Barack and Mitt, take note

Gregory Nemec

View full image

Political campaigns are increasingly using voter targeting: sending tailored messages to voters based on their ethnicity, religion, or special interest. But new research suggests that it doesn’t really work. A new paper expected to appear in the Journal of Politics implies that voters rarely prefer targeted messages to general messages—and that they don’t take kindly to off-target messages.

Eitan Hersh, a political scientist at Yale, and Brian Schaffner, a political scientist at the University of Massachusetts–Amherst, created a fictional congressional candidate named Williams. As part of a survey, they showed the “candidate’s” messages to four groups of potential voters—people in union households, gun owners, born-again Christians, and Latinos—and showed off-target messages to voters who didn’t belong to any of these groups.

Some group members received a generic message (“Williams pledges to work hard on behalf of the middle class”) and others a targeted message (“Williams pledges to represent the interests of Latinos in Congress”). The voters seemed unswayed by targeted messaging: when Latinos were asked to rate a candidate on a scale of 0 to 100, those who got a targeted message rated the candidate about the same as did those who got the generic message.

The most conspicuous finding: mistakes in targeting are costly. Non-Latinos who got the Latino message tended to rate the candidate 25 points lower than did those who got the generic message. “This is not trivial,” says Hersh. He estimates that 25 percent of those targeted as Latino do not identify themselves that way. He has looked himself up on a voter database used by campaigns and found he was listed as an unmarried Protestant. Actually, Hersh is married and Jewish.