Recent changes in how schools report the share of economically disadvantaged students have made it difficult to compare student poverty across time and states. Model Estimates of Poverty in Schools (MEPS) 2.0 is an update to the original MEPS 1.0 K–12, school-level measure of the share of students from households earning up to 100 percent of the federal poverty level that is comparable across states and time and reflects, as closely as possible, the students who attend each school. The measure is now available for school years 2009–10 through 2022–23, and new years of data will be updated annually via the Urban Institute’s Education Data Portal.

Why This Matters

MEPS was created in 2022 to provide a consistent measure of student poverty in the wake of the Community Eligibility Provision’s (CEP’s) growing popularity and the associated decline in accuracy and reliability of the free and reduced-price lunch (FRPL) measure. As universal free meal programs continue to expand, MEPS provides school-level estimates of students from households at or below the federal poverty level, supporting cross-state and longitudinal research and helping policymakers identify schools serving students most in need.

Key Takeaways

• At least 97 percent of US schools each year have MEPS estimates.
• Estimates are available for school years 2009–10 through 2022–23, with new years of data added annually.
• MEPS is publicly available for download via the Education Data Portal.

How We Did It

We built on the previous MEPS 1.0 methodology, making improvements and changes attributable to the lack of data previously used in the estimation process. MEPS 2.0 uses a multilevel (mixed effects) linear regression model to produce comparable estimates and to handle correlated and nonindependent hierarchical data from schools within school districts within states. We use aggregated geographic district-level free lunch and direct certification shares to predict adjusted poverty estimates from the US Census Bureau’s Small Area Income and Poverty Estimates in each available year. We then apply the estimated parameters from this district-level model to the school-level data to predict school-level poverty.