A while back I linked to a couple of studies in which scary public-health messages had the opposite of their intended effect: These anti-drinking and anti-smoking ads made people want to drink and smoke more. I didn't think much about the notion that a message can have "an" effect. But human beings aren't all alike, so I should have.
In this study Steffen Nestler and Boris Egloff tried two versions of a lifestyle-risk message on 229 women and 68 men. Both texts urged readers to cut down on caffeine, but one was laid back while the other contained frightening mentions of cancer and said the volunteers' age group was the one with the highest risk. I'd expect this "scared straight" message would boomerang, and make people less inclined to avoid caffeine. That did happen—but not to everyone.
Before the experiment, the researchers had given the group a personality test to find people who tend to avoid thoughts of danger (by, for example, shifting their attention to something else, or telling themselves "that can't happen to me"). As you can read here, it was only these people, the "high cognitive avoiders," on whom the "scared straight" message backfired. People without that kind of defensiveness, on the other hand, were much more likely to accept the alarming message, and vow to drink in less caffeine.
So Nestler and Egloff argue that public-health messages should be tailored to different audiences. They frame that as a point about innate personality differences, but it has also been raised in the context of social-class differences in health behavior. Perhaps "scared straight" messages make more sense to their creators (upper-middle-class people) than to the rest of society.