An Information-Theoretic Approach to Modeling Binary Choices: Estimating Willingness to Pay for Recreation Site Attributes

This paper applies the new maximum likelihood-minimum power divergence (ML-MPD) binary response estimator developed by Mittelhammer and Judge (2011) to model the underlying behavioral decision process that leads to a person’s willingness to pay for recreation site attributes. Empirical choice probabilities, willingness to pay (WTP) measures for recreation site attributes, and marginal probability effects of decision-maker characteristics are estimated based on a real stated-preference on-site contingent valuation data, collected at the Caribbean National Forest in Puerto Rico. For comparison purposes, the linear probit model and the Kriström/Ayer’s estimators are implemented. The ML-MPD method yields a significantly lower mean WTP estimate ($27.80) to attend the recreation sites compared to WTP measures obtained from the fully parametric ($120) and fully non-parametric ($97) approaches. We argue, based on the decision context and demographics of decision makers visiting the recreation sites, that the ML-MPD approach suggests a more defensible representation of the underlying data-generating process and economic decision-making behavior. As such, the ML-MPD estimator suggests future potential for improved econometric analyses of discrete behavioral decision choice processes.