Misallocation in Open Economy

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Misallocation in Open Economy Maria D. Tito and Ruoying Wang 2021-007 Please cite this paper as: Tito, Maria D., and Ruoying Wang (2021). “Misallocation in Open Economy,” Finance and Economics Discussion Series 2021-007. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2021.007. 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.

Misallocation in Open Economy Maria D. Tito∗ Ruoying Wang†‡ January 15, 2021 Abstract Thispaperestimatestheimpactofreducingexportandimporttariffsonfirminputchoices. Inpresenceofborrowingconstraints,lowerexporttariffsfacilitatethereallocationofcapitaland laborinputsacrossfirms,whileadeclineinimporttariffseithertightensimportcompetitionor increasestheavailabilityofimportedinputs;allthreemechanismssuggestthatahigherdegree ofopennessshouldbeassociatedwithlowermisallocation. Toanalyzetheempiricalrelationship between openness and input misallocation, we draw on the annual surveys conducted by the ChineseNationalBureauofStatistics(NBS)between1998and2007. Fromthesurveys,weconstructfirm-levelmeasuresofinputmisallocationthatcontrolforfirmheterogeneity;weidentify shocks to openness using industry tariff levels and firm trade shares. We find that firm facing higher tariffs in either import or export markets make less optimal input choices. We further decompose our analysis between input and output tariffs: our results suggest that the labor reallocation mainly occurs because of lower input tariffs, while the selection effect induced by changes in output tariffs does not necessarily cause more distorted firms to exit and, therefore, tends to have an insignificant effect on input allocation. Finally, we calculate the contribution oftariffchangestowardsaggregatemisallocationandproductivity: ourresultsindicatethatthe impactoffirm-leveltariffreductionsonaggregatemisallocationandproductivitywasmarginal in our sample period, but the presence of sizeable interactions between trade shocks and misallocation at the sector level suggests that our result should be interpreted as a lower bound of the overall effect. Key words: Openness, Misallocation, Export Tariffs, Input and Output Tariffs. JEL classification: F14. ∗FederalReserveBoard. Contact: maria.d.tito@frb.gov. †LinkedIn. ‡Theviewspresentedinthispaperrepresentthoseoftheauthorsanddonotnecessarilycoincidewiththoseof theFederalReserveSystem. 1

1 Introduction An extensive literature documents productivity gains following tariff reductions.1. More recently, empirical work has delved into the channels for the realization of those gains, focusing on increasing returns to scale, on the self-selection of more efficient firms into exporting, or on the role of output vs. input tariffs.2 This paper explores a new channel for the realization of productivity gains fromtrade, theextenttowhichtariffsaffectaffectinput reallocationandinduce productivity gains. Whilethefactthatthereallocationofresourcesacrossfirmsmayresultinproductivitygainsiswelldocumented since the seminal work by Hsieh and Klenow [2009], there is little empirical evidence on the role that trade shocks have on resource misallocation, the central focus of our paper.3 Specifically, we estimate the effect of tariffs in export and import markets on labor and capital allocation decisions at the firm level. Other work suggests that lower tariffs induce lower misallocation. In particular, Tito and Wang [2017] offer a framework that ties export shocks to misallocation inpresenceoffinancialfrictions;inthatframework,ahigherdegreeofopennessinexportmarkets— which is associated with lower export tariffs—eases borrowing constraints and induce capital and laborchoicesclosertotheoptimalequilibrium. AmitiandKonings[2007]andGoldbergetal.[2010] identifies the effect of import tariffs on productivity, a connection that intuitively also relates to misallocation: A decline in import tariffs could either tighten import competition, forcing the least productive firms to shrink or exit the industry, or increases the availability and quality of imported inputs. Theformerchannelwouldreducemisallocationiftheshrinking/exitingfirmswerealsothose facing higher frictions, while the latter could facilitate the reallocation of inputs across firms. All told,bothchannelsintuitivelyimplythatlowertariffsshouldbeassociatedwithlowermisallocation. Toanalyzetheempiricalrelationshipbetweenopennessandinputmisallocation, wedrawonthe annual surveys conducted by the Chinese National Bureau of Statistics (NBS) between 1998 and 2007. We construct firm-level measures of input misallocation that control for firm heterogeneity; our measures show that misallocation in labor and capital markets, on average, moved down over 1SeePavcnik[2002]forChileandTybout[2003]forasurveyoftradeliberalizationreformsindevelopingcountries. 2SeeTyboutetal.[1991],TyboutandWestbrook[1995],HeadandRies[1999]andHeadandRies[2001]on theroleofincreasingreturnstoscale. Trefler[2004],Badinger[2007],andBadinger[2008]analyzegainsduetothe self-selectionofefficientfirmsintoexporting. AmitiandKonings[2007]disentangletheeffectsofoutputandinput tariffsonproductivity. 3AnexceptionistheworkbyCaggeseetal.[2019]thatreliesonfirm-specificdemandshockstoexplorefiring decisionsinpresenceoffinancialfrictions. Tradedata,intheircase,isusedtoidentifydemandshocksinconjunctionwithexchangerates. Theyfindthatconstrainedfirmsthatsufferanexchangerateappreciationshockfiremore short-tenuredthanlong-tenuredworkers,incomparisontofinanciallyunconstrainedfirms. Asshort-tenuredworkers,onaverage,havesteeperproductivityprofilesandlowerfiringcoststhanlong-tenuredworkers,theiranalysis suggeststhatappreciationshocks—whichproxiesanegativeliberalizationshockinexportmarkets,butapositive liberalizationshockonimportmarkets—increasemisallocation. 2

our sample period, a time when Chinese firms also experienced tariff declines in export and import markets. Usingfirm-levelimportandexporttariffs,calculatedfromtradesharesandindustrytariffs, we identify firms that have been exposed to large shocks—that is, firms facing above the median tariffs in export and import markets—and look at the impact of those shocks on input choices and misallocation. Our results confirm the intuition that openness is negatively correlated with misallocation: We find that firm facing higher tariffs in either import or export markets—synonym of lower openness—make less optimal input choices and, thus, experience higher misallocation. In particular, firm facing above-the-median export tariffs experience 2 percent of a sd higher capital frictions, while firms facing above-the-median import tariffs experience 1.5 percent of a sd higher frictions in labor markets. With endogeneity likely affecting the relationships between import tariffs and firm-level characteristics, we resort to an instrumental variable (IV) strategy, first proposed by Brandt et al. [2017], that uses the WTO schedules China agreed to follow upon accession to the WTO. That schedule, which sets the maximum tariff that can be charged for each products and has a compliance rate of around 97% after 2002, is known only in September 2001 and more likely to be exogenous. Our estimates of the effect of trade shocks on measures of misallocation are little changed after adopting this IV strategy. Following Amiti and Konings [2007], we further decompose the effect of import tariffs into the effects associated with input and output tariffs. We find that labor reallocation mainly occurs becauseoflowerinputtariffs,whiletheselectioneffectinducedbychangesinoutputtariffsdoesnot necessarilycausemoredistortedfirmstoexitand,therefore,tendstohaveaninsignificantimpacton inputallocation. Allourresultsarerobusttoaspecificationthatusesthemeasureofmisallocation, proposedbyPetrinandSivadasan[2013],thatreliesonthedeviationbetweenmarginalbenefitsand costs associated with labor and capital choices. Finally, we calculate the contribution of tariff changes towards aggregate misallocation and productivity. We confirm that our measures of misallocation are negatively correlated to sector-level productivity,butourestimatesimplythattheimpactoffirm-leveltariffreductionsonaggregatemisallocation and productivity was marginal during our sample period. The presence of more sizeable interactionsbetweentradeshocksandmisallocationinsector-levelregressions,however,suggestthat our firm-level estimates could be interpreted as a lower bound of the overall effect of trade shocks on misallocation. This paper extend Tito and Wang [2017]’s analysis to offer a more complete picture on the relationship between trade openness and misallocation at the firm level. Our work complements 3

Bai et al. [2019]. In that paper, the authors characterize the effects of trade openness in presence of exogenous frictions in output and input markets; in particular, they highlight that, after trade liberalization, average productivity could be lower if frictions offset true firm’s productivity and cause highly distorted/less productive firms to expand after opening to trade. Their prediction is confirmed in a quantification with Chinese data. The fact that our work, instead, highlights that trade liberalization could have the opposite effect on misallocation—that is, trade opening tends to decrease misallocation and increase aggregate productivity—mainly stems out of two differences. First, our analysis identifies the effect of trade on misallocation in continuing firms.4 Second, and likely most important, our analysis allows for an endogenous response of firm-level distortions to trade shocks, following Tito and Wang [2017], thus suggesting that if frictions themselves respond to trade shocks, the effect on aggregate productivity could still be positive, as in Melitz [2003], even in presence of friction. Therestofthepaperisorganizedasfollows. Section2presentsthedata,andsection3describes our empirical strategy, shows our results, and discusses the implication for aggregate misallocation and productivity. 2 Empirical Analysis 2.1 Data The empirical analysis draws on the Annual Survey of Industry (AIS) conducted by China’s National Bureau of Statistics. This dataset collects the balance sheet information of all state-owned enterprises and of non-state-owned firms with revenues above five million RMB (∼ USD 700,000) in the industrial sector. Our data extract is restricted to manufacturing firms sampled between 1998 and 2007; it contains 2,226,109 observations (here an observation is a firm-year combination). The survey collects data on revenues, employment, investments, and material purchases. We follow Brandt et al. [2012] to construct a real capital stock series from investments; moreover, we use their deflators for gross output, input, and capital. Following Yu [2015], we exclude all firms with fewer than 8 employees and with long-term assets above the total reported assets. After also dropping those firms with missing observations, we are left with a working sample of 1,214,513 observations. 4Oursector-levelregressionssuggestthattheeffectoftradeshocksonmisallocationcouldbedifferentinexiting firms,butfurtherempiricalworkisneededtocorroboratethiscorrelation. 4

Inaddition,wecombinebalance-sheetinformationwithexportandimportcustomsdatafor2000 to 2007. Using matching techniques similar to Yu [2015], we are able to match around 50 percent of the total number of observations. Finally, tariff data used to construct measures of market access are downloaded from the World Integrated Trade Solution (WITS) database. 2.2 Measuring Distortions: Firm-Level Measures Hsieh and Klenow [2009] show that frictions in input markets induce within-sector variation in the marginal revenue products of labor and capital across firms. Thus, within-sector measures of dispersion of marginal products proxy for the presence of distortions in a sector. Following a similar intuition, we propose firm-level measures of distortions that exploit the deviation of firmlevel outcomes from sectoral aggregates, an approach motivated by Tito and Wang [2017]. We construct our measures in two steps. First, we normalize the firm-level input product by the sector return and take log-s; for the labor return, for example, PistYist lnλ =ln List ist PstYst Lst This log-normalization conveniently shifts the distribution of relative labor products around zero. Second, we consider the deviation of relative labor returns from zero by constructing its absolute value; in fact, a sector with zero deviations across firms would approximate the frictionless equilibrium. In absence of heterogeneous frictions, the labor return of each individual firm would coincide with the sector return–i.e., lnλ = 0 for all firms in sector s at time t. Positive and negative deist viations of individual returns from zero reveal the presence of heterogeneous wedges affecting labor choices. In particular, firms with lnλ >0 demand less labor than the average firm in the sector, ist whilefirmswithlnλ <0demandmorelaborthantheaveragefirm; iftherewerenoothersources ist of firm heterogeneity, those differences in demand would be entirely explained by the presence of distortions. Thus, |lnλ | identifies the deviation from the sectoral averages and captures firm-level ist frictions in the labor market. Similarly, the log product of capital relative to the sector aggregate, lnκ ist PistYist lnκ =ln Kist ist PstYst Kst measures, in absolute value, distortions in capital markets. 5

Haltiwanger et al. [2018] point out that the measures of frictions based on Hsieh and Klenow [2009], however, require restrictive assumptions on the demand and supply system and may reflect factors other than distortions. In particular, we find that firm heterogeneity in productivity, labor shares,ormarkupscreatespuriousdispersionacrossfirminputproducts,and,thus,inourempirical analysis, we controls for those differences to effectively capture the presence of frictions within sectors.5 As an additional source of validation, we rely on Petrin and Sivadasan [2013]’s measure of misallocation, “input gaps”, calculated as the gap between a firm’s marginal input product and its marginal cost, (cid:12) (cid:12) Gj =(cid:12)MPj −p (cid:12), j =L,K ist (cid:12) ist ist(cid:12) As for our baseline measures, Petrin and Sivadasan [2013]’s gaps identify misallocation from firms’ suboptimal input choices. Though Petrin and Sivadasan [2013]’s gaps are based on less restrictive assumptions than Hsieh and Klenow [2009]’s, sources of firm heterogeneity could also inflate them. Inadirectcomparisonacrossthetwomeasuresofmisallocations,wefindthattheyarepositively correlated: inparticular,wefindacorrelationof0.25betweenmeasuresoffrictionsinlabormarkets, GL and|lnλ |,andacorrelationof0.46betweenmeasuresoffrictionsincapitalmarkets,GK and ist ist ist |lnκ|.6 Thepositivecorrelationsuggeststhatbothmeasuresarecapturing,atleastinpart,common factors driving suboptimal input choices. While our analysis includes results on both measures, we rely on |lnλ | and |lnκ | in our baseline specification as their scale-free characterization calls for ist ist an easier interpretation.7 Preliminary Evidence: Frictions and Trade Openness On average, firm-level frictions declined between 1998 and 2007. Figures 1 and 2 summarize the evolutionoftheaveragewithin-industrydistortionsinlaborandcapitalmarkets. Weconstructthese estimates by regressing |lnλ | and |lnκ | on time dummies after including controls that extract ist ist firm-leveldifferencesinproductivity,laborshares,andmark-ups; eachpointestimaterepresentsthe average distortion in a particular year relative to 1998, the base year. The figures suggest some differences across input markets. While distortions in the labor market trended down in the early part of the sample but moved up after 2004, frictions in capital markets declined throughout the 5SeeTitoandWang[2017]formoredetailsontheroleoffirmobservables. 6Weproxymarginalinputproductswithaverageproducts,themarginalwagewiththeaveragewage,andthe marginalrentalrateofcapitalwiththeaverageinterestrate. 7|lnλist|and|lnκist|capturepercentabsolutedeviationfromthesectoroutcomes;theinterpretationforGL ist andGK ,instead,dependsontheunitofmeasurement. ist 6

sample period: our proxy suggests that capital distortions were 40 percent lower in 2007 relative to 1998. Figures A1 and A2 highlight qualitatively similar patterns for Petrin and Sivadasan [2013]’s measures of distortions; a main difference applies to GL (figure A1), which displays a much larger ist increase over the last years in the sample, suggesting a rise in labor distortions relative to 1998. Over the same period, Chinese firms experienced notable tariff declines, despite tariff reductions wereimplementedsincethemid-nineties. Bothexportandimporttariffscontinuedtodecreaseafter 1998 (see figure A3); the decline accelerated around 2001, after China’s entry into the World Trade Organization. Export tariffs fell 3 percentage points between 2001 and 2004, from 12.4 percent in 2001 to 9.8 percent in 2004; they continued declining to 8.9 percent by 2007, the last year in our sample. The decline is more pronounced for import tariffs, which plunged around 4 percentage points between 2001 and 2002–from 12.81 percent in 2001 to 8.21 percent in 2002–and continued to slide to 4.75 percent by the end of our sample period. While the reduction in tariffs broadly aligns with the pattern of distortions, table 1 takes a first lookatthefirm-levelrelationshipbetweenfrictionsandopenness,proxiedbyfirmexportandimport status, in a specification that controls only for sector-year and province-year fixed effects. With the exception of the effect of import status on Petrin and Sivadasan [2013]’s labor gap, we find that firms that either export or import display lower misallocation relative to non-traders. In particular, looking within an industry, exporting firms display 4-to-12 percent of a standard deviation lower labor distortions and 4-to-7 percent of a standard deviation lower capital distortions relative to firms not engaged in foreign markets; labor misallocation at an importer is 2 percent of a standard deviationlowerandcapitalmisallocationis9-to-19percentofasdlowerthanatanon-traderwithin the same industry. This preliminary analysis is only suggestive of a relationship between frictions in input markets andtradeopennessduetotheendogeneityoffirmcharacteristicsandimportantomittedvariables— such as firm productivity, demand heterogeneity, and size differences—which may induce different input choices across firms. Next section explores an empirical strategy more robust to identifying a causal relationship. 7

3 Regression Analysis 3.1 Trade Shocks and Misallocation To identify the impact of trade openness on input choices, our main specification relates firm-level measures of misallocation to shocks to openness in import and export markets, y =β +β ·Export Shock +β ·Import Shock +γX +D +D +ε (1) ist 0 1 ist 2 ist ist i st ist where y denotes a measure of firm-level distortion. In absence of frictions, measures of distortions describedinsection2.1wouldbezeroaftercontrollingforsourcesoffirmheterogeneity;positiveand negative deviations from zero are associated with frictions affecting labor and/or capital choices. β and β are our main coefficients of interest; they investigate the impact of shocks in import 1 2 and export markets on our measures of misallocation. We conjecture that shocks to openness could either affect input choices in presence of exogenous frictions or have a direct impact on the source of misallocation, a case when frictions would endogenously respond to trade shocks. To quantify shocks to openness, we rely on export and import tariffs that firms face and follow two steps to construct our main regressors. First, we create firm-specific tariffs for all firms in our sample. We assign to non-exporters/non-importers the tariff of the industry in which they are classified. For firms operating in foreign markets, instead, we compute trade-weighted tariffs, using as weights the firm-level trade shares in the first year of foreign presence.8 In particular, our firm-level tariffs for firm i in year t, Exp. Tariff = (cid:88) Xk i,firstyear exp.τk+ Ds i,firstyear exp.τs ist (cid:80) Xk +Ds t (cid:80) Xk +Ds t k k i,firstyear i,firstyear k i,firstyear i,firstyear Imp. Tariff = (cid:88) Mk i,firstyear imp.τk+ Ds i,firstyear imp.τs ist (cid:80) Mk +Ds t (cid:80) Mk +Ds t k k i,firstyear i,firstyear k i,firstyear i,firstyear where exp.τs and imp.τs denote the sector-level export and import tariffs; Xk , Mk , t t i,firstyear i,firstyear and Dk represent exports, imports, and domestic production in the first year of operation. i,firstyear Second,weexploitthewithin-sectortariffdistributiontoidentifylargershocks; inparticular,inour baseline regression, we construct an indicator identifying firms facing tariffs above median of the 8Ourexport/importsharesarerelativetototalproductionand,thus,accountforthetariffoftheindustryin whichafirmisclassified,aswithdomesticproducers. 8

tariff distribution, Tariffs Above Median .9 Our firm-level indicators of above-the-median tariffs ist appropriately capture differences in openness across firms: we find a correlation of -.42 with firm export status and of -.64 with firm import status. Under this characterization, we expect β > 0 1 andβ >0, thatis, firmsfacingabove-the-mediantariffsmakeinputchoicescharacterizedbylarger 2 deviations from what is optimal. Our baseline regression includes a large set of firm, sector-time, and province-time fixed effects; in particular, the presence of firm dummies implies that we identify the effect of openness on misallocation only for continuing firms. Inaddition,ourspecificationalsoincludescontrolsoffirmheterogeneitythataffectinputchoices and that may cause spurious correlation between measures of openness and measures of distortions. In particular, we include the profit margin, lnψ , to control for differences in mark-ups, and we ist use TFP and the firm size proxies to control for heterogeneity in productivity.10 ist Finally, Tito and Wang [2017] show that older firms tend to face less binding constraints and make capital and labor choices closer to the optimal allocation; thus, in our specification, we also include firm’s age to isolate the role of openness in resource allocation.11 3.2 Results Thissectionssummarizestheestimatesformodel(1). Table2analyzestheeffectofhighertariffsin import and export markets on firm input allocation, measured by |lnλ | and |lnκ |. Our sample ist ist excludes firms in processing zones as those firms face different tariff regimes. The coefficients on the tariff indicators are positive across all column, but the effect of import shocks is significant only for labor choices, and the effect of export shocks tends to be significant only for capital choices. Thoseresultsarerobusttotheinclusionsofcontrolsforfirmheterogeneityandfirmage. Intermsof magnitudes, we rely on the estimates in columns (3) and (6): firm facing above-the-median export tariffs experience 2 percent of a sd higher capital frictions, while firms facing above-the-median import tariffs experience 1.5 percent of a sd higher frictions in labor markets. Other controls tend to display expected signs. In particular, more productive firms face higher |lnλ | and |lnκ |: this finding underlines that fact that firms with higher productivity demand ist ist more labor and capital and, thus, have a higher labor and capital products relative to the sector. Bigger firms, instead, tend to face lower frictions in input markets; this result is consistent with a 9Amongotherrobustnessanalysis,wealsoinvestigateshocksassociatedwiththe75thpercentiles. 10HsiehandSong[2015]showthattheprofitmarginisaproxyformark-ups. 11Wecalculatefirmageasthedifferencebetweenthesampleyearandyearofbirth. 9

modelwhereenteringinforeignmarketsisisomorphictoarevenueshock. Lowerfrictionsapplyalso to older firms, as suggested by Tito and Wang [2017]. Finally, the profit margin exhibits opposite effects on labor and capital choices, likely reflecting some patterns of substitutability across inputs: while a higher profit margin is associated with larger deviations from optimal labor choices, the profit margin is negatively correlated with distortions in capital markets. Results that rely on Petrin and Sivadasan [2013]’s measures of misallocation, are shown in table A1. The relation between tariffs and input gaps continues to display the expected positive sign but, with significant coefficients across all columns, appears even stronger than in our baseline. The magnitudes remain roughly similar: Relying on the estimates on columns (3) and (6), which conditiononasetofcontrolsanalogoustotable2,wefindthatfirmsfacingabove-the-medianexport tariffs experience 1.6 percent of sd higher labor and 2.8 percent of a sd higher capital distortions, while facing above-the-median import tariffs is associated with 2.4 percent of a sd higher output distortions and 3.2 percent of a sd higher capital distortions. Other differences from the baseline resultsariseovertheeffectofsomeofthecontrols: wefindthatprofitmarginisnegativelycorrelated with input gaps across all columns and that the effect of firm size switches sign between the labor gap and capital gap specification. The effects of total factor productivity and firm age, instead, are consistent across the two specifications: productivity is positively correlated with Petrin and Sivadasan [2013]’s measures of misallocation—a correlation that captures the higher input demand atmoreproductivefirms—andolderfirmsdisplaylowermisallocation,inthespiritofTitoandWang [2017]. The impact of larger tariff shocks—proxied by an indicator equal to 1 for firms facing tariffs above the75th percentile ofthe tariffdistribution, on measuresof misallocation—isshown in tables A2 and A3 and is not significantly different from our baseline results. Table A3 highlights that the effectoflargerexportshocksislessrobustonlaborchoicesbutcontinuestobeimportantforcapital choices. Instrumental Variable Strategy Our strategy of proxying trade shocks with firm-level tariffs is, however, not robust to concerns of endogeneity. While export tariffs are more likely to be exogenous since they are set by foreign governments, the political economy literature argues that firms with larger political power may influence import tariff schedules and tend to display different characteristics. Thus, this section develops an instrumental variable (IV) strategy for import tariffs. Following Brandt et al. [2017], we use the import tariff schedule that China agrees to follow upon its accession to the WTO in Dec 10

2001 as an instrument for applied tariffs. The schedule, which sets the maximum tariff that can be charged for each products and has a compliance rate around 97% after 2002, is known only in September 2001 and, therefore, more likely to be exogenous. With above-the-median import tariffs as our main dependent variable, we similarly construct firm-level import tariffs based on the WTO schedulesandidentifyfirmsfacingabove-the-medianWTOschedules. Asinourbaseline,ourmodel includesthefull setoffirm-level controlsaswellasfirm, sector-time, and province-timefixedeffects that also capture political economy factors that are firm-specific or that vary at the sector-time and province-time level. Table 3 shows the second stage results with |lnλ | and |lnκ | as our dependent variables.12 ist ist Our findings on the relation between import shocks and input choices remain robust: import tariffs continue to be positively correlated with measures of misallocation, although their impact is significant only in labor markets. Furthermore, the magnitudes of the effects of both import and export tariff are unchanged relative to our baseline estimates as are the effect of other controls. The results that use Petrin and Sivadasan [2013]’s gaps and adopt the IV strategy are shown in table A5. In this specification, the effect of import tariffs on misallocation is somewhat larger than in our baseline: firms facing above-the-median import tariffs experience 5.6 percent of sd higher labor and 6.7 percent of a sd higher capital distortions, conditioning on all other controls (columns (3) and (6)). These results suggest that the OLS estimates in table A1 may be biased downwards. If the omitted variable in the baseline specification captures the firm political influence, the higher coefficientonimportshocksimpliesthatfirmsinvestingmoreinlobbyingorotherpoliticallyoriented activitiesfacelessdistortionsininputmarketunderapositivecorrelationbetweenpoliticalinfluence andtheprobabilitytoattainhigherimporttariffs. Thischaracterizationsuggestthattheunobserved relation between political influence and misallocation resembles that at larger firms, an assumption that is consistent with the fact that lobbying firms are more likely to be larger than non-lobbying firms.13 The same type of bias does not apply to the results in table 3 because our measures of misallocation, |lnλ | and |lnκ |, are normalized by sector-level outcomes: lobbying on import ist ist tariffs would likely benefits all firms in the sector, thus shifting the average sector-level distribution of input products. In table A5, the effect of above-the-median export tariffs and of other controls is unchanged relative to the results in table A1. Looking into the Effect of Import Tariffs: Output and Input Tariffs 12FirststageresultsarereportedintableA4. ImporttariffsarepositivelycorrelatedwithWTOimportschedules;thetablealsoreportsthefirst-stageF-statistic,whichiswellabovetherecommendedthresholdcharacterizing weakinstruments. 13See,forexample,BorghesiandChang[2015]. 11

Theimpactofimporttariffshocksoninputallocationandproductivitycouldreflectshockstooutput tariffsand/orinputtariffs. Thosetwosourcesofvariationunderscoredifferentchannelsfortheeffect. In particular, empirical studies of the impact of trade liberalization on productivity suggest that reductions in output tariffs spur productivity via increased competition, while lower input tariff may affect productivity through the expansion in the variety of intermediate inputs available for production or the access to higher-quality inputs. We speculate that those mechanisms imply that lower tariffs are associated with lower misallocation; in fact, lower output tariffs may force the most constrained firms to shrink or exit the industry, while access to more or higher-quality intermediate inputs may affect the within-firm/within-sector allocation of inputs. This section provides evidence on these two channels. We first document general trends of output and input tariffs. Using the concordance developed by Brandt et al. [2017], we map the HS product classification to the four-digit China Industrial Classification (CIC) system; we, then, construct firm-level output tariffs by rescaling import tariff with (first-year) export and production values weights, Output Tariff = (cid:88) Xk i,firstyear τk+ Ds i,firstyear τs ist (cid:80) Xk +Ds t (cid:80) Xk +Ds t k k i,firstyear i,firstyear k i,firstyear i,firstyear where k denotes a 4-digit CIC sector, τk represent the import tariff in sector k, Xk and t i,firstyear Dk capture firm i exports and domestic production in sector k in the first year of operation. i,firstyear This formula extends Yu [2015]’s definition by taking into account firms’ domestic production; as with import and export tariffs, we consider domestic production to construct a tariff also for firms non directly involved in foreign markets. The input tariffs are weighted averages of import tariffs, rescaled by industry input shares and using first-year import and production values as weights, Input Tariff = (cid:88) Mk i,firstyear ωj,k τk+ (cid:88) D i,firstyear ωj,s τs ist (cid:80) Mk +D 2002 t (cid:80) Mk +D 2002 t k,j k i,firstyear i,firstyear j k i,firstyear i,firstyear whereMk denotesfirmimportsinindustryk forthefirstyearofpresenceinimportmarkets, i,firstyear andωj,k representstheshareofinputsfromindustryj usedinindustryk. Inputsharesareobtained from China’s 2002 input-output table. Figure A4 summarizes the average evolution of output and inputtariffsoveroursampleperiod.14 Inthelate1990s,outputtariffswere,onaverage,higherthan input tariffs, but the gap between the two measures has shrunk overtime: output tariffs fell from 17.1 percent in 1998 to 5.7 percent in 2007, while the decline in input tariffs, from 13.5 percent in 14Weusetradesectorsharestoconstructtariffspriorto2000,thefirstyearinourcustomdatabase. 12

1998 to 5.8 percent in 2007, has been a little more moderate. Table A6 documents a large degree of heterogeneityacrosssectors: thelargevariabilityoftheoutputandinputtariffdistributionsnotably shrinks over time, as sectors facing higher tariffs experience larger declines; by 2007, output tariffs range between 2 and 22 percent, while input tariffs span the 1-to-12 percent range. Moving back to our regression analysis, table 4 reports the effect of output and input tariffs on measures of misallocation; as in our baseline, we focus on the impact of above-the-median tariffs. Ouranalysisrevealsthatwithin-sectorreallocationofinputsisprimarilylinkedtoinputtariffshocks; their impact is significant for measures of labor and capital misallocation: using the estimates in columns (3) and (6), firms facing above-the-median input tariffs experience 2.1 percent of sd higher laborand1.8percentofasdhighercapitaldistortions. Outputtariffsarepositivelycorrelatedwith measures of frictions in input market, but their effect is not significant: we interpret this result as indicating that the impact of tighter import competition, associated with higher output tariffs, is either reflected mostly in sector-level measures of misallocation or exiting/shrinking firms are not necessarily those that face higher frictions in input markets. The dominant role of input tariff is also confirmed in our alternative specification that uses Petrin and Sivadasan [2013]’s measures of misallocation, shown in table A7. While output tariffs tend to be significant across all columns, the impact of input tariffs is significantly larger: above-the-median output tariffs are associated with 2.1 percent of a sd higher labor friction and 1 percent of a sd higher capital frictions compared with 5.5 percent of sd higher labor and 6.6 percent of a sd higher capital frictions for above-the-median input tariffs. Theeffectofabove-the-medianexporttariffsinbothtables4andA7remainconsistentwithour baseline. Concerns of endogeneity we described earlier also apply to output and input tariff. We develop a similar strategy that relies on the 2001 WTO schedules to construct instruments for output and input tariffs. We present the IV results in tables 5 and A8. Table 5 confirms that shocks to input tariffs tend to significantly affect the allocation of labor inputs; the effect on capital choices is, instead, not significant once we control also for firm size and age, after having included all other firm characteristics. In table A8, both higher output and input tariffs are associated with higher misallocation,buttheeffectofinputtariffsremainsmorethandoublethatofoutputtariff. Relative to the specification shown in table A7, the point estimates on above-the-median output tariffs are higher,consistentwithourintuitionthatthecoefficientsonimporttariffstendtobebiasedtowards zero; however, the difference in this case is not statistically significant.15 15Weexpecttheomittedvariablebiasrelatedtopoliticalinfluencetohaveadifferenteffectoninputtariffs. 13

3.3 Implications for Aggregate Misallocation and Productivity Our analysis so far has highlighted that trade shocks have a significant effect on the allocation of inputs,withalowerdegreeofopennessinducinghighermisallocationininputmarkets. Thissection investigates how the impact of trade shocks on firm-level input allocation translates into aggregate misallocation and productivity. To answer this question, we first look at the relationship between misallocationandproductivity. Whiletheeffectofmisallocationonproductivityiswell-knownsince the seminal work by Hsieh and Klenow [2009], we quantify that effect in our data. Table 6 looks at the sector-level relation between misallocation and total factor productivity (TFP). We find that capital misallocation has a negative and significant impact on TFP, a result that is robust to the inclusion of sector-level controls, such as age and the profit margin. Relying on the estimates in column(4)tocharacterizethemagnitudes,aone-standard-deviationdeclineincapitalmisallocation is associated with a 7 percent of a sd increase in productivity. The effect of labor misallocation on productivity tends, instead, not to be significantly different from zero, with a point estimate that shows a negative sign only in the column (1).16 Thus, combining our firm-level estimates on misallocationwithaggregateeffectsonproductivity, ourresultsimplythatfacingabovethemedian tariffs lowers productivity by 0.15 percent of a sd through firm-level input choices. Thismagnitude,however,impliesthattheshareoffirmsfacingabove-the-mediantariffsdeclined from one to zero over our sample period. To precisely calibrate the effect on productivity over the period, we look at the actual change in the share of firms facing above-the-median tariffs across sectors. Theshareoffirmsfacingabove-the-medianexporttariffsdeclined,onaverage,2percentage points(pp)between2001and2007;overthesameperiod,a2ppdeclinealsooccurredintheshareof firmsfacingabove-the-medianoutputtariffs, whiletheshareoffirmsfacingabove-the-medianinput tariffs declined 4 pp. Thus, using the estimates in table 5, we find that shocks to openness reduced labormisallocationby0.06percentandcapitalmisallocationby0.02percentbetween2001and2007; those effects map to an increase in productivity of 0.004 percent.17 While those magnitudes suggest that the effect of facing above-the-median tariffs was marginal on both aggregate misallocation and Continuingtoassumeanegativecorrelationbetweenpoliticalinfluenceandmeasuresofmisallocation,politically connectedfirmsarelikelytopetitionforlowertariffsininputmarkets. However,dependingonthenumberofinputsusedintheproductionprocess,lobbyingonoutputtariffsmightbeaneasierinvestment. Wefindsomeevidenceofupwardbiasoninputtariffcoefficientsintheresultsforlabormisallocation—tablesA7andA8;thedifferencebetweenthetwosetofestimatesis,however,notsignificant. Furthermore,thebiasseemtogotheotherway forcapitalmisallocation. 16Thisresultcouldaccountforthefactthatpositivetradeshocksmightnotlowermisallocationiffocusingon theextensivemarginofentryandexit,consistentwiththefindingsinBaietal.[2019]. 17UsingtheestimatesfromtableA8,wefindthatshockstoopennessreducedlabormisallocationby3.5percent andcapitalmisallocationby0.7percent;however,PetrinandSivadasan[2013]’smeasureoflabormisallocationwas littlechange,whiletheirmeasureofcapitalmisallocationshowed20percentdecline. 14

productivity, our estimates, however, focus on firm-level effects and abstract from declines in the mediantariff, aswellasotherchangesinthedistributionoftariffs.18 Wearguethatthoseestimates likely represent a lower bound of the effects of trade shocks on aggregate misallocation. Toprovidesomesuggestiveevidenceofthebroaderinteractionbetweentariffsandmisallocation, we looked at the relationship between sector-level measures of misallocation and tariffs. Table A10 presents our results for import and export tariffs, where we continue to use WTO schedules to instrument import tariffs.19 While export tariffs do not have a significant effect on sector level misallocation, higher import tariff are associated with higher misallocation; in particular, a onestandard-deviation higher import tariffs lower labor misallocation by 30 percent of a sd (column (1)). Within our sample, with import tariffs declining 6 percentage points between 2001 and 2007, the decline accounts for nearly all of the decline in labor misallocation.20 The decline in import tariffs also contributed to lower capital misallocation if looking at the results using the Petrin and Sivadasan[2013]’smeasure(column(4)): aone-standard-deviationhigherimporttariffslowercapital misallocationby15percentofasd(column(4)),accountingformorethanhalfofthedeclineinGK. The fact that the sector-level effects tend to be much larger than the firm-level effects suggests that ourfirm-levelresultspreciselyidentifyandisolateonlypartoftherelationbetweentradeshocksand misallocation, confirming our intuition that the effect we document should be considered a lower bound. 4 Conclusions This paper estimates the impact of openness on the reallocation of inputs across firms, a channel for the realization of the gains from trade. We find that firms’ input choices significantly respond to trade shocks quantified by firm-level tariffs in export and import markets. In particular, we find that firm facing above the median tariffs in import market tend to face larger distortions in labor markets, while the effect of above-the-median export tariffs is mainly directed at capital allocation. The import tariff effect is robust to an instrumental variable strategy that relies on the schedules ChinaagreedtofollowuponaccessiontotheWTO.Ourdecompositionoftheeffectofimportshocks into output and input tariffs indicates a larger role for input tariffs: in our baseline specification, 18Overthesameperiod,laborandcapitalmisallocationdeclined3percentand10percent,respectively,while totalfactorproductivityrosearound25percent. 19Wedonotincludeoutputandinputtariffsseparatelyduetotheirhighcollinearityatthesectorlevel. 20Whilecolumn(2)alsoindicatesasignificantinteractionbetweenimporttariffsandlabormisallocationusing PetrinandSivadasan[2013]’smeasure,thisresultisnotconsistentwiththefactthatGL roseoveroursampleperiod. 15

output tariffs do not significantly affect input choices, while input tariffs significantly reduce labor misallocation. All told, facing above the median tariff lower productivity, but we estimate that the impact of firm-level tariff reductions on productivity through misallocation was marginal in our sample period. The channel that we identify, however, operates only via firm-level choices and, looking at sector-level correlations, likely represent a lower bound of overall effect of trade shocks on misallocation and productivity. 16

References Mary Amiti and Jozef Konings. Trade liberalization, intermediate inputs, and productivity: Evidence from Indonesia. The American Economic Review, pages 1611–1638, 2007. Harald Badinger. Market size, trade, competition and productivity: Evidence from OECD manufacturing industries. Applied Economics, 39(17):2143–2157, 2007. Harald Badinger. Trade policy and productivity. European Economic Review, 52(5):867–891, 2008. YanBai,KeyuJin,andDanLu. Misallocationundertradeliberalization. Technicalreport,National Bureau of Economic Research, 2019. Richard Borghesi and Kiyoung Chang. The determinants of effective corporate lobbying. Journal of Economics and Finance, 39(3):606–624, 2015. Loren Brandt, Johannes Van Biesebroeck, and Yifan Zhang. Creative accounting or creative destruction? firm-level productivity growth in Chinese manufacturing. Journal of Development Economics, 97(2):339–351, 2012. Loren Brandt, Johannes Van Biesebroeck, Luhang Wang, and Yifan Zhang. WTO accession and performance of Chinese manufacturing firms. American Economic Review, 107(9):2784–2820, 2017. Andrea Caggese, Vicente Cun˜at, and Daniel Metzger. Firing the wrong workers: Financing constraints and labor misallocation. Journal of Financial Economics, 133(3):589–607, 2019. Pinelopi Koujianou Goldberg, Amit Kumar Khandelwal, Nina Pavcnik, and Petia Topalova. Imported intermediate inputs and domestic product growth: Evidence from India. The Quarterly journal of economics, 125(4):1727–1767, 2010. John Haltiwanger, Robert Kulick, and Chad Syverson. Misallocation measures: The distortion that ate the residual. Technical report, National Bureau of Economic Research, 2018. Keith Head and John Ries. Rationalization effects of tariff reductions. Journal of International economics, 47(2):295–320, 1999. Keith Head and John Ries. Increasing returns versus national product differentiation as an explanation for the pattern of US-Canada trade. American Economic Review, 91(4):858–876, 2001. Chang-Tai Hsieh and Peter J Klenow. Misallocation and manufacturing TFP in China and India. The Quarterly Journal of Economics, 124(4):1403–1448, 2009. Chang-Tai Hsieh and Zheng Michael Song. Grasp the large, let go of the small: The transformation of the state sector in China. Technical report, National Bureau of Economic Research, 2015. Marc J Melitz. The impact of trade on intra-industry reallocations and aggregate industry productivity. econometrica, 71(6):1695–1725, 2003. Nina Pavcnik. Trade liberalization, exit, and productivity improvements: Evidence from Chilean plants. The Review of Economic Studies, 69(1):245–276, 2002. 17

Amil Petrin and Jagadeesh Sivadasan. Estimating lost output from allocative inefficiency, with an application to Chile and firing costs. Review of Economics and Statistics, 95(1):286–301, 2013. MariaD.TitoandRuoyingWang. Exportingandfrictionsininputmarkets: EvidencefromChinese data. Technical Report 077, Board of Governors of the Federal Reserve System (US), 2017. Daniel Trefler. The long and short of the Canada-US free trade agreement. American Economic Review, 94(4):870–895, 2004. James Tybout, Jamie De Melo, and Vittorio Corbo. The effects of trade reforms on scale and technical efficiency: New evidence from Chile. Journal of International economics, 31(3-4):231– 250, 1991. James R Tybout. Plant-and firm-level evidence on new trade theories. Handbook of international trade, 1(1):388–415, 2003. James R Tybout and M Daniel Westbrook. Trade liberalization and the dimensions of efficiency change in Mexican manufacturing industries. Journal of International Economics, 39(1-2):53–78, 1995. Miaojie Yu. Processing trade, tariff reductions and firm productivity: Evidence from Chinese firms. The Economic Journal, 125(585):943–988, 2015. 18

Figure 1: Dispersion across Labor Returns Figure 2: Dispersion across Capital Returns 19

Table 1: Misallocation, Exporters, and Importers: Descriptive Evidence (1) (2) (3) (4) Variables |lnλ| GL |lnκ| GK Export -0.028*** -2.954*** -0.065*** -0.149*** (0.003) (0.113) (0.004) (0.018) Import -0.011*** 4.925*** -0.082*** -0.708*** (0.003) (0.185) (0.004) (0.019) Sector-Year y y y y Prov-Year y y y y Obs. 1,181,051 1,181,051 1,181,051 1,181,051 R2 0.043 0.100 0.026 0.062 ln age lnλ: log return to labor relative to the sector. GL: labor gap from Petrin and Sivadasan [2013]. lnκ: log return to capital relative to the sector. GK: capital gap from Petrin and Sivadasan [2013]. Export : export status for firm i at time t. t Import : import status for firm i at time t. t Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: OLS regressions, 1998-2007. Firm-level clustered standard errors are reported in parenthesis. 20

Table 2: Misallocation, Export Tariffs, and Import Tariffs (1) (2) (3) (4) (5) (6) Variables |lnλ| |lnκ| ExpTariffsAboveMedian 0.005 0.005 0.003 0.026*** 0.025*** 0.018*** (0.004) (0.004) (0.004) (0.005) (0.005) (0.005) ImpTariffsAboveMedian 0.012*** 0.013*** 0.011*** 0.013*** 0.012** 0.004 (0.004) (0.004) (0.004) (0.005) (0.005) (0.005) TFP 0.066*** 0.066*** 0.066*** 0.267*** 0.268*** 0.309*** (0.002) (0.002) (0.002) (0.003) (0.003) (0.003) lnψ 0.011*** 0.012*** -0.014*** -0.014*** (0.001) (0.001) (0.001) (0.001) lnAge -0.031*** -0.193*** (0.003) (0.004) lnK -0.027*** (0.001) lnEmpl -0.146*** (0.003) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 1,205,375 1,205,375 1,205,375 1,205,375 1,205,375 1,205,375 R2 0.015 0.015 0.016 0.057 0.058 0.073 NumberofFirmIDs 406,170 406,170 406,170 406,170 406,170 406,170 lnλ: log return to labor relative to the sector. lnκ: log return to capital relative to the sector. Exp Tariffs Above Median: dummy equal to one if firm export tariff is above the 50th percentile within an industry. Imp Tariffs Above Median: dummy equal to one if firm import tariff is above the 50th percentile within an industry. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnψ: profit margin. ln Age: log firm age. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 21

Table 3: Misallocation and Tariffs: Instrumenting Import Tariffs with WTO Schedules (1) (2) (3) (4) (5) (6) Variables |lnλ| |lnκ| ExpTariffsAboveMedian 0.005 0.005 0.003 0.026*** 0.025*** 0.018*** (0.004) (0.004) (0.004) (0.005) (0.005) (0.005) ImpTariffsAboveMedian 0.013*** 0.013*** 0.010** 0.013** 0.013*** 0.000 (0.004) (0.004) (0.004) (0.005) (0.005) (0.005) TFP 0.066*** 0.066*** 0.066*** 0.267*** 0.268*** 0.309*** (0.002) (0.002) (0.002) (0.003) (0.003) (0.003) lnψ 0.011*** 0.012*** -0.014*** -0.014*** (0.001) (0.001) (0.001) (0.001) lnAge -0.031*** -0.193*** (0.003) (0.004) lnK -0.027*** (0.001) lnEmpl -0.146*** (0.003) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 1,205,375 1,205,375 1,205,375 1,205,375 1,205,375 1,205,375 R2 0.015 0.015 0.016 0.057 0.058 0.073 NumberofFirmIDs 406,170 406,170 406,170 406,170 406,170 406,170 lnλ: log return to labor relative to the sector. lnκ: log return to capital relative to the sector. Exp Tariffs Above Median: dummy equal to one if firm export tariff is above the 50th percentile within an industry. Imp Tariffs Above Median: dummy equal to one if firm import tariff is above the 50th percentile within an industry. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnψ: profit margin. ln Age: log firm age. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 22

Table 4: Misallocation and Tariffs: Input and Output Tariffs (1) (2) (3) (4) (5) (6) Variables |lnλ| |lnκ| ExpTariffsAboveMedian 0.004 0.004 0.002 0.023*** 0.022*** 0.017*** (0.004) (0.004) (0.004) (0.005) (0.005) (0.005) OutTariffsAboveMedian 0.006 0.006 0.005 0.012** 0.012** 0.004 (0.005) (0.005) (0.005) (0.006) (0.006) (0.006) InpTariffsAboveMedian 0.019*** 0.020*** 0.015*** 0.031*** 0.031*** 0.014** (0.005) (0.005) (0.005) (0.006) (0.006) (0.006) TFP 0.066*** 0.066*** 0.066*** 0.268*** 0.268*** 0.309*** (0.002) (0.002) (0.002) (0.003) (0.003) (0.003) lnψ 0.011*** 0.012*** -0.014*** -0.014*** (0.001) (0.001) (0.001) (0.001) lnAge -0.029*** -0.193*** (0.003) (0.004) lnK -0.027*** (0.001) lnEmpl -0.146*** (0.003) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 1,214,513 1,214,513 1,214,513 1,214,513 1,214,513 1,214,513 R2 0.015 0.015 0.016 0.057 0.058 0.073 NumberofFirmIDs 409,213 409,213 409,213 409,213 409,213 409,213 lnλ: log return to labor relative to the sector. lnκ: log return to capital relative to the sector. Exp Tariffs Above Median: dummy equal to one if firm export tariff is above the 50th percentile within an industry. Out Tariffs Above Median: dummy equal to one if firm output tariff is above the 50th percentile within an industry. Inp Tariffs Above Median: dummy equal to one if firm input tariff is above the 50th percentile within an industry. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnψ: profit margin. ln Age: log firm age. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 23

Table 5: Misallocation and Tariffs: Instrumenting Input and Output Tariffs (1) (2) (3) (4) (5) (6) Variables |lnλ| |lnκ| ExpTariffsAboveMedian 0.002 0.002 0.001 0.024*** 0.023*** 0.019*** (0.004) (0.004) (0.004) (0.005) (0.005) (0.005) OutTariffsAboveMedian 0.008 0.008 0.007 0.009 0.009 -0.002 (0.007) (0.007) (0.007) (0.008) (0.008) (0.008) InpTariffsAboveMedian 0.032*** 0.032*** 0.028*** 0.022*** 0.022*** 0.001 (0.006) (0.006) (0.006) (0.007) (0.007) (0.007) TFP 0.066*** 0.066*** 0.066*** 0.268*** 0.268*** 0.309*** (0.002) (0.002) (0.002) (0.003) (0.003) (0.003) lnψ 0.011*** 0.012*** -0.014*** -0.014*** (0.001) (0.001) (0.001) (0.001) lnAge -0.029*** -0.193*** (0.003) (0.004) lnK -0.027*** (0.001) lnEmpl -0.146*** (0.003) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 1,214,513 1,214,513 1,214,513 1,214,513 1,214,513 1,214,513 NumberofFirmIDs 409,213 409,213 409,213 409,213 409,213 409,213 lnλ: log return to labor relative to the sector. lnκ: log return to capital relative to the sector. Exp Tariffs Above Median: dummy equal to one if firm export tariff is above the 50th percentile within an industry. Out Tariffs Above Median: dummy equal to one if firm output tariff is above the 50th percentile within an industry. Inp Tariffs Above Median: dummy equal to one if firm input tariff is above the 50th percentile within an industry. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnψ: profit margin. ln Age: log firm age. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 24

Table 6: Aggregate Misallocation and TFP (1) (2) (3) (4) Variables TotalFactorProductivity Avg|lnλ| -0.023 0.078 0.057 (0.055) (0.054) (0.047) Avg|lnκ| -0.178*** -0.206*** -0.200*** (0.052) (0.052) (0.048) Avglnψ 0.110*** (0.029) AvgAge -0.334*** (0.046) Year y y y y SectorFE y y y y Obs. 4,180 4,180 4,180 4,180 R2 0.902 0.904 0.904 0.916 NumberofIndustries 423 423 423 423 TFP: average sector-level total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon- Petrin methodology. Avg |lnλ|: average dispersion across absolute relative labor returns within sector. Avg |lnκ|: average dispersion across absolute relative capital returns within sector. Avg lnψ: average profit margin within sector. Avg Age: average firm age within sector. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: Sector-level FE regressions. Standard errors, clustered at the sector levels, are reported in parenthesis. 25

A Additional Empirical Results Figure A1: Dispersion across Labor Returns Figure A2: Dispersion across Capital Returns 26

Figure A3: Export and Import Tariffs, 1998–2007 Figure A4: Output and Input Tariffs, 1998–2007 27

Table A1: Misallocation and Tariffs: Petrin and Sivadasan [2013]’s Measure (1) (2) (3) (4) (5) (6) Variables GL GK ExpTariffsAboveMedian 0.346** 0.342** 0.398** 0.130*** 0.129*** 0.105*** (0.163) (0.163) (0.164) (0.018) (0.018) (0.017) ImpTariffsAboveMedian 0.510*** 0.505*** 0.578*** 0.155*** 0.154*** 0.120*** (0.168) (0.168) (0.168) (0.017) (0.017) (0.016) TFP 14.810*** 14.819*** 14.833*** 2.443*** 2.445*** 2.624*** (0.088) (0.088) (0.089) (0.012) (0.012) (0.013) lnψ -0.333*** -0.349*** -0.051*** -0.049*** (0.023) (0.023) (0.003) (0.003) lnAge -0.261** -0.643*** (0.125) (0.017) lnK 1.156*** (0.065) lnEmpl -0.606*** (0.011) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 914,553 914,553 914,553 914,553 914,553 914,553 R2 0.228 0.229 0.230 0.247 0.248 0.262 NumberofFirmIDs 329,719 329,719 329,719 329,719 329,719 329,719 GL: labor gap from Petrin and Sivadasan [2013]. GK: capital gap from Petrin and Sivadasan [2013]. Exp Tariffs Above Median: dummy equal to one if firm export tariff is above the 50th percentile within an industry. Imp Tariffs Above Median: dummy equal to one if firm import tariff is above the 50th percentile within an industry. lnψ: profit margin. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 28

Table A2: Misallocation and Tariffs, Tariffs above the 75th Percentile (1) (2) (3) (4) (5) (6) Variables |lnλ| |lnκ| ExpTariffsAbove75th Pctile 0.002 0.002 0.001 0.019*** 0.018*** 0.013*** (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) ImpTariffsAbove75th Pctile 0.013*** 0.013*** 0.011*** 0.013*** 0.013*** 0.004 (0.004) (0.004) (0.004) (0.005) (0.005) (0.005) TFP 0.066*** 0.066*** 0.066*** 0.267*** 0.268*** 0.309*** (0.002) (0.002) (0.002) (0.003) (0.003) (0.003) lnψ 0.011*** 0.012*** -0.014*** -0.014*** (0.001) (0.001) (0.001) (0.001) lnAge -0.031*** -0.193*** (0.003) (0.004) lnK -0.027*** (0.001) lnEmpl -0.146*** (0.003) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 1,205,375 1,205,375 1,205,375 1,205,375 1,205,375 1,205,375 R2 0.015 0.015 0.016 0.057 0.058 0.073 NumberofFirmIDs 406,170 406,170 406,170 406,170 406,170 406,170 lnλ: log return to labor relative to the sector. lnκ: log return to capital relative to the sector. Exp Tariffs Above 75th Pctile: dummy equal to one if firm export tariff is above the 75th percentile within an industry. Imp Tariffs Above 75th Pctile: dummy equal to one if firm import tariff is above the 75th percentile within an industry. lnψ: profit margin. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 29

Table A3: Misallocation and Tariffs, Tariffs above the 75th Percentile, Petrin and Sivadasan [2013]’s Measure (1) (2) (3) (4) (5) (6) Variables GL GK ExpTariffsAbove75th Pctile 0.221 0.219 0.263* 0.104*** 0.104*** 0.091*** (0.142) (0.142) (0.142) (0.018) (0.018) (0.017) ImpTariffsAbove75th Pctile 0.517*** 0.512*** 0.585*** 0.157*** 0.156*** 0.122*** (0.167) (0.167) (0.167) (0.017) (0.017) (0.016) TFP 14.809*** 14.818*** 14.832*** 2.443*** 2.445*** 2.624*** (0.088) (0.088) (0.089) (0.012) (0.012) (0.013) lnψ -0.333*** -0.349*** -0.051*** -0.049*** (0.023) (0.023) (0.003) (0.003) lnAge -0.263** -0.644*** (0.125) (0.017) lnK 1.156*** (0.065) lnEmpl -0.606*** (0.011) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 914,553 914,553 914,553 914,553 914,553 914,553 R2 0.228 0.229 0.230 0.247 0.248 0.262 NumberofFirmIDs 329,719 329,719 329,719 329,719 329,719 329,719 GL: labor gap from Petrin and Sivadasan [2013]. GK: capital gap from Petrin and Sivadasan [2013]. Exp Tariffs Above 75th Pctile: dummy equal to one if firm export tariff is above the 75th percentile within an industry. Imp Tariffs Above 75th Pctile: dummy equal to one if firm import tariff is above the 75th percentile within an industry. lnψ: profit margin. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 30

Table A4: Instrumenting Import Tariffs with WTO Schedules: First Stage (1) (2) (3) (4) Variables ImpTariffsAboveMedian WTOImpSched. AboveMedian 0.012*** 0.013*** 0.011*** 0.013*** (0.004) (0.004) (0.004) (0.005) TFP 0.066*** 0.066*** 0.066*** 0.267*** (0.002) (0.002) (0.002) (0.003) lnψ 0.011*** 0.012*** -0.014*** (0.001) (0.001) (0.001) lnAge -0.031*** -0.193*** (0.003) (0.004) lnK -0.027*** (0.001) lnEmpl -0.146*** (0.003) Sector-Year y y y y Prov-Year y y y y FirmFE y y y y F-stat 12,464 12,464 12,459 12,453 Obs. 1,205,375 1,205,375 1,205,375 1,205,375 R2 0.015 0.015 0.016 0.057 NumberofFirmIDs 406,170 406,170 406,170 406,170 Imp Tariffs Above Median: dummy equal to one if firm import tariff is above the 50th percentile within an industry. WTO Imp Schedules Above Median: dummy equal to one if firm import tariff is above the 50th percentile within an industry. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnψ: profit margin. ln Age: log firm age. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: First-stage regression for tables 3 and A5, 2000-2006. Columns (1)-(2) are first stage estimates for columns (1)-(2) and (4)-(5), column (3) contains the first stage estimate for column (3), and column (4) reports the first-stage estimates for column (6). Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 31

Table A5: Misallocation and Tariffs, Petrin and Sivadasan [2013]’s Measure: Instrumenting Import Tariffs with WTO Schedules (1) (2) (3) (4) (5) (6) Variables GL GK ExpTariffsAboveMedian 0.311* 0.307* 0.358** 0.120*** 0.119*** 0.099*** (0.164) (0.164) (0.164) (0.018) (0.018) (0.018) ImpTariffsAboveMedian 1.204*** 1.208*** 1.375*** 0.352*** 0.353*** 0.251*** (0.380) (0.380) (0.381) (0.038) (0.038) (0.037) TFP 14.814*** 14.823*** 14.837*** 2.444*** 2.446*** 2.624*** (0.088) (0.088) (0.089) (0.012) (0.012) (0.013) lnψ -0.332*** -0.349*** -0.051*** -0.049*** (0.023) (0.023) (0.003) (0.003) lnAge -0.262** -0.643*** (0.125) (0.017) lnK 1.161*** (0.065) lnEmpl -0.605*** (0.011) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 914,553 914,553 914,553 914,553 914,553 914,553 R2 0.015 0.015 0.016 0.057 0.058 0.073 NumberofFirmIDs 329,719 329,719 329,719 329,719 329,719 329,719 GL: labor gap from Petrin and Sivadasan [2013]. GK: capital gap from Petrin and Sivadasan [2013]. Exp Tariffs Above Median: dummy equal to one if firm export tariff is above the 50th percentile within an industry. Imp Tariffs Above Median: dummy equal to one if firm import tariff is above the 50th percentile within an industry. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnψ: profit margin. ln Age: log firm age. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 32

Table A6: Input and Output Tariffs across Sectors, 1998 and 2007 1998 2007 Industry Output Input Output Input Food 35.78 15.56 13.63 8.07 Beverages 41.74 6.93 18.63 3.28 Tobacco 55.39 16.42 21.74 9.81 Textile 25.32 22.20 9.13 9.86 Apparel 33.61 8.02 14.34 4.25 Leather 12.93 22.02 10.97 12.07 Wood 12.26 8.41 3.91 2.90 Furniture 22.00 2.94 2.50 1.38 Paper 16.63 12.77 6.08 6.11 Printing 1.73 4.15 3.17 2.03 Recreational 19.65 8.54 7.78 3.82 Petroleum&Coal 7.10 3.86 5.57 2.15 Chemicals 12.35 12.52 6.56 7.07 Pharmaceuticals 10.78 2.91 5.32 1.42 SyntheticFibers 19.29 20.44 4.76 8.46 Rubber 18.54 15.80 10.13 7.86 Plastics 17.73 15.69 8.40 8.00 Clay,Stone,andGlass 17.00 6.40 10.64 3.80 Metals 9.56 10.69 5.52 5.82 MetalProducts 12.75 9.44 8.39 5.23 Machinery 13.26 11.79 7.71 5.58 Equipment 12.51 6.22 6.88 3.07 TransportationEq 26.30 16.01 10.05 7.64 ElectronicProducts 15.22 9.53 7.75 4.45 Computer&CommEq 14.97 9.63 3.96 3.62 OtherMfg 27.12 8.96 16.03 4.99 Source: WITSand2002Input-OutputTables,1998and2007. 33

Table A7: Misallocation and Tariffs, Petrin and Sivadasan [2013]’s Measure: Output and Input Tariffs (1) (2) (3) (4) (5) (6) Variables GL GK ExpTariffsAboveMedian 0.220 0.216 0.262 0.109*** 0.108*** 0.091*** (0.163) (0.162) (0.163) (0.018) (0.018) (0.018) OutTariffsAboveMedian 0.506*** 0.510*** 0.533*** 0.058*** 0.059*** 0.035* (0.192) (0.191) (0.192) (0.020) (0.020) (0.020) InpTariffsAboveMedian 1.191*** 1.181*** 1.347*** 0.311*** 0.310*** 0.245*** (0.196) (0.196) (0.197) (0.022) (0.022) (0.021) TFP 14.786*** 14.796*** 14.812*** 2.452*** 2.453*** 2.632*** (0.088) (0.088) (0.088) (0.012) (0.012) (0.013) lnψ -0.331*** -0.348*** -0.051*** -0.049*** (0.022) (0.022) (0.003) (0.003) lnAge -0.278** -0.646*** (0.125) (0.017) lnK 1.165*** (0.065) lnEmpl -0.607*** (0.011) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 919,339 919,339 919,339 919,339 919,339 919,339 R2 0.228 0.228 0.229 0.247 0.248 0.262 NumberofFirmIDs 331,509 331,509 331,509 331,509 331,509 331,509 GL: labor gap from Petrin and Sivadasan [2013]. GK: capital gap from Petrin and Sivadasan [2013]. Exp Tariffs Above Median: dummy equal to one if firm export tariff is above the 50th percentile within an industry. Out Tariffs Above Median: dummy equal to one if firm output tariff is above the 50th percentile within an industry. Inp Tariffs Above Median: dummy equal to one if firm input tariff is above the 50th percentile within an industry. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnψ: profit margin. ln Age: log firm age. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 34

Table A8: Misallocation and Tariffs, Petrin and Sivadasan [2013]’s Measure: Instrumenting Output and Input Tariffs (1) (2) (3) (4) (5) (6) Variables GL GK ExpTariffsAboveMedian 0.182 0.177 0.238 0.103*** 0.102*** 0.087*** (0.155) (0.155) (0.156) (0.019) (0.019) (0.018) OutTariffsAboveMedian 0.674*** 0.677*** 0.708*** 0.092*** 0.092*** 0.056** (0.254) (0.254) (0.255) (0.029) (0.029) (0.028) InpTariffsAboveMedian 1.148*** 1.146*** 1.312*** 0.338*** 0.338*** 0.260*** (0.207) (0.207) (0.207) (0.024) (0.024) (0.024) TFP 14.965*** 14.975*** 15.039*** 2.508*** 2.509*** 2.684*** (0.083) (0.083) (0.084) (0.013) (0.013) (0.013) lnψ -0.323*** -0.341*** -0.053*** -0.051*** (0.021) (0.021) (0.003) (0.003) lnAge -0.316*** -0.655*** (0.118) (0.018) lnK 1.307*** (0.059) lnEmpl -0.612*** (0.011) Sector-Year y y y y y y Prov-Year y y y y y y FirmFE y y y y y y Obs. 1,084,155 1,084,155 1,084,155 935,474 935,474 935,474 NumberofFirmIDs 373,191 373,191 373,191 334,394 334,394 334,394 GL: labor gap from Petrin and Sivadasan [2013]. GK: capital gap from Petrin and Sivadasan [2013]. Exp Tariffs Above Median: dummy equal to one if firm export tariff is above the 50th percentile within an industry. Out Tariffs Above Median: dummy equal to one if firm output tariff is above the 50th percentile within an industry. Inp Tariffs Above Median: dummy equal to one if firm input tariff is above the 50th percentile within an industry. TFP: total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon-Petrin methodology. lnψ: profit margin. ln Age: log firm age. lnK: log capital. lnEmpl: log employment. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: FE firm-level regressions, 2000-2006. Our sample excludes processing exporters and importers. Standard errors are clustered at the firm level. 35

Table A9: Aggregate Misallocation and TFP, Petrin and Sivadasan [2013]’s measures (1) (2) (3) (4) Variables TFP AvgGL 0.010*** 0.009*** 0.008*** (0.002) (0.002) (0.002) AvgGK 0.040*** 0.014 0.000 (0.012) (0.013) (0.012) Avglnψ 0.079*** (0.028) AvgAge -0.321*** (0.057) Year y y y y SectorFE y y y y Obs. 3,453 3,453 3,453 3,453 R2 0.905 0.901 0.905 0.914 NumberofIndustries 350 350 350 350 TFP: average sector-level total factor productivity, calculated according to the Wooldrige (2009) extension to the Levinshon- Petrin methodology. Avg GL: average dispersion across labor gap (Petrin and Sivadasan, 2013) measure. Avg GK: average dispersion across capital gap (Petrin and Sivadasan, 2013) measure. Avg lnψ: average profit margin within sector. Avg Age: average firm age within sector. Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: Sector-level FE regressions, 1998-2007. Standard errors, clustered at the sector levels, are reported in parenthesis. 36

Table A10: Tariffs and Aggregate Misallocation (1) (2) (3) (4) Labor Capital Variables Avg|lnλ| AvgGL Avg|lnκ| AvgGK ImpTariffs 0.005*** 0.147** 0.001 0.018*** (0.001) (0.070) (0.001) (0.006) ExpTariffs 0.000 -0.054 -0.001 -0.002 (0.001) (0.039) (0.001) (0.002) Controls* y y y y Year y y y y SectorFE y y y y Obs. 2,421 2,421 2,421 2,421 R2 0.162 0.461 0.245 0.548 NumberofIndustries 349 349 349 349 *Controls include TFP, average age, capital stock, employment, and profit margin (sector-level averages). Avg |lnλ|: average dispersion across absolute relative labor returns within sector. Avg GL: average dispersion across labor gap (Petrin and Sivadasan, 2013) measure. Avg |lnκ|: average dispersion across absolute relative capital returns within sector. Avg GK: average dispersion across capital gap (Petrin and Sivadasan, 2013) measure. Imp Tariffs: import tariffs (HS schedules matched to CIC classification). Exp Tariffs: export tariffs (HS schedules matched to CIC classification). Legend: ∗∗∗ significant at 1%, ∗∗ at 5%, ∗ at 10%. Notes: Sector-level FE regressions, 1998-2007. All columns include year dummies and sector-level averages of firm characteristics. Standard errors, clustered at the sector levels, are reported in parenthesis. 37