• Description

Most researchers and policymakers rely on the share of students eligible for free and reduced-price meals when describing student socioeconomic background in schools. But shares of students receiving free and reduced-price meals, and other measures related to the distribution of school meals, vary by state and across time because of changes in school meal eligibility criteria.

In this report, we describe the development of a new measure: Model Estimates of Poverty in Schools (MEPS). This measure estimates the school-level share of students from households with incomes at or below the federal poverty level between fall 2013 and fall 2018. The MEPS measure aims to be comparable across states and over time and to broadly align with the school's enrolled population (as opposed to a neighborhood measure).

We find that MEPS broadly aligns with aggregate state measures of student poverty and are strongly correlated with geographic district poverty as measured by the Small Area Income and Poverty Estimates (SAIPE) program. We also find that MEPS can under- or overestimate poverty shares for certain districts. In particular, we find that our model underestimates school-level poverty for districts enrolling high shares of Black students. To correct for this, we produce modified MEPS, a second measure that mechanically adjusts our estimate to align with geographic district poverty rates. Because of wider margins of error for districts with small populations in the SAIPE data, we recommend using modified MEPS only for analysis of geographic districts with more than 65,000 residents.

These statistical estimates should be used primarily by researchers. MEPS could be useful for those conducting research across states or years or for policymakers who want to understand how a school's socioeconomic characteristics may have changed over time. But these estimates are not appropriate for allocating resources within a state or district or for other uses when having a true count, rather than a model estimate, is required.