Memorandum to the Science Advisory Board Environmental Economics Advisory Committee Additional Explication of Methods for Measuring Non-use Values: a Contingent Valuation Study of Groundwater Cleanup


MEMORANDUM TO THE SCIENCE ADVISORY BOARD
ENVIRONMENTAL ECONOMICS ADVISORY COMMITTEE

ADDITIONAL EXPLICATION OF
METHODS FOR MEASURING NON-USE VALUES:
A CONTINGENT VALUATION STUDY OF
GROUNDWATER CLEANUP

Gary H. McClelland
William D. Schulze

Jeffrey K. Lazo

Center for Economic Analysis
University of Colorado
Boulder, CO 80309

June 29, 1993

-------
DISCLAIMER

Although prepared under EPA Cooperative Agreement CR815183, this report has
neither been reviewed nor approved by the U.S. Environmental Protection Agency for
publication as an EPA report. The contents do not necessarily reflect the views or policies of
the U.S. Environmental Protection Agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.

-------
TABLE OF CONTENTS

1.	Introduction"	1

2.	Commodity Definition	3
2A. A Theoretical and Empirical Analysis of Use Values	7
2B. A Theoretical and Empirical Look at Non-Use Values	16

Theoretical Issues:	16

Psychological Issues:	21

Further Evidence on Non-Use Values	24

Pre-Testing of the Commodity	30
The NOAA Panel and Procedures Used In The Groundwater Study. 37

3.	Issues in the Use of the Benefit Estimates	40
3A. Treatment of Non-Respondents	41
3B. Sample Bias	43
3C. Extent of Market	44
3D. Estimation of National Benefits	46
3E. Order of Magnitude for Benefits and Costs	48

4.	Conclusions	49

5.	References	51

6.	Appendices	55
6A. POST-TEST QUESTION ON TEMPORAL ASPECTS OF VALUATION	55
6B. VIEWGRAPHS USED BY PROFESSOR WILLIAM D. SCHULZE	56
6C. VIEWGRAPHS USED BY PROFESSOR GARY H. MCCLELLAND	57

-------
LIST OF FIGURES

FIGURE 1 EFFECT OF PROVISION OF SUBSTITUTE COMMODITIES	8
FIGURE 2 PREDICTED WTP AS FUNCTION OF DEPENDENCE ON GROUNDWATER 14

FIGURE 3 DERIVED WATER DEMAND CURVE AND ELASTICITIES	15

FIGURE 4 WTP AS FUNCTION OF COMPLETION OF CLEANUP	29

FIGURE 5 COMMUNITY SIZE COMPARED TO POLITICAL UNIT	45

FIGURE 6 COMMUNITY SIZE AS FUNCTION OF RADIUS OF COMMUNITY	46

FIGURE 7 NONUSE BENEFITS AS FUNCTION OF EXTENT OF MARKET	48

LIST OF TABLES

TABLE 1 RANKINGS OF OPTIONS	11

TABLE 2 SELF-REPORTED EFFECTS OF CONTEXT	13

TABLE 3 OPTION RATINGS RELATED TO BENEFIT COMPONENTS	22

TABLE 4 COMPONENT ALLOCATION OF TOTAL WTP	23

TABLE 5 TEMPORAL ASPECTS OF CLEANUP VALUATION	28

TABLE 6 IMPLICIT RATES OF TIME PREFERENCE	29

TABLE 7 MEAN REDUCED WTP AS A FUNCTION OF TIME TO RETURN	42

TABLE 8 NONUSE VALUE BENEFITS AS A FUNCTION OF MARKET SIZE	47

-------
SAB MEMO

1. Introduction

This memo is written in response to a number of remaining questions
raised both by the Environmental Economics Advisory Committee (EEAC) as
part of its review of the McClelland et al. report on the benefits of
groundwater cleanup and tiy reviewers. We also include the view graphs
used in our presentation to the committee as an Appendix. Prior to
discussing specific theoretical and empirical questions raised by the EEAC,
we provide some background information with respect to the purpose of the
study under review.

The motivating question the groundwater study is designed to address
is simply: Do any non-use benefits derive from corrective actions regarding
groundwater contamination and if so, how large might they be? The
contingent valuation study of the benefits of groundwater cleanup
undertaken by the research team at the University of Colorado followed
naturally from two prior contingent valuation studies. These previous
studies were undertaken to examine methodological issues in using
contingent valuation to measure the benefits from improving visibility in the
Eastern United States. (Schulze, et al. March 1990; McClelland, et al. June
1991).

Results from these prior studies indicated, among other things, that
1) for a familiar commodity such as visibility, information had little effect on
total values and 2) that embedding posed serious problems for
disaggregating and interpreting respondent's stated values. Groundwater

-------
presented a more challenging commodity for valuation in that, as shown in
pre-testing by Mitchell and Carson (1989), people know very little about
groundwater and people reject an existence value scenario where
groundwater would never be used i.e., they did not believe that clean
groundwater would not be used ("... it is too difficult to overcome people's
beliefs about future use by others to design a scenario that would only
capture stewardship [existence] values.", p.85, Mitchell and Carson, 1989).

This scenario rejection problem severely limited the types of
scenarios we could consider. For example, it would have been desirable to
use a scenario in which groundwater was already contaminated and other
surface water sources had been substituted. This would have allowed us to
ask for existence values for cleanup of groundwater directly. However, the
problem with, this scenario is that as soon as cleanup occurs, many
respondents would assume that the water would be available for immediate
use and include use values in their valuation. Again note that Mitchell and
Carson were unable to convince pretest respondents that clean groundwater
would not be used (".. many participants are unwilling to believe that there is
no likelihood of future use in the relatively near future, despite specific
assurances to the contrary.", p. 54, Mitchell and Carson, 1989). Given this
problem, the ability of the CVM to estimate non-use values as a separate
category from use values becomes difficult. However, groundwater does
provide an excellent commodity to test the methodological limits of
contingent valuation. Thus, our study should be viewed as exploratoiy in
nature. Both the Office of Solid Waste and the Office of Policy, Planning and
Evaluation recognized the experimental nature of the study and gave us
complete intellectual freedom in its conduct.

2

-------
The remainder of this memo discusses two issues, the definition of
the commodity and the application of the study to the estimation of national
benefits.

2. commodity Definition

The EEAC raised the issue of the definition of the commodity being
valued by respondents in the national groundwater study. In response to
USEPA's mission for the study, the benefits of primary interest are those
deriving from complete groundwater cleanup with emphasis on non-use
values. Complete groundwater cleanup can be viewed as a hedonic
commodity which potentially consists of a vector of services including

1)	reduced risk for present and future generations if the
contaminated groundwater were potentially available for
consumption

2)	clean water for the current generation (providing use and
altruistic values) if the contaminated groundwater is not believed
to be available for current consumption

3)	clean water for future generations (bequest values) if the
contaminated groundwater is not believed to be available for
future consumption

4)	clean water even if such water is never used (existence values).

In practice the process of disaggregating such values has proven to be

difficult in empirical analysis (See Section 3.3 pp. 33-49, of the report
which discusses embedding). Further, serious theoretical questions have

3

-------
been raised concerning the possibility of double counting if altruistic and
bequest values are naively included in a benefits assessment (see Chapter 2).

Given the charge from EPA, the first task was to examine the
conditions under which non-use values are present and how they might be
measured. First consider the difference between Expert benefits and
Subjective benefits. Expert benefits can be defined as the benefits experts
believe to exist. They are usually calculated as a value of a life times an
expert assessment of risk reduction times the exposed population. This
value obviously excludes at least some categories of non-use values and would
limit consideration to the first category listed above for services provided by
complete cleanup. Subjective benefits on the other hand are defined as the
values potentially exposed people themselves and others place on
environmental cleanup. These values will be based on perceived risks and
will include non-use values if they are present. Clearly, however, to have
subjective non-use values, consumers must know that a groundwater
problem exists.

The distinction between Expert and Subjective benefits raises
important theoretical questions in welfare economics. Use of expert
assessments in public policy decision making may violate consumer
sovereignty, but such judgments may be based on different (and possibly
more complete) information. However, to obtain non-use values one must
generally measure subjective values. In fact substantial subjective values have
been shown to exist for LULU, NIMBY or BANANA1 sites in a large number of
studies using property values and/or contingent values in situations where
expert risks and values are very small but where the public has become

^LULU = locally undesirable land use: NIMBY= not in rny backyard; BANANA= build
absolutely nothing anywhere near anybody.

4

-------
alarmed (See for example V.K. Smith and W.H. Desvousges, 'The Value of
Avoiding a LULU: Hazardous Waste Disposal Sites" ReStat.. 1986 and
McClelland, Schulze and Hurd, 'The Effect of Risk Beliefs on Property
Values: A Case Study of a Hazardous Waste Site." Risk Analysis. 1990).

Given the goal of measuring subjective values, a number of underlying
factors had to be accounted for to develop a scenario for groundwater
valuation. In real world contamination scenarios, offsite groundwater
contamination of wells assures public awareness of a problem. Further,
regardless of how small the expert assessment of risk may be, to our
knowledge, public officials have never knowingly allowed contaminated
water from a NIMBY site into a public water supply. This effectively rides
out category (1) of services as described above.

Further in our own survey pre-testing and in real world situations,
attempts at risk communication have been surprisingly ineffective. For
example, in survey pre-t esting, risk information was given to respondents
on a risk ladder which showed that the relative risk of drinking the
contaminated water was extremely small. This information had no effect on
76% of subjects, raised values for 15% and lowered values for only 9% of
the respondents (see Table 2 below). At the same time new research
showed that the conventional wisdom in risk communication which
suggested that a complex risk ladder was required proved to be false. Roth
et al.. state that a comparison of two risks across (i.e. risk of x-rays compared
to risk from drinking contaminated groundwater) fared much better than
the previous literature had suggested (Roth, Morgan, Fischhoff. Lave and

2USEPA at one time attempted to have experts from the Center for Disease Control evaluate
risks at Superfund sites and inform local residents tn person of the magnitude (very small) of
those risks. These experts were in nearly all cases "run out of town on a rail", i.e., they were met
with complete disbelief and hostility.

5

-------
Bostrom, 1990). Given that the risk ladder appeared to raise rather than
lower values, we decided to replace the risk ladder with a direct
comparison between risk from medical x-rays and risk from drinking
contaminated groundwater. This fortuitous research development
considerably simplified the resulting final survey instrument.

Thus, the scenario used in the groundwater survey describes the
hypothetical case of offsite well contamination where public officials have
eliminated the contaminated groundwater from the public water supply.

This implies that values from this study fall In categories (2), (3) and (4)
listed above but should most appropriately be applied only to sites with
actual or potential offsite well contamination. In other words, it is highly
likely that consumers living near a site with offsite well contamination will
have heard about the site. This is a precondition which must be fulfilled for
the existence of subjective non-use values such as existence value, but is
likely to severely limit the number of sites to which the benefits should be
applied.

The inclusion of other alternative options to complete cleanup has the
following justification: as discussed by Fischhoff and Furby (1988), survey
respondents are likely to use their own default assumptions with respect to
a scenario if information is not provided in the survey instrument. In the
case of groundwater contamination people may fear that no groundwater will
be available for themselves or future generations without complete cleanup if
this is the only option discussed, producing unrealistically high values. The
groundwater survey thus specifies several default alternatives to eliminate
the fear of "no water" without complete cleanup. From an economic as
opposed to a psychological perspective these default alternatives enter the
individual's decision making process as substitutes to the complete cleanup

6

-------
program. These substitutes include (in the final design): 1) home
treatment, 2) public treatment, and 3) containment. The no substitute
alternative of water rationing (40% in the base scenario) is also included as a
scenario for consideration. Since our hypothetical scenario involved a public
water supply, and because of the high cost both In convenience and on a
$/1000 gallon basis bottled water was not included as a substitute. This
substitute would however be appropriate for the entirely different situation
of contamination of private wells.

2A. A Theoretical and Empirical Analysis of Use Values Obtained in the Study

We begin a formal analysis by first only considering use values. Figure
1 shows a household demand curve for current use of water. At a price of
Pwper thousand gallons, Qe is the initial equilibrium quantity of water used.

This corresponds to the situation presented respondents prior to
discovering that the groundwater used in their public water supply has been
contaminated by their local landfill. In Figure 1, Qs represents the amount
of water (60% of Qe) still available from surface water sources, given that
public officials take contaminated groundwater off line. If the only available
alternative to complete cleanup of the groundwater is a shortage, then the
willingness to pay for use value is the same as that to avoid the potential
shortage. This is the sum of the shaded areas "a" and "b" shown in the figure
(assuming that the price of water remains the same at Pw). If however, a

substitute source of water from public treatment of the contaminated water
is available at a price like PT, the willingness to pay for use value for

complete cleanup is reduced by the availability of the substitute to the
shaded area "b" shown in Figure 1. It is important to note that the use value

7

-------
FIGURE 1: EFFECT OF PROVISION OF SUBSTITUTE COMMODITIES

of complete cleanup (shaded area "b") is conditioned on both the availability
of surface water Qs (or conversely the magnitude of the potential shortage)
and on the next best (cheapest) substitute available at price PT.

We can formalize the arguments made above as follows. Let
P = price of water

Q = quantity of water consumed by the household
Pw = price of untreated water

8

-------
PT= price of treated water (PT>PW)

Qe = amount of water consumed at price Pw

Qt = amount of water consumed at price PT

Qs = amount of surface water available to households given that

contaminated of groundwater is not used
c = consumption of a composite commodity
Y° = income

The analysis will continue to ignore non-use values for the moment so
that the role of a substitute for complete cleanup, e.g. public treatment, can
be evaluated in deriving use value for complete cleanup. Note that for
simplicity we consider the case of a perfect substitute. The utility function
U(C,Q) thus depends only on the composite commodity, consumption C, and
on water consumed Q. Since the survey values were generally less than half
of one-percent of income, we further assume a constant marginal utility of

r

consumption (money) so 9U/9C = Uc = a constant. Thus, the utility function
can be approximated as

(1) U » U° + Uc • (C - C°) + W(Q)

where U° is a constant, Co is the initial consumption level and W(Q) is the
separable utility (as a result of a fixed marginal utility of consumption)

i

derived from consuming water. Note that fixing Uc does not imply that

dU/dQ = BW/dQ is fixed. To obtain the willingness to pay for complete
cleanup take the utility of complete cleanup which provides 9e
but with a reduction in consumption equal to WTP (willingness to pay) for
complete cleanup,

9

-------
(2) U(Y° "PwQe - - , Qe).

and set it equal to the utility which obtains in the substitute situation where
public treatment is the chosen second best alternative,

(3) U(Y° PwQt	" (Qt Qs)» 9t)-

Note that in (3) the household pays a fixed marginal cost, PT-Pw, for the
portion of water which is treated, Qt-Qs* and pays the old price Pwfor the
surface water portton of water used, Q§. If marginal costs of public
treatment are constant (actually about $1.60 per thousand gallons for
charcoal filtration)3 and marginal costs of surface water provision are also
constant, then this formulation provides the appropriate welfare measure.
Setting (2) equal to (3) and using (1) yields

(4) WTP

,[W!9^W!92i_pw(g

c*Ot)J + [(p-r-W) .(Ot - Qs>] •

Where Qt corresponds to the intersection of Pxwith the demand curve in
Figure 1, the first term in square brackets in (4) above corresponds to the

^We seriously considered providing this figure In the survey but this would have made it
impossible to ask for a WTP for public treatment since the price of public treatment would have
been known. Further we did not have available the costs of other alternatives which might be
appropriate for inclusion in future research. Our public treatment alternative was similar to a
real situation. The town of Commerce City near Denver, Colorado, is located next to the Rocky
Mountain Arsenal. The town's wells were contaminated (at a very low risk, however) and the
Army and USEPA were forced to construct a water treatment plant producing 100 million
gallons of water per month of charcoal filtered groundwater. The plant cost about $ 12 million
and operating costs are about 60* per 1000 gallons. Using a capital recoveiy factor of. 10
Implies a total cost of treatment including capital and operating costs of $1 .60/ 1000 gallons.
Commerce City gets an additional 23 million gallons per month from surface water sources.
Thus, the town is about SO% dependent on groundwater. This solution satisfied the local
residents who were outraged that their groundwater had been contaminated.

10

-------
triangular shaded area to the right of Qj (in the area marked b) and the

second term corresponds to the remainder of the area marked b. to the left
of*.

Obviously the survey design could have been simplified if only one
substitute option, public treatment, was presented to respondents as the
next best substitute. However, the next best substitute is complicated both
by the varied nature of respondent preferences and by the additional
consideration of non-use values. Looking at Table 1, public treatment is
ranked second to complete cleanup by the "average" respondent. But, only
56% of respondents ranked complete cleanup highest, and only about one
quarter of all respondents had complete cleanup ranked first and public
treatment ranked second. In fact, many respondents who ranked complete
cleanup first, ranked containment second (about 31 percent of the total
sample). Thus, deletion of substitute cleanup options was not deemed
feasible based on our pretest results which had revealed a similar preference
pattern. Note also that many households ranked public treatment which
does nothing to cleanup groundwater first, so existence values apparently
did not matter greatly and their value given for complete cleanup would not

TABLE 1: RANKINGS OF OPTIONS TO RESPOND TO GROUNDWATER
CONTAMINATION FROM VERSION D

RANKING

OPTION

MEAN LEVEL OF SATISFACTION

1

COMPLETE CLWP

4.35

2

PUBLIC TREATMENT

3.77

3

RATION - 10%

3.65

4

CONTAINMENT

3.40

5

HOME TREATMENT

2.89

6

RATION - 40%

2.61

7

RATION - 70%

2.35

(there Is no statistical difference between the mean level of satisfaction for any of the options
between respondents answering Version D and other versions - other versions did not rank
10% or 7 0% dependence on water)

11

-------
likely reflect an existence value. By thinking about an alternative substitute
solution such as public treatment, respondents can think about whether or
not existence values are important. Further justification for retaining a
variety of substitutes can be obtained by looking at Table 2 which reports the
results of our debriefing questionnaire used in pre-testing to determine
what information was useful to respondents in constructing values.

The information and context presented in the survey had the effect of
lowering the mean value for complete cleanup in our December 1990 pre-
testing from $20.22 to $12.20. Table 2 helps answer the question, why did
values fall? Looking at the "Lowered Value" rows in Table 2 shows that the
largest self-reported impacts of information which served to lower value
were associated with substitutes: the information on the water supply
treatment option (public treatment, Q-55) caused 34°A of respondents to
lower values, information on private options (home treatment, Q-53) caused
24% to lower values, and buy water (from another community Q-51) caused
20% to lower values4. In all of these cases the offsetting number of
individuals who increased values consists of a much smaller group. Thus, we
concluded that the most important information presented in the survey was
information on substitute commodities since the evidence strongly suggests
that the right hand tail of high values in the skewed distribution of bids was
in great part eliminated by information on substitutes just as economic
theory would predict i.e., respondents informed of substitutes reported area
"b" rather than the sum of area "a" and "b" in Figure 1.

^The buy water option was replaced by a containment option in the final survey design because
EPA was interested In obtaining a value for containment In situations where complete cleanup
was infeasible. Containment received higher rankings and values than the "buy water" option
did in pretesting.

12

-------
TABLE 2: SELF-REPORTED EFFECTS OF CONTEXT DECEMBER 1990 PRETEST

Self-reported Effects of Context from
December 1990 Pretest (Summary of responses to Q-47
through Q-56, Appendix B)

Bacentgge

Self-

reported
effect

9*47

Pers
exp.

9-48

Def. of
gwater.

949

Speed of
gwater.

940 901

Water Buy water
bill option

No effect

75%

82%

90%

77%

67%

Lowered value

0%

3%

8%

8%

20%

Raised value

25%

15%

2%

15%

13%

Facentege

Self-

reported
effect

&82

Water
cons.

9*3

Private
options

9«4

Dis-
counting

9«

W. S. T.
option

Q46

Risk

commun.

No effect

72%

66%

79%

61%

76%

Lowered value



13% 24% 10.5%

34%

9%

Raised value

15%

11%

10.5%

5%

15%

The survey was thus designed so that if complete cleanup were not
funded, respondents would likely assume a cheaper alternative would be
funded (such as public treatment, or, in the final survey design,
containment). However, respondents were left to implicitly guess at the
price of these substitutes for the reasons noted in Footnote 3.

The small value obtained for a complete cleanup when a 70??
dependence on groundwater was used as the scenario also suggests that
respondents were assuming substitute alternatives were available. Figure 2
shows values taken from Version D of the survey which asked for
willingness to pay for complete groundwater cleanup as a function of the '

13

-------
FIGURE 2: PREDICTED WTP AS FUNCTION OF DEPENDENCE ON
GROUNDWATER

$/Mo 15 r

10

10 20 30 40 90 00 70 80

PERCENT OF WATER FROM GROUNDWATER

level of dependence on groundwater for domestic water supply (10%, 40%,
and 70% dependence). Note that even as the dependence on groundwater
increases to 70%, the total bid remains relatively small. Demand studies
(cited below) suggest that this level of potential shortage should move into
an almost totally inelastic region of the demand curve. Willingness to pay
values should explode unless respondents assume substitutes for a water
shortage are available.

From the information presented in Figure 2 we can derive an implied
demand curve for water use. Figure 3 shows the water demand curve
implied by Figure 2 where the marginal willingness to pay for water above its
current price is taken as the slope of the bid function. The data presented
in Figure 2 are converted to $/1000 gallons per month based on an average
annual water usage in the United States of about 8000 gallons/month which
we assume applies to our average respondent. The average U.S. price is
about $1.20/ 1000 gallons. These estimates are taken from Michael

14

-------
Nieswiadomy "Estimating Residential Water Demand" Water Resources
Research. 1992. The midpoints of the water quantity changes along with
the incremental marginal willingness to pay above the current average price
of residential water are plotted in Figure 3. The first step down, from 8000
to 6000 gallons/mo. implies an arc elasticity of -.4. This estimated
elasticity of water of -0.40 for moving from a 8000 to 6000 gallon per month
dependence on surface water falls well within the range of water elasticities
found in the empirical literature on water demands. Note that Nieswiadomy
estimated demand elasticities as low as -0.11 and Howe (WRR. 1982)
estimated demand elasticities as high as -0.57. Further, the more inelastic
estimates are for winter demand and the less inelastic are found for

FIGURE 3: DERIVED WATER DEMAND CURVE AND ELASTICITIES

^Figure 2 implies that households would on average bid an additional $1.35/1000 gallons of
water (based on average water use) to avoid moving from a 10% to a 4096 "shortage" and
$2.57/ 1000 gallons to avoid moving from a 40% to a70% "shortage."

15

-------
summer demand. The next step shown in Figure 3 from 6000 to 3600
gallons per month implies a very low total price for water of $3.77/1000
gallons. Based on existing demand studies we conclude that respondents
seem to be limiting their marginal WTP by considering substitutes.

2B. A Theoretical and Empirical Look at Non-Use Values

Empirically very little is known about the nature of non-use values. In
general, respondents to CV studies will attribute a large share of values to
non-use categories if asked to split up their values (e.g., Greenley, Walsh and
Young, 198 1) Altruistic and bequest values have been challenged
theoretically and one empirical study (Madariaga and McConnell 1987) has
shown that this challenge has at least some validity. Based on our own verbal
protocols and pretest results which provide some indication as to the nature
of preferences for non-use values, Chapter 2 of the report attempted to
explore a number of possible theoretical models of non-use values. Given
this situation, and given the severity of embedding problems (as
documented in Chapter 3), we employed three separate approaches in the .
study to attempt to partition total value into use and non-use components.
Each method has obvious problems and advantages. In what follows we first
consider theoretical issues relevant to measuring non-use values and then
turn to psychological issues relating to how individuals think about non-use
values.

Theoretical Issues:

To begin the discussion we construct a first order model which
assumes separable utility both between use and non-use and between

16

-------
categories of non-use such as altruistic, bequest, and existence motives.
Thus, we initially assume that utility can be written as

(5) U(C,Q) + A + B + E

where A is the altruistic utility derived from providing clean groundwater to
the present generation (friends and neighbors) B is bequest utility derived
from providing clean groundwater for future generations, and E is utility
derived from knowledge that groundwater itself exists in an uncontaminated
state independent of use. Following the theoretical analysis for use values
provided above, assume that complete cleanup provides levels of non-use
related utility Ac> Bc, and Ec while public treatment provides levels of non-
use related utility Ap, and Ep. Then, to adjust the argument for inclusion
of non-use values we only need to add 4, Bc and Ec to expression (2) of the
preceding section which describes the utility of complete cleanup and AT,
Br and Et to expression (3) which describes the utility of public treatment.
To insure the legitimacy of including values derived from A and B we assume
that these utilities derive solely from paternalistic altruism (See Chapter 2).
Note that this is a very simplistic analysis since we do not explore either the
nature of altruism or the functional dependence of A, B, and E on quantities
of clean water available for use or in the ground over time. As shown in
Chapter 3 these are in fact complex issues. Following the same procedure
used in the preceding section and using equation (1), the willingness to pay
for complete cleanup (WTPC) conditioned on a next best substitute of public
treatment is

rr\ wttt TFiTx^ Bc-Br Ec-Er

(6) WTPC - WTP + -q- + -jjr- + -jjr-

(a)
-------
where WTP can now be interpreted as the separable willingness to pay for
public treatment and is identical to the expression shown on the right hand
side of equation (4) above. Thus, the availability of a substitute option lowers
the use value in this more general case as it did in the case of use value
alone. However, the term WTP implicitly now includes not only use value
but also the cost of providing the joint non-use benefits which accrue from

Arp Et

public treatment which are equal to + u^+uy These i3enef^s 'f

providing non-use values for the public treatment option are correctly .
proxied by their costs contained in the WTP term (i.e. see equation (4)
which includes the term PT - P\v) in valuing complete cleanup. Thus, the

substitute option also serves to potentially reduce these values as well.

Terms (a), (b), and (c) in equation (6) are, respectively the separable
incremental willingness to pay for altruistic, bequest, and existence value as
provided by complete cleanup over and above those provided by public
treatment.

The question of the separability of non-use values which allows this
very simple structure is, of course, a purely empirical question about which
little is known and which is likely to be confounded with the embedding
problems which are discussed at length in Chapter 3. Usually pure
existence values (term c above) are assumed to be separable (Freeman
1992). This separability is assumed to make indirect market based
valuation impossible. Thus, such values also leave no behavioral trail (see
also Report to NOAA of the Panel on Contingent Valuation, January 1993).
The separability of altruistic and bequest values has had even less discussion
in the literature. We view equation (6) as a first order approximation only to
begin to explore the issue of identifying component values.

1 8

-------
The first of three methods employed in the study for attempting to
partition values as shown in equation (6) is use of a scenario difference. In
particular a subsample in the survey was asked both for the value of
complete cleanup and for their value for public treatment. We then
estimated non-use values by subtracting the public treatment value from the
complete cleanup value which is the same as subtracting WTP from both
sides of (6). Referring to equation (6) it is likely that E-p = O for public

treatment since nothing is done about subsurface groundwater
contamination so WTPC-WTP will contain a complete measure of existence
value. It is likely that = 0 since future generations (at least in the far

future) will benefit little from a treatment plant constructed today compared
to complete cleanup which insures clean water is available for all time.
Finally it is likely that A^ - Ap since both complete cleanup and treatment

protect the current population. Thus, WTPC-WTP is likely to exclude
altruistic benefits. This difference (under our assumptions that = O, Rr =
O and & = Ap) can be expressed as .

( 7) WTPC - WTP = |r + §r

Separability is not necessary for this approach to work. However, the
resulting measure will likely exclude altruistic values.

The second approach was to ask respondents to split up their bid for
complete cleanup into four categories. The obvious problem from (6) is that
the WTP term incorporates the joint cost of providing& Bp and Ex as the

appropriate proxy for benefits which is the cost of public treatment.

However, under our assumption that Ej = 0, By = 0 and AT = Ap, (6) can be

rewritten as

-------
Be Ec

(8) WTPC = WTP + + u|

where as noted above WTP is in part a cost based measure of the value of
W AT

providing jp and 777. This raises the question of how respondents will split
uc uc

up the WTP term. This is the standard allocation of joint cost problem
which has no obvious solution. One hypothesis would be that WTP is
allocated by respondents in proportion to the relative magnitudes of W/U'c
and Ac/U^. Note that WTP £ W/V'c + Ac/U^ since costs are less than benefits
for use and altruistic values. Thus, we see that the presence of substitutes
complicates the percent split approach even where separable utility is
assumed. However, the percent split approach is most likely to be valid for
the case of existence valvies where separability is most plausible and no
other substitute (including containment since values were comparable to
public treatment) was viewed as assuring clean groundwater independent of
use.

The third approach was to reduce the dependence on groundwater
from 40?? to a low level (1 0%) and extrapolate WTPC to a no dependence
scenario. We felt we could not use a 0% dependence scenario directly since
this approach had been rejected by the Mitchell and Carson pretest
respondents (i.e., "someone, sometime will use the water no matter what
the survey says"). Presumably a zero percent dependence sets & = Ap = O as

well as WTP = O. Thus, (6) would now take the form

Be Ec
(9) WTPC = ^ + 0|

-------
providing a direct estimate of bequest and existence values. A problem with
this approach in addition to its exclusion of altruistic values is that Bcmay

be conditional on surface water dependence.

Psvcho-ogicaLIssues:

The discussion above details the theoretical assumptions and
implications of value partitioning. This section is concerned with the
question of whether individuals can actually partition the values in a
meaningful way such that they can be stated in response to contingent
valuation questions. This is a question of the psychological process of
respondents' value construction using the information and context in the
survey instrument. These psychological questions can be defined by two
primary concerns: First, do people have enough information to be able to
partition the values they are providing and, second, have people thought
carefully enough about the issues presented to allow partitioning?

In pre-testing the survey instrument this process was investigated
largely in terms of what information and context individuals used in value
construction. In pre-testing, respondents overwhelmingly rejected a
proposed fund for future use. The option offered did not fit individuals'
psychological model of how the world works and thus many respondents
simply refused to state a value for the program. The survey instrument
provided alternative scenarios to educate respondents and to get them to
think about future generations and about existence values that may obtain
from the proposed cleanup scenario. The rankings of individuals for these
options suggest that respondents were thinking about the varying aspects of
the programs in a manner consistent with value construction based on the
hedonic attributes different options provided.

-------
as Table 3 shows, the options including more attributes or services as
listed on pages 2 and 3 are rated higher by respondents suggesting both that
the survey provided enough information regarding the commodity being
valued and that many individuals thought carefully about the issues.

TABLE 3: OPTION RATINGS RELATED TO BENEFIT COMPONENTS

OPTION
(MEAN RATING)

OPTION BUN1MTS

COMPLETE CLEANUP
(4.35)

Respondent's Use (Use Value)
Other's Use (Altruistic Value)
Future Use (Bequest Value)
Existence Value

CONTAINMENT
(3.45)

Respondent's Use (Use Value)
Other's Use (Altruistic Value)
Future Use (Bequest Value)
Some Existence Value
less certain

PUBLIC TREATMENT

(3.74)

Respondent's Use (Use Value)
Other's Use (Altruistic Value)

HOME TREATMENT
(2.81)

Respondent's Use (Use Value)

Table 3 indicates that individuals were implicitly able to distinguish
the components of the value of complete groundwater cleanup in their
preferences since they were able to rate the different options based on
these components. The value partitioning question asked individuals to
directly partition these values as a portion of total value. Some critics of
this approach have argued that when faced with such a cognitive task that
individuals will simply allocate an equal percent to each of the categories
offered (i.e. 2596 into each of four categories offered totaling to 100?A).

Table 4 shows the component allocation for a random subsample from
the survey respondents (every 50th respondent in the econometric data
set). Considerable differences in partitioning of values are shown for

-------
TABLE 4: PERCENT COMPONENT ALLOCATION OF TOTAL WTP FOR
COMPLETE GROUNDWATER CLEANUP (RANDOM SUBSAMPLE)

CBS

USE

ALTRUIST

BEQUEST EXISTENCE

100

150

200

250

300

350

400

450

100

500

550

600

650

700

750

800

850

900

950

100

1000

1050

1100

1150

1200

100

1250

1300

100

1350

1400

1450

1500

1550

1600

1650

1700

100

1750

100

IsOO

1850

1900

1950

2000

100

different individuals. For observations 50 and 100, for example, all value is
allocated to existence. For individual 950 all value is allocated to current
use. Others, such as 200 and 1000, allocate evenly between use, altruistic
and bequest values (all consumptive options) and place a zero value on non-
consumptive existence value. The table also shows that roughly one out of

-------
five individuals could be claimed to be simply dividing 100% by the four
categories offered. The majority of the respondents do however seem to be
making a conscious allocation between the four components This type of
partitioning may only provide only a very rough indication of the relative
importance of the components of value. However, if serious doubts exist as
to the validity of certain components (such as altruistic and bequest values
which may represent double counting), value partitioning allows such values
to be discarded to obtain a conservative estimate of total value. Note that the
component split questions followed the total question and can be viewed as
debriefing questions which simply provide additional information6.

Further Evidence on Non-IJse Values

Previous CV studies of groundwater values from private wells have
almost always assumed that non-use values were incorporated in the stated
values. For example Edwards (1988) states:

"Even households with a zero probability of future demand for
groundwater on Cape Cod have positive option prices. This benefit is
attributable exclusively to the bequest motivation." (Edwards, p. 484),
as well as.

"A second surprising result is the small size of option value relative to
option price (1-2% or less)" (Edwards, p. 486),
and finally,

"A third interesting result is the strong influence of bequest motives
on total willingness-to-pay. Equity issues not withstanding individuals
appear to be willing to pay substantial amounts of money annually to

protect groundwater for use by future generations." (Edwards, p. 486).

®The issue of reading ahead was raised by the committee. Having observed hundreds of self-
administered pretests where reading ahead was allowed, we can state that this "problem" is
very rare. Occasionally, however, a test respondent will flip ahead to see how much la left to do.

-------
Poe and Bishop in their recent study of contamination of private wells
reach similar conclusions, as do most other researchers. A nrlorl. since our
study deals with contamination of public rather than private wells, it seems
even more likely that our study would include non-use values.

To demonstrate that individuals have non-use values for groundwater
cleanup remains an empirical issue for our study. As described above, the
groundwater survey included variations of the survey instrument to provide
three alternate approaches for estimating non-use values. (1) Scenario
Difference Approach: Version C asked for respondent's value for a public
treatment option which would only clean up water as used. The public
treatment option mostly captures use value and clearly excludes existence
value so the difference between the value for total cleanup and public
treatment approximates (but likely underestimates) non-use values. (2)
Component Allocation Approach: All versions of the survey asked for the
value of complete groundwater cleanup and how respondents' values were
allocated between categories of use and non-use values. (3) Extrapolation
Approach: Version D of the survey asked respondents how much they would
value complete cleanup if they depended on groundwater for 10??, 40%, or
70% of their total water supply. By modeling each individual's values as a
quadratic (or linear) function of the percent of water shortfall, the intercept
predicts the bid when there is no dependency on groundwater. The
intercept thus estimates non-use value for groundwater cleanup. The mean
estimates for non-use values (bequest and existence values combined) are
$2.81, $3.49 and $3.54 (quadratic, linear is $2.89) per household/per
month for the scenario difference, percent splits and extrapolation
approaches respectively.

-------
Additional support for the contention that non-use values exist and
were measured for groundwater can be found in the pre-testing of the
survey instrument. During this pre-testing, verbal protocols and
retrospective reports were used to examine individuals' cognitive processes
while completing the survey Instrument. These verbal protocols and
retrospective reports provide direct evidence that individuals hold non-use
values for groundwater:

..of course the immediate concern would be for me, my family
and future generations (which includes my family). What other
things would I be concerned about... if it wasn't for you, your
family and future generations."

"I think that they are probably going to have to have places to
grow food where they know it will be safe. And places with
water that they know will be safe. I don't think it is going to be
anything like it is today"

".. .we probably ruined the water so we should do something to fix
it. I was thinking we should do something for our kids. I have a
kid."

"Contamination problems, the main thing I was concerned with
is the fact that this solves problems permanently as long as no
more contaminants get into the ground. With people paying a
large amount of money they are going to be after them to make
sure the water stays clean instead of treating it as you use and
keep polluting it??? . . .Mostly in terms of the future so you know
you are not procrastinating the situation."

"I feel like its important that something is done, sometime
somebody is going to have to pay for it. It may as well be this
generation because I think people are more than/ they should be
willing to have clean water."

"I don't trust the people other than the people involved at this
time will have the same level of commitment to keeping
groundwater clean and pure..."

"... you never know if something is set up if it will actually happen
and you don't know if people will use the money for what it was
meant for."

-------
These individuals display paternalistic or non-paternalistic altruism in
deriving their values for future generations in addition to a concern simply
that the groundwater be cleaned up. In contrast, some individuals display
no particular concern for future generations:

"I am not one to think long term, unfortunately

As Table 4 showed above, individuals are different and thus allocate
their values differently between use and non-use values. As these quotes
from the verbal protocols and retrospective reports show, these value
allocations are based on different concerns for future generations and the
environment. These concerns are direct evidence of the presence of non-
use values for groundwater cleanup.

Finally a fourth approach for empirically estimating non-use values was
derived from post-testing of the survey instrument in Denver during the
Spring of 1993. The same survey was used but with additional questions
added to deal with the question of market size for applying the household
benefits and timing issues of the completion of groundwater cleanup.
Following the valuation questions (which remained identical to those in the
mail survey7 ) individuals were asked how long they had assumed
groundwater cleanup would take. Following this, individuals were asked
how much their value would change if the cleanup was not completed until
10, 30 or 100 years in the future.

7For the individuals answering Version A of the Denver post-test (where monthly WTP for
groundwater cleanup were elicited) the geometric mean monthly WTP was $7.32 (n = 36). We did
not perform regression analysis (or Box-Cox transformation regression) on the post-test due to
the sample size. The geometric mean Is the anti-log of the mean logfWIP). This would be the
effect of a Box-Cox transformation with a transformation coefficient of zero. This value of
$7.32 can be compared to the $7.01 predicted wip from the regression of the national mail
sample.

-------
As Table 5 shows the mean expected time to cleanup was 7.77 years.
The willingness to pay for cleanup fell as the time to completion of cleanup
was extended indefinitely into the future as would be expected.

TABLE 5: TEMPORAL ASPECTS OF CLEANUP VALUATION

MEAN
(STD DEV)
n

EXPECTED YEARS TO COMPLETION
OF CLEANUP

7.77 years

(5.40)

PERCENT OF BASE WTP IF
CLEANUP COMPLETED IN TEN
YEARS

82.03%
(33.50)
69

PERCENT OF BASE WTP IF
CLEANUP COMPLETED IN THIRTY
YEARS

55.36%
(40,58)
69

PERCENT OF EASE WTP IF
CLEANUP COMPLETED IN ONE
HUNDRED YEARS

36.30%

(39.94)

69 I

Figure 4 displays the relation between WTP and time to completion.
As cleanup is delayed beyond thirty years, the function flattens out at a
positive level and does not approach zero which it would if individuals were
simply exponentially discounting use values over time.

From this graph it appears that individuals are discounting some
portion of the total value and that after approximately thirty years the value
remains relatively stable. It is reasonable to assume that individuals will
perceive a declining use value as the time to completion is drawn out but
that if bequest and existence value are based on very long term
considerations or on moral considerations these values will not depreciate
to zero as the time to completion is delayed. The positive asymptotic value
supported by the relatively horizontal portion of the graph can be

2 8

-------
considered a non-use portion of the total value (possibly consisting of
bequest and existence values).

FIGURE 4:

PERCENT OF BASE WTP AS A FUNCTION
OF EXPECTED COMPLETION OF CLEANUP

100 -1

90"

80"

70"

00 -

80"

40*

30 -

20 "

10 -

Oi
0

t r

20 30 40 BO 60 70 80

YEARS TO COMPLETION

90 100

Table 6 indicates the implicit rates of discount the data imply based
on the total stated value. The rate of discount after thirty years is close to
zero (0.60?! per year).

TABLE 6: IMPLICIT RATES OF TIME PREFERENCE

Between 7.77 yrs and 10 yrs

8.89%/yr.

Between 10 yrs and 30 yrs

1.97%/yr.

Between 30 yrs and 100 yrs

0.60%/yr.

Since virtually all respondents will be dead within the next hundred
years it is reasonable to assume that any values for cleaning up groundwater
in one hundred years are not use values (either for oneself or for others now

-------
alive which we label altruistic values). The mean percent allocated to non-
use in the national groundwater survey was 44.20A. In post-testing, the
mean percent of total WTP individuals stated they are willing to pay for
cleanup with completion in one hundred years is 36.3% of total value. If
there is some loss of bequest or existence value during the cleanup delay
then this approach would underestimate non-use values to some degree. For
the post-test subjects there was not a statistically significant difference
between the non-use WTP using the component allocation approach or the
WTP for complete cleanup with a one hundred year completion date8.

Pre-Testlng of the Commodity

Our pre-testing of the commodity began with the design of a lengthy
pre-test survey instrument containing a variety of test approaches for
attempting to partition non-use from use values. In designing this pretest
instrument we carefully evaluated the prior study by Mitchell and Carson
(1989). This study only went so far as to conduct focus groups in which
subjects were exposed to information on ground, water and hypothetical
groundwater scenarios. No preliminary survey appears to have been pre-
tested. Rather, based on general reactions to information and scenarios in
the focus group sessions, a first draft survey was prepared for possible later
pre-testing. Much valuable information was obtained from the focus groups.
Our own detailed reading of the transcripts suggested two conclusions: (1)
most participants knew little about the scientific aspects of groundwater and
(2) respondents strongly rejected scenarios in which groundwater would
supposedly never be used. Respondents plausibly rejected the notion that

8For Version A (monthly WTP): t = -1.65, p = .11, n = 34: for Version B (lump sum WTP): t= 1.04, p
= .31, n = 27

-------
clean groundwater would never, ever be used by anyone, anytime in spite of
numerous guarantees made now. Mitchell and Carson came to the same
conclusions. However, surprisingly, in their draft survey, they still
attempted to construct a scenario in which groundwater would never be
used, providing assurances that government would deny access to
groundwater below a newly proposed landfill . Note that in our own pre-
testing respondents uniformly rejected government assurances of future
action, e.g., "If it is not done now we doubt it will ever happen." We found
this scenario to be unlikely to succeed based on Mitchell and Carson's own
focus groups and it should be noted that they expressed reservations
themselves. Further, Mitchell and Carson chose not to continue the study.
We did view their scenario as a backup strategy in case other approaches
failed.

Thus, since respondents rejected a scenario which would allow direct
estimation of non-use values, we were forced to attempt to partition values.
Note that we defined a specified believable current use-value so that
respondents would not substitute their own uncontrolled assumptions about
use as we believe they would have in the Mitchell-Carson scenario. Only two
approaches are available for partitioning values: (1) asking respondents to
directly partition values and (2) use of scenario differences. Both methods
have a long history in contingent valuation studies dating back to the late
1970s (reported in studies such as Brookshire et al. 1979, Greenley, Walsh
and Young, 1982, Tolley et aL . 1985, and Mitchell and Carson, 1987). We
should also note that both the extrapolation approach and the temporal
delay approaches described above are variants of the scenario difference
approach. Thus, a major goal in pre-testing was to solve three potential
problems with the scenario differences approach.

-------
First, respondents may not understand the differences in scenarios
well enough to provide different values. Thus, pre-testing should be able to
show demonstrable value differences.

Second, one hypothesis explaining the embedding problem which
could confound the scenario difference approach is the warm glow
phenomenon (Andreoni, 1989). If the first scenario gets an extra warm
glow value attached to it and the second does not, bias could result. For
example two questions might ask for the value of partial environmental
cleanup and then for the value of complete cleanup. The difference, then,
should value incremental aspects of cleanup of interest to researchers.
However the difference will be reduced if the first bid incorporates the pure
separable satisfaction of contributing to a good cause (warm glow) while the
second does not. Thus, we have specifically attempted to remove warm
glow values with a disembedding question9.

Third, scenario rejection often occurs when multiple scenarios are
presented. If some people prefer scenario A to B, they will bid for A but
then feel no obligation to bid for B, even if it has some value to them as an
alternative. A referendum approach can exacerbate this phenomenon in that
people are unlikely to be willing to vote for a less preferred alternative.

To attempt to address these and other issues we began pre-testing
with a first phase which obtained 10 verbal protocols which recorded the
detailed thoughts of respondents as they filled out two versions of the pre-
test instrument. To our knowledge this was the first application of verbal
protocols to the design of a CV instrument. The transcripts from these
were only briefly extracted in the report because they are comparable in

9 Mitchell and Carson in their draft groundwater survey also include adisembeddlng question
of slightly different design.

-------
length to the report itself. Verbal protocols have proven to be much more
revealing of the detailed thought processes of respondents than focus
groups. Thus, the verbal protocols gave us substantial insights into the
detailed thought processes of respondents, processes which are masked in
group settings. For example, they showed a strong tendency by respondents
to override assumptions made in the design of the survey instrument with
their own beliefs if those beliefs contradicted those implicit in the survey
instrument. These beliefs about the way the world works have been termed
mental models by psychologists (see for example, Bostrom, FIschhoff and
Morgan, 1992). Subsequently, the survey was designed to avoid
contradictions between survey design assumptions and respondent's mental
models (e.g., government can't be trusted).

We further found that the "commodity" as economists would term it
was, of course, nothing of the sort in the minds of respondents. Rather,
they viewed the survey as presenting a nroblem for which a number of
alternative solutions were presented. Many economists might view
contaminated groundwater as a small reduction in the total quantity of all
clean groundwater. In contrast, residents near a NIMBY site view
contaminated groundwater as a separable commodity or problem in and of
itself which directly provides negative utility. Overwhelming psychological
evidence supports this latter view (see for example McClelland, Schulze and
Hurd, 1990 and Bostrom, et al. 1992). In this sense the final version of the
groundwater survey closely follows the kind of process people and real
communities follow for solving problems. First, citizens find out about the
problem and then alternative solutions are evaluated, compared and ranked.
In this sense the original version of the survey (with a series of referenda on
different solutions used to explore alternative valuation strategies) was

-------
viewed as unrealistic by respondents. The sequence of referenda created a
scenario rejection problem which was clearly demonstrated in the second
phase of pre-testing which employed self-administered surveys for 80
respondents. We discovered that when a respondent had voted yes on a
particular solution, and when a following alternative was lower ranked (but
quite likely still "acceptable") there was a strong tendency to vote "no" on
the lower ranked option (scenario reject) even though the respondent still
had a positive value. This caused us to drop the referendum format
inasmuch as we wanted to obtain multiple values from each individual to
allow use of the scenario difference approach.

Thus, in the third phase (117 respondents) we tested a format which
had respondents consider all alternative solutions (by rating each of them)
before valuing complete cleanup. This design worked very well because it
now conformed to what respondents viewed as a natural cognitive flow (1)
present problem, (2) consider alternative solutions for the problem, (3)
value one solution which might be "chosen" by the community, and (4) value
an alternative solution in case the initial "choice" is infeasible. However,
dropping the referendum format in and of itself was not sufficient to entirely
eliminate scenario rejection for the second valued scenario. The problem
demonstrated in phases I and II of pre-testing was an inherent reluctance to
put a value on a less valued commodity. To explain this psychologically,
imagine that a consumer prefers a Lexus to a Volkswagen. A Volkswagen
sales person then asks "What would you pay for this Volkswagen." A natural
response would be for the consumer to say "nothing," i.e., reject the
scenario. Another way of asking this question which helps greatly to
overcome scenario rejection is to ask for a relative value, i.e., "thinking about
the Lexus I know you want, how much of what you would pay for the Lexus

-------
would you pay for the Volkswagen." This relative comparison focuses the
respondent immediately on the features of the Volkswagen relative to the
features of the Lexus and on missing attributes of the Volkswagen. The value
of the Lexus forms an anchor from which the respondent can adjust his or
her value down based on missing features or characteristics. A large
literature exists in psychology on anchoring and adjustment, much of it
addressed to cognitive error introduced by respondent use of inappropriate
anchors (see for example Tversky and Kahneman, 1974, Carlson, 1990,
Northcraft and Neal, 1987, and Kahneman, 1992). In our survey design,
however, we deliberately established an appropriate anchor (complete
cleanup) as the basis for adjustment. The pre-testing in phase III of this
approach was successful beyond our hopes in that we found little evidence of
scenario rejection in debriefing subjects and had a very high response rate
for obtaining relative values.

It should be explained that in both the second and third phases of pre-
testing with self-administered surveys, respondents were given a debriefing
survey, essentially a survey about the survey, which produced information
such as that presented in Table 2 above, which convinced us of the
importance of substitute options. Note that Table 2 strongly suggests that
the most important information in the NIMBY context of our survey was
information about multiple substitutes, not information about risk or the
scientific aspects of groundwater. The 197 debriefing surveys provided
guidance on design issues both from quantitative information as well as from
qualitative information taken from open-ended questions. In essence, our
survey was presented for comment to 207 reviewers (including the 10
individuals who provided verbal protocols) whose opinions we took very
seriously in designing the survey instrument. It is our view that major

-------
problems in past CV studies have resulted from the failure to design
instruments which were ' understandable to and accepted by respondents as
opposed to experts in survey design. Since experts (such as ourselves) often
have expectations which are inconsistent with the actual views of
respondents, we consider the process described above to be indispensable
in developing a defensible contingent valuation instrument. To our
knowledge this type of cognitive design process has never before been
implemented in a CV study. We took a psychological approach to attempt to
overcome what have appeared to be intractable problems in obtaining
component values (see for example, Tolley, et aL, 1985, who found large
order effects which are attributable in our view to warm glow and scenario
rejection problems).

Further, it is our view that component values must be identified if
problems such as the potential double counting of altruistic values are ever
to be addressed. For example, the NOAA sponsored Alaska study, by only
obtaining an option price, is open to the charge of double counting altruistic
and bequest values as well as the inclusion of warm glow values. The
approach we have attempted at least allows for the possibility of discarding
such values to obtain a lower bound estimate.

Finally, the only possible test of the reliability of the measurement of
non-use values is internal consistency between different ways of obtaining
non-use values within a CV study. Since no other approach has yet to be
proven, internal consistency is the only game in town. Thus, we employed
four methods in our survey designs (percent splits, a scenario difference
between complete cleanup and public treatment, extrapolating dependence
on groundwater to zero, and temporal delay of cleanup) to attempt to
provide internal cross validation.

-------
A Comparison Df Recommendations Of The NOAA Panel On Centingent
Valuation With Procedures Used In The Groundwater Study.

Motivated by controversy over the Exxon Valdez damage assessment,
NOAA recently commissioned a "Panel on Contingent Valuation" which
concluded that contingent valuation provides a reliable measure of non-use
values (passive use in their terminology) as long as a lengthy list of specific
criteria are satisfied. Those most relevant to our study include:

(1) Use of probability sampling

(2) Minimization of non-response

(3) Use of face-to-face interviews or telephone rather than mail-
Surveys

(4) Careful pre-testing of the questionnaire

(5) Conservative design

(6) Use of Willingness to Pay as opposed to Willingness to Accept
measures of value

(7) Use of a referendum format for eliciting values

(8) Accurate description of the program or policy

(9) Photographs should be pre-tested

(10) Reminders of substitute commodities should be included

(11) "A "No-answer" should be included for valuation questions

(12) Follow-up questions should be provided to valuation questions

(13) Checks of understanding and perceptions of the survey by
respondents should be made

(14) Respondents should be reminded of alternative expenditure
possibilities

(15) Values derived from a warm glow or dislike of big-business
should be deflected.

-------
(16) The timing cf remediation should be clear to respondents
In particular the panel argued that they would view as unreliable a study
which

• had a high non-response rate

• showed inadequate responsiveness to the scope of the insult

• showed a lack of task understanding by respondents

• had many respondents who did not believe the restoration
scenario

• contained votes on the referendum question not explained by
economic value motives

Given the highly specific nature of their detailed list of recommendations for
survey design and implementation, the NOAA panel qualified their
recommendations as follows: "We think it is fair to describe such
information as reliable by the standards that seem to be implicit in similar
contexts, like market analysis for new and innovative products and the
assessment of other damages normally allowed in court proceedings. As in
all such cases, the more closely the guidelines are followed, the more
reliable the result will be. It is not necessary, however, that every single
injunction be completely obeyed; inferences accepted in other contexts are
not perfect either." (tit.p.42)

Since our groundwater study was designed and implemented much
prior to the release of the NOAA panel report on Jan. 1 1th, 1993, and since
several of the panel's recommendations are controversial, no CV study
accomplished to date including the one reported here complies fully with all
of their recommendations. While our study followed the vast majority of the
recommendations, it is useful to point out specific differences between our
study and the panels' recommendations. These relate to recommendations

-------
(3), (7) and (11) as listed above. We know of no literature which provides
Quantitative statistical support for recommendation (11), the provision of a
"no answer" option for respondents—arguments can be made on both sides.
But without quantification, this point remains nothing more than speculation
by the panel. Our speculation differs from the panel, but we feel that this
issue merits little attention. More evidence does exist on points (3) and (7),
although that evidence suggests more of a healthy debate than overwhelming
support for a particular position.

Point (3) argues that in person or telephone surveys are inherently
superior to mail surveys Since the only member on the panel with
extensive survey data collection expertise was Howard Schuman, this
recommendation likely represents the position of one individual who is well
known for taking a polar view on an admittedly controversial subject. Since
USEPA has extensively employed mail surveys in their valuation research,
they requested a specific evaluation of this recommendation from Dr. Donald
Dillman, Professor of Sociology at Washington State University, who is also
now Chief Scientist of the U.S. Census. His response has been circulated to
the committee and speaks for itself. We feel that further research on the
issue of mail versus telephone versus in-person interviews is warranted. Our
own view is that all methods can provide useful information if carefully
implemented.

The other major difference between the groundwater study and the
NOAA panels recommendations relates to point (7), use of a referendum
format to obtain values. Prior to the panels' report, the most commonly used
argument in favor of the referendum format was that policy makers would
find the approach appealing. The primary objection was that many people
don't vote, and would find the referendum notion difficult, having never

-------
experienced one. Natural resource damages accrue to non-voters as well as
voters. The panel's recommendation appears, again, to be based on the fact
that their main focus was on the NOAA damage study rather than on the
published, refereed, literature which suggests that the referendum approach
produces larger values than those obtained by traditional open ended or
payment card valuation questions (see, for example Seller, Stoll, Chavas,
1985 and Schuzle, McClelland, et. al. 1989). Further, it has been shown in
the experimental literature that the iterative hypothetical bidding process
(used in the NOAA sponsored study by Mitchell and Carson) produces bids
higher than those from an actual incentive compatible auction using real
transactions (Coursey, Schulze, Hovis, 1987). This study showed that a
direct question asking for maximum hypothetical willingness to pay was an
accurate predictor of real willingness to pay. The payment card approach
was chosen both because it produces conservative values and because it
reduces scenario rejection as noted above. The context for our valuation
question was precisely patterned after the well known Smith public good
auction which has been shown to be incentive compatible both in the
laboratory and in practice (public television uses the Smith auction to collect
funds from local stations to support national programming such as McNeil
Lehrer). The key feature is that money is returned if the public good is not
funded.

3. Issues in the Use of the Benefit Estimates

The question arises as to how the benefit estimates from the
groundwater survey can be used for determining benefits from groundwater
cleanup for the nation. Specifically the question arises as to the utilization of
the estimates for determining the national benefits from remedial actions

-------
with respect to contaminated groundwater under RCRA. Prior to attempting
to examine the potential for aggregating benefit estimates from the
groundwater survey we need to address some substantive issues not fully
explicated in the McClelland et al. report. These include: 1) non-response
'bias, 2) sampling bias and 3) the question of "extent of market" or
community size to which these individual benefit estimates should be
applied.

3A. Treatment of Non-Respondents

The mean non-use value of $2.81 to $3.54 derived in Table 7.21 (p.

193) using three independent approaches obtain for those individuals in the
regression sample: 1983 individuals. This group comprises a subsample of
the original sample of 4533 individuals (see Table 6.2, p. 143). For purposes
of applying these mean non-use benefit estimates we need to determine if
the "non-regression sample" population has different values and if so how to
treat these for aggregating population benefits. As illustrated in Table 6.4 (p.

146) there is evidence that those individuals who answered fewer questions
in the groundwater survey (and thus self-selected out of the regression

sample by item non-response) had lower willingness to pay values than those
in the regression sample. These individuals stated a positive mean raw
willingness to pay amount equal to roughly one half of the mean raw
willingness to pay for those in the regression sample. A reasonable
treatment of non-respondents would thus be to consider an upper bound
approach in which non-respondents have a WTP equal to one half that of
respondents and a lower bound approach which would be to assume that
non-respondents have a zero value.

-------
A second way of examining the issue of non-response bias asks
whether individuals who do not respond have lower values. A priori it might
be expected that individuals with lower values do not respond at all to the
survey or, if they do, may delay their response until they have been
"prodded" sufficiently by follow-up postcards and additional reminders. The
results from the groundwater survey are actually counter-intuitive to this
supposition. Table 7 presents the respondents mean willingness to pay as a
function of how long it took for individuals to return the survey instrument
after the first mailing.

TABLE 7: MEAN REDUCED WTP AS A FUNCTION OF TIME TO RETURN

U1&P0NSU DATE

MEAN REDUCED WTP

(STD DEV)
n

EARLY

11.42

Oct. 31- Nov. 15

(25.79)

1204

MIDDLE

11.81

Nov. 16- Dec. 8

(26.29)

906

LATE

14.85

Dec. 12- Jan. 7

(41.29)

The surveys were date stamped upon receipt. By matching date of
receipt with mean reduced willingness to pay we are able to test for any
change in mean WTP as the length of time to return increases. As can be
seen from Table 6 the mean WTP appears to increase over time. This
essentially rejects the notion that individuals with less interest or concern,
and thus less willingness to pay, for the commodity will take longer to
respond. The increase in mean WTP is not statistically significant however
(F = 0.58, df = 2,180, n.s.). The variance in bidding does increase
dramatically in the last group. This result suggests that our

-------
recommendation in the report that non-respondents should be assumed to
have a zero value is likely to underestimate values.

3B. Sample Bias

All known methods of survey data collection produce sample bias.
Unfortunately, the best approach for reducing sampling bias is the use of
personal interviewers. This led Schuman to argue against the use of mail
surveys for contingent valuation in his contribution to the evaluation of the
contingent valuation method by a panel commissioned by the National
Oceanic and Atmospheric Administration (NOAA, 1993). However, use of a
sample-based approach which relies on costly personal interviews would
severely limit the use of CV and further prohibit benefit transfers in many
settings 10. The sample-based approach involves obtaining as complete and
unbiased a sample as possible for use in computing a mean value for the
benefit assessment.

In contrast to the sample-based approach, a model-based approach,
which uses econometric modeling to estimate a benefit function can be cost
effective when applied to many situations. In this approach the admittedly
biased sample of, for example, a mail survey is used to estimate a model of
household willingness to pay, and relatively unbiased census tract data is
used to predict benefits using the estimated model. Thus, in applying a

10Boyle and Bergstrom ( 1992) describe benefit transfers as "the transfer of existing estimates
of non-market values to a new study, which is different from the study for which the values
were originally estimated... [T]his Is simply the application of secondary data to a new policy
issue." The transfer of values to new scenarios has received increasing attention particularly
in situations where financial and time limitations preclude the individual site specific studies.
Issues in benefit transfers related to the topic of water quality benefits was recently examined
In a series of articles (Smith 1992, Desvousges et al. 1992, Brookshlre and Nelll 1992. Boyle and
Bergstrom 1992, Loomis 1992, Atkinson et al. 1992). Some of the topics that were examined
included regression analysis, obtaining and analyzing data, and the evaluation of policy
changes that occur from the original study site to the site where the benefit transfer has
occurred.

4 3

-------
model-based approach, the investigator can correct for biases such as the
tendency to obtain more responses from older, wealthier respondents.
Obviously this procedure raises a number of statistical questions and
requires a number of assumptions to be valid.

An ongoing statistical debate has, thus, arisen concerning sample-
based versus model-based approaches (see Smith, 1983, Sarnadal, 1978,
and Hansen et al., 1983) Unfortunately. Schuman did not consider model
based approaches in his evaluation of the CVM. Given the significant
differences in the costs of the two approaches, there are substantial'
potential benefits associated with research that can further justify the use of
model-based methods. The model-based approach we recommended for
application of our study was relatively unsophisticated and much more could
be done in future research.

3C. Extent of Market

Once a function predicting household values has been estimated using
a model-based approach, the question arises as to the relevant market area
over which to apply the model. Given that the groundwater survey asked
individuals for their willingness to pay for groundwater cleanup for their
community, the appropriate market is the individual's community. To
support this contention one must first show that individuals have little or no
value for groundwater cleanup in communities other than their own.

As discussed in the groundwater report, Versions B of the survey
pursued the question of how much individuals are willing to pay for
groundwater cleanup in other communities. As would be expected
individuals valued local cleanup vastly higher than groundwater cleanup
outside of their community. Mean non-use values for national groundwater

-------
cleanup were less than $1.00 per household with a modal value of zero.
Given our analysis of skewed error in hypothetical values we conclude that
national non-use values are not significantly different from zero.

This leaves the question of appropriate community size. The
individual's concept of community size is an empirical issue which we
examined in post-testing of the groundwater survey in March of 1993.
Using the national mail groundwater instrument, we modified the survey
following the initial valuation question to include questions about the
community size respondents assumed in their answers. Figure 5 shows the
distribution of respondent's concept of community size related to the local
political unit. Based on this research we concluded that the appropriate
market is the smallest local political unit, such as a town or city.

FIGURE 5: COMMUNITY SIZE COMPARED TO POLITICAL UNIT

49%

Leaa thin Political unit Greater than

Community Size

Figure 6 shows a distance measure of community. The bimodal
distribution results from a lower mode consistent with the concept of a
"neighborhood" (1 to 1.9 miles radius) and the upper mode consists of

-------
households who primarily view their community as consisting of their
smallest local political unit (4 to 7.4 miles radius). Note that since post-
testing was conducted in Denver, the smallest political unit is generally
larger than those that exist in the Eastern portions of the U.S. where
townships and/or boroughs often consist of smaller areas.

FIGURE 6: COMMUNITY SIZE AS FUNCTION OF RADIUS OF COMMUNITY

* 1

30 "

1 20"

10 -

0 "

Mile Radius N

3D. Estimation of National Benefits

Industrial Economics, Incorporated (IEc) has used the results from
the McClelland et al. study to calculate non-use benefits of the RCRA
corrective action rule. This process involved several steps including:

• estimate individual's total WTP based on site characteristics at
relevant sites

• estimate non-use values based on the individual total WTP using the
extrapolation approach

• calculate the present value of the monthly willingness to pay (over
ten years as in McClelland et al.) using an annual discount rate of 4
percent

36.1%

-------
• aggregate individual values across the number of households in the
place where the site is located

• calculate a national benefit estimate for non-use values from
groundwater cleanup using sample multipliers

• conduct a sensitivity analysis with respect to assumptions including:

1) population demographics, 2) size of contaminate-on plume, 3)
dependence on: groundwater, 4) contamination scenario, 5) ,

effectiveness of the remedy, 6) non-response biases, 7) the extent
of market and 8) the number of sites where non-use benefits apply.

Using the most reasonable assumptions (using only sites where
remediation achieves drinkable water, only sites with off-site groundwater
contamination, counting benefits as accruing when remediation is complete,
using the extrapolation approach of McClelland et al.), IEc estimated non-
use benefits from groundwater cleanup as a function of community size as
shown in Table 8. The base estimate derived by IEc is $276 million (1992
dollars) using a market size of "place of residence." From the survey post-
testing, the average place of residence size that appears appropriate for use
in aggregating across the individuals community is slightly less than a 5 km
radius.

TABLE 8: TOTAL NONUSE VALUE BENEFITS FROM COMPLETE
GROUNDWATER CLEANUP AS A FUNCTION OF MARKET SIZE

MARKET SIZE

TOTAL NONUSE BENEFITS
(millions of 1992 dollars)

PLACE OF RESIDENCE

276

5 KM

283

10 KM

1,012

Assuming a zero value when the market radius is zero we can graph
total non-use values as a function of market size in terms of kilometers
radius. Thus Figure 7 graphs total non-use benefits as a function of the

-------
extent of market (in kilometers radius) and demonstrate the sensitivity of
the analysis to market size. Further sensitivity analysis shows that the total
national estimate ranges from $134 million to $5,559 million. The $134
million is derived using lower bound assumptions (non-respondents value
equals zero, only counting sites with offsite well contamination, households
within 5 km, benefits accruing only when remediation is complete). Using
all upper bound assumptions (10 km radius extent of market, benefits
accruing during remediation, non-respondents having a WTP equal to 1 /2 of
respondents and including sites without offsite contamination) leads to the
$5g559 estimate.

FIGURE 7:

TOTAL NONUSE BENEFITS AS A
FUNCTION OF EXTENT OF MARKET

3E. Order of Magnitude for Benefits and Costs

Using place of residence as the best estimate of community size we
obtain a total non-use benefit estimate of $276 million. How accurate does

-------
such an estimate have to be to be useful for policy purposes? Consider the
following case: Total program costs are believed to be two hundred million
dollars. Since the benefit estimate of $276 million is sensitive to a variety of
assumptions about market size, treatment of non-respondents, and other
factors, serious questions might arise as to whether or not benefits cover
costs. On the other hand if costs are reliably estimated to be ten billion
dollars, then costs are very likely to exceed benefits (at least from non-use
values). The necessary precision of the benefit estimate in this case needs
only to be within an order of magnitude.

4. Conclusion

We would like to thank the committee for the opportunity you have
given us to consider your thoughtful and provocative questions. Investigators
in the policy arena are only rarely given a chance to carefully reevaluate their
own research.

Table 9 summarizes the empirical results of the four approaches for
estimating non-use values used in the national mail survey and post-testing
of the survey instrument. Each of these approaches for estimating non-use
values for groundwater cleanup derive from different cognitive tasks. While
we feel each of these approaches include existence values they likely include
varying degrees of bequest value. Altruistic values are excluded from all of
these methods. For comparison across methods, non-use values range from
a low of 3696 to a high of 50% of total value for complete groundwater
cleanup.

As to the question the research was asked to answer "Do non-use
values exist for groundwater cleanup?" we believe (in spite of our own initial
skepticism) that they do. As to the related question of the magnitude of

-------
TABLE 9 COMPARISON OF DIFFERENT APPROACHES TO ESTIMATING

NON-USE VALUES METHOD (PREDICTED VALUES)11

MEAN NON-USE VALUE
(STD DEV)
n

NON-USE VALUE AS
PERCENT OF TOTAL
VALUE

PERCENT SPLITS

3.49
(3.97)

1126

49.8%

SCENARIO
DIFFERENCES

(3.11)

345-

41.1%)

EXTRAPOLATION

(QUADRATIC)

3.54

(5.86)
344

49.2%

(LINEAR)

2.79
(4.60)

355

38.8%

TEMPO-
VALUATION

2.33
(5.61)
68

36.3%

non-use values, we conclude that a substantial portion of total value for
groundwater cleanup at NIMBY sites derives from non-use motives since
each of the four methods we have employed to attempt to partition total
value yields substantial non-use value. It should be stressed that it is unlikely
that people who do not know of the existence of contaminated water could
plausibly hold non-use values. However, since each of these four methods
for partitioning values is inherently imperfect, we do not feel that we have
achieved great precision in this allocation of values. In our view, the need
for precision depends greatly on the relative magnitude of benefits and
costs.

1 *To adjust the temporal valuation from the post-test values to predicted values as used in the
other non-use value estimation approaches, reduced willingness to pay was adjusted by the
percent that willingness to pay decreased in using the Box-Cox transformation to generate
predicted values in the national mail survey. The percent of total value Is calculated as the
mean non-use value divided by the mean total value for each of the four subgroups used in the
different approaches.

-------
5. References

Andreoni, James, "Giving with Impure Altruism: Applications to Charity and
Ricardian Equivalence" Journal of Political Economy, vol.97, no.6,

1447-1458, 1989

Atkinson, S.C., T.D.Crocker, and J.F. Shogren, Bayestan Exchangeability,

Benefit Transfer, and Research Efficiency, "Water Resources Research
March, 1992, Vol. 28:3, pp. 715-722.

Bostrom, Ann, Baruch Fischhoff, and M. Granger Morgan, "Characterizing
Mental Models of Hazardous Processes: A Methodology and an
Application to Radon." Journal of Social Issues, Vol. 48, No. 4, 1992,
pp. 85-100.

Boyle, K.J. and J.C. Bergstrom, 1992. "Benefit Transfer Studies: Myths,
Pragmatism, and Idealism," Water Resources Research, Vol. 28:3,
pp. 657-663.

Boyle, Kevin J. "A Review of Contingent-Valuation Studies of the Benefits of
Groundwater Protection" Center for Economics Research, Research
Triangle Institute, Research Triangle Park, NC, draft report to USEPA,
April 1993

Brookshire, D. and H. Neill, 1992. "Benefit Transfers: Conceptual and

Empirical Issues," 2. "Benefit Transfer Studies: Myths, Pragmatism,
and Idealism," Water Resources Research Vol. 28:3, pp. 651-656.

Brookshire, David S., Ralph C. d'Arge, "Experiments in Valuing Non-

marketed Goods: A Case Study of Alternative Benefit Measures of Air
Pollution Control in the South Coast Air Basin of southern California,"
in Methods Development for Assessing Tradeoffs in Environmental
Management, vol. 2 EPA-60076-79-0016, Washington D. C., NTIS,

1979

Carlson, Bruce W., "Anchoring and Adjustment in Judgments Under Risk,"
Journal of Experimental Psychology, Vol. 16:4, 1990. p 665.

Coursey, D., J. Hovis, and W.D. Schulze, "On the Supposed Disparity Between
Willingness to Accept and Willingness to Pay Measures of Value,"
Quarterly Journal of Economics, 679-690, 1987.

Desvousges, W. H., M.C. Naughton, and G.R. Parsons, "Benefit Transfer:
Conceptual Problems in Estimating Water Quality Benefits Using
Existing Studies," Water Resources Research Vol. 28:3, pp. 675-
683.

-------
Edwards, Steven F. "Option Prices for Groundwater Protection", Journal of
Environmental Economics and Management, 15, 475-487, 1988

Fischhoff, B. and L. Furby "Measuring Values: A Conceptual Framework for
Interpreting Transactions with Special Reference to Contingent
Valuation of Visibility," Journal of Risk and Uncertainty, Vol. 1:147-
184. 1988

Freeman, M.A., "Non-Use Values in Natural Resource Damage Assessment"
Natural Resources Damages: l aw and Economics. 1992

Greenley, Douglas A., Walsh, Richard G. and Young, Robert A., "Option
Value: Empirical Evidence from a Case Study of Recreation and
Water Quality", Quarterly Journal of Economics, November 1981,
Vol. 96. No. 4, 657-673.

Greenley, Douglas A., Walsh, Richard G. and Young, Robert A.,

Economic Benefits of Improved Water Quality: Public
Perceptlo ris of Opton and Preservation Values. Westview Press.
Boulder, Colorado, 1982

Hansen, M. H., W.G. Madow and B.J. Tepping, "An Evaluation of model-

dependent and probability-sampling inferences in sample Surveys,"
Journal of the American Statistical Association , 78:776-807.

Howe, Charles W. 'The Impact of Price on Residential Water Demand: Some
New Insights" Water Resources Research, Vol. 18, No. 4, PP. 713-716,
August 1982

IEc, "Chapter 11: Non-Use Benefits of Ground Water Remediation" internal
working draft, 17 December 1992, Industrial Economics,

Incorporated, Cambridge, MA

Kahneman, Daniel, "Reference Points, Anchors, Norms, and Mixed

Feelings," Organizational Behavior and Human Decision Processes, Vol
51, 1992, pp. 296-312.

Laze, Jeffrey K., William D. Schulze and Gary H. McClelland, "Economic

Theory and Psychology of Non-Use Values" working paper, University
of Colorado, 1993

Loomis, J. B., 'The Evolution of a More Rigorous Approach to Benefit

Transfer: Benefit Function Transfer," Water Resources Research,
March, 1992, Vol. 28:3, pp. 701-705.

Madariaga, B. and K.J. McConnell, May 1987, "Exploring Existence Values",
Water Resources Research, Vol. 23:5:936-941.

McClelland, Gary H., William D. Schulze, Jeffrey K. Laze, Donald M. Waldman,
James K. Doyle, Steven R. Elliott and Julie R. Irwin "Methods for

-------
Measuring Non-Use Values: A Contingent Valuation Study of
Groundwater Cleanup", October 1992 Draft report to USEPA

McClelland, Gary, William Schulze, and Brian Hurd, 'The Effect of Risk
Beliefs on Property Values: A Case Study of a Hazardous Waste Site,"
Risk Analysis, 1990

McClelland, Gary, William Schulze, Donald Waldman, Julie Irwin, David

Schenk, Thomas Stewart, Leland Deck. Mark Thayer, 'Valuing Eastern
Visibility A Field Test of the Contingent Valuation Method, USEPA
Cooperative Agreement CR-8 15183, June 1991.

Mitchell, Robert Cameron and Richard T. Carson "Existence Values for

Groundwater Protection", Draft final report to USEPA, Resources for
the Future, May 1989

Nieswiadomy, Michael "Estimating Urban Residential Water Demand: Effects
of Price Structure, Conservation, and Education" Water Resources
Research, Vol. 28,' No. 3, pp. 609-615, March 1992

Northcraft, Gregory B, and Margaret A. Neal, "Experts, Amateurs, and Real
Estate: An Anchoring-and-Adjustment Perspective on Property
Pricing Decisions," Organizational Behavior and Human Decision
Processes 39, 84-97, 1987.

Poe, Gregory L. and Richard C. Bishop, "Application of a Convolutions
Approach to Measuring Differences in Benefit Measures from
Dichotomous Choice Contingent Valuation." Selected paper, American
Agricultural Economics Association Annual Meetings, 1992

Roth, Emilie, M. Granger Morgan, Baruch FischhofF, Lester Lave, and Ann
Bostrom 'What Do We Know About Making Risk Comparisons?" Risk
Analysis, Vol. 10, No. 3, pp. 375-387, 1990

Sarnadal, C. E., 1978. "Design-based and model based inference in survey
sampling." Scandinavian J. of Statistics, 5:25-52.

Schulze, W. D., D.J. Schenk, S.R. Elliott, R.R. Boyce, J.R. Irwin, T. Stewart, P.
Slovic, L. Deck and M. Thayer. Field and Laboratory Experiments on
the Reliability of the Continent Valuation Method USEPA
Cooperative Agreement #CR-812054, May, 1989.

Schulze, William, Gary McClelland, Donald Waldman, Julie Irwin, Thomas
Stewart, Leland Deck. Paul Slovic, Mary Thayer, "Valuing Visibility A
Field Test of the Contingent Valuation Method, USEPA Report,
Cooperative Agreement CR-812054, March 1990.

Seller, Christine, John R. Stoll, Jean-Paul Chavas, "Validation of Empirical
Measures of Welfare Champs: A Comparison of Non-market
Techniques." Land Economics Vol. No. 2, Mary 1985.

-------
Smith, T.M.F. 1983. "On the validity of inferences on non-random
samples." J. of the Royal Stat. Sot.. 146, series A:394-403.

Smith, V.K. and W.H. Desvousges, "The Value of Avoiding a LULU: Hazardous
Waste Disposal Sites" ReStat, 1986

Smith, V. K., 1992, "On Separating Defensible Benefit Transfers From

"Smoke and Mirrors , Water Resources Research^ Vol. 28:3, pp.685-
693.

Tolley, George S., and Alan Randall, with G. Blomquist, M. Brien, R. Fabian,
M. Grenchik, G. Fishelson, A. Frankel, J. Hoehn, A. Kelly, R Krumm, E.
Mensah, and T. Smith, Establishing and Valuing the Effects of
Improved Visibility in the Eastern United States, final report to the
U.S. Environmental Protection Agency, 1985

Tversky, A., and Kahneman, D., "Judgment Under Uncertainty Heuristics
and Biases" Science, 1985, pp. 1124-1131.

-------
6. Appendices

6A. POST-TEST gUESTION ON TEMPORAL ASPECTS OF VALUATION

Q19. Each site requires different levels and types of cleanup technology.
Some sites will be cleaned up quickly while others may takes years
to clean up due to different types of contaminants and varying soil
and water conditions. When you indicated how much you are willing
to pay for cleanup, how long did you expect the cleanup to take?

Q20 Of the dollar amount you stated you are willing to pay on your
monthly water bill for the next tens years for complete cleanup,
what percent would you be willing to pay if you were now informed
that cleanup would be completed ten (1 O) years from now?

0% 5% 10% 20% 25% 30% 40% 60% 60% 70% 75% 60% 90% 100% 125% 150% 200% 250%+

t \ f t t t MORE THAN

NONE ONE FOURTH HALF AS THREE FOURTHS PAY THE ONE AND A

As MUCH MUCH AS MUCH SAME HALF TIMES T*°AND *

AS MUCH HALF ™E8
AS MUCH

Q20 "...cleanup would be completed thirty (30) years from now?"
Q21 "...cleanup would be completed one hundred (100) years from now?"

-------
6B. VIEWGRAPHS USED IN PRESENTATION TO THE COMMITTEE BY
PROFESSOR WILLIAM D. SCHULZE

-------
OUTLINE OF RESPONSE

1 ) MISSION

2 ) COMMODITY, CONTEXT,

INFORMATION, DESIGN AND
THEORETICAL QUESTIONS

3 ) MARKET SIZE

4 ) DATA ANALYSIS

-------
MISSION

Question:

Do any non-use benefits derive from
corrective actions regarding groundwater "
contamination? If so, how large might they
be?

Prior studies of air quality undertaken to
satisfy OMB concerns:

• Denver Brown Cloud: 8 survey design
variants

East Coast Visibility: 2 survey
design variants

Familiar vs. Exotic Commodity

-------
TABLE I

SUMMARY OF BROWN CLOUD SURVEY DESIGN FEATURES

BASE

THREE

VOTNG

FREQ
DIST.

3 COMMODITY
COMPARISON

VERSION

G H
(WTP)j(Choice)

RESPONSE

FRAME

WTP

x "

x i

WTA

X ! X

HEALTH vs.
VISIBILITY

3 Questions

X Split

FORM OF
THE VALUE
QUESTIONS

Std. CVM

x : x

x i x

x :

Voting

Choice

DESC RIPTION

OF CHANGE IN
AIR QULAITY

Average Air
Quality Change

X 1 X

X :

Freq. Distribution
of AirQual. Change

CONTEXT/

INFORMATION

CONTW

Health
Information

Extra Context

Minimal Context

X i X

-------
CONCLUSIONS FROM PREVIOUS

STUDIES

For a familiar commodity:

Voting context had no effect on
values

• Information had little effect on total
values

Embedding disaggretation problems
are serious

-------
COMMODITY: Non-use values for
groundwater cleanup

• Prior pretesting effort by Mitchell
and Carson (1989) showed this to be
a very difficult commodity. They
found:

1 ) People know little about

groundwater

2 ) People rejected an existence value

scenario where groundwater would
never be used

Groundwater cleanup provided a
perfect commodity to test the
methodological limits of contingent
valuation

Complete intellectual freedom
provided by OSW and OPPE

Limited budget for study

-------
COMMODITY DEFINITION

USEPA wanted benefits of complete
groundwater cleanup

• Containment was proposed as a
backup technology where complete
cleanup was technologically impossible

Complete cleanup provides a vector
of services:

(1) In s o m e cases it provides clean
water for use by the present
generation (use and altruistic
value)

(2) Clean water for use by future
generations (bequest value)

(3) Knowledge that "mother earth" is
not contaminated (existence value
or moral value)

Disaggregation of such values has

proven difficult.

-------
UNDER WHAT CIRCUMSTANCES MIGHT NON-
USE VALUES ACTUALLY EXIST?

Contrast (1) Expert Benefits
with (2) Subjective Benefits

(1 ^ Expert Benefits are defined as the

benefits experts believe to exist
(e.g. Value of Life X Expert
Assessment of Risk Reduction X
Exposed Population). This
measure excludes non-use values.

m Subjective Benefits are defined as
the values potentially exposed
populations themselves place on
environmental cleanup. These
will be based on perceived risks
and may include non-use values.

Consumer Sovereignty vs. Expert
Assessments

To obtain non-use values one must
measure subjective values.

-------
SUBJECTIVE VALUES

• Substantial subjective values have
been shown to exist for NIMBY sites
in a large number of studies using
property values and/or contingent
values where expert risks are very
small

(1)V.K. Smith and W.H. Desvousges
"The Value of Avoiding a LULU:
Hazardous Waste Disposal Sites"
ReStat 1986.

(2) McClelland, Schulze, Hurd, "The
Effect of Risk Beliefs on Property
Values: A Case Study of a
Hazardous Waste Site," Risk
Analysis 1990.

-------
GROUNDWATER SCENARIO DEVELOPMENT

(1) Offsite groundwater contamination
of wells assures public awareness

(2) Public officials have never
knowingly allowed contaminated
water from a NIMBY site into a
public water supply for fear of
public outrage no matter how
small the expert assessment of risk

(3) Risk information in the pretest
survey caused 15% of pretest
subjects to raise value for
cleanup, 9% to lower value and
had no effect on 76%.

(4) Risk communication has been
surprisingly ineffective in real
world NIMBY situations

(5) Values from this study should
most appropriately be applied to
sites with actual or potential
offsite contamination of wells.

-------
DEFAULT ASSUMPTIONS IF COMPLETE
CLEANUP IS NOT FUNDED

Very large values obtained in a
previous contingent valuation study
of groundwater (Steven Edwards,

"Option Prices for Groundwater
Protection," .TEEM, 1988).

Default Assumptions (Fischoff and
Furby, 1988)

People may fear no groundwater will
be available for themselves or future
generations (No substitutes)

Specify several default alternatives
(substitutes for complete cleanup of
groundwater available to current
and/or future generations) to eliminate
fear of "no water" such as

1 ) Home treatment

2 ) Public treatment (most favored)

3 ) Shortage not 10070 (surface water)

4 ) Containment.

-------
PROVISION OF SUBSTITUTE
COMMODITIES

CONSIDER THE EFFECT OF SUBSTITUTES ON
USE VALUE (DEMAND FOR WATER)

A1 = CONSUMER SURPLUS WITHOUT A SUBSTITUTE

A2 = CONSUMER SURPLUS WITH A SUBSITUTE OF HOME
TREATMENT

• CONSUMER SURPLUS LOWER STILL IF
Pp IS ASSUMED TO BE LESS THAN Ph

• ALTRUISTIC AND BEQUEST VALUES
ALSO LOWERED IF RESPONDENTS
ASSUME FUTURE GENERATIONS HAVE
SUBSTITUTES

-------
IMPLIED DEMAND FOR WATER

The survey was designed so that if complete
cleanup is not funded, respondents would
likely assume a cheaper alternative would
be funded (such as public treatment) or as a
last resort, home treatment. The small
value obtained for cleanup with a 70%
dependence on groundwater suggests that
respondents were assuming substitute
alternatives were available

PREDICTED WTP
as a function of precent of water

O 10 20 30 40 80 60 70 80

PERCENT OF WATER FROM GROUNDWATER

-------
WHAT IF YOU DEPENDED LESS OR MORE ON GROUNDWATER

Dependency on groundwater is different for every location at which
contamination has occurred. Some areas use groundwater for all of their
domestic water supply while others use none. To plan new groundwater cleanup
programs that could cost you money, " decision makers want to learn how much
clean groundwater is worth to people like you in these different situations.

WHERE GROUNDWATER SUPPLIES 10% OF DOMESTIC WATER

Q16 Consider an imaginary leaking landfill identical to that described above
except that now groundwater supplies 10*A of the domestic water supply
instead of 40%. Remembering that, on average, households use half of
their domestic water outdoors, one third in the bathroom and the rest in
the kitchen how satisfied are you with water rationing as an option where
water use would have to be cut by 10%?

NOT AT AU EXTREMELY

SATISFIED SATISFIED

1 2 3 4 5 6 7 (3.64)

Q17 What would a complete cleanup program like that described in Q6
be worth to your household if your imaginary community faced a
groundwater problem where 10% of the local domestic water supply .
comes from groundwater which was contaminated and could not be
used without treatment? In answering you should assume that:

¦ The hypothetical situation is now one in which only 10% of the
water you use in your community comes from groundwater
resources. The other 90% of your water comes from surface water
sources such as lakes and streams.

¦ The complete cleanup program is identical to the program

described in the previous section.

Now, of the dollar amount you would have paid just for complete
groundwater cleanup when faced with 40% of your water supply
contaminated, what percent would you still be willing to pay for
complete groundwater cleanup if faced with 10% of your water supply
coming from contaminated groundwater?

(46.5170)

NONE 90ME HALF MOST ALL

0% 10% 20% 30% 40% 50% 600/0 70% 80% 900/0 100%

-------
WHERE GROUNDWATER SUPPLIES 70%_OF.DOMESTlC WATER

Q18 Consider an imaginary leaking landfill identical to that described above
except that now groundwater supplies 70% of the domestic water supply
instead of 40°/0. Remembering that, on average, households use half of
their domestic water outdoors, one third in the bathroom and the rest in
the kitchen how satisfied are you with water rationing as an option where
water use would have to be cut by 70%?

Q19 What would a complete cleanup program like that described in Q6
be worth to your household if your imaginary community faced a
groundwater problem where 70% of the local domestic water supply
comes from groundwater which was contaminated and could not be
used without treatment? In answering you should assume that:

¦ The hypothetical situation is now one in which 70% of the water you
use in your community comes from groundwater resources. The
other 30°/0 of your water comes from surface water sources such as
lakes and streams.

¦ The complete cleanup program is identical to the program
described in the previous section.

Now, of the dollar amount you would have paid just for complete
groundwater cleanup when faced with 40% of your water supply
contaminated, what percent would you be willing to pay for complete
groundwater cleanup if faced with 70% of your water supply coming
from contaminated groundwater? (Circle the best per cent response)

(166.24%)

SAME TWICE 3 TIMES MORE THAN

10OO/. 125% 150% 175% 200% 250% 300740 3 5 0 % 400%+

NOT AT ALL
SATISFIED
1 2

EXTREMELY

SATISFIED

3 4 5 6 7 (2.35)

AS MUCH

4X AS MUCH

-------
gallons/mo

• Average household price and consumption

from Michael Nieswiadomy " Estimating
Residential Water Demand" WRR (1992)

• Range of Demand Elasticity Estimates

-.11 Nieswiadomy (op. cit. 1992)
-.57 Howe WRR (1982)

-------
IMPACT OF INFORMATION/CONTEXT

• Mean household value fell from $20.22 to
$12.20 per month

• Small samples, n« 40 each

• Debriefings used to explain value
decrease (Table 4.2)

• Conclusion: Information on substitutes
lowered values

• Emphasized substitutes in final survey
design

• Conservative choices

-------
TABLE 4.2 SELF-REPORTED EFFECTS OF CONTEXT - DECEMBER 1990
PRETEST

Self-reported Effects of Context from
December 1990-Retest (Summary of responses to Q-47
through Q-56, Appendix B)

Percentage

Self-

reported
effect

9-47
Pers.
exp.

9-48

Def. of
gwater.

9-49

Speed of
gwater.

9-50

Water
bill

9-51

Buy water
option

No effect

75%

82%

90%

77%

67%

Lowered
value

20%

Raised
value

25%

15%

Percentage

15%

13%

Self-

reported
effect

9-62

Water

cons.

Q-53

Private

options

9-54

Dis-
counting

9-55

W. S. T.
option

9-56

Risk

commun.

No effect

72%

66%

79%

61%

7 6 %

Lowered
value

13%

24%

10.5740

34%

Raised
value

15%

11740

10.5%

15740

-------
VALUE PARTITIONING

(1) Theoretical Issues (% splits)

• Separable Utility Function

• Constant Marginal Utility of Money

(2) Psychological Issues

• Do people have enough information
to partition

• Have people thought carefully
enough about issues to allow
partitioning

-------
PORTIONING:
THEORETICAL ISSUES

Ci = Consumption by generation 1
Wi = water consumption by generation 1
Z = amount of contaminated groundwater
U2 = utility of next generation or others

Separable Utility Function:

(1) DMClWlZ, U2) = V(Ci) + F (Wi) -D(Z) + A(U2)

-------
Willingness to pay for complete cleanup (WTP)
which provides AWi>0, AZ<0 and AU2>0 is
determined by:

(2) Ul(Ci - WTP, Wi + AW, Z+AZ, IJ2 + AU2) =

Ui (Ci,Wi,Z,U2) or

(3) -AV = AF - AD + AA

With constant marginal utility of money
(consumption) i.e., 3V/3Ci= constant:

,_AF_ — AA

av/acv+ w/aci' + lav/ac

Use Existence Bequest

Value Value or

Altruistic
Value

• Constant marginal utility of money is
plausible since estimates of total WTP
are about 1/4 of 1 % of income

• Separability is an empirical question

but non-use values are usually
assumed to be separable which makes
market based measurement impossible
(M.A. Freeman, "Non-Use Values in
Natural Resource Damage Assessment"
Natural Resources Damages: Law and
Economics. 1992)

-------
PARTITIONING

PSYCHOLOGICAL ISSUES

• Pretest respondents rejected fund for
future use

• Also had rejected existence value
scenario

• Alternative scenarios used to educate
respondents and get them to think about
future generations, others' and existence
value.

Option(Rating)

Benefits

Complete Cleanup
(4.35)

Respondents' Use,
Others' Use, Future
Use, Protection of
Earth

Containment
(3.45)

Same as above but
less certain

Public Treatment
(3.74)

Respondents' Use,
Others' Use

Home Treatment
(2.81)

Respondents' Use

-------
COMPONENT ALLOCATION OF TOTAL
WILLINGNESS TO PAY FOR COMPLETE
GROUNDWATER CLEANUP (RANDOM SUBSAMPLE

USE

ALTRUIST

BEQUEST

EXISTENCE

100

150

200

250

300

350

400

450

1 00

500

550

600

650

700

750

800

850

900

950

100

1000

1050

1100

1150

1200

100

1250

1300

100

1350

1400

1450

1500

1550

1600

1650

1700

100

1750

100

1800

1850

1900

1950

2000

100

-------
OTHER DESIGN ISSUES

Did not ask for water bill

1 ) Small positive impact in self-

reported effect of context on
values (Table 4.2)

2 ) Many people didn't know the

answer

• Did not allow" don't know" as a
response

1) Little data is available to
support one position or
another

2) DK response provides an easy out
to difficult questions

3) Creates econometric problems by
encouraging missing observations

4) Failure to include DK may
encourage "bad answers"

5) Need testing.

-------
DESIGN ISSUES (CONT.)

• Payment card

1) Used approximate geometrically
increasing values where upper
limit is chosen not to truncate
values

2) Rowe et al. study (1993) using
payment card values of the form
(1 + X)nforn = 0,1, . . N values
shows no effect of varying X
unless the Nth value (last value)
truncates rhs of value
distribution.

-------
DESIGN ISSUES (CONT.)

. Referendum with dichotomous choice

vs. payment card

1) Brown Cloud study showed no
impact of referendum context
alone (open ended value question)

2) Values very similar in
preliminary pretest and final
pretest of groundwater survey.

3) Need more research but doubt
large differences will be found
between payment card and
dichotomous choice because they
are similar cognitive tasks.

However, surprises are not
uncommon.

4) We know of no data based
evidence to pick one approach
over the other.

-------
6C. VIEWGRAPHS USED IN PRESENTATION TO THE COMMITTEE BY
PROFESSOR GARY H. MCCLELLAND

-------
Resolution of
Statistical Issues

• Interdisciplinary confusions
about mathematically
equivalent terms and
procedures

• Draft report for a different
audience

• Appropriate Level of
Precision

— Sensitivity Analysis

— Policy context,
comparison to costs

-------
Oversampling of
NPL Households

• EXPOSED dummy variable not
significant in original
analysis

• Test of Equality for Separate
Regressions: F(l 5,1 967)=0.96

• Search for any pairwise
differences on any variables

-------
Exposed vs Non-
Exposed Differences

— Demographically similar

INCOME, KIDS, AGE, EDUC, GENDER
no sign, cliffs
marginally more WHITE

— USE, more likely to use groundwater or
be aware of it

— SOURCES, more aware of sources of
local gwater contamination

— RECYCLES, more likely to be a recycler

— COMPLETE, MEANNCOM, & RESPONS are
lower in exposed, (attitude change as a
function of experience?)

— Predicted WTP

Exposed: $7.77 Non-Exposed: $6.90
n.s.

-------
RESPONS

Responsibility Variable

Strong predictor of WTP
(t= 1 8.6)

Exogenous or Endogenous

Not available for policy

Sensitivity to its omission

-------
Omitting RESPONS

(Box-Cox)

X'. .15 to .13

R2: .30 to .18

Other variables:

— OTHENV no longer sign.

— MEANNCOM now sign.

— no sign changes

Predicted WTP: $7.01 to
$6.48

RESPONS quadratic effect

-------
Box-Cox Estimation

WTPA—1

WTF = ——— if X*0

= log(y) if X - 0

Handling WTP=0

— including implies X > 0

— prior empirical results
and theory suggested X< o

— so replaced WTP=0 with
WTP = e

— examine sensitivity to

choice of £

Variability of estimate of X

-------
FIGURE 6.6 REDUCED WTP FOR COMPLETE GROUNDWATER CLEANUP

NATIONAL MAIL SURVEY

REDUCED WTP FREQUENCIES (LOG SCALE)

WTP MIDPOINT

FIGURE 6.9: WTP FOR NATIONAL CLEANUP PROGRAM
NATIONAL MAIL SURVEY

NATIONAL WTP FREQUENCIES (LOG SCALE)

» ¦ i ^1 i i r

0 1 2 4 8 16 32 64 128 256 5121024

WTP MIDPOINT

-------
Box-Cox Sensitivity

Zero
set
to

Mean
WTP

Gee.
Mean

[95%C.I.]

Pred.
Mean
WTP

Income
Coef.

10$

11.711

4.49

0.15

[.1 25,.165]

7.01

3.07

1«

11.704

3.71

0.21

_ [.19 S,.235]

7.15

2.99

•U

11.703

3.07

0.24

[.225,.255]

7.23

2.99

-------
TABLE 7.7: LINEAR REGRESSION ON BOX-COX TRANSFORMATION OF

REDWTP (a = 0.15)

Analysis of Variance

Sum of Mean

Source DF Squares Square F Value Prob>F

Modal

2S 290 SI.93129 1162.0772S

34.272 0.0001

Error

1957 663 S7.

S1024 33.90777

C Total

1982 9S409.44153

Root MSE

5.82304

R-square

0.3045

Dep Mean

6.90664

Adj R-sq

0.2956

c.v.

84.31072

Parai

neter Estimates

Paraaatar

Standard T for H O :

Variable DF

Estimate

Error Para

¦ecer-0

Prob > IT!

XVTERCZP 1

-5.832276

1.4531s9s7

-4.014

0.0001

XMCOMEVD 1

0.026445

0.00432236

6.118

0.0001

KIDS 1

0.016927

0.31086890

0.054

0.9566

AGE 1

-0.040563

0.01002827

-4.04s

0.0001

WHITE 1

0.776703

0.4S832938

1.69S

0.0903

EDuc l

0.382S80

0.0808S742

4.732

0.0001

GENDER i

0.100170

0.29853243

0.336

0.7373

NORTHEAS 1

-0.3S2492

0.7s477114

-0.467

0.640S

NEWYORK 1

0.938023

0.68434316

1.371

0.1706

MXDATLAN 1

0.4S3276

0.666910S9

0.680

0.4968

SOUTH 1

-1.129931

0.63122688

-1.790

0.0736

LAKES 1

0.151642

0.6079S37S

0.249

0.8031

SOUTHWES

-0.706929

0.682S2417

-1.036

0.3004

MOUNTAIN 1

0.3208S1

0.824743SS

0.389

0.6973

WEST

-0.195098

0.71671498

-0.272

0.7855

NORTHWES

-0.043883

0.81344447

-0.054

0.9570

LANDFILL

-0.S4S102

0.32416310

-1.682

0.0928

EXPOSED 1

-0.229842

0.22S302S8

-1.020

0.3078

USE *

0.8S0450

0.27S83740

3.083

0.0021

SOURCES ?

0.424435

0.12S82177 ,

3.373

0.0008

RECYCLES

0.08S447

0.06S23121

1.310

0.1904

OTHENV 1

-0.606282

0.17686973

-3.428

0.0006

GRNDWTR 1

0.469827

0.13647S64

3.443

0.0006

COMPLETE 1

0.35824S

0.077S2901

4.621

0.0001

MEANNCOM 1

0.223914

0.12073641

1.8SS

0.0638

RESPONS 1

1.S92300

0.08s80801

18.557

0.0001

Variable Label N Mean std Dev

REDWTP reduced wtp 231S 11.5783S85 2S.9979281

PREDWTP pred in dollars 1983 7.0077342 5.2925489

-------
FIGURE 1:

400 i

RESIDUALS FROM REGRESSION ON
UNTRANSFORMED WTP

-21-16 -Is -12-9 -6 -s o Slllt 18 U 21
RESIDUAL

FIGURE 2:

600 *1

RESIDUALS FROM REGRESSION ON
BOX-COX TRANSFORMED WTP

-21 -IS -18 .IS -0 4 -* O 3 • 9 12 Is IS 21

RESIDUAL

-------
Version D

(1 0%, 40%, 70% Shortages)

• Separate analysis for this
group?

— Yes, and it made no
difference on percent splits

. Correlated errors from
multiple responses

WTP = /30 + (3x(%Short) + P2(%Shortf + e,

Estimated Po for each person

3 eqns and 3 unknowns for each
person

Solves correlated error problem

-------
Tllff 7.21: COMPARISON OP DIFFERENT APPROACHES TO

ESTIMATING NON-USE VALOM (for predicted values)

METHOD

BEQUEST PUIS
EXISTENCE

ffSSDARD

WVU1HW

FEKCJiN 1 SPLITS

3.49

1126

SCENARIO
DIFFERENCES

2.81

xii

345

EXTRAPOLATION

3.54

5.86

344

-------
Estimate Reliability
within Respondent

$o vs. %splits

Mean cliff = $.08
n = 3 5 4
t = 0.7 8 n.s.

Scenario Diff vs. %splits

Mean cliff = $-1.62
n = 337

t = -2.93 p < .005

— public treatment option
may contain bequest value

— therefore, scenario
difference may underestimate
non-use value

-------
a. ao

(0 IS

10
s-
•'

WTP AS FUNCTION OF SHORTAGE
UNBAR AND QUADRATIC

PREDICTED AND RAW VALUES

COMPARISON OF DIFFERENT APPROACHES TO ESTIMATING NON-U8B
VALUESJEOR PREDICTED VALUES)

METHOD

MEAN

ESTIMATED
NON-USE VALUE

STANDARD
DEVIATION

PERCENT
SPLITS

3.94

3.97

1126

SCENARIO
DIFFERENCES

2.81

3.11

345

fcXTRAPOLATION
(QUADRATIC)

3.54

5.86

344

(LDVEAR)

2.89

4.64

344

-------
YOUR COMMUNITY

Q1 8 Different people have different ideas about
the size of their community. Among the
different descriptions of the size of a
community that are listed below, please
circle the letter next to the one that most
closely describes what you think of as
defining the size of your own community.

A. Just my block.

B. Just my block and the next two blocks
in any direction.

C. The area that I could drive from my
house to the edge of in five minutes in
city traffic (not at rush hour).

D. The area that I could drive from my "
house to the edge of in fifteen

minutes in city traffic (not at rush
hour).

E The area that I could drive from my
house to the edge of in thirty minutes
in city traffic (not at rush hour).

F. The entire city in which I live
G The entire county in which I live

H. The entire state of Colorado.

I. The entire nation.

J. Other. Please describe

-------
Community Size Definitions

for Groundwater WTP

-------
Community Size
Definitions for
Groundwater WTP

Community Size

-------
1200

TOTAL NONUSE BENEFITS AS A
FUlfCTKHf OP EXTENT OF MARKET

las 466
EXTENT OF MARKET

RADIUS (KM)

MARKET SIZE

TOTAL NONUSE
BENEFITS

(minions of 1992 doOan)

PLACE OF RESIDENCE

276

5 KM

283

10 KM 1,012

-------