Persuasion modeling

Persuasion modeling predicts how movable an individual voter's views are, how likely they are to be swayed toward or away from a candidate or position, rather than whether they will vote at all. It is built from voter file data combined with survey responses, past contact history, and consumer or demographic data, and is one of the two core targeting models used in modern campaigns, alongside turnout modeling.

Campaigns commonly cross these two models to sort voters into groups: persuadable voters, base supporters who need only to be turned out, opposition voters unlikely to be worth contacting, and low-propensity voters who are both unlikely to vote and hard to persuade. A more rigorous version uses uplift modeling, which estimates the incremental effect of contacting a specific voter with a specific message, rather than their raw likelihood of support, so outreach targets people whose behavior contact can actually change.

Persuasion models draw on the same underlying data as microtargeting and raise the same debates about privacy and the line between tailored communication and manipulation; regulators and platforms increasingly treat political targeting data as sensitive. A common pitfall is confusing a high persuasion score with a high turnout probability — a highly persuadable voter who does not vote has no effect on the outcome, which is why campaigns model the two dimensions separately.