Specifying Future Behavior When Assessing Risk Perceptions: Implications for Measurement and Theory.

Background. Many theories assert that high perceived risk motivates health behavior change; the empirical literature shows mixed findings. Purpose. To determine whether, for whom, and under what circumstances specifying a future behavior when assessing perceived risk (i.e., "conditioning" risk perception items on behavior) improves data quality and strengthens the perceived risk-intentions/behavior relationship. Methods. Internet panel participants (N = 787, 58.8% no college experience, 44.4% racial/ethnic minority, 43.7% men, 67.3% aged 18-49 years, 59.0% nonadherent to physical activity guidelines) answered 8 colon cancer perceived risk items in a within-subjects design. Participants answered 4 types of risk perception items: absolute and comparative perceived likelihood and absolute and comparative feelings of risk. Participants answered each type of item twice: once conditioned on not engaging in physical activity and once unconditioned. Results. Compared to unconditioned items, conditioned items elicited fewer "don't know" (DK) responses (OR = 0.80; 95% CI, 0.68-0.93), higher risk perceptions (b = 0.55; 95% CI, 0.49-0.61) and stronger positive correlations with intentions (zSteiger = 5.46, P < 0.001) and behavior (zSteiger = 5.10, P < 0.001). The effect of conditioning was more pronounced for perceived likelihood than feelings of risk items (OR = 2.21; 95% CI, 1.63-3.01 and b = 0.14; 95% CI, 0.08-0.20 for DK responding and risk perception magnitude, respectively). The effect on risk perception magnitude (except absolute feelings of risk) was higher among people with higher health literacy (χ2(3) = 8.11, P = 0.04). Conclusions. Researchers who examine whether perceived risk motivates precautionary behavior should consider conditioning risk perception items on behavior to increase the validity of the statistical conclusions they draw and to gain insight into the nature of perceived risk and its relation to behavior.