Quanto ci si può fidare delle stime ottenute con i Choice Models?

In this study we compare two specifications of logit models for the analysis of discrete choices from the same sample. The first specification is a conditional logit model, while the second is a more flexible panel model with random parameters that does not imply the restrictive assumption of independence of irrelevant alternatives and allows the researcher to derive individual estimates of WTP. The estimates of the two models appear different with different interpretative approach. This gives us the opportunity to examine in some depth the consequences of various hypotheses on the final estimation results of similar analysis.