Reporting willingness to pay estimates from random parameters models

Abstract

Value estimates such as willingness to pay and their confidence intervals are commonly calculated from estimated random-parameters logit models. That parameters in these models have a distribution within the population as well as a sampling distribution poses a challenge for how to best help an audience understand variation in willingness to pay. We detail this issue, discuss the ways in which willingness to pay and related distributional information have been conveyed in the literature, and review and propose several ways to meaningfully but relatively succinctly convey both sources of variation to audiences.