LawInfo Forum

Contingent Valuation of Natural Resource Damages

Lewis Owen Amack

March 16, 1994


TABLE OF CONTENTS

I. Introduction
II. Historical Background of Natural Resource Contingent Valuation
III. A Boolean Approach to Natural Resource Valuation
IV. Contingent Valuation Methodology
A. Survey Instrument Design and Development
B. Survey Administration
1. Sampling
2. Response Rate
C. Survey Interpretation
V. Contingent Valuation Shortcomings and Solutions
A. Nonacceptance
B. Costliness
1. Feasibility
2. Social Costs
i. Overestimation
ii. Underestimation
C. Accuracy
D. Representativeness
1. Market Size
2. Sampling Bias
E. Hypotheticality
1. Realism
2. Budgeting
3. Expert Panels
F. Replicability
1. Pollution Incident Chronology
2. Administration Methodology
G. Survey Effects
H. Media Effects
I. Interviewer Effects
J. Survey Bias
K. Strategic Bias
L. Embedding
M. Self-reinforcement
N. Centering
O. Ordinality
P. Additivity
1. Respondent Additivity
2. Aggregate Additivity
Q. Dispersion
R. Income Effects
S. Nonpristineness
T. Ex post facto Questioning
U. Double Counting
VI. Recommendations
VII. Research Suggestions
VIII. Conclusion

I. INTRODUCTION

Contingent valuation (CV) is a survey method for assessing monetary value that is particularly useful for items which have no objective market price. CV technology has been around for decades,1 but it was not applied to environmental damage assessments until passage of the Comprehensive Environmental Response, Compensation, and Liability Act2 (CERCLA) and the Oil Pollution Act (OPA).3 CV has gained greater legal recognition during the last few years, as courts have increasingly acknowledged the need to consider all environmental assets, including publiclyowned natural resources for which market valuation is inadequate.

However, critics contend that CV data should only be admissible in court if its validity4 and reliability problems are resolved. Moreover, courts may reject poorly conducted CV studies.5 Therefore, public trustees and professional
[
Top | Next Page ]


1Although the primary purpose of CV in natural resource damage assessments is the measurement of nonuse values, CV was originally developed to measure use values. David S. Brookshire et al., Estimating Option Prices and Existence Values for Wildlife Resources, 59 LAND ECON. 1 (1983) (cited in Note, Ask a Silly Question...: Contingent Valuation of Natural Resource Damages, 105 HARVARD L R. 1981 (1992))
242 U.S.C. 96019657 (1982 and 1986). CERCLA is also designated as Superfund because, pursuant to the Superfund Amendments and Reauthorization Act of 1986 (SARA), CERCLA claims may be presented to the Hazardous Wastes Superfund-a multibilliondollar federal monetary reserve set aside for financing environmental cleanups-when responsible parties cannot be identified or default on their liabilities.
333 U.S.C. 27012761 (1990). Statutes providing for environmental damages, in chronological order of adoption, were the TransAlaska Pipeline Act, 43 U.S.C. 1653 (1973); the Deepwater Port Act, 33 U.S.C. 15011524 (197*; the Clean Water Act, 33 U.S.C. 12511387 (1982) (1977 Amendments); the Outer Continental Shelf Lands Act, 43 U.S.C. 13011356 (1978 Amendments); CERCLA (1980); the Marine Protection, Research, and Sanctuaries Act, 16 U.S.C. 1443 (1988 Amendments); the National Park System authorities, 16 U.S.C. 19jj (1990); and the Oil Pollution Act (1990) (superseding the Deepwater Port and Outer Continental Shelf Lands Acts, as well as the oil spill provisions of the Clean Water Act). The Oil Pollution Act of 1990 was enacted in the wake of the Exxon Valdez incident and specifically provides for damages associated with oil pollution within navigable waters of the United States. CV, which first appeared in 1986 DOI CERCLA regulations, has thus far been employed primarily in Superfund and OPA cases.
4CV has been criticized in terms of criterion, theoretical, and convergent validity. Id. at 6263.
5Thus far, the only natural resource damage litigation involving CV was State of Idaho v. Southern Refrigerated Transport, Inc., 1991 WL 22479, 48 (D. Idaho), concerning a chemical spill which caused a fish kill. Idaho presented a CV survey which was performed for power planning prior to the spill, not for damage assessment purposes. The court rejected the CV as applied to fish valuation without explanation.


survey researchers who wish to use CV for natural resource damage assessments should be familiar with established CV guidelines. This paper will address the reported weaknesses of CV, and suggest strategies for counteracting these deficiencies while enhancing CV validity and reliability.

II. HISTORICAL BACKGROUND OF NATURAL RESOURCE CONTINGENT VALUATION

CERCLA and OPA specify generally that when an environmental pollution incident occurs, designated public trustees6 shall (1) use the best available procedures7 for assessment8 of response costs9 and natural resource losses,10 both direct and indirect, "including, but not limited to, replacement value, use value, and ability of the ecosystem or resources to recover,11 as well as "the costs of damage assessment"12; (2) take legal action13 against potentially
[
Previous Page | Top | Next Page ]


6Under CERCLA, 42 U.S.C. 9607(f)(2)(A) concerns federal trustees and 42 U.S.C. 9607(f)(2)(B) addresses state trustees. Under OPA, 33 U.S.C. 2706(c)(1) concerns federal trustees and 33 U.S.C. 2706(c)(2) concerns state trustees.
742 U.S.C. 9651(c)(2)(B). Not explicit in OPA.
8Under CERCLA, type A assessments, 42 U.S.C. 9651(c)(2)(A), are generally used for small spills and utilize monetary tables or computer programs (such as the Natural Resource Damage Assessment Model for Coastal and Marine Environments (NRDAM/CME)). Type B assessments, 42 U.S.C. 9651(c)(2)(B), are generally used for largescale pollution incidents, and involve more extensive studies and field work. Under OPA, CV technology is limited to Comprehensive Damage Assessments (analogous to a Type B assessments). Report of the NOAA Panel on Contingent Valuation, United States Department of Commerce, National Oceanic and Atmospheric Administration (Rockville, MD), 59 Fed. Reg. 1062, 44 (January 7, 1994) [hereinafter NOAA] (to be codified at 15 C.F.R. 990).
9For CERCLA, see 42 U.S.C. 9607(a)(4)(A)(B). The primary component of response cost is cleanup or removal of the hazardous substance. 42 U.S.C. 9604(a)(1). For OPA, see 33 U.S.C. 2702(b)(1) (removal costs).
10For CERCLA, see 42 U. S.C . 9607(a)(4)(C). For OPA, see 33 U. S .C . 2702(b)(2)(A) (natural resources) and 33 U.S.C. 2702(b)(2)(C) (subsistence use).
1142 U.S.C. 9651(c)(2)(B). For OPA, see 33 U.S.C. 2706(d)(1)(A) and 33 U.S.C. 2706(d)(1)(B) (specifically addressing diminution in value of natural resources prior to restoration).
12For CERCLA, see 42 U.S.C. 9607(f)(1). This section also specifies that sums received by the trustees shall be used "only to restore, replace, or acquire the equivalent ... natural resources." For OPA, see 33 U.S.C. 2706(d)(1)(C).
13The goals of legal action are not only restitution for pollutioncaused damages, but also incentivizing prevention of spillage and leakage of hazardous materials, optimizing economic efficiency and consumer welfare, and internalization of the full social or external costs of pollution by responsible parties. Thomas A. Campbell, Economic Valuation of Injury to Natural Resources, 6 W.T.R. NAT. RES. & ENV. 28 (1992). Of all current legal approaches, CV under CERCLA and OPA offers the most expansive opportunities for natural resource damages recovery. Nevertheless, trustees may employ other damage award strategies, including additional state and federal statutes, as well as state common


responsible parties (PRPs)14; and (3) coordinate cleanup, restoration,15 and any natural resource acquisitions.16 The responsibility for determining the methodology for natural resource damages assessments17 is delegated by the President.18 The accuracy of the trustees' damage figures is rebuttably presumed only if DOI or NOAA assessment guidelines are followed.19 In 1989, the Court of Appeals for the District of Columbia20 found that DOl's assessment guidelines21 were undervaluing22 natural resource damages,23 and suggested that contingent
[
Previous Page | Top | Next Page ]


law tort, parens patriae, or public trust remedies. Some state laws also allow the use of CV for damage claims. On the other hand, the common law of many states may proscribes consideration of existence or intrinsic values on speculativeness grounds. James L. Nicoll, Marine Pollunon and Natural Resource Damages: The MultiMillion Dollar Damage Award and Beyond, 5 U. OF SAN FRAN. MARITIME L.J. 323 (1993).
14PRPs are liable for both direct and indirect environmental injuries. Furthermore, "[t]he measure of damages ... shall not be limited by the sums which can be used to restore or replace [natural] resources." 42 U.S.C. 9607(f)(1). Under OPA, "responsible party" is defined at 33 U.S.C. 2701(32).
15The feasibility of restoration is directly related to the valuation of the damaged resources. Policymakers who prefer restoration favor CV, because CV measures total compensable value, including losses in consumer welfare, governmental fees, and economic rent. Because CV encompasses more than use values, it reduces the likelihood of a gross disproportionality finding (see infra notes 27 &42). NOAA, supra note 8, at 49.
1642 U.S.C. 9607(f) (CERCLA) and 33 U.S.C. 2706(c) (OPA).
17As specified at 42 U.S.C. 9651 (c)(l) (CERCLA) and 33 U.S.C. 2706(e)(1) (OPA), pursuant to a National Contingency Plan, 42 U.S.C. 9605 (CERCLA) and 33 U.S.C. 1321(d) (OPA).
18CERCLA's damage assessment methods are determined by the Department of Interior (DOI), while OPA's are formulated by the National Oceanic and Atmospheric Administration (NOAA), a division of the Department of Commerce, in consultation with the Environmental Protection Agency (EPA) and the Fish and Wildlife Service (FWS).
1942 U.S.C. 9607(f)(2)(C) (CERCLA) and 33 U.S.C. 2706(e)(2) (OPA).
20According to Gregory G. Garre, CERCLA, Natural Resource Damage Assessments, and the D. C. Circuit's Review of Agency Statutory Interpretations under Chevron, 58 GEO. WASH. L. REV. 932 (1990)' Chief Judge Patricia M. Wald and Circuit Judges Spottswood W. Robinson, III and Abner J. Mikva were insufficiently deferential toward DOI interpretations of CERCLA ambiguities, instead allowing their own activist policy views to influence their judgments concerning statutory construction and Congressional intent.
21The challenged guidelines were promulgated in August 1986 and February 1988, the latter after SARA's enactment.
22The federal government is itself responsible for some of the undervaluation, since the prices which it sets for natural resource utilization, including park admission fees; timber, water, and grazing rights; and fishing and hunting licensure are not designed to cover all costs, but to encourage public use. Denis Swords, Note, Ohio v. Department of Interior: A Contingent Step Forward for Environmentalists, 51 LA. L.R. 1347, 1364 (1991). The difference between WTP and actual fees paid to the government, known as the economic rent, is the public cost of resource undervaluation. NOAA, supra note 8, at 29.


valuation may be the "best available procedure" for assessment, especially where restoration costing is inapplicable or impractical.24 Furthermore, the court announced that "option and existence values may represent 'passive' use,25 but they nonetheless reflect utility derived by humans from a resource, and thus, prima facie, ought to be included in a damage assessment."26 Consequently, in 1991 the DOI declared the measure of damages to be restoration cost27 plus services lost during recovery,28 and that "[i]f nonuse values29 are significant, the only way to quantify these values explicitly is to use the contingent valuation methodology."30
[
Previous Page | Top | Next Page ]


23The DOI's 1986 Type A assessment guidelines were stricken down in Colorado v. Department of Interior, 880 F.2d 481 (1st Cir. 1989) (damage assessments must include both replacement and lost use values), and the 1986 Type B guidelines were invalidated in a companion case, Ohio v. Department of Interior, 880 F.2d 432, 442 (D.C. Cir. 1989) (favoring a restoration cost baseline, rejecting the "lesser of" restoration, replacement, or diminution in value standard and the hierarchy of assessments approach (which required market costs if available, else "comparable sales" prices if determinable, else nonmarket valuation methods); recognizing that market values fail to capture all openaccess (i.e., nonmarket) values of natural resources; and upholding the public ownership rule (i.e., limiting coverage to publiclyowned property) and the "committed use" requirement (i.e., assessing only uses which were documented as current or anticipated at the time of the pollution event, as opposed to highly speculative uses) when computing diminution of value during restoration).
24According to 43 C.F.R. 11.83(d)(1) (1990), CV should be limited to circumstances "when neither the market price nor the appraisal methodology is appropriate." In such circumstances, CV should be used to measure both use and nonuse values, since "[CV] is just as valid a method to estimate use as other methods." 51 Fed. Reg. 27674,27719 (1986).
25This denotation of option and existence values may originally have been erroneous, since "passive use" is sometimes considered synonymous with "nonconsumptive use," whereas option and existence are "nonuse" values. Swords, supra note 22, at 1367. Nevertheless, NOAA currently prefers "passive use value" to "nonuse value" when referring to the combination of option and existence values. NOAA, supra note 8, at 45.
26Ohio v. DOI, 880 F.2d at 464.
27However, restoration cost is zero if restoration is not performed because of costs grossly disproportionate to damages, technical unfeasibility, or natural recovery. One heuristic is that restoration cost is grossly disproportionate if more than twice the value of natural resource damages.
28This restoration cost approach is becoming increasingly prevalent and now also applies to the Clean Water Act. Barry N. Breen, Natural Resource Damages: Hazardous Wastes, Superfund, and Toxic Substances, C778 ALIABA 403, 408 (October 29, 1992).
29Nonuse values, including existence and option value, are distinct from nonconsumptive use values. They were first described in Burton A. Weisbrod, CollectiveConsumpnon Services of IndividualConsumption Goods, 78 Q.J. ECON. 471, 47273 (1964) and John V. Krutilla, Conservation Reconsidered, 57 AM. ECON. REV. 777, 77881 (1967).
30Natural Resource Damage Assessments, 56 Fed. Reg. 19,759 (April 29, 1991).


III. A BOOLEAN APPROACH TO NATURAL RESOURCE VALUATION

From a systemic perspective,31 there are three major types of natural resource valuesuse existence, and intrinsic.32 Use value measures the consumptive value of tangible natural resources -such as fish, fowl, timber, water - as well as nonconsumptive or recreational uses of natural resources such as fishing, boating, birdwatching, backpacking, and picnicking. Nonuse value, which consists of option and existence value, represents the marginal utility of a natural resource beyond its use value. The components of existence value include33: (1) option value, the value of retaining resources for future use, which is often distinguished from existence value because of its quasi-use quality; (2) quasi-option value, a subcategory of option value, defined as the estimated value of potential but as yet undiscovered future uses34; (3) vicarious value, the value of knowing that a natural resource is protected; (4) intertemporal value, the value of natural resources to future generations35; and (5) bequeathment value, the value of endowing a natural resource to posterity.36
[
Previous Page | Top | Next Page ]


31The systemic or ecological economics perspective is a hierarchical model which recognizes both subjective human values and objective biological values. Michael W. Jones, Comment, Natural Resource Damage Assessments for Oil Spills: Policy Considerations Underlying the Evolution of the Department of Regulations, I VILL. ENVTL. L.J. 491 (1990).
32Use, existence, and intrinsic values may be viewed on a continuum. At one extreme are use values, which are anthropocentric, dualistic, and economic. At the other extreme, intrinsic values represent biocentrism, holism, and "deep ecology. " At the center of the continuum are existence values, which are neither anthropocentric nor biocentric, and represent "shallow ecology." Id.
33Existence value, which is also known as preservation value or vicarious value, may be denoted as the value to a human of knowing that a resource which he never intends to consume is protected. However, existence value is also frequently defined as including option and other longterm human use values (e.g., intertemporal, bequeathment), as well as nonuse values. Therefore, hereinafter existence value can be assumed to include option and quasioption values, unless otherwise indicated. In addition to CERCLA and OPA, existence value is recognized in the Endangered Species Act, the National Interests Lands Conservation Act of 1980, and the Wilderness Act of 1964, 16 U.S.C. 1131 1136 (1988).
34According to Eric D. Olson, Natural Resource Damages in the Wake of the Ohio and Colorado Decisions: Where Do We Go From Here ?, 19 ENVTL. L. REP. 10551, 10555 (1989), CV studies should separately account for option, vicarious, and intertemporal existence values.
35For instance, a declared goal of the Alaskan National Interests Lands Conservation Act of 1980, 16 U.S.C. 3101 (a) (1982), is the preservation of land and water for "future generations. "
36A CV of Grand Canyon visibility found nonuse value to be 38% bequest (i.e., bequeathment) value, 31% existence (vicarious and intertemporal) value, and 31% option (and quasioption) value. William


Intrinsic value recognizes the valuation of the environment from a nonhuman perspective.37 Because intrinsic value is expressed in nonhuman terms, a monetary equivalent is highly theoretical.38

While a satisfactory intrinsic valuation approach has yet to be developed, several techniques are available to measure use values, including restoration or replacement cost,39 market valuation,40 and behavioral use valuation.41 Most of these techniques are supply-sided or costbased, and rely upon market prices. On the other hand, existence value can only be measured with demand-sided, value-based nonmarket methods. Of the nonmarket methods, CV is most effective in capturing the full utility of natural resources. Although less convenient and more subjective than marketoriented techniques, CV more
[
Previous Page | Top | Next Page ]


D. Schulze et al., The Economic Benefits Resources Preserving Visibility in the National Parklands of the Southwest, 23 NAT. Resources. 149, 149 (1983).
37A principle concern of the Endangered Species Act, 16 U.S.C. 15311544 (1973), is the intrinsic value of flora and fauna. Existence and intrinsic values are collectively referred to as openaccess values.
38The intrinsic value of natural resources was first suggested in 1972. Christopher Stone, Should Trees Have Standing? Toward Legal Rights for Natural Objects, 45 S. CAL. L. REV. 450 (1972). Nevertheless, intrinsic values have never been explicitly incorporated into a natural damage award, although some courts have at least given titular acknowledgment to its potential legitimacy as a remedy. See, e.g., Justice Douglas dissenting opinion in Sierra Club v. Morton, 405 U.S. 727, 74143, 92 S.Ct. 1361, 136970 (1972).
390ne measure of replacement cost is the value of land equivalent to the prespill conditions of the polluted site.
40Diminution in property value is a common form of market valuation. However, market prices tend to vastly underestimate both use and nonuse values, because they are distorted by imperfect knowledge and markets, and generally fail to include such factors as externalities, public use, and consumer surplus.
41Behavioral use valuation techniques, which are necessary when market prices must be indirectly imputed, are nonmarket techniques based upon observations of human behavior or consumer surveys, and include travel cost studies, hedonic pricing, the factor income or reverse value added method, unit day measurement, and CV. Ellen Louderbough, Note, The Role of Science in Valuing Natural Resources After State of Ohio v. Department of Interior, 32 NAT. RES. J. 137 (1992); 43 C.F.R. 11.83(d) (1986). The most common nonmarket methods are CV, travel costing, and hedonic pricing. Of these methods, only CV can measure nonuse values. Travel cost analysis considers the actual costs historically incurred by visitors to the actual (or a comparable) site, as well as the opportunity costs (i.e., expected value of alternative pursuits) and environmental impact costs of travelers, and is probably the most accurate method of evaluating recreational use values. Hedonic pricing, on the other hand, has thus far only been effective in correlating local environmental conditions with private property values, not in assessing damages to public property.


accurately reflects the full economic worth of natural resources, at least from the perspective of human consumers.42

IV. CONTINGENT VALUATION METHODOLOGY

A CV has three major procedural components: (1) survey instrument design and development, (2) survey administration, and (3) survey interpretation.

A. Survey Instrument Design and Development

The four basic elements of CV survey design are: (1) description of a market43; (2) description of a natural resource; (3) value elicitation questions; and (4) validation and background questions to verify comprehension and acceptance of the scenario, and to elicit socioeconomic and attitudinal characteristics for interpretation of the variation in valuation responses. 44

The design and development phase involves deciding between restoration and prevention, 45 willingness to pay (WTP) and willingness to accept (WTA), 46 and single commodity CV v. multiple valuations47; pretesting48; defining scenarios49; selecting photographic materials; constructing questions50; and
[
Previous Page | Top | Next Page ]


42Furthermore, CV is the preferred test for gross disproportionality of restoration costs. Frank Cross, Natural Resource Damage Valuation 42 VAND. L. REV. 269, 338 (1989). The "grossly disproportionate" standard originated in Puerto Rico v. SS Zoe Colocotroni, 628 F.2d 652, 67576 (1st Cir. 1980), cert. denied, 450 U.S. 912 (1981).
43CV is a type of "constructed market" methodology. Christine M. Augustyniak, Economic Valuation of Services Provided by Natural Resources: Putting a Price on the Priceless, 45 Baylor L.R. 389, 399 (1993).
44NOAA supra note 8, at 33.
45The Exxon Valdez study employed a prevention plan. See Richard T. Carson et al., A Contingent Valuation Study of Lost Passive Use Values Resulting from the Exxon Valdez Oil Spill (Report to the Attorney General of Alaska), 2.1, 12 (November 10, 1992).
46Although WTP produces a small downward bias, it substantially narrows the confidence intervals of CV. Id. at 5.11, 117. WTP was used in the Exxon Valdez survey. Id. at 1.3.2, 8. See infra p. 22 & n.138.
47Within the time constraints of an interview, a single commodity can be described in greater detail if other commodities were excluded. Therefore, Prince William Sound was chosen as the only CV scenario for the Exxon Valdez survey. Id. at 2.4, 1920.
48Survey instrument development for the Exxon Valdez study was an iterative procedure involving multiregional pilot surveys and field tests. Id. at 1.4, 9. Particularly important concerns to address during pretesting are question ordinality effects, comprehension, and scenario believability.
49The Exxon Valdez trustees used exploratory, moderator-led focus groups in six states during scenario development. Id. at 1.3.2, 7.


refinement of the survey content.51 The CV survey involves the creation of a hypothetical market for those natural resources which must be assessed. For example, the location and environmental effects of a possible beach oil spill might be described, as well as an oil spill prevention program. The respondent (R) might then be asked how much he would be willing to allow gasoline or home heating oil prices to increase, or how much he would be willing to pay in higher taxes, in order to finance the prevention program.52 Alternatively, participants might be asked to bid for natural resources, or respond to a "take or leave it" referendum format.53 However, a singlecommodity WTP referendum approach is recommended because it most realistically simulates actual voting behavior.54

B. Survey Administration

The principal concerns at this stage are (1) sampling methodology and (2) response rate.

1. Sampling

CV requires surveying a sample of consumers within the population universe affected by the pollution incident. The size of the universe is typically directly related to the size of the natural resource damages. Thus, the universe for the eleven million gallon ExxonValdez oil spill was the entire population of the United States,55 whereas the universe of either the 600,000-gallon January
[
Previous Page | Top | Next Page ]


50For example, referenda CV figures may be determined from median WTP findings in pilot studies. Id. at 3.4, 58.
51survey instruments should ( I) present comprehensible, plausible scenarios which mirror the actual injuries, (2) measure utility losses concordant with economic theory, and (3) maintain impartiality. Id. at 2.2, 13.
52Augustyniak, supra note 43, at 399.
53Swords, supra note 22, at 1354.
54The construct validity of a CV can be further enhanced if it is portrayed as a democratic decision or jury award rather than a survey or opinion poll.
55carson. supra note 45. at 1.4. 10.


1991 U. S. Oil Spill in Tacoma or the 210,000-gallon February 1991 spill at Texaco's Anacortes refinery might be confined to the state of Washington.

2. Response Rate

The response rate is highly dependent upon the modality of administration. There are three common modes of CV administration: mailin questionnaires, telephonic surveys, and inperson interviewing.56 Mail surveying is the least expensive mode, but the minimal acceptable response rate of seventy percent is rarely achieved.57 Because mailings also have less interrogative flexibility than other modes, they are only recommended for pretesting or mixedmode surveys.58 On the other hand, telephonic interviewing is far less costly than inperson interviewing, yet equal in interrogative flexibility and less vulnerable to interviewer effects59 In addition, random-digit dialing is almost as effective as probabilistic sampling.60 Nevertheless, the recommended survey modality is facetoface intrerviewing,61 because an interviewer can provide photographs of the damage scenario, establish rapport, and monitor the responsiveness of interviewees.62
[
Previous Page | Top | Next Page ]


56The Exxon Valdez CV survey involved 1,043 personal interviews, and had a 75.2% response rate. Id. at 4.9, 75. The average interview time was 42 minutes, the range was 19 minutes through 2 1/2 hours, and the 95% confidence interval was 25 through 70 minutes. Interview cost ranged from $50 to $600. Id. at 4.8, 74.
57NOAA supra note 8, at 36.
58In mixedmode surveys, mail responses can be crossvalidated (1) against other survey modalities and (2) by subsampling. Report of the NOAA Panel on Contingent Valuation (58 Fed. Reg. 4601), 47 (January II, 1993) [hereinafter NOAA].
59Telephone surveys can cut interviewing costs by as much as fifty percent. Id. at 48.
60NOAA, supra note 8, at 76.
61Id.
62In contrast, the effects of respondent inattention, distraction, and noncooperation are more pronounced in telephone interviewing. NOAA, supra note 58, at 48.


C. Survey Interpretation

The description of aggregate contingent value63 should be informative- not simply a product of average elicited value and population size. Among the concerns of CV statistical analysis are: (1) selection of a CV estimation methodology,64 (2) extrapolation of missing and implausible values,65 (3) developing a weighting scheme,66 (4) determining criteria for elimination of outliers,67 and (5) sensitivity analysis,68 and (6) explaining relationships between independent variables and CV.69 Although mean WTP more accurately
[
Previous Page | Top | Next Page ]


63The CV of a natural resource is determined by aggregating survey findings across the entire relevant human population. As with recreational use values, willingness to pay can be aggregated "for all users in the area." 44 Fed. Reg. 72912,72958 (1979).
64WTP can be described as a mean, median, maximum likelihood estimate (see Carson, supra note 45, at 5.6, 96), valuation function (see Id. at 5.9, 104), or in terms of probabilistic confidence intervals (see Id. at 5.7, 99). For instance, a median WTP of $31 per household was reported by the Exxon Valdez study. Id. at 1.5, 11. The "lower bound" of aggregate WTP for "lost passive use" was $2.8 billion, or $31 multiplied by 90.8 million households. The 95% confidence range was $2.4 billion <= CV <= $3.2 billion. Id. at 5.14, 123.
65Where predictor variable data is missing or apparently erroneous, a value is generally interpolated through multiple regression estimation. For example, in the Exxon Valdez study, 15% of Rs either provided no answer for income, or a value that was highly improbable. Id. at 5.9.1, 105106 & n.90. An improbable response might be an income of $75,000 for a 20yearold student, or $ 10,000 for a dualincome couple in a high rent area. Such figures are changed to the most likely true value. Where it is impractical to rectify a missing or implausible value for a variable which is part of the valuation function, the entire observationi.e., all data for that interviewee must be eliminated from the model.
66A weighting scheme is used for nonresponse adjustment and post-stratification with respect to the most significant CV prediction variables. For example, region, age, race, household size, and marital status of the Exxon Valdez survey sample were adjusted so that they more accurately reflected the 1990 census data for the relevant population. Id. at 4.10, 7677. Also, the 2.7% non-English speaking households were eliminated due to sampling and cost limitations. Id at 4.10, 77 & n.69.
67A simple method of outlier analysis is to eliminate all observations whose reported CV deviates from its valuationfunctionestimated CV by more than 4 standard deviations.
68Some of the sensitivity analyses performed by the Exxon Valdez researchers, and their conditional median WTPs (in parentheses) were: (1) "Not sure" responses for WTP were changed from "no" to the median of $31 ($33). (2) Eliminating Rs whose reported CV was "not at all serious" ($33). (3) (a) Omitting the 10% of Rs who were uncooperative, "extremely" influenced by onlookers, distracted, "extremely" bored, or low in comprehension ($34). (b) Further deleting the 18% of Rs who were uninformed, "very" distracted, or "slightly" bored ($38). (4)(a) Dropping the 24% of Rs who thought that oil companies should pay ($44). (b) Further removing Rs opposed to any tax, and Rs who thought that the money would be wasted ($47). (5) Discarding Rs who based their CV on a spill of a different magnitude from the scenario ($28). (6) Excluding Rs who doubted the effectiveness of the prevention plan ($43). (7) Culling protest responses ($38). (8) Extracting Rs whose WTP was 1% or more of their annual income ($30). Carson, supra note 45, at 5.10, 112 & 5.11, 113115.
69For example, the Exxon Valdez valuation model adjusted for protest responses, perceptions of hypothetical damage prevention differing in magnitude from the actual oil spill, and variation in perceived effectiveness of the spill prevention plan. Id. at 5.10, 112.


represents consumer behavior, the more conservative median is preferred for CV aggregation, especially when WTP data is heavily skewed. Outlier analysis, used to identify extreme observations for elimination from the computational model, generally requires multiple regression analysis, sometimes augmented by graphical depictions of multivariate CV curves. There are many ways to mathematically or graphically depict the relationships between CV and background or attitudinal data.70 Cross-tabulation is recommended, because it summarizes data in a format that is easy to understand.71 Variables that might be cross-tabulated against CV include income, distance from site, prior site visitation, prior knowledge of injury,72 belief in the scenario, understanding of the restoration or prevention plan, attitudes toward big business,73 and attitudes toward the environment.74

CV surveys have been found to produce reasonably consistent and replicable results75 that are roughly comparable to travel cost76 and hedonic77 analyses.78 Yet even after random sampling of the relevant population; meticulous survey design; skilled, unbiased interviewers; and multifactor econometric analysis of survey results, additional problems affecting CV validity and reliability may need to be resolved.
[
Previous Page | Top | Next Page ]


70A mathematical model might predict WTP as a function of population demographics, baseline environmental conditions, and discharge characteristics.
71NOAA supra note 8, at 41.
72For example, over ninety percent of RS in the Exxon Valdez survey were aware of the spill. carson, supra note 45, at 1.5, 11.
73"Protest voting'' occurs among RS with hostility toward big business, oil companies, or-conversely- environmentalism.
74carson, supra note 45, at 1.5, 11; NOAA, supra note 58, at 34.
75Brookshire, supra note 1, at 1.
76Richard C. bishop et al., Toward Total Economic Valuation of Great Lakes Fishery Resources, 116 TRANSACTIONS OF THE AMERICAN FISHERIES SOCIETY, 339, 343 (1987).
77David S. Brookshire et al., Valuing Public Goods: A Comparison of Survey and Hedonic Approaches, 72 AMERICAN ECONOMICS REVIEW 165 (1982).
78Yet travel and hedonic pricing techniques fail to include existence values.


V. CONTINGENT VALUATION SHORTCOMINGS AND SOLUTIONS

The major criticisms of CV are its unreliability, unverifiability,79 potentially excessive valuations.80 Some opponents of contingent valuation argue that C:V findings should be inadmissible81 because CV is hearsay,82 CV methodology is experimental,83 or because CV fails to meet the stringent common law standards for damage assessment.84 Others contend that because CV lacks reliability, it should be limited to longterm harms to unique, well known resources.85 Below are some of the shortcomings of CV, and recommendations for improvement.
[
Previous Page | Top | Next Page ]


79Although external criterionrelated validation of CV is impossible (see NOAA, supra note 58, at 6), discriminant validation and calibration is possible, and reallife referenda may be useful for construct or content validation. DAVID B. MAGLEBY, DIRECT LEGISLATION (Johns Hopkins Press, 1984) (cited in NOAA, supra note 58, at 25).
80Contrarily, there are several reasons to believe that CV may underestimate nonuse values: (1) CV generally measures total value, not just nonuse value; (2) population growth and resource depletion will make natural resources increasingly scarce and valuable in the future; (3) nonuse value of private property adjacent to public lands is generally excluded in CERCLA and OPA damage assessments. NOAA, supra note 8, at 13.
81Charles J. Cicchetti & Neil Peck, Assessing Natural Resource Damages: The Case Against Contingent Valuation Survey Methods, 4 NAT. RESOURCES & ENV. 6, 8 (1989) (suggesting that expert witness testimony of natural resource economists concerning nonuse values is preferable to CV surveying).
82Note, supra note 1, at 1999.
83According to Fed. R. Evid. 703, an expert's methods must meet generally accepted standards of the scientific community. Therefore, for a CV project to be admissible, not only must CV itself be generally acceptable to relevant scientific communities, but the specific CV project must meet community standards.
84Swords, supra note 22, at 1360; JOHN G. FLEMING, THE LAW OF TORTS 205, 213 (7th ed. 1987) (because of the general rule of tort law to exclude losses which are uncertain or difficult to calculate, nonuse values have traditionally been unavailable in tort actions). But see NOAA, supra note 8, at 60 (there is no requirement in tort law that damages be proven with mathematical precision). Moreover, CERCLA was intended to alter common law rules. Ohio v. DOI, 880 F.2d at 45556 (cited in Michael A. Brown et al., Court of Appeals for the District of Columbia Circuit Issues Decision on Natural Resource Damage Assessment Regulations 377 PLI/LIT. 245 (Sept. 12, 1989)).
85Natural Resource Damage Assessments, 56 Fed. Reg. 19752, 19760 (1991); Krutilla, supra note 29, at 77980 (only unique, irreplaceable, unsubstitutable natural resources are worthy of nonuse value recovery). However, NOAA rejects the view that a natural resource must be unique or wellknown in order to have significant nonuse values. Nevertheless, recognition is given to "resources of special significance," such as historical landmarks, archaeological sites, and endangered species. NOAA, supra note 8, at 82.


A. Nonacceptance

CV is rejected by many economists,86 in part because of its questionable validity.87 Nevertheless, whether a technology such as CV should be employed is not predicated on the proportion of professionals who advocate that technology. New, state-of-the-art techniques may be legally mandated even when the preponderance of professionals have not yet adopted the technology.88

B. Costliness

Two major cost issues associated with CV should be considered: (A) the feasibility of CV, and (B) the social costs of CV.

1. Feasibility

If all non-trivial nonuse values are awardable, then CV might be required for almost every natural resource damage assessment.89 Yet costs can be prohibitive for all but the most massive spills, since a CV study should be performed by an independent survey research organization90 and requires pretesting, a sample size of at least one thousand, and statistical analysis.91 Consequently, even for massive spills, trustees may opt for quick and dirty assessments, not only to avoid the expense of a CV survey but also to expedite settlement92 and restoration.93
[
Previous Page | Top | Next Page ]


86Krutilla, supra note 29, at 778 (cv not favored by most economists).
87However, all of the experiments cited herein which challenge the validity or reliability of CV fail to meet NOAA guidelines. NOAA, supra note 8, at 73.
88"Reasonable prudence may require a standard of practice which is higher than that exercised by the relevant professional community." Gates v. Jensen, 92 Wash.2d 246, 595 P.2d 919 (1979).
89Swords, supra note 22, at 1366 (arguing that even when CV is employed, it should be limited to those values which are not determinable by more reliable methods of damage assessment).
90NOAA, supra note 8, at 37.
91Because there is always a risk of not recouping all assessment costs, a useful heuristic is that Cv should not be conducted if its cost is likely to exceed estimated nonuse values. Note, supra note 1, at 1982. Furthermore, OPA requires that the benefits of a damage assessment outweigh its costs, and that the damage estimates exceed the costs of assessment. 43 C.F.R. 11.14(ee) (1992).
92Thomas A. Campbell, Natural Resource Damage Assessments: A Glance Backward and a Look Forward (Symposium Issue. Oil Pollution Act Rulemaking), 45 BAYLOR L.R. 221, 230 (1993).
93NOAA, supra note 8, at 44.


Prior to deciding whether to conduct a full-scale CV survey, one or more of the following screening techniques may be employed: (1) A small sample94 can be surveyed with protocols that minimize costs, and only if the resultant aggregate CV is at least twice as large as the estimated full survey costs would a full survey be recommended. (2) A CV study may only be performed if its cost will be less than the product of the number of households in the affected population multiplied by a predetermined figure such as $5. (3) Feasibility can be appraised from the damages and survey costs of similar CV studies.95

This feasibility problem can also be ameliorated through standardized damage relief determination or "benefits transfer."96 The federal government could finance benchmark97 or reference studies of common pollution scenarios, supervised by expert panels.98 Although expensive, such a project would be a worthwhile investment, because meta-analysis99 of such CV findings could lead to standardization of nonuse value measurement.100 The existence and option values of most spills might then be obtained from assessment tables,101 computerized multiple regression models, or compensation formulas.102 However, before applying benefits transfer, several factors should be
[
Previous Page | Top | Next Page ]


94Small pretest surveys and focus groups are subject to self-selection bias. Carson, supra note 45, at 2.3, 16. Nevertheless, probability sampling is unnecessary so long as population heterogeneity is considered. NOAA, supra note 8, at 86.
95Id. at 44.
96"Benefits transfer" is the use of applicable studies in new settings. Augustyniak, supra note 43, at 401 n.46.
97NOAA, supra note 8, at 61.
98NOAA, supra note 58, at 23.
99See generally Richard G. Walsh et al., Benefits Transfer of Outdoor Recreation Demand Studies: 19681988, 28 WATER RESOURCES RESEARCH 707 (1992) (cited in Carson, supra note 45, at 2.4, 22 n.26) (meta-analyzing 129 CVs of recreational assets).
100The CV of a spill might then be determined by asking Rs, "Would you pay (much more, more, about the same, less, much less) to prevent this spill than you would to prevent Standard Spill A?" "Would you pay an amount to avoid this spill that is between the amounts you would pay to avoid Standard Spill B and Standard Spill C? If so, is the amount (much closer to B than C, closer to B than C, halfway between B and C, closer to C than B, much closer to C than B)?" Id. at 38.
101NOAA, supra note 8, at 22.
102A compensation formula is best for small losses. Id. at 61.


considered: (1) the comparability of the (a) relevant population, (b) natural resources, (c) change in quantity of resources, and (d) change in quality of resources; and (2) the quality of the previous study.103 When there is no qualified, comparable, preexisting study, CV is recommended over benefits transfer, unless CV is unfeasible.104

2. Social Costs

Since CV is expensive, its costs must be passed on in the form of higher consumer prices. Furthermore, the social cost of CV is directly related to its unreliability.105

i. Overestimation

When CV studies overestimate damages, they inflate the contingent liability faced by PRPs. As a result, insurance coverage for potential OPA and CERCLA claims will be more expensive. Furthermore, because some PRPs will decide to litigate rather than pay an exorbitant assessment,106 inaccurate CVs will generate wasteful litigation costs. All of these costs will be transferred to consumers.

ii. Underestimation

On the other hand, when CV underestimates damages, the public must pay the difference in the form of a poorer environment, and polluters have less incentive to curb environmental damage.107 In other words, minimization of the social costs of CV requires maximization of CV validity and reliability.
[
Previous Page | Top | Next Page ]


103Id. at 87.
104As an alternative to computerized or tabular estimations, RPs could request a Comprehensive Damage Assessment, including CV-but only at their own expense. Id. at 2223.
105The consequences of under or overestimates of damages are known as "misallocation effects." Id. at 49.
106Note, supra note 1, at 1992.
107NOAA, supra note 8, at 53.


C. Accuracy

Whether CV systematically overvalues or undervalues natural resources is uncertain, because of the unworkability of external criterionrelated CV validation.108 However, accuracy is undoubtedly compromised if the chronology of recovery or restoration is not considered,109 or if the scenario is not effectively identical to the actual damage scene. CV inaccuracy can also be addressed through conservative assessment procedures, which tend to underestimate CV and make it more difficult for defendant RPs to overcome the rebuttal presumption that damage assessments are accurate.

D. Representativeness110

Two criticisms of CV studies are that (1) the market size is excessively large111 and (2) sampling is biased.112

1. Market Size

A critic might contend that a person in Florida who is unlikely to ever visit Alaska would not suffer a measurable loss from an incident like the ExxonValdez oil spill, and is not a part of the relevant population. Granted, the option
[
Previous Page | Top | Next Page ]


108Nevertheless, other validity and reliability checks can be persuasive. For example, the Exxon Valdez study conducted pilot studies and a "tracking" study which demonstrated that mean WTP was stable among several replications over the course of one year. Carson, supra note 45, at 1.5, 11 & 5.12, 119. See also supra note 79.
109Diminution of nonuse value, especially bequeathment value, may be negligible where recovery or restoration is rapid. See A. Myron Freeman, Nonuse Values in NRDA, in VALUING NATURAL ASSETS 46, 5253 (Raymond J. Kopp & V. Kerry Smith eds., 1992) (nonuse values trivial for resources which recover quickly and completely). Nevertheless, even transitory natural resource damage incidents may involve the loss of critically important individuals within a species and cause multigenerational or steady state (i.e., postrecovery) effects. In addition, interim losses may be significant, and there may be longterm esthetic and psychological effects. Therefore, rapid reclamation should not preclude CV. NOAA, supra note 8, at 50.
110See Christine Cartwright, Note, Natural Resource Damage Assessment: The Exxon Valdez Oil Spill and its Implications, 17 RUTGERS COMPUTER& TECH. L.J. 451, 465 (1991).
111Raymond J. Kopp et al., Natural Resource Damages: The Economics Have Shifted After Ohio v. Department of Interior, 20 ENVTL. . L. REP. (Envt. L. Inst.) 10, 130 (1990) (the relevant population experiencing measurable loss for environmental damages is much smaller than population included in most CV studies).

112Swords, supra note 22, at 136 l (suggesting that even random selection is biased, because respondents, like juries, should be subjected to voir dire).


and quasi-option values of a natural resource may have some positive association with the geographical proximity of the consumer to the damage location. Even vicarious, intertemporal, and bequeathment values should have at least some inverse relationship to the distance of the consumer from the pollution site. Nevertheless, future circumstances cannot be foretold, so nonuse values should apply to people who currently have no plans to ever visit the pollution area, and even to people previously unaware of the natural resource's existence.113

Furthermore, geographical factors should be measured, because the relationship between geographical proximity and natural resource valuation is often uncertain. Possible independent variables representing the geographical proximity of respondent's primary residence might be: (1) distance from the relevant site, (2) county, (3) state, or (4) region. If such factors are included in the analysis, it should be possible to explain the nature of the relationship between geographical proximity and existence value.

2. Sampling Bias

Sampling must be random or probabilistic within the relevant population in order to minimize self-selection114 and other systematic biases. However, even when random sampling is applied, nonresponse bias must be considered. Therefore, a followup analysis of nonrespondents is recommended in order to assess differences from respondents.115 To further minimize sampling bias, cluster or stratified sampling should be combined with probabilistic sampling.116
[
Previous Page | Top | Next Page ]


113However, "[u]ndersampling and even zero sampling ... may be appropriate" for a population subgroup whose cv is predicted to be de minimis. NOAA, supra note 58, at 16.
114Self-selection bias is also minimized by inperson interviewing, since attrition is negligible after interviewing commences. NOAA, supra note 8, at 7576.
115Id. at 75.
116NOAA, supra note 58, at 49. The Exxon Valdez study employed standardized multistage area probability sampling techniques. Carson, supra note 45, at 1.4, l0.


E. Hypotheticality117

Respondents' expressed attitudes do not necessarily reflect their actual behavior.118 Since no actual monetary exchange is involved, there is no risk to a survey participant in providing an excessively high valuation. There is "no cost to being wrong, land] no incentive to undertake the mental effort to be accurate."119 Nevertheless, there is substantial indirect evidence that the preponderance of CV responses are reasonably rational. For example, CV respondents consistently provide much lower figures for WTP than WTA for natural resources.120 Similarly, consumers generally spend less for a resource than they are willing to pay, since most resources have surplus value. Furthermore, if Rs expect their answers to have few if any consequences to their lifestyles or environmental resources, then undervaluation is as likely as overvaluation.

In any event, the hypotheticality problem can be mitigated by: (1) Realism. In order to obtain more informed opinion, provide respondents with unbiased data about natural resources and their economic significance, both as a prelude to the questioning and by integrating information into the interrogatories. (2) Budgeting. Construct the survey as a budgeting exercise, so that R must allocate his actual financial resources for all of his essential needs,
[
Previous Page | Top | Next Page ]


117See Frank B. Cross, Restoring Restoration For Natural Resource Damages, 24 U. TOL. L. REV. 319, 329 (1993)
118Kalle Seip & Jon Strand, Willingness to Pay for [Environmental Goods in Norway: A Contingent Valuation Study with Real Payment (SAF Center for Applied Research, Economics Dept., University of Oslo) (cited in NOAA, supra note 58, at 7) (selfreported WTP for membership in Norwegian environmental organization significantly higher than actual duespaying behavior). Put cf. John W. Duffield & David A. Patterson, Field Testing Existence Values: An Instream Flow Trust Fund for Montana Rivers (American Economic Association, New Orleans) (1991) (actual and hypothetical WTP Montana Nature Conservancy to protect rare fish small enough that discounted CV could be used for conservative WTP estimate).
119A. Myrick Freeman, Approaches to Measuring Public Goods Demands, 61 AM. J. AG. ECON. 915, 916 (1979)
120Nancy E. Bockstael & Kenneth E. McConnell, Calculating Equivalent and Compensating Variation for Natural Resource Facilities, 56 LAND ECONOMICS 56, 61 (1980).


for environmental concerns in general, and for the specific environmental resources which the survey is seeking to measure.121 (3) Expert panels composed of natural resource economists122 with diverse perspectives, could be employed to interpret and augment the results of consumer surveys. Alternatively, where largescale consumer surveys are unfeasible and the relevant parties are amenable, expert panels could substitute for public polling.

F. Replicability

When multiple CV trials are administered, the findings are sometimes inconsistent. At least two factors explain this unreliability: (1 ) Pollution incident chronology. A sensational incident that receives a lot of journalistic attention may be more highly valued at the peak of its exposure than a year later. Valuation may also be related to the extent of restoration. Therefore, a site that was recently damaged may be more highly valued than the same site after it has recovered substantially. (2) Administration methodology. If (a) a sufficiently large sample123 of respondents is randomly selected from the relevant population, (b) the same wellconstructed survey items are utilized during each replication, and (c) the interviewers are skilled, unbiased questioners, then 95% of CV results should deviate from the true population CV by less than three percent.

G. Survey Effects124

The sites of natural resource damages are frequently unknown to the respondents prior to the CV survey. Consider the poser, "If a tree falls in a
[
Previous Page | Top | Next Page ]


121Aconsumers' budget is most conveniently formulated on a monthly or annual basis. The amount which an individual is willing to spend monthly or annually to protect against natural resource damages might be expressed as a percentage of gross, beforetax income. Furthermore, a resource protection expenditure model should incorporate consumers' willingness to make longterm investments against future environmental degradation.
122Cicchetti & Peck, supra note 81, at 8.
123A typical target sample size is 1000.
124See Cross, supra note 117, at 330.


forest but no one hears it, does it make a sound?" Some critics argue that a specific natural resource is of no value to individuals who are not aware of its existence, and that CV surveys generate natural resource values where none would otherwise existthat the survey questions themselves generate resource values.

However, the fact that someone is unaware of a resource or that he makes no direct monetary outlays for environmental purposes does not indicate that relevant natural resources are of no value to him. The tree does make a sound, and even if no one hears it, the repercussions are unpredictable and infinite. Although the option value may be larger when R is already familiar with the natural resource, the existence value reflects the value of its nonuse, and nonuse does not require familiarity. A CV survey merely helps some respondents to realize and articulate the value of environmental assets.

Another argument is that the extent of information provided to Rs is directly related to their CV responses. However, reliability and validity are also directly related to the amount of scenario detail presented to Rs. Furthermore, Rs should receive sufficient information to minimize any biases or doubts about the hypothetical incident.125 Finally, there are limits to the ability of Rs to internalize the conveyed information.126

H. Media Effects127

A similar argument is that CV is directly related to respondents' familiarity with the pollution incident, or similar incidents.128 Thus, media exposure of a
[
Previous Page | Top | Next Page ]


125NOAA, supra note 58, at 15
126Id. at 15.
127See NOAA, supra note 8, at 10.
128The mean difference between pre- and post-test knowledge is inversely related to the extent of media attention given to the pollution incident. If there is little or no media exposure, the unsurveyed segment of the population may have no knowledge of the incident. Consequently, NOAA has no firm position on whether to use pretest or posttest knowledge levels when extrapolating CV to the entire relevant population. NOAA, supra note 8, at 43.


discharge incident may inflate CV responses. However, any such "controversy effects are diminished by the long lead time between the incident and the CV study.129

Additionally, CV may provide an incentive for corporations with potential pollution liability to keep the public in the dark through surreptitiousness and misinformation.130 Ironically, there is no better argument for CV. CV provides a valuable public service by educating citizens and counteracting secretiveness among polluters. Additionally, the fact that greater media attention to environmental pollution leads to higher valuation suggests that natural resources are undervalued.

I. Interviewer Effects

Responses may be affected by the presence of interviewers, who may subtly transmit their preferences to respondents. If Rs sense that certain responses are favored, they are likely to oblige the questioner.131 Consequently, interviewers should be completely independent from the litigation parties,132 or even blinded as to the purposes of the study,133 and should assure Rs of confidentiality.134 Smallscale telephonic surveys can be conducted using interviewers who are unaware of survey purposes, in order to determine whether inperson interviewing causes systematically biased CV ratings. Additionally, interviewer effects can be pretested, so that interviewers with significantly disparate results can be excluded,135 and during survey
[
Previous Page | Top | Next Page ]


129Id. at 72. The lag between an environmental discharge and its CV survey will typically exceed one year.
130Existence valuation creates a perverse incentive to keep the public ignorant about the natural world. As a result, many species like the snail darter may never be brought to public attention. Cross, supra note 42, at, at 291 (cited in Pilar Okun, The Revised National Resource Damage Assessment Rules: Computation for Compensation and Restoration, 70 WASH. U.L.Q. 959 (1992)).
131This phenomenon is known as "social desirability bias." NOAA, supra note 8, at 77.
132Id. at 8.
133Carson, supra note 45, at 2.6, 24..
134NOAA, supra note 8, at 86.
135NOAA, supra note 58, at 30.


administration, supervisors should perform quality control checks on interviewee samples by telephone.136 A far more costly approach to quality assurance is for both sides to conduct independent, unbiased CV surveys.

J. Survey Bias

Survey questions must be neutrally worded, so that the respondents are not subjected to tendentious influences. Also, where semantical effects are uncertain, the option which is more likely to generate a conservative contingent valuation should be chosen.137 For example, a WTP scenario, wherein R is asked how much he would be willing to pay to prevent an environmental disaster, is recommended instead of a WTA scenario,138 wherein R is asked how much money he would accept for allowing an environmental disaster to occur, since WTP elicits smaller CV amounts than WTA.139

K. Strategic Bias

People may tailor their answers to correspond with their personal political philosophies or circumstances.140 As a result, a free rider might overstate values, hoping that his opinion might lead to greater expenditures on natural resources.141 On the other hand, a wise use advocate might value all nonuse values at zero. Therefore, it is crucial that the actual purpose of the CV survey not be divulged to respondents. Most importantly, respondents should not be
[
Previous Page | Top | Next Page ]


136Carson, supra note 45, at 4.7, 7374. 137NOAA, supra note 8, at 11.
138Nevertheless, WTA would be preferable, since it more accurately reflects the role of trustees as agents for public property interests than WTP. Another cv measure is the Marshallian surplus, a consumer welfare measure which is generally greater than WTP but less than WTA. NOAA allows any of these three for measuring natural resource damages, but only recommends WTP for CV. Id. at 87.
139Robert G. Cummings et al., VALUING ENVIRONMENTAL GOODS: AN ASSESSMENT OF THE CONTINGENT VALUATION METHOD, 3536, 44 (1986).
140For instance, a respondent may assume that he will never have another opportunity to contribute to an environmental cleanup, and therefore decide to overspend. NOAA, supra note 58, at 58.
141Paul A. Samuelson, The Pure Theory of Public Expenditures, 36 REV. ECON. & STAT. 386,38788 (1954).


aware that litigation may be involved, or who might be liable for damages.142 Strategic bias can also be reduced through a referendum format.143

L. Embedding144

The fact that some respondents fail to proportionately value different quantities of the same resource is often cited as evidence of the unreliability of CV.145 Sequential embedding146 is exemplified by an interviewee who places a roughly equivalent value on 200,000 birds as his earlier valuation of 2,000 birds.147 Scalar embedding was demonstrated in a 1991 experiment, where three groups were each asked to place a value on protecting a different wilderness area. A fourth group was asked to place a value on protecting all three areas combined. The valuations were found to be roughly equivalent in all four groups.148 To some extent, embedding may reflect an "it's the thought that counts" ideation, whereby a respondent is willing to donate a customary monetary sum for a philanthropic purpose, irrespective of the magnitude of the environmental problem.149 However, the principal cause of embedding is perceived diminishing marginal utility of environmentallybeneficial expenditures.150
[
Previous Page | Top | Next Page ]


142NOAA, supra note 8, at 9.
143Id. at 75.
144Daniel Kahneman & Jack L. Knetsch, Valuing Public Goods: The Purchase of Moral Satisfaction, 22 J. ENVTL. ECON. & MGMT. 57 (1992) (cited in Note, supra note 1, at 1989 ) (embedding as symbolic bias)+ See also NOAA, supra note 58, at 11 12 (embedding as an example of inconsistency in rational choice).
145Cross, supra note 117, at 330.
146Sequential embedding occurs when an R is asked to evaluate a series of goods. WTP declines with each successive item, because of substitution and income effects. On the other hand, WTA increases sequentially, since goods become more valuable with increasing scarcity. Assuming that WTA is uniformly higher than WTP, the closest available approximation to WTA is a single, nonsequential WTP elicitation. See Carson, supra note 45, at 2.4, 19.
147William H. Desvousges et al., Measuring Natural Resource Damages with Contingent Valuation: Tests of Validity and Reliability, in CONTINGENT VALUATION: A CRITICAL ASSESSMENT (Cambridge Economics, Inc.), 1523 (Apr. 2, 1992) (cited in Note, supra note 1, at 1988 n.47).
148Peter A. Diamond et al., Does Contingent Valuation Measure Preferences? Experimental Evidence, in CONTINGENT VALUATION, note 147 (cited in Note, supra note 1, at 1988 n.46).
149See Self-Reinforcement, infra p. 24, which elaborates on this symbolic bias or warm glow effect.
150NOAA, supra note 8, at 69.


Sequential embedding can be eliminated by eliciting only one commodity valuation from each R. In addition, scalar embedding can be reduced by clarifying with Rs that they are being asked to evaluate the specified site, not all damage sites or all environmental causes.151 Also, explanations for errant monetary responses can be elucidated through indepth followup inquiry. Finally, embeddinginduced anomalies can be reduced by eliminating spurious responses from the analysis.152

M. Self-reinforcement

One cause of embedding may be the "warm glow" of altruism.153 According to this theory, contributing a symbolic, affordable payment to a "good cause," irrespective of its magnitude, has a transactional value154 and is self-approbatory.155 Since CV tends not to rise proportionately with the scale of damages, the extent to which CV underestimates natural resource losses is directly related to the magnitude of the losses. Therefore, selfreinforcement is more problematic for smallscale damages than the typical largescale incidents for which CV surveys are feasible.

N. Centering

If respondents are instructed to select one of a set of values, they will tend to choose a figure in the middle of the range.156 This problem can be avoided by offering only one value for the R to accept or reject, or by using
[
Previous Page | Top | Next Page ]


151Because scalar embedding is an inclusiveness problem, it is also inversely related to the amount of contextual detail about the scenario provided to the respondent. See Carson, supra note 45, at 2.4, 21.
152Removal of outliers or extreme values from a statistical model is common in econometric analysis. The rationale is that deviant values are the effects of relatively rare phenomena which are not significant enough to demand explanation. See supra p. I I & note 67.
153Altruistic euphoria is strongest if R believes that his payment will go to a charitable organization. Carson, supra note 45, at 2.9, 30 n.30.
154NOAA, supra note 58, at 35.
155Id. at 26.
156Cicchetti & Peck, supra note 81, at 9.


open-ended valuation queries. A double-bounded dichotomous choice mechanism157 is also unlikely to have significant centering effects.

O. Ordinality

The sequence of questioning may affect CV results.158 For example, one survey asked respondents to place a value on natural resources at several well known locations. When the value of improving visibility in the Grand Canyon was the first item listed, its valuation was three times larger than when it was listed third.159 To minimize ordering effects, (1) questions should be pretested,160 or (2) subgroups of the survey sample should be questioned in different sequences in order to control for ordinality.

P. Additivity

There are two types of additivity with which CV users must contend: (1) respondent additivity and (2) aggregate additivity.

1. Respondent Additivity

The value placed on one natural resource item, if projected across all similar items, may amount to an unrealistic proportion of R's discretionary income. For instance, an interviewee might be willing to pay five dollars annually to prevent the extinction of a species. This could be construed as a willingness to pay $250,000 annually, since there are 50,000 endangered speciesfar beyond the means of a person with a $50,000 salary.161

2. Aggregate Additivity

The aggregate contingent value of a natural resource may a multibillion dollar figure. Such large contingent valuations may appear to vastly exceed the
[
Previous Page | Top | Next Page ]


157See infra note 193.
158Charles C. Harris et al., Improving the Contingent Valuation Method: A Psychological Perspective, 17 J. ENV. ECON. & MGT. 213, 214 (1989).
159GEORGE S. TOLLEY & A. RANDALL, Establishing and Valuing the Effects of Improved Visibility in the Eastern United States (report to the EPA) (1983).
160Cf. Cicchetti & Peck, supra note 81, at 9. 161Note, supra note 1, at 1987 (1992).


true value of the natural resource, or may vastly exceed what responsible parties (RPs) are able to pay. For example, a series of contingent valuations of whooping cranes generated mean annual willingness to pay values of $5 $149. The aggregate nationwide values, ranging from 0.5 to 14.9 billion dollars,162 all vastly exceed the total FWS endangered species budget of $30 million.

A carefully designed CV survey should minimize such additivity problems. Additivity effects simply demonstrate that many consumers are unaccustomed to natural resource valuation, and need more information. One method to reduce additivity as well as embedding is the "topdown disaggregation" survey approach (TDD).163 Most CV surveys are "single focus," asking R to place a value on one natural resource scenario in isolation from all substitutes and budgetary considerations.164 TDD, on the other hand, provides a list of social (and environmental) programs, and asks the respondent to assign a value to each program. If the list of social programs is combined with a list of common budgetary items, it should be possible for R to compose a reasonable facsimile of his actual budget, and to keep his contingent valuation within the budget.165

Q. Dispersion

The large standard deviation or variance of contingent valuations relative to their magnitude may suggest unreliability, but is primarily the result of each R's unique perspective on the extent to which natural resource protection should be prioritized, and the amounts that people are willing to spend for
[
Previous Page | Top | Next Page ]


162James M. Bowker & John R. Stoll, Use of Dichotomous Choice Nonmarket Methods to Value the Whooping Crane Resource, 70 AM. J. AGRIC. ECON. 372, 380 (1988).
163Note, supra note 1, at 1988. Topdown approaches are not recommended by NOAA, because CV commodity descriptions generally lack specificity relative to other potential budgetary choices. NOAA, supra note 8, at 69.
164In a study of marine oil spills in Alaska, the contingent valuation was 290 times larger with a single focus survey than with a TDD. Michael A. Kemp & Christopher Maxwell, Exploring a Budget Context for Contingent Valuation Estimates, in CONTINGENT VALUATION, supra note 147 (cited in Note, supra note 1, at 1988 n.49).
165The employment of budgeting within contingent valuation might be termed "contingent budgeting."


environmental protection vary widely. Natural resource CV tends to have a bimodal distribution, and the proportion of respondents who are unwilling to pay anything is often as high as forty percent.166 Such rightskewedness,167 along with a high coefficient of variation,168 is also found when market goods are contingently valued.169 Some of the variability can be reduced by asking respondents to give a precise figure, rounded to the nearest penny instead of the nearest dollar. For example, many respondents who indicate a zero valuation are assuming that the smallest expenditure figure is one dollar. If respondents are aware that they can indicate as little as a penny, many more will provide a nonzero figure. On the other hand, extremely large figures can be reduced with TDD or a budgetary approach.

R. Income Effects170

Income is positively correlated with willingness to pay. A poor person may be unable to afford any environmental protection, and may never be able to visit environmentally valuable areas. On the other hand, a wealthy person who visits natural resourcerich areas frequently may be willing and able to pay much larger sums for their protection than any middleclass person. Three suggestions for handling income effects are: (1) use median WTP as a baseline or floor value, (2) control WTP statistically for income, and (3) use willingness to sell or WTA. Although not currently recommended by NOAA, WTA is not biased by the respondent's income, and may more accurately reflect longrange existence values than WTP.
[
Previous Page | Top | Next Page ]


166P.A. Diamond & J.A. Hausman, On Contingent Valuation Measurement of Nonuse Values, in CONTINGENT VALUATION, supra note 147 (cited in NOAA, supra note 58, at 17).
167Also known as positive skewness or righttailedness.
168The coefficient of variation is computed as the ratio of the standard deviation to the mean.
169Cf. NOAA, supra note 8, at 54.
170See Cross, supra note 42, at 336.


S. Nonpristineness

One erroneous contention is that CV can only be applied to pristine environmental scenarios. Actually, CV can be reliably measured for settings which are partially degraded or developed, if baseline conditions are delineated in sufficient detail.171

T. Ex post facto Questioning

Ex ante measurement of CV minimizes bias. However, the preponderance of CV surveys will inevitably be performed after the environmental damage has occurred. Furthermore, because the upward bias of ex post facto data is manageably small and proportional to the size of the pollution incident, NOM considers ex post facto CV studies acceptable.172

U. Double Counting

If other assessment methods are used to supplement CV, there is a danger that the total value of the natural resources may be overstated. Double counting is particularly likely if hedonic pricing or travel costing are combined with CV,173 or if use and nonuse measures are not clearly distinguished. Therefore, where existence and option values are significant, CV should measure all use and nonuse values for the surveyed population. If additional use value measurements are employed, then the nonuse value components of CV may be ascertainable. I However, aggregate CV may need to be augmented by the supernormal use values of parties who are not adequately represented by the survey sample.174
[
Previous Page | Top | Next Page ]


171NOAA, supra note 8, at 70.
172Id. at 7273.
173Cicchetti & Peck, supra note 81, at 9.
174Use values especially among subsistence users, fishermen,? and loggers in close proximity to the damaged area-are likely to be underestimated in a CV survey.


VI. RECOMMENDATIONS

Methodological research is urgently needed to improve CV techniques.175 In the meantime, trustees for natural resource damage claims should consider the following CV guidelines176: (1) Sufficient time should elapse after the discharge incident to allow Rs to regard restoration as plausible.177 (2) Where uncertainty exists, the more conservative CV approach is recommended.178 (3) Advance approval of the CV design should be obtained from all litigants.179 (4) Environmental injury scenarios should accurately and understandably depict actual damages. (5) The restoration or prevention program should be chronologically detailed and credible. (6) The choice mechanism and payment vehicle should be realistic.180 (7) CV discriminant validity should be pretested using scenarios which differ by the scope of damages. For example, each of three split samples181 or focus groups182 can be provided with a different scenario, and group results can be compared using pairwise statistical methods.183 The scenarios are valid if less than 95% of Rs indicate that there are meaningful value differences among the scenarios.184 (8) Photographic depictions of the damage scenario185 should be informative but not overly
[
Previous Page | Top | Next Page ]


175NOAA, supra note 8, at 74.
176Many of these suggestions come from the NOAA "blue ribbon" CV panel, which was cochaired by Nobel Laureates Kenneth Arrow and Robert Solow.
177Cf. NOAA, supra note 8, at 72.
178To counter any arguments that CV overestimates damages, NOAA proposes a default calibration factor of 50% to be multiplied against WTP when calculating aggregate CV. Id. at 7374.
179NOAA, supra note 58, at 3536.
180NOAA, supra note 8, at 65. For WTP, the payment vehicle may be taxes, price increases, or bids (see Id. at 88). In the Exxon Valdez study, an oil price payment vehicle was found to produce significantly higher WTP than a tax payment vehicle. Carson, supra note 45, at 5.12, 117 n.98.
181Sample group sizes need be no larger than 60.
182A typical focus group is composed of eight or twelve participants whose interactions are observed by researchers from behind oneway mirrors and videotaped for future reference. Carson, supra note 45, at 2.3, 15.
183Scope tests should use dichotomous variables and analysis of covariance. NOAA, supra note 8, at 39.
184Id. at 40.
185NOAA, supra note 58, at 32.


graphic.186 (9) The value elicitation question should measure WTP to prevent described injuries, or (as a less desirable alternative) WTP to restore injured resources to baseline or comparable conditions. (10) R bias should be assessed, especially concerning such issues as (a) the fairness of financial burdens, (b) program feasibility and efficiency, and (c) tax aversion.187 Subsequently, during sample screening or outlier analysis, Rs with unreasonable responses-such as those who express an unwillingness to pay for an environmental cleanup of any size or who invariantly provide unrealistically large values-may be excluded from CV analysis.188 (11) Undamaged substitutes to the relevant natural resource commodities should be suggested prior to value elicitation.189 (12) Rs should be reminded of their budgetary constraints.190 (13) A referendum,191 voting,192 or dichotomous choice mechanism193 should be employed. (14) Three subsamples should each be provided with a different, predetermined CV figure.194 (15) If a no-answer option is exercised, then the respondent should be nondirectively requested to explain his choice-for example, (a) indifference, (b)
[
Previous Page | Top | Next Page ]


186A photo of a dead, oil-covered sea otter may be realistic, but more neutral, less provocative illustrations are recommended. NOAA, supra note 8, at 4142.
187NOAA, supra note 58, at 22.
188NOAA, supra note 8, at 87.
189NOAA, supra note 58, at 32. Substitutes should be presented in the form of a moderatelysized list of alternative public goods as similar as possible to the damaged resources. The list should also contain the annualized or lump sum price of each commodity. Id. at 59.
190NOAA, supra note 8, at 86.
191The referendum approach is preferable to an openended WTP format because it: (l) simulates actual voting behavior, thereby reducing the likelihood of strategic behavior (Id. at 66); (2) simulates market behavior where the buyer is oblivious to his maximum WTP and bases his decision to purchase on the specified price; and (3) enhances scenario credibility, because Rs are more likely to believe that the program will actually be instituted if they vote for it. Id. at 3536.
192A referendum or voting format can be made more realistic by arranging for Rs to deposit ballots in a sealed box or enter their choice on a computer mimicking the anonymity of a voting booth. NOAA, supra note 58, at 48. NOAA currently is undecided as to whether a no opinion or no answer option should be included. NOAA, supra note 8, at 76.
193An example of a doublehounded dichotomous choice WTP elicitation framework is as follows: (l) If R votes yes to $30, then ask whether R would be willing to pay $60. (2) On the other hand, if R votes no to $30, ask whether R would be willing to pay $10. Carson, supra note 45, at 5.5, 9192. Such intervalcensored survival data (e.g., WTP < $10, $10 <= WTP <= $30, $30 <= WTP <= $60, WTP > $60) is more biased than a singlevote format, but enhances the accuracy of WTP estimates. Id. at 94.
194The subsamples must be econometrically combined in order to determine the mean or median CV.


need for more time or information, (c) desire for a different decisionmaking format, or (d) tedium.195 (16) A "lump sum" value should be utilized, as opposed to an annual installment approach.196 (17) Interim197and steady state198 losses should be differentiated, and Rs should be tested to ensure that they are aware of the differences.199 (18) After value elicitation, Rs should be reminded of alternative expenditure possibilities and given an opportunity to change their bids or votes.200 (19) Followup questions should be employed whenever helpful for response clarification.201 (20) At the end of the interview, Rs should be nondirectively questioned concerning the credibility of (a) the injury description, including (i) interim and (ii) permanent losses; (b) the prevention (or restoration) program and its chronology; and (c) the referendum.202 (21) The final report should include (a) the rationale for each procedural decision-especially the choice between a prevention or restoration scenario,203 (b) the environmental scenario, (c) sampling methodology, (d) response rates, (d) all questions in
[
Previous Page | Top | Next Page ]


195NOAA, supra note 58, at 33.
196The lump sum approach is more conservative and avoids the difficulties of multiyear annuities, such as estimation of subjective discount rates and present values. NOAA, supra note 8, at 41. In addition, the lump sum approach avoids recontracting problems i.e., unwillingness to make longterm commitments, or expectations that terms on installment payments can be modified. Carson, supra note 45, at 2.8, 30. However, for longlasting damages, where CV must be measured longitudinally, discounting is preferable to temporal averaging. NOAA, supra note 58, at 33. U.S. Treasury note rates should be used for discounting. NOAA, supra note 8, at 79.
197A CV survey can employ three recovery scenarios: (1) immediate or instantaneous restoration, which is a hypothetical construct; (2) accelerated restoration, which requires restoration technology; and (3) natural restoration or nonintervention. Where restoration technology is employed, interim losses accrue during the interval between immediate and accelerated restoration. Otherwise, interim losses accrue between immediate and natural restoration. However, there are problems with this method, because many Rs find immediate restoration implausible, or feel that losses should be paid by polluters. NOAA, supra note 58, at 5253. Interim losses are best measured immediately after restoration, or periodically if restoration is a longterm process. Id. at 5455.
198Id. Steady state or permanent losses should be assessed when restoration is complete.
199Id. at 35.
200NOAA, supra note 8, at 86.
201For example, "Why did you vote yes?" (1) worthwhile, (2) don't know, or (3) oil companies should pay. NOAA, supra note 58, at 33. Followup questioning is especially important for openended CV questions when unusually low or high responses are received. NOAA, supra note 8, at 65.
202NOAA, supra note 58, at 57.
203NOAA, supra note 8, at 8586.


sequence, and (d) any other communications with respondents.204 (22) The party contracting for a CV study should assume that it will carry the burden of proving the CV's reliability.205

VII. RESEARCH SUGGESTIONS

The following are some ideas for future CV investigations: (1) If sufficient governmental research funding can be obtained for benefits transfer projects, then ex ante CV studies should be performed in order to develop a set of CV benchmarks. (2) If benefits transfer is to be fully achieved, then CV databases should be regularly metaanalyzed. (3) CV methodology-especially the wording and sequence of interview interrogatories-should be standardized, so that the results of any survey can be readily compared with other surveys. (4) CV statistical models should be continuously refined so that they will be applicable to an increasingly broad range of environmental damage situations. For example, a multivariate regression model might incorporate (a) scope variable(s), which measure the magnitude of the spill; (b) dichotomous or dummy variables for pollutant type, such as oil or acetone; (c) and binary or dummy variables for various environmental types, such as seashore, river, or desert. Ultimately, statistical models should be able to incorporate all significant environmental factors, including unique biomes and each endangered or threatened species. (5) Existing CV studies should be subjected to outlier analysis so that a wider range of contingent valuation responses can be explained. (6) An historical econometric analysis of public land valuation in relation to population growth, natural resource depletion, and other factors might assist in predicting future natural resource values. Such future valuation might serve as a surrogate measure or baseline calibration of nonuse values. (7)
[
Previous Page | Top | Next Page ]


204NOAA, supra note 58, at 30.
205Id. at 36.


TDD or budgetary approaches should be refined to maximize economic detail within the constraints of R patience and tolerance. (8) CV reliability might be enhanced through more detailed itemization of current as opposed to the current approach which lumps all relevant natural resources into one sum.206 (9) mean and median approaches to CV should be compared, so that conditional median or mean CV data might be utilized in aggregate CV computation. (10) A semi parametric estimator should be developed for doublebounded interval survival data.207 (11) Income and substitution elasticities should be used to extrapolate from WTP.208

VIII. CONCLUSION

Although as yet there is no legal precedent for the application of contingent valuation to natural resource damage awards, a safe assumption is that contingent valuation surveys, which closely follow the advice of the NOAA panel study are likely to be admissible in court Factfinders are also more likely to adopt CV findings if the data is thoroughly analyzed statistically, and if the results are presented so as to be understandable and salient to lay audiences. In order to maximize benefits transfer, accumulated CV data should be meta analyzed, and mathematical models should be developed which can estimate the contingent value of environmental damages according to magnitude, pollutant ecosystem type, and affected biota.
[
Previous Page | Top ]


206Cf. NOAA, supra note 58, at 39.
207Carson, supra note 45, at 5.13], 122.
208Id.