Trade and Labor Reallocation with Heterogeneous ......3 Allowing for globalization to impact plants...
Transcript of Trade and Labor Reallocation with Heterogeneous ......3 Allowing for globalization to impact plants...
1
TradeandLaborReallocationwithHeterogeneousEnforcementofLaborRegulations
RitaK.Almeida
TheWorldBankandIZA
JenniferP.PooleUniversityofCalifornia,SantaCruz
September2012
PRELIMINARYANDINCOMPLETE
Abstract:Thispaperrevisitsthequestionofhowtradeopennessaffectslabormarketoutcomesinadeveloping country setting.Weexplore the fact thatplants face varyingdegreesof exposure toglobalmarketsand to theenforcementof labormarket regulations, and relyonBrazil’s currencycrisis in 1999 as an exogenous source of variation in industry‐specific real exchange rates andhence, access to foreign markets. Using administrative data on employers matched to theiremployees andon the enforcement of labor regulations at the city level overBrazil’smain crisisperiod,we document that theway trade openness affects labormarket outcomes for plants andworkersdependsonthestringencyofdefactolabormarketregulations.Inparticular,weshowthatafter a trade shockplants facing stricter enforcement of the labor lawdecrease job creation andincrease job destruction by more than plants facing looser enforcement. Consistent with ourpredictions, this effect is strongest among small, labor‐intensive, non‐exporting plants, forwhichlabor regulations are most binding. We also note a stronger impact of enforcement on youngerworkers. Finally, our findings are consistent with the hypothesis that increased regulatoryenforcement limits the plant‐level productivity gains associated with trade openness. The latterimpliesthatincreasingtheflexibilityofdejureregulationswillallowforbroaderaccesstothegainsfromtrade.
Keywords: Globalization; Enforcement; Labor Market Regulations; Linked Employer‐EmployeeData.JEL:F16;J6;J8.
OurspecialthankstoPauloFurtadodeCastroforhelpwiththeRAISandSECEXdataandtotheDepartmentofEconomicsandAcademicComputingServicesattheUniversityofCalifornia,SanDiegoforassistancewithdata access.We also thank theBrazilianMinistry of Labor for providing data on the enforcement of laborregulationsand important informationabouttheprocessofenforcement,especiallyEdgarBrandao,SandraBrandao,andMarceloCampos.Forhelpfulcommentsandsuggestions,wearegratefultoKevinMilliganandPetiaTopalova,aswellasworkshopparticipantsattheFederalReserveBankofSanFrancisco,UCIrvine,andvarious conferences. Financial support from the UCSC Academic Senate Committee on Research isacknowledged.AaronColeandKunLiprovidedexcellentresearchassistance. CorrespondingAuthor: 437Engineering2,DepartmentofEconomics,UniversityofCalifornia,SantaCruz,(831)459‐5397,[email protected].
2
1. Introduction
Akeyargumentinfavorofliberalizingtraderelationsisthatfactorscanreallocatetomoreefficient
uses,allowingforenhancedproductivity,incomegrowth,andwelfare(Pavcnik2002;Feyrer2009;
Broda and Weinstein 2006). Early studies in many developing countries, however, found little
impact of trade liberalization on plant‐level employment andwages (Currie and Harrison 1997;
Feliciano 2001). More recent work offers mixed results on the causal link from exporting to
productivityandevidenceofslowlabormarketadjustmenttotradereform.Apotentialexplanation
forthesevariousfindingsarerestrictivelabormarketregulations,whichinhibitthereallocationof
workers, limiting the extent to which plants can benefit from increased openness (Freund and
Bolaky2008;Kaplan2009).
Inthispaper,werevisitthequestionoftheimpactoftradeliberalizationonlaborreallocationina
developing country by exploring the fact that plants vary in the degree of exposure to global
marketsandthatdefactolaborregulationsareheterogeneouswithincountries.Werelyondetailed
administrative data from Brazil covering the currency’s devaluation episode. Ourmain reduced‐
form specification relates exogenous industry‐specific exchange rate shocks to plant andworker
outcomes over time, differentially for plants located in distinct labor market regulatory
environments. Our findings show that more stringent de facto regulations reinforce the
contractionary labor market effects of trade openness for small, labor‐intensive, non‐exporting
plants.Overall,domesticplantsinstrictly‐enforcedareasincreasejobdestructionanddecreasejob
creationbymorethanotherwiseidenticaldomesticplantsinweakly‐enforcedareas.Fromapolicy
standpoint,ourwork thereforecontributes toanunderstandingof jobsecurity inan increasingly
globalizedworld.Increasingtheflexibilityofdejureregulationswillstimulatejobcreationandoffer
broaderaccesstothegainsfromtrade.
Wecontributetoagrowingbodyofworkinseveralways.First,themicro‐dataavailableforBrazil
arerichandappropriatetostudytheeffectsoftradeliberalizationonlaborturnover.Weexploita
matchedemployer‐employeedatabasecoveringtheformal‐sectorlaborforce,incombinationwith
information on the plant’s exposure to global markets. The data allow us to analyze total
employment at the plant level, and also to trace the movement of workers across different
employers in response to the trade shock. Furthermore, it permits the decomposition of labor
turnover into changes along theextensivemargin (the accessionand separationofworkers) and
alongtheintensivemargin(hoursworkedandtemporarycontracts).Theabilitytomatchworkers
3
totheiremployersisanintegralpartofouranalysis,asrecentevidencepointstoasortingeffectof
globalization.1For instance,as in themodeldescribed inHelpman, Itskhoki,andRedding(2010),
when there are complementaritiesbetweenplantproductivity andworker ability, plantshave an
incentivetoscreenforworkersbelowagivenabilitylevel.Higherproductivityexportersscreentoa
higher ability threshold and will thus have a workforce of higher average ability than non‐
exporters.2Becauseglobalization increasesplant selection intoexportingas inMelitz (2003)and
theincentivestoscreenforhigh‐qualitymatches,itisimportanttoaccountforheterogeneityinthe
quality of the worker‐plant match in determining the effects of globalization on labor market
outcomes(Woodcock2011;Krishna,Poole,andSenses2011).Inoursetting,anotherwiseidentical
workermayhaveahigherprobabilityofseparationfrom(oralowerprobabilityofaccessionto)a
high productivity plant than from (and to) a low productivity plant, when such worker‐plant
production complementarities exist. This diversity offers disparate predictions for the effects of
trade liberalization onworker turnover at exporting (high productivity) and non‐exporting (low
productivity) plants. Our reduced‐form specification builds on the existing literature in this
dimension. Notably, our preferred specification uses information on worker‐level labor market
outcomes,separatelyforexportinganddomesticplants,andincludesplant‐workermatch‐specific
effects to allow for the possibility ofworker sorting and unobservable plant,worker, andmatch
heterogeneity.
Whilewearenot the first to investigate the impactof tradeby theplant’smodeof globalization
(e.g.,AmitiandDavis(2012)),wearenotawareofanypaperallowingforglobalizationto impact
plantsdifferentlydependingontheirexposuretolabormarketregulatoryenforcement.3Brazilhas
oneof themostrestrictive labormarketregulatoryframeworksintheworld(Botero,Djankov,La
1Verhoogen(2008)documentsaskill‐upgradinginMexicanexportingplantsafterthe1994pesodevaluation.Bustos(2011)looksattheBrazilianreductionintariffsaspartoftheMercosurregionalfreetradeagreementandfindsthatArgentineanplantsabovethemediansizeupgradeskills,whileplantsbelowthemediansizedowngradeskills.2Changesintheskillcompositionoftheworkforceatexportersrelativetonon‐exportersarealsopresentinanumberofotherrecenttrademodels(seeforexample,Yeaple(2005)andDavidson,Matusz,andShevchenko(2008)).3 Allowing for globalization to impact plants differently depending on exposure to regulatory enforcementmayhelptoprovideanexplanationforthemixedresultsintheliteratureonthecausalimpactofexportingonplant‐level productivity. Clerides, Lach, and Tybout (1998) argue that the positive relationship betweenexportingandproductivityisaresultoftheself‐selectionofplantsintoexports.Inmostcases,plantsdonotbecomemoreproductivebyexporting.Bycontrast,usingplant‐leveldatafromSlovenia,DeLoecker(2007),controllingforself‐selection,findsevidencethatexportersbecomemoreproductive.
4
Porta,andLopezdeSilanes2004;AlmeidaandCarneiro2012).4However,thesizeoftheinformal
labor force suggests that enforcement isweak in some areas, hinting at a gap between the laws
statedonthebooks(dejureregulations)andtheireffectiveimplementation(defactoregulations).
Therefore, contrary to previous studies which rely on cross‐country or across‐state variation in
existing de jure labor regulations (e.g., Besley and Burgess (2004) and Autor, Kerr, and Kugler
(2007)),we explore the fact thatBrazilian employers are exposed to varyingdegrees ofde facto
labor regulations, via Ministry of Labor inspections. Especially in a developing country context
where enforcement is not homogeneous, we argue exploring time series and within‐country
variationinregulatoryenforcementoffersabettermeasureofaplant’strueflexibilityinadjusting
labortoshocksthanlookingatvariationsindejureregulations.5Wethusinvestigatethedifferential
impactof globalizationonworker turnover amongotherwise identicalplantsandworkers facing
differentdefactoenforcementofthelaborlaw.6
Finally, incontrast tomostof the literature investigating the impactof trade liberalizationon the
real economy using potentially endogenous tariff changes7, we explore the Brazilian currency’s
strong devaluation in January 1999 as a large and unanticipated exogenous shock to both
employersandworkers.8FollowingGoldberg(2004),weconstructtrade‐weightedindustry‐specific
real exchange rates in order to capture changes in industry competitiveness over time. The
economy‐widerealexchangeratedepreciated32%from1996to2001,witha23%dropoccurring
between December 1998 and January 1999 alone (see Figure 3.1; Muendler 2003). However,
though all industries suffered exchange rate declines over this time period, some enduredmore
severe shocks than others, as measured by trade‐weighted real exchange rates. We rely on this 4There isanextensive literature fordevelopingcountriesanalyzing therelationshipbetween labormarketregulations and labor market outcomes (e.g., Kugler (1999), Kugler and Kugler (2009), Ahsan and Pages(2009),PetrinandSivadasan(2006),andseveralotherstudiescitedinHeckmanandPages(2004)).5Todate, fewpapershaveexploredbothwithin‐countryandtimeseriesmeasuresofenforcement.NotableexceptionsincludeCaballero,Cowan,Engel,andMicco(2004)andAlmeidaandCarneiro(2012).6 Currie and Harrison (1997) rule out labor market regulations as an explanation for their insignificantfinding of the effects of trade reform on employment levels, and suggest that despite formal labormarketbarriers there is littleenforcementwhich leavesregulations ineffective.UnlikeCurrieandHarrison(1997),our city‐level data on Ministry of Labor inspections allow us to capture exactly this variation in within‐countrycompliancewithlabormarketregulations.7Politicaleconomyfactorsintariffformationandadjustmenthavebeennotedbyanumberofauthors.See,forexample,OlarreagaandSoloaga(1998)forthecaseofBrazil’sregionalfreetradearea,Mercosur.Infact,asprotectionistpressuresgrewintheaftermathoftheintroductionofanewcurrencyin1994,averagetariffsmarginallyincreasebeginningin1995.SeeSection2.2forfurtherdiscussion.8 Other papers using currency shocks as exogenous sources of variation to investigate international traderelationshipsincludeVerhoogen(2008),whousesMexico’s1994pesodevaluationtoexploretherelationshipbetweentradeandinequality,andBrambilla,Lederman,andPorto(forthcoming),whouseBrazil’scurrencycrisisasashocktoArgentineanexporters.
5
industry‐levelvariationinrealexchangeratesovertimetoexogenouslyidentifytheeffectofBrazil’s
increasedglobalizationonemploymentandlaborturnoverattheplantandworkerlevel.
Tosummarize,weareinterestedinanalyzingtheeffectoftradeliberalizationonlaborreallocation
withinBrazil.Weexploreacrossindustryandovertimevariationinrealexchangeratesinorderto
capture changes in industry competitiveness, in combinationwith city and time variation in the
enforcementoflabormarketregulations,facingexportersandnon‐exporters.Animportantconcern
relatestotheexogeneityofthevariationintheenforcementoflaborregulationsacrosscities.Atany
givenpoint in time,enforcementof laborregulationsat thecity level isnot likely toberandomly
distributed.On theonehand,enforcementmaybestronger incitieswithhigherviolationsof the
law.Ontheotherhand,citieswithbetterinstitutionscouldhavestricterenforcement.Althoughitis
unclear how these patterns may impact worker reallocation, a potential bias may still exist. To
minimize this concern, we note that our empiricalmethodology follows the program evaluation
literature and relates exogenous real exchange rate changes to plant andworker outcomes over
time,differentially forplants located invariable labormarketregulatoryenvironments.Ourmain
coefficient of interest is the differential effect of changing enforcement on plants that, all else
constant, are exposed to different exogenous industry‐specific trade shocks. Therefore, our
reduced‐form specification relates annual changes in theprobability of inspection in a given city
(whichweproxyforusingtotallaborinspectionsper100plants)andannualchangesinindustry‐
specific real exchange rates, with annual changes in labor market outcomes. We run this
specificationseparatelyforexportingandnon‐exportingplants.
Onecouldstillquestiontheexogeneityofchangesinenforcementatthecitylevel.Totheextentthat
these changes correlatewith changes in labormarket outcomes, our estimates for the effects of
enforcementmaybebiased.Weemphasize,however,thatourfocusisontheinteractionbetween
exogenouschangesinindustry‐specificrealexchangeratesandchangesinthedegreeofregulatory
enforcement, as is customary in the program evaluation literature. For our main coefficient of
interest tobebiased, itmustbe thatplants in industriesexposed togreaterdepreciationsand in
citiesexposedtogreaterdefactoenforcementalsohavesystematicallydifferentlaborturnover,for
someunobservedreasons.Onepossibilityisthatindustriesareregionally‐concentrated,suchthat
theindustriesexperiencingthemostseveredepreciationsarelocatedinthecitiesexperiencingthe
greatest increases in enforcement (i.e., growing cities that may also have more dynamic labor
markets).Ourreduced‐formestimationincludesstate‐specificyeardummies,whichwearguehelps
6
to correct for some of this bias. In ongoingwork,we also include city‐specific year dummies in
ordertocontrolforthepossibilitythatdifferencesinlaborflowsovertimeincertaincitiesmaybe
drivingthedifferencesinlaborturnoverwefind.
Webeginouranalysisattheplant‐level, investigatingthedifferentialimpactoftradeopennesson
plantsizeforplantslocatedinheavily‐inspectedcitiesrelativetoplantslocatedinweakly‐inspected
cities.Wethenconsideramoredisaggregatedworker‐levelanalysis.Thisallowsustodecompose
how plants adjust labor in response to currency devaluations and how enforcement influences
these adjustments—along the extensivemargin (hiring and firing) or along the intensivemargin
(changesinhoursworkedorbetweenfull‐timeandtemporarycontracts).Meanwhile,ourworker‐
levelanalysishelpstoaddressthepossibilityofworkersortinginthelaborreallocationprocess.
We now briefly discuss the main empirical predictions in our reduced‐form model. With a
devaluation of the Brazilian currency (the real), imports into Brazil become more expensive,
improving the competitiveness and enhancing the profitability of Brazilian plants selling in the
domestic market. To the extent that profits and employment growth are correlated at the plant
level, a currency depreciation is expected to increase employment for the average plant in the
country.9Thesamedepreciationdifferentiallyimprovesconditionsforexporters,asBrazil’strading
partnersneedfewercurrencyunits topurchaseBraziliangoods.Weexpectthisenhancedforeign
marketaccess todifferentially increaseemploymentatBrazil’sexportingplants relative toplants
producingonlyforthedomesticmarket.10
Ourhypothesisrelieson theextent towhich laborregulations “bite”; that is,whetherregulations
are enforced. An increase in the enforcement of labor market regulations through more labor
inspectionsisexpectedtodirectlyimpactthecompliancewithlaborregulationsthroughthehiring
andfiringof formal labor.11However,thedirectionoftheeffectofenforcementonemploymentis
9 Revenga (1992) uses the sharp appreciation of the U.S. dollar during the early 1980s to demonstratesignificantemploymentreductionsforimport‐competingindustries.Ribeiro,etal(2004)considerthecaseofBrazil,documentingtheimportanceoftheexchangerateforjobcreation.Interestingly,BurgessandKnetter(1998)evaluateemploymentresponsestoexchangerateshocksattheindustry‐levelacrosstheG‐7countries,andargue that country‐leveldifferences in the response to theexchange rate shockmaybeattributable tovariationinlabormarketregulations.10 Goldberg and Tracy (2003) demonstrate that the employment declines associated with a U.S. dollarappreciationgrowstrongerasindustriesincreaseinexportorientation.11CardosoandLage(2007)showthatinspectionsareprimarilylinkedtostricterenforcementofmandatoryseverancepayments,mandatedhealthandsafetyregulations,andtotheworker’sformalregistration.Bertola,Boeri,andCazes(2000)suggestthatdifferencesinenforcementacrosscountries,relatedforexampletothe
7
ambiguous. On the one hand, stricter enforcement of labor regulations raises the cost of formal
workers. As such, plants facing stricter enforcement will have increased difficulties in adjusting
labor.Ontheotherhand,thestricterenforcementoflaborregulationsalsoincreasesjobquality,in
termsofcompliancewithmandatedbenefitsfortheworker.Forthisreason,wemayfindincreases
inemployment inmoreheavily‐enforcedcities, as formalemploymentbecomesamoreattractive
optionandformalworkregistrationincreases.
In this paper,we focus on the differential impact of openness on labormarket outcomes across
plants located in cities with varying degrees of regulatory enforcement. Our main results are
consistentwiththeviewthattheextenttowhichtradeaffectslabormarketoutcomesdependson
thedefactodegreeofstringencyofthelaborregulationsfacedbyplants.Inparticular,plantsfacing
stricter enforcement of the labor laws increase employment by less than plants facing fewer
inspections with the expansionary trade shock. Moreover, conditional on the (unobservable and
time‐invariant) plant‐worker match, and several time‐varying worker, plant, and sector
characteristics,wenotethatopennessisassociatedwithanincreaseintheprobabilityofhiringat
exportersandadecreaseintheprobabilityofhiringatdomestically‐orientedplants,asispredicted
bynewheterogeneousfirmtrademodels.Theresultssuggestthatenforcementinfluencesmainly
non‐exporting plants along the extensive margin, but has little effect on adjustment along the
intensivemargin(ascapturedbyhoursworkedandworkerswithfull‐timecontracts).Wenotealso
that our findings aremainly concentrated among small, labor‐intensive, non‐exportingplants for
whichlaborregulationsarelikelytobemostrestrictive.
Themagnitudesofourestimatesseemtobeplausible.Evaluatingtheeffectonworkersandplants
located in municipalities at themean level of inspections, a 10 percent depreciation of the real
increases job match creation at exporting plants by 4.1% and decreases job creation at non‐
exporters by 7.3%. Meanwhile, the impact for domestic plants varies depending on the level of
enforcementtheplantfaces—fordomesticplants locatedinmunicipalitiestthe10thpercentileof
inspections, a10percentdepreciationof the real decreases theprobabilityofhirebyonly6.0%,
efficiencyofacountry’slegalsystem,areasorevenmoreimportant,thandifferencesindejureregulations.Forexample,Caballero,Cowan,Engel,andMicco(2010)exploreapanelof60countriesaroundtheworldandfindthatlaborregulationshaveadverseeffectsonjobturnoverandplants’speedofadjustmenttoshocks,butonlyincountrieswithastrongruleoflawandgovernmentefficiency(takenasmeasuresofenforcementofregulations). However, as with many cross‐country studies, the limited time series variation in laborregulationsandmeasuresofenforcementposeschallengesforidentification.
8
while workers matched with domestic plants located in municipalities in the 90th percentile of
inspectionsexperienceadecreaseinjobmatchcreationofaround9.0%.
Our results strongly suggest that more stringent de facto regulations limit job creation with
enhanced trade openness. Overall, small, labor‐intensive, non‐exporters separate from more
workers and hire fewerwith increased enforcement of regulations.We show that this increased
enforcementof laborregulations isalsoassociatedwith lowerplant‐levelproductivity,asproxied
byplant‐levelaveragewages.Itseemsclearfromourdatathatstrictlabormarketinstitutionslimit
thepossibility forplant‐level productivity andprofitability gains associatedwith tradeopenness.
Fromapolicystandpoint,ourworkthereforecontributestoanunderstandingofjobsecurityinan
increasingly globalized world. Increasing the flexibility of de jure regulations will offer broader
accesstothegainsfromtradeandenhancedjobcreation.
Inadditiontotheworkcitedabove,ourpaperrelateswithanumberofdifferentliteratures.First,
ourresearchiscloselylinkedtoagrowingbodyofstructuralmodelslinkingtradeandlabormarket
policies, such as firing costs. In the model presented by Cosar, Guner, and Tybout (2011), tariff
liberalizationsincreasefirm‐leveljobturnover,andreductionsinfiringcostsreinforcetheimpactof
globalization further increasing job turnover. Kambourov (2009) presents a model in which
liberalizingtradeinarestrictedlabormarketenvironmentisassociatedwithslowerinter‐sectoral
labormarketreallocation, loweroutput,andreducedproductivity.Weseethesestructuralpapers
ascomplementarytoourreduced‐formframeworkdesignedtoidentifythecausal implicationsof
tradeopennessonlaborreallocationinthepresenceofacompletesetoflabormarketregulations.
Second, our research is related to a set of empirical papers on productmarket liberalizations in
different labor market environments. Aghion, Burgess, Redding, and Zilibotti (2008) show that
India’s deregulation of the License Raj (control over entry and production in the manufacturing
industry) led todifferential ratesofgrowthacross industries located instateswithpro‐employer
labormarketinstitutionsrelativetoindustrieslocatedinstateswithpro‐workerlaborinstitutions.
Similarly,Hasan,Mitra,andRamaswamy(2007)alsodistinguishIndia’sstatesbytheextentoflabor
market restrictions, and analyze the impact of India’s 1991 trade reform on labor demand. The
authorsfindsupportiveevidencefortheinteractionoftradereformandlaborregulations;thatis,
theimpactoftradereformonlabordemandislargerinstateswithmoreflexiblelaborinstitutions.
Using the same data, Topalova (2010) demonstrates that India’s trade liberalization negatively
9
impacted poverty and per capita expenditures predominantly in states with less flexible labor
markets. Also relevant to our study is Freund andBolaky (2008)who argue that trade can only
improvelivingstandardsinflexibleeconomies.Inparticular,theirfindingsonhiringandfiringcosts
suggest that the positive effects of openness are reducedwhen labor regulations are excessive.12
Thebenefitoflinkedemployer‐employeedataallowsustomovebeyondtheindustrylevelandstate
level, to compute individual‐level accessions and separations, as well as to incorporate plant,
individual,andmatchheterogeneity.Moreover,aswepreviouslymention,exploitingvariationinde
factolaborregulationsoffersamorecompletemeasureoflabormarketflexibilitythanvariationin
dejurelaborregulationsalone.
Thepaperproceedsasfollows.Inthenextsection,weprovidebackgroundinformationonthe1988
Constitutionalreform,whichestablishedthecurrentlabormarketregulatoryframeworkinBrazil,
therecentevolutionintheenforcementoftheselaborlawsconductedbytheMinistryofLabor,and
themainfeaturesofBrazil’srecentglobalization.InSection3,weoutlineourmaindatasourcesand
offersomesimpledescriptivestatistics.Section4discussestheconceptualframeworkbehindour
mainempiricalstrategyandproposesasimpledifference‐in‐differencereduced‐formspecification
fortheempiricalwork.InSection5,wepresentourmainfindings.Section5alsooffersevidenceon
theheterogeneityofourmainfindings,consideringtheindustry’stechnologicalintensity,plantsize,
and theageof theworker,andaggregate implicationsof theenforcementof labor regulationson
plant‐levelproductivity.WeconcludeinSection6byhighlightingthemainpolicyimplications.
2. PolicyBackground
Thelate1980stotheearly2000smarkedaperiodofsubstantialmarket‐orientedreforminBrazil.
Of particular relevance to ourwork are the establishment of a new Constitution in 1988,which
offered increased employment protections for workers, the liberalizing trade policy reforms
beginningin1987,andtheimplementationofanewcurrency,thereal, in1994.Subsequently,the
currencyexperiencedasevereandunanticipateddevaluationinJanuary1999.Eachofthesepolicy
changesmayhavecontributedtochanginglaborcostsandlaborreallocation.
2.1. LabormarketregulationswithinBrazil
12Eslava,Haltiwanger,Kugler,andKugler(2010)considerthecaseofColombia’spro‐marketreformsofthe1990s. The authors find that allowing for frictionless factor adjustment would lead to substantialimprovementsinefficiencyoverthereformperiod.
10
The1988Constitutionalreform The Brazilian Federal Constitution of 1988 imposed high
labor costs to plants and was very favorable to workers. First, it reduced themaximumweekly
workingperiodfrom48to44hours.Second,itincreasedtheovertimewagepremiumfrom20%to
50%oftheregularwage.Third,themaximumnumberofhoursforacontinuousworkshiftdropped
from8to6hours.Fourth,maternity leave increasedfrom3to4months.Finally, it increasedone
month’svacationtimepayfrom1to4/3ofamonthlywage.
Followingthe1988changes in the laborcode, thecostof labor toemployers increased.First, the
employer’spayrollcontribution increasedfrom18%to20%.Second, thepenaltyontheplant for
dismissingtheworkerwithoutcauseincreasedfrom10%to40%ofthetotalcontributionstothe
severancefund,FundodeGarantiadoTempodeServiço(FGTS).13EmployersinBrazilmustalsogive
advancenoticetoworkersinordertoterminateemployment.Duringthisinterimperiod,workers
aregranteduptotwohoursperday(25%ofaregularworkingday)tosearchforanewjob.14
Enforcementoflaborregulations Thesede jure labor regulationsare effective throughout the
country. However, as theMinistry of Labor is chargedwith enforcing compliancewith the labor
regulations, there issignificantheterogeneitybothwithincountryandover time in termsofhow
bindingisthelaw.15Giventhegeographicscopeofthecountry,enforcementisfirstdecentralizedto
thestate levelwith themain laboroffices(delegacias) located in thestatecapital.Enforcement is
thenfurtherdecentralizedtothelocallevelwithineachstate,dependingonthesizeofthestate.For
example, in 2001 the state of Sao Paulo had 24 local labor offices (subdelegacias)while smaller
stateslikeAcrehadonlytheoneofficecoincidingwiththedelegaciainthestatecapital.
Throughout the late1990sandearly2000s, labor inspectionsbecamemorefrequentasthe large
publicdeficit led theBraziliangovernment tosearch foralternativeways tocollect taxrevenue.16
13Iftheworkerisdismissedwithoutjustification(withtheexceptionofworkersonaprobationaryperiod),theplantisfinedandhastopaytheworker40%oftheFGTScontributions.14Someplantsvoluntarilychoosetograntworkersthefullmonthlywagewithoutrequiringwork.BarrosandCorseuil(2004)findthattherearelargeproductivitylossesduringthisperiod.15AcomprehensiveexplanationoftheenforcementofthelaborregulationsystemanditsimportanceinBrazilisgiveninCardosoandLage(2007).16Aninspectioncanbetriggeredeitherbyarandomplantaudit,orbyareport(oftenanonymous)ofnon‐compliancewiththelaw.Workers,unions,thepublicprosecutor’soffice,oreventhepolicecanmakereports.Inpractice,almostallofthetargetedplantsareformalplantsbecauseitisdifficulttovisitaplantthatisnotregistered,sincetherearenorecordsofitsactivity.
11
Wenotetimevariationinthelocationoflocallaborofficesasnewsubdelegaciasopenovertime.For
instance, in 1996 in addition to the delegacia in the state capital, the state of Bahia had 6
subdelegacias. By 2001, Bahia had 8 subdelegacias. Because of this, the average distance to the
nearestMinistryofLaborofficedecreasedbyabout5%between1996and2001. Inaddition, the
averagenumberofinspectionsinthemanufacturingsectorpermunicipalityincreasedfrom13.2in
1996 to 14.3 in 2001.As inspectors reachedoutwith increased intensity, themediannumber of
inspections also increased, suggesting a leftward shift of the distribution of inspections across
municipalities.
Most of the inspections and subsequent fines for infractions in Brazil are to ensure plants’
compliance with the worker's formal registration in the Ministry of Labor, contributions to the
severancepayfund(FGTS),compliancewiththeminimumwage,andwiththemaximumworking
lengths. Evasion of one of these dimensions accounted formore than 40% of all fines issued in
2001.Themonetaryamountofthefinesiseconomicallysignificantandmaybeissuedperworker
oritmaybeindexedtotheplant’ssize.Forexample,in2001values,aplantisfined216reais(or
approximatelyUSD100)foreachworkerwithoutacarteiradetrabalho,formalworkauthorization.
Considering that, at 2001prices, the federalminimumwagewas222 reais, non‐compliancewith
worker registration is non‐trivial, implying a penalty of approximately one monthly wage per
worker.
Plants weigh the costs and benefits of complying with this strict labor regulation. They decide
whethertohireformally,informally,orformallybutwithoutfullycomplyingwithspecificfeatures
of the labor code (e.g., avoiding the provision of specificmandated benefits, such as health and
safetyconditions,oravoidingpaymentstosocialsecurity).Theexpectedcostofevadingthelawisa
functionofthemonetaryvalueofthepenalties(finesandlossofreputation)andoftheprobability
ofbeingcaught.Inturn,theprobabilityofbeingcaughtdependsontheplant’scharacteristics(such
assize,globalizationstatus,andlegalstatus)andonthedegreeofenforcementofregulationinthe
citywheretheplantislocated.17
2.2. Brazil’sglobalization
17Asinspectorsfaceaperformance‐basedpayscheme,theyoftenlookforcaseswherethepenaltyislikelytobelarge.Assuch,thereisastrongcorrelationbetweenthesizeofthefirm,asaproxyforthevisibilityofthefirm,andthenumberofinspections(CardosoandLage2007).
12
Policyreforms Thesecondhalfof the20thcentury inBrazilwascharacterizedby tight import
substitutionindustrializationpoliciesdesignedtoprotectthedomesticmanufacturingsectorfrom
foreigncompetition.Beyondhightariffrates,substantialnon‐tariffbarrierscharacterizedBrazilian
tradepolicyduring this timeperiod.The latterhalfof the1980sand thebeginningof the1990s,
however,witnessedsweepingchangesinBraziliantradepolicy.Thisoccurredintwophases.First,
averageadvaloremfinalgoodstariffratesfellfrom58%in1987to32%in1989.Thesereformshad
little impact on import competition however, as non‐tariff barriers remained highly restrictive.
Second,between1990and1993,thefederalgovernmentabolishedallremainingnon‐tariffbarriers
inheritedfromtheimportsubstitutioneraandannouncedascheduleforthereductionofnominal
tariffsoverthenextfouryears(MoreiraandCorrea1998).Effectiveratesofprotectionfellbyover
70%injustfouryears—fromapproximately48%,onaverage,in1990to14%,onaverage,in1994
(Kume,Piani,andSouza2003).
In1994,afterdecadesofhighinflationandseveralunsuccessfulstabilizationattempts,theBrazilian
government succeeded with a macroeconomic stabilization plan (Plano Real), designed to help
correctalargefiscaldeficitandlastinglyendhyperinflation.Thenewcurrency,thereal,waspegged
totheU.S.dollar,andbeganatparityonJuly1,1994.Officially, therealwasset toacrawlingpeg
whichpermittedthecurrencytodepreciateatacontrolledrateagainsttheU.S.dollar.However,as
the country’s persistent effort to control inflation paid off, the real exchange rate actually
appreciated in its first months. In response, the government partially reversed trade reforms in
1995aftermanufacturingindustrieslostcompetitivenessduetothereal’sappreciation.18
Currencycrisis Despite efforts to control public spending and raise tax revenues, Brazil’s
fiscaldeficitsremainedhighandcontinuedtogrow.Meanwhile,persistentcurrentaccountdeficits
placedsignificantpressureonthepeggedexchangerateandgovernmentreserves,leadinginvestors
towithdraw funds from Brazil. In response to the financial crises in Asia in 1997 and Russia in
1998,theauthoritiesraisedinterestratestoencouragedomesticsavingsandinvestment.However,
as debt service obligations increased, investor panic persisted. Dollar reserves fell from
approximately $58 billion in 1996 to $43 billion in 1998. Inmid‐January 1999, capital outflows
acceleratedfurtherwhentheGovernoroftheStateofMinasGeraisdeclaredamoratoriumonthe
state’s debt payments to thenational government, triggering the government’s announcement of
18Averageadvaloremtariffsclimbedslightlyinsubsequentyears—fromanaverageof12.2%in1994toanaverageof14.4%in2001.
13
the end of the crawling peg, allowing the real to float against the U.S. dollar (Gruben and Kiser
1999).Overnight,thenominalexchangeratedevaluedby9%againsttheU.S.dollar,andbytheend
ofthemonth,therealhaddepreciatedby25%(seeFigure3.1).
3. Data
Our main data are administrative records from Brazil for formal sector workers linked to their
employers. We match these data by the plant’s municipality with city‐level information on the
enforcementoflabormarketregulations,bytheplant’ssectorwithinformationonindustry‐specific
real exchange rates, and by the employer tax identifier to information on exposure to global
markets.ThesampleperiodforanalysiscoversBrazil’smaincurrencycrisisperiod,between1996
and2001.Thisexogenousshocktoplantandworkeroutcomesallowsustouncoverthedifferential
impact of increased exposure to trade on labor reallocation depending on the degree to which
plantsfaceregulatoryenforcement.
Matchedemployer‐employeeadministrativedata Weusedata collectedby theBrazilianLabor
Ministry,whichrequiresbylawthatallregisteredestablishmentsreportontheirformalworkforce
ineachyear.19ThisinformationhasbeencollectedintheadministrativerecordsRelaçãoAnualde
InformaçõesSociais (RAIS) since 1986. For our analysis, however,weuse data fromRAIS for the
years 1996 through 2001, when we also have complementary information on regulatory
enforcement,exchangerates,andtheemployer’sglobalizationstatus.
The main benefit of the RAIS database is that both plants and workers are uniquely identified
allowing us to trace workers over time and across different plants. The data also include the
industryandmunicipalityofeachplant.20Otherrelevantvariablesofinterestincludetheworker’s
monthofaccessiontoandthemonthofseparationfromthejob,weeklyhoursworked,andthetype
19Our analysis is restricted to the effect of trade liberalizationon formal labor reallocation. It is plausible,however, that in citieswithweaker enforcementandhencemore flexible laboradjustment, alongboth theformal and informal margin, plants may be more likely to make adjustments along the informal margin.Therefore, our findings on formal labor adjustment do not capture total labor adjustment. It is not clear,however, how the resultswould change.GoldbergandPavcnik (2003), Paz (2012), andMenezes‐Filho andMuendler(2011)findmixedresultsontheimpactoftradeliberalizationontheinformalsectorinBrazil.Also,totheextentthatenforcementincreasesthecostof labor,wemaynoteshiftsawayfromlabor(bothformalandinformal)towardscapital.20The industrial classification available inRAIS is the4‐digitNationalClassificationofEconomicActivities(CNAE).
14
of employment contract (temporary versus permanent), as well as detailed information on the
worker’shumancapital, includingheroccupation,education, tenureat theplant,gender,andage.
WedefineaworkerashiredtoemployerjduringyeartifRAISreportsanon‐missingvalueforthe
monthofaccession.Wedefineaworkerasfiredfromemployerjduringyeart ifRAISreportsthe
workernolongeremployedatfirmjonDecember31ofyeart.
WerestrictobservationsinRAISasfollows.Wedrawa1%randomsampleofthecompletelistof
workers ever to appear in the national records and retrieve their complete formal sector
employmenthistory.Weincludeonlymanufacturingsector21workersbetweentheagesof15and
64years,withapositiveaveragemonthlywage,andemployedinprivate‐sector jobs.Toensurea
more precise identification of worker andmatch fixed effects, we further restrict the sample to
thoseworkerswhoareinthedataforatleasttwoyears.
Enforcementdata We also explore administrative city‐level data on the enforcementof labor
regulations,alsocollectedbytheBrazilianMinistryofLabor.Dataforthenumberofinspectorvisits
are available by city and 1‐digit sector for the years 1996, 1998, 2000, and 2002. We use the
informationonvisitsby inspectors tomanufacturingplantsonly.Forouranalysis,we interpolate
averagevaluesforthemissingyearsandmatchtheenforcementdatatotheRAISdatabytheplant’s
municipal location. This information identifies plants and workers facing varying degrees of
regulatoryenforcement.
Weproxy thedegreeof regulatoryenforcementwith the intensityof labor inspectionsat thecity
level.Inparticular,ourmainmeasureofenforcement,designedtocapturetheprobabilityofavisit
bylaborinspectorstoplantswithinacity,isthelogarithmofthenumberoflaborinspectionsatthe
city level (plusone)per100plants in thecitybasedonRAIS.Thisscaledmeasureof inspections
helpstocontrolforimportantdifferencesacrosscities.Moreover,theimpactofsuchameasurewill
reflectthedirecteffectofinspections,aswellasplants’perceivedthreatofinspections(eveninthe
absenceofplant‐levelinspections)basedoninspectionsatneighboringplants.
Table3.1 reports thenationwide increase inenforcementof labor regulationsbetween1996and
2001.Theproportionofcitieswithatleast1manufacturinginspectionrosefrom33%in1996to
21ThemanufacturingsectorconsistsofCNAE2‐digitcodes15‐37. In futuredraftsofthepaper,wehopetoincorporateothersectorsforacomparativeanalysis,aswell.
15
52%in2001.Thiscorrespondstoan increase in theaveragenumberof inspectionsacrosscities,
mostnotablybetween1998and2000,whentheaveragenumberofinspectionsincreasedby15%.
Asthenumberofinspectionsatthecityleveliscorrelatedwiththesizeofthecity(i.e.,population,
labor force, andnumberofplants), ourpreferredmeasuredocuments thenumberof inspections
per100registeredplantsasisreportedincolumn(3).Thedatareportincreasesinthenumberof
inspections per plant and per worker, as inspectors intensify the enforcement process to reach
additionalplants,workers,andcities.22
We also note significant heterogeneity in the within‐country variation in the intensity of
enforcementacrosscitiesinBrazil,asisdepictedinFigure3.2.Theleftpanelillustratesthenumber
ofinspectionsperBraziliancityin1998,withdarkershadesportrayinghighernumbers.Theright
paneldepictsthesamestatistictwoyearslaterintheyear2000.Weremarkonthevariationacross
municipalities and over time. First,we observe the darkest areas of themap in the high‐income
SouthernandSoutheasternregionsofthecountry.Wealsonoticeadarkeningofthemapbetween
1998and2000asenforcementspreadstofurtherpartsofthecountry.23Intheanalysisthatfollows,
weincludeinteractivestate‐yeardummiesintoourdifference‐in‐differenceapproachtoaccountfor
thespatialpatternofinspections.
Industry‐specificexchangerates Weconstructtrade‐weightedindustry‐specificrealexchange
ratesbasedonbilateralrealexchangeratedatafromtheInternationalMonetaryFundandbilateral
tradeflowsbycommoditymadeavailablebytheNationalBureauofEconomicResearch(Feenstra,
Lipsey,Deng,Ma,andMo2004).24Wematchtheindustry‐specificrealexchangeratestotheRAIS
data by the plant’s 4‐digit industrial classification, in order to identify plants and workers in
industrieswithdifferentialglobalizationexperiences.25
22Indoingso,thedistributionofinspectionsacrosscitiesshiftsleft.Mostnotably,thecityofSaoPauloreportsthemaximumnumberofinspectionsineveryyear,aswouldbeexpectedgiventhecity’slargepopulationandlabor force size. However, Sao Paulo city also sees its number of inspections reduced by a third over thecourseofthefiveyears.23 The same qualitative patterns hold for the number of inspections and the number of inspections perworker.24Trade flowsareorganizedbyStandardIndustrialTradeClassification(SITC)codes.Wematchthe4‐digitSITC revision 2 codes to the 4‐digit CNAE codes available in RAIS using publicly available concordances(http://www.econ.ucsd.edu/muendler/html/brazil.html#brazsec).25AswediscussinSection2.2,averagetariffrateswererelativelyflatoveroursampleperiod.Variationsintherealexchangerate,therefore,provideamorerealisticmeasureofchangesintradeopennessduringoursampleperiod.
16
Aswaspreviouslynoted,Brazil’saggregaterealexchangeratedevaluedinJanuary1999,increasing
therelativepriceofBrazilianimports.However,theaggregateexchangeratemaybelesseffectiveat
capturing true changes in industry competitiveness, induced by changes in specific bilateral
exchangerates,ifparticulartradingpartnersareofparticularimportanceforparticularindustries.
Thatis,movementsinthedollar/real,peso/real,andeuro/realexchangeratesmayhavedifferent
implicationsfordifferentindustries,dependingontheindustry’stradewiththeU.S.,Argentina,and
Europe, respectively. Therefore, following Goldberg (2004), we calculate the trade‐weighted real
exchangerateasfollows:
. 5 ∗∑
.5 ∗∑
∗
wheretindexestime,kindexesindustry,andcindexescountry,suchthatthebilateralrealexchange
rate, ,denotedintermsofforeigncurrencyunitsperreal,isweightedbyindustry‐specificand
time‐varyingexportshares(∑
)andimportshares(∑
).26
A decrease in the value of this index implies a real depreciation of the Brazilian real in trade‐
weightedtermsforindustryk.Acrossallindustries,theaverageindexdecreasedfrom0.97in1996
to0.62in2001,withthemostdramaticdropofroughly30%occurringbetween1998and1999.As
Figure3.3illustratesintheleftandrightpanels,respectively,thereisalsosubstantialheterogeneity
acrossindustriesinboththelevelofandannualchangesintherealexchangerate.Thoughthemean
exchangerateisvaluedat0.66in1999intheaftermathofthecrisis,themanufactureofotherfood
products has a substantially lower trade‐weighted real exchange rate at 0.54, while the trade‐
weightedrealexchangerateintheindustrywhichmanufacturesstrings,cables,andothercordsis
far higher at 0.80. Similarly,while all sectors experienced sharp exchange rate declines between
1998 and 1999, some sufferedmore than others. Non‐ferrousmetalmanufacturing endured the
steepestannualdepreciationof40percentagepoints,whilesugarmanufacturingfacedamere16
percentagepointdecline.
Exposuretoglobalmarkets We also investigate information on the firm’s degree of global
engagement,ascapturedbytotalexportsales.WerelyoncomplementarydatafromtheBrazilian 26FollowingCampaandGoldberg(2001),welagthetradesharesoneperiodtoavoidissuesofendogeneitybetweentradeandtheexchangerate.
17
CustomsOffice(SECEX)tocreateasingleindicatorforthefirm’sglobalizationstatus.Information
onfirm‐levelexporttransactionsisavailablefromSECEX,whichrecordsalllegally‐registeredfirms
inBrazilwithatleastoneexporttransactioninagivenyear.Wedenoteexporterstobethosefirms
thatexportedanypositivedollaramountat anypointduring the1996 to2001 timeperiod.This
time‐invariant indicator isdesigned tominimizepotential endogeneity concerns surrounding the
exportdecisionpost‐devaluation.
3.1. Descriptivestatistics
We report detailed descriptive statistics in Table 3.2. Column (1) reports statistics for our final
sampleofformal‐sectormanufacturingworkers,aswellastheplants,cities,andindustriesinwhich
theywork.Column(2)reportssummarystatisticsforthesampleofexportingplants,whilecolumn
(3) reports statistics for domestic plants. The final sample has 313,940 worker‐plant‐year
observations,with81,254workersemployed in53,011plantsallowing for122,017worker‐plant
matches,covering2,769municipalitiesand2404‐digitindustriesthroughoutthesampleperiodof
1996to2001.
Approximately29%ofmanufacturingworkerswerehired to anewemployerduringour sample
period,while25%wereseparatedfromtheiremployer.Thedatareportthatonly2%ofworkersare
employedwithtemporarycontracts.Acrossemployers,workersaveraged43.5hoursperweek.The
averageageofaworkeris32years.Themajorityofthemanufacturinglaborforcehaslessthana
highschooleducation,whileabout25%haveatleastahighschooleducation,andonly7%havea
tertiary education. Roughly a third of themanufacturing labor force is employed in skilled blue
collarprofessions,suchasmachineoperators.Another25%areinwhitecollarprofessions—7%in
secretarial and sales positions and 17% in professional,managerial, and technical positions. Ten
percent of the formalmanufacturing sector is employed in unskilled blue collar jobs,most often
foundintheconstructionandservicesectors.Theaverageplantemploys59workers,andpaysan
averageannualwageof2,970reais.Theaveragedurationofaworker‐plantmatchinthesampleis2
years and 9months. The averagemunicipality in our sample faces approximately 35 inspections
duringthe1996to2001sampleperiod.27Theaverageindustryhasatrade‐weightedrealexchange
rateindexof0.62andemploysover18,000workers,ofwhichabout1inevery5areunionized.
27ThisnumberstandsincontrasttotheaveragenumberofinspectionsacrossallBrazilianmunicipalities(seeTable3.1).Our1%randomsamplecoversonlyregisteredfirmswhichare,onaverage,larger.Therefore,this
18
Ofthe53,011plants,roughlyaquarterareexporters.However,these13,891plantsrepresentover
halfofthetotalnumberofobservationsandworker‐plantmatches,largelybecauseexportingplants
employmoreworkersonaverage(at161workersascomparedto27workersfordomesticplants)
and because the average worker‐plant match duration is roughly 4 months longer at exporting
plantsthanatdomesticplants.SimilartoMenezes‐FilhoandMuendler(2011),ourdatareportthat
accession rates are lower at exporters than at non‐exporters—24% as compared to 34%,
respectively. We also find that separation rates are lower at exporting plants for our sample of
workers. Our RAIS matched data sample report the common finding in the literature that, on
average,exportersaremoreskill‐intensiveandpayhigherwages(e.g.,BernardandJensen(1995)).
Almost 40% of the manufacturing labor force at exporters is high‐skilled, as defined by those
workerswithatleastahighschooleducation,whilebycomparison25%oftheworkforceishigh‐
skilledatplantsservingthedomesticmarket.Theaverageannualwagepaidbyexportingplantsis
4,907reais,ascomparedto2,541reaisatdomesticplants.Exportersareonlyrepresentedinabout
53%ofthe2,769municipalitiescoveredbyourformalsectordata.28Combinedwiththeirgreater
visibility due to the higher total employment numbers, the data indicate that on average the
municipalities in which exporters are located are more heavily enforced than those in which
domesticplantsarelocated.Onaverage,exportingplantsface61manufacturinginspectionswhile
domestic plants face, on average, 39 inspections. By contrast, exporters and non‐exporters are
represented across almost all industries in Brazil. For this reason, we see little variation across
plant‐typeintheaveragetrade‐weightedrealexchangerateorinunionizationrates.
4. EmpiricalModel
Our goal in this paper is to uncover how trade openness affects labor market reallocation. We
consider thedevaluationof thereal in1999as themainexogenous tradeshockandargue thata
similar trade shock impacts plants differentially based on their exposure to the enforcement of
laborregulationsandtheirmodeofglobalization.Webeginwiththefollowingframeworkinmind:
(1)
istobeexpected,asthesefirmsarenaturallymoreexposedtoenforcementthansmallerfirms(CardosoandLage2007).28 See Aguayo‐Tellez, Muendler, and Poole (2010) for further information on the spatial distribution ofexportingplants.
19
wherejindexestheplant,kindexestheplant’s4‐digitindustry,andtindexestime.Werelateplant‐
level outcomes ( ), such as total plant employment, to time‐varying, plant characteristics ( )
such as average worker tenure at the plant, and the age, gender, educational, and occupational
compositionoftheplant,andtime‐varying, industrycharacteristics( )suchastheunionization
rate29, industry employment, average worker tenure in the industry, and the age, gender,
educational, and occupational composition of the industry. The specification also includes plant
fixed effects ( to capture time‐invariant factors, such as the plant’s unobserved underlying
productivity, technology, ormanagement style,whichmay influence both a plant’s selection into
exportingandplant‐levellabormarketadjustment,andstate‐specificyeardummies( tocontrol
fortheaverageeffectonlaborturnoverofBrazil’smanypolicyreformsoverthistimeperiod.
Importantly, among the time‐varying, industry‐specific characteristics is the trade‐weighted
industry‐specific real exchange rate ( which serves as an exogenous shock to trade
openness.Ourbasicargumentisbasedonthefactthatwhentherealdepreciates,thepriceofgoods
typically imported into Brazil will rise, improving the competitiveness and increasing profits of
Brazilianplants.Totheextentthatplantprofitsandemploymentgrowtharecorrelated,weexpect
that a devaluation of theBrazilian realwill increase employment for the averageplant. The first
column of Appendix Table A.1 reports results from the estimation of equation (1) where the
dependent variable is the logarithm of plant employment. Consistent with the literature (e.g.,
Revenga (1992)), a depreciation of the trade‐weighted real exchange rate is associated with
increasesinemploymentfortheaverageplant.
Equation (1), however, considers only the industry‐time shock of the exchange rate devaluation.
Brazil’s large informal sector suggests significant evasion of Ministry of Labor regulations. The
implications of increased trade openness, via a real exchange rate depreciation, for formal labor
turnover depend on the degree to which plants are exposed to labor market regulatory
enforcement.Wehypothesizethattwoidenticalplantswillresponddifferentlytochangesintrade
opennessdependingonthedefactoregulationstheyface.Forthisreason,weadaptequation(1)as
follows:
29 As labor turnover may be restricted in heavily‐unionized industries, we control for the time‐varyingunionizationrate,basedonBrazilianhouseholdsurveys(PNAD).
20
∗ (2)
where m now indexes the city (munícipio). represents time‐varying, municipality‐level
enforcement of labor regulations, as captured byMinistry of Labor inspections.We beginwith a
measureofenforcementasthelogarithmofthenumberofinspectionsatthecitylevel(plusone).
, ourmain coefficient of interest, captures thedifferential impact of a trade shockonplants in
strictly‐enforced municipalities relative to weakly‐enforced municipalities. In response to an
expansionarytradeshock,suchasBrazil’scurrencydevaluation,plantswishtoexpandemployment
( 0).However,plantsinheavily‐inspectedcitiesmaybedifferentiallyrestrictedfromadjusting
labor ( 0)—as the cost of a formal worker increases, strictly‐enforced plants will increase
employmentby less thanweakly‐enforcedplants—ormayadjust formal laborby relativelymore
( 0),asformalworkregistrationsincrease.
Onecouldworrythatthelevelofenforcementoflabormarketregulationsisnotexogenoustoplant
outcomes( ).Inparticular,enforcementmaybestricterincitieswhereviolationsofthelabor
lawsaremorefrequentorincitieswhereinstitutionsaremoredeveloped.Moreover,enforcement
may be stronger formore visible (i.e., larger andmore globalized) plants. As violations of labor
laws, better institutions, and plant size and type are likely also correlated with labor market
outcomes, tominimizethisconcern,weemphasizethatequation(2)relateschangesovertimein
the enforcement of labor market regulations to labor turnover. Importantly, as in much of the
programevaluationliterature,ouridentificationstrategyreliesonhowtheexogenoustradeshock
impactsplantturnoverdifferentiallybasedontheplant’sexposuretoregulatoryenforcement.
Column(2)ofAppendixTableA.1reportsresultsfromequation(2).Ourdatalargelyconfirmthese
predictions. The exogenous real exchange rate depreciation increases plant‐level employment.
However, consistent with findings in Almeida and Carneiro (2009), increases in regulatory
enforcementatthecityleveltendtodecreaseformalplantsize,suggestingthatincreasesinthecost
of formal workers dominate any potential impact from increased compliance with mandated
benefits and formal work registrations. In this paper, we are interested in the interaction term
reflecting the differential impact of globalization on plants located in strictly‐enforced
municipalities ( ). Our results report that plants in heavily‐inspected cities are restricted from
expandingemploymentwithadepreciationofthecurrency.
21
In the absence of plant‐level information on inspections, our analysis intends to capture the
probabilitythataplant is inspected.Thisallowsforthedirecteffectof inspections,aswellasthe
indirect effect of a neighboring plant’s inspections. To the extent that large cities have many
inspections,butalsomanyplantstobeinspected, incolumn(3)ofAppendixTableA.1,weadjust
our main enforcement variable to control for the size of the city. Specifically, our preferred
enforcement variable characterizes inspections per 100 plants in the city. Evaluated at the 10th
percentileofinspections,a10percentdepreciationincreasesemploymentby1.6%,whilethesame
devaluationincreasesemploymentbyonly1.0%atplantslocatedincitiesatthe90thpercentileof
inspections. These plant‐level results highlight our main predictions—that strict labor market
institutionslimitplants’laboradjustmentinresponsetoshocks.
4.1. Worker‐levelemploymenttransitions
When facinga tradeshock,expandingplantscanadjustalong theextensivemarginby increasing
hiring, decreasing firing, or both, as well as along the intensivemargin by increasing the hours
worked for existing employees or switching from temporary to permanent contracts. Moreover,
enforcement also likely influences adjustment along each of these margins. In order to better
understandthesemechanisms,ourmainreduced‐formequationfocusesonaworker‐levelanalysis.
Inparticular,weaugmentequation(2)asfollows:
∗
(3)
where i indexes the worker and represents worker‐level labor market outcomes, such as
employmenttransitionsandhoursworkedperweek.Wecharacterizeemploymenttransitionswith
threevariables:anindicatorvariablethattakesthevalueoneifamatchbetweenworkeriandplant
jiscreatedattimet(i.e.,ifthereisaplant‐yearaccession);anindicatorvariablethattakesthevalue
one if amatch between worker i and plant j is destroyed at time t (i.e., if there is a plant‐year
separation);andan indicatorvariablethat takesthevalueonewhenworker i isemployedwitha
full‐timecontractinplantjattimet.Allothervariablesaredefinedaspreviouslydiscussed. are
time‐varyingworker‐levelcharacteristics(suchastheworker’stenureattheplant,theworker’sage
(and age squared), education, and occupation) and is a worker‐plant (time‐invariant) match
effect.
22
Wearguethattime‐invariantworker‐plantmatcheffectsareimportantbecausewhenworker‐plant
productioncomplementaritiesexist(asinnewtrademodels),highproductivityplantswillscreen
forhighabilityworkers.Thismayleadtothesortingofhighabilityworkersintohighproductivity
plants.30Inoursetting,thisimpliesthatfollowingatradeliberalization,otherwiseidenticalworkers
may have a higher probability of separation from (or a lower probability of accession to) a high
productivityplantthanfrom(andto)alowproductivityplant.Forthisreason,wereplacetheplant
fixedeffectsfromequation(2)withworker‐plantmatch‐specificfixedeffects( ),whichallowfor
time‐invariant,unobservablematchquality(associatedwiththepotentialforworkersorting)inthe
laborreallocationprocess.31
Ourmainspecificationinequation(3)relatesexogenouschangesinindustry‐specificrealexchange
rates with match‐specific outcomes, between 1996 and 2001, differentially for worker‐plant
matches in strictly‐enforced areas. In other words, we explore a difference‐in‐difference
methodologytoidentifytheeffectsofopennessonlaborturnover.Themaincoefficientofinterestin
equation (3) is , which captures the differential effect of stricter enforcement for workers
employedinplantsexposedtovaryingrealexchangeratechanges.
Inaddition,wearguethatthe implicationsofarealexchangeratedevaluationareheterogeneous
acrossplanttypesandtherefore,considerequation(3)separatelyforexportinganddomestically‐
orientedplants.Ourprioristhatopennessallowsplantsbestplacedtocompeteabroadtoexpand
and those in import‐competing industries to relatively contract. We hypothesize that the
expansionaryeffect(increaseinhiringanddecreaseinfiring)oftheexchangerateshock( )will
belargerforexportingplantsthanforplantsservingonlythedomesticmarket,asforeignmarket
accessimproves.
Asintheplant‐levelanalysis,thetheoreticalpredictionsfor areambiguous.Ontheonehand,the
stricterenforcementof laborregulationsraises thecostof formalworkers.Assuch,plants facing
strictenforcementwillhaveincreaseddifficultiesinadjustinglabor.Ontheotherhand,thestricter
30Krishna,Poole,andSenses(2011)documenttheimportanceofworker‐firmcomplementarityinthelaborreallocation process post‐liberalization using matched employer‐employee data for Brazil. Their results,controllingforthenon‐randomassignmentofworkerstofirms,suggestsastrongbiasinplant‐levelanalyses,asexportersdifferentiallyincreasematchqualityrelativetonon‐exporterspost‐liberalization.31Asneithertheplantnortheworkervarywithinamatch, thematch‐specificeffectsalsocontrol fortime‐invariant,unobservableplantheterogeneityandtime‐invariant,unobservableworkerheterogeneity.
23
enforcementoflaborregulationsalsoincreasesjobquality,intermsofcompliancewithmandated
benefitsfortheworker.Forthisreason,wemightexpectincreasesinemploymentinmoreheavily‐
enforced cities, as formal employment becomes a more attractive option and formal work
registrationincreases.Weempiricallytestthisambiguitygivenourstrongpredictionsontheimpact
of trade openness on employment for exporters relative to non‐exporters. Moreover, we further
hypothesize that labormarket regulations on formal employment are less binding for exporting
firms,andthusexpecttheeffectsofregulatoryenforcementtobelessimportantforplantsexposed
toglobalmarkets.32
Asmany of the covariates in equation (3) are also dummy variables, we choose to estimate the
equation using a linear probability model. Compared to a probit analysis, the linear probability
model has the advantage of allowing for a straightforward interpretation of the regression
coefficients.33 To take into account theoccurrence of repeated observationsof individualswithin
city‐sectors,weclustertherobuststandarderrorsatthecity‐sectorlevel,thoughourmainresults
arerobusttoclusteringatthematchlevel,aswell.
5. MainResults
Table 5.1 reports our main estimates where the dependent variable is the worker‐plant‐year
accession.Incolumns(1)and(2)ofTable5.1,weestimatevariantsofequation(3)forallworker‐
plant matches in our sample. As discussed, the specification controls for unobservable, time‐
invariant worker‐plant match quality, observable, time‐varying worker, plant, and industry
32CardosoandLage(2007)arguethattheintegrationoffirmsininternationaltradeandtheneedtocomplywithinternationalqualitystandardsimplicitlyforcefirmstocomplywithlaborregulations.ThisisreinforcedinHarrisonandScorse(2003),whoreport thatexportersand foreignfirms in Indonesiaaremore likelytocomplywithlaborregulations.Inaddition,BloomandVanReenen(2010)showthatlabormarketregulationsare negatively correlated with the quality of management practices across countries. At the same time,multinational and exporting firms tend to be better managed across all countries, suggesting the betterinstitutionalenvironmentatexportingplantsoffersenhancedcompliance.33 It is well‐known that in the extreme case of a fully saturated model (i.e., one where all independentvariablesarediscretevariablesformutually‐exhaustivecategories),thelinearprobabilitymodeliscompletelygeneralandthefittedprobabilitiesliewithintheinterval[0,1].Whenlookingataccessions,separations,andfulltimecontracts,ourdependentvariablewillbeadummyvariableandapplyingleastsquareswillnotyieldthemostefficientestimator.Therefore, inongoingwork,weareestimatingourmainmodelsusingaprobitanalysisforrobustness.
24
characteristics, and state‐year dummies.34 The point estimates in columns (1) and (2) are never
statisticallysignificant,butthesignsareinformative.Inparticular,thecoefficientssuggestthatfor
otherwiseidenticalworkers,plants,andmatches,adepreciationoftherealexchangerateincreases
aworker’sprobabilityofhireastheaverageplantexpands.
Weargue,however,thattradeshocksandregulatoryenforcementhavedifferenteffectsdepending
on the plant’s mode of globalization. As real exchange rate depreciations increase the
competitivenessofexportingplantsinforeignmarkets,weanticipateanexpansionofemployment
atexportersrelativetonon‐exporters.InthefinaltwocolumnsofTable5.1,wereportcoefficients
forthesetofexportingplantsandnon‐exportingplants,respectively.Wedenoteexportersasthose
firms thatexportedat anypointduring theperiod1996 to2001.This time‐invariant indicator is
designedtominimizepotentialendogeneityconcernssurroundingtheglobalizationdecisionpost‐
devaluation.35
Aspredictedbynewheterogeneousfirmtrademodels,adepreciationoftherealincreaseshiringat
exporters,anddecreaseshiringatplantsproducingforthedomesticmarket.Moreover,consistent
with the literature, the way in which enforcement impacts hiring is different depending on the
globalizationstatusoftheplant.Notably,enforcementhasnostatisticalimpactonexportingplants,
inlinewithresultsinHarrisonandScorse(2003)forIndonesiathatglobalizedfirmsareinternally‐
enforced and more likely to follow existing labor regulations. By contrast, domestic plants in
strictly‐enforcedmunicipalitiesdecreasehiringbymorethanotherwiseidenticaldomesticplantsin
weakly‐enforcedmunicipalities,asthecostofformalworkersincreasesfortheseplants.Theimpact
of trade openness differs for plantswith the samemode of globalization but varying degrees of
34 In particular,we include at theworker level, theworker’s age (andage squared), tenure at theplant inmonths,education(astwodummyvariables—atleasthighschoolandmorethanhighschoolwherelessthanhighschool istheomittedcategory)andoccupation(asthreedummyvariables—skilledbluecollarworker,unskilledwhitecollarworker,andprofessional/managerialworkerwhereunskilledbluecollarworkeristheomitted category). At the plant level, we include average plant wages, plant employment, average workertenure at the plant, and the age, gender, educational, and occupational composition of the plant.We alsoincludethefollowingindustrycharacteristics:theindustryunionizationrate,industryemployment,averageworkertenureintheindustry,andtheage,gender,educational,andoccupationalcompositionoftheindustry.35 In unreported regressions,we test the robustnessof our results to the endogeneity of the export statusindicator. Specifically,we categorize plants based on the industry’s import penetration. This industry‐levelcategorization of the plant’s globalization status largely confirms ourmain findings. Alternatively, we alsoclassifyplantsbasedonthecity’smarketaccess(intheformofthreeindicators:distancetothestatecapital,distanceto the federalcapital,andan indexof transportationcosts fromthecity tothenearestcapitalcityregardlessofstate).Acrossallmeasures,ourmainfindingsarerobusttoadefinitionofglobalizationbasedonthecity’smarketaccess.
25
exposuretodefactolabormarketregulations.
Themagnitudesofourestimatesseemtobeplausible.Evaluatingtheeffectonworkersandplants
located in municipalities at themean level of inspections, a 10 percent depreciation of the real
increasesthehiringprobabilityatexportingplantsby4.1%anddecreasesthehiringprobabilityat
domestic plants by 7.3%. The impact for domestic plants varies depending on the level of
enforcementtheplantfaces—fordomesticplantslocatedinmunicipalitiesatthe10thpercentileof
inspections, a10percentdepreciationof the realdecreases theprobabilityofhirebyonly6.0%,
while workers matched with domestic plants located in municipalities in the 90th percentile of
inspectionsexperienceadecreaseintheaccessionprobabilityofaround9.0%.
InTable5.2,wereportestimatesforequation(3),bytheplant’smodeofglobalization,wherethe
maindependentvariablerepresentstheworker’sseparationfromtheplant.Wehypothesizethata
depreciationoftherealexchangeratedecreasestheprobabilityofseparationfortheaverageplant,
ascompetitivenessincreases.ThisisconfirmedinTable5.2;thepointestimateon ispositive,but
statisticallyinsignificant.Wealsonotethat,aspredicted,theeffectisdrivenbydecreasesinfiringat
expandingexportingplants.Thoughstatistically insignificant, theresults fornon‐exportingplants
point to an increase in firing relative to exporting plants in response to a real exchange rate
depreciation.
Weremindthereaderthatincreasesininspectionsambiguouslyrelatetoseparations.Ontheone
hand,more enforcement increases the cost of firing and thusmaydecrease firings.On the other
hand,withincreasedenforcement,existinglaborbecomesmoreexpensiveandtheplantmayresort
to firingworkers in the short run to overcome the increased cost. Once again,we anticipate the
impact of regulatory enforcement to be stronger for non‐exporting firms where regulations are
most binding. Our data from Brazil support these claims. Exporters in strictly‐enforced
municipalitiesrespondnodifferentlytoanexpansionarytradeshockthandoidenticalexportersin
weakly‐enforced cities. However, non‐exporting plants in strictly‐enforced municipalities
differentially increase firing as compared to similar non‐exporters facing weaker enforcement.
Domesticplantslocatedincitiesatthe10thpercentileofinspectionscontractbyincreasingfiringby
approximately2.0%inresponsetoa10percentdevaluation,whiletheprobabilitythatamatchis
destroyedatsimilardomesticplantslocatedincitiesatthe90thpercentileofinspectionsincreases
bycloseto5.0%inresponsetothesametradeshock.
26
Takentogether,Tables5.1and5.2suggestthatstrongerenforcementoflaborregulationsinfluence
labor turnover along the extensivemargin for non‐exporting plants through increased firing and
decreased hiring. That is, contracting non‐exporters decrease job creation and increase job
destruction even further due to increases in the cost of formal employment in strictly‐regulated
areas. These results offer important implications for policy. Job security in an increasingly
globalizedworldreceivesconsiderableattentionfromacademics,policymakers,andthemedia.Our
results confirm recent trademodels inwhichnon‐exporting plants contract in response to trade
reform. More importantly, our data imply that labor market regulations reinforce these
contractionaryeffectsoftradereformforsmall,constrainednon‐exportingplants.
Expandingandcontractingplantsmayalsoadjustalong the intensivemargin.For instance,when
facedwitha tradeshock,employersmayadjust thehoursworked forexistingemployeesorshift
workers between full‐time and temporary contracts. Although the labor law limits a continuous
work shift to 6 hours, limits theweeklyworking period to 44 hours, andmandates increases in
overtimepay, theeffectsof the labor lawsonplantswill likelydifferdependingon thedegreeof
enforcement.Similarly, the increasedcostof formal laborassociatedwithstricterenforcementof
laborregulationsmayleadplantstoshifttowardstheuseoftemporarycontractsoverpermanent
contracts.Thelattercouldhelpemployersovercomethelong‐termrelationshipsofmorerestrictive
employmentcontracts.Wenextconsidertheseadjustmentsforallplantsandbytheplant’smodeof
globalization.
Table5.3reportsresultsfromtheestimationofequation(3)wherethedependentvariable isthe
logarithmofhoursworkedperweekforallplantsandbytheplant’sexportstatus.Table5.4reports
coefficientsfortheestimationofequation(3)wherethedependentvariableisanindicatorvariable
ifworkeriisemployedwithafull‐timecontractinplantj intimet.AcrossTables5.3and5.4,our
data suggest little variation in the intensive margin in response to exchange rate shocks and
regulatory enforcement for both exporters and non‐exporters. The point estimates in Table 5.3
providesomesuggestiveevidencefortheideathatexportersdifferentiallyexpandhoursandnon‐
exporters differentially contract hours. In Table 5.4, the evidence points to an increased use of
temporary contracts to help smooth the shock from exchange rate changes, particularly at non‐
exportingplants.However,acrossallplanttypes,thedegreeofregulatoryenforcement,contraryto
predictions,doesnotinfluencehoursworkedorcontracttypes.
27
5.1. HeterogeneityoftheResults
Our main results provide suggestive evidence that non‐exporters facing strict labor market
regulations differentially decrease job creation and differentially increase job destruction in
responsetotradeopenness,ascomparedtosimilarplantslocatedinareaswithweaklabormarket
regulatoryenforcement.Wedonotfindanysignificantimpactofenforcementonadjustmentsalong
the intensive margin (hours and contract type) for either exporters or non‐exporters. In this
section,weconsidertheheterogeneityoftheseresultsbasedonindustry‐leveldifferencessuchas
thesector’stechnologicalintensity,basedonplant‐leveldifferencessuchasemployment,andbased
onworker‐leveldifferencessuchastheworker’sagegroup.Giventhelackofstatisticalevidencein
supportofadjustmentsalongtheintensivemargin,werestricttheanalysisforwardtotheimpacton
jobcreationandjobdestruction.Theresultsforhoursandcontracttypeareavailableuponrequest.
TechnologicalIntensity Ourmainargumentrestsonthefactthatatradeshockwillreallocate
factors of production towards more efficient use. We consider labor as the relevant factor of
production in this paper.However, the same shockmay also influence adjustments in the short‐
term in terms of capital and other physical materials factors of production. To ensure that the
effects of labor market regulatory enforcement we find in Tables 5.1 and 5.2 reflect plants’
constraintsinadjustinglaborintheshort‐run,wesplitourmainsampleintosectorsdependingon
the technological intensity. Our assumption here, consistent with much of the literature, is that
sectorsrelyingontechnologyarerelativelycapital‐intensive.Therefore,lowtechnologysectorsare
assumedtobemorelabor‐intensive.Forthisreason,weanticipatethatourmainfindingsaredriven
byplants in low‐tech(labor‐intensive) industries.Werelyondata fromtheWorldBanktodefine
sectors’technologicalintensity.Examplesofhigh‐techindustriesare:petroleumrefining,chemical
manufacturing, and automobile manufacturing. Examples of low‐tech industries are: food and
beverage,textile,andwoodmanufacturing.
Wereportcoefficientsfromtheestimationofequation(3)bythesector’stechnologicalintensityin
Table5.5.Inthetoppanel,thedependentvariablerepresentsanindicatorforjobcreation,whilein
thebottompanel,weuseajobdestructionindicatorasthedependentvariable.Theresultsmostly
confirm our hypotheses. The result that domestic plants in strictly‐enforcedmunicipalities limit
hiring by more in response to a trade shock is driven by domestic plants in low‐tech (labor‐
28
intensive) industries. Table 5.2’s finding that domestic plants located in strictly‐enforced areas
increase jobdestructionbymore is also largelydrivenbydomesticplants in low‐tech industries.
Themaininteractionparameterofinterestisnegativeandstatisticallysignificantatthe10%level
ofsignificance.
PlantSize AsisemphasizedinCardosoandLage(2007),thereisastrongcorrelationbetweenthe
sizeofthefirm,asaproxyforthevisibilityofthefirm,andthenumberofinspections.Theresultsin
Kugler (2004) reinforce this finding. The author reports Colombian labormarket reforms had a
greaterimpactonworkersinlargerfirms.Moreover,itisnowwell‐establishedintheinternational
economicsliteraturethatexportersandnon‐exportersdiffersubstantiallyintermsofproductivity
andsize,amongotherattributes(BernardandJensen1995).Forthesereasons,wenextexplorethe
heterogeneityofourmainfindingsbythesizeoftheplant.Wedefineatime‐invariantlargeplant
indicatorequaltooneforthoseplantswithaverageemploymentbetween1996and2001greater
than themedian value.36We argue that comparing similarly‐sized plants helps tominimize any
possibleselectionbiasassociatedwiththeplant’sglobalizationstatus.
Table5.6displaysresultsfromtheestimationofequation(3),bythesizeoftheplant,forallplants
andbytheplant’smodeofglobalization,wherethedependentvariableinthetoppanelisaworker‐
plantjobcreation.First,wenotethatthepositive,butinsignificant,interactioncoefficientincolumn
(1)ofTable5.1 isdrivenby the impactonbelow‐mediansizedplants.Furthermore, it is evident
thatthiseffectiswhollyconcentratedamongsmall,non‐exportingplants.Thismakessenseasthese
are the plants for which we argue labor market regulations are most restrictive. Due to their
visibility,asissuggestedbyCardosoandLage(2007),largeplantsaremorelikelyalreadyinspected
andaremorelikelytofollowexistingregulations.Assuch,smallnon‐exporterswhoexperiencedan
increaseinenforcementoveroursampleperioddemonstratesignificantdifferencesinjobcreation
and job destruction in response to trade reform as compared to similar plants facing less labor
marketregulatoryenforcement.
In addition, we note that when comparing large exporters to large non‐exporters, the main
coefficientsonthetrade‐weightedrealexchangerateremainstatisticallysignificantandofthesame
signincomparisontoTable5.1.Consideringsimilarly‐sizedplantshelpstominimizepotentialbias
36Medianemploymentacrossallplants is9employees.Wemaintain this cutoff thresholdacrossallplant‐types.
29
associated with selection into exporting post‐trade reform. The results for worker‐plant job
destructionarereported inthebottompanel.Ourconclusionsare in linewiththose fromthe job
creationresultsinthetoppanel.
WorkerAge Our main findings show that, following a trade shock, labor adjustment (and
particularly at non‐exporting plants) varies depending on the degree of enforcement of labor
marketregulations.However,inadditiontothismaineffect,thecompositionofemploymentisalso
likelytobeaffectedbythestringencyofenforcementof laborregulations.Wehypothesizethatin
environments facing strict enforcement, those already employed are more likely to remain
employed,whilenewentrantsorre‐entrantsintothelaborforce—asislikelythecasewithyounger
workers—are less likely to be hired. An implication of this is thatwithin plants after a currency
depreciation, we expect to observe more older workers in the workforce. Table 5.7 reports
estimates for equation (3), dividing the sample by the age of the worker and the plant’s
globalizationstatus.Wedefineworkersas“older”whentheyareolderthan32(themeanworker
ageinthesample)and“younger”whentheyare32orless.
As predicted, we see that increases in de facto labor regulations decrease the hiring of young
workers at non‐exportingplants. The sign on themain interaction term therefore shows that, in
response to a real exchange rate depreciation, non‐exporting plants in strictly‐enforced
municipalitiesdifferentiallydecreasehiringofyouthworkersinparticular.Similarly,theresultthat
non‐exporters in strictly‐enforcedmunicipalitiesdifferentially increase jobdestruction relative to
non‐exportersinweakly‐enforcedareasisdrivenbytheimpactonyoungworkers.
5.3. AggregateImplications
Akeyargumentinfavoroftradeliberalizingreformsisthatfactorscanreallocatetomoreefficient
uses,allowingforenhancedproductivityandgrowth.However, inthispaperwedemonstratethat
the efficient reallocation of labor in response to trade shocks is inhibitedby strictde facto labor
marketregulations.Inthissection,weinvestigatetheextenttowhichdampenedlaborreallocation
alsorestrictsthewithinplantproductivitygainsassociatedwithtradeopenness.
Thereareafewpotentialchannelslinkingincreasedenforcementtolowerplant‐levelproductivity.
Forinstance,theinabilityofplantstoadjusttochangingconditionsandtoreallocatefromdeclining
30
todynamicsectorsmayreduceplant‐levelproductivity.Inaddition,moreregulationsmayprevent
plants from introducing new goods or investing inmore complexproduction technologieswhich
may have higher value‐added, but also face more volatile demand and thus require greater
adjustments.Finally,giventhehighcostsofdismissalsinareaswithstrictemploymentprotection,
employersmay now be forced to retain unproductiveworkers theywould have otherwise fired.
Also,giventheexpectationofajob‐for‐life,employeesmaynowhavelessincentivetoexerteffort,
thusloweringtheirplantandworkerproductivity.
On theotherhand,wecanalso imagineeffects in theotherdirection; that is, stricter regulations
increasingplant‐levelproductivity.As labormarketregulations increasethecostsassociatedwith
formalemployment,plantsmayalsoraisethebarforthequalityofworkerstheyarewillingtohire,
given the increased costs, and consequently increase plant‐level productivity. Moreover, the
expectationofa long‐termrelationshipmayincrease investments inplant‐specific training,which
neithertheemployernortheworkerwouldbewillingtoincuriftherelationshipwasshort‐term.
Finally, businesses may switch away from hiring workers and use mechanized technologies to
replaceworkers,whichmayraiseproductivityfortheremainingworkers.
Intheabsenceofdirectdataonplant‐levelproductivityandprofitability,werelyoninformationon
plant‐levelaveragewagesundertheassumptionthatincreasesinproductivityandprofitabilitywill
be positively associated with increases in plant‐average wages when plants share rents with
workers.37 In Table 5.8, we present coefficients from the estimation of equation (2), where the
dependentvariableistheplant‐levelaveragewageasaproxyforplant‐levelproductivity.Acrossall
plant‐types,adepreciationsignificantlyincreaseswithin‐plantproductivity(consistentwithstudies
like Pavcnik (2002)). Moreover, globalization increases productivity and profitability bymore at
exporting plants relative to plants serving only the domestic market, also consistent with the
literature (e.g., Trefler (2004)). We also find a statistically significant and negative impact of
increased enforcementonplant‐level productivity, asproxiedbyplant‐averagewages, suggesting
that the first effect pertaining to restricted labor reallocation dominates any potential positive
impact of enforcement on firm productivity via increases in investments in training or physical
capital.
37Forinstance,inamodelliketheoneinAmitiandDavis(2012),whereworkers’wagesaredirectlylinkedtofirmprofitsthrougha“fairwage”mechanism.
31
Our focus, however, is on the interaction term, where our predictions are confirmed. Across all
plant‐types,strictenforcementoflaborregulationslimitspotentialwithin‐plantproductivitygains
(duetotheefficientreallocationofworkers)associatedwithtradeopenness,asproxiedbyplant‐
average wages. Regulatory enforcement inhibits productivity gains for both exporting and non‐
exportingplants.
6. ConclusionsandPolicyImplications
Economists have long debated the effects of trade liberalization on labor market outcomes in
developingcountries.Earlystudieshavefoundlittleimpactonplant‐levelemploymentchanges.We
argueapotentialexplanationrelates tohowrestrictive labormarket regulationsare in inhibiting
the reallocation of workers. In this paper, we revisit the question of the impact of trade
liberalization on labor reallocation using data for Brazil. Brazil is an especially interesting case
studygiventhestringencyofthedejure labormarketregulationsinthecountry(seeBotero,etal
(2004)). Furthermore, the topic is also at the forefront of economic policy discussions as the
countryconsidersnewwaysoffosteringindustrialproductivityandofcreatingamorecompetitive
workforce.38 Finally, the size andgeographicheterogeneity of the country also creates significant
variationwithinthecountryontheenforcementofthelaborlaw.
Weexplorethefactthatwithincountries,plantsvaryinthedegreeofexposuretoglobalmarkets
and in the incidence of de facto labor regulations they face. We use a difference‐in‐difference
methodologytoidentifytheeffectsofopennessonlaborturnover,forfirmswithdifferentdegrees
ofexposuretotradeanddefactolaborregulations.Inparticular,weanalyzetheimpactofincreased
exposuretotrade,followingBrazil’scurrencycrisisin1999,anddiscerntheimpactdependingon
theplant’sexposuretoglobalmarketsandtotheenforcementoflabormarketregulationsbasedon
theplant’smunicipallocation.
Weshowthat,inBrazil,theextenttowhichtradeaffectslabormarketoutcomesdependsonthede
facto degree of stringency of the labor regulations faced by plants. Conditional on the
(unobservable) time‐invariant plant‐worker match, and several time‐varying worker, plant, and
38Forexample,theBraziliangovernmenthasrecentlylaunchedaprogramofincentivestopromoteindustrialgrowthandcompetitiveness.Theprogramproposesanexemptionfroma20%socialsecuritylevyonworkerpayrollsforcertainsectors.Eligiblesectorsincludeautomotives,textiles,footwearandplastics(seeFinancialTimes2012).
32
sectorcharacteristics,wenotethat,asispredictedbynewheterogeneousfirmtrademodels,trade
openness is associated with an increase in the probability of hiring at exporting plants and a
decrease in the probability of hiring at domestically‐oriented plants. Furthermore, we find that
labor inspections largely influence labor adjustment along the extensive margin at small, labor‐
intensive, non‐exporting firms for which labor regulations are likely most binding. This is an
especiallyinterestingfindingforpolicymakers,giventhecurrentchallengeofrevampingindustrial
growth,throughamorecompetitivelaborforce,inthefaceofaglobalizingworld.
Insummary,ourresultsstronglysuggestthatmorestringentdefactoregulationslimitjobcreation
with enhanced trade openness. Overall, stricter labor enforcement increases job destruction and
decreases jobcreationat firmsmost restrictedby labor regulations.Wealsoshowthis increased
enforcementisassociatedwithlowerproductivitygainspost‐tradereform.Ourworkthereforemay
offer an explanation for themixed results in the literature on the causal impact of exporting on
plant‐level productivity. From a policy standpoint, our work also suggests that increasing the
flexibilityofde jure laborregulationswillallowfor increased jobcreationandthusofferbroader
accesstothegainsfromtrade.
33
ReferencesAghion, Philippe, Robin Burgess, Stephen J. Redding, and Fabrizio Zilibotti. 2008. “The Unequal
Effects of Liberalization: Evidence from Dismantling the License Raj in India,” AmericanEconomicReview,98(4),pp.1397–1412.
Aguayo‐Tellez, Ernesto, Marc‐Andreas Muendler, and Jennifer P. Poole, 2010. “Globalization and
Formal‐SectorMigrationinBrazil,”WorldDevelopment,38(6),pp.840‐856.Ahsan,Ahmad andCarmenPages, 2009. “AreAll LaborRegulationsEqual? Evidence from Indian
Manufacturing,”JournalofComparativeEconomics,37(1),pp.62‐75.Almeida, Rita and Pedro Carneiro, 2012. “Enforcement of Labor Regulation and Informality”
AmericanEconomicJournal:AppliedEconomics,4(3),pp.64‐89.Amiti,MaryandDonaldR.Davis,2012.“Trade,Firms,andWages:TheoryandEvidence,”Reviewof
EconomicStudies,79,pp.1‐36.Autor, David, William R. Kerr, and Adriana Kugler, 2007. “Does Employment Protection Reduce
Productivity?EvidencefromUSStates,”TheEconomicJournal,117,pp.189‐217.Barros, Ricardo P. and Carlos H. Corseuil, 2004. “The Impact of Regulations on Brazilian Labor
Market Performance,” in James J. Heckman and Carmen Pages, Eds., Law andEmployment:LessonsfromLatinAmericaandtheCaribbean,Chicago:TheUniversityofChicagoPress.
Bernard,AndrewB.andJ.BradfordJensen,1995.“Exporters,JobsandWagesinU.S.Manufacturing,
1976‐87,”BrookingsPapersonEconomicActivity:Microeconomics,pp.67‐112.Bertola,Giuseppe,TitoBoeri,andSandrineCazes,2000.“EmploymentProtectioninIndustrialized
Countries:TheCaseforNewIndicators,”InternationalLabourReview,139(1),pp.57‐72.Besley,TimothyandRobinBurgess,2004. “CanLaborRegulationHinderEconomicPerformance?
EvidencefromIndia,”TheQuarterlyJournalofEconomics,CXIX,pp.91‐134.Bloom,Nicholasand JohnVanReenen,2010. “WhyDoManagementPracticesDifferacrossFirms
andCountries?”JournalofEconomicPerspectives,24(1),pp.203‐224.Botero, Juan C., Simeon Djankov, Rafael La Porta, and Florencio Lopez de Silanes, 2004. “The
RegulationofLabor,”TheQuarterlyJournalofEconomics,119(4),pp.1339‐1382.Brambilla, Irene, Daniel Lederman, and Guido Porto, forthcoming. “Exports, Export Destinations,
andSkills,”AmericanEconomicReview.Broda, Christian and David E. Weinstein, 2006. “Globalization and the Gains from Variety," The
QuarterlyJournalofEconomics,121(2),pp.541‐585.Burgess, Simon M. and Michael M. Knetter, 1998. “An International Comparison of Employment
Adjustment toExchangeRateFluctuations,”Reviewof InternationalEconomics,6 (1),pp.151‐163.
34
Bustos, Paula, 2011. “The Impact of Trade Liberalization on Skill Upgrading: Evidence fromArgentina,”unpublishedmanuscript.
Davidson, Carl, Steven Matusz, and Andrei Shevchenko, 2008. “Globalization and Firm‐Level
AdjustmentwithImperfectLaborMarkets,”JournalofInternationalEconomics,75(2),pp.295‐309.
Caballero,RicardoJ.,KevinN.Cowan,EduardoM.R.A.Engel,andAlejandroMicco,2010.“Effective
Labor Regulation and Microeconomic Flexibility,” Cowles Foundation Discussion Paper No.1480.
Campa, JoseManuelandLindaS.Goldberg,2001. “EmploymentversusWageAdjustmentand the
U.S.Dollar,”TheReviewofEconomicsandStatistics,83(3),pp.477‐489.Cardoso, Adalberto and Telma Lage, 2007.AsNormaseosFactos, Editora FGV, Rio de Janeiro,
Brasil.Clerides,SofronisK.,SaulLach,and JamesR.Tybout,1998. “IsLearningByExporting Important?
Micro‐Dynamic Evidence From Colombia, Mexico, and Morocco,” The Quarterly Journal ofEconomics,113(3),pp.903‐947.
Cosar,A.Kerem,NezihGuner,andJamesR.Tybout,2011.“FirmDynamics,JobTurnover,andWage
DistributionsinanOpenEconomy,”unpublishedmanuscript.Currie,JanetandAnnE.Harrison,1997.“SharingtheCosts:TheImpactofTradeReformonCapital
andLaborinMorocco,”JournalofLaborEconomics,15(3),pp.44‐71.DeLoecker,Jan,2007.“Doexportsgeneratehigherproductivity?EvidencefromSlovenia,”Journal
ofInternationalEconomics,73(1),pp.69‐98.Eslava,Marcela, JohnHaltiwanger,AdrianaKugler,andMauriceKugler,2010.“FactorAdjustments
after Deregulation: Panel Evidence from Colombian Plants,” The Review of Economics andStatistics,92(2),pp.378‐391.
Feenstra,RobertC.,RobertE.Lipsey,HaiyanDeng,AlysonC.Ma,andHengyongMo,2004.“World
TradeFlows:1962‐2000,”NBERWorkingPaperNo.11040.Feliciano, Zadia M., 2001. “Workers and Trade Liberalization: The Impact of Trade Reforms in
MexicoonWagesandEmployment,”IndustrialandLaborRelationsReview,55(1),pp.95‐115.Feyrer, James, 2009. “Trade and Income—Exploiting Time Series in Geography,” unpublished
manuscript.TheFinancialTimes,2012.“BrazilUnveilsMeasurestoBoostIndustry,”April3,2012.Freund, Caroline and Bineswaree Bolaky, 2008. “Trade, regulations, and income,” Journal of
DevelopmentEconomics,87(2),pp.309‐321.Goldberg,LindaS.,2004.“Industry‐specificexchangerates fortheUnitedStates,”EconomicPolicy
Review,FederalReserveBankofNewYork,May,pp.1‐16.
35
Goldberg,LindaS.andJosephTracy,2003.“ExchangeRatesandLocalLaborMarkets,”inRobertC.
Feenstra, Ed., The Impact of International Trade onWages, Chicago: The University ofChicagoPress.
Goldberg, Pinelopi and Nina Pavcnik, 2003. “The Response of the Informal Sector to Trade
Liberalization,”JournalofDevelopmentEconomics,72,pp.463‐496.Gonzaga, Gustavo,NaercioAquinoMenezes‐Filho, and Cristina Terra, 2006. “Trade Liberalization
andtheEvolutionofSkillEarningsinBrazil,”JournalofInternationalEconomics,68(2),pp.345‐367.
Gruben,WilliamC.andSherryKiser,1999. “Brazil: TheFirstFinancialCrisisof1999,”Southwest
Economy,pp.13‐14.Harrison, Ann and Jason Scorse, 2003. “Globalization’s Impact on Compliance with Labor
Standards,”BrookingsTradeForum,pp.45‐96.Hasan,Rana,DevashishMitra,andK.VRamaswamy,2007.“TradeReforms,LaborRegulations,and
Labor‐Demand Elasticities: Empirical Evidence from India," The Review of Economics andStatistics,89(3),pp.466‐481.
Heckman,JamesJ.andCarmenPages,2004.LawandEmployment:LessonsfromLatinAmerica
andtheCaribbean,Chicago:TheUniversityofChicagoPress.Helpman,Elhanan,OlegItskhoki,andStephenRedding,2010.“InequalityandUnemploymentina
GlobalEconomy,”Econometrica,78(4),pp.1239‐1283.Kambourov,Gueorgui,2009.“LaborMarketRegulationsandtheSectoralReallocationofWorkers:
TheCaseofTradeReforms,”TheReviewofEconomicStudies,76(4),pp.1321‐1358.Kaplan,DavidS.,2009. “JobCreationandLaborReforminLatinAmerica,” JournalofComparative
Economics,37,pp.91‐105.Krishna, Pravin, Jennifer P. Poole, andMine Zeynep Senses, 2011. “WageEffects of TradeReform
withEndogenousWorkerMobility,”NBERWorkingPaperNo.17256.Kugler,Adriana,1999.“TheImpactofFiringCostsonTurnoverandUnemployment:Evidencefrom
theColombianLaborMarketReform,”InternationalTaxandPublicFinanceJournal,6(3).Kugler,Adriana,2004.“TheEffectofJobSecurityRegulationsonLaborMarketFlexibility:Evidence
fromtheColombianLaborMarketReform,”inJamesJ.HeckmanandCarmenPages,Eds.,LawandEmployment:LessonsfromLatinAmericaandtheCaribbean,Chicago:TheUniversityofChicagoPress.
Kugler, Adriana andMaurice Kugler, 2009. “LaborMarket Effects of Payroll Taxes in Developing
Countries: Evidence from Colombia,”EconomicDevelopment and Cultural Change, 57 (2), pp.335‐358.
Kume, Honorio, Guida Piani, and Carlos Frederico Braz de Souza, 2003. “A Polıtica Brasileira de
36
Importacao no Perıodo 1987‐98: Descricao e Avaliacao,” in Carlos Henrique Corseuil andHonorio Kume, eds., A abertura comercial brasileira nos anos 1990: Impactos sobreempregoesalários,RiodeJaneiro:MTEandIPEA,chapter1,9–37.
Melitz,MarcJ.,2003.“TheImpactofTradeonIntra‐IndustryReallocationsandAggregateIndustryProductivity,”Econometrica,71(6),pp.1695‐1725.
Menezes‐Filho,NaercioAquinoandMarc‐AndreasMuendler,2011.“LaborAllocationinResponseto
TradeReform,”unpublishedmanuscript.Moreira,MauricioMesquitaandCorrea,PauloGuilherme,1998."Afirstlookattheimpactsoftrade
liberalizationonBrazilianmanufacturingindustry,"WorldDevelopment,26(10),pp.1859‐1874.
Muendler,Marc‐Andreas,2003.“NominalandRealExchangeRateSeriesforBrazil,1986‐2001,”
unpublishedmanuscript.Olarreaga,MarceloandIsidroSoloaga,1998.“EndogenousTariffFormation:TheCaseofMercosur,”
WorldBankEconomicReview,12(2),pp.297‐320.Pavcnik, Nina, 2002. “Trade Liberalization, Exit, and Productivity Improvements: Evidence from
ChileanPlants,”TheReviewofEconomicStudies,69,pp.245‐276.Paz, Lourenco, 2012. “The Impacts of Trade Liberalization on Informal Labor Markets: A
TheoreticalandEmpiricalEvaluationoftheBrazilianCase,”unpublishedmanuscript.Petrin, A. and J. Sivadasan, 2006. “Job Security Does Affect Economic Efficiency: Theory, A New
Statistic,andEvidencefromChile”,NBERWorkingPapers12757.Revenga, Ana L., 1992. “Exporting Jobs?: The Impact of Import Competition on Employment and
WagesinU.S.Manufacturing,”TheQuarterlyJournalofEconomics,107(1),pp.255‐284.Ribeiro,EduardoPontual,CarlosH.Conseuil,DanielSantos,PauloFurtado,BrunuAmorim,Luciana
Servo, and Andree Souza, 2004. “Trade Liberalization, the Exchange Rate, and Job Flows inBrazil,”TheJournalofPolicyReform,74(4),pp.209‐223.
Topalova,Petia,2010.“Factor ImmobilityandRegional ImpactsofTradeLiberalization: Evidence
onPovertyfromIndia,”AmericanEconomicJournal:AppliedEconomics,2,pp.1‐41.Trefler, Daniel, 2004. “The Long and Short of the Canada‐U.S. Free Trade Agreement,” American
EconomicReview,94,pp.870‐895.Verhoogen, Eric, 2008. “Trade, Quality Upgrading and Wage Inequality in the Mexican
ManufacturingSector,”TheQuarterlyJournalofEconomics,123(2),pp.489‐530.Yeaple,StephenRoss,2005.“ASimpleModelofFirmHeterogeneity,InternationalTradeandWages,”
JournalofInternationalEconomics,65,pp.1‐20.Woodcock,SimonD.,2011.“MatchEffects,”unpublishedmanuscript.
1
Figure3.1:NominalandRealExchangeRateSeriesforBrazil,1994–2001
0.0
0.5
1.0
1.5
2.0
2.5
3.0Jul‐94
Nov‐94
Mar‐95
Jul‐95
Nov‐95
Mar‐96
Jul‐96
Nov‐96
Mar‐97
Jul‐97
Nov‐97
Mar‐98
Jul‐98
Nov‐98
Mar‐99
Jul‐99
Nov‐99
Mar‐00
Jul‐00
Nov‐00
Mar‐01
Jul‐01
Nov‐01
NominalExchangeRate(R/$) RealExchangeRate(R/$)
Source:Muendler(2003).
Figure 3.2: Enforcement Intensity by Municipality, 1998 and 2000
1998 2000
® ®
Source: Authors’ calculations based on administrative data from the Brazilian Ministry of Labor (1996-2001).Note: This figure reports the number of inspections per Brazilian municipality, with darker shades representing higher numbersof inspections. The map on the left is for the year 1998, while the map on the right is for the year 2000.
Figure 3.3: Industry Variation in Trade-Weighted RER (Levels and Changes), 1999
Trade-Weighted RER Changes in Trade-Weighted RER
02
46
810
Den
sity
.55 .6 .65 .7 .75 .8Trade−Weighted Real Exchange Rate
®
05
1015
2025
Den
sity
−.4 −.35 −.3 −.25 −.2 −.15Change in Trade−Weighted RER
®Source: Authors’ calculations based on bilateral real exchange rate data from the IMF and trade flows data from the NBER(1998-1999).Note: This figure illustrates industry-level heterogeneity in the level of the trade-weighted real exchange rate (left panel) andin the annual change of the trade-weighted real exchange rate (right panel).
Table3.1:EnforcementData,1996‐2001
ShareofCitiesInspected
AverageNumberof
InspectionsineachCity
AverageNumberofInspectionsPer100RegisteredPlants
AverageNumberofInspectionsPer10,000RegisteredWorkers
1996 0.33 13.2 2.16 14.831997 0.44 13.0 2.19 16.901998 0.38 12.8 2.39 20.901999 0.50 13.8 2.62 21.552000 0.43 14.8 2.68 22.912001 0.52 14.3 2.28 16.42
Source:Authors'calculationsbasedonadministrativedatafromtheBrazilianMinistryofLabor(1996‐2001).Note: This table reports different statistics at the city level between 1996 and 2001. Column (1) reports theshare of cities that have at least one manufacturing labor inspection. Column (2) reports the average numberof labor inspections in each city. Column (3) reports the average number of inspections per 100 registeredplants in the city and column (4) reports the average number of labor inspections per 1,000 registeredworkers("comcarteira ").
Table3.2:DescriptiveStatistics,1996‐2001
AllPlants ExportersNon‐
Exporters(1) (2) (3)
DependentVariableShareofworkers:Hired 0.29 0.24 0.34Fired 0.25 0.22 0.28TemporaryContract 0.02 0.02 0.02Average:HoursPerWeek 43.5 43.4 43.6
Worker‐levelCovariatesAge 32 32 31Shareofworkers:LessthanHighSchool 0.68 0.62 0.74HighSchool 0.25 0.28 0.22MorethanHighSchool 0.07 0.10 0.03
UnskilledBlueCollar 0.10 0.10 0.10SkilledBlueCollar 0.66 0.64 0.68OtherWhiteCollar 0.07 0.07 0.07ProfessionalorTechnical 0.17 0.19 0.15
Plant‐levelCovariatesEmployment 59 161 27AverageWage(inlogs) 8.00 8.50 7.84AverageMatchDuration(inyears) 2.57 2.73 2.40
Municipality‐levelCovariatesInspections 35.1 60.8 38.5
Industry‐levelCovariatesTrade‐weightedRER 0.62 0.62 0.62Employment 18,328 18,755 18,912UnionizationRate 0.21 0.21 0.21
NumberofObservations 313,940 171,012 142,928NumberofWorkers 81,254 46,181 45,543NumberofPlants 53,011 13,891 39,120NumberofMatches 122,017 62,547 59,470NumberofMunicipalities 2,769 1,466 2,578NumberofIndustries 240 237 236Source:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralexchangerates,NBERtradeflows,PNAD,andSECEX(1996‐2001).
Note: This table reports descriptive statistics for the main variables used in our empirical work. We report on
worker‐level variables (averages across workers), plant‐level variables (averages across plants), municipality‐level
variables (averages across municipalities), and industry‐level variables (averages across industries). Column (1)
reports statistics for all plants in our manufacturing sample. Column (2) reports statistics for the set of exporting
plants (defined to be those firms exporting any positive dollar amount at any point during the 1996 to 2001 sample
period)andcolumn(3)reportsstatisticsforthegroupofnon‐exporters.
Table5.1:Trade,Enforcement,andJobCreationDep.Variable:MatchCreation
All All ExportersNon‐
ExportersTRER*Enforcement ‐0.011 ‐0.011 0.042*
(0.015) (0.021) (0.018)Trade‐weightedRER ‐0.020 ‐0.103 ‐0.356** 0.522**
(0.042) (0.080) (0.103) (0.112)NumberofObs. 300,857 300,857 165,176 135,681Plant‐YearControls YES YES YES YESCity‐YearControls YES YES YES YESSector‐YearControls YES YES YES YESWorker‐YearControls YES YES YES YESState‐YearDummies YES YES YES YESMatchFixedEffects YES YES YES YESSource:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangerates,NBERtradeflows,andSECEX(1996‐2001).
Note: This table reports coefficients from the ordinary least squares estimation of equation (3) in the paper, where the dependentvariable is an indicator variable which takes the value one if a match between worker i and plant j is created in time t , for all plantsand by the plant's export status. ** denotes significance at the 1% level; * denotes significance at the 5% level. Robust standarderrors, clustered at the city‐industry level, are reported in parentheses. All regressions also include the following city controls: log(inspections), log (number of plants), and an interaction between the trade‐weighted real exchange rate and log (number of plants).Unreported covariates at the worker level include the worker’s age (and age squared), tenure at the plant in months, education (astwo dummy variables—at least high school and more than high school where less than high school is the omitted category) andoccupation (as three dummy variables—skilled blue collar worker, unskilled white collar worker, and professional/managerialworker where unskilled blue collar worker is the omitted category). At the plant level, we include average plant wages, plantemployment, average worker tenure at the plant, and the age, gender, educational, and occupational composition of the plant. Wealso include the following industry characteristics: the industry unionization rate, industry employment, average worker tenure intheindustry,andtheage,gender,educational,andoccupationalcompositionoftheindustry.
Table5.2:Trade,Enforcement,andJobDestructionDep.Variable:MatchDestruction
All All ExportersNon‐
ExportersTRER*Enforcement ‐0.005 0.003 ‐0.041*
(0.012) (0.016) (0.017)Trade‐weightedRER 0.027 0.114 0.194* ‐0.124
(0.036) (0.065) (0.083) (0.094)NumberofObs. 300,857 300,857 165,176 135,681Plant‐YearControls YES YES YES YESCity‐YearControls YES YES YES YESSector‐YearControls YES YES YES YESWorker‐YearControls YES YES YES YESState‐YearDummies YES YES YES YESMatchFixedEffects YES YES YES YESSource:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangerates,NBERtradeflows,andSECEX(1996‐2001).
Note: This table reports coefficients from the ordinary least squares estimation of equation (3) in the paper, where the dependentvariable is an indicator variable which takes the value one if a match between worker i and plant j is destroyed in time t , for allplants and by the plant's export status. ** denotes significance at the 1% level; * denotes significance at the 5% level. Robuststandard errors, clustered at the city‐industry level, are reported in parentheses. All regressions also include the following citycontrols: log (inspections), log (number of plants), and an interaction between the trade‐weighted real exchange rate and log(number of plants). Unreported covariates at the worker level include the worker’s age (and age squared), tenure at the plant inmonths, education (as two dummy variables—at least high school and more than high school where less than high school is theomitted category) and occupation (as three dummy variables—skilled blue collar worker, unskilled white collar worker, andprofessional/managerial worker where unskilled blue collar worker is the omitted category). At the plant level, we include averageplant wages, plant employment, average worker tenure at the plant, and the age, gender, educational, and occupational compositionof the plant. We also include the following industry characteristics: the industry unionization rate, industry employment, averageworkertenureintheindustry,andtheage,gender,educational,andoccupationalcompositionoftheindustry.
Table5.3:Trade,Enforcement,andHoursDep.Variable:Log(Hours)
All All ExportersNon‐
ExportersTRER*Enforcement 0.004 0.000 0.007
(0.003) (0.003) (0.005)Trade‐weightedRER ‐0.003 0.013 ‐0.011 0.036
(0.011) (0.014) (0.015) (0.025)NumberofObs. 300,857 300,857 165,176 135,681Plant‐YearControls YES YES YES YESCity‐YearControls YES YES YES YESSector‐YearControls YES YES YES YESWorker‐YearControls YES YES YES YESState‐YearDummies YES YES YES YESMatchFixedEffects YES YES YES YESSource:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangeratesandNBERtradeflows,andSECEX(1996‐2001).
Note: This table reports coefficients from the ordinary least squares estimation of equation (3) in the paper, where the dependentvariable is the logarithm of hours worked per week for worker i in plant j at time t, for all plants and by the plant's export status. **denotes significance at the 1% level; * denotes significance at the 5% level. Robust standard errors, clustered at the city‐industrylevel, are reported in parentheses. All regressions also include the following city controls: log (inspections), log (number of plants),and an interaction between the trade‐weighted real exchange rate and log (number of plants). Unreported covariates at the workerlevel include the worker’s age (and age squared), tenure at the plant in months, education (as two dummy variables—at least highschool and more than high school where less than high school is the omitted category) and occupation (as three dummyvariables—skilled blue collar worker, unskilled white collar worker, and professional/managerial worker where unskilled bluecollar worker is the omitted category). At the plant level, we include average plant wages, plant employment, average worker tenureat the plant, and the age, gender, educational, and occupational composition of the plant. We also include the following industrycharacteristics: the industry unionization rate, industry employment, average worker tenure in the industry, and the age, gender,educational,andoccupationalcompositionoftheindustry.
Table5.4:Trade,Enforcement,andContractTypeDep.Variable:Full‐TimeContract
All All ExportersNon‐
ExportersTRER*Enforcement 0.003 ‐0.002 0.009
(0.005) (0.007) (0.006)Trade‐weightedRER 0.022 0.025 ‐0.007 0.060*
(0.012) (0.022) (0.030) (0.028)NumberofObs. 300,857 300,857 165,176 135,681Plant‐YearControls YES YES YES YESCity‐YearControls YES YES YES YESSector‐YearControls YES YES YES YESWorker‐YearControls YES YES YES YESState‐YearDummies YES YES YES YESMatchFixedEffects YES YES YES YESSource:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangeratesandNBERtradeflows,andSECEX(1996‐2001).
Note: This table reports coefficients from the ordinary least squares estimation of equation (3) in the paper, where the dependentvariable is an indicator variable if worker i is employed with a full‐time contract in plant j at time t, for all plants and by the plant'sexport status. ** denotes significance at the 1% level; * denotes significance at the 5% level. Robust standard errors, clustered at thecity‐industry level, are reported in parentheses. All regressions also include the following city controls: log (inspections), log(number of plants), and an interaction between the trade‐weighted real exchange rate and log (number of plants). Unreportedcovariates at the worker level include the worker’s age (and age squared), tenure at the plant in months, education (as two dummyvariables—at least high school and more than high school where less than high school is the omitted category) and occupation (asthree dummy variables—skilled blue collar worker, unskilled white collar worker, and professional/managerial worker whereunskilled blue collar worker is the omitted category). At the plant level, we include average plant wages, plant employment, averageworker tenure at the plant, and the age, gender, educational, and occupational composition of the plant. We also include thefollowing industry characteristics: the industry unionization rate, industry employment, average worker tenure in the industry, andtheage,gender,educational,andoccupationalcompositionoftheindustry.
Table5.5:Trade,Enforcement,andTurnover,BySectorTech‐Intensity
Dep.Variable:MatchCreation
All ExportersNon‐
ExportersAll Exporters
Non‐Exporters
TRER*Enforcement 0.015 0.020 0.042* ‐0.038 ‐0.028 0.032(0.016) (0.023) (0.021) (0.029) (0.035) (0.036)
Trade‐weightedRER 0.086 ‐0.276 0.547** 0.009 ‐0.117 0.599**(0.101) (0.142) (0.135) (0.125) (0.146) (0.203)
NumberofObs. 208,270 102,615 105,655 92,593 62,567 30,026
Dep.Variable:MatchDestruction
All ExportersNon‐
ExportersAll Exporters
Non‐Exporters
TRER*Enforcement ‐0.017 ‐0.013 ‐0.036 0.018 0.020 ‐0.046(0.014) (0.018) (0.020) (0.023) (0.027) (0.032)
Trade‐weightedRER 0.099 0.202 ‐0.051 ‐0.088 ‐0.016 ‐0.389*(0.081) (0.110) (0.113) (0.106) (0.130) (0.170)
NumberofObs. 208,270 102,615 105,655 92,593 62,567 30,026Plant‐YearControls YES YES YES YES YES YESCity‐YearControls YES YES YES YES YES YESSector‐YearControls YES YES YES YES YES YESWorker‐YearControls YES YES YES YES YES YESState‐YearDummies YES YES YES YES YES YESMatchFixedEffects YES YES YES YES YES YES
Low‐TechSectors High‐TechSectors
Source:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangerates,NBERtradeflows,andSECEX(1996‐2001).
Note:Thistablereportscoefficientsfromtheordinaryleastsquaresestimationofequation(3)inthepaper,wherethedependentvariableinthetoppanelisanindicatorvariable
whichtakesthevalueoneifamatchbetweenworkeriandplantjiscreatedintimetandthedependentvariableinthebottompanelisanindicatorvariableequaltooneifamatch
betweenworkeri andplantj isdestroyedintimet ,forallplants,bytheplant'sexportstatus,andbythesector'stech‐intensity.**denotessignificanceatthe1%level;*denotessignificanceatthe5%level.Robuststandarderrors,clusteredatthecity‐industrylevel,arereportedinparentheses.Allregressionsalsoincludethefollowingcitycontrols:log
(inspections),log(numberofplants),andaninteractionbetweenthetrade‐weightedrealexchangerateandlog(numberofplants).Unreportedcovariatesattheworkerlevel
includetheworker’sage(andagesquared),tenureattheplantinmonths,education(astwodummyvariables—atleasthighschoolandmorethanhighschoolwherelessthan
highschoolistheomittedcategory)andoccupation(asthreedummyvariables—skilledbluecollarworker,unskilledwhitecollarworker,andprofessional/managerialworker
whereunskilledbluecollarworkeristheomittedcategory).Attheplantlevel,weincludeaverageplantwages,plantemployment,averageworkertenureattheplant,andtheage,
gender,educational,andoccupationalcompositionoftheplant.Wealsoincludethefollowingindustrycharacteristics:theindustryunionizationrate,industryemployment,
averageworkertenureintheindustry,andtheage,gender,educational,andoccupationalcompositionoftheindustry.
Table5.6:Trade,Enforcement,andTurnover,ByPlantSize
Dep.Variable:MatchCreation
All ExportersNon‐
ExportersAll Exporters
Non‐Exporters
TRER*Enforcement 0.101** 0.042 0.112** ‐0.007 ‐0.014 0.033(0.032) (0.139) (0.033) (0.016) (0.021) (0.021)
Trade‐weightedRER 0.695** 0.368 0.775** ‐0.107 ‐0.366** 0.532**(0.193) (0.786) (0.201) (0.084) (0.103) (0.129)
NumberofObs. 39,097 2,010 37,087 261,778 163,176 98,602
Dep.Variable:MatchDestruction
All ExportersNon‐
ExportersAll Exporters
Non‐Exporters
TRER*Enforcement ‐0.066* ‐0.038 ‐0.069* ‐0.002 0.006 ‐0.030(0.029) (0.114) (0.030) (0.013) (0.016) (0.020)
Trade‐weightedRER ‐0.059 ‐0.055 ‐0.079 0.101 0.200* ‐0.150(0.171) (0.577) (0.179) (0.069) (0.084) (0.111)
NumberofObs. 39,097 2,010 37,087 261,778 163,176 98,602Plant‐YearControls YES YES YES YES YES YESCity‐YearControls YES YES YES YES YES YESSector‐YearControls YES YES YES YES YES YESWorker‐YearControls YES YES YES YES YES YESState‐YearDummies YES YES YES YES YES YESMatchFixedEffects YES YES YES YES YES YES
SmallPlants
Source:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangerates,NBERtradeflows,andSECEX(1996‐2001).
LargePlants
Note:Thistablereportscoefficientsfromtheordinaryleastsquaresestimationofequation(3)inthepaper,wherethedependentvariableinthetoppanelisanindicatorvariable
whichtakesthevalueoneifamatchbetweenworkeriandplantjiscreatedintimetandthedependentvariableinthebottompanelisanindicatorvariableequaltooneifamatch
betweenworkeriandplantjisdestroyedintimet,forallplants,bytheplant'sexportstatus,andbytheplant'ssize.**denotessignificanceatthe1%level;*denotessignificance
atthe5%level.Robuststandarderrors,clusteredatthecity‐industrylevel,arereportedinparentheses.Allregressionsalsoincludethefollowingcitycontrols:log(inspections),
log(numberofplants),andaninteractionbetweenthetrade‐weightedrealexchangerateandlog(numberofplants).Unreportedcovariatesattheworkerlevelincludethe
worker’sage(andagesquared),tenureattheplantinmonths,education(astwodummyvariables—atleasthighschoolandmorethanhighschoolwherelessthanhighschoolis
theomittedcategory)andoccupation(asthreedummyvariables—skilledbluecollarworker,unskilledwhitecollarworker,andprofessional/managerialworkerwhereunskilled
bluecollarworkeristheomittedcategory).Attheplantlevel,weincludeaverageplantwages,plantemployment,averageworkertenureattheplant,andtheage,gender,
educational,andoccupationalcompositionoftheplant.Wealsoincludethefollowingindustrycharacteristics:theindustryunionizationrate,industryemployment,average
workertenureintheindustry,andtheage,gender,educational,andoccupationalcompositionoftheindustry.
Table5.7:Trade,Enforcement,andTurnover,ByWorkerAge
Dep.Variable:MatchCreation
All ExportersNon‐
ExportersAll Exporters
Non‐Exporters
TRER*Enforcement 0.005 ‐0.009 0.056* ‐0.005 0.003 0.048(0.020) (0.028) (0.025) (0.017) (0.021) (0.025)
Trade‐weightedRER ‐0.058 ‐0.422** 0.477** ‐0.009 ‐0.230* 0.753**(0.110) (0.147) (0.155) (0.089) (0.103) (0.144)
NumberofObs. 173,642 92,230 81,412 127,220 72,951 54,269
Dep.Variable:MatchDestruction
All ExportersNon‐
ExportersAll Exporters
Non‐Exporters
TRER*Enforcement ‐0.018 ‐0.007 ‐0.055* ‐0.003 0.013 ‐0.044(0.017) (0.022) (0.024) (0.014) (0.018) (0.023)
Trade‐weightedRER 0.113 0.229 ‐0.118 0.030 0.115 ‐0.233(0.092) (0.119) (0.133) (0.079) (0.098) (0.128)
NumberofObs. 173,642 92,230 81,412 127,220 72,951 54,269Plant‐YearControls YES YES YES YES YES YESCity‐YearControls YES YES YES YES YES YESSector‐YearControls YES YES YES YES YES YESWorker‐YearControls YES YES YES YES YES YESState‐YearDummies YES YES YES YES YES YESMatchFixedEffects YES YES YES YES YES YES
YoungWorkers OlderWorkers
Source:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangerates,NBERtradeflows,andSECEX(1996‐2001).
Note:Thistablereportscoefficientsfromtheordinaryleastsquaresestimationofequation(3)inthepaper,wherethedependentvariableinthetoppanelisanindicatorvariable
whichtakesthevalueoneifamatchbetweenworkeriandplantjiscreatedintimetandthedependentvariableinthebottompanelisanindicatorvariableequaltooneifamatch
betweenworkeriandplantjisdestroyedintimet,forallplants,bytheplant'sexportstatus,andbytheworker'sage.**denotessignificanceatthe1%level;*denotessignificance
atthe5%level.Robuststandarderrors,clusteredatthecity‐industrylevel,arereportedinparentheses.Allregressionsalsoincludethefollowingcitycontrols:log(inspections),
log(numberofplants),andaninteractionbetweenthetrade‐weightedrealexchangerateandlog(numberofplants).Unreportedcovariatesattheworkerlevelincludethe
worker’sage(andagesquared),tenureattheplantinmonths,education(astwodummyvariables—atleasthighschoolandmorethanhighschoolwherelessthanhighschoolis
theomittedcategory)andoccupation(asthreedummyvariables—skilledbluecollarworker,unskilledwhitecollarworker,andprofessional/managerialworkerwhereunskilled
bluecollarworkeristheomittedcategory).Attheplantlevel,weincludeaverageplantwages,plantemployment,averageworkertenureattheplant,andtheage,gender,
educational,andoccupationalcompositionoftheplant.Wealsoincludethefollowingindustrycharacteristics:theindustryunionizationrate,industryemployment,average
workertenureintheindustry,andtheage,gender,educational,andoccupationalcompositionoftheindustry.
Table5.8:Trade,Enforcement,andPlant‐LevelWagesDep.Variable:Log(AverageWage)
All ExportersNon‐
ExportersTRER*Enforcement 0.032** 0.020* 0.032**
(0.006) (0.010) (0.007)Trade‐weightedRER ‐0.123** ‐0.170** ‐0.153**
(0.036) (0.060) (0.043)NumberofObs. 364,555 86,249 278,306Plant‐YearControls YES YES YESCity‐YearControls YES YES YESSector‐YearControls YES YES YESState‐YearDummies YES YES YESPlantFixedEffects YES YES YESSource:Authors calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangeratesandNBERtradeflows,andSECEX(1996‐2001).
Note: This table reports coefficients from the ordinary least squares estimation of equation (2) in the paper,
where the dependent variable is the logarithm of plant‐level average wages, for all plants and by the plant's
export status. ** denotes significance at the 1% level; * denotes significance at the 5% level. Robust standard
errors, clustered at the city‐industry level, are reported in parentheses. Unreported covariates at the plant‐level
include average worker tenure at the plant, the age, gender, educational, and occupational composition of the
plant, as well as city‐level enforcement, and industry‐level unionization rate, industry employment, average
workertenureintheindustry,andtheage,gender,educational,andoccupationalcompositionoftheindustry.
TableA.1:Trade,Enforcement,andPlant‐LevelEmployment
Dep.Variable:Log(Employment)
Log(Inspections)
Log(InspectionsPer100Plants)
TRER*Enforcement 0.055** 0.034*(0.005) (0.014)
Trade‐weightedRER ‐0.119** ‐0.430** ‐0.171**(0.044) (0.049) (0.048)
NumberofObs. 367,978 367,978 367,978Plant‐YearControls YES YES YESCity‐YearControls YES YES YESSector‐YearControls YES YES YESState‐YearDummies YES YES YESPlantFixedEffects YES YES YESSource:Authors'calculationsbasedonRAIS,MinistryofLaboradministrativedataoninspections,IMFbilateralrealexchangerates,andNBERtradeflows(1996‐2001).
Note: This table reports coefficients from the ordinary least squares estimation of equations (1) and (2) in the paper,where the dependent variable is the logarithm of plant‐level employment. In columns (1) and (2), enforcement ismeasured as the logarithm of the number of inspections in the city (plus one). In column (3), enforcement is measured asthe logarithm of the number of inspections in the city (plus one) per 100 plants in the city. ** denotes significance at the1% level; * denotes significance at the 5% level. Robust standard errors, clustered at the city‐industry level, are reportedin parentheses. Unreported covariates at the plant‐level include average worker tenure at the plant, the age, gender,educational, and occupational composition of the plant, as well as city‐level enforcement, and industry‐level unionizationrate, industry employment, average worker tenure in the industry, and the age, gender, educational, and occupationalcompositionoftheindustry.