Mark Prell is a senior economist at the Economic Research Service (ERS), US Department of Agriculture (USDA), where he coordinates interagency activities that support the Census-FNS-ERS Joint Project.

 

Erik Scherpf is an economist at the Economic Research Service (ERS), US Department of Agriculture (USDA), where he leads and conducts statistical research using the project’s administrative and linked data.

 
 
Two statistical agencies—the US Census Bureau and US Department of Agriculture’s (USDA’s) Economic Research Service (ERS)—have formed a strategic partnership with USDA’s Food and Nutrition Service (FNS), which administers USDA’s food and nutrition assistance programs. The Census-FNS-ERS Joint Project is a long-term effort to acquire state-level administrative data for USDA nutrition assistance programs and to link those data to Census Bureau surveys and other administrative files. The project leverages the strengths of multiple data sources, obtaining statistical results that could not be achieved using any one source by itself.

Results of the project include specialized statistical tables and data visualizations that help states administer USDA nutrition assistance programs, as well as findings on policy-relevant issues concerning program access, targeting of benefits, and participation patterns.

The Census-FNS-ERS joint project acquires data from two of USDA’s largest food assistance programs—the Supplemental Nutrition Assistance Program (SNAP, formerly known as Food Stamps) and the Special Supplemental Nutrition Assistance Program for Women, Infants, and Children (WIC). Last year, SNAP served about 42 million low-income people per month at an annual cost of $68 billion to taxpayers. WIC serves low-income women, infants, and children at nutritional risk. In 2016, WIC served an estimated 48% of all infants born in the US. The program’s annual cost is about $5.6 billion.

The joint project is a national undertaking. State SNAP and WIC agencies collect and store the administrative data for the programs. State agencies can participate by sending their confidential microdata (involving millions of records that must be maintained in a secure data environment) to the Census Bureau in exchange for state-specific tables and data visualizations. The project began in 2012 with three states and has grown to 23 SNAP agencies and 11 WIC agencies.

Any data source has strengths and limitations. The states’ SNAP and WIC data are known for their accuracy. However, these administrative records exclude people who do not participate in the program. The Census Bureau’s survey data have a complementary strength because they include both program participants and nonparticipants. Survey data also include a rich set of socioeconomic variables, such as annual income—critical for certain analyses. Survey-based data on program outcomes, such as health and food security, further expand research possibilities. Matching these survey and administrative data enhances the ability to conduct research previously impossible.

The project is working to scale up by acquiring and linking data from more states, agencies, and programs. The Census Bureau acquires state administrative data on Temporary Assistance to Needy Families (TANF). The bureau has also developed data-sharing agreements with the US Department of Veterans Affairs (VA), US Department of Housing and Urban Development (HUD), and Medicaid and Medicare programs in the US Department of Health and Human Services (HHS). Data on veterans, housing, and health and human services programs can be linked to state administrative records for USDA nutrition assistance programs, to surveys, and to the Decennial Census. In addition to acquiring more data, the joint project is making data more available to academic researchers through the Federal Statistical Research Data Centers (FSRDCs).

States that participate in the project receive specialized tabulations designed to address a key policy-relevant issue: How effectively does SNAP (or WIC) reach the people the program is designed to serve? Using the linked data and a methodology developed in an ERS report, the bureau estimates the SNAP access rate—the percentage of people who receive program benefits (according to administrative data) among those estimated to be eligible for SNAP (based on survey data); an analogous WIC access rate is provided to state WIC agencies. SNAP and WIC access rates are estimated at the state level, at highly granular geographic levels within a state, and for demographic subgroups.

Prior to this project, it was not possible for states to obtain these results. A related product, which is newly developed, is a SNAP Profile—a web-based, interactive data visualization tool that provides access rates and other SNAP information; a SNAP Profile for New York State provides a first case.

Using the linked data, ERS researchers address policy issues such as program access, targeting of benefits, and participation patterns. For example, using linked data and careful modeling of membership in SNAP households, an ERS report found that SNAP benefits are targeted more toward those in deep poverty (with annual income less than 50% of the poverty line) than is detected using survey data alone.

The Census-FNS-ERS project exemplifies how a partnership between statistical and program agencies, can succeed by serving the mission of each partner. The project brings together data, resources, and expertise to create unique statistical products that are valued by policymakers and other stakeholders in the programs.