Childcare Facility Closure and Exacerbated Gender Inequality in Parenting Time during the COVID-19 Pandemic

Abstract

The unprecedented large-scale childcare facility closure during the COVID-19 pandemic led to a dramatic increase in the childcare burden at home, which is shouldered disproportionately by women more than men. Leveraging anonymized mobile tracking data and nationally representative time-use survey data from the USA, this study adopts a quasi-experimental approach to examine the impact of childcare facility closure on the gendered division of household childcare time. It further investigates whether this impact varies according to respondent education, family income, and employment status. Results show an expanding gender gap in parenting time on child education with young children during the pandemic. The gender gap expanded even more in places and months with more childcare facility closures, but this gendered effect is only evident among parents with lower education and family income. Our findings call for institutional support during similar public crises to mitigate the potentially negative impact on gender equality.

Keywords

childcare COVID-19 family gender socio-economic status time use

Introduction

The COVID-19 pandemic brought profound changes to people’s everyday life, including unprecedented large-scale school and childcare facility closures that dramatically increased home childcare burdens. This particularly impacted women, who already bore the brunt of unpaid, invisible care work before the pandemic (Aguiar and Hurst, 2007; Bianchi et al., 2000). Childcare accessibility plays a crucial role in household and workplace gender equality (Hook, 2010; Ruppanner et al., 2021). It is thus extremely important to investigate the impact of the pandemic and its related shocks on the gendered division of household childcare work. Such investigation will not only enhance our understanding of the determinants of domestic labor division in times with suddenly reduced external childcare support, but also inform policy measures to mitigate the impact of a public crisis on gender inequality.

Recent research has revealed preliminary evidence of gender disparities in care work during the pandemic (e.g. Augustine and Prickett, 2022; Farre et al., 2022; Sevilla and Smith, 2020). However, many studies use ad hoc surveys with the specific purpose of understanding situations during the pandemic. Such surveys could not reflect how the pandemic may have exacerbated the pre-existing gender inequality due to a lack of comparable pre-pandemic data. More importantly, no study has yet directly investigated the impact of childcare facility closures on the gendered childcare division, partly due to the challenges in accurately measuring the operational statuses of childcare facilities.

Leveraging an innovative, anonymized mobile tracking dataset and linking it to US nationally representative time-use survey data, this study bridges this gap by utilizing a direct measure of childcare facility closure across time and space and examining its impact on gender disparities in household childcare work. It also investigates the potential intersectionality between gender and socio-economic status (SES) in shaping inequality during a crisis. The following research questions are examined:

RQ1: Did the pre-existing gender gap in childcare time increase during the COVID-19 pandemic?

RQ2: If yes, did it increase more when and where more childcare facilities were closed?

RQ3: Does the potentially gendered effect of childcare facility closure also vary by socio-economic factors such as education, family income, and employment status?

Background

Gender Disparities in Care Work

Gender inequality in the distribution of unpaid care work is a prevailing and persistent global phenomenon. Despite increased labor force participation, women still bear most unpaid care work (Bianchi et al., 2000; Craig and Mullan, 2011; Sayer and Gornick, 2012). Research has examined the gender division of household work through various theoretical lenses (Brines, 1994; Craig and Mullan, 2011). The time availability perspective posits unpaid household work is rationally allocated among household members based on time left after paid work; gender disparities in unpaid household work thus result from gender inequality in paid work (Bianchi et al., 2000; Hiller, 1984). The relative resources perspective argues household work distribution reflects domestic power dynamics: the partner with more resources, such as higher education, income, or career prestige, has more bargaining power to avoid or negotiate away household work (Brines, 1994; Evertsson and Nermo, 2004; Ross, 1987).

The gender ideology perspective, rooted in sociology, links household work division to gender ideologies, socialization, and performance (Greenstein, 1996; Poortman and van der Lippe, 2009; Schneider, 2011). It explains cases when women maintain domestic work time despite earning more and having a higher social status than their partners. This is especially true for childcare: when work hours increase, mothers would prefer to reduce leisure and sleep time rather than childcare time (Craig, 2007).

Consistent with the relative resource perspective, highly educated mothers see increased child education time from fathers in Spain and the UK (Gimenez-Nadal and Molina, 2013). US women with higher incomes also spend less time on housework (Gupta, 2007). However, echoing the gender ideology perspective, gender inequality in childcare persists among college-educated parents, and college-educated mothers allocate more time to childcare than other mothers in Australia (Craig, 2006). Evidence from the USA also shows the gender gap in unpaid work even increases when women earn more than their partners (Bertrand et al., 2015).

The Role of Institutional Childcare Infrastructure

The gendered division of care work is not only rooted in individual-level dynamics. Institutional constraints and opportunities, such as parental leave policies and childcare infrastructures, are also important. Cross-national evidence shows that women share less housework where public childcare is abundant (Hook, 2010). Countries with robust childcare support and well-paid leaves witness smaller differences between mothers and childless women in employment and working hours (Boeckmann et al., 2015). In Germany, daycare subsidies nearly doubled labor supply for mothers with small children (Domeij and Klein, 2013). Policies promoting paternity leave contribute to a more egalitarian, non-traditional household labor division (Patnaik, 2019) and reduce the within-household gender wage gap (Andersen, 2018).

In the USA, although informal childcare provided by unlicensed, noncustodial caregivers forms a crucial source of childcare support, formal childcare facilities still play an important role and shape household work divisions (Laughlin, 2010). Expensive childcare makes more mothers leave the labor market, especially single and low-income mothers (Han and Waldfogel, 2001; Landivar et al., 2022b). States with longer school days observe increased labor force participation from mothers (Ruppanner et al., 2019). Mothers who relocate to states with more expensive childcare also experience lower odds of employment (Landivar et al., 2021).

While these studies primarily focus on women’s employment rather than the gendered childcare division as the outcome, the underlying implication is that mothers’ labor force participation is more significantly impacted by household childcare needs than that of fathers. Therefore, the loss of external childcare support would likely exacerbate gender inequality at home and work. The sudden changes in external childcare availability during the COVID-19 pandemic thus provide a unique opportunity to study the impact of a severe shock to childcare needs.

The Gendered Impact of the Pandemic

When the external childcare infrastructure weakens due to the pandemic and its subsequent social-distancing measures, the increased household childcare responsibilities are likely disproportionately borne by women. Recent literature has indeed revealed global gender disparities in childcare obligations during the COVID-19 crisis. In the USA, women shouldered a heavier load of childcare than men, regardless of mothers’ employment status (Adams-Prassl et al., 2020; Augustine and Prickett, 2022; Zamarro and Prados, 2021). Similar patterns are found globally, including in the UK (Sevilla and Smith, 2020), Canada and Australia (Johnston et al., 2020), Spain (Farre et al., 2022), India (Chauhan, 2021), Nigeria (Obioma et al., 2023), and South Africa (Obioma et al., 2023; Parry and Gordon, 2021). Moreover, this unequal share of childcare is associated with reduced working hours and higher chances of unemployment for women (Zamarro and Prados, 2021). Mothers working from home with young children experienced significantly reduced work hours while fathers were nearly immune to such arrangements (Collins et al., 2020; Landivar et al., 2020). Mothers were three times more likely to report unemployment due to childcare issues than fathers (Heggeness and Fields, 2020).

The gendered impact of the pandemic may also intersect with SES: women with lower education were more likely to report reduced work hours whereas higher-educated women reported increased work hours (Fan and Moen, 2022). It is not clear, however, if the pandemic’s impact on childcare time division also differs for women with different SES backgrounds.

Regarding the specific impact of childcare facility and school closures, research shows that having fewer in-person instruction days at elementary schools reduced mothers’ labor force participation relative to fathers and women without children (Landivar et al., 2022a). The gender gap in labor force participation rates also increased more in US states where elementary schools offered primarily remote instruction in 2020 (Collins et al., 2021). For those with young children, the loss of full-time childcare increased the risk of unemployment for mothers but not fathers (Petts et al., 2021).

These studies, however, suffer from important limitations. First, many use ad hoc surveys from 2020, and are thus unable to isolate the pandemic effect from the long-existing gender inequality in care work. Second, even studies with pre-pandemic data tend to simply compare situations before and after the pandemic; no study has directly linked childcare facility closures to parental time use to investigate the immediate policy impact. Childcare facility closure varied greatly across places and times due to differential risk levels of COVID-19 contagion and responses from local governments (Lee and Parolin, 2021). Research utilizing information on childcare facility closure is still limited due to challenges in accurately documenting facility operation statuses across time and space. Consequently, we have a limited understanding of the impact of childcare infrastructure loss on households with younger children.

This study addresses these gaps by utilizing a novel dataset with anonymized mobile tracking information to measure changes in childcare facility traffic and merging it with US nationally representative time-use surveys from 2011 to 2020. Through quasi-experimental two-way fixed effects (TWFE) models, we examine changes in men’s and women’s time spent on childcare activities, in particular child education. We propose the following hypotheses corresponding to the three research questions:

H1: The pre-existing gender difference in household childcare time increased during the COVID-19 pandemic.

H2: The gender difference in household childcare time increased more when and where more childcare facilities were closed during the pandemic.

H3: The gendered effect of childcare facility closure on household childcare time is more evident among those with lower education, lower family income, and not employed.

Methods

Data

This study leverages two data sources. To examine household childcare time division, we utilize the American Time Use Survey (ATUS), a national survey conducted monthly by the US Census Bureau since 2003. The ATUS is a repeated cross-sectional survey to gather nationally representative data on how US adults allocate their time to various work and life activities. The ATUS sample is drawn from the Current Population Survey (CPS), covering the US civilian population in occupied households. The ATUS first selects households through stratified sampling based on race/ethnicity, the presence and age of children, and the number of adults in households. An eligible household member who is at least 15 years old is then randomly selected as the designated respondent. The ATUS interviews each respondent once to collect a 24-hour retrospective time diary (Bureau of Labor Statistics, 2022).

This study uses ATUS 2011–2020 and limits the sample to respondents between 18 and 55 who live with their spouse or unmarried partner of the opposite sex and have at least one child under the age of six living in the same household, resulting in 13,531 cases across 10 years. The 2011–2019 data are used to compare with the 2020 data to demonstrate changes before and during the pandemic. The 2020 ATUS was disrupted by the pandemic and was suspended from mid-March to mid-May (Bureau of Labor Statistics, 2022). We thus only keep data from 10 May to 31 December each year, when the 2020 ATUS data are available, resulting in 8447 cases. We further exclude responses on weekends and national holidays, resulting in 5497 cases in the final analytical sample, including 491 cases from 2020 and 5006 cases from 2011–2019.

To measure childcare facility closure across time and space, we utilize a novel dataset, the US Database of Childcare Closures, which uses aggregated Global Positioning System (GPS) tracking data of over 40 million mobile phone users in the USA provided by SafeGraph (Parolin and Lee, 2021). The SafeGraph data contain location-specific information from opted-in devices with users’ consent. Anonymized geo-coded information is collected at over four million ‘points-of-interest’ (POI) across the USA and aggregated to county and state levels. To protect privacy, these data are strictly anonymized, aggregated, and further infused with randomized noise to ensure that they cannot be linked back to individuals or devices (SafeGraph, 2022). The SafeGraph POI data have been used to study mobility and facility closures during the pandemic and have shown high consistency with administrative data (Hansen et al., 2022; Lee and Parolin, 2021; Parolin and Lee, 2021). Childcare facilities are included as POI and identified through the North American Industry Classification System code 62441, enabling researchers to track in-person visits to these locations before and throughout the pandemic. The database includes 85,328 total childcare facilities across 2228 US counties, covering approximately 78% of all licensed childcare institutions in the country (Lee and Parolin, 2021).

Measures

Independent Variable

We define a childcare facility as ‘closed’ when its monthly in-person visits decreased by 75% or more in 2020 compared with the same month in 2019. The percentage of closed childcare facilities (PCC) is then calculated by county and month in 2020. We then match the county–month PCC information to the ATUS sample. For respondents without county information, we match them with the state–month PCC. All ATUS respondents in the analytical sample are successfully matched to PCC data. The final sample contains information on 381 counties/states across 80 months. In a robustness check, we redo all analyses with state–month PCC and achieve very similar results (presented in the Online Appendix).

We observe a high level of facility closure throughout 2020. In Figure 1, we present a boxplot on the distribution of county-level percentages of closed childcare facilities for ATUS respondents by month. There are considerable spatial and temporal variations: the highest PCC was found in May, but still ranged from near zero to over 40% across counties. In December, on average about 9% of childcare facilities were closed, ranging from 0 to over 25% across counties. These variations across time and space provide a unique opportunity to estimate the impact of changes in institutional childcare support.

Figure 1.

Boxplot on the distribution of county percentages of closed childcare facilities by month in 2020.

Dependent Variable

The ATUS includes an extensive list of childcare-related activities, such as checking homework, helping with school projects, giving children medicine, dropping off and picking up children at babysitter’s, and so on. This study focuses on daily time spent on overall childcare and education-related activities. For overall childcare activities, we use a variable that covers activities related to: (1) caring for and helping household children; (2) household children’s education; and (3) household children’s health. For education-related activities, we use a variable that covers activities related to household children’s education only. In 2020, new categories were added to child education-related activities to accommodate changes during the pandemic. These include activities like helping children set up and sign in to online classes and attending online school meetings. While these new categories could increase reported time on child education simply because they were omitted in previous years, it should not bias our estimation because time spent on these activities should be minimum pre-pandemic. Moreover, our main focus is on gender disparities in the increased time on child education and how childcare facility closure is related to such disparities, and this relies more on comparisons across different counties and months in 2020.

We also report time on non-education-related childcare activities by subtracting time on education-related activities from the overall childcare time. For details of activities included in the time-use variables, please refer to the ATUS documentation.

Other Variables

Additional variables include respondent sex (female = 1, male = 0), age, race (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian, and Other/Mixed), education (college degree or above = 1; no college degree = 0), employment status (currently employed = 1; unemployed or not in the labor market = 0), family total annual income (coded into 16 levels ranging from ‘Less than $5000’ to ‘$150,000 and over’ and treated as a continuous variable), the total number of children under 18 in the household, and the year–month indicators. These variables are included in all models unless otherwise specified.

Analytical Approaches

This study uses linear regression models with county and year–month fixed effects. The general model is:

T i m e_{i t c} = β X_{i t c} + α_{c} + γ_{t} + ϵ_{i t c}

where $α_{c}$ is the unit (county) fixed effects and $γ_{t}$ is the time (year–month) fixed effects, $T i m e_{i t c}$ is the time spent on childcare activities for individual i in county c at time t, $X_{i t c}$ is a vector of predictors at interest for individual i in county c at time t, $β$ is a vector of coefficients at interest, and $ϵ_{i t c}$ is the error term. This is a generalized form of Difference-in-Difference models (DiD). County fixed effects are included to control for county-level time-invariant heterogeneities; year–month fixed effects are included to control for time-varying factors that have common trends across US counties. We cluster standard errors at the county level and include final ATUS survey weights in all models; we consider alternative settings in the robustness check as shown in the Online Appendix.

To examine RQ1, whether the gender gap in childcare time increased during the pandemic, we use the following model:

T i m e_{i t c} = β_{1} F e m a l e_{i} + β_{2} Y e a r 2020 X F e m a l e_{i} + α_{c} + γ_{t} + ϵ_{i t c}

Here, $β_{1} F e m a l e_{i}$ represents the overall gender gap in childcare time before the pandemic (Year2020 = 0), and the interaction term $β_{2} Y e a r 2020 X F e m a l e$ indicates whether the gender gap has changed in the year 2020. A significantly positive coefficient $β_{2}$ thus means that the gender gap increased in 2020, indicating that women shouldered more of the additional childcare time during the pandemic. Note that the main effect of the year 2020 has been absorbed into the year–month fixed effect. This model thus isolates the effect of the pandemic by controlling for county and time heterogeneities at the same time.

To examine RQ2, whether the gender gap increased more when and where more childcare facilities were closed, we use the following model:

\begin{matrix} T i m e_{i t c} = β_{1} F e m a l e_{i} + β_{2} P C C_{t c} + β_{3} Y e a r 2020 X F e m a l e_{i} + β_{4} P C C_{t c} X F e m a l e_{i} \\ + β_{5} Y e a r 2020 X P C C_{t c} + β_{6} Y e a r 2020 X F e m a l e_{i} X P C C_{t c} + α_{c} + γ_{t} + ϵ_{i t c} \end{matrix}

While we need to include all main effects and two-way interactions in the model, we are mainly interested in the three-way interaction term, $β_{6} Y e a r 2020 X F e m a l e_{i} X P C C_{t c},$ which reflects how childcare facility closure affects the changes in the gender gap in childcare time before and during the pandemic. Note that the two-way interaction term $β_{3} Y e a r 2020 X F e m a l e_{i}$ now represents the change in the gender gap during the pandemic when and where no childcare facility is closed (PCC = 0), and a significantly positive coefficient of the three-way interaction term ( $β_{3}$ )means that the gender gap increased more in places and times when more childcare facilities are closed.

To examine RQ3, whether the potentially gendered effect of childcare facility closure varies according to other socio-economic factors, we further split the sample by respondent education, family income, and employment status, and conduct the same set of analyses in each subsample. We also tested for significant differences across the subsamples through models with four-way interaction terms. We then visualize the results via marginal effects plots to present the group differences more clearly.

It is worth noting that in RQ2 and RQ3, we have a continuous treatment variable – the percentage of closed childcare facilities. Scholars have pointed out that TWFE DiD settings with continuous treatment variables require different assumptions from the conventional DiD models with binary treatment variables (Callaway et al., 2021). Beyond the regular parallel trends assumption, we further assume homogenous treatment effect functions across counties with different levels of childcare facility closure. In other words, parents would respond in the same way in adjusting their parenting time to marginal changes in percentages of closed childcare facilities in counties with a higher or lower level of childcare facility closure.

Robustness Check

We conducted various robustness checks by redoing all analyses (1) without weights; (2) clustering standard errors at the state instead of county level; (3) using state–month instead of county–month PCC; (4) without control variables; (5) with a placebo test assuming 2018 is the pandemic year; (6) with county-level random slopes; and (7) with leads specifications to test the parallel trends assumption. All analyses show very similar results with the main analyses regarding the effect directions, magnitudes, and significance levels for variables of interest. These results are presented in the Online Appendix. Specifically, we treat each year as separate dummy variables and estimate whether the effect of PCC varies from year to year, with 2019 as a reference year. In other words, we test whether there are heterogenous temporal trends across different levels of PCC. Results are shown in Table S10 and Figure S1. Overall, only the year 2020 shows a large difference from the year 2019, supporting the parallel trends assumption.

Results

Descriptive Data and Trends

Table 1 presents the summary statistics of the main variables by gender and time period. Before 2020, men typically spent 87.07 minutes and women spent 166.04 minutes daily on overall childcare activities. Most time is spent on non-education-related activities. Men on average only spent 3.12 minutes and women spent 7.79 minutes daily on child education. These striking gender differences echo previous literature on unequal household childcare labor division. Notably, in 2020, overall childcare time rose by 14.39 minutes for men (to 101.46 minutes) and 26.53 minutes for women (to 192.57 minutes). Time spent on child education increased by 17.19 minutes for women (to 24.98 minutes) but decreased for men (to 1.15 minutes). We further discuss these descriptive changes in the results section.

Table 1.

Unweighted descriptive statistics.

Variable	Before 2020			Year 2020
Variable	Male,N = 2405^a	Female,N = 2601^a	p-value^b	Male,N = 234^a	Female,N = 257^a	p-value^b
Race			0.068			0.2
White	1679 (70%)	1800 (69%)		156 (67%)	179 (70%)
Black	137 (5.7%)	109 (4.2%)		8 (3.4%)	13 (5.1%)
Hispanic	376 (16%)	454 (17%)		34 (15%)	42 (16%)
Asian	176 (7.3%)	192 (7.4%)		27 (12%)	19 (7.4%)
Other/Mixed	37 (1.5%)	46 (1.8%)		9 (3.8%)	4 (1.6%)
Age	35.83 (6.31)	33.36 (5.81)	<0.001	36.60 (5.99)	33.70 (5.23)	<0.001
Education			0.001			0.3
No college degree	1244 (52%)	1227 (47%)		110 (47%)	110 (43%)
College or above	1161 (48%)	1374 (53%)		124 (53%)	147 (57%)
Family income	12.47 (3.45)	12.19 (3.55)	0.002	13.64 (2.59)	13.34 (2.90)	0.4
Employment status			<0.001			<0.001
No	162 (6.7%)	1038 (40%)		18 (7.7%)	101 (39%)
Yes	2243 (93%)	1563 (60%)		216 (92%)	156 (61%)
Total number of children in household	2.15 (1.05)	2.11 (1.02)	0.12	2.08 (0.95)	2.19 (1.17)	0.6
Daily time on overall childcare (min)	87.07 (103.12)	166.04 (137.15)	<0.001	101.46 (126.79)	192.57 (160.43)	<0.001
Daily time on child education (min)	3.12 (14.79)	7.79 (29.15)	<0.001	1.15 (8.69)	24.98 (74.46)	<0.001
Daily time on non-education childcare (min)	83.95 (102.00)	158.26 (133.78)	<0.001	100.31 (125.91)	167.59 (144.42)	<0.001

Notes: N = 5497.

N (column %) for categorical variables; Mean (SD) for continuous variables.

Pearson’s Chi-squared test for categorical variables; Wilcoxon rank sum test for continuous variables.

It is worth noting that the socio-economic characteristics of respondents shifted over time. In 2020, the percentages of respondents with a college degree or above increased for both men (from 48% to 53%) and women (from 53% to 57%), and the average level of family income also slightly increased for both genders. These changes may indicate that those who managed to complete the survey during the pandemic had relatively higher SES.

The gender disparities are again clearly shown in Figure 2, which presents the overall trends of parents’ time spent on all childcare, education, and non-education activities across the past 10 years. While there were fluctuations across the 10 years, when the pandemic hit, we observe a significant increase in overall childcare time for women from 2019 to 2020. The additional burden is not shared by men: in fact, their overall childcare time slightly decreased by seven minutes from 2019 to 2020. Time on child education increased more dramatically from 12 minutes to 32.5 minutes a day for women but decreased from 4.8 minutes to 1 minute for men. Time on non-education childcare activities did not change much from 2019 to 2020.

Figure 2.

Average daily time spent on overall childcare and child education, 2011–2020. (a) Daily time on overall childcare. (b) Daily time on overall education. (c) Daily time on non-education childcare.

The descriptive data indicate that the increase in childcare time during the pandemic mainly concentrates on child education-related activities, which are typically more intensive and stressful (Musick et al., 2016; Negraia et al., 2018). Therefore, we focus on education-related childcare time and present analyses from the two-way fixed effects models.

The Gendered Effect of The Pandemic

Table 2 presents results from regression models predicting daily time on child education with two-way fixed effects. Model 1 presents results from the pre-pandemic sample, and model 2 presents results from the 2020 sample. Models 3 and 4 present results from the full sample.

Table 2.

Two-way fixed effects models predicting time on child education.

	(1)Before 2020	(2)Year 2020	(3)Full sample	(4)Full sample
Female	4.19 *** (0.84)	29.78 ** (9.70)	4.81 *** (0.90)	5.80 *** (1.45)
Female X Year 2020			23.25 ** (8.94)	1.10(10.67)
PCC				0.03(0.14)
Female X PCC				–0.13(0.12)
Year 2020 X PCC				0.78(0.55)
Female X Year 2020 X PCC				2.75(1.88)
Race (Reference = White)
Black	2.23(2.28)	–12.68(10.73)	3.34(3.20)	3.20(3.10)
Hispanic	–1.08(1.58)	–4.83(6.54)	–1.53(1.49)	–1.40(1.45)
Asian	6.71 ** (2.08)	17.15(12.47)	10.26 ** (3.26)	10.72 *** (3.15)
Other/Mixed	–3.93(2.42)	3.95(10.16)	–1.35(3.10)	0.30(3.49)
College or above	–2.66 ** (0.96)	–6.09(8.17)	–3.25 * (1.28)	–3.30 ** (1.15)
Family income	0.35 * (0.17)	0.05(0.83)	0.32(0.17)	0.31(0.17)
Employed	–7.82 *** (1.39)	5.00(9.81)	–6.52 *** (1.29)	–6.78 *** (1.29)
Age	0.18 * (0.08)	0.19(0.43)	0.14(0.08)	0.14(0.08)
Number of children in household	4.21 *** (0.81)	9.20(5.39)	4.63 *** (0.80)	4.55 *** (0.83)
N	5006	491	5497	5497
R²	0.17	0.48	0.16	0.21
Adjusted R²	0.08	0.18	0.09	0.14
Within R²	0.07	0.06	0.07	0.10

Notes: total N = 5497. PCC = Percentage of Childcare Closure at the county–month level. Linear regression models with county and year–month fixed effects. Model 1 uses cases before 2020; Model 2 uses cases in 2020; Models 3 and 4 use the full sample. ATUS final weights are used in all models. Standard errors are clustered at the county level.

p < 0.05; **p < 0.01; ***p < 0.001.

In Model 1, we observe a significant pre-pandemic gender gap: women spent 4.19 more minutes than men daily on child education (b = 4.19, SE = 0.84). This gender gap increased dramatically in Model 2: in 2020, women spent 29.78 more minutes than men on child education daily (b = 29.78, SE = 9.70). In Model 3 using the full sample, we include an interaction term between female and the year 2020. The significant interaction effect shows that the pre-existing gender gap is significantly increased by 23.25 minutes in 2020 (b = 23.25, SE = 8.94).

In Model 4, we further include a three-way interaction term between respondent sex, the year 2020, and the indicator on childcare facility closure. We also include all main effects and lower-order two-way interactions in the models unless they are absorbed by the year–month fixed effects. The coefficient of the three-way interaction term (Female X Year 2020 x PCC) thus reflects whether the gendered impact of the pandemic varies according to the percentage of closed childcare facilities in a given county and month during the pandemic.

In Model 4, we do not observe a significant three-way interaction effect (b = 2.75, SE = 1.88). However, the impact of childcare facility closure could differ across socio-economic groups, and an overall insignificant interaction effect may be masking the underlying heterogeneity. Therefore, we next repeat the analyses in subgroups by education, family income, and employment status.

The Gendered Effect of Childcare Closure across Socio-economic Groups

In Table 3, we repeat the analyses as in Model 4 of Table 2 using different subsamples. In Panel A, we split the sample according to respondent education, and this time we observe a significant three-way interaction effect between respondent sex, the year 2020, and the percentage of childcare facility closure among those without a college degree (b = 6.95, SE = 3.27). This reflects a significantly gendered effect of childcare closure among those with lower education: a one percentage point increase in closed childcare facilities is related to 6.95 more minutes in the gender gap in child education time. This effect is not found among those with a college degree or above (b = −0.67, SE = 0.72). We test the difference between these two subsamples by including a four-way interaction term in the full sample (Female X Year 2020 X PCC X Education) and find a significant difference. In other words, while childcare facility closure increases the gender gap in time spent on child education, this gendered effect only exists among those without a college degree.

Table 3.

Two-way fixed effects models predicting time on child education by education, family income, and employment status.

	Subsample 1	Subsample 2	Difference test
Panel A: Split by Education	No college degree	College or above
Female X Year2020 X PCC	6.95 * (3.27)	–0.67 (0.72)	*
N	2691	2806
Panel B: Split by Family Income	Low family income	High family income
Female X Year2020 X PCC	6.43 * (3.07)	1.55 (1.47)	*
N	2089	3408
Panel C: Split by Employment Status	Not employed	Employed
Female X Year2020 X PCC	5.10 (4.35)	1.49 (1.49)
N	1319	4178

Notes: total N = 5497. PCC = Percentage of Childcare Closure at the county–month level. Linear regression models predicting daily time on child education with county and year–month fixed effects. ATUS final weights are used in all models. Standard errors are clustered at the county level. Individual-level control variables (education, family income, employment status, age of respondent, and total number of children in household) are included in all models (unless used to split the sample) but omitted from the table. For each three-way interaction term, all main effects and lower-order two-way interaction terms are also included in the models but omitted from the table. The difference test is conducted via a four-way interaction term.

p < 0.05; **p < 0.01; ***p < 0.001.

We further present these results in Figure 3a, which shows the marginal effect of monthly childcare facility closure in 2020 on daily time on child education by respondent sex and education. The left plot shows that childcare facility closure in 2020 does not affect time spent on child education before 2020, which is as expected since the treatment had not happened then. The right plot clearly shows a strong effect on females with no college degree and much weaker effects on the other three groups. Consequently, in counties and months with more childcare facility closures, we observe a larger gender gap in time on child education among parents without a college degree but not those with higher education.

Figure 3.

Predicted daily time on child education by monthly county average percentage of closed childcare facilities in 2020. (a) By sex and education. (b) By sex and family income.

Similarly, in Panel B of Table 3, we split the sample by family income. We define those with a family income at level 12 or below (⩽ $59,999) as low income, and those with a family income at level 13 or above (>$59,999) as high income. Results again show a significant three-way interaction among those with a lower family income (b = 6.43, SE = 3.07). This means for every one percentage point increase in closed childcare facilities, the expanding gender gap in daily time on child education during the pandemic further increases by 6.43 minutes. This effect, however, is much smaller in magnitude and not significant among those with a higher family income (b = 1.55, SE = 1.47). We also test this difference across the family income spectrum by including a four-way interaction term (Female X Year 2020 X PCC X Family Income, where family income is treated as a continuous variable) in the full sample and find a significant four-way interaction, indicating that childcare facility closure has a particularly gendered effect among low-income families.

We also present these results in Figure 3b. The right plot again clearly shows differential effects of childcare facility closure in the year 2020 by respondent sex and family income. The effect is greatest among women with low family income, followed by women with high family income; the effect on men, conversely, is negligible regardless of family income. Therefore, as more childcare facilities close, we observe a larger increase in the gender gap in time spent on child education, and this gendered effect is especially evident among low-income parents.

Last, in Panel C of Table 3, we observe a larger coefficient of the three-way interaction term for the group not working than the employed group, indicating that childcare facility closure may have a larger impact on the gendered division of child education time among those unemployed or not in the labor market than those employed. This difference across the two subsamples, however, is not statistically significant.

Discussions and Conclusions

This study utilizes a novel dataset on childcare facility closure during the COVID-19 pandemic to examine how changes in institutional childcare support impact gender inequality in households with young children. Our findings not only extend research on the gendered care work division and the role of external childcare infrastructure, but also provide insights to improve policy responses to school and childcare facility closures during a public emergency.

First, we highlight the heightened childcare burden for women during the pandemic, particularly in child education. Women’s daily time on child education rose from 12 minutes in 2019 to 32.5 minutes in 2020, reflecting the increased homeschooling needs due to the large-scale childcare closure. Conversely, while men already spent much less time on childcare than women pre-pandemic, they spent even less time in 2020 on child education activities. Further analyses indeed show exacerbated gender inequality in caregivers’ time spent on child education after controlling for individual factors such as race, education, income, and employment status. The gender gap expanded from about five minutes daily to nearly half an hour in 2020. Consistent with recent studies (Del Boca et al., 2020; Sevilla and Smith, 2020), our analyses provide further evidence that the additional burden of child home education during the pandemic was almost entirely shouldered by women. This gendered division could seriously limit women’s opportunities for employment and exacerbate household and labor market inequality.

Second, we found childcare facility closure significantly widened this gender gap among lower-SES groups. In counties and months with more childcare facilities closure, the gender gap in child education time also increased more; yet this gendered effect is only evident among those with lower education and income. On average, for lower-SES respondents, a 10-percentage point increase in childcare closure expands the gender gap in daily child education time by over one hour. Considering that during the toughest times of the pandemic, in some counties over 40% of childcare facilities were closed, this impact is non-negligible.

Notably, childcare facility closure does not significantly impact men’s time on child education regardless of their education level or income. Conversely, it has a much greater impact on lower-SES women than higher-SES women. This finding may reflect that women with better socio-economic resources could seek alternative arrangements for child education during a crisis, while lower-SES women have to shoulder the additional childcare burden on their own. Additionally, given that informal childcare is still a major form of childcare arrangement in the USA (Laughlin, 2010) and that lower-SES families are more likely to rely on informal childcare (Susman-Stillman and Banghart, 2008), by only capturing the licensed childcare supply, this study may underestimate the impact of the pandemic. How the informal childcare supply has changed during the pandemic and impacted the gender division of childcare work for households with different SES backgrounds merits further research.

Our finding is also likely a result of the huge disparities in access to essential resources for home learning. Recent research shows that about one in 10 US children lacks adequate access to the internet and a computer for virtual learning, ranging from 3.9% for children of parents with graduate degrees to 20.3% for children of parents without high school diplomas (Friedman et al., 2021). High-SES families provide better access to online education, and therefore parents saved time on child education even when in-person childcare support became unavailable. In contrast, lower-SES families may be unable to afford a proper virtual-learning environment, and the burden of child education was shouldered by mothers when in-person facilities were closed.

This study is not without limitations. The SafeGraph dataset does not cover all US childcare facilities; it identifies childcare centers based on their online presence, and as Lee and Parolin (2021) have noted, we cannot estimate whether these facilities differ from those not in the sample. Still, this dataset is the broadest available regarding childcare facility closures during the pandemic. Furthermore, while the SafeGraph dataset contains information on 2228 US counties, the ATUS respondents in the analytical sample are only from a small sample of US counties. Nevertheless, our aim is not a representative estimation for all US counties. Instead, the counties are used as aggregation units to measure the percentages of closed childcare facilities. Moreover, the ATUS 2020 respondents have higher education and family income than respondents before 2020. Considering that lower-SES women are more severely impacted by the pandemic, our findings may underestimate the impact of childcare facility closure on the gendered childcare time division. Last, while our study adopts quasi-experimental approaches and attempts to control for confounding factors, it does not assert causation. Childcare facility closure could be associated with unobserved conditions across time and place, and readers should be cautious when interpreting the findings.

Despite limitations, this study deepens our understanding of how the COVID-19 pandemic and its response measures profoundly impacted gender inequality. It also adds to the theoretical and empirical literature on the gendered division of domestic labor. Our findings extend research on childcare access and cost to scenarios of public crisis when external childcare infrastructure suddenly retreated. We show the sudden loss of institutional childcare support exacerbates household gender inequality. Notably, the rising gender divide in childcare time may further impact women’s employment and well-being, contributing to a larger gender gap in unemployment and reduced paid work (Collins et al., 2020). Research also uncovers the deterioration of mental health for mothers of young children during the pandemic (Zamarro and Prados, 2021), and strong correlations between psychological distress and childcare time (Cheng et al., 2021). These facts suggest a potential setback in gender equality during the unprecedented times. It is therefore critical to understand the profound impact of the pandemic and provide adequate institutional support to accompany any policies leading to childcare infrastructure losses. This study calls for special policy support for women, especially lower-SES women, amid large-scale childcare facility and school closures during similar public crises.

Supplemental Material

sj-docx-1-soc-10.1177_00380385231224433 – Supplemental material for Childcare Facility Closure and Exacerbated Gender Inequality in Parenting Time during the COVID-19 Pandemic

Supplemental material, sj-docx-1-soc-10.1177_00380385231224433 for Childcare Facility Closure and Exacerbated Gender Inequality in Parenting Time during the COVID-19 Pandemic by Ran Liu and Siyun Gan in Sociology

Footnotes

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: the authors gratefully acknowledge support from the University of Wisconsin-Madison, Office of the Vice Chancellor for Research and Graduate Education with funding from the Wisconsin Alumni Research Foundation.

ORCID iD

Ran Liu

Supplemental material

Supplemental material for this article is available online.

Ran Liu is Assistant Professor of Educational Policy Studies at the University of Wisconsin-Madison. Her research focuses on gender values, inequalities, and stereotypes in education, households, and school-to-work linkages. She is also interested in applying innovative methods and data to provide new insights into sociological research. Her recent articles appear in European Sociological Review, Sex Roles, and Educational Researcher. She received her PhD in Sociology from the University of Pennsylvania. Before joining UW-Madison, she worked as a People Research Scientist at Meta to tackle workplace diversity issues.

Siyun Gan is a PhD student in the Department of Educational Policy Studies at the University of Wisconsin-Madison. Her current research focuses on the intersection of stratification beliefs, perceived competitiveness in educational environments, and their roles in the process of educational stratification. Additionally, Siyun is engaged in studying educational mobility patterns within East Asia as well as among Asian immigrants in the United States. She received her MA degree in International Education from New York University before joining UW-Madison.

References

Adams-Prassl

Boneva

Golin

, et al. (2020) Inequality in the impact of the Coronavirus shock: Evidence from real time surveys. Journal of Public Economics 189: 104245.

Aguiar

Hurst

(2007) Measuring trends in leisure: The allocation of time over five decades. The Quarterly Journal of Economics 122(3): 969–1006.

Andersen

(2018) Paternity leave and the motherhood penalty: New causal evidence. Journal of Marriage and Family 80(5): 1125–1143.

Augustine

Prickett

(2022) Gender disparities in increased parenting time during the COVID-19 pandemic: A research note. Demography 59(4): 1233–1247.

Bertrand

Kamenica

Pan

(2015) Gender identity and relative income within households. The Quarterly Journal of Economics 130(2): 571–614.

Bianchi

Milkie

Sayer

, et al. (2000) Is anyone doing the housework? Trends in the gender division of household labor. Social Forces 79(1): 191–228.

Boeckmann

Misra

Budig

(2015) Cultural and institutional factors shaping mothers’ employment and working hours in postindustrial countries. Social Forces 93(4): 1301–1333.

Brines

(1994) Economic dependency, gender, and the division of labor at home. American Journal of Sociology 100(3): 652–688.

Bureau of Labor Statistics (2022) American Time Use Survey user’s guide: Understanding ATUS 2003 to 2020. Available at: https://www.bls.gov/tus/atususersguide.pdf (accessed 13 December 2022).

10.

Callaway

Goodman-Bacon

Sant’Anna

PHC

(2021) Difference-in-differences with a continuous treatment. arXiv: 2107.02637.

11.

Chauhan

(2021) Gendering COVID-19: Impact of the pandemic on women’s burden of unpaid work in India. Gender Issues 38: 395–419.

12.

Cheng

Mendolia

Paloyo

, et al. (2021) Working parents, financial insecurity, and childcare: Mental health in the time of COVID-19 in the UK. Review of Economics of the Household 19(1): 123–144.

13.

Collins

Landivar

Ruppanner

, et al. (2020) COVID-19 and the gender gap in work hours. Gender, Work, and Organization 28(S1): 101–112.

14.

Collins

Ruppanner

Christin Landivar

, et al. (2021) The gendered consequences of a weak infrastructure of care: School reopening plans and parents’ employment during the COVID-19 pandemic. Gender & Society 35(2): 180–193.

15.

Craig

(2006) Parental education, time in paid work and time with children: An Australian time-diary analysis. The British Journal of Sociology 57(4): 553–575.

16.

Craig

(2007) How employed mothers in Australia find time for both market work and childcare. Lifestyles Family and Economic Issues 28(1): 69–87.

17.

Craig

Mullan

(2011) How mothers and fathers share childcare. American Sociological Review 76(6): 834–861.

18.

Del Boca

Oggero

Profeta

, et al. (2020) Women’s and men’s work, housework and childcare, before and during COVID-19. Review of Economics of the Household 18: 1001–1017.

19.

Domeij

Klein

(2013) Should day care be subsidized? The Review of Economic Studies 80(2): 568–595.

20.

Evertsson

Nermo

(2004) Dependence within families and the division of labor: Comparing Sweden and the United States. Journal of Marriage and Family 66(5): 1272–1286.

21.

Fan

Moen

(2022) Working more, less or the same during COVID-19? A mixed method, intersectional analysis of remote workers. Work and Occupations 49(2): 143–186.

22.

Farre

Fawaz

Gonzalez

, et al. (2022) Gender inequality in paid and unpaid work during COVID-19 times. The Review of Income and Wealth 68(2): 323–347.

23.

Friedman

York

Mokdad

, et al. (2021) U.S. children ‘learning online’ during COVID-19 without the internet or a computer: Visualizing the gradient by race/ethnicity and parental educational attainment. Socius: Sociological Research for a Dynamic World 7: 237802312199260.

24.

Gimenez-Nadal

Molina

(2013) Parents’ education as a determinant of educational childcare time. Journal of Population Economics 26(2): 719–749.

25.

Greenstein

(1996) Husbands’ participation in domestic labor: Interactive effects of wives’ and husbands’ gender ideologies. Journal of Marriage and the Family 58(3): 585.

26.

Gupta

(2007) Autonomy, dependence, or display? The relationship between married women’s earnings and housework. Journal of Marriage and Family 69(2): 399–417.

27.

Han

Waldfogel

(2001) Childcare costs and women’s employment: A comparison of single and married mothers with pre-school-aged children. Social Science Quarterly 82(3): 552–568.

28.

Hansen

Sabia

Schaller

(2022) Schools, Job Flexibility, and Married Women’s Labor Supply: Evidence From the COVID-19 Pandemic. Cambridge, MA: National Bureau of Economic Research.

29.

Heggeness

Fields

(2020) Parents juggle work and childcare during pandemic. Available at: https://www.census.gov/library/stories/2020/08/parents-juggle-work-and-child-care-during-pandemic.html (accessed 3 July 2022).

30.

Hiller

(1984) Power dependence and division of family work. Sex Roles 10(11–12): 1003–1019.

31.

Hook

(2010) Gender inequality in the welfare state: Sex segregation in housework, 1965–2003. American Journal of Sociology 115(5): 1480–1523.

32.

Johnston

Mohammed

van der Linden

(2020) Evidence of exacerbated gender inequality in childcare obligations in Canada and Australia during the COVID-19 pandemic. Politics & Gender 16(4): 1131–1141.

33.

Landivar

Ruppanner

Rouse

, et al. (2022a) Research note: School reopenings during the COVID-19 pandemic and implications for gender and racial equity. Demography 59(1): 1–12.

34.

Landivar

Ruppanner

Scarborough

(2021) Are states created equal? moving to a state with more expensive childcare reduces mothers’ odds of employment. Demography 58(2): 451–470.

35.

Landivar

Ruppanner

Scarborough

, et al. (2020) Early signs indicate that covid-19 is exacerbating gender inequality in the labor force. Socius: Sociological Research for a Dynamic World 6: 2378023120947997.

36.

Landivar

Scarborough

Collins

, et al. (2022b) Do high childcare costs and low access to Head Start and childcare subsidies limit mothers’ employment? A state-level analysis. Social Science Research 102: 102627.

37.

Laughlin

(2010) Who’s minding the kids? Childcare arrangements. US Census Bureau Current Population Reports, August 2010. Available at: https://www.census.gov/library/publications/2013/demo/p70-135.html (accessed 13 December 2022).

38.

Lee

Parolin

(2021) The care burden during covid-19: A national database of childcare closures in the United States. Socius: Sociological Research for a Dynamic World 7: 237802312110320.

39.

Musick

Meier

Flood

(2016) How parents fare. American Sociological Review 81(5): 1069–1095.

40.

Negraia

Augustine

Prickett

(2018) Gender disparities in parenting time across activities, child ages, and educational groups. Journal of Family Issues 39(11): 3006–3028.

41.

Obioma

Jaga

Raina

, et al. (2023) Gendered share of housework and the COVID-19 pandemic: Examining self-ratings and speculation of others in Germany, India, Nigeria, and South Africa. Journal of Social Issues 79(3): 907–934.

42.

Parolin

Lee

(2021) Large socio-economic, geographic and demographic disparities exist in exposure to school closures. Nature Human Behaviour 5(4): 522–528.

43.

Parry

Gordon

(2021) The shadow pandemic: Inequitable gendered impacts of COVID-19 in South Africa. Gender, Work & Organization 28(2): 795–806.

44.

Patnaik

(2019) Reserving time for daddy: The consequences of fathers’ quotas. Journal of Labor Economics 37(4): 1009–1059.

45.

Petts

Carlson

Pepin

(2021) A gendered pandemic: Childcare, homeschooling, and parents’ employment during COVID-19. Gender, Work & Organization 28(S2): 515–534.

46.

Poortman

A-R

van der Lippe

(2009) Attitudes toward housework and childcare and the gendered division of labor. Journal of Marriage and Family 71(3): 526–541.

47.

Ross

(1987) The division of labor at home. Social Forces 65(3): 816.

48.

Ruppanner

Collins

Landivar

, et al. (2021) Norms, childcare costs, and maternal employment. Gender & Society 35(6): 910–939.

49.

Ruppanner

Moller

Sayer

(2019) Expensive childcare and short school days = lower maternal employment and more time in childcare? Evidence from the American time use survey. Socius: Sociological Research for a Dynamic World 5: 1–14.

50.

SafeGraph (2022) SafeGraph’s response to congressional inquiry on user privacy. Available at: https://www.safegraph.com/blog/safegraph-response-to-congressional-inquiry-on-user-privacy (accessed 13 December 2022).

51.

Sayer

Gornick

(2012) Cross-national variation in the influence of employment hours on childcare time. European Sociological Review 28(4): 421–442.

52.

Schneider

(2011) Market earnings and household work: New tests of gender performance theory. Journal of Marriage and Family 73(4): 845–860.

53.

Sevilla

Smith

(2020) Baby steps: The gender division of childcare during the COVID-19 pandemic. Oxford Review of Economic Policy 36(S1): S169–S186.

54.

Susman-Stillman

Banghart

(2008) Demographics of family, friend, and neighbor childcare in the United States. National Center for Children in Poverty. https://www.nccp.org/publication/demographics-of-family-friend-and-neighbor-child-care-in-the-united-states/ (accessed 13 December 2022).

55.

Zamarro

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.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.18 MB