Impacts of the COVID-19 pandemic on the child care sector: Evidence from North Carolina

Abstract

This study provides a comprehensive, census-level evaluation of the impacts of the COVID-19 pandemic on the county child care market in a large and diverse state, North Carolina, and the disproportionate impacts of the pandemic on different types of providers and communities. We use county-level panel data from 2016 to 2020 and a difference-in-differences design to isolate the effects of the pandemic from unobservable seasonal trends in enrollments and closures. We found that the COVID-19 pandemic reduced county-level child care enrollment by 40% and the number of providers by 2% as of December 2020. Heterogeneity analyses revealed that the family child care sector experienced not only less severe reductions in enrollment and closure than center providers, but also a small growth in the number of family providers. Declines in enrollment were most substantial for preschool-aged children. There was a significant drop in the number of 5-star providers and an increase in the number of lower-quality providers. Provider closures were more concentrated in communities with a higher percentage of Hispanic residents. Higher-SES communities experienced larger drops in enrollment as well as provider closures. Implications for child development and future research and policies are discussed.

1. Introduction

Stable, high-quality child care is critical for young children's learning and development, particularly those from socioeconomically, racially, and geographically disadvantaged groups (Magnuson & Duncan, 2016). Yet the COVID-19 pandemic created acute challenges for the child care sector, with sharp drops in enrollment, widespread provider closures, and increasing needs for health, safety, and remote learning (Weiland et al., 2021). Although many studies have described the impacts of the pandemic on child care, most of them rely on survey data or simple cross-sectional differences measured before and after the pandemic hit in 2020. In turn, these studies are subject to bias from low response rates, poor sampling schemes, or other confounding events. Given the scientific limitations of the prior work and the high stakes of COVID-related policy decisions, there is an urgent need for rigorous evidence and independent research on the pandemic's impacts on the child care sector.

The purpose of this study is to provide a more accurate estimate of the impacts of the pandemic on child care enrollment and provider closure using administrative data and a quasi-experimental approach (i.e., differences-in-differences). Specifically, we use county-level records of enrollment and opening status of licensed providers in a large and diverse state, North Carolina (NC). We define the onset of COVID-19 in March 2020 when NC mandated its first Stay-At-Home-Order (SAHO) and use data from 2 seasonal time points (February and December) in the years leading up to, and during the pandemic (2016–2020). Because enrollment and provider closure are potentially influenced by both supply and demand, our investigation captures the combined effects of the pandemic on child care demand and supply.

Another goal of this study is to estimate the impacts of the pandemic on different types of child care providers and communities. Combining the NC data with county characteristics from the American Community Survey (ACS), we provide 1 of the first regression-adjusted estimates of any potentially widening inequities across an array of provider and county characteristics, including family- or center-based care, age groups served, provider quality, communities of color, county socioeconomic status, and urbanicity. In doing so, this study helps to identify which child care providers and communities are the most in need of immediate and long-term investments.

North Carolina provides a useful context because they adopted their SAHO on March 30, 2020, and lifted it on May 22, 2020, which is the typical length of SAHOs in most states (Raifman et al., 2020). North Carolina also has a growth trajectory of COVID-19 cases that closely mirrors the national trends during the study period.1 A unique advantage of the NC data is that a majority of the providers reported their enrollment and opening status during the COVID-19 pandemic, resulting in a much higher response rate than prior studies (∼ 90%).2 It is also true that NC is often considered a national leader in child care, with more robust state investments and a stronger child care infrastructure to “weather the storm” than many other states, which may limit our study's generalizability. Nevertheless, our study's aims and findings address a critical gap in our understanding of COVID-19′s impacts on child care providers, and the communities in which they are located.

1.1. Impacts of the COVID-19 pandemic on the child care sector

The COVID-19 pandemic and nationwide lockdowns have led to unprecedented challenges to the child care sector by affecting both the demand for and supply of child care, aligned with the overall trends in unemployment and virtual work in the US (Cascio, 2021; Weiland et al., 2021). Increases in unemployment and labor market exits that lowered families’ incomes, along with increases in telecommuting, both reduced child care demand (Albanesi & Kim, 2021; Brynjolfsson et al., 2020; Chetty, Friedman, Hendren, Stepner & Team, 2020). Parents might also become more reluctant to enroll in child care due to fears for safety and health (Cascio, 2021; Weiland et al., 2021). On the supply side, the variety of state mandates and guidelines—including temporary closure in the early months of the pandemic—restricted child care capacity but also increased safety procedures, reducing revenues while raising operating costs. Families thus have less access to child care at a price they are willing to pay (Cascio, 2021; Weiland et al., 2021). Additionally, small businesses, especially those that provide in-person services, experienced significantly larger drops in revenue and higher rates of closure than larger firms (Bartik et al., 2020; Bloom, Fletcher & Yeh, 2021; Chetty et al., 2020).

Weiland et al. (2021) conducted a systematic review of the literature on COVID's short-term impacts on the child care sector, and reported that the decline in birth to age 5 enrollment and the number of open providers ranged from 20% to 50% at different time points and in different states. Lee and Parolin (2021)) used anonymized cellphone location data and estimated the national monthly change in total visits to child care establishments in 2020 and 2021, compared to the same month in 2019. Similar to the survey results, they found that 70% of child care centers were closed or had reduced capacity (i.e., less visits) in April 2020. This rate gradually declined in later months of the pandemic and stayed at around 40% from December 2020 to April 2021.

Although there exist many reports of the damaging impacts of the pandemic on the child care sector, they are largely correlational, or use highly selected samples through the use of non-random, unrepresentative sampling approaches with low response rates. Additionally, most studies only compare the cross-sectional differences before and after the pandemic hit in 2020 (e.g., Barnett & Jung, 2021; Bassok, Markowitz, Smith & Kiscaden, 2020). According to the Economic Census, (2020a, 2020b), the number of child care establishments decreased from 768,521 in 2012 to 667,168 in 2018, with an annual rate of decrease ranging from 1% to 9%.3 This suggests that simple pre-post estimates of COVID's impacts on the child care sector may overestimate its effects without accounting for the seasonal changes and existing trends in the sector.

To the best of our knowledge, Ali, Herbst and Makridis (2021) is the only study that accounted for state characteristics and time trends when estimating the effects of COVID-19 on the child care sector, using online job postings as a measure of labor demand. The authors used a difference-in-differences design, exploiting variation in states’ timing of SAHOs, and found a 16% decrease in child care job postings per day. This finding remained robust when accounting for parental demand as measured by online child care searches, suggesting that the child care sector did experience a decrease in supply in the short term.

1.2. Disproportionate impacts on different providers

While the pandemic has affected the entire child care sector, its impacts have likely been unequally borne by certain types of providers. Particularly, the pandemic may have had a more profound effect on family child care providers, who had limited access to the Paycheck Protection Program and thus faced more severe financial instability than center providers (Weiland et al., 2021). On the other hand, family child care providers typically offer more flexible services and can quickly adapt their practices to meet the changing conditions and parental needs (National Survey of Early Care & Education Project Team, 2016a), a big advantage during a crisis. Furthermore, parents may have preferred enrolling their children in family child care settings for smaller group sizes and lower total enrollment than centers (Porter, Bromer, Melvin, Ragonese-Barnes & Molloy, 2020). Indeed, national and state surveys have consistently reported that family child care providers were more likely than child care centers to stay open during the pandemic (Department of Health & Human Services, 2020; Porter et al., 2020). Lower tuition costs at family providers may also be more tenable for parents experiencing financial struggles. We examine how the totality of such countervailing hypotheses impacted the enrollment and availability of these 2 types of care.

The types of care that are the most costly to provide may also be hit the hardest by the pandemic. In particular, providers serving infants and toddlers may be disproportionately affected due to the high labor cost of serving this age group (Tekin, 2021). Parents may be more likely to keep their infants and toddlers at home for safety. Higher quality providers tend to have higher operating costs, and therefore may be at a greater risk of closure than lower-quality providers when enrollment drops. Parents are usually more sensitive to convenience and costs than quality when making child care decisions (National Survey of Early Care & Education Project Team, 2014), so those who experienced financial difficulties may settle for lower-quality but less expensive care. Any potential decline in the supply of high-quality providers is concerning, particularly for disadvantaged communities, which have already had limited high-quality child care options (Bassok & Galdo, 2016). Understanding which providers are most impacted by the pandemic can help policymakers create targeted plans to assist the most financially vulnerable providers in the child care sector.

1.3. Disproportionate impacts on different communities

Disparities in ECE availability, affordability, and quality across communities in different socioeconomic strata and geographical locations existed long before the pandemic. For example, there are fewer high-quality ECE opportunities in low-income communities (Bassok & Galdo, 2016; Bassok, Fitzpatrick & Loeb, 2011; Li-Grining & Coley, 2006; National Survey of Early Care & Education Project Team, 2016b) and in communities of color (Bassok & Galdo, 2016; Hatfield, Lower, Cassidy & Faldowski, 2015), where providers are less likely to participate in state Quality Rating and Information Systems (Jenkins, Duer & Connors, 2021). Rural areas also have fewer child care slots per demanded slot compared with urban areas (Henning-Smith & Kozhimannil, 2016).

The pandemic has likely exacerbated these pre-existing economic, racial, and geographical inequities. Studies indicate that child care enrollment drops were more pronounced among low-income, Black, Hispanic, and dual language learner children (Barnett & Jung, 2021; Bassok & Shapiro, 2021; Weiland et al., 2021). Moreover, children of color were more likely to experience child care closure in their communities compared to White children (Cascio, 2021; Lee & Parolin, 2021). However, no study has provided regression-adjusted estimates to quantify the differential impacts of the pandemic on different types of providers and communities, as we do here.

1.4. North carolina's state response prior to December 2020

In NC's state-level response to COVID-19, the Governor signed Executive Order 121, which officially mandated the first state SAHO on March 30, 2020. The state lifted restrictions throughout the year during different reopening phases. Child care services were initially mandated to close for regular business, and then allowed to open to serve all children on May 20, 2020, as long as they followed health and safety guidelines (Executive Order 141). With a new wave of COVID-19 infections hitting NC in the last months of 2020, the Governor called for a modified SAHO on December 11, 2020, requiring that residents stay at home during overnight hours.

According to a report of state actions by the Bipartisan Policy Center (2020), NC child care providers received grant disbursements totaling $118 million from the Coronavirus Aid, Relief, and Economic Security (CARES) Act. The funding per pupil was on the higher end of the financial support awards granted to states (i.e., relative to NC's 0- to 4-year-old population).4 North Carolina estimated that 70% of this funding was used to support child care providers who had temporarily closed or seen a decrease in enrollment. Another 25% of the funding was distributed to cover expenses to provide child care for essential workers. The CARES Act funding and another $60 million of state supplemental funding supported all child care providers, open or closed, for April through June. North Carolina also disbursed 2 operational grants to support providers that were open during the pandemic. Emergency child care centers that opened in April for essential workers received between $500 and $30,000 per month opened, or between $359 and $2500 for family child care homes. In September, the Governor announced $35 million in operational grants to providers offering in-person services during August, September, and October (House Bill 1105, 2020). Overall, NC has been relatively responsive in the child care sector during the pandemic.

1.5. Present study

This study evaluates the impacts of the COVID-19 pandemic on the county-level child care market in the first year of the pandemic (2020) in NC, a large and diverse state that offered substantial supports to providers during the crisis. We take advantage of the existing administrative data before and during the pandemic and use a difference-in-differences approach to provide more accurate estimates of the impacts than prior studies. To date, documentations of the damaging impacts of COVID-19 on the child care sector are mostly descriptive, using survey data with small samples of providers and relatively low response rates. Furthermore, no study has provided regression-adjusted estimates on the differential impacts of the pandemic on different types of providers and communities. Our study fills these gaps by exploring the following research questions:

• 1.What were the impacts of the COVID-19 pandemic on the child care market (i.e., number of open providers and enrollment) at the county level in 2020?
• 2.Did the impacts vary by provider type (home/center), age group (infants and toddlers, preschoolers, school-age children), and provider quality (measured by quality ratings)?
• 3.Did the impacts vary by county characteristics (i.e., percentage of Black residents, percentage of Hispanic residents, socioeconomic status, and urbanicity)?

2. Method

2.1. Data

Our outcome data come from 2 main sources. For years 2016 to 2019, we use monthly administrative data of all licensed child care providers in NC serving ages 0 to 12. The data were collected by the Division of Child Development and Early Education (DCDEE), from the NC Department of Health and Human Services, and included detailed information about enrollment, capacity, license type, and the quality rating of all licensed providers. This information was collected through monitoring visits by the DCDEE at least 1 or 2 times each year.

Since these visits were interrupted from March to September 2020 due to the pandemic, our 2020 data come from a daily survey of providers that was initially collected by the DCDEE and then continued by a child care resource and referral agency on a weekly basis. These data were collected at the provider level and provided to us at the county level (n = 100) at 2 time points: February and December 2020.

To combine our 2 sources of data, we aggregate the 2016–2019 provider data to the county- and year-level and keep the February and December data in each year to match the 2020 data. Our final dataset captures each of the 100 counties from 2016–2020, with 2 time points (i.e., February and December) per county and year, giving us a total sample size of 1000 county-month-year observations.

The primary concern from combining the 2 data sources is that the provider survey in 2020 was voluntary, so not all providers responded to the survey. However, because monitoring visits were conducted until the beginning of March 2020, we are able to cross-validate our 2 sources of data using February 2020 data. We confirm that these 2 sources overlap substantially in February 2020, suggesting a much higher response rate from providers than reported in prior studies (e.g., Barnett et al., 2021; Bassok et al., 2020). Specifically, by comparing the number of providers in the survey data with the statistics in the administrative data, we find that the average response rate across counties is 93%. Fifty percent of the counties have a response rate of nearly 100%, with the lowest response rate in a county being 70%.5 Because the DCDEE strongly encouraged providers to report enrollment and some of the relief funding was tied to opening status, providers in NC might have been more likely to respond to the survey. The responses rates are high for both center-based and home providers, although they are slightly higher for home providers. More details are available in Appendix A.

Child care providers included in our analyses are all licensed providers. Characteristics of these providers are presented in Appendix Table B1, based on the February 2020 administrative data. About 3 out of 4 providers were center-based, while the remaining were family home providers. The average center enrolled 53 children, had a maximum capacity of 81 children, and with 9 employees. The average home provider enrolled 5 children, had a maximum capacity of 7 children, with only 1 employee. While almost all home providers were categorized as private providers, center-based facilities included private centers (53.2%), public schools (26.7%), religious-sponsored organizations (11.9%), Head Start (6.4%), and others (1.8%).6

County demographic characteristics for our heterogeneity analyses come from the ACS 2018 5-year-estimates, the latest available ACS estimates when the study was conducted. We include the percentage of the population who are Black, the percentage of the population who are Hispanic, urbanicity, and a socioeconomic status (SES) index that combines the following measures: the percentage of the population with a Bachelor's degree or higher, poverty rate, unemployment rate, the percentage of families with a female householder, the percentage of households receiving SNAP benefits (food stamps), and the (logged) median household income. We create the SES index by averaging the standardized values of the 6 measures following Fahle et al. (2021). In our heterogeneity analyses by county characteristics, we divide counties into SES quartiles and distinguish between counties that have a Black or Hispanic population above and below the state mean (20% for Black population and 7% for Hispanic).

Our outcome variables are child enrollment and number of providers at the county level. Table 1 displays the descriptive statistics of the county averages of the enrollment and number of providers for February and December 2020 (similar statistics for previous years are presented in Appendix Table B2). Taking the raw differences between the February and December 2020 data, we find that the pandemic was associated with a 41% decrease in county-level enrollment and a 6% decrease in the number of providers. Fig. 1 presents the county-level variation in the decrease in enrollment and number of providers. While the majority of the counties experienced a decline in enrollment ranging from 20% to 60%, some counties experienced substantially larger drops in enrollment. There are also some variations in provider closure across counties, with most counties having less than 10% of provider closure. Additionally, the drop in enrollment was larger among center providers and 3- to 5-year-olds. Provider closures appear to affect family providers and providers of lower quality more. Below, we show that our regression-adjusted DID estimates provide more accurate estimates; not accounting for seasonal fluctuations in enrollment and closure will result in overestimation and misrepresentation of the impacts of the pandemic.

Table 1.

County-level descriptive statistics, February and December 2020.

Variable	February 2020 mean	December 2020 mean	Difference	% change
Total Enrollment	2150.5	1270.3	−880.2	−40.9%
	(3867.73)	(2264.75)
Number of providers	52.4	49.4	−3.1	−5.9%
	(80.54)	(75.78)
Centers vs family homes
Enrollment in child care centres	2075.7	1217.1	−858.6	−41.4%
	(3750.42)	(2184.79)
Enrollment in family home care	74.8	53.2	−21.5	−28.8%
	(128.53)	(91.11)
Number of centre providers	38.8	36.8	−2.0	−5.2%
	(56.48)	(54.04)
Number of family providers	13.6	12.6	−1.1	−8.1%
	(25.31)	(22.96)
Ages
Enrollment 0–2 y.o.	678.3	437.5	−240.8	−35.5%
	(1421.18)	(847.58)
Enrollment 3–5 y.o.	1190.5	607.8	−582.7	−48.9%
	(2018.86)	(1077.28)
Enrollment school-aged children	281.7	225.1	−56.6	−20.1%
	(457.80)	(356.62)
Quality
Number of 1-star providers	2.3	2.1	−0.2	−8.8%
	(5.96)	(5.52)
Number of 2-stars providers	1.2	1.2	−0.1	−8.1%
	(2.57)	(2.49)
Number of 3-stars providers	11.3	11.0	−0.2	−1.8%
	(18.03)	(17.50)
Number of 4-stars providers	9.8	9.6	−0.2	−2.0%
	(17.61)	(17.07)
Number of 5-stars providers	20.3	19.1	−1.1	−5.4%
	(28.65)	(27.67)
Observations	100	100	100	100

Note. Standard deviations in parenthesis.

Fig. 1.

Change in (A) Enrollment and (B) Number of Providers by County between February and December 2020.

2.2. Empirical strategy

To measure the effects of the COVID-19 pandemic, we employ a difference-in-differences (DID) model, exploiting 2 levels of temporal variations. The first difference compares the county-level enrollment and number of providers before (February) and after (December) the pandemic in 2020. The second difference compares the February-December changes in 2020 with the same months in pre-pandemic years (2016–2019). In other words, we adjust the raw difference between February and December 2020 by controlling for the changes between February and December in previous years. Our model is:

$$\begin{array}{ccc}\hfill Y_{cmy}& =& \sum _{t=2016}^{t=2018}{\beta }_y\left(December*I\left[year=t\right]\right)\hfill \\ & & +{\beta }_{2020}\left(December*I\left[year=2020\right]\right)\hfill \\ & & +\gamma Decembe{r}_m+\sum _{t=2016,t\ne 2019}^{t=2020}\delta \left(I\left[year=t\right]\right)+{\epsilon }_{cmy}\hfill \end{array}$$ (1)

where $Y_{cmy}$ denotes an outcome measure for county c in month m and year y. December represents an indicator for the month of December (February is omitted). I[year=t] are indicator variables for pre-treatment years 2016 to 2018. Year 2019 is omitted as the reference year. ${\beta }_{2020}$is our coefficient of interest and can be interpreted as the difference in the averages across counties in outcome Y between December and February 2020 relative to the difference between December and February in 2019, our reference year, accounting for the differences in years 2016 to 2018. Standard errors are clustered at the county level. For county heterogeneity analyses, we split the sample by county characteristics and run the same models in Eq. (1) on the subsamples of counties, where the outcomes are aggregated for a subset of providers.

We rely on the assumption that changes in outcomes in February over time reflect the same trend as outcomes measured in December would show over time, had the pandemic not occurred in December 2020. In other words, we assume that differences between February and December are constant in the 2016 to 2019 period (i.e., the parallel-trends assumption). Then, we attribute the differences in outcomes between February and December 2020, once accounting for the time trends in previous years, to the COVID-19 pandemic. Empirically, we observe that the parallel-trends assumption is met when the coefficients ${\beta }_y$in Eq. (1) are equal to zero. As shown in Fig. 2 , using 2019 as the reference year, we found that the ${\beta }_y$coefficients are indeed not significantly different from zero in years prior to 2019, suggesting no systematic pre-trends in previous years. The estimate in 2020 marks a clear break from previous years because of the pandemic.

Fig. 2.

Changes in (A) Enrollment and (B) Number of Providers between February and December by Year, Relative to the 2019 February-December Change (All Providers and by Centers and Homes).

Notes: The figures illustrate the βy coefficients (i.e., December-February difference in a given year) relative to the December-February difference in 2019 (the reference year) from Eq. (1).

3. Results

3.1. Overall effects of the COVID-19 pandemic on the child care sector

We first present the DID estimates of the COVID-19 pandemic on county-level enrollment and the number of providers (Table 2 ). Columns 1 and 3 show the results using only the first difference, i.e., the difference between February and December 2020. Columns 2 and 4 show the results using our preferred specification, which includes previous years and was estimated using Eq. (1). On average, we found a reduction of 872 enrolled children per county between February and December 2020, relative to 2019. This represents a 40.6% decrease in child care enrollment from a baseline of 2150 enrolled children per county before the pandemic. Although we found large drops in enrollment, the number of providers were only mildly affected. On average, only 1 provider per county closed in this time period, equivalent to a 2.2% decrease in the number of providers available in a county. The simple difference estimate between February and December 2020 in Table 1 suggests that, on average, there was a 6% decrease in the number of providers in each county. As shown in Table 2, failing to account for seasonal differences would thus lead to an overestimation of the pandemic's effects on provider closures by almost 3 times.

Table 2.

Main effects of the COVID-19 pandemic on total enrollment and number of providers.

	Total enrollment		Number of providers
	(1)	(2)	(3)	(4)
	Yr 2020	Yr 2016–2020	Yr 2020	Yr 2016–2020
Panel A. All providers
December	−880.15⁎⁎	−8.14	−3.06⁎⁎	−1.91⁎⁎
	(162.996)	(15.447)	(0.546)	(0.338)
December*2020 coefficient		−872.010⁎⁎		−1.150*
		(170.896)		(0.511)
February 2020 mean	2150.50	2150.50	52.42	52.42
Effect in% (A/B)	−40.9%	−40.6%	−5.8%	−2.2%
Panel B. Centers
December	−858.61⁎⁎	−0.680	−2.01⁎⁎	−0.34
	(159.236)	(15.976)	(0.323)	(0.231)
December*2020 coefficient		−857.930⁎⁎		−1.670⁎⁎
		(168.589)		(0.467)
February 2020 mean	2075.72	2075.72	38.79	38.79
Effect in% (A/B)	−41.4%	−41.3%	−5.2%	−4.3%
Panel C. Homes
December	−21.54⁎⁎	−7.46⁎⁎	−1.05⁎⁎	−1.570⁎⁎
	(4.496)	(1.755)	(0.263)	(0.309)
December*2020 coefficient		−14.080⁎⁎		0.520*
		(3.462)		(0.220)
February 2020 mean	74.78	74.78		13.63
Effect in% (A/B)	−28.8%	−18.8%	−7.7%	3.8%
Observations	200	1000	200	1000

Note. Standard errors in parentheses. Columns (1) and (3) show the results of regressions that were estimated using only 2020 data with a dummy variable that indicates the mo of December (i.e., it represents the simple difference in the outcome between December and February 2020). Columns (2) and (4) correspond to the DID strategy, which includes 2016–2019 data to adjust for pre-trends.

⁺P < 0.10.

^⁎P < 0.05.

^⁎⁎P < 0.01.

3.2. Heterogeneity by provider characteristics

Our heterogeneity analyses reveal substantial differences in how different types of providers were affected by the COVID-19 pandemic (Table 3 ). Panel A of Table 3 presents the heterogeneous effects of COVID-19 on enrollment by provider type (i.e., child care centers vs homes) and age groups, respectively, and then by age groups within each provider type. We found that, while enrollment in child care centers decreased by an average of 858 children per county (corresponding to a 41.3% reduction), enrollment in family child care homes decreased by an average of 14 children per county (a 18.8% reduction). Enrollment reductions affected providers serving children from all age groups. However, we found steeper reductions in the enrollment of 3- to 5-year-old children compared with the enrollment of infants and toddlers and school-aged children (−49.3% vs −35.0% vs −29.9%, respectively, P < 0.01). Further analyses by age groups within centers and homes reveal that these reductions were primarily driven by center-based providers rather than family homes. Although the differences in enrollment reductions between centers and homes were not statistically significant for infants and toddlers and school-aged children, center-based providers serving preschoolers experienced a reduction of 50.1%, while home-based providers serving the same age group only experienced a 7.4% reduction in their enrollment.

Table 3.

Effects of the COVID-19 pandemic on enrollment and number of providers by provider characteristics.

Panel A. Enrollment
						0–2 y.o.		3–4–5 y.o.		School-aged
	Centres	Homes	0–2 y.o.	3–4–5 y.o.	School-aged	Centres	Homes	Centres	Homes	Centres	Homes
A) December*2020 coefficient	−857.930⁎⁎	−14.080⁎⁎	−237.500⁎⁎	−586.410⁎⁎	−48.100⁎⁎	−227.640⁎⁎	−9.860⁎⁎	−584.730⁎⁎	−1.680	−45.560⁎⁎	−2.540+
	(168.589)	(3.462)	(60.662)	(102.058)	(13.500)	(58.109)	(2.958)	(101.853)	(1.067)	(13.525)	(1.408)
B) February 2020 mean	2075.72	74.78	678.31	1190.52	281.67	645.32	32.99	1167.67	22.85	262.73	18.94
C) Effect in% (A/B)	−41.3%	−18.8%	−35.0%	−49.3%	−17.08%	−35.3%	−29.9%	−50.1%	−7.4%	−17.3%	−13.4%
D) Diff. P-value	0.000		0.000			0.261		0.000		0.672
Observations	1000	1000	1000	1000	1000	1000	1000	1000	1000	1000	1000
Panel B. Number of Providers
						1–2 stars		3–4 stars		5-stars
	Centres	Homes	1–2 stars	3–4 stars	5-stars	Centres	Homes	Centres	Homes	Centres	Homes
A) December*2020 coefficient	−1.670⁎⁎	0.520*	0.040	1.060⁎⁎	−0.900⁎⁎	0.020	0.020	0.480+	0.580⁎⁎	−0.920⁎⁎	0.020
	(0.467)	(0.220)	(0.183)	(0.304)	(0.269)	(0.116)	(0.113)	(0.274)	(0.220)	(0.258)	(0.103)
B) February 2020 mean	38.79	13.63	3.51	23.34	20.29	1.09	2.42	14.15	9.19	18.60	1.69
C) Effect in% (A/B)	−4.3%	3.8%	1.1%	4.5%	−4.4%	1.8%	0.8%	3.4%	6.3%	−5.0%	1.2%
D) Diff. P-value	0.000		0.000			0.922		0.395		0.334
Observations	1000	1000	1000	1000	1000	1000	1000	1000	1000	1000	1000

Note. Standard errors in parentheses. Regressions were estimated using Eq. (1).

⁺P < 0.10.

^⁎P < 0.05.

^⁎⁎P < 0.01.

Panel B of Table 3 shows the differential effects of COVID-19 on the number of providers in each county by provider type, quality rating, respectively, and then by quality ratings within each provider type. An average of 1.67 center-based providers were closed by December relative to February 2020 (a 4.3% decrease) after we control for time trends in previous years, while the number of family child care providers increased by 3.8%. This is in stark contrast to the simple difference estimate, which suggests a 7.7% decrease in the number of family child care providers.

As to the quality of providers defined by the QRIS rating, we found that closures affected 5-star (the highest tier in the QRIS) providers only, also different from the simple pre-post comparisons (Table 1, bottom panel). There was a decrease of 1 5-star provider on average per county, corresponding to a 4.4% reduction.7 Additionally, we found that the number of 3- and 4-star providers increased by 4.5%. The reduction in 5-star providers also appears to be driven by center-based providers, although the difference between centers and homes was not statistically significant.

3.3. Heterogeneity by county characteristics

We then investigate whether the effects of the COVID-19 pandemic varied across different county characteristics (Table 4 ). Panel A of Table 4 presents the differential effects on county-level enrollment by urbanicity, the percentage of Black residents, the percentage of Hispanic residents, and quartiles of the SES index. We found no significant differences in enrollment between counties based on their urbanicity or the percentage of Black and Hispanic residents. Reductions in enrollment were around 40% throughout our specifications. The pandemic did differentially affect enrollment in counties of varying SES, but enrollment reductions always ranged between 36.0% and 44.3% across SES quartiles. Counties in the top SES quartile experienced the largest drop in enrollment (−44.3%, P < 0.05).

Table 4.

Effects of COVID-19 on enrollment and number of providers by county characteristics.

Panel A. Enrollment
	Urban	Rural	Less than 20% Black	More than 20% Black	Less than 7% Hispanic	More than 7% Hispanic	1st SES Quartile	2nd SES Quartile	3rd SES Quartile	4th SES Quartile
A) December*2020 coefficient	−336.516⁎⁎	−1824.000⁎⁎	−633.727⁎⁎	−1163.244⁎⁎	−386.672⁎⁎	−1542.238⁎⁎	−427.040⁎⁎	−641.200⁎⁎	−736.280⁎⁎	−1683.520*
	(46.018)	(429.997)	(89.572)	(363.339)	(65.011)	(377.593)	(79.427)	(151.208)	(187.392)	(625.330)
B) February 2020 mean	859.81	4445.06	1548.31	2886.51	983.14	3762.57	1078.92	1678.72	2046.36	3798.00
C) Effect in% (A/B)	−39.1%	−41.0%	−40.9%	−40.3%	−39.3%	−41.0%	−39.6%	−38.2%	−36.0%	−44.3%
D) Diff. P-value	0.863		0.964		0.890		0.000
Observations	640	360	550	450	580	420	250	250	250	250
Panel B. Number of Providers
	Urban	Rural	Less than 20% Black	More than 20% Black	Less than 7% Hispanic	More than 7% Hispanic	1st SES Quartile	2nd SES Quartile	3rd SES Quartile	4th SES Quartile
A) December*2020 coefficient	−0.266	−2.722*	−0.527	−1.911+	0.172	−2.976⁎⁎	−0.680	1.520*	−1.960*	−3.480*
	(0.286)	(1.307)	(0.511)	(0.950)	(0.389)	(1.041)	(0.581)	(0.549)	(0.729)	(1.636)
B) February 2020 mean	24.98	101.19	36.09	72.38	26.67	87.98	31.24	45.16	50.84	82.44
C) Effect in% (A/B)	−1.1%	−2.7%	−1.5%	−2.6%	0.7%	−3.4%	−2.2%	3.4%	−3.9%	−4.2%
D) Diff. P-value	0.344		0.540		0.033		0.000
Observations	640	360	550	450	580	420	250	250	250	250

Note. Standard errors in parentheses. Regressions were estimated using Eq. (1). Race and ethnicity variables were categorized as below and above the state mean (20% for Black and 7% for Hispanic population). The socioeconomic status (SES) index is an average of the standardized values of the following 6 measures: percentage of the population with a Bachelor's degree or higher, poverty rate, unemployment rate, percentage of families with a female householder, percentage of households receiving SNAP benefits (food stamps), and (logged) median household income (Fahle et al., 2021).

⁺P < 0.10.

^⁎P < 0.05.

^⁎⁎P < 0.01.

Panel B of Table 4 displays the variation in the effects of the pandemic on the number of providers by the same set of county characteristics. We found that although rural counties and counties with a higher percentage of Black residents seemed to have experienced slightly more provider closures (−2.7% and −2.6%, respectively) than urban counties and counties with a lower percentage of Black residents, the differences across these communities were not significant. However, counties with higher percentage of Hispanic residents had significantly more provider closures than counties with less Hispanic residents (−3.4% vs 0.7%, P < 0.05). In terms of the SES index, counties in the higher SES index quartiles (i.e., the third- and fourth- quartiles) experienced significantly more provider closures than counties in the lower SES index quartiles (−3.9% and −4.2%, respectively).8

4. Discussion

Given the constraints in collecting reliable real-time data, the total effects of the COVID-19 pandemic on child development and educational inequality may never be fully captured. However, there are ways that policymakers can make use of existing resources and data to obtain credible estimates of the impacts of the pandemic on the child care sector, which plays an essential role in supporting children's development and parental employment. Our study makes important progress in examining how much the child care market suffered from the pandemic, and which providers and communities were disproportionately impacted by the pandemic. We demonstrate the use of detailed, county-level administrative data in a large and diverse state to more accurately document how this emergency “shock” to NC's child care system affected child care enrollment, child care closures, and its differences across communities and sectors, accounting for temporal trends across counties. The dramatic shifts caused by the pandemic are relevant for the early care and education system for the uncertain future, and therefore its impacts have meaningful implications for early childhood research more broadly.

We found that the pandemic reduced child care enrollment across the state by 41% and reduced the total number of providers by 2% by December 2020. This decline in enrollment is large and similar with most state estimates, even after accounting for seasonal differences; however, our estimate on provider closure is much smaller than those reported in other studies (Weiland et al., 2021). The smaller estimate on provider closure is expected because our DID design removes the effects of seasonal trends in provider entry and exit in the child care market from the effects of the pandemic. The operational grants offered by the state of NC may have also helped providers to stay open. Although we cannot completely parse out the pandemic's impacts on child care demand and supply, the slight decrease in the number of providers, in stark contrast with the large drop in enrollment, appears to suggest that the supply of child care, as shown by the number of open providers, recovered rather quickly a few months into the pandemic, while the demand for child care, as shown by enrollment, might be slower to respond with lingering fears of COVID and increased telecommuting. This is consistent with the pattern observed by Ali et al. (2021), who found that the demand for child care labor among providers (an indicator for increase in supply) began to recover quickly after the lift of the SAHOs.

We further found that enrollment reductions and closures varied by provider type and by community characteristics. In particular, family child care providers fared much better than center providers through the pandemic; they experienced only a 19% decline in enrollment, and the total number of family care providers increased by 4%. This is in contrast to child care centers, where we found a 41% decrease in enrollment and a 4% decrease in the number of providers. This aligns with recent descriptive studies showing that child care homes were more likely to remain open during COVID-19 than child care centers (Porter et al., 2020; Weiland et al., 2021) and during economic crises as well (Brown & Herbst, 2021).

These findings are likely attributable to the well-known, but historically underappreciated, strengths of family care providers – small groups with flexible schedules, a wider range of care hours, substantial child care experience, and nimbleness to adapt routines and practices (Gerstenblatt, Faulkner, Lee, Doan & Travis, 2014; Gormley, 2007; Morrisey & Banghart, 2007; National Survey of Early Care & Education Project Team, 2015; National Survey of Early Care & Education Project Team, 2016a). Such features became even more important during the pandemic, when families could depend on small “pods” to reduce the risk of contagion, and work alternate hours as the SAHOs, local risks, and other school closures changed their everyday lives. Additionally, because family child care is usually more affordable than centers, parents may have switched to this type of care when facing more financial struggles such as pay cuts and job loss (Brown & Herbst, 2021; Morrisey & Banghart, 2007). The family child care sector also provides an opportunity for individuals who lost their jobs to earn income (by directly providing child care at home) when outside alternative employment opportunities are limited (Brown & Herbst, 2021; Katz & Krueger, 2017).

We also found that declines in enrollment were most substantial for preschool-aged children, showing a 49% decrease from levels prior to the pandemic. This is in contrast to a 35% decrease in the enrollment of infants and toddlers, and a 17% decrease for school-aged children. It is noteworthy that while the enrollment of preschool-aged children dropped substantially for child care centers (50%), family child care homes only experienced an 7% decrease in this age group. This further implies that parents may be more willing to enroll their preschoolers in child care homes for smaller group sizes and cheaper tuition.

With respect to provider quality, our findings reveal that there was a statistically significant 4% decrease in 5-star providers and a proportional increase in providers with lower quality ratings. The decrease in the 5-star providers is primarily driven by center providers, whereas the overall quality of child care homes improved. This is consistent with our prediction that higher quality providers, particularly center-based ones, are more likely to close due to the increased cost of maintaining high-quality staff and facilities, and decreased demand for quality when families are under more substantial economic constraints.

The COVID-19 pandemic appears to have exacerbated inequities in terms of child care closures among communities with a higher percentage of Hispanic residents, aligning with other recent studies documenting that non-White families were more likely to experience child care closures than White families (Cascio, 2021; Lee & Parolin, 2021). If such closures become permanent, this is worrisome because it would compound the extant disparities in access to quality care in Hispanic and Latino communities (Bassok & Galdo, 2016; Hatfield et al., 2015), which are more often spatially isolated and have low levels of educational attainment and poor public health (Lichter & Johnson, 2021). This accumulation of inequality in Latino communities and lower availability of formal, licensed early child care opportunities may lead families to turn to informal child care arrangements (Ansari, 2017; Meyers & Jordan, 2006; Sandstrom, Giesen & Chaudry, 2012), which are often of lower developmental quality than formal care (Fuller, Kagan, Loeb & Chang, 2004). Communities with a higher percentage of Black residents experienced a slightly higher level of center closure than those with a lower percentage, but the difference was not statistically significant. However, our data only captured the impacts of the pandemic in the first year of the pandemic. Some evidence suggests that the gap in access to child care among Black children began to exacerbate at the start of 2021 instead of the first months of the pandemic (Cascio, 2021). Ongoing research monitoring such possibilities will be critically important.

The magnitudes of the pandemic's effects on enrollment and closures were generally similar across SES quartiles, with higher-SES communities impacted slightly more. This is consistent with reports that found low-income children experienced fewer center closures in the early months of the pandemic (Cascio, 2021; Weiland et al., 2021). One explanation is that Head Start centers and other public programs targeting low-income children were more likely to stay open during the pandemic (Weiland et al., 2021). Another explanation is that families with higher SES may be more likely to keep their children at home because of their financial resources (i.e., hire a private caregiver), or the greater job flexibility that comes with jobs that require more education (i.e., the ability to telework) (Chen, Byrne & Vélez, 2021). This may imply that demand was somewhat more elastic in higher-SES communities in response to the pandemic, where high contagion would make such families more likely to not enroll in out-of-home care. Providers in rural and urban communities were also similarly affected by the pandemic, although rural communities experienced slightly larger enrollment drops and closures. Overall, we see some important differences in enrollment and closures by community characteristics, but not as dramatic in magnitude as some have feared (e.g., Malik, Hamm, Lee, Davis & Sojourner, 2020).

4.1. Implications for child development and policy

The reductions in enrollment and provider availability we find are large enough that COVID-19 could generate long-lasting and serious consequences for children if their learning and developmental needs are not properly addressed. With fewer child care options, parents may have challenges transitioning to home production and remote learning in the short term. Early care at home may lead to losses of learning opportunities and disruptions to child development if they are substituted by increased screen time, as some studies suggest (Nores, 2020). This would further exacerbate the extant inequalities in parents financial and time investments in children's development since higher income families spent more time on learning with their children during the pandemic than parents in low-income families (Barnett, Jung & Nores, 2020; Duncan & Murnane, 2011). Indeed, all the factors reviewed here and reporting throughout the pandemic show that parents of low-income families and children of color were more likely to be essential workers, more likely to contract COVID-19, and less likely to telecommute, leaving few hours of the day to devote to children's home learning (Office of Human Services Policy, 2021). In turn, the accumulating evidence suggests that the care children have received at home may be worse than it was before the pandemic; young children require substantial care and supervision, and this needs to be responsive and emotionally warm for positive development, yet parents facing pandemic-related stress and myriad mental health challenges struggle with such provision (McCoy et al., 2021; Patrick et al., 2020; Yoshikawa et al., 2020). Monitoring the availability and quality of child care, as well as improving the affordability and accessibility of these care across communities will be critical for states and localities to offset such potential consequences to development, educational equity, and family well-being.

If the center-based closures we find here were to become permanent, or if COVID-19 contagion concerns were to persist, this may lead to changes in parents’ preferences and child care selection in the longer term. The increased number of family home providers—and the characteristics of such care with elevated importance during a pandemic for family health and well-being (e.g., smaller groups, greater flexibility)—may lead parents to choose more family home care. In turn, family child care providers will need increased professional supports from states and local systems, such as QRIS, to promote children's development. Family home providers that are financially supported and receive professional development targeting their attitudes (i.e., professional motivation, beliefs about childrearing) and practices are more likely to have positive associations with children's early academic and socioemotional outcomes (Forry et al., 2013). Additionally, because family home providers are not often subjected to nutrition and physical activity requirements or have limited training in these areas, children enrolled at these providers can have poor diet quality, low physical activity, and increased risk of obesity (Black, Matvienko-Sikar & Kearney, 2017; Francis, Shodeinde, Black & Allen, 2018; Trost, Messner, Fitzgerald & Roths, 2009). Comprehensive professional supports for family providers will be central to ensuring children's healthy development in these settings.

Our study also demonstrates how state policymakers can utilize existing administrative processes to collect data from providers and child care subsidy recipients on their enrollment and attendance to monitor changes in child care use, both in later stages of the pandemic and when COVID-19 becomes endemic. Such data are regularly collected and could be used to obtain more accurate estimates with designs such as DID when answering questions empirically and in real time to prevent inequities (i.e., regular monitoring of communities with lower child care supply), or to address problems as soon as they arise (i.e., targeted supports for communities with greater care provider loss or challenges). Each of the data sources we used were publicly available, also enabling local researchers and professional organizations to keep track of early learning opportunities in key localities and for specific subgroups of families or providers. The continually changing circumstances of the pandemic will require continued work of this type both across the country and around the world (McCoy et al., 2021).

4.2. Limitations

We note several limitations to our study. First, the differences in our 2 data sources may create measurement error in our outcome variables, as well as bias caused by non-respondents in the provider-reported survey data. Specifically, the non-respondents in the December 2020 survey data may include providers who closed and those who were open but did not report. It is thus unclear which direction our estimates may be biased. Second, although our DID design provides a more accurate estimate of COVID-19′s impacts on child care providers, we are unable to control for time-varying county characteristics due to data unavailability. Our results should therefore be interpreted with caution. Because we only use county-level data from NC, our results may not generalize widely to other states, or represent the local conditions within NC and across states. Additionally, more fine-grained geographic measures can help to examine local level impacts for communities compared with the relatively coarse county level characteristics we use in our study.

CRediT authorship contribution statement

Qing Zhang: Conceptualization, Methodology, Resources, Data Curation, Writing - Original Draft, Writing - Review & Editing. Maria Sauval: Methodology, Formal Analysis, Data Curation, Writing - Original Draft, Visualization. Jade Marcus Jenkins: Methodology, Writing - Original Draft, Writing - Review & Editing.

Appendix A. Description of Data Sources

The North Carolina Division of Child Development and Early Education Statistical Reports are updated through monitoring visits that stopped at the beginning of March 2020. Although filling out the COVID-19 Child Care Provider Survey was not mandatory, it was strongly recommended, and most providers continued reporting their data daily through this tool. Because we have February 2020 data from both sources, we are able to assess the response rate from providers and the overlap between the 2 sources in our outcome variables. We first compare the average enrollment and number of providers reported in the 2 sources. We find small discrepancies between these 2 sources – the provider survey has only 11% lower enrollment and 9% fewer providers than the Statistical Report data. Fig. A1 illustrates the total enrollment and provider data from our 2 sources of data.

We also check the county-by-county differences in enrollment and number of providers in the 2 sources. The within-county comparison shows similar levels of enrollment and number of providers between the 2 sources. On average, counties in the provider survey have 8% lower enrollment and 7% fewer providers than reported in the Statistical Report data. Because the Statistical Report data include all licensed providers, our check suggests that an average of 93% of providers responded to the provider survey in February, a much higher response rate from providers than reported in most prior studies. Nevertheless, we recognize the potential bias resulting from the non-respondents and discuss this limitation in the main text (Appendix B, Fig. A1 ).

Fig. A1.

Average Enrollment and Number of Providers using the DCDEE Statistical Reports and the COVID-19 Child Care Provider Survey. North Carolina, 2018–2020.

Appendix B. Additional Tables and Figures

(Table B1, Table B2, Table B3, Table B4).

Table B1.

Characteristics of North Carolina licensed child care providers (February 2020).

	All	Centers	Homes
Number of providers
Total	5779	4417	1362
%	100%	76.43%	23.57%
Characteristics (average per facility)
Enrollment	42.0	53.2	5.5
	(43.9)	(44.5)	(2.4)
Maximum capacity	63.8	81.1	7.6
	(66.1)	(62.7)	(1.4)
Number of employees	7.0	8.8	1.1
	(7.9)	(8.2)	(0.5)
Age group served (average per facility)
0–2 yr-olds	11.7	14.6	2.4
	(17.6)	(19.3)	(1.5)
3–5 yr-olds	20.7	26.5	1.7
	(26.8)	(28.1)	(1.4)
School-aged children	9.6	12.1	1.4
	(19.3)	(21.5)	(1.7)
Category of operation
Private	64.2%	53.2%	99.9%
Public School	20.4%	26.7%	0%
Religious sponsored	9.1%	11.9%	0%
Head Start	4.9%	6.4%	0.1%
Other	1.4%	1.8%	0%
QRIS
1 star	4.0%	2.2%	10.0%
2 stars	2.1%	0.4%	7.9%
3 stars	19.4%	16.3%	29.2%
4 stars	25.7%	21.8%	38.5%
5 stars	39.5%	47.9%	12.3%
Missing QRIS data	9.3%	11.5%	2.1%

Note. Standard deviations in parenthesis.

Table B2.

County-level descriptive statistics, February and December 2016–2019.

	2016			2017			2018			2019
	Feb	Dec	% change	Feb	Dec	% change	Feb	Dec	% change	Feb	Dec	% change
Total Enrollment	2435.1	2430.4	−0.2%	2451.7	2421.2	−1.2%	2431.2	2395.7	−1.5%	2398.8	2390.7	−0.3%
	(4277.77)	(4242.76)		(4273.8)	(4252.77)		(4270)	(4241.72)		(4285.77)	(4354.05)
Number of providers	68.2	66.2	−2.9%	65.3	62.6	−4.0%	62.2	60.1	−3.4%	59.8	57.9	−3.2%
	(104.42)	(101.21)		(99.67)	(96.42)		(96.03)	(92.8)		(92.91)	(90.97)
Enrollment in child care centers	2327.4	2328.3	0.0%	2351.9	2328.3	−1.0%	2340.2	2311.5	−1.2%	2315.7	2315	0.0%
	(4114.55)	(4087.81)		(4119.6)	(4110.72)		(4131.19)	(4108.43)		(4154.67)	(4237.35)
Enrollment in family home care	107.6	102.2	−5.1%	99.9	92.8	−7.0%	91	84.2	−7.5%	83.1	75.6	−9.0%
	(175.32)	(167.99)		(166.79)	(153.94)		(151)	(144.89)		(143.32)	(129.52)
Number of center providers	47	46.5	−1.1%	46.2	45.3	−1.7%	45.2	44.4	−1.6%	44.5	44.1	−0.7%
	(69.98)	(68.52)		(67.79)	(67.16)		(67)	(65.82)		(66.4)	(67.03)
Number of family providers	21.3	19.7	−7.5%	19.1	17.3	−9.4%	17.1	15.6	−8.2%	15.3	13.8	−10.4%
	(36.45)	(34.6)		(33.69)	(31.03)		(30.79)	(28.63)		(28.15)	(25.58)
Enrollment 0–2 y.o.	653.5	667.1	2.1%	671.8	661.7	−1.5%	660.4	669.6	1.4%	673	669.7	−0.5%
	(1280.16)	(1318.47)		(1319.27)	(1308.95)		(1306.33)	(1362.37)		(1386.53)	(1401.34)
Enrollment 3–5y.o.	1158.8	1148.4	−0.9%	1165.8	1160.7	−0.4%	1175.4	1153.5	−1.9%	1161.2	1164.9	0.3%
	(1851.97)	(1821.56)		(1853.38)	(1865.37)		(1892.4)	(1862.6)		(1915.75)	(1972.08)
Enrollment school-age	622.8	614.9	−1.3%	614.2	598.9	−2.5%	595.3	572.6	−3.8%	564.5	556	−1.5%
	(1245.11)	(1237.46)		(1233.47)	(1214.35)		(1206.63)	(1142.85)		(1119.15)	(1117.84)
Number of 1-star providers	4	3.4	−12.7%	3.4	3	−11.9%	2.9	2.2	−21.0%	2.3	2.3	0.0%
	(9.22)	(7.79)		(7.58)	(6.93)		(6.87)	(5.74)		(5.73)	(5.95)
Number of 2-stars providers	2.4	2.1	−12.6%	1.9	1.7	−10.8%	1.7	1.6	−5.9%	1.6	1.3	−19.2%
	(4.23)	(3.91)		(3.61)	(3.34)		(3.38)	(3.04)		(2.92)	(2.6)
Number of 3-stars providers	14.8	14.3	−3.4%	14.3	13.4	−6.3%	13.1	12.1	−7.7%	12	11.3	−5.9%
	(22.83)	(21.84)		(21.87)	(20.9)		(20.26)	(19.18)		(18.95)	(18.58)
Number of 4-stars providers	18	17.5	−2.8%	17.4	16.6	−4.0%	16.5	15.8	−4.2%	15.8	14.9	−5.7%
	(31.11)	(29.97)		(30.18)	(29.68)		(29.2)	(27.68)		(27.78)	(26.84)
Number of 5-stars providers	23.8	23.7	−0.4%	23.6	22.9	−3.0%	23	22.9	−0.4%	22.9	22.7	−0.9%
	(36.43)	(35.88)		(35.24)	(33.3)		(33.24)	(34.03)		(34.61)	(34.01)

Note. Standard deviations in parenthesis.

Table B3.

Heterogeneity: effects of the Covid-19 pandemic by urbanicity, age, and provider quality.

Panel A. Enrollment by Age and Urbanicity
	0–2 y.o.		3–4–5 y.o.		School-aged
	Rural	Urban	Rural	Urban	Rural	Urban
A) December*2020 coefficient	−73.594⁎⁎	−528.889⁎⁎	−251.984⁎⁎	−1180.944⁎⁎	−10.937	−114.167⁎⁎
	(11.588)	(158.326)	(33.902)	(251.559)	(9.010)	(31.511)
B) February 2020 mean	235.55	1465.44	503.11	2412.58	121.16	567.03
C) Effect in% (A/B)	−31.2%	−36.1%	−50.1%	−49.0%	−9.0%	−20.1%
D) Diff. P-value	0.681		0.927		0.232
Observations	640	360	640	360	640	360
Panel B. Number of Providers by Provider Quality and Urbanicity
	1- or 2-star providers		3- or 4-star providers		5-star providers
	Rural	Urban	Rural	Urban	Rural	Urban
A) December*2020 coefficient	0.109	−0.083	0.766*	1.583*	−0.500+	−1.611⁎⁎
	(0.087)	(0.493)	(0.309)	(0.646)	(0.284)	(0.542)
B) February 2020 mean	1.41	7.25	11.02	45.25	10.09	38.42
C) Effect in% (A/B)	7.8%	−1.2%	7.0%	3.5%	−5.0%	−4.2%
D) Diff. P-value	0.331		0.274		0.809
Observations	640	360	640	360	640	360

Note. Standard errors in parentheses. Regressions were estimated using Eq. (1).

⁺P < 0.10.

^⁎P < 0.05.

^⁎⁎P < 0.01.

Table B4.

Heterogeneity: effects of the COVID-19 pandemic on the number of providers by county sees and provider quality.

	1- or 2-star providers				3- or 4-star providers				5-star providers
	1st SES Quartile	2nd SES Quartile	3rd SES Quartile	4th SES Quartile	1st SES Quartile	2nd SES Quartile	3rd SES Quartile	4th SES Quartile	1st SES Quartile	2nd SES Quartile	3rd SES Quartile	4th SES Quartile
A) December*2020 coefficient	0.160	0.400+	−0.160	−0.240	0.880	3.120⁎⁎	−0.480	0.720	−0.520	−0.080	−1.280*	−1.720*
	(0.127)	(0.220)	(0.247)	(0.658)	(0.553)	(0.674)	(0.460)	(0.553)	(0.492)	(0.466)	(0.541)	(0.635)
B) February 2020 mean	0.96	2.68	2.64	7.76	15.36	19.04	23.96	35.00	12.00	17.80	19.92	31.44
C) Effect in% (A/B)	16.7%	14.9%	−6.1%	−3.1%	5.7%	16.4%	−2.0%	2.1%	−4.3%	−0.5%	−6.4%	−5.5%
D) Diff. P-value	0.339				0.000				0.219
Observations	250	250	250	250	250	250	250	250	250	250	250	250

Note. Standard errors in parentheses. Regressions were estimated using Eq. (1).

⁺P < 0.10.

^⁎P < 0.05.

^⁎⁎P < 0.01.