Recent trends in employment, earnings, and time use among college-educated married mothers

Starting in the late 1950s and peaking in the early 2000s, women’s employment–population ratios (hereafter E–P ratios) markedly increased in the United States. At the same time, men’s E–P ratios steadily fell.¹ In this article, we decompose labor supply from 1975 to 2019 by sex, marital status, education, and the presence and ages of children.² We use the March supplement of the Current Population Survey (CPS) data from 1975 to 2019 and American Time Use Survey (ATUS) data from 2003 to 2019. We consider recent trends in labor supply at both the extensive and intensive margins.³ We use the E–P ratio as our measure of the extensive margin of labor supply and calculate the number of employed people (those who have worked at least 500 hours in the prior year) as a percentage of the civilian noninstitutional population. For the intensive margin of labor supply, we report average annual hours worked for those who are employed.

We focus on the 2000–19 period and document several new findings, all of which are statistically significant.⁴ First, the E–P ratio did not decrease for married women in that time, but it did for most other comparable groups (notably, men of all education groups, high school-educated women, and nonmarried women). More importantly, labor supply increased among college-educated married mothers during this period at both the extensive and intensive margins. This increase was especially pronounced for college-educated married mothers with young children (those whose youngest child is below the age of 5), compared with those with older children (those whose youngest child is between the ages of 5 and 14 years old). For most other groups, we did not see increases in labor supply during this period.

We are among the first to document a divergence in E–P ratios and hours worked among subgroups of married women beginning in the year 2000.⁵ Education levels, the presence of children, and the ages of those children all affect E–P ratios and hours worked: high school-educated married women had a decrease in E–P ratios during this period, and the decreases were larger for mothers than nonmothers (married women without children in the home). In addition, hours worked for high school-educated married mothers have been constant since 2007, while E–P ratios and hours worked for college-educated married mothers have been increasing since the mid-1970s.

We also compare real hourly earnings for various groups of married women.⁶ Our analysis suggests that, from 2000 to 2019, the real hourly earnings of college-educated women have increased at a faster rate than those of other cohorts.⁷ This pattern is more pronounced among college-educated married mothers (especially those with young children) than nonmothers. As a result, the real earnings gap between college-educated nonmothers and mothers with young children closed in 2004. Since then, real hourly earnings among married college-educated mothers with young children have surpassed those of their childless counterparts.

Lastly, we turn to the ATUS to document important trends in time allocation between married men and women from 2003 (when the ATUS was first available) to 2019, focusing on the time spent on childcare and unpaid domestic work. Between 2004 and 2018 (as measured using 3-year moving averages), college-educated married men have increased their time on childcare, whereas college-educated married women have reduced their childcare time; all of these changes are statistically significant. We also document decreases in time spent on unpaid domestic work among college-educated men and women.

Data sources

For our analysis, we draw on two nationally representative data sources—the CPS and the ATUS. The samples of each dataset are described in detail below.

CPS data

We use hours and earnings data from the March supplement of the CPS from 1975 to 2019.⁸ The CPS data also contain detailed information about sex, age, marital status, years of education, and number and ages of children in the household. Marital status is categorized as either married or nonmarried.⁹ We classify workers as college-educated if they have more than 12 years of formal education and high school-educated if they have 12 years or fewer of education. We compare workers with and without children, and with children of different ages. We classify those with children as anyone with at least one child under the age of 15 in the household (i.e., one or more children). Those with young children are those whose youngest child is under the age of 5 (when formal primary schooling usually starts for children). We also consider households with older children, in which the youngest child is between the ages of 5 and 14. Women and men with no children include those who never had children and those whose youngest child is 15 or older.¹⁰

We restrict the CPS sample to people between the ages of 25 and 64. For each year, the E–P ratio represents the number of employed people (those who have worked at least 500 hours in the prior year) in a demographic category as a percentage of the civilian noninstitutional population in the same demographic group.¹¹ For an estimate of hours worked, we calculate average annual hours worked for those who are employed. To calculate this value, we multiply usual hours worked per week over all jobs in the prior year by the actual number of weeks worked last year.

We use CPS data to generate average real hourly earnings for different demographic groups each year. We take total annual earnings from all jobs and divide it by the number of annual hours worked to derive a measure of hourly earnings.¹² Earnings are reported in 2012 dollars, using the Personal Consumption Expenditures (PCE) price index. We use March CPS sample weights for all of our estimates.

American Time Use Survey data

The American Time Use Survey (ATUS) is produced by the U.S. Bureau of Labor Statistics (BLS) and measures the time that people spend doing paid work, childcare, and other nonmarket activities, such as unpaid domestic work and leisure. The ATUS is a federal survey in which people are randomly selected from a subset of households that have completed their eighth month of interviews for the CPS. Survey respondents are interviewed once about how they spent their time during the previous day. Our sample includes those who report zero and nonzero work hours, but we restrict our sample to people between the ages of 25 and 64. We use the definition of unpaid domestic work from the American Heritage Time Use Survey (AHTUS); it includes a broad range of household activities.¹³ We report average minutes per day spent on different activities, averaged over all days of each week. We decompose the results for various demographic groups, again using sample weights.

Results

In this section, we present key findings on trends in employment, real hourly earnings, and time use patterns.

Employment

We begin by examining employment trends at both the extensive and intensive margins.

Employment–population ratios (extensive margin)

We first highlight a few important trends in E–P ratios, with a particular focus on the 2000–19 period. In chart 1, we plot E–P ratios for women and men by marital status. As is well recognized in the literature on labor market fluctuations, employment is strongly procyclical: E–P ratios went down during the recessions in the mid-1970s, the early 1980s, the early 1990s, the early 2000s, and the Great Recession of the late 2000s. These cyclical movements of E–P ratios often make it more complicated to distill long-term upward or downward trends. Following the Great Recession, for example, E–P ratios took nearly a decade to recover to their pre-2007 levels for all subgroups.

Nevertheless, even without using a sophisticated filtering method, chart 1 shows that men’s E–P ratios have no clear trend until around 2000, when they started to decline. For both married and nonmarried men, E–P ratios were lower in 2006 than they were in 2000, several years after the recession of the early 2000s. Also, the E–P ratio of nonmarried men in 2019 (a decade after the end of the Great Recession) was still lower than its prerecession level. As for women, E–P ratios steadily increased until 2000 for both married and nonmarried groups and, since then, have declined for nonmarried women but have stagnated for married women. In the 2000–2019 period, fitted linear regression lines reveal a flat trend for married women but negative slope coefficients for the other three groups.¹⁴ Nonmarried men experienced a particularly large 5.5-percentage point reduction in the E–P ratio, from 0.803 in 2000 to 0.748 in 2019.¹⁵ Meanwhile, married women showed distinct labor supply behavior from 2000 to 2019.

In chart 2, we decompose E–P ratios for groups of married women across education levels. The employment gap between college- and high school-educated married women has widened, particularly since 2000. From 2000 to 2019, while high school-educated women experienced a sizeable reduction in the E–P ratio of 4.9 percentage points (from 0.605 in 2000 to 0.556 in 2019), the ratio slightly increased for college-educated women, by 1.1 percentage points (from 0.721 in 2000 to 0.733 in 2019).¹⁶

We next decompose E–P ratios of married women by the presence and age of children (in chart 3). In the recent period (2000–2019), we observe increased E–P ratios among married women with young children (a 3.6-percentage point increase from 0.586 to 0.622).¹⁷ For the same period, we observe very few changes in E–P ratios for married women with no children or those with older children.

In chart 4, we further decompose E–P ratios among married women by education and the presence and ages of children from 2000 to 2019. Among college-educated married women, mothers with young children had an increase in the E–P ratio of 6.6 percentage points during this period, whereas nonmothers and those with older children had small changes (under 2 percentage points). On the contrary, among those with a high school education, E–P ratios decreased for all subgroups since 2000 but more rapidly for mothers compared with nonmothers (approximately 8 percentage points for mothers compared with 2 percentage points for nonmothers).

Thus, the divergence of recent E–P ratios for married women of different education levels shown in chart 2 is mostly attributable to motherhood. For nonmothers, employment gaps between college- and high school-educated married women have remained relatively constant throughout the sample period. The rapid increase in E–P ratios among college-educated mothers was particularly pronounced for those with young children, showing the largest employment gain—a 6.6-percentage point increase—from 0.616 in 2000 to 0.682 in 2019.¹⁸ As a result, the employment gap between college-educated nonmothers and college-educated mothers with young children was reduced from 14 percentage points in 2000 to 6 percentage points in 2019. This decrease contrasts with the employment gap between college-educated nonmothers and college-educated mothers with older children, which was nearly zero in 2000 and 1.6 percentage points in 2019. Thus, overall trends in E–P ratios among married women from 2000 to 2019 depend on their education levels and the presence and ages of children. In particular, the labor market experience among college-educated married mothers (particularly those with young children) has been somewhat unique in that E–P ratios have increased in recent decades.

Because our focus is on married women in this article, we do not display the results for various subgroups of men (though we report some results for married and nonmarried men as a point of comparison). Nevertheless, when we reproduce charts 2 through 4 using the married men’s sample, the results show that E–P ratios have generally decreased for all six subgroups in the recent period.¹⁹

Hours worked (intensive margin)

We now shift our focus to the intensive margin of labor supply among married women.²⁰ We report hours worked, for those with positive hours, in charts 5 through 8.

Chart 5 displays average annual hours worked by sex and by marital status. Throughout the entire period from 1975 to 2019, married men have consistently worked the longest hours, followed by nonmarried men as the next highest group. Nonmarried women, however, have supplied more labor hours than married women. Aside from procyclical adjustments in hours worked, all groups generally increased annual work hours between the mid-1970s and 1999. However, since then, while annual hours worked have decreased for both married and nonmarried men (by approximately 3.0 percentage points) and remained relatively constant for nonmarried women, they continue to rise for married women (with a 3.6-percentage point increase from 2000 to 2019).²¹

Charts 6 through 8 decompose the upward trend in annual hours among groups of married women. As before, we classify married women by education levels and the presence and age of children. We find that the group heterogeneity observed in E–P ratios are qualitatively preserved in hours worked since 2000. In chart 6, for example, although annual hours worked increased for both college and high school groups, the gap between the two groups slightly widened between 2000 and 2019.

In chart 7, hours worked increased for both nonmothers and mothers, but they increased at a faster rate for mothers. Among married mothers, there was a slightly more rapid increase in hours worked for those with young children (5.4 percentage points for the 2000–2019 period) compared with those with older children (3.5 percentage points).

When education and children are considered in chart 8, the divergent behavior for hours worked between college and high school groups appears more evident among mothers than nonmothers. Although annual hours continued to increase from 2000 to 2019 for college-educated mothers, they started to fall and stagnate for high school-educated mothers in 2007. Because of the rapid increase in hours worked among college-educated mothers (especially those with young children), the gap in annual hours between college-educated nonmothers and mothers with younger children fell by almost half, from 182 hours in 2000 to 96 hours in 2019. For those with older children, the reduction was similar, from 102 to 60 hours over the same period.

Real hourly earnings

In this section, we compare trends in real hourly earnings across different demographic groups. Chart 9 plots real hourly earnings (reported in log 2012 dollars) during the 1975–2019 period for married women and married men by education and by children. For married women, chart 9 corroborates the well-known rise in women’s real earnings, which is particularly pronounced among college-educated women. Over the period as a whole, real hourly earnings of college-educated married women rose at a rate of 0.13 per decade (in log dollars); the change is both economically and statistically significant. This strong upward trend in real hourly earnings makes it complicated to tease out trend movements from cyclical adjustments because real earnings are procyclical and the impacts of economic downturns on nominal earnings are heterogeneous.²²

Some interesting patterns emerge regarding group heterogeneity in trend movements in married women’s real hourly earnings. First, focusing on the period from 2000 to 2019, except for the period of the Great Recession and its aftermath (2007–13), real hourly earnings consistently increased for all subgroups of college-educated married women.²³ However, for high school-educated women, real hourly earnings stagnated in 2002 (several years before the Great Recession) until 2013, after which they began to show an upward trend. As a result, earning gaps between college- and high school-educated married women have further widened. Second, among college-educated married women, real hourly earnings have grown faster for mothers (at a rate of 0.25 from 2000 to 2019) than nonmothers (at a rate of 0.18).²⁴ Because of this faster earnings growth, the real earnings gap between nonmothers and mothers with young children closed entirely in 2004 and has reversed since then. A similar crossover occurred between nonmothers and mothers with older children starting in 2013. Although not reported here, the real earnings gap between college-educated fathers and college-educated mothers narrowed at a faster rate compared with the earnings gap between college-educated nonfathers and college-educated nonmothers. Third, for high school-educated married mothers, there was no earning catch-up or crossover observed during the sample period. This lack of catch-up suggests that real earning trends among married women differ with the presence and age of children and with education level.

As for married men, real hourly earnings appear to have generally stagnated from 1975 to 2013. Similar to women’s hourly earnings, men’s real hourly earnings have also increased for both high school- and college-educated men since 2013, albeit at a faster rate among college-educated men. In contrast to college-educated married women, trend movements in real hourly earnings were similar between college-educated fathers and nonfathers.

Time use

Among the demographic groups studied in this article, college-educated married mothers had the greatest gains in E–P ratios, hours worked, and real hourly earnings in recent decades, compared with various groups of married men who experienced declines in E–P ratios and hours worked, along with stagnant earnings. Since participation and hours decisions are jointly made within a household, we turn to the ATUS data to explore how many minutes people spent on childcare and unpaid domestic work per day (averaged over all days of the week).²⁵ To maintain consistency with our previous analysis, we compare married men and women by education and by the presence and age(s) of children.

Chart 10 shows the trends in time spent on childcare per day. Married women dedicated more time to childcare than married men. Regardless of sex, those with young children allocated the highest amount of time to childcare.

College-educated mothers with young children reduced their childcare time by 5 percent, from 198 minutes per day in 2004 to 188 in 2018. At the same time, college-educated mothers with older children reduced their childcare time by 8 percent, from 82 to 75 minutes.²⁶ This pattern suggests that, to some extent, the increased time spent working for this group was associated with a modest reduction in hours devoted to childcare. On the other hand, college-educated fathers with young children increased their childcare time by about 10 percent, from 94 minutes per day in 2004 to 104 in 2018. College-educated fathers with older children did not have a statistically significant change. This trend is more pronounced for college-educated fathers than high school-educated fathers. Thus, there has been some substitution of childcare from college-educated mothers to college-educated fathers, although college-educated mothers still spend nearly twice as much time on childcare (an average of 188 minutes per day compared with 104 minutes per day for college-educated fathers in 2018).

In chart 11, we plot the time spent on unpaid domestic work for married women and men with different ages of children. Among college-educated married mothers, regardless of the ages of children, there was a steady decline in the number of minutes spent per day on domestic work from 2004 to 2018, resulting in a 15-percent reduction.²⁷ College-educated women with older children spent more time on domestic work than those with young children (approximately 20 more minutes per day). As with childcare, women spent nearly twice as much time per day on domestic work than men. Similar to college-educated married mothers, college-educated married fathers with young children reduced their time in unpaid domestic work by approximately 12 percent from 2004 to 2018.²⁸ To summarize, college-educated married mothers are now working more and are spending slightly less time on childcare and unpaid domestic work than they did in 2004.²⁹

Summary

We document that married women’s labor supply has remained fairly steady since 2000 while men and nonmarried women have reduced their labor supply. Using data from the Current Population Survey and the American Time Use Survey, we examine recent trends in employment–population ratios, annual hours worked, real hourly earnings, and time allocation among married men and women. We disaggregate groups of married women and men based on education levels and the presence and ages of children.

Our most important finding is that only college-educated mothers (especially those with young children) have increased their employment since 2000, while all other groups—college-educated nonmothers, high school-educated married women (mothers and nonmothers), and all subgroups of men—reduced their employment. Among high school-educated married women, mothers have reduced their employment more rapidly compared with nonmothers. These observations are qualitatively similar when analyzing data on hours worked: college-educated married mothers have increased their annual hours worked since 2000, while all other groups of married women experience either small increases or stagnation in annual hours.

In addition, college-educated married mothers, especially those with young children, have experienced the most rapid hourly earnings growth over the entire sample period (1975–2019). As a result, the gap in real hourly earnings between college-educated nonmothers and mothers with young children completely closed in 2004. Since then, college-educated mothers with young children have enjoyed the highest hourly earnings among all groups of married women.

By examining the time use of married men and women, we find that college-educated women spent more time on paid work and less time on childcare and housework in the late 2010s compared with the early 2000s. Concurrently, college-educated men with young children spent more time on childcare in the late 2010s compared with the early 2000s, suggesting some substitution of childcare from college-educated mothers to college-educated fathers.

Appendix: Standard errors associated with the data

For the sake of completeness, this appendix provides the standard errors associated with the measurements used in the article.