Panel Paper: New Evidence-Based Tools to Improve Performance of Public Assistance Programs: Developments from Collaborative Partnerships Between the Census Bureau and State Agencies

Thursday, November 3, 2016 : 10:20 AM
Dupont (Washington Hilton)

*Names in bold indicate Presenter

Benjamin Cerf-Harris, Mark A. Leach and Rachel M. Shattuck, U.S. Census Bureau


As a part of broad efforts to research and test operations for the 2020 decennial census, the Census Bureau is collaborating with state agencies and the United States Department of Agriculture’s Food and Nutrition Service to acquire individual-level administrative records data from state public assistance programs. Administrative records (AR) – particularly those from public assistance programs – offer the potential to improve decennial operations and the coverage of households and persons that historically are difficult to enumerate, such as children and those in poverty. As a benefit to state agencies for providing AR data, the Census Bureau’s Center for Administrative Records Research and Applications (CARRA) is developing a new set of tools that enable a state to evaluate the performance of their programs in new ways.

State partners supply ARs from programs such as SNAP, WIC, or TANF to the Census Bureau. We use probability record linkage techniques to assign a unique person identifier called the Protected Identification Key (PIK) to each program participant using personal information such as social security number and address. Once PIKs are assigned, the AR data is stripped of all personal identification information and the data are used for analysis. We then link the ARs to individual-level responses to the American Community Survey (ACS).

The linked data enable us to model which survey respondents are eligible for program participation and also to observe which eligible survey respondents actually participate in the program. We then estimate eligibility rates in the population and participation rates within the eligible population. We use detailed demographic and economic information in the ACS to profile and compare the eligible, participating, and eligible-but-not-participating populations. And we also calculate monthly rates of entry, exit, and churn. If the sample size of linked data allow, profiles of each population may be compared at state and county levels. These unique data support analyses that may assist state agencies in their efforts to improve program performance.

In addition to tabulations, we provide data visualization tools to convey findings in more intuitive and user-friendly ways. A data visualization tool may include interactive charts or maps that illustrate eligibility and participation rates and demographic profiles by county or over time.

These innovative tools have the potential to help states take evidence-based actions to target outreach and improve customer access and outcomes. Our presentation will provide an overview of data sharing, record linkage, and eligibility modeling. We will begin with a discussion of some of the challenges and successes CARRA has encountered in developing collaborative relationships with states. We will then show examples of tabulations and give a live demonstration of the data visualization tools that we recently have delivered to state agencies.