The Numbers Behind 'The Persuadables'
What is a persuadable voter?
We considered voters to be persuadable when they had a predicted partisanship in the middle range—between 30 and 70, on a scale where 0 is Republican and 100 Democrat—and at least a 70 percent likelihood of casting a ballot.
Where does this data come from?
We partnered with Clarity Campaign Labs and TargetSmart Communications, two Democratic-aligned, Washington, D.C. firms that specialize in collecting and analyzing information on voters. TargetSmart works with the Democratic National Committee and provided core components of the data infrastructure to the 2012 Obama campaign. Clarity is a targeting, analytics, and polling firm that services Democratic and progressive clients across the United States, but is not working directly with any of the Senate candidates profiled here.
Where do they get their numbers?
The underlying information is from voter-registration records assembled and continuously updated by TargetSmart. In many cases, we have tracked changes to local electoral rolls through the end of last week, accounting for new and changed registrants. TargetSmart augments this with material from the Census, other public sources like hunting licenses, and consumer marketers. In cases where information is not available—only some states, for example, require voters to include their race or ethnicity when registering—Clarity develops statistical models that predict individual attributes. Other models we used predict the percentage likelihood than an individual voter will own a gun, have a college degree, or regularly attend church.
What do you mean by “total vote”?
That’s our projection of the total votes that will be cast in each state. It’s based primarily on multiplying a state’s total number of registrants by the average probability Clarity has calculated that each one will cast a ballot this year. (In most cases, the resultant total vote is a number between the actual 2010 and 2012 turnout figures, while reflecting population growth and the differing intensity of each state’s campaign environment.)
What are these different categories of voters?
We sorted each state’s electorate into three groups in much the same way a candidate would when developing a campaign plan.
(To pare down to plausible targets for the campaigns, we excluded those whom Clarity predicts have a less than 10 percent likelihood of voting.) Our primary tool for this triage was Clarity’s statistical models predicting the partisanship of individual voters.
How did you calculate each party’s base?
We labeled those voters with strong partisan loyalties and a high likelihood of voting this year as each party’s base.
How about in states, especially in the South, where a lot of registered Democrats never vote for Democrats for national office anymore?
Clarity’s models reflect a voter’s party registration, where available, but account for the interplay of hundreds of other variables, too, anticipating voting behavior rather than vestigial attachments.
So what about GOTV (“get out the vote”) targets?
Again we used the 70 percent threshold to weed out those who are already probable to turn out. Typically campaigns will focus their get-out-the-vote efforts on those in a middle range—effectively, those who do vote, but do so unreliably.
How do you know that those voters are really susceptible to being persuaded by both sides?
We don’t, but we think this is a good way of estimating. Candidates in each state have access to their own internal polling, modeling, and research that further refines who a “persuadable voter” is given the particular dynamics of their races. But our method allows us to quantify and define such people in a uniform language, so they can be compared across multiple states.