The Small Area Income and Poverty Estimates (SAIPE) program of the U.S. Census Bureau produces model-based estimates of the number of school-age children in poverty in states, counties and school districts. This research explores the use of free and reduced-price lunch (FRPL) eligibility counts for estimating poverty in school districts. Through regression analysis we estimate a positive relationship between FRPL data and Census 2000 poverty estimates with a median prediction error of 30 percent. The high degree of prediction error suggests the FRPL data are not sufficiently precise for formal use in producing school district poverty estimates at this time.