166 



Fishery Bulletin 105(2) 



wage rate). Additionally, anglers prefer sites that offer 

 the possibility of catching and keeping more striped 

 bass. Finally, with regard to the number of MRFSS 

 sites aggregated into a site as defined in this study, 

 anglers preferred to visit counties that contain more 

 sites. Table 3 gives the mean CV for a one-fish increase 

 in the catch-and-keep rate. A one-fish increase in the 

 catch-and-keep rate is equivalent to the net benefits 

 of an improvement in angling quality large enough 

 to increase the keep rate by one fish or a regulation 

 that allows increasing keep rates. A quality change 

 significant enough to change KRATE by one fish would 

 be unrealistic in the short term, considering the striped 

 bass stock size distribution inferred from the catch 

 rate and KRATE estimates (Table 1). Because KRATE 

 incorporates the five-year average probability of catch- 

 ing a striped bass large enough to keep, this one-fish 

 increase in KRATE models an angler's willingness to 

 pay for a one-fish increase in the bag limit or an angler's 

 willingness to pay for a special license allowing the 

 retention of one striped bass more than the current 

 two-fish limit. 



This result supports Parson and Hauber's (1998) 

 and Whitehead and Haab's (1999) results that there 

 is indeed little difference in the definition of choice 

 sets with the use of a distance metric. To examine the 

 significance of the difference in welfare estimates, and 

 not just the magnitude, 95% confidence intervals were 

 calculated around each welfare measure (Krinsky and 

 Robb, 19861. In fact, the mean of the smallest choice set 

 is almost entirely contained within the 95% confidence 

 interval of the next smallest choice set, and the entire 

 lower bound for the smallest choice set is contained 

 in the next smallest choice set. This is demonstrated 

 graphically in Figure 1. From Figure 1, however, it 



appears that as the choice sets are truncated past the 

 150-mile threshold, welfare estimates rise — a similar 

 result to that of Parson and Hauber (1998). Unfortu- 

 nately, the aggregation strategy necessary when using 

 the MRFSS data precludes an examination of a dis- 

 tance-based cut-off as small as that used in Parsons 

 and Hauber's study (1998). 



Conclusions 



In general, as choice sets are restricted, the coefficient on 

 cost goes up, its absolute value goes down, and its stan- 

 dard error goes up, but only slightly, until the point is 

 reached where the aggregation strategy begins to impose 

 an artificial restriction on the choice set with this data 



