What Happened™ in 2022

Voter registration data is public and routinely compiled by political parties and voter file companies, including Catalist.

Registration data on voter files. Generally, states publish voter rolls that include basic data such as names, addresses, voting precincts, age, and gender. However, voter registration laws and practices vary widely by state. For instance, some states require voters to proactively register and re-register for elections after moving. Meanwhile, other states automatically register voters as they interact with state offices, such as the department of motor vehicles, and also assist voters with moving their official address when they move in-county or in-state. 

Age data on voter files. Voting is an age-based right and 41 states report dates of birth, birth years or age on their public voter files. For jurisdictions that do not report age data, age is often represented via commercial or other public data and can also be estimated through modeling.

Generational cohorts in electoral analysis. Catalist tends to rely on birth years and generational cohorts for its analysis. By contrast, most polls divide voters into specific age buckets, such as 18 to 29 or 65+, meaning different generational cohorts will fit into different age categories over time. Normally, these approaches do not yield very different conclusions, but the Baby Boomer generation is very large, so the raw number of people moving into the 65+ category over the past several cycles has increased. This does not mean that older voters are becoming generally more engaged overall — but rather that more people are moving into that age category. Similarly, voters between the ages of 18 and 29 enter the electorate unevenly as presidential election years have higher turnout but only occur every 4 years, which makes apples-to-apples comparisons among the youngest voters across midterms more difficult. 

Gender data on voter files. Gender is well-represented on voter files, with 30 states providing gender data on their public voter files. In states without such data, gender is represented via commercial data as well as modeled based on surveys, naming conventions, and other inputs. A few states allow voters to identify as non-binary when they register to vote or get a drivers license, but these data are not reported on voter files.

Marriage data on voter files. Marriage data is modeled based on name and address conventions, polling, and other factors. 

Population density. Population density is derived from Census data and calculated at the Census tract level. Catalist counts the bottom 25% of tracts as rural, the middle 50% as suburban and the top 25% as urban, relative to each state and county.

Education. Voter files can reliably model aggregate education levels thanks to robust Census data and the degree to which college- and non-college graduates geographically sort themselves. Education is not a direct stand-in for class, income or wealth. A college educated teacher with a great deal of student debt, for instance, may have lower income and wealth than a non-college graduate who owns a small business. 

Race data on voter files. Seven states collect self-identified race data on their voter registration forms and include that data on their public voter files. These states include Alabama, Florida, Georgia, Louisiana, North Carolina, South Carolina and Tennessee. Additionally, voter files include historical self-reported race data from Kentucky and Mississippi, which no longer include this data on their public file. Why these states and not others? Surprisingly, there is no comprehensive historical record as to why this is the case in each state, however some of these states used to hold segregated primary elections which were ruled unconstitutional by the Supreme Court in 1944. During the Civil Rights Era, these data were useful for measuring the impact of the Voting Rights Act in covered jurisdictions, but were not mandated by the act itself.¹These practices are discussed in Hacking the Electorate by Eitan Hersch (2015). Today, these data remain valuable for precisely estimating the racial composition of the electorate; some state legislators have suggested adding race data to their registration forms to help shed light on how representative their electorates are. When self-identified data is available, Catalist prioritizes that information over modeled data, including when a voter identifies their race in a state that collects race data then moves to a state that does not. When self-reported data is not available, Catalist’s Race Model is used. It relies on Census data, naming conventions, poll results and other factors to estimate and assign a race to a given record. Modeling race on voter files has allowed campaigns to reach more voters of color outside of so called “majority-minority” precincts, an important development as suburbs have become more diverse.