2020 Report Methodology
Nonprofit Voters and Turnout: Voters analyzed through this program were engaged by participating nonprofits through either paper or digital means. These “voter contacts” include voter registrations (including updating their registration, checking registration status, and applying to register in the state for the first time), pledge-to-vote cards, and absentee ballot requests. In total, 35k voters submitted analyzable information through these channels, and 25k of those were matched to the Catalist voter file. By matching to the voter file, we are able to determine turnout, age, and sex. This data was supplemented with Catalist’s marital status, race, income, education, and vote propensity models.
Comparable Voters: We used Catalist data files to construct a dataset of comparable voters with matching geographic and demographic composition as the individuals in our nonprofit voter dataset. To do this, we assigned each of our nonprofit voters an expected turnout based on the turnout of their registered counterparts from Catalist, matching them by county, sex, race, age, and marital status. This process ensures our comparisons account for the unique demographic distribution of the nonprofit voters we engaged. For income, vote propensity, and education level, we also constructed a dataset of comparable voters with additional matching characteristics.
Turnout Boost: Boosts were calculated by finding the difference between the average turnout of nonprofit voters themselves and the average turnout of comparable voters (the expected turnout of a demographically and geographically similar population, modeled as noted above).
Paper & Digital: Paper contact methods included paper registrations, pledges, vote by mail application forms, and sign-in sheets. Digital contact methods included digital forms of the above plus the use of a portal to check registration status. About 41% of all contacts were collected through paper means and 59% via digital ones.
Catalist Voter File: Federal law requires all states to maintain publically accessible voter rolls that include name, address, and whether they voted in past elections. Catalist, using additional data to model voter demographics, supplements the state rolls to create their voter files.
Education Level: Likely College vs likely Non-College was a comparison of education level provided by Catalist modeling on a scale of 0 to 100, where 0 to 50 represents voters who likely don’t have college degrees and 51 to 100 represents those who likely do.
Propensity Score: Propensity is a likelihood of voting score modeled by Catalist. Its use is popular among campaigns to target high-likelihood voters. 0 to 50 are less likely to vote and 51 to 100 are scored as more likely to vote. This is calculated based on prior vote history and other factors.
Among other things, the survey asked what services the nonprofit offered, budget size, number of people served, and staff size. We then created groupings of these organizations based on the distribution of results. Sites also self-selected whether they did in-person, digital, or both kinds of voter engagement as well as who conducted the voter engagement and when/where it took place. By cross-referencing site survey data with voter contacts made by those sites, we were able to analyze voter turnout and demographics specific to different site groupings like health centers or food pantries. Method of voting determinations in the Practitioner’s Report came from Catalist reported data of how people voted in the 2020 election.
In addition to our online survey, we conducted follow-up interviews with 5 participating nonprofits. These follow-ups went deeper into their initial responses, providing context and insights that were outside the scope of the survey. They also told us about successful strategies and gained new insights for engaging particular communities. Their experiences are featured in the accompanying case studies as well as the quotations and photos throughout the report.
Lastly, Get Out the Vote (GOTV) numbers by type of outreach were determined by the total number of program participants who received mailers, texts, live-scripted phone calls, or voicemails from Nonprofit VOTE or participating sites.