Less Work, More Labor: School Closures and Work Hours During the COVID-19 Pandemic in Austria

This article explores the gendered impact of school closures on paid work hours during the COVID-19 pandemic in Austria. Using data from the Austrian Corona Panel Project (ACPP) covering generalized school closures from March 2020 to April 2021, the study examines adjustments in work hours by gender and parental status. The descriptive data show general reductions in work time, especially in the first months. From July 2020 onward, however, mothers reduced work hours more than fathers when schools were closed – and they increased time spent on childcare, while fathers reduced theirs. Using OLS and fixed effects models, the study confirms that mothers reduced their work hours during school closures more than any other group. In contrast, fathers reduced their work hours the least – even less than individuals without children. Finally, there is some evidence that school closures capture policy stringency in high-incidence phases of the COVID-19 pandemic. HIGHLIGHTS In Austria, mothers reduced paid work time more than fathers in response to pandemic school closures. In contrast, fathers reduced their work time even less than individuals without children. School closures thus triggered a gendered labor market response among parents. The additional unpaid care work burden on women is a potential mechanism for these effects. COVID-19 policy responses may have exacerbated existing gender differences in the labor market.


INTRODUCTION
The COVID-19 pandemic and the subsequent increase in caregiving needs due to school and daycare closures threatened to reinforce traditional gender roles between men and women regarding the division of paid and unpaid labor (Heintz, Staab, and Turquet 2021).The initial European response to the COVID-19 pandemic was an aggressive public health policy in the form of school closures, and Austria presents a particularly salient case.It can be assumed that this served to reduce uncertainty and thus suppressed a labor supply response of households.However, infection risks for the elderly inhibited alternative informal childcare arrangements (for instance, through grandparents).Combined with a higher number of school closures this may well have precipitated decisions regarding labor supply adjustment.Furthermore, intrahousehold bargaining may lead to symmetric or asymmetric reductions in labor supply for women and men.Whether parents reduced work time more than childless workers, and whether these reductions differ by gender are the research questions we aim to answer in this article.
We investigate the effect of school closures over the course of the COVID-19 pandemic on work time by parental status and gender in Austria.Austria is a particularly interesting case, because social norms around mothering are rather conservative, marked by tenuous labor market attachment of women even before the onset of the crisis with high female part-time rates 1 and one of the highest motherhood pay penalties in comparison to other European countries and the United States (Kleven et al. 2019).Furthermore, Austria was affected relatively early and severely by the pandemic, and policy measures aiming to contain the spread of COVID-19 centered mainly on school closures, along with stores and restaurants.This led to a fairly volatile policy of repeated closures and re-openings of schools.Austria reacted with a strict lockdown to its early first wave after "seeding" the virus across Europe (NYT 2020), closing schools and daycare facilities for children of all ages, non-essential stores, and restaurants.Even though the second wave in November 2020 was more severe than the first wave in terms of the number of cases, the political response was more irresolute.Offices continued to operate virtually unrestricted, 2 while schools and daycares closed again, then swiftly opened in the beginning of December, only to close again after the Christmas break.This back-and-forth, while possibly dubious from a policymaking point of view, introduces a welcome variance from a research perspective.This article thus investigates the effects of the radical rationing of childcare availability through school and daycare closures, combined with blocked informal childcare arrangements, on women's and men's labor supply.The literature is inconclusive whether the pandemic led to a stronger decrease of employment or paid work hours of women relative to men in high-income regions like the United States and Europe (Amuedo-Dorantes, Kaushal, and Muchow 2020; Alon et al. 2022).The feminist economic literature explores the channels giving rise to this potential difference in parental labor market effects of the pandemic, focusing in particular on unpaid household and care work and the double-edged sword of working from home (Kabeer, Razavi, and van der Meulen Rodgers 2021;Corsi and ˙Ilkkaracan 2022).This article expands the literature by using high-quality, high-frequency representative survey data from the Austrian Corona Panel Project (ACPP; Kittel et al. 2021), which comprises twentytwo waves covering the period from March 2020 to April 2021.The ACPP includes a wide variety of questions relating to pandemic life, including information on attitudes and behaviors.Our main variables of interest are paid weekly work hours, data on policy measures that concern work hours, and socioeconomic variables, all of which are included in all waves and therefore offer detailed insight.We augment this dataset with data on school closures from the Organisation for Economic Co-operation and Development (OECD 2021), as well as data on different COVID-19 containment measures from the European Centre for Disease Prevention and Control (2022).
Our descriptive results indicate that -after an initial shock at the beginning of the pandemic in March 2020 -work hours stabilized.However, they also show that the difference in work time between mothers and fathers increased during school closures, while mothers increased time spent on childcare in these times and fathers decreased it.OLS and individual-level fixed effects regression models confirm this finding, showing that on average mothers reduced their paid working time by economically and statistically significant 31.5 percent or 8.1 h per week during school closures.In contrast, fathers reduced their work hours by merely 14.4 percent during times of school closures -less than individuals without children.
Furthermore, our findings support the hypothesis that school closures capture the intensity of the COVID-19 crisis and that it might in fact have been the main policy variable in explaining changes in labor market outcomes due to containment measures, since school closures also affect the work time of childless women and men when controlling for the closure of non-essential stores.Splitting school closures into a dummy variable with two values, one for under 14-year-olds and one for over 14-yearolds, supports this interpretation: School closures only for over 14-year-olds affect the work time of every group less than closures for all children, so more intensive phases of the pandemic.However, school closures for all, including children under the age of 14 again affect mothers the most.This study's contribution to the existing literature is thus twofold: First, we investigate the change in work time due to school closures from the beginning of the pandemic up to April 2021, covering the entire period of the pandemic with generalized school closures in Austria.This allows us to study medium-run effects of increased childcare burden on parental labor market outcomes during the COVID-19 crisis.Second, we show that although policy measures such as short-time work schemes likely stabilized employment and work hours in Austria, school closures still appear to reduce mothers' work hours to a much larger degree than fathers'.We thus confirm that school closures invoke a gendered labor market response for parents.In addition, we provide preliminary evidence that school closures may be a useful measure for the tightness of policy response to COVID-19.

LITERATURE REVIEW
This article speaks to three distinct strands of empirical literature.First, an extensive literature investigates the effect of childbearing on women's (and to a small extent, men's) labor market outcomes.Second, another empirical strand estimates the effects of price and quantity variations in the childcare market on parents' labor market outcomes.School and daycare closures arguably form an extreme case of rationing childcare availability.Third, an extensive and rapidly growing literature investigates the effects of the COVID-19 pandemic on paid and unpaid work of women and men.
We embed our research, first, in an extensive empirical literature, which investigates the effect of childbirth -and thus the need for additional household production in the form of childcare work -on women's labor market outcomes.Labor market participation rates and work hours are particularly relevant for this article.Both drop significantly after the birth of the first child for mothers (Jacobsen, Pearce, and Rosenbloom 1999;Cristia 2008).The effect of children -and much less childcare availability -on men's labor market outcomes is less well-studied (Lundberg 2005); notable exceptions are Shelly Lundberg and Elaina Rose (2002), Shelly Lundberg (2005), and Hyung-Jai Choi, Jutta M. Joesch, and Shelly Lundberg (2008).These researchers mostly find positive effects -in contrast to the effect for women, having children raises men's work hours.The large literature on motherhood pay penalties shows that these effects extend to income (Kleven et al. 2019).This literature thus suggests that an increased need for household work in childcare may increase intrahousehold specialization.
Second, an extensive empirical literature investigates the effect of childcare costs on women's labor supply decisions directly (for an overview, see Del Boca [2015]).Estimating the participation decision using reducedform or structural models, it finds a robust negative relationship between childcare costs and maternal labor force participation.However, this assumes not only the availability of informal childcare arrangements for every child (Heckman 1974), but also sufficient supply of formal childcare (Powell 2002;Lokshin 2004;Fitzpatrick 2010Fitzpatrick , 2012)).
Third and most relevant for this article is the empirical literature on the gender differences in paid work resulting from school closures during the pandemic.As for most COVID-19-related topics, this literature expanded and continuous to grow rapidly.Some very early articles investigate research questions similar to those in our article for the United States, with mixed results.Using a discrete choice participation equation, Felipe Lozano Rojas et al. (2020) find no evidence that school closures affected unemployment based on weekly unemployment claims data in a time fixed effects model.Misty L. Heggeness (2020) finds effects on care leave (but not unemployment) of women (but not men) using the Current Population Survey (CPS) in a difference-in-differences (DiD) model for an "early closing" and a "late closing" group of states.Caitlyn Collins et al. (2021a) estimate a logistic regression on a unique data set of elementary school closures and find that mothers reduced their labor market participation more than fathers, with an even larger gap in regions where teaching was largely remote.
Another set of articles uses work hours, a more granular dependent variable than the binary labor market participation or unemployment.This research indicates that in the United States especially mothers of young children seem to have reduced their work time due to limited availability of childcare and schools.Catalina Amuedo-Dorantes, Neeraj Kaushal, and Ashley N. Muchow (2020) use CPS data and a DiD model and find that school closures in the United States reduced parents' weekly work hours between 11 and 15 percent.This effect is larger for young mothers.Charlene Marie Kalenkoski and Sabrina Wulff Pabilonia (2022) confirm this result for the self-employed, using the same data and method.Scott Barkowski, Joanne Song McLaughlin, and Yinlin Da (2021) find no evidence for a gender difference in the reduction of work hours, although their data show that parents of children younger than 13 reduced their working hours more than parents with older children.Collins et al. (2021b) show that in particular mothers of young children reduced their weekly work hours more than fathers.Gema Zamarro and María J. Prados (2021) find that women reduced work hours more than men and that they were more likely to have become sole caregivers in a representative COVID-19 survey for the United States.
Whether these findings transfer to cases in continental Europe is not clear a priori.Sebastian Dullien and Bettina Kohlrausch (2021) find that during the first COVID-19 wave in Germany, a loss of 1.1 percent of aggregate work hours can be attributed to school closures at the macro level.However, in their two-wave survey 20 percent of parents of school-age children state that they reduced their work time due to childcare duties.Titan Alon et al. (2022) provide evidence from six different countries, including Germany and the Netherlands, that -in contrast to previous crises -women were more affected than men by the COVID-19 recession.
They establish a negative correlation between the severity of school closure measures and the change in overall labor supply and provide evidence for larger gender gaps in work hours for parents of school-age children.For Turkey, İpek İlkkaracan and Emel Memiş (2021) find that the gender gap in paid work hours narrowed during the initial phase of the pandemic, yet women's unpaid work increased more than men's, especially for those with school-age children.
A rich feminist economic literature explores the channels giving rise to this potential difference in parental labor market effects of the pandemic, focusing in particular on unpaid household and care work and the doubleedged sword of working from home.It documents that the shock to time required for unpaid work affected women more than men (Kabeer, Razavi, and van der Meulen Rodgers 2021).Men did increase their unpaid work time, which reduced the relative childcare differential to women due to men's low base level.But, women also increased their unpaid work time from their already-high base level, often leaving the absolute gap largely unchanged (Craig and Churchill 2021;Croda and Grossbard 2021;Heintz, Staab, and Turquet 2021;˙Ilkkaracan and Memiş 2021).At the same time, work-from-home directives rendered home schooling by parents a theoretical possibility.While working from home increases flexibility and the possibility of reconciling paid and unpaid work, the net effect on women's labor market outcomes is unclear, since it may also limit their career opportunities and entrench a gendered division of labor (Seiz 2021;Corsi and ˙Ilkkaracan 2022;Johnson 2022).In Austrian households where mothers and fathers work the same number of paid weekly hours, mothers provided more unpaid care hours during the pandemic (Berghammer 2022).Additionally, children were more likely to remain in the room where mothers worked (Derndorfer et al. 2021).Against this background, it is crucial to understand that policies are not per se gender neutral, and an extensive feminist economic literature debates these effects for the COVID-19 pandemic (Zuazu 2022).
The literature is thus inconclusive whether the pandemic led to a stronger decrease of paid work time for women relative to men.This article adds evidence from Austria using high-quality, high-frequency representative survey data, showing that school closures affected mothers' paid work time more than any others group, while fathers reduced theirs the least.

THE COVID-19 PANDEMIC IN AUSTRIA
Austria is a particularly interesting case for investigating the links between school and daycare closures and labor supply, since its policy response has been highly volatile.Austria was one of the earliest affected countries, as a lax initial policy response to COVID-19 cases in the winter tourism town Ischgl in Tyrol, a western province of Austria, likely contributed to spreading of the virus throughout Europe (Correa-Martínez et al. 2020;Kreidl et al. 2020).Policy measures were then tightened substantially, and a hard lockdown consisting, among other measures, of school and shop closures reduced the seven-day rolling average of registered daily new infections throughout April and May, with effects lasting roughly until August (see Figure 1).
However, with very few limitations in place until after the fall vacation week in early November 2020, Austria experienced an unprecedented surge of infections in November and December, with new infections reaching the highest level world-wide.A second lockdown in December 2020 and the beginning of January 2021 brought new COVID-19 cases back below the EU average.Yet, winter sport facilities -in particular ski resorts -remained open throughout, as did stores, and new infections were once again rising steeply in Austria until approximately April 2021.Public support for policies during the pandemic has since waned, with less than half of the surveyed individuals approving of the government's handling; in 2021, criticism was split roughly equally between those considering the measures to be too lax and those deeming them too strict (Profil 2021).
School closures played a key role in restricting social contact in the Austrian policy mix. 3 After the first case of a COVID-19 infection in Austria was documented on February 25, 2020, schools and daycare facilities, as  Finally, in comparison to other high-income countries, Austria experienced an especially deep recession.This is not only due to its dependency on tourism but also because of the number of days in which non-essential stores were closed, which was high due to the strict lockdown in the first wave and the large number of COVID-19 cases in the second (Huber and Picek 2021).Although stimulus was weak in international comparison (especially in comparison to the United States; OECD 2021), it was sizeable for European standards (Ederer 2021).Austria passed two Corona aid packages in quick succession in March 2020 (amounting to four billion Euros and 38 billion Euros, which was roughly 10 percent of the 2020 GDP), and another stimulus package in September 2020 (worth 13 billion Euros or 3.5 percent of GDP).For employees, support included payments to businesses for funding short-time work, lowering the tax rate in the first income tax bracket, one-off increases in unemployment benefits, and two additional one-off payments in child support (Budgetdienst 2020).These aid packages and labor market measures led to available income dropping less than GDP ( − 1.9 versus − 5.5 percent, respectively; Statistik Austria 2021) relative to 2019. 4  The COVID-19 aid packages contained concrete measures affecting work time.Especially short-time work was heavily used in Austria, which saved up to 1.2 million jobs in 2020 (AMS 2021).By December 2020, Austria had spent 5.5 billion euros, 1.5 percent of its GDP, on short-time work.While women make up almost half of the employees on short-time work schemes, they receive only about 40 percent of the payouts (Hehenberger and Pixer 2021).Furthermore, Austria extended paid leave for parents during the pandemic.Each parent was entitled to one week of paid leave due to school closures, one week of care leave for sick children, and up to four weeks of "special care leave" per (school) year. 5

DATA
In order to investigate the effects of school and daycare closures on labor market outcomes, we augment the panel data of the Austrian Corona Panel Project (ACPP; Kittel et al. 2021) by data on school closures from the OECD (2021), as well as data on COVID-19 containment measures from the European Center for Disease Prevention and Control (2022).The ACPP contains twenty-two waves between March 2020 and April 2021, which were conducted weekly from late March to June 2020, and monthly thereafter.It is representative for Austria based on age, education, gender, and region.
The ACPP covers a broad set of questions relating to pandemic life, including information on attitudes and behaviors.Every wave comprises roughly 1,500 individuals in an unbalanced panel. 6For the purpose of this article, we use paid weekly work hours as the dependent variable.Our controls include socioeconomic variables (age, children, education, gender, household size, income, and migration background) as well as work time variables (the ability to work from home, furlough, and shorttime work).We supplement the ACPP with data on school closures -our main explanatory variable -and COVID-19 containment measures such as the closures of non-essential shops and restaurants (European Centre for Disease Prevention and Control 2022).
We restrict our data to individuals between 18 and 65 years of age.Since paid work hours are only surveyed for employees and/or self-employed, the estimations further exclude the unemployed, pensioners, students, persons in military or community service, parental or educational leave, and those not in the labor force.Weekly work hours are reported as current average paid hours worked per week including overtime.They are top-coded at eighty hours per week.This yields a total of 17,892 observations of 1,702 individuals.School closures are our main explanatory variable.Using data from the OECD (2021), we define both school closures and parental status for children age 14 or younger, since we expect older children to require less supervision. 7For our controls, we use a categorical group dummy based on gender and whether a child under the age of 14 lives in the same household, which thus differentiates between mothers, fathers, and women and men without children.We consider three age cohorts (18 to 30 years, 31 to 50 years, and 51 to 65 years) and three education categories (less than a secondary degree, secondary school degree or equivalent, and tertiary degree).Migration background is defined as the person themselves or at least one parent, having been born outside of Austria.Low household income is a dummy variable with a cutoff at 2,700 Euros of net monthly household income.In order to control for other factors which may affect work time in the estimation of the intensive margin, we leverage the ACPP's many pandemic-and work-related questions.They permit us to include two dummy variables for the employees' states during the respective waves: a variable capturing whether the worker reported being furloughed, that is, involuntarily having to reduce vacation time; and a variable whether the survey respondent was in short-time work, which was used extensively in Austria during the pandemic compared to Anglo-Saxon countries (Adams-Prassl et al. 2020;AMS 2021).We also control for the ability to work from home by including a person-level dummy variable, which indicates whether the individual worked from home at any point during the pandemic.Finally, we attempt to control for unobservable pandemic-related factors by including a dummy variable for waves in which non-essential shops and/or restaurants were closed (European Centre for Disease Prevention and Control 2022).This variable comprises three values that are defined as (0) no measures, (1) some non-essential shops or restaurants are closed, (2) some non-essential shops and restaurants are closed, and (3) all non-essential shops and restaurants are closed. 8 Table 1 provides an overview of the summary statistics of our variables for the full sample, as well as for the subsamples of mothers, fathers, and women and men without children.Mothers and fathers make up 13.3 and 15.9 percent of the full sample, respectively.Paid weekly work hours between March 2020 and April 2021 are on average highest for fathers (at about 35.9 h per week), and lowest for mothers (25.7 h).Median weekly work hours range from twenty-five for mothers to forty for both fathers and childless men.As is to be expected, the majority of parents are between 31 and 50 years of age, while individuals without children are more evenly distributed between 31 and 65 years of age.62 percent of the sample's individuals hold less than a secondary degree; this share is higher for men and lower for women.Mothers hold the highest share of university degrees in the sample at 21.8 percent.About 30 percent of individuals in the sample have a migration background; parents are more likely to have a migration background, at about 36 percent for fathers and 38 percent for mothers.The average household size is around three persons, with parents unsurprisingly living in larger households.The income threshold roughly splits the sample in half, except for childless women, who are more likely to live in a low-income household.The ability to work from home is coded as the realized ability to work from home -so if individuals have ever worked from home, we assume that they have the ability to work from home.These numbers are larger for women with around 60 percent, than for men, with around 55 percent.Between 15 and 19 percent report being on short-time work, and around 6 percent being furloughed.
At the bottom of Table 1, we include two sets of descriptive data that are relevant to our argument, but cannot be used in the econometric analysis due to data restrictions.First, mean weekly work hours during the observation period dropped from the retrospectively surveyed prepandemic period (February 2020) for all groups.Mothers' average paid work hours fell the most during the pandemic, by eight hours.While the gap between men with and without children remained virtually unchanged, the gap widened between the two groups of women.As noted above, these data cannot be used in the econometric analysis below due to sampling issues.Second, out of the twenty-two ACPP waves in our sample, five contain time-use data including time spent on childcare in a normal day.On average, mothers spend twice as much time on childcare than fathers, at 7.2 h per day.Interestingly, this relation roughly holds for individuals without children, where women report spending more than twice as many hours on childcare as men (0.7 versus 0.3 h per day).These data cannot be used in the econometric analysis due to the limited number of observations.
Figure 3 shows the density of the average paid weekly work hours for women and men between March 2020 and April 2021.It provides visual evidence of bunching, which is particularly noticeable at full time work of forty hours for women and especially for men.Furthermore, women are more likely than men to report working part time.
Figure 4 shows the monthly difference in average working hours of parents by gender, color-coded by the state of school closures over time, showing the rate at which the survey waves were conducted (more frequently in the beginning and monthly after July 2020). 9Open schools for children under 14 appear to be associated with a smaller difference in work hours between fathers and mothers.This is especially salient for the second half of 2020, after the initial shock of the pandemic in March and April 2020 had worn off over the summer months. 10Contrasting this evidence against the less fine-grained changes in time spent on unpaid childcare shown in Figure A2 in the Appendix -which shows that during school closures women on average increased their hours spent caring for children while men decreased theirs -suggests that one channel through which school closures affect work time may be through additional time spent on childcare.

EMPIRICAL STRATEGY
We investigate the gendered link between work time and school closures through a multivariate analysis using the following model, for which we first estimate the effect of school closures on weekly work hours in an OLS regression, and second incorporate individual-level fixed effects: where the dependent variable is the inverse hyperbolic sine (IHS)transformed paid weekly work hours WH it of individual i in wave t.This transformation of our dependent variable, arcsinh(WH ) = log(WH + √ W H 2 + 1), approximates the natural logarithm and thus provides a semi-elasticity for the change in work hours while preserving zero-valued observations (Bellemare and Wichman 2020).These are especially relevant for our analysis since some parents might reduce their work hours to zero during school closures.
Regarding the explanatory variables, SC t denotes school closures and Group i is a categorical four-group variable differentiating between mothers and fathers (with children younger than 14 in the household) and childless women and men.SC t × Group i is the interaction of school closures and the respective group.The control vector X i contains our time-invariant demographic control variables age, education, income, household size, and migration background.The control vector Z it varies over time and allows us to control for pandemic-related factors through dummy variables for non-essential shop and restaurant closures.We are also able to include variables which comprise measures that directly affect work time in our control vector Z it .That is, whether the individual was in short-time work or furloughed in each wave and whether the individual has the ability to work from home.
To estimate the gender gap in work hours more precisely and to leverage the panel structure of the ACPP data, we not only use OLS, but also a person-level fixed effects estimation.This permits us to study the adjustments in weekly work hours of an individual during the COVID-19 pandemic, while controlling for individual characteristics that remained unchanged over its course by demeaning the variables using a within-estimator.Thus, observed and unobserved time-invariant variables like gender or parental status, but also for example constant personal preferences, are subsumed in the individual-level fixed effects.However, through interaction terms we can still differentiate between the four groups.
Our identification comes from the repeated opening and closing of schools, and therefore from a variation in time.In order to still be able to leverage this variation, we do not include wave fixed effects.A Hausman test suggests that a fixed effects model is the correct model choice for our specification. 11

RESULTS
Table 2 shows our main results with the OLS estimation in column (1) and the individual-level fixed effects estimation in column (2). 12The OLS results indicate that school closures are generally negatively correlated with paid work time, with the reduction amounting to about 17.3 percent.Furthermore, taking the interaction with school closures into account, mothers additionally reduce weekly work hours more than all other groups, by about 16.8 percent, relative to the base group of childless men.This is especially notable when compared to fathers, who reduce work time even less than all other groups.The OLS model thus points to a substantially gendered effect to school closures among parents.
Our work time controls short-time work and furlough capture reductions in weekly work hours, as intended.The ability to work from home is insignificant; this is not surprising since we would expect working from home only to reduce work hours if it is coupled with more unpaid work in childcare.The required closure of non-essential shops and restaurants does not reduce work time statistically significantly.This indicates that our school closures variable captures policy responses to the pandemic very well.The estimated coefficients of the control variables age, education, household size, migration background, and income are shown in Table A1 in the Appendix.
The results of the fixed effects model in column (2) confirm that all groups reduce their paid work time in periods when schools are closed.Our main finding is again that mothers reduce their work time the most out of all four groups during school closures, while work time reduction for fathers is the smallest.Mothers on average decrease their paid weekly work hours by economically and statistically significant 31.5 percent in periods with school closures.Since mothers on average work roughly 25.7 h per week, this amounts to a reduction of about 8.1 weekly work hours.Fathers reduce their paid work hours by about 14.4 percent.Since childless men and women also show lower work hours in these waves -in fact even larger reductions than fathers -this indicates again that the school closures variable also captures indirect policy effects. 13 Additional evidence for the hypothesis that school closures may be better at flagging intensive phases of the COVID-19 pandemic than our control variable for non-essential shops and restaurant closures is that none of the groups reduced their work hours statistically significantly during periods when non-essential shops and/or restaurants were closed.However, shorttime work and being furloughed, both policy instruments designed to reduce working hours in times of low economic activity and in order to reduce unemployment as discussed earlier, have the expected negative signs and explain some of the variation in work time.The ability to work from home is not contained in the fixed effects analysis since it is timeinvariant and its effect is thus subsumed in the fixed effects.Taken together with the descriptive evidence on the difference in time spent on childcare, these findings provide indications for the channels through which school closure may be linked to a reduction in work hours.Concretely, it appears that this is not due to the ability work from home by itself, but rather whether individuals spent more time on unpaid labor like childcare.
Our findings fit well with the existing literature.In the short run, Collins et al. (2021a) find that mothers of younger children reduce their work hours by 1.5 to 2 h per week, while they do not find significant reductions for fathers.While Amuedo-Dorantes, Kaushal, and Muchow (2020) report work time reductions for both mothers and fathers from the beginning of the pandemic to May 2020, they show that mothers reduced their work hours a lot more than fathers (by 30 versus 11 percent, respectively).Our data covers a longer time period and our findings corroborate these results for Austria.
Yet, naturally, our results should be interpreted with caution.First, we are unable to account for the effect of expectations in our estimates, although they likely play an important mediating role in determining work time reductions.Given the volatility of policy measures in Austria, which swung between hard lockdowns and almost complete openings, expectations may well have been unstable.If high uncertainty leads to a weaker reduction of work time, then our estimated effects are likely to be conservative.Second, the short-time work schemes and the special care leave implemented in the COVID-19 pandemic in Austria might have stabilized employment and work hours despite additional childcare duties for parents.This would also suggest that our estimates for the effect of school closures are conservative.

ROBUSTNESS CHECKS
We check the robustness of our results by, first, differentiating our main explanatory variable school closures into schools closed for all children and schools closed only for over 14-year-olds, in order to explore whether the school closure variable may capture the strictness of containment measures.And second, we investigate the effect of school closures on labor supply at the extensive margin, that is labor force participation and employment (which amounts to reducing working hours to zero and dropping out of the labor force).To investigate in more detail whether school closures in fact capture the stringency of COVID-19 policy measures, we add a third value to our previously binary school closures dummy variable, namely school closures for over 14-year-olds (OECD 2021).The school closures variable may now take the values 0 (open for all), 1 (open for under 14-year-olds), or 2 (closed for all).School closures for over 14-year-olds were more extensive than those for under 14-year-olds.Our hypothesis is that parents of children under 14 years old should not be as affected by school closures for older children.
The results in Table 3 show that school closures for all children do in fact impact every group more severely than when schools were only closed for children above 14.In line with our hypothesis, we find that when schools are open for children under 14 but closed for older children, mothers' work hours are not affected.However, when schools are closed for all, including children under 14, they reduce their work time again by around 31.7 percent.Fathers' work hours decrease by 4.7 percent when schools are open for under 14-year-olds and 15.4 percent when they are closed for all.For both childless women and men, school closures for all and school closures only for over 14-year-olds have a statistically significant impact, which supports our hypothesis that the school closures variable may capture policy stringency in high incidence phases of the COVID-19 pandemic. 14 Finally, the coefficients for our control variables are qualitatively robust to this change in our main explanatory variable (see Table A2 in the Appendix for full results including control variables).Lastly, we examine the effect of school closures on the extensive margin of labor supply.Concretely, we specify the reduced-form logit model: where y is either labor force participation of individual i or whether they are employed or not (the difference being, of course, that the unemployed are included in the labor force but not in employment).The explanatory variables SC t , Group i , SC t × Group i and the control vectors X i and Z it are the same as in Equation (1).For our estimations of the extensive margin, we exclude persons under 18 and over 65 years of age (as in the main analysis), as well as persons in military or community service.Our dependent variable labor force participation is a dummy variable indicating whether the person reports to be working for pay (employed or self-employed) or to be actively looking for a job during the respective wave.When our dependent variable is employment, the unemployed are excluded.In the full sample, labor force participation is 87 percent and the employment rate is 76 percent between March 2020 and April 2021.Both are lower among women than among men, but especially so for women without children under the age of 14 (81 and 68 percent respectively).
The results in Table 4 show that women's labor force participation and employment rate is lower than men's also when controlling for covariates, and that there is an additional negative effect for mothers.Age, education, migration background, and income all have the expected signs; on average, a larger household leads to a lower probability of being in the labor force or employed.We do not observe statistically significant effects for nonessential shop and school closures or the interaction term of school closures with any of the groups on the probability of being in the labor force or being employed.We thus do not find an additional effect of school closures on the labor force participation of mothers.One possible interpretation of this finding is that while school closures did lead to mothers' reducing their work time, COVID-19 policy measures in Austria succeeded in keeping them in the labor force and in employment.

CONCLUSION
This article investigates the impact of school closures on paid work time during the COVID-19 pandemic in Austria.We use high-frequency survey data from the ACPP from March 2020 to April 2021 and data on school closures by the OECD (2021) to test for gender differences in the effect of school closures for mothers and fathers of children under 14 -that is, children with high care needs -on paid weekly work hours.
We find descriptive evidence that both women and men reduced their paid work time due to the COVID-19 crisis, especially in the first months of the pandemic in spring 2020.However, after work hours stabilized around July 2020, mothers reduced work hours more than fathers in periods with mandatory school closures.At the same time, mothers increased their time spent on unpaid care work while fathers decreased theirs.
This descriptive finding is corroborated by both OLS and individuallevel fixed effects models.The OLS estimation controls for a host of socioeconomic factors including age, education, migration status, and household size; other factors which may potentially reduce work time in the pandemic, that is, short-time work, furloughs, and the ability to work from home; as well as non-essential store and restaurant closures, which should account for the stringency of COVID-19 policy measures and thus the intensity of the pandemic.
The fixed effects models control for work-time variables and nonessential shops and restaurant closures in addition to the fixed effects parameters capturing all time-invariant characteristics of individuals, and show that women in general reduce their work hours more than men.This effect is predominantly driven by mothers, whose weekly work hours fell by an economically and statistically significant 31.5 percent on average during periods with school closures, or approximately 8.1 h.In contrast, fathers reduced their work hours by merely 14.4 percent during times of school closures -even less than individuals without children.We therefore confirm that school closures prompt a gendered labor market response among parents, which we surmise might have been channeled through increased unpaid work in the form of childcare.
However, we also find an effect of school closures on the work time of childless women and men, which leads us to conjecture that school closures may in fact capture indirect policy effects and thus represent the tightness of COVID-19 measures better than shop and restaurant closures.Splitting school closures into a dummy variable with two values, one for all children and one for over 14-year-olds, supports this hypothesis: School closures for over 14-year-olds affect all groups less, whereas school closures for all children including under 14-year-olds mainly affect their mothers.
Estimating a logit model for labor force participation and employment (that is, the extreme form of hours reduced to zero and a change in labor market status) shows robust gender and parental effects, but fails to confirm the effect of school closures.This may be due to pandemic policy in Austria, which was aimed at maintaining employment mainly through short-time work.
Our results thus strongly suggest that the additional childcare responsibilities impacted paid work time differently by gender.Since working from home could not be squared with monitoring children in the medium run (Derndorfer et al. 2021), mothers appear to have reduced their work time, while fathers' work time was affected to a much smaller degree after the initial shock phase.Especially in the medium run, the COVID-19 pandemic thus appears to have reinforced the traditional division of paid and unpaid labor within households in Austria.This development, if diagnosed correctly, may have ramifications for gender differences in economic outcomes, ranging from the gender pay gap to the gender pension gap and to the representation of women in top positions, as the lower work time of women relative to men is an important explanatory factor for all of these economic disadvantages for women.Our results also provide some indication that the COVID-19 policy response in Austria may have exacerbated these trends -by using school closures as an indirect way of promoting limited workplace closures, policymakers forced women to stay at home to care for their children.Policies based on the evidence presented here, in contrast, would focus on counteracting these trends, in order to mitigate the well-documented negative long-run effects of weaker labor market attachment of women.Chief among such equity promoting policies is the restoration of reliable and safe childcare through schools and daycare facilities.
Since we are at the beginning of understanding the COVID-19 pandemic and its economic impacts, many questions remain open.Investigating the concrete mechanisms through which parents form choices around the allocation of work time in the household, especially with regard to expectations for the development of the pandemic and school closures.Second, placing our findings in an international comparison would be a natural avenue for future research, to answer whether work time effects of school closures extend to countries beyond those covered by the literature so far.Especially in a European context, it would be interesting to ask whether the Austrian policy response to the COVID-19 pandemic and its gendered impact on work time was unique.Finally, investigating the distributional effects of these work time choices through formal modeling may yield interesting insights into the long-run consequences of the COVID-19 pandemic.

Figure 1
Figure 1 Daily confirmed cases in Austria, Europe, the United States, and the world (seven-day rolling average) from March 2020 to March 2021 Source: Center for Systems Science and Engineering (CSSE) at Johns Hopkins University.

Figure 2
Figure 2 School and shop closures during the COVID-19 pandemic by ACPP waves Note: Date shown on the x-axis is the end date of the respective wave.Sources: OECD (2021); European Center for Disease Prevention and Control (2022).
Figure 2 summarizes the highly volatile profile of repeated opening and closing of Austrian schools.

Figure 3
Figure 3 Density of paid weekly hours worked by gender between March 2020 and April 2021 Sources: Own calculations, ACPP, Kittel et al. (2021).

Figure 4
Figure 4 Difference in paid work hours between fathers and mothers by month, March 2020 to April 2021 Sources: Own calculations, ACPP, Kittel et al. (2021).

Figure A2
Figure A2 Mean number of hours spent on childcare by mothers and fathers in the ACPP (weighted) Sources: Own calculations, ACPP, Kittel et al. (2021).

Table 1
Summary statistics for different groups in the ACPP betweenMarch 2020  and April 2021 (weighted)

Childcare per day (Mean, in hours)
This table contains data on individuals that were employed throughout March 2020 to April 2021.The data contained in the last five rows cannot be used in the econometric analysis due to data restrictions.

Table 2
School closures and weekly paid work hours of mothers, fathers, and women and men without children This table shows the results of an ordinary least squares and fixed effects regression on IHStransformed paid weekly work hours.Standard errors in brackets, clustered at the group-level.Base categories for the dummy variables are schools open, men without children, all stores and restaurants open, not being on short-time work, not being furloughed, not being able to work from home, the age group 18-30, less than high school education, monthly household income above 2,700e, and no migration background.

Table 3
School closures for different age-groups and weekly paid work hours

Table 3
Continued This table shows the results of an OLS and fixed effects regression on IHS-transformed paid weekly work hours.Standard errors in brackets, clustered at the group-level.Base categories for the dummy variables are schools open, men without children, all stores and restaurants open, not being on short-time work, not being furloughed, not being able to work from home, the age group 18-30, less than high school education, monthly household income above 2,700e, and no migration background.* * * , * * , * denote significance at the 1, 5, and 10 percent levels, respectively.

Table 4
The extensive margin of effect of school closures on labor force participation

Table 4
Kittel et al. (2021) shows the results of logistic regressions with labor force participation and employment as the dependent variables.Robust standard errors in brackets.Base categories for the dummy variables are schools open, men without children, all stores and restaurants open, not being on short-time work, not being furloughed, not being able to work from home, the age group 18-30, less than high school education, monthly household income above 2,700e, no migration background.*** , * * , * denote significance at the 1, 5, and 10 percent levels, respectively.Sources: Own calculations, data: ACPP,Kittel et al. (2021). Notes: PhD candidate at the Department of Economics at the Free University of Berlin.She earned an MSc from the University of Economics in Vienna and has been part of the graduate school "Political Economy of Inequality" at the Institute for Socioeconomics of the University of Duisburg-Essen.Gema and María J. Prados.2021."Gender Differences in Couples' Division of Childcare, Work and Mental Health During COVID-19."Review of Economics of the Household 19(1): 11-40.Zuazu, Izaskun.2022."Caring the Care Sector: Contributions of Feminist Macroeconomics in the Post-COVID-19 Era." Mimeo, 1-18.
Miriam Rehm, PhD is Professor of Socioeconomics at the University Duisburg-Essen with a focus on empirical inequality research.ARTICLE Zamarro,

Table A1
Effect of school closures on paid weekly work hours (main results full table) (Continued).

Table A1 Continued
Kittel et al. (2021) results of an OLS and fixed effects regression on IHS-transformed paid weekly work hours.Standard errors in brackets, clustered at the group-level.Base categories for the dummy variables are schools open, men without children, all stores and restaurants open, not being on short-time work, not being furloughed, not being able to work from home, the age group 18-30, less than high school education, monthly household income above 2,700e, and no migration background.*** , * * , * denote significance at the 1, 5, and 10 percent levels, respectively.Sources: Own calculations, data: ACPP,Kittel et al. (2021). Notes:

Table A2
Effects of school closures for different age-groups on the weekly paid work hours of different groups (robustness check full table) (Continued).

Table A2 Continued
Kittel et al. (2021) results of an OLS and fixed effects regression on IHS-transformed paid weekly work hours.Standard errors in brackets, clustered at the group-level.Base categories for the dummy variables are schools open, men without children, all stores and restaurants open, not being on short-time work, not being furloughed, not being able to work from home, the age group 18-30, less than high school education, monthly household income above 2,700e, and no migration background.*** , * * , * denote significance at the 1, 5, and 10 percent levels, respectively.Sources: Own calculations, data: ACPP,Kittel et al. (2021). Notes: