Full text of DOI:10.1111/j.1744-7976.2001.tb00302.x

Status of Benefits Transfer in the United States
and Canada: Comment
Dustin D. French and Fred J. Hitzhusen
The Ohio State University, Department of Agricultural, Environmental,
and Development Economic
Bergstrom and De Civita (BD) have recently discussed issues related to benefit transfer (BT)
in this journal. In the paper, the authors classify major types of BT and the testing procedures
for BT results. However, in constructing their classification scheme, they ignore the inter-
dependence that exists among the major types of BT. In addition, the description and applic-
ability of benefit transfer testing procedures raise some concern. More specifically, the
authors present the academic view of BT testing that is not consistent with the practical uses
of BT stated in their introduction.
This comment has two objectives: first, to show that a BT scheme may fall into two or
more of the major BT categories, and second, to point out limitations and conceptual difficul-
ties with the BT testing procedure, and suggest a more practical methodology of BT testing
not mentioned in Bergstrom and De Civita.
Major Types of Benefit Transfer
Bergstrom and De Civita state that BT methods can be divided into three major groups: fixed
value transfer, expert judgement, and value estimator models. While there is no doubt that
this broad classification captures most if not all the BT schemes, there remains a question as
to the relation of these three sets. The authors present these groups as implicitly disjoint.
However, there are examples of BT schemes which simply do not fit into a single category.
For instance, BD describe an example of a fixed value transfer where “the total benefits
of fishing at the policy site may be estimated as the product of some standard value per fish-
ing day at the study site and total fishing days at the policy site”. However, in choosing an
appropriate study site a researcher would need to consider the similarities shared with the policy
site. Since there are no formalized rules for a policy site selection, the researcher is ultimately
bound to use an expert judgement approach.
Another example that illustrates the dependence between the major types of BT is the
bayesian technique. The bayesian technique, as described, for example, by Atkinson et al
(1992), implicitly incorporates an expert opinion in choosing a prior. On the other hand, by
BD’s classification, the inference procedure itself would fall under a value estimator model.
Thus, this methodology is a combination of expert judgement and value estimator models.
There are perhaps other examples that illustrate this interdependence, but the general conclu-
sion is that a BT scheme may involve a combination of characteristics from two or more of the
major types. Recognition of this fact is important for better classification of future BT studies.
Testing BT
The BD paper draws a distinction between value surface tests (VST) and convergent validity
tests (CVT). In the value surface test, “study site data are combined either by pooling raw data
Canadian Journal of Agricultural Economics 49 (2001) 259–261
2
59

2
60
CANADIAN JOURNAL OF AGRICULTURAL ECONOMICS
across sites or by aggregating reported results across sites (as done in meta-analysis studies)”.
On the other hand, CVT examines the difference in means or model coefficients. In the BD
paper, the second approach is illustrated with Loomis (1992), which uses a Chow test to
examine travel cost model coefficients for types of fishing among various northwest states.
While the authors consider CVT and VST as two separate classes, both are actually
applications of meta-analysis. This becomes more apparent when one looks at a definition of
MA. Hunt (1996) describes MA as an approach that seeks to pinpoint reasons why different
studies of similar effects come up with divergent findings. In essence, the Chow test in Loomis
(1992) is a form of MA, because it seeks to determine if states and fishing type are influential
factors in a BT procedure. Therefore what BD present as BT testing procedures (i.e., CVT
and VST) is nothing more than classes of MA. This conclusion highlights two fundamental
problems with the suggested testing techniques.
First, MA requires information or data available on the policy site(s) beforehand. For
instance, Loomis used MA to test appropriateness of BT for salmon fishing in Oregon and
Washington and freshwater fishing in Oregon and Idaho. In that study, the author had the luxury
of four different data sets from which he could estimate travel cost models and perform Chow
tests to determine appropriateness of BT.
In real life, practitioners are more often faced with a single set of data or results when
using BT. Thus verification of the transfer by Chow or other tests requires collecting primary
data, which compromises the fundamental least-cost principle of BT. After all, if the primary
data are available, why perform BT in the first place? While BD have a disclaimer before the
description of CVT and VST stating, “a condition for undertaking these tests is that primary
study results must exist for every site in the test”, they fail to address the need for a BT test-
ing scheme that maintains the rationale of BT.
A more useful but perhaps less conclusive approach not mentioned in BD is to use exist-
ing study sites to simulate a BT situation. More specifically, before applying a BT procedure
to derive results for a policy site with incomplete data, a researcher randomly selects one of
the study sites to be used as a “policy” site and then applies the BT procedure to the remaining
(n–1) sites in order to predict benefits at the nth (missing) site. The BT results can be compared
to the actual values of parameters thus providing an estimate of the method’s accuracy, at
least within the study site sample. The use of this “with-versus-without” approach allows one
to determine the sensitivity and robustness of the transfer procedure without extensive data
collection and reveal the likely errors which may occur when the methodology is fully imple-
mented. Such a technique has been utilized but has not been established as a convention.
The second problem has to do with the notion of VST as presented in BD. More specif-
ically, the VST approach seems to be fundamentally different from the value estimator model.
For example, BD illustrate VST with the paper by Boyle et al (1994), which estimates the nat-
ural logarithm of willingness-to-pay as a function of physical, biological, and demographic
characteristics to explain the variation of 46 or more ground water studies. While the inten-
tion of Boyle et al (1994) was to explain differences among studies, the function that they
estimate could be re-calibrated with explanatory data collected at the policy site to infer BT
values. Thus, a question arises as to whether the VST and a value estimator model are one
and the same? In other words, the VST approach suggested as a BT testing procedure turns
out to be a BT procedure by itself, which further undermines its applicability.
In summary, the BD article has identified major types of BT and suggested two testing
procedures. However, the suggested classification is somewhat imprecise. A BT scheme may

COMMENT
261
possess characteristics from two or more of the major types presented. The authors also pre-
sented the academic view on BT testing that heavily relies on primary data thus contradicting
their own introductory statement: “in many situations, it may not be practical for an agency
or organization to collect primary data on which to base economic value estimates”. The
“with-vs.-without” approach may provide an alternative testing technique while not requiring
additional data. Finally, the VST testing procedure itself appears to be a BT methodology,
which raises questions regarding its redundancy and applicability.
ACKNOWLEDGMENT
Review provided by Dmitry Vedenov (The Ohio State University).
REFERENCES
Bergstrom, J. C. and P. De Civita. 1999. Status of benefits transfer in the United States and Canada:
A review. Canadian Journal of Agricultural Economics 47: 79–87.
Boyle, K. J., G. L. Poe and J. C. Bergstrom. 1994. What do we know about groundwater values?
Preliminary implications from a meta analysis of contingent-valuation studies. American Journal of
Agricultural Economics 76(5): 1055–61.
French, D. D. and F. J. Hitzhusen. 2000. Toward improved codification, protocol, and application of
benefit transfer methods. Paper presented at Southern Economic Association Annual Conference,
Washington, DC. 10 November 2000.
Hunt, M. 1997. How Science Takes Stock: The Story of Meta-Analysis. New York: Russell Sage
Foundation, 1997.
Loomis, J. B. 1992. The evolution of a more rigorous approach to benefit transfer: benefit function
transfer. Water Resources Research 28(3): 701–05.