Using modeled voter data to target voters can provide huge benefits and advantages for a political campaign. If you have ever managed or worked on a political campaign it is likely you had to make decisions about voter targeting. In other words, what voters should receive the candidate’s first biography mail piece? What voters should be informed about the candidate’s position on public schools, fracking, or government spending? What voters should be a high priority in your get-out-the-vote universe?
These questions, and choosing the right voters for each universe, are best answered when a campaign is armed with an accurate, current and enhanced database of registered voters. However, a database of registered voters that includes modeled voter data, voter data mining or “microtargeting” predictive scores for each individual voter can take voter targeting to a whole new level.
The process of predictive modeling a population of individual voters starts with a large quantitative survey of 2,500 to 5,000 individuals. The sample size must be large so that machine learning software can have enough demographic examples of what, for example, a Donald Trump or Hillary Clinton supporter looks like in terms of age, gender, registration identification, martial status, etc. The survey data is used to build statistical models that predict each voter’s opinion of an issue or their candidate preference. The predictive modeling software evaluates each respondent’s individual demographic and data variables, and finds other voters who share similar characteristics.
Once the model is built, it will predict a confidence score for each individual within the population. This confidence score shows us who is in support or opposition of a specific candidate or issue. To ensure that the predictive model is correct, cross-validation and a post-modeling survey should be fielded. There is nothing more terrifying to a campaign decision maker than dropping a mail piece to a universe of voters who support the other side of an issue or policy position.
Once the process is completed, the modeled voter data can be used for much more effective voter targeting. A typical survey of voters will inform the campaign of the demographics of voters who are undecided in choosing a candidate. By selecting voters from the database of registered voters using the demographics of undecided voters, and then using the modeled voter data to determine what issues or positions those individual voters strongly support, targeting becomes much easier. Yet the huge benefits that really come from using modeled voter data are realized in the efficiency of the campaign’s spend, voter contact efforts and human resources. The bottom line is that using modeled voter data means that more voters will receive the right message at the right time. At the same time, the campaign will reduce its error rate of contacting the wrong voter with the wrong message, and therefore will be much more efficient with its resources.