IT Shields: Technology Adoption and Economic Resilience during the COVID-19 Pandemic

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) IT Shields: Technology Adoption and Economic Resilience during the COVID-19 Pandemic Myrto Oikonomou, Nicola Pierri, Yannick Timmer 2023-010 Please cite this paper as: Oikonomou, Myrto, Nicola Pierri, and Yannick Timmer (2023). “IT Shields: Technology Adoption and Economic Resilience during the COVID-19 Pandemic,” Finance and EconomicsDiscussionSeries2023-010. Washington: BoardofGovernorsoftheFederalReserve System, https://doi.org/10.17016/FEDS.2023.010. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

IT Shields: Technology Adoption and Economic Resilience during the COVID-19 Pandemic Myrto Oikonomou† Nicola Pierri ‡ Yannick Timmer§ Abstract We study the labor market effects of information technology (IT) during the onset of the COVID-19 pandemic, using data on IT adoption covering almost three million establishments in the US. We find that in areas where firms had adopted more IT before the pandemic, the unemployment rate rose less in response to social distancing. IT shields allindividuals,regardlessofgenderandrace,exceptthosewiththelowesteducationalattainment.Instrumentalvariableestimates–leveraginghistoricalroutineemploymentshare asaboosterofITadoption–confirmIThadacausalimpactonfosteringlabormarkets’resilience.Additionalevidencesuggeststhisshieldingeffectisduetotheeasinessofworkingfrom-homeandtostrongercreationofdigitaljobsinhighITareas. JELCodes: E24,O33 Keywords: Unemployment Rate, Technology, IT Adoption, Inequality, Skill-Biased Technical Change *This paper was prepared as background material for the US Article IV 2020 and for the WEO October 2020 Chapter 2. We are grateful to Nigel Chalk, Damiano Sandri, Anke Weber, and presentation participants at the CESifoConferenceontheEconomicsofDigitization,PSDRNAnnualConferenceonFirmsResilience,Innovation andTechnology,andtheIMFfortheirinsightfulcomments. Co-authorYannickTimmerworkedonthisproject priortoemploymentattheFederalReserveBoard, whileemployedbytheInternationalMonetaryFund(IMF). Data sources were obtained under purview of IMF licenses. The views expressed in the paper are those of the authorsanddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,itsManagement,theFederal ReserveBoardortheFederalReserveSystem. †InternationalMonetaryFund.Email:moikonomou@imf.org ‡InternationalMonetaryFund.Email:npierri@imf.org §FederalReserveBoard.Email:yannick.timmer@frb.gov

1 Introduction As COVID-19 spread across the world and the United States in 2020, people greatly reduced their mobility, stayed more at home, and spent less time producing and consuming products and services that require face-to-face interactions. These changes, caused by both voluntary behavior and various mitigation policies, have also severely damaged the economy. What are the labor market consequences of lockdowns and mobility restrictions? And can information technology(IT)mitigatetheseadverseeffects? Foreveryone? Thispaperanalyzestheinterplaybetweenthesuddendeclineinmobility, itseffectonthe labormarket,andfirms’adoptionofITintheUS.Itreliesonseveraldatasources,andinparticularonsurveydatacoveringsoftwareandhardwarepurchasesin2016foralmostthreemillion establishmentsindifferentindustries. Firm-level IT adoption can strengthen or dampen the effect of mobility on economic outcomes in several ways. On the one hand, IT adoption can cushion the impact of the pandemic by facilitating work-from-home or contact-less interactions [Bloom, 2020; Brynjolfsson et al., 2020; Papanikolaou and Schmidt, 2020] and by increasing online sales. IT adoption can facilitate online job search, which may be particularly important when physical mobility is reduced. Availability of IT investments and capabilities may also spur the creation of new digital-intensivejobs. Ontheotherhand,thepandemicmayreinforcethesubstitutionoflabor withtechnologyforex-anteheavyITadopters[ChernoffandWarman,2020]. High-technology adopting firms may be more inclined to automate processes when the pandemic spreads as humanswouldbeatriskofcontractingthevirus. WefindthatITadoptionsignificantlyshieldsworkersfromtheeconomicconsequencesof the pandemic. Figure 1 illustrates the increase in the unemployment rate between February andApril2020foreachUSstateandthedeclineinmobilityduringthesameperiod. Inlow-IT adoption states, there is a strong correlation between the drop in mobility and the rise in the unemploymentrate. Conversely,mobilityisnotassociatedwithrisingunemploymentratesin states with higher IT adoption. An event study empirical design confirms this finding and illustrate that states hit more harshly by the pandemic and states with more IT adoption were not experience different pre-pandemic trend in unemployment. We present further evidence relyingonindividual-leveldatafromtheCPS(CurrentPopulationSurvey)respondentsandusingwithin-state(MSA-level)variationinITadoptionwhilecontrollingforarichsetofvarious otherpotentialconfoundingfactors,suchasthepre-pandemicindustryandoccupationofthe 1

respondent. We find that respondents living in MSAs with a larger drop in mobility are more likelytobeunemployedduringSpring2020(controllingforpre-pandemicunemployment),but theimpactofmobilityislesspronouncedamongMSAswhereITwasadoptedmoreintensely. Importantly, we provide causal estimates on the mitigating role of firms’ IT adoption on local labor markets thanks to an instrumental variable approach. IT adoption can be correlated(andcausedby)severallocalcharacteristics,suchasavailabilityofhumancapital. While wecontrolforvariouspotentialconfoundingfactors,suchasthelevelofeducation,wecannot ruleoutthatunobservablecharacteristicsaredrivingthemitigatingimpactofIT.Wethusfollow Autoretal.[2003]byinstrumentingregional-levelITadoptionbyitshistoricalroutineemployment share. In regions where historically more routine workers were employed, IT adoption has been faster and stronger when the price of IT equipment fell and routine workers could be replaced by technology. Because of path-dependency, even today IT adoption is higher in areas where historically the routine employment share was higher than in other regions. InstrumentalvariableregressionsconfirmourOLSestimates: theimpactofthemobilitydropon unemploymentprobabilityislowerinareaswhereITisadoptedmoreintenselybyfirms. This points toward IT playing a causal role in mitigating adverse employment outcomes during a pandemic. WequantifytheeffectofITadoptionrelativetoacounterfactualscenarioinwhichthepandemichadhittheworldfiveyearsearlier. Thedigitaleconomyasashareofemploymentgrew byaround10%relativetofiveyearsbefore(seesubsection5.2fordetails).Combiningthisnumberwithourregressionresults,wefindthattheunemploymentratewouldhavebeenaround2 percentagepointshigherduringAprilandMay2020ifITadoptionwouldhavebeenatthelevel of2015. Insteadofanunemploymentrateof14%theunemploymentratewouldhavereached 16%.1 Therecentliterature(seesection2forabriefreview)hasarguedthattheeconomicconsequencesofCOVID-19–especiallyatitsonset–weresignificantlymoresevereformoreeconomically vulnerable individuals, such as women, racial minorities, immigrants, and individuals withlowereducationalattainment. ITadoptionmay alsohaveaheterogeneousimpactalong thosedimensions. Forinstance,informationtechnologycanbeacomplementforskilledlabor, 1Thisback-of-the-envelopecalculationshouldbetakenwithagrainofsalt,ascomputingtheaggregateeffects fromcross-sectionalheterogeneityisdifficult. OurspecificationdoesnotallowustotakepotentialgeneralequilibriumeffectsintoaccountthatwouldaffecttheaggregateconsequencesofITadoptionandinsteadonlycaptures thepartialequilibriumeffectscomingthroughIT. 2

whileitmaysubstituteunskilledlabor. IftheCOVID-19shockpromotesfurtherautomationof productionprocesses,andmoresoformoreITintensivecompanies,thenitmaydifferentially impactwomenormenaccordingtowhichindustryissubjecttothegreatestchanges(e.g.manufacturing sector predominantly employs male workers). Minorities have been experiencing COVIDdeathsandinfectionsathigherrates[Kirby,2020];anoccupationaldistributionskewed towards occupations requiring in-person contacts is a main potential culprit. Therefore, IT adoption, by facilitating the delivery of contactless services and goods, may help individuals employedintheseriskyoccupations. The effect of IT adoption in shielding workers is consistent across most groups. We show thatbothmalesandfemalesaswellasindividualsofdifferentracesbenefitfromITadoption. However, we find a striking difference in the way IT adoption shields individuals with heterogeneouslevelsofeducationalattainment. Individualswithhigh-andmediumlevelsofeducationsignificantlybenefitfromITadoption, whileindividualswithloweducationalattainment (those who did not complete high school) are not shielded by IT. These findings suggest that the COVID-19 pandemic increases inequality across educational groups through skill-biased technicalchange. Thisisconsistentwithevidencefrompastrecessionswhenlow-skilledindividuals were disproportionately affected, which further reduced complementary IT skills and persistentlywidenedinequality[Heathcoteetal.,2020]. Finally,weinvestigatetheroleofdifferentchannelsinexplainingtheshieldingeffectofIT. WefindthatlocalITadoptionisstronglycorrelatedwithmeasuresofthefeasibilityofworking from home [Dingel and Neiman, 2020]. We find that local IT adoption and the ability to work fromhomearebothindependentlyshieldingtheeconomyfromalocalmobilityshock,butthe roleoflocalITissignificantlyreducedwhenlocalworking-from-homeabilityiscontrolledfor. ThissuggeststhatpartoftheshieldingimpactofITisduetohighITfirmshavingfacinglower disruptionintheshifttoworkfromhome,butotherforcesarealsoatplay. Conversely,wefind nosignificantroleforlocalfirms’accesstoe-commercetechnologies. Localunemploymentcanbeimpactedbybothjobdestructionandjobcreation. Thepandemic depressed job creation both because of lower labor demand and because mobility restrictionslimitedfirmsandworkersabilitytomeetinperson,potentiallyworseninglabormarket search frictions. To shed more light on the importance of IT adoption for the job creation margin, we study how online job postings respond to the pandemic and to local IT adoption. WefindthatduringSpring2020,jobpostingsdeclinedmoreinMSAsthatsufferedalargerdrop 3

inmobility,butthisdeclinewaslesspronouncedinhigh-ITMSAs.ThisshieldingimpactofITis presentonlyforjobpostingsthatrelatetodigital-intensiveoccupationsandnotforotherjobs. Thus, a further reason why IT shielded local labor markets from the impact of the pandemic wasthatitprotectedfirms’abilitytocreatevacanciesfordigitaljobs. Theseresultssuggestthat ITadoptionimprovedfirms’abilitytoadjustjobcreationinaflexibleanddynamicmannerand highlight the role that local IT played in facilitating the transition to a more digital economy duringtheearlystagesofthepandemic. Wefinallytestfortheimportanceoflocaldemandspillover–whichcanbeasourceofgeneral equilibrium effects at the local level–by testing whether IT shields local-level employment in tradableornon-tradableindustries[MianandSufi,2014]. WefindnomitigatingroleoflocalIT fornon-tradableindustries,suggestingaminorroleforsuchspillovers. Theremainderofthepaperisstructuredasfollows. Insection2wepresentabriefliterature review. Insection3wedescribethedata. Insection4weillustratestate-levelpatterns. Insection5wepresentevidence(includingIVestimates)onthemitigatingroleofITusingindividuallevel data. In section 6 we investigate the potential channels through which IT shields and in section7weconclude. 2 Related Literature TheliteratureontheeconomiccrisistriggeredbytheCOVID-19pandemichasexpandedvery rapidly. Foranearlyreviewofthisliterature, seeChapter2ofthe2020OctoberWEO(IMF)or Brodeuretal.[2020]. Some authors have argued that voluntary social distancinghas had a more important role than lockdowns in disrupting economic activities [Allcott et al., 2020; Bartik et al., 2020; Kahn et al., 2020; Maloney and Taskin, 2020]. This literature documents that people’s mobility and economic activity in the US contracted before lockdowns [Chetty et al., 2020] and that lifting lockdowns led to a limited rebound in mobility [Dave et al., 2020] and economic activity (Cajner et al. [2020] is an exception). Goolsbee and Syverson [2020] find small differences in visits to nearby retail establishments by people that faced different regulatory restrictions because of being located in different counties. Similar results are documented in Chen et al. [2020] that expand the analysis to Europe and find no robust evidence of the impact of lockdowns on several high-frequency indicators of economic activities. The importance of voluntary so- 4

cialdistancingisalsohighlightedbythecaseofSwedenthat—despiteavoidingstrictlockdown measures—hasexperiencedsimilar(thoughslightlysmaller)declinesinmobilityandeconomic activitiestocomparablecountries[Andersonetal.,2020;Chenetal.,2020]. Whilenotthefocus ofthispaper,ourresultsalsosuggestthatvoluntarysocialdistancingratherthandejurerestrictionsaremostlyresponsibleforthedeclineinmobility. Some papers have documented that more economically vulnerable individuals—such as thosewithlowerincomeandeducationalattainment[Cajneretal.,2020;Chettyetal.,2020;Shibata,2020],minorities[Fairlieetal.,2020],immigrants[BorjasandCassidy,2020],andwomen [Alonetal.,2020;DelBocaetal.,2020;PapanikolaouandSchmidt,2020]—havebeenimpacted more harshly during the early phases of the COVID-19 pandemic, both in the US and other countries [Alstadsæter et al., 2020; Béland et al., 2020]. One reason is that lower-paid workers areoftenunabletoperformtheirjobswhileworkingfromhome[DingelandNeiman,2020;Gottliebetal.,2020]. Thispointstoapotentialwideningofinequality[MongeyandWeinberg,2020; Palominoetal.,2020]. Wealsoshowthatthedeclineinmobilityhasraisedtheunemployment rateforethnicminoritiesaswellaslow-educatedindividualsmoststrongly, therebywidening inequality.However,weaddanadditionalelementtothedebate.WeshowthatITadoptioncan shieldvariousmembersofsociety,regardlessoftheirgenderorrace,fromthemobility-induced COVID-19shock.Oneexceptionislow-educatedindividualsforwhichwedonotfindshielding byITadoption. InareaswherefirmsareheavyITadopters,theincreaseinoverallinequalitycanbedampened. However,intheseareasonlyhighlyeducatedindividualsbenefitfromthehigherex-ante IT adoption, not lowly educated ones. Therefore, in these areas, the COVID-induced mobility shock,raisesthistypeofinequalityevenmorethaninlowITadoptingareas. TheclosestpapertooursisChiouandTucker[2020],whichstudytheimpactofthediffusion ofhigh-speedInternetonanindividual’sabilitytoself-isolateduringthepandemic. Theyalso focus on the US and find that, while income is correlated with the ability of social distancing, thediffusionofhigh-speedinternetexplainsmostofthisincomeeffect. AlargeliteraturehasalsostudiedtheimplicationsofITadoptionforvariousoutcomes,such asproductivityandlocalwages(seeforinstance,Akermanetal.[2015];Autoretal.[2003];BrynjolfssonandHitt[2003];Bloometal.[2012];Beaudryetal.[2010];Bresnahanetal.[2002];Bloom andPierri[2018];Formanetal.[2012];McElheranandForman[2019];BessenandRighi[2019]). WestudytheroleofITasamitigatingfactorfortheCOVID-19shock. Closertousistherefore 5

PierriandTimmer [2020]thatshowthatITadoptioninfinancewasamitigatingfactorduring theGlobalFinancialCrisis. ITadoptionhasbeenconsideredanimportantskill-biasedtechnologicalchange[Acemoglu and Autor, 2011]. While IT is often a complement for highly skilled workers, it can often substitute the work of less-skilled workers. In previous recessions, less-skilled workers have been alsohardhitbyeconomicconditions, whichreinforcedthetrendofskill-biasedtechnological changeHeathcoteetal.[2020]. Finally, there was been a growing body of literature that studies how COVID-19 impacted labormarketstappingonhigh-frequencydatafromonlinejobboards(seeforexampleHensvik et al. [2021], Bellatin and Galassi [2022], Soh et al. [2022], Adrjan et al. [2021] and Marinescu etal.[2020]). Theuseofonlinejobplatformdatapredatesthepandemic. Wecontributetothis literaturebyexploringtheimpactofthepandemicononlinevacancies,andtheroleoflocalIT adoptionfordigitalandnon-digitaljobpostings. 3 Data Sources ITadoption We construct a set of measures of local-level IT adoption building on an establishmentsurveyonITbudgetperemployeebyCiTBDsAberdeen(previouslyknownas“Harte Hanks”)for2016. Weaccessdataonmorethan2,800,000establishments,e,inallstatesinthe US.2WetakethelogoftheITbudgetperemployeeIT andestimatethefollowingregressions: e IT =δ+α +θ +(cid:178) (1) e g(e) ind(e) i whereα isafixedeffectforthegeographicalunitweareinterestedin,i.e.stateorMSA.θ g ind isanindustry(2-digit)fixedeffect. α isusedasourmeasureofITadoptionfortherespective g geographical unit. The fixed effect can be interpreted as the average log of the IT budget per employeeinanestablishmentinagivengeographicunit,conditionalonitsindustry.Wecontrol for industry fixed effects to ensure that our measure of IT adoption is not solely driven by the factthatsomeindustriesareheavierITadoptersandlocatedinregionswhereunemployment behaveddifferentlyduringtheCOVID-19pandemicthaninothersduetoreasonsotherthanIT adoptionoftheestablishments. 2WhiletheITdataareattheestablishmentlevel,weusefirmsandestablishmentsinterchangeablyintherestof paper. 6

Other data sources We use the Current Population Survey (CPS) to assess the effect of the lockdownonthelabormarket[Floodetal.,2021].TheCPSisasurveythatistheprimarysource of monthly labor force statistics in the US. We construct the unemployment rate at different levels of aggregation, i.e. MSA, state, and national levels. The mobility data are coming from Googlemobilityreports. GoogleCommunityMobilityReportsdatausethelocationhistoryof usersondifferenttypesofactivities,suchasretailandrecreation,todocumenthowthenumber ofvisitsandthelengthofstayatvariouslocationschangedcomparedtoapre-COVIDbaseline. ThedatacapturetheGPSlocationofindividualsatvariousplaces,suchasretailandrecreation, workplaces, transit station, parks, etc.. The data are made available as disaggregated as the county level for the US and are reported as an index compared to the pre-COVID 19 period (January-February). LockdowndataareobtainedthroughKeystoneandtheiroriginalsourcearethestatewebpages. Lockdown data are based on 11 non-pharmaceutical intervention (NPI) dummy variables,i.e. (i)theclosingofpublicvenues,(ii)banofgatheringsize500-101,(iii)banofgatheringsize100-26,(iv)banofgatheringsize25-11,(v)banofgatheringsize10-0,(vi)fulllockdown, (vii)non-essentialservicesclosure,(viii)banofreligiousgatherings(ix)schoolclosure,(x)shelter in place, and (xi) social distancing. The dummy variablestake the value one ifthe specific NPIisinplaceandzeroifnot. Foreachstateonagivenday,wetaketheaverageacrossthe11 lockdowndummiessothatalockdownof100%referstohavingall11NPIsinplaceatagiven time. Werelyonadditionalstandarddatasourcesforlocal-levelcharacteristics. Theseinclude theAmericanCommunitySurveyforlocalsocio-demographiccharacteristics,theCountyBusiness Patterns and Quarterly Workforce Indicators for local level industrial composition of the workforce,andOccupationalEmploymentandWageStatisticsdataforlocalleveloccupations. We use high-frequency online job postings data from Indeed, a leading job postings platform. Using data from online job boards has become a common practice in a wide range of labor market studies as these data provide rich information on the characteristics of the jobs posted(includinggranularregionalinformation,anddetailedoccupationalinformation).3 The useofonlinejobpostingsdatahasbecomeprevalentalsointherecentliteraturethatstudiesthe labormarketimpactoftheCOVID-19shockasthesedataareavailableatveryhighfrequency. 3For example, Hershbein and Kahn [2018] use online vacancy postings to document how skill requirements changedinresponsetotheGlobalFinancialCrisisshock, MarinescuandWolthoff [2020]employdatafroman onlinejobboardtostudywhathigh-wagejobpostingsimplyforjobsearch, whileBrownandMatsa[2020]use similardatatoanalyzehowhousingmarketconditionsimpactedjobsearchbehaviourduringtheGreatRecession. 7

FromtheIndeeddatabaseweobtaindetailedjobtitlesoftheindividualjobpostingsaswell asinformationonthe regionanddateofeachpostingfromJanuary 2019 onwards. WeaggregatejobpostingsattheMSAlevelandatmonthlyfrequencyandweemployaseriesofmatching algorithms to map Indeed job titles into 4-digit 2008 International Standard Classification ofOccupations(ISCO-08)occupationcodes. Weobtainapproximately60millionvacanciesfor 2019and2020.Toclassifyoccupationsintodigitalandnon-digital,wefollowcloselyMuroetal. [2017] and Soh et al. [2022] and compute a ranking of occupation codes by their digital content based on O*NET. Specifically, we create a digital score for each occupation based on two measuresoftheO*NET2019vintage: (i)ameasureoftheoverallknowledgeofcomputersand electronics required by a job and (ii) a measure of the importance of working with computers forajob. Thesetwomeasuresaimtocapturethelevelandimportanceofdigitalskillsperoccupation. Weclassifyoccupationsasdigitaliftheirscoreisabovethe50th percentileofthedigital scoredistribution,withtheremainingoccupationsclassifiedasnon-digital.4 4 Mobility, IT and Unemployment across US States In this section, we ask whether the impact of the onset of the COVID pandemic on US states’ labormarketsisaffectedbylocalfirmITadoption. Figure 1 shows that the extent of job losses is correlated with the decline in mobility only in those states where their firms utilize a relatively low level of IT. In states where firms are relativelystrongadoptersofinformationtechnology,theincreaseinunemploymentshowslittle relationship to the degree to which mobility fell. For instance, both Colorado and Nevada experiencedadeclineinmobilityof(abitmorethan)40%. However,theincreaseoftheunemploymentratewastwiceaslargeinNevada,whichisalow-ITadoptionstatethaninColorado, whichisahigh-ITadoptionstate. AnanalogouspatternemergesforthecorrelationbetweenthestringencyoflockdownpoliciesandtheincreaseintheunemploymentrateovertheperiodbetweenFebruarytoApril2020. 4ExamplesofISCO-08occupationcodesatthebottomdecileofthedigitalscoredistributionincludehomebasedpersonalcareworkers, bricklayersandrelatedworkers, carpentersandjoiners. Examplesofoccupations atthetopdecileincludewebtechnicians,systemsadministrators,informationandcommunicationstechnology service managers and software developers. Examples of occupations in the middle 10% include psychologists, employmentagentsandcontractorsandnursingassociateprofessionals. Sinceourdigitalscoresarebasedona pre-pandemicvintageofO*NET,theoccupationalrankingdoesnotreflectchangesindigitalizationwithinoccupationcodesthatmayhaveoccurredduringthepandemic. FormoredetailsonourmethodologyseeSohetal. [2022]). 8

Thereisapositivecorrelationbetweentheseverityofmitigationpoliciesandtheincreaseofunemploymentonlyamonglow-ITadoptionstates(FigureA1). These results suggest that more IT-oriented states appear better able to shift quickly to a sociallydistantenvironmentand,indoingso,maintaintheirworkforce. Totestforthedifferencebetweenhigh-andlow-ITstatesintheresponseofunemployment ratetothemobilitydecline,weestimatethefollowingequation: ∆UR =α+β ∆Mobility +β IT +β ∆Mobility ∗IT +X (cid:48)σ+(X ∗Mobility ) (cid:48)γ+(cid:178) (2) s 1 s 2 s 3 s s s s s s where ∆UR is the change in the unemployment rate in state s between April and February s 2020. ∆Mobility istheaveragedeclineinmobilityinstatesinAprilandIT isadummythat s s indicateswhetherastateisabovethemedianintermsofITadoptionandzeroifitisbelowthe median. X includes the level and the interaction between mobility and GDP per capita, the s populationdensityandthemanufacturingshareofthestateascontrolvariablesintheregressions. β whichisourmaincoefficientofinterestisequivalenttotestingthedifferenceinthe 3 slopebetweenhighandlowITadoptingstatesinFigure1. Table1reportstheresults.WefirstestimateasimplifiedversionofEquation2thatregresses thechangeintheunemploymentrateontheITadoptiondummy. AhigherlevelofITadoption is associated with a lower increase in the unemployment rate: a state in which firms adopt IT more strongly saw a 1.8 percentage points weaker increase in the unemployment rate relative tostateswherefirmsarenotadoptingITasheavily. Column(2)thenshowsthatonaverage,alargerdropinmobilityisassociatedwithastronger increaseintheunemploymentrate. A10percentagepointsstrongerdropinmobilityisassociatedwitha1.5percentagepointsstrongerincreaseintheunemploymentrate. Column (3) reports estimates of our full specification, which includes the interaction betweentheITdummyandthechangeinmobility. Thecoefficientontheinteractionispositive and statistically significant. The coefficient on ∆Mobility indicates the correlation between thechangeinmobilityandtheincreaseintheunemploymentrateforlowITstates. Thecoefficientisnowmuchlargerthanincolumn(2)whichreflectedtheaverageeffectacrossbothhigh andlowITadopters. ForlowITadopters,a10percentagepointslargerdeclineinmobilitywas associatedwitha5percentagepointslargerincreaseintheunemploymentrate. Forinstance, in the case of Michigan mobility declined by around 40% while in Ohio mobility declined by 9

30%; both are low IT states. Ohio saw its unemployment rate rising by around 13 percentage points while Michigan’s unemployment rate rose by approximately 18 percentage points, a 5 percentagepointsdifferencewithrespecttoa10percentagepointsdifferenceinthedeclinein mobility(seeFigure1). Thecoefficientontheinteractionispositive, whichindicatesthatinhighITstatestheimpact of mobility on unemployment is more muted. The point estimate of the interaction is 0.463,closeinabsolutevaluetothecoefficientonthemobilitycoefficient.Thisindicatesasmall or negligible impact of mobility in high IT states; the sum of the coefficient (-0.505+0.463=- 0.042)reflectstheslopeofhighITadoptersinFigure1. ApotentialexplanationforwhyhighITstatesexhibitaweakercorrelationbetweenmobility andtheunemploymentcouldbethatthesestatesaredifferentfromlowITonesforsomeother reasons. This problem is known as omitted variable bias. For instance, states in which firms adopt more technology may just be more economically developed and thus more resilient to economic shocks. Hence, in column (4) we include the GDP per capita, the population density, and the manufacturing share of the state as control variables in the regressions. We also includetheinteractionofeachcontrolwiththemobilitydrop:inthiswayweallowstateswhich are richer, more educated, or less dense to be affected by the pandemic differently. We then focus our attention to the coefficient of the interaction between IT adoption and mobility. If thiscoefficientweretodeclinesubstantiallyandloseitsstatisticalsignificance,wewouldinfer thattheestimatedimpactofITadoptionasamitigatingfactorisprobablydrivenbyspurious correlation. However,thecoefficientontheinteractionincolumn(4)remainsalmostidentical. Becauseofthesmallsamplesize(N=51),itisdifficulttoincludeamuchrichersetofcontrols. Nonetheless, our results suggest that such key demographic factors are not the drivers of the mitigatingimpactofITontherisingunemploymentrate. We also investigate how the results change when we vary the cutoff for labeling a state as high or low IT. As illustrated by Table A3 (column 3 in particular), states in the top quartile of theITdistributionareshieldedfromtheimpactofmobilitychanges,whilestatesinthemiddle orthebottomoftheITadoptiondistributionarenot. 4.1 EventStudyDesign Acomplementaryapproachtoanalyzethedata,istorelyonthepaneldimensionandestimate thefollowingeventstudy(two-wayfixedeffects)specification: 10

UR s,t =α s +α t + (cid:88) 1(t =τ)·∆Mobility s ·(cid:161)β τ +β τ,3 ∗IT s (cid:162)+(cid:178) s,t (3) τ(cid:54)=τ∗ whereUR istheunemploymentrateinstatesinmontht,whileIT isthecontinuouspres,t s pandemicITadoptioninthe same state, and∆Mobility is themobility shock (theaverage s,t change in mobility in April and May). Both IT and ∆Mobility are standardized for ease s s,t of interpretation. α and α are state and month fixed effects, which allow us to control for s t time-invariantlocalcharacteristicsandnational-leveltime-varyingshocks. Thecoefficientsβ τ capturetheimpactofthechangeinmobilityonunemploymentrateinthemonthτ,(τ∗isthe omittedmonth,February2020)whilethecoefficientsβ τ,3 capturetheshieldingimpactoflocal ITadoption. The estimated coefficients are reported, together with 95% confidence intervals, in Figure 2.5 Panel (a) illustrates that states which were hit more harshly by the pandemic were not onadifferentpathbeforeFebruary2020,butexperiencedasharperincreaseintheunemployment rate. However, as illustrated by Panel (b), the impact of the shock is smaller for states wherefirmsadoptedmoreITbeforethepandemic.Tovisualizesuchheterogeneity,Tovisualize the heterogeneity in the response of unemployment in high versus low IT states, Panel (c) reportstheestimatedimpactovertimeofaone-standarddeviationmobilitydropinastateabove andbelowthestandardsizedITmean. Thisalternativespecification,whichallowsustocontrol for local observable and unobservable (fixed over time) characteristics through fixed effects, confirms the findings of the previous subsection and highlights the absence of pre-pandemic differentialtrends. 5 Evidence from Individual-Level Data The state-level analysis suggests firm IT adoption can partially shield the local economy from theimpactofthepandemic. Whileinsightful,thisanalysishasimportantdrawbacks: thesmall sample size limits our ability to control for other potential confounding factors, in analyzing whichworkersaremoreprotectedbyITadoption. We therefore use individual-level data from CPS to control for respondent- and local-level 5ThemodelisestimatedbyOLS.RecenteconometricliteraturehashighlightedthatOLScanprovidebiased estimatesoftwo-wayfixedeffectswhenthetimeoftheshockortreatmentisdifferentacrossunits[Goodman- Bacon,2021;CallawayandSant’Anna,2021]. ThisislikelytobeaminorconcerninoursettingasallMSAare impactedatthesametime,buttheintensityoftheshockisdifferent. 11

characteristics. WealsocomputelocalITadoptionatafinergeographicallevel(MSA),inorder tomeasuremorepreciselytechnologyadoptionfortheindividual’srelevantlabormarket. Thisanalysisreliesonthefollowinglinearprobabilitymodel: Unemployed =α+β ∆Mobility +β IT +β ∆Mobility ∗IT i,t 1 msa(i),t 2 msa(i) 3 msa(i,t) msa(i) (4) +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α +(cid:178) i msa(i) msa(i) msa(i),t s(i) i,t where Unemployed is a dummy that equals one if the individual is unemployed, but i,t in the labor force, in a month t, where t is either April or May 2020, the height of the unemployment rate during the pandemic. The variableUnemployed is zero if the individual is i,t employedinmonth t. ∆Mobility isthechangeinmobilityintheMSAwheretheindimsa(i),t viduallives, and IT isthelevelofITadoptionintheMSAwheretheindividuali lives. X msa(i) capturesMSA-levelcontrolsandincludesthelevelandinteractionbetweenmobilityandGDP percapita,theshareofminorities,theshareofpeoplewithathreeyearBachelor’sdegreeand theunemploymentrateinFebruary 2020. Z areindividuallevelcontrols. α arestatefixed i s(i) effects. StandarderrorsareclusteredattheMSAlevelandtheregressionsareweightedbythe assignedweightoftherespondent. Thisspecificationthuscomparesworkerswiththesamesocio-demographiccharacteristics, living in different cities which are similar in various characteristics–and are within the same state–buthavedifferentdegreesofpre-pandemicfirmITadoption.6 Table 2 shows the results based on a pooled linear regression across individuals reporting their employment status in either April or/and May (Table A1 reports the results of the same equation using a probit model). These results illustrate the same pattern documented by the state-levelanalysis. Column(1)showsthatastrongerdeclineinmobilityinanMSAisassociatedonaveragewithalargerprobabilityofapersonreportingtobeunemployed.Ahigherlevel ofITadoptionisassociatedwithalowerprobabilityofbeingunemployedinAprilandMayof 2020. Column (2) shows that the probability of being unemployed in April and May is higher forrespondentslivinginMSAswhichexperiencedlargermobilitydeclines,butITadoptionof companies mitigates this impact. The increase in the probability of being unemployed asso- 6AsthepanelcomponentofCPSislimited,andrespondentsarenotnecessarilyreportingtheiremployment statusinconsecutivemonths,wedonotincludeindividualfixedeffects.Inarobustnessexercise,describedbelow, wefocusonlyonindividualswhowereemployedbeforethepandemic. 12

ciated with a large drop in mobility (one standard deviation, equal to 10 pp) is 2.4 percentage pointsinalow-ITMSA.AonestandarddeviationlargerlevelofITadoptioninanMSAreduces the increase in the probability by 0.7 percentage points to 1.7 percentage points. Column (3) shows that the coefficient remains stable and statistically significant after controlling for the interactionofthemobilityintheMSAandvariousMSA-levelcharacteristicssuchaspercapita income,theshareofpeoplewithathreeyearBachelor’sdegree,theshareofminorities,andthe unemploymentrateinFebruary. In column (4) we saturate the specification with additional fixed effects. The fixed effects includeindividualfixedeffectsbasedongender,race,andeducationlevel,aswellasstatefixed effects. Theinclusionofstatefixedeffectsimpliesthatcomparingtwoindividualslivingwithin thesamestatebutindifferentMSAsaredifferentiallyaffectedbyamobilitydeclineduetodifferentlevelsofITadoptionintheMSA.Theresultholdswhencomparingindividualswithalso thesamegenderorrace,orwithinthesameeducationlevel. Moreover,thecoefficientontheinteractionbetweenmobilityandITremainsstableafterincludingtheseadditionalsetsoffixedeffects,buttheR-squaredincreasesfrom0.418%to3.8%. TheincreaseintheR-squaredconfirmsthattheadditionalcontrolvariablesarehighlyimportantforexplainingtheemploymentstatusoftheindividualbutevenaftercontrollingforthese characteristics the level of IT adoption in the MSA remains a significant predictor of whether thepersonwasunemployed.ThecoefficientonITturnsfromnegativetopositiveassoonaswe includetheinteractionterm.Thisflipinthecoefficientispurelymechanical.Thecoefficienton ITcanbeinterpretedasthehypotheticaleffectofITontheprobabilityofbeingunemployedin anMSAwheremobilityhasnotchanged.AsmobilitydeclinedstronglyinallMSAs,theeffectof ITontheprobabilityofbeingunemployedisnotinterpretable(andthereforeomittedinmost ofthefollowingexercises). Robustness We conduct several robustness tests, reported in Table A2, all of which confirm our main findings. Column (1) shows the baseline equation for reference (similar to column (3) of Table 2). In column (2) we replace our measure of IT adoption with the share of highspeed internet that is available in the MSA. The interaction is, as for our IT measure, positive andstatisticallysignificant, butonlyatthe5%level. Incolumn(3)wereplaceourcontinuous measureofITwithadummythattakesthevalueoneiffirmsintheMSAareabove-medianIT adopters and zero if firms in the MSA are below median IT adopters. Again, the coefficient is 13

positiveandstatisticallysignificant. Column(4)replacesthebaselineITmeasure, logITbudget per employee, with another measure that has been used commonly in the literature, also fromtheHarteAberdeed/Hanksdataset, namelytheratioofpersonalcomputersperemployees[Bloometal.,2012]. Next,wesubstituteourleft-hand-sidevariable,thedummyindicating whetherthepersonisunemployed,tocaptureabroadermeasureofunemployment. Ourbaseline unemployment rate is the U-3 unemployment rate, which is the official one. It takes into account people who are jobless but actively seek employment. In column (5) instead, we use the U-6 unemployment rate definition that accounts for anyone who has been seeking employment for at least 12 months but left discouraged without being able to secure a job. This measurealsoincludesanyonewhohasgonebacktoschool,becomedisabled,andpeoplewho areunderemployedorworkingpart-timehours.Ourresultsremainrobusttousingthisbroader unemploymentmeasure. TofurthercontrolfordifferencesinlocaleconomicstructureacrossdifferentMSAs,weadd asetofcontrolsfortheshareofemploymentindifferentoccupationsanddifferentindustries attheMSA-level(wefocusonthelargest2-digitNAICSsectorsandthelargest2-digitSOC2018 occupationswhichaccountedformorethanonethirdofnationalemploymentin2019). Both theleveloftheindustryandoccupationemploymentsharesaswellastheirinteractionwiththe mobilityshockareaddedascontrols. Theinclusionsofsuchcontrolshaslimitedimpactonthe estimatedshieldingeffectoflocalITadoption,asreportedbycolumn(6). WethenfocusonrespondentsthatwereintheCPSalsoinFebruary2020,toinvestigatethe impactofmobilityandlocalITamongindividualsthatwereemployedinthatmonth. (Inthis way, the empirical specification investigates the impact of local mobility and IT on the probability that an individual becomes unemployed.) The estimating sample shrinks considerably both because of the rotating panel structure of the survey and because only about 60% of respondentswereemployedinFebruary2020. WefindthattheshieldingimpactofITispresent amongtheworkerswhowereactuallyemployedbeforethepandemic(column7). Focusingon therespondentswhoworkedinFebruary2020, wecanalsoincludetwosetsoffixedeffectsto controlforthe(4digit)occupationandindustryinthatmonth. Theinclusionofsuchcontrols does not change the estimated shielding effect of IT, mitigating the concern that the local IT effectiscapturingdifferencesinthelocalsectoralandoccupationalmix. WefinallyinvestigatewhethertheresultsarealsorobusttoMSA-levelaggregation. Weconstruct MSA-level unemployment rate and we aggregate all individual-level controls of Equa- 14

tion 4. We then estimate an MSA-level version of Equation 4 where we regress the change in unemploymentratebetweenFebruaryandAprilorMay2020onthechangeinmobility,MSAlevelIT,andtheinteractionbetweenthetwovariableswhilecontrollingfortheothercovariates. Results,presentedinTableA4,areinlinewiththerespondent-levelestimates:MSAswheremobilitydroppedmoreexperiencedastrongerincreaseinunemploymentrate,butlesssoifthey havefirmsthatadoptedmoreITbeforethepandemic. 5.1 InstrumentalVariableApproach IT adoption can be correlated with many other local characteristics. For instance, in areas where the complementarities between workers’ human capital and IT adoption are higher, more IT is adopted more intensely [Beaudry et al., 2010]. In our regression analysis, we control for various characteristics that are likely correlated with IT adoption–such as the share of the population with a bachelor’s degree or the industry composition– and our results are insensitive to the inclusion of these controls. However, it is difficult to completely rule out the presence of unobserved confounding factors which are correlated with IT and also limit the economicharmofthepandemic. Suchfactorscouldbiasourestimates. Wethereforeadoptaninstrumentalvariableapproach,relyingoncharacteristicsofthelocallabormarketthatpredatetheoriginsofthedigitalrevolution,i.e. whencomputersbecame widely available for the local adoption of IT. When computer equipment prices started falling strongly, it became more and more attractive to replace routine workers with IT equipment. Duringtheendofthe20thcentury,USregionsthatwerehistoricallyspecializedinroutineintensive occupations (e.g. butchers or payroll and timekeeping clerks) indeed experienced a largerworkplacecomputeruseafter1980[AutorandDorn,2013]. We closely follow Autor and Dorn [2013] who argue that the measure of historical routine employment shares can be seen as an exogenous shifter of IT adoption, as they are unlikely toaffectemploymentoutcomestodaythroughotherchannelsotherthantechnology. Wetest whetherhistoricalvariationinroutinetasksharesattheregionallevelpredictsITadoptionjust before the Covid-19 pandemic. To measure routine tasks the job task requirements from the fourtheditionoftheUSDepartmentofLabor’sDictionaryofOccupationalTitles(DOT)(USDepartmentofLabor1977)aremergedtotheircorrespondingCensusoccupationclassifications [Autoretal.,2003]. Thenforeachcommutingzone,aroutineemploymentshareiscreated. We directlytakethedatafromAutoretal.[2015]onthecommutingzonelevelandapplytheshare 15

ofroutineworktoeachcountywithinthatcommutingzoneandthenaverageacrossMSAs. Figure3showsthatthereisastrongpositivecorrelationbetweentheemploymentsharein routinetasksin1980andthelevelofITadoptionjustbeforethepandemic.Undertheexclusion restrictionthattheoccupationalstructurein1980affectstheemploymentoutcomesduringthe pandemic only through higher IT adoption and not through other channels, we can use the shareofroutineemploymentinaregionasaninstrumentforITadoptionbeforethepandemic, whichallowsustoestimatethecausaleffectofITadoptiononemploymentoutcomes. Were-estimatethelinearprobabilitymodel: Unemployed =α+β ∆Mobility +β IT +β ∆Mobility ∗IT +α +(cid:178) i,t 1 msa(i),t 2 msa(i) 3 msa(i),t msa(i) s(i) i,t (5) whileinstrumentingtheendogenousvariablesIT andMobility ∗IT withthe msa(i) msa(i,t) msa(i) excludedinstrumentsRoutine and∆Mobility ∗Routine . msa(i) msa(i,t) msa(i) Weperformestimationviatwo-stagesleastsquare. Theestimatesforthecoefficientofinterests (β , which refers to the interaction term between IT and the change in mobility) are 3 reportedinTable7. Incolumn(1)wereporttheOLSestimate. Incolumns(2)-(5)weestimate the2SLSspecificationwithtwoendogenousvariablesandvaryingsaturationofthemodelswith controlsandfixedeffects. The coefficient on the interaction term between IT and mobility is positive in all specifications, confirming our previous result that IT adoption can mitigate the adverse economic consequencesinresponsetoamobilitydecline. However,thecoefficientissmallerintheOLS specification than in the IV estimates, although not statistically different, as shown in the row P−value=OLS. As we have two endogenous variables, the conventional first-stage F-stage statistic is not appropriate to test for the strength of the instrument [Angrist and Pischke, 2008]. Instead, we reporttheSandersonandWindmeijer [2016]F-statisticsformodelswithmultipleendogenous variablestotestforweakinstruments. ThetwoF-statisticsforthefirststageforITitselfandthe interactionrangebetween7and30. Incolumns(3)and(4)theF-statsforbothfirststagesare allabove15,abovetherule-of-thumbthresholdof10. Inconclusion,theIVestimatesconfirmthatITadoptionhasacausalimpactonmitigating theadverseemploymentoutcomesinresponsetorestrictionsinmobility. Therefore,thefindingthatlabormarketsinstatesorMSAswherefirmsadoptedmoreITwerealsomoreresilient 16

tothepandemicisnotmainlydrivenbythepresenceofunobservedconfoundingfactors. 5.2 Counterfactual InaninterviewwithTheEconomist,BillGatesarguedthat“if[thepandemic]wouldhavecome 5 years earlier that would have been a disaster", referring to the economic damage due to a “crappy online experience". Other commentators have also highlighted that if the pandemic hadhappenedinthepast–evenintherecentpast–theabilityofcompaniesandworkertoquickly scale the use of working-from-home, contactless delivery, and other remedies needed to respondtosocialdistancingwouldhavebeensignificantlylessdeveloped. Theimprovementsin IT,internetinfrastructure,thewidespreaduseofsmartphonesanddeliveryapps,havebeenof greathelp. We can use our estimates to compute the counterfactual labor market consequences that would have occurred given a lower level of IT adoption. To perform such an exercise, we reestimate Equation 4 without normalizing the measure of IT adoption; non-normalized coefficients are expressed in terms of IT expenses per employee (rather than in terms of crosssectionalstandarddeviationasinsection4andsection5). BureauofEconomicAnalysis[2019] revealsthat“since2010,digitaleconomyrealgrossoutputgrowthaveraged2.5percentperyear.”, whilethegrowthrateofthelaborforceisabout0.5percentperyear.7 Thus,weassumethatIT adoptiongrowsat2percentagepointsperyear,andwas,therefore,approximately10%smaller 5yearsago. WealsoassumethatthegrowthrateofITishomogeneousacrossallMSAs. Under the assumptions described above, we can estimate the counterfactual probability thatanindividuali isunemployedas: Unemployed i,t =α+β (cid:99)1 ∆Mobility msa(i),t +β (cid:99)2 ∗0.9∗I(cid:225)T msa(i) +β (cid:99)3 ∆Mobility msa(i,t) ∗0.9∗I(cid:225)T msa(i) +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α i msa(i) msa(i) msa(i),t s(i) (6) 7Expensesininformationtechnologyarethemainbutnottheonlycomponentofthedigitaleconomy,asdefinedbytheBEA.BureauofEconomicAnalysis[2019]specifiesthat“BEAincludesinthedigitaleconomytheentire informationandcommunicationstechnologies(ICT)sectoraswellasthedigital-enablinginfrastructureneededfor acomputernetworktoexistandoperate,thedigitaltransactionsthattakeplaceusingthatsystem(“e-commerce”), andthecontentthatdigitaleconomyuserscreateandaccess(“digitalmedia”)”.However,aslongaseithertheother partsofthedigitaleconomygrowatthesamerateasITadoption,ortheyaresimilarlycorrelatedtounemployment,wecanstillequatethegrowthrateofITexpensestooneofthemorebroadlydefinedmeasuresof“digital economy”. 17

where the “hat” signs highlight that the IT adoption measure and the coefficients are not normalized. The estimated counterfactual unemployment rate (average between April and May 2020) underthe2015ITadoptionis16%versustheobserved14%. Itistherefore2percentagepoints (or14.3%)higherthanwhatwasobservedinthedata. Theestimatesfromalinearmodelmay overestimate the counterfactual impact of a large change in IT adoption if non-linearities are important. It is therefore reassuring that using a probit model (instead of a linear probability model)providesthesameresults. ThisfindingillustratestheimportanceofinvestmentsinIT adoptiontobuildaneconomythatisnotonlyfaster-growingbutalsomoreresilienttoshocks. Thisback-of-the-envelopecalculationshouldbetreatedwithcaution. AlthoughourIVestimateandourcoefficientsaftercontrollingforvariousobservablecharacteristicsarerelatively stable,wecannotcompletelyruleoutpotentialexclusionrestrictionviolationsorthatotherunobservable or omitted characteristics which are correlated with IT spending partially bias our coefficient of interest. Moreover, this type of calculation assumes that there are no spillover effects from the adoption of IT. If, for example, IT spending in one region makes not only the regionitselfmoreresilient,butalsootherregionsthatdonotadoptITasstronglymoreresilient – for instance via a smaller decline in aggregate demand– our estimate would provide a lower bound for the total effect of how much IT shields unemployment losses. See also Nakamura andSteinsson[2018]foranin-depthdiscussionofthecaveatsofextrapolatingaggregateeffects fromcross-sectionalregressions. 5.3 ITandInequality DoesITshieldallworkersfromtheimpactofthepandemic? Wetestwhetherthemitigatingeffectoflocalfirms’ITadoptiononworkers’labormarketoutcomesdependsontheircharacteristics,suchasgender,race,andeducationalattainment. Tothisaim,weestimatethefollowing linearprobabilitymodel: 18

Unemployed =α+β ∆Mobility ∗A +β ∆Mobility ∗(1−A ) i,t 1 msa(i),t i 2 msa(i),t i +β IT ∗A +β IT ∗(1−A ) 3 msa(i) i 4 msa(i) i +β ∆Mobility ∗IT ∗A (7) 5 msa(i),t msa(i) i +β ∆Mobility ∗IT ∗(1−A ) 6 msa(i),t msa(i) i +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α +(cid:178) i msa(i) msa(i) msa(i),t s(i) i,t whichissimilartoEquation4exceptfortheadditionoftheinteractiontermsdisciplinedby A ,whichisadummyvariableequaltooneifrespondenti belongstoacertaincategory.Inpari ticular,weestimateEquation7forthreedifferentcharacteristics:gender,race,andeducational attainment. First, we estimate the regression equation for gender, where A = 1 is one if the i respondent is male and A =0 if the respondent is female. Second, we estimate the equation i forethnicitywhere A =1iftherespondentiswhiteand A =0iftherespondentisnon-white. i i Third, A =1 ifthe individualhasahigh- or mediumlevel ofeducation(highschool or more) i and A =0 if the individual has no high school degree. (Observation where the relevant catei gorical variable is missing are dropped.) The remainingvariables are defined asabove, where thevectorZincludesthevariouscategoriesasdummies. Table 3 presents the results for β and β . The coefficient is positive for males, females, 5 6 whites,non-whites,andhigh/mediumeducation. Onlyinthecaseoflow-educationindividuals,wedonotfindamitigatingimpactofITontheeffectofmobilityontheprobabilityofbeing unemployed. The coefficient β and β are also plotted in Figure 4. Interestingly, the effect is largest for 5 6 femalesandnon-whiteindividuals. Theseareamongtheindividualswhicharemosthitduring thefirstphaseofthepandemicandITadoptionhasmoreroomtomitigatetheshockforthese individualsratherthanforexamplehighly-educatedoneswhoseunemploymentrateshavenot respondedasstronglytothedeclineinmobility. Loweducatedindividuals,however,although hitveryharshlyfromthepandemicarenotshieldedbyfirmIT. Overall,eventhoughITadoptionmay—intheaggregate—significantlyshieldlabormarkers againsttheeffectsoftheCOVID-19pandemic,itmayalsocontributetowideninginequalityby increasingeconomicdisparitiesbetweenhigh-andlow-educatedindividuals. 19

6 Channels In this section we analyze various channels through which IT adoption could have mitigated the adverse consequences of social distancing. In particular, we test whether IT adoption is associatedwithbetterabilitytoworkfromhome,highere-commerceactivity,greaterresilience ofjobcreation,aswellasareallocationoflabordemandfromnon-digitaltodigitaljobsduring Spring2020. 6.1 IT,Working-From-Home,andE-commerce One potential reason why low-educated individuals are not shielded by IT adoption is due to skill-biasedtechnologicalchange. Moreskilledworkershavelargercomplementaritieswithinformationtechnologiescomparedtolower-educatedworkersforwhichITmayevensubstitute theirwork. High-skilledindividualshavebeenabletoswitchtoworkfromhomewithlittleadjustmentnecessary. DingelandNeiman[2020]showthataround1/3ofallworkerscandojobs fromhome,ofwhichmostofthemarehigher-educatedworkers. OnepotentialexplanationforourresultsisthereforethatITadoptionandwork-from-home abilitiesarehighlycorrelatedandthereasonwhyindividualslivinginareaswherefirmsadopt ITmoreheavilyarealsoareaswheremorepeoplecanworkfromhome. Indeed,Figure5shows thereisahighcorrelationbetweentheshareofjobsthatcanbedonefromhomeinanMSAand ITadoption. We re-estimate Equation 4 substituting the IT measure with the share of jobs that can be donefromhometotestwhethertheworkfromhomeabilitiescanalsoshieldworkersfromthe declineinmobility. Table 5 shows the results. The results for WFH mirror those of IT, in line with the results byBaietal.[2021]. IndividualslivinginMSAswhereWFHismorefeasiblearelesslikelytobe unemployed for a given decline in mobility than individuals who live in areas where WFH is not as widely possible. Column (3) shows the results with both interactions, between IT and mobilityandbetweenWFHandmobility. Bothcoefficientsremainstatisticallysignificant,but thecoefficientdeclinesinbothcases. The fact that the coefficient on the interaction between IT and mobility declines once the interactionbetweenWFHandmobilityisincludedintheregressionsuggeststhatWFHisone channelthroughwhichITshieldsworkersfromtheeconomicconsequencesofthepandemic. 20

However,thecoefficientontheinteractionremainsstatisticallysignificant,suggestingthatteleworking does not seem to be the only channel through which IT has a mitigating effect and otherchannelsthroughwhichITadoptionmitigatestheconsequencesofsocialdistancingare atwork. Another potential channel for the shielding effect of local IT adoption could be that ITsavvy firms have better e-commerce capabilities and thus can more promptly expand online sales. TheAberdeensurveycontainsmoredetailedinformationonthepresenceofspecificecommerce technology for about 1% of the sample establishments. For these establishments, weknowwhetherornottheyhaveadoptedae-commercerelatedtechnologyin2016. WeconstructanMSA-levelmeasureofe-commercepresencebyestimatingthesameregressionused toestimatethebaselinemeasureofIT(Equation1).8 Wethenaugmentthebaselineindividual-levelspecificationwiththeMSA-levelmeasureof e-commerce prevalence and its interaction with the change in mobility. Results are reported in Table 5. We do not find evidence in favor of a significant shielding impact of pre-COVID ecommercetechnologiesonlocallabormarkets. Theempiricalirrelevanceofthischannelmay besurprisinggiventheriseofonlinesalesduringtheonsetofthepandemic.Weconjecturetwo reasons that could justify this finding. First, it may be easier for IT-savvy firms to start selling onlineoncethepandemichit,eveniftheydidnotdosobefore. Second,asweshowbelow,the shieldingimpactofITisparticularlyimportantintradableindustries,butnotinnon-tradable ones. Firmsintradableindustries, likemanymanufacturingorminingindustries, tendtosell tootherbusinessratherthanconsumers,thuslimitingtheimportanceofe-commerce. 6.2 IT,OnlineVacancies,andDigitalJobs The abrupt skyrocketing of the unemployment rate at the onset of the pandemic indicates a severe increase in (temporary) layoffs. While job destruction was a key driver of the unemployment rate, depressed job creation was another important margin of adjustment. Labor demandcollapsedintheSpringof2020asfirmsrespondedtomobilityrestrictionsandtheextraordinary degree of uncertainty by severely restricting vacancies. Mobility restrictions may 8Thatis,weestimatetheMSA-levelfixedeffectα fromthelinearprobabilitymodel g(e) Ecommerce =δ+α +θ +(cid:178) (8) e g(e) ind(e) i whereEcommerce isanindicatorvariableflaggingthepresenceofe-commercetechnology. e 21

have affected firms’ ability to create new posting not only due to the contraction in aggregate demand that they entailed but also by exacerbating search frictions as in-person interactions wererarer. However,theimpactoftheCOVID-19shockonjobcreationmayhavebeenasymmetric across regions with different degrees of IT adoption. IT-adopters may have benefited fromhigherqualityandmorereadilyavailabledigitalinfrastructureandgreaterdegreeofdigital preparedness which could allow firms to flexibly adapt their working practices and shield job creation. In contrast, lagging regions may have suffered from lower digital infrastructure and may have struggled to digitalize their business models and work practices with stronger negativeeffectsforjobcreation. We test whether part of the shielding impact of IT comes from enhancing the resilience of job creation. We focus on online job posting as online job search became an increasingly importantwaythroughwhichfirmspostedvacanciesinthepresenceofmobilityrestrictions.9 Weestimatethefollowinglinearregression: ∆JobPosting =α+β ∆Mobility +β IT +β ∆Mobility ∗IT msa 1 msa 2 msa 3 msa msa (9) +X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+(cid:178) msa msa msa msa where∆JobPosting istheaveragechangeintheloglevelofvacanciesbetweenFebrumsa ary2020andMayorApril2020attheMSAlevel. ∆Mobility istheaveragedeclineinmomsa bility between February 2020 and April or May 2020 in each MSA and IT measures local msa IT adoption. The coefficient β captures the impact of the mobility drop on vacancy postings 1 whileβ capturestheshieldingimpactoflocalITadoption. 3 Results are reported in Table 6. Aggregate online job postings dropped more in areas that suffered a larger decline in mobility during Spring 2020, as reported in column (1). However, consistentlywiththeresultspresentedinsection5,thenegativeimpactofthemobilitydropon job postings is mitigated in areas with stronger pre-COVID IT adoption (as shown in column (2)). TheseresultslendsupporttothehypothesisthattheshieldingimpactoffirmITadoption onlocallabormarketsisdrivenalsobyincreasingtheresilienceofjobcreation,ratherthanonly becauseoflessseverejobdestruction. Toshedfurtherlightonthismechanism, wetestwhethershieldingtookplaceforoccupationscharacterizedbyskillsthatarecomplimentarytoIT,orwhetherITadoptionbenefitedless 9Unfortunatelyourdatadonotprovideinformationonwhetherthejoblistingresultedinafinalhiring. 22

digitally-savvyoccupationsaswell. Were-estimateEquation9separatelyfortwocategoriesof vacancies:ondigitalandonnon-digitaloccupations(seesection3). WefindthatlocalITadoptionshieldedtheimpactofthepandemiconlyondigitaljobs, asillustratedbycomparingthe coefficientontheinteractionbetweenITandmobilityincolumns(4)ofTable6,whichrefersto vacanciesfordigitaljobs,totheoneincolumn(6),whichreferstonon-digitalvacancies. While bothtypesofjobpostingareimpactedbylocalmobility(assuggestedbythepositivecoefficient onthechangeinmobilityincolumns(3)and(5)),theinteractioncoefficientisstatisticallydifferent than zero only for digital job postings. In fact, columns (7) and (8) illustrate that the shareofdigitalvacanciesovertotaljobpostingsincreasedinareasmorehitbythepandemic, andevenmoresoinareasthatalsohadahigherdegreeofpre-pandemicITadoption. ThesefindingssuggestthatanimportantreasonwhyITshieldedlocallabormarketsfrom theimpactofthepandemicisbecauseitfacilitatedandamplifiedtheexpansionofthedigital economy, helping firms to create more digital jobs. In areas hit more harshly from the pandemic, the transition to a more digital-intense economy was stronger. Importantly, in places where pre-pandemic IT adoption was higher, there was an even stronger shift in the demand towards more digitally-intensive jobs which absorbed in part the negative impact of mobility restrictionsonjobcreationduringSpring2020. Event-study design A complementary approach to analyze the data, is to rely on the panel dimensionandestimatethefollowingeventstudy(two-wayfixedeffects)specification: JobPosting msa,t =α msa +α t + (cid:88) 1(t =τ)·∆Mobility msa ·(cid:161)β τ +β τ,3 ∗IT msa (cid:162)+(cid:178) msa,t (10) τ(cid:54)=τ∗ which differs from Equation 3 only in that the unit of observation is an MSA rather than a state, and the dependent variable is the log level of online vacancies in that MSA in month t. Results, illustrated in graphical form by Figure 6, confirm the results of the cross-sectional regression (Equation 9). MSAs in which mobility dropped more severely experienced a larger declineinonlinejobpostings,howevertheimpactofmobilitywasreducedinareaswherefirms hadadoptedITmoreintensivelypre-pandemic. Thefigurealsoshowsthatthedynamicsofjob postings before the pandemic were similar for areas more or less hit by the drop in mobility. This absence of a pre-trend mitigates the concern that other confounding shocks are driving 23

theresultsandsuggeststheparallel-trendassumptionisnotviolated. Inparticular, wewould expectthatregionswithdifferentiallevelsofITadoptionwouldhaveperformedsimilarlyduringthespringof2020ifthepandemicwouldnothavehittheworld. 6.3 DemandSpillovers Our empirical investigation relies on variables measured at the local level. Therefore, part of the shielding impact of IT could come from general equilibrium effects impacting local markets. Demand spillovers are a channel of particular importance when analyzing local impact of shocks [Mian and Sufi, 2014]. For instance, if some firms are shielded because of IT and thuscanmaintaintheirworkforce,othernearbyfirmsthatsellproductsorservicestotheemployeesofthelattermayalsobenefit. Togaugetheimportanceofsuchdemandspillovers,we studythedynamicsofemploymentintradableversusnon-tradableindustriesinSpring2020. We compute MSA monthly employment by industry collapsing CPS data, and use tradable vs non-tradable industry classification from Mian and Sufi [2014]. In line with the results of our baselinespecification,wefindthatthechangeoftotalemploymentfromFebruary2020toApril andMay2020wasmorenegativeforMSAswhichexperiencedamoreseveredropinmobility, butlesssowhenITwasadoptedpre-pandemic.However,theshieldingroleofITispresentonly for tradable industries, and not for non-tradable ones.10 These results (reported by Table A5) suggestthatdemandspilloversplayaminorroleinexplainingtheshieldingimpactofITduring theonsetofthepandemic. 7 Conclusion Inthispaper,weshowthattechnologyadoptioncanactasanimportantmitigatingfactorwhen the economy is hit by a shock, and therefore our results contribute to the question of how to buildamoreresilientsociety[Brunnermeier,2021]. The dampening effect of IT adoption has important implications for the implementation oflockdownpolicies. Ourresultsimplythatthecostofthesocialdistancingislowerinplaces wherefirmsadoptITmoreheavily,reducingapotentialtrade-offbetweenhealthandtheeconomy. This implication is relevant independently of whether individuals willingly reduce their 10Note that tradable and non-tradable is not a partition of all industries according to Mian and Sufi [2014]’s classification.Sototalemploymentislargerthanthesumoftradableandnon-tradableemployment. 24

mobilityorarecompelledtodosobymorerestrictivepolicies. However,eveninhigh-ITareas,noteveryoneisshieldedfromtheeconomicconsequences of lockdowns. While IT protects people of different races and both women and men, IT does notshieldlow-skilledworkersfromtheeconomicconsequencesoftheCOVID-19shock. Overthelastdecades,low-skilledindividualshavealreadysufferedfromtheconsequences ofskill-biasedtechnologicalchange,whichseemstobereinforcedbytheCOVID-19pandemic. The large burden of the COVID-19 pandemic, which falls hardest on the less-skilled, may not only have negative economic, but also indirect health consequences over and above the directimpactofthepandemic[CaseandDeaton,2020]. Ourfindingsspeaktotheimportanceof policiestargetedtoimprovedigitalskillsfortheless-educatedpopulation,inordertopromote inclusivegrowthandwell-being. 25

References Acemoglu, D., and D. Autor, Skills, tasks and technologies: Implications for employment and earnings,2011. Adrjan,P.,G.Ciminelli,A.Judes,M.Koelle,C.Schwellnus,andT.Sinclair,Willitstayorwillitgo? AnalysingdevelopmentsinteleworkduringCOVID-19usingonlinejobpostingsdata,OECD ProductivityWorkingPapers30,OECDPublishing,doi:10.1787/aed3816e-en,2021. Akerman,A.,I.Gaarder,andM.Mogstad,Theskillcomplementarityofbroadbandinternet,The QuarterlyJournalofEconomics,130(4),1781–1824,2015. Allcott,H.,L.Boxell,J.Conway,M.Gentzkow,M.Thaler,andD.Y.Yang,Polarizationandpublichealth: Partisandifferencesinsocialdistancingduringthecoronaviruspandemic,NBER WorkingPaper,(w26946),2020. Alon, T., M. Doepke, J. Olmstead-Rumsey, M. Tertilt, et al., This time it’s different: The role of women’semploymentinapandemicrecession,Tech.rep.,UniversityofBonnandUniversity ofMannheim,Germany,2020. Alstadsæter,A.,J.B.Bjørkheim,W.Kopczuk,andA.Økland,Norwegiananduspoliciesalleviate business vulnerability due to the covid-19 shock equally well, Unpublished Working Paper, 2020. Anderson, R. M., H. Heesterbeek, D. Klinkenberg, and T. D. Hollingsworth, How will countrybased mitigation measures influence the course of the covid-19 epidemic?, The Lancet, 395(10228),931–934,2020. Angrist,J.D.,andJ.-S.Pischke,Mostlyharmlesseconometrics,Princetonuniversitypress,2008. Autor,D.,andD.Dorn,Thegrowthoflow-skillservicejobsandthepolarizationoftheuslabor market,Americaneconomicreview,103(5),1553–97,2013. Autor, D. H., F. Levy, and R. J. Murnane, The skill content of recent technological change: An empiricalexploration,TheQuarterlyjournalofeconomics,118(4),1279–1333,2003. Autor,D.H.,D.Dorn,andG.H.Hanson,Untanglingtradeandtechnology: Evidencefromlocal labourmarkets,TheEconomicJournal,125(584),621–646,2015. 26

Bai,J.J.,E.Brynjolfsson,W.Jin,S.Steffen,andC.Wan,Digitalresilience:Howwork-from-home feasibilityaffectsfirmperformance,Tech.rep.,NationalBureauofEconomicResearch,2021. Bartik,A.W.,M.Bertrand,Z.B.Cullen,E.L.Glaeser,M.Luca,andC.T.Stanton,Howaresmall businessesadjustingtocovid-19? earlyevidencefromasurvey,Tech.rep.,NationalBureauof EconomicResearch,2020. Beaudry,P.,M.Doms,andE.Lewis,Shouldthepersonalcomputerbeconsideredatechnological revolution? evidence from us metropolitan areas, Journal of Political Economy, 118(5), 988–1036,2010. Béland, L.-P., A. Brodeur, and T. Wright, The short-term economic consequences of covid-19: exposuretodisease,remoteworkandgovernmentresponse,2020. Bellatin, A., and G. Galassi, What COVID-19 May Leave Behind: Technology-Related Job PostingsinCanada,StaffWorkingPapers22-17,BankofCanada,2022. Bessen, J. E., and C. Righi, Shocking technology: What happens when firms make large it investments?,BostonUniv.SchoolofLaw,LawandEconomicsResearchPaper,(19-6),2019. Bloom,N.,Thebrightfutureofworkingfromhome,SIEPRblog,2020. Bloom, N., and N. Pierri, Cloud computing is helping smaller, newer firms compete, Harvard BusinessReview,2018. Bloom, N., R. Sadun, and J. Van Reenen, Americans do it better: Us multinationals and the productivitymiracle,AmericanEconomicReview,102(1),167–201,2012. Borjas,G.J.,andH.Cassidy,Theadverseeffectofthecovid-19labormarketshockonimmigrant employment,Tech.rep.,NationalBureauofEconomicResearch,2020. Bresnahan, T. F., E. Brynjolfsson, and L. M. Hitt, Information technology, workplace organization, and the demand for skilled labor: Firm-level evidence, The Quarterly Journal of Economics,117(1),339–376,2002. Brodeur,A.,D.M.Gray,A.Islam,andS.Bhuiyan,Aliteraturereviewoftheeconomicsofcovid- 19,2020. 27

Brown,J.,andD.A.Matsa,Lockedinbyleverage: Jobsearchduringthehousingcrisis,Journal ofFinancialEconomics,136(3),623–648,doi:10.1016/j.jfineco.2019.11,2020. Brunnermeier,M.K.,Theresilientsociety,2021. Brynjolfsson, E., and L. M. Hitt, Computing productivity: Firm-level evidence, Review of EconomicsandStatistics,85(4),793–808,2003. Brynjolfsson, E., J. J. Horton, A. Ozimek, D. Rock, G. Sharma, and H.-Y. TuYe, Covid-19 and remote work: An early look at us data, Tech. rep., National Bureau of Economic Research, 2020. BureauofEconomicAnalysis,.,Measuringthedigitaleconomy: Anupdateincorporatingdata fromthe2018comprehensiveupdateoftheindustryeconomicaccounts,2019. Cajner, T., L. D. Crane, R. A. Decker, J. Grigsby, A. Hamins-Puertolas, E. Hurst, C. Kurz, and A.Yildirmaz,Theuslabormarketduringthebeginningofthepandemicrecession,Tech.rep., NationalBureauofEconomicResearch,2020. Callaway,B.,andP.H.Sant’Anna,Difference-in-differenceswithmultipletimeperiods,Journal ofEconometrics,225(2),200–230,2021. Case, A., and A. Deaton, Deaths of Despair and the Future of Capitalism, Princeton University Press,2020. Chen,S.,D.Igan,N.Pierri,andA.F.Presbitero,Trackingtheeconomicimpactofcovid-19and mitigationpoliciesineuropeandtheunitedstates,Tech.rep.,WorkingPaper,2020. Chernoff,A.W.,andC.Warman,Covid-19andimplicationsforautomation,Tech.rep.,National BureauofEconomicResearch,2020. Chetty, R., J. N. Friedman, N. Hendren, M. Stepner, et al., How did covid-19 and stabilization policies affect spending and employment? a new real-time economic tracker based on privatesectordata,Tech.rep.,NationalBureauofEconomicResearch,2020. Chiou, L., and C. Tucker, Social distancing, internet access and inequality, Tech. rep., National BureauofEconomicResearch,2020. 28

Dave, D., A. I. Friedson, K. Matsuzawa, and J. J. Sabia, When do shelter-in-place orders fight covid-19best?policyheterogeneityacrossstatesandadoptiontime,EconomicInquiry,2020. Del Boca, D., N. Oggero, P. Profeta, and M. Rossi, Women’s work, housework and childcare, beforeandduringcovid-19,2020. Dingel,J.I.,andB.Neiman,Howmanyjobscanbedoneathome?,Tech.rep.,2020. Fairlie, R.W., K.Couch, andH.Xu, Theimpactsofcovid-19onminorityunemployment: First evidence from april 2020 cps microdata, Tech. rep., National Bureau of Economic Research, 2020. Flood,S.,M.King,R.Rodgers,S.Ruggles,J.R.Warren,andM.Westberry,Integratedpublicuse microdataseries,currentpopulationsurvey: Version9.0[dataset].minneapolis,mn: Ipums, 2021,2021. Forman, C., A. Goldfarb, and S. Greenstein, The internet and local wages: A puzzle, American EconomicReview,102(1),556–75,2012. Goodman-Bacon, A., Difference-in-differences with variation in treatment timing, Journal of Econometrics,225(2),254–277,2021. Goolsbee,A.,andC.Syverson,Fear,lockdown,anddiversion: Comparingdriversofpandemic economicdecline2020,Tech.rep.,NationalBureauofEconomicResearch,2020. Gottlieb, C., J. Grobovšek, and M. Poschke, Working from home across countries, Covid Economics,1(8),71–91,2020. Heathcote, J., F. Perri, G. L. Violante, et al., The rise of us earnings inequality: Does the cycle drivethetrend?,Tech.rep.,FederalReserveBankofMinneapolis,2020. Hensvik,L.,T.LeBarbanchon,andR.Rathelot,JobsearchduringtheCOVID-19crisis,Journal ofPublicEconomics,194(C),doi:10.1016/j.jpubeco.2020.10,2021. Hershbein, B., and L. B. Kahn, Do recessions accelerate routine-biased technological change? evidence from vacancy postings, American Economic Review, 108(7), 1737–72, doi:10.1257/ aer.20161570,2018. 29

Kahn, L. B., F. Lange, and D. G. Wiczer, Labor demand in the time of covid-19: Evidence from vacancypostingsanduiclaims,Tech.rep.,NationalBureauofEconomicResearch,2020. Kirby,T.,Evidencemountsonthedisproportionateeffectofcovid-19onethnicminorities,The LancetRespiratoryMedicine,8(6),547–548,2020. Maloney, W., and T. Taskin, Determinants of social distancing and economic activity during covid-19: Aglobalview,2020. Marinescu,I.,andR.Wolthoff,OpeningtheBlackBoxoftheMatchingFunction: ThePowerof Words,JournalofLaborEconomics,38(2),535–568,doi:10.1086/705903,2020. Marinescu, I. E., D. Skandalis, and D. Zhao, Job search, job posting and unemployment insurance during the covid-19 crisis, SSRN Scholarly Paper ID 3664265, Social Science Research Network,Rochester,NY,2020. McElheran,K.,andC.Forman,Firmorganizationinthedigitalage: Ituseandverticaltransactionsinusmanufacturing,AvailableatSSRN3396116,2019. Mian,A.,andA.Sufi,Whatexplainsthe2007–2009dropinemployment?,Econometrica,82(6), 2197–2223,2014. Mongey, S., and A. Weinberg, Characteristics of workers in low work-from-home and high personal-proximityoccupations,BeckerFriedmanInstituteforEconomicWhitePaper,2020. Muro,M.,S.Liu,J.Whiton,andS.Kulkarni,Digitalizationandtheamericanworkforce,BrookingsPaper,2017. Nakamura, E., and J. Steinsson, Identification in macroeconomics, Journal of Economic Perspectives,32(3),59–86,2018. Palomino, J.C., J.G.Rodriguez, andR.Sebastian, Wageinequalityandpovertyeffectsoflockdownandsocialdistancingineurope,AvailableatSSRN3615615,2020. Papanikolaou,D.,andL.D.Schmidt,Workingremotelyandthesupply-sideimpactofcovid-19, Tech.rep.,NationalBureauofEconomicResearch,2020. Pierri, N., and Y. Timmer, Tech in fin before fintech: Blessing or curse for financial stability?, Tech.rep.,CESifoGroupMunich,2020. 30

Sanderson, E., and F. Windmeijer, A weak instrument f-test in linear iv models with multiple endogenousvariables,Journalofeconometrics,190(2),212–221,2016. Shibata,I.,Unemploymentintoday’srecessioncomparedtotheglobalfinancialcrisis,2020. Soh,J.,M.Oikonomou,C.Pizzinelli,I.Shibata,andM.MendesTavares,DidtheCOVID-19Recession Increase the Demand for Digital Occupations in the United States? Evidence from Employment and Vacancies Data, IMF Working Papers 2022/195, International Monetary Fund,2022. 31

Figure1: UnemploymentandMobilityintheUS ThisfigureplotsthechangeintheunemploymentratebetweenFebruaryandAprilbystateontheaveragechange inmobilityinretail,recreationandtransitstationinApril. ThereddiamondsrepresentstateswhereITadoption isabovethemedianandthebluetrianglesrepresentstateswhereITadoptionisbelowthemedian. Theredline shows the linear fit for high-IT state and the blue line shows the linear fit for low IT states. See section 3 and section4formoredetails. 32

Figure2: Unemployment,Mobility,andITintheUS:Event-Studydesign (a)Coefficientson∆Mobility 1 0 1- 2- 3- (b)Coefficientson∆Mobility×IT Jun2019 Sep2019 Dec2019 Mar2020 Jun2020 1 5. 0 5.- Jun2019 Sep2019 Dec2019 Mar2020 Jun2020 (c)ImpactofaMobilityDropforHighvsLowITStates 4 3 2 1 0 Jun2019 Sep2019 Dec2019 Mar2020 Jun2020 High IT Low IT TheFigureshowsestimatesofEquation3.Panel(a)reportsestimatesofβτand95%confidenceintervals;Panel(b)reportsestimatesofβ 3,τ and95%confidenceintervals;Panel(c)reports−σ(∆Mobilitys)·(βτ+/−β 3,τσ(ITs))whereσ(∆Mobilitys)andσ(ITs)arethestandard deviationofthemobilitychangeandoftheState-levelITadoption. 33

Figure3: ITAdoptionandRoutineWork This figure is a binscatter that plots the level of IT adoption in an MSA on the vertical axis against the routine employmentshareinanMSAonthehorizontalaxis.Seesection3andsubsection5.1formoredetails. 34

Figure4: MitigatingImpactofITacrossIndividuals Thisfigureplotsthecoefficientandthe90%confidenceintervalofβ andβ fromEquation7: 5 6 Unemployed =α+β ∆Mobility ∗A +β ∆Mobility ∗(1−A ) i,t 1 msa(i),t i 2 msa(i),t i +β IT ∗A +β IT ∗(1−A ) 3 msa(i) i 4 msa(i) i +β ∆Mobility ∗IT ∗A 5 msa(i),t msa(i) i +β ∆Mobility ∗IT ∗(1−A ) 6 msa(i),t msa(i) i +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α +(cid:178) i msa(i) msa(i) msa(i),t s(i) i,t whereUnemployed isadummyvariablethattakesthevalueoneiftheindividuali isunemploymentinmonth i,t t (April/May2020)andzeroiftheindividualisemployed.∆Mobility isthechangeinmobilityinmontht msa(i),t relativetothepre-COVIDbaseline.IT istheaveragelevelofITadoptionintheMSA.A aredummyvariables msa(i) i categorizing the respondent according to gender, race, and education subgroups. X includes GDP per capita, populationdensityandtheminorityshare.Seesection3andsection5formoredetails. 35

Figure5: ITAdoptionandWork-from-Homeability ThisfigureplotsthelevelofITadoptioninanMSAonthehorizontalaxisagainsttheshareofjobsthatcanbedone fromhomeontheverticalaxis.TheshareofjobsthatcanbedonefromhomearetakenfromDingelandNeiman [2020].Seesection3andsection5formoredetails. 36

Figure6: VacancyPostings,Mobility,andITintheUS:Event-Studydesign (a)Coefficientson∆Mobility 51. 1. 50. 0 50.- (b)Coefficientson∆Mobility×IT Jun2019 Sep2019 Dec2019 Mar2020 Jun2020 20. 0 20.- 40.- Jun2019 Sep2019 Dec2019 Mar2020 Jun2020 (c)ImpactofaMobilityDropforHighvsLowITMSAs 50. 0 50.- 1.- 51.- Jun2019 Sep2019 Dec2019 Mar2020 Jun2020 High IT Low IT TheFigureshowsestimatesofEquation10. Panel(a)reportsestimatesofβτ, Panel(b)reportsestimatesofβ 3,τ andPanel(c)reports −σ(∆Mobilitymsa)·(βτ+/−β 3,τσ(ITmsa))whereσ(∆Mobilitymsa)andσ(ITmsa)arethestandarddeviationsofthemobilitychangeand oftheMSA-levelITadoption. 37

Table1: Unemployment,MobilityandIT:State-levelRegressions Dependentvariable: ∆UnemploymentRate (1) (2) (3) (4) ∗ ∗∗∗ ∗∗∗ IT -0.0180 0.134 0.142 (0.010) (0.037) (0.033) ∆Mobility -0.148 ∗∗ -0.505 ∗∗∗ -0.622 (0.070) (0.102) (0.377) ∆Mobility×IT 0.463 ∗∗∗ 0.476 ∗∗∗ (0.116) (0.105) R-squared 0.0575 0.116 0.478 0.598 N 51 51 51 51 Controls No No No Yes ResultsofestimatingEquation2: ∆UR =α+β ∆Mobility +β IT +β ∆Mobility ∗IT +X (cid:48)σ+(X ∗Mobility ) (cid:48)γ+(cid:178) s 1 s 2 s 3 s s s s s s where∆UR isthechangeintheunemploymentrateinstatesbetweenAprilandFebruary. ∆Mobility isthe s s averagedeclineinmobilityinstatesinApril.IT isadummythatindicateswhetherastateisabovethemedianin s termsofITadoptionandzeroifitisbelowthemedian. Xincludesthelevelandtheinteractionbetweenmobility andGDPpercapita,thepopulationdensityandthemanufacturingshareofthestateascontrolvariablesinthe regressions. Robuststandarderrorsarereportedinparentheses. *p<0.1,**p<0.05,***p<0.01. Seesection4 formoredetails. 38

Table2: Unemployment,MobilityandIT:Individual-levelRegressions Dependentvariable: Unemployed (1) (2) (3) (4) ∆Mobility -0.181 ∗∗∗ -0.239 ∗∗∗ -0.742 0.0236 (0.031) (0.037) (1.559) (1.358) ∗∗∗ ∗∗ ∗∗∗ IT -0.00697 0.0187 0.0193 0.0292 (0.005) (0.007) (0.009) (0.011) ∆Mobility×IT 0.0699 ∗∗∗ 0.0656 ∗∗ 0.0677 ∗∗∗ (0.023) (0.032) (0.025) R-squared 0.00346 0.00418 0.0293 0.0384 N 71812 71812 71812 71812 Controls No No Yes Yes StateFEs No No No Yes ResultsofestimatingEquation4: Unemployed =α+β ∆Mobility +β IT +β ∆Mobility ∗IT i,t 1 msa(i),t 2 msa(i) 3 msa(i,t) msa(i) +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α +(cid:178) i msa(i) msa(i) msa(i),t s(i) i,t whereUnemployed isadummythatequalsoneiftheindividualisunemployedinmontht,wheret(April/May i,t 2020)andzerootherwise. ∆Mobility isthechangeinmobilityintheMSAwheretheindividuallivesand msa(i),t IT isthelevelofITadoptionintheMSAwhereindividualilives. Z areindividuallevelcontrols. X msa(i) i msa(i) areMSA-levelcontrols,includingthelevelandtheinteractionbetweenmobilityandGDPpercapita,theshareof minorities,theshareofpeoplewithathreeyearBachelor’sdegree,andtheunemploymentrateinFebruary2020. α arestatefixedeffects. StandarderrorsareclusteredattheMSAlevel. Theregressionsareweightedbythe s(i) assignedweightoftherespondent.*p<0.1,**p<0.05,***p<0.01.Seesection3andsection5formoredetails. 39

Table3: Unemployment,MobilityandIT Dependentvariable:Unemployed (1) (2) (3) (4) (5) (6) ∆Mobility×IT×Male 0.0306 ∗ 0.0494 ∗ (0.017) (0.025) ∆Mobility×IT×Female 0.0684 ∗∗∗ 0.0894 ∗∗∗ (0.019) (0.028) ∆Mobility×IT×White 0.0346 ∗∗ 0.0610 ∗∗ (0.017) (0.027) ∆Mobility×IT×Non-White 0.0577 ∗ 0.0909 ∗∗∗ (0.030) (0.035) ∆Mobility×IT×High/MedEduc 0.0520 ∗∗∗ 0.0712 ∗∗∗ (0.016) (0.025) ∆Mobility×IT×LowEduc -0.0324 0.0122 (0.049) (0.054) R-squared 0.0204 0.0386 0.0206 0.0388 0.0208 0.0386 N 71812 71812 71812 71812 71812 71812 Controls No Yes No Yes No Yes FEs Yes Yes Yes Yes Yes Yes ResultsofestimatingEquation7: Unemployed =α+ i,t +β ∆Mobility ∗A +β ∆Mobility ∗(1−A ) 1 msa(i),t i 2 msa(i),t i +β IT ∗A +β IT ∗A 3 msa(i) i 4 msa(i) i +β ∆Mobility ∗IT ∗(1−A ) 5 msa(i),t msa(i) i +β ∆Mobility ∗IT ∗(1−A ) 6 msa(i),t msa(i) i +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α +(cid:178) i msa(i) msa(i) msa(i),t s(i) i,t whereUnemployed isadummyvariablethattakesthevalueoneiftheindividuali isunemploymentinmonth i,t t (April/May2020)andzeroiftheindividualisemployed.∆Mobility isthechangeinmobilityinmontht msa(i),t relativetothepre-COVIDbaseline. IT istheaveragelevelofITadoptionintheMSA. A andB aredummy msa(i) i i variablesforgender,race,andeducationsubgroups. X capturesMSA-levelcontrols,includingthelevelandthe interactionbetweenmobilityandGDPpercapita,theshareofminorities,theshareofpeoplewithathreeyear Bachelor’sdegree, andtheunemploymentrateinFebruary2020. Theregressionsareweightedbytheassigned weightoftherespondent.*p<0.1,**p<0.05,***p<0.01.Seesection3andsection5formoredetails. 40

Table4: InstrumentalVariableApproach Dependentvariable:Unemployed (1) (2) (3) (4) (5) ∆Mobility -0.246 ∗∗∗ -0.230 ∗∗ -0.237 ∗∗ -0.168 ∗∗∗ -0.165 ∗∗∗ (0.039) (0.098) (0.101) (0.048) (0.047) ∗∗∗ IT -0.0192 -0.00590 -0.00404 -0.00596 -0.00524 (0.007) (0.018) (0.019) (0.010) (0.010) IT*∆Mobility 0.0710 ∗∗∗ 0.188 0.223 ∗ 0.102 ∗ 0.0981 ∗ (0.024) (0.117) (0.134) (0.059) (0.058) R-squared 0.00418 -0.00469 -0.00830 0.0111 0.0217 N 71812 51111 51111 51111 51111 F-statIT 29.59 28.13 15.63 15.69 F-statInt. 9.189 7.468 24.62 24.58 P-value=OLS 0.317 0.255 0.600 0.641 Instrument Routine1980 Routine1980 Routine1980 Routine1980 Controls PreUR PreUR +Demographics (cid:88) (cid:88) StateFE Resultsofa2SLSestimationof Unemployed =α+β ∆Mobility +β IT +β ∆Mobility ∗IT +(cid:178) i,t 1 msa(i),t 2 msa(i) 3 msa(i,t) msa(i) i,t whereUnemployed isadummythatequalsoneiftheindividualisunemployedinmontht,wheret(April/May i,t 2020)andzerootherwise. ∆Mobility isthechangeinmobilityintheMSAwheretheindividuallivesand msa(i),t IT isthelevelofITadoptionintheMSAwhereindividualilives.TheendogenousregressorIT isinstrumsa(i) msa(i) mentedwiththeroutineemploymentsharein1980,andtheendogenousregressorIT ∗∆Mobility msa(i) msa(i),t isinstrumentedwiththeproductoftheroutineemploymentsharein1980andthedeclineinmobility. Standard errorsareclusteredattheMSAlevel. Theregressionsareweightedbytheassignedweightoftherespondent. * p<0.1,**p<0.05,***p<0.01.Seesection3andsubsection5.1formoredetails. 41

Table5: Unemployment,Mobility,Teleworkingabilities,E-commerceandIT Dependentvariable: Unemployed (1) (2) (3) (4) (5) ∆Mobility×IT 0.0677 ∗∗∗ 0.0539 ∗∗ 0.0929 ∗∗∗ (0.025) (0.025) (0.030) ∆Mobility×Teleworking 1.100 ∗∗ 1.002 ∗∗ (0.517) (0.506) ∆Mobility×E-commerce 0.0113 0.0196 (0.021) (0.020) R-squared 0.0384 0.0385 0.0387 0.0373 0.0376 N 71812 71812 71812 62276 62276 Controls Yes Yes Yes Yes Yes StateFEs Yes Yes Yes Yes Yes Resultsofestimatingthefollowingequation: Unemployed =α+β ∆Mobility +β IT +β ∆Mobility ∗IT i,t 1 msa(i),t 2 msa(i) 3 msa(i,t) msa(i) +β W +β ∆Mobility ∗W 4 msa(i) 5 msa(i,t) msa(i) +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α +(cid:178) i msa(i) msa(i) msa(i),t s(i) i,t whereUnemployed isadummythatequalsoneiftheindividualisunemployedinmontht,wheret(April/May i,t 2020) and zero otherwise. ∆Mobility is the change in mobility in the MSA where the individual lives. msa(i),t IT is the level of IT adoption in the MSA where individual i lives. W is either the share of jobs that msa(i) msa(i) canbedonefromhomeintheMSAwhereindividualilives, takenfromDingelandNeiman[2020](columns2 and3)ortheshareofestablishmentsthatusee-commercetechnologiesaccordingto2016Aberdeensurvey,after controllingforestablishment’sindustry,(columns4and5). Z areindividuallevelcontrols. X areMSAlevel i msa(i) controls. α arestatefixedeffects. StandarderrorsareclusteredattheMSAlevel. Theregressionsareweighted s(i) bytheassignedweightoftherespondent. *p<0.1,**p<0.05,***p<0.01. Seesection3andsection5formore details. 42

Table6: Vacancies,MobilityandIT ∆TotalVacancies ∆DigitalVacancies ∆Non-DigitalVacancies ∆ShareofDigitalVacancies (1) (2) (3) (4) (5) (6) (7) (8) ∆Mobility 0.553 ∗∗∗ 5.227 0.284 ∗∗ 8.148 0.825 ∗∗∗ 9.182 -0.541 ∗∗∗ -1.033 (0.132) (8.992) (0.130) (9.658) (0.087) (10.126) (0.141) (9.060) ∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗ ∗ ∗ IT -0.0259 -0.125 -0.0213 -0.128 -0.0395 -0.0633 0.0182 -0.0647 (0.014) (0.038) (0.016) (0.044) (0.009) (0.035) (0.011) (0.035) ∆Mobility×IT -0.410 ∗∗∗ -0.420 ∗∗ -0.171 -0.248 ∗ (0.154) (0.171) (0.137) (0.137) R-squared 0.400 0.580 0.184 0.402 0.576 0.667 0.339 0.422 N 250 250 250 250 250 250 250 250 Controls No Yes No Yes No Yes No Yes ResultsofestimatingEquation9: ∆JobPosting =α+β ∆Mobility +β IT +β ∆Mobility ∗IT msa 1 msa 2 msa 3 msa msa +X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+(cid:178) msa msa msa msa where∆JobPosting istheaveragechangeintheloglevelofvacanciesbetweenFebruary2020andMayorApril msa 2020ineachMSA.∆Mobility istheaveragedeclineinmobilitybetweenFebruary2020andAprilorMay2020 msa andIT measuresITadoptionattheMSAlevel. X includesthelevelandtheinteractionbetweenmobilityand msa variousMSA-levelcharacteristicssuchasGDPpercapita,theshareofpeoplewithathreeyearBachelor’sdegree, theshareofminoritiesandtheunemploymentrateinFebruary2020.Columns(1)and(2)reportestimationresults forthechangeinthelogleveloftotalvacancies,columns(3)and(4)reportresultsforthesamespecificationbut focusingondigitalvacancies,columns(5)and(6)reportresultsonnon-digitalvacanciesandcolumns(7)and(8) reportresultsforthechangeintheshareofdigitalvacancies.RegressionsareweightedbytheMSApre-COVID-19 employmentshares. Robuststandarderrorsarereportedinparentheses. *p<0.1,**p<0.05,***p<0.01. See section3formoredetails. 43

ONLINE APPENDIX—NOT FOR PUBLICATION TableA1: Unemployment,MobilityandIT:Probit Dependentvariable: Unemployed (1) (2) (3) (4) ∆Mobility -0.840 ∗∗∗ -1.115 ∗∗∗ -4.285 -0.555 (0.147) (0.165) (7.616) (6.912) ∗∗ ∗ ∗∗∗ IT -0.0324 0.0937 0.0893 0.154 (0.022) (0.037) (0.046) (0.056) ∆Mobility×IT 0.328 ∗∗∗ 0.292 ∗∗ 0.350 ∗∗∗ (0.105) (0.147) (0.128) N 71812 71812 71812 71812 Controls No No Yes Yes StateFEs No No No Yes ResultsofestimatingEquation4withProbit: Unemployed =α+β ∆Mobility +β IT +β ∆Mobility ∗IT i,t 1 msa(i),t 2 msa(i) 3 msa(i,t) msa(i) +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α +(cid:178) i msa(i) msa(i) msa(i),t s(i) i,t whereUnemployed isadummythatequalsoneiftheindividualisunemployedinmontht,wheret(April/May i,t 2020)andzerootherwise. ∆Mobility isthechangeinmobilityintheMSAwheretheindividuallivesand msa(i),t IT isthelevelofITadoptionintheMSAwhereindividualilives. Z areindividuallevelcontrols. X are msa(i) i msa(i) MSAlevelcontrols.α arestatefixedeffects.StandarderrorsareclusteredattheMSAlevel.Theregressionsare s(i) weightedbytheassignedweightoftherespondent. *p<0.1,**p<0.05,***p<0.01. Seesection3andsection5 formoredetails.

TableA2: Unemployment,MobilityandIT:Robustness Dependentvariable:Unemployed (1) (2) (3) (4) (5) (6) (7) (8) ∆Mobility×IT 0.0655∗∗∗ 0.00497∗∗ 0.129∗∗∗ 0.0518∗∗∗ 0.0572∗∗ 0.0776∗∗∗ 0.0794∗∗∗ 0.0733∗∗∗ (0.016) (0.002) (0.043) (0.019) (0.022) (0.025) (0.026) (0.024) R-squared 0.0198 0.0195 0.0198 0.0198 0.0245 0.0211 0.0221 0.186 N 68923 68923 68923 68923 68923 68923 24680 24653 Controls No No No No No No No No FEs Yes Yes Yes Yes Yes Yes Yes Yes Specification Baseline High-SpeedInternet HighIT PCs/Emp U6Unemployment Ind&Occcontrols EmployedFeb2020 Occ/IndFeb2020 Resultsofestimatingthefollowingequation: Unemployed =α+β ∆Mobility +β IT +β ∆Mobility ∗IT i,t 1 msa(i),t 2 msa(i) 3 msa(i,t) msa(i) +Z (cid:48)δ+X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+α +(cid:178) i msa(i) msa(i) msa(i),t s(i) i,t whereUnemployed isadummythatequalsoneiftheindividualisunemployedinmontht,wheret(April/May i,t 2020) and zero otherwise. Column (1) is the baseline specification. Column (2) replaces our baseline IT measurewiththeshareofpeoplewhohaveaccesstohigh-speedinternetinthegivenMSA.Column(3)definesthe ITvariableasadummythatequalsoneiftheMSAhasanabove-medianITadoptionandzerootherwise. Column(4)replacestheITmeasurewithameasureoftheshareofpersonalcomputersperemployee. Column(5) classifiesindividualsasunemployedaccordingtotheU6unemploymentrate. Column(6)includestheleveland theinteractionbetweenmobilityandtheemploymentsharesofthelargestpre-COVID-19industryandoccupationcategoriesasadditionalcontrolvariablesintheregression.Theindustryandoccupationemploymentshares accountformorethan1/3oftotalemploymentin2019. Column(7)includesonlyrespondentsthatwereinthe surveyalsoinFebruary2020andwereemployed. Column(8)includesonlyrespondentsthatwereinthesurvey alsoinFebruary2020andwereemployedandalsoaddsfixedeffectsfortheindustryandtheoccupationofthe respondentinthatmonth. StandarderrorsareclusteredattheMSAlevel. Theregressionsareweightedbythe assignedweightoftherespondent.*p<0.1,**p<0.05,***p<0.01.Seesection3andsection5formoredetails.

TableA3: Unemployment,MobilityandIT:State-levelRegressionswithdifferentcutoffs Dependentvariable:∆UnemploymentRate (1) (2) (3) ∆Mobility×AboveMedianIT -0.0423 (0.056) ∆Mobility×BelowMedianIT -0.505 ∗∗∗ (0.102) ∆Mobility×Top33%IT -0.0560 (0.077) ∆Mobility×Bottom66%IT -0.350 ∗∗∗ (0.100) ∆Mobility×Top25%IT -0.0340 (0.105) ∆Mobility×75%to25%IT -0.353 ∗∗ (0.133) ∆Mobility×Bottom25%IT -0.291 ∗∗∗ (0.105) R-squared 0.478 0.330 0.348 N 51 51 51 Resultsofestimatingtheequation: ∆UR =α+γIT +β ∆Mobility ∗IT +β ∆Mobility ∗(1−IT )+(cid:178) s s high s s low s s s where∆UR isthechangeintheunemploymentrateinstatesbetweenAprilandFebruaryinstates.∆Mobility s s istheaveragedeclineinmobilityinstatesinApril. Incolumn(1)IT isadummythatindicateswhetherastate s isabovethemedianintermsofITadoptionandzeroifitisbelowthemedian. Incolumn(2)isadummythat indicateswhetherastateisinthetoptercileintermsofITadoptionandzerootherwise. Incolumn(3),instead, asetofthreedummiesareinteractedwiththechangeinmobility: adummyforstatesinthetopquartileofIT adoption,adummyforstatesinthebottomquartileofITadoption,andadummyforallstateabovethebottom quartileandbelowthetopquartile. Robuststandarderrorsarereportedinparentheses. *p<0.1,**p<0.05,*** p<0.01.Seesection4formoredetails.

TableA4: Unemployment,MobilityandIT:MSA-level Dependentvariable: ∆UnemploymentRate (1) (2) (3) ∆Mobility -0.191 ∗∗∗ -0.251 ∗∗∗ -2.930 (0.031) (0.041) (2.263) ∗∗ ∗∗ ∗∗ IT -0.00928 0.0166 0.0266 (0.004) (0.008) (0.011) ∆Mobility×IT 0.0687 ∗∗∗ 0.103 ∗∗∗ (0.024) (0.035) R-squared 0.140 0.164 0.210 N 508 508 508 Controls No No Yes ResultsofestimatingEquation4attheMSAlevel ∆UR =α+β ∆Mobility +β IT +β ∆Mobility ∗IT msa,t 1 msa,t 2 msa 3 msa,t msa +X (cid:48) σ+(X ∗Mobility ) (cid:48)γ+(cid:178) msa msa msa,t msa,t where∆UR istheMSA‘schangeintheunemploymentratebetweenFebruary2020andmontht,wheret is msa,t AprilorMay2020. ∆Mobility isthechangeinMSA-levelmobilityoverthesameperiod. IT isthelevel msa,t msa ofITadoption. X includesthelevelandtheinteractionbetweenmobilityandvariousMSA-levelcharacteristics includingGDPcapitaincome,theshareofpeoplewithathreeyearBachelor’sdegreeandtheshareofminorities pre-pandemic. RegressionsareweightedbytheMSA’spre-COVID-19employmentshare. Robuststandarderrors arereportedinparenthesis.*p<0.1,**p<0.05,***p<0.01.Seesection3andsection5formoredetails.

TableA5: Employment,MobilityandIT:MSA-level Dependentvariable: ∆Employment Total TradableIndustries Non-TradableIndustries (1) (2) (3) ∆Mobility 0.516 ∗∗∗ 0.235 0.613 ∗∗∗ (0.127) (0.281) (0.218) ∗∗∗ ∗ IT -0.0713 -0.163 -0.0423 (0.025) (0.086) (0.050) ∆Mobility×IT -0.212 ∗∗∗ -0.326 ∗ -0.0606 (0.065) (0.195) (0.133) R-squared 0.0693 0.0240 0.0461 N 513 463 506 Resultsofestimatingthefollowingequation: ∆Employment =α+β ∆Mobility +β IT +β ∆Mobility ∗IT +(cid:178) msa,t 1 msa 2 msa 3 msa msa msa,t where∆Employment isthechangein(log)employmentineachMSAbetweenFebruary2020andAprilor msa,t May2020. ∆Mobility isthechangeinmobilityoverthesameperiodandIT isthelevelofITadoptionat msa msa theMSA.RegressionsareweightedbytheMSA’spre-covidemploymentshare.Robuststandarderrorsarereported inparenthesis.*p<0.1,**p<0.05,***p<0.01.Seesection6formoredetails.

FigureA1: UnemploymentandLockdownStringencyintheUS ThisfigureplotsthechangeintheunemploymentratebetweenFebruaryandAprilbystateontheaverageLockdownstringencyindex(accordingtoKeystone)overthesameperiod.ThereddiamondsarestateswhereITadoptionisabovethemedianandthebluediamondsarestateswhereITadoptionisbelowthemedian. Theredline shows the linear fit for high-IT state and the blue line shows the linear fit for low IT states. See section 3 and section4formoredetails.