standard errors for the medians and truncated means were 

 estimated by bootstrapping with 200 replications. This 

 statistical procedure is described in Duffield and Patterson 

 (1991). Bootstrapping was used because there is no analytic 

 formula for a standard error of the truncated mean. Although an 

 approximate asymptotic formula exists for the standard error for 

 the median, bootstrapping will probably give a more reliable 

 estimate, particularly for small samples. 



Four specific aggregation schemes are presented in Tables 14-16. 

 These different aggregations test the a priori assumptions that 

 the following differences exist: 



(1) That residents and nonresidents exhibit different 

 current trip net economic values, 



(2) That trips to the Central and Pacific Flyways will 

 exhibit differing values, 



(3) That early and late season (first and last) trips will 

 show differing net economic values, 



(4) That trips with hypothetically changed conditions (Twice 

 the birds and more liberal regulations, or half the birds 

 and more conservative regulations) will have net economic 

 values which differ from the current trip values. 



Table 14 shows the estimated values for hunters current trip (an 

 actual specific trip which the hunters made during the season) 

 for the entire sample as well as for resident and nonresident 

 subsamples. As would be expected, nonresident values per trip 

 (329.48) were significantly higher than resident values (126.21). 

 When adjusted for the different average length of trips taken by 

 the two groups nonresident values remained higher at $100.15 per 

 day vs. $78.88 per day for residents. 



One interesting figure shown in Table 14 is the median trip value 

 and standard error for the nonresident subsample. While measures 

 of the median of CVM distributions tend to be significantly lower 

 than measures of mean willingness to pay, this particular model 

 shows just the opposite. The reason for this anomaly is that 

 even at the highest bid level asked ($500) nearly 50% of the 

 nonresidents responded "yes" as to their willingness to pay that 

 amount. The effect of this is that the Truncated mean WTP was 

 extremely conservatively estimated while the median was 

 accurately estimated. Since the variance of the median estimate 

 is based in a large part upon the unobserved upper tail of the 

 distribution, while the median is accurately estimated its 

 variance is very imprecisely estimated. 



25 



