costs of production. The use of the Pareto

distribution is empirically supported in awidenumberof studies (Eaton,Kortum

and Kramarz 2011, Di Giovanni, Levchenko, and Rancier 2011), although recent

research has extended analysis to include a log-normal productivity distribution

(Head,Mayer,andThoenig,2014andFernandesetal.2015).

For this chapter, wemove forwardwith the Pareto assumption due to its

easeofuseandreasonablystrongsupportinthedata.Precisely,weassumethatin

theexportingcountry,Nfirms,someofwhichwillfailimmediatelyuponentry,have

paidafixedcostofentryandarepotentialexporters.Afterpayingtheirentrycosts,

these N firms draw a random value of marginal costs, c, from the following

distribution:

(16) 𝐺 𝑐 = Pr 𝐶 < 𝑐 = !!!

!,

where, 𝑐! is the maximum value of costs in this distribution, and 𝑘 is the

distribution’sshapeparameter.Afterfirmsdrawtheirvalueofmarginalcosts,they

make decisions regardingwhichmarkets to serve, and howmuch to sell to each

market.

Assuming that𝑐! is nonbinding (above the cutoff values for exporting),

aggregateexportsassumingCESpreferencesisderivedusingthefollowing:

(17) 𝑉 = 𝑁 𝜎𝐴𝑡!! 𝜏𝑐 !!!𝑔(𝑐)𝑑𝑐!∗

! ,

where𝑐∗ ≡ !!

!!!!!

!!!!is the cost cutoff for exporting in CES. Imposing the Pareto

distributionfrom(16)andintegrating,wehave:

(18) 𝑉 = 𝑁𝐴!

!!!𝑡!!!

!!!𝜏!!𝐾!"#,

26

where𝐾!"# is a constant of model parameters. In (18), we again see a distinct

difference between the effects of tariffs and the effects of iceberg costs. First, the

effectsoftariffsaremorepronouncedandalsoafunctionoftheshapeparameterof

exporters, and thedestination-marketdemandelasticity. In contrast, theelasticity

oftradewithrespecttoicebergcostsissimplytheshapeparameteroftheexporter.

Forlineardemand,theelasticityofdemandisendogenous.However,usinga

similar procedure for linear demands, we can show that under the Pareto

assumptiontheaggregatevalueoftradehasasimilarform:

(19) 𝑉 = 𝑁𝐴!!!𝑡!(!!!)𝜏!!𝐾!"#,

where𝐾!"#isafunctionofmodelparametersrelatedtothelinearcase.In(19),we

seehowfirm-leveltradeaggregatestotheobservedbilateraltradevalueundertwo

commonmodeling assumptions. On its face, the relationship between trade costs

and tariffs and aggregate trade value are observationally equivalent between the

twomodels.That is, ineachmodel, tradecostsare raised to the shapeparameter

andtariffsareraisedtothepowerofatariffelasticity.However,theinterpretation

ofthetariffelasticityitselfderivesfromdifferentmicro-foundations.20

As a final note on both the simplicity and the limitations of the canonical

firm-heterogeneity models, we can solve for the average exporter trade flow,

conditionalonsuccesstothatmarket.Thisstatistic iscrucialsinceitrepresentsthe

averagefirmsizeofthesetofsurvivingfirms,whichisthesetoffirmsthatcomprise

reported trade data.Within the CESmodel of exporting, average export revenue,

conditionalonexporting,iswrittenas:

(20) 𝐸 𝑣|𝑣 > 0 = !"!!(!!!)

𝐹! .

In(20),wehaveastarkrelationshipbetweenkeyparametersofthemodelandthe

averagesizeofanexporterintheexportmarket.Indeed,withintheCEScontext,the

onlytwoplacesthatcouldyieldwithin-marketvariationinaverageexportsizeare

theelasticityofsubstitutionandthefixedcostofexporting.Fernandesetal.(2015)

20Indeed,intermsofthedirecteffectoftariffsonthevalueoftrade,thetariffelasticityunderCES,𝑘 !!!!

,couldbemoreresponsivetofirmheterogeneinty,k,inhighlydifferentiatedgoods(1 < 𝜎 < 2),whencomparedwiththetariffelasticityunderlineardemand,𝑘 + 1.

27

use thisstarkresult tomotivateusinga log-normaldistributionofproductivity to

bringinendogenouseffectsofdistanceandmarketsizetobettermatchthemodelto

data.

Withinthelinearmodel,averageexportrevenue,conditionalonexporting,is:

(21) 𝐸 𝑣|𝑣 > 0 = !!

!!"(!!!).

IncontrastwiththeCEScasein(20),weseein(21)abitmoredestination-market

influenceintermsoftheaveragesizeoftheexporter.However,aftercontrollingfor

tariffs and destination-market fixed effects, all remaining variation is captured by

theexporter’sshapeparameter.Putdifferently,thereisnobilateralvariationinthis

statistic,asispossibleintheCEScase.

2.6 Multi-ProductFirms

Akeyextensionfromthecanonicalfirm-heterogeneitymodelistheabilityoffirms

toexporttomorethanonedestination,andwithmorethanoneproduct.Indeed,as

described in section one, the biggest exporters tend to be those that sell many

products along with shipping to many destinations. Sending products to many

destinations is a fairly straightforwardextensionof theMelitz-typemodels—from

thesameunitcostfunction,totalproductionissplitacrossmultiplelocations.There

may be multiple locations to serve, and separate fixed costs per location, but

profitabilitycanberankedandfirmsenterdifferentmarketsaccordingly.

Inthecaseofmulti-productfirms,thereareindeedsomesimilarities,and,in

a pinch, researchers could easily re-label destinations as products and gain some

limited intuition about serving different products. However, this approach is

undesirableforafewimportantreasons.First,firmsoftenhaveaprimaryproduct,

oracorecompetency,inwhichtheyspecialize.Inmanycases,thisproductmayhave

ahigherrevealedqualityoradifferentcoststructurethanotherproductsthatthey

sell.AnearlypapertofocusontheseissuesisBernard,Redding,andSchott(2011),

wherefirmsreceiveadrawofproductqualityatthedestination-productlevelanda

firm-specific draw of productivity. From this point, themodel proceeds inMelitz

28

style, using a CES demand systemwithmarket-product-firm quality shifters. The

authorsshowthatfirmsexportingmanyproductstendtoservemoredestinations,

andhave higher scale to each destination. Further, in response to declining trade

costs globally, enhanced competition forces firms to drop their least-profitable

products.21

On this lastpoint, the second important extension related tomulti-product

firmsis thatthecostsofproducingonevarietymaybe linkedineithervariableof

fixedtermstoothervarieties,therebycreatingalinkagebetweenthedecisiontosell

oneproduct to themarket and thedecision to sell otherproducts to the sameor

othermarkets. In these types ofmodels, trade shocks can force different firms to

rationalize the varieties they sell based on the efficiency of scope. Arkolakis,

Ganapati,andMuendler(2015)proposeaCES-basedmodel, inwhichfirmshavea

core-competency—the product that the firm produces most efficiently. Other

productscanbeproduced,butatalowerefficiency.Further,therearediseconomies

ofscopeinproducingmanyvarieties,whichdependonhowfarawayfromthecore

competencyoneproduces.UsingBraziliandata,theirworkestimatesthatproducts

furtherfromafirm'scorecompetencyincurhigherunitcosts,butfacelowermarket

accesscosts.

Third,differentproductssold to thesamemarketraise the issueofwithin-

firm cannibalization of varieties. That is, when selling differentiated but similar

varieties to the same market, each variety steals some demand from the other

within-firmvariety. Feenstra andMa (2008) andEckel andNeary (2010)provide

examplesofthiseffectintwo-countrymodels.Dhingra(2013)evaluatestheeffectof

tariffreductionsonthenumberofproductsofferedwithineachfirmandtheoverall

processefficiencyofthefirm.Inhermodel,thereisatradeoffbetweenmorebrands

cannibalizing one another and increases in the profitability of process innovation

with more brands. Global tariff cuts increase market size, which expands both

processandproductinnovationforexportingfirms.Incontrast,non-exportingfirms

21Goldbergetal.(2010)discussthenon-reallocationoffirmsandproductsinresponsetoregulatoryreforminIndia.

29

face intensified competition and cut back on the number of products.

Cannibalization has also been viewed through the lens of lowermark-ups, due to

within- and across-firm competition. Spearot (2013a) and Mayer, Melitz, and

Ottaviano(2016)usethebasicframeworkofMelitzandOttaviano(2008)toallow

for multi-product firms selling to export destinations. In Mayer, Melitz, and

Ottaviano(2016),firmsfocusontheirbest-performingproductsinmarketswiththe

toughestcompetition.Theyfindstrongevidenceforthispredictionusingfirm-level

tradedatafromFrance.22

Finally,thenewestareaofworkevaluatesthelinkagesbetweenthebiggest

multinationals and the scope of the products they sell around the world. As

describedbyYeaple (2013),Dupont,amajorU.S.-basedmultinational,operates in

over 70 countries and sells products ranging from food tomotor vehicle parts to

industrialchemicals(amongmanyothers).Atamorerefinedlevel,recentworkin

Tintelnot(2016)providesananalyticalframeworktoestimatetheeffectsofexport-

platformforeigndirectinvestment,wherefirmsinvestinadistantlocationanduse

thatlocationasaproductionsourcetoexporttoothermarkets.

2.7 ExportingandInvestment

The canonicalmodelof selection into exporting canalsobe extended to allow for

firm-levelinvestmentinproductivityandquality.AsinDhingra(2013)above,from

a neoclassical perspective, the incentives to invest are larger when market size

increases. In the traditional firm-heterogeneity context, this has been studied

through two primary mechanisms: quality and productivity investment. In

Verhoogen(2008),firmsinvestinhigherlevelsofquality,wherereachingahigher

qualitylevelrequiresafixedcostofinvestment.Similarly,Bustos(2011)examines

productivity investment, where a higher productivity requires a fixed cost of

investment. Interestingly, these two problems in their most basic form are

isomorphic when using a CES revenue function, where higher-productivity firms22InSpearot(2013a),themulti-productdimensionisusedtojustifythechoiceoffixedeffects,withfirmsenteringattheHS4levelbutproducingvarietiesacrossmanyHS10products.

30

invest, whether for productivity or quality. Further, when trade costs fall,

investment increases by adding less-productive firms that previously did not find

investment profitable. As derived generally in Mrázová and Neary (2015), this

results from super-modularity between productivity (or quality) and the costs of

reachingtheforeignmarket.

Arkolakis (2010) presents another extension to the canonical firm-

heterogeneity framework, in this case through investment in marketing

expenditure.Inhismodel,firmsinvestinmarketingsubjecttoaconvexmarketing

cost, where marketing is more costly as you reach more consumers. Similar to

Verhoogen(2008)andBustos(2011),reducedcostsoftradeincreaseinvestmentin

marketing, which increases trade through a “new consumers” margin of trade.

However, different from the canonical model, the elasticity with respect to trade

costs ishigher for small firms, ason themargin it is less costly for these firms to

reachnewconsumers in response toa tradeshock. Interestingly, this response to

tradecostsisqualitativelysimilartothenon-CESmodelasdescribedin2.4.

Anotherformofinvestmentandself-selectionisinvestmentinbecomingan

importer.Similartoinvestinginproductivity,becominganimporterisoptimalifit

reduces costs or increases competition among suppliers. Thus, if firms can

overcomethefixedcostofbecominganimporter,firmsmakethisinvestment.This

processismodeledinAmiti,Itskhoki,andKonings(2014),whodevelopamodelto

study exchange rate pass-through in an environmentwhere firms can select into

exportingandimporting.Interestingly,intheirmodel,self-selectionintoimporting

and exporting can mute the effects of exchange rate shocks. For example, an

exchange rate appreciation in the country in question will have counteracting

effectsontheimportandexportsidesofthebusiness.

Finally, non-CES models have been used to a lesser degree to study the

impact of investment on exporting and firmperformance. Again, as characterized

generally inMrázováandNeary(2015),thetypicalselectionpatternsfoundinthe

CES-based literature—that high-productivity firms engage in investment—do not

necessarily follow in an environment with non-constant elasticities, since it

becomespossiblethattheprofitfunctionisnolongersupermodularinproductivity

31

andcostsof trade.TworecentexamplesareRodriguez-Lopez (2014)andSpearot

(2013b). In the former, the investment is in offshoring, in a model based on a

translog expenditure function. In the latter, firms invest in capitalwithin a linear

demand framework. In both cases, there is an inverse-U-shaped relationship

betweenproductivityandtheprobabilityofinvestment.Theintuitionisthatunder

common, non-CES demand systems, revenues will be bounded or the absolute

elasticitydecreasesinquantitywillbesuchthat,onthemargin,investmentwillnot

beprofitableforlargefirms.

Thediscussionoffirmheterogeneityandinvestmentnaturallyleadstoissues

ofexportergrowthanddynamics.Asexportingisaparticularlycostlyendeavor,itis

notastretchtosuggestthatthedecisiontoexportisonlythebeginningofacomplex

setofdecisions thatrequireexcellentmanagement toensuresuccess.To thisend,

wenowdiscussexportdynamics,and,inparticular,howexportersgroworfailafter

reachingbeyondtheirdomesticborders.

3 FirmsandExportDynamics

In section one, we identified some key empirical findings that are difficult to

reconcilewith the classic trademodel. In section twowedescribedhowselection

between exporters and non-exporters is at the heart of the literature on firms in

trade. Here, we elaborate further on some of the new, policy-relevant, questions

firm-level studies are able to answer, using theoreticalmodels similar to those in

sectiontwotoexplainsomeoftheempiricalfindingsdescribedinsectionone.

3.1 Self-Selectionvs.Learning-by-Exporting

Whilethestylizedfactsfromsectiononemakeitclearthatexportersaredifferent

fromnon-exporters,manystudiesdonotexplainthemechanismbehindthisfinding.

For instance, did the difference in firm characteristics exist before exporting,

perhaps due to intrinsic characteristics of the firms, or do they result from the

exporting decision itself? In the first case, the difference in firm characteristics

32

existswhether or not firms export and the difference shows up only because the

more-productivefirmsfinditprofitabletoselectintoexporting.This“self-selection”

ofexporterswascentraltothetheoreticalmechanismspresentedinsectiontwoof

thischapter.Inthesecondcase,therearenoneorfewerdifferencesbetweenfirms

beforeexportingoccurs,andthedifferenceshowsupbecause theactofexporting

helps exporters learn new, more-efficient production techniques. Naturally, the

literature refers to this explanation as “learning-by-exporting.” The distinction

between these two different mechanisms is important, as identifying any causal

relationshipwill leadtoverydifferentpolicyrecommendations.Forasummaryof

thefindingsinthepapersdiscussedhere,seeTable3.1.

One of the first studies to test for evidence of learning-by-exporting is

Bernard and Jensen (1999).23Using U.S. data, the authors find that there are

differencesbetweenexportersandnon-exportersandthatthedifferencesprecede

theactofexporting: firms thatwillexport, comparedwith those thatwillnot,are

alreadylargerintermsofemploymentandshipments,payhigherwages,andhave

higherproductivity.Additionally, thedifferences inproductivity andwagegrowth

remainunchangedafter exporting.Tomeasureproductivity, the authors calculate

total factor productivity (TFP) using the residual of an estimated Cobb-Douglas

production function.Theythenregressvariousperformancemeasuresonchanges

in export status with some controls for initial-firm characteristics. While

productivitydoesnotgrow,theydofindevidencethatfirmsurvivalincreaseswith

exporting.

Clerides,Lach,andTybout(1998)findsimilarresultsusingaverydifferent

empirical strategy, in which they evaluate whether productivity trajectories for

firms in Colombia, Mexico, and Morocco improve after exporting. The authors

developamodelwithendogenousandexogenousexplanationsforexportingstatus

andproductivity;intheirmodel,marginalcostisendogenousbecausetheexporting

decisionmayaffectmarginal cost, anddemandshifters (foreign income, exchange

rates, pricesof otherproducts) are exogenous to the firm.To test for evidenceof23For a thorough summary of the learning-by-exporting vs. self-selection empirical literaturecovering33countries,seeWagner(2007).

33

learning-by-exporting theauthors simultaneouslyestimatebothanautoregressive

cost function and the choice to participate in exporting. They find that cost and

productivitytrajectoriesdonotchangeafterfirmsenterforeignmarkets,andthey

concludethatthepositiveassociationbetweenexportingandproductivityispurely

drivenbyself-selectionofmoreproductivefirmsintoexporting.24

While most studies find no evidence of learning-by-exporting, some find

mixedevidenceandothers findevidence inspecial cases.Aw,Chung,andRoberts

(2000), using a methodology similar to that of Bernard and Jensen (1999), find

mixed evidence of learning-by-exporting: In Korea, TFP does not increase when

firmsexport,but itdoesinTaiwan.Theauthorsrestrictthedatatofive industries

that have high export participation rates and compare the productivity of similar

plantsthatdifferinexportstatus.Theyconcludethattheevidencesupportsneither

thelearning-by-exportinghypothesisnortheself-selectionhypothesis.Casas,Díez,

and González (2015) find weak evidence of learning-by-exporting. Firms that

becomeexporters(“Entrants”)experienceanincreaseinTFP;evidenceoflearning,

however, disappears once the authors control for firm size. Other studies find

stronger evidence of learning-by-exporting in transitional and in least-developed

economies. De Loecker (2007), using data from Slovenia and matched sampling

techniques, finds that exportersbecomemoreproductiveafterexportingand that

the effect is stronger for those firms exporting to high-income regions.He argues

thatfirmsineconomiesintransition,suchasSlovenia,havethemosttolearnfrom

exporting. Van Biesebroeck (2005), using data from several least-developed

countries in Africa and three different econometricmethodologies, finds that the

productivityadvantageincreasesforfirmsafterexporting.VanBiesebroeckargues

thatlearning-by-exportingexistsbecause,unlikeotherstudies,hissampleincludes

smalldomesticeconomieswithpoorlyfunctioningcreditmarkets.

Finally, in recent work, Atkin, Khandelwal, and Osman (2016) provide

experimentalevidenceonlearning-by-exportinginarandomizedcontroltrialwith

Eygptian rug manufacturers. Specifically, after randomizing orders from an24Forotherpapersthattest foranddon’t findevidenceof learning-by-exporting,seeIsgut(2001),usingdatafromColombia,andBernardandJensen(2004),usingU.S.data.

34

international rug intermediary across small rug makers, they evaluate in a

laboratory setting whether those receiving export access becamemore skilled at

making high-quality rugs. Employing external evaluators of rug quality, they find

strongevidenceinfavoroflearning-by-exporting.

Table3.1:SummaryofLearning-By-ExportingLiterature

Paper Learning-By-Exporting?

Data

BernardandJensen(1999) No UnitedStates

Clerides,Lach,andTybout(1998)

No Colombia,Mexico,andMorocco

Isgut(2001) No Colombia

BernardandJensen(2004) No UnitedStates

Aw,Chung,andRoberts(2000) No Korea

Aw,Chung,andRoberts(2000)

Yes Taiwan

Casas,Díez,andGonzález(2015)

Weak Colombia

VanBiesebroeck(2005) Yes Burundi,Cameroon,Coted’Ivoire,Ethiopia,Ghana,Kenya,Tanzania,Zambia,andZimbabwe

DeLoecker(2007)

Yes Slovenia

Atkin,Khandelwal,andOsman(2016)

Yes EgyptianRugManufacturers

3.2 AggregateProductivityImprovements

Evenifexportingdoesnotleadtoimprovementsinproductivityatthefirmlevel,it

is nevertheless possible that exporting leads to productivity improvements at the

sector level.Here, it is importanttonotetheroleof importcompetition indriving

self-selection. That is, in isolation, lower tariffs in an export market reduce the

averageproductivityoffirmsthatcanreachthemarket(sinceitisnoweasiertosell

35

profitably in that market). However, during liberalization episodes that involve

cutting tariffs globally, the surge in imports with resulting higher wages due to

increased export demand leads to a market with lower domestic margins and,

consequently, drives out the least-productive firms. That is, trade liberalization

increasesaverageproductivitythroughself-selection.

Aw, Chung, and Roberts (2000), Pavcnik (2002), and Bernard and Jensen

(2004) all find evidence that productivity indeed rises at the industry level after

tradeliberalization,astheleast-productivefirmsexitthemarketandresourcesare

reallocatedtomore-productivefirms.Clerides,Lach,andTybout(1998)arguethat

firms may not benefit exclusively from exporting; that is, firms may learn from

exporting,butdomestic-onlyfirmsarenotexcludedfromthecostreductions.Thus,

theynot only find, asmentioned above, no evidenceof learning-by-exporting, but

they also find that increases in export activity decrease average variable costs

withinaregion.

3.3 TheIntensiveMarginvs.theExtensiveMarginofTrade

While productivity differences may explain why few firms export and why few

exporters dominate export value, it does not fully explain the export dynamics

observedinthedata.Forexample,isexportgrowthledbytheintensivemargin,as

exportsper firm increase,or is exportgrowth ledby theextensivemargin, as the

number of firms exporting increases? As shown in section two of this chapter,

changes in trade costs most clearly impact the extensive margins of trade when

using the Pareto distribution to represent firm heterogeneity. In themodel, fixed

export costs determine whether or not a firm exports, and variable trade costs

(tariffs/transportation costs) affect prices and, thus, sales abroad.However, since

lowervariabletradecostsandlowerfixedexportcostsdrawinfirmsonthemargin

to export, the impact of these changes on exports per firm at the country level is

ambiguous, and export growth is driven exclusively by the new exporters.25This

25Seesectiontwoforadiscussionofthedeterminantsofaverageexportflows.

36

finding contrastswith thoseof previousmodels,where trade liberalizationwould

affectonly the intensivemargin(that is, through theproductionchoicesofa fixed

numberofexporters).

Recent empirical work has identified several important facts about the

intensivemargin, in contrastwithour theoreticalmodel above. First, the already-

mentionedworkbyFernandesetal.(2015)usesdatafromtheWorldBankcovering

50 developing countries to evaluate trade growth on the intensive and extensive

margin. They develop a Melitz-style model of exporting, but with a log-normal

distribution of productivity. They estimate the model using maximum likelihood,

and find that half of the variation in exports occurs along the intensive margin.

Second, Lawless (2010) usesU.S. Census Bureau data and finds that distance—in

contrast with other gravity variables such as language, internal geography, and

importcostbarriers—lowerstheintensivemargin.Finally,Das,Roberts,andTybout

(2007),usingColombiandata, findthatreductionsintradecosts,enhancedexport

promotion,andmoderatedepreciationoftheexportingcountry’sexchangerateall

resultinincreasedtradeontheintensivemargin.26

Whiletheseempiricalfindingsmaketheargumentthattheintensivemargin

shouldnotbe ignored, theevidencenonetheless impliesthattheextensivemargin

explainsmostoftheexportdynamicsseeninthedata.Akeyreason,asmentionedin

sectionone,istheimportanceofnewexportersonfutureexportgrowth.Eatonetal.

(2007), for example, find that new exporters contribute little to overall revenues

when entering the exportmarket, butwill account for almost half of total export

growth within a decade. In addition to Eaton et al. (2007), Mayer and Ottaviano

(2008)andBernardetal.(2007)alsoarguethattheextensivemarginismuchmore

important. Mayer and Ottaviano (2008) find empirical evidence suggesting that

distance and other trade barriers correlatedwith distance reduce the number of

exporters, but not average exports per firm. Bernard et al. (2007) find evidence

using U.S. data for the importance of the extensive margin; they argue that

26Thisresult isespeciallystrong insectorswith firmsthatareclustered far fromtheexportentrymargin.

37

economies of scale and sunk costs may lead firms to expand the number of

products/destinations.

Therelativeimportanceoftheextensiveandintensivemarginsoftradeisthe

focus of severalmodels. Here, we focus on two: Helpman,Melitz, and Rubinstein

(2008) and Das, Roberts, and Tybout (2007). Helpman, Melitz, and Rubinstein

(2008) develop a model that decomposes trade flows into the intensive and

extensivemarginsoftrade,includingthepossibilityofzerotradeflows.Theyargue

thatstandardgravityregressionscaptureonlytheintensivemarginoftradeandfail

totakeintoaccountselection.Toaccountfortheextensivemargin,theyuseafirst-

stage probit to account for selection into trade, and then a second-stage gravity

regression toestimate trade flowsconditionalon these flowsbeingpositive.They

find that theprobabilityof exportingbehaves ina similarway to tradevolume in

responsetogravityfactorsandthattheextensivemarginmayexplainhighertrade

volumes when barriers are lowered. In an alternative model for the margins of

trade,Das,Roberts,andTybout(2007)estimatetheirmodelusingaBayesianMonte

Carlo Markov chain estimator and firm-level data on three Colombian

manufacturingindustries:basicchemicals, leatherproducts,andknittedfabrics. In

contrastwiththeirfindingsfortheintensivemargin,theextensivemarginislikely

to increasewhenmany firmsareclusterednear theirexport-entry thresholdsand

either expectations about future market conditions improve or firms experience

favorable shifts in the exchange rate. The effect ismuted, asmentioned above, if

firmsarefarfromtheentrythresholdandmostoftheadjustmentswill takeplace

through the intensive margin. Finally, the authors find that export promotion

policies that subsidizeexportervariable costswill havea larger impactonexport

salesthanpoliciessubsidizingfixedexportercosts.Theauthorsarguethatthelatter

policy is likely to bring in firms on the export-entrymargin that will likely have

relativelylowexportsales.

Oneaspectabsent fromtheabovediscussion is thatof financingentry into

exportmarkets.We save that discussion until section four,wherewe discuss the

important issues of trade finance. Instead, we now end the firms and export

dynamics discussion by summarizing the literature that seeks to understandwhy

38

firmsstartsmallwhentheyfirstexport(andthengrow)andtheliteraturefocusing

onfirmsthatfailatexporting.

3.4 WhyFirmsStartSmall

There is another strand of the literature that focuses on firm dynamics, but that

doesnotdrawacleardistinctionbetweenthenumberofexporters (theextensive

margin)andexportvalueperfirm(theintensivemargin).Thisliteraturefocuseson

the finding thatexporters tend tostartsmall, in termsofsales,whenentering the

export market and then, conditional on surviving, increase sales. These firms, as

mentionedinsectionone,playanimportantroleinexportgrowth.Studiesseeking

to reconcile these facts incorporate elements of both the extensive and intensive

marginsoftrade.Ingeneral,thesepapersarguethatfirmslearnaboutthesuccessof

theirproductonlyafterexportingasmallamountandthatthosefirmsfindingthat

theyarenotcompetitiveabroadexittheexportmarketveryquickly(theextensive

margin). In contrast, successful exporters subsequently increase export sales (the

intensivemargin) andmay even increase the number of products or destinations

(theextensivemargin).Albornozetal.(2012)refertothisas“sequentialexporting.”

Intheirmodel,exportprofitabilityisuncertain,butcorrelatedovertimeandacross

destinations. Their model may explain why some new exporters give up shortly

after entry, despite having paid high fixed export costs, and others increase sales

andexpandtootherdestinations.Theirmodelalsorationalizeswhyfirmsmaystart

in smaller markets first, only expanding to larger markets after realizing export

quality.TheyfindevidencetosupporttheirmechanismsusingdatafromArgentina.

Conditional on survival, export growth will be higher after the first year than in

subsequentyears (23percentagepointshigher),newexportersaremore likely to

enter othermarkets than continuous exporters (4.8 percentage points), and new

exportersaremore likelytoexit theexportmarketthancontinuousexporters(29

percentagepoints).

There are several other models that explain some of these new-exporter

facts. In recent work, Eaton et al. (2014) develop a search-and-learning model

39

wheresuccessinexportingrevealsinformationaboutdemandforaproduct,anda

successful exporterwill subsequently search formorebuyers. The studyuses the

methodofsimulatedmomentstoreplicatekeypatternsinColombiancustomsdata

andcalculatetheeffectoftradecostsandlearningeffectsonexporterbehavior.The

authorsquantifytheroleofseveralfrictionsandfindthatfornewexportersthecost

for finding one client in theUnited States every two years is $1,405, and rises to

$51,471to findone inoneyear.Onceaclient is found, thecostof findinganother

one drops to $106 and $3,898, respectively. On average, only one out of five

potential foreign clients a firm meets will result in a successful partnership. In

anotherexample,RauchandWatson(2003)focusmoreonimportersindeveloped

countriestryingtoidentifyfirmsindevelopingcountriesabletosupplylargeorders.

In their purely theoretical paper, firms in developed countries start with small

ordersunderuncertaintytolearninformationaboutthesupplier’scapabilities,and

exportssalesgrowforsuccessfulpartnerships.

Finally,Morales,Sheu,andZahler(2014)usetheideaof“extendedgravity”

to explainwhy firms often export tomarkets that are similar tomarkets already

served.Inthismodel,bilateraltradeliberalizationincreasesexportsnotonlytothe

partner’s market, but also to other, similar markets. The argument is that entry

coststoother,similarmarketsarelowerafterenteringoneofthemarkets,sotrade

liberalizationwithonecountrycanhaveanimpactonother,similarmarkets(even

withouttradeliberalizationinthoseothermarkets)through“extendedgravity.”The

authorsusematched,firm-leveldatafromChileinthechemicalssectortomeasure

theimportanceofgravityandextendedgravityandfindthatfirmsaremorelikelyto

export tocountries thatborderorare inthesamecontinentasaprevioustrading

partner. The authors estimate that for Chilean firms: (1) market entry costs are

loweredby$22,930whennewmarketsborderexistingmarkets, (2)marketentry

costs of entering a similar country in South America are between $16,350 and

$18,970,and(3)marketentrycostsoutsideofSouthAmericaarebetween$94,860

and$101,990.

3.5 ExportSurvival

40

Whilemost of the learning literature focuses on the dynamics of successful, new

exporters,astrandoftheliteraturefocusesonthefactthatmostnewexportersdo

not export beyond one year. 27 This “export survival” literature attempts to

understandwhatmakesanexportersuccessfulandwhathappenstothefirmwhen

itisnot.Thisquestionisimportant,since,asmentionedabove,severalpapersfind

that thecostsofenteringexportmarketsarequitehighandthatexportgrowth is

ledbytheextensivemargin.Additionally,first-timeexporters,whichtendtoexport

smallamountsthefirsttimetheyexport,likelyexperienceanegativeprofitbecause

ofmarketentrycosts,andthislossinturnmayaffectdomesticperformance.Mora

(2016), for example, finds that financially constrained unsuccessful exporters are

more likely to go out of business and experience lower domestic revenue and

revenue growth than similar successful exporters and firms that havenever tried

exporting.Thefirm-levelliteraturefindsthatexportingtoclosermarketsimproves

export survival (Esteve-Pérez et al. 2007) and export success increases with the

numberofexporters, suggestingwithin-sectorexternalities (Cadotetal.2013and

Stirbat,Record,andNghardsaysone2013).

As mentioned, financial constraints are one of the determinants of an

exporter’s success and, therefore, play a significant role in shaping the overall

international trade flows. In the next section, we look precisely at how financial

factorsaffecttheexporters’performanceandhowexportersmanagetofinancetheir

operations.

4 InternationalTradeandFinance

Traditionally, the international economics literature has focused on the role of

physicalcapitalindeterminingthepatternsoftrade.Incontrast,theroleoffinancial

capital has been mostly overlooked. Still, the workhorse trade models, including

thosementioned insectiontwo(Melitz2003,Bernardetal.2003), focusedonthe27Here,wefocusonfirm-levelstudies,buttherearestudiesthatlookatthesurvivalofcountry-levelexportproductlinesanddestinations(forexample,seeBesedešandPrusa2006).

41

decisions made by individual firms to participate in international markets and

provided the setting to incorporate financial elements into the firm’s decision-

makinganalysis.Whyisthisanalysisrelevant?Thereareanumberofreasonswhy

firmsengagedininternationaltradedeservetobeanalyzedthroughadifferentlens

fromthetypical(domestic)corporatefinanceconsiderations.Firmsininternational

trade face situationswhere the timebetween the shipmentof theproductand its

payment is significantly longer than for firms operating in the domestic market,

creating special working capital needs.28Additionally, in the case of a contract

breach—either the importing firm’s not paying for the delivered goods or the

exporter’s not sending the goods according to the agreed-upon specifications—

thereisanincreaseinriskduetopotentiallitigationinaforeignjurisdiction.

Theliteraturecanbe(broadly)brokenintotwoparts.First,abranchofthe

literaturestudieshow firms'exportingoutcomesareaffectedbycountry-, sector-,

and firm-level financial factors. Second, another branch of the literature focuses

directlyonthefinancingofspecifictradeflows,theso-calledtradefinanceandtrade

credit.29Thisliteraturelooksatthedeterminantsofhowagiventradeflowis,inone

wayoranother,financed.Wedescribethesedifferentanalysesnext.30

28FoleyandManova(2015)reportthatinternationalshippinganddeliverytypicallytakes60dayslonger than domestic transactions. This paper also provides an excellent survey of the recentliteratureintersectingcorporatefinanceandinternationaltradeandinvestment.29Typically, the term“international trade finance”refers to thesetof contractsbywhichexporterandimporteragreeonthespecifictermsandconditionsregulatingtheirtransaction.Still,sometimesthe term “trade finance” is reserved for the financing of a transaction via the intermediation of abank,whiletradecreditreferstothefinancingofonefirmbytheother.Whileitliesbeyondthescopeof this article to provide a detailed description, there is an abundant literature on trade credit,includingPetersenandRajan(1997),BurkartandEllingsen(2004),Love,Preve,andSarria-Allende(2007),andKlapper,Laeven,andRajan(2012),tomentionjustafew.30The literature on trade finance has also focused on the dynamics of international trade duringfinancial crises,especiallyduring theGreatTradeCollapse (GTC) that tookplaceduring the2008–2009globalfinancialcrisis,whenthefallininternationaltradefarexceededthatofeconomicoutput.SomepapersexplaintheGTCessentiallyastheresultofdemandshocksandcompositionaleffects—seeEatonetal.(2016),Bems,Johnson,andYi(2010),Behrenseta.l(2013),Brincogneetal.(2012),andLevchencko,Lewis,andTesar (2010).Anothergroupofpapersemphasizes instead theroleoffinancialfactors,suchasreducedaccesstocapital—seeAmitiandWeinstein(2011),Paravisinietal.(2015), Chor and Manova (2012), Ahn, Amiti, andWeinstein (2011), and Berman, de Souza, andMayer. (2013). The general consensus is that demand and compositional issues played a primaryrole,whilethetradefinanceconsiderationshadasecondary,butstillsizeable,effect.

42

4.1 FinancialConsiderationsthatShapeInternationalTrade

Thereareseveralwaysbywhichfinancecanaffecttheresultingpatternsoftrade.

For instance, theoveralldevelopmentof the financialsystem,asourceof financial

capital, establishes cross-country differences that enable firms from certain

countries tohaveanadvantage.Similarly, sincesomesectorshavestrongerneeds

for external funding, countries with more-developed financial systems facilitate

accesstocapitalandtherebyenhancetheabilityoffirmsincertainsectorstoenter

externalmarkets.Further,afirm’sfinancialhealthalsoaffectsaccesstobothcapital

and internationalmarkets. The literature studies the interaction between finance

andtradeatallof thesedifferent levels.Wedescribethemain findingsbelowand

summarizethemainfactorsconsideredbytheliteratureinTable4.1.

4.1.1 AggregateFinancialFactors

Onegroupofpapersstudiesthelinkbetweenthedevelopmentlevelofacountry’s

financial system, the different sectoral external financial needs, and the resulting

firms that enter internationalmarkets.31Beck (2002) considers amodelwith two

sectors,ahomogeneousgoodandadifferentiatedgood(witheconomiesof scale).

The model also includes a loan market with both asymmetric information and

searchcosts.Critically,firmslocatedinafinanciallymore-developedeconomyface

lower search costs, implying greater capital access.Themodelpredicts that these

financially developed economies have a comparative advantage in the

differentiated-goods industry. Beck (2003) uses data for 56 countries and 36

industries and finds evidence that financially more-developed countries export

relatively more in financially dependent industries. Manova (2008) provides

additional evidence, supporting these results.Usingdata for 91 countries and the

31A country’s level of financial development is usually proxied by variables such as credit to theprivatesectorasashareofGDP,thevalueofthecapital(equity)marketsasashareofGDP,ortheratio of the financial system liabilities to GDP. Further, a sector’s financial vulnerability is usuallymeasuredby external fundingneeds and tangible assets, followingRajan andZingales (1998) andClaessensandLaeven(2003),respectively.

43

1980–1997 period, she finds that opening equity markets to foreign capital

increasesexports,especiallyinsectorswithhigherexternalfinancingneeds.

Manova (2013) also studies how financial frictions affect aggregate trade

flows,but she incorporateselementsof thecurrentworkhorse firm-heterogeneity

trademodel (Melitz).Thepaperspecificallypresentsamulti-country,multi-sector

model,withheterogeneousfirmsintermsofproductivity,heterogeneoussectorsin

terms of external financing needs and collateralizable assets, and heterogeneous

countries in terms of the probability of contract enforcement.32In this setup,

dependingon the firm’sneed for external funding, credit constraints can increase

theexporter’sproductivitycutofforevenreduceexportsbelowtheirfirstbestlevel.

She then applies the model to a large dataset of 107 countries and 27 sectors

spanning1985 to1995,and finds thataround20–25percentof the totaleffectof

credit constraintson trade isdue todecreasedoutput,one-thirdof the remaining

(pure trade) effect is due to decreased entry into the export market, and the

remainingtwo-thirdsareduetodecreasedexportsales.Further,countriesthatare

financially more developed have less-acute credit constraints—therefore, they

export more in sectors with higher external financing needs, export to more

countries,andexportmoreproducts.

4.1.2 Firm-LevelFinancialFactors

Anothergroupofpapersfocusesonhowfirm-levelfinancialconsiderationsaffecta

firm’sinvolvementintheexportmarket.Greenaway,Guariglia,andKnelier(2007)

examine whether a firm’s financial health affects its export-market participation,

wherefinancialhealth ismeasuredaseithera firm’s liquidity(theratioofcurrent

assetsminuscurrentliabilitiesovertotalassets)orleverage(theratioofshort-term

32Inmostmodels,thereisaclear(oftenone-to-one)correspondencebetweenafirm’sproductivityand export performancewith its ability to obtain external capital. This link becomes imperfect inChaney (2016), where firms draw productivity levels and liquidity levels. Since export costs arefinancedthroughdomesticsalesandtheliquidityendowment,someproductivefirmsmaynotexportiftheydrawalowliquiditylevel,whilesomelow-productivityfirmsmayexportiftheyreceiveahighliquiditydraw.

44

debttocurrentassets).ThepaperusesdataonU.K.manufacturingfirmsfor1992–

2003andfindsthatexporterspresentbetterfinancialhealththannon-exporters.33

However, these differences are essentially driven by continuous exporters that,

consistentwith the stylized facts presented in section one, tend to be larger than

entrantexporters.Newexportersactuallypresentpoorerfinancialhealth,possibly

duetothesunkcostsincurredtostartexporting.Further,inasimilarfashiontothe

discussion on self-election vs. learning-by-doing of section three, the paper also

findsnoevidencethat firmswithbetterexantefinancialhealtharemore likelyto

startexporting,andstrongevidencethatparticipationinexportmarketsimproves

firms'expostfinancialhealth.34Theauthorsthereforeconcludethatfinancialhealth

canbeseenasanoutcomeratherthanadeterminantofexportentry.

RecentworkbyManovaandYu(2016)showshowcreditconstraintsaffect

not just a firm’s choice to export, but its exportmode. There are three types of

export modes in China: ordinary trade, import-and-assembly processing trade

(where the processing firm sources and pays for imported inputs), and pure-

assembly processing trade (where the processing firm receives foreign inputs for

free).Whileprofitability increases frompureassemblytoprocessingwith imports

toordinarytrade,more-profitabletraderegimesrequiremoreworkingcapital.The

paper finds that financially healthier firms conduct more ordinary trade than

processing trade and more import-and-assembly than pure assembly, and that

financial health improvements are followed by reallocations towards more

profitablemodes.Theimpactoffirmandsectorfinancialhealthandvulnerabilityis

biggerinChineseprovinceswithweakerfinancialsystems.Further,thepaperalso

finds that the role of firm financial health as a determinant of the export mode

choice is independent of firm size, age, productivity, ownership structure,33Severalotherpapersalso find thatexporterspresentbetter financialhealth thannon-exporters.For example, Muûls (2015) finds that among Belgian firms, credit ratings are correlated withexporter(and importer)statusandexported(and imported)values.BermanandHéricourt (2010)usefirm-leveldatafromnineemergingeconomiesandalsofindthatfinancialhealthcorrelateswithexport status, but it does not increase the probability of remaining an exporter nor the size ofexports.MinettiandZhu(2011)usedataonItalianfirmsandfindthatcreditrationingaffectsboththeextensiveandtheintensivemarginsofexporting.34This is in contrast to Bellone et al. (2010), who, using data on French firms, find that firmsenjoyingbetterfinancialhealtharemorelikelytobecomeexporters.

45

production technology, and tariffs on imported inputs; and that its effect is

economicallylargerelativetothatoffirmproductivity.

4.2 HowAreTradeFlowsFinanced?

Sinceshippinggoodsinternationallytakessignificanttime,exportersandimporters

mustagreenotonlyonthepriceandquantityofatransaction,butalsoonwhowill

financethelagbetweenproductionanddelivery.Indeed,thefirmthatfinancesthe

transactionwillbeartherisk.Theseconsiderationstranslate intoessentiallythree

alternativecontractualsetups:anopenaccountsystemwheretheexporterprovides

the financing (and bears the risk) by shipping the goods and waiting for the

importertoreceivethemandonlythenreceivespayment,acash-in-advancesystem

where the importer provides the financing (and bears the risk) by paying to the

exporter upfront, and a letter-of-credit systemwhere the parties are financed by

theirrespectivebanks.35AccordingtoareportbytheIMF(2009),theopenaccount

systemistheonemostcommonlyused,andcomprises42percentoftransactions,

whilethebankfinancingandthecashinadvancesystemsamountto36percentand

22percent,respectively.

Thereareahandfulofrecentpapersthatstudyhowexportersandimporters

choose a specific type of contract to execute their transaction. In general, the

literatureidentifiesthedegreeofcontractenforceabilityandthefinancingcostsin

bothcountriesasthekeyfactorsforthecontractualchoice.For instance,Schmidt-

Eisenlohr(2013)presentsamodelwherethepartyfromthecountrywiththelower

financing costs and weaker contract enforcement finances the transaction.When

both parties are located in countries with weak contractual enforcement, bank

financing is optimal since it resolves the commitment issues on both sides. The

35Banking finance actually includes other products, such as documentary collections, pre-exportfinance,andsupplychainfinance.AreportfromtheBIS(2014)describesthesealternativemethods.The same report estimates that trade finance directly supports about one-third of global trade,around6.5–8trilliondollarsperyear,with lettersofcreditcoveringaboutone-sixthof total trade.Figure1presentsaschematicrepresentationoftheworkingsofaletterofcreditsystem,takenfromAhn(2015).

46

model also implies that trade finance costs are proportional to the value of the

exportedgoods,similartotheicebergcostmentionedinsectionone.Severalofthe

model’spredictionsarevalidatedusingadatasetthatcovers150countriesbetween

1980and2004.Inparticular,thepaperfindsthattwocountriestradelesswitheach

otheriftheirfinancingcostsarehigherandthatthiseffectislargerthegreaterthe

distance(proxyfortimetoship)betweenthetwocountries.

Antràs and Foley (2015) present another model of trade finance where

cross-countrydifferences incontractualenforcementdrivethecontractualchoices

madebytheparties.Thestaticversionofthemodelpredictsthatopenaccountsor

cash-in-advance are used (over letters of credit) as long as the banks from the

importer’s country can pursue claims against importers more effectively than

exporterscan.Additionally,thetheoryalsopredictsthattheeffectofthecontractual

environment on the financing choice is stronger the farther away the importer is

from the exporter. The dynamic version of themodel considers that a fraction of

importers are not trustworthy, but the exporter learns about the importer’s type

and,throughrepeatedinteractions,mayofferpost-shipmentfinancingterms.When

theytakethesepredictionstothedata(ofaU.S.-basedfoodexporter),theyfindthat

indeed cash-in-advance and letters of credit are used most frequently with

customers located in countries with relatively weak contractual enforcement

measures.Still,theeffectofweakcontractualenforcementisreducedforimporters

incloseproximitytotheexporter.Further,theyalsofindthatasarelationshipwith

an importerdevelops, itbecomesmore likely the firmswillmovetoopenaccount

terms. This last result has important implications for policy, since it implies that

developingatradingrelationshipcanbecomeasourceofcapitalforfirmslocatedin

countrieswithweakcontractualenvironments.

Otherpapersfocusdirectlyontradefinanceprovidedbybanks.Forinstance,

onatheoreticallevel,Ahn(2011)modelstheuseoflettersofcreditinasetupwhere

firmsneedcapitalfrombanks,which, inturn,needtoinvesttoobtaininformation

about both the importer and the exporter. Since international trade flows are

smaller thandomestic flows, banks invest less in the former type of transactions,

making international traderelativelyriskier.Onanempirical level,Niepmannand

47

Schmidt-Eisenlohr (2016) provide evidence on the role of letters of credit for

exporting,usingdataonU.S.banks’tradefinanceclaims.Thepaper’sbaselineresult

isthataone-standard-deviationnegativeshocktothecountry-levelsupplyoftrade

finance leads to an average 1.5 percentage point decrease in export growth.

Moreover,giventhehighdegreeofconcentrationinthebusiness,ashocktoasingle

large bank can affect U.S. aggregate export growth, and, since banks specialize in

certain markets, a shock to a given bank will have heterogeneous effects across

destinationcountries.

5-ConclusionandDirectionsforFutureResearch

Ourunderstandingoffirmsininternationaltradehasexpandedgreatlyoverthelast

20years.However, there ismuchtobedoneto furtherourunderstandingofhow

firms,andinparticularbigfirms,operateintheglobaleconomy.Below,weaddress

afewareasthatcouldbethenextfrontierofresearchinthisarea.

Althoughmanyofouranecdotesandempiricalanalyseshaveclearlypointed

to the critical role of big firms in international trade, the ability of the theory to

capturethefullextentobservedinthedataisstilllimited.Forexample,alargeshare

ofempiricalmodelsismotivatedundertheassumptionofmonopolisticcompetition,

whereas, in reality, models of empirical industrial organization are likely more

appropriate. Indeed, the entry of firms into newmarkets is not smooth—a small

tariff change will likely not affect firm behavior much, although a large trade

agreement or devaluation could have large effects onmarket entry. Atkeson and

Burstein (2008), Eaton, Kortum, and Sotelo (2015), and Gaubert and Itskhoki

(2015) pioneered early work studying the impact of shocks on discrete entry

decisions,wherepotentiallylargefirmsenterbasedonanorderingofproductivity.

Continuingthislineofresearchtostudyallaspectsoffirmdecisionsinresponseto

international shocks will greatly enhance our understanding of firms in

internationalcommerce.

48

Wealsodiscussedvarious elementsof firmdynamics,manyofwhichhave

not been settled by empirical work, and future work would benefit from an

introductionofelementsrecentlyidentifiedininternationaltrade.Forexample,the

debate between learning-by-exporting and self-selection would benefit from

studying the increasingly important role of intermediate goods in international

trade.36Vertical specialization may allow firms to produce items more efficiently

andmayallowfirmstoproduceitemsthatwouldnothavebeenpossiblewithoutan

expansion of global supply chains. Likewise, since vertical specialization may

increase the activity of firms in the global economy, it may affect the extensive

marginasmorefirmsexport,andmayaffecttheintensivemarginasfirmsareable

toexportmore.Thus,understandingthefirmdecisiontoimportintermediategoods

andtospecializeinvariouspartsoftheglobalsupplychainmayprovideinsightin

thediscussionoflearning-by-exportingversusself-selectionandinthediscussionof

the extensive versus intensive margins of trade. Another potential direction of

futureresearchinunderstandingfirmdynamicsistheroleofuncertaintyinlimiting

exports. For example,Handley and Limao (2015) find that firmbehavior changes

significantly in the presence of policy uncertainty and firmsmay choose to delay

entering a foreignmarket until conditions improve. Likewise, firms in developing

countries that tend to have higher export failure rates may also delay entering

markets.Infact,itmaybethatuncertaintylimitstheactivityoffirmsabroadmore

than other trade costs and that policymakers should spend resources not only in

ordertoincreasemarketaccessbutalsotoloweruncertainty(forexample,helping

potential exporters find partners abroad) or lowering fixed costs associatedwith

exporting.

Finally, there are several directions in which the literature on trade and

finance can move forward. First, trade finance theory could improve greatly by

incorporating new layers of heterogeneity. Specifically, allowing for the optimal

financing choice to depend on, for example, the product being traded, could shed

new light into theanalysis—asthecharacteristicsof thegoodsaffect themodeby

36SeeChapter11ofthishandbookforathoroughdiscussionofintermediariesintrade.

49

which they are transported, it is reasonable to think that they also affect the

payment choices, thus potentially creating a link between trade finance and the

mode of transportation.Moreover, the literature could also incorporate elements

usuallyconsidered inmodelsofopeneconomymacro likeexchangeratevariation

and expectations, as these surely play a significant role in determiningwhether a

giventransactionwillbeprofitable.Further,thechoiceofthecurrencyinwhichthe

transactionwill takeplacealsoappears tobeextremely relevant, yetvery little is

known about it.37Last, but not least, probably the most important need of this

literature is access to datasets on the different payment choices to enable

identificationofnewstylizedfactsandtotestfuturetheories.

Table4.1:MainFinancialVariablesAffectingTrade

Country-level variables

− Credit to the private sector to GDP ratio − Value of listed shares to GDP ratio − Financial sector liquid liabilities to GDP ratio − Indices of contract enforcement and expropriation

risk

Sector-level variables

− External financing dependence − Asset tangibility ratio − Inventory to sales ratio − R&D to sales ratio

Firm-level variables

− Leverage ratio − Liquidity ratio − Credit rationing

37Casas et al. (2016) show the importance of currency invoicing in the reaction of trade flows toexchangeratevariations.

50

Figure4.2:LetterofCreditSystem(Ahn2015)

Exporter Importer

Importer’sBank(IssuingBank)

Exporter’sBank(AdvisingBank)

(6)Payment

(5)Payment

(4)Payment

(1)Contract

(3)Confirmationagreement

(3)Shipmentofgoods

(2)LetterofCreditrequest/issuance

(2)LetterofCreditconfirmation

51

References: Ahn,J.2011.ATheoryofDomesticandInternationalTradeFinance.IMFWorkingPapers,11/262.Ahn,J.2015.UnderstandingTradeFinance:TheoryandEvidencefromTransaction-LevelData.InternationalMonetaryFund,Washington,DC.manuscript.Ahn,JaeBin,MaryAmiti,andDavidE.Weinstein.2011.TradeFinanceandtheGreatTradeCollapse.AmericanEconomicReview:PapersandProceedings101(3):298–302.Ahn,J.,&McQuoid,A.F.(2015).Capacityconstrainedexporters:IdentifyingIncreasingMarginalCost.manuscript.Albornoz,F.,H.F.C.Pardo,G.Corcos,andE.Ornelas.2012.SequentialExporting.JournalofInternationalEconomics88(1):17–31.Amiti,M.,andD.E.Weinstein.2011.ExportsandFinancialShocks.QuarterlyJournalofEconomics126(4):1841–1877.Amiti,Mary,OlegItskhoki,andJozefKonings.2014.Importers,Exporters,andExchangeRateDisconnect.AmericanEconomicReview104(7):1942–1978.Antràs,P.,andF.C.Foley.2015.PoultryinMotion:AStudyofInternationalTradeFinancePractices.JournalofPoliticalEconomy123(4):853–901.Arkolakis,Costas.2010.MarketPenetrationCostsandtheNewConsumersMargininInternationalTrade.JournalofPoliticalEconomy118(6):1151–1199.Arkolakis,C.,S.Ganapati,andM.A.Muendler.2015.TheExtensiveMarginofExportingProducts:AFirm-LevelAnalysis.mimeo.Atkeson,Andrew,andArielBurstein.2008.Pricing-to-Market,TradeCosts,andInternationalRelativePrices.AmericanEconomicReview98(5):1998–2031.Atkin,D.,A.K.Khandelwal,andA.Osman.2016.ExportingandFirmPerformance:EvidencefromaRandomizedTrial.NationalBureauofEconomicResearchWorkingPaperNo.20690Aw,B.Y.,S.Chung,andM.J.Roberts.2000.ProductivityandTurnoverintheExportMarket:Micro-LevelEvidencefromtheRepublicofKoreaandTaiwan(China).TheWorldBankEconomicReview14(1):65–90.

52

Beck,T.2002.FinancialDevelopmentandInternationalTrade:IsThereaLink?JournalofInternationalEconomics57(1):107–131.Beck,T.2003.FinancialDependenceandInternationalTrade.ReviewofInternationalEconomics11(2):296–316.Behrens,K.,G.Corcos,andG.Mion.2013.TradeCrisis?WhatTradeCrisis?ReviewofEconomicsandStatistics95(2):702–709.Bellone,F.,P.Musso,L.Nesta,andS.Schiavo.2010.FinancialConstraintsandFirmExportBehaviour.TheWorldEconomy33(3):347–373.Bems,R.,R.C.Johnson,andK.M.Yi.2010.DemandSpilloversandtheCollapseofTradeintheGlobalRecession.IMFEconomicReview58(2):295–326.Berman,N.,andJ.Héricourt.2010.FinancialFactorsandtheMarginsofTrade:EvidencefromCross-CountryFirm-LevelData.JournalofDevelopmentEconomics93(2):206–217.BermanN,deSousaJ,MartinP,MayerT.2013.TimetoShipduringFinancialCrises.InNBERInternationalSeminaronMacroeconomics2012,225–60.Chicago:Univ.ChicagoPress.Bernhofen,D.M.,andBrown,J.C.2004.ADirectTestoftheTheoryofComparativeAdvantage:theCaseofJapan.JournalofPoliticalEconomy,112(1):48-67.Bernard,A.B.,andJ.B.Jensen.1999.ExceptionalExporterPerformance:Cause,Effect,orBoth?JournalofInternationalEconomics47(1):1–25.Bernard,A.B.,andJ.B.Jensen,2004.ExportingandProductivityintheUSA.OxfordReviewofEconomicPolicy20(3):343–357.Bernard,A.B.,Eaton,J.,Jensen,J.B.,andKortum,S.2003.PlantsandProductivityinInternationalTrade.AmericanEconomicReview,93(4):1268-1290.Bernard,A.B.,J.B.Jensen,S.J.Redding,andP.K.Schott.2007.FirmsinInternationalTrade.JournalofEconomicPerspectives21(3):105–130.Bernard,A.B.,S.J.Redding,andP.K.Schott.2011.Multi-ProductFirmsandTradeLiberalization.QuarterlyJournalofEconomics,126(3):1271–1318.Bernhofen,D.M.,andBrown,J.C.2004.ADirectTestoftheTheoryofComparativeAdvantage:theCaseofJapan.JournalofPoliticalEconomy,112(1):48-67.

53

Bertoletti,Paolo,EtroFederico,andInaSimonovska.2016.InternationalTradewithIndirectAdditivity.NationalBureauofEconomicResearchWorkingPaperNo.21984.Besedeš,T.,andT.J.Prusa.2006.ProductDifferentiationandDurationofU.S.ImportTrade.JournalofInternationalEconomics70(2):339–358.BIS2014.TradeFinance:DevelopmentsandIssues.CGFSPapersNo.50.Blum,B.S.,Claro,S.,andHorstmann,I.J.2013.Occasionalandperennialexporters.JournalofInternationalEconomics,90(1):65-74.Brander,James,andPaulKrugman1983A‘ReciprocalDumping’ModelofInternationalTrade.JournalofInternationalEconomics15(3):313–321.Brander,JamesA.,andJ.SpencerBarbara1985.ExportSubsidiesandInternationalMarketShareRivalry.JournalofInternationalEconomics18(1):83–100.Bricongne,J.C.,L.Fontagné,G.Gaulier,D.Taglioni,andV.Vicard.2012.FirmsandtheGlobalCrisis:FrenchExportsintheTurmoil.JournalofInternationalEconomics87(1):134–146.Brooks,E.L.2006.WhyDon'tFirmsExportMore?ProductQualityandColombianPlants.JournalofDevelopmentEconomics80(1):160–178.Burkart,M.,andT.Ellingsen.2004.In-KindFinance:ATheoryofTradeCredit.AmericanEconomicReview94(3):569–590.Bustos,P.2011.TradeLiberalization,Exports,andTechnologyUpgrading:EvidenceontheImpactofMERCOSURonArgentinianFirms.AmericanEconomicReview101(1):304–340.Cadot,O.,L.Iacovone,M.D.Pierola,andF.Rauch.2013.SuccessandFailureofAfricanExporters.JournalofDevelopmentEconomics101:284–296.Casas,C.,F.Díez,andA.González.2015.ProductivityandExportMarketParticipation:EvidencefromColombia.mimeo.Casas,C.,F.Díez,F.,andA.González.2016.HeterogeneousExporters:QuantitativeDifferencesandQualitativeSimilarities.mimeo.Casas,C.,F.Díez,G.Gopinath,andP.Gourinchas.2016.DominantCurrencyParadigm:ANewModelforSmallOpenEconomies.NationalBureauofEconomicResearchWorkingPaperNo.22943.

54

Chaney,Thomas.2008.DistortedGravity:TheIntensiveandExtensiveMarginsofInternationalTrade.AmericanEconomicReview98(4):1707–1721.Chaney,T.2016.LiquidityConstrainedExporters.JournalofEconomicDynamicsandControl72:141–154.Chor,D.,K.Manova.2012.OfftheCliffandBack?CreditConditionsandInternationalTradeduringtheGlobalFinancialCrisis.JournalofInternationalEconomics87(1):117–133.Claessens,S.,andL.Laeven.2003.FinancialDevelopment,PropertyRights,andGrowth.TheJournalofFinance58(6):2401–2436.Clerides,S.K.,S.Lach,andJ.R,Tybout.1998.IsLearningbyExportingImportant?Micro-DynamicEvidencefromColombia,Mexico,andMorocco.QuarterlyJournalofEconomics113(3):903–947.Das,S.,Roberts,M.J.,andJ.R.Tybout.2007.MarketEntryCosts,ProducerHeterogeneity,andExportDynamics.Econometrica75(3):837–873.Davis,D.R.,andD.E.Weinstein.2001.TheFactorContentofTrade.NationalBureauofEconomicResearchWorkingPaperNo.8637.DeLoecker,J.2007.DoExportsGenerateHigherProductivity?EvidencefromSlovenia.JournalofInternationalEconomics73(1):69–98.Deardorff,A.V.1984.TestingTradeTheoriesandPredictingTradeFlowsinR.W.JonesandP.B.Kenen(eds.)HandbookofInternationalEconomics,Volume1:467-517.DiGiovanni,J.,A.Levchenko,andR.Ranciere.2011.PowerLawsinFirmSizeandOpennesstoTrade:MeasurementandImplications.JournalofInternationalEconomics85(1):42–52.Dhingra,S.2013.TradingAwayWideBrandsforCheapBrands.AmericanEconomicReview103(6):2554–2584.Eaton,J.,M.Eslava,C.J.Krizan,M.Kugler,andJ.Tybout.2014.ASearchandLearningModelofExportDynamics.mimeo.Eaton,J.,M.Eslava,M.Kugler,andJ.Tybout.2007.ExportDynamicsinColombia:Firm-LevelEvidence.NationalBureauofEconomicResearchWorkingPaperNo.13531.Eaton,Jonathan,andGeneM.Grossman1986.OptimalTradeandIndustrialPolicyunderOligopoly.QuarterlyJournalofEconomics101(2):383–406.

55

Eaton,J.,S.Kortum,andF.Kramarz.2011.AnAnatomyofInternationalTrade:EvidencefromFrenchFirms.Econometrica79(5):1453–1498.Eaton,J.,S.Kortum,B.Neiman,andJ.Romalis.2016.TradeandtheGlobalRecessionAmericanEconomicReview106(11):3401–3438.Eaton,Jonathan,SamuelKortum,andSebastianSotelo.2015.InternationalTrade:LinkingMicroandMacro.AdvancesinEconomicsandEconometrics:TenthWorldCongress,VolumeII,AppliedEconomics.Eckel,C.,andJ.P.Neary.2010.Multi-ProductFirmsandFlexibleManufacturingintheGlobalEconomy.ReviewofEconomicStudies77(1):188–217.Esteve-Pérez,S.,J.A.Mánez-Castillejo,M.E.Rochina-Barrachina,andJ.A.Sanchis-Llopis.2007.ASurvivalAnalysisofManufacturingFirmsinExportMarketsin(eds.)J.M.Arauzo-CarodandM.C.Manjon-Antolin,Entrepreneurship,IndustrialLocation,andEconomicGrowth.EdwardElgarPublishingLimited.Feenstra,RobertC.2014.RestoringtheProductVarietyandPro-CompetitiveGainsfromTradewithHeterogeneousFirmsandBoundedProductivity.NationalBureauofEconomicResearchWorkingPaperNo.19833.Feenstra,R.,andH.Ma.2008.OptimalChoiceofProductScopeforMultiproductFirms,in(eds)ElhananHelpman,DaliaMarin,andThierryVerdier,TheOrganizationofFirmsinaGlobalEconomy.HarvardUniversityPress.Fernandes,A.M.,P.J.Klenow,S.Meleshchuk,.D.Pierola,andA.Rodríguez-Clare.2015.TheIntensiveMargininTrade:MovingBeyondPareto.mimeo.Foley,C.F.,andManova,K.(2015).InternationalTrade,MultinationalActivity,andCorporateFinance.AnnualReviewofEconomics7(1):119–146.Freund,C.,andM.D.Pierola.2016.TheOriginsandDynamicsofExportSuperstars.IDBWorkingPaperNo.741.Gandhi,A.,S.Navarro,andD.A.Rivers.2016.OntheIdentificationofProductionFunctions:HowHeterogeneousIsProductivity?mimeo.Gaubert,Cecile,andOlegItskhoki.2015.GranularComparativeAdvantage.UCBerkeley,workingpaper.Greenaway,D.,A.Guariglia,andR.Kneller.2007.FinancialFactorsandExportingDecisions.JournalofInternationalEconomics73(2):377–395.

56

Goldberg,P.K.,A.K.Khandelwal,N.Pavcnik,andP.Topalova.2010.MultiproductFirmsandProductTurnoverintheDevelopingWorld:EvidencefromIndia.TheReviewofEconomicsandStatistics92(4):1042–1049.Handley,K.,andN.Limao.2015.TradeandInvestmentunderPolicyUncertainty:TheoryandFirmEvidence.AmericanEconomicJournal:EconomicPolicy7(4):189–222.Head,Keith,ThierryMayer,andMathiasThoenig.2014.WelfareandTradewithoutPareto.AmericanEconomicReview104(5):310–316.Helpman,E.,M.Melitz,andY.Rubinstein.2008.EstimatingTradeFlows:TradingPartnersandTradingVolumes.QuarterlyJournalofEconomics123(2):441–487.Helpman,E.,Melitz,M.J.,andYeaple,S.R.2004.ExportversusFDIwithHeterogeneousFirms.AmericanEconomicReview,94(1):300-316.IMF.2009.SustainingtheRecovery.WorldEconomicOutlook.InternationalMonetaryFund(October).Irarrazabal,A.,A.Moxnes,andL.Opromolla.2015.TheTipoftheIceberg:AQuantitativeFrameworkforEstimatingTradeCosts.ReviewofEconomicsandStatistics97(4):777–792.Isgut,A.2001.What'sDifferentAboutExporters?EvidencefromColombianManufacturing.JournalofDevelopmentStudies37(5):57–82.Johnson,HarryG.1966.ANoteonTariffValuationBases,EconomicEfficiency,andtheEffectsofPreferences.JournalofPoliticalEconomy74(4):401–402.Klapper,L.,L.Laeven,andR.Rajan.2012.TradeCreditContracts.ReviewofFinancialStudies25(3):838–867.Krugman,P.1979.IncreasingReturns,MonopolisticCompetition,andInternationalTrade.JournalofInternationalEconomics9(4):469–479.Krugman,P.1980.ScaleEconomies,ProductDifferentiation,andthePatternofTrade.AmericanEconomicReview70(5):950–959.Lawless,M.2010.DeconstructingGravity:TradeCostsandExtensiveandIntensiveMargins.CanadianJournalofEconomics/Revuecanadienned'économique43(4):1149–1172.Leamer,E.E.1980.TheLeontiefParadox,Reconsidered.JournalofPoliticalEconomy88(3):495–503.

57

Leontief,W.1953.DomesticProductionandForeignTrade;theAmericanCapitalPositionRe-Examined.ProceedingsoftheAmericanphilosophicalSociety97(4):332–349.Levchenko,A.,L.Lewis,andL.Tesar.2010.TheCollapseofInternationalTradeduringthe2008-2009Crisis:InSearchoftheSmokingGun.IMFEconomicReview,58(2):214-253.Love,I.,Preve,L.A.,andSarria-Allende,V.(2007).TradeCreditandBankCredit:EvidencefromRecentFinancialCrises.JournalofFinancialEconomics83(2):453–469.Manova,K.2008.CreditConstraints,EquityMarketLiberalizationsandInternationalTrade.JournalofInternationalEconomics76(1):33–47.Manova,K.2013.CreditConstraints,HeterogeneousFirms,andInternationalTrade.ReviewofEconomicStudies80(2):711–744.Manova,K.,andZ.Yu.2016.HowFirmsExport:Processingvs.OrdinaryTradewithFinancialFrictions.JournalofInternationalEconomics100:120–137.Mayer,T.,M.J.Melitz,andG.I.Ottaviano.2016.ProductMixandFirmProductivityResponsestoTradeCompetition.NationalBureauofEconomicResearchWorkingPaperNo.22433.Mayer,T.,andG.I.Ottaviano.2008.TheHappyFew:theInternationalisationofEuropeanFirms.Intereconomics43(3):135–148.McQuoid,A.,andRubini,L.2014.TheOpportunityCostofExporting.Unpublishedmanuscript.Melitz,MarcJ.2003.TheImpactofTradeonIntra-IndustryReallocationsandAggregateIndustryProductivity.Econometrica71(6):1695–1725.Melitz,MarcJ.,andGianmarcoI.P.Ottaviano2008.MarketSize,Trade,andProductivity.ReviewofEconomicStudies75(1):295–316.Minetti,R.,andS.C.Zhu,S.C.2011.CreditConstraintsandFirmExport:MicroeconomicEvidencefromItaly.JournalofInternationalEconomics83(2):109–125.Mora,J.2016.ExportFailureandItsConsequences:EvidencefromColombianExporters.mimeo.

58

Morales,E.,G.Sheu,andA.Zahler.2014.GravityandExtendedGravity:UsingMomentInequalitiestoEstimateaModelofExportEntry.NationalBureauofEconomicResearchWorkingPaperNo.19916.Morrow,P.,andD.Trefler.2016.Endowments,FactorPrices,andSkill-BiasedTechnology:ImportingDevelopmentAccountingintoHOV.mimeo.Mrázová,Monika,andJ.PeterNeary.2015.SelectionEffectswithHeterogeneousFirms.workingpaper.Muûls,M.2015.Exporters,ImportersandCreditConstraints.JournalofInternationalEconomics95(2):333–343.Niepmann,F.,andT.Schmidt-Eisenlohr.2016.NoGuarantees,NoTrade:HowBanksAffectExportPatterns.InternationalFinanceDiscussionPapersNo.1158.BoardofGovernorsoftheFederalReserveSystem.OECD-WTO.2015.AidforTradeataGlance2015:ReducingTradeCostsforInclusive,SustainableGrowth.Paravisini,D.,V.Rappoport,P.Schnabl,andD.Wolfenzon.2015.DissectingtheEffectofCreditSupplyonTrade:EvidencefromMatchedCredit-ExportData.ReviewofEconomicStudies82(1):333–359.Pavcnik,N.2002.TradeLiberalization,Exit,andProductivityImprovements:EvidencefromChileanPlants.ReviewofEconomicStudies69(1):245–276.Petersen,M.A.,andR.G.Rajan,1997.TradeCredit:TheoriesandEvidence.ReviewofFinancialStudies10(3):661–691.Rajan,R.G.,andL.Zingales.1998.FinancialDependenceandGrowth.AmericanEconomicReview88(3):559–586.Rauch,J.E.,andJ.Watson.2003.StartingSmallinanUnfamiliarEnvironment.InternationalJournalofIndustrialOrganization21(7):1021–1042.Rodriguez-Lopez,J.A.2011.PricesandExchangeRates:ATheoryofDisconnect.TheReviewofEconomicStudies,78:1135-1177.Schmidt-Eisenlohr,T.2013.TowardsaTheoryofTradeFinance.JournalofInternationalEconomics91(1):96–112.Soderbery,A.2014.MarketSize,Structure,andAccess:TradewithCapacityConstraints.EuropeanEconomicReview,70:276-298.

59

Spearot,A.C.2012.FirmHeterogeneity,NewInvestmentandAcquisitions.TheJournalofIndustrialEconomics,60(1):1-45. Spearot,AlanC.2013a.VariableDemandElasticitiesandTariffLiberalization.JournalofInternationalEconomics89(1):26–41.Spearot,AlanC.2013b.MarketAccess,Investment,andHeterogeneousFirms.InternationalEconomicReview54(2):601–627.Stirbat,L.,R.Record,andK.Nghardsaysone.2013.DeterminantsofExportSurvivalintheLaoPDR.WorldBankPolicyResearchWorkingPaperNo.6301.Tintelnot,F.2016.GlobalProductionwithExportPlatforms.NationalBureauofEconomicResearchWorkingPaperNo.22236.Trefler,D.1993.InternationalFactorPriceDifferences:LeontiefWasRight!JournalofPoliticalEconomy101(6):961–987.Trefler,D.1995.TheCaseoftheMissingTradeandOtherMysteries.AmericanEconomicReview85(5):1029–1046.U.S.CensusBureau.2013.AProfileofU.S.ImportingandExportingCompanies,2010-2011.USCensusBureauNews.VanBiesebroeck,J.2005.ExportingRaisesProductivityinSub-SaharanAfricanManufacturingFirms.JournalofInternationalEconomics67(2):373–391.Vanek,J.1968.TheFactorProportionsTheory:Then—factorCase.Kyklos,21(4):749-756.Verhoogen,E.A.2008.Trade,QualityUpgradingandWageInequalityintheMexicanManufacturingSector.QuarterlyJournalofEconomics123(2):489–530.Wagner,J.2007.ExportsandProductivity:ASurveyoftheEvidencefromFirm-LevelData.TheWorldEconomy30(1):60–82.Yeaple, S. R. 2013. Scale, Scope, and the International Expansion Strategies ofMultiproduct Firms. National Bureau of Economic Research Working Paper No.19166.