whether they produce differential rates of change for the disadvantaged population. For example, a focused campaign to reach Native American populations to stimulate active care seeking for diabetes might begin operating soon after a national campaign with the same objective. Is there evidence that the focused campaign produces higher rates of care seeking among Native Americans than the national campaign, particularly in the context of relative costs per person reached?

4. Comparison across campaigns that differ in diversity strategy, but also differ in other important aspects, such as in their behavioral foci (tobacco versus drugs; blood pressure versus cholesterol). This evidence would be useful, but would depend on offering a credible argument that the other differences between the comparison campaigns were not so large as to confound the diversity differences. The credibility of claims based on this evidence would be greater if it was based on multiple comparisons rather than on just one comparison.

The discussion about theory (Chapter 2) describes the basis for justifying the use of different message strategies for different subgroups. It argues that different strategies were justified when there was evidence that for specified groups, different sets of beliefs were predictive of their behavior. For example, for the youth antitobacco campaign, one subgroup’s discussion about beginning to smoke might be related to its belief that a person’s athletic endurance would be damaged by smoking. Another group might exhibit a stronger association of concerns about parental disapproval of their child starting to smoke. A campaign might develop messages that embodied each of those ideas, then they might be tested with members of both subgroups. The test could be done in a constrained way: asking subgroup members to evaluate the ads and