DISMISSED: METHODOLOGY
Data Sources
To create the school district border dataset, EdBuild used the folowing data sources.
• School district boundaries: geography for school district borders for the 2017-18 school year come from the US Census Bureau, Education Demographic and Geographic Estimates Program (EDGE), Composite School District Boundaries File.
• School district revenues: Revenues from federal, state, and local sources for the 2016-17 school year come from the Census, Annual Survey of School System Finances (F33).
The following subtractions were made from total state and local revenues for each school district:
1. Because it can contribute to large fluctuations in district revenues from year to year, we exclude revenue for capital from the calculation of state revenues.
2. Similarly, we exclude money generated from the sale of property from local revenues, because it too can contribute to large fluctuations in revenues.
3. In just under 2,000 districts, revenues received by local school districts include monies that are passed through to charter schools that are not a part of the local school district but are instead operated by charter local education agencies (charter LEAs). This artificially inflates the revenues in these local school districts because they include money for students educated outside of the district who are not counted in enrollment totals. To address this, we subtract from state and local revenues a proportional share (based on the percent of each districts’ revenues that come from local, state and federal sources) of the total amount of money sent to outside charter LEAs—an expenditure category included in the F33 survey.
4. In Arkansas, large portions of districts’ revenues that should be considered local are categorized as state revenues. The value of this misattribution for each district is described in the F33 documentation as C24, Census state, NCES local revenue. Before analysis, the value of C24 is subtracted from state revenues and added to local revenues for the state of Arkansas.
5. In Texas, many districts report exorbitantly high per-pupil revenues. This is in part because of the policy and procedures for recapturing and redistributing local revenues raised by property-wealthy districts in the state. In the F33 survey, recapture is reported as expenditure code L12. Because these monies are included in the state revenue for other, receiving districts, we subtract a districts’ L12 expenditures from their local revenues for the state of Texas.
See the F33 Survey Documentation and File Layout for state-specific notes relation to education finance data.
• School district enrollments and racial composition: School district enrollment characteristics for the 2016-17 school year come from the US Department of Education, National Center for Education Statistics, Common Core of Data (CCD).
• School district school-age poverty rates: School district-level data on poverty rates among relevant school-age children in 2017 come from the Census, Small Area Income and Poverty Estimates (SAIPE) .
• School district community indicators: school district-level data on median owner-occupied property value and median household income for the 2016-17 school year come from the US Department of Education, National Center for Education Statistics, Education Demographic and Geographic Estimates (EDGE).
Methodology
To begin, EdBuild conducted a spatial analysis of all unified districts in the nation. This process identified all pairs of school district neighbors that share a land border (districts whose shared border exists entirely along a large body of water were not considered to be neighbors). Pairs were then excluded from this neighbor list if their shared boundary was less than 500 feet or if the districts are in different states.
Each neighbor pair was identified by their shared school district border and joined to the above described data from the SAIPE, CCD, and ACS. Then we made the following calculations.
Percent nonwhite calculations: The proportion of students enrolled in a district that are nonwhite was calculated by dividing the number of nonwhite students by the total enrollment within a given district.
Revenue calculations: Per-pupil state and local revenues were calculated by dividing the state and local revenues (adjusted to exclude the monies described above) by fall enrolment counts as reported in the F33 survey. A school district’s total revenue per pupil as displayed in the map on the website and in the report’s tables and text is the sum of its state revenue per pupil and local revenue per pupil.
The revenue figures are not cost adjusted as the analysis focuses on differences between neighboring districts, which are assumed to have the same cost of living.
School District Exclusions
EdBuild employed several exclusion criteria in compiling our borders dataset. Our analysis includes only districts that meet our standard requirements for a finance-based analysis. EdBuild excluded districts that are of types 5 (vocational or special education), 6 (nonoperating) or 7 (educational service agency) in the F33 data. If F33 school type is missing, EdBuild excluded districts that are of types 4 (regional education service agency), 5 (state agency), 6 (federal agency), 7 (charter agency) or 8 (other education agency) based on Common Core of Data excepting 26 type 7 (charter agency) districts that are the sole education provider for a geographic area. Further removed were all districts with missing or zero total enrollments, all districts with missing or zero operational schools, and all districts with missing revenues.
Districts with very low revenues (<$500) and very high revenues (>$100,000) were also excluded.
Geographically, EdBuild excluded any districts from the US territories. Further, because EdBuild only identitfies within-state school district neighbors, Hawaii and the District of Columbia were excluded as they each have only one school district.
Edbuild also excluded all elementary and secondary school districts from our dataset, leaving 10,548 unified districts. There are three types of school districts: unified, elementary, and secondary. Thirty states and the District of Columbia have only unified districts. Unified districts are geographically distinct, while elementary and secondary districts overlap. The analysis was confined to unified school district pairs to avoid comparing resources across districts of different types which may have very different structures and needs.
EdBuild also removed all districts with the urbanicity rural, remote. This urbanicity classification comes from the US Department of Education, National Center for Education Statistics, Common Core of Data (CCD) which uses an urban-centric locale assignment system. A rural, remote district is a Census-defined rural territory that is more than 25 miles from an urbanized area and is also more than 10 miles from an urban cluster. EdBuild removed these 1,958 districts from our dataset. Finally, EdBuild removed 260 districts because they had a student density of less than or equal to one student per square mile. These exclusions were made since these districts may have reason to resource differently than their more populous neighbors and they have unique geographic constraints due to the extremely low student density.
This resulted in a dataset that contains 8,330 districts and 18,857 pairs of district neighbors.
Analysis
For each school district pair in our dataset, EdBuild calculated the following:
1. The percentage point difference in percent of nonwhite students
2. The percentage point difference in poverty rate
3. The absolute and percent difference in local revenue per pupil
4. The absolute and percent difference in state revenue per pupil
5. The absolute and percent difference in total revenue per pupil, with and without impact aid
a. To calculate the difference between neighboring districts' total revenue, impact aid per pupil was added to total revenue per pupil for each district. Impact aid is federal general aid to "assist local school districts that have lost property tax revenue due to the presence of tax-exempt Federal property, or that have experienced inreased expenditures due to the eneorllment of federally connected children, including children living on Inidian lands." EdBuild included impact aid in our difference calculations to ensure that the total revenye gap between districts was not driven solely from the presence of tax-exempt property or enrollment on Indian lands.
b. Impact aid is not included in all reported total revenue and total revenue difference figures in the report text and tables or on the website. It was only included to find the district pairs categorized below.
6. The absolute and percent difference in median household income
7. The absolute and percent difference in median property value
EdBuild then categorized district pairs into 6 tiers:
1. Revenue gap of at least 10%: district pairs with at least a 10 percent difference in total revenue per pupil including impact aid.
2. Revenue gap of at least 20%: district pairs with at leaste a 20 percent difference in total revenue per pupil including impact aid.
3. Race gap of at least 25 percentage points: district pairs with at least a 25 percentage point difference in the percent of nonwhite students.
4. Race gap of at least 50 percentage points: district pairs with at least a 50 percentage point difference in the percent of nonwhite students.
5. Divisive borders: district pairs with at least a 25 percentage point difference in the percent of nonwhite students and at least a 10 percent difference in total revenue per pupil including impact aid.
6. Deeply divisive borders: district pairs with at least a 50 percentage point difference in the percent of nonwhite students and at least a 20 percent difference in total revenue per pupil including impact aid.
For the national and state analysis, EdBuild grouped district pairs by the categories outlined above and calculated the following:
• Divisive borders: the number o pairs included in the category
• Disadvantaged districts: the number of unique districts included in the category which are more nonwhite and receive less revenue
• Advantaged districts: the number of unique districts included in the category which are less nonwhite and receive more revenue
• Students in disadvantaged districts: the total number of students enrolled in the category which are more ninwhite and receive less revenue
• Students in advantaged districts:the total number of students enrolled in the category which are less nonwhite and receive more revenue
• Average enrollment in disadvantaged districts: the average number of students enrolled in each district in the category which are more nonwhite and receive less revenue
• Average enrollment in advantaged districts: the average number of students enrolled in each district in the category which are less nonwhite and receive more revenue
• Number of pairs in the same county: the number of school district neighbors in each category where both districts are in the same county
• Percent of pairs in the same county: the percent of school district neighbors in each category where both districts are in the same county
• Average local revenue per pupil difference: the average difference in local revenue per pupil between pairs in the category (both in absolute dollars and as a percent difference)
• Average state revenue per pupil difference: the average difference in state revenue per pupil between pairs in the category (both in absolute dollars and as a percent difference)
• Average total revenue per pupil difference: the average difference in total revenue, including impact aid, between pairs in the category (both in absolute dollars and as a percent difference)
• Average poverty rate: the average poverty rate of the districts included in the category
• Average MHI: the average median household income for the districts included in the category (both in absolute dollars and as a percent difference)
• Average MPV: average median property value for the districts included in the category (both in absolute dollars and as a percent difference)