What is a discrete choice experiment?
A DCE reveals how people make trade-offs and how much they value each attribute of a product, service, or policy.
DCEs are the standard preference elicitation method across health economics, transport planning, environmental valuation, and marketing. This article explains the method from first principles.
Knowledge Base -> Foundations -> Methods & Academic
Ben White, 07.07.2026
What is a DCE?
A discrete choice experiment is a survey-based method for measuring preferences. Respondents see a series of choice sets, each containing two or more alternatives described by a set of attributes at varying levels. They choose their preferred alternative from each set.
The pattern of choices across the experiment reveals the trade-offs respondents are willing to make - how much efficacy they will sacrifice for reduced side effects, how much they value travel time savings versus cost, or whether they prefer a policy that improves water quality over one that reduces flood risk.
Why DCEs are the gold standard
DCEs are grounded in random utility theory, developed by Daniel McFadden - Nobel Prize in Economics, 2000. The theoretical foundation distinguishes DCEs from simple rating or ranking tasks and makes the resulting estimates statistically defensible in regulatory submissions and peer-reviewed publications.
The method forces realistic trade-offs. Unlike rating scales, which allow respondents to rate everything as important, DCEs require choosing between alternatives with different combinations of attributes. This mimics real decision-making and produces preference estimates with lower hypothetical bias.
Running your first DCE in SurveyEngine
Step 1: Define the decision context. What decision will the preference estimates inform? The answer determines the attributes, levels, target population, required sample size, and model specification.
Step 2: Select attributes and levels. Typically 4-6 attributes with 2-4 levels each. Attributes must be meaningful to respondents, policy-relevant, and variable enough to drive trade-offs. Levels must span the realistic range without being implausible.
Step 3: Generate the experimental design. A D-efficient design is a good choice as it minimises parameter uncertainty for a given number of choice sets. SurveyEngine's built-in design generator handles this.
Step 4: Estimate the choice model. The multinomial logit model is the starting point. Mixed logit models account for preference heterogeneity and are standard for most applications. Apollo in R is the most widely used estimation package, which is built in to the SurveyEngine Platform.
A worked example - commuter mode choice
A government transport agency needs to estimate the value of travel time savings and reliability improvements for a rail investment appraisal. A DCE is designed with five attributes: in-vehicle travel time, frequency of delays, ticket price, walking time to station, and service frequency.
The study is conducted with 400 rail commuters. Mixed logit estimation produces value of travel time, value of reliability, and value of frequency estimates that feed directly into the cost-benefit analysis. The results are consistent with international benchmarks and are accepted by the Treasury appraisal unit.
References
Louviere, J., Hensher, D. and Swait, J. (2000). Stated Choice Methods. Cambridge University Press.
Ready to run your first discrete choice experiment? Open a free SurveyEngine account and follow the step-by-step tutorial.
Or Contact us at support@surveyengine.com — we're glad to help.