Measuring the Incidence and Effects of Unemployment ...

Measuring the Incidence and Effects of Unemployment Insurance Benefits During the COVID-19 Crisis Evidence from Administrative Records from California Alex Bell, CPL UCLA Geoff Schnorr, CPL UCLA & UC DAVIS Till von Wachter, CPL UCLA, CEPR, NBER

KeynotePresentationLED/LEHDConferenceApril30,2021

Unemployment During the COVID-19 Crisis in the U.S. & California

2

PublichealthmeasuresinresponsetoCOVID-19ledtostaggeringnumberofUIclaims•  Over50millionworkersfiledforUIbenefitsinU.S.,about10millioninCalifornia

CrisisalsoledtolargeshiftinthecompositionofUIclaimantstowardslow-wageworkers•  Asubstantialincreaseinclaimsfromloweducated,younger,non-white,andwomen

FastmovingnatureofcrisisputmoreemphasisonweeklyUIclaimsdata•  Somelimitationsofthatkeysourceofinformationbecameapparent

Inresponse,federalgovernmentsubstantiallyincreasesUIbenefitpayments•  Ledtoadebateonwhetherextendingsuchgenerousbenefitsiscounterproductive

è ImportanttorevisitwhatwecanlearnfromUIclaimsandwhatweknowabouttheeffectofUIclaimsonlaborsupply

Unemployment Insurance Extensions during COVID-19 Crisis in U.S.

3

NewMajorPolicyResponses:

•  LargeincreaseinbenefitlevelsthroughFederalPandemicUnemploymentCompensation(FPUC,$600/week),LostWagesAssistance(LWA,$300/week)

•  Thisledtoriseinincomereplacementratetoover100%formanyworkers•  PartlyoutofconcernforworkdisincentivesprogramexpiredendofJuly•  Commondebateinrecessions:doUIextensionsdomoreharmthangood?

•  ExtendedUIcoverageforself-employedandnotcoveredlowincomeworkersthroughPandemicUnemploymentAssistance(PUA)program

MoreCommonMajorPolicyResponses:•  Extendbenefitdurationfinancedbyfederalgovernment•  FullyfinanceWorkSharing(Short-TimeCompensation)benefits•  Providefundsforstates’administrationofUIbenefits

UI Claims during COVID-19 Crisis & Labor Supply Effects

4

Step1:AnalyzedifferencesinUIclaimsduringCOVID-19inCalifornia

•  ExploitunusualaccesstomicrorecordsfromCaliforniaUIsystem

•  Allowustoanalyzereal-timedifferencesbyeducation,demographics,industry

•  Allowustomeasure(forthefirsttime!)forwhichdatesindividualsreceivedUI

Step2:EstimateeffectofUIbenefitlevelsonlaborsupplyusingkinkedbenefitschedule

•  ExploitidiosyncraticfeaturesofUIbenefitsandchangesduringthecrisis

•  CompareeffectduringCOVID-19witheffectsinyearsbeforecrisis&GreatRecession

PreliminaryBottomLine:

1.   UImicrodatacanprovidesubstantiallymoreinformationthatpublisheddata

2.   HigherUIbenefitsprolongedunemployment,butmaynothavereducedemployment

Unique Administrative Data on Unemployment Insurance (UI) Claims

5

UIClaimsData:

Informationontimingofinitialclaims&benefitreceipt

Informationondemographics,residence,industry,prioremployer

Sample:

ConsiderallUIclaimsoccurringduring2020-2021

Lateron,bringindatagoingbacktotheyear2000

Whyisthisunique:

1.  Canrevisitsomekeymeasurementissues

2.  Provideinformationthatwasnotavailablebefore

3.  Canassociateclaimbehaviortocircumstancesofindividualandcommunity

BreakingDownInitialClaimsinCalifornia–ChurnintheUISystem

InCA,initialclaimsstillatpeakofGreatRecessionearly2021.AdditionalclaimoccursifaworkerwasonUIandleavesUIforinterveningemployment.MostinitialclaimantsareadditionalclaimantsinCAsinceFall2020.PUAplayedanimportantroleinSpringandSummerof2020.

Source:CaliforniaPolicyLab,MarchUIReport

(Note:U.S.DOL’s“InitialClaims”CombinesNewInitialClaimsandAdditionClaims.)

StockofUIRecipientsvs.NumberofWeeksCertifiedinCalifornia

Source:CaliforniaPolicyLab,MarchUIReport

ThestockofUIrecipientsintheU.S.istypicallymeasuredbythenumberofpeoplewho“certify”forbenefits–i.e.,thosethatconfirmatabi-weeklylevelthattheyarestillunemployed.

Thisisonlyagoodmeasureifa)claimantsroughlycertifywhentheyareunemployed,b)claimantsdonotcertifyretroactivelyformanyweeks.

Forthefirsttime,CPLcalculatedthestockbytheweekofactualunemployment,whichevolvesquitedifferentfromthenumberofcertifications.

(NotethatinCA,certificationisbi-weekly,sotheno.ofpeoplecertifyingisroughlyhalfofno.ofcertificationsinabsenceofretroactiveclaims.)

More UI Claims Among More Vulnerable Workers

Over1in2workerswithahighschooldegree,youngworkers,Blackworkers;over1in2womenhavefiledaUIclaimfromMarch2020toApril2021.StatewidethefractionofworkersfilingaregularUIclaimwas30%,48.2%includingPUA.

Incontrast,thefractionamongmatureworkersorthosewithaBachelor’sdegreearesmaller.TheriseinclaimsbymorevulnerableworkersispartlyexplainedbyalargeinitialamountofclaimsfromAccommodation&FoodServicesandRetailTradeindustries.

8

ACloserLookAtOurPreferredMeasureDuringtheCOVID-19Crisis

Seelowpre-crisisrecipiencyrates.Byendoftheyear,recipiencyratearound90%.Throughout,therewasa30-40percentagepointdifferenceinrecipiencyratebetweenhighestandlowestcounty.

Source:CaliforniaPolicyLab,FebruaryDataPoint

Taking a Closer Look at the Recovery HowhavedifferentneighborhoodsinLACountyfaredsincethepeakofthecrisis?

Inglewood,Comptonareaslowertorecover.

Futureworkwilldirectlyanalyzegeographicpatternsandcorrelations.

10

RecipiencyRateDifferencesbyCountyinCalifornia

Recipiencyratesofcountiesvaryinpredictablefashion.Countiesthatarepoorer,havelessbroadbandaccess,aremoreHispanic,andhavefewerEnglishspeakershavelowerrecipiency.ThesameholdsattheCensusTractlevelaswell.

Source:CaliforniaPolicyLab,FebruaryDataPoint

FPUC Helps UI Claimants Avoid Near-Poverty Level Benefit Levels

TheaverageWeeklyBenefitAmount(WBA)forregularUIbenefitsfluctuatedbetween$300-330duringthecrisis.

LowerWBAimplieslowerpriorearnings,sinceWBAisapproximately50%ofprioraverageweeklyearnings.

WithoutFPUC,$330isbelow30%ofMedianFamilyIncomeinCA,andthuswouldbeconsidered“ExtremelyLowIncome”byHUDstandards.

With$300LWAorPACpayment,totalbenefitsriseabove“VeryLowIncome”level,butstillfarbelow“LowIncome”threshold.

12

Estimating the Effect on Labor Supply During COVID-19 Crisis

13

ApproachNo.1:Exploitthefactthatbenefitschedulehasakinkinit

•  InCalifornia,benefitsare50%ofearningsinabaseperiodupto$450,creatingkink

•  Baseearningsarethehighestquarterearningsof5mostrecentcompletedquarters

èEstimateastandard‘RegressionKinkDesign’(RKD)1.  Assesswhetherindividualsjustaboveandbelowthebenefitkinkarecomparable2.  SeewhethertherateofexitingUIalsoexhibitsakinkwhenbenefits‘kink’

ApproachNo.2:Exploitthat$300/weekLWAbenefitonlypaidifUIbenefits>$100/week

•  Compareunemploymentofworkerswithjust$100UIbenefitsvs.thosejustbelow$100

•  Hardertouseturnon/offofbenefitsupplementsforallworkersbecauseoftimetrends

Approach 1: Kinked UI Benefit Schedule in California (& other states)

14

0

200

400

600

800

1000

1200

WBA

0 20000 40000 60000Highest Quarter Wages

Pre-Pandemic

Pandemic

Clear Labor Supply Responses Around Benefit Kink (2011-2019)

15

.58

.6

.62

.64

.66

.68

8-w

eek

Surv

ival

-5000 0 5000Highest Quarter Wages Relative to Kink

Benefitsincreaseby50centsper$1priorearnings Benefitsflatat$450

Clear Labor Supply Responses Around Benefit Kink

16

.58

.6

.62

.64

.66

.68

8-w

eek

Surv

ival


Benefitsincreaseby50centsper$1priorearnings Benefitsflatat$450

Sharpdifferenceinslopesàeffectof(lackof)addedbenefits

counterfactualwithnokinkinbenefits

Benefit Effects from 2006 to 2020 Expressed in Percent Terms To Account for Differences in Benefit Levels

17

0

.2

.4

.6

.8

1

% D

urat

ion

Incr

ease

from

1%

WBA

Incr

ease

2006 2008 2010 2012 2014 2016 2018 2020year

MedianUSelasticity≈ .38(Till’ssurvey)

1%increaseinbenefitsà≈ .25-.3%durationincrease(doesnotseemtovaryw/recessions)

Approach 2: Discontinuity in Receipt of $300/week LWA Benefit

18

For6weeksstartingonJuly26th,individualscouldreceiveadditional$300/weekfromLostWagesAssistance(LWA).ButonlyiftheirWeeklyBenefitAmount(WBA)wasatleast$100.FigureclearlyshowsthatthoseworkersreceivingLWAhadalowerrateofUIweeklyexit.(BasedonaverageexitratesforindividualsreceivingbenefitsendofAugust.)

Approach 2: Discontinuity in Receipt of $300/week LWA Benefit

19

SizeofLaborSupplyEffect:

•  At$100WeeklyBenefitsAmount,benefitsincreaseby300%!

•  Averageunemploymentdurationincreasedby~20%,orabout5weeks

è Estimateimpliessimilarmagnitudeasregressionkinkdesign!

Interpretation:

1.   ClearevidencethatUIbenefitsraisedunemploymentdurationsduringthecrisis

2.   Onlyaconcerniftheseworkersgetworsejobs–workinprogress

3.   MayhavepreventedinfectionswithCOVID-19–workinprogress

4.   NotevidencethatUIbenefitsloweredtotalemployment

Conclusion

20

UIDataAllowedustoObtainSomeKeyInsightsIntotheUnemploymentCrisis

•  NovelinsightsintoarangeofaspectsofUIsystem

Analyzedincidence&laborsupplyeffectsofUnemploymentInsuranceinCOVID-19Crisis

•  UseduniqueadministrativemicrorecordscoveringallCaliforniaforover20years

•  UsedRegressionKink&DiscontinuityDesignstoobtaincausaleffectofUIbenefits

OverallourCurrentFindingsIndicate:

1.   Individual-levelUIdataallowedtosubstantiallyimprovemeasurementofincidence,adequacy,andrecipiencyofUIduringthecrisis

2.   IncreasedUIbenefitsduringthecrisislikelyincreaseddurationofUIbenefitreceipt,thesameorlesscomparedtobeforethecrisis

21

BONUSSlides–RKD

Comparison of Magnitudes with Respect to Prior Literature

22

Differenceinourpaper:MostU.S.papersuseUIdurationinweeks

Thatsaid,comparedtothetypicalbenefitelasticity,ourestimatesareonthelowerend:

•  LiteratureSurvey(Schmieder&vonWachter2018):U.S.medianelasticity0.38

•  RKD5U.S.states1976-84(Landais2015):0.21-0.7

•  RKDMissouri2003-2013(Cardetal.2015):Expansion0.35,Recession0.65-0.9

•  U.S.1985-2000(KroftandNotowidgdo2016):0.6(Recession0.3)

Variationofbenefitelasticitywithunemploymentrate:

•  Mixedevidence-Kroft&Notowidigdo(declining)vs.Cardet.al.(increasing)

•  Effectofbenefitdurationclosetoa-cyclicalinSchmieder,vonWachter,Bender(2012)

èOurestimatesareofplausiblemagnitude&provideimportantupdatetoliterature

No Discernible Change in Density of UI Beneficiaries Around Kink

23

.021

.022

.023

.024

.025

Den

sity

of C

laim

s

6000 8000 10000 12000 14000 16000Highest Quarter Wages

Pre-Pandemic

Pandemic

Labor Supply Effect at Kink has Declined During COVID-19 Crisis

24

.58

.6

.62

.64

.66

.68

8-w

eek

Surv

ival


Benefitsincreaseby50centsper$1priorearnings

Downwardpivotmaybe:1)  Behavioralresponses

to+$6002)  Disproportionate

impactofcrisisonlowerearners,etc.

Benefitsflatat$450+$600


25

.58

.6

.62

.64

.66

.68

8-w

eek

Surv

ival









26

.58

.6

.62

.64

.66

.68

8-w

eek

Surv

ival








Responsetolostbenefitsmuchsmallerthanearlier


27

.58

.6

.62

.64

.66

.68

8-w

eek

Surv

ival






ResponsetolostbenefitsmuchsmallerthanearlieràButalsocomingfromalargerbase…letuslookatthenumbers



Robustness with Respect to Changes in Bandwidth

28

-.00008

-.00006

-.00004

-.00002

0

.00002

.00004

.00006

.00008

Red

uced

-For

m E

stim

ate

0 2000 4000 6000 8000 10000HQW bandwidth

Role of Recall for Chances of Reemployment [BONUS]

29

Differences in Labor Supply Effects at Kink by Recall Status [BONUS]

30

-.00001

0

.00001

.00002

.00003

Diff

eren

ce in

RF

Estim

ate

for t

hose

exp

ectin

g re

call

2000 2005 2010 2015 2020year

Interpretation:Generallylowerbehavioralresponseforthoseexpectingrecall(lessnegativeslopearoundkink)

31

BONUSSlides–RKD

Robustness with Respect to Changes in Bandwidth

-.1

0

.1

.2

.3

RD

Est

imat

e

0 50 100 150Bandwidth

Variation in Effect by Calendar Week

0

.1

.2

.3

RD

Est

imat

e

Aug 1

Aug 8

Aug 15

Aug 22

Aug 29

Sep 5

Week of Unemployment

Measuring the Incidence and Effects of Unemployment ...

Documents