The Social Division of Care Work Time Over Half a Century Open Access

Abstract

This study introduces a demographic framework to analyze the social division of care work time, defined as the sum of paid and unpaid care work time provided to children and adults in a population. Combining data from the American Heritage Time Use Survey (AHTUS) and the Current Population Survey Annual Social and Economic Supplement (CPS-ASEC), we focus on routine interactive care and analyze how the volume and social division of this care work has evolved in the United States over a half century (1965–2018). Results reveal relative stability in the division of care work across domains (paid vs. unpaid and child vs. adult) but substantial change across social groups (by gender and race). The share of total care work provided by paid caregivers remained stable, challenging expectations about defamilialization, whereas the share of total care work going to adults increased over time. Gender and race inequality in total care work time experienced notable declines. Analyses show that these changes are driven by men's increased involvement in unpaid childcare and non-White women's declined involvement in some paid care jobs, respectively. Our framework provides new tools to examine how demographic, social, and economic changes impact the social organization of care work time.

Introduction

Observers are increasingly recognizing that the work of providing routine care to children and adults is undergoing important transformations. This care work involves interactive, support, and supervisory activities, such as feeding, teaching, cooking meals, or being on call to provide care if needed.¹ Researchers have often argued that this work is increasingly provided by paid caregivers instead of unpaid family caregivers, that population aging is shifting the kinds of care work that will be needed, or that shifts in social stratification are changing the extent to which women and particularly non-White women are overrepresented among care providers (Duffy 2011; Folbre 2001, 2012; Fraser 2016; Glenn 2010). For instance, Folbre (2012:41) claimed that “the combination of fertility decline and the aging of the population, and health trends suggest that the need for unpaid care of adults will increase relative to unpaid care of children.”

However, evidence supporting these claims about social change remains inconclusive. Current research sheds light on shifts within specific domains of care work—in paid care, unpaid childcare, or unpaid adult care—but we lack a unified analysis of the collective impact of these changes on the social division of care work. For instance, arguments highlighting the growing significance of paid care hinge on data showing an increase in paid care jobs (Duffy 2011; Dwyer 2013; Hartmann et al. 2018). Yet, the expansion of such jobs does not inherently imply greater importance of paid care over unpaid care: unpaid care might rise alongside paid care. Similarly, men's increased involvement in unpaid childcare (Altintas and Sullivan 2017; Craig and Mullan 2011; Hook 2006; Raley et al. 2012; Sayer et al. 2004) does not necessarily signal a decline in women's overrepresentation in care work activities, given that paid care, which is on the rise, continues to be female-dominated (Duffy 2011; Folbre 2012; International Labour Organization 2018). Claims regarding the impacts of population aging or shifts in racial stratification encounter similar limitations.

Determining whether and how changes in one domain of care impact the general social division of care work activities requires an integrated analysis of the different domains of care work, but few scholars have bridged care domains. Folbre (e.g., 2001, 2012, 2021) has most significantly contributed to a unified approach. Her 2012 book proposes the concept of care sector to bring together all care activities of paid and unpaid caregivers. This concept has motivated valuable parallel analyses of paid and unpaid care provision, but it has not yet been used explicitly to quantitatively analyze changes in the social division of care work. Other studies have empirically merged only some domains of care work. Ophir, for instance, developed the concept of care life expectancy to analyze unpaid caregiving for children and adults over the life course (Ophir 2022; Ophir and Polos 2022), but these analyses did not include paid care work.

In this study, we introduce a demographic framework to examine changes in the social division of care work. Our approach combines employment and time-use survey data to estimate the quantity of total care work time, defined as the sum of time spent providing care to children and adults by all paid and unpaid caregivers in a population on an average day. We propose classic demographic methods to describe how total care work time is distributed across domains (paid vs. unpaid and child vs. adult care) and across social groups (by gender and race). Implementing this approach necessitates delineating a cohesive set of care activities that can be consistently identified and quantified over time across paid and unpaid care work domains. As a result, our main analysis focuses on a subset of care activities: routine interactive care performed with children aged 0–12 and with adults of any age who require support with activities of daily living.² Because this subset of care activities misses important aspects of care provision, such as supervisory care, results from the main analysis do not reflect trends for other types of care work. Supplementary analyses using alternative, broader definitions of care work are discussed in the Results section. For simplicity, we use the shorthand “care work” to refer to routine interactive care work and differentiate between the focus in the main analysis and broader definitions of care work whenever necessary.

Leveraging data from the American Heritage Time Use Survey (AHTUS) and the Current Population Survey Annual Social and Economic Supplement (CPS-ASEC), we provide an integrated demographic analysis of the social division of care work time in the United States over half a century. We find that time devoted to interactive routine care work has increased but that the distribution of care work time across domains has been relatively stable. Unpaid childcare consistently constitutes the largest domain, followed by unpaid adult care, paid childcare, and paid adult care. However, this stability hides important changes, such as increases in paid childcare and unpaid adult care and substantial declines in gender and racial inequality. Notably, the overrepresentation of women in routine interactive care work time has nearly halved, largely driven by increases in men's unpaid childcare. The overrepresentation of non-White women has also declined, chiefly because of their decreased participation in some paid care jobs.

The Changing Social Division of Care Work in the United States

At the turn of the twentieth century, the male breadwinner–female housewife paradigm assigned women the duty of fulfilling the care needs of children and adult family members (Glenn 2010). Women's unpaid labor was the primary mode of care provision, given that access to paid care services was uncommon except among affluent families who hired domestic workers (Duffy 2011; Glenn 1992; Rollins 1985; Romero 1992). The social obligation to provide care applied to all women, but racial and economic hierarchies heavily shaped the types of care work women provided. Systemic racism and class oppression coerced non-White women and White working-class women to provide paid care for White affluent families, limiting and devaluing the unpaid care they provided for their own families and communities (Davis 1981; Glenn 2010; Nadasen 2012; Romero and Pérez 2016). Devotion to unpaid care work was a cornerstone of White femininity and was reinforced by social institutions that restricted engagement in employment (Folbre 1994). This social paradigm was characterized by a demographic regime of high fertility and relatively shorter lifespans, which emphasized childcare demands relative to adult care (Glenn 2010).

The latter half of the twentieth century witnessed substantial social, cultural, economic, and demographic transformations, prompting scholars to suggest that the social division of care work was shifting. Three shifts in this reconfiguration are frequently mentioned: (1) the defamilialization of care work, resulting from the transfer of care work from unpaid family members to paid caregivers; (2) the heightened significance of care for adults, considering the context of aging populations and declining birth rates; and (3) decreasing gender and racial inequality in care work involvement owing to the expansion of economic opportunities. As noted earlier, these claims of social change evoke the overall division of care work, but thus far evidence has solely analyzed isolated domains of care. Consequently, the claims lack robust substantiation.

The Defamilialization of Care Work

Defamilialization refers to the extent to which individuals can uphold adequate living standards independent of family relationships (Lister 1994; Orloff 2009). Applied to care work, defamilialization means reducing individuals’ reliance on family members to obtain care support and expanding the availability of paid caregiver services (Estévez-Abe and Naldini 2016; Lewis and Giullari 2005). Defamilialization can be driven by multiple factors, including economic dynamics influencing paid care supply and demand and social policies promoting public or privately funded paid care services (Leitner 2003; Saraceno 2016).³

In the United States, claims about the defamilialization of care work often highlight the remarkable growth of care jobs (Buera and Kaboski 2012; Duffy 2011; Dwyer 2013; Hartmann et al. 2018; U.S. Bureau of Labor Statistics 2024). The expansion of paid care provision is typically associated with the transformation of women's economic and family roles (Duffy 2011). Women have become more integrated into the workforce, reducing their availability for unpaid care work and prompting families to seek paid alternatives. Other processes have also contributed to this trend: growing inequality and international migration have increased the availability of paid caregivers (Cortés and Tessada 2011; Duffy 2020; Milkman et al. 1998), and increased residential mobility has further reduced the ability to rely on unpaid caregivers (Bea and Chy 2022; Glasgow 2000; Margolis and Verdery 2019). Policy changes have also incentivized paid care provision: Medicaid's expansion of home- and community-based care has increased access to paid long-term care home services (Murray et al. 2021; Watts et al. 2020), and programs such as Head Start, childcare subsidies, and childcare tax credits expand access to paid childcare services (Chaudry et al. 2017).

However, evidence also suggests that the transition from unpaid to paid caregiving might not be as extensive as anticipated. For instance, the rise of intensive mothering⁴ has amplified social expectations regarding mothers’ involvement in unpaid childcare, increasing maternal investments in childcare alongside growing engagement in paid work (e.g., Bianchi 2000; Sayer et al. 2004). This trend is noticeable even among parents who purchase childcare services, indicating that paid and unpaid childcare might function complementarily rather than substitute for each other (Folbre and Bittman 2004). Additionally, significant barriers persist in accessing paid care services. The lack of large-scale public investment in care services has resulted in a private paid care industry that is prohibitively expensive for most of those who need it and that creates precarious and low-paying jobs, affecting workforce recruitment and the quality of care (Kashen and Novello 2021; Poo 2016). Paid long-term care remains unaffordable for most families not qualified for Medicaid, leading them to rely on unpaid caregivers or leave care needs unmet (Brown and Finkelstein 2011; Johnson 2016; Kaye et al. 2010). Similarly, the affordability of paid childcare remains limited, with childcare subsidies reaching only a minority of eligible families (Landivar et al. 2023; Ullrich et al. 2019).

Population Aging

Claims that population aging shifts the distribution of care work toward adult care have typically showcased the growth of the older adult population and their associated care needs (Bloom et al. 2015; Christensen et al. 2009; Ophir and Polos 2022). Between 1960 and 2020, the proportion of the U.S. population older than 65 surged from 9% to nearly 17%, and projections indicate that its share might constitute almost 25% by 2050. In contrast, the percentage of children younger than 18 decreased from 36% in 1960 to 22% in 2020 (U.S. Census Bureau 2024). The combination of fertility decline and the increase in the older population, along with the rise in dementia and other cognitive conditions (Fishman 2017; Larson and Kanga 2008), suggests that adult care might represent a growing share of overall care work (e.g., Folbre 2012; Ophir 2022; Ophir and Polos 2022).

However, the impact of these demographic shifts remains ambiguous. Increases in longevity have also led to longer healthy lifespans (Banks et al. 2016; Beltrán-Sánchez et al. 2015: figure 1; Dosman et al. 2006; Lee 2014), indicating that a larger older-adult population does not necessarily equate to a proportionate rise in long-term care needs. Simultaneously, estimates reveal that nearly half of older adults with care needs experience some unmet need (Beach and Schulz 2016; Gibson et al. 2006; Patterson 2022; Potter 2019; Rahman et al. 2022). The pattern of unmet needs aligns with the trend of older adults today being more likely to live alone and experience social isolation (Anderson and Hussey 2000; Cudjoe et al. 2020). Ophir (2022) found that European women aged 50 or older spend a similar number of years providing unpaid child and unpaid adult care, indicating the continuing significance of childcare in a context of population aging.

Gender Inequality in Care Work

Scholars have widely acknowledged the persistent feminization of care work while also identifying important shifts in men's engagement in care provision. Notably, research indicates significant changes in residential and nonresidential fathers’ involvement in unpaid childcare (Altintas and Sullivan 2017; Berger and Langton 2011; Craig and Mullan 2011; Hook 2006; Kim and Meyer 2014; Raley et al. 2012; Sayer et al. 2004). For instance, Altintas and Sullivan (2017) revealed a nearly quadrupled increase in fathers’ involvement since the 1970s across the United States, Canada, the United Kingdom, and Australia. Factors such as women's employment, evolving gender norms, and changes in child-rearing attitudes are crucial drivers behind men's increased participation in unpaid caregiving (Hook and Wolfe 2012; Raley et al. 2012; Sullivan et al. 2014). Additionally, there are modest indications of men's increased involvement in paid care work, particularly among non-White men (Dill and Duffy 2022; Dill and Hodges 2019; Hodges 2020; Wingfield 2009).

The extent to which these changes have translated into a reduction of women's overrepresentation in care work activities remains uncertain. The increased participation of men in unpaid childcare has not supplanted women's unpaid childcare because of the concurrent rise of intensive mothering (Hays 1996; Macdonald 2011). Furthermore, research has not indicated large changes in men's participation in adult care or in paid care, which might be growing domains of care. Grigoryeva (2017) found no change between 1995 and 2010 in the extent to which daughters are more likely than sons to provide unpaid adult care for their parents, and other studies found only small increases (National Alliance for Caregiving and AARP 2021). Modest increases in men's engagement in paid care have not notably altered the persistent hyperfeminization, particularly in jobs providing routine interactive care, where women continue to be overrepresented (Duffy 2011; Folbre 2012; Hodges 2020). For instance, the proportion of women among childcare workers scarcely shifted from 96.8% to 94.3% in the period 1995–2022 (U.S. Bureau of Labour Statistics 2022). Scholars have argued that this enduring feminization results from low pay and gender norms designating women as default and superior caregivers (Dill and Hodges 2019; Duffy 2011; Glenn 2010).

Racial Inequality in Care Work

The evidence of declining overrepresentation of non-White women in care work stems from studies documenting shifts in labor market discrimination and segregation. In the first half of the twentieth century, domestic work was among the few job options available to non-White women (Branch 2011). In 1900, nearly 75% of employed Black women were in either domestic service or agriculture (Duffy 2011:24). Domestic work was also prevalent among recently arrived European immigrants in the Northeast, Hispanics in the South, and Asian populations in the West (Glenn 2002). The latter half of the twentieth century witnessed economic diversification, triggering a substantial exodus from domestic work (Duffy 2011; Glenn 2002).

However, persisting job discrimination continued to confine non-White women primarily to care roles (Glenn 1992). Non-White women were often perceived as “suitable” for (and obligated to assume) lower tier precarious care jobs, while White women occupied managerial and higher status ones (Branch 2011; Dill and Duffy 2022; Duffy 2005, 2011; Glenn 2010; Romero and Pérez 2016; Wooten and Branch 2012). As Glenn (1992) put it, by the 1990s, non-White women had moved out of White women's houses but continued to work for White women in hospitals, schools, and government services. Thus, non-White women remain overrepresented in many low-paying care jobs (Bleiweis et al. 2021; Dill and Duffy 2022). For instance, non-White women represent 37% of the care workers in nursing homes and home health care but constitute only 11% of the workforce.⁵ Overall, despite occupational diversification, the overrepresentation of non-White women in paid care and in overall care might remain high, especially if paid care is expanding as a share of overall care work.

Changes in non-White women's involvement in unpaid care can also impact racial inequality in overall care work, although there is limited research on this aspect. The prevalence and intensity of unpaid care vary based on several compositional elements, including the number of family and community members requiring care, employment, household structure, and access to paid care services (Cross 2017; Dukhovnov and Zagheni 2015, 2019; Gennetian and Rodrigues 2021; Ice 2023; Margolis and Verdery 2019; Park et al. 2019; Sweeney and Raley 2014). Structural racism and racial health inequities might increase adult care burdens for non-White families but could also result in shorter lifespans that reduce long-term care demand (Boen 2020; Boen and Hummer 2019; Geronimus et al. 2006; Phelan and Link 2015; Wrigley-Field 2020).

Summary and Hypotheses

Existing evidence provides incomplete and mixed support for common assertions about changes in the social division of routine care work. Research documenting changes in one care domain cannot conclusively determine whether these changes impact the overall distribution of care work. Here, we summarize the possible expectations that emerge from the literature review.

Defamilialization

• Hypothesis 1a: The expansion of paid care services has increased the proportion of care work delivered by paid caregivers relative to unpaid caregivers.

• Hypothesis 1b: Although the expansion of paid care services has augmented the volume of paid care work, concurrent increases in unpaid child, adult care, or both have constrained the degree to which this expansion has led to paid caregivers contributing a larger proportion of total care work.

Population Aging

• Hypothesis 2a: As the population ages, the time dedicated to adult care has increased and has led to a greater proportion of total care work being directed toward adults versus children.

• Hypothesis 2b: Population aging has led to modest increments in adult care work, with no significant alterations in the relative contribution of adult care to overall care work, perhaps because of a slower-than-anticipated rise in demand for adult care, shortcomings in meeting escalating demands for adult care, or the continued significance of unpaid childcare.

Overrepresentation of Women

• Hypothesis 3a: The disproportionate representation of women in overall care work has diminished owing to men's increased participation in unpaid childcare.

• Hypothesis 3b: The overrepresentation of women in overall care work has remained unchanged because the escalation in paid care services and the increasing demand for adult care have counteracted the impact of men's increased involvement in unpaid childcare.

Overrepresentation of Non-White Women

• Hypothesis 4a: The disproportionate representation of non-White women in paid care and in overall care work has declined with the diversification of economic opportunity.

• Hypothesis 4b: The expansion of the paid care sector and the continuation of labor market discrimination have led to small or no declines in the disproportionate representation of non-White women in paid care and in overall care.

A Demographic Framework to Study the Social Division of Care Work

We propose an analytic framework to provide concrete empirical evidence to examine claims concerning shifts in the social division of care work. This aim requires delineating a cohesive set of care activities that can be consistently identified across time, irrespective of the performer, recipient, or location of these tasks. It also requires a common metric for meaningfully gauging the extent or volume of care work for both paid and unpaid caregiving. We describe our framework as it applies to the subset of care activities we focus on: routine interactive care work. However, this framework can be adapted to other definitions of care work.

Our framework conceptualizes the social division of time spent on care activities in a given population. The first quantity of interest is total care work time, which is defined as the sum of work hours that paid and unpaid caregivers spend providing care to children and adults on an average day in a given population. Formally, this quantity can be described as follows:

where T is the average number of hours spent on care on an average day in year y, and N is the number of caregivers on an average day in year y. Subscripts i and j classify care work according to domains, where i distinguishes between paid and unpaid care, and j distinguishes between adult care and childcare. Subscript g classifies care according to caregivers’ social groups, which we define by gender and race. Analyses by gender define two groups: women and men. Analyses by race also define two groups: non-White and White women.⁶

Computing this quantity would be easy if nationally representative surveys collected information about paid and unpaid care work, but most surveys focus on either paid or unpaid care (Folbre 2024; Folbre et al. 2023).⁷ To overcome this obstacle, we propose relying on two common survey types: employment and time-use surveys. Employment surveys can yield estimates of the number of paid caregivers and the average time they spend providing care on an average day, and time-use surveys can yield estimates of the number of unpaid caregivers and the average time they spend providing unpaid care on an average day. Further, information about care recipients can be leveraged to classify paid and unpaid caregivers into child and adult care. Time-use surveys typically include information about the age of the care recipient, and in employment surveys, information about the occupations can be leveraged to distinguish caregivers who provide childcare from those who provide adult care. For instance, daycare workers care for children, whereas nursing home assistants care for adults.

After making computations for the employment and time use surveys, we merge the estimates to calculate total care work time and compute key quantities of interest to examine shifts in the social division of care work time. Claims about defamilialization can be assessed by examining changes in the proportion of total care work time provided by paid caregivers; the impact of population aging can be assessed by examining changes in the proportion of total care work time dedicated to adults; and changes in gender and racial inequality can be assessed by examining changes in the ratio of total care work time provided by women versus men and by non-White women versus White women. Each of these quantities can be computed for total care work time and specific domains. For instance, we can evaluate the defamilialization of total care work or of childcare alone or compute the overrepresentation of non-White women in total and in paid care work separately. Additionally, we can use these quantities to analyze how changes in specific care work domains (e.g., increases in the share of total care work provided by paid caregivers) impact the social division of care work across social groups (e.g., increasing the overrepresentation of non-White women versus White women).

This demographic framework offers clear and direct metrics to address fundamental yet unresolved questions concerning the social division of care work. To our knowledge, this empirical endeavor is the first attempt to integrate paid and unpaid child and adult care work time in a unified accounting framework.

Data and Sample

We use data from the 1965–2018 AHTUS (Fisher et al. 2018) and the 1968–2018 CPS-ASEC (Flood et al. 2021). The AHTUS compiles time-use surveys, and we use these data to obtain estimates about unpaid care work time. The CPS-ASEC is a nationally representative household survey that collects information on employment status and includes detailed occupational and industry information that can produce estimates about paid care work time. To maximize comparability across datasets, we use data from the following years to track changes over 50 years: the mid-1960s (AHTUS 1965, CPS-ASEC 1968), the mid-1970s (AHTUS and CPS-ASEC 1975), the mid-1980s (AHTUS and CPS-ASEC 1985), the mid-1990s (AHTUS and CPS-ASEC 1995), the mid-2000s (AHTUS and CPS-ASEC 2005), and the late 2010s (AHTUS and CPS-ASEC 2018).

The analytic sample comprises the adult population, defined as respondents aged 18 or older at the time of the interview. This age restriction excludes care provided by children. The final analytic samples include 32,253 observations in the AHTUS and 686,018 observations in the CPS-ASEC.

Measures

We describe the measures for our main analysis focused on routine interactive care work. The online appendix describes alternative measures of care work used in supplementary analyses.

Unpaid care work is measured using harmonized activity codes available in the AHTUS. Respondents are classified as unpaid caregivers if they reported spending time providing routine interactive care for children or adults as a primary activity. This care might be provided for individuals in or outside the household, and codes distinguish between care provided for children and adults. Table S1 lists the codes used for this measure (all tables and the figure designated with an “S” appear in the online appendix). For those providing unpaid care work, unpaid care work time indicates the number of hours respondents reported providing routine interactive care as a primary activity. We focus on primary activity data because of between-survey comparability limitations with secondary and supervisory activity data. We discuss the limitations of focusing on primary activity data and provide supplementary analyses with broader measures of unpaid care in the online appendix.

Paid care work is measured using detailed occupational and industry codes in the CPS-ASEC. Respondents are classified as paid caregivers if their main job is a routine interactive care job.⁸ These occupations include childcare workers, nurse aides in nursing homes, and domestic workers. In supplementary analyses, we remove domestic workers from the analysis because this job category might include workers who do not provide interactive care, and we also report analyses with broader definitions of paid care jobs (see the online appendix). Paid care work time is measured using information on usual weekly hours at the job among paid care workers.⁹ This measure assumes that all the time on the job involves care provision, which likely overestimates the time effectively spent on interactive care. Also note that survey-based job measures likely underrepresent the informal care economy; we return to this limitation in the Discussion section.

Although our goal is to consistently identify the set of tasks related to providing routine interactive care by paid and unpaid caregivers, survey instruments and measuring conventions create unavoidable errors, mismatches, and inconsistencies. Some care activities are captured more accurately in time-use surveys than in employment surveys, and vice versa. The inability to include secondary and supervisory time-use data impacts our ability to capture unpaid supervisory care, which constitutes a substantial time investment for unpaid caregivers. The lack of detailed work-time data for paid care workers means that we likely overestimate the time effectively spent on interactive care and that supervisory care is included in our measure of paid care time. Additionally, including the category of domestic workers is important because it has historically been a key care job, but survey measures do not allow us to identify which domestic workers actually provide interactive childcare and adult care. We aim to use the imperfect data as carefully as possible to obtain reasonable estimates, but readers should keep these data limitations in mind. A detailed discussion of measurement limitations and supplementary analyses is available in the online appendix.

Analytic Approach

We use demographic methods to describe the social division of routine interactive care work time across domains and social groups and to test hypotheses about social change. The first step is to construct measures describing shifts in the composition of care work time across domains (paid vs. unpaid and child vs. adult). Defamilialization is measured with the proportion of care work time provided by paid caregivers, and the impact of population aging is measured with the proportion of care work time provided to adults.¹⁰ Both are calculated for overall care work time and for specific care domains. These quantities will be used to evaluate Hypotheses 1a–1b and 2a–2b.

The second step is to analyze changes in gender and racial inequality. Women's overrepresentation compares the volume of care work women and men provide, and non-White women's overrepresentation compares the volume of care work non-White women and White women provide. We adjust both ratios for the differences in group size to assess the volume of care work time in relation to the size of the group. Formally, women's overrepresentation ratio is calculated as follows:

where WO stands for women's overrepresentation in year y. T is the average number of hours spent on care, and N is the number of individuals. Subscripts i and j classify care work according to domains, where i distinguishes between paid and unpaid care and j distinguishes between adult care and childcare. Thus, the numerator is the care work time women provide adjusted by the number of women in the adult population, and the denominator is the care work time men provide adjusted by the number of men in the adult population. A ratio of 1 indicates that women and men provide the same amount of care work time, and a ratio above 1 indicates that women provide more care work than men. For instance, a ratio of 1.5 indicates that for every 1 hour of men's care work, women provide 1.5 hours. We compute all these measures for each year and each domain of care work.

Next, we use classic decomposition methods (Kitagawa 1955) to analyze the drivers of change in gender and racial inequality between 1965 and 2018. We decompose the change into parts due to (a) changes in the composition of care work across domains (i.e., changes in the relative size of paid/unpaid child/adult care) and (b) changes in the ratios of overrepresentation within each care domain (e.g., changes in the extent to which unpaid childcare is done by women more than men). Following Kitagawa's notation, this decomposition can be formally written as follows:

where, respectively, r.. and R.. are the ratios of overrepresentation for the last and first year of observation (i.e., 2018 and 1965); n.. and N.. are the total care work for the last and first observation years; and $rij$⁠, $Rij$⁠, $nij$⁠, and $Nij$ are the domain-specific ratios of overrepresentation and the care work time in paid and unpaid adult care and childcare corresponding to the first and last years of observation. This decomposition will be used to test Hypotheses 3a–3b and 4a–4b.

Results

Table 1 shows the characteristics of the AHTUS and CPS-ASEC samples; all statistics are weighted to represent population-level quantities. The composition of the two samples is generally similar, as expected. Table 1 also reports the percentage of the population providing routine interactive paid and unpaid care for children and adults, as well as the average hours caregivers spent providing this care on an average day. Unsurprisingly, a much larger share of the adult population provided unpaid care than paid care: 24% to 38% provided unpaid childcare, and 8% to 18% provided unpaid adult care, whereas the percentage providing paid childcare or adult care oscillated around 3% and 1%, respectively. Conversely, paid caregivers spent, on average, more hours providing care than unpaid caregivers: 5 versus 1–2 hours on a typical day.

Changes in the Composition of Total Care Work Time Across Domains

Figure 1 describes trends in the volume and social division of total routine interactive care work time, and Table 2 provides the corresponding quantities. To adjust for changes in population size, we measure total care work time in hours per 1,000 individuals. Panel a of Figure 1 reveals that total care work time increased by 26% between the 1960s and the 2010s, from 757 hours of care work per 1,000 individuals on an average day in 1965 to 951 hours per 1,000 individuals in 2018. This change was driven by increases in all care work domains except for paid adult care. Unpaid adult care and paid childcare increased the fastest (120% and 42%, respectively), whereas the increase in unpaid childcare was more moderate (9%). The volume of work time dedicated to providing care was quite large. By comparison, the amount of work time to run the entire U.S. retail sector in 2018 was 492 hours per 1,000 individuals.

Panel b of Figure 1 reveals remarkable stability in the social division of routine interactive care work across domains, contradicting the expectations of social change discussed earlier. Over the half century studied, unpaid childcare consistently constituted the largest domain of routine interactive care work, followed by unpaid adult care, paid childcare, and paid adult care. Table 2 reports defamilialization and aging proportions to evaluate Hypotheses 1a–1b and 2a–2b. Overall, we find little support for the defamilialization of care work but clear evidence for the impact of population aging. The proportion of routine interactive care work provided by paid caregivers barely increased from .19 to .20 between 1965 and 2018, indicating that the growth in paid care services did not translate into a greater share of total care work time provided by paid care workers. Childcare became more defamilialized, consistent with claims about the increasing relevance of paid care provision, but adult care became less defamilialized. Paid caregivers provided 17% of total childcare in 1965, compared with 21% by 2018 (representing a 24% increase), whereas their share of adult care provision decreased from 29% to 15% (a 47% decrease) over the same period.¹¹ Supplementary analyses excluding domestic workers show a modest increase in defamilialization in overall care work and in paid adult care work, which we discuss later in the article and in the online appendix. On the whole, the results support Hypothesis 1b, with evidence pointing to increases in unpaid adult care and childcare containing the extent to which care work became defamilialized.

For population aging, the results provide clear evidence that a greater share of total routine interactive care work time is being allocated to adults instead of children. Adults received 16% of total care work time in 1965 but 23% in 2018—a 46% increase. This increase is due to the remarkable growth in unpaid adult care (which more than doubled) outpacing the growth in unpaid childcare, translating into a larger share of unpaid care going to adults over time. Trends in paid care did not contribute to the growth in the share of overall care work going to adults because the share of paid care delivered to adults versus children decreased over this period (from 24% to 18%). In all, the results support Hypothesis 2a: a greater proportion of total care work time was directed toward adults.

Changes in Gender and Racial Inequality

Did gender inequality in total care work time change between 1965 and 2018? Table 3 reports trends and decomposition results to answer this question. Panel A shows that women's overrepresentation in routine interactive care work time nearly halved over the study period. In 1965, women completed 4.46 hours of care work for every 1 hour men performed, compared with 2.46 hours by 2018. The overrepresentation of women declined across all domains of care, although more strongly in paid adult care and unpaid childcare. Panel B reports decomposition results to examine the extent to which this decline in gender inequality can be attributed to shifts in the composition of care work time across domains versus changes in the overrepresentation ratios within specific domains of care work. This analysis shows that 46% of the decline in gender inequality can be attributed to the decline in women's overrepresentation in unpaid childcare. Additionally, the decline in overrepresentation in paid child and adult care and the change in the division of care work across domains account for some of the decline in gender inequality. These results do not support Hypothesis 3b, which expected the growth of paid care work to offset reductions in women's overrepresentation in unpaid childcare work, but are consistent with the results in Table 2 showing little change toward defamilialization in overall care work and only modest increases in childcare defamilialization. Although a greater share of childcare in 2018 was provided by paid caregivers who were disproportionately women, unpaid childcare continued to constitute a very large domain of care work, and men's increased involvement in this domain played a key role in reducing women's overrepresentation in overall routine interactive care. Additionally, the overrepresentation of women in paid care work declined over the study period, although it remained high. Therefore, the results support Hypothesis 3a, which anticipated substantial declines in gender inequality driven by men's increased involvement in unpaid childcare.

In supplementary analyses, we further probed the influence of changes in unpaid childcare on the overall decline in women's overrepresentation in total care work time. First, we evaluated the possibility that this rapid decline could be partly driven by men's involvement in joint childcare rather than solo childcare. To do this, we ran analyses deflating unpaid childcare in joint episodes and found that our results did not substantially change (see Table S4). Additionally, we examined whether the influence of unpaid childcare was related to intensive mothering increasing unpaid childcare time. To this end, we ran analyses holding constant the average time caregivers spent providing unpaid childcare to the 1965 level and found that the influence of unpaid childcare remained important but diminished, consistent with the influence of intensive mothering (see Table S5). Overall, the results indicate that increases in men's involvement in unpaid childcare had a genuine and sizable influence in reducing gender inequality in total routine interactive care work time.

Table 4 reports results for the analysis of racial inequality in routine interactive care work. Panel A shows that non-White women's overrepresentation declined between the 1960s and 2010s. In 1965, non-White women provided 1.42 hours of care work for every 1 hour of care White women provided, and this rate declined to 0.87 by 2018. Panel B decomposes this change into parts due to changes in composition and ratios. The results indicate that 54% and 27% of the decline in racial inequality can be attributed to declines in the overrepresentation rate in paid childcare and paid adult care, respectively, whereas changes in unpaid care contribute less to explaining the decline. The decline in racial inequality in paid child and adult care is largely driven by changes in the category of maid and housekeeping jobs, the only paid care job in which non-White women's overrepresentation is in decline (see Figure S1). These results are generally consistent with Hypothesis 4a, which posited that expanded economic opportunity would lead to declines in racial inequality in care work. However, the increasing overrepresentation of non-White women across most paid care jobs suggests that if these jobs grow in the future, the decline in racial inequality might stall or even reverse.

Sensitivity to Changes in the Definition of Care Work

We performed supplementary analyses to evaluate the sensitivity of our conclusions to different operationalizations of care work. These analyses considered, for example, secondary and supervisory activity time-use data, measures of support care work, and broader definitions of paid care. Overall, our conclusions regarding Hypotheses 1–4 remained similar. One noticeable deviation resulted from excluding domestic workers from paid care work: we found more evidence for the defamilialization of total care work time and a smaller decline in the overrepresentation of non-White women. For full results and a discussion of all supplementary analyses, see the online appendix.

Discussion

This study proposes a demographic framework to analyze social change in care provision. Combining time-use and labor-force surveys, we introduce novel population-level estimates for the volume and social division of care work time across domains (paid vs. unpaid and adult vs. child) and across social groups (by gender and race). Our analysis, which focuses on routine interactive care, reveals considerable stability and change. We find limited evidence of defamilialization, but we find support for the increasing relevance of adult care work. Gender and racial inequality in routine interactive care work appreciably declined. Although women remained overrepresented as caregivers across all domains throughout the study period, this overrepresentation nearly halved since the 1960s. In 2018, non-White women were no longer overrepresented in routine interactive care work, but they remained overrepresented in paid adult care work.

These results are consistent with expectations of substantial social change propelled by population aging and men's involvement in care work. However, several factors indicate that these shifts could have been even larger. The estimated increase in unpaid adult care is likely an underestimate because adult care time is often underreported in time-use surveys (Suh 2016)¹² and because the evidence about unmet needs suggests that the demand for adult care has increased more than provision (Beach and Schulz 2016; Patterson 2022; Potter 2019). Furthermore, increases in unpaid and paid childcare have partially offset the impact of increased adult care on the overall distribution of care work time. With baby boomers turning 70 and 80 in the coming years, the shift toward adult care will likely be even more pronounced in the future, especially if unpaid child time investments stall or efforts to address unmet adult care needs improve. The impact of changes in men's involvement in care work was substantial, too, decreasing the overrepresentation of women by half. Men became more involved in all care domains, but especially in unpaid childcare and paid adult care. These results are consistent with research documenting fathers’ increased involvement in childcare (e.g., Altintas and Sullivan 2017; Sayer et al. 2004) and with the involvement of men, especially non-White men, in some paid care jobs (Budig et al. 2019; Wingfield 2009). Had intensive motherhood not increased women's unpaid childcare investments over this period, the decline in gender inequality would have been even larger.

The results are less consistent with expectations of substantial shifts in defamilialization and in racial inequality. The lack of evidence for defamilialization could be seen as inconsistent with research documenting the growth of the paid care sector (Buera and Kaboski 2012; Duffy 2011; Dwyer 2013). However, the key to reconciling these findings is that the growth in paid care coincided with substantial growth in unpaid care due to intensive mothering and population aging. Additionally, concurrent increases in paid and unpaid care indicate that paid care can both complement and substitute for unpaid care (Folbre and Bittman 2004). It is also possible that the lack of evidence for defamilialization is related to the inclusion of domestic workers. We might overestimate the amount of paid care work in the earlier period because domestic work was a large job category then, and we do not know the share of domestic workers who provide routine interactive care. Results excluding domestic workers show a moderate increase in defamilialization (see the online appendix). With respect to the overrepresentation of non-White women, the results indicate that the modest decline over this period began tapering off. Among paid adult care jobs experiencing growth, the overrepresentation of non-White women is exacerbating, suggesting that future expansions of the long-term care industry could stall or even reverse the estimated decline. This trend is consistent with research documenting the continuing concentration of non-White women in some paid care jobs (Dill and Duffy 2022; Dwyer 2013; Hodges 2020).

Our analyses have several limitations. We use the gold-standard data sources that have been extensively employed to study unpaid and paid care work, but these datasets have important shortcomings. Unpaid care measures are likely underestimated because time-use surveys have not consistently tracked secondary and supervisory forms of care work, and unpaid adult care is underreported. Paid care measures are likely under- and overestimated for different reasons. Labor force surveys likely underestimate the size of the informal economy, which plays a well-known important role in paid care services (Duffy 2005; Shih et al. 2022; U.S. Department of Labor 2011). However, our analyses assume that all paid work hours constitute care provision, which overestimates the amount of care effectively provided during the workday. We conducted extensive sensitivity analyses with available data, but we cannot entirely rule out whether these limitations biased our estimates. Hence, our study reinforces the need to improve survey data about care work (Folbre et al. 2023)—more specifically, the need for improved measures about care work needs (met and unmet), supervisory care, unpaid adult care activities, and paid care job responsibilities, including in the informal economy. Our analysis of racial inequality is also limited due to the lack of detailed and consistent measures of race and ethnicity in the study period, restricting our focus to inequalities between White and non-White women. Our conclusions about the overrepresentation of non-White women could be different if could have incorporated measures of Hispanic/Latinx and immigrant status, given that paid care jobs are common among Hispanic women (Duffy 2011; Wright and Dwyer 2003).¹³ Lastly, the analytic approach considers one axis of inequality at a time (gender and then race) instead of examining the interrelatedness between the two. Future studies should adopt approaches that can better capture intersectional dimensions of changes in the social division of care work.

Notwithstanding these limitations, our analysis provides novel evidence to evaluate common claims about changes in the social division of care work. Our framework overcomes the fragmentation of the empirical literature on care work and presents an integrated description of changes in paid and unpaid care of adults and children. Future studies can apply this framework to investigate variations across regions or countries and examine detailed drivers of specific changes in total care work time. In sum, this study reinforces the importance of care work as a central sphere of human labor and the need to challenge its systematic invisibility and devaluation.

Acknowledgments

This research was supported by the Russell Sage Foundation Visiting Scholar Award and was conducted using the facilities of the University of Pennsylvania Population Studies Center (R24 HD044964). This research received support from the Population Research Training Grant (NIH T32 HD007242) awarded to the Population Studies Center at the University of Pennsylvania by the National Institutes of Health's Eunice Kennedy Shriver National Institute of Child Health and Human Development. We are grateful to Liana Sayer, Philip Cohen, and Paula Fomby for their comments and suggestions on earlier versions of this research.

Notes

References