based on the standard double log model . 



The statistical inference assumptions relate to the test 

 statistics (such as the t-statistic and F-statistic) used to 

 evaluate the "fit," or appropriateness, of the model. These 

 include homoscedasticity , independence of error terms 

 (autocorrelation), normal properties of the dependent variable , 

 and the assumption that no linear relationship exists between any 

 two or more of the independent variables (multicollinearity ) . 

 Generally, the major problem with estimating a travel cost model 

 using cross-sectional data has been heteroscedasticity . 



A final assumption of ordinary least squares regression is 

 that the variables are correctly measured in the survey data. In 

 the TCM model, measurement error is a potential problem, 

 especially with respect to reported travel costs and/or 

 distances . 



A related problem is that the TCM model requires zonal 

 aggregation. The costs of defining population zones on other 

 than political boundaries (county, state) are prohibitive. This 

 aggregation will inevitably result in some survey distances (even 

 if the latter are correct) that do not represent population- 

 weighted averages. The methods used to define zones and validate 

 distances are described later. 



Calculation of Benefits From the Per Capita Demand Equation 



Once the per capita demand equation of the form in Equation 

 1 is estimated using OLS regression, benefits can be calculated 

 in several ways. First, the per capita curve can be integrated 

 for each zone of origin (or individual observation) between the 



9 



