ijems | scientific article TheEffectsofJobRetentionSchemes onEmploymentPreservation duringthe covid-19Epidemic inEuroAreaCountries anton rop rop.anton@yahoo.com Inthispaper,weanalysetheeffectsofdifferentjobretention(jr) schemestake-upsonthepreservationofemploymentduringthe co- vid-19pandemicineuroareacountries.Wefindthat jr schemesin euroareacountrieshelpedreducejoblossesduringthepandemic.The mosteffectiveinpreservingemploymentweretake-upsofthemost extensivelyupdatedpre-existingshort-timework(stw)schemesthat weremoregenerousandincludednonstandardworkers.However,the impactof jr schemeswaslessthantheoverallemploymentpreser- vationachieved.IncontrasttotheGreatRecession,macroeconomic measuresofeconomicsupportalsohelpedpreservejobsduringthe pandemicaswell.Correspondingdifferencesinsectoralemployment preservationeffectsshowthatsuchmacroeconomicsupportledto morejobsbeingkept,especiallyinthegroupofvulnerableservicesec- tors. KeyWords:covid-19pandemic,jobretentionschemes,short-time workschemes,macroeconomicmeasures https://emuni.si/ISSN/2232-6022/17.115-140.pdf introduction Jobretention(jr)schemesfeatureamongthekeyinstrumentsformit- igating the effects of the lockdowns on employment and social hard- ship introduced or extended by different countries in response to the crisis. Such schemes provided strong income support to workers with reducedworkinghours,reducedincomelosses,bolsteredaggregatede- mand,andsignificantlyloweredthenumberofjobsatriskofbeingter- minatedduetoliquidityconstraints(oecd 2021). jr schemes can take the form of short-time work (stw) schemes thatdirectlysubsidisehoursnotworked,suchasGermany’sKurzarbeit or France’s Activité partielle. They can also include wage subsidy (ws) volume 17 | 2024 | number 1 | 115–140 [116] AntonRop schemes that subsidise hours worked, and can in addition be used to top up the earnings of workers with reduced hours, such as the Netherlands’ Emergency Bridging Measure (Noodmatregel Overbrug- ging Werkgelegenheid, now ) or the JobKeeper Payment in Australia. A crucial aspect of all these jr schemes is that employees keep their contracts with the employer even if their work is fully suspended. Ac- cordingtoanoecdassessment(oecd2020a),inthesecondquarterof 2020,whentake-upratespeaked, jrschemeswerebeingimplemented in almost all oecd countries, covering around 60 millionworkers. In comparison, during the Great Recession, jr schemes only included some6millionworkers, eventhough16 oecd countrieslaunched jr schemesorhadimplementedtheschemesalreadyearlyoninthecrisis and7 oecd countrieshadintroducednewschemesduringthatperiod (HijzenandVenn2011). However, research on the impact of jr schemes on employment preservationinthepandemichasbroughtmixedresults.Adams-Prassl et al. (2020) find that in Germany, with a well-established short-time work(stw)scheme,34ofemployeesinworkattheonsetofthepan- demichadbeenaskedtoreducetheirhourstobenefitfromthisscheme. InApril2020,only5ofGermanworkershadlosttheirjobscompared tothe usa andthe uk wheretherespectivefigureswere20and17 ofindividuals(Adams-Prassletal.2020).Similarly,theJobKeeperPay- ment scheme in Australia is estimated to have saved one in five jobs (BishopandDay2020).Incontrast, oecd analysisbasedonthetake- up¹ of jr schemes in oecd countries shows jr schemes had a rela- tively small effect on employment (compared to data on employment inhoursdecline),accordingtowhichremovingthe jr schemeswould haveledtoadropinemploymentofbetween6and11(oecd 2021). Besidesthehighcosts,anintriguingaspectofrunningintensive jr schemesmightbethe‘deadweighteffects,’namely,theriskofsupport- ingjobsthatactuallydonotneedsupport(oecd 2021).Thus,govern- mentscouldbereluctanttousetheexistingsupportschemesoncethey discovertheirlimitedreach.Thismightexplainwhyinmost eu (devel- oped) countries scheme cover grew strongly in the second quarter of 2020 buteasedconsiderablyalreadybythenextquarterto fallbehind the second quarter level in almost every country. This was also seen ¹Take-up rates refer to actual use and are calculated as a share of total employees in short-timework(oecd 2020a). ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [117] duringthesecondandthirdwavesofepidemicwhenthepick-upineco- nomic activity continued, even though in the most of these countries schemecoverremainedavailableuntilthemiddleof2022. Inadditiontojrschemes,eucountriessoughttoreducetheharm- ful effects of the lockdowns on economic activity, employment, and socialhardshipbylaunchingotherpowerfulfiscalandmonetarymea- sures,whichhadstronganddifferenteffectsinvarioussectorsofeco- nomicactivityandhencealsoonemployment.Giventhat jr schemes were simultaneously operating alongside those policy measures, eval- uating the effects of jr schemes by only considering firm-based em- piricalevidenceonthepure take-up of jr schemes’covercouldbebi- ased.Moreover,whileevaluatingtheemploymentpreservationeffects of such take-ups, one must also consider the sectoral different inten- sity of the impact of the economic support measures. Sectoral use of jrschemesduringthecovid-19pandemicisquiteunlikethatseenin the Great Recession. During the first three waves of the epidemic, jr schemesaffectedemploymentacrossmanysectorsandtypesoffirms, whereasintheGreatRecessionalmost80ofjrschemetake-upswere concentratedinmanufacturing(oecd 2020a). In this paper, we analyse the effects of different job retention (jr) schemes’take-ups onemployment preservation duringthe covid-19 pandemicineuroareacountriesconsideringthecompleteportfolioof policymeasuresandsectoraleffectsasacrucialnon-policy-relatedfac- tor.Wefindthat jr schemesintheeuroareacountrieshelpedreduce job losses during the pandemic. The most effective in retaining em- ployment weretake-ups ofthemostextensivelyupdated,pre-existing short-timework(stw)schemesthatweregenerousandincludednon- standard workers. However, the impact of jr schemes take-ups was lessthantheachievedlevelofemploymentthatwaspreserved.Corre- spondingdifferencesinsectoralemploymentpreservationeffectsshow that macroeconomic support eased the loss of employment especially inthegroupofvulnerableservicesectors. Our study complements previous research (Hijzen and Venn 2011; Aiyar and Dao 2021) since assessments of what determines the size andquarterlydynamicsofjrschemestake-upssupport,andtheirem- ploymentpreservationeffectsbysectorsandbasedonmacroeconomic dataarerareoronlypartial.Thestudycontributestotheliteratureon the implementation and effectiveness of various jr scheme take-ups indifferentcountriesregardingseveralcovid-19wavesandsectors.In volume 17 | 2024 | number 1 [118] AntonRop addition, the effects of other macroeconomic (non-jr schemes) mea- sures on employment as well as potential sectoral differences in such macroeconomicemploymentsupportarealsopresented. literature review Although in the Great Recession jr schemes covered ten times fewer workers thanintherecentpandemic,theirimplementationandeffec- tiveness soonbecameasubjectofacademicresearch.Asearlyas2011, theoecdconductedadetailedanalysisofjrschemes’impactandrole duringtheGreatRecession(HijzenandVenn2011).Thestudydescribes the characteristics of the schemes implemented (albeit, it deals solely withstwschemes)andevaluatestheireffectivenessinpreservingem- ploymentintheshortandlongrun(inbustandrecovery).Itunderlines two important potential shortcomings of these schemes. First, it as- sessedthattheimpactonjobswassmallerthanthepotentialnumber ofjobssaved,indicatingweaktargetingand,second,thattheschemes led to greater labour segmentation if limited to workers holding per- manentcontracts. Similar ambition and results may be seen in a study by Boeri and Bruecker(2011).Theyfoundthat stw helpedreducejoblossesduring theGreatRecession.Still,accordingtotheirmacroeconomicestimates, thenumberofjobssavedwaslessthanthefull-timeequivalentjobsin- volved in these programmes, in some cases pointing to sizeable dead- weightcostsentailingthesamemoralhazardproblemsasthosearising withtheprovisionofunemployment insurance.Workers andemploy- ersmightcolludetoextractpaymentsfromthestateevenwhenincen- tives for reductions in hours would notbe required to avoid layoffs as thefirmwasnolongerfacinganegativedemandshock. The performance of jr schemes during the recovery period of the Great RecessionepisodewasanalysedinbyHijzenandMartin (2013). They foundthat stw raisesthe output elasticity ofworking time and helpspreservejobsinthesizeablecontextofarecessionbymakingem- ploymentlesselasticwithrespecttooutput.Akeyfindingwasthatthe timingof stw wascrucial. One can also find several papers dealing with jr schemes’ per- formance during the Great Recession in specific countries. See, for example, Bellmann, Gerner, and Upward (2015) regarding Germany, Calavrezo,DuhautoisandWalkoviak(2009) analysingthesituation in France, and Siegenthaler and Kohl (2019), describing the Swiss expe- ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [119] riences with jr schemes during the Great Recession and afterwards. Asthepandemicdeveloped,manystudiestrackedtheinitialimpact ofthecovid-19inducedcrisisontheusaandEuropeancountriesre- gardingemployment,hoursworked,andincome(InternationalLabour Organization 2020; 2021; Zimpelmann et al. 2021; Cotofan et al. 2021; Andertonetal.2020).GangopadhyayaandGarrett(2020)comparedthe levelofunemploymentinbothcrises:theGreatRecessionandcovid- 19. They found that during the Great Recession unemployment in the usa reached 10, while during the pandemic unemployment spiked at12.8.Andertonetal.(2020)analysedthecovid-19pandemic’s impactontheeuroarealabourmarketfromtheperspectiveofthecu- mulative contribution of four specific economic shocks to changes in total hours worked and the labour force: a technology or productivity shock,ashockinthelaboursupply(viaashocktolabourforcepartic- ipation), a shock giving rise to an increase in the demand for labour, andawagebargainingshock. The oecd (2020a) analyses the jr schemes that oecd countries reliedonduringthe(firstwaveof)thecovid-19pandemic.Theoecd estimates that stw schemes typically allow reduced working time at zero cost to firms, while ws schemes generally permit larger reduc- tionsinlabourcoststhanstwschemes,yetareassociatedwithgreater fiscal costs or weaker income protection for workers. Due to the bet- ter targeting of stw subsidies to firms likely to experience financial difficulties, they are probably more effective at saving jobs than ws schemes. Accordingto oecd simulations based on the single-hit sce- nario, stw subsidiesreducetheshareofjobsatriskby10percentage points from 22, whereas this is only 7 percentage points under ws. Asmallersectionofthestudyalsodiscussesthesectoraldimensionof the jr schemes’effects. Another study for g20 countries (oecd 2020b) finds that diverse working arrangements offered less security and were concentrated in affected sectors. Workers in a range of employment forms that vary from a full-time wage and salary work under a permanent contract – such as self-employed workers, those on temporary, on-call or part- time contracts, and informal economy workers – have been very vul- nerabletothejobandincomelossestriggeredbythepandemic. An oecd (2020c) study stresses that the sectors most directly af- fectedbythecovid-19containmentmeasuresaccountforaround40 oftotalemploymentandthesesectorsemployalargeshareofnonstan- volume 17 | 2024 | number 1 [120] AntonRop dardworkers,i.e.,part-timeworkers,self-employed,andworkershired underfixed-term contracts. Relative topermanentemployees, tempo- rary workers have a higher risk of losing their jobs and less chance of beingenrolledinshort-timeworkschemes. The oecd (2021) devotes a separate chapter to the jr schemes in place during the first three waves of epidemic. The paper tackles the sizeandvolatilityofthe jr schemestake-ups, dealswithsectoraldif- ferencesintake-ups,aswellasthedependenceofemploymentsupport onthesizeofworkersincome. Several studies of jr scheme effectiveness have looked at pro- grammesinparticularcountries,yettheirresultsarealsoinconclusive. Smaller estimates than expected are also evident in an imf study for Germany (Aiyar and Dao 2021), whereas estimated effects of the Job- Keeper Payment scheme in Australia show just the opposite – much higher effects, with the jr scheme being estimated to have saved the jobofoneinfiveemployees(BishopandDay2020).Resultsalsodiffer significantlyforstudiesofthesameschemeandcountry,suchasstud- ies ofthe PaycheckProtectionProgram (ppp)used inthe usa (Autor etal.2020;HubbardandStrain2020). data, descriptive statistics of main variables and hypotheses Data We can generally capture the observations made in the previous sec- tionsinthefollowingequation: Jobpreservation = f(jr schemes,macroeconomiceffects oftheportfoliooffiscalandmonetary measures,sectors)( 1 ) Weexplainthesourcesofdataandconstruction ofthemainvariables below. Ouranalysisisbasedonquarterlydatafromthefirstquarterof2019 until the second quarter of 2021 (q1 2019–q2 2021) for 19 euro area countriesand9sectors.Foreachsectorandeuroareacountry,wetake seasonallyadjusted dataonemployment inhours andemployment in persons from Eurostat (https://ec.europa.eu/eurostat). We normalise employmentdatabysettingtheaverageemploymentlevelachievedin 2019foreachcountryas1.Thisallowsustoconstructtheemployment ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [121] preservationindicatorasaratioofemploymentinpersonsandemploy- mentinhours,whichservesasthemaindependentvariableinthere- search.Withtheemploymentpreservationindicator,foreachquarter, country and sector, we measure the level of employees per number of hoursusedrelativetotheaveragein2019soastocapturetherelative level of employees who remained in employment despite the decline inthenumberofworkinghours.Forinstance,inthesecondquarterof 2020,i.e.duringthefirstcovid-19wave,theratioforGermanyis1.10. ThisindicatesthatinthisquarterGermanyrecordedalevelofemployed personsthatwas10higherthanthelevelobservedinworkinghours. It is thus evident that the number of employees in Germany dropped considerablylessthanthenumberofworkinghours.Thiswasactually thecaseforalleuroareacountries. To assess the impacts of different jr schemes, we use the oecd classification whereby countries use five types of jr schemes, name- ly,besidesthefour stw schemesalso ws (oecd 2020a).Accordingto the oecd study,23countrieswithapre-existing stw schemerapidly adjustedtheirstwschemetocopewiththecovid-19crisis(oecd 2020a). They applied different combinations of three key changes: (1) simplifying access and extending coverage; (2) extending coverage to non-permanent workers; and (3) making them more generous. Boeri andBruecker(2011)arguethatmakingthebenefitsmoregenerouspro- vides the subject workers with stronger support whilegranting access fornonstandardjobsmeansthatbettertargetingcanbeachievedsince workersholdingnonstandardjobs–i.e.theself-employedandworkers withtemporaryorpart-timedependentemployment–areveryvulner- able to job and income losses.² However, employers have little or zero incentivetouse stw fornonstandardjobsastheyknowtheseworkers can be fired at little or no cost, meaning access for nonstandard jobs shouldprobablybecombinedwithmoregenerous stw benefits. A number of countries have introduced temporary ws in response tothecovid-19crisisthatcanbeusedbyfirmsforhoursworked(like standard wage subsidies) as well as for hours not worked (like stw ²On average, across the oecd countries, the sectors most directly affected by the covid-19containmentmeasuresaccountforaround40oftotalemployment.These sectorsemployalargeproportionof‘nonstandardworkers,’i.e.part-timeworkers,self- employed and workers hired under fixed-term contracts. This proportion is generally highestinentertainmentindustries,hotelsandrestaurants(oecd 2020c). volume 17 | 2024 | number 1 [122] AntonRop table1 TypesofjrSchemesUsedduringthecovid-19PandemicintheEuro Area Stw,Leastupdated,pre-existing stw scheme:Increased accessandcoveragewithmoregenerousbenefits Austria,Belgium,Luxem- bourg,SlovakRepublic Stw,Updated,pre-existing stw scheme:increasedaccess andcoverageandaccessforworkersinnonstandardjobs Italy,Portugal Stw,Mostextensivelyupdated,pre-existing stw scheme: increasedaccess,coverage,benefitgenerosityandaccess forworkersinnonstandardjobs Germany,Spain,Finland, France Stwn,New(notpreviouslyexisting) stw scheme Greece,Lithuania,Latvia, Slovenia,Cyprus Ws,Newwagesubsidyscheme Estonia,Ireland, Netherlands,Malta notes Basedondatafrom oecd (2020a). schemes), e.g., Australia, Canada, Estonia, Ireland, New Zealand. ws arereservedforfirms experiencingasignificantdeclineinrevenue.In somecountries,thesizeoftheactualsubsidyonlydependsonthewage bill (before programme participation) and not the decline in business activity(oecd 2021). Table 1 presents different types of jr schemes used during the covid-19crisisintheeuroarea.Itrevealsimportantcross-country differences in the jr schemes used: 10 countries that adjusted their pre-existing stw schemes;5countrieswithnew stw schemes,and4 countrieswithwagesubsidyschemes. Data on the total portfolio of economic support measures were collectedfromtheOxfordcovid-19GovernmentResponseTracker (GitHub2022).TheOxford covid-19GovernmentResponseTrack- er (oxcgrt)provides a systematic setofcross-national, longitudinal measures of government responses for more than 180 countries since 1stJanuary2020(Haleetal.2021).Atpresent,itincludes19policyindi- catorscoveringclosureandcontainment,healthandeconomicpolicies. Tomake iteasierto describe government responses inaggregate, ox- cgrt calculates simple indices that combine individual indicators to provide an overall measure of the intensity of government response acrossafamilyofindicators.Theseindicesare:(1) gri (allcategories); (2) stringency index (containment and closure policies sometimes re- ferredtoaslockdownpolicies);(3) chi (containmentandclosureand healthpolicies);and(4) esi (economicsupportmeasures).The esi in- dexiscomposedofeconomicpolicyresponseindicatorswhichinclude ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [123] income support, debt/contract relief for households, fiscal measures and giving international support indicators.³ We used the esi index as an aggregate measure of the economic support for the period q1 2019–q2 2021for19euroareacountries. Tobeabletodeterminehowmuchthemacroeconomiceffectsofthe portfolio of fiscal and monetary measures (used to mitigate the dam- agecausedbythelockdownmeasures)helpedpreserveemploymentin addition to the actual jr scheme take-up effects, we must control for macroeconomic effects and their sectoral dimension on the trajectory of jr take-ups.Lockdowneffectsandthecorrespondingemployment lossvaried considerably between sectors. Hence, we use employment dataforthea10sectionsofthebroadnacestructureofeucountries. Lockdown measures were quantified by using the corresponding stringencyindexwhichembracesallindicatorsoncontainmentandclo- sure policies (school closure, workplace closure, cancellation of public events, restrictions on the size of gatherings, halting of public trans- port, stay-at home requirement, limitations on internal movement, restrictionsoninternationaltravel),constructedandpublishedonthe Oxford covid-19GovernmentResponseTracker(Haleetal.2021). Specific npi indicators (restrictions on the size of gatherings and school closure) used while constructing the instruments are collected fromthesamesource. DescriptiveStatisticsofMainVariablesandOperativeHypotheses Table 2 presents descriptive statistics of the most important model variablesforeuroareacountries:employmentpreservationratio,nor- malised employment in persons, normalised employment in hours, government’seconomicsupportmeasures,andgovernment’scontain- mentmeasuresforeachquarterfrom q1 2020until q2 2021. The levels of employment in persons and employment in hours in euro area countries were at their lowest in the second quarter of 2020,andwhilethemaximumdropintheaveragelevelofemployment perpersonpercountrydidnotexceed7.6(Spain),unemploymentin hours dropped substantially more, notably on average by 12.7 com- paredtotheaveragefor2019,toreachamaximumdecreaseof27.2in the case of Greece. For the entire period under observation, the nor- malised level of employment in persons was higher than the level of ³ThewaycompositeindicesarecalculatedisdescribedinHaleetal.(2021). volume 17 | 2024 | number 1 [124] AntonRop table2 DescriptiveStatistics Item  q  q  q  q  q  q Mean Employmentpreservationratio . . . . . . Employmentinperson = . . . . . . Employmentinhours = . . . . . . (a) Economicsupportmeasures . . . . . . stwn_take-up . . . . . . ws_take-up . . . . . . stw_take-up . . . . . . stw_take-up . . . . . . stw_take-up . . . . . . (b) stringency . . . . . . gatherings . . . . . . school . . . . . . sd Employmentpreservationratio . . . . . . Employmentinperson = . . . . . . Employmentinhours = . . . . . . (a) Economicsupportmeasures . . . . . . stwn_take-up . . . . . . ws_take-up . . . . . . stw_take-up . . . . . . stw_take-up . . . . . . stw_take-up . . . . . . (b) stringency . . . . . . gatherings . . . . . . school . . . . . . Continuedonthenextpage employment in hours. This difference is most pronounced in the sec- ondquarterof2020andshrinksslowlyafterwards. The employment preservation ratio was at its highest during the peak of both epidemic waves. Still, there is quite a high cross-country heterogeneity in the preservation ratios, reflecting differences in the intensityofpolicyresponsestothepandemicandthesectoralcompo- sitionoftheeconomies(Andertonetal.2020). The jr supportmeasureswereattheirlowestin q1 2020,increased considerablyinq22020andremainedatelevatedlevelsfortheremain- ing quarters of the observation period. For all types of jr schemes, ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [125] table2 Continuedfromthepreviouspage Item  q  q  q  q  q  q min Employmentpreservationratio . . . . . . Employmentinperson = . . . . . . Employmentinhours = . . . . . . (a) Economicsupportmeasures . . . . . . stwn_take-up . . . . . . ws_take-up . . . . . . stw_take-up . . . . . . stw_take-up . . . . . . stw_take-up . . . . . . (b) stringency . . . . . . gatherings . . . . . . school . . . . . . max Employmentpreservationratio . . . . . . Employmentinperson = . . . . . . Employmentinhours = . . . . . . (a) Economicsupportmeasures .      stwn_take-up . . . . . . ws_take-up . . . . . . stw_take-up . . . . . . stw_take-up . . . . . . stw_take-up . . . . . . (b) stringency . . . . . . gatherings . . . . . . school . . . . . . notes Row headings are as follows: (a) government’s support measures, (b) gov- ernment’s containment measures. Based on data from Eurostat (https://ec.europa.eu /eurostat),oecd(2000a),andOxfordcovid-19GovernmentResponseTracker(GitHub 2022).Employmentvariablesarenormalisedonthebasis 2019=1;indexofeconomicsup- port variable, Stringency index, School closing and Gatherings are used as defined in the Oxfordcovid-19GovernmentResponseTracker(GitHub2022).Typesofjrschemesde- finedassuggestedinthe oecd (2020a)study. thehighestleveloftake-upswasreachedinthesecondquarterof2020 while afterwards they declined and stayed quite stable. In the second quarterof2020,thelevelofschemeuse(take-ups)reachedamaximum level of 31 (Italy) for scheme type stw2 (a pre-existing stw scheme withupdatedaccessandcoverageaswellasaccessforworkersholding nonstandard jobs) with a mean value of 25 of scheme take-ups. The volume 17 | 2024 | number 1 [126] AntonRop lowest level of take-ups was reached in all quarters in those countries withnew stw schemes(stwn),eventhoughtheyhadanaveragelevel (7.8) of take-up in q2 2020 with a corresponding minimum level of almost2.8(Latvia). The containment and closure policies index (stringency, measured from 1 to 100) is at its lowest in q1 2020 and its highest in q2 2021. Overall, the mean value increased after q3 2020 and remained high through the other periods. A similar pattern occurred with the indi- cator Restriction on the size of gatherings (ordinal values, 1–5), while the indicator School closure (ordinal values 1–5) was at its highest in q2 2020(2.37),butlaterrelaxed. Againstthisbackgroundandtheliteraturereviewed,wetestthefol- lowinghypotheses: 1 Changesintheemploymentpreservation ratioovertimecanbe to a larger degree explained by changes in jr scheme take-up rates. 2 Changes in the preservation indicator over time are also influ- enced by changes in other support measures (fiscal and mone- tary)thatgovernmentshaveimplementedduringthepandemic. 3 Among different jr schemes, the most effective at preserv- ing employment levels were take-ups of already existing stw schemesthathadbeenmostextensivelyupdated. model In normal times, Okun’s law⁴ suggests that employment in persons depends on employment in hours, the cyclical phase of economic ac- tivity (Burggraeve, de Walque, and Zimmer 2015), as well as sectoral andcountrycharacteristics(Crivelli,Furceri,andToujas-Bernaté2012). However, inthepandemic,policymakershavesupportedemployment preservation (our dependent variable) directly by using jr schemes andindirectlythroughmacroeconomicpolicysupporttotheeconomy. We start the description of the equation composing our model of employmentpreservationwiththeequationforemploymenttoappro- priately encompass the relationship between employment in persons and employment in hours (Burggraeve et al. 2015). Explanatory vari- ables of the model for employment in persons are therefore employ- ⁴Okun’sLawisanempiricallyobservedrelationshipbetweenunemploymentandlosses inacountry’sproduction(Prachowny1993). ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [127] mentinhours, jr schemetake-uprates,economicsupportmeasures, andtimefixedeffects.Toencompasslargedifferencesinthepotential effects of economic support between sectors, economic support vari- able effects are specified separately for five groups of sectors. Corre- spondingexplanatoryvariablesaredefinedasaproductofthesectoral groupindicatorvariableandeconomicsupportvariable.Timefixedef- fects areincludedtoaccountforanyothertime-specificeffects onthe employment in persons variable thatmighthave affected penetration ratesinthecountriesunderstudy. If ep ijt is employment in persons and eh ijt employment in hours, jr it ·dum_jr ik thetake-upofjrscheme k(forcountryitake-upjr it and schemedum_jr ik ),es it ·dum_es il sectoraleconomicsupport(forcountry ieconomicsupportes it andsectordum_es il ),dum t timeindicatorsem- bracingpotentialother(undisclosed)yetsystematicfactors’effectson persons employed dynamics, U i unobservable country effects, U j un- observable sector effects andε ijt the error term, then the conceptual versionofthemodelforemploymentmaybeformallywrittenas ep ijt = F(eh ijt ,jr it ·dum_jr ik ,es it ·dum_es il , dum t ,U i ,U j ,ε ijt ), (2) where index i stands for country, j for sector, t for time, k for type of schemeandlforsectoralgroup. Regarding the specification of the function F,itisassumedthat thereisalineardependenceofep ijt ontheelasticityofeh ijt andthein- crementsofothervariablesstatedsuchthatthecompletespecification of the estimable operative version of the model for ep ijt isthe follow- ing: ep ijt = eh α ijt ·exp( k β k jr it ·dum_jr ik + l γ l es it ·dum_es il + δ t dum t +U i +U j +ε ijt ). (3) Sectoral groups are defined by sectors of the a10 sections of the broadest nace sectoral classification. These groups of sectors are defined according to the potential extent of their lockdown expo- sure (manufacturing, construction, utilities, vulnerable services, non- vulnerable services, public sector). Types of jr schemes are specified accordingtotheclassificationusedin oecd (2020a). Weanalysetheperiod q1 2019–q2 2021. Theperiodisextendedto volume 17 | 2024 | number 1 [128] AntonRop thebeginningof2019toidentifytheeffectsofmodificationsmadeto jr schemes at the start of the epidemic (14 euro area countries mod- ified an already existing stw scheme in q1 2020, as well as potential other systematic (but undisclosed) time-specific impacts on employ- mentduringtheepidemicepisode(parametersσ t )aswellastoincrease the accuracy of the estimated dependence of the employment preser- vationindicator. Sinceforestimatedrelation(3)parameterαdidnotsignificantlydif- ferfrom1,⁵theestimablestartingoperativeversionofthemodelspec- ificationforemploymentpreservationisdefinedasfollows: log ep ijt eh ijt = k β k jr it ·dum_jr ik + l γ l es it ·dum_es il + t σ t dum t +U i +U j +ε ijt ,( 4 ) where index i stands for country, j for sector, t for time, k for type of schemeandlforsectoralgroup. Thisstartingversionofthemodelisestimatedandanalysedinthree steps;ineachstep,specificationofthepreviousstepisfurthersimpli- fiedtoallowspecificcharacteristicsofthemodeltobeanalysed. In step one, the starting version of the model specification (4) is usedtocheckthepotentialexistenceofspecifictimeeffectsinfluencing employmentpreservationintheepidemicepisode. Insteptwo,themodelisestimatedinitsbasicspecificationas log ep ijt eh ijt = k β k jr it ·dum_jr ik + l γ l es it ·dum_es il +U i +U j +ε ijt .( 5 ) Notably, the basic specification differs from the starting specifica- tiononlyin(missing)timedummies.Sinceitencompassesboththeo- reticallyimportantfactors–thetake-upofdifferent jr schemeeffects aswellasthesectoralmacroeconomiceffects,adiscussionofthebasic modelestimatesrepresentsthecoreoftheanalysisinthispaper. Arobustness checkofthemainbasicmodelconclusionsismadein step three when the model is estimated without any explicit specifi- cation of the sectoral differences, therefore formally in the following specification: ⁵Correspondingestimatesareavailablefromtheauthoruponrequest. ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [129] table3 HausmanTest Model χ 2 P Startingmodel Asymptoticassumptionsviolated Basicmodel . . Robustmodel . . notes Hausmantestvaluesandsignificance;thedataforthestartingmodelviolatethe asymptoticassumptionsofthetest. log ep ijt eh ijt = k β k jr it ·dum_jr ik +γ l es it +U i +U j +ε ijt .( 6 ) Thepresentedstepwisesimplificationofthemodelspecificationen- ables the explicit focusing on (testing of) the crucial questions (hy- potheses)ofthestudyembracedinthestatedpaperresearchquestion. results Themodel(4)isestimatedonpaneldata(wheretheobservationunitis country,sectorinaquarter)for19eurocountriesand9sectorsinthe period q1 2019–q2 2021.Duetomissingdata,thereare1,330complete observations. Becauseofthehighpossibilitythatunobservableindividualeffects for country and sectors are present,⁶ a fixed effects estimator should beusedasitexcludescountryandsectortime-invariantvariables’im- pacts and gives consistent parameter estimates. Nonetheless, Haus- man’s test is conducted to test for the presence of fixed effects and whetherthemoreefficientrandomeffectsestimatorcouldalsobeused. Table3presentsvaluesofHausman’stestforallthreemodelspecifica- tions analysed (starting specification, basic specification, robust spec- ification).Hausman’s test does notenablethe use ofa randomeffects estimatorinanymodelvariantandthusallthreemodelsareestimated withfixedeffects. The possible endogeneity of economic support measures as well as the jr schemetake-upratesleadsustorunaninstrumentalversionof thefixedeffectsregression(theinstrumentalestimatorgmmisused).⁷ ⁶Crivelli et al. (2012) suggest a set of determinants of cross-country variations of employment growth consisting of the following variables: (a) Structural and Policy Variables (labour market policies, product market policies, and government size), (b) Productmarketregulations,likelabourmarketregulations,(c)governmentsize,(d) macroeconomicvariables,and(e)demographicvariables. ⁷Economicsupportmeasuresandtheassociatedjrschemetake-upsarehighlyendoge- volume 17 | 2024 | number 1 [130] AntonRop table4 StartingModelEstimates Explanatoryvariables Coefficient t-stat P support(–) . . . support_con(–) –. –. . support_vul(–) .*** . . support_nvul(–) –. –. . support_uti(–) –. –. . support_pub(–) –. –. . takeup_ws .*** , . takeup_stw .*** . . takeup_stw .*** . . takeup_stw .*** . . takeup_stwn .*** . . dum_stwo .*** . . dum_q . . . dum_q . . . dum_q . . . dum_q –. –. . dum_q –. –. . dum_q –. –. . dum_q –. –. . dum_q –. –. . dum_q –. –. . _cons –.*** –. . Continuedonthenextpage The instruments used are a stringency index, lockdown variables forpublicgatheringsandschoolclosures,dummiesforsectorsandthe typeof jr schemesaswellasthecombination(products)ofthesevari- ables. We used instrument variables representing pandemic contain- ment measures as they are defined in relation to the state of the pan- nous to labour market conditions since they were mainly used to alleviate the short- term effects of the covid policy measures constraining social mobility on employ- ment and temporary unemployment (Bole, Prašnikar, and Rop 2021). For instance, firms tend to place workers in jr schemes when the underlying conditions are poor and,correspondingly,reducetheshareoftheworkforcein jr schemeswhenbusiness conditions improve. Such pro-cyclical behaviour strongly biases the estimate of our variableofinterestbecausetheunobservablebusinessconditionswouldbepartofthe residualandnegativelycorrelatedwiththe jr schemetake-upvariable(forGermany, seeAiyarandDao2021). ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [131] table4 Continuedfromthepreviouspage Explanatoryvariables P Andersoncanoncorrelationtestofunderidentification . SarganHansentestofoveridentification . notes Calculations based on Eurostat (https://ec.europa.eu/eurostat) and Oxford covid-19GovernmentResponseTracker(GitHub2022).Thedependentvariableisem- ploymentinpersonsperemploymentinhours,normalisedsothattheaveragein2019is 1. Explanatory variables are: support_con(–1)– economic supportpolicy index multiplied byadummyforconstruction,lag1;support_vul(–1)– economicsupportpolicyindexmul- tipliedbyadummyforvulnerable,lag1;support_nvul(–1)–economicsupportpolicyindex multipliedbyadummyfornon-vulnerable,lag1;support_uti(–1)–economicsupportpol- icy index multiplied by a dummy for utilities, lag1; support_pub (–1) – economic support policyindexmultipliedbyadummyforpublicsector,lag1;takeup_ws–take-upsmultiplied by a dummy for a ws scheme; takeup_stw3 – take-ups multiplied by a dummy for an up- dated stw scheme (increased access and coverage; increased generosity; increased access for workers in non-standard jobs); takeup_stw2 – take-ups multiplied by a dummy for an updated stw scheme(increasedaccessandcoverage;increasedaccessfor workersinnon- standardjobs);takeup_stw1–take-upsmultipliedbyadummyforanupdatedstw scheme (increased access and coverage; increased generosity); takeup_stwn – take-ups multiplied byadummyforanew stw scheme;dum_stwo–adummyforapre-covid stw scheme. Sargan-Hansenoveridentificationtest(significance);Andersontestofunderidentification (significance).***,**,*significant,respectivelyat0.01,0.05and0.1. demicandnotthestateofthelabourmarketandtheeconomy,butmay well impactthe levelofthegovernment’s economicsupport measures and jr schemes. Table4presentsestimatesofthemodelinitsstartingspecification (5)foreuroareacountries.TheSargan-HansenandAndersontestscon- firmthattheinstruments’ qualityisacceptable.Thecoefficientsof jr schemetake-uptypesaresignificantandhavetheexpectedsign,while among the sectoral economic support variables only the support for vulnerablesectorsissignificant.Othershavetheexpectedsignbutare not significantly different from the corresponding effect of manufac- turing, which represents the basis of the sectoral economic support variablescomparisonandwhichinitselfisnon-significant.Giventhat alltimedummiesinthecovid-19periodarenon-significantandare not significantly different from the time dummies in the pre-covid- 19period,⁸therearenootherdecisivefactorsmitigatingemployment lossesduringthelockdownepisode. Table 5 displays estimates of the model in its basic specification. It ⁸Correspondingestimatesareavailablefromtheauthoruponrequest. volume 17 | 2024 | number 1 [132] AntonRop table5 BasicModelEstimates Explanatoryvariables Coefficient t-stat P support(–) . . . support_con(–) . . . support_vul(–) . .*** . support_nvul(–) . –. . support_uti(–) –. –. . support_pub(–) –. –. . takeup_ws . .*** . takeup_stw . .*** . takeup_stw . .*** . takeup_stw . .*** . takeup_stwn . .*** . dum_stwo . .*** . Cons –. –.*** . Andersoncanoncorrelationtestofunderidentification . SarganHansentestofoveridentification . notes N = 5130. The dependent variable is employment in persons per employment in hours,normalisedsothattheaveragein2019is1.Explanatoryvariablesare:support(–1) – economic support policy index, lag1; support_con(–1) – economic support policy index multiplied by a dummy for construction,lag1; support_vul(–1)– economic supportpolicy index multiplied by a dummy for vulnerable, lag1; support_nvul (–1) – economic support policy index multipliedby a dummyfor non-vulnerable, lag1; support_uti(–1) – economic supportpolicyindexmultipliedbyadummyforutilities,lag1;support_pub(–1)–economic support policy index multiplied by a dummy for public sector, lag1; takeup_ws – take-ups multiplied by a dummy for a ws scheme; takeup_stw3 – take-ups multiplied by a dummy foranupdatedstw scheme(increasedaccessandcoverage;increasedgenerosity;increased accessforworkersinnon-standardjobs);takeup_stw2–take-upsmultipliedbyadummyfor anupdatedschemestw(increasedaccessandcoverage;increasedaccessforworkersinnon- standardjobs);takeup_stw1–take-upsmultipliedbyadummyforanupdatedstw scheme (increased access and coverage; increased generosity); takeup_stwn – take-ups multiplied by a dummy for a new stw scheme; dum_stwo a dummy for a pre covid stw scheme. Sargan-Hansenoveridentificationtest(significance);Andersontestofunderidentification (significance);***,**,*significant,respectivelyat0.01,0.05and0.1. differs from the model’s starting specification in the absence of time dummies. The model is estimated with instrumentalised fixed effects. In estimating the basic model, the same instruments are used as for estimatingthestartingmodel. The coefficients of the jr scheme take-up rates are highly signif- icant and have the expected sign. A positive impact was especially strong in pre-existing stw schemes with increased access and cov- ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [133] table6 EffectsofEmploymentPreservationPolicy Period Actualvalues Allmodelmeasures Macroec.measures q  . . . q  . . . q  . . . q  . . . q  . . . q  . . . Average . . . notes Employment preservation; basic model simulation of employment preservation effects; actual values;totalsimulatedeffects; simulationof theeffects of only themacroe- conomicsupportmeasures. erage, increased generosity, and increased access for workers holding non-standardjobs. Theeffectsofthemacroeconomiceconomicsupportbywayofmit- igated sectoral employment loss are positive for manufacturing, con- struction,vulnerableandnon-vulnerableservices,butonlysignificant forthe group ofvulnerable sectors. To reveal the structure of policy contributions to the retained em- ployment levels, table 6 presents actual values of employment in per- son per working hours, model-simulated common effects of the jr schemesandthemacroeconomicsupportmeasures,aswellasthecon- tributionofonlythemacroeconomicsupportmeasures.Averagevalues forallsectorsandcountriesaregiven. Employment preservation effects were quite volatile in the first threewavesoftheepidemic.Theyreachedtheirpeakin q2 2020when slightlymorethan10oftheemployedwerenotworking.Thebiggest contributiontosuchemploymentpreservationeffectswasmadebythe takingupofjrschemes.Still,thecontributionmadebythemacroeco- nomicmeasurestocurbingemploymentlosseswasalsonotnegligible. Itwassmallonlyinthefirsttwoquartersofthecovid-19pandemic. After that, thecontribution was quite sizeable; in the wholeperiod of the first three epidemic waves, it accounted for around one-quarter (0.011/0.048=0.23)ofthetotalmitigationeffects. Overall, it may be concluded that in line with hypotheses 1 and 2 changes in the employment preservation ratio over time can be to a largerdegreeexplainedbychangesin jr schemetake-upratesandby changesinothergovernmentsupportmeasures. volume 17 | 2024 | number 1 [134] AntonRop table7 MacroeconomicSupportEffectsonEmploymentPreservation–Sectoral Differences Sector support_con support_vul support_nvul support_uti support_pub support_man . .*** –. –. –. support_con .*** –. –. –. support_vul –.*** –.*** –.*** support_nvul –.* –. support_uti . notes Differencesinmacroeconomicsupporteffectsonemploymentpreservation;sec- toral differences (column sector less row sector item) are multiplied by 1,000; macroeco- nomicsupporteffectsonemploymentpreservationin:manufacturing(support_man),con- struction (support_con), vulnerable service sectors (support_vul), nonvulnerable service sectors(support_nvul),utilities(support_uti),andpublicsector(support_pub);***,**,*sig- nificantat0.01,0.05,and0.1,respectively. The estimated effects of macroeconomic support on sectoral em- ployment preservation presented in table 5 also enable a comparison ofthoseeffectsbetweensectors.Correspondingdifferencesinsectoral employment preservation effects are given in table 7. Macroeconomic supportisshowntohavemitigatedemploymentlossbyfarthemostin thegroupofvulnerableservicesectors.Theemployment-preservation impact of economic support measures for the vulnerable service sec- tors is several times (significant at p = 0.00) larger than in the other sectors.However,asshownintable7,theeffectsofthosesectors(with oneexception)didnotsignificantlyexceedtheeffectsinanyothersec- tor; only the effects in non-vulnerable service sectors differ from the effectsinutilitiesbythelowestmarginofsignificance(p=0.10). We may conclude that the evidence presented in table 8 confirms thattheimpactofeconomicsupportmeasuresonemploymentpreser- vationvariesacrosssectors,especiallybetweenvulnerablesectorsand others. Estimatesofthebasicmodel(intable5)alsoindicatethattheeffec- tiveness ofdifferent types of jr schemetake-ups varies considerably. To enable a more detailed comparison, table 8 presents differences in the effects on employment preservation of the analysed types of jr schemetake-ups. The evidence presented in table 8 shows that the most successful countrieshavebeenthosewithaprevious stw schemewhichtheyex- tended most extensively by applying all three key changes to it (the mostextensivelyupdated stw schemes–denotedbytakeup_stw3): • simplifyingaccessandextendingcoverage; ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [135] table8 EmploymentPreservationEffects–Differencesbetween jr Schemes Take-Ups takeup_stw takeup_stw takeup_stw takeup_stwn dum_stwo takeup_ws .*** .*** .*** . –.*** takeup_stw –.*** –.*** –.*** –.*** takeup_stw .*** –.** –.*** takeup_stw –.*** –.*** takeup_stwn –. notes Differencesinemploymentpreservationeffectsfortypesof jr schemetake-ups; effect of jr scheme take-ups in column less the effect of a jr scheme in row item; anal- ysed types of jr schemes: takeup_ws – take-ups multiplied by a dummy for ws scheme; takeup_stw3 – take-ups multiplied by a dummy for an updated existing stw scheme (in- creased access and coverage; increased generosity; increased access for workers in non- standardjobs);takeup_stw2–take-upsmultipliedbyadummyforanupdatedexistingstw scheme(increasedaccessandcoverage;increasedaccessforworkersinnon-standardjobs); takeup_stw1 – take-ups multiplied by a dummy for an updated existing stw scheme (in- creasedaccessandcoverage;increasedgenerosity);takeup_stwn–take-upsmultipliedbya dummyforanew stw scheme;dum_stwoadummyforapre-covid stw scheme;***,**, *significant,respectivelyat0.01,0.05and0.1. • extendingcoveragetonon-permanentworkers;and • makingtheirbenefitsmoregenerous. The take-ups of this jr scheme had the strongest impact on em- ployment preservation, outperforming other types of jr schemes in thecovid-19pandemicby25–70.Thelowesteffectonemployment preservation was seen for take-ups of the jr scheme ws, which sup- ported employment bysubsidising all employees in the firm (denoted by takeup_ws). For every 1 of take-ups, this type of jr scheme re- duced employment losses almost four times less effectively than the mostsuccessful stw scheme. Onemayconcludethattheempiricalevidencegivenintable9val- idates the last (3rd) research hypothesis, namely, that among the dif- ferentjrschemesinthecovid-19pandemicthemosteffectiveat preserving employment levels were take-ups of already existing stw schemesthathadbeenmostextensivelyupdated. robustness test Tochecktherobustnessoftheestimatedbasicmodelwithregardtoits estimates, consistencyand lessons, themodel is also estimated in the robustspecification(6).Themodelisestimatedinasimplifiedversion withoutanexplicitsectoraldimension.Again,weemployedaninstru- mentalversion ofthefixedeffects regression (instrumental estimator volume 17 | 2024 | number 1 [136] AntonRop table9RobustModelEstimates Explanatoryvariables Coefficient t-stat P support(–) .*** . . takeup_ws .*** . . takeup_stw .*** . . takeup_stw .*** . . takeup_stw .*** . . takeup_stwn .*** . . dum_stwo .*** . . Cons –.*** –. . Andersoncanoncorrelationtestofunderidentification . Sargan-Hansentestofoveridentification . notes Thedependentvariableisemploymentinpersonsperemploymentinhours,nor- malisedsothattheaveragein2019is1.Explanatoryvariablesare:support(–1)–economic supportpolicyindex,lag1;takeup_ws–take-upsmultipliedbyadummyfora ws scheme; takeup_stw3–take-upsmultipliedbyadummyforanupdated stw scheme(increasedac- cessandcoverage;increasedgenerosity;increasedaccessforworkersinnon-standardjobs); takeup_stw2–take-upsmultipliedbyadummyforanupdated stw scheme(increasedac- cessandcoverage;increasedaccessforworkersinnon-standardjobs);takeup_stw1–take- upsmultipliedbyadummyforanupdated stw scheme(increasedaccessandcoverage;in- creasedgenerosity);takeup_stwn–take-upsmultipliedbyadummyforanewstwscheme; dum_stwoadummyforapre-covid stw scheme.Sargan-Hansenoveridentificationtest (significance); Anderson test of under identification (significance); ***, **, * significant, re- spectivelyat0.01,0.05and0.1. gmm).Instrumentsusedinestimatingtherobustmodelareagaincon- structed according to the same principles as for the previous model variants.Theestimatedmodelispresentedintable9. Themacroeconomiceffectsonemploymentpreservationarehighly significantandlargerthanthesimpleaverageofthecorrespondingsec- toraleffectsinthebasicmodel.Namely,thesimpleaverageofthesec- toraleffectsinthebasicmodel(seetable5)was0.00016( p=0.06)ver- sus0.00019( p=0.00)intherobustversionofthemodel. This fact further confirms the sectoral differences in the economic supportmeasureseffectsand,inparticular,thesizeandimportanceof the vulnerable service sectors support for the success in limiting the lossofjobsduringthecovid-19pandemic,asalreadyseenintable7. conclusion The covid-19pandemichashadanunprecedentedimpactonthe labour market across world economies. The key instruments for mit- ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [137] igating the effects of the lockdowns on employment and social hard- shipthatdifferentcountriesintroducedorextendedinresponsetothe crisis include different job retention (jr) schemes. jr schemes were implementedinalleuroareacountries,althoughdifferentcountriesin- troduced orextendedarangeof jr schemes.Usingdataforeuroarea countries, this paper has analysed the effects of various jr scheme take-ups on employment preservation during the covid pandemic. Toassesstheimpactsofdifferent jr schemes,weusedthe oecd clas- sification whereby the countries use five types of jr schemes (oecd 2020a). Ourpapersupportsliteraturefindings(HijzenandVenn2011;Boeri andBruecker2011;HijzenandMartin2013;oecd2020a;2020b;2020c; 2021) that jr schemes have been the most important instrument for reducingthelossofemploymentfollowingtheimpactsofthenonphar- maceuticalinterventionsduringthe covid-19crisis.Suchschemes were able to relatively successfully limit excessive layoffs in the situ- ation of a temporary reduction in business activity. Our results also showthatcountries(France,Germany,Spain,Finland)whichextended aprevious stw schemebyincreasingitsaccess,coverageandgeneros- ity, and also integrated workers holding non-standard jobs (denoted by takeup_stw3) into the scheme had the most successful jr scheme take-ups. Our study reveals that jr schemes contributed less than the over- allemploymentpreservationachievedduringdifferentepidemicwaves, and that other macroeconomic measures (non-jr schemes) contrib- uted around one-quarter to the employment preserved. Correspond- ingdifferencesinsectoralemployment preservation effects showthat macroeconomic support mitigated the loss of jobs by far the most in thegroupofvulnerableservicessectors,wherethecorrespondingnon- pharmaceuticalintervention (npi)losseswerethehighest,andwhich was the crucial driver of the high indirectnet effects in other sectoral groups(Bole,Prašnikar,andRop2021).Bettertargetingbyusing stw fornon-standardjobs(i.e.self-employedworkers andthoseintempo- rary or part-time dependent employment) and providing more gen- erous benefits have no doubt helped to improve the situation (oecd 2020b). Still, since employers have little or zero incentive to use stw fornon-standardjobswhentheyknowthattheseworkerscanbefired at little or no cost, and governments are reluctant to subsidise these jobsduetothemoralhazardproblem(BoeriandBruecker2011),other volume 17 | 2024 | number 1 [138] AntonRop macroeconomic measures (non-jr schemes) mightalso do a good job atpreservingjobs,inparticular,inthesesectorsoftheeconomy. references Adams-Prassl,A.,T.Boneva,M.Golina,andC.Rauh.2020.‘Inequality in the Impact of the Coronavirus Shock: Evidence from Real Time Surveys.’JournalofPublicEconomics189:104245. Aiyar,S.,andDao,M.(2021).TheEffectivenessofJob-RetentionSche- mes:covid-19EvidencefromtheGermanStates(imfWorkingPa- perNo.2021/241).Washington, dc. Anderton,R.,V.Botelho,A.Consolo,A.D.DaSilva,C.Foroni,M.Mohr, andL.Vivian.2020.‘TheImpactofthecovid-19Pandemiconthe EuroAreaLabourMarket.’ ecb EconomicBulletin (8). Autor,D.,D.L.Cho,M.Crane,B.Goldar,J.Lutz,W.B.Montes,D.Rat- ner,W.B.Peterman,D.Villar,andA.Yildirmaz.2020.‘AnEvaluation of the Paycheck Protection Program Using Administrative Payroll Microdata.’ nber Working Paper Series w29972, National Bureau ofEconomicResearch,Cambridge, ma. Bellmann, L., H. D. Gerner, and R. Upward. 2015. ‘The Response of GermanEstablishmenttothe2008–2009EconomicCrisis.’In Com- plexity and Geographical Economics, edited by P. Commendatore, S. Kayam,andI.Kubin,165–207.Cham:Springer. Bole,V.,J.Prašnikar,andA.Rop.2021.‘SupportforThoseNotAffected: HowMacroeconomicPoliciesHaveShaped covid’sImpactonSec- toral Activity.’ Paper presented at the Economic and Business Re- viewConferenceand seb lu DoctoralConference,Ljubljana,3De- cember. Bishop,J.,andI.Day.2020.‘HowManyJobsDidJobKeeperKeep?’Re- searchDiscussionPaper rdp 2020-079, Reserve Bank ofAustralia, Sydney. Boeri, T., and H. Bruecker. 2011. ‘Short-Time Work Benefits Revisited: SomeLessonsfromtheGreatRecession.’EconomicPolicy26:697–765. Burggraeve, K., G. H. de Walque, and H. Zimmer. 2015. ‘The Relation- shipbetweenEconomicGrowthandEmployment.’EconomicReview 1:32–52. Calavrezo,O.,R.Duhautois,andE.Walkoviak.2009.‘Short-TimeCom- pensation and Establishment Survival: An Empirical Analysis with FrenchData.’PaperpresentedatConferenceonComparative Anal- ysisofEnterpriseData,Tokyo,2–4October. Cotofan,M.,J.E.DeNeve,M.Golin,M.Kaats,andG.Ward.2021.‘Work and Well-Being during covid-19: Impact, Inequalities, Resilience, and the Future of Work.’ In WorldHappinessReport 2021,edited by ijems TheEffectsofJobRetentionSchemesonEmploymentPreservation [139] J.F.Helliwell,R.Layard,J.D.Sachs,J.-E.DeNeve,L.B.Aknin,and S.Wang,153–90.UniversityofOxford:WellbeingResearchCentre. CrivelliC.,D.Furceri,andJ.Toujas-Bernaté.2012.‘CanPoliciesAffect EmploymentIntensityofGrowth? ACross-Country Analysis.’ imf WorkingPaper wp/12/218,InternationalMonetaryFund,Washing- ton, dc. Gangopadhyaya, A., and B. Garrett. 2020. ‘Unemployment, Health In- surance,andthecovid-19Recession.’http://dx.doi.org/10.2139 /ssrn.3568489 GitHub. 2022. ‘Methodology for Calculating Indices.’ https://github .com/OxCGRT/covid-policy-tracker/blob/master/documentation /index_methodology.md Hale,T.,N.Angrist,R.Goldszmidt,B.Kira,A.Petherick,T.E.Phillips,S. Webster,E.Cameron-Blake,L.Hallas,S.Majumdar,andH.Tatlow. 2021.‘AGlobalPanelDatabaseofPandemicPolicies(Oxfordcovid- 19GovernmentResponseTracker).’ NatureHumanBehaviour5:529– 38. HijzenA.,andS.Martin.2013.‘TheRoleofShort-TimeWorkSchemes during the Global Financial Crisis and Early Recovery: A Cross- Country Analysis.’ iza Discussion Paper 7291, Institute of Labor Economics,Bonn. Hijzen, A., and D. Venn. 2011. ‘The Role of Short-Time Work Schemes duringthe2008-09Recession.’oecdSocial,EmploymentandMi- grationWorkingPapers115, oecd,Paris. Hubbard, R. G., and M. R. Strain. 2020. ‘Has the Paycheck Protection ProgramSucceeded?’nber WorkingPaperSeriesw28032,National BureauofEconomicResearch,Cambridge, ma. International Labour Organization. 2020.ilo Monitor: covid-19and theWorldofWork.Geneva:InternationalLabourOrganization. ———. 2021.ilo Monitor: covid-19 and the World of Work. Geneva: InternationalLabourOrganization. oecd.2020a. JobRetentionSchemesduringthe covid-19Lockdownand Beyond.Paris: oecd. ———.2020b.TheImpactofthe covid-19PandemiconJobsandIncomes in g20Economies. Paris: oecd. ———. 2020c.DistributionalRisksAssociatedwith Non-Standard Work: StylizedFactsandPolicyConsiderations.Paris: oecd. ———. 2021.oecdEmploymentOutlook2021:Navigatingthe covid-19 CrisisandRecovery.Paris: oecd. Prachowny,M.F.J.1993.‘Okun’sLaw:TheoreticalFoundationsandRe- visedEstimates.’TheReviewofEconomicsandStatistics75(2):331–6. Siegenthaler,M.,andD.Kopp.2019.‘Short-TimeWorkandUnemploy- volume 17 | 2024 | number 1 [140] AntonRop mentinandaftertheGreatRecession.’kofWorkingPapers19-462, kof SwissEconomicInstitute,Zürich. Zimpelmann, C., H. M. von Gaudecker, R. Holler, L. Janys, and B. Si- flinge.2021.‘DriversofWorkingHoursandHouseholdIncomeDy- namicsduringthe covid-19Pandemic:TheCaseoftheNether- lands.’ econtribute DiscussionPaper093,UniversityofBonnand UniversityofCologne. ijems