Gentner: Sensitivity of angler benefit estimates to the definition of substitute sites considered by tfie angler 



165 



looks at the aggregation of sites. On average there are 

 almost 44 MRFSS intercept sites aggregated into the 

 definition of a "site" used in the present study and a 

 maximum of 138 sites and a minimum of five sites in a 

 county. On average, anglers spend 4.2 hours on the wa- 

 ter and 73% own the boat they are fishing from. Finally, 

 this is a fairly experienced group, with an average of 

 almost 24 years of saltwater fishing experience. 



The final portion of this analysis yet to be discussed 

 is the definition of distance-based choice sets. Haab 

 and Hicks (1997) and Parsons and Hauber (1998) found 

 that welfare estimates change little beyond a certain 

 threshold. Up to a point, limiting an angler's choice set 

 by using a distance-based metric only increases the 

 realism of the choice that anglers consider in reality. 

 Hicks et al. (1999) used such a designation in their 

 analyses. Both included all sites within 150 miles, if the 

 angler lived within 30 miles of the site selected, and all 

 sites within 400 miles otherwise. The extreme rational 

 limit for a one-day trip is probably 400 miles one-way. 

 That distance translates into a 10 hour or 6 hour and 

 40 minute one-way travel time at 40 and 60 miles per 

 hour, respectively. Whitehead and Haab (1999) used 

 definitions of distance-based choice sets that ranged 

 from 180 miles one-way to 360 miles one-way (3-6 

 hours one-way at 60 miles per hour), realizing that the 

 360 mile cut-off is likely not very realistic. These defini- 

 tions probably drive the small difference in parameters 

 across specifications, because eliminating sites outside 

 of what anglers are really considering should have little 

 effect (Whitehead and Haab, 1999). 



To examine the definition of choice sets, this study 

 examined much smaller cut-offs than those found in the 

 previous literature that focused on saltwater angling. 

 The cut-offs included the following: a full unrestricted 

 choice set, and 300-, 250-, 200-, 150-, and 100-mile dis- 

 tance cut-offs. Initially, a 50-mile one-way distance cut- 

 off was included, but, because of the site aggregation 



strategy used, the only substitutes left in the choice set 

 at the 50-mile cut-off were closer than the site that was 

 chosen, for most anglers. If it is possible to estimate the 

 model on an individual site basis, it would be possible to 

 run smaller distance cutoffs without encountering this 

 problem. In order not to also drop observations when 

 applying these cut-offs, if an individual was observed 

 to make a choice outside of the cut-off, that observation 

 is retained. If all substitute sites for that individual 

 are also outside of the cut-off, the next nearest site is 

 included in that angler's choice set. Therefore, anglers 

 have at least one substitute left in their choice set, no 

 matter how restrictive the cut-off becomes. The aver- 

 age number of sites in each choice set is 63, 37.2, 30.5, 

 23.3, 16.9, and 10.6 for the six choice sets, from least 

 restrictive to most restrictive, respectively. In percent- 

 age terms, these restrictions on the choice set eliminate 

 between 409f and 937c of the available sites. 



Estimation of the confidence intervals around these 

 welfare estimates is calculated by taking 1000 random 

 draws from a multivariate normal distribution parame- 

 terized by the vector of estimated parameters and their 

 covariance matrix. Sorting these draws from highest 

 to lowest and removing the upper and lower 2.5% and 

 59c, respectively, construct 95% and 90% confidence 

 intervals (Krinsky and Robb, 1986). 



Results 



Table 2 contains the parameter estimates and standard 

 errors of the six conditional logit models. All six models 

 strongly reject the hypothesis that the coefficients are 

 simultaneously equal to zero. Also, all coefficients in all 

 models are statistically significant at the 99% level or 

 better. In general, anglers prefer sites that are closer 

 to home, both in terms of the cost of driving and the 

 time cost (travel time multiplied by the individual's 



