The Spoils of War: Trade Shocks during WWI and Spain's Regional Development

62 Pages·2017·5.88 MB·English

∗
SpoilsofWar: TradeShocks&SegmentedLaborMarketsinSpainduringWWI
SimonFuchs(AtlantaFed)
December2021
Abstract
Howdoesdomesticfactormobilityshapethewelfareeffectsoftrade? Tostudythis,Iemployanevent-
study design to show that during WWI a large trade shock caused uneven labor reallocation, wage and
price growth across Spanish provinces. Employing a fully estimated spatial equilibrium model featuring
imperfectspatialandsectorallabormobility, IshowthatduringWWIincreasesinconsumerpricesoffset
nominal income gains. Afterwards, the effects of persistent labor reallocation increased welfare by 2.93
pc. Industrial centers benefited from sectoral reallocation, while spatial flows disseminated gains across
provinces.Loweringthespatialmobilityfrictionsdecreasescountervailingpriceeffects.
JELclassification: D5,F11,F12,F15,F16,N9,N14,R12,R13
Keywords: GainsfromTrade,LaborMobility,EconomicGeography
∗Fortheircomments,IamgratefultoKonradAdler,TrebAllen,AndyBernard,EmilyBlanchard,ToniBraun,AlbertCarreras,Thomas
Chaney,KeremCos¸ar,KlausDesmet,ChrisEdmond,PatrickFève,JimFeyrer,SharatGanapati,ChristianHellwig,RobJohnson,Brian
Kovak,TimLee,ThierryMayer,RoryMcGee,MartíMestieri,NinaPavcnik,FranckPortier,BRavikumar,VincentRebeyrol,JuanRubio-
Ramírez, Mohammed Saleh, Chris Snyder, François de Soyres, Bob Staiger, Robert Ulbricht, Nikolaus Wolf as well as seminar and
workshopparticipantsinmanyplaces.IalsowanttothankJavierSilvestreforsharinghisdataoninternalmigrationinSpainandJulio
Martinez-Galarragaforsharingtheirdataonconsumerpricesaswellasprovidinghelpfulcomments. Finally,Iacknowledgefinancial
supportviaERCgrantNo337272-FiNet.Allremainingerrorsaremyown.Theviewsexpressedhereinarethoseoftheauthorandnot
necessarilythoseoftheFederalReserveBankofAtlantaortheFederalReserveSystem.
1 Introduction
In recent years, a rapidly growing empirical literature has documented the uneven incidence and effect of
tradeshocksacrosslocallabormarketswithincountries(Autoretal.,2016;Topalova,2010;Kovak,2013;Dix-
CarneiroandKovak,2017;JaravelandSager,2019;McCaigandPavcnik,2018). Thesestudieshaveempha-
sized the distributional consequences of uneven trade shocks on employment, wages and consumer prices
across locations and occupations. However, they often employ a regression-based approach that implicitly
treatsindividuallabormarketsasindependentobservations. Thisabstractsfromtherichnetworkstructure
that-vialabormobilityacrosssectorsandspace-connectslocallabormarketsanddetermineshowshocks
affect factor allocation and consumer prices in general equilibrium. While some studies have incorporated
this feature1, important questions remain unanswered: What is the implication of spatially uneven shocks
across connected local labor markets? What is the qualitative role of labor reallocation in determining the
gainsfromtrade? Andwhatistherelativequantitativeimportanceofsectoralcomparedtospatialmobility?
Inthispaper,Iexplorehowlocallabormarketsthatareconnectedviaimperfectlabormobilityshapethe
aggregatewelfareconsequencesoftradeshockswithunevenincidence. Iarguethatwhenlabormobilityis
impeded by sectoral and spatial mobility frictions, the uneven incidence of a trade shock across local labor
marketsmatters. Ontheonehand, improvementsinallocativeefficiencymayariseandincreasegainsfrom
trade. Ontheotherhand,theunevenincidenceofatradeshockacrossconnectedlocallabormarketsmight
causeheightenedcompetitionforalimitedpoolofworkers,inducingonlylimitedreallocationoflabor,and
thuslimitgainsfromtrade.
Toillustratetheseopposingeffectsofuneventradeshocks,considerastylizedexampleofasimpleecon-
omy with two local labor markets (i,j). Labor is imperfectly mobile and supply in i is increasing in local
wages, w, but decreasing in wages elsewhere, w . Labor demand is decreasing in wages in location i, but
i j
increasinginaparameterthatrepresentsdemandshifts, e.2 Let ρ and ρ betheelasticityoflabordemand
i i j
with regard to demand shifts, and ψ and ψ be the own-wage and cross-wage elasticity of labor supply,
ii ij
respectively. Now, consider a (small) demand shift in both i and j (dlne > 0,dlne > 0), and solving for
i j
wageandemploymentchangesthatsatisfylabormarketclearinginbothiandj,weobtain,3
dlnw = α dlne +α dlne dln(cid:96) = β dlne +β dlne
i 1 i 2 j i 1 i 2 j
where α ∝ ρ, α ∝ ψ ρ . Furthermore, β and β are linear combinations of the reduced-form effect of
1 i 2 ij j 1 2
thedemandshockonwagesacrosslocallabormarkets,wheretheweightsaregivenbytheownandcross-
wageelasticityoflaborsupply. Thereduced-formsystemclarifiestheroleofunevenshocks: First,giventhe
assumptionsabove,bothα andβ arepositive,implyingthatthedirecteffectofanincreaseinlocaldemand
1 1
istoincreasewagesandlaborallocations. Ifademandshockdisproportionatelyaffectshigherproductivity
sectors or locations, then efficiency gains arise. Second, α is positive and β is negative, implying that the
2 2
indirecteffectofdemandshockselsewhereinduceswagepressureandreduceslaborallocations. Thestrength
of this channel depends on how connected the two local labor markets are as measured by the cross-wage
elasticity ψ . Therefore, if trade shocks only affect a small set of tightly connected local labor markets, the
ij
consequenceisheightenedcompetitionforalimitedpoolofworkers,wagepressureandlimitedreallocation.
1Monteetal.(2018)exploretheimpactofproductivityshocksacrosslocallabormarketsthatareconnectedbycommutinglinkages
inoutputandinputmarkets(tradeandmigrationfrictions). Caliendoetal.(2019)characterizethedynamicevolutionofthespatial
equilibriumincorporatingmigrationlinkages.Adaoetal.(2019)revisittheimplicationsofthe’Chinashock’employingareduced-form
systemthattakesgeneralequilibriumfeedbackintoaccount.
2ForsimplicityIassumethatlabordemandisindependentofwageselsewhere. Thisamountstoassumingthatoutputmarketsare
completelysegmentedbetweeniandj. Thiscanberelaxedandthequalitativepredictionswillholdregardlessaslongastheindirect
impactofwageselsewhereonlabordemanddonotexceedinmagnitudetheindirecteffectsonlaborsupply.
3DerivationscanbefoundintheonlineappendixSectionB.1
1
Increasedwages,inturn,maypass-throughintoincreasesinconsumerprices,offsettingincomegains. Inthe
aggregate,unevenwageandpricegrowthtogetherwithonlylimitedefficiencygainsmightbetheresult.4
Inpractice,studyingtheeffectofatradeshockacrossconnectedlabormarketsischallengingfortwodis-
tinctreasons.First,withoutknowledgeabouttheconnectednessoflabormarkets,itisdifficulttodisentangle
thedirectandtheindirecteffect,especiallywhenashockiscorrelatedacrosslocallabormarkets. Therefore,
itiscrucialtomodelandestimatethelinkagesbetweenlocallabormarketsinasufficientlyrich,yettractable
way. Second,observedchangesinwagesandemploymentmightbedrivenbyunobservedproductivityim-
provements that could be correlated with an observed trade shock. To disentangle the impact of demand
shocks from other confounders, an exogenous trade shock is needed that affects local labor markets in an
observablyunevenmanner.
Toovercomethesechallengesandstudytheeffectsoftradeshocksacrossconnectedlocallabormarkets,
thisstudycombinesthreedifferentelements: First,Iexamineanaturalexperiment,whereaplausiblyexoge-
noustradeshockwithdiscerniblespatialandsectoralasymmetriesaffectsacountrywithhighlysegmented
local labor markets. The reduced-form evidence illustrates how an international trade shock that affects
closelyconnectedlocallabormarketscancauselocalizedwagepressurethatfeedsintoconsumerprices,but
alsoreallocationoffactorstowardshighproductivityindustries.Second,Idevelopandestimateaneconomic
geographymodelwherelocallabormarketsareconnected. Theworkerfacesareallocationchoicesubjectto
switchingcoststhatimpedebothspatialandsectoralmobility. Thiscreatesatractablelaborsupplysystem
that links local labor markets. I furthermore show how the historical setting can be used to estimate the
model. Third, Ishowthatchangesinlaborflowsacrossspaceandsectorscanbeusedasasufficientstatis-
tic to measure and decompose gains from trade taking spatial and sectoral labor reallocation into account.
Thismethodologycleanlydistinguishesbetweenimprovementsintheworker’scurrentlocallabormarket-
cross-sectionalwelfareimprovements-andimprovementsintheworker’soptionvaluetorelocatetoother
locallabormarkets-dynamicgainsfromtrade. SimulatingWWIshockunderdifferentdegreesof(spatial)
labormarketsegmentationallowsmetoquantifythesensitivityofgainsfromtradetofactorimmobility.
Attheheartofthisstudyistheanalysisofahistoricalnaturalexperiment:Aninternationaltradedemand
shocktotheSpanisheconomythatwascausedbytheparticipationofSpain’skeytradingpartnersinthefirst
WorldWar(1914-1918).Spain,however,remainedneutralthroughoutthisperiod,butwasindirectlyaffected
viatradelinkages.5 Thisuniquehistoricalepisodefeaturesaplausiblyexogenousandlargetradeshock(cp.
Figure1),withhighlyunevenincidenceacrosssectorsandacrossspace. Spainatthetimewasmarkedbya
highdegreelabormarketsegmentation,however,despitethatthetradeshockinducedsectoralreallocation
(cp. Figure 2), causing at the same time a dramatic increase in consumer prices (cp. Figure 3). A detailed
analysisofthisperiodisonlypossible,becauseofauniquespatialpanelthatbringstogetherforthefirsttime
hand-collecteddataontraderecords, detailedemploymentsurveysacrossallSpanishprovinces, aswellas
dataonconsumerprices,coveringtheperiodbetween1910-1920.
Ibeginbyexaminingthehistoricalevidence: First,Iestablishtheexogeneityoftheshock. Iexaminethe
historicaltraderecordsandshowthatanincreaseinexportsisassociatedwithdemandfactorsinbelligerent
countriesthatcoincidewiththeoutbreakofthewar. Thisexportshockwasmarkedbysectoralasymmetry
and the composition was in line with products that would have been needed for the sustained war effort.
Second, I then examine the shock’s impact on wages, labor allocations and consumer prices. The shock
4TheonlineappendixSectionB.2providesanextensionofthismodelforanarbitrarynumberoflocallabormarkets.Inthatsetting
theeffectonwagesandemploymentacrosslocallabormarketscanbewrittenintermsofadirectandindirect(generalequilibrium)
effect. Furthermore,whilethegeneralequilibriumadjustmentsmightbedifficulttoexpressinclosed-form,anapproximatereduced-
formcharacterizationintermsofownandcross-wagelaborsupplyelasticitiesisfeasible.
5TheshockwascausedbycircumstancesexternaltotheSpanisheconomy,specificallyreducedindustrialcapacityduetothelarge-
scalemobilizationrequiredforthewareffortaswellasheightenedwarneedsinbelligerentcountries,particularlyFrance,whileSpain
remainedneutralthroughouttheconflict.
2
Figure1: AggregateTradeLevels
Notes:Aggregateexports(inmillionpesetas)bywhetherdestinationcountryisbelligerent.Belligerentcountriesareprimarybelligerentcountrieswhere
tradewasnotdisruptedbythefrontlineitself,i.e. i.e. France,ItalyandtheUnitedKingdom. Thenon-belligerentcountriesexcludetheUnitedStates
andotherlaterparticipantsofWWI.TheblueshadedareaindicatestheperiodofWWI.Thesourcedataarethedigitizedproduct-destinationleveltrade
statistics.
was highly uneven across space and sectors, and its direct effect induced limited spatial labor reallocation,
but highly uneven wage and price growth. I also show that indirect exposure to the trade shock further
exacerbated wage pressure: Provinces that are more closely located to other provinces that were heavily
affectedbythetradeshock,experiencedadditionalwageandpricepressure.
Thepresenceofindirecteffectspointstowardstheimportanceofspatiallinkagesacrossgoodsandfactor
markets. Inordertoquantifythewelfareeffectsofatradeshockinthepresenceofsuchlinkages,thispaper
takesanexplicitlystructuralapproach. Iincorporateimperfectlabormobilityacrossspaceandsectorsinan
otherwisestandardeconomicgeographymodel. Thechallengeistodevelopaframeworkforlabormobility
that is sufficiently rich to capture the details of how segmented labor markets interact, both characterizing
theflowsbetweensectorsandacrossspace, butstillremainstractabletobeempiricallyestimated. Thekey
innovationistorelyonasequentialformulationoftheworkers’relocationchoicethatseparatesspatialfrom
sectoraladjustmentsandthusmodelscomplexinteractionsacrossspaceandsectorsinatractableway.
To characterize the welfare implications of a trade shock under imperfect labor mobility, I extend the
sufficient statistic approach by Arkolakis et al. (2012) and Ossa (2015) and derive a closed-form formula
for the gains from trade that takes domestic re-allocation into account. The model implies an inverse and
invertible relationship between the share of workers that remain in a location or sector and the spatial or
sectoralex-anteexpectedutilityassociatedwiththereallocationchoice.6 Therefore,observingthechangein
theshareofworkersthatremainacrosscomparativestaticsisasufficientstatisticfromwhichcommonlyused
aggregatewelfaremeasurescanbeconstructed. Thesequentialchoiceimpliesaseparabilitybetweenspatial
and sectoral mobility which in turn allows us to decompose and attribute welfare changes to changes in
sectoralandspatialmobility. Thismethodologyallowsforasharpdecompositionofwelfareimprovements
into different qualitative channel: Improvements in the current local labor market to which the worker is
attachedsignifycross-sectionalstaticgains. Increasesinsectoralmobilitypindownincreasesintheoption-
value of sectoral relocation, while increases in spatial mobility are associated with increases in the option-
6Conceptuallythisapproachisrelatedtothenotionthat(conditional)choiceprobabilitiesareinformativeofchoice-specificvalue
functions,asiscommonlyusedintheestimationofthedynamicdiscretechoicemodels(HotzandMiller,1993).Inthissetting,remain
sharesaredirectlyrelatedtotheoption-valueoftheworker’smobilitychoice. Thisoption-valueofthereallocationchoiceisrelatedto
whatMcFadden(1977)termedthesocialsurplusfunctionwhichisacommonlyusedwelfaremeasure.
3
Figure2: AggregateCompositionoftheEconomy
Notes:Primary(agriculture),secondary(industry/manufacturing)andtertiary(services)shareoftotalemploymentinSpainbetween1877-1930.Dataseries
isconstructedfromcensusdata,usingthecalculationsinHarrison(1978)fortheyears1877-1900,andfollowsowncalculationsfortheperiodbetween1900-
1930.TheblueshadedareaindicatestheperiodofWWI.Furtherinformationonhowaconsistentdataseriesisconstructedfromcensusdataisprovidedin
theonlineappendix.
value of spatial relocation. These latter two channels are associated with the worker’s mobility choice and
constitutedynamicgains.
Toimplementthewelfareanalysistwoingredientsareneeded:First,thisapproachrequiresknowledgeof
laborflowsintheshockedandnon-shockedcounterfactualscenario,whichinturnrequiresafullyestimated
modeltosimulatethecounterfactualscenario. Second,thewelfareformularequiresestimatesofthespatial
andsectorallaborsupplyelasticitiesaswellasthetradeelasticitytobeconstructed.Alargepartofthisstudy
isthereforededicatedtoexploitingthehistoricalsettingtoobtaincredibleestimatesofkeyelasticiesandto
fullycalibratethemodel,includingadetailedefforttoestimatemobilityfrictionsacrossspaceandsectors.
The estimation of the model proceeds in four steps. In a first step, I derive from the model a structural
reduced-formthattracesoutnominalincomegainsasafunctionofthespatialexposuretotheWWIshock.
Implementingthisregressiondesignallowsmetoestimatedomestrictradecosts.Inasecondstep,Iinvertthe
modeltolocation-sectorspecificfundamentalswhichcorrespondtomarket-shareshifters. Thesensitivityof
thesemarketshareshifterswithregardto(exogenous)changesintheinputcostpinsdownthetradeelasticity.
I exploit the differential impact of the trade shock across locations and sectors to isolate wage changes and
thereby identify the trade elasticity. I then turn towards estimating the parameters that determine labor
flows in the model. Typically, the estimation of mobility frictions would require data on flows of workers
across space and sectors. However, in historical settings this is rarely available. In this setting, the census
providesadditionaldataonthestockofresidentsdissectedbytheirplaceofbirthin1920and1930. Ibegin
byexploitingthisinformationtoestimatea(spatial)gravityregression,whichidentifiesthespatialdecayof
migrationflowsaswellastheaverageout-migrationshareacrossallprovinces. Inordertoestimatesectoral
switching costs and labor supply elasticities I fit the model to changes in labor allocations at the province-
sectorlevelfrombeforetoafterthewar. Achallengeisthatmigrationdecisionsweremadeduringthewar
based on wage dynamics that are not directly observable at the province-sector level. To overcome this, I
invert the model to back out baseline productivities and then feed in the observed aggregate WWI trade
shock to simulate - conditional on a guess for the migration frictions that are being estimated - labor flows
andwagesthatsolvethelabormarketequilibriumconditionsacrossallprovince-sectorunits. Thefrictions
are being estimated by minimizing the distance of the simulated labor allocations and the observed labor
allocationsjustafterthewar.
4
Figure3: EvolutionoftheSpanishCPI
Notes:ConsumerpriceindexaccordingtoBallesterosDoncel(1997)between1890and1936.Normalizedto100in1913.Blue-shadedareaindicatesWWI
period.
AsaresultoftheestimationIobtainpredictedsectoralandspatiallaborflowsthatareconsistentwiththe
observed changes in province-sector specific labor allocation. Both sectoral and spatial frictions are highly
prohibitive. However, the implied reallocation strongly suggest that spatial frictions dominate sectoral ad-
justmentfrictions,with76percentoftheadjustmenthappeningacrosssectorswithinprovinces,ratherthan
betweenprovinces.
With the fully estimated model in hand, it is possible to quantify and decompose the welfare effects of
theWWIshockacrossSpain. Todoso,Ifirstsimulatelaborflowsinthenon-shockedscenario,byusingthe
calibrated model to simulate the sectoral and spatial reallocation if external trade and productivity would
haveremainedatthe1914level. Withthelaborflowsintheshockedandnon-shockedscenarion,thewelfare
effectscanbequantified. Thisisdonebothforthescenariowherethelabormarketclearingwage(andthere-
fore the prices of domestic tradeables and non-tradeables) is generated feeding in the WWI shock and for
thescenariowheretheWWIshockjustdissipatedandexportlevelslevelofftoalowerpost-warexportsce-
nario.Thisexerciseallowsustoexaminethedynamicsofthegainsfromtradefromatemporarytradeshock:
Whiletheshockpersistedincreasesinconsumerpricesentirelyoffsetanygainsfromtrade. Aftertheshock
dissipated,thereallocationalgainsincreasedwelfareby2.93percent,howeverthegainswerehighlyuneven
across provinces. Lowering the spatial segmentation, increases reallocative gains and decreases offsetting
priceeffects.
Inafinalstep,Itraceoutthewelfareeffectsfordifferentdegreesof(spatial)labormarketsegmentation.
Byvaryingthespatialmobilitycostandrecalculatingthelaborflowsinboththeshockedandnon-shocked
scenarioIcantraceouttherelationshipbetweenlabormarketsegmentationandgainsfromtrade. Notsur-
prisingly, as labor markets become more integrated the gains from trade increase. However, the exercise
shines a role on the qualitative and quantitative importance of spatial mobility. Interestingly, the marginal
gainsareequallysharedbetweenincreasesinreallocationandlessenedcountervailingpricepressure. This
reinforces the insight from the theoretical model, that uneven trade shocks cause price pressure and that
factormobilityplaysanessentialroleinmitigatingthis.
Related literature. My paper is related to a number of different strands of research. First, there is a long-
standingliteratureininternationaltradeexaminingtheimplicationsofalackoffactormobility,goingback
atleasttothecanonicalanalysisusingthespecificfactormodel(Jones,1971;Mayer,1974;Mussa,1974,1982).
5
Mussa (1982) in particular pointed out that factor immobility leads to differential income gains across sec-
tors with different factor endowments. More specifically related to labor adjustments, a number of papers
have further examined the interaction between dynamic labor adjustments and external trade shocks with
Matsuyama (1992) developing a first tractable analysis, and with a more recent set of papers exploring the
phenomenonquantitatively(TombeandZhu,2019;Kambourov,2009;Artucetal.,2007;Dix-Carneiro,2010;
Dix-CarneiroandKovak,2017;Kovak,2013;Caliendoetal.,2015;FajgelbaumandRedding,2014;Fan,2019;
Adao et al., 2019; Monte et al., 2018; Caliendo et al., 2019). What is less explored in this literature is the in-
teractionbetweenconnectedlocallabormarkets,unevenshockincidenceandconsumerprices.7 Thispaper
fillsthisgapbyprovidingbothreduced-formevidencefromauniquehistoricalnaturalexperimentaswellas
acompletequantitativeanalysisoftheinteractionbetweenlabormarketsegmentation,uneventradeshocks
andconsumerprices.
Second,mypapercontributestotheliteratureoncharacterizinggainsoftradeusingsufficientstatistics.
Recent contributions sought to extend the initial work (Arkolakis et al., 2012) to allow for multiple sectors
with different trade elasticities (Ossa, 2015), or workers with heterogeneous productivities across sectors
(Galle et al., 2017; Kim and Vogel, 2020; Lee, 2020). This paper contributes to this literature by characteriz-
ing gains from trade taking into account the imperfect reallocation of workers across domestic local labor
markets8andhighlightingthatdatalaborflowscanbeusedtoconstructasufficientstatistictodoso.9
Third,thepaperaddstotheliteratureonSpanisheconomichistorybyshowingthattheWWIshockhad
animportantimpactontheSpanisheconomybyreallocatingfactorsacrossspaceandsectorstoprovidethe
preconditions for an economic take-off in the 1920s. As such it is a middle ground between two opposing
viewsintheliterature. Thetraditionalview, representedbyRoldanandDelgado(1973), interpretsthewar
asalargeturningpointforeconomicdevelopment. Themodernview,representedbyPradosdelaEscosura
(2016) emphasises that the shock actually decreased real GDP and instead he points towards the 1920s as
a much more important decade for Spain’s development. My analysis provides a middle ground between
thesetwoopposingviews,pointingtowardssubstantialreallocationandnominalincomegains,buttracing
outsubstantialcountervailingpriceeffectsthataredrivenbyreallocationcostsinthelabormarket,leading
to positive but somewhat modest welfare gains despite a historically large demand shock to the Spanish
economy.
Outline. Theremainderofthepaperisstructuredasfollows. Section2describesthehistoricalbackground
and the various data sources as well as the construction of the data set that underlies most of the analysis.
Section3givesreducedformevidenceonthetradeshockanditseffectlocallabormarkets.Section4describes
the theoretical model, the estimation of the model as well as the welfare quantification. Finally, Section 5
concludes.
7Whiletheliteraturegenerallydoesnotfocusontheinteractionbetweenlocallabormarkets, thestudiesbyHelm(2020);Adao
etal.(2019)arenotableexceptions. InHelm(2020),theauthorexploitsemploymentspilloversbetweenlocallabormarketstoestimate
agglomeration effects. In Adao et al. (2019), the authors revisit the reduced-form analysis of Autor et al. (2013), but introduce an
estimationframeworkthatexplicitlytakeslabormarketlinkagesandgeneralequilibriumresponsesintoaccount. Indeedthestylized
modelinthenextSectioncanbeseenasasimplifiedversionoftheirframework,butthefocusoftheiranalysisabstractsfromefficiency
gainsandhowthoseareconditionedbytheunevenessoftheshock.
8KimandVogel(2020)conductasimilaranalysis,takingtheimperfectreallocationofworkersintoaccountwhencalculatingthe
welfareeffectoftheChinashockacrossUSlabormarkets. However, byabstractingfrombilateralreallocationintheirsetting, they
cannotcapturetherichinteractionsbetweencloselyconnectedlabormarketsthatisthecenterpieceofthisstudy. Intheirempirical
implementation,theyfurthermoreabstractfromtheimpactonconsumerprices.
9Thepaperisalsorelatedandaddstoagrowingliteraturecharacterisingthewelfareimplicationsoffactormisallocation, going
back at least to Harberger’s initial analysis (Harberger, 1964), with Baqaee and Farhi (2020) offering a characterisation of the effect
ofmicroeconomicshocksininefficienteconomies,HornbeckandRotemberg(2019)studyingtheimplicationsofmisallocationonthe
welfareeffectoftransportationimprovementsintheUShistoricalcontextandZarate(2021)doingsoinanurbancontemporarycontext.
6
2 Data and Historical Background
Beforeturningtowardstheempiricalanalysis,thissectionwillprovidebackgroundonthehistoricalcontext
andthekeydatasourcesthatunderlietheanalysis. IwillfirstdescribekeyfeaturesoftheSpanisheconomy
justbeforetheoutbreakofWWI,focusinginparticularonthespatialandsectoralorganizationoftheSpanish
economy,externaltradeandthesegmentationofdomesticlabormarkets. Iwillthenproceedbyintroducing
thehistoricalspatialdatasetthatwillbeusedintherestofthepaper.
2.1 TheSpanisheconomyatthebeginningofthe20thcentury.
At the beginning of the 20th century, Spain remained at a relatively low level of industrial development.10
According to census data, in 1900 roughly 70pc of the working population worked in agriculture and only
12.5pc worked in manufacturing. Industrialization only proceeded slowly, with the industrial sector only
growing marginally in total employment by 3pc, adding a little bit less than 40,000 jobs nation-wide in the
first decade of the century. At that time, the largest share of the industrial sector was made up of sectors
associatedwithprimarygoods,suchastheexploitationofminesortheproductionofconstructionmaterial.
Spatial distribution of economic activity. In terms of the spatial distribution of the population, most of
the population was still concentrated in predominantly rural and agricultural areas such as Andalucía11 or
Castilla y León.12 Major urban centers such as Oviedo, Valencia, Bilbao, Madrid and Barcelona concen-
tratemostoftheindustrialactivityascanbeseenbythemapinFigure4indicatingthespatialdistribution
of manufacturing employment. The industrial structure of those urban centers was heterogeneous. For
example, Barcelona was highly specialized in the cotton textile industry, while Valencia specialized in gar-
ments. Because of natural endowments mining and associated downstream industries dominated Oviedo
and Jaen. The Basque country had an early advantage in the heavy metal industries, featuring numerous
Martin-Siemensopen-hearthfurnacesforsteelproductionaswellasotherfixedinstallations.
Internal migration. Up until the 1920s, the Spanish economy was marked by perennially low levels of
internalmigration,withnetmigrationneveramountingtomorethan5pcthepopulationatadecennialrate
(Silvestre, 2005). Explanations focus mainly on an insufficient release of agricultural works to urban areas,
driven either by supply based factors - such as low agricultural productivity and demographic dynamism
- or demand based factors - such as the lack of pull of industry and services until at least WWI.13 Either
explanationisperfectlyconsistentwiththepointofviewthatsubstantialpushorpullfactorswererequired
toovercometheeconomic,linguistic,orsociologicalbarriersthatimpededspatialandsectoralmobility.
Externalmarkets. Finally,intermsofexternalmarkets,attheendofthe19thcentury,(former)coloniesand
otherLatinAmericanmarketsplayedaparticularlyimportantrole,whileafterthelossofthecoloniesSpain’s
10Aftermissingthefirstwaveoftheindustrialrevolutioninthefirsthalfofthe19thcentury(Harrison,1978),theSpanisheconomy
underwentaperiodofrapidindustrializationinthesecondhalfofthe19thcentury,fueledbymarketintegrationduetotheexpansionof
therailroadnetworkwhichinturnresultedinthedevolutionofindustrialcapacitytotheperipheralprovinceswiththecottonindustry
inCataloniaandMetallurgyintheBasquecountrydevelopingespeciallyrapidly(Nadal,1975). However,industrializationsooncame
toanearlyhaltwiththecensusdatashowinglittleincreaseinindustrialemploymentfrom1887onwards.Thisisalsomirroredbyvery
lowGDPperheadgrowthratesaveraging0.6percentbetween1883-1913(PradosdelaEscosura,2017).Someauthorsattributethelow
levelsofgrowthtolimiteddemandformanufacturinggoodsdomesticallyaswellaslittlecapacitytocompetewithgoodsfromcountries
suchasGermany,FranceandtheUKthataremoreadvancedintermsoftheirindustrialization(Harrison,1978).
11Andalucíacompriseseightprovinces:Almería,Cádiz,Córdoba,Granada,Huelva,Jaén,MálagaandSeville,withmajorindustrial
activitylocatedinSevilleandMiningemploymentinHuelva
12CastillayLeóncomprisesnineprovinces: Ávila,Burgos,León,Palencia,Salamanca,Segovia,Soria,ValladolidandZamorawith
majorindustrialactivitycenteredinValladolid.
13Foracompletediscussionandreferencesofdemand-basedandsupply-basedexplanationseeSection2inSilvestre(2005).
7
Figure4: SpatialDistributionofManufacturingEmployment
Notes:Mapoftotalmanufacturingandminingemploymentbyprovincein1910(excludingCanaryIslandsandNorthAfricanpossessions).Sourcedatais
the1910census.
exportsshiftedmoretowardsEuropeancountrieswithFranceandGreatBritaintakingupthebiggestshare
ofexports(comparetheright-hand-sideinFigure5). Mostoftheexportswererawmaterialsoragricultural
productsconsistentwiththelowdevelopmentalstatusofSpainatthetimeasdepictedontheleft-hand-side
inFigure5. Ingeneral,Spainranatradedeficitformostofthebeginningofthe20thcenturyexceptforthe
shortperiodunderconsiderationinthispaper.
2.2 AspatialdatasetforSpainbetween1910-1920.
To examine the impact of WWI on both trade flows and local labor markets, I construct a regionally disag-
gregateddatasetforSpainbetween1910-1920thatcovershandcollectedinformationonwages,employment
levels,pricesandexportsacrosslocallabormarkets. Thisdatasetallowsmeforthefirsttimetoanalyzethe
impact of the trade shock taking both external trade and internal labor reallocation into account. I rely on
sixprincipaldatasourcesthattogetherdescribemanufacturingandagriculturalemployment,externaltrade,
migrationpatterns,consumerprices,thetransportationnetworkandthehousingmarket.14
Manufacturing employment. I obtain disaggregated information regarding wages and labor quantities
across local labor markets. At the beginning of the 20th century, the plight of the working class and their
working conditions became a more prominent political issue in Spain. In order to better understand and
14Seetheonlineappendixfordetailedinformationonreferencesfordatasourcesanddetailsondataconstruction.
8
Figure5: TopExportSectorsandDestinations(1910,1915,1916)
Notes:Aggregateexports(inmillionpesetas)bysector;aggregateexports(inmillionpesetas)bydestinationcountry.Exportsreportedfortopsevensectors
andtopsixdestinationsrespectivelyaccordingtotheirrankin1915.Thesourcedataarethedigitizedproduct-destinationleveltradestatistics,asdiscussed
intheonlineappendix.
tracktheworkingconditionstheInstituteforSocialReform-anentitythatwouldlatermorphintothemin-
istryoflabor-startedconductinglarge-scalesurveysonworkingconditionswiththefirstannualreportbeing
released in 1907. The institute continued to publish yearly reports covering the whole period of 1910-1920.
The surveys were conducted at all public firms and large private enterprises in cities that are larger than
20,000inhabitants(Casanovas2004). Theycovered23differentindustries15 and48differentprovinces.16 In
theannualreports,theinstitutionreportedwages,workinghours,andnumberofemployeesacrosslocalla-
bormarkets. Theresultsareavailableintwodifferentformats. Ontheonehand,industry-specificresultsare
availableacrossthemoregeographicallyaggregatedunitofregions,ontheotherhand,provincialwagesare
reported but with the industry-specificresults missing. Additionally, the Ministry of Labor laterpublished
a compilation that offers a more complete picture across local labor markets with employment and wages
beingreportedacrossprovince-sectorpairsfortheyears1914,1920and1925(MinisteriodeTrabajo,1927).
Agriculturalemployment. Iaugmenttheindustrysurveywithadditionaldatafromthecensus. Whilethe
industrysurveycoversalargerangeofthemanufacturingsector,itdoesnotgivefurtherinformationonthe
remainingeconomy.Asmentionedbefore,acrucialfeatureoftheSpanisheconomywasthelargeagricultural
sector. Toaccountforthat, Idigitizedtheoccupation-provincespecificSectionofthecensusfor1900, 1910,
1920, and 1930. I use the 1920 data on agricultural employment to augment the 1920 data. For the 1914
data,Iusethe1910province-specificagriculturalemploymentdataandextrapolatebycalculatingprovince-
specificfertilitytrendsuntil1914. Finally,IusedatacontainedintheofficialSpanishstatisticalyearbookson
province-specificagriculturalmeanwagesfor1915and1920.
Externaltrade. Iobtaineddetaileddataregardingexportsandimportsfromannualtraderecordsreleased
by the Spanish custom agency. I digitized the trade statistics for the years 1910-1919. For those years, the
15The industries included are called: Books, Ceramics, Chemicals, Construction, Decoration, Electricity, Food, Forrest, Furniture,
Garments,Glass,Leather,MetalWorks,Metallurgy,Mines,Paper,Public,PublicIndustry,Textiles,Tobacco,Transport,Varias,Wood.
16Thecensusfor1910lists49differentprovinces. Theymostlycorrespondtothemodernadministrativeunitscalledprovincias-
provinces-whichareinturnroughlytheNUTS3leveladministrativeunitsofSpain. Therearesomeminordifferences,e.g. inhow
differentoff-continentaladministrativeunitsarebeingtreated. FormyanalysisIdroptheCanaryislandsfromthesamplesincetheir
distancefromthemainlandmakesithardtoarguethattheyaresimilarlyintegratedasotherprovinces.
9