Stated preference vs revealed preference in transport research
What people say vs. what they do.
Stated preference (SP) data captures hypothetical choices, revealed preference (RP) data captures actual behaviour. Both are necessary for robust transport modelling.
The differences between stated and revealed preference data in transport research, when each is appropriate, and how they are combined in joint RP/SP models.
Knowledge Base -> Foundations -> Transport
Ben White, 07.07.2026
The limits of observed behaviour
Revealed preference data is observed behaviour - mode choices, route choices, departure time decisions recorded from travel diaries, smart card data, or GPS tracking. It has the advantage of representing real decisions with real consequences, which eliminates hypothetical bias.
Its limitation is that it only covers options that actually exist and vary in the real world. You cannot estimate the value of a new rail line from RP data if no rail line exists yet. Nor can you estimate the response to a pricing policy that has never been implemented.
Why RP/SP combination is the gold standard
SP data overcomes the constraint of observed variation. By designing choice experiments with hypothetical, yet realistic alternatives, researchers can estimate preferences for new modes, new pricing structures, or policy scenarios that have not yet been implemented. This is essential for forecasting demand for new infrastructure.
The disadvantage is hypothetical bias - respondents may not behave in reality as they say they would in a survey. Joint RP/SP models address this by anchoring SP estimates to observed RP behaviour, using the RP data to scale and validate the SP estimates.
Setting up RP/SP data collection in SurveyEngine
Step 1: Identify what RP data is available. Smart card data, travel diaries, GPS traces, and mode choice surveys are all potential RP data sources. The more detailed the RP data, the more precisely the SP scale parameter can be estimated.
Step 2: Design the SP experiment around the RP context. The SP alternatives should be grounded in the respondent's actual travel context - use their reported current journey as the reference point for the SP pivot design.
Step 3: Collect both data sources from the same respondents. Joint estimation requires matched RP and SP data. The simplest approach is to collect RP information at the beginning of the same survey that contains the SP experiment.
Step 4: Estimate a joint RP/SP model. The joint model estimates a single set of preference parameters but allows different scale parameters for the RP and SP data, accounting for the difference in error variance between real and hypothetical choices.
Worked example - urban mode choice with RP/SP combination
A city transport authority needs to forecast mode shift to a proposed new light rail line. No comparable light rail exists in the city, so RP data cannot be used alone. A SP experiment is designed with the light rail as a new alternative, with attributes set at their proposed values.
The survey collects respondents' current commute mode, journey time, cost, and frequency of delays as RP data, followed by a SP experiment presenting the light rail alongside the current modes. Joint RP/SP estimation using Apollo in SurveyEngine produces preference parameters anchored to observed behaviour. Mode split forecasts are produced for three service frequency and pricing scenarios.
References
Hensher, D.A., Rose, J.M. and Greene, W.H. (2015). Applied Choice Analysis, 2nd edition. Cambridge University Press.
Louviere, J., Hensher, D. and Swait, J. (2000). Stated Choice Methods. Cambridge University Press.
Ready to set up an RP/SP combination study? Log in to SurveyEngine and configure your pivot design.
Or Contact us at support@surveyengine.com — we're glad to help.