A political candidate running for regional public office asked a famous political psychologist what kind of television ads she should air in three heavily contested districts: positive ones emphasizing her own record, or negative ones attacking her opponent’s record.
“You’re in luck,“ said the psychologist. “I have a new theory of persuasion that addresses exactly this question. I just published a paper containing four large studies that all strongly support the theory and show that participants are on average more persuaded by attack ads than by positive ones.“
Convinced by the psychologist’s arguments and his confident demeanor, the candidate’s campaign ran carefully tailored attack ads in all three districts. She proceeded to lose the race by a landslide, with exit surveys placing much of the blame on the negative tone of her ads.
As part of the campaign post-mortem, the candidate asked the psychologist what he thought had gone wrong.
“Oh, different things,“ said the psychologist. “In hindsight, the first district was probably too educated; I could see how attack ads might turn off highly educated voters. In the second district““and I’m not going to tiptoe around the issue here—I think the problem was sexism. You have a lot of low-SES working-class men in that district who probably didn’t respond well to a female candidate publicly criticizing a male opponent. And in the third district, I think the ads you aired were just too over the top. You want to highlight your opponent’s flaws subtly, not make him sound like a cartoon villain.“
“That all sounds reasonable enough,“ said the candidate. “But I’m a bit perplexed that you didn’t mention any of these subtleties ahead of time, when they might have been more helpful.“
“Well,“ said the psychologist. “That would have been very hard to do. The theory is true in general, you see. But every situation is different.“
Situations like this are what makes mediation and moderation analysis so valuable statistical techniques.