Utility Survey Techniques

DCE (Discrete Choice Experiment)


1. Description

DCE (Discrete Choice Experiment) uses exactly the same principles as Conjoint Analysis (CA) with a more structured guideline to generating the hypothetical scenarios to be used in the elicitation process.[1][2][3][4] DCE can be regarded a framework for eliciting utilities from relevant stakeholders with roots in the random utility theory and a strong foundation in behavioural psychology. In DCE the most important characteristics of a situation are defined and labelled as attributes. Then, each attribute is assigned levels which can be cardinal, ordinal, or categorical. The attributes and levels are then systematically varied to explore all potential configurations of attributes. These are later reduced via fractional factorial designs, where the optimal design would be orthogonal. This results in hypothetical situations, which are then compiled into choice sets that contain two or more hypothetical scenarios. Stakeholders will select the most attractive scenario from the choice set, and it is assumed their selection has the highest utility out of the options provided. From this, it is possible to analyse the value each attribute via logistic regression.

2. Evaluation

2.1 Principle
  • DCE is a stated preference method.
  • DCE uses conjoint analysis technique and has foundations in the random utility theory and statistical experimental design.
  • The choice behaviour elicited under hypothetical circumstances may not truly reflect stakeholder behaviour in real life situations.
  • The description of a hypothetical scenario is reduced to a specific number of attributes, which may not accurately represent real life situations and decision-making because these attributes may not be the only driving criteria for a particular problem.
  • Another drawback is that there have been questions raised over its internal validity, consistency and test-retest reliability. [4]
  • There is a high degree of transparency and all the steps involved in the process are disclosed.
  • DCE incorporates stakeholders' value judgements on each attribute or criterion.
  • The practicality of its application is limited because respondents may experience fatigue, thus affecting to what extent the number of choice sets can be completed, and the number of scenarios presented within each choice set.
  • A compromise between being overly comprehensive and being too specific is usually necessary.

2.2 Features
  • The attributes in DCE represent benefits and risks, and so it is capable of handling multiple benefits and risks simultaneously.
  • DCE can incorporate time dimension as one of the attributes.
  • The coefficient for each attribute is tested for statistical significance and it is possible to calculate marginal rates of substitution between a selected attribute and other attributes.

2.3 Visualisation
  • There is no commonly used visual representation of the method.

2.4 Assessability and accessibility
  • The results are not easily interpretable from the perspective of a non-statistician or those unfamiliar with the concept of experimental designs.
  • Some real life situations may not be accurately reduced and represented by a set of attributes.
  • DCE can be used to investigate how specific attributes may be viewed differently by different stakeholders.

This method was tested in the Rimonabant case study.

3. References

[1] Ryan M, Gerard K, Amaya-Amaya M. Using Discrete Choice Experiments to Value Health and Health Care. Dordrecht, The Netherlands: Springer; 2008.
[2] Ryan M, Hughes J. Using Conjoint Analysis to Assess Women's Preferences for Miscarriage Management. Health Econ 1997;6(3):261-73.
[3] Ryan M, Bate A, Eastmond CJ, Ludbrook A. Use of discrete choice experiments to elicit preferences. Quality in Health Care 2001 Sep;10:I55-I60.
[4] Drummond MF, Sculpher M, Torrance G, O'Brien B, Stoddart G. Methods for the Economic Evaluation of Health Care Programmes. 3 ed. Oxford: Oxford University Press; 2005.