ABSTRACT
Accurate information on who is eligible to vote—and who is registered—is essential to voting rights enforcement, election reporting, and scholarship on political participation. Yet, the data used to measure eligibility and registration vary in important ways based on how the data were collected and for what purpose. This article compares three federal data sources on citizenship and voter registration, alongside state voter roll data aggregated by a private vendor (L2). We assess their strengths and limitations, helping researchers make informed choices. We review existing research on how individual preferences, policies, and politics shape registration rates across geography and demographic groups. We also explain why some uncertainty in counts of eligible and registered voters is unavoidable. Our side-by-side comparison reveals striking inconsistencies. The American Community Survey overstates the citizen voting-age population. The Current Population Survey and L2 voter file counts of registered voters show a very weak correlation across states. Counts for small geographic areas vary widely. Because we are largely unable to disentangle the source of differences we observe, we recommend using multiple data sources. We also call for more attention to how voter rolls are maintained at the local level—key to reducing uncertainty.