SIMULATION AS A NECESSARY STEP IN THE DESIGN OF STATED PREFERENCE EXPERIMENTS

This paper focuses its attention on one of the steps suggested in the literature when a Stated Preference design is made: the simulation of responses. Although this step is broadly used and recommended, little attention has been given when reporting its results. The simulation procedure consists of adopting some reasonable coefficients related to the attributes in the utility function, that combined with the design to be tested and a random error term, permit simulation of the responses that people would give. In a second step the simulated responses and the design are used to estimate the coefficients. A good design should retrieve the set of coefficients used to carry out the simulation. Results show that both the absolute size of the recovered coefficients and the substitution rate between them depended on the size of the error term. It seems that adequate pilot surveys are far more worthwhile than simulation when testing an experimental design. (A) For the covering abstract see ITRD E107985.