Simulating College Student Eligibility for SNAP to Close the Take-Up Gap

• In Event: Leveraging Data and Partnerships to Help States Navigate SNAP Policy Changes

Abstract

Across the country, many college students from underserved backgrounds struggle to meet their basic needs. The SNAP program can address food insecurity enabling students to focus on their studies. Connecting more eligible college students to SNAP has been a major focus in California for the past five-ten years. State agencies and higher education leaders have partnered with the California Policy Lab (CPL) to help understand the problem and identify solutions to close the take-up gap.

CPL has built a unique data linkage of California Department of Social Services (CDSS) SNAP participation data with higher education from all three higher education systems in California (California Community Colleges, the University of California, and California State University) and financial aid data from the California Student Aid Commission. Using these data, we published a report in June 2024, which found that a considerable number of students are eligible for SNAP but not currently receiving it. In Fall 2019, 20% of all CCC students and 33% of UC undergraduate students were eligible. However, the majority of eligible students did not receive benefits – only 26% of eligible CCC students and 22% of eligible UC undergraduates were actually enrolled in SNAP. Since publishing this report, we have executed a data use agreement with the California State University and are extending our analysis to CSU students.

Colleges and universities are invested in reaching these eligible, non-participants, particularly in light of emerging evidence that it improves educational outcomes (Chirikov & Rothstein, 2026). However, the data they have available to identify these students is more limited than the rich (but de-identified) linked dataset available to CPL. Identifying likely-eligible students using the same linked data would require burdensome new data sharing agreements.

To help higher education partners overcome these barriers, this study explores: How accurate are each of the unlinked datasets in predicting which students are eligible for SNAP, overall and by demographic group? Is the data available to any college sufficient to identify most likely-eligible students who are not currently receiving SNAP?
We answer these questions by simulating a version of our eligibility measure that uses only the data available to each of our partners to understand the extent to which they could estimate eligibility using data already available to them. We test different models, including a “variable matching” approach and three machine learning models.  We find that the individual datasets perform sufficiently well at estimating eligibility for colleges and universities to leverage their own data now to target outreach, rather than wait for inter-agency agreements. We find that for the UC system, for example, the decision tree model performs best, with 89% precision, 75% recall, 91% accuracy, and 0.82 correlation.

Based on these findings, CPL is providing code and technical assistance to systems leaders, colleges, and universities who are interested in leveraging their own data for targeted SNAP outreach to reduce the take-up gap in this population and support their students’ basic needs.

Author

• Jennifer Hogg, University of California, Berkeley
  Presenting Author