Premium
Measuring Subjective Probabilities: The Effect of Response Mode on the Use of Focal Responses, Validity, and Respondents’ Evaluations
Author(s) -
Bruine de Bruin Wändi,
Carman Katherine G.
Publication year - 2018
Publication title -
risk analysis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.972
H-Index - 130
eISSN - 1539-6924
pISSN - 0272-4332
DOI - 10.1111/risa.13138
Subject(s) - psychology , scale (ratio) , response bias , external validity , mode (computer interface) , sample (material) , social psychology , criterion validity , population , applied psychology , computer assisted web interviewing , statistics , psychometrics , construct validity , developmental psychology , computer science , medicine , environmental health , geography , mathematics , chemistry , cartography , chromatography , operating system
Subjective probabilities are central to risk assessment, decision making, and risk communication efforts. Surveys measuring probability judgments have traditionally used open‐ended response modes, asking participants to generate a response between 0% and 100%. A typical finding is the seemingly excessive use of 50%, perhaps as an expression of “I don't know.” In an online survey with a nationally representative sample of the Dutch population, we examined the effect of response modes on the use of 50% and other focal responses, predictive validity, and respondents’ survey evaluations. Respondents assessed the probability of dying, getting the flu, and experiencing other health‐related events. They were randomly assigned to a traditional open‐ended response mode, a visual linear scale ranging from 0% to 100%, or a version of that visual linear scale on which a magnifier emerged after clicking on it. We found that, compared to the open‐ended response mode, the visual linear and magnifier scale each reduced the use of 50%, 0%, and 100% responses, especially among respondents with low numeracy. Responses given with each response mode were valid, in terms of significant correlations with health behavior and outcomes. Where differences emerged, the visual scales seemed to have slightly better validity than the open‐ended response mode. Both high‐numerate and low‐numerate respondents’ evaluations of the surveys were highest for the visual linear scale. Our results have implications for subjective probability elicitation and survey design.