The US economy lost a net of 8 million jobs between February 2020 and April 2021. Agreement is growing that people not actively seeking employment (inadequate labor supply) have been playing a major role in the slow recovery, as evidenced by factors including record job openings, the largest wage increases in decades, and other signs of a tighter labor market than would generally be expected given the still low levels of employment. Why has labor supply been slow to return? There are many candidate explanations, including the ongoing worry among some adults about getting COVID-19 if they return to the workplace, the increased availability and generosity of unemployment insurance benefits, and challenges for working parents associated with closed schools and inadequate childcare, among others. Diagnosing the sources of ongoing weak labor supply is important to inform the types of policies that are needed now to speed the recovery.
In this analysis, we quantify the effect of childcare challenges on the labor market by examining how much of the overall decline in employment can be explained by excess job loss among parents, and mothers specifically. We do this by constructing counterfactual employment rates, or employment-to-population ratios, as well as labor force participation rates that assign to parents with young children the percent change in employment and labor force participation rates experienced by comparable people without young children. The results of this exercise imply that differential job loss among parents, or even mothers specifically, accounts for a negligible share of aggregate job loss and could even have led to a small increase in jobs between the first quarters of 2020 and 2021.
This analysis demonstrates that despite the widespread challenges that parents across the country have faced from ongoing school and daycare closures, excess employment declines among parents of young children are not a driver of continuing low employment levels. In fact, while women with young children have left the workforce at a slightly higher rate than other women, men with young children have left the workforce at a lower rate than men without. Overall, the employment rates of parents of young children have declined by 4.5 percent as compared with 5.2 percent among people who are not parents of young children (the decline is also smaller for parents of young children when adjusting for age differences between the two groups).
Furthermore, this conclusion holds even if we consider only the excess declines in employment among mothers with young children without accounting for the offsetting effect among men. The reason why is twofold. First, women with young children account for only 12 percent of the US workforce. Second, mothers with young children left the workforce by only slightly more than comparable women without young children. Combining these two facts means that any childcare issues that have pushed mothers out of the workforce account for a negligible share of the overall reduction in employment since the beginning of the pandemic adjusting only for age differences. The estimated amount of the overall decline in employment that can be explained by challenges particular to mothers of young children is similar if we adjust the comparisons between mothers of young children and other women to control for age and education.
School closings and ongoing childcare challenges have been a tremendous source of stress for parents during the pandemic. They are also likely to have lasting, negative impacts on the learning and social development of children. In this analysis, however, our focus is on the one specific, empirical question: How has parenting affected the aggregate employment numbers over the course of the pandemic? Instead, parents of young children have suffered about equally as others in the widespread and, in many cases, damaging employment losses that have occurred throughout the economy.
Employment declines among parents
Women with young children have experienced the greatest rate of job loss over the past year, which is descriptively consistent with the hypothesis that school closures and childcare struggles have lowered women's work during the pandemic. Employment counts from the Current Population Survey in January, February, and March of 2020 and 2021 confirm that between the first quarter of 2020 and the first quarter of 2021, employment rates fell more for women with young children (defined as any own child under 13 in the household, including adopted children and stepchildren) than for all other women. Among all women, the employment rate fell by 5.7 percent (3.9 percentage points) for those with at least one child under 13, as compared with 5.0 percent (2.6 percentage points) for those without a child under 13.
The pattern is reversed among men, with larger declines among men without young children than among men with young children. The employment rate fell by 3.3 percent (3.0 percentage points) for men with at least one child under 13, as compared with a decline of 5.3 percent (3.2 percentage points) for men without a child under 13. Though explaining this finding is beyond the scope of this analysis, one possible explanation is the substitutability of labor supply between parents.
The gap in employment declines between mothers with young children and other women is driven by women without a bachelor's degree. The figure shows changes in the employment rates for individuals with any child under 13 and those without, separately for men and women with and without a bachelor's degree. As can be seen in the figure, among women with a bachelor's degree, there is no difference in the percent change in employment between mothers of young children and other women. But, among women without a bachelor's degree, women with young children experienced a larger decline in employment, as compared with women without young children. This affected group comprised 6 percent of the overall employment in the first quarter of 2020.
These comparisons are noteworthy, but they are not necessarily indicative of an effect of childcare struggles, school closures, or other factors associated with parenting on employment declines, even for mothers without a bachelor's degree. The reason is because women with and without young children differ on other dimensions that are related to employment outcomes, such as age, marital status, income, race and ethnicity, and industry. We next turn to a counterfactual analysis to estimate how much of the decline in employment—both overall and for mothers in particular—is attributable to having a young child in the household (and the associated challenges), adjusting for a range of factors.
Counterfactual analysis: What if the pandemic labor market experience of parents with young children was like that of otherwise similar people without young children?
The heart of our analysis is a counterfactual: What would have happened to aggregate employment rates if parents with young children experienced the same rate of employment decline as individuals without young children? This provides a reasonable estimate of the potential role of factors that affect families with young children, especially school closures and lack of childcare.
The motivation for this exercise is the observation that the fact that millions of mothers have lost jobs or left the labor force is not, by itself, evidence that childcare, school closings, or other child-related reasons are to blame. Hypothetically, if employment rates or labor force participation rates fell by the same proportion for similar people with and without young children, the primary causes of the decline are likely to be factors other than childcare challenges or school closures. In such a case, the source of the decline in employment among parents with young children would likely be something that was affecting everyone in a similar manner, such as workplaces being closed, jobs being undesirably unsafe, or unemployment insurance benefits being more generous and available.
A very simple, naive counterfactual exercise is to ask what would have happened to the overall employment rate if the employment of mothers with young children changed in the same way it did for women without young children. That is, what if women with young children experienced only a 5.0 percent decline in employment, instead of the 5.7 percent decline they actually did? Given that mothers with young children were 12 percent of total employment at the beginning of 2020, this would have resulted in the aggregate employment rate falling by 0.05 percentage point, less than the 3.1 percentage points it actually did. If we apply this counterfactual calculation to both mothers and fathers with young children, the decline in employment would have been even larger than actually observed.
Our baseline counterfactual accounts for some basic demographic differences between people with and without a child under 13. Specifically, we collapse the data into cells defined by sex, four age groups, and whether someone has a bachelor's degree. For each sex-age-education cell, we calculate what employment would have been if those with a child under 13 experienced the same percent change in employment as those in the same sex-age-education cell but who did not have a child under age 13. This allows us to compare parents of young children to otherwise similar individuals. We use the same methodology for labor force participation.
Table 1 reports the results of this baseline counterfactual calculation. This calculation implies that the effect of any excess impact of the pandemic on parents with young children can explain none of the decline in aggregate employment rates. In fact, this calculation implies there would have been even more jobs loss if parents of young children experienced the same employment decline as similar people without a young child, on account of fathers with young children experiencing relatively less employment loss than other men. If we only do the counterfactual calculation for women, the excess effect of the pandemic on mothers with young children can explain 2 percent of the aggregate employment rate change. Looking at the full childcare situation facing families, the sign flips, meaning the excess job loss among parents of young children can explain -2 percent of the aggregate employment-to-population decline.
|Table 1 Change in employment and labor force participation rates, observed and simulated under counterfactual scenario assuming no disproportionate effect on adults with children under age 13, 2020Q1-2021Q1|
|Employment-to-population rate||Labor force participation rate|
|Actual decline in outcome (percent)|
|Percentage point decline in outcome, total||-3.10||-1.69|
|Change in outcome under baseline counterfactual (p.p.)|
|Percent of decline explained by differential outcomes of people with young children|
|p.p. = percentage point
Note: Under baseline counterfactual, individuals with a child under age 13 are assigned the percent change in the employment rate or labor force participation rate as individuals without a child under age 13 within the same sex, educational attainment (bachelor's degree vs. not), and age (16-24, 25-39, 40-54, 55+) group.
Sources: IPUMS-CPS, University of Minnesota, www.ipums.org; authors' calculations.
Table 2 presents the results of numerous alternative counterfactual constructions, in addition to the baseline counterfactual and the naive unadjusted counterfactual, in order to examine the robustness of the baseline counterfactual finding. The top panel reports results from changing the age cutoff of children, looking alternately at parents of children less than age 6 or less than age 18. As shown in the top panel of the table, the qualitative conclusions are not dependent on age cutoff. If we consider the excess impact on mothers with children less than 6, the share of the aggregate employment decline that can be explained is 0 percent under the baseline counterfactual; if we consider the excess impact on mothers with children less than 18, the share that can be explained is 1 percent under the baseline counterfactual.
|Table 2 Change (in p.p.) from actual outcomes in employment and labor force participation rates, simulated under alternative counterfactual approaches, 2020Q1-2021Q1|
|Employment-to-Population Rate||Labor Force Participation Rate|
|Alternative age thresholds for children|
|Child under age 6||-0.01||-0.07||-0.02||-0.06|
|Child under age 18||0.03||-0.19||0.06||-0.10|
|Alternative demographic adjustment|
|No demographic adjustment (naïve counterfactual)||0.05||-0.11||0.04||-0.09|
|Adjusting for age||0.10||-0.03||0.10||0.01|
|Baseline: Adjusting for age and education||0.06||-0.07||0.07||-0.01|
|Adjusting for age, education, and income||0.05||-0.07||0.05||-0.02|
|Adjusting for age, education, and marital status||0.08||0.04||0.07||0.00|
|Adjusting for age, education, and race/Hispanic origin||0.05||-0.08||0.07||-0.01|
|Adjusting for age, education, and industry||0.00||-0.04||0.00||-0.01|
|Calculate counterfactual for parents with child under age 13 based on change for men with no child under age 13||0.08||-0.05||0.10||0.02|
|Memo: Percentage point change in outcome||-3.10||-1.69|
|p.p = percentage point
Sources: IPUMS-CPS, University of Minnesota, www.ipums.org; authors' calculations.
Next, we construct the counterfactual changes with different sets of demographic adjustments. We redefine the sets of demographic characteristics used for comparisons to alternately exclude education, to include marital status, to include income group, to include race/ethnicity, and to include industry. This allows us to compare parents of young children to people who are similar along various different dimensions.
To the extent that mothers with young children tend to work in a different set of industries than women without young children—even conditional on age and college-degree attainment—they would have experienced a differential degree of job loss. Constructing the counterfactual employment numbers adjusting for age, educational attainment, and industry leads to a simulated change in employment of 0.00 percentage point. But, constructing the counterfactual with an adjustment for industry is complicated by the problem of assigning industry to people who are out of work. The Current Population Survey asks some out-of-work respondents about the industry of their most recent job. However, this information is missing for 5 percent of unemployed workers and 98 percent of workers not in the labor force in our 2021 sample. Our counterfactual analysis that conditions on industry is thus essentially unable to account for employment loss that results in labor force nonparticipation.
For each different demographic adjustment we consider, the simulated change in the employment rate from assigning mothers with young children the corresponding change in employment of comparable women without young children is very small, below 0.10 percentage points across all counterfactuals, explaining at most 3 percent of the total decline in the overall employment rate.
One concern with these counterfactuals might be that to the extent that all women—including those without young children—disproportionately experienced caregiving burdens that reduced their employment, assigning the employment changes of women without young children to the employment changes of women with young children still incorporates caregiving burdens into the simulated effect. To address this concern, we consider a different counterfactual approach that assigns women with young children the change in employment experienced by similar men without young children. As with the results of the other alternative counterfactuals reported in table 2, this counterfactual still leads to the conclusion that differential employment changes for mothers with young children explain very little of the total change in employment. While our discussion has primarily focused on the effects of parenting on employment, the same series of counterfactual exercises show similar results for labor force participation. Our results indicate that differential declines in labor force participation for parents of young children explains essentially none of the 1.7 percentage point decline in the labor force participation rate between the first quarter of 2020 and the first quarter of 2021. For mothers of young children, differential declines in labor force participation can explain, at most, 0.10 percentage point, or 6 percent, of the total decline. Results for full-time employment are similarly small, and, if anything, generally indicate that full-time employment would have been slightly lower if parents of young children experienced the same change in full-time employment rates as individuals without young children.
We also tried to find direct evidence for the impacts of school closures. One method was to compare states that had above and below median rates of in-person schooling (according to this tracker) as of early May 2021. States with below median in-person schooling had larger percent reductions in employment rates for almost all groups by sex and educational attainment—whether or not they had young children—consistent with these states either having more serious problems with the virus or a greater level of mandatory or culturally induced social distancing. The differences between people with and without young children were not, however, systematically different in states with below median in-person schooling.
We also estimated regression models to obtain the conditional estimated impact of having a young child on employment rates for men and women in places with below median rates of in-person schooling as of early May 2021. The estimated regression coefficients do not show an excess negative effect on employment for men or women with young children in places with relatively low rates of in-person instruction. These results bolster our confidence in the results of our counterfactual analysis.
Finally, we also examined the patterns of employment losses specific to parents in the 2001 and 2007–09 recessions. We found that they were similar to the patterns observed over the past year. This suggests that there has not been something very different about how the pandemic and associated school and daycare closings have differentially affected the labor market outcomes of parents with young children relative to other workers who have also experienced the adverse impacts of the combination of the pandemic and associated recession.
Our examination of data on employment declines among parents with young children and others over the course of the pandemic suggests that overall, parents of young children did not leave the workforce substantially more than other comparable individuals. We constructed counterfactual estimates of what employment declines would have been if mothers and fathers with young children experienced the same change in employment as comparable people without young children. These estimates indicate that a negligible share of the overall decline in employment can be attributed to challenges specific to parents with young children.
While school closures and ongoing childcare challenges have substantially burdened parents and children alike, they do not appear to be a meaningful driver of the slow employment recovery. This means that the factors responsible for the slow employment recovery and depressed labor supply are issues that are not exclusively related to the struggles of working parents, such as the continued concern about the threat of getting COVID-19 at work or expanded unemployment insurance benefits and eligibility.
Furthermore, the fact that aggregate job losses do not appear to be explained by excess job losses among mothers with young children (after accounting for other factors like age and education) does not mean that mothers with young children have not been especially burdened over the past year. The fact that these women did not disproportionately retreat from the workforce in substantially larger numbers, while shouldering increased childcare and educational responsibilities for their children, might very well be an indication of excess burden that represents a shortcoming of the safety net established to respond to the COVID-19 crisis.
Melissa S. Kearney is the Neil Moskowitz Professor of Economics at the University of Maryland and director of the Aspen Economic Strategy Group.
This analysis was updated on June 21, 2021 to address the fact that the industry variable is missing for the vast majority of non-workers in the IPUMS data used in this analysis; that data issue makes models adjusting for industry unreliable.
The data underlying this analysis are available here. These data include a subsample of the IPUMS CPS data, which are cited as Sarah Flood, Miriam King, Renae Rodgers, Steven Ruggles and J. Robert Warren. Integrated Public Use Microdata Series, Current Population Survey: Version 8.0 [dataset]. Minneapolis, MN: IPUMS, 2020. https://doi.org/10.18128/D030.V8.0. The IPUMS CPS data are intended for replicating this publication only. Individuals analyzing the data for other purposes must submit a separate data extract request directly via IPUMS CPS. Individuals are not to redistribute the data without permission. Contact [email protected] for redistribution requests.