• No results found

State-Level Expenditure and Program Policy Data

3.4 Model

3.5.2 State-Level Expenditure and Program Policy Data

In this analysis, the primary interest is the state-level amount of spending relative to the target population. All expenditure measures throughout this essay are given as real 2014 dollars based on the personal consumption expenditures deflator (excluding food and energy) obtained via the U.S. Bureau of Economic Analysis. In estimation, the expenditures are expressed as spending per capita by state, which is defined as children 21See Duffy and Sastry (2014) for details regarding achievement tests in the PSID-CDS. In addition to these early educational achievement measures, CDS also collected information on a child’s self-concept in math and reading, that is, measures of one’s perceptions of own ability in these subjects. An additional measure of intelligence in the CDS is the WISC digit span raw score, which represents the total for the longest forward-sequence and backward-sequence span of digits the child can recall after hearing (Wechsler Intelligence Scale for Children–Revised: The WISC Digit Span Test for Short-Term Memory,

c

Psychological Corporation). Other achievement measures for high school students include grade point average, SAT and ACT scores.

living below the U.S. Census poverty threshold for means-tested expenditure categories, or all children under age 18 for broader spending categories such as education.

3.5.2.1 AFDC/TANF and CCDF Childcare Assistance Data

During the AFDC years 1991-1996, state-level expenditure data, including pre-PRWORA childcare spending, are available via the Background Material and Data on Programs within the Jurisdiction of the Committee on Ways and Means (Green Book). For the TANF years, detailed expenditure data are available for every fiscal year via the U.S. Department of Health and Human Services (USDHHS) website. The Green Books also provide childcare caseload data in certain years, though caseloads during the TANF era only correspond to cash assistance cases. The only indications of the number of TANF childcare cases come from a measure of the percent of active TANF cases with subsidized child care in the Characteristics and Financial Circumstances of TANF Recipients, 2000- present, and state TANF reports on Separate State Programs and Maintenance of Effort (MOE) activity, which are available for years 2005 and 2010.22 For earlier years of AFDC childcare assistance, state-level expenditure or caseload data are much less available, though the total level of childcare spending is also much lower before 1991. (See Appendix F for more details and additional data sources.)

From 1999 to present, CCDF program data are available for both expenditure and caseloads by state via the USDHHS website, and earlier data (including predecessor programs) are available for years 1991-1998 via Green Book publications. Given that some states allocate a generous proportion of TANF expenditure toward CCDF block grant 22While state reports for MOE activity are presumably collected in other years, only two have been available recently via government websites or select Internet archive sites. In some states, the total MOE spending for child care matches the total state expenditure in official budget reports, which means that the childcare caseloads in those states should be representative.

transfers, the TANF program can be a significant contributor to the variance of CCDF programming. For example, in Wisconsin since 1997, the proportion of TANF transfers to CCDF relative to total CCDF expenditure implies that TANF contributes toward about 8,700 children receiving child care on average per month out of a total of about 26,000. Another data source for childcare assistance participation is a series of USDHHS briefs providing estimates of childcare eligibility and receipt. In 2012, for instance, around 14 million children were eligible for subsidized care according to simulations under federal rules, almost 9 million of those were eligible under more strict state rules, and the number of children served was a little over 2 million (15 percent of those federally eligible).23

While there is little information about specific childcare policies within the TANF pro- gram, the Urban Institute maintains the CCDF Policy Rules Database in coordination with USDHHS. These policies include information on the amount of copays and eligibility for copay exemption (for example, some states exempt TANF participants), the presence of family-size adjustments for copays, whether relatives are authorized to provide subsi- dized in-home care, or the length of eligibility redetermination periods. Another Urban Institute project is the Welfare Rules Database, which does not offer childcare-specific policy rules, yet does have state-level TANF policy information such as the size of child- care earnings disregards, whether states impose rules on school attendance or parental 23Take up for childcare subsidies has been historically low. Evidence from certain pilot AFDC ex- periments provided early evidence that childcare subsidies can incentivize low-income families to use center-based child care, yet the costs born by families after subsidization was still a significant burden, as was the prospect of transitioning out of public assistance (Committee on Ways and Means, U.S. House of Representatives, 2000; Michalopoulos, 2010). Also, an increasing complication of using child- care assistance when eligible is that low-wage workers frequently experience variable scheduling subject to changes and difficult hours, such as contract or temp-to-hire work (Henly and Lambert, 2005; Katz and Krueger, 2016). Even though TANF conditions assistance on mothers working, the percent of active TANF cases receiving subsidized child care has stayed close to 10 percent over the last decade.

involvement, and whether transitional child care after TANF participation is limited by income or time.

3.5.2.2 Other State-Level Controls

In order to study the effect of childcare expenditure and policy on family behavior, it is important to control for other variation within states over time that may also be correlated with outcomes related to child care and education. In terms of additional program ex- penditures, the most relevant for low-income families are the Social Services Block Grant (SSBG), Head Start, elementary/secondary public education, and total public welfare.

SSBG funds are primarily directed toward child-wellbeing services such as foster care or protection from abuse, yet the program allows states to fund a wide variety of social services, which also includes providing for childcare assistance. Before the welfare reform era, SSBG (or Title XX), was the leading funder of government subsidized child care, especially for AFDC recipients. Today, SSBG childcare makes up almost 20 percent of total SSBG spending, which amounts to less than 3 percent of the total spending from TANF and CCDF combined. Head Start, meanwhile, is a large program that increased dramatically during the 1990s (along with total childcare spending) and is comparable in magnitude to total CCDF spending. In the analysis, I ignore SSBG spending given the small size of the program and its shrinking role after welfare reform, and I assume that Head Start spending varies consistently across states and thus control only for the aggregate increases in Head Start allocations over time.

In addition to controlling for state-level expenditures as child development inputs, controls for potentially correlated outcomes, such as high school graduation rates, can further identify the relationship between childcare-specific spending and later education

outcomes. In more recent years, the National Center for Education Statistics (NCES) has begun measuring high school graduation using a 4-year adjusted cohort graduation rate (ACGR), which is reported in the Common Core of Data. The income gap for high school completion has narrowed considerably in the last two decades, as shown in Figure 3.6. However, for economically disadvantaged students as of 2012, the ACGR can range from 58 percent in Nevada to 85 percent in Indiana.