Loss aversion, reference dependence and diminishing sensitivity in choice experiments
Introduction
The design of policies and interventions to influence decisions requires an understanding of how decisions are made. For example, attempting to move people from their current set of choices to a new set of choices depends, according to standard economic theory, on the changes in costs and benefits. Economic theory assumes an attribute of a good or service is valued independently of context and that individuals have well-formed preferences (Hess et al., 2018; Shiell et al., 2000). Thus, choice tasks should generate attribute rankings that are consistent regardless of how they were elicited, but this is not observed in practice. Prospect theory (Kahneman and Tversky, 1979) recognises that reference points influence choices, and movements away from the reference point are evaluated in terms of gains and losses relative to this point. In a series of experiments, individuals were found to be more sensitive to losses than to gains when evaluating their options, and were generally risk-seeking when outcomes were losses, and risk-averse when outcomes were gains. This implies an s-shaped value function, where losses lead to larger reductions in value than equivalent sized gains. Using prospect theory, Thaler (1980) introduced the concept of the endowment effect that also helped explain the importance of the difference in value between losses and gains. If individuals view losses as out-of-pocket costs, and gains as forgone opportunity costs, then a good taken away from an individual's endowment will be weighted more heavily than the same good added to their existing endowment. Samuelson and Zeckhauser (1988) discuss status quo bias: decision makers choose the alternative that was their current situation more frequently than predicted by a rational choice model. They cite a range of examples from observations and experiments that support that status quo biases are ubiquitous and significant, and that the bias increases with the number of choice alternatives. The design of their experiments allows them to conclude that status quo bias is generally present, not just as a result of loss aversion (Samuelson and Zeckhauser, 1988). De Borger and Fosgerau (De Borger and Fosgerau, 2008) introduced a theoretical framework that maps choice pairs to different welfare measures and used this to compare four commonly used valuation measures that represent gains and losses in two dimensions from a reference point. They find loss aversion, reference dependence and diminishing sensitivity in an empirical application using data from a survey of binary choices where respondents trade off money and time. They also find that “mistakes”, defined as choosing a dominated alternative, are much more common when the reference scenario is dominated. In other words, respondents are more likely to choose their status quo, even if it is a dominated scenario.
The aim of this paper is to examine the effect of the status quo in the context of a discrete choice experiment (DCE). We present an application that integrates a scenario describing each respondent's current situation (their status quo) consisting largely of qualitative attributes into a DCE. We use data from a DCE of job preferences among nurses, where respondents were asked to choose their preferred alternative from a set of hypothetical scenarios, then asked to choose again from the same hypothetical scenarios and their current situation. The ‘current situation’ is an additional alternative reconstructed for each respondent from answers to questions that were phrased with the same description of attributes and levels used in the DCE, which offers a simple way to integrate any reference scenario into discrete choice experiments without loss of data.
Section snippets
Background and related literature
A status quo/opt-out is increasingly being included in DCEs (Mandeville et al., 2014; Lancsar and Louviere, 2008). It is necessary for ‘realism’ by recognising that individuals can postpone choice or decide not to choose at all, and for predictions of demand and market shares (Pedersen et al., 2011; Scarpa et al., 2005). In addition, a ‘status quo’ alternative is required for welfare measurement that examines welfare changes compared to the current state of the world. Previous studies in health
DCE design and data
The data come from a DCE examining job preferences of nurses. The full methods of the nurses’ DCE are in Scott et al. (2015). Briefly, the DCE was completed by 990 nurses based in Victoria, Australia in 2008. Attributes included earnings, hours worked, public or private sector employment, autonomy, shift type, processes to deal with violence and bullying, and patient to nurse ratio (Table 1). SAS software was used to create an orthogonal, balanced, 100% efficient design, and choice pairs were
Attribute and choice descriptive statistics
Table 2 shows the distribution of gains or losses (of any size) relative to the status quo for each attribute, where gains indicate that the hypothetical job is better. For five out of seven attributes, the distribution is symmetric with roughly equal proportions of gains and losses appearing in the experiment. The distribution is skewed towards losses for autonomy, because only a small percentage of respondents (11%) report a “poor” level of autonomy in their current job. Therefore, it is less
Discussion
When presented with a choice, people prefer what they know or own even if the attributes of the new alternative are better (Kahneman and Tversky, 1979; Thaler, 1980; Samuelson and Zeckhauser, 1988; Tversky and Kahneman, 1991). This has been found across many decision-making contexts in experimental and non-experimental studies and is cited as the reason for the empirical difference between measures of willingness to pay and willingness to accept.
We tested for the effect of an explicit status
Author contributions
Anthony Scott: Conceptualization; Methodology; Formal Analysis; Writing – Original Draft and Review & Editing; Funding acquisition.
Julia Witt: Methodology; Software; Formal analysis; Investigation; Writing – Original Draft and Review & Editing
Funding
This research was funded by an Australian Research Council (ARC) Linkage Grant (LP0669209) and the Department of Health, State Government of Victoria (Victoria).
Declaration of competing interest
None.
References (38)
- et al.
Simulation error in maximum likelihood estimation of discrete choice models
J. Choice Model.
(2019) - et al.
Can healthcare choice be predicted using stated preference data?
Soc. Sci. Med.
(2020) - et al.
The trade-off between money and travel time: a test of the theory of reference-dependent preferences
J. Urban Econ.
(2008) - et al.
Correlation and scale in mixed logit models
J. Choice Model.
(2017) - et al.
Asymmetric preference formation in willingness to pay estimates in discrete choice models
Transport. Res. E Logist. Transport. Rev.
(2008) - et al.
A framework for capturing heterogeneity, heteroskedasticity, non-linearity, reference dependence and design artefacts in value of time research
Transport. Res. Part B
(2017) - et al.
Analyzing loss aversion and diminishing sensitivity in a freight transport stated choice experiment
Transport. Res. Pol. Pract.
(2010) - et al.
Does the includsion of a cost attribute in forced and unforced choices matter? Results from a web survey applying the discrete choice experiment
J. Choice Model.
(2011) Eliciting GP's preferences for pecuniary and non-pecuniary job characteristics
J. Health Econ.
(2001)- et al.
Eliciting preferences of the community for out of hours care provided by general practitioners: a stated preference discrete choice experiment
Soc. Sci. Med.
(2003)