When the risks are low: the impact of absolute and comparative information on disturbance and understanding in US and UK samples

Abstract

When Klein, W. M. gave participants absolute and comparative risk information (crossed experimentally) they were more disturbed by being above than below average, but not by being at higher rather than lower risk. The current experiment tests whether Klein's findings extend to situations involving lower risk figures more typical of genuine health risks, assesses participants’ understanding of the information, and directly compares responses of US and UK samples. Participants were presented with hypothetical information about comparative and absolute risks of deep vein thrombosis. There was a main effect of absolute risk information on disturbance and precaution intentions in the US sample, but no effects of comparative information on these measures in either sample. Understanding was poor among participants receiving both pieces of risk information. Future studies should include measures of understanding to establish whether people are failing to understand what they are told or failing to respond systematically to what they understand. Practically, the findings caution against providing comparative risk information when communicating low risk figures.

Keywords:

• Risk perception
• absolute risk
• comparative risk
• optimistic bias
• deep vein thrombois

Acknowledgement

We are grateful to Paul Norman and Paul Sparks for their helpful comments on an earlier draft.

Notes

See, for example, http://www.healthatoz.com/healthatoz/Atoz/default.html, under “diseases and conditions”; www.vhihealthe.com/calculators/burnrisk/burnrisk_quiz.jsp, under “calculators and quizzes”; www.aad.org/skinrisk.html.

Analysis of the combined (transformed) ratings produced similar findings: a significant absolute risk × sample interaction, F(1, 341) = 4.2, p = 0.041, and a main effect of sample that approached significance, F(1, 341) = 3.5, p = 0.061. Adding covariates of age, gender or occupation (student or non student) to the analysis of the disturbance ratings did not change the findings, suggesting that the effects of sample are not a result of these differences between samples. Indeed, in no analysis was a covariate significant. The same was true when these covariates were added to analyses of the combined ratings.

CL data were analysed using 2 (absolute risk: higher or lower) × 2 (sample: US or UK) between-participants Analysis of Variance. In the US most participants (71%) rated themselves as below average in the baseline conditions, whereas in the UK most rated themselves as average (54%). AL data were analysed using 2 (absolute risk: higher or lower) × 2 (sample: US or UK) × 2 (target: self or average other) Analysis of Variance for mixed designs.

CL data were analysed using 2 (absolute risk: higher or lower) × 2 (comparative risk: above average or below average) × 2 (sample: US or UK) between-participants Analysis of Variance. AL data were analysed using 2 (absolute risk: higher or lower) × 2 (comparative risk: above average or below average) × 2 (sample: US or UK) × 2 (target: self or average other) Analysis of Variance for mixed designs.

Of the 117 participants in the above average conditions, 44% rated their risk as below average and 32% rated their risk as average. Means were significantly below 4 in all but two cells of the design: UK, above average, higher, M = 3.9 and lower M = 3.8 (one sample t-tests).

Significantly more of the sample were inaccurate than were accurate, z = −4.5, p < 0.001. There was no association between sample and accuracy (US = 39%, UK = 40% accurate), χ²(1, N = 232)<1, p = 0.85 Analysis of the disturbance ratings used 2 (absolute risk: higher or lower) × 2 (comparative risk: above average or below average) × 2 (sample: US or UK) between-participants Analysis of Variance.

We are grateful to an anonymous reviewer for this suggestion.