On the use of flexible mixing distributions in WTP space: an induced value choice experiment

Abstract

In this study, we use data from an induced value choice experiment to compare estimates from mixed logit models in willingness to pay (WTP) space using different parameter distributional assumptions. Specifically, we test differences in WTP estimates when using flexible parameter mixing distributions (i.e. Legendre polynomials, step functions and splines) and conventional parameter distributions (normal and lognormal). Similar WTP estimates are obtained. However, we observe that WTP estimates are statistically different from the induced value when conventional distributions are assumed, but they are not when more flexible distributions are assumed. This suggests that flexible distributions can provide more reliable WTP estimates.