Willingness-to-pay for snowmobile recreation: travel cost method models with and without post-season resurvey of trip count

ABSTRACT

Annual willingness-to-pay (WTP) for snowmobile recreation in Idaho is estimated using the travel cost method (TCM). On-site snow conditions are important and erratic, thus we collect two measures of the annual trip count, an in-season survey of the expected count, and a post-season survey of the actual count. Two variants of the TCM model are estimated. Using the post-season actual trip count data and the ‘traditional’ TCM model, WTP increases from $41 to $91 per person per trip as the fraction of the wage rate that is used to value the opportunity cost of travel time is increased from 1/4 to one. Using the preferred short-run decision TCM model, WTP increases from $53 to $194 as the snowmobiler ratings of off-trail snow conditions vary from worst to best. WTP estimates using the in-season expected trip count data and the traditional TCM model are much higher (triple) than those found using the post-season actual trip count data, and the confidence intervals are much larger.

KEYWORDS:

• Count data
• off-trail snow conditions
• post-season resurvey
• short-run decision model
• snowmobiling

JEL CLASSIFICATION:

• Q26
• H41

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ An additional assumption universally required by the travel cost method model specification is that travel does not enter the utility function in a positive or negative fashion (Walsh, Sanders, and McKean 1990).

² The demand functions (and past participation estimates) show that both participation and the value of snowmobiling vary directly with the quality of the snow. Snowfall in Valley County is erratic. In the year of this study, snowfall was 81 inches which is 64% of the twenty-year average. In the prior year, snowfall was 129 inches, but two years prior, snowfall was only 44 inches. Snowfall in 2014 was 59 inches.

³ The Akaike Information Criterion with small sample adjustment is calculated as AIC = 2[(m – log likelihood) + (m)(m + 1)/(n – m – 1)], and the Bayesian Information Criterion is calculated as BIC = −2 log likelihood + (m)[ln(n)], where n is sample size and m is the number of parameters in the model. Smaller values for AIC and BIC indicate a preferred model.