LRDTRDIS = log of round trip distance from origin i to site j. 

 LYRSHUNT = log of number of years individual i has hunted site j. 

 LSUCS = log of deer success ratio at site j. 



The adjusted R^ statistic and F-statistc are quite high for 

 a TCM model, indicating that the model provides a good fit to 

 the data. All of the estimated coefficients except success are 

 significant at the 95% level and have the expected sign. 

 Success is significant at the 90% level. The key parameter, for 

 purposes of consumer surplus estimation, is the slope 

 coefficient on distance. The very high t-statistic for this 

 parameter indicates that it is precisely estimated. 



None of the site characteristic variables other than success 

 were significant, and most of the demographic variables (such as 

 income) were also not significantly correlated to per capita 

 trips. The sample with substitute information was limited almost 

 entirely to residents and was too small to use for the general 

 model . 



In addition to statistical significance and consistency with 

 the theoretical model, the TCM model estimate can be evaluated on 

 how well it predicts. As the model is estimated on per capita 

 trips, accurate prediction of total trips is critical for the 

 consumer surplus estimate. The model badly overestimates total 

 trips for the sample: 7,180 predicted trips versus 2,010 actual 

 trips. Examination of trip prediction at the observation level 

 revealed that most of the overprediction is for origins either 

 very close to or very far from the hunting sites. 



Because of the prediction problems, the model was tested for 



16 



