Loss Aversion, Reference Dependence and Diminishing Sensitivity in Choice Experiments

Melbourne Institute Working Paper No. 16/15

Date: 2015

Author(s):

Anthony Scott
Julia Witt

Abstract

In the discrete choice experiment literature, it has been argued that the choice sets from which respondents choose should include an unforced choice because this is more realistic and accounts for status quo bias. However, we propose a much stronger set of arguments for preferring to use unforced choices where relevant. These relate to the concepts of loss aversion, reference dependence and diminishing sensitivity from prospect theory. We use data from a discrete choice experiment of different types of jobs for nurses, where the introduction of a third alternative, representing the respondent’s current job, changes the reference point, which is different for each respondent. The increased salience of the reference point, in turn, changes the size of any losses or gains when comparing Job A or Job B with their current situation, and since losses are valued more than gains, this affects the marginal utility of each attribute. This has implications for policy conclusions based on willingness to pay. Including an unforced choice is necessary (when appropriate) not only for the purposes of ‘realism’, but also because different marginal utilities are produced due to loss aversion, reference dependence and diminishing sensitivity.