Parental Employment at the Onset of the Pandemic: Effects of Lockdowns and Government Policies

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) Parental Employment at the Onset of the Pandemic: Effects of Lockdowns and Government Policies Kabir Dasgupta, Linda Kirkpatrick, and Alexander Plum 2024-012 Please cite this paper as: Dasgupta, Kabir, Linda Kirkpatrick, and Alexander Plum (2024). “Parental Employment at the Onset of the Pandemic: Effects of Lockdowns and Government Policies,” Finance and Economics Discussion Series 2024-012. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2024.012. 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.

Parental Employment at the Onset of the Pandemic: Effects of Lockdowns and Government Policies Kabir Dasgupta*, Linda Kirkpatrick†, Alexander Plum‡ March 12, 2024 Abstract The COVID-19 pandemic had disproportionate impacts on women’s employment, especially for mothers with school-age and younger children. However,theimpactslikelyvarieddependingonthetypeofpolicyresponse adopted by various governments. New Zealand presents a unique policy setting in which one of the strictest lockdown restrictions was combined withagenerouswagesubsidyschemetosecureemployment. Weutilizetax records to compare employment patterns of parents from the pandemic period(treatmentgroup)tosimilarparentsfromarecentpre-pandemicperiod (controlgroup). Formotherswhoseyoungestchildisagedbetweenoneand 12, we find a 1-2-percentage point decline in the likelihood of being employed in the first six months of the pandemic; for fathers, we hardly see any significant changes in employment. Additionally, the decline in mothers’employmentratesismainlydrivenbythosenotemployedinthemonth before the lockdown. We also find similar employment patterns for future parentswhohadnochildrenduringtheevaluationperiod. Thisindicatesthat theadverselabourmarketimpactsarenotuniquelyexperiencedbymothers, butbywomeningeneral. JELCode: D10;D13;E24 Keywords: Pandemic;Employment;Parentalgap;Administrativedata. *Senior Economist, Consumer and Community Affairs, Federal Reserve Board, Washington, DC,UnitedStates. Correspondingauthor: kabir.dasgupta@frb.gov. FederalReserveBoard,1875 IStNW,Washington,DC20006,UnitedStatesofAmerica. †SeniorResearchOfficer,NewZealandWorkResearchInstitute,AucklandUniversityofTechnology,Auckland,NewZealand ‡SeniorResearchFellow,NewZealandWorkResearchInstitute,AucklandUniversityofTechnology,Auckland,NewZealand Theresultsandopinionsexpressedinthispaperreflecttheviewsoftheauthorsandshouldnotbe attributedtotheFederalReserveBoard,FederalReserveSystemorStatisticsNewZealand.

1 Introduction TheCOVID-19pandemictriggeredsharpdeclinesineconomicactivitiesandemployment all over the world (e.g., Chetty et al., 2020; Wynne and Balke, 1993). In contrast to previous recessions, the pandemic initially had a larger economic impact on women compared to men (e.g., Albanesi and Kim, 2021; Kugler et al., 2023; Bluedorn et al., 2023). We contribute to the existing literature that documents gender gap in the effects of the pandemic on labour market outcomes by focusing on different-sex parents from New Zealand, which represents a unique policy-settingwhencomparedtoothermorewidelystudiedcountries. The disproportionate labour market outcomes experienced by women, especially by mothers with school-age or younger children, can be attributed to several economic reasons. First, high-contact service industries (such as hospitality or tourism) typically represented by a higher share of female workers, saw larger employmentdeclinesduetotheeconomicshutdowns(Alonetal.,2022). Second, the pandemic prompted school and daycare closures, thereby resulting in an unequaldistributionofhouseholdactivitiesandchildcarebetweenmenandwomen. Sincemotherstypicallyspendmoretimeincaringforchildrenthanfathersdo,the increaseinhome-basedchildcareneedsconstrainedwomen’sabilitytoworkmore thanmen’s(Alonetal.,2022,p.84).1 Third,temporaryandpart-timeemployment is much more prevalent among women than among men. Non-standard employment contracts are usually at a much greater risk of being terminated during an economicdownturn(Petrongolo,2004). Investigatingtheeconomicimpactofthe pandemic across six different countries, Alon et al. (2022, p. 86) conclude that the “recession is a she-cession, that is, declines in employment and hours worked are larger among women.”2 However, it is also worth noting that the “COVID- 19 crisis she-cessions were short-lived” (Bluedorn et al., 2023) and were mainly 1For example, Goldin (2022) calculated for the US that childcare time for college-graduated women(includingtimeforeducation)grewfrom8.7hoursperweekjustshortlybeforethepandemicto17.3hoursattheonsetofthepandemic. 2The authors use micro survey data from the following six countries: the United States, Canada,Germany,theNetherlands,Spain,andtheUnitedKingdom 1

observedduringthebeginningofthelockdownperiod. The policy response from the New Zealand government presents a particularly interesting case to study the impact of the pandemic on parental employment. The government implemented one of the strictest lockdowns compared to other countries in the Western world. Apart from a few exceptions like essential workers, everyone had to be isolated within their own “household bubble”, while in-person interactions with people from outside a residential unit were highly restricted. In addition, all non-essential firm sites had to be closed and switched to remote working. At the same time, the government rolled out high-trust schemes including a generous wage subsidy program extended to firm owners to prevent job losses and business closures. As a plausible result of the New Zealand government’s policy measures, the effect of the pandemic on the country’s labour market has been much less severe relative to other countries which focused more onpandemic-relatedreliefpackagesandcashtransferprogramsdirectlyprovided to the consumers. This is highlighted in Figure 1, which shows that the drop in NewZealand’semploymentrateaftertheonsetofthepandemicwassmallercomparedtosomeothermajoreconomiesfromEurope,NorthAmerica,andAustralia. However, despite the mild labour market implications in New Zealand, female workersweredisproportionatelyaffectedduringthepandemic(Kidoetal.,2021). Given New Zealand’s distinct policy setting, we contribute to the pandemic literaturebyexploringemploymentgapsbetweenmenandwomenexperiencedin the island nation during the pandemic. Since changes in the overall employment duringtheinitialpandemicperiodwerelesslikelytobedrivenbybusiness-related economic challenges owing to the support of the generous wage subsidy scheme, ouranalysisprovidespolicy-relevantinsightsintotheunderlyingmechanismsbehind women’s labour market experiences. We particularly focus on parents to explore the pandemic’s effect on the gender gap in employment outcomes. Our studyisfurthermotivatedbytheuseofdetailedadministrativedatathatallowsus to examine the impacts more objectively relative to the other studies in relevant internationalliterature. 2

Figure1: Employmentrateacrosscountries-2015-2023 Note:DataaccessedfromOrganisationforEconomicCo-operationandDevelopment(OECD). OECD(2024),Employmentrate(indicator).doi:10.1787/1de68a9b-en(Accessedon20January2024) Several studies have shown that the impact of the pandemic was most noticeable within the first six months for most countries (e.g., Goldin, 2022; Bluedorn et al., 2023). As such, our study looks at parents’ employment during the initial monthsofthepost-lockdownperiodwhenpeoplewerefacedwiththedecisionof balancing their time between employment and unpaid work, including childcare andfamilyresponsibilities(Chengetal.,2021;Alonetal.,2020a,b). Weconsider parents whose youngest child is of ages between 1 and 12, as parental time constraints are possibly more binding for that age group than for older children who areteenagersoryoungadults(DelBocaetal.,2014;WikleandCullen,2023). As a possible explanation for the gendered division of labour reallocation between paid work and childcare responsibilities during the pandemic, studies have found evidence of shifts in perceptions, attitudes, and beliefs about gender roles (Danzer et al., 2021; Boring and Moroni, 2023). For example, based on a sample of 1000 individuals from the French working population, Boring and Moroni (2023)showthattherewasaconsiderableincreaseintheshareofmen,withchildren aged 12 and below, who believe in traditional gender roles based on a spe- 3

ciallydesignedsurveyonbeliefsaboutgenderroles. Theauthorsfindasignificant 14-15-percentagepointincreaseintheshareoffatherswhoagreewithstatements like “A man’s job is to earn money; a woman’s job is to look after the home and family” and “All in all, family life suffers when the woman has a fulltime job” comparedtopre-lockdownshares. We utilize Statistics New Zealand’s administrative data hub–the Integrated Data Infrastructure (IDI)–to investigate the effect of the COVID-19 lockdown in New Zealand on parental employment. Our identification strategy compares a sampleofopposite-sexparentslivingwithinthesame“householdbubble”during the pandemic to a similar sample of couples identified from a pre-pandemic era. Weemployadynamicframeworktotrackhowmothers’andfathers’employment evolved over a period around the lockdown month, March 2020–spanning from five months before to five months after the lockdown was implemented. Similarly, for the sample of parents identified from a recent pre-pandemic era (control period), we incorporate a dynamic setting centred around a ‘placebo’ lockdown month of March 2019. Our empirical approach allows us to control for seasonal variations in employment trends that could additionally influence people’s labour marketoutcomes. Ourresultsshowthatrelativetothepre-pandemicera,thepandemicshutdown was followed by a statistically significant decline of 1-2 percentage points in the employment propensity for mothers. However, for most fathers, we do not detect any significant differences in the likelihood of being employed between the treatmentandcontrolperiods. Furtherstratificationrevealssubstantialdropinthe likelihood of being employed for mothers who were non-employed in the month prior to the lockdown period compared to similarly situated mothers from one year prior. To further understand whether the employment declines among mothersdifferfromthechangesexperiencedbywomenwithoutchildren,westudythe employment patterns of future parents who were observed to bear their first child in and after 2021–i.e., at least a year later following the onset of the pandemic. Interestingly, the results are largely comparable to that of the actual mothers, in- 4

dicating that the drop in employment is not uniquely experienced by mothers but instead, also by other women who did not have children during the period of our analysis. The rest of the paper is organized as follows: Section 2 provides background informationontheonsetofthepandemicinNewZealand;Section3describesthe datausedandprovidesdescriptivestatistics;Section4discussesouridentification strategy;Section5presentsourresultsandthelastSection6concludes. 2 The New Zealand context OnMarch23rd 2020,theNewZealandPrimeMinisterdeclaredastrictnationwide lockdown from March 26th onward.3 This announcement came 24 days after the first case of COVID-19 was reported in New Zealand. New Zealand had a particularly strict lockdown relative to other Western countries during the beginning of the pandemic. According to Mathieu et al. (2020), as of the first week of April 2020, the value of the stringency index for New Zealand was 96.3. For context: the stringency index is based on a scale of 0-100 with a higher score indicating a stricter response. In comparison, the respective value stood at 79.6 for the United Kingdomand72.7fortheUnitedStates. Apart from a few essential services, such as the supply of food and healthcare, most on-site business activities and professional services, including child daycare centres, schools, colleges, and universities, were closed. Furthermore, according to the government’s lockdown guidelines, New Zealand residents were required to stay within household-level isolation “bubbles”. People could only leave their houses for groceries, healthcare needs, and exercise in their immediate neighbourhood. These rules applied to everyone except for those identified as “essentialworkers”,suchashealthcareandgroceryworkers. Moreover,childcare was available for free for essential workers with children aged up to 13.4 Aided 3Seehttps://covid19.govt.nz/about-our-covid-19-response/history-of-the -covid-19-alert-system/;AccessedonMarch15,2023. 4Seehttps://www.education.govt.nz/news/childcare-available-again-for-w 5

by the government’s strict border restrictions imposed on international travel, the lockdownrestrictionswereeventuallyliftedonJune8,2020. In response to the possibility of mass unemployment resulting from the pandemic-induced containment measures, the New Zealand government introduced a large-scale Wage Subsidy Scheme.5 The primary objective of the expansionaryfiscalpolicywastohelpfirmowners,includingself-employedindividuals, retaintheirbusinessesbyfinanciallysupportingtheirstaff. Thewagesubsidyschemeprovidedrapidup-frontpaymentstobusinessesthat wereaffectedbytheCOVID-19restrictions.6 Anemployerwaseligibletoreceive financialsupportiftheirrevenuewasatleast30percentlowerintheprior30days compared to a similar period in the year earlier. The government paid out a flat weeklyrateof$585.80perpersontofull-timeworkersand$350.00perpersonto part-time employees. On average, the full-time rate was around 58 percent of the median weekly earnings in 2019 (Maani, 2021). With total funding amounting to around $13.9 billion, the wage subsidy scheme accounted for almost 4.3 percent of the nation’s GDP at the time and supported over 60 percent of the employed workforce(Kidoetal.,2021). 7 3 Data and descriptive statistics 3.1 Data preparation This research aims to understand how parental employment was affected at the onsetofthepandemicinNewZealand. Ouridentificationstrategytracksparental orkers-in-alert-level-4-businesses-and-services/;AccessedonApril12,2023. 5Seehttps://www.workandincome.govt.nz/covid-19/previous-payments/wage-s ubsidy-extension.html;RetrievedonMarch21,2023. 6Retrieved from Ministry of Social Development. See https://www.msd.govt.nz/abou t-msd-and-our-work/work-programmes/wage-subsidy-integrity/index.html; Accessedon19January2024. 7RetrievedfromInternationalMonetaryFund’sinformationoncountrywisepolicyresponses toCOVID-19. Seehttps://www.imf.org/en/Topics/imf-and-covid19/Policy-Respo nses-to-COVID-19;AccessedonApril12,2023. 6

employmentspanningfromfivemonthspriortothelockdownandfivemonthsafterthelockdownandcomparesthoseemploymentpatternstothetrendsobserved for similarly situated parents over the same monthly periods from just a year earlier. Tothatend,ourdataisdividedintotwoperiods-thepre-pandemicperiod(or thecontrolperiod)thatspansfromOctober2018untilAugust2019;andthepandemic period (or treatment period) spanning from October 2019 through August 2020. Since the pandemic-induced lockdown was enacted and swiftly enforced in the month of March 2020, our analysis is centred around that month (t =0) in both the control and the treatment periods. Specifically, while March 2020 is the actual lockdown month (the treatment month), we consider March 2019 as our ‘placebo’lockdownmonth. Forouranalysis,weutilizedatafromtheIntegratedDataInfrastructure(IDI)– alarge-scaledatabasehostedbyStatisticsNewZealand(StatsNZ).TheIDIholds a wide range of administrative data collected from different ministries and public agenciessuchasInlandRevenue,theDepartmentofInternalAffairs,theMinistry of Education, the Census, etc. Information is collected at the individual level and individualscanbelinkedusingauniqueconfidentializedidentifier. To identify the population of interest, we begin with the Department of Internal Affairs’ (DIA) birth records. The DIA birth register documents all births in NZ and contains information such as the child’s birth date and gender, as well as identifiers of their parents. This enables us to identify all children born to each parent couple. Additionally, parental identifiers allow individual linkage to other datasets,suchastheirmonthlytaxrecords,whichprovidelabourmarketinformation. We incorporate several steps to refine our sample in order to evaluate an unbiased estimate of the effect lockdown on parental employment gaps. Overall, oursampleincludesparentswhoseyoungestchildisagedbetween1and12years old in February 2019/2020.8 To avoid confounding influences from unobserved 8Inouranalysis,weperformourregressionforeachageyearofthechildseparately. Notethat achildthatis,forexample,oneyearoldinFebruary2020canbebetween12and23monthsold (andsoon). 7

individual-level preferences, we first exclude existing mothers who subsequently gavebirthtoanotherchildwithinthenexttwoyearsfromtheperiodofevaluation as their labour market decision might differ from mothers who are not expecting any further children. We then restrict the sample to couples who have, in total, less than four children. We apply this condition since larger families may have different socio-economic conditions and labour market preferences compared to smaller family sizes (Cools et al., 2017). We also exclude half-siblings born to differentparentstoreducepossibleconfoundinginfluencesoffamily-specificunobservedcomplicationsarisingfromparentalseparations. Next, we use the personal details table to incorporate demographic information. The personal details table is prepared by Stats NZ based on the information retrievedfromvariouspopulation-leveladministrativedatasourcesincludedinthe IDI. The table documents individuals’ demographic information, including birth date, deceased date, ethnicity, and sex. We use information from the personal details table to control for observable characteristics and further homogenise our sampleofparentsinthetreatmentandthecontrolperiods. We use the deceased date from the personal details table to remove observationswhereatleastoneoftheparentswasdeceasedafterthebirthofthelastchild within the following two years. Next, we use the parental birth date and restrict the sample to mothers aged between 20 and 40 and fathers aged between 20 and 45atthebirthofthelastchild. Ethnicity information from the personal details table is used to create indicators of ethnic identity. Notably, an individual may identify with multiple ethnicities. In New Zealand, ethnicity is often prioritised, i.e. if an individual identifies as both Asian and European, they are noted as Asian, which is prioritised over European. It is worth noting that outcomes may vary substantially across various ethnicities due to cultural, social, and economic differences (Harris et al., 2006; Barnettetal.,2004). Althoughcontrollingforethnicidentityinregressionmodels can capture some of the ethnic differences, there may still be unobserved drivers of ethnic disparities that may be correlated with individuals’ labour market and 8

health outcomes. Failure to account for such unaccounted heterogeneities may contaminatecausalmechanisms. Assuch,toensuregreatercomparability,werestrict our analysis to families where both parents identify only as NZ European ethnicity and have no other ethnic identity. NZ Europeans are the largest ethnic groupinNZ. Next,eachfamilyinoursampleneedstobecomprisedofparentsandchildren belongingtoasingle“householdbubble”sothattheoutcomescanbeattributedto choices derived from family-level interactions. This requires us to focus on parentswhoarenotseparatedduringtheperiodsunderevaluation. Onepossibleway to achieve this is to use the DIA’s marriage records to restrict our sample to married parents. However, there are two limitations to this approach. First, the share of couples who are married is relatively small in New Zealand. This is largely due to the large prevalence of de facto partnerships. In New Zealand, couples in de facto relationships have similar legal rights as married couples, including but not limited to regulations governing access to welfare support, immigration policies and uptake of health services. However, de facto partnerships are not administratively recorded. Second, even though we can observe the incidence of marriages and divorces until mid-2022, New Zealand law requires a mandatory two-year separation period before someone can seek a dissolution order from the court. Therefore, it is likely that some parents who appear to be married in the datamayactuallybeseparatedordonotliveinthesamehousehold. Our analysis uses the address notification dataset as an alternative indicator of single ”household bubbles”. Stats NZ prepares the dataset and uses multiple administrative sources to identify an individual’s residential location. It provides location information at different geographic levels, with the most granular being onthemeshblocklevel.9 Wefocusonlyonparentswhoresideinthesamelocation throughoutourstudyperiods. Theaddressnotificationdatasetisalsousedtoderivetwoadditionalvariables 9MeshblockisthesmallestgeographicalareainNewZealandstandardgeographicclassification,representingroughly30to60dwellings. Seehttps://vhin.co.nz/guides/geograph ic-information-in-idi/;AccessedonApril2,2023. 9

in our empirical analysis. First, we create a geographic marker of a family’s residentiallocationtoaccountforpossibledifferencesinlabourmarketeffectsacross regions with varying levels of population density. This is done by classifying regions into five categories including, Auckland, Wellington, Canterbury (includes Christchurch city), the rest of the North Island, and the rest of the South Island. Second,Atkinsonetal.(2019)calculatedasocialdeprivationindexforeachmeshblock using the 2018 Census to represent the economic conditions of families residingineachlocation. Theindexrangesbetween1and10,with1beingtheleast deprived and 10 the most deprived. We aggregate this information to form three groups: Index1-3,4-6and7andabove. Lastly,thesamplewaslimitedtoparentswhowerephysicallypresentinNew Zealand to participate in the labour market. We use the Ministry of Business, Innovation&Employment’s(MBIE)Immigrationandvisadatasets. Thedatasetincludes administrative data on the movement of individuals across New Zealand’s border including migrants, international visitors, and New Zealand citizens. We exclude families where at least one of the members was outside NZ for a minimumof90days. Weonlyremoveobservationswherethebeginningorendofthe overseasspellfallswithinourperiodofinteresttoadditionallyavoidthepossibility of at least one parent being employed overseas. Overseas employment is not capturedintheIDIdatabase,butmayaffectourregressionestimates. Our key research objective focuses on parents’ labour market implications from the pandemic-induced lockdown by examining parents’ employment status.10 Parents are linked to the Inland Revenue Employer Monthly Schedule (IR- EMS), which provides monthly information for seven different income sources, includingwagesandsalaries. Parents(motherorfather)areconsideredemployed iftheyreceiveincomefromwagesandsalaries. TheIR-EMSdatasetdoesnotin- 10Wedonotanalysehowearningsareaffectedforseveralreasons. First, wedonothavesufficientdataonhoursworked,andthesemighthavechangedsubstantiallyduringtheonsetofthe pandemic. Furthermore,employersaffectedbytheCOVID-19restrictionsreceivedupfrontfinancialaidfromtheWageSubsidySchemeandpassedpaymentsontostaffinwages–however,itis notpossibletoidentifywhichemployeereceivedmoneyfromthescheme. 10

clude information from self-employment. This is collected in a separate dataset; however, the information is only available on the annual level and refers to the fiscal year, which ends in March. We use the relevant IR-IR3 data set to identify income from self-employment for both parents. We then removed families where atleastoneparentearned$15000perfinancialyear(in2020NZ$terms),assuming that income from self-employment is a major income source.11 However, our findingsarerobusttoloweringorincreasingtheincomethreshold. Ourfinalsample consists of 71424 families in the treatment period and 72510 families in the controlperiod(seealsoTableA1). 3.2 Descriptive statistics Figure2showsparentalemploymentratebyageoftheyoungestchildinFebruary 2020, the month before the nationwide lockdown was implemented. We observe that the share of mothers who are employed increases with age of the youngest child. For example, about 55% of mothers whose youngest child is one year old received income from wages and salaries in February 2020. This share increases to 71% for mothers whose youngest child is seven years old. However, beyond this age, the increase in mothers’ labour market participation is only marginal (e.g., 72% of mothers whose youngest child is 12 years old are employed). This suggests that among mothers, the return into employment as children get older plays a major role. Among fathers, there appears to be no visible trend. In contrast, on average, fathers’ labour market participation declines slightly as age of theyoungestchildincreases. Figure3showsparentalemploymentrates,separatelyformothers(leftgraph) andfathers(rightgraph),forthecontrolperiod(“2019”)andthetreatmentperiod (“2020”). The employment rate is indexed at February 2019 and 2020, respectively. For mothers, we observe almost identical employment patterns in the five 11As an example, for parents observed between October 2019 and August 2020, we look at income from self-employment in March 2020 (which refers to income from April 2019-March 2020)andinMarch2021(whichreferstoincomefromApril2020-March2021). 11

Figure2: Parentalemploymentbytheyoungestchild’sage Note:IDIandauthorscalculations. months before the start of the pandemic. However, following the onset of the pandemiclockdown,weobservesignificantdifferencesinemploymentrates-the employment rate for mothers in the control period is much higher compared to mothersinthetreatmentperiod. Forfathers,weobservealmostidenticalemployment patterns in the five months before the start of the pandemic. However, and in stark contrast to post-March trends observed for the mothers, we observe no differencesintheoverallemploymentrateoffathers. 4 Empirical approach Asalreadyhighlighted,ouridentificationstrategyincludestwosetsofparents: the pandemic sample (or treatment group) and the pre-pandemic sample (or control group). For each family in the treatment group, we track monthly employment patterns of parents from October 2019 to August 2020. The lockdown month March 2020 is used to divide observations into pre-treatment and post-treatment 12

Figure3: Parentalemploymentaroundthelockdown Notes:IDIandauthorscalculations. EmploymentrateisindexedatFebruary2019andFebruary2020,respectively. The parent’syoungestchildisbetween1and12yearsoldinFebruary2019andFebruary2020,respectively. periods. Similarly, employment trends of each parent in the control group are observed for the months between October 2018 and August 2019, with the control period centred around March 2019 as the placebo lockdown month. While the control group can be assumed to be unaffected by the pandemic during the period they are examined, our empirical approach allows us to control for possible seasonal variations that could additionally affect parental employment during the pandemic. Theempiricalanalysisformothersandfathersisrunseparately. Forourbaselinespecification,weestimate: y =α+β .Post +β .(Post ×Pandemic)+X’ β +µ +u (1) it 1 t 2 t i it 3 i it suchthat  1 ifmontht ≥March2019/2020 Post = t 0 ifmontht <March2019/2020 and  1 ifparenti∈pandemicsample Pandemic = i 0 ifparenti∈pre-pandemicsample The binary outcome variable y is equal to 1 if a parent (mother/ father) re- 13

ceives income from wages and salaries in the respective month, 0 otherwise. The parameterofinterestβ measurestheaveragedifferenceinemploymentoutcomes 2 betweenparentsinthepandemicsamplecomparedtothepre-pandemicsample.12 We control for time-varying covariates, which include age (in years) of each parent, the deprivation index at the meshblock level and the region of residence. We also control for individual fixed effects µ. Lastly, u denotes an idiosyncratic eri it rorterm. Thismodelallowsforsinglecoefficientstobeestimatedforeachparent, foreachchildagecategorybetweenoneand12years. While the model in equation (1) allows us to estimate the average change in employment over pre- and post-lockdown months, we also estimate a dynamic specification to examine the monthly trend of the differences in the employment outcomes between parents in the pre-pandemic and the pandemic sample. More specifically,weestimate: +5 +5 y it =ρ+ ∑ γ k T k + ∑ θ k T ×Pandemic i +X’ it β 3 +µ i +u it (2) k=−5(̸=−1) k=−5(̸=−1) where the parameters represented by γ estimate the average likelihood of a k parent being employed in each month (k) relative to the lockdown or the placebo month, for each parental sample, respectively. The parameters θ estimates the k difference between the employment outcomes of parents in each of the two samplesforeachmonth(k)relativetothelockdownortheplacebomonth. Themonth ofFebruary(i.e. k=−1)fromthepandemicandpre-pandemicperiodistreatedas ourreferenceperiodandthereforedroppedfromouranalysis. Thisdynamicanalysisallowsustoempiricallytesttheparalleltrendsassumptionandverifywhether there are any significant anticipatory effects prior to the lockdown month. In all ourempiricalspecifications,weclusterourstandarderrorsattheindividuallevel. 12As we apply a fixed effects model, there is no parameter referring to whether the parent belongedtothepandemicorpre-pandemicsample. 14

Figure4: Pandemic’simpactonparentalemployment Note: IDIandauthorscalculations. Thegraphshowsforfathers(leftpanel)andmothers(rightpanel)thelikelihoodto beemployed(andthecorresponding95%confidenceinterval)duringtheonsetofthepandemic(t≥0)bytheageofthe youngestchild. 5 Results We first estimate parental employment status using equation (1), separately for fathersandmothersandforeachindividualagegroupoftheyoungestchild(aged from one to 12). Figure 4 plots the estimated regression coefficients of interest ˆ (β ) from equation (1) and the corresponding 95% confidence interval. The left 2 (right)panelreferstochangesinfathers’(mothers’)employmentduringthetreatment period relative to the control period. Coefficient estimates are presented in TableA2. TheleftpanelofFigure4showsthatinmostcases,theemploymentpropensity of fathers in the pandemic sample did not differ significantly from that of fathers in the pre-pandemic sample during the post-March months. That is, on average, 15

fathers did not experience any significant decline in their likelihood of being employedduringthefirstsixmonthsofthepandemic(March-October2020). For mothers (right panel), the pattern appears to be quite different. Mothers inthepandemicsampleexperienceasignificantdeclineinthelikelihoodofbeing employedattheonsetofthepandemicwhencomparedtosimilarlysituatedmothers in the pre-pandemic sample. On average, the magnitude of the decline ranges betweenoneandtwopercentagepointsformotherswhoseyoungestchildisaged between 1 and 9 and those who are aged 12. For mothers whose youngest child is either ten or eleven years old, the estimated effect on the likelihood of being employedisnotsignificantlydifferentfromzero,althoughthecoefficientremains negative,similartoallotherchildages. Overall, our baseline findings indicate that mothers’ labour market participation, especially among those with younger children, declined during the onset of the pandemic while fathers experienced no change. The differences in the estimated effect on each parent’s probability of being employed varied across age of the youngest child, which indicates that parental labor market participation or non-work time allocation conversely may not vary monotonically across child ages. To test how the effects of the pandemic-induced lockdown on parents’ employment propensity evolved over time (from the pre-pandemic months to the post-lockdown period), we perform an event analysis with individuals fixed effects (see Eq 2). We run separate regressions for each parent-child age combination. This estimates a monthly series of ten coefficients (θ in Eq 2) to represent k the dynamic effects of the pandemic on parental employment. Specifically, the coefficients represent the difference in employment propensity between the pandemic sample and corresponding pre-pandemic sample for the four months prior tolockdownperiod,thelockdownmonthandfivemonthspost-lockdown. Table A3 (Table A4) presents the coefficients for mothers (fathers) by age of the youngest child. With the exception of a few observations, both the employmentpropensityformothersandfathersbetweenthecontrolandpandemicperiod 16

do not vary significantly when estimated relative to the reference period (February).13 Focusing on the initial months of the pandemic, for most child ages (except for when the youngest child is 4 years old) we do not observe any statistically significant variation in the employment likelihood for fathers from the treatment group– this is in line with that of Figure 4. For mothers we find that in the first month of the pandemic (t = 0), employment propensity did not change significantly. Thisisnotsurprisinggiventhatthelockdownwasdeclaredaroundtheend of March (March 26th). However, we observe that from April onwards, there was asignificantdropinemploymentformothersinthepandemicsample. Moreover, the magnitude of the coefficients do not vary largely across the post-lockdown months until August. Thus, we do not find any indications that the drop in mothersemploymentprobabilityintensifiedovertime. 5.1 Exiting or not-entering employment The findings so far indicate that during the onset of the pandemic, mothers employment propensity declined while fathers were not significantly affected. Two possibilities may drive our overall results: (i) previously employed mothers exiting employment, or (ii) previously non-employed mothers staying out of the labour force. We empirically test whether one or both groups drive our overall findings. As discussed earlier, the aim of the wage subsidy scheme was to secure employment and prevent large-scale business closures due to the lockdown restrictions. We split our sample by employment status in the month prior to the lockdownandtheplacebomonth,i.e.,February2019andFebruary2020,respectively. We create a binary indicator equal to 1 if either parent were employed at least for one month between March and August of the pre-pandemic and pandemic years 13Two exceptions are observed for fathers whose youngest child was aged four and mothers whose youngest child was aged seven. The pre-period coefficients for most months in both the casesarestatisticallydifferentfromzerowhencomparedtothereferencemonth. 17

of 2019 and 2020 respectively, 0 otherwise.14 Table A5 shows the share of fathers/mothers who were employed at least once in the post-March months for the treatment and control group, differentiated by the employment status in the pre-lockdown/placebo month. For parents who were employed in the month of February in 2019 and 2020, the share of individuals who were employed for at least one month in the post-March period were similar between pandemic and pre-pandemic group. This was observed for both mothers and fathers across all child ages. However, among the corresponding fathers who were not employed in the month prior to the lockdown or the placebo month, we see a small drop in the post-period employment share in the treatment period compared to the control period. In the majority of those cases, the drop is below 2 percentage points. For mothers, we observe relatively larger drops among those who were notemployedinFebruary2020comparedtothenon-employedmothersinFebruary 2019. The difference between the two groups ranges between two and seven percentagepoints. Weaddtoourdescriptiveanalysisbyrunninglinearprobabilitymodelstoestimate the likelihood of being employed in at least one of the months in the postlockdown period, differentiated by the employment status in the month February. Inallourregressions,wecontrolforparentalage,region,deprivationindex,number of siblings and gender of the youngest child. We add a dummy variable to denote the treatment (or pandemic) period. We again run separate regressions for eachparent-childagecombination. Theseriesofregressionsareestimatedbythe employmentstatusinthemonthofFebruary(priortothelockdownandthecorrespondingplacebomonth). Wereporttheregressioncoefficientsofinterestderived fromatotalof48regressionsinFigure5. For fathers (top panel), the magnitude of the post-period coefficient appears to be economically small and not significantly different from zero at any conven- 14To test the robustness of our marker, we repeated this analysis and measured the outcome variable for (a) whether being employed in August 2019 and August 2020, respectively, and (b) thenumberofmonthsemployedinthet ≥0. Thefindingsdonotdifferqualitativelyandtables canbeobtainedfromtheauthorsuponrequest. 18

Figure5: Employmentprospectsbyinitialemploymentstatus Note:IDIandauthorscalculations.Thegraphshowsforfathers(toppanel)andmothers(bottompanel),differentiatedby beingemployedinFebruary(leftpanel)ornon-employedinFebruary(rightpanel)thecoefficient(andthecorresponding 95%confidenceinterval)forthetreatmentperiodtobeemployedatleastonceinthepost-periodbytheageoftheyoungest child. 19

tionallevel. Thestatisticallyinsignificantfindingspersistregardlessofthefathers’ employment status in the month of February t =−1 from the pandemic and the pre-pandemic years. For mothers who were employed at month t =−1, we also observethatthelikelihoodofbeingemployedforatleastonemonthintheperiod t ≥0doesnotdiffersignificantlybetweenthecontrolandtreatmentsamples. In comparison, when we look at the mothers who were not employed at t = −1, we see a statistically significant drop in their employment probability in the post-period. The negative effect varies between two and seven percentage points for those in the treatment period compared to the control period. This effect is statistically significant at the 5 percent level, for seven out of the 12 child age categories, and at the 10 percent level for two further two child age categories. The magnitude of this difference is significant - the average share of mothers who were not employed in February 2019, but had a job for at least one month between March and August 2019, was 17 percent. Relative to that share, a three percentage-pointdeclineinthelikelihoodofbeingemployedinthepost-lockdown monthsimpliesadropofalmost18percent. Our additional analysis provides empirical evidence indicating that the declineinemploymentobservedamongmothersduringtheonsetofthepandemicis largely driven by an increase of mothers not returning to or entering employment andlesssobymothersexitingemployment.15 5.2 Employment patterns of future parents The empirical evidence provided thus far show that during the onset of the pandemic, mothers’ labour supply significantly declined compared to similarly situated mothers a year prior; however, we do not find any changes for fathers. One 15To explore further mechanisms, we also perform disaggregated analysis by parental education level and prior industry characteristics (e.g., essential versus non-essential sectors). In the disaggregatedanalyses,thesamplesizewasreducedsubstantiallytoprovideconsistentstatistical evidence. However,wedonotfindanystrongevidenceofvariationsinkeyfindingslargelyhold acrosseducationallevelsandacrossindustries. Theadditionalresultsthatarenotprovidedforthe sakeofbrevityareavailableuponfurtherrequest. 20

limitation is that we cannot say whether this pattern is uniquely experienced by mothers or whether the decline in employment is also observed among women without children. To shed some light on this aspect, we construct a sample of futureparentswhodidnothaveanychildduringtheperiodsstudiedinouranalysis. This sample contains couples who had their first child in 2021 or 2022. With the exception of being parents, we estimate similar empirical specifications adopted earlierinouranalysistofutureparents. We calculate the employment share for the five months before and after lockdown and the placebo month to provide context for how lockdown affected the labour supply of future parents. Figure 6 show their employment patterns are largely similar to the trends observed for actual parents in our sample. With the onset of the pandemic, we only observe a visible employment gap between the controlandtreatmentgroupforthesampleoffuturemothers. Figure6: Employmentoffutureparentsaroundthelockdown Notes:IDIandauthorscalculations.EmploymentrateisindexedatFebruary2019,resp2020.Theparent’syoungestchild isbetween-2and-1yearsoldinFebruary2019,resp.2020. We repeat our regressions to estimate the magnitude of the change in labour supply of future parents. The coefficients of the interaction effect for the fixed effects linear probability model can be found in Figure 4. Similarly, we do not observeanysignificantdeclineinemploymentlikelihoodforfuturefathersforthe months March until August between the control and the treatment group. For futuremothers,thedeclinerangesbetweenoneandtwopercentagepointsforthose affectedbythepandemiccomparedtothecontrolperiod. Moreover,whenswitch- 21

ing to the dynamic event analysis (Tables A3 and A4), we do not observe any significant differences in the pre-period for both groups of future mothers and for futurefathers. However,fromApril2020onwards,thereisasignificantdeclinein thepost-lockdownmonthsobservedonlyforfuturemothers. Themarginaleffects appear to be similar in size when compared to coefficients derived for mothers in our earlier analysis. Further stratification reveals that the number of future mothers not employed in February and not being employed for at least one month in the post-March period (t ≥ 0) is significantly larger (at the 10% level) for those experiencing the pandemic compared to the future mothers in the control group. These findings indicate that the drop in employment is not uniquely experienced bymothersbutthepatternseemstobecommonacrosswomeningeneral. 6 Conclusion The employment effects triggered by the COVID-19 pandemic are unlike those observed in earlier economic recessions in the recent past. The economic downturn resulting from the pandemic had disproportionate effects on women’s labour supply. As per the existing international literature, these effects seem to be more pronounced for mothers with school-aged and younger children (Goldin, 2022; Alon et al., 2022). The adverse impact of the pandemic on women’s labour supply has been attributed to several reasons including large-scale job losses in sectors and occupations that have higher shares of female workers and unequal distribution of household activities. The combined effect of these factors alongside different policies adopted by governments to address a potential economic crisis mayhavehadvaryingimpactsonwomen’semploymentoutcomes. New Zealand is one of the few examples where most of the government’s pandemic-related economic resources was devoted to financing a generous wage subsidyschemethatsupportedfirmstosecureemploymentofexistingemployees during the lockdown period. This is in contrast to pandemic-related policies targeted to consumers directly in the form of transfer payments or stimulus checks 22

and expansion of social welfare policies designed for economically vulnerable families. We find empirical support indicating that the New Zealand government’s policyresponsewascomparativelymoreeffectiveinminimizingtheoverallemployment decline resulting from the lockdown-induced reduction in economic activities. Specifically, our findings provide additional context to the the cross-country comparisonbasedonOECDdatapresentedearlierinthestudy. Wealsofindsuggestive evidence that the employment effects observed for New Zealand mothers were more likely to be driven by changes in family dynamics and/or individuallevel choices than by business-related effects and firm closures. This is because firm closures during the pandemic have been found to have affected both male and female labour supply in other countries. For instance, evidence from the US shows that while women experienced larger decline in employment outcomes, employment rates dropped for men (with and without children) too. However in caseofNewZealand,wedidnotfindanyrelevanteffectsonfathers’employment propensity during the post-lockdown implementation period. In general, the employmenteffectsinouranalysisseemtobesmallerinsizecomparedtothelabour market evidence documented in the current international literature for both men (fathers)andwomen(mothers). Our detailed analysis also provides evidence that the relationship between child age and parental labor supply may not be monotonic. For example, we find no relevant effects for mothers when their youngest child is aged between 10 and 11. Such variations in the link between child age and parental labor market outcomescastdoubtonexpectationsthatchildcareneedseaseforparentsaschildren grow older and become more self-sufficient. As such, our analysis paves the way for future research to explore, with greater detail, the evolution of parent-child interactions and the possible effects of child welfare policies that may influence thoseinteractions. Our analysis also shows that future mothers who were childless at the time of ouranalysisalsoexperiencedcomparabledeclinesinemployment. Thisisincon- 23

trast to the findings that show that mothers with especially younger children experienced larger labour market impacts compared to mothers with older children. We also find evidence that the decline in mothers’ employment rate was largely driven by non-employed mothers delaying their return to employment rather than employedmothersleavingthelabourforce. Finally,ourempiricalspecification andtheuseofdetailedadministrativedata allow us to address some of the potential concerns associated with the relevant empiricalstudiesinthisspace. Firstly,webelievethedataenabledustoidentifya ‘householdbubble’(orafamilyunit)withmoreprecisioncomparedtootherstudies. This is achieved with the help of several detailed administrative data sources with information on birth records and personal life events, address notifications, and international border movements. Unlike studies which rely on large-scale surveys, we are able to estimate the employment gap between fathers and mothers based on individuals belonging to the same family unit. Secondly, our control groupincludedsimilarlysituatedparentsfromarecentpre-pandemicperiodrather than comparing non-parents who are matched with parents based on observable characteristics. Comparing parents to non-parents may involve selectivity issues as labour supply decisions may vary across families conditional on the presence ofachild,individuals’childbearingintentions,andotherunobservedpreferences. Additionally, our empirical model allows us to control for seasonal variations in employment that could generate additional variations in the post-lockdown employmenttrends. Overall, our focus on New Zealand provides an interesting and alternative insight to a well-documented pandemic knowledge base, given its distinct policy setting and detailed data availability. By mitigating the risk of business closures, it is possible that policies that were more specifically designed to prevent mass layoffs in the economy may have reduced the size of the employment gaps observedbetweenmenandwomenduringthepandemic. 24

References S. Albanesi and J. Kim. The gendered impact of the covid-19 recession on the us labormarket. Technicalreport,NationalBureauofEconomicResearch,2021. T. Alon, M. Doepke, J. Olmstead-Rumsey, and M. Tertilt. The impact of the coronavirus pandemic on gender equality. Covid Economics Vetted and Real- TimePapers,4:62–85,2020a. T.Alon,M.Doepke,J.Olmstead-Rumsey,andM.Tertilt. Thistimeit’sdifferent: the role of women’s employment in a pandemic recession. Technical report, NationalBureauofEconomicResearch,2020b. T. Alon, S. Coskun, M. Doepke, D. Koll, and M. Tertilt. From mancession to shecession: Women’s employment in regular and pandemic recessions. NBER MacroeconomicsAnnual,36(1):83–151,2022. J. Atkinson, C. Salmond, and P. Crampton. Nzdep2018 index of deprivation, final research report, december 2020. University of Otago: Wellington, New Zealand,5541:1–65,2019. R. Barnett, G. Moon, and R. Kearns. Social inequality and ethnic differences in smokinginnewzealand. Socialscience&medicine,59(1):129–143,2004. J. Bluedorn, F. Caselli, N.-J. Hansen, I. Shibata, and M. M. Tavares. Gender and employment in the covid-19 recession: Cross-country evidence on “shecessions”. LabourEconomics,81:102308,2023. A. Boring and G. Moroni. Turning back the clock: Beliefs about gender roles duringlockdown. LabourEconomics,page102363,2023. Z.Cheng,S.Mendolia,A.R.Paloyo,D.A.Savage,andM.Tani.Workingparents, financial insecurity, and childcare: mental health in the time of covid-19 in the uk. ReviewofEconomicsoftheHousehold,19:123–144,2021. 25

R. Chetty, J. N. Friedman, M. Stepner, et al. The economic impacts of covid-19: Evidence from a new public database built using private sector data. Technical report,nationalBureauofeconomicresearch,2020. S. Cools, S. Markussen, and M. Strøm. Children and careers: How family size affects parents’ labor market outcomes in the long run. Demography, 54(5): 1773–1793,2017. N. Danzer, M. Huebener, C. K. Spiess, A. Pape, N. A. Siegel, and G. Wagner. Crackingunderpressure? genderroleattitudestowardmaternalemploymentin timesofapandemic. 2021. D. Del Boca, C. Flinn, and M. Wiswall. Household choices and child development. ReviewofEconomicStudies,81(1):137–185,2014. C.Goldin. Understandingtheeconomicimpactofcovid-19onwomen. Technical report,NationalBureauofEconomicResearch,2022. R. Harris, M. Tobias, M. Jeffreys, K. Waldegrave, S. Karlsen, and J. Nazroo. Effects of self-reported racial discrimination and deprivation on ma¯ori health and inequalities in new zealand: cross-sectional study. The Lancet, 367(9527): 2005–2009,2006. Y. Kido, S. Kothari, and Y. C. Wong. New zealand: Selected issues. Technical ReportCountryReportNo.21-89,InternationalMonetaryFund,2021. M. Kugler, M. Viollaz, D. Duque, I. Gaddis, D. Newhouse, A. Palacios-Lopez, and M. Weber. How did the covid-19 crisis affect different types of workers in thedevelopingworld? WorldDevelopment,page106331,2023. S. A. Maani. Covid-19: government responses to labour market disruptions and economic impacts. In Management Perspectives on the Covid-19 Crisis, pages 26–43.EdwardElgarPublishing,2021. 26

E. Mathieu, H. Ritchie, L. Rodés-Guirao, C. Appel, C. Giattino, J. Hasell, B. Macdonald, S. Dattani, D. Beltekian, E. Ortiz-Ospina, and M. Roser. Coronavirus pandemic (covid-19). Our World in Data, 2020. https://ourworldindata.org/coronavirus. B. Petrongolo. Gender segregation in employment contracts. Journal of the EuropeanEconomicAssociation,2(2-3):331–345,2004. J.WikleandC.Cullen. Thedevelopmentalcourseofparentaltimeinvestmentsin childrenfrominfancytolateadolescence. SocialSciences,12(2):92,2023. M. A. Wynne and N. S. Balke. Recessions and recoveries. Federal Reserve Bank ofDallas,EconomicReview,firstquarter,pages1–18,1993. 27

A Disclaimer The results in this paper are not official statistics, they have been created for researchpurposesfromtheIntegratedDataInfrastructure(IDI),managedbyStatistics New Zealand. The opinions, findings, recommendations, and conclusions expressedinthispaperarethoseoftheauthors,notStatisticsNZ. TheresultsarebasedinpartontaxdatasuppliedbyInlandRevenuetoStatistics NZ under the Tax Administration Act 1994. This tax data must be used only for statistical purposes, and no individual information may be published or disclosed in any other form, or provided to Inland Revenue for administrative or regulatory purposes. Any person who has had access to the unit record data has certifiedthattheyhavebeenshown,haveread,andhaveunderstoodsection81of theTaxAdministrationAct1994,whichrelatestosecrecy. Anydiscussionofdata limitationsorweaknessesisinthecontextofusingtheIDIforstatisticalpurposes, andisnotrelatedtothedata’sabilitytosupportInlandRevenue’scoreoperational requirements. Access to the anonymised data used in this study was provided by Statistics NZinaccordancewithsecurityandconfidentialityprovisionsoftheStatisticsAct 1975. Only people authorised by the Statistics Act 1975 are allowed to see data about a particular person, household, business, or organisation, and the results in this paper have been confidentialised to protect these groups from identification. Careful consideration has been given to the privacy, security, and confidentiality issuesassociatedwithusingadministrativeandsurveydataintheIDI. FurtherdetailcanbefoundinthePrivacyimpactassessmentfortheIntegrated DataInfrastructureavailablefromwww.stats.govt.nz. I

TableA1: Numberoffamiliesperchild’sage Child’sage† Controlperiod‡ Treatmentperiod∗ 1 9855 9498 2 8388 8181 3 6483 6420 4 5688 5598 5 5397 5235 6 5277 5166 7 5496 5160 8 5361 5409 9 5250 5343 10 5217 5184 11 5127 5163 12 4971 5067 Total 72510 71424 Note:IDIandauthorscalculations.†asmeasuredinFebruary2019,resp2020. ‡ControlperiodisfromOctober2018untilAugust2019. ∗Treatmentperiodis fromOctober2019untilAugust2020. II

Table A2: Coefficients on the parental employment probability during the onset ofthepandemic Fathers Mothers Child’sage† Coefficient StdErr Obs Coefficient StdErr Obs -2 -0.001 (0.002) 17457 -0.016*** (0.004) 17457 -1 0.002 (0.003) 15945 -0.011** (0.004) 15945 1 -0.001 (0.002) 19353 -0.013*** (0.004) 19353 2 0.001 (0.002) 16569 -0.018*** (0.004) 16569 3 -0.001 (0.003) 12903 -0.018*** (0.004) 12903 4 -0.007*** (0.003) 11286 -0.008* (0.004) 11286 5 0.001 (0.003) 10632 -0.008* (0.004) 10632 6 0.001 (0.003) 10443 -0.017*** (0.004) 10443 7 -0.002 (0.003) 10656 -0.015*** (0.004) 10656 8 -0.001 (0.003) 10770 -0.007** (0.003) 10770 9 -0.001 (0.003) 10593 -0.012*** (0.003) 10593 10 0.003 (0.003) 10401 -0.004 (0.003) 10401 11 -0.003 (0.003) 10290 -0.005 (0.003) 10290 12 -0.002 (0.003) 10038 -0.008** (0.003) 10038 Note: IDIandauthorscalculations. † asmeasuredinFebruary2019, resp2020. *, **, and***signifystatistical significanceatthe10,5,and1percent-levels,respectively. III

TableA3: Fixedeffectseventstudymodel: fathersemployment Child’sage t -2 -1 1 2 3 4 5 6 7 8 9 10 11 12 -5 0.002 -0.004 -0.002 -0.002 -0.001 0.005 0.000 0.001 -0.001 -0.004 0.004 -0.005 -0.001 0.004 (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) (0.003) (0.003) (0.004) -4 0.003 -0.001 -0.001 -0.000 -0.001 0.006** -0.001 0.000 0.002 -0.001 0.005 -0.003 -0.002 0.004 (0.003) (0.003) (0.002) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) -3 0.004 0.001 -0.002 0.000 -0.000 0.006** 0.000 0.001 0.004 0.004 0.006* 0.000 -0.001 0.003 (0.003) (0.003) (0.002) (0.002) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) -2 -0.003 -0.002 -0.004** -0.001 -0.001 0.004** -0.001 -0.001 0.002 -0.003 0.000 -0.002 -0.002 0.000 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) -1 referencecategory 0 0.001 0.002 0.001 0.005** 0.003 0.003 0.003 0.005* 0.001 0.006** 0.002 0.001 0.001 0.003 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.003) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) 1 -0.001 0.000 -0.002 -0.001 -0.004 -0.004 -0.002 -0.001 -0.002 0.000 0.000 -0.001 -0.004 0.001 (0.003) (0.003) (0.002) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) 2 -0.002 -0.002 -0.002 -0.002 0.000 -0.004 -0.001 0.000 -0.002 -0.003 -0.000 -0.000 -0.002 -0.003 (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) 3 0.002 0.002 -0.004 0.003 -0.001 -0.002 0.004 0.003 -0.001 -0.003 0.003 0.005 -0.004 0.000 (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) 4 0.001 0.000 -0.006* -0.001 -0.005 -0.005 0.002 -0.001 -0.000 -0.006 0.004 0.002 -0.009** -0.001 (0.003) (0.004) (0.003) (0.003) (0.003) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) 5 -0.001 0.001 -0.006** -0.002 -0.002 -0.007 0.001 -0.001 -0.001 -0.004 0.002 0.000 -0.008* 0.001 (0.003) (0.004) (0.003) (0.003) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) Obs 17457 15945 19353 16569 12903 11286 10632 10443 10656 10770 10593 10401 10290 10038 Note:*,**,and***signifystatisticalsignificanceatthe10,5,and1percent-levels,respectively. IV

TableA4: Fixedeffectseventstudymodel: mothersemployment Child’sage t -2 -1 1 2 3 4 5 6 7 8 9 10 11 12 -5 0.003 0.001 -0.000 0.001 0.003 -0.005 -0.005 -0.003 0.011** 0.006 -0.006 -0.003 0.002 0.003 (0.005) (0.006) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.004) (0.005) (0.004) (0.004) (0.004) (0.004) -4 0.004 0.003 -0.002 0.002 0.003 -0.004 -0.006 -0.008* 0.010** 0.000 -0.002 -0.004 0.004 0.006 (0.005) (0.005) (0.005) (0.004) (0.005) (0.005) (0.005) (0.005) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) -3 0.007 0.007 -0.001 -0.000 0.003 -0.004 -0.006 -0.004 0.009** -0.001 0.002 -0.003 0.005 0.005 (0.004) (0.005) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) -2 0.002 0.001 0.002 0.001 0.005 -0.006* -0.005 -0.003 0.008** 0.002 -0.004 0.001 -0.003 0.000 (0.003) (0.004) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) -1 referencecategory 0 -0.001 0.006 0.001 0.002 -0.001 0.003 -0.000 -0.003 0.006* -0.000 -0.001 0.003 0.004 0.002 (0.003) (0.004) (0.003) (0.003) (0.004) (0.003) (0.004) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) 1 -0.014*** -0.010** -0.013*** -0.016*** -0.015*** -0.016*** -0.012*** -0.015*** -0.003 -0.006 -0.012*** -0.004 -0.003 -0.005 (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) 2 -0.016*** -0.013** -0.018*** -0.026*** -0.021*** -0.017*** -0.014*** -0.025*** -0.008** -0.011*** -0.016*** -0.010*** -0.005 -0.004 (0.005) (0.005) (0.004) (0.005) (0.005) (0.005) (0.005) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) 3 -0.012** -0.011** -0.015*** -0.018*** -0.016*** -0.012** -0.015*** -0.024*** -0.008* -0.006 -0.015*** -0.007 -0.004 -0.006 (0.005) (0.005) (0.004) (0.005) (0.005) (0.005) (0.005) (0.005) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) 4 -0.019*** -0.013** -0.017*** -0.025*** -0.016*** -0.014** -0.020*** -0.025*** -0.015*** -0.005 -0.022*** -0.010** -0.006 -0.007 (0.006) (0.006) (0.005) (0.005) (0.006) (0.006) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.004) (0.005) 5 -0.014** -0.013** -0.018*** -0.021*** -0.021*** -0.013** -0.012** -0.028*** -0.014*** -0.006 -0.022*** -0.009* -0.008* -0.009* (0.006) (0.006) (0.005) (0.006) (0.006) (0.006) (0.006) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) Obs 17457 15945 19353 16569 12903 11286 10632 10443 10656 10770 10593 10401 10290 10038 Note:*,**,and***signifystatisticalsignificanceatthe10,5,and1percent-levels,respectively. V

TableA5: Employmentsharesbyemploymentstatus Fathers Mothers Employedatt=−1 Non-employedatt=−1 Employedatt=−1 Non-employedatt=−1 Child’sage† Controlperiod Treatmentperiod ∆ Controlperiod Treatmentperiod ∆ Controlperiod Treatmentperiod ∆ Controlperiod Treatmentperiod ∆ 1 0.994 0.994 0.000 0.136 0.107 -0.029 0.989 0.986 -0.003 0.190 0.162 -0.028 2 0.994 0.995 0.000 0.130 0.122 -0.008 0.983 0.981 -0.002 0.194 0.158 -0.036 3 0.996 0.995 -0.001 0.111 0.109 -0.002 0.986 0.982 -0.004 0.182 0.162 -0.020 4 0.996 0.997 0.001 0.115 0.101 -0.014 0.987 0.990 0.002 0.199 0.151 -0.047 5 0.998 0.994 -0.004 0.117 0.102 -0.015 0.990 0.991 0.002 0.190 0.164 -0.025 6 0.995 0.996 0.001 0.103 0.087 -0.017 0.995 0.991 -0.004 0.210 0.138 -0.072 7 0.996 0.993 -0.003 0.133 0.092 -0.041 0.994 0.994 0.000 0.155 0.149 -0.006 8 0.996 0.998 0.002 0.100 0.097 -0.004 0.996 0.995 -0.001 0.162 0.137 -0.025 9 0.994 0.994 -0.001 0.085 0.097 0.011 0.997 0.994 -0.002 0.181 0.133 -0.048 10 0.995 0.994 -0.001 0.088 0.084 -0.004 0.994 0.995 0.001 0.158 0.136 -0.022 11 0.997 0.996 -0.001 0.093 0.089 -0.003 0.996 0.994 -0.001 0.131 0.134 0.003 12 0.996 0.995 -0.001 0.077 0.079 0.002 0.995 0.995 0.000 0.157 0.128 -0.029 Note:IDIandauthorscalculations.†asmeasuredinFebruary2019,resp2020.Thetableshowstheshareoffathers/motherswhowereemployedforatleastonemonthinthemonthsMarchuntilAugust(t≥0),differentiatedbytheemploymentstatusinFebruary2019,resp.2020(t=−1).∆is thedifferencebetweenthetreatmentandthecontrolperiod. VI