Using Classical Inference Methods to reveal individual-specific parameter estimates to avoid the potential complexities of WTP derived from population moments

Abstract

nference estimation methods for logit models with Bayesian methods and suggested that the latter are more appealing on grounds of relative simplicity in estimation and in producing individual observation parameter estimates instead of population distributions. It is argued that one particularly appealing feature of the Bayesian approach is the ability to derive individual-specific willingness to pay measures that are claimed to be less problematic than the classical approaches in terms of extreme values and signs. This paper takes a close look at this claim by deriving both population derived WTP measures and individual-specific values based on the classical ‘mixed logit’ model. We show that the population approach may undervalue the willingness to pay substantially; however individual parameters derived using conditional distributions can be obtained from classical inference methods, offering the same posterior information associated with the Bayesian view. The technique is no more difficult to apply than the Bayesian approach – indeed the individual specific estimates are a by-product of the parameter estimation process. Our results suggest that while extreme values and unexpected signs cannot be ruled out (nor can they in the Bayesian framework), the overall superiority of the Bayesian method is overstated.