Which Type of Risk Information to Use for Whom? Moderating Role of Outcome-Relevant Involvement in the Effects of Statistical and Exemplified Risk Information on Risk Perceptions

Abstract

The extant empirical research examining the effectiveness of statistical and exemplar-based health information is largely inconsistent. Under the premise that the inconsistency may be due to an unacknowledged moderator (O’Keefe, 2002), this study examined a moderating role of outcome-relevant involvement (Johnson & Eagly, 1989) in the effects of statistical and exemplified risk information on risk perception. Consistent with predictions based on elaboration likelihood model (Petty & Cacioppo, 1984), findings from an experiment (N = 237) concerning alcohol consumption risks showed that statistical risk information predicted risk perceptions of individuals with high, rather than low, involvement, while exemplified risk information predicted risk perceptions of those with low, rather than high, involvement. Moreover, statistical risk information contributed to negative attitude toward drinking via increased risk perception only for highly involved individuals, while exemplified risk information influenced the attitude through the same mechanism only for individuals with low involvement. Theoretical and practical implications for health risk communication are discussed.

Notes

¹ Additional analyses with drinkers only were conducted for all four hypotheses. The results show that the exclusion of nondrinkers did not change the findings: The interaction between statistical risk and involvement was significant and in the predicted direction (H1), F(1, 161)= 5.20, p =.02, η_p2 = .03, and that between exemplified risk and involvement was also significant and in the predicted direction (H2), F(1, 161)= 7.68, p <.01, η_p2 = .05. The results for H3 and H4 were consistent for this subgroup as well. More specifically, for H3, consistent with findings based on the full sample, the overall test of the moderated mediation effect showed a nonsignificant effect (effect = −.09, SE = .06, 95% Confidence Interval [−.21, .01]). However, the predicted indirect effect of statistical evidence on attitude via risk perception was observed among those who have high outcome-relevant involvement (indirect effect = −.07, SE = .04, 95% Confidence Interval [−.17, −.0004]), while no such indirect effect was found among those who have low involvement (indirect effect = .01, SE = .03, 95% Confidence Interval [−.05, .08]). With regard to H4, consistent with the findings based on the full sample, the moderated mediation effect of exemplified evidence was statistically significant (moderated mediation effect = .13, SE = .07, 95% Confidence Interval [.02, .28]). In addition, the predicted indirect effect of exemplified information was present when involvement was low (indirect effect = −.09, SE = .05, 95% Confidence Interval [−.21, −.01]), while no indirect effect was observed when involvement was high (indirect effect = .03, SE = .04, 95% Confidence Interval [−.03, .12]).