Evaluating the Simplified Conjoint Expected Risk Model: Comparing the Use of Objective and Subjective Information

Abstract

The simplified Conjoint Expected Risk (CER) model by Holtgrave and Weber posits that perceived risk is a linear combination of the subjective judgments of the probabilities of harm, benefit, and status quo, and the expected harm and benefit of an activity. It modifies Luce and Weber's original CER model--that uses objective information to evaluate financial gambles--to accommodate activities such as health/technology activities where values of the model variables are subjective. If the simplified model is a valid modification of the original model, its performance should not be sensitive to the use of subjective information. However, because people may evaluate information differently when objective information is provided to them than when they generate information on their own, the performance of the simplified CER model may not be robust to the source of model-variable information. We compared the use of objective and subjective information, and results indicate that the estimates of the simplified CER model parameters and the proportion of variance in risk judgments accounted for by the model are similar under these two conditions. Thus, the simplified CER model is viable with activities for which harm and benefit information is subjective.

A classification model for continuous responses: Identifying risk perception groups on health‐related activities

“…This risk perception data has been used in Carlstrom et al. (2000) and Palmer et al. (2001) but with different analysis purposes.…”

“…The risk perception data from Carlstrom et al. (2000) is used to illustrate the advantages of the C‐DINA model for identifying groups by type of risk (attribute profiles), considering two latent attributes: internal attributes (skills) (if depends on the person) and external attributes (skills) (if depends on the environment) and then using these attributes to specify a Q‐matrix for the risk perception data. It can be a limitation on the results of the application.…”

“…We consider that is possible to use a new approach for the risk perception study, since in both types of scales with a discrete and continuous response, the risk perception is regarded as a subjective judgment of a person and thus, it is a latent variable, required to estimate from the outcomes of these questionnaires. In a different way than IRT models, where latent traits are continuous and responses are discrete, or in common cognitive diagnosis models (CDMs) (von Davier & Lee, 2019) where latent trait and response are discrete, in our approach to study the risk perception, considering data from Carlstrom et al (2000), we assume that latent traits are discrete and responses are continuous.…”

“…The other approach for study risk perception is using a continuous based scale value in [0, 100] and then, based on observed empirical regularities in risk judgments, different axiomatic measures of perceived risk are used, and, for example, the conjoint expected risk (CER) model suggested by Luce and Weber (1986) and developed for Holtgrave and Weber (1993) can be considered. Specifically, we consider data from Carlstrom et al (2000) and Palmer et al (2001) where participants complete two forms of a simplified CER questionnaire in which some health and financial activities are described and then the respondents give an overall judgment of their risk perception for each activity. All judgments are made on a scale from 0 to 100 and are assumed with continuous support.…”

“…Specifically, we consider data from Carlstrom et al. (2000) and Palmer et al. (2001) where participants complete two forms of a simplified CER questionnaire in which some health and financial activities are described and then the respondents give an overall judgment of their risk perception for each activity.…”