Examining disparities in work–family time-use patterns among workers with caregiving responsibilities in the context of COVID-19

The COVID-19 pandemic fundamentally disrupted the lives of American working families, forcing millions of families to navigate unprecedented challenges in balancing paid work and unpaid caregiving responsibilities. School and daycare closures sharply increased caregiving demands, and unemployment soared, peaking at 14.8 percent in April 2020.¹ For those who were able to keep their jobs, work arrangements shifted considerably during the lockdowns. The work-from-home rate surged to a peak of 38.1 percent in 2021, up from 23.7 percent in the prepandemic years of 2018 and 2019.² Together, these changes altered not only how much time families devoted to paid work and caregiving but also altered how those responsibilities were distributed across the day.

These disruptions did not affect all workers equally. Women disproportionately shouldered increased caregiving burdens during the pandemic, with numerous studies documenting that mothers spent substantially more time on childcare than fathers.³ Racial and ethnic minority workers were more likely to be employed in frontline and inflexible jobs with limited access to remote work and schedule control.⁴ Occupational differences in job flexibility and stability further shaped workers’ ability to adapt to pandemic-related disruptions. As a result, the pandemic raised concerns about whether existing disparities in work–family time use, categorized by sex, race and ethnicity, and occupation, widened, narrowed, or persisted in the years that followed.

Most existing research on work–family time use during the pandemic has focused on its immediate impacts during the early lockdown period and has emphasized changes in total time spent on paid work and caregiving.⁵ Less is known about how these patterns evolved beyond 2020, particularly during the reopening and postpandemic period. In addition, many studies have focused primarily on how much time is spent on work and caregiving, which reflects the overall magnitude or scope of dual demands.⁶ Far less attention has been paid to changes in the structure of daily time use, such as the extent to which paid work and caregiving time are divided into multiple episodes or the share of paid work time that is spent multitasking. Yet prior research shows that these dimensions are central to understanding work–family strain and changing working conditions.⁷

This article addresses these research gaps by examining work–family time-use patterns from 2018 to 2023 using nationally representative data from the Integrated Public Use Microdata Series (IPUMS) of the American Time Use Survey (ATUS).⁸ To capture both the amount and organization of work and caregiving, the analysis focuses on three dimensions of daily time use: total work–family burden (total time spent on paid work and caregiving), time fragmentation (the number of work and caregiving episodes across the day), and multitasking (providing caregiving while engaged in paid work). This article documents overall trends across the prepandemic, early pandemic, and postpandemic periods and assesses whether patterns differ by sex, race and ethnicity, and occupation. This article asks the following questions:

1. How did overall levels of total work–family burden, fragmentation, and multitasking change over this period?

2. Did these patterns differ for men and women, and, within each sex, did racial and ethnic and occupational gaps in work–family time use widen, narrow, or remain stable over time?

The results show that the total work–family burden remained relatively stable across the study period, despite major disruptions to employment and caregiving arrangements. In contrast, the structure of daily time use changed substantially. Time fragmentation declined sharply in 2020 and remained below prepandemic levels through 2023, indicating a sustained shift toward more consolidated daily schedules. Multitasking increased markedly during the early pandemic years but returned to approximate prepandemic levels by 2022–23.

Across all years, women experienced consistently higher levels of fragmentation and multitasking than men, even when total time commitments were similar. Racial and ethnic disparities varied by outcome and sex. Differences in total work–family burden narrowed between Black men and White men before temporarily reversing during the early pandemic years. Among women, multitasking increased sharply among Black women during the pandemic period. Occupational differences were comparatively modest and largely explained by compositional factors such as education, work-from-home access, and caregiving intensity.

Together, these findings indicate that work–family inequality during the pandemic was reflected not only in total time demands but also, and more persistently, in the organization of work and caregiving time. The results highlight the value of time-use data for capturing how the pandemic shaped everyday constraints among U.S. working families.

Data, methods, and sample

This article uses data from the 2018–23 ATUS, obtained from the IPUMS-ATUS data extract system.⁹ ATUS is a nationally representative, federally administered time-diary survey that collects detailed information on individuals’ primary and secondary activities over a 24-hour period. The survey covers the civilian noninstitutional population 15 years of age and older in the United States, with an average annual sample size of approximately 8,000 to 9,000 respondents.¹⁰ This article uses the cross-sectional survey to enable trend analysis of work–family time-use patterns in the context of the COVID-19 pandemic.

The analytic sample is restricted to employed adults 18 to 64 years of age who reported caregiving responsibilities for children 17 years of age or younger and adults 65 years of age and older. Because ATUS does not indicate whether the diary day reflects a respondent’s typical workday, both weekdays and weekends are included. The sample is restricted to respondents who reported at least 1 hour of paid work on the diary day to ensure that the analysis focuses on those likely to face dual demands in balancing paid work and family responsibilities and to reduce bias from differences in employment or caregiving status. Individuals with more than 3 hours of missing diary time are excluded. The final analytic sample consists of 7,863 respondents.

Because the analytic sample is restricted to employed individuals, changes in employment over time may affect sample composition. Among individuals 18 to 64 years of age, employment declined from 74.5 percent in 2019 to 70.6 percent in 2020 and recovered in subsequent years, reaching 77.4 percent in 2023. The decline was larger among women than men (from 68.3 percent to 65.1 percent for women, compared with 81.0 percent to 76.2 percent for men from 2019 to 2020). Differences were also evident across racial and ethnic groups, with larger declines among Black and Hispanic individuals, particularly among women. For example, employment among Black women declined from 70.7 percent in 2019 to 60.2 percent in 2020, while Hispanic women experienced a decline from 61.5 percent to 60.7 percent. The occupational composition of employed workers also shifted during this period. Among employed men, the share in professional and managerial occupations rose from 39.0 percent in 2019 to 42.3 percent in 2020, and among employed women from 47.7 percent to 50.6 percent, reflecting disproportionately larger employment losses in service, sales, and other occupations. These patterns indicate that the composition of the analytic sample may shift over time, particularly during the pandemic period.¹¹

Data collection for ATUS was temporarily suspended from March 18 to May 9, 2020, resulting in partial-year data for 2020. Following guidance from the U.S. Bureau of Labor Statistics, this article uses data collected from May 10 through December 31, 2020, and full-year data for all other years. Because of the disruption in data collection, ATUS sample weights for 2020 were adjusted to reflect only the days when data were collected, totaling 313 days instead of 366, and used a revised weighting method (the ATUS 2020 final weight, TU20FWGT).¹² As a result, estimates for 2020 are not directly comparable with those from other years and should be interpreted with caution.¹³ 

Tables 1, 2, 3, and 4 report the following descriptive statistics: total burden, fragmentation, and multitasking; the mean daily minutes spent in total burden and its components (paid work, caregiving, and household activities); the average number of daily episodes for fragmentation and its components; the distribution of fragmentation across categories of the number of daily episodes; and the distribution and mean levels of multitasking for the full sample and among multitaskers.

Table 5 summarizes sample characteristics by sex. Survey weights were applied in all analyses to produce nationally representative estimates. Men account for 53.1 percent of the sample and women for 46.9 percent. The average age is similar across sexes (42.26 years of age for men and 41.78 years of age for women). Racial and ethnic compositions are broadly comparable, although women are slightly more likely to be non-Hispanic Black. Educational attainment differs modestly by sex, with women more likely to hold a bachelor’s degree or higher (51.3 percent for women, compared with 44.1 percent for men), and men are more concentrated in service, sales, and other occupations (52.9 percent for men, compared with 43.7 percent for women).

Family characteristics also vary by sex. Women are more likely to have no partner (28.5 percent, compared with 15.0 percent) and, conditional on having a partner, are more likely to have an employed partner. Women are also somewhat more likely to provide eldercare and slightly less likely to have very young children (0 to 5 years of age) in the household.

In terms of work and family time use, total burden is similar across the sexes (approximately 643 minutes per day). However, the composition differs between the sexes: men spend more time in paid work, while women spend more time in caregiving and household activities. Women also experience higher levels of fragmentation (10.51 episodes, compared with 7.77 episodes per day) and multitasking (10.40 percent, compared with 5.32 percent of paid work time). Consistent with this, a larger share of women is in the highest fragmentation category (the 42.2 percent who report 10 or more episodes) and in the highest multitasking category. These descriptive patterns are consistent with the regression-adjusted results presented in this article.

Table 6 presents descriptive differences in work–family time use by sex, race, and ethnicity and by sex and occupation. Across all racial and ethnic groups, women exhibit higher levels of fragmentation and multitasking than men, although total work–family burden is broadly similar. Fragmentation is particularly high among women across groups, especially among non-Hispanic White and Hispanic women. Multitasking also varies substantially, with the highest levels observed among men and among women in the “other” racial and ethnic category.

Differences by occupation show a similar pattern for different sexes. Within both professional and managerial occupations and service, sales, and other occupations, women experience more fragmented daily schedules and higher levels of multitasking than men, while total burden remains comparable across groups. Notably, multitasking is highest among women in professional and managerial occupations.

These descriptive patterns provide context for the interaction analyses reported in the regression models. However, table 6 indicates that some subgroup categories, particularly for racial and ethnic minority groups, are based on relatively small sample sizes. This contributes to larger standard errors in interaction models, which helps explain why several interaction terms are not statistically significant despite sizable coefficient estimates.

Regression variables

This section of the article discusses the variables used in the regressions. The following dependent variables are used: total work–family burden, fragmentation of work and care activities, and multitasking. The independent variables used are sex, race and ethnicity, and occupational status. This section ends with a discussion of the covariates.

Dependent variables

Three measures are used to capture work–family time-use patterns.¹⁴ First, total work–family burden (referred to as total burden afterward) is measured as the total number of minutes per day spent in primary paid work and primary caregiving activities, including both childcare and eldercare. This measure reflects the overall time demands associated with combining work and family responsibilities. Consistent with prior research, time spent traveling for work, caregiving, and household activities is included. Secondary caregiving is not included because it occurs simultaneously with primary paid work activity and would otherwise result in double counting of time.¹⁵

Second, fragmentation of work and care activities (referred to as fragmentation afterward) is measured as the total number of paid work and caregiving episodes across the diary day. This combined measure captures the overall division of time across work and family responsibilities. Higher values indicate more frequent interruptions in work and caregiving activities, capturing the extent to which work and caregiving time are divided into shorter, discrete segments rather than concentrated in longer, uninterrupted blocks.

Consistent with the definition of total burden, fragmentation is based on primary activity episodes and excludes secondary caregiving, which occurs simultaneously with primary activities and does not constitute a separate episode. Because ATUS work activities are often recorded as relatively large time blocks, short interruptions of less than 15 minutes may not be captured; therefore, the fragmentation measure may understate the true extent of fragmentation.

Fragmentation may affect paid work and caregiving through distinct mechanisms. In paid work, many tasks require sustained attention and continuity, such that interruptions impose cognitive switching costs, reduce efficiency, and lengthen task completion time. Fragmented work schedules may therefore lower productivity and increase work-related strain. In contrast, caregiving activities are inherently more episodic, consisting of discrete tasks such as feeding, supervising, and bathing. As a result, fragmentation may be less detrimental to caregiving productivity. However, fragmentation can still affect caregiving by increasing time pressure, reducing the ability to respond consistently to needs, and contributing to stress. To provide additional descriptive context, the distributions of work fragmentation (the number of paid work episodes) and caregiving fragmentation (the number of caregiving and household episodes) are also reported separately in the descriptive tables. The regression analyses use the combined fragmentation measure to capture overall fragmentation and simplify interpretation.

Third, multitasking is measured as the percentage of paid work time during which caregiving occurs as a secondary activity, capturing the degree to which caregiving responsibilities overlap with paid work rather than occurring at separate times. This measure is constructed using the ATUS secondary childcare measure and a constructed measure of secondary eldercare during paid work episodes. While ATUS does not collect a general set of secondary activities except secondary childcare, it does include measures of secondary eldercare that capture time spent providing care to adults during other primary activities. While secondary childcare is collected from all respondents, secondary eldercare is collected only from respondents who report providing eldercare in the prior months. As a result, secondary eldercare is conditionally observed and reflects multitasking among respondents identified as caregivers. Thus, the measure may not capture infrequent caregiving episodes occurring on the diary day. In addition, because ATUS records secondary childcare only for children under age 13, multitasking involving older children cannot be directly observed. However, caregiving for younger children is most likely to require continuous supervision and, therefore, to coincide with paid work.

Independent variables

The primary predictors include year (2018, 2019, 2020, 2021, 2022, and 2023), sex (male or female), race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or other, which includes Asian American and Pacific Islander, Native American, and two or more races), and occupation status (sales, service, and other or professional and managerial).¹⁶

Covariates

Control variables include the following: age (a continuous variable), education level (high school or less, associate’s degree or some college, or college graduate or more), spousal or partner employment status (categorized as having no spouse or partner, a spouse or partner not working, or a spouse or partner working), the number of an individual’s own children under 18 in the household (categorized as 0, 1, 2, and 3 or more), the number of elderly individuals needing care (categorized as 0, 1, and 2 or more), the age of the youngest child (categorized as 5 years of age or under, 6 to 17 years of age, and adult or none), working from home (binary variable), day of the diary day (weekday or weekend), and state of residence.

Methodology 

Regression models are used to estimate changes in work–family time use and differences across population groups. Models are based on the following distributional characteristics: (1) an ordinary least squares (OLS) linear regression is used for total work–family burden (continuous), (2) a negative binomial regression is used for fragmentation (nonnegative count of daily episodes and exhibits overdispersion), and (3) a Tobit regression is used for multitasking (continuous but a large share of zero values).¹⁷ In light of prior research on time-diary data, robustness checks are conducted using alternative estimators, including OLS and a two-part model that separately estimates the probability of engaging in multitasking and the intensity of multitasking among those who report any multitasking.¹⁸ The substantive conclusions are consistent across model specifications.

All models include a series of year indicator variables (2018–23) to account for time-specific variation. Differences by sex, race and ethnicity, and occupation are examined using models that include interactions between year and each group indicator. All reported estimates are derived from fully adjusted models that account for demographic, family, and job characteristics; results from unadjusted models show similar patterns and are not shown.

To assess racial and ethnic and occupational differences, analyses are conducted separately for men and women. This stratified approach allows for the examination of whether patterns of work–family time use by race and ethnicity and occupation differ by sex rather than assuming uniform effects across the sexes. Prior research shows that work and caregiving responsibilities are highly specific to an individual’s sex and that racial, ethnic, and occupational disparities in time use may operate differently for men and women.

Results

This section describes regression-adjusted patterns of work–family time use from 2018 to 2023. First, overall trends are presented. Second, differences in work–family time-use patterns between the sexes are considered. Third, differences for racial, ethnic, and occupational disparities are estimated separately for men and women. Predicted values are shown in the charts, with corresponding regression estimates reported in the tables.

Overall trends in work–family time-use patterns, 2018–23

Chart 1 presents regression-adjusted predicted values of total burden, fragmentation, and multitasking from 2018 to 2023. After adjusting for demographic, family, and job characteristics, total burden shows little change over the study period, ranging from roughly 640 to 650 minutes per day in the prepandemic years, and shows no statistically meaningful departures during the pandemic or reopening years. Predicted levels from 2020 to 2023 differed from 2018 by less than 10 minutes per day, indicating that the combined time devoted to paid work and caregiving neither substantially increased nor decreased despite major disruptions to work and family life.

In contrast, the structure of daily time use changed more noticeably. Fragmentation declined sharply in 2020, with the predicted number of work and caregiving episodes falling by approximately 20 percent relative to 2018. Although fragmentation increased somewhat after 2020, predicted levels from 2021 to 2023 remained 8 percent to 12 percent below prepandemic levels, suggesting a persistent shift toward more consolidated daily schedules.

Multitasking during paid work followed a different trajectory. The predicted share of paid work time spent multitasking with caregiving increased by roughly 4 to 5 percentage points in 2020 relative to 2018; among workers who multitasked at all, the share rose by roughly 30 percentage points. From 2022 to 2023, multitasking levels had moved closer to prepandemic values, but they remained modestly higher than in 2018. Taken together, these results indicate that, while total work–family time commitments remained largely unchanged, the organization of work and caregiving time shifted substantially during the pandemic, with some structural changes persisting into the postpandemic period. Table 7 reports the corresponding regression estimates.

Differences by sex in work–family time-use patterns, 2018–23

Chart 2 shows regression-adjusted predicted values of work–family time use by sex from 2018 to 2023, and, again, table 7 reports the corresponding regression estimates. Consistent with the overall trends, predicted total burden for men and women remained similar throughout the period, averaging approximately from 640 to 655 minutes per day for both sexes. No substantial differences by sex are observed in total burden in the prepandemic years, and the predicted values do not indicate the emergence or widening of a gap between the sexes in total work–family time use during the pandemic or postpandemic years.

Differences by sex are more pronounced in the structure of daily time use. Across all years, women exhibited higher levels of fragmentation than men, averaging roughly 8.5 to 11.5 daily work and caregiving episodes, compared with about 6.5 to 8.5 episodes among men. Although fragmentation declined sharply in 2020 for both sexes, falling by about 2 episodes per day, the gap between the sexes of approximately 2 to 3 episodes persisted through 2023, and there was no evidence of convergence.

A similar pattern is observed for multitasking during paid work. Across all model specifications, predicted multitasking levels were consistently higher for women than for men in every year. Two-part model results indicate that differences by sex in multitasking reflect both higher participation and greater intensity among women. Prior to the pandemic, women spent approximately 11 percent to 12 percent of paid work time multitasking with caregiving, compared with about 6 percent among men. Multitasking increased sharply in 2020 for both sexes, peaking at roughly 17 percent for women and 10 percent for men, before declining toward prepandemic levels in subsequent years. Despite these shifts, the gap between the sexes remained relatively stable over time.

Taken together, these results suggest that disparities between the sexes in work–family time use are concentrated in how time is organized rather than in the total amount of time spent on paid work and caregiving. Pandemic-related shifts affected men and women in broadly similar ways, without substantially narrowing or widening existing differences between the sexes in time fragmentation or multitasking.

Racial and ethnic differences in work–family time-use patterns, by sex, 2018–23

Chart 3 presents regression-adjusted predicted trends of work–family time use by race and ethnicity that are estimated separately for men and women. Racial and ethnic differences in total burden were modest in 2018 and remained relatively stable for women throughout the study period. Across all years, predicted total burden for women in different racial and ethnic groups clustered closely, generally ranging from approximately 615 to 675 minutes per day, and there were no statistically significant differences across groups. Among men, racial differences in total burden varied over time. In 2018, Black men spent substantially less time in paid work and caregiving than White men, with predicted values of roughly 580 minutes per day for Black men, compared with about 650 minutes for White men. This difference narrowed sharply by 2019, and, in some subsequent years, Black men exhibited levels comparable with or exceeding those of White men. As shown in chart 3, predicted total burden among Black men increased in 2019 and 2021 to levels comparable with, and in some years exceeding, those of White men, indicating a temporary convergence and partial reversal of the Black–White gap. By 2022 and 2023, this convergence attenuated and predicted differences returned toward prepandemic levels.

Chart 3 shows a sharp decline in fragmentation in 2020 for both men and women across all racial and ethnic groups, indicating fewer daily transitions between paid work and caregiving during the first year of the pandemic. Among men, predicted fragmentation fell from approximately 8 to 9 episodes per day in 2018 and 2019 to about 5.5 to 7 episodes in 2020, and fragmentation partially rebounded to about 6.5 to 8 episodes by 2023. Among women, fragmentation declined from approximately 10 to 12 episodes before the pandemic to about 7 to 9 episodes in 2020, with partial recovery thereafter. Racial and ethnic differences in fragmentation were generally small, typically within half an episode per day, and remained relatively stable over time for both sexes.

In contrast, multitasking exhibited pronounced racial and ethnic divergence among women. In 2018, Black and Hispanic women engaged in substantially less multitasking than White women, with predicted multitasking levels of approximately 3 percent to 6 percent of paid work time among Black and Hispanic women, compared with about 14 percent among White women. Table 8 indicates that these disparities are primarily driven by differences in the likelihood of engaging in multitasking rather than differences in intensity, conditional on multitasking. Table 8 reports the corresponding regression estimates.

This pattern shifted sharply during the pandemic. As shown in chart 3, multitasking among Black women increased dramatically in 2020 and 2021, reaching levels of approximately 22 to 23 percent, comparable with or exceeding those of White women. Table 8 suggests that this increase reflects both a rise in the likelihood of multitasking and, to a lesser extent, increases in multitasking intensity. Although multitasking declined after 2021, predicted levels among Black women remained elevated through 2023 relative to prepandemic differences. Among men, multitasking rose in 2020 for White, Hispanic, and other-race men, from roughly 3.5 to 6.5 percent to roughly 9 to 13 percent, while among Black men it fell from 7.1 percent to 4.8 percent. Apart from the decline among Black men, racial and ethnic differences among men were generally not statistically significant across specifications.

Taken together, these patterns indicate that the pandemic was associated with broad declines in fragmentation across racial and ethnic groups, a temporary convergence in total work–family burden among men, and a marked intensification of multitasking among Black women. These results underscore the sex-specific and racialized nature of changes in work–family time use during and after the pandemic.

Occupational differences in work–family time-use patterns, by sex, 2018–23

Chart 4 shows regression-adjusted predicted values of work–family time use by occupation, estimated separately for men and women. Occupational differences in total burden were modest for both men and women. In 2018, predicted total burden differed by fewer than 10 minutes per day between professional and managerial workers and those in service, sales, and other occupations for both men and women, and these differences were not statistically significant. Across subsequent years, occupation-specific differences in total burden remained small and showed no consistent widening during the pandemic.

Chart 4 shows a pronounced decline in fragmentation in 2020 for both men and women across occupational groups, reflecting fewer daily transitions between paid work and caregiving during the first year of the pandemic. For both sexes in both occupational groups, fragmentation fell substantially in 2020 relative to 2018 and remained below prepandemic levels through 2023, recovering only partially. Occupational differences in fragmentation were generally small. For men, fragmentation was similar for workers in both professional and managerial occupations and service, sales, and other occupations. Among women, fragmentation was modestly higher among professional and managerial workers in some years, but these differences were not statistically significant except in 2021, indicating that fragmentation declined more among professional workers relative to service, sales, and other workers in that year. Table 9 reports the corresponding regression estimates, and occupational patterns in multitasking are shown in chart 4.

In 2018, before the pandemic, multitasking levels did not differ significantly by occupation for either men or women. During the pandemic, multitasking increased temporarily in 2020 across both occupational groups and sexes, before returning to prepandemic levels. Chart 4 suggests larger differences in multitasking between professional and managerial workers and those in service, sales, and other occupations, particularly among women during 2020 and 2021, but these differences decline in subsequent years.

The magnitude and statistical significance of occupational differences in multitasking are broadly consistent across model specifications, with limited evidence of systematic disparities. Among women, professional and managerial workers exhibit higher multitasking levels in 2020 in OLS models and in the intensive margin of the two-part model, although this pattern is not observed in the Tobit specification. In contrast, the extensive margin (the likelihood of multitasking) shows little occupational difference for either sex. Among men, occupational differences in multitasking are generally small and statistically insignificant across specifications.

Supplemental analyses

Supplementary analyses suggest that much of the observed occupational disparity in multitasking is associated with compositional differences. Chart 5 presents a stepwise decomposition of occupational differences in multitasking by sequentially adding groups of covariates to the model. Chart 5 illustrates how the coefficient for professional and managerial occupations changes as demographic, family-related, and work-related characteristics are introduced. Results indicate that differences in education and work-from-home arrangements account for most occupational differences in multitasking for both men and women. Caregiving intensity, including the number of an individual’s own children who are in the household and the number of elderly needing care, further attenuates remaining differences only among women, indicating that family demands play a distinct role in shaping women’s multitasking patterns. Because the magnitude of attenuation may vary with alternative ordering of covariates, these results should be interpreted as descriptive. However, the overall pattern remains consistent across alternative specifications.

Overall, occupational differences in work–family time-use patterns were modest for both men and women over the 2018–23 period. Total work–family burden varied little by occupation and remained relatively stable over time. Fragmentation declined sharply in 2020 across occupational groups for both sexes and recovered only partially in subsequent years, with limited and short-lived occupational differences, particularly among women. Multitasking increased temporarily during the early period of the pandemic for both occupational groups and sexes but returned to prepandemic levels thereafter, and there were no statistically significant occupational disparities in fully adjusted models.

Across analyses by sex, race and ethnicity, and occupation, regression-adjusted results consistently show that the COVID-19 pandemic was associated with changes in the structure of daily work–family time use, while total work–family burden remained largely stable. Differences across population groups persisted over time, indicating continuity in underlying patterns of inequality in how paid work and caregiving are organized.

Summary and discussion