DEPARTMENT OF ECONOMICS Uppsala University
Bachelor’s Thesis in Economics (15 ECTS) Semester and year: VT21
Author: Emma Genlott Supervisor: Arizo Karimi
The effects of school closures due to Covid-19 on parental labor supply: evidence from the United States
Date of submission: 2021 / 06 / 04
2
Acknowledgements
I want to sincerely express my gratitude towards my supervisor Arizo Karimi, Associate professor at the Economics Department at Uppsala University, who has provided continuous guidance and feedback throughout the process which this thesis has immensely benefited from. The skills which I have acquired through this work serve a great thanks to Arizos’
dedication and enthusiasm in the topic and the methodology, and all effort behind it.
3
Abstract
The spread of Covid-19 led to social restrictions of various kinds, of which closing schools was one. This paper studies the effect of school closures on parental labor supply. To this end, I use repeated cross-sectional data on households at the monthly level from the US Current Population Survey (CPS), and employ a difference-in-differences methodology where I compare the labor market outcomes for parents to school-aged children that require supervision with parents to slightly older children, before and after March 2020. The results show that there is a significant reduction in the labor supply of parents to younger children as a result of school closures, and that the effects are larger for mothers than for fathers.
Keywords: childcare, covid-19, school closures, parental labor supply, United States
4
Table of contents
1. Introduction 5
2. Previous literature 9
3. Conceptual background 11
4. Data 12
4.1. Effects on the labor market 12
4.2. Data on school closures 14
5. Research design 14
6. Results 17
6.1. Main findings 18
6.2. Findings in wage differences 22
6.3. Alternative variables 23
7. Discussion 24
Reference list 26
Appendix 28
5 1. Introduction
The spread of Covid-19 began in the end of 2019 and accelerated in the beginning of 2020. On the 11th of March 2020 the situation was declared a pandemic by the World Health Organization (WHO). To curb the spread of the virus, governments around the world implemented a variety of restrictions on social gatherings, stay-in-place orders, lockdowns and school closures. While these restrictions may be effective in preventing the transmission and spread of the virus in the population, they may impose challenges for businesses as well as individuals in both the immediate and long-term, not to mention the large economic consequences. The trade-off that countries have had to face between reducing the spread of the virus and economic outcomes has not been an obvious choice and countries have chosen different approaches. A heavily debated topic in the media has been the closing of schools - for several reasons. It has been questioned whether the school closures actually contribute to reducing the spread of the virus, since children are less likely to spread the disease (Davies et al., 2020). This was shown by Vlachos et al (2021) who concluded that closing schools has a small effect on the spread of the virus. They find that, among parents, exposure to open rather than closed schools resulted in a small increase in PCR-confirmed infections. Meanwhile, school closures have had large social and economic impacts on both parents and children affected. Having children switch from normal schooling to distance learning is expected to have economic implications in terms of parent’s labor supply, employment and family income, as a change of that magnitude requires responses from the household and the adoption might be comprehensive according to Alon et al. (2020). Nevertheless, schools did close in the US, as well as in big parts of the rest of the world, and children switched to distance learning, and all that amounts to.
6 The aim of this thesis is to analyze whether the labor supply of parents with school aged children have changed as an effect of the school closures by examining employment, income and weekly work hours.
During the past year the media has reported that job loss due to the restrictions on businesses and social gatherings have and will affect women relatively more than men (BBC, 2021; Fastighetsfolket, 2021; SVT 2021).1 This is supported by empirical evidence (UN, 2020;
Alon et al. 2020; Abi Adams-Prassl et al. 2020; Sevilla & Smith, 2020) and is partly explained by the fact that women tend to be concentrated in service sectors which have suffered more from the stay-at-home orders and other restrictions, as work cannot be done from home.
Another explanation is that women are observed to be the main caretakers of children in the household, and would therefore bear the brunt of having to care for children whose schools have closed - thus enhancing inequality on the labor market.
Blau & Robins (1991) conclude that young mothers tend to adjust their labor supply and childcare to economic and demographic changes to a higher degree than men. Blau &
Robins (1991) analyze fertility, employment and childcare decisions of young women over time, and state that young mothers are in a volatile phase when it comes to labor supply. The availability of babysitters and other arrangements of childcare has shown to affect how often young mothers enter and exit the labor market. Berger & Black (1992) show that young mothers who receive childcare support are more likely to be employed than mothers who do not receive child care subsidies. The analysis focused on young single mothers in Kentucky with low income and the effect on employment with regards to child care subsidies. The results
1https://www.bbc.com/worklife/article/20201021-why-this-recession-disproportionately-affects-women (BBC, 2020) accessed 2021-04-21 https://blog.dol.gov/2021/03/19/5-facts-about-the-state-of-the-gender-pay-gap (BLS, 2020) accessed 2021-04-15
https://fastighetsfolket.se/2021/03/04/kvinnor-med-arbetarjobb-drabbas-hardast-i-
pandemin/?fbclid=IwAR1VwArzMC8szlouc23TV1E4KZbrXjtxYMhC4ZyIkB4l_OfEP81E_mM2fs4 (Fastighetsfolket, 2020) accessed 2021-04-15
https://www.svt.se/nyheter/inrikes/coronaeffekten-dubbelt-sa-manga-kvinnor-blev-sjuka-pa-jobbet-
2020?fbclid=IwAR2DSlTh2AGU3yShm9ga_97Pqc3nmzFGJI9dQDAxixUoAKSR_KXwOGPdLKE (SVT, 2021) accessed 2021-04-17 https://www.un.org/sexualviolenceinconflict/wp-content/uploads/2020/06/report/policy-brief-the-impact-of-covid-19-on-women/policy- brief-the-impact-of-covid-19-on-women-en-1.pdf (UN, 2020) accessed 2021-04-17
7 that Berger & Black (1992) present are in line with the expectation of mothers being the main caretakers of children and therefore become more affected when schools are closed.
Additionally, Lefebvre and Merrigan (2008) find that the availability of childcare does have an impact on the labor supply of mothers with young children in Canada, using quasi-natural variation in childcare subsidies. Similarly, the employment and labor supply response to having children has been shown to be significantly negative for women (Angrist and Evans 1998;
Cruces and Galiani 2007). 2
With this in mind it’s therefore highly relevant to investigate if and how the pandemic has affected households as a whole in the form of labor supply, and the ratio between men and women’s labor supply.
To obtain causal estimates of the effect of school closures on parental labor supply, I make use of repeated cross-sectional micro-level data on households from the Current Population Survey (CPS), provided by the US Bureau of Labor Statistics (BLS), and exploit variation in exposure to school closures across households based on the age of children.
Specifically, I assume that children of younger ages require adult supervision, and parents of young children are thus more exposed to school closures compared to parents of older children.
I then compare the outcomes of households with younger and older children, before and after the schools closed in their respective states (March 2020 for all states in the analysis sample).
This thesis contributes to gender research on child care by using a large sample with the help of CPS and by focusing on the US on a state level. It contributes in the sense that eventual differences between states will become clear, thus illuminating impacts of the pandemic for the equality in household work between men and women with different
2The gender pay gap is well known, as women in the US tend to have less pay than men who conduct the same job (Bureau of Labour Statistics, 2020). Empirical evidence points to that apart from getting less pay than men, women conduct more unpaid work than men in the form of household duties and childcare (Alon et al. 2020; Abi Adams-Prassl et al. 2020).
8 socioeconomic status. The pandemic has not yet reached its end and experts claim that future pandemics are probable. As such, this thesis provides knowledge not only in the wake of Covid- 19, but also for future challenges. The results of this thesis can be used both for learning more about school closures and equality labor patterns in the US during a pandemic, but also as a backdrop for future actions concerning decisions of providing childcare packages or other methods for ensuring women’s continued labor supply in the pursuit of an equal labor and childcare choice within households.
I find that the school closures did have an impact on the labor supply as the employment rate was reduced at the time of the school closures. The impact on mothers is larger than on fathers as the effect on employment is more than twice as large. Looking at the total income for families within the treatment group, there is evidence of a declining income since the time of the school closures with a downward trend, whilst the working hours of those still employed also is declining. There are evident effects on the labor supply of parents as an effect of school closures.
The findings from this thesis are in line with the perception of mothers being the leading child care taker in the households, as several women were no longer employed after the school closures - which is the main sign for what happens on the labor market.
The remaining part of the paper is organized as follows, section 2 describes relevant previous research which is important as a backdrop of this thesis, section 3 concludes the conceptual background, section 4 presents the data and section 5 goes through the research design. Following, section 6 presents the results of the study and section 7 concludes and discusses the result.
9 2. Previous literature
Previous studies on the topic of the effects of the Covid-19 virus are all quite recent papers, mostly from 2020, and some of them are working papers. Even if the pandemic is still ongoing a pile of research has been made that shed light on the implications. Some studies try to foresee the future impacts of the pandemic, these estimations are interesting to evaluate further down the path now that Covid-19 has affected the world for over a year. Estimates as such provide a sense of what was expected to happen in the beginning of the pandemic, where this study hopefully will provide new evidence concerning if these estimations were accurate.
Alon et al. (2020) study employment and childcare in the US before the Covid-19 pandemic, and try to foresee the possible impacts of it. The authors suggest that the impacts on the employment of women due to the pandemic are likely to be worse than the impacts of a recession. This is due to the restrictions in form of lockdowns that follow the pandemic, which most countries have imposed in order to reduce the infectivity. Employees are asked, and some imposed, to work from home to the highest grade possible, but for some industries this is an impossible task. Examples of industries as such are retail and leisure industries, where the labor force mostly consists of women. Along with school and nursery closures it is predicted that women will suffer from a higher need of child care in the homes.
Abi Adams-Prassl et al. (2020) show in their real time survey, conducted in March- April 2020 (in quite an early stage of the pandemic), that countries differ in the effects that Covid-19 has on the labor market. The effects also differ within countries, the impacts are unequal and the study concludes that women in the same way as less educated individuals are more affected by the crisis, as well as employees who cannot conduct their work from home.
The study by Abi Adams-Prassl et al. (2020) shows that the percentage of work that can be done from home is a predictor of the likelihood of losing a job, and in both the United States
10 and the United Kingdom women and employees without a college degree are significantly more likely to lose their jobs.
Sevilla and Smith (2020) study the division of child care work in the household between men and women in the UK before and during Covid-19. Before the pandemic there was a clear gender gap, where women in couples did 65% of childcare work in the household, which equates to a 30 percentage points gap between men and women. The study of Sevilla and Smith (2020) concludes that Covid-19 has had an impact on childcare in such a way that parents spend on average 40 hours more/week on childcare in comparison to when external providers of childcare were accessible. With the implications of the pandemic women spend on average 10 hours more than men on childcare weekly, and are more likely to lose their job. Sevilla A and Smith S (2020) also state that women's share of childcare work load is less affected by their employment status than men´s, in other words; men are more likely to fluctuate in providing childcare, if unemployed there is more time to spend on childcare, but if employed not as much time, in relation to women who are not as sensitive to these fluctuations.
Previous studies have estimated the effect of school closures on female labor supply using the CPS, with the same variation that I exploit in this thesis. Heggeness (2020) conducts a difference-in-differences analysis with data from the CPS and with data on school closures which are divided into early and late school closures. The result shown by Heggeness (2020) concludes that women with children of school age in states that had early closures of schools were 68% more likely to have a job without working in comparison to late closure states. It is also shown that mothers took a week off from work as a result of school closures, and fathers decreased their working hours with 0.53 hours per week. I contribute to these studies partly by the timing of the analysis that allows us to see further trends after the school closures, as well as the use of other outcome variables to represent the labor supply.
11 3. Conceptual background
Becker (1981;1965) is mostly known for the standard household utility model. In the model the family within the household maximizes the utility function of the household. The utility function has the constraint of 24 hours on a day, with regards to minimum hours of sleep and necessary childcare. The utility function includes the choice of labor, leisure, childcare and sleep, where the household all together maximize the utility. A standpoint in the model is that the rational parent will choose to work and pay for external childcare if the pay is higher than the cost of the childcare. This might not always be true as different parents may have different valuations of doing childcare, one might value it higher than someone else and this of course affects their choice. With this basic household utility model by Becker (1981) we can apply the implications of Covid-19, where the parents with young children no longer have the same option of paying for childcare, as nurseries and schools have closed. Becker’s (1981) model assumes that there is one decision making parent in the household - which is not entirely applicable to the real time US. To elaborate on this model researchers have studied how the bargaining works within the household when there is more than one decision maker in the family (Manser and Brown 1980; Lundberg and Pollak 1996), and this is what can be expected to happen within households as Covid-19 caused a pandemic.
Lundberg and Pollak (1996) find that parents, or adults, within a household bargain for resources, and there is evidence that shifting resources affects the consumption of the household. They also find that when mothers are in control of resources, they tend to (on average) spend more resources on children’s education and various household goods. This is confirmed by several other studies that have been conducted (Quisumbing and Maluccio, 2000;Rubalcava et al., 2004;Rangel, 2006,Rosero and Schady, 2007,Nunley and Seals, 2011;Heggeness, 2020a). These studies and the basic utility household model provides a good standpoint from which the situation during Covid-19 will be assessed; households maximize
12 one mutual utility function and there is bargaining within the household. In other words, it can be expected that the household will jointly make decisions on resources and consumption, and more importantly for this thesis, labor and childcare when Covid-19 leads to school closures throughout the US. Parents face a trade-off between childcare and work along with the bargaining decision of the labor supply.
With the empirical background provided it is a logical suspicion that in the current situation of this worldwide crisis mothers will lean more towards childcare and spend both more time and resources on children and domestic duties than fathers (Sevilla and Smith, 2020). Empirical studies also show that wives are the lower earning spouse in the household in around 70% of all households with married couples (Winkler et al., 2005; Murray- Close and Heggeness, 2019). Concluded by Bennett is that the spouse with the lowest pay will be much more likely to leave the labor market to partake in full time childcare before the higher earning spouse, e.g.; on average the male (Bennett, 2020).
4. Data
The data that I collect and use in this thesis is the date of the school closures on state level, household and individual data of socio-demographics and labor market information from the Current Population Survey. I present the variables in Table 1 in the appendix and give a short explanation of them as well in the following section that presents the data.
4.1 Effects on the labor market
The data on households in the US used in this thesis to form the differences-in- differences approach is gathered from the United States Census Bureau (CB). The CB is part of the U.S department of commerce which is part of the U.S Government, and is the nation's leading provider of quality data about its citizens and economy. One of the surveys conducted by the bureau along with the U.S Bureau of Labor Statistics (BLS) is the Current Population
13 Survey, (CPS). The CPS is a part of the Integrated Public Use Microdata Series, from now on called IPUMS, and the survey collects monthly data from around 70 000 individuals each month and has been doing so since 1962. The CPS is mostly conducted via telephone and includes over 200 questions. The data collected is monthly data, where the years chosen are 2019 and 2020. As 2021 is not yet over, it is not available.
To measure the effect of school closures on parents' labor supply I use three outcome variables. As I am interested in the effects on the labor supply I use a variable for employment status, which is a dummy variable that takes the value zero if the person is not employed, and one if the person is employed. This makes a good distinction to see if parents have gone from employed to not employed, and allows analyzing heterogeneity across genders, as employment status is available at the individual level (as opposed to the household level). I also make use of the outcome variable of family income, as this is a variable that is speaking of the family's situation when schools close. This variable includes the incomes of all members of the household aged 15 and older. The third outcome variable I study is hours worked per week, which is reported separately for all individuals in the household. This is a suitable complementary outcome variable, if parents are still employed, hours worked will be able to tell if the working hours have had to change due to school closures. As this variable is individual there may be a difference in how women and men change their hours worked when forced to do full time child care.
I control for a second order polynomial in age, education level (dummies for three levels of education, compulsory, high school and college plus), and the number of children in the household. I also control for the number of adults in the household. For a more detailed explanation of all the variables, as mentioned these can be observed in Table 1, Appendix.
14 4.2 Data on school closures
When the WHO declared the situation as a pandemic it did not take long until news arrived that schools were closing. School closures had been on the agenda as Covid-19 spread and became a reality shortly after the enunciation from the WHO. On March 16 it is reported that 22 states shut down all K-12 schools to contain the virus, and not long after the remaining states followed. K-12 includes elementary school, junior high school and senior high school.
Ages affected by these are in other words children who are 5-18 years old. Data on the school closures is gathered from Education week, where information on exact dates for state level school closures are available. Closures of schools were also taken at the school district level, but as data is on state level the state level closures are used in this thesis.
5. Research Design
The research design I use in this thesis is the difference-in-differences (DID) design.
DID is widely used in economics and is conducted by the use of observational data, in the absence of true experimental data. When a fraction of a population is exposed to a ‘treatment’
the effect of the treatment can be measured by comparing a control group that is not affected by the treatment effect. Under the assumption that the outcomes of the treatment- and control groups would have evolved similarly in the absence of the treatment, the difference-in- differences estimate captures the causal effect of the treatment.
I compare the labor market outcomes of parents to school-aged children that require supervision with parents to slightly older children, before and after March 2020. This comparison is made between households, in all states before and after the school closures. The time periods will be “before” which will refer to before the school closures and “post” which will refer to after the school closures. D(i,t) =1 denotes that individual i has a child within
15 school age (treated) and D(i,t)=0 if they don't (control). The treated group consists of parents of all ages in all states with at least one child in the household who is under the age of 12.
I have decided to confine the treatment group to parents with children under the age of 12, instead of including all children who were affected by the school closures (namely 5-18 years old) as all children cannot be expected to demand the same amount of child care. At the age of 18 children often have acquired the skills of cooking simple meals as well as taking care of themselves and taking responsibility for school work from home. In the CPS the variable family income measures all incomes from all individuals over 15, suggesting that from the age of 15 children might have a job for themselves. Taking this into consideration the age limit has to be drawn below 15 years of age. Children often mature in their teens, they become more self-running and independent and are not in the dire need for child care any longer. With this in mind, 12 years of age seems a not too arbitrary limit to draw between children who urge more and less childcare, just prior to the teenage years. The dependent variable Y(i,t) is the labor supply of parents, measured with three outcome variables. To assess the parallel trends assumption, I estimate a dynamic difference-in-differences
specification, where I interact the treatment indicator for indicators of each month before and after the school closures. Specifically I estimate:
!!" = $# + $$ & $$((!"× *!) + $%*!+ & $&(!"+ $',!+ -!"
(
)*+$' (
)* +$'
where the tau:s are dummy variables taking the value one for each calendar month since the school closure, such that tau = {-14 ,..., 8}, and tau = -1 is the omitted category. Thus, I compare the evolution in the outcome variables relative to February 2020, i.e., the calendar month before the school closures came into effect.
(1)
16 If the parallel trends assumption holds, there should be no statistically significant differences in the outcomes across households with differently aged children (i.e., between treated and control) before the school closure event.
The results are divided into focusing on two-partnered households as well as single parent households. The summarizing statistics on the construction of the household (Appendix Table 1C) show that 89% of the households within the sample are two-partnered households, and further almost 82% of the households consist of one male and one female. The thesis focuses on two-partnered heterosexual households, not simply because it is the most common family construction in the sample but because the focus lies in identifying if there is a bargaining between women and men in households.
As these three outcome variables will give implications on how the school closures have affected parents with at least one child below the age of 12 years it gives a sense of what has happened to the labor market. This is the aim of the thesis, but I also include complementary data from the Annual Social and Economic Supplement (ASEC) of the Current Population Survey (CPS), which annually reports on more detailed questions than the monthly CPS.
Variables that I use from ASEC is total income and income from wages, to analyze if the income from wages has decreased for parents and if there is a difference between men and women.
The ASEC is conducted in March every year, in other words the same month as the school closures. It can be expected that the effect of the school closures that might become evident in lowered income for men and women has somewhat of a time lag effect, parents might be able to cope with the situation with children in the beginning of the school closures with help from example grandmothers and grandfathers, and therefore might be able to keep the income steady for a while - which might become too difficult at some point and the income will reduce at this time. Nevertheless, it is of interest to see if there are any indications of such
17 an effect, and a way of getting a grip on the separate income of parents. I conduct a static difference-in-differences, measuring the income from wages of individuals with an interaction term that includes women that are within the treated group.
!!"= $#+ $$('()*) + $%(,-./*.0) + $& (1(2/3) + $(('()* × ,-./*.0 × 1(2/3) + $(5!"+ 6!"
In regression equation (2) the dummy Post is time period 1, i.e. after the school closure in the state, and Treated as a dummy for all parents with children within school age and Woman is a dummy for being a woman. $0 is the intercept, $4 is the DID-estimator which includes the interaction term for Woman, Post and Treated, and u is the error term. X is a vector of control variables, including a second order polynomial of age, education level and number of children and adults in the household. The standard errors are clustered on the state level. Further detailed information from the regression is presented in appendix Table 3, and the results of the analysis are presented in the following section.
6. Results
This section presents the results from the estimates. The general finding in the thesis is that the school closures seem to have affected women to a greater extent than men in terms of labor supply, especially when analyzing the variable employed, where several mothers have gone from employed to unemployed than men. For the variable family income there is a downward trend for the total incomes of the household. For the outcome variable hours worked there is a larger gap for men than for women. For the static difference-in-differences the result is significant for the interaction term, however the analysis is presented and discussed below.
From my estimations I conclude that parental labor supply is affected by school closures.
(2)
18 6.1 Main findings
The main results from the DID estimation show that there are changes on the labor market that have taken place along with the school closures in the United States. I start by presenting the outcome variable employed as it may be the most easily interpretable one.
Employed takes on the value 0 if the parent is unemployed and 1 if the parent is currently employed, and therefore makes a good starting point in presenting the results. The result shows that there is a significant drop in employment at the time of the school closure, meaning that the employment rate has gone down. For men there is a significant drop in employment and for women the effect is more than twice as large. This is presented in figure 1.A and 1.B below, separately for men and women.
The graph shows a downward trend for employment after the school closures for men, where 0 equals March 2020.
19 The graph shows a downward trend for employment after the school closures for women,
where 0 equals March 2020.
This serves as an indicator that both female and male parents' employment have suffered due to school closures, women’s more so. In figure 1.B there are close to no significant differences from 0 in the pre-period, i.e., before the school closures, which supports the key identifying assumption of the estimation. The other conspicuous decrease in employment that is visible in the graphs for both women and men, indicates that a big change was present in July 2019. The BLS reports that the unemployment for less than 5 weeks increased by 240,000 individuals in July, this may be the cause. However, the difference is not significant from 0 in figure 1.B and therefore the assumption stands supported.
The second outcome variable is family income, the total income of the household. In the graph below (Figure 2.A) it is clear that the family income of the households has decreased since the school closure, a rather steep decrease that spreads over months and looks quite
20 continuous. This variable is not split into gender whilst analyzing two partnered households, as the variable includes the total income of the whole family. 3
The graph shows a downward trend for family income after the school closures for households, where 0 equals March 2020.
Figure 2.A shows the families situation, even if both parents are still able to work they might not be able to work as much, or one parent does not work resulting in a lower total income for the household. However, it is evident that there is a downward facing trend that is not present before the school closures. In figure 2.A there are no significant differences from 0 in the pre- period, i.e., before the school closures, which supports the key identifying assumption.
The third outcome variable that I analyze is hours worked for individuals, figure 3A and 3B. The results show that there is a decline in the hours worked both for women and men.
Although, the drop is bigger for men than for women. This may be explained by the fact that the variable hours worked only includes individuals who are still employed, and as we concluded above, several women were unemployed in relation to men. This may be a reason
3 Although, when family income is analyzed specific on gender, the graphs show differences for women and men – it is concluded that this depended on single households in the sample and these estimations are thus not presented.
21 for the smaller drop in female hours worked, women who are still in the labor force did not reduce their working hours to the same extent as men, but at the same time several women became unemployed.
The graph shows a downward trend in hours worked for men after the school closures, where 0 equals March 2020.
The graph shows a downward trend in hours worked for women after the school closures, where 0 equals March 2020.
22 There might be a difference in this variable outcome between two partnered households and single households. When taking the standpoint of the bargaining household it is concluded and evident that the female to a greater extent becomes unemployed. Nevertheless, in single households, this may not be possible as there is only one source of income and mothers have to juggle it all to supply for their families. As concluded, there are more single mother households than single father households in this sample, and further analysis could tell us if the effects we observe on two partnered households are the same for single households, an analysis not conducted in this thesis.
Similarly to figures 2A and 2B that show employment it is evident that there is a drop around July for men and June for women in the number of hours that the individuals work.
This may be due to June and July being the summer months, it may depend on an ambulant presence of Covid-19 or a point where employers simply had to lay off staff. There is no clear trend for the hours worked for either men or women, but the male hours seem a little more balanced.
6.2 Findings in wage differences
Presented in appendix Table 2 are some evidence of the relationship firstly between gender and having children. What the table shows on the 99% significance level is that there is a negative effect on wage for women. Women without children earn less than men without children. When looking at parents, this study shows that men with children have a positive effect on the income, and for women with children there is a negative effect on the income in general. Therefore, I use the static difference-in-differences estimation to analyze the wages separately for women and men. As mentioned the only sample where this information is provided is within the ASEC which is conducted in March, and as schools closed down in March it is a little tricky to estimate. However, the output shows that there is a negative effect for women after school
23 closures who are within the treated group, although the effect is not significant when controlling for age, a polynomial of age, education and number of children. The output from the estimation is presented in the appendix (Table 2) as it still may give an indication of the effect (negative).
6.3 Alternative variables
This thesis aims to analyze the labor supply outcomes and does so with the help of above presented results and variables. Other variables that I have found can strengthen the results and are thus presented in this section. Nowork is a variable that takes on the value 1 if the person has a job but did not go to work the previous week, and 0 if the person did go to work. Nowork is coded as a dummy with the help of the CPS variable EMPSTAT that takes on several values for various employment status. The results for nowork are in line with the main results, we see a spike in June, July and August 2019 for both men and women, the trend stabilizes and after the school closure the trend continues upwards rather sharply. The effect for men is larger in the first two months than for women, but the long-term effect for women is larger. Figures are found in the appendix; (Figure 4.A) and (Figure 4.B).
I make use of a variable from the CPS that is only available from May 2020 as the variable is a dummy for teleworking for pay due to Covid-19, covidtelew. Even if May is not in direct connection to the school closures the variable still serves as informative on the situation of the households. The variable equals 1 if the individual has not worked remotely, and 2 if the person has worked remotely. Only a few more women in the sample have worked from home than men. The more interesting part though, is that many more workers, either male or female, showed not to have worked remotely for pay due to Covid-19. A possible explanation to this is simply because the job does not allow teleworking. It is interesting to see that a great share of the individuals in the sample do not work from home, for
24 reasons that are unknown to us. The share of individuals who did work remotely for pay due to Covid-19 can be expected to not be as affected in their work situation as those who did not work from home.
7. Discussion
I analyze how school closures due to Covid-19 have impacted the labor supply of parents in the US. I use cross sectional repeated micro data from IPUMS-CPS along with data of school closures in the United States that is collected from Education week. The evidence found in the OLS when looking at individuals show that women’s income is affected negatively. The results from the difference in difference estimation show that there is a significant reduction in labor supply for parents, especially for women in which the results of employment are more than twice as large as for men. Not controlled for in this thesis are other social distancing measures, which may also impact the income of the families as well as employment. The results of this thesis are speaking of a gender inequality when it comes to labor supply as women in two-partnered bargaining households have to a higher extent than men suffered from unemployment due to the school closures. This certainly begs the question of how the bargaining works if the parents have the same income, namely the same wage. Is the woman regardlessly harder hit by school closures, or are the effects equal? This is not measured in the thesis, but would be an interesting follow up study that would give more insight on the bargaining households. Also shown in this thesis is that women's wages are affected to a higher extent by having children after the school closures than men, also in line with the rest of the results.
The results that are presented in this thesis serves as evidence that the outcomes on the labor market are negatively affected by the school closures in the United States, and that mothers are suffering more from the closures, probably due to childcare, bargaining and lower
25 wages. This thesis analyzes the US, and it can be argued that the evidence can be generalized to other parts of the world. In many countries the family is constructed in the same way, in many countries the female is considered the main child caretaker, and in many countries women have lower wages than men. I suggest that further research can be conducted on the single households, to investigate whether single mothers were harder hit by the school closures than single fathers, or if this difference only exists within the bargaining household with one mother and one father. An interesting addition is analyzing the labor supply of parents at the time of the reopening of schools – to see if the effect is immediate and positive. Further, analysis might be proven useful if conducted in smaller samples, where state specific characteristics and socio-economic status are included to investigate whether any specific group suffers more from extended childcare duties due to school closures. As Cajner et al.
(2020) conclude in the beginning of the pandemic, the employment losses were disproportionally concentrated, it would be of interest to see if the concentration of employment losses due to school closures might also be disproportionate. As Covid-19 is still a pressuring crisis and this study is conducted in the second year of the pandemic, I highly suggest a follow up where even longer-term effects on the labor market might be visible, e.g., if women come back to the labor market when the pandemic passes, or if the effect is persistent.
26 Reference list
Abi Adams-Prassl, Teodora Boneva, Marta Golin, Christopher Rauh, “Inequality in the impact of the coronavirus shock: Evidence from real time surveys”, Journal of Public Economics, Volume 189, 2020, 104245, ISSN 0047-2727,
Alon, Titan, Matthias Doepke, Jane Olmstead-Rumsey and Michèle Tertilt (2020) “The impact of Covid-19 on gender equality”, CEPR COVID Economics and Real Time Papers, 4, 62-86.
Angrist, Joshua D., and William N. Evans. 1998. "Children and their Parents' Labor Supply:
Evidence from Exogenous Variation in Family Size." American Economic Review, 88(3): 450- 77.
Becker, G.S. "A Theory of the Allocation of Time," in G.S. Becker, The Economic Approach to Human Behavior, pp. 89-114 (also Econ. J. 1965:493-517).
Becker, G.S. A Treatise on the Family. Cambridge, MA: Harvard University Press 1981.
Berger, M. C., & Black, D. A. (1992). Child care subsidies, quality of care, and the labor supply of low-income, single mothers. The Review of Economics and Statistics, 74(4), 635–642.
Blau, D.M., Robins, P.K. Child care demand and labor supply of young mothers over time.
Demography 28, 333–351 (1991). https://doi.org/10.2307/2061460
Cajner, Tomaz, Crane, Leland D., Decker, Ryan A., Grigsby, John, Hamins-Puertolas, Adrian,Hurst, Erik, Kurz, Christopher, Yildirmaz, Ahu, 2020.The US Labor Market During the Beginning of the Pandemic Recession (National Bureau of Economic Research).
27 Davies, N. G., Klepac, P., Liu, Y., Prem, K., Jit, M., & Eggo, R. M. (2020). Age-dependent effects in the transmission and control of COVID-19 epidemics. Nature Medicine, 26(8), 1205–
1211. https://doi.org/10.1038/s41591-020-0962-9
Lundberg, S., & Pollak, R. A. (1996). Bargaining and distribution in marriage. Journal of Economic Perspectives, 10(4), 139–158.
Manser, M., & Brown, M. (1980). Marriage and household decision-making: A bargaining analysis. International Economic Review, 21(1), 31–44.
Murray-Close, M., & Heggeness, M. L. (2019). Manning up and womaning down: How husbands and wives report earnings when she earns more. Opportunity and Inclusive Growth Institute. Working Paper No. 28. Minneapolis, MN: Federal Reserve Bank of Minneapolis.
https://doi.org/10.21034/iwp. 28. Accessed 15 Oct 2020.
Sevilla, A., & Smith, S. (2020). Baby steps: The gender division of childcare during the COVID-19 pandemic.” CEPR Discussion Paper No. DP14804. London: Centre for Economic Policy Research.
Vlachos et al. (2021). The effects of school closures on SARS-CoV-2 among parents and teachers. PNAS Paper No. 2020834118
Winkler, A. E., McBride, T. D., & Andrews, C. (2005). Wives who outearn their husbands: A transitory or persistent phenomenon for couples?. Demography, 42(3), 523–535.
28 APPENDIX
Table 1: summary statistics of data and variable definitions
Variable name Definition Description Mean S.D Mean S.D
(Men) (Men) (Women) (Women) A. Person & household variables
Age Individual’s age Years 39.48 8.78 36.76 8.30
No. children Number of own children Children in household 1.87 1.0 1.81 0.99 in household
No. mothers Number of mothers in Mothers in household 1.05 0.36
household
No. fathers Number of fathers in Fathers in household 0.92 0.35 household
Sex Gender Male = 1, Female = 2
High school Educated from high school Dummy = 1 if educ = 73 0.27 0.44 0.23 0.42
College Educated from college Dummy = 1 if educ = 81 0.24 0.43 0.27 0.44
educ = 91 educ = 92
Higher education Higher education than Dummy = 1 if educ = 111 0.39 0.49 0.41 0.49
college educ = 123
educ = 124 educ = 125
29
Variable name Definition Description Mean S.D Mean S.D
(Men) (Men) (Women) (Women) B. Labor market outcomes
Employed Employed or Dummy = 1 if empstat = 10 0.29 0.45 0.29 0.45
unemployed if empstat = 12
Family income Total income for all Categorized into groups 10.99 3.93 10.0 3.93 members aged 15 or over with equal differences
in the household
Hours worked Hours worked last week Total number of hours 37.46 12.75 35.39 11.30 worked, all hours spent
at work, thus not including childcare work
Not in labor force In the labor force, e.g., Dummy = 1 if empstat = 32 0.29 0.46 0.24 0.43 can be unemployed if empstat = 34
but looking for work if empstat = 36
No work Did not work last
week, but is employed Dummy = 1 if empstat = 12 0.01 0.10 0.01 0.10
30
Variable name Definition Mean S.D Mean S.D
(Men) (Men) (Women) (Women) C. Dummy & control variables
Education Education takes on several different values depending 45.27 36.55 42.66 38.94 on what type of education the individual has completed.
Highschool Dummy created from education, where the dummy takes 0.26 0.44 0.20 0.40 value 1 if the individual has completed high school and
no further education.
College Dummy created from education, where the dummy takes 0.24 0.43 0.14 0.35
value 1 if the individual has completer college and no further education.
Highereduc Dummy created from education, where the dummy takes 0.39 0.49 0.26 0.23 value 1 if the individual has completed any higher
education then college, namely bachelor, mag, masters, PhD and so on.
Lowincome Dummy created from family income where the dummy 0.21 0.40 0.27 0.46
takes value 1 if the total family income is between
$500- $24,999
Midincome Dummy created from family income where the dummy 0.35 0.47 0.41 0.49
takes value 1 if the total family income is between
$25,000-$59,999
Highincome Dummy created from family income where the dummy 0.44 0.49 0.31 0.46
takes value 1 if the total family income is between
$59,999 – any higher total income.
31
Variable name Definition Description Freq %
D. Household construction
Single mother One mother One mother lives in the household 35,570 8.74%
Single father One father One father lives in the household 7,364 1.81%
Two partner Two parents Two parents live in the household 363,241 89.24%
Hetero partner Male & female One mother & one father live in 332,375 81.65%
Parents the household
Male partner Two-partnered Two fathers live in the household 277 0.07%
household, male
Female partner Two-partnered Two mothers live in the household 5,760 1.42%
Household, female
Married Married & spouse Married = 1 if spouse present 312,86 76.89%
is present
32
33 Table 2: regression output static difference-in-differences
--- income Variable wage --- 1.woman 60243.8***
(96894.34) 1.treated 683491.0**
(196818.66) 1.woman#treated -1348038.5***
(158921.78) 1.post 294945.9**
(103395.62) 1.woman#post 22572.8 (119640.00) 1.treated#post 405803.0**
(122883.07) 1.woman#treated#post -208362.0 (174326.99) age -1845810.7***
(129859.39) age2 6985.2***
(1170.86) educ -792632.5***
(31892.63) highereduc 34420943.3***
(979376.58) cons 101987541.1***
(558582.25) --- N 338,060 ---
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.00
