186 July 1989
Western Journal of Agricultural Economics
procedures, and the federal guidelines. Ob-
viously, some adjustment for the omission of
travel time is required; however the precise
level is not known and would vary for each
study site. The statistical effect of the travel
time cost variable could be improved if spec-
ified as a continuous variable in dollars per
hour rather than as a qualitative variable in-
dicating presence or absence of the adjustment.
With respect to the adjustment for use of in-
dividual observations in TCM studies, some
economists argue that values from zonal stud-
ies should be increased rather than decreasing
values from individual observation studies be-
cause of the dampening effect of the aggrega-
tion problem in the zonal approach (Mc-
Connell and Bockstael). Finally, limitation of
the sample to in-state residents originates in
the institutional constraints of the researcher.
The precise level of adjustment for sample
truncation would vary with the actual origin
of the user population of each site.
The regression results indicate other prime
candidates for adjustment not considered by
the earlier work. Benefit estimates from TCM
studies omitting an effective cross-price term
for substitution could be decreased about 30%
according to the regression results. If the be-
havior-based TCM becomes the accepted
standard for benefit estimation, then the CVM
estimates Ofintended willingness to pay would
be increased by an average of 20-25%. The
results suggest that benefit estimates from CVM
studies using dichotomous choice questions
may be closer to TCM benefit estimates, per-
haps requiring about half as much adjustment.
However, benefit estimates from CVM studies
asking open-ended willingness-to-pay ques-
tions could be increased by 10-15% based on
the preliminary regression results considered
here. These are but a few of the possible ad-
justments that should be considered in apply-
ing the Sorg and Loomis approach of making
adjustments before presenting statistical sum-
maries of the data in policy applications.
An important question raised by the Forest
Service in applying the data to policy decisions
is whether the benefit estimates from other
public and private recreation sites are appli-
cable to Forest Service resources. The insig-
nificant coefficient for study sites administered
by the agency suggest that there may be no
appreciable difference. Apparently, the benefit
estimates from the literature review apply to
valuation of the agency’s recreation program.
In theory, benefit estimates for a forest lacking
data can be predicted by inserting appropriate
values of explanatory variables into the regres-
sions. Unfortunately, an insufficient number
of studies have been completed to obtain more
than a few estimates of value by this method.
The agency requires benefit estimates for 19
national recreation-use categories in nine For-
est Regions for a total of 171. However, only
three of the 19 national recreation-use cate-
gories and four of the nine Forest Regions are
significant in the models fitted to data from
the study sites (table 3). The other regions may
not differ significantly from the average and
thus cannot have significant coefficients, or
possibly sample size for these regions is too
small.
The specialized activity variable could pro-
vide a rough indication of the benefit for some
activities with few studies. For example, the
benefit of sightseeing and off-road driving, the
largest single recreation activity with 27.6% of
total output, would be $20 per day [= 39 —
(27.6 × .679)] based on the TCM equation.
This compares favorably to the mean of $20
for six studies of this activity (table 1). It seems
likely that the agency will need to rely on a
combination of several approaches until a
greater number of studies of most recreation
activities have been completed (McCollum et
al.; Bergstrom and Cordell).
Finally, these results should be considered
tentative and subject to revision with more
complete specification of the model. Sensitiv-
ity analysis omitting various combinations of
variables from the final equation significantly
changes the coefficients of those remaining (as
in Atkinson and Crocker; Smith and Kaoru).
This suggests that leaving important variables
out of the final equations may attribute too
much of the variation in benefit estimation to
the differences in method that are included.
Nonetheless, the equations in table 3 include
many possibly important variables and pro-
vide a basis for eliminating some of them as
serious candidates for new research. The task
remains to discover how far these results can
be generalized. The importance of continued
research is illustrated by the conceptual and
empirical difficulties associated with estima-
tion and the potential importance of recreation
benefit in the economic assessment of pro-
grams such as forest recreation.