How Do We Identify and Count Charter Schools?
The National Alliance serves as a repository for charter school data. Data are collected from a variety of sources including:
- State data files containing enrollment, directory, and performance information
- Federal datasets, such as Common Core of Data (CCD), the Civil Rights Data Collection (CRDC), and EdFacts
- State and local partners, and in some cases, individual schools
In addition to quality control and verification, the data are processed and categorized based on the type of record.
How Do We Categorize Schools?
Each record receives a Modified Count code. A single record—typically represented as a row or unique identification number in a dataset—is not necessarily school. For example, the record may represent a program within a school, or a campus associated with another school. These record types can vary between federal and state datasets, and appear frequently among charter school records. The Modified Count code indicates whether the school is a campus, charter holder, program, or indeed a school. Without accounting for, and ideally rectifying, these discrepancies, one may generate inaccurate counts of charter schools. charter school openings and closures, and other important metrics. The National Alliance’s Modified Count code allows researchers to consistently classify records across data sources and school years. Overtime, we have seen an increasing number of campus records. From 2005-06 to 2018-19, the number of campuses as the total number of records doubled from 4 to 8 percent.
Charter Record Type by Year
Some states appear disproportionately affected by the use of campuses and holder records. States with some of the largest charter school sectors, such as D.C., Texas, and Florida, also have some of the highest proportions of non-school records.
States With More than 5% of Records as Non-Schools
How Do We Account For Changing School Identification Codes?
Using an identification system developed by the National Alliance, data are merged together into our Database of Charter Schools. This system allows us to identify schools across time, regardless of whether state or federal IDs have changed. Because state and federal IDs contain embedded information, a change to any of the underlying information contained within the ID means that a school’s ID will change even if the school remains operational. The creation or elimination of state or federal IDs are not necessarily a result of changes to a school’s operational status. Accounting for changes due to these factors is important for properly accounting for school openings and closings, particularly for charter schools. Between the 2005-06 and 2018-19 school years, charter schools changed their federal IDs (NCES ID) almost 1,600 times, and in some cases two or more times. This means that out of all charter records across that time, 14% saw their NCES ID change, despite not actually closing. Table 4 shows the top 10 states with the most NCES ID changes as a proportion of total records. Reasons for these changes vary depending on state contexts. In some cases, charter schools can change authorizers, which may result in a NCES change. In the case of Louisiana, numerous schools saw their NCES change when they moved from the Recovery School District (RSD) to the Orleans Parish School Board. Conversions, mergers, and expansions are other factors that can lead to an NCES change.
Top 10 States With the Most NCES Changes
|States||Total Unique Records||Number of Changes||Percent Changed|
How Is Demographic Data Collected?
Demographic data for student racial/ethnic backgrounds were collected using a combination of CCD and state-level datasets. Demographic data for “Other/Two or More Races”, “Native American/Alaskan Native”, and “Pacific Islander/Hawaiian Native” classifications were summed into a single category labeled “Other”, along with any missing demographic enrollment, for the purposes of this report. As individual groups, these populations constituted low percentages of the overall population in both charter and district schools with a relatively steady trend, although it should be noted that “Other/Two or More Races” alone reached greater than 4% of the population for the first time in 2018-19.
Demographic data for students with disabilities (SWDs), economically disadvantaged students (EDSs), and low English proficiency students (LEPSs) were collected using EdFacts datasets from the 2009-10 through the 2017-18 school years. Using EdFacts provided a more consistent and up-to-date report of SWD, EDS, and LEPS data for the years examined, compared to other datasets available at the time of this draft. Although EdFacts reflects the population of students participating in state assessments, rather than total student enrollment, careful review by the NAPCS Data and Research team determined that EdFacts participation data was a valid and representative sample of the overall student population. EdFacts data did not significantly skew the prevalence rates for the special populations in question. The data presented in this report was found by averaging the values of the Math EdFacts and ELA EdFacts data in each appropriate year.
How Do We Distinguish School Districts?
Because charter schools are often coded as their own districts on the national level, we assign all charter schools a geographic school district in order to find school and enrollment share. We acquire addresses for every public school (charter and district) from CCD and our own collection process. We then use Geocodio to find latitude and longitude values and use ArcGIS to plot them on public school district shapefiles. This allows us to see which geographic school districts charter schools would fall into if they were a part of the public school system.
In our enrollment share analysis, we eliminate virtual schools. (Most are tagged with a virtual flag from CCD, but we also do a manual check for schools with virtual identifiers in their name, such as virtual, online, digital and remote. Furthermore, we check schools with populations far exceeding what we would expect based on location as this can often indicate virtual enrollment.)
We derive locale data from CCD as well.
How Is Poverty Data Collected?
Census tract information is taken using methods similar to district enrollment share. After excluding virtual schools, we use addresses from CCD, longitude/latitude from Geocodio, and census tract shapefiles from the Census Bureau. Schools are plotted using ArcGIS.
Once a school has been plotted in a census tract, we take 2018 5-year estimation data from the American Community Survey to determine if a census tract is high-poverty, using the threshold of 20% or more of the population above the poverty line. This allows us to say how many students are enrolled in and how many schools are located in a high-poverty area.
Measuring School Finances
Data on school finances are obtained from the 2006-07 through 2016-17 National Public Education Financial Survey. The survey contains an indicator for whether the district has charter schools. To isolate the financing of each sector, we only use data from districts that have only charter schools or only district schools. Because of this, we do not have the universe of charter or district schools. We also exclude states for which we do not have data on at least 50 percent of charter schools. Still, we account for approximately 60 percent of charter schools, though we notably are missing data from California, New York, and Florida, three states with some of the largest charter school sectors. We also adjust for inflation using 2016-17 constant dollars.