Destination Choice Experiment

Observed travel flows are revealed-preference data in which destination attributes are correlated and confounded, so it is hard to isolate the effect of any single feature, and you cannot study configurations that do not exist. A choice experiment is stated-preference: the analyst designs hypothetical destinations and manipulates their attributes independently, so each feature's effect on choice can be estimated separately and causally. The cost is hypothetical bias, since respondents are not actually travelling, which is why design quality and realism matter so much.

If price is one of the attributes and enters the utility function linearly, the marginal willingness to pay for any other attribute is the negative of that attribute's coefficient divided by the price coefficient. This works because both coefficients are measured on the same utility scale, so dividing cancels the scale and leaves a money value. For example, the ratio of the climate coefficient to the price coefficient tells you how much extra travellers would pay for a better climate, all else equal.

The full factorial of all attribute-by-level combinations is usually enormous, far more than any respondent could evaluate. Orthogonal or statistically efficient fractional designs select a small subset of choice tasks that still let you estimate each attribute's effect independently and with good statistical precision. This independence is exactly what allows the later causal interpretation: because attributes were varied independently by design, their estimated effects are not confounded with one another.