Conjoint analysis. The cost variable: an Achilles' heel?

Abstract

This paper seeks to enlighten the readers on the potential complexities involved in including cost variables in conjoint analysis, with the aim of emphasising that interpretation of implicit WTP values should be tackled with caution. To illustrate the potential pitfalls, a large data set from a recent Danish study is applied. The data consists of 1991 interviews in which participants are required to perform three discrete choice tasks regarding choice of hospitals, and three choice tasks involving health-care systems in general. Model comparisons are performed which test the effect of (1) the cost range applied and (2) the effect of including a dummy variable to represent the utility associated with payment per se. A wider cost range including higher payments is associated with lower parameter weights associated with the payment variable, and thus increased WTP values. Including a dummy variable to explain utility associated with payment per se has significant effects on the model incurring some of the other variables to become insignificant, and others to change sign. Results suggest that inclusion of a two-dimensional structure to explain the relationship between cost and utility may avoid erroneous conclusions and give rise to significant changes in implicit WTP estimates.