Our rigorous, data-driven approach helps us do more with less and provide the civic engagement community with crucial information.
The Voter Participation Center’s groundbreaking research played an instrumental role in identifying women’s marital status as an indicator of voter registration and turnout—and, later, the potential electoral power of the Rising American Electorate.
We know that marital status, in and of itself, is a fundamental factor in registration and voting behaviors for women. But in order to reach unmarried women regardless of their location, we had to come up with a list-based approach that could reliably identify whether a woman was married or unmarried—leading us to develop our own marital status model, because the ones that already existed weren’t accurate or complete enough for our purposes.
Over time, we realized that other groups — people of color and millennials — were equally disengaged and underrepresented, and we developed models that improved the accuracy of our ability to target the members of these populations who were most likely to take action: to register or re-register to vote, to vote by mail, to turn out on Election Day, or to take action on a policy issue.
These statistical models, which go well beyond the modeling used in many political and advocacy campaigns, enable us to target our voter registration, mobilization, and advocacy efforts more effectively: response modeling for registration and vote-by-mail, issues modeling to target voters who share the same values, and advocacy modeling to predict who will take action on a public policy debate.
One of the things we realized in this process was how significant a factor mobility is in voting behavior — finding, for example, that a voter who moves during an election cycle is 12 times more likely not to vote than the average consistent voter. In response, we focused additional efforts on re-registering voters who had moved, and developed a new statistical model to achieve better predictions about movers’ likelihood to return a voter registration form.
Because we’re also able to track the response rates for our efforts, we’re constantly improving our statistical models, and creating new ones to make our programs even better. And we’re making our statistical models available to our partner organizations, so our research is benefitting the whole civic engagement community.
We’ve also made immense strides in list-based voter registration, using modern commercial direct-marketing techniques to improve civic engagement among the Rising American Electorate.
One of our fundamental advances has been our list-based, universal approach to registration that focuses more on demographics than on particularly dense geographic data.
We realized from the outset that it would be essential to increase the quality and value of our lists in order to increase voter participation, cost efficiency, and effectiveness. And over the past decade, we’ve made enormous progress—hard-won, detail-by-detail progress—in converting low-quality commercial data into high-quality lists of unregistered Americans.
Our secret is to merge together and compare multiple commercial data sources along with historical records to build a list. Then, we apply statistical modeling algorithms to identify and remove records that are unlikely to represent a real person, that are likely to already be registered to vote, or that are reported as deceased by the Social Security Administration’s Death Master File.
While no process is 100% foolproof, we’re as thorough as we can be in filtering out inaccurate information—and we’re constantly striving to improve our methods and accuracy. Our investments in list enhancement have enabled us to achieve dramatically-higher ROI by excluding people who shouldn’t be mailed and getting better response rates from those who are mailed.