A Guide to Understanding State Child Care Subsidy Programs Through Analysis of Public and Non-public Use Datasets
Publication Date: August 2009
This is an excerpt from the full report
SECTION I: INTRODUCTION
Purpose of the Guidebook
This guide has been developed to assist researchers in using public and non-public use U.S. Census Bureau data as well as public and non-public use state administrative data—both separately and in combinations—to analyze state-specific child care subsidy take-up and employment of child care subsidy recipients.
Through analysis of survey data gathered by the Census Bureau and administrative data generated by state child care subsidy and other programs, researchers can help states answer two fundamental questions about their subsidy programs.
- Who does and does not take up the child care subsidy?
- What is the effect of child care subsidy program participation on parental employment?
This guidebook outlines example analyses that states can tailor for their use in addressing these questions.
Evolution and structure of the guidebook
This guidebook has been developed by members of the research team for a study of Employment and TANF Outcomes for Families Receiving Subsidies in Illinois, Maryland, and Texas. The guide was originally planned to share methodology and examples of analyses from this pioneering study, the first to merge Non-Public Use American Community Survey (NP ACS)data from the Census Bureau with non-public use administrative data for child care subsidy, Unemployment Insurance wage records, and Temporary Assistance for Needy Families. The American Community Survey does not ask respondents about their child care utilization patterns or any publicly subsidized child care assistance they may have received in the past year. Therefore, to obtain the additional attributes needed to answer the research questions of who does and does not take up child care subsidies, and what is the effect of child care subsidy program participation on parental employment, merging the non-public use ACS with administrative data sources was required. The guidebook refers to this as the All Child Care Eligibles study.
Non-public use ACS data contain uniquely identifying information such as names and addresses. The Census Bureau’s Data Integration Division has developed a matching system whereby records in the ACS files can be linked with individual records in state administrative data bases. The system uses probabilistic matching to assign a Protected Identity Key (PIK) to survey and administrative records. The algorithm for the probabilistic match uses names, addresses, and demographic characteristics to output a set of matches, probable matches, possible matches, and non-matches. Through the PIK process and other safeguards, the study team protected the confidentiality of non-public use data from both the Census Bureau and the participating states. The steps involved in requesting and receiving Census Bureau permission to use and link to the NP ACS and in accessing the merged data in the Bureau’s Research Data Centers—while ultimately successful—proved too involved and protracted to describe in a guidebook.
The team therefore decided to use the guidebook to share more accessible methodologies and example analyses using the Public Use Microdata Sample of the American Community Survey (PUMS ACS). With all the same variables as the non-public use ACS, except the identifying information, the PUMS ACS can answer state-specific, point-in-time questions about the number and characteristics of families eligible to receive a child care subsidy. See Section III below. Section III also explains use of publicly available Child Care and Development Fund state sample data to answer descriptive, point-in-time questions about the number and characteristics of eligible families actually receiving a child care subsidy.
Section IV offers examples of methodologies and analyses that use non-public state administrative data (but not non-public Census data) to explore point-in-time and longitudinal questions about subsidy use and its effects. This section draws on the experience of various members of the research team in an earlier multi-state study, Child Care Subsidy Use and Employment Outcomes of TANF Mothers during the Early Years of Welfare Reform: A Three- State Study (2004). This prior study relied exclusively on non-public use, linked, individuallevel, administrative data in Illinois, Maryland, and Massachusetts to follow a series of TANF entry cohorts over time. Individuals in the TANF population were defined as eligible for a child care subsidy if they were working, had a child under the age of 13, and had a household income below the child care subsidy state eligibility ceiling. Child care subsidy program data were used to identify those who received the subsidy and those who did not. Unemployment Insurance (UI) wage record data were used to identify those who are eligible for the child care subsidy because they had found employment and had wages below the child care subsidy state eligibility ceiling. UI data were also used to examine individuals’ employment patterns. The guidebook refers to this earlier report as the TANF Child Care Eligibles study.
Accessing Public and Non-Public Use Datasets
A number of datasets are available to explore these general policy questions at the state level. As noted, public use data files, with personal identifiers removed, are sufficient for point-in-time descriptive analyses that explore the basic characteristics of a population.
- The Public Use Microdata Sample of the American Community Survey (PUMS ACS) is available for download from the Census Bureau website ( http://factfinder.census.gov/home/en/acs_pums_2001.html) in comma-delimited, PC SAS and UNIX SAS formats.
- Similarly, the Child Care and Development Fund state sample data (ACF-801) is widely available via the Child Care & Early Education Research Connections website ( http://www.researchconnections.org/discover/index.jsp). Search “ACF” > “Datasets” > “Any Date” > “Go.”
Non-public use data complete with intact personal identifiers such as Social Security numbers, names and addresses, are ordinarily essential when researchers need to merge data files at the individual level. Understandably, confidentiality issues make this information much more difficult to acquire.
- The Non-Public Use American Community Survey (NP-ACS) data are available only through arrangements with the Census Bureau and its Research Data Centers. For more information on how to access private Census data, visit the Bureau’s Center for Economic Studies website ( www.ces.census.gov/index.php/ces/researchprogram).
- Confidential state administrative micro-data are largely available only through agreements with state officials. Formal agreements with state agencies are necessary to ensure data security and controlled data accessibility, understanding variable measurements, as well as the development of a process for sharing of research results prior to public dissemination. The non-summary child care subsidy data, Unemployment Insurance wage record data, and Temporary Assistance for Needy Families (TANF) data cited in this guidebook were all obtained through data-sharing agreements with the respective states.