Exporters of Services: A Look at U.S. Exporters outside of the Manufacturing Sector

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Exporters of Services: A Look at U.S. Exporters outside of the Manufacturing Sector Maria D. Tito 2019-063 Please cite this paper as: Tito,MariaD.(2019). “ExportersofServices: ALookatU.S.ExportersoutsideoftheManufacturingSector,”FinanceandEconomicsDiscussionSeries2019-063. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2019.063. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Exporters of Services: A Look at U.S. Exporters outside of the Manufacturing Sector Maria D. Tito∗† August 8, 2019 Abstract UsingtransactiondatafortheU.S.,thispaperpresentsaseriesofstylizedfactsonexporters in services industries. We find that most of the basic facts on manufacturing exporters extend totheservicessectorswiththreeimportantdifferences. First,theparticipationrateofservices firms in foreign markets is much lower than that of manufacturing firms. Second, the size premia at services exporters are significantly higher than those among manufacturers. Third, the survival rates of services exporters tend to be lower than that of manufacturing exporters. Allthreefactsarecompatiblewiththehypothesisthatfirmsinservicessectorsfacelargertrade costs. Asimplecalibrationsuggeststhatservicesfirmsfacetwo-to-three-timehigherfixedcosts than manufacturing exporters. Key words: Exporters of Services, Firm Heterogeneity, Extensive and Intensive Margins, Trade Costs. JEL classification: F14. ∗FederalReserveBoard. Contact: maria.d.tito@frb.gov. †Theviewspresentedinthispaperrepresentthoseoftheauthorsanddonotnecessarilycoincidewiththoseof theFederalReserveSystem. 1

1 Introduction Servicestradehasgrownrapidlyoverthelastdecadeandnowaccountsforabout30percentofU.S. exports. However, despite the increasing importance of the U.S. services sector in foreign markets, empirical work has mostly focused on manufacturing because of service-sector data limitations.1 WhileanumberofrecentcontributionshasestablishednewfactsonEuropeanservicesfirmsengaged in foreign markets, little is known on services exporters located in the United States, the world’s largest exporter of services.2 Among the exceptions, Gervais and Jensen [2019] use data on the distributionofserviceswithintheUnitedStatestoindirectlydeterminetheworldwide“tradability” of services. Complementing the existing literature on trade in services, our paper establishes a new series of stylized facts on U.S. exporters of services: In addition to the standard margins of trade (countries and firms), our contribution dissects the customer margin and identifies its influence on firm heterogeneity. We rely on the Compustat customer segment data, a source of unique firm-to-firm transactions across all sectors in the U.S. economy. Exploiting the customer name and the market segment of the buyer, we construct a foreign indicator that differentiates between domestic and foreign transactions. Our analysis focuses on the broadly defined services sector, including most industries outside of goods production: specifically, our definition encompasses wholesale and retail trade, transportation and warehousing, business and personal services. With no direct information on products, we characterize services exports as the exports of firms classified in services industries; while our definition may combine the export of different types of products, the availability on firmto-firm transactions partly addresses that limitation. Ourresultsindicatethatmostofthebasicfactsdocumentedinthetradeliteratureformanufacturing exporters also apply to services industries; there are, however, three important distinctions. First, the participation rate of services firms in foreign market is lower than that of manufacturing firms. In our data, the share of exporters in services is, on average, 15 percentage points smaller than that of manufacturing firms, with large variation across 3-digit sectors. Second, the size premia at exporters of services are significantly higher than those among manufacturers. While there is no significant difference in the market value of equity between exporters and domestic firms in 1SeeFrancoisandHoekman[2010]foranextensiveliteraturereview;Jensen[2011]focusesontheroleofservicesinU.S.trade. 2See,amongothers,AriuandMion[2010]andAriu[2016]onBelgium;KelleandKleinert[2010]onGermany; WalterandDell’mour[2010]onAustria;BreinlichandCriscuolo[2011]ontheU.K.;Gaulieretal.[2011]onFrance; GrubljesicandDamijan[2011]onSlovenia;Halleretal.[2014]onFinland,France,Ireland,andSlovenia;andFedericoandTosti[2017]onItaly. 2

manufacturing, exporters of services are 30 percent bigger than non-exporter with respect to that metric; other dimensions are, instead, not significantly different between the two groups. Third, the survival rates of services exporters tend to be lower than that of manufacturing exporters. Only 62 percent of the new services exporters continue to export the year after entry, while the 1-year survival probability among manufacturing exporters is 71 percent. Over time, the gap between the two sectors tends to shrink. All three facts point to higher trade costs for services firms. With limited data that quantify trade costs in services, we combine our stylized facts with sector-level estimates of the elasticity of substitution from Gervais and Jensen [2019] to calibrate differences in trade costs across sectors. We find that the variation in export participation alone implies that fixed export costs for services firms are between 1.4 and 1.7 times as large as those for manufacturers. If taking into account also differences in the elasticity of substitutions, our estimates suggest a two-to-threefold divergence. Our paper contributes to the growing literature on services trade, recently reviewed by Francois andHoekman[2010]. Withafocusonfirm-levelevidence,ourpaperiscloselyrelatedtorecentwork on services exporters for a group of European countries.3 An important novelty of our analysis is the decomposition of services exports along the customer margin, which has recently been receiving more attention in the literature on goods trade.4 The rest of the paper is organized as follows. Section 2 describes our data. Section 3 introduces new stylized facts on U.S. services exporters. Section 4 present a trade cost calibration, and section 5 concludes. 2 Data Our analysis relies on a unique dataset, the Compustat customer segment, which collects the filings of public firms on their major customers–that is, customers that account for at least 10 percent of theirtotalsales–, incompliancewithStatementNo. 14(1976)andtheStatementNo. 131(1997)of the Financial Accounting Standards. Thus, the Compustat customer segment is a source of unique firm-to-firm transactions across all sectors of the economy. To identify foreign transactions, we rely on two main pieces of information from the dataset: 3RecentworkonexportersofservicesincludeAriuandMion[2010]andAriu[2016]onBelgium;Kelleand Kleinert[2010]onGermany;WalterandDell’mour[2010]onAustria;BreinlichandCriscuolo[2011]ontheU.K.; Gaulieretal.[2011]onFrance;GrubljesicandDamijan[2011]onSlovenia;Halleretal.[2014]onFinland,France, Ireland,andSlovenia;andFedericoandTosti[2017]onItaly. 4Amongrecentcontributions,seeBernardetal.[2018b]andCarballoetal.[2018]. 3

the customer name and the market segment of the buyer.5 A brief description of our methodology follows. First,wematchthereportedcustomernamestoCompustatfirms. Toaddresstheproblemofnonstandardizedcustomernames,weadoptasimilarstrategytoFeeandThomas[2004]. Afterexcluding all customers with unreported names and those identified as governments or geographic regions, we run a text-matching program requiring the letters in the customer name to be sequentially present in a potential match. To ensure matching accuracy, we manually review the matched pairs: if there are multiple potential matches, and we cannot identify a unique match by looking at information on firm web sites or Google, we exclude all these possible firm-customer pairs from the sample. The namematchingprocedureresultsin23,833firm-customeror74,353firm-customer-yearobservations. Ofthematchedsample,weusethecustomer’sheadquarterstoproxyforthefirm’sphysicallocation in order to differentiate between domestic and foreign transactions. Second, we complement the name matching strategy with additional geographic imputations based on the customer’s name or the market segment. Finally, for the largest unmatched transactions, we look at publicly available information to identify the foreign status of the customer. Overall, we are able to assign a foreign status indicator to 449,015 firm-customer-year transactions– that is, to over 84 percent of the total number of observations. Because of a large number of zero dollartransactions,theshareisevenlargerintermsofvalue: weidentifytheforeignstatusformore than 90 percent of the observations.6 With the inclusion of a foreign indicator, the Compustat customer segment parallels traditional data on firm-level U.S. exports with two important distinctions. First, the dataset includes annual transactionsofU.S.firmswithdomesticandforeigncustomers.7 Second,thedatasetisnotrestricted to manufacturing firms but includes transactions of firms in all sectors of the economy. In our analysis, we focus on a broad definition of the service sector, considering most industries outside of goods production: retail and wholesale trade, transportation and warehousing, business services, and personal services. However, the restrictions that identify the data–the exclusive inclusion of public firms and of large transactions–may affect the interpretation and the generalizabity of our results. Table A1 compares the distribution of firms by sector in our data to the economy-wide distribution from County Business Patterns data. The composition of our sample is skewed towards larger firms: 5Forcustomersidentifiedasgeographicregions,thedatasetrecordsthelocationwheretheshipmentisdirected to. 6Thereportedvalueiszeroforabout15,000transactions. 7InformationondomestictransactionsiscollectedbytheU.S.CensusBureauonlyevery5yearsviatheCommodityFlowSurveyandisrestrictedtoselectedsectors. 4

enterprises with at least 500 employees represent more than 50 percent of the firms in our sample vs. 1-2 percent in the entire economy. While our data place more emphasis on large enterprises, the divergence in composition appears more contained if considering that large firms account for the largest fraction of overall employment, as detailed in table A2.8 The skewness of our sample suggests that a straightforward comparison to other results for services firms in the literature might not be accurate. Thus, in our analysis, we’ll be using the manufacturing sector as a reference point; in particular, we’ll be looking at differences of services firms with respect to manufacturing firms in our sample and compare those differences to other available data. 3 The Margins of U.S. Firm-Level Exports in Services Followingtheexistingliteratureontradeatthefirm-level,weanalyzestaticanddynamicaspectsof export flows in services and contrasts their features to the manufacturing sector as well as to other contributions on services trade. In our analysis, we characterize services exports as the exports of firms in the services sectors: with NAICS codes reflecting the activity that generates the largest share of total revenues, our definition should mainly capture transactions of services. Exploiting the disaggregation over the customer margin, we partly address the concern that the composition of exports is skewed towards a group of products that mostly includes goods or other services. Anothermissingpieceofinformationistheidentificationoftheprecisemodeofexporting. Withthe imputation of foreign status of transaction partly from firms’ location, it is likely that the majoity of our dataset covers the cross-border supply of services (known as mode 1 in GATS-speak).9 Next section describes the main elements that contribute to the cross-sectional variation in U.S. services exports; section 3.2 explores entry, exit, and survival of services firms in foreign markets. 3.1 Cross-Sectional Variation in U.S. Services Exports Inthissection, wedescribethecross-sectionalfeaturesofU.S.servicesexporters. Table1highlights that firm exporting is a relatively rare activity in services industries, in line with the prediction in the trade literature on heterogeneous firms following Melitz [2003]. Replicating the analysis in Bernard et al. [2007] and Bernard et al. [2018a], column (2) reports the average share of firms in a given industry that export; we calculate the average for firms with reported transactions between 8Firmswithatleast500employeesaccountforabout99percentoftotalemploymentinourdata. 9SampsonandSnape[1985]classifythedifferentmodesofsupplyinservicestrade;thatclassificationhasbeen subsequentlyincludedinthedesignoftheGeneralAgreementonTradeinServices(GATS). 5

2003 and 2007. On average, the share of exporters in 2-digit NAICS sectors ranges between 15 percent in retail trade industries and 35 percent in the transportation sector, with the variation likely reflecting differences in trade costs across sectors.10 Within 2-digit sectors, the variation is even larger: industries, such as general merchandise stores, gasoline stations, and social assistance services, report no exporters, while in other sectors–the examples here are water transportation, accommodation services, and repair & maintenance services–at least half of all firms are exporters. Overall, the shares of services exporters in our data is roughly in line with the export participation rates of German business services firms reported by Vogel and Wagner [2010], which focus on a sample of large firms.11 Relative to other contributions in the literature, the participation rate of U.S. firms is well above that of most studies on European services exporters, with most of the difference likely explained by the fact that the Compustat customer segment database is skewed toward larger firms and bigger transactions.12 Foramoredirectcomparison,tableA3reportsexporters’statisticsforthemanufacturingsector: we find that 41 percent of firms in manufacturing are exporters, a number somewhat above the average export participation of services firms within Compustat.13 Comparing the shares of goods andservicesexporters, similardiscrepanciesalsoappliestoEuropeandata, asshowninHalleretal. [2014] and Ariu [2016]. Thedifferencesinparticipationrelativetothemanufacturingsectorpartlydisappearwhenlooking at export values. Column (3) emphasizes that the average share of exports in firm shipments averages between 18 and 32 percent for 2-digit NAICS sectors, values which are below the export shareformanufacturing(35percent). Arelativelyhighersimilarityinthefractionofshipmentssent to foreign markets between manufacturing and services exporters could stem from the prevalence of goodsintheproductcompositionofservicesexports. Inparticular, Halleretal.[2014]findthatthe sharesofservicesexportsinoverallexportsbyservicesfirmsrangefrom18percentinFinlandto42 percent in Ireland.14 While product codes are not directly available in our data, we examined the NAICScodesofcustomerstodifferentiatebetweenexportsofgoodsandexportsofservices.15 Table 10Inamoreextensivedefinition,tradecostsencompasstheintrinsiclowertradabilityofsomeservices. Inparticular,servicestendtobenon-storableand,thus,frequentlyrequireeithertheproviderorthecustomertoreachthe otherparty. 11UsingdataonGermanenterpriseswithtotalsalesabove250,000euros,VogelandWagner[2010]reportthat theshareofexportersinallenterpriseswasabout14percentin2003andabout16percentin2005. 12Theexportparticipationratesamongservicesfirmsvarybetween0.14percentforGermanfirms(Kelleand Kleinert[2010])and50percentforSlovenianfirms(Halleretal.[2014]). 13Usingmicro-leveldataonU.S.establishments,Bernardetal.[2018a]documentthat37percentofmanufacturingfirmswereexportersin2007. 14Halleretal.[2014]alsoreportthatservicesaccountfor75percentofoverallexportsbyservicesfirm. However, theyexplainthatFrancecouldbeanoutlierbecausetheyaremissingdataonwholesaleandretailtrade,which typicallyhavelowerexportparticipationrates. 15Weassigncustomers’NAICScodesfollowingasimilarproceduretotheidentificationoftheforeignstatusof 6

A4 confirms that almost half of all export transactions are between services firms and customers in good-producing sectors; the share, however, is somewhat smaller in terms of value, 36 percent, but still points to the fact that firms in services industries are also exporters of goods.16 Nextweinvestigatethecontributionoftheintensiveandextensivemarginstothecross-sectional differences between manufacturing and services. Following Bernard et al. [2009], we decompose firm-level export flows, X , as follows ft X =n c d x¯ ft ft ft ft ft where n represents the number of countries, c denotes the number of customers, d indicates ft ft ft the density of trade–that is, the share of customer-country combinations with positive trade, d ≡ ft πft –, and x¯ is the average exports of firm f at time t across customer-country combinations nftcft ft with positive trade, x¯ ≡ Xft. This decomposition identifies three extensive (number of countries, ft πft number of customers, and density) and one intensive (average value) margins. Table 2 summarizes the distribution of services exporters across each margin; for comparison, table A5 reports analogous statistics for manufacturing exporters.17 Services and manufacturing exportersappearremarkablysimilaralongtheextensivemargins,withonlyminordifferencestowards thetopofthedistribution. Themedianexportertradeswithonecustomerlocatedinasinglecountry. Theaveragevalueofexporttransactions,instead,carriesthebulkofthedifferences: transactionsat servicesexportersare, onaverage, smallerthanthoseofmanufacturingfirms, withmoremeaningful differences for the top traders. Part of this divergence reflects the distinct characteristics of the distribution of exporters within each sector. Figures A1 and A2 reveal that the size distribution of manufacturing exporters, measured either in terms of employment or in terms of sales, has a fatter right tail. If trade costs between the two industry groups were similar, higher size of manufacturers at the top of the distribution would directly translate into larger export transactions. However, differencesinparticipationalsosuggestthattradecostshaveahigherincidenceonservicesexports, contributing to the differences along the intensive margin. Table 3 rephrases the dominant role of the intensive margin in the variation of exports across firms. Following Bernard et al. [2009], we separately regress the log of each margin against the log of total exports, lnX ; the coefficient on lnX in each regression represents the share of ft ft export variation explained by each margin. The intensive margin is the biggest source of variation, transactions. Inparticular,wewereabletoimputeNAICScodesfor119,725firm-customer-yearobservations. 16WeclassifyNAICS11(agriculture&forestry),NAICS21(mining),andNAICS31-33(manufacturing)as good-producingsectors. 17Sector-leveldetailsforeach2-digitservicesectorisavailableintableA7. 7

accounting for more than 80 percent of the variability in firm-level export flows. Recently, Breinlich and Criscuolo [2011] and Federico and Tosti [2017] have also documented that the intensive margin explains most of the firm-level variation in services exports. In our data, a quantitatively similar decompositionalsoappliestomanufacturingfirms, asshownintableA6; thisresultdiffersfromthe findings on the exports of goods by U.S. firms: in particular, Bernard et al. stress the importance of the extensive margins.18 While the restriction to public firms and larger transactions in our data mayhavesomeeffectontheOLSresultsformanufacturing,ourfindingsmayalsobeinpartrelated to the application of the OLS decomposition at the customer level. Despite making a small contribution to the cross-sectional variation in exports, the customer margin is a novel feature of our data that we’ll exploit next to document a set of facts on exporterscustomers’ relationships. Figure 1 explores the distribution of services exporters across customers andconfirmsthattherarenatureoftheexportingactivityextendstoothermarginsofparticipation ininternationaltrade. Whilethemajorityoffirmsexportstoasinglecustomer, thosefirmsaccount for about 7 percent of export value and 10 percent of employment. Firms that export to 6 or more customers, instead, represent more than 50 percent of the total export value and of employment. Table 4 provides further details on the customer dimension. We classify the relationship between an exporting firm and its customer into four categories: one-to-one, including exporters and customers that have a single connection; many-to-one, referring to the group of exporters that has multiple connections and the set of customers with a single connection; one-to-many, denoting exporters with a unique connection and customers with multiple connections; and many-to-many, capturing exporters and customers with multiple connections. One-to-many matches account for more than 50 percent of aggregate services trade in our data, confirming the dominant role of a small group of exporters in shaping trade patterns through their extensive connections. In manufacturing, one-to-many connections also account for a large share of matches and exports (table A8).19 While larger exporters tend to be well connected and sell to a variety of customers, figure 2 implies that smaller firms tend to be less connected are able to reach only the most important customers. After classifying each firm by the number of foreign market connections, we find that a 1 percent increase in the number of customers per exporter is associated with a 0.3 percent decline in the average number of connections among the customers–the slope of the fitted regression line in figure 2. The degree of negative assortativity in services trade in our data is just a little below 18Eatonetal.[2011],instead,findthattheintensivemarginadjustmentsalsodominatefirm-levelvariationin exportflowsacrosstheFrenchmanufacturingfirms. 19Bernardetal.[2018b]reportamoresignificantroleofmany-to-manymatches,which,inNorwegiandata,accountfortwo-thirdsofaggregatetrade. 8

an analogous estimate for the manufacturing sector; Bernard et al. [2018b], instead, finds quite a weaker degree of negative assortativity between Norwegian manufacturing exporters and their customers. The similarity between manufacturing and services in our data, however, hides large sectoral differences, which arise because of the size of the set of potential contacts and the magnitude of relationship-specific costs: Table A9 shows that, conditional on the number of customers per exporter, the average number of export connections across customers declines faster in personal and business services than in wholesale and retail trade, pointing to higher concentration and lower transaction costs in the trade sectors. The negative degree assortativity over the number of connections coexists with positive assortative matching on firm size.20 Figures 3 and 4 describes the sorting patterns between exporters and customers. Ranking firms according to their average size–measured by employment in figure 3 and total sales in 4–, we find that large exporters of services match with large customers. Dragusanu [2014], Benguria [2015], and Sugita et al. [2016] document qualitatively similar findings for the relationships between foreign exporters and U.S. buyers in the manufacturing sector. In our data, our estimatespointtostrongersortingforservicesindustriesrelativetothemanufacturingsector,which could arise because of either technological differences or cross-sectoral variation in search costs. Firm Characteristics The literature on heterogeneous firms in trade has documented that exporters in manufacturing are different from non-exporters. Table 5 explores the margins of systematic differences between exportersandnon-exportersinservicesindustriesinthespiritofBernardetal. (2007)and(2018a). Each row of the table shows the implied average percent difference between exporters and nonexporters, estimated in a regression of firm characteristics against a dummy variable capturing the firm export status. Starting with the results in column (1), we confirm that exporters tend to be larger–41 percent larger in terms of employment, 52 percent in terms of shipments, and 87 percent in terms of market capitalization; another important margin that distinguishes exporters from nonexporters are capital expenditures, which are 79 percent higher at exporting firms than at nonexporters. Differences in terms of labor productivity or capital intensity are, instead, not significant incolumn(1). Column(2)suggeststhatunobservedheterogeneityacrosssectorsdampensexporters’ premia: we find that differences relative to non-exporters tend to be magnified after controlling for industryfixedeffects. Finally, column(3)investigatestheimpactofdifferencesinsize, measuredby 20Bernardetal.[2018b]provesthatthenegativedegreeassortativityisconsistentwithpositiveassortative matchingontheintensivemargin. 9

employment,onperfomancepremiainforeignmarkets: inourmostrestrictedspecification,exporters are significantly different from non-exporters not only in terms of revenue, market capitalization, and capital expenditures, but also as to output per worker and capital intensity. TableA10includesmanufacturingfirmsinouranalysis. Allspecificationsincludetimedummies, sector fixed effects, and a control for size (log Employment), as in column (3) of table 5. The interactionbetweentheexportdummyandtheservicessectordummyidentifiestheaveragepercent differenceofservicesexportpremiarelativetopremiaofexportersinmanufacturing. Wefindthatthe premia at services exporters are significantly higher than those of manufacturing exporters in terms of market valuation: Exporters of services are 30 percent bigger than non-exporters, while there is no significant difference between exporters and domestic firms in manufacturing along the same dimension. Looking at other characteristics, being an exporter is not associated with significant differences between the two industry groups.21 In sum, our analysis suggests that exporters of services are different from non-exporters in services and manufacturing exporters. 3.2 Time-Series Variation in U.S. Services Exports ThechangeinaggregateU.S.exportsbetweenyeartandyeart−1canbedecomposedinto3margins: the increase due to new firm entry in foreign markets, the decrease due to the exit of existing exporters, and the expansion/contraction of exports at continuing firms. This section analyzes the contributions of each margin and the characteristics among entering, exiting, and continuing exporters. Figure 5 offers an overview of export participation in each sector between 2000 and 2016. Confirming the cross-sectional results reported in tables 1 and A3, we find that export participation is higher in manufacturing than in services sectors in each year of our sample. The evolution of foreign participation appears roughly similar across sectors: all sectors share an upward trend over the years in our sample. In particular, export participation more than doubled in wholesale trade and business services between 2000 and 2016; the share of exporters also rose in other sectors, with increases ranging between 11 and 28 percentage points. Table 6 decomposes overall participation in foreign markets into entry (entrants), exit (exiting), and survival (continuing firms). In all sectors, the share of entrants is larger than the share of exiting exporters, a finding that translates into the steady increase in participation shown in figure 5. Exit and entry rates tend to be higher across services firms relative to the manufacturing sector, 21Ourestimatesofexporterpremiainthemanufacturingsectorsareinlinewithwhathasbeenreportedby Bernardetal.[2007]andBernardetal.[2018a]. 10

resulting in higher turnover rates in services industries. The difference in turnover rates between manufacturing and services exporters occurs during a period of faster growth in the exports of services; Ariu [2016] documents a similar finding for Belgian exporters between 1995 and 2005. The faster growth and higher turnover of services exports are coupled with larger exit rates among entrants: 16 percent of firms in services industries leave the foreign markets the year after entry vs. 13 percent of manufacturing firms. Table 7 highlights the characteristics of entrants and exiting firms. New and exiting firms representasmallshareoftotalservicesexports–7percentand9percent,respectively,asin?;thenumber of countries, the number of customers, and density among entrants and exiting firms are similar to those of the average exporter. Table 8 characterizes a systematic comparison of entrants/exiting firms relative to continuing exporters. We focus on the results of columns (3) for entrants and (6) for exiting firms, the specifications that include year dummies, sector fixed effects and a firm size (measured by employment). While entrants tend to be smaller–with insignificant export premia after controlling for firm size–exiting firms tend to be worse performer: firms that exit from foreign marketsarenotonlysmaller,butalsolessproductiveandlesscapitalintensiverelativetocontinuing exporters. Our analysis so far suggests that continuing exporters account for the bulk of services exports. Given their importance, we’ll next look at survival in more details. Figure 6 shows the survival probabilities t years after starting to export: only 62 percent of the new services exporters continue to export the year after entry. The survival probability is noticeably higher for manufacturing firms,at71percent,a10-percentage-pointdifferencethatcarriesforwardforafewyearsafterentry. Similarly, Ariu [2016] finds that trading services is much riskier than trading goods. The survival probabilities for services exporters exponentially decline but catch up some to the rates of survival among manufacturers: 20 years after entry, only 1 percent of manufacturing exporters continue to serve the foreign markets vs. 0.5 percent of services exporters. We conclude our analysis with an investigation on the growth rates of services exports. Table 9 decomposes yearly growth into the contribution of the extensive margins and that of the intensive margin. In addition to entry and exit–summarized in column (2)–column (3) identifies a second extensive margin, changes in exports due to the addition or dropping of countries and customers. Finally,column(4)showsthecontributiontogrowthofexpanding/contractingtransactionsbetween exporters and their continuing customers. Services exports grew at about 4.5 percent per year over 2000-2016, with faster growth in the earlier part of our sample. Averaging at 4.2 percent per year between 2000 and 2016, the intensive margin contributed more than 90 percent to total export 11

growth; thetwoextensivemargins, instead, averageupatabout0.1percentperyear. Ananalogous time-seriesdecompositionofexportgrowthinmanufacturingyieldssimilarresults,asshownintable A11.22 4 Trade Costs: Services vs. Manufacturing The stylized facts that we have presented point to large differences in trade costs between manufacturing andservices. Taking stock of ournew evidence, this section offers a calibrationof trade costs based on the empirical evidence. Inatheoreticalframework`alaMelitz[2003],foreignsalesatsuccessfulexporters,netofvariable costs,arerequiredtocoverthefixedcostsofexporting. Exploitingthepropertythatvariableprofits are fraction of total sale, the condition for exporting requires that r i ≥f (1) σ where r denotes total revenue of firm i, σ represent the demand elasticity, and f the fixed cost of i exporting. Exploiting the empirical distribution of total sales and the share of exporters, we can, thus, easily identify the fixed costs of exporting. In particular, the condition (1) implies that the share of exporters, s , coincides with the firms whose revenues are above σ·f, x s =1−G(σ·f) (2) x where G(·) denotes the size distribution of firms in equilibrium. Under the assumption that G(·) is Pareto with slope parameter κ, condition (2) becomes s =(σ·f)−κ (3) x Assuming sectoral heterogeneity, the ratio of the fixed costs of exporting in services to the fixed costs in manufacturing takes the following expression f σ (sx)−1/κs s = m s (4) f m σ s (sx)−1/κm s 22Bernardetal.[2009]alsohighlightsthedominantroleoftheintensivemargininthetimeseriesdecomposition ofU.S.manufacturingexports. 12

Expression (4) implies that the variation in export participation across sectors is magnified by differencesinthedistributionoffirmsanddifferencesintheelasticitiesofsubstitution. Inparticular, Gervais and Jensen [2019] estimates that firms in manufacturing face an elasticity of substitution of 8.14 vs. an average of 5.88 for services firms; this discrepancy alone implies that trade costs would be 1.5 times larger for services firms. To quantify the contribution of differences in distribution, we calibrate the slope parameter of the Pareto distribution to match the data on firm sales. In particular,inaParetodistribution,theratiobetweentheunconditionalandtheconditionalaverage of firm sales equals the participation cut-off µ x =(sx)−1/κ (5) µ where µ indicate the unconditional sales average, µ is the average conditional on export status. x With premia of µx =1.68 in manufacturing and of µx =1.87 in services, the scale parameters are µ µ κ =1.72 m κ =3.03 s Relative to the literature, which estimate that the slope parameters are close to one for manufacturers, our estimates are consistent with the empirical estimates of the exporter premia.23 Our analysis, however, shows that the revenue export premium is not significantly different across the two sectors; therefore, in our calibration, we’ll assume that the two distributions share the same slope parameter, which we set to be κ ≤ κ ≤ κ . With this range of possible values for κ, the m s difference in participation implies that the fixed costs of exporting are between 1.4 and 1.7 times larger for services firms than for manufacturers. Overall, factoring in differences in participation as well as of elasticities, fixed costs of exporting are about twice as large in the services sector than in manufacturing. Survival probabilities also offer additional insights into the magnitude of trade costs. In particular, Albornoz et al. [2016] notes that differences in survival probabilities between services and manufacturingexportersrelatetothevariationinsunkrelativetofixedcostsbetweenthetwosectors andinmarket-specificcharacteristics. Withbroadsimilaritiesinthedistributionofcountriesacross firms, we abstract from the effect of market features and attribute the entire difference in survival probabilities to the relative variation of sunk-to-fixed costs. In particular, a lower probability of 23Inparticular,Axtell[2001]estimatesthattheslopeparameterof1.024,whileLuttmer[2007]recoversκ = 1.065. Thoseparameters,however,arenotcompatiblewithempiricallyconsistentsizepremiaamongexporters. 13

survival in services is associated with a lower ratio of sunk-to-fixed costs relative to manufacturers. Exploiting the conditions on the ratio of fixed costs that we have derived above, this implies that the sunk costs faced by services firms are less-than-twice as large than those among manufacturers. Estimates in the literature have generally suggested that sunk costs are substantially above fixed costs;withoutamorearticulatedframework,ourcalibrationdoesnothaveanydirectimplicationon theabsolutesizeofthosecosts,anditshouldbeinterpretedonlyinarelativesense,asacomparison between manufacturing and services exporters.24 5 Conclusion ThispaperpresentsaseriesofnovelfactsonU.S.exportersofservices. Ouranalysisshowsthatmost of the basic facts on exporters in manufacturing extend to services sectors, with three important distinctions. First, the participation rate of services firms in foreign markets is much lower than that of manufacturing firms. Second, the size premia at services exporters are significantly higher than those among manufacturers. Third, the survival rates of services exporters tend to be lower than that of manufacturing exporters. While our results partly rely on the composition of our data, which includes only major transactions of larger firms, the characteristics of U.S. services exporters we document are useful to infer some features of services trade, such as trade costs. Using a simple calibration, we find that services firms face two-to-three-time higher fixed costs than manufacturing exporters. These estimates are an important step to better understand and quantify the variation in the response of firms across sectors to changes in trade policy. 24Dasetal.[2007]andMoralesetal.[2014]estimatethatsunkcostsaresubstantiallyhigherthanfixedcosts; Albornozetal.[2016]findtheopposite. 14

Figure 1: Distribution of Customers across Exporters, Services Industries Figure 2: Firm-Customer Matching in Foreign Markets, Services Industries 15

Figure 3: Assortative Matching in Services Industries: Employment Ranking Figure 4: Assortative Matching in Services Industries: Sales Ranking 16

Figure 5: Exporters’ Share, 2000-2016 Figure 6: Evolution of the Share of Survivors, Manufacturing vs. Services 17

Table 1: Firm Exporting in Services (1) (2) (3) Percent Fractionof AvgShareofExports NAICS Sector ofFirms Exporters inTotalShipments 42 Wholesale Trade 100.0 0.27 0.32 423 DurableGoods 60.1 0.30 0.28 424 NondurableGoods 37.9 0.23 0.20 425 Agents&Brokers 1.2 0.12 0.50 44-45 Retail Trade 100.0 0.15 0.20 441 MotorVehicle&PartsDealers 7.1 0.17 0.14 442 FurnitureStores 3.1 0.00 0.00 443 ApplianceStore 5.3 0.32 0.70 444 Building&Garden 2.4 0.00 0.00 445 Food&BeverageStores 10.2 0.20 0.04 446 HealthStores 8.9 0.07 0.26 447 GasolineStations 2.4 0.00 0.00 448 ClothingStores 21.2 0.31 0.15 451 Sporting&Hobby 7.6 0.25 0.02 452 GeneralMerchandise 10.1 0.00 0.00 453 MiscellaneousStores 4.7 0.08 0.10 454 NonstoreRetailers 16.9 0.09 0.20 48-49 Transportation & Warehousing 100.0 0.35 0.23 481 AirTransportation 19.5 0.40 0.42 482 RailTransportation 5.6 0.16 0.08 483 WaterTransportation 22.0 0.50 0.45 484 TruckTransportation 14.1 0.20 0.13 485 PassengerTransportation 2.2 0.00 0.00 486 PipelineTransportation 18.7 0.13 0.08 488 SupportActivities 11.7 0.43 0.24 492 Couriers 5.0 0.31 0.47 51-56 Business Services 100.0 0.29 0.30 511 PublishingIndustries 9.7 0.40 0.34 512 MotionPicture 1.0 0.33 0.32 515 Broadcasting 1.9 0.39 0.31 517 Telecommunications 5.9 0.31 0.40 518 DataProcessing 1.9 0.32 0.26 519 OtherInformationServices 3.7 0.31 0.38 522 CreditIntermediation 21.3 0.18 0.22 523 Securities&OtherInvestments 5.5 0.33 0.30 524 InsuranceCarriers 5.6 0.37 0.27 525 Funds&Trusts 21.3 0.17 0.15 531 RealEstate 6.1 0.25 0.24 532 Rental&Leasing 1.1 0.38 0.43 533 LessorsofNonfinancialIntangibleAssets 1.5 0.40 0.43 541 ProfessionalServices 9.1 0.29 0.28 561 Administrative&SupportServices 3.1 0.16 0.26 562 WasteManagement&Remediation 1.0 0.17 0.19 61-81 Personal Services 100.0 0.26 0.18 611 EducationServices 7.6 0.19 0.20 621 AmbulatoryHealthCare 24.4 0.10 0.22 622 Hospitals 4.8 0.02 0.30 623 Nursing&ResidentialCare 4.6 0.07 0.14 624 SocialAssistance 0.9 0.00 0.00 711 PerformingArts 3.0 0.30 0.31 713 Amusement&Gambling 12.3 0.32 0.14 721 Accomodation 11.2 0.50 0.21 722 FoodServices 23.7 0.36 0.26 811 Repair&Maintenance 2.1 0.70 0.04 812 Personal&Laundry 5.2 0.26 0.04 Source: S&P Global Market Intelligence, Compustat North America, 2003-2007 Notes: Column 1 summarizes the average distribution of firms across services industries. Column 2 reports the average share of firms in each industry that export. Column 3 reports the average share of exports in total shipments across all exporters. Percentages in the third column do not sum exactly to 100 because of som1e8omitted sectors and rounding.

Table 2: Anatomy of Services Exports Services Centiles NumCountries NumCust. Density Avg. Exports 1 1 1 0.8 0.1 10 1 1 1.0 1.9 25 1 1 1.0 9.2 50 1 1 1.0 56.6 75 2 1 1.0 318.1 90 4 2 1.0 1,433.0 99 11 5 1.0 14,074.5 Source: S&P Global Market Intelligence, Compustat North America, 2003-2007. Note: Decomposition of U.S. services exports across customers along extensive (number of countries, number of customers, customer density) and intensive (average export values) margins. Services sectors include wholesale and retail trade, transportation and warehousing, business and personal services. Table 3: OLS Decomposition (1) (2) (3) (4) Customers Countries Density Intensive Exports 0.045*** 0.084*** -0.003*** 0.874*** (0.002) (0.003) (0.0003) (0.004) Obs. 10,807 10,807 10,807 10,807 Source: S&P Global Market Intelligence, Compustat North America. Legend: ∗∗∗ significant at 1%. Note: OLS Decomposition of services exports across customers along extensive and intensive margins. Each specification also includes sector-time dummies. Table 4: Type of Matches, Services Exporters One-to-one Many-to-one One-to-many Many-to-many ShareofMatches 11.7 8.7 52.3 27.3 ShareofValue 4.1 5.4 50.7 39.8 Source: S&P Global Market Intelligence, Compustat North America, 2003- 2007. Note: Distribution of firm-customer matches by type. 19

Table 5: Exporter Premia in Services (1) (2) (3) Dep. Variable ExporterPremia logL 0.415*** 0.727*** logR 0.520*** 0.787*** 0.119* logMktVal 0.875*** 0.928*** 0.351*** logY/L 0.098 0.064 0.122* logCapEx 0.795*** 1.007*** 0.300*** logK/L 0.264 0.157** 0.251*** Industry, Industryand AddVars YearFE YearFE YearFE,logL Source: S&P Global Market Intelligence, Compustat North America. log L: Employment (in log-s). log R: Revenues (in log-s). log Mkt Val: Market Valuation (in log-s). log Y/L: Real output per worker (in log-s). log CapEx: Capital Expenditure (in log-s). log K/L: Capital per worker (in log-s). Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: Average percent differences between exporters and non-exporters in a regression of firm characteristics on an export dummy. Table 6: Exit and Entry in Export Markets Sector Entrants Exiting Continuing ExitbyEntrants Manufacturing 11.8% 10.2% 79.6% 13.0% Wholesale 12.5% 10.2% 79.2% 15.2% Retail 23.0% 19.9% 61.9% 20.9% Transp. &Warehous. 13.5% 10.0% 78.2% 12.7% BusinessSvcs 15.4% 12.7% 74.2% 16.0% PersonalSvcs 18.1% 12.7% 71.9% 14.9% Source: S&P Global Market Intelligence, Compustat North America. Note: The table reports the share of new exporters over the total number of exporters (entrants), the share of firms that will not export the next year (exiting), the share of firms that were already exporting the previous year (continuing), and the share of entrants that will not export the following year (exit by entrants). Table 7: Characteristics of Entrants and Exiting Firms, Services Entrants Exiting ShareofFirms 15.5% 12.6% ShareofExportValue 6.8% 9.4% Avg. NumCustomers 1 1 Avg. Num. Countries 2 2 Density 1.0 1.0 Source: S&P Global Market Intelligence, Compustat North America. Note: Characteristics of firms entering and exiting the foreign market. 20

Table 8: Export Premia in Services: Entrants, Exiting, and Continuing Firms (1) (2) (3) (4) (5) (6) Dep. Variable Entrants Exiting logL -0.108 -0.305*** - -0.591*** -0.612*** logR -0.239*** -0.340*** -0.078* -0.744*** -0.701*** -0.185** logMktVal -0.168** -0.205** 0.037 -0.926*** -0.881*** -0.394*** logY/L -0.135* -0.040 -0.083* -0.160** -0.093* -0.189** logCapEx -0.348*** -0.364*** -0.086 -0.782*** -0.705*** -0.152 logK/L -0.154 0.022 -0.028 -0.179 -0.110 -0.227** Industry, Industryand Industry, Industryand AddVars YearFE YearFE YearFE YearFE,logL YearFE YearFE,logL Source: S&P Global Market Intelligence, Compustat North America. log L: Employment (in log-s). log R: Revenues (in log-s). log Mkt Val: Market Valuation (in log-s). log Y/L: Real output per worker (in log-s). log CapEx: Capital Expenditure (in log-s). log K/L: Capital per worker (in log-s). Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: Average percent differences relative to the group of continuing exporters in a regression of firm characteristics on an entry/exit dummy. Table 9: Time Series Decompositions, Services Exports (1) (2) (3) (4) Total Net Net Net Year Growth Entry Customer Intensive 2001 12.7 -0.4 -2.3 15.4 2002 6.9 1.2 -1.0 6.6 2003 2.2 -1.1 -2.4 5.8 2004 12.6 1.2 14.3 -2.9 2005 12.8 5.3 2.7 4.7 2006 3.2 -4.2 -7.1 14.5 2007 3.0 -1.6 -1.6 6.3 2008 2.0 -0.2 -0.9 3.1 2009 1.1 0.9 5.3 -5.1 2010 1.6 -0.4 -5.9 7.8 2011 5.3 3.3 2.7 -0.7 2012 4.3 -1.6 3.7 2.1 2013 2.8 1.4 -0.5 1.8 2014 0.5 -2.4 -1.4 4.3 2015 4.2 1.0 -1.4 4.7 2016 2.8 0.1 -0.5 3.3 Source: S&P Global Market Intelligence, Compustat North America. Notes: Changes in U.S. services exports and decomposition along extensive (net entry and net customer) and intensive margins. 21

References U.S. Census Bureau, County Business Patterns. https://www.census.gov/programs-surveys/ cbp.html. Accessed July 15, 2019. S & P Global. Compustat North America, Wharton Research Data Services (WRDS). https: //wrds-www.wharton.upenn.edu/pages/support/data-overview/. Accessed July 15, 2019. Facundo Albornoz, Sebasti´an Fanelli, and Juan Carlos Hallak. Survival in export markets. Journal of International Economics, 102:262–281, 2016. Andrea Ariu. Services versus goods trade: a firm-level comparison. Review of World Economics, 152(1):19–41, 2016. AndreaAriuandGiordanoMion. Tradeinservices: ITandtaskcontent. National Bank of Belgium working paper, (200), 2010. Robert L Axtell. Zipf distribution of US firm sizes. science, 293(5536):1818–1820, 2001. Felipe Benguria. The matching and sorting of exporting and importing firms: theory and evidence. Available at SSRN, https://ssrn.com/abstract=2638925, 2015. Andrew B Bernard, J Bradford Jensen, and Peter K Schott. Importers, exporters, and multinationals: A portrait of firms in the US that trade goods. Andrew B Bernard, J Bradford Jensen, Stephen J Redding, and Peter K Schott. Firms in international trade. Journal of Economic perspectives, 21(3):105–130, 2007. Andrew B Bernard, J Bradford Jensen, Stephen J Redding, and Peter K Schott. The margins of US trade. American Economic Review, 99(2):487–93, 2009. Andrew B Bernard, J Bradford Jensen, Stephen J Redding, and Peter K Schott. Global firms. Journal of Economic Literature, 56(2):565–619, 2018a. AndrewBBernard,AndreasMoxnes,andKarenHeleneUlltveit-Moe. Two-sidedheterogeneityand trade. Review of Economics and Statistics, 100(3):424–439, 2018b. Holger Breinlich and Chiara Criscuolo. International trade in services: A portrait of importers and exporters. Journal of International Economics, 84(2):188–206, 2011. Jer´onimo Carballo, Gianmarco IP Ottaviano, and Christian Volpe Martincus. The buyer margins of firms’ exports. Journal of International Economics, 112:33–49, 2018. SanghamitraDas,MarkJRoberts,andJamesRTybout.Marketentrycosts,producerheterogeneity, and export dynamics. Econometrica, 75(3):837–873, 2007. Raluca Dragusanu. Firm-to-firm matching along the supply chain. Technical report, Mimeo, 2014. Jonathan Eaton, Samuel Kortum, and Francis Kramarz. An anatomy of international trade: Evidence from French firms. Econometrica, 79(5):1453–1498, 2011. 22

Stefano Federico and Enrico Tosti. Exporters and importers of services: Firm-level evidence on Italy. The World Economy, 40(10):2078–2096, 2017. CEdwardFeeandShawnThomas. Sourcesofgainsinhorizontalmergers: evidencefromcustomer, supplier, and rival firms. Journal of Financial Economics, 74(3):423–460, 2004. Joseph Francois and Bernard Hoekman. Services trade and policy. Journal of economic literature, 48(3):642–92, 2010. Guillaume Gaulier, Emmanuel Milet, and Daniel Mirza. French firms in international service trade. Economie & Statistique, 2011. Antoine Gervais and J Bradford Jensen. The tradability of services: Geographic concentration and trade costs. Journal of International Economics, 118:331–350, 2019. Tanja Grubljesic and Joze Damijan. Differences in export behavior of services and manufacturing firmsinSlovenia. Economic and Business Review for Central and South-Eastern Europe,13(1/2): 77, 2011. Stefanie A Haller, Joˇze Damijan, Ville Kaitila, Cˇrt Kostevc, Mika Maliranta, Emmanuel Milet, Daniel Mirza, and Matija Rojec. Trading firms in the services sectors: comparable evidence from four EU countries. Review of World Economics, 150(3):471–505, 2014. J Bradford Jensen. Global trade in services: fear, facts, and offshoring. Peterson Institute for International Economics Washington, DC, 2011. Markus Kelle and J¨orn Kleinert. German firms in service trade. Technical report, Economics Working Paper, 2010. Erzo GJ Luttmer. Selection, growth, and the size distribution of firms. The Quarterly Journal of Economics, 122(3):1103–1144, 2007. Marc J Melitz. The impact of trade on intra-industry reallocations and aggregate industry productivity. econometrica, 71(6):1695–1725, 2003. Eduardo Morales, Gloria Sheu, and Andr´es Zahler. Gravity and extended gravity: Using moment inequalities to estimate a model of export entry. Technical report, National Bureau of Economic Research, 2014. Gary P Sampson and Richard H Snape. Identifying the issues in trade in services. World Economy, 8(2):171–182, 1985. Yoichi Sugita, Kensuke Teshima, and Enrique Seira. Assortative matching of exporters and importers. Institute of Developing Economies (IDE), 2016. Alexander Vogel and Joachim Wagner. Exports and profitability-first evidence for german business services enterprises. Applied Economics Quarterly, 56(1):7, 2010. P. Walter and R. Dell’mour. Firm Level Analysis of International Trade in Services : the Case of Austria. IFC Working Papers No.4. Bank for International Settlements., 2010. 23

A Additional Figures and Tables Figure A1: Distribution of Employment at Exporters, Manufacturing vs. Services Figure A2: Distribution of Total Sales at Exporters, Manufacturing vs. Services 24

Table A1: Distribution of Firms by Sector: Compustat vs. CBP Compustat NumEmpl Manufacturing Wholesale Retail Trans. &Warehous. Business Personal 0-4 1.2 2.8 0.7 9.8 4.6 0.5 5-9 1.3 0.8 0.9 1.7 0.9 0.4 10-19 2.7 1.0 0.9 2.1 3.2 0.8 20-99 10.1 12.1 5.7 4.2 12.8 4.2 101-499 19.3 11.2 10.2 7.4 24.6 10.1 500+ 65.6 72.0 81.6 74.8 53.9 84.0 CountyBusinessPatterns NumEmpl Manufacturing Wholesale Retail Trans. &Warehous. Business Personal 0-4 39.6 56.0 58.7 63.6 70.7 53.6 5-9 18.2 17.2 20.4 14.2 13.6 20.2 10-19 15.1 11.7 11.2 9.5 7.4 12.7 20-99 18.9 11.2 7.9 8.9 6.0 10.8 101-499 5.3 2.6 1.3 2.2 1.5 2.0 500+ 2.9 1.3 0.4 1.5 0.8 0.7 Source: County Business Patterns (CBP) and S&P Global Market Intelligence, Compustat North America. Note: Average distribution of firms within manufacturing and services sectors, 2002, 2007, and 2012. Table A2: Distribution of Employment by Sector, County Business Patterns NumEmployees Manufacturing Wholesale Retail Trans. &Warehous. Business Personal 0-4 1.7 5.4 5.0 4.1 6.2 4.8 5-9 2.8 6.5 6.5 4.1 5.2 6.8 10-19 4.7 9.0 7.2 5.5 5.7 8.7 20-99 17.2 23.2 14.0 14.3 13.4 21.0 101-499 18.1 17.2 8.2 11.9 13.2 16.1 500+ 55.6 38.6 58.9 60.1 56.2 42.5 Source: County Business Patterns (CBP). Note: Average distribution of employment within manufacturing and services sectors, 2002, 2007, and 2012. 25

Table A3: Firm Exporting in Manufacturing (1) (2) (3) Percent Fractionof AvgShareofExports NAICS Sector ofFirms Exporters inTotalShipments 31-33 Manufacturing 100.0 0.41 0.35 311 Food 3.8 0.32 0.26 312 Beverage&Tobacco 1.8 0.46 0.30 313 TextileMills 0.4 0.26 0.11 314 TextileProd 0.1 0.22 0.76 315 Apparel 1.9 0.31 0.35 316 Leather 0.7 0.44 0.17 321 Wood 1.2 0.31 0.38 322 Paper 2.0 0.40 0.54 323 Printing 0.9 0.15 0.36 324 Petroleum&Coal 1.6 0.48 0.39 325 Chemicals 25.3 0.38 0.40 326 Plastics&Rubber 2.2 0.36 0.28 327 NonmetallicProd 1.2 0.50 0.35 331 PrimaryMetals 2.7 0.48 0.34 332 FabricatedMetals 2.8 0.38 0.29 333 Machinery 8.0 0.53 0.42 334 Electronics 27.1 0.53 0.40 335 ElectricalEq 3.4 0.43 0.43 336 TransportationEq 5.5 0.56 0.36 337 Furniture 1.0 0.38 0.12 339 Miscellaneous 6.4 0.44 0.28 Source: S&P Global Market Intelligence, Compustat North America, 2003- 2007. Notes: Column 1 summarizes the average distribution of firms across manufacturing industries. Column 2 reports the average share of firms in each industry that export. Column 3 reports the average share of exports in total shipments across all exporters. Percentages in the third column do not sum exactly to 100 because of rounding. Table A4: Exports of Services Firms: Goods vs. Services CustomersinServices ShareofTransactions 55.2% ShareofValue 64.0% Source: S&P Global Market Intelligence, Compustat North America, 2003-2007. Note: Share of exports of firms in services industries to customers in services industries. 26

Table A5: Anatomy of Manufacturing Exports Manufacturing Centiles NumCountries NumCust. Density Avg. Exports 1 1 1 0.7 0.1 10 1 1 1.0 1.4 25 1 1 1.0 7.5 50 1 1 1.0 61.1 75 3 1 1.0 468.8 90 6 2 1.0 2,112.1 99 10 6 1.0 21,594.9 Source: S&P Global Market Intelligence, Compustat North America, 2003-2007. Note: Decomposition of U.S. manufacturing exports across customers along extensive (number of countries, number of customers, customer density) and intensive (average export values) margins. Table A6: OLS Decomposition, Manufacturing (1) (2) (3) (4) Customers Countries Density Intensive Exports 0.060*** 0.064*** -0.006*** 0.881*** (0.002) (0.002) (0.000) (0.003) Obs. 20,215 20,215 20,215 20,215 Source: S&P Global Market Intelligence, Compustat North America. Legend: ∗∗∗ significant at 1%. Note: OLS Decomposition of manufacturing exports across customers along extensive and intensive margins. Each specification also includes sector-time dummies. 27

Table A7: Anatomy of Services Exports, Sectoral Detail Wholesale Trade Centiles NumCountries NumCust. Density Avg. Exports 1 1 1 0.9 0.1 10 1 1 1.0 2.9 25 1 1 1.0 18.0 50 1 1 1.0 160.4 75 2 1 1.0 623.7 90 4 2 1.0 2,288.0 99 8 4 1.0 18,276.8 Retail Trade Centiles NumCountries NumCust. Density Avg. Exports 1 1 1 0.9 0.1 10 1 1 1.0 5.9 25 1 1 1.0 26.0 50 1 1 1.0 177.9 75 2 1 1.0 686.5 90 4 2 1.0 2,647.0 99 7 5 1.0 31,339.7 Transportation & Warehousing Centiles NumCountries NumCust. Density Avg. Exports 1 1 1 0.7 0.7 10 1 1 1.0 8.3 25 1 1 1.0 32.7 50 1 1 1.0 87.7 75 1 2 1.0 373.0 90 4 3 1.0 2,050.6 99 10 8 1.0 12,295.0 Business Services Centiles NumCountries NumCust. Density Avg. Exports 1 1 1 0.9 0.1 10 1 1 1.0 1.7 25 1 1 1.0 7.9 50 1 1 1.0 45.5 75 2 1 1.0 246.6 90 4 2 1.0 1,139.0 99 11 5 1.0 14,020.2 Personal Services Centiles NumCountries NumCust. Density Avg. Exports 1 1 1 0.8 0.1 10 1 1 1.0 1.1 25 1 1 1.0 5.5 50 1 1 1.0 35.1 75 1 1 1.0 350.3 90 3 2 1.0 2,000.0 99 10 11 1.0 4,840.0 Source: S&P Global Market Intelligence, Compustat North America, 2003-2007. Note: Sectoral decomposition of U.S. services exports across customers along extensive (number of countries, number of customers, customer density) and intensive (average export values) margins. 28

Table A8: Type of Matches, Manufacturing Exporters One-to-one Many-to-one One-to-many Many-to-many ShareofMatches 6.8 7.0 46.3 39.8 ShareofValue 4.1 6.8 43.0 46.1 Source: S&P Global Market Intelligence, Compustat North America, 2003- 2007. Note: Distribution of firm-customer matches by type. Table A9: Exporter-Customer Interactions Sector Coefficient Manufacturing -0.41 Wholesale -0.24 Retail -0.18 Transp. &Warehous. -0.42 BusinessSvcs -0.38 Personal -0.35 Source: S&P Global Market Intelligence, Compustat North America, 2003-2007. Note: OLS estimates from a regression of the average number of exporters per customers on the number of customers per exporter. 29

Table A10: Exporter Premia, Manufacturing vs. Services (1) (2) (3) (4) (5) VARIABLES logR logMktVal logY/L logCapEx logK/L Export 0.147*** 0.019 0.198*** 0.131*** 0.150*** (0.027) (0.060) (0.038) (0.035) (0.044) Export*Services -0.078 0.310*** -0.148 0.105 0.024 (0.072) (0.101) (0.088) (0.082) (0.091) Obs. 28,833 28,833 28,833 28,833 28,833 R2 0.847 0.626 0.238 0.748 0.398 Source: S&P Global Market Intelligence, Compustat North America. log R: Revenues (in log-s). log Mkt Val: Market Valuation (in log-s). log Y/L: Real output per worker (in log-s). log CapEx: Capital Expenditure (in log-s). log K/L: Capital per worker (in log-s). Export: dummy indicator for exporters. Services: dummy indicator for firms in the services sectors (NAICS 42, 44-45, 48-49, 51-56, and 61-81). Legend: ∗∗∗ significant at 1% and ∗∗ at 5%. Notes: Average percent differences between exporters and non-exporters in a regression of firm characteristics on an export dummy. All specification include time dummies, sector dummies, and a control for size (log Employment). 30

Table A11: Time Series Decompositions, Manufacturing Exports (1) (2) (3) (4) Total Net Net Net Year Growth Entry Customer Intensive 2001 -3.6 -4.3 -10.7 11.4 2002 11.1 -2.0 7.6 5.5 2003 -1.1 -2.1 -6.9 8.0 2004 6.2 0.6 8.8 -3.2 2005 12.4 -0.2 10.9 1.8 2006 11.7 5.4 -9.1 15.4 2007 0.1 -3.1 -2.1 5.2 2008 3.5 -0.4 2.9 1.0 2009 -4.4 -0.7 -2.7 -1.0 2010 6.4 1.5 1.3 3.6 2011 4.2 -2.2 2.9 3.4 2012 3.2 0.4 1.8 1.0 2013 5.0 0.4 1.5 3.1 2014 1.9 0.8 -4.4 5.5 2015 1.3 -0.5 1.3 0.5 2016 2.5 -0.1 -0.1 2.6 Source: S&P Global Market Intelligence, Compustat North America. Notes: Changes in U.S. manufacturing exports and decomposition along extensive (net entry and net customer) and intensive margins. 31