Research

"Fear of Crime"-Survey Konstanz 2002

Context Effects in Measuring Subjective Probabilities

Subjective probabilities are of great interest for many survey researchers. Measuring those subjective probabilities is problematic because there are predictable biases in the way people make such estimates, especially when they do so in percent terms. Much evidence exists that the use of a relative-frequency scale promotes more accurate probabilistic reasoning. However, phrasing questions on the subjective single-event probabilities in which survey researchers are often interested in frequency terms is difficult. We therefore test the possibility that survey respondents can be taught to apply frequentistic reasoning to their estimates of single-event probabilities in percent terms through a prior frequency-judgment task. Specifically, we examine the effect of several kinds of base rate estimations on a later single-event probability judgment in the context of a mail-survey experiment. The results, obtained in a mail survey of approximately 1,100 residents of a German city on their fear of crime, indicate that such a reasoning transfer can occur.

Publications

Elisabeth Coutts, Rainer Schnell (2005): Context effects in the measurement of subjective probabilities in surveys, Paper presented at the first international conference of the European Survey Research Association, Barcelona, July 2005.

Elisabeth Coutts, Rainer Schnell (2000). „Not very likely… I’d say 70%: Learning effects in the estimation of subjective probabilities". Paper presented at the Fifth International Conference on Logic and Methodology, Cologne, Oktober 2000

Elisabeth Coutts-Heller (20002): Context effects in the measurement of subjective probabilities in surveys, submitted for a diploma, University of Konstanz.