Using American Community Survey Data to Expand Access to the School Meals Programs (2012)
Chapter: Appendix D: American Community Survey (ACS) and Small Area Income and Poverty Estimates (SAIPE)-Like Tabulations Requested from the U.S. Census Bureau
This appendix presents the panel’s request to the Census Bureau for eligibility estimates for the school meals programs from the American Community Survey (ACS). The request first provided specifications for deriving direct ACS estimates based on the panel’s research, as described in Appendix B. Then, it described the level of detail desired for Small Area Income and Poverty Estimates (SAIPE)-like model-based estimates. Although the specifications for the latter estimates were adequate for the purposes of this study, the panel suggests that if SAIPE-like modelbased estimates and their standard errors are needed in the future, the request should include the estimated covariance of the two percentages in addition to their standard errors. This will facilitate the calculation of standard errors of derived estimates such as percentage eligible for free or reduced-price and the blended reimbursement rate.
The goal is to obtain estimates for the percentages of public school students eligible for free, reduced-price, and full-price school meals for all school districts in the country and for the school attendance areas in the five case study districts. We would also like estimates for the total number of public school students associated with those school districts or school attendance areas. This is an estimate for potential enrollment and will be used for evaluation. If there is a disclosure issue in a geographic region, the total number of public school students is the variable that should be suppressed. We would like to receive standard errors for all estimates.
Tabulations for school districts are requested from all five ACS 1-year releases; all three 3-year releases; and the one 5-year release for 2005, 2006, 2007, 2008, and 2009, as allowed under disclosure restrictions based on the size of each district. Most likely the tabulations for school attendance areas will be possible only using the 5-year ACS.
For all school districts, the most recent geographic boundaries, as recorded in the Topologically Integrated Geographic Encoding and Referencing (TIGER) database, should be used. In the case study districts, the School Attendance Boundary Information System (SABINS) has provided digitized boundaries and linkages to the National Center for Education Statistics’ (NCES’) Common Core of Data (CCD) for school attendance areas.
Identifying Public School Students
ACS person-level variables are used to define public school students:
- AGEP less than or equal to 19 (defines children);
- SCH = 2 (attended public school in last 3 months);
- SCHL less than 16 (no high school diploma or general educational development [GED] credential, no college);
- REL not equal to 16 or 17 (eliminate group quarters students); and
- SCHG =11 through 14 (grade level), used to assign students to school districts or school attendance areas by comparison with grade composition in school (from CCD or case study districts).
Measuring Eligibility for Schools Meals Programs
ACS variables are used to categorize each student according to eligibility (FREE, REDUCED PRICE, FULL PRICE):
- If REL = 14 (foster child), then FREE.
- If FS = 1 (household receiving food stamps), then FREE.
- If no other person in student’s HH has:
AGEP > 19 and REL = 11, 12, or 15 (no adult nonrelatives who are not unmarried partners), then:
If any person in HH has PAP greater than zero (someone receiving public assistance income), then FREE.
____________
1 This change allows us to include prekindergarten students if a school includes prekindergarten among its grades.
Otherwise
Compute lunch index as ratio of HINCP (household income) to poverty guidelines2 (expressed as a percentage) associated with household size of NP and issued in year of ACS year. Any foster children in the household and their income should be subtracted from NP and HINCP, respectively.
If lunch index less than or equal to 130, then FREE;
If lunch index greater than 130 and less than or equal to 185, then REDUCED PRICE;
If lunch index greater than 185, then FULL PRICE.
Otherwise (there are adult nonrelatives in the household who are not unmarried partners):
If student has REL less than or equal to 10 or equal to 13 (student part of core family):
If any person in HH with REL less than or equal to 10 or REL = 13 has PAP > 0, then FREE (someone in core family receiving public assistance income).
Otherwise
Compute sum of PINCP for all persons in HH with REL less than or equal to 10 or equal to 13. Compute count of such persons. Compute lunch index as ratio of sum of PINCP to poverty guidelines associated with number of people (using guidelines appropriate to year of ACS).
If lunch index less than or equal to 130, then FREE;
If lunch index greater than 130 and less than or equal to 185, then REDUCED PRICE;
If lunch index greater than 185, then FULL PRICE.
If student has REL = 11, 12, or 15 (student part of second economic unit):
____________
2 See http://www.fns.usda.gov/cnd/Governance/notices/iegs/IEGs.htm for the poverty guidelines associated with the school meals programs.
If any person in HH with REL = 11, 12, or 15 has PAP > 0, then FREE (someone in second economic unit received public assistance income).
Otherwise
Compute sum of PINCP for all persons in HH with REL = 11, 12, or 15. Compute the count of such persons. Compute lunch index as ratio of sum of PINCP to poverty guidelines associated with number of people (using guidelines appropriate to year of ACS).
If lunch index less than or equal to 130, then FREE;
If lunch index greater than 130 and less than or equal to 185, then REDUCED PRICE;
If lunch index greater than 185, then FULL PRICE.
The goal is to obtain SAIPE-like estimates of the percentages of public school students who are eligible for free and for reduced-price school meals, and their standard errors. The ACS-based estimates for these quantities are defined above.
The panel would like to have SAIPE-like estimates for all school districts (and states and counties) in the country for each year 2005-2009. We would also like estimates for the school attendance areas in the five case study districts. We assume that the latter would be derived using a shares-based approach that is similar to the approach used to derive school district estimates.
The panel realizes that preparing these tabulations requires adaptation of the current SAIPE model, and the time frame of the study is short. Nonetheless, the SAIPE-like estimates will provide a proof of concept for using small-area estimates of eligibility for schools and school districts.
