Developing and Evaluating Methods for Using American Community Survey Data to Support the School Meals Programs: Interim Report (2010)

Previous chapters have described the technical details of the work to be done by the panel to develop methods using American Community Survey (ACS) and other data for estimating claiming percentages for reimbursement by the U.S. Department of Agriculture (USDA) for universally free school meals served under a new Provision 4. This chapter describes the panel’s approach in making sure the proposed methods can be implemented in practice. It comments on agreements and reviews needed at various steps of the estimation process to ensure operational feasibility. It describes these steps in terms of the two main factors that influence feasibility: (1) geographic detail of estimates (school district versus individual school or group of schools) and (2) ACS direct estimates versus model-based estimates.

An additional complexity is introduced by the adjustments discussed in Chapter 5, not only those needed to better reflect eligibility criteria, enhance timeliness, and capture the impact of charter and magnet schools,¹ but also those needed to reflect students’ participation and the meals that will be served under Provision 4. Details associated with these adjustments and the data needed to make them are as yet not fully specified. Information provided by the case study districts, the Census Bureau, and the Food and Nutrition Service (FNS) will help the panel develop recommendations for how adjustments can best and most easily be imple-

mented. The approaches discussed below will require one or more interagency agreements between the Census Bureau and FNS that will address schedules for activities, resources that will be devoted to those activities, data use restrictions, and other issues ultimately determining whether recommended methods and procedures are operationally feasible.

As noted in Chapter 3, the Census Bureau maintains up-to-date school district boundary information and already provides special tabulations of the ACS for school districts. Furthermore, it provides estimates from its Small Area Income and Poverty Estimates (SAIPE) Program for all school districts included in its geographical database. Hence, it should be relatively straightforward for the Census Bureau to prepare estimates for school districts according to specifications from the panel.

The direct ACS-based estimates of eligibility described in Chapter 5, possibly with the indicated adjustments, are similar to data products already produced by the Census Bureau. These estimates would be based on 1-year, 3-year, and (eventually) 5-year ACS data. As noted in Chapter 5, 1-year ACS estimates will be publicly available only for school districts with a population greater than 65,000 (approximate school enrollment of 11,700). Three-year ACS estimates will be available for all school districts with a total population greater than 20,000 (approximate school enrollment of 3,600). Beginning in late 2010, 5-year estimates will be available for all school districts.

The panel’s work aims to result in model-based estimates of eligibility that are of sufficient accuracy to be considered for use in developing claiming percentages for determining reimbursements under Provision 4. The current SAIPE process, on which the panel’s work will be based, produces model-based estimates of poor school-age children that are publicly released for all school districts in the country. Should FNS decide that model-based estimates of eligibility could be used in developing claiming percentages for determining reimbursements to school districts under Provision 4, an interagency agreement with the Census Bureau would be needed for regularly providing estimates for the multiples of the poverty thresholds that are used for free and reduced-price school meals. Work would be needed by the Census Bureau to incorporate the panel’s meth-

odology, potentially including adjustments, into a production system for acquiring and preparing data and deriving estimates. Model-based estimates would need to be reviewed by the Census Bureau’s disclosure review board to make sure public release would not jeopardize the confidentiality of ACS respondents. Should such an approach be feasible, with no confidentiality concerns, estimates would be publicly available for all school districts.

The challenge associated with providing estimates for schools or groups of schools in districts for which Provision 4 would not be adopted for the entire district for financial or other reasons is that the Census Bureau does not maintain boundary information at this level of geographic detail. Hence, the panel will need to propose a process, most likely an annual process, by which school districts provide to the Census Bureau school attendance-area boundary information and possibly auxiliary information, such as counts of students directly certified, and the Census Bureau provides to the districts ACS direct or model-based estimates using the methods developed by the panel. Among the issues that are not yet clear are the cost to school districts and how geographic issues (such as split blocks) could be resolved. These details will be considered by the panel with input from the Census Bureau, FNS, and the case study districts. Another issue that is not yet clear is how many school districts might participate in the process with the Census Bureau. Based on the simple tabulations presented in Chapter 4, it is likely to be a minority—perhaps a small minority—of districts.

The Census Bureau has indicated that if school districts were to provide sufficiently accurate digitized school attendance-area boundaries or lists of the census blocks and block groups associated with school attendance areas (or both), it would be able to provide special ACS tabulations of students eligible for free and reduced-price school meals for those areas. The estimates for school attendance areas or for groups of schools would be based on 1-year, 3-year, or 5-year ACS data, depending on the population size of the attendance area for a school or group of schools. School-level detail is most likely to be available only from the 5-year ACS data. Other details that would need to be worked out include how the adjustments described in Chapter 5 would be incorporated into the estimates.

Obtaining model-based estimates for schools or groups of schools would also require accurate digitized school attendance-area boundaries or lists of the census blocks and block groups associated with school attendance areas. As the panel works through the development of model-based approaches, it will need to specify procedures for producing estimates for school attendance areas only for districts that request such estimates. This need will affect the details associated with how estimates are derived, as this is different from the current SAIPE process that produces estimates for the entire universe of areas at each geographic level (states, counties, and school districts).

Once the details of the estimation methodology have been worked out, the Census Bureau would need to incorporate the panel’s methodology into a production system. The bureau’s disclosure review board would need to review school-level estimates. However, the current SAIPE process produces estimates for some small districts that have only a single school, so incorporating estimates for selected additional school attendance areas should not significantly change the risk of disclosure.

The panel will need to work with the Census Bureau to determine whether it would be efficient to add school attendance-area estimates to the school district-level product described above. If not, the panel will need to develop details of other approaches with input from FNS and the Census Bureau. The panel looks forward to working with the case study school districts, FNS, and the Census Bureau to determine methods that are operationally feasible for producing estimates that could be considered in a new Provision 4 to expand the reach of the school meals programs and minimize administrative burden for schools and families.

This page intentionally left blank.