### Childcare

#### Demand estimated by

#### Supply scenarios

### Overlay DSHS COVID-19 Metrics

#### On/Off

#### Metric

## Supply of child care seats during the COVID-19 pandemic

### Data

### Methdology

TWC launched a new application in April, 2020, to help collect data on the supply of child care in the state of Texas. Data on child care seat availability and average daily attendance (ADA) are collected at the provider level. During the month of August, 2020, the time period in which attendance data was most recently collected, 33% (3,614 of 10,928) of open child care providers reported ADA. To estimate the supply of child care in August, we extrapolate ADA for the two-thirds of child care providers which did not report ADA.

First, we merge ADA attendance data with publicly available data from Child Care Licensing to include additional variables such as the type of care, subsidy status, age groups served, and licensed capacity. Providers are then categorized according to the following dimensions, which are predictive of ADA:

- Type of care (center vs homes)
- Subsidy acceptance status
- Age groups served
- Licensed capacity
- County’s total population

Next, we match two providers that have exactly the same dimensions for the first four variables and are in counties with similar populations. For each group, we compute the average and standard deviation of ADA. In this step, we only used the providers – within each group – that reported ADA. We use the mean ADA for the group as our estimate for the ADA for the providers in the same group that are open but do not report ADA. This extrapolation results in complete data for all providers in all counties.

Finally, we created three supply scenarios specific to each county based on current enrollment for each county:

- Low scenario: Current enrollment as the ratio of licensed capacity (ADA/licensed capacity).
- Medium scenario: the mid point between low and high scenarios.
- High scenario: ADA + 1 standard deviation divided by licensed capacity.

At this time, we are unable to integrate the data on additional available slots to estimate supply. Currently, two-thirds of providers are not reporting ADA nor availability data. Extrapolating availability data would require two extrapolations and would greatly increase the uncertainty of our estimates. Additionally, a significant portion of providers are reporting available slots that would put them well above licensed capacity and we need to further understand why this would be the case. Nevertheless, we were able to utilize data on available slots to validate our "high" scenario.

## Demand for child care among current workforce estimates

### Data

- ACS Households and Families 1-year estimates (2018) table
- ACS Housing and Demographics 5-year estimates (2018) table
- ACS Housing and Demographis 1-year estimates (2018) table
- Data USA Industry and Employment data

### Methodology

Demand for child care estimates are created utilizing household population data from the American Community Survey (ACS), workforce industry and occupation employment data from Data USA, and SafeGraph cell phone data. Demand for child care is estimated using both industry and occupation estimates. Please note that workforce estimates by industry are higher than estimates by occupation, so we consider estimates by industry the upper bound of potential demand estimates. See the appendix for industries and occupations included.

First, we take a county’s population and total number of households to estimate the average number of people per household. We assume that households from our industries of interest follow these typical figures and estimate the total number of households with workers in our industries of interest.

Next, we estimate the percent of the populatin under 13, by assuming a uniform distribution of child ages younger than 18 and applying that to the proportion of the population youth than 18-years in each county. We use this to estimate the number of children under age 12 per 100 households. We adjust the estimate of the number of children under age 12 per 100 households according to whether or not households contain other non-working adults or older sibling that can provide care for some children in the household. To do this we use Current Population Survey (Census and Bureau of Labor Statistics) data on the proportion of workers who do not have other adults/ older children in the home or who are single parents.

Finally, we adjust the estimates to reflect the reduced operating capacity of certain industries and occupations as of August, 2020. Estimates are adjusted by using foot traffic data to indicate overall foot activity. In June, overall foor activit was already around 72%. In some locations, it was even higher (i.e., Harris County, 77%). This higher level of foot activity is possible, for example, if there is little monitoring of activity. Thus, we considered scenarios in which capacity were at 65% and 85%.

To estimate the number of workers in scenarios for client capacity, we use SafeGraph cell phone data that provides information about the number of visits to all businesses and the number of workers based on cell phone activity. Specifically, we combine data for two periods. The first period extends from January 1, 2019 to April 30, 2019. The second period extends from January 1, 2020 to April 30, 2020. We estimate the number of clients according to the number of "short" visits to the locations. We define short visits are visits that last at most three hours. We estimate the number of workers according to the number of "long" visits to the locations. Long visits are visits that last at least three hours.

We compute these figures for the two periods described above. Then, we take the ratio of (1) for 2020 to 2019. In the 2019 period, there was no pandemic, so we assume that the businesses were operating at full capacity. In the 2020 period, there was no pandemic in the first two months of the year, but the situation starts to change in early March 2020. Therefore, after the pandemic started, we observe weeks in which the number of clients were at 65% and 85% of full capacity relative to the same week in 2019. We use these data to estimate the fraction of workers that corresponds to different levels of operating capacity.

If we base capacity at 65% and 85% capacity, then we project that we will have 76% and 88% work activity (relative to the same week in 2019). The middle point is 82%, which we take as our preferred estimate for the portion of the workforce in non-essential industries that are not fully operational.