Skip to main content

Child care attendance plummeted well before state-ordered shutdowns

Analysis shows that COVID-19 child care closures went into effect after child care attendance had already fallen

September 28, 2022


Ryan Nunn Assistant Vice President, Community Development and Engagement
Vanessa Palmer Data Director, Center for Indian Country Development
Child care attendance plummeted well before state ordered shutdowns
Yann Roulet/EyeEm/Getty Images

Article Highlights

  • Closure mandates did not drive initial child care attendance declines
  • COVID-19 appears to have diminishing influence over attendance
  • Attendance remains below pre-pandemic levels
Child care attendance plummeted well before state-ordered shutdowns

In early 2020, in the opening stages of the COVID-19 pandemic, the child care sector contracted dramatically. Unclear at the time was the extent to which the enrollment drop was driven by labor market disruption and fear of COVID-19 exposure or state policy actions such as stay-at-home orders and child care shutdowns. To disentangle these factors, we examined state-level child-care-closure-policy information and COVID-19 wastewater monitoring data alongside new, confidential data on child care attendance.1 We find that the steep decline in child care attendance in the spring of 2020 began well before states implemented stay-at-home orders and child care shutdowns. This suggests that those policies could not have caused the shift, since it was already well under way. As the pandemic moves toward patterns of endemic spikes and declines, it is essential to understand the relationship between state policy actions and child care attendance, and any implications for parental labor force participation. As of the spring of 2022, the data show a child care sector with persistently lower attendance than before the pandemic but considerably less sensitivity to spikes in COVID-19.

Closures came after attendance plummeted

We began by aligning average daily child care attendance per center with the timing of state shutdown orders, looking only at the 19 states that eventually implemented child care shutdowns. (For further details on our methodology and a list of the 19 states, see the appendix.) These states closed programs between March and June 2020, with closures ranging in duration from a week in Alabama to three months in Massachusetts.2 See Figure 1. (In this figure, you can toggle between showing results for just the 19 states that implemented shutdowns and for all states that declared an emergency and/or implemented a stay-at-home order.) Because shutdowns were often brief—a median of 66 days—we used daily attendance data provided by Procare Solutions, a child care management software provider, to compare the precise timing of the various factors in play.

Because of the variation in the timing of state policies, we organized the data in terms of days before and after a given policy change rather than calendar dates.3 We find that average attendance started to fall steeply about three weeks before state-mandated child care shutdowns (yellow line, Figure 1). When those measures were put in place, average attendance was already near its low point.


Daily child care attendance relative to timing of state-level policy changes

What might account for this pattern? One possibility is that general stay-at-home orders, rather than child care-specific shutdowns, were the driving force. That would be consistent with prior research finding that stay-at-home orders substantially reduced online job postings for center-based child care providers. However, as with child care closures, we find that the marked drop in child care attendance preceded stay-at-home orders (orange line, Figure 1) and could not have been caused by them.4

Another possibility is that center administrators and parents responded directly to the anticipated or ongoing rise of COVID-19 in their communities—as reflected in emergency declarations, employers’ introduction of work-from-home arrangements, and COVID-19 spread—rather than policy actions prompted by COVID-19. Examining attendance relative to states’ initial emergency declaration dates supports this relationship (red line, Figure 1). To better understand the role of community COVID-19 burden, we employ a comparable measure over time and space: SARS-CoV-2 genome copy concentration in wastewater as publicly provided by Biobot, a wastewater monitoring firm.5 In Figure 2, we overlay average attendance with COVID-19 viral level.

Child care attendance fell from about 60 children per center in late February 2020 to fewer than 20 children per center in early April 2020. This drop-off varied across states, ranging from a 51 percent decline in Nebraska to a 97 percent decline in Massachusetts over this period. Drops in attendance mirrored increases in COVID-19 viral presence (Figure 2). However, the relationship in time between the two is not definitive due to the relatively small number of states with wastewater surveillance in place at this early stage of the pandemic. COVID-19 viral levels spiked in late March 2020 and remained high for about a month, subsiding as the summer began. By contrast, child care attendance picked up slowly from April through June.* Pooling all states and days across the period from February through June 2020, we find that closure, COVID-19 viral level, and observed attendance are correlated as might be expected: On those days on which a state had a child care shutdown in effect, COVID-19 viral level was, on average, over two and a half times higher and child care attendance was less than half the level prevailing when no shutdown was in effect.


Weekly child care attendance and COVID-19 viral level relative to state-level policy dates

Child care matters for children and parents

These trends had and continue to have severe implications for the well-being of children and parents. The full effects on children are not yet understood, but early indications (for older children) are that substantial learning loss occurred when schools were closed. One study found that parental reports of problem behavior in younger children were not more common when parents took on more child care duties. However, parental COVID-19 infection itself was strongly associated with behavioral problems.

As was clear to any working parent of young children, the pandemic caused a severe time crunch for parents. The burden of unanticipated child care fell disproportionately on mothers, leading many to exit the labor force. Research by economists Lauren Russell and Chuxuan Sun has indicated that some of the labor force exits may be due to policy actions such as child care shutdown orders. But any such effect would have had to occur after the initial attendance declines, given the patterns shown in figures 1 and 2.

We provide additional insight into these dynamics by comparing labor force participation of men and women with and without children during the April to June 2020 period when child care shutdowns were in effect. We then separate the analysis by states that never issued child care shutdowns and those that did. Labor force participation plummeted for all groups, but the decline was sharpest for mothers: 10.5 percentage points lower in April 2020 than in February 2020. Comparing ever-closed versus never-closed states, we find that declines in labor force participation were greater in ever-closed states for all groups. See Figure 3. However, these differences in declines approached or reached statistical significance only for individuals without children. Overall, we do not interpret this pattern of results as clearly suggesting any effect of closure orders on parental labor force participation.

Labor force participation in states that ever or never issued child care closure mandates

Source: Authors’ calculations using information as noted in the appendix, restricted to individuals ages 25 through 54.

Child care attendance remains below pre-pandemic levels

The child care sector remains in a depressed state compared to pre-pandemic activity. Average attendance per center in March 2022 stood at only 85 percent of that in March 2019 (Figure 2, zoomed out). While we find that attendance in states that closed at some point was lower than in never-closed states before the pandemic, it does not appear that differences in attendance between ever-closed states and never-closed states are significantly larger or smaller today than before the pandemic.

The relationship among COVID-19 viral levels, child care attendance, and state-level child care closure policies has changed over time, as has states’ capacity to assess it.6 Spikes in COVID-19 viral levels in winter 2020, fall 2021, and winter 2021 coincided with visible dips in child care attendance. But these later decreases were comparatively more modest than in the early months of the pandemic.

Future COVID-19 variants or other infectious diseases could change this situation and give rise to policy actions like the shutdowns implemented in the spring of 2020. As policymakers consider such possibilities—and the need for further investments in public health and the resiliency of the child care sector—they should do so with a clear understanding of the pandemic’s past and present.

The authors thank Procare Solutions for data provided; Rob Grunewald for support in compiling information on child care closure policies; and Chris Herbst, Lauren Russell, and Brian Shurley for insightful feedback on an earlier draft.

Appendix: About the data

Information on states’ child care closure policies was compiled from the COVID-19 US State Policy (CUSP) database, the Hunt Institute’s State Child Care Actions database, the U.S. Department of Health and Human Services National Snapshot of State Agency Approaches to Child Care During the COVID-19 Pandemic report, and states’ governmental websites. Emergency declaration and stay-at-home order dates were drawn from CUSP. As of the date of publication, Biobot, a wastewater monitoring firm, claims to offer the largest publicly available dataset on SARS-CoV-2 concentrations in wastewater. This measure of COVID-19 viral presence is given as the three-sample, population-weighted average of weekly normalized concentrations in effective genome copies per milliliter of wastewater. In our analysis, we used an average at the state-week level. Our child care attendance measure, given at the state-day level, was total attendance divided by the number of centers reporting attendance on that day. We included weekends and holidays, except where explicitly noted, but omitted those dates on which fewer than three centers in a state reported attendance. As a measure of labor market participation, we used data at the state-month level from the Bureau of Labor Statistics Current Population Survey (CPS), accessed via IPUMS-CPS. The CPS records employment data from the week containing the twelfth day of each month, making April 2020 the first month for which employment data are available after child care closures began taking effect in March. Thus, to shed light on how child care shutdown decisions correlate with labor market outcomes, we analyzed state-level labor outcomes by states’ ever-closed and never-closed status rather than shutdown dates. The 19 states we identified as having implemented child care shutdowns at some point in the pandemic were Alabama, Delaware, Hawaii, Illinois, Kentucky, Maryland, Massachusetts, Michigan, New Jersey, New Mexico, North Carolina, Ohio, Oregon, Pennsylvania, Rhode Island, Texas, Vermont, West Virginia, and Wyoming. States’ child care shutdown orders went into effect during an 18-day window in the latter part of March 2020; reopening dates spanned an 87-day window.

Note: Limitations of our analysis

* Our analysis has several meaningful limitations. Not all states had wastewater monitoring programs in place during the early stages of the pandemic, placing constraints on the times and geographies in which COVID-19 viral level can be used as a measure. While wastewater data represent COVID-19 prevalence in a particular geography, they do not characterize the resulting degree of local system-level stress. For example, the same COVID-19 viral level may reflect a case burden that one community’s health care system could accommodate, but that was overwhelming to another’s. Thus, the social burden and perceived risk of COVID-19 in different communities at similar viral measurements may have varied from place to place in ways that influenced both labor and child care choices. In addition, while the proprietary child care attendance dataset we used provides a new view of how families experienced the pandemic, it does not capture all child care attendance nationally (and, in particular, does not tend to capture home-based care). If a center shuts down entirely, it falls out of our sample. Depending on how large the surviving child care centers are, this could skew our results (likely making them an underestimate of the true attendance decline).


1 The analysis includes proprietary, de-identified data from Procare Solutions.

2 Providers approved to serve essential workers with young children were excepted from closures. Rhode Island was the only state that closed child care programs without exception.

3 We calculate the unweighted average across states, rather than weighting by state population size or total attendance. See the appendix for further details.

4 These findings are potentially consistent, given that job postings and child care attendance are different variables with different timing. For example, it may be that policy actions like stay-at-home orders cause a reduction in postings that comes after the initial reduction in attendance.

5 By contrast, COVID-19 case counts are often inconsistent and may not be fully representative of a community due to variation in testing availability, take-up, and reporting over the course of the pandemic.

6 Many more states now have COVID-19 wastewater monitoring in place, and early-stage efforts to use these data in developing real-time surveillance and risk assessment programs are active. For the month of July 2022, publicly available data were available for all but four states and the District of Columbia through Biobot and for all but five states through the Centers for Disease Control and Prevention.

Ryan Nunn
Assistant Vice President, Community Development and Engagement
Ryan Nunn is an assistant vice president in the Minneapolis Fed’s Community Development and Engagement Department. Leading the Bank’s applied research function, Ryan works to improve outcomes for low- and moderate-income communities with the help of better evidence and analysis.
Vanessa Palmer
Data Director, Center for Indian Country Development

Vanessa Palmer is the data director for the Minneapolis Fed’s Center for Indian Country Development (CICD), where she leads efforts to collect, harmonize, and sustainably manage research-ready data in support of economic self-determination in Indian Country. In addition, she uses statistical tools and data visualization to support CICD’s applied research work.