McNeil River State Game Sanctuary permit lottery applicant preferences and marginal willingness to pay for permit application: a best-worst discrete choice experiment

Abstract

This study applies data from a web-based survey administered to 2016-2018 McNeil River State Game Sanctuary permit lottery applicants to examine preferences and marginal willingness-to-pay (WTP) for an application contingent upon marginal interval increases among specific attributes of a bear viewing experience. A best-worst discrete choice experiment (BWDCE) was used to elicit respondent data, which consisted of eight individual choice tasks using a Balanced Incomplete Block Design (BIBD) in Sawtooth Software. Each choice task was comprised of five attributes: permit application price, odds of winning a permit, number of bears viewed daily during visit to the Sanctuary, cubs being present, and most common type of bear feeding activity viewed while at the Sanctuary. Each attribute was decomposed into two to four varying levels across choice tasks, depending on the attribute in question. The findings suggest that lottery permit applicants have a significant desire to view bears fishing for salmon, and to see cubs. These results imply a clear desire of applicants to visit the Sanctuary in high season. As expected, respondents also stand to obtain a positive effect on personal utility of increased odds of winning a permit, and to a lesser extent, view a larger number of bears while at the Sanctuary, and therefore have a positive mWTP to both of these characteristics as well. The price coefficient in both the preference parameter utility model and the mWTP model is negative, as expected, but not large in magnitude which may be attributed to the sample being wealthier than average and/or the forgone permit application price is viewed as a wildlife conservation donation. The main model used for analysis is the mixed (random parameters) logit (MXL), and the preference parameters estimated are then used to estimate mWTP in WTP-space using Stata. Results using a multinomial logit (MNL) and conditional logit (CL) are also presented for comparison and affirmation that MXL is better suited for the data in order to allow for preference heterogeneity and random parameters, rather than fixed parameters.