In the Driver's Seat: Pandemic Fiscal Stimulus and Light Vehicles

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) In the Driver’s Seat: Pandemic Fiscal Stimulus and Light Vehicles Jack Dunbar, Christopher Kurz, Geng Li, and Maria D. Tito 2024-013 Please cite this paper as: Dunbar, Jack, Christopher Kurz, Geng Li, and Maria D. Tito (2024). “In the Driver’s Seat: Pandemic Fiscal Stimulus and Light Vehicles,” Finance and Economics Discussion Series 2024-013. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2024.013. 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.

In the Driver’s Seat: Pandemic Fiscal Stimulus and Light Vehicles* JackDunbar† ChristopherKurz‡ GengLi§ MariaD.Tito¶ February27,2024 Abstract Thispaperexplorestheimpactoftwofiscalprograms,theEconomicImpactPayments and the Paycheck Protection Program, on vehicle purchases and relates our findings to post-pandemic price pressures. We find that receiving a stimulus check increasedtheprobabilityofpurchasingnewvehicles. Inaddition,thedisbursementof funds from the Paycheck Protection Program was associated with a rise in local new carregistrations. Ourestimatesindicatethatthesetwoprogramsaccountforaboost of 1 3/4 million units—or 12 percent—to new car sales in 2020. Furthermore, the inducedboostinsalescoincidedwiththepresenceofsignificantproductionconstraints and exacerbated an inventory drawdown, thereby contributing to the rapid increase innewvehiclepricesthatprevailedinthesubsequentyears. Keywords: DiscretionaryFiscalPolicy,LightVehiclePurchases,Inflation. JELclassification: E21,E31,G31,G51,H24,H31. *WewouldliketothankTomazCajner,LisaDettling,GlennFollette,JakeOrchard,NickTurner,and otherFederalReservecolleaguesfortheirinsightfulsuggestionsandcomments.Wewouldalsoliketothank MichaelGreenforexcellentresearchassistance.TheviewspresentedinthispaperrepresentthoseoftheauthorsanddonotnecessarilycoincidewiththoseoftheFederalReserveBoard,theFederalReserveSystem, oritsstaff. †UniversityofPennsylvania. ‡FederalReserveBoard. §FederalReserveBoard. ¶FederalReserveBoard.Correspondingauthor.Contact:maria.d.tito@frb.gov.

1 Introduction TheCOVID-relatedlockdownsledtoabruptinterruptionstobroadeconomicactivitiesin thesecondquarterof2020. Surprisingly,lightvehiclesalesupontheonsetofthepandemic outbreakandinsubsequentquartersremainedrelativelyresilient. Thispaperattemptsto explore the reasons behind this resilience and relates the strong vehicle sales to significantly higher vehicle prices consumers encountered following the pandemic. We choose tofocusonvehicledemandbecauselightvehiclesalestendtobe“highbeta,”thatis,highly procyclical and responsive to changes in economic conditions. In addition, the motor vehiclesectorisanindustryblessedwiththepreponderanceofwell-measuredanddetailed data. Onewaytoillustratethepandemicresilienceoflightvehiclesalesistoestimateacounterfactual level of sales based on the standard measures of macroeconomic activity. As shown in figure 1, the realized drop in vehicle purchases during the second quarter of 2020 (the blue line) was far less pronounced than what could be predicted by the regular determinantsofvehiclesales,suchasGDPgrowth,populationgrowth,theunemployment rate, and gasoline prices (the red line).1 Replacing GDP with disposable personal income (DPI)growththatdoesnotincludethefiscalsupporthouseholdsreceivedduringthistime predictsanevenlargercontractioninsales(thegreenline).2 Indeed,fromtheonsetofthepandemicoutbreakthroughMarch2021,thefederalgovernment enacted several fiscal packages with the purpose of providing economic relief toconsumers,smallbusinesses,andotherentitiesthathadbeenadverselyaffectedbythe pandemic. Suchfiscalsupportmayhaveboostedautosalesduringaperiodwhendemand wouldotherwisebeevenmoresubdued. Wefocusontwodiscretionaryfiscalprograms— the Economic Impact Payments (EIP) and the Paycheck Protection Program (PPP)—and evaluate their potential effects on light vehicle purchases using household- and countylevel data. In particular, drawing on data from the Consumer Expenditure Survey (CE) and on new car registrations, we estimate that those two discretionary fiscal programs 1Therelativelymoderatedropinvehiclesaleswasevenmoreremarkableifconsideringsocialdistancing measuresthatemergedduringthepandemic,factorsthatarenotcapturedbyourmodelofprojection. 2OurmeasureofdisposableincomeexcludesthesupportfromtheEconomicIncomePayments(EIP), fromthePaycheckProtectionProgram(PPP)aswellasfromtheUnemploymentInsurance(UI). 1

boosted sales in 2020 by 1 3/4 million units. We then analyze the effect of resilient sales on vehicle inventories and detailed vehicle prices data, as the boost in sales mirrored a comparable decline in inventory levels because it occurred during a period of significant productionconstraints. Wefindthatstrongsaleswereassociatedwithanincreaseinvehiclepriceinflationof1.3percentagepoints(20percent),contributingtotherapidpickupin pricesthatprevailedoverthefollowingyear. Ourpaperaddstothevast,unsettledliteraturethatstudiestheeffectsofdiscretionary fiscalpolicyoneconomicactivityandothermacroeconomicvariables. Severalcontributions— such as Auerbach [2002] and Taylor [2011, 2022]—point to a largely limited stabilization impact of fiscal policy in recent decades. Moreover, Shapiro and Slemrod [2009] and Orchardetal.[2023]alsofindonlyamodesteffectofthe2008stimuluspaymentsonoverall spending. By contrast, Parker et al. [2013] document a significant effect of the 2008 stimuluspaymentsonconsumption,with2/3oftheconsumptionboostmaterializingthrough vehiclepurchases. Fiscal stimulus’ effects on vehicle sales also remain an active debate with respect to their magnitude and durability. On the one hand, the seminal contribution by Mian and Sufi [2012] suggests that the Car Allowance Rebate System—a program implemented in 2009 that was directly targeting light vehicle sales—led to a temporary boost of sales that was reversed in the following months. On the other hand, Hicks et al. [2012] point to a somewhat larger effect with little change in purchasing patterns following the end of the program. Finally,Orchardetal.[2023]suggestasmallboosttomotorvehicleexpenditures fromthe2008stimuluspayments. Ourpaperrevisitsthelinkbetweendiscretionaryfiscalprogramsandlightvehiclepurchases,takingadvantageoftheunprecedentedmagnitudeoffiscalsupportenactedinresponse to the pandemic recession. Indeed, the literature remaining inconclusive on this question may reflect the relatively moderate sizes of previous stimulus programs. The effect of the first round of EIPs on consumption has been also analyzed by Parker et al. [2022], which characterizes the MPCs for various consumption categories by exploiting the variation in the amount, the receipt, and the timing of receipt of the 2020 stimulus checks;inparticular,theydocumentanegligibleimpactontransportation,acategorythat 2

mainly includes purchases of vehicles. While our paper separately analyzes the various roundsofEconomicImpactPayments,itidentifieswhetherthereceiptofastimuluscheck altertheprobabilityofpurchasingavehicle,ratherthantheeffectonthethedollarvalue; in fact, our analysis of units is likely to be less affected by confounding factors and other incentives that, instead, could shift consumer expenditure towards vehicles of different values.3 Ourresultssuggestthattheeffectmaterializesthroughadditionalnewandused vehicle purchases. More importantly, our work adds two pieces of novel evidence. First, we document that the PPP, a program that primarily targeted small businesses, had also implications on household expenditures.4 Second, we rely on novel micro-level data on prices,sales,andinventorylevelstoinfertheinflationimplicationsofrobustvehiclesales.5 2 Data Thispaperleverageshousehold-levelinformationonconsumervehiclepurchasebehavior fromtheConsumerExpenditureSurvey(CE),detailedcounty-levelvehiclesalesdatafrom R.L.Polk&Co. (Polk),andvehiclemodel-levelsales,inventories,andpricesfromInforma BusinessMedia,Inc.. TheCEisconductedquarterlybytheBureauofLaborStatisticsandcollectsdetailedinformationonU.S.householdexpendituresinadditiontosocioeconomicanddemographic characteristics. Inparticular,theCEcollectsinformationonvehiclepurchasesand,importantly,thereceiptoffiscalstimuluschecks. MotorvehiclepurchasesreflectedintheCEdataarebroadlyconsistentwiththeaggregatesalesdata. Forexample,asshowninfigure2,thenumberofnewvehicleoutlaysfrom the survey (in blue) tracks the trend in the Polk data on retail purchases (in black) well, thoughthelevelissomewhatlower.6 3Forexample,Hoekstraetal.[2017]documentthattheCarAllowanceRebateSystemshiftedpurchases towardslessexpensivecars,whichwerealsofuelefficient. 4MostoftheliteratureonthePPPexploresitslabormarketimplications.See,forexample,Autoretal. [2022a]andAutoretal.[2022b]. 5Orchardetal.[2023]alsocharacterizedthepriceimplicationsassociatedwiththe2008stimuluschecks, buttheyrelyonaggregatepricedata.Theimpliedeffecttheydocumentissmallercomparedtoours;the differenceislikelytheresultofsignificantlymoresevereproductionconstraintsin2020and2021relativeto thoseduringtheGreatRecession. 6WeusetheCEsurveyweightstoestimatetheaggregatesales.Therehavebeenseveralpapers(see, 3

The vehicle prices in the CE data are also consistent with national trends. As seen in figure 3, the average new vehicle prices from the survey (the dark blue line) follows the trendintheJ.D.PowerandAssociates(JDPower)averagenewtransactionprice(thelight blueline)closely. Our analysis also takes advantage of registration data. When an auto or light truck is purchased, it is eventually registered with the state in which ownership occurs.7 Polk collects registration data from each state at the zip code and county level, and it records informationonregistrationtype(personal/business/lease).8 WelinkdataonPPPloansto smallfirmstothecountyregistrationinformationfromPolktoexplorethepotentialeffect ofPPPonvehiclesales. ThePPPloandatasourcefromtheSmallBusinessAdministration andareaggregateduptothecountylevel.9 Inaddition,InformaBusinessMedia,Inc. collectsdetailedmakeandmodel-leveldata onprices,inventories,andsales. Thesalesandinventoriesaremonthlydata,whileprices (specifically, manufacturer’s suggested retail prices or MSRPs) are available at an annual frequency. Each series contains information on vehicle make and model. Using these data, we explore the potential effect of robust sales during a period of vehicle production constraintsonsubsequentpriceincreases. 3 Empirical Analysis WenowanalyzehowvehicledemandrespondedtoEIPstimuluschecksandthePPPloan program. Our estimates suggest that these two programs together boosted sales by 1 3/4 millionunitsandcontributedtothepricepressuresinthemotorvehiclemarket. amongothers,Battistin,2003;Passeroetal.,2014)thathavedocumentedCEtendstounderestimateaggregatedexpenditures,suchasthoseintheNationalIncomeandProductAccounts.Thesurveyalsocollects dataonusedvehicles,whichwewillbeusinginouranalysis. 7Afterpurchasinganeworusedvehicle,statelawrequiresthatthevehicleberegisteredatthestate’s departmentofmotorvehicleswithin30days. 8Thisinformationhasbeenusedinseveralpapers,includingMianandSufi[2012] 9ThePPPdataareavailableathttps://data.sba.gov/dataset/ppp-foia. 4

3.1 EconomicImpactPayments BetweenMarch2020andMarch2021,thefederalgovernmentenactedthreefiscalpackages thatestablisheddirectpaymentstoeligibleindividuals. Thosepayments—–rangingfrom $600 to $1400 for individuals or from $1200 to $2800 for married couples jointly filing, plusanadditional$500to$1400foreachqualifyingdependent—–werebroadlyavailable toindividualswithincomesbelowspecificthresholdsandwerephasedoutbeyondthose thresholds.10 Across the three waves and all recipients, the total fiscal stimulus totaled $837.5billion,aheretoforeunprecedentedfiscalsupport. The CE data cover a large portion of the EIP program, with an estimated aggregate amountofstimuluscheckpaymentsnear$600billion. Finally,thetimingoftransfers(Figure 4) and the size of stimulus payments for the median household and those at the 25th and75thpercentile(Figure5)arebroadlyconsistentwiththeofficialdata. How did households use the stimulus checks? According to the CE survey, up to 60 percent of households mostly used this sizable income supplement to increase spending, while smaller shares reported paying down debt or adding to savings. Figure 6 plots proportion of the three uses of the stimulus checks (along with the fraction of missing responses). Thepropensitytospendstimulusfunds,whichappearstobesomewhatlarger than what was reported in the Household Pulse Survey, declined slightly for later waves ofpayments.11 Toevaluatehowthestimuluschecksmayhaveboostedhouseholdvehicledemand,we estimatealinearprobabilitymodelthatrelateswhetherahouseholdpurchasedacarwith 10Specifically,thestimuluspaymentswereinitiallyavailabletoindividualswithadjustedgrossincome (AGI)upto$75,000,headofhouseholdfilerswithAGIupto$112,500,ormarriedcouplesfilingjointlywith AGIupto$150,000;stimuluspaymentsweregraduallyphased-outabovethosethresholds.Inlaterwaves, thephase-outcriteriaweretightened. 11AccordingtotheCensusBureau’sHouseholdPulseSurvey,upto20percentofhouseholdsusedthe stimuluscheckstoincreasespendingwiththelargemajorityreportingpayingdowndebt.Otherworkdocumentedthedirectallocationofthestimulusfund.Forexample,Coibionetal.[2020]andArmantieretal. [2020]estimatethatforWave1directstimulusallocations,householdsspentabout30to40percent,saved about30to35percent,andpaiddowndebtwithabout30to35percent.Armantieretal.[2021]furtherindicatethathouseholdsplannedtospendlessoftheirWave2andWave3fiscalstimuluspayments,instead focusingmoreonsavinganddebtreduction. 5

thereceiptofstimulusfunds, y = α+β·Stimulus +γX +ε (1) it,t+1 it it it wherey isequalto1forhouseholdipurchasingacarinquartertort+1,Stimulus is it,t+1 it adummythatindicateswhetherthehouseholdreceivedastimuluscheck,andX denotes it anarrayofdemographicandincomecharacteristics.12 Multipleroundsofstimuluschecksweremailedtohouseholdsinboth2020and2021, butahouseholdonlyremainsintheCEsampleforatmostfourconsecutivequarters. As aresult,wefocusontwocohortsofCEhouseholdstoassessthepotentialboosttovehicle salesrelatedtothesestimuluschecks. Wekeeponlythehouseholdsthathaveparticipated inallfourquarterlysurveysinoursample. Thefirstwaveofhouseholdsstartedthesurvey ineitherthefirstorthesecondquarterof2020. Wecorrelatetheirreportedstatusofreceiving stimulus check between April and July with their new vehicle purchases during the second and third quarter of that year. The second wave of households started the survey in either the first or the second quarter of 2021. For these households, we correlate their receiving stimulus checks between December 2020 and May 2021 with their new vehicle purchasesinthefirstandthesecondquarterof2021.13 Theidentificationof β,ourcoefficientofinterest,reliesonthecross-sectionalvariation across households regarding the receipt of stimulus checks and their vehicle purchasing behavior. Indeed,theCEsampledisplaysvariabilityinthereceiptofstimulusfundsacross households, even below the income thresholds at which households were eligible for the payments. In particular, we find that only around 75 percent of households with income below$75,000reportedtohavereceivedthefirststimuluschecksin2020andonlyaround 85 percent reported having received the second check in 2021. By contrast, according to theprogram,allhouseholdsinthisgroupshouldhavereceivedthosechecksaroundthose 12Specifically,wecontrolforafter-taxincome(inlog-s),adummyifincomehasbeentop-coded,thesize ofthehousehold,andvariousobservables(age,race,gender,education,maritalstatus)fortheheadofthe householdineachwave. 13Thereweretworoundsofstimuluschecksmailedduringthistime.ThefirstroundwasprimarilydistributedbetweenlateDecembertolateJanuary,withmostindividualsreceiving$600.Thesecondroundwas primarilydistributedbetweenMarchandMay,withmostindividualsreceiving$1,400.Becausetheywere distributedwithlittletimeapart,wedonotseparatelyassesstherespectivepotentialvehiclesalesboost. 6

timelinesirrespectiveofotherfamilycharacteristics. We argue that two factors may help account for this discrepancy. First, as reported by the Government Accountability Office 2002, “[N]on [tax] filers, first-time filers, unbanked/underbanked, mixed immigrant status families, those with limited internet access, and those experiencing homelessness were likely to experience difficulties with receiving timely payments.” In fact, members of this group, while they did not file tax returnsin2018or2019,werestillrequiredtofileaformwiththeIRSbyNovember21,2020 to receive the stimulus in 2020, or to file a 2020 tax return in 2021 to receive it in 2021. In addition,householdsmayhavenotcorrectlyreportedthereceiptofstimulusfundsinthe survey. Thelatterfactorwouldimplythatthestimulusvariableismeasuredwithsomeerror,biasingusagainstfindinganyeffect. However,ourresultsarenotsolelytheoutcome of reporting errors: in fact, a falsification exercise that assigns a “stimulus treatment” to householdswithincomebelow$75,000,reportedintableA1,suggestsahigherpropensity topurchaseavehicleonlyinthepost-pandemicperiod. Accordingly,ourestimatescanbe interpretedaslowerboundsofthetrueeffects. OLSestimatesofmodel(1)arereportedintable1. Columns1–3lookatthe2020wave, while columns 4-6 report the results for the 2021 waves. We find that receiving a stimulus check generally boosted the probability of purchasing a car. In particular, during the first wave, receiving the stimulus checks increased the probability of purchasing a car by 3.9 percentage points, an effect that is both statistically and economically significant. In our CE sample, 11.4 percent of individuals that received a stimulus check purchased, on average, a vehicle over two consecutive quarters. Thus, our estimate suggests that the stimulus checks are associated with a one-third boost of the vehicle-purchase likelihood. Moving to columns (2) and (3), we estimate the likelihood of purchasing new and used cars separately. We find that, while the coefficient estimate is slightly larger for used car purchases, theybothimplyafairlysimilarboosttothevehicle-purchaseprobability: a40 percentincreasefornewvehiclebuyersvs. aone-thirdboostforusedcarpurchases.14 Duringthe2021stimuluswaves,receivingastimuluscheckcontinuestobeassociated 14Inoursample,4.1percentofhouseholdswhoreceiveda2020stimuluscheckboughtanewcar,onaverage,over2020q2-2020q3,and7.8percentofhouseholdswhoreceiveda2020stimuluscheckboughtaused carduringthesameperiod. 7

withanincreaseintheprobabilityofpurchasingavehicle(column4),withaneffectmaterializing only through used car purchase (column 6), while the coefficient on new vehicle purchasesisinsignificant,althoughwithanegativesign(column5). Thefactthatthebulk oftheeffectofthe2021stimuluschecksoccurredthroughused—ratherthannew—vehicle purchases is consistent with the emergence of significant production constraints in early 2021. Indeed, as new vehicle availability quickly deteriorated, demand for used vehicles soared,andvehicleshoppersweremorelikelytoexitthenewvehiclemarket.15 To quantify the impact of the EIP on sales, we consider the counterfactual of vehicle purchaseshadthatprogramnotbeenimplemented: withabout70percentofhouseholds inoursamplehavingreceivedastimuluscheck, ourestimatessuggestthatnewretailvehicles sales would have been lower cumulatively by about 1 1/4 million units in 2020, an effectthatappearstohavebeenfairlyshort-lived.16 Inotherwords,ourestimateaccounts foratleast40percentofthedifferencebetweenrealizednewvehiclesalesandthepredictionsofourheuristicmodelsreportedinfigure1.17 3.2 PaycheckProtectionProgram(PPP) The Paycheck Protection Program provided forgivable loans to small firms and was establishedundertheCoronavirusAid,Relief,andEconomicSecurityActof2020(CARES). Roughly12millionloansofnearly800billiondollarsweremadeduringthePPPprogram’s lifecycle,whichlastedfromApril2020untilMarchof2021.18 Theaverageloanpublished by the Small Business Administration (SBA), most of which were subsequently forgiven, was68thousanddollars,withtheloanamountvaryingsubstantivelybasedonthetypeof lender and the industry of the recipient. For example, in 2021 a PPP recipient in mining 15AccordingtoconsumerresearchpublishedbyKelleyBlueBookinMay2021,37percentofshoppers plannedtopostponetheirpurchase.Furthermore,amongthosethatremainedin-market,23percentwere consideringashiftfromnewtoused.Inthesubsequentwave,publishedinSeptember2021,thosepercentageshadrisento48percentand38percent,respectively. 16AccordingtothePolkdata,newretailvehiclepurchaseswere4.4millionunitsover2020q2and2020q3. Inourcounterfactual,iftherecipientsofstimuluschecksdidnotseeanincreaseinthecorrespondingprobabilityofpurchasinganewcar,thatwouldtranslateinto4.4*(0.017/0.041)*0.7=1.26millionunitslessover thosetwoquarters,as70percentofthesamplereceivedthefirstroundEIPchecks. 17Whilesalesreached11.2millionunitsinQ2,ourmodelspredictsaleswouldhavereachedbetween8.3 and9.8millionunitsinthatquarter,thussuggestinganadditionaldropbetween11/2and3millionunits. 18SeeU.S.SmallBusinessAssociationathttps://www.sba.gov/funding-programs/loans/covid-19-reliefoptions/paycheck-protection-programforinformationontheprogramandtherelateddata. 8

sector received an average loan of 110 thousand dollars in PPP funds while a firm in the agricultureandforestrysectorreceivedadisbursementofroughly20thousanddollars. Given that the majority of firms are small firms, the coverage of organizations—and hence owners—by PPP was quite extensive. Indeed, Decker et al. [2021] find that PPPeligiblefirmsandentitiesaccountforroughly60percentofprivatepayrollemployment.19 Accordingly, the possible implications of PPP-loan disbursement on consumer behavior, eitherdirectlythroughbusinessownerpurchasesorindirectlythroughcontinuedemploymentfundedbytheprogramcouldbesubstantial. WhilewedonothaveaccesstodatathatlinksthereceiptofPPPfundswithindividualleveloutcomes,acutofthehouseholdexpendituredatabytypeofemployerforthehead ofhouseholdpointstoaninterestingpattern. Asshownintable2,theprobabilityofvehiclepurchasesinhouseholdsthereferencepersonofwhichwasself-employedsignificantly increased after the pandemic relative to the 2017-2019 period.20 The change in vehicle purchasing behavior appears largely driven by new car purchases, with the probability rising to 2.2 percent in the post-pandemic from 1.3 percent, although the probability of purchasing a used vehicle also edges up between the two periods.21 By contrast, households working in private companies, in the federal, state, or local government, or in a familybusiness,experiencedalowerprobabilityofpurchasingavehicle. To investigate whether PPP boosted light vehicle sales, as hinted by the pattern in the CE occupation data, we merge loan data from the Small Business Administration, aggregatedatthecountylevel,withdataontheuniverseofnewvehicleregistrations.22 OuranalysisfocusesontheApriltoDecember2020period,thetimewhenissuedPPP loanswereuntargeted.23 Specifically,wecorrelate(log)changesincounty-levelper-capita 19ToqualifyforPPPloanforgiveness,businesseshadtospendthefundsonspecificeligibleexpenses—primarily,employeepaychecks.Formoredetailsandathoroughanalysisoftheimpactoftheprogram,seeAutoretal.[2022a],whofindPPPhadamaterialeffect,boostingJune2020payrollemploymentby about2.3million. 20AccordingtoAutoretal.[2022a],non-employerbusinesses,agroupthatincludestheself-employed, received$43billionintheinitialPPPtranches. 21Whileaself-employmentindicatorisastrongpredictorofvehiclepurchases,addingthatcontrolto model(1)doesnotchangetheresultsdiscussedintheprevioussection. 22MonthlyregistrationdatafromPolkandthesalesdatafromInformaBusinessMedia,Inc.alignalmost perfectlywithacorrelationof0.979fromJanuary2015untilAugust2023. 23AllsmallbusinesseswereeligibleforthePPPprogramthroughDecember2020.In2021,anewtranche oftheprogramtargetedfirmsthathadalreadyreceivedafirstPPPloanandhadexperiencedsignificant revenuelossesoverthecourseofthepandemic. 9

registrationsbetweenAprilandDecember2020,normalizedbythegrowthinregistrations observed in the same period of 2019, to county-level (log) PPP aid per capita that was disbursed, Regs Regs log c,Dec.2020 − log c,Dec.2019 = α + α logPPPLoans + d + u (2) 0 1 c s c Regs Regs c,Apr.2020 c,Apr.2019 whereRegs denotesthenewvehicleregistrationpercapitaincounty c attime t and c,t logPPPLoans denotes the cumulative PPP loans per capita approved for self-employed c applicants in county c between April and December 2020. As our main dependent variable, logPPPLoans , is analogous to a change in PPP aid relative to the period before c the program was announced, our specification effectively controls for county-level unobserved heterogeneity; furthermore, model (2) also includes state dummies, d , to absorb s commonchangesacrossstatesovertheperiodofanalysis.24 Our analysis can be illustrated in the scatter plot in figure 7: we find that counties receiving more PPP loans experienced a faster recovery in total registrations. Related regression results are reported in table 3.25 Controlling for state fixed effects, we find that a one-standard-deviationincreaseinper-capitaPPPloans—whichcorrespondtoanincrease ofabout$2,000percapita—isassociatedwitha10percentofastandarddeviationincrease inlogregistrationsrelativetothecomparisonperiod. To shed light on the mechanism behind this increase, we decompose new vehicle registration into personal and non-rental business purchases. We find that PPP loans had an impact on both personal registrations (column (2) of table 3) and business registrations (column (3)), with the coefficient for business registration somewhat higher although not statisticallydifferentfromthebaselineeffectorthatonpersonalregistrations. Columns(4)-(6)oftable3estimatethepatternofcumulativeregistrationsrelativetoa 24Duringthepandemic,severalstatesdelayedregistrationrequirementsduetothein-personnatureof manyDMVvisits.Thisdelaycouldresultinameasurementissuebetweenactualsalesandregistration,despitetheirusuallyhighcorrelations.Fortunately,ouranalysisaveragesoverseveralmonths,soanyslippage wouldnotbesignificant.Moreover,atamonthlyfrequency,vehicleregistrationsdonotdivergefromsales inamannerconsistentwithmismeasurementaffectingexclusivelyregistrations. 25Weexcludeleasesbecausetravelrestrictions,firstimplementedtowardtheendofMarch2020anddifferentiallyrevisedacrossstatesandcountiesovertime,significantlyimpactedthesigningofleaseagreements,especiallyamongcommercialunits. 10

counterfactual scenario. In particular, we compare the (log) cumulative registrations betweenAprilandDecember2020tothehypotheticalscenariowhereregistrationsremained attheApril2020lowoverthatperiod;wecontinuetonormalizethedependentvariableby the2019valuestocontrolforseasonalpatterns–thatis,thedependentvariablesincolumns (4)-(6)isthefollowing ∑12 Regs ∑12 Regs log (cid:104) m=4 c,m,2020 (cid:105) −log (cid:104) m=4 c,m,2019 (cid:105) 9·Regs 9·Regs c,Apr.2020 c,Apr.2019 Thecoefficientestimatesforthisspecificationremainclosetothecorrespondingresultsin columns (1)-(3) even though they capture the entire path of the recovery in vehicle sales thatoccurredinAprilandthesubsequentmonthsrelativetothecounterfactualofanApril low. Intermsofmagnitudes,wefindthataone-standard-deviationincreaseinper-capita PPP loans explains 2.5 percent of a standard deviation higher sales over the April to December2020periodrelativetothecounterfactual. Recalltheaforementionedincreaseinthelikelihoodofnew-vehiclepurchasesbyselfemployed individuals from table 2. Similarly, the county-level PPP regression results in table4indicatethatcountiesreceivingmorePPPloansdirectedtowardstheself-employed alsoexperiencedafasterrecoveryintotalregistrations. Wecanalsotranslatetheempiricalresultsintothenumberoflightvehiclesalesengendered by the PPP program. The per-capita (not per loan) PPP disbursement was $ 2,054. Using the coefficient from column (4) of table 3—our preferred specification—PPP fundingaccountsforroughlya500,000unitincreaseinannualizedregistrationsin2020.26 Thus, togetherwiththeEconomicImpactPayments,thesetwofiscalsupportprogramsboosted vehiclesalesbyasmuchas13/4millionunits. 26Ideally,wewouldliketocapturetheincreaseinlightvehiclepurchasesfromthedisbursementlogpercapitaPPPfundsascapturedinourregressionframework.AveragePPPfundspercapitawere$2,000.Resultingly,wewillcalculatethechangeinregistrationsfroma$2000increaseinPPPfunds,whichisequivalenttoamovementfromaboutthe5thtothe50thpercentileinthePPPpercapitadistribution(amovementfrom697dollarspercapitato2737dollars).Takingthe0.053coefficientfromcolumn4,ourpreferred specification,fromtabletable3andlogsofthedollarvalues,wewouldgetanincreaseof:0.053(7.59)- 0.053(6.547)=0.072.Thedifferenceingrowthinnewvehicleregistrationsbetween2019and2020relativeto ∑12 Regs ∑12 Regs Aprilisgivenbylog m=4 m,2020 −log m=4 m,2019 =0.48.SothePPPportionofthegainsis0.07/.48,or 9·Regs 9·Regs Apr.2020 Apr.2019 15percent.Ifthegrowthinsalesrelativeto2019betweenApriltoDecemberof2020was2.25millionunits, then15percentofthatincreaseatanannualrateisnearly500,000units. 11

OurresultsofapositiveassociationbetweenPPPfundsandpersonalregistrationsdocuments an unintended channel for the impact of fiscal intervention aimed at businesses– —thatis,thePPPloanswerenotonlysystematicallyusedtodirectlyacquirebusinessvehicles, but they also may have induced spillovers on personal purchases.27 Although we do not control for other fiscal programs that were enacted around the same time or other county-level time-varying characteristics, our estimates could be interpreted causally under the condition that the disbursement of PPP loans is uncorrelated with county-level characteristics influencing the changes in vehicle registrations; the assumption of the orthogonalityofthePPPfundstoomittedandunobservedcounty-levelfactorsisverylikely to hold due to the untargeted nature of the program over the time horizon of our analysis.28 3.3 ImplicationsforSubsequentVehiclePriceSurge All told, we estimate that the two fiscal programs we analyzed contributed to boost sales by 1 3/4 million units of sales in 2020. This boost of sales is particularly notable when compared with the inventory dynamics over the same period. In fact, inventory levels declined from 3.5 million units in 2019 to 1.1 million units in 2021. Had those programs not being enacted, our estimates imply that the counterfactual level of inventories would havestayedaround2.85millionunitsbytheendof2021.29 Inotherwords, ourestimates suggestthatthefiscalprogramsexplainabout70percentofthedropininventoryrelative to the pre-pandemic period. The declines in inventory, in turn, contributed to the price pressuresthatmaterializedinrecentyears. Infact,asshowninfigure8,newvehicleprices tendtorisefollowinglowinventorylevelsinpreviousperiods. Toquantifytherelationshipbetweenpricesandinventorylevels,weadoptthefollow- 27Thisresultisbroadlyconsistentwiththecominglingofpersonalandbusinessfinancesareparticularly salientforsmallbusinessesandstartups.SeeRobbandRobinson[2014],whichemploystheKauffmanFirm Surveytostudythecapitalstructurechoicesofentrepreneurs. 28Table3reportstheresultsthatanalyzePPPloansacrossallbusinessesandtheirimplicationsonnew vehicleregistrations. 29Inamarketwithnoproductionconstraints,theincreaseinsaledoesnotnecessarilymapontodeclines ininventory.However,nearlyallvehiclemanufacturersencounteredsubstantialsupplyconstraintsin2021, leavingthemessentiallyunabletoadjustoutputinresponsetosales. 12

ingmodel, logPrice = γ +γ Inventory +δX +d +d +ζ (3) vt 0 1 v,t−1 v,t v t vt whichrelates(log)MSRPsofvehiclevinyearttoinventorylevels(inthousandofunits)of vehiclevinyeart−1. Ourspecificationreliesonlaggedinventorylevelstocontrolforthe differentialreactionsofautomakersofprioritizingtheproduction—withensuingeffectson inventory levels–of more expensive models in response to semiconductor shortages. We augment our model to control for sales, a proxy for demand conditions, and the number of models per vehicle, to capture additional heterogeneity across vehicles. Furthermore, we add time and vehicle dummies to absorb aggregate shocks and unobservable timeinvariantvehiclecharacteristics. Table 5 reports our results. Columns (1)-(3) explore the relationship between prices and inventory levels using an OLS regression model; according to the estimates in those columns, a decline of 100,000 units in inventory is associated with large price increases, in the order of 0.5 percent; the coefficient estimate is little changed when switching from contemporaneous to lagged inventory levels, our preferred regressor—columns (2) and (3)—or after controlling for year fixed effects—column (3). Unobserved vehicle heterogeneity explain a large part of the negative association between inventory and prices: In fact,movingtotheFEregressionsincolumns(4)-(6),thecoefficientestimatesonlaggedinventorylevelsappearoneorderofmagnitudesmaller. Ourpreferredspecification,which includes all controls, is reported in column (5) and implies that a decline in inventory of 100,000 units is associated with an increase in vehicle prices of 0.07 percent.30 With the reliance on MSRPs, our results should be interpreted as a lower bound on the true relationship between inventory and prices; in fact, the decline in inventory took place at a timeofdecliningincentivesandrapidgrowthindealers’margins,whicharenotcaptured by our data. With this caveat in mind, a drop in inventory levels of 1 3/4 million units— themagnitudeoftheimpactofstimulusprogramsonsales—addedalmost1.3percentage 30Column6alsoinvestigateswhethertherehavebeenchangesintherelationshipbetweenprices,sales, andinventoryinthepost-pandemicperiod.Whiletheinteractionsbetweenthepost-pandemicdummyand eachofsalesandinventorylevelsarepositive,theyremainstatisticallyinsignificant,likelypointingtoatoo short-timeframetoidentifychanges.Furthermore,thecoefficientonlaggedinventoryisnotstatistically differentfromwhatisreportedincolumn(5),ourpreferredspecification. 13

points to the price increases over the subsequent years. As vehicle prices inflation rose almost 61/2 percentage points between the pre-pandemic and the post-pandemic period, ourestimatesaccountfor20percentofthepick-upinpricesinthelastfewyears. 4 Conclusion Bybringingnovelmotorvehicledata—ongeographicpatternsofconsumption, onprices andinventories,andonhouseholdbehavior—tobear,wehavefoundevidencethatseveral of the pandemic-era stimulus programs that were targeted at businesses and consumers notonlyhelpedmaintainautodemand,butcontributedtopricepressures. Specifically,we find that receiving a stimulus check in 2020 increased the probability of purchasing new vehicles by 1.7 percent, while subsequent waves boosted only the probability of used car purchases. Similarly,lookingatcounty-leveldata,thedisbursementoffundsfromthePaycheckProtectionProgramwasassociatedwithariseinnewcarregistrations;interestingly, theresultseemsrobustforbothpersonalandbusinessregistrations, alikelyindicationof thecominglingoffinancesforsmallbusinesses. Taken together, the fiscal programs account for a boost of 1 3/4 million units to sales in2020. Importantly,theseresultsstandincontrasttoTaylor[2022],whichdocumentsan insignificant impact on aggregate consumption. While our findings are evidence of the importanceofthefiscalprogramsinsupportingthepost-pandemicrecoveryineconomic activity, the boost in sales coincided with the presence of constraints in production—– rangingfromtemporarystoppagesattheonsetofthepandemictotheemergenceofinput shortages in 2021. As a result, the increase in sales following the enactment of the fiscal stimulus was sustained by a drawdown in inventory and ultimately fostered the emergence of new vehicle pricing pressures. Indeed, the stimulus programs not only boosted salesbutalsoexplainaround70percentoftheinventorydropthroughtheendof2021and added1.3percentagepointstothe price increasesoverthenextyears, contributingtothe pick-upinnewvehiclepricesrelativetothepre-pandemicperiod. There are several directions for fruitful future research. Detailed data on individual PPP loan recipients should be able to more directly link small businesses to vehicle pur- 14

chases and other consumption patterns. Also, this research omits other fiscal programs, suchasthechildtaxcredit,thatcouldalsoprovidetheimpetusforlightvehiclepurchases. Inaddition,thereisvalueinbetterunderstandingtheimplicationsofhousehold-leveleconomic behavior if there were no such provision of aid. To that end, exploring the heterogeneitywithintheConsumerExpenditureSurveyshouldfacilitatesuchwork. References Olivier Armantier, Leo Goldman, Gizem Kos¸ar, Jessica Lu, Rachel Pomerantz, Wilbert Van der Klaauw, et al. How have households used their stimulus payments and how wouldtheyspendthenext? Technicalreport,FederalReserveBankofNewYork,2020. Olivier Armantier, Leo Goldman, Gizem Kos¸ar, and Wilbert Van der Klaauw. An update on how households are using stimulus checks. Technical report, Federal Reserve Bank ofNewYork,2021. Alan Auerbach. Is there a role for discretionary fiscal policy? Working Paper 9306, National Bureau of Economic Research, November 2002. URL http://www.nber.org/ papers/w9306. David Autor, David Cho, Leland D Crane, Mita Goldar, Byron Lutz, Joshua Montes, William B Peterman, David Ratner, Daniel Villar, and Ahu Yildirmaz. The $800 billion paycheck protection program: Where did the money go and why did it go there? JournalofEconomicPerspectives,36(2):55–80,2022a. David Autor, David Cho, Leland D Crane, Mita Goldar, Byron Lutz, Joshua Montes, William B Peterman, David Ratner, Daniel Villar, and Ahu Yildirmaz. An evaluation of the paycheck protection program using administrative payroll microdata. Journal of PublicEconomics,211:104664,2022b. Erich Battistin. Errors in survey reports of consumption expenditures. Technical report, IFSWorkingPapers,2003. BureauofLaborStatistics. ConsumerExpenditureSurvey,2022. Olivier Coibion, Yuriy Gorodnichenko, and Michael Weber. How did us consumers use their stimulus payments? Technical report, National Bureau of Economic Research, 2020. Ryan A. Decker, Robert J. Kurtzman, Byron F. Lutz, and Christopher J. Nekarda. Across the universe: Policy support for employment and revenue in the pandemic recession. 15

AEA Papers and Proceedings, 111:267–71, May 2021. doi: 10.1257/pandp.20211058. URL https://www.aeaweb.org/articles?id=10.1257/pandp.20211058. Michael Hicks, Nalitra Thaiprasert, et al. A first look at the’cash for clunkers’ program. EconomicsBulletin,32(1):567–573,2012. Mark Hoekstra, Steven L Puller, and Jeremy West. Cash for corollas: When stimulus reducesspending. AmericanEconomicJournal: AppliedEconomics,9(3):1–35,2017. Informa Business Media, Inc. Wards Intelligence Data Query, 2022. URL https:// wardsintelligence.informa.com/data-query-tool. J.D. Power and Associates. Incentive Spending Report, 2021. URL http://www.jdpower. com/solutions/power-information-network-pin. Atif Mian and Amir Sufi. The effects of fiscal stimulus: Evidence from the 2009 cash for clunkersprogram. TheQuarterlyjournalofeconomics,127(3):1107–1142,2012. GovernmentAccountabilityOffice. Stimuluschecks: DirectpaymentstoindividualsduringtheCOVID-19pandemic. Technicalreport,GAO-22-106044,2002. Jacob Orchard, Valerie A Ramey, and Johannes F Wieland. Micro mpcs and macro counterfactuals: thecaseofthe2008rebates. Technicalreport,NationalBureauofEconomic Research,2023. Jonathan A Parker, Nicholas S Souleles, David S Johnson, and Robert McClelland. Consumer spending and the economic stimulus payments of 2008. American Economic Review,103(6):2530–2553,2013. Jonathan A Parker, Jake Schild, Laura Erhard, and David Johnson. Household spending responses to the economic impact payments of 2020: Evidence from the consumer expendituresurvey. Technicalreport,NationalBureauofEconomicResearch,2022. William Passero, Thesia I Garner, and Clinton McCully. Understanding the relationship: CE survey and PCE. In Improving the measurement of consumer expenditures, pages 181– 203.UniversityofChicagoPress,2014. R.L. Polk & Co. New Vehicle Registration Data, 2020. URL https://www.ihs.com/btp/ polk.html. Alicia M Robb and David T Robinson. The capital structure decisions of new firms. The ReviewofFinancialStudies,27(1):153–179,2014. MatthewDShapiroandJoelSlemrod. Didthe2008taxrebatesstimulatespending? AmericanEconomicReview,99(2):374–379,2009. 16

SmallBusinessAdministration. PPPFOIADatabase,2021. URLhttps://data.sba.gov/ en/dataset/ppp-foia. JohnBTaylor. Anempiricalanalysisoftherevivaloffiscalactivisminthe2000s. Journalof EconomicLiterature,49(3):686–702,2011. JohnBTaylor. Theeffectoftherescueplansandtheneedforpoliciestoincreaseeconomic growth. JournalofPolicyModeling,44(4):768–779,2022. 17

Tables Table1: EIPsandVehiclePurchases (1) (2) (3) (4) (5) (6) VehiclePurchases 2020StimulusChecks 2021StimulusChecks Variables Total New Used Total New Used Stimulus 0.039** 0.017* 0.027* 0.022 -0.009 0.032** (0.017) (0.009) (0.015) (0.018) (0.011) (0.015) AdditionalControls y y y y y y Obs. 1,893 1,893 1,893 1,974 1,974 1,974 R2 0.041 0.025 0.032 0.038 0.029 0.043 Source:BureauofLaborStatistics(BLS). Total:dummyequalto1forpurchaseofavehicleinquartertort+1. New:dummyequalto1forpurchaseofanewvehicleinquartertort+1 Used:dummyequalto1forpurchaseofausedvehicleinquartertort+1. Stimulus:dummyequalto1forthereceiptofstimuluscheckinquartert. Legend:∗∗∗significantat1%,∗∗at5%,∗at10%. Notes:LinearProbabilityModel.Thefirstthreecolumnslookattheimpactof thestimuluschecksdisbursedin2020(firstwave),whilecolumns(4)-(6)documenttheimpactofthestimuluschecksdisbursedin2021(secondandthird wave).Allcolumnsincludeafter-taxincome(inlog-s),adummyifincomehas beentopcoded,sizeofthehousehold,andvariousdemographics(age,race,gender,education,maritalstatus)oftheheadofthehousehold.Robuststandard errorsarereportedinparenthesis. 18

Table2: VehiclePurchases: Self-Employedvs. Others TypeofEmployer1 2017-2019Avg.2 2020-2021Avg.3 Self-employed 5.3 6.4 New 1.3 2.2 Used 4.0 4.3 Others 5.7 5.3 New 1.5 1.4 Used 4.2 4.0 Source:BLS. 1Fortheheadofhousehold. 2Averagefor2017–2019. 3Averagefor2020–2021. Notes:Carpurchasingbehaviorforself-employedcompared withothergroups.Valuesareinpercent.Othersincludehouseholdsemployedinaprivatecompany;inthefederal,state,or localgovernment;orinafamilybusiness/farm.Averagesare weightedbysurveyweights. Table3: PPPLoansandNewVehicleRegistrations (1) (2) (3) (4) (5) (6) LogChange,Apr.toDec. LogCum.Change,Apr.toDec. withApr.2020Counterfactual Variables Total Personal Business Total Personal Business logPPPLoans 0.068* 0.070* 0.112** 0.053** 0.051** 0.019 (0.036) (0.041) (0.048) (0.022) (0.023) (0.032) StateFE y y y y y y Obs. 2,124 2,124 2,124 2,124 2,124 2,124 R2 0.588 0.579 0.179 0.803 0.812 0.296 Source:R.L.Polk&Co.;SmallBusinessAdministration. LogChange,AprtoDec2020:logchangeinregistrationpercapitabetweenApril andDecember2020relativetothesamechangein2019. LogCum.Change,Apr.toDec.2020:logchangeintotalregistrationoverApril- December2020vs.counterfactualregistrationsthatremainedattheApril2020level overthatperiodrelativetothesamechangein2019. logPPPLoans:logPPPloansapprovedpercapita. Legend:∗∗∗significantat1%,∗∗at5%,∗at10%. Notes:County-levelregressions.Robuststandarderrors,clusteredatthecounty level,arereportedinparenthesis. 19

Table4: PPPLoansandNewVehicleRegistrations,Self-EmployedIndividuals (1) (2) (3) (4) (5) (6) LogChange,Apr.toDec. LogCum.Change,Apr.toDec. withApr.2020Counterfactual Variables Total Personal Business Total Personal Business logPPPLoans 0.042** 0.046** 0.058** 0.077*** 0.078*** 0.054*** (0.017) (0.019) (0.028) (0.011) (0.011) (0.020) StateFE y y y y y y Obs. 2,113 2,113 2,113 2,113 2,113 2,113 R2 0.587 0.579 0.178 0.812 0.820 0.300 Source:R.L.Polk&Co.;SmallBusinessAdministration. LogChange,AprtoDec2020:logchangeinregistrationpercapitabetweenApril andDecember2020relativetothesamechangein2019. LogCum.Change,Apr.toDec.2020:logchangeintotalregistrationoverApril- December2020vs.counterfactualregistrationsthatremainedattheApril2020level overthatperiodrelativetothesamechangein2019. logPPPLoans:logPPPloansapprovedforself-employedindividualspercapita. Legend:∗∗∗significantat1%,∗∗at5%,∗at10%. Notes:County-levelregressions.Robuststandarderrors,clusteredatthecounty level,arereportedinparenthesis. Table5: PricesandInventoryLevels (1) (2) (3) (4) (5) (6) Variables logPricet Inventoryt -0.513*** (0.083) Inventory t−1 -0.476*** -0.462*** -0.054*** -0.074*** -0.055*** (0.084) (0.085) (0.013) (0.015) (0.019) Post2020*Inventory t−1 0.027 (0.031) Salest 0.015 0.012 (0.017) (0.018) Post2020*Salest 0.006 (0.011) Num.Models 0.003** 0.003** (0.001) (0.001) YearFE n n y y y y VehicleFE n n n y y y Obs. 2,079 1,511 1,511 1,511 1,499 1,499 R2 0.044 0.051 0.060 0.224 0.240 0.242 NumberofVehicleIDs 386 386 386 Source:InformaBusinessMedia,Inc.. logPricet:manufacturers’recommendedstickerpricesforvehicles. Inventoryt:inventorylevels,inthousandunits,attimet. Salest:sales,inthousandunits,attimet. Post2020:dummyequalto1for2021and2022. Num.Models:numberofmodelspervehicleID. Legend:∗∗∗significantat1%,∗∗at5%,∗at10%. Notes:OLSregressionsincolumns(1)-(3),FEregressionsincolumns(4)-(6).Robuststandard errors,clusteredatthevehicleIDlevel,arereportedinparenthesis. 20

Figures Figure1: LightVehicleSales 21

Figure2: MotorVehiclePurchases Source: BureauofLaborStatistics;R.L.Polk&Co.. Figure3: MotorVehiclePrices Source: BureauofLaborStatistics;J.D.PowerandAssociates. 22

Figure4: EconomicImpactPayments Source: BureauofLaborStatistics. Note: Aggregatereceiptsofstimuluschecksacrossallhouseholdsbythemonthofreporting. Figure5: EconomicImpactPaymentsbyQuartile Source: BureauofLaborStatistics. Note: Percentilesinthedistributionofthevalueofstimuluschecksacrossallhouseholdsbythe monthofreporting. 23

Figure6: HowHouseholdsUsedEconomicImpactPayments Source: BureauofLaborStatistics. Figure7: TheImpactofPPPLoansonNewVehicleRegistrations Source: R.L.Polk&Co.;SmallBusinessAdministration. Note: Relationshipbetween(log)PPPloanspercapitaandchangesin(log)registrationatthe countylevel. EachdotrepresentsaU.S.county,anditssizeisproportionaltoitspopulation. The redlinedenotestheregressionline,whiletheshadedgrayarearepresentsthe95-percentconfidenceinterval. Topandbottomonepercentofobservationshavebeendropped. 24

Figure8: TheRelationshipbetweenNewVehiclePricesandInventoryLevels 25

A Additional Tables TableA1: Pre-andPost-PandemicVehiclePurchases (1) (2) (3) VehiclePurchases Variables Total New Used Pandemic -0.005** -0.003** -0.003 (0.002) (0.001) (0.002) StimulusAssignment -0.006*** -0.008*** 0.000 (0.002) (0.001) (0.002) StimulusAssignmentxPandemic 0.002 0.001 0.002 (0.003) (0.002) (0.002) AdditionalControls y y y Obs. 191,317 191,317 191,317 R2 0.012 0.006 0.012 Source:BLS. Total:dummyequalto1forpurchaseofavehicleinquartert. New:dummyequalto1forpurchaseofanewvehicleinquartert. Used:dummyequalto1forpurchaseofausedvehicleinquartert. Pandemic:dummyequalto1for2020andfollowingyears. StimulusAssignment:dummyequalto1forhouseholdswhosebeforetaxincomeisequalto$75,000orless. Legend:∗∗∗significantat1%,∗∗at5%,∗at10%. Notes:LinearProbabilityModel.Allcolumnsincludeadummyfor thestimuluswaveanditsinteractionwiththestimulusreceipt,aftertaxincome(inlog-s),adummyifincomehasbeentopcoded,sizeof thehousehold,andvariousobservables(age,race,gender,education, maritalstatus)oftheheadofthehousehold.Robuststandarderrors, clusteredatthehouseholdlevel,arereportedinparenthesis. 26