Impact of violations of model assumptions: the role of lexicography, inconsistent choices and changing preferences in model estimation

摘要

Stated choice experiments are commonly used in a range of disciplines, including transport,to elicit responses about the preference behaviour of a sample population. Whilst there aremany econometric models that may be used to analyse such stated preference behaviour, theyare all based on specific assumptions made by the analyst in relation to how individualrespondents make their choices. The two main assumptions are that respondents are bothconsistent (they do not change their behaviour whilst completing the survey) and also full9traders (their behaviour is based on trading off between the full set of alternatives). Inpractice however, it may be difficult to satisfy these assumptions. Accepting that a share ofrespondents may exhibit alternative behaviour, the question becomes what impact such ‘non12conformist’ behaviour may have on estimated models. In an attempt to answer this question,the present paper presents an in our opinion unprecedented in-depth analysis using simulateddata to study the impact of such behaviour on model estimates.