Risk Ratings that Do Not Measure Probabilities

Survey questions regarding assessed survival chances are an often-used example of a risk rating scale for eliciting a probability assessment. The responses to such questions do exhibit several properties of probabilities, but differ in some key respects, resulting in relationships which are not only inconsistent with accurate beliefs, but also in which precision is sacrificed for ease of use. The Health and Retirement Study, for example, uses a 0 to 10 scale to measure self-assessed survival probabilities to a particular age. Transformation of these responses for use as a probability results in some patterns that are consistent with a model of imperfect information, or a monotonic transformation of imperfectly perceived risks, but more subtle analysis reveals inconsistencies with either of these theories, suggesting the scale is inappropriate for use as a probability measure. The age-related effects for female respondents are the most salient results that are inconsistent with use of the survey's response scale as representing a probability.

Subjective expectations/Retirement/Smoking/Health

11962