Are models and respondents talking the same language: evidence from stated and inferred discontinuous preferences in a choice experiment valuing public goods?

In the choice experiment framework, it is assumed that respondents consider all the attributes when making their choices. However, there is evidence that respondents may not consider all the attributes. This study has proved that in our Choice Experiment survey evaluating public goods, there is significant correlation between the stated preferences declared in a continuous scale by the respondents and the parameter estimates inferred from the models. The correlation has been tested for the coefficient of variation estimated in a Random Parameter Logit and the probability of ignoring derived from a 2K latent class model