Introduction
Over the last twenty-five years economic applications attempting to value environmental
change and assets have expanded exponentially. While cost ^ benefit analysis (CBA)
and monetary valuation remain only one element in government decisions, and often
not a primary one, they remain ever present and often important in specific contexts.
Thus, cases are easily found of large projects, taxes, and government decisions or
policy justifications involving valuation of this sort. In the environmental context
CBA is evident from global analyses, such as the Stern report, through to local
projects. Environmental economists have been highly vocal in their advocacy of the
CBA approach and of the spread of its individual methods, if critical of some applications.
Their main success has been the contingent valuation method (CVM), which has
supplied numbers for government agencies in North America [especially via benefit
transfer (see Spash and Vatn, 2006)], has been picked up by ecologists to produce their
`ecosystem values', is accepted as legal evidence in courts of the USA, and is often
in the background when monetary numbers are produced for environmental change. In
this paper I offer a reflective overview of the conduct and practice of contingent
valuation in the environmental context.
In trying to address the problem of how to achieve an efficient allocation of
resources, mainstream economists call upon the preferences of individuals to inform
their policy recommendations. These preferences have traditionally been assessed by
analysis of actual market demand to obtain a welfare measure of a proposed project or
policy change. However, because many environmental aspects have no explicit market
price they have been given a pseudomarket or shadow price. This requires either
assessing the welfare changes in markets which are related to the environmental aspect
Contingent valuation design and data treatment:
if you can't shoot the messenger, change the message
Clive L Spash
Commonwealth Scientific and Research Organisation, Sustainable Ecosystems Division,
Canberra, Australian Capital Territory, Australia; e-mail: clive.spash@csiro.au;
website: http://www.clivespash.org
Received 12 September 2007
Environment and Planning C: Government and Policy 2008, volume 26, pages 34 ^ 53
Abstract. The contingent valuation method has become an established and major part of the toolbox
used to produce monetary values for evaluating environmental changes. It has been used to inform
everything from the value of ecosystem services to cultural heritage to loss of life. The method has
been highly controversial at various stages but despite this, or perhaps due to the publicity, it has
grown in scope and scale. Numerous occurrences of `bias' and `anomalies' in results have been
addressed by improved design, so providing guidance on perfected approaches to making sure
respondents reveal preferences in accord with theoretical expectations. That respondents may not
wish to and often fail to conform is seen as a challenge for the design team to be more ingenious with
their incentive mechanisms which get respondents to act `rationally'. Failing this, data can be
classified and treated to derive `conservative' results. I document in this paper how whole areas of
evidence from contingent valuation have been removed from consideration by design, with respondents
expected to conform to an idealised rational agent model or to suffer branding and exclusion as
having the `wrong motives'. While the method is then susceptible to manipulation (eg to meet
sponsors' requirements), if used more scientifically it also holds the potential to reveal fundamental
flaws in economic theory and ways to advance that same theory.
doi:10.1068/cav4
of interest or designing hypothetical markets. The hypothetical market approach is
typically and most commonly applied using the CVM to obtain stated (as opposed
to expressed or observed) preferences.(1) This method was largely developed during the
late 1970s and early 1980s by researchers at the University of Wyoming as an experimental
technique, but was catapulted into the limelight by use as evidence in the courts
of the USA.(2)
In theory, the aim is to assess compensating and equivalent surplus measures of
economic welfare (Hanley and Spash, 1993). In practice, numbers are normally produced
with little attention to such theoretical abstractions. Basically, the CVM asks
individuals to give their maximum willingness to pay (WTP) or minimum willingness
to accept (WTA) compensation for an environmental quality or quantity change. A
popular alternative is to offer a specific amount to be, hypothetically, paid or received
and to ask if this is acceptable or not (ie dichotomous choice).
CVM surveys are claimed to access a range of value categories (ie option, existence,
and bequest values), besides those which relate to the direct use of a resource as
traditionally considered in economics. The collective term for these values in the courts
of the USA is passive use value, although a more theoretically correct term is indirect
use value.(3) The combined effect, of a simple approach and of the large numbers
obtained by including the indirect use value, has been to make CVM studies popular
amongst those trying to show that ecosystems and species loss and environmental
degradation should be taken more seriously in policy circles.
Mainstream economists argue that the figures have the credibility of a `market vote'
from members of the general public which provides political acceptability by conforming
with free-market democratic ideals. Stated-preference methods are backed by detailed,
expert-produced, government-sponsored guides (eg, Arrow et al, 1993; Bateman et al,
2002). This adds an air of rigour and objectivity. In fact, practitioners must make numerous
difficult decisions about survey design, conduct, and analysis. As Hanemann's (1994)
apologia for the CVM states:
``The method, though simple in its directness, is in fact difficult to implement
without falling into various types of design problems that require effort, skill and
imagination to resolve'' (page 38).
The general contention of stated-preference practitioners is that all problems can
be designed away. That is, the basic economic theory is taken for granted and defended
against the occurrence of empirical evidence which shows flaws in that theory. Fundamentally
mainstream economics assert value monism (Norton and Noonan, 2007),
while the underlying message coming from the CVM is of value pluralism (Spash,
2000a). Those economists who fear a harbinger of the need for change in orthodox
thinking may ridicule and dismiss the method. Alternatively, others believe the CVM
has the potential to spread market valuation and have devoted their careers to perfecting
design to remove `biases' and `anomalies'. This has involved preventing some forms
of CVM design, which in effect is an attempt (which has had some considerable
success) at killing off whole areas of research. A more subtle approach has been to
(1) A sister stated preference technique is choice modelling, formerly choice experiments, which is a
strong minority pursuit amongst environmental economists while CVM is applied by noneconomists
and economists alike. In Australia the notoriety and bad press surrounding the Kakadu
CVM study meant only choice experiments survived. Much of what follows applies to all stated
preference methods but I restrict the explicit argument to CVM for brevity.
(2) Cases concerned clean up of industrial-waste sites under Superfund legislation and later the
high-profile Exxon Valdez oil spill case under natural-resource damage-assessment legislation.
(3) The term nonuse value has been used by some authors, particularly in the UK, and is highly
misleading as all economic values, as defined under microeconomic theory, are related to the
usefulness of an entity or objectöthat is, utility for human ends.
Contingent valuation design and data treatment 35
change the message arising from CVM studies by developing various forms of `treating'
data. This raises the potential to meet the expectations of those paying for the studies,
so that the `right' numbers may be forthcoming. I follow Vatn (2004) in arguing for a
middle path between outright rejection and conformist acceptance of the CVM. There
is then a clear case for listening to all that this rare (in economics) primary data source
has to offer.
These issues are addressed in two main sections. First, the approach taken to
survey design and specifically some major debates in the literature are discussed. In
this section three issues are presented: the institutional context of contingent values
and whether they are more political votes or market prices, the emphasis on design for
`conservative' numbers which has excluded WTA formats, and the role of information in
forming and informing preferences. These are selected as areas of persistent debate
which are highly relevant to the contention that observing what humanity does in
practice is far from the primary goal. The second main section then addresses what
economists do with the data they have gathered. Here, I define approaches taken to data
classification, its treatment, and its analysis. The aim is to identify the type of approaches
which are seen as acceptable practice and to identify how this can lead to unscientific
data manipulation. I conclude with some reflections upon the potential for CVM to
play a positive role in the development of some new economic thought about environmental
values, if only the data were gathered openly and were accepted at face value.
Targeting unwanted messengers
A variety of attempts have been made to control for various problems or `biases' by
redesign and refinement of the survey instrument. Most notable amongst these is the
National Oceanic and Atmospheric Administration (NOAA) report (Arrow et al,
1993), arising in the aftermath of the Exxon Valdez oil spill in Alaska and the ensuing
litigation which involved CVM as court evidence. This report involved two Nobelprize-winning
economists from different sides of the court case and set many CVM
industry standards which have persisted through to today as best practice for addressing
problems. While the NOAA report is taken as a major reference point by most
CVM practitioners, there are a large number and range of problems referred to as
`biases' in the literature over the last three decades, extending to a list of at least
twenty.(4) Rather than conduct a comprehensive review of these biases, the objective
here is to draw out key issues, and three illustrative main areas are chosen to do so:
choice format, conservative design, and information provision.
Political vote or market payment
A key change in the 1990s concerning survey design was a shift away from open-ended
bid questions towards the dichotomous-choice format, the main reason being the
NOAA panel recommendation that CVM surveys be essentially self-contained referenda
in which respondents vote to tax themselves (Arrow et al, 1993, page 20). However,
the single dichotomous-choice referendum recommended by NOAA was soon criticised
as exacerbating several problems (eg `yea saying' even for high bids, and mixing protests
with zeros). Bidding games and iterative bidding arose as preferred approaches by many,
while others moved to choice experiments. Choice experiments have been regarded as a
natural progression and have been described as a generalisation of dichotomous-choice
(4) These include the following factors cited as `bias': strategic behaviour, nonresponse, starting
point, range, relation, importance, position, interviewer, sponsor, theoretical misspecification,
symbolic misspecification, part ^ whole, metric misspecification, problem of provision, payment
vehicle, property rights, method of provision, budget constraint, elicitation question, and question
order.
36 C L Spash
contingent valuationöthat is, where a sequence of dichotomous-choice valuation
questions are asked involving two or more goods (Hanley et al, 2001, page 448).
While some claim that dichotomous choice only supplies the `truth' (Hanemann,
1994, page 23), dichotomous choice, in its various forms, is far from problem free. The
results are sensitive to the analyst's selection of bids from which respondents must
choose, and the functional form used for mean estimation adds to variability in results
while there is no clear theoretical reason to choose between forms. Bateman et al
(1999) found that dichotomous choice shows anchoring in single-bounded approaches
and starting-point bias in multiple-bounded cases and
``that the bid amount presented to respondents dominates their responses, to the
virtual exclusion of other socio-economic and preference factors'' (pages 530 ^ 532).
Meanwhile, the problem of `yea saying' creates overestimation. Dichotomous choice
has been found to exceed open-ended bids and to have a greater standard erroröthat
is, substantial variability (Desvouges et al, 1993).
However, debate about the technical differences between the original open-ended
and more recent dichotomous-choice formats can be misleading as a selection guide.
Design of a dichotomous-choice survey requires selecting the bids to be offered to
respondents, which, as noted above, are key to the results. An open-ended approach
is often used to establish the initial bids required for dichotomous choice, so even
dichotomous-choice practitioners then rely upon open-ended formats. On a practical
basis the choice of format is affected by the sample size because dichotomous choice
requires a larger sample to estimate the bid from a probit or logit model. The available
survey budget is then clearly important. Thus, design is not purely motivated by
theoretical concerns.
Indeed, rather than being a purely technical matter, the format selected may have
more to do with the analyst's view on political economy. In defending the CVM against
its critics, Hanemann (1994) follows the line of ``valuation as voting in a referendum''
(page 22), uses this to dismiss CVM work that fails to use ``voting'' formats (page 25),
which basically means all earlier studies, and defends lack of public knowledge about
what they are asked to value on the basis that, under voting, ``prior experience or
training are irrelevant'' (page 29). Dichotomous choice, he claims, is advantageous
because posted prices are normal in the USA rather than bargaining; it is easier
to answer because you do not need to think of a number, so there is less tendency to
anchor on costs and no strategic bias (page 23). These points are far from lacking in
contention. For example, there is no reason to expect cost to act as an anchor in openended
formats as opposed to dichotomous choice, and this can only arise in the former
if this is explicit in the information pack (so easily avoided). More generally, the extent
to which the institutional context of the USA can be generalised is limited and the
acceptance of markets as well as referenda and their structure is also culturally
specific. For example, Willis (1995) notes that voting on hypothicated taxes is totally
outside UK experience. Even within the USA, Hanemann (1994) notes that ``there are
few cases where local governments actually set environmental quality'' (page 20), which
means his advocacy of the NOAA approach of a referendum on taxes is actually
promotion of an unrealistic and unfamiliar scenario to respondents.
Confusingly, there is also inconsistency in describing what exactly dichotomous
choice does represent. Bateman et al (2002, page 139), which includes Hanemann,
seek to justify the approach as `normal' because it represents the same choice as
making a supermarket purchase.(5) This again seems to be adoption of a specific
(5) Actually, many supermarket purchases are habitual, but this explanation, which is common in
social psychology and marketing, appears hard for many economists to grasp.
Contingent valuation design and data treatment 37
context and institutional framing to produce an overgeneralisation for purposes of
rhetorical persuasion. In many countries, haggling is normal in the market place.
Actually, even where single posted prices exist the model fails to hold because people
can and do shop around. The growth of ebay also shows that people can and do, even
in the USA, haggle and state maximum bids and participate in iterative auctions. Thus,
there seems to be a desire for a model of political economy to justify design practice
but a lack of clarity over the relevant model. Is the CVM meant to be a market choice,
a referendum vote, or some artificial hybrid, and, if it is a market choice then which
model of the market should it represent and why?
One thing arising from this discussion is the potential for alternative institutional
arrangements for articulating values (Vatn, 2005). In effect, economists have stumbled
blind into the area of political science and deliberative democracy. Some have then
pointed out the potential for monetary valuation to occur in various forms from
straight CVM surveys to those involving variants on citizens' juries (see Niemeyer
and Spash, 2001). Moving in this direction raises issues of how to handle individual
versus group deliberation and the whole social and political context of environmental
valuation (Spash, 2007). However, the aim of mainstream economists is not to engage
in a debate over ``deliberative monetary valuation'', but rather to control the process so
that studies limit their design to achieve conformity with expectations from a narrowly
interpreted theory (Spash, 2008).
Design for conservative numbers: WTP versus WTA
The tendency of certain designs to consistently push the numbers in the same direction
leads to the potential for manipulating outcomes through design choices. For example,
Bateman et al (1999) cite the desire for ``conservative numbers'' (page 535) as the
reason a UK government agency rejected their dichotomous choice results in favour
of open-ended ones. A similar issue also arose during the conduct of the UK's largest
ever CVM study (DETR, 1999), and will be discussed further below. The idea of design
for conservative numbers comes from the NOAA panel. Their primary concern in this
regard was the long-recognised potential for WTA formats to far exceed those from
WTP (Knetsch and Sinden, 1984; Thayer, 1981).
Twenty years of empirical studies, including actual goods in experimental situations,
have found people systematically value losses, measured by WTA, two to four
times more than otherwise-commensurate gains, measured by WTP (Knetsch, 2005,
page 96). Attempts at explaining the difference within a neoclassical economic framework
have included income effects, strategic behaviour, and goods having imperfect
substitutes (Hanemann, 1991; Shogren et al, 1994). However, these accounts fail where
the conditions required by standard theory are violated (Knetsch, 2005), and cannot
explain the large and consistent variations in CVM surveys. An alternative explanation
is the psychological impact of `ownership' of a good actually changing the demand
function and indeed causing it to become kinked.(6) This has been termed by economists
an ``endowment effect'' (Thaler, 1980; Knetsch and Sinden, 1984), with the
associated psychological explanation being loss aversion (Kahneman et al, 1991).
Experimental results show large differences between people's valuations of gains and
losses, and are inconsistent with economic-preference-order axioms of completeness,
transitivity, and dominance. That is, economic analyses and predictions of consumer
behaviour are largely based on theories inconsistent with actual choices (Knetsch, 1995).
(6) The WTP ^ WTA divergence has also been given explanations relating to ethical motives. These
include feelings of responsibility for harm, dissociation of entities or changes from the market place
and trade-offs (ie that there is a legitimacy problem), and the activation of social norms [see the
review by Brown and Gregory (1999)].
38 C L Spash
The NOAA Panel recommendation of exclusive WTP usage effectively removed this
troublesome aspect of behavioural findings from further consideration.
Indeed, the argument has been put forward that WTP is a good substitute for WTA
because it is less problematic, is preferred by business, and causes less protesting
(Hanley and Shogren, 2005). In fact, the choice requires an implicit decision over the
ruling property rights. The WTP framing assumes polluters, for example, have the right
to use the environment for their own ends. This denies the fact that property rights
may be contested so that the general public, as in the case of environmental damages,
must assert their position. The determining factor is then the reference state which
people associate with a given environmental change (Knetsch, 1994), and their feeling
of psychological ownership rather than legal entitlements (Knetsch, 2005). However,
under CVM these factors are predetermined by effectively killing all WTA messengers
as a principle of design. For example, Hanley and Shogren (2005, pages 16 ^ 17)
describe in successful terms the replacement of WTA by WTP in the UK aggregates
study (DETR, 1999). The process involved a complaint by the industrial polluters
against the WTA format, government back down to commission a new study, and
the newly commissioned experts adopting WTP to produce `conservative' numbers. A
few points are missing from this account. In the expert discussions of design, Hanemann
argued that people are unfamiliar with the WTA format, which seems strange given that
every payment by a purchasing agent is matched by an acceptance of monetary
compensation on the part of the selling agentöthat is, every market transaction
involves a simultaneous WTA and WTP. The design alone apparently proved inadequate
at producing the stated desire for `conservative numbers', perhaps due to the dichotomous-choice
format. Thus, a 25% discount rate was used in the final analysis to
substantially reduce the damage estimates (DETR, 1999).
The concept of `conservative design' seems to be aimed at justifying some highly
dubious practices and imposing general rules regardless of context. For example, there
is no reason to expect WTP to diverge from WTA under the loss aversion theory if
there is no aversion to lossöfor example, when the person accepting payment is in the
business of selling and that is indeed their purpose. Unfortunately, some then make
the tortuous tautological argument that if only people were educated in the ways of the
market and given the right incentives (carrots and sticks) then they would conform and
produce the required results consistent with market theory (List, 2003). After all, ``evidence
exists suggesting that people can learn to act rationally'' (Hanley and Shogren,
2005, page 29), and the problem is apparently that behaviour goes ``unpunished by
market discipline'' (Shogren and Hayes, 1997, page 243). In contrast, researching compensation
as a serious issue shows how people differentiate between types of exchange
relationships and monetary framing removes the realms of justice and care (Claro,
2007). Monetary framing is then an implicit message to respondents, as to how the issue
is expected to be treated, which will not be universally acceptable.
Informing and forming preferences
In environmental valuation, information has largely been discussed in terms of whether
providing more information on an issue is likely to increase WTP. On the basis of
microeconomic theory, respondents are assumed to have preformed preferences. The
assumption is that objective information relevant to the valuation of an entity can be
provided by the interviewer in different quantities with the aim of merely informing the
respondent.
Consider then the problem of embedding or part ^ whole bias. Kahneman and
Knetsch (1992), amongst others, have raised concerns about the extent to which the
CVM claims to measure a value for a specified environmental change but in fact
Contingent valuation design and data treatment 39
proves insensitive to changes in scope or scale of that change. They attribute this to a
moral motivation diverging from standard economic explanations.(7) A widely cited
example is that of Desvouges et al (1993) where WTP estimates to protect 2000 birds
were not statistically different from those to protect ten times as many birds. A
variation on this type of insensitivity is where the component parts of an individual's
valuation, if assessed separately, are found to exceed the valuation placed upon the
whole (Bateman et al, 1997). Carson et al (2001) believe design and information
provision are to blame:
``Poorly executed survey design and administration procedures appear to be a
primary cause of problems in studies not exhibiting sensitivity to scope. None of
the commonly cited studies with scope insensitivity bears much resemblance to the
current state-of-the-art CV surveys where respondents are presented with a substantial
amount of information about the good they are asked to value in a manner
which facilitates their comprehension of the material'' (page 183).
The suggestion is that all such studies exhibiting embedding problems should be
ignored (shoot the messenger) because information supply was somehow insubstantial
and so failed to facilitate comprehension. As noted earlier, Hanemann (1994) has
argued that under voting ``prior experience or training are irrelevant'' (page 29), which
rather conflicts with the need noted above by Carson et al (2001) for ``a substantial
amount of information'' to be taken onboard by respondents.
The NOAA panel recommended that the CVM aims for a level of information
provision ``at least as high as that which the average voter brings to a real referendum on
the provision of a specific public good'', and use ``follow-up questions''on understanding
(Arrow et al, 1993, page 4607). Their guidelines state that:
``Adequate information must be provided to respondents about the environmental
program that is offered. It must be defined in a way that is relevant to damage
assessment'' (page 4608).
However, the meaning of being `adequately' informed with `relevant' information is
vague and the methods by which individuals assimilate and process information, and
to what ends, is unexplained.
Addressing these issues would require attention to cognitive and social psychology,
while opening up consumer preferences to question on grounds of motivation. That is,
poorly informed bids are unequal to well-informed ones and may be motivated by a
variety of undesirable factors as far as good judgment might be concerned (eg total
disinterest, inattention, lack of engagement, malicious intent). Modern economics has
closed down research on motivation, preferring to take preferences as sacrosanct.
Indeed, economists conforming to orthodox political economy regard motivation as
irrelevant.
``The answer from economic theory is very clear: it is utility whatever its source that
matters for total value. Motives are essentially irrelevant from the perspective of
economic theory (although policy makers may care) and acceptance of consumer
sovereignty is one of the most enshrined principals [sic] of economics'' (Carson
et al, 2001, page 177).
For such economists, the mythically sovereign consumer can make choices on whatever
grounds they like. Information provision then poses rather a major challenge.
(7) The embedding and part ^ whole problems have also been linked to warm-glow giving because
individuals then are regarded as deriving utility from the act of giving itself rather than from the
consequences of that giving. There is also evidence from experimental work which argues that
the embedding problem may be a symptom of a broader problem in economic theory rather than
merely an issue for CVM surveys (see Bateman et al, 1997).
40 C L Spash
Survey design means bounding the information set in terms of the issues to be
presentedötheir format and detail. Recognition that information must be selected and
adapted to the survey design means the information issue is often conceptualised as a
`framing' problem (eg Boyle, 1989). A separation is then attempted between objective
data and the way in which information is delivered. Changes in the former are
expected to impact perceptions and valuations while if the latter do so this is regarded
as a bias. Objective data are meant to describe the `commodity' to be valued while the
framing is merely the method of obtaining a WTP or WTA measure.
This division assumes environmental change can be defined `objectively'. Yet, even
describing changes in common-market commodities can involve disagreements
between different individuals over the condition of commodities, their qualities, and the
definition of their attributes (Spash, 2002). A divergence between actual and perceived
conditions then becomes harder to define and reliance falls upon subjective perception.
The problem is compounded for complex environmental issues. The situation is one
where the CVM practitioner has no practical anchor for accuracy and relies upon
individual perceptions, which means value variations can reflect differences in perceptions
of the hypothesised commodity (Cummings et al, 1986, pages 57 ^ 58), rather than
objectively defined characteristics.
Munro and Hanley (1999) conclude that CVM practitioners face ``major problems''
which they describe in orthodox economic terms as determining ``the optimal extent
of information provision'' and ``what is `true and accurate' information?'' In a more
worried tone they note:
``Putting responsibility onto the researcher for the supply of information creates
ethical dilemmas. To the extent that subjects view interviews as unbiased sources,
the possibility of altering the information set gives the unscrupulous researcher the
opportunity to produce CV results which reflect the experimenter's preferences
rather than those of the respondents'' (1999, page 277).
What these authors fail to notice is the extent to which totally scrupulous economists
may just have a different world view from other people and may hence impose their
personally preferred perspective on their respondents.
That economic analysts may conceive of issues in ways which diverge widely from
those of respondents has been termed an issue of differential encoding and decoding by
Burgess et al (2000). For example, environmental issues concern value conflicts and
one of those relates to disagreements over the role of markets and how environmental
issues should be addressed. Economists take as given that all things are commensurable
and that trade-offs are normal, even if implicit. This contrasts to common
evidence to the contrary (Aldred, 2002; 2006). So, for example, concerns over justice
and rights can then conflict with the very assumptions behind survey design. Fairness
has also been noted to be a common underlying concern of respondents (Jorgensen
et al, 2001).
Response motives based upon such `moral' perspectives have been described as a
failure of the survey to engage with respondents (Ajzen et al, 1996). The argument
goes that the motive to process information is determined by an individual's ability
to understand the issue and their perception of the personal relevance of the issue to
them. If their motivation is high then they enter central processing mode and scrutinise
and evaluate information with regard to the substance of the argument. A low motivation
means that peripheral processing mode is entered and moods and subtle cues
become determining factors in responses so that individuals are susceptible to stimulation
of a moral perspective (Ajzen et al, 1996). Spash (2002) offers a different
explanation and conceptual model where ethical beliefs, such as utilitarianism or
rights, are likely to be key motivations. Along with environmental attitudes and social
Contingent valuation design and data treatment 41
norms, such ethical beliefs can feed into the reasoning over whether to make a
monetary trade-off. Contrary to Ajzen et al (1996), evidence supports a model where
those in central processing mode actually call upon ethical positions to value environmental
changes (Spash, 2002; 2006). Interestingly then, CVM practitioners treat such
noneconomic motivation as bias, and recommend censoring and removing what they
designate as ``ethical protesting'' (Bateman et al, 2002, page 276; Hanley et al, 2001,
page 451).
Extracting the desired message
The treatment of data can be regarded as comprising three stages which take place
after data collection. In the first stage the raw data on bids are assigned to a particular
`classification group'. In the second stage decisions are made as to how to treat
particular classification groupsöfor example, removing or imputing values. In the
third stage data is analysed with the possible testing of bid curves and calculation of
aggregate results. The following sections concentrate on the first two stages and draw
out implications of these for the third stage. Figure 1 provides a guide to the first-stage
categorisation of CVM results to be described next.
Classification groups
The most common classification in treatment of CVM data is between positive and
zero bids. This is a key distinction affecting the mean or median value. At first sight the
distinction seems almost trivial and rather self-evident. However, the literature has
developed from simply accepting all zeros as the same to questioning the validity of
responses and how bids should be classified. This is extremely significant given the
Unit
response
No Yes
WTP/WTA
in principle
Noneconomic
motivation
Unknown
motivation
Item
nonresponse
Don't
know
No
Protest
general
Zero
value
Noneconomic
motivation
Yes WTP/
WTA specific
Item
nonresponse
Don't
know
Zero
value
Valid
zero
Protest
specific
Positive
bid
Invalid
response
Valid
response
Noneconomic
motivation
Outliers Noneconomic
motivation
Figure 1. Possible response categories for contingent valuation method surveys.
42 C L Spash
large number of such bids, so that any outcome will be highly susceptible to classification
and treatment decisions. Furthermore, positive or zero bids may also be reclassified into
one another. As Hanemann (1994) explains, with respect to the use of follow-up questions,
``This information can be exploited in the data analysis. One can monitor for the
misunderstandings, measure statistically how they affected respondents' willingness-to-pay,
and adjust accordingly. For example, if a subject who voted `yes'
appeared to be valuing something different than the survey intended, this case
can be dropped or the `yes' converted to a `no' '' (page 24).
A payment-in-principle question now often precedes the actual payment question
creating a separate class of nonbiddersöthat is, those who have no apparent intention
to pay at all. When the payment question is then asked, those bidding zero are a class
who would pay-in-principle but not in this specific case. A zero bid may be further
classified as a protest zero and may thus be assigned to a separate group. One reason
for the payment in principle question would seem to be the desire to separate out
potential protest bidders concerned about the idea of monetary valuation from those
who find this particular circumstance (eg institutional setting) problematic. That is, two
classes of protest bidder are then identified, the general and specific, as shown in
figure 1.
A neglected distinction is worth making here in order to clarify nonresponse
research. First, a class of refusals to participate in the hypothetical market can be
classified as unit nonresponse. That is where a potential respondent refuses to take
part in a survey. Such refusals give no information on the motives underlying responses
to the hypothetical market. These refusals are assumed to be unrelated to the bid
question because they occur prior to the market scenario or central content of the
survey being introduced (this cannot be inferred for nonresponse postal surveys).
Motives for unit nonresponse are likely to be general, such as lack of interest in the
overall survey topic (Morton-Williams, 1993).
Second is item nonresponse where a respondent participates in the survey but
refuses to answer a question. This is of particular interest when the payment question
is left unanswered while other questions are filled in. There has been a tendency to
refer to nonresponse without making such a difference clear. For example, Loomis
(1987) fails to make any distinction when he is in fact referring specifically to unit
nonresponse. In general, unit nonresponse has received more attention than item nonresponse
with one measure of a survey's success being the proportion of those sampled
who responded. However, unit nonresponse is in fact unrelated to design or content if
the survey has not even been administered to or, for postal surveys, has not been read
by the respondent. In contrast, item nonresponse can provide insight into respondents'
motives and has implications for the estimation of the monetary values which will be
associated with an environmental entity.
In order to investigate the motives behind bid item nonresponse clear categories of
nonresponse are required. Unfortunately, most contingent valuation work tends to
treat as identical refusing to bid (item nonresponse) with stating `don't know' with
giving a zero bid. All are often regarded as zero bids and no separation of categories
may even appear in the survey. The main exception is to classify some respondents as
protestors.
Protest bids are zero bids given for reasons other than a zero value being placed on
the environmental change in question. Yet the classification of protest bidding remains
largely a matter for the analyst. For example, a respondent may refuse any amount of
compensation for loss of an environmental entity which they regard as unique or which
they feel should be protected at all costs. Respondents may refuse to state a WTP
amount because they reject the survey's implicit institutional approach to the problem,
Contingent valuation design and data treatment 43
or because they have an ethical objection to the language of trade-offs used to frame
the issue. They may hold preferences which reject some types of trade-offs such as a
lexicographic preference (see Spash, 1998; 2000b; Spash and Hanley, 1995). Whether an
individual is classified as a protest bidder or as somebody who places no value on the
environmental change in question depends upon the extent to which the analyst can be
bothered to design and include follow-up questions and the way in which they separate
out, using their own judgment, classes of valid from invalid responses.
Positive bids have been far more neglected in terms of subclassification despite the
obvious requirement that these bids be real intentions. That a positive amount is given
has seemed more acceptable and less pressing to explain than zero bids. One reason is
the aim of the survey being to elicit a positive WTP/WTA measure, which means that
a considerable number of zeros are seen as a problem whereas nonzero bids are
desirable.(8) Exxon funded work addressing the meaning of positive bids using verbal
protocol analysis (Schkade and Payne, 1993; 1994) suggested multiple reasons underlying
value formation. Construction of responses during the CVM process was then
found to rely on grounds ``other than trading off a change of wealth against a change in
provision of an environmental good'' (Schkade and Payne, 1994, page 103), although
the sample size (N ˆ 105) for this work was small. Vadjnal and O'Connor (1994) used
follow-up questions to classify bid reasons. They found that 63% of respondents gave
responses which indicated that the meaning of the values given was inconsistent with
economic interpretations (N ˆ 240). Debriefing interviews revealed five categories via
content analysis. These referred to uniqueness, sense of place and personal memories,
community value, aesthetic qualities beyond price, and protecting the environment.
Perhaps the most in-depth analysis of positive bids is that of Burgess, Clark, and
Harrison who conducted discussion groups as follow-up procedures on an independently
conducted CVM study (Burgess et al, 1998; 2000; Clark et al, 2000). They found
respondents had problems in conceptualising the worth of the environmental change
they had supposedly valued, they suffered part ^ whole problems, and expressed incommensurability
with respect to values for `nature'. They take their results to indicate that
the CVM is ineffective at capturing complex cultural values for nature and landscape.
In addition, respondents would have preferred a forum for engaging in collective
deliberation with scientists and others claiming responsibility for management of the
local environment.
In a series of studies, I show how noneconomic motivation plays a key role in
explaining WTP for environmental changes (Spash, 1997; 2000b; 2000c; 2002; 2006;
Spash and Hanley, 1995). This includes analysis of positive bidders via follow-up
questions on ethical beliefs, attitudes, and, more recently (Spash et al, 2008), social
norms and perceived behavioural control. These studies have tended towards reasonably
large samples (N ˆ 700 to 1000) and have been conducted in different countries and
contexts. The basic and consistent finding is that a large proportion of positive bidders
claim motivations which are inconsistent with economic theory. That is, they specifically
identify themselves with rights-based beliefs, explicitly rejecting consequentialist
reasoning, and often express preferences consistent with lexicographic orderings. There
are those respondents who also give reasons consistent with economic expectations but
they seem to be in the order of only half the sample. The divergence in motivations
means that, while some may regard their bids as a value in exchange for an environmental
change, others may regard themselves as giving a charitable contribution,
(8) Early researchers did apparently show some concern about positive bid motivation, largely
because they could not explain the large remainder term in their results, until they conceived of
the category of indirect use values (personal communication with Ralph C d'Arge).
44 C L Spash
making a political gesture, or supporting deeply held nonutilitarian ethical beliefs. The
resulting bids cannot then be simply aggregated as equivalent measures of value.
Yet, despite the evidence for mixed motives underlying positive bids, the majority of
CVM literature ignores any requirement to question the monetary values being produced.
Positive bids are generally accepted at face value as trade prices. The only
concern seems to be that they do not appear `too large' relative to the mean bid, and
that they seem `reasonable' in light of the respondents' income level. If respondents fail
to meet these expectations they may be classified as `outliers'.
The outlier should not be confused with the protest bidder. These are individuals
who bid large amounts and so have a strong influence on the sample mean bid. In
economic terms, this should only be regarded as a problem when the bid is unlikely to
occur because the individual lacks the income to pay or would actually accept a much
lower amount than a stated WTA. In such cases the respondent would be regarded as
acting strategically in order to influence the aggregate bid outcome. Economists tend
to look actively for strategic behaviour and like to explain many varying motives as
merely strategic. In practice, such strategic play in response to a short one-off survey is
uncommon. For example, an extensive review of CVM studies in the health-care
literature found no instances of strategic bias (Klose, 1999). Indeed, CVM practitioners
often make reference to strategic behaviour without citing any empirical case studies
and merely take its existence as given; instead, they talk of the implications and
concentrate upon alternative `incentive compatible' bid designs (eg Carson et al, 2001;
Hanley and Shogren, 2005).
Treatment of classification groups
The recommended solution to unit nonresponse is to impute a value by weighting the
sample. Some assume that unit nonresponse is a design issue but, having failed to
achieve good design, the researcher can infer missing results.
``The best solution is to minimize nonresponse in the first place, by designing a data
collection process which encourages high participation rates for all groups. This
aside the researcher must rely on what she [sic] knows about the population and the
sample, and on what she [sic] knows or can assume about the response process, in
order to infer something about the characteristics of the non-respondents'' (Hanemann
and Kanninen, 1999, page 404).
If the nonrespondents are randomly spread throughout the population the sample
results should be unaffected, but if they are correlated with factors which correlate
with WTP/WTA then results will be affected. If, say, older males refuse to fill in the
survey and these are random refusals within the group of older males, then a weighting
may be employed. However, refusals seem unlikely to be random within either the
general population or classes. Indeed, the key problem with unit nonresponse is a
lack of data on the respondents and their motivations.
Item nonresponse has gained the most attention with respect to WTP questions,
but can affect any other factor and often occurs for income questions. The sensitivity of
questions is also highly culturally specificöfor example, women revealing their age,
and men admitting they have no education or cannot read and write. This means that
condemning a survey as poor can be erroneous, because of, say, a low correlation of
bids with income where income data are highly unreliable, because in some societies
people never tell strangers their actual income. Where there is item nonresponse the
result is self-evident in reduced sample size but perhaps more worrying and difficult to
recognise is the giving of false information due to social normsöfor example, women
stating a lower than actual age, and people reducing their stated over actual income.
Contingent valuation design and data treatment 45
Bid item nonresponse is of particular concern due to the obvious potential impact
on aggregated results (typically median bids are less than mean bids due to left-hand
truncation of the distributionöthat is, a large frequency of zero bids). Systematic
differences between respondents and nonrespondents will bias aggregated WTP/WTA
results. Hanemann and Kanninen (1999) again offer some less than helpful advice
followed by strong warnings concerning their recommendation to impute values for
nonrespondents:
``The only way to prevent this bias is to collect a complete sample at the start. Short
of this, the most common practical remedy for item non-response is imputation ... .
Imputation is a dangerous art ... . As indicated, the imputation procedure chosen
is based upon the instincts of the survey researcher about the behaviour of
non-respondents. The validity of the imputations is not verifiable unless actual
responses are eventually obtained from non-respondents ... . Unfortunately, in the
case of non-response, the researcher is unable to verify assumptions'' (pages 405 ^ 406).
The use of the payment in principle question has been justified as avoiding ``socialdesirability
bias'' and ameliorating the implication that all respondents are expected to
pay for a good thing (Green and Tunstall, 1999, page 244). This is seen by some as
conforming to the `conservative' design principles of the NOAA panel (Bateman et al,
1999, page 517)öthat is, avoiding aspects which might increase the valuation of an
environmental change. However, no work has been done on testing for differences
between refusals in principle from zero bids. Analysts' treatment of these classes varies
depending upon the format of the CVM question. For open-ended questions the two
categories may be lumped together on the assumption that they all represent zero bids,
while conventional analyses of dichotomous choice omit refusal to pay in principle
(Bateman et al, 1999).
The treatment of `protest bids' is clearly problematic. Their omission from the
mean calculation can be undertaken on a large scale. For example, Kenyon and Hanley
(2005) conducted a stratified sample of nine towns' results in 336 responses to their
CVM survey; of these, 29% were classified as protest bids and were removed from the
dataset and any further analysis. Even more dramatically, Meyerhoff (2005) excluded
56% of his CVM sample as protestors on the basis of follow-up questions ``because
they do not represent true economic values'' (page 56). Less draconian reduction of
the sample can still bias results, for example, the 8 ^ 12% used across samples by
Desvouges et al (1993). Jorgensen et al (1999) state that:
``From a psychometric viewpoint, the exclusion of responses necessary to meet such
assumptions creates serious concerns about the lack of theoretical validity of the CV
method. One must feel suspicious of a methodology if, before or after questioning,
rules are made for omitting responses'' (page 137).
They note the problematic practice of removing protest zeros while leaving identically
motivated positive bids. Jorgensen and Syme (2000) point out that where protest
responses are a reaction to the act of paying they are unlikely to be removed by
changing CVM survey design. They go on to show evidence for protests being related
to WTP, via attitudes, and to demographic factors such as income. They note that
standard censoring of protest responses in their study would bias the sample towards
those favouring environmental public goods and with higher incomes.
Treatment of outliers is also of concern due to lack of motivational research and
thus inadequate justification behind extreme bids being regarded as unwarranted
and then censored. However, the lack of evidence and motivational research has not
prevented truncation of bid distributions on the basis of a randomly selected distance
from the mean or some percentage of bid relative to income. For example, Schkade
and Payne (1994, page 96) screened their nonzero WTP responses for outliers and were
46 C L Spash
happy that there were none on the basis that no respondent was prepared to pay more
than 25% of their incomeöthat is, paying up to a quarter of ones income for one
environmental change seemed reasonable to them. Boyle et al (1996) used the same exclusion
threshold plus any amount equal to or greater than US $10 000 plus regression
diagnostics. The choice of these threshold criteria is totally arbitrary.
Informal procedures and inconsistency
Desvouges et al (1987) note that ``using largely informal procedures, analysts have
screened contingent valuation data sets to eliminate protest bids and to identify/delete
influential observations'' (page 252) and later that ``Nearly all analyses of contingent
valuation surveys have used some judgmental procedure to eliminate some bids from
the full sample of responses'' (page 255). Boyle et al (1996) state that ``most contingent
valuation studies employing open-ended questions generally do some censoring of
response data by removing protest zeros, trimming high observations and statistically
searching for outliers'' (page 386). Indeed, leading practitioners are found to recommend
various data-censoring practices but mostly with highly arbitrary justifications.
As Boyle and Bergstrom (1999) note: ``The fundamental concern remains; no established
theoretical or established protocols exist for excluding responses'' (page 198).
Thus, there are a wide range of potential practices from which the analyst can choose.
As part of a `best practice standard', Mitchell and Carson (1989) suggest that in
evaluating a CVM study the reviewer should ask the following questions:
``What procedures were used to identify and handle outliers and protest responses?
Is sufficient information provided about the cases dropped to permit a judgement
about the validity of the procedure?'' (page 302).
Unfortunately, defining the meaning of `sufficient information' has remained problematic.
For example, the suggestion of Hanemann (cited in Schkade and Payne, 1994 is:
``that respondents should be debriefed after they have valued a good to identify
different motives. The researcher would then consider dropping those people who
have the `wrong motives' from the sample'' (page 106).
What exactly constitutes a `wrong motive' is left unspecified.
Detailing the way in which data have been treated and deleted should allow an independent
reviewer to obtain identical results after following the procedure described,
given the same dataset prior to manipulation. However, the provision of such detail is
extremely rare, and several inconsistencies in current practice can be identified which
affect the reliability and validity of results. More specifically, there appear to be three
main inconsistencies underlying current data-manipulation practice: varying the type
and extent of manipulation, using different classification systems for the same bids, and
employing inconsistent motive rationales to determine data treatment.
The first inconsistency arises when one analyst (analyst A) simply engages in
data manipulation to a greater or lesser extent than another (analyst B). Analyst A
might choose to avoid data manipulation completely while analyst B applied both
removal and imputation of responses; or analyst A might choose to impute for item
nonresponse and to remove outliers while analyst B chooses to impute for item nonresponse
but to keep outliers. Such differences in practice are important because, for
example, outliers may be defined as comprising from less than 10% to over 50% of the
bids, depending on the information content of the contingent market and whether
the bid format is WTP or WTA (Edwards and Anderson, 1987). Obviously, the variation
in potential results from the same data will be large.
Second, an individual bid can be classified in more than one way. That is, the same
bid can potentially be placed into one of several classifications. Inconsistency will then
arise due to the differential treatment of bid groups during the analysis of data.
Contingent valuation design and data treatment 47
Outliers, for example, may be defined as being a different standard deviation from the
mean by different analysts. Thus, a bid may be regarded either as an outlier, and
removed, or as a valid bid. Another example is bid classification as a protest zero.
One practice is to treat the protest zero as a final classification group and then to
remove bids in that group from the analysis (Freeman, 1986; Hanley, 1989). However,
Mitchell and Carson (1989, page 268) further classify protest zeros as item nonresponse,
together with `don't knows', `refusals', and responses that fail to meet a standard of
minimal consistency. Some researchers would then impute positive values for item nonresponse
while others may treat them as zero bids. Thus, the same response by an
individual may either be removed from analysis (under the first classification scheme)
or have some value attributed to it by the analyst (under the second approach).
The third inconsistency concerns differences in the rationale underlying the classification
of bids. The problem here relates back to the lack of explicit conceptualisation
as to what makes a valid bid or protest. For example, Hanemann (1994) is happy to use
debriefing questions to
``identify whether subjects were inattentive or unfocussed and offered hasty or
ill-considered responses, and these can be discarded if desired'' (page 28).
However, a few sentences later he criticises others for rejecting CVM survey results
that are failing to reflect economic preferences and asks the key questions:
``But, what are `true economic preference'? If a subject responds thoughtfully to a
question about voting to raise taxes for a public good, by what criterion is that not
a valid preference?'' (page 28).
Similarly, Carson and Groves (2007) talk much of truthful preferences and express
a belief in a ``true WTP'' in order to justify incentive compatible design. This is defined
as follows:
``question formats are incentive compatible, in the sense that a truthful response to
the actual question asked constitutes an optimal strategy for the agent'' (page 184).
So there is then an assumption that respondents go around optimising things and
conform to the economists' model of human psychology, otherwise they will presumably
show the wrong motives and be classified as `untruthful' and `irrational'. Yet the
authors themselves seem to be at least slightly aware of the blinkered vision of humanity
this offers. As Carson and Groves (2007) seem to lament in a later remark,
``there are norms that seem to totally or partially override considerations of rationality/
self-interest, which is why many individuals would return a lost wallet'' (page 205).
Presumably an orthodox economist would not, so beware!
There has been a slowly increasing awareness of the importance of psychological
processes relating to how respondents interpret survey questions and relating to their
decision-making processes leading to a stated bid value (Spash, 2006). This has meant
realising that social psychology regards preferences of the sort being addressed in the CVM
survey as constructed. Orthodox economists conducting CVM studies claim their only
concern is whether preferences are stable regardless of how they are reached. However,
as has been shown, they are still ready to reject respondents' stated preferences on the basis
that they have the `wrong motives' or appear inadequately `thoughtful', and spend much
time concerned with getting `truthful preferences' using incentive-compatible designs.
There is then a contradiction within the mainstream economic position of claiming no
concern for personal motivation in how individuals rank, make orderings, or maximise
welfare (eg Becker, 1993, page 386) and the practice found in the design and conduct
of CVM surveys, where motivation is central to classification and analysis of data.
48 C L Spash
Conclusions
Economists need to start becoming far more realistic as to how decisions are reached
both at the individual and societal levels. In conducting and reflecting upon this review
of economic practice in contingent valuation of the environment, over the last quarter
century, the role of a restricted world view on the part of the expert analysts becomes
self-evident. Orthodox economists have become evangelic in their eagerness to convert
all to their specific market model of political economy. However, markets may be
totally inappropriate methods of social organisation with respect to many environmental
issues and their management. Thus, to find respondents rejecting an approach
which imposes hypothetical market trading is hardly surprising. What may be more
surprising are the attempts by economists to explain away empirical evidence and
to develop technical guidelines to remove fundamental methodological issues. Thus,
for example, rejection of WTA formats is a rather blatant attempt to avoid issues
concerning compensation, property rights, ethics, and loss aversion.
Existing research reveals an interesting array of information about individuals'
attitudes and beliefs, which often challenges the economic analysts' narrow psychological
model. This is particularly problematic for orthodox economists who find
respondents expressing concerns which contradict their theoretical preconceptions.
Thus, attention has been paid to the protest zero bidders because they diminish the
democratic credibility of the process. An explanation that they are acting `strategically'
is theoretically convenient. Less easy to explain and given far less attention are the
contradictory motives of positive bidders. Once having admitted that motives are
important to determining whether a respondent's bid is `valid' and how it should be
`treated', the door is opened to a range of questions from cognitive psychology to
applied philosophy. The classification of large proportions of respondents as acting
outside the model of economically rational behaviour is one result. However, rather
than seeing this as data to be `lost', this `irrational' behaviour is some of the most
interesting in terms of insights into what motivates individuals.
The implications run on two levels. While CVM practitioners seem ready to accept
some data classification and analysis on a motivational basis, there is no consensus as
to the grounds for doing so. There are concerns as to how economic analysts should
treat data which they manipulate and a pretence that all preferences are equal due to
consumer sovereignty. The danger is that wide variations in practice, without clear
explanation and sensitivity analysis, will continue to allow results to be changed by
many orders of magnitude. More fundamentally, even with a standard approach to
data treatment and open explanation, the message from respondents is that current
economic theory is an incorrect interpretation of their behaviour. Thus, expanding
upon the economic interpretation to include a wider range of motives to act is
required.
There is then a fundamental methodological divergence between the approach of
CVM practitioners who try to find evidence in support of their a priori model in the
neoclassical economic tradition, and those who allow the data to inform them as to
the possible variety of human motivation and behaviour. If there is a future for the
CVM, and stated preference methods in general, then this lies in developing empirical
investigation into plural environmental values. In contrast, much existing CVM work
appears to be an exercise in producing the one `true' price given by `true' preferences.
Orthodox economic assumptions of how humans operate as rational fully informed
utility-maximising agents with well-formed preferences confront a reality of fallible
complex social beings with plural values and mixed motives. Rather than pay attention
to these motives as they impact upon respondents' stated WTP and WTA, the contention
of this paper is that many CVM economists have engaged themselves in the design
Contingent valuation design and data treatment 49
of an institutional process to make respondents conform to a narrow behavioural and
market model. This means removing protestors, classifying people as strategic players,
ignoring nonutilitarian ethics, using sticks and carrots under `incentive compatible'
design, and generally manipulating data and approaches to get the desired results.
Those results are increasingly in danger of becoming the production of an acceptable
number which supports predetermined goals for a funding agent or government
agency. Breaking the cycle is necessary if economists are ever going to join those
attempting to offer insights into the relationship between human well-being and the
environment.
References
Ajzen I, Brown T C, Rosenthal L H, 1996, ``Information bias in contingent valuation: effects
of personal relevance, quality of information and motivational orientation'' Journal of
Environmental Economics and Management 30 43 ^ 57
Aldred J, 2002, ``Cost ^ benefit analysis, incommensurability and rough equality'' Environmental
Values 11 27 ^ 47
Aldred J, 2006, ``Incommensurability and monetary valuation'' Land Economics 82 141 ^ 161
Arrow K, Solow R, Portney P R, Leamer E, Radner R, Schuman H, 1993, ``Report of the NOAA
Panel on Contingent Valuation'', Resources for the Future,Washington, DC
Bateman I, Munro A, Rhodes B, Starmer C, Sugden R, 1997, ``Does part ^ whole bias exist?:
an experimental investigation'' Economic Journal 107 322 ^ 332
Bateman I J, Langford I H, Rasbash J, 1999, ``Willingness-to-pay question format effects in
contingent valuation studies'', in Valuing Environmental Preferences: Theory and Practice of
the Contingent Valuation Method in the US, EU and Developing Countries Eds I J Bateman,
K G Willis (Oxford University Press, Oxford) pp 511 ^ 539
Bateman I J, Carson R T, Day B, Hanemann M, Hanley N, Hett T, Jones-Lee M, Loomes G,
Mourato S, Ozdemioglu E, Pearce D W, Sugden R, Swanson J (Eds), 2002 Economic Valuation
With Stated Preference Techniques: A Manual (Edward Elgar, Cheltenham, Glos)
Becker G S, 1993, ``Nobel lecture: the economic way of looking at behaviour'' Journal of Political
Economy 101 385 ^ 409
Boyle K, 1989, ``Commodity specification and the framing of contingent valuation questions''
Land Economics 65 57 ^ 63
Boyle K J, Bergstrom J C, 1999, ``Doubt, doubts, and doubters: the genesis of a new research
agenda?'', inValuing Environmental Preferences:Theory and Practice ofthe Contingent Valuation
Method in the US, EU and Developing Countries Eds I J Bateman, KG Willis (Oxford University
Press, Oxford) pp 183 ^ 206
Boyle K J, Reed Johnson F, McCollum D W, Desvouges W H, Dunford R W, Hudson S P, 1996,
``Valuing public goods: discrete versus continuous contingent valuation responses'' Land
Economics 72 381 ^ 396
Brown T C, Gregory R, 1999, ``Why the WTA ^ WTP disparity matters'' Ecological Economics
28 323 ^ 335
Burgess J, Clark J, Harrison C M, 1998, ``Respondents' evaluations of a CV survey: a case study
based on an economic valuation of the wildlife enhancement scheme, Pevensey levels in
East Sussex''Area 30 19 ^ 27
Burgess J, Clark J, Harrison C M, 2000, ``Culture, communication, and the information problem
in contingent valuation surveys: a case study of a Wildlife Enhancement Scheme'' Environment
and Planning C: Government and Policy 18 505 ^ 524
Carson R T, Groves T, 2007, ``Incentive and informational properties of preference questions''
Environmental and Resource Economics 37 181 ^ 210
Carson R T, Flores N E, Meade N F, 2001, ``Contingent valuation: controversies and evidence''
Environmental and Resource Economics 19 173 ^ 210
Clark J, Burgess J, Harrison C M, 2000, ````I struggled with this money business'': respondents'
perspectives on contingent valuation'' Ecological Economics 33 45 ^ 62
Claro E, 2007, ``Exchange relationships and the environment: the acceptability of compensation
in the siting of waste disposal facilities'' Environmental Values 16 187 ^ 208
Cummings R G, Brookshire D S, Schulze W D (Eds), 1986 Valuing Environmental Goods:
An Assessment of the Contingent Valuation Method (Rowman and Allanheld, Totowa, NJ)
50 C L Spash
DesvougesW H, SmithV K, Fisher A,1987,``Option price estimates for water quality improvements:
a contingent valuation study for the Monongahela river'' Journal of Environmental Economics
and Management 14 248 ^ 267
Desvouges W H, Johnson F R, Dunford R W, Boyle K J, Hudson S P, Wilson K N, 1993,
``Measuring natural resource damages with contingent valuation: tests of validity and
reliability'', in Contingent Valuation: A Critical Assessment Ed. J A Hausman (North-Holland,
Amsterdam) pp 91 ^ 159
DETR,1999 The Environmental Costsand Benefitsofthe Supplyof Aggregates: Phase 2 (Department
of the Environment, Transport and the Regions, London)
Edwards S F, Anderson G D, 1987, ``Overlooked biases in contingent valuation surveys: some
considerations'' Land Economics 63 168 ^ 178
Freeman A M, 1986, ``On assessing the state of the art of the contingent valuation method of
valuing environmental changes'', in Valuing Environmental Goods: An Assessment of the
Contingent Valuation Method Eds R G Cummings, D S Brookshire,W D Schulze (Rowman
and Allanheld, Totowa, NJ) pp 148 ^ 161
Green C,Tunstall S,1999,``A psychological perspective'', inValuing Environmental Preference:Theory
and Practice of the Contingent Valuation Method in the US, EU and Developing Countries
Eds I Bateman, K Willis (Oxford University Press, Oxford) pp 207 ^ 257
Hanemann W M, 1991, ``Willingness to pay and willingness to accept: how much can they differ''
American Economic Review 81 635 ^ 647
Hanemann W M, 1994, ``Valuing the environment through contingent valuation'' Journal of
Economic Perspectives 8 19 ^ 43
Hanemann M, Kanninen B, 1999, ``The statistical analysis of discrete-response CV data'', in
Valuing Environmental Preferences: Theory and Practice of the Contingent Valuation Method
in the US, EU, and Developing Countries Eds I J Bateman. K G Willis (Oxford University
Press, Oxford) pp 302 ^ 441
Hanley N D, 1989, ``Valuing non-market goods using contingent valuation'' Journal of Economic
Surveys 3 235 ^ 252
Hanley N, Shogren J F, 2005, ``Is cost ^ benefit analysis anomaly-proof?'' Environmental and
Resource Economics 32 13 ^ 34
Hanley N, Spash C L,1993 Cost ^ Benefit Analysisand the Environment (Edward Elgar, Cheltenham,
Glos)
Hanley N D, Mourato S, Wright R E, 2001, ``Choice modelling approaches: a superior alternative
for environmental valuation?'' Journal of Economic Surveys 15 435 ^ 462
Jorgensen B S, Syme G J, 2000, ``Protest responses and willingness to pay: attitude toward paying
for stormwater pollution abatement'' Ecological Economics 33 251 ^ 265
Jorgensen B S, Syme G J, Bishop B J, Nancarrow B E, 1999, ``Protest responses in contingent
valuation'' Environmental and Resource Economics 14 131 ^ 150
Jorgensen B S,Wilson M A, Heberlein T A, 2001, ``Fairness in the contingent valuation of
environmental public goods: attitude toward paying for environmental improvements at two
levels of scope'' Ecological Economics 36 133 ^ 148
Kahneman D, Knetsch J L, 1992, ``Valuing public goods: the purchase of moral satisfaction''
Journal of Environmental Economics and Management 22 57 ^ 70
Kahneman D, Knetsch J L, Thaler R H, 1991, ``Anomalies: the endowment effect, loss aversion,
and status-quo bias'' Journal of Economic Perspectives 5(1) 193 ^ 206
Kenyon W, Hanley N D, 2005,``Three approaches to valuing nature: forest floodplain restoration'',
in Alternatives for Environmental Valuation Eds M Getzner, C L Spash, S Stagl (Routledge,
London) pp 209 ^ 224
Klose T, 1999, ``The contingent valuation method in health care'' Health Policy 47 97 ^ 123
Knetsch J L, 1994, ``Environmental valuation: some problems of wrong questions and misleading
answers'' Environmental Values 3 351 ^ 368
Knetsch J L, 1995, ``Asymmetric valuation of gains and losses and preference order assumptions''
Economic Inquiry 33(1) 134 ^ 141
Knetsch J L, 2005, ``Gains, losses, and the US EPA economic analyses guidelines: a hazardous
product?'' Environmental and Resource Economics 32 91 ^ 112
Knetsch J L, Sinden J A, 1984, ``Willingness to pay and compensation demanded: experimental
evidence of an unexpected disparity in measures of value'' Quarterly Journal of Economics
99 507 ^ 521
List J A, 2003, ``Does market experience eliminate market anomalies?'' Quarterly Journal of
Economics 118 41 ^ 71
Contingent valuation design and data treatment 51
Loomis J B, 1987, ``Expanding contingent value sample estimates to aggregate benefit estimates:
current practices and proposed solution'' Land Economics 63 396 ^ 402
Meyerhoff J, 2005, ``Non-use values and attitudes: wetlands threatened by climate change'',
in Alternatives for Environmental Valuation Eds M Getzner, C L Spash, S Stagl (Routledge,
London) pp 51 ^ 68
Mitchell R C, Carson R T, 1989 Using Surveys to Value Public Goods: The Contingent Valuation
Method (Resources for the Future,Washington, DC)
Morton-Williams J, 1993 Interviewer Approaches (Dartmouth, Aldershot, Hants)
Munro A, Hanley N D, 1999, ``Information, uncertainty, and contingent valuation'', in Valuing
Environmental Preferences: Theory and Practice of the Contingent Valuation Method in the US,
EU and Developing Countries Eds I J Bateman, K G Willis (Oxford University Press, Oxford)
pp 258 ^ 279
Niemeyer S, Spash C L, 2001, ``Environmental valuation analysis, public deliberation, and their
pragmatic syntheses: a critical appraisal'' Environment and Planning C: Government and Policy
19 567 ^ 585
Norton B G, Noonan D, 2007, ``Ecology and valuation: big changes needed'' Ecological Economics
63 664 ^ 675
Schkade D A, Payne J W, 1993, ``Where do the numbers come from?: how people respond to
contingentvaluationquestions'',inContingentValuation:ACriticalAssessmentEd.JAHausman
(Elsevier, Amsterdam) pp 271 ^ 293
Schkade D A, Payne J W, 1994, ``How people respond to contingent valuation questions: a
verbal protocol analysis of willingness to pay for an environmental regulation'' Journal of
Environmental Economics and Management 26(1) 88 ^ 109
Shogren J F, Hayes D J, 1997, ``Resolving differences in willingness to pay and willingness to
accept: reply''American Economic Review 87(1) 241 ^ 244
Shogren J F, Shin S Y, Hayes D J, Kliebenstein J B, 1994, ``Resolving differences in willingness
to pay and willingness to accept''American Economic Review 84(1) 255 ^ 270
Spash C L, 1997, ``Ethics and environmental attitudes with implications for economic valuation''
Journal of Environmental Management 50 403 ^ 416
Spash C L, 1998, ``Investigating individual motives for environmental action: lexicographic
preferences, beliefs and attitudes'', in Ecological Sustainability and Integrity: Concepts and
Approaches Eds J Lemons, L Westra, R Goodland (Kluwer, Dordrecht) pp 46 ^ 62
Spash C L, 2000a, ``Multiple value expression in contingent valuation: economics and ethics''
Environmental Science and Technology 34 1433 ^ 1438
Spash C L, 2000b, ``Ecosystems, contingent valuation and ethics: the case of wetlands re-creation''
Ecological Economics 34 195 ^ 215
Spash C L, 2000c,``Ethical motives and charitable contributions in contingent valuation: empirical
evidence from social psychology and economics'' Environmental Values 9 453 ^ 479
Spash C L, 2002, ``Informing and forming preferences in environmental valuation: coral reef
biodiversity'' Journal of Economic Psychology 23 665 ^ 687
Spash C L, 2006, ``Non-economic motivation for contingent values: rights and attitudinal beliefs
in the willingness to pay for environmental improvements'' Land Economics 82 602 ^ 622
Spash C L, 2007, ``Deliberative monetary valuation (DMV) issues in combining economic and
political processes to value environmental change'' Ecological Economics 63 690 ^ 699
Spash C L, 2008, ``Deliberative monetary valuation (DMV) and evidence for a new theory of
value'' Land Economics forthcoming
Spash C L, Hanley N, 1995, ``Preferences, information and biodiversity preservation'' Ecological
Economics 12 191 ^ 208
Spash C L,Vatn A, 2006, ``Transferring environmental value estimates: issues and alternatives''
Ecological Economics 60 379 ^ 388
Spash C L, Urama K, Burton R, KenyonW, Shannon P, Hill G, 2008,``Motives behind willingness to
pay for improving biodiversity in a water ecosystem: economics, ethics and social psychology''
Ecological Economics forthcoming
Thaler R, 1980, ``Toward a positive theory of consumer choice'' Journal of Economic Behavior
and Organization 1 39 ^ 60
Thayer M A, 1981, ``Contingent valuation techniques for assessing environmental impacts: further
evidence'' Journal of Environmental Economics and Management 8 27 ^ 44
Vadnjal D, O'Connor M, 1994, ``What is the value of Rangitoto Island'', Department of Land
Economy, University of Cambridge, and Department of Economics, University of Auckland
Vatn A, 2004, ``Environmental valuation and rationality'' Land Economics 80 1 ^ 18
52 C L Spash
Vatn A, 2005 Institutions and the Environment (Edward Elgar, Cheltenham, Glos)
Willis K, 1995, ``Contingent valuation in a policy context: the National Oceanic and Atmospheric
Administration report and its implications for the use of contingent valuation methods in
policy analysis in Britain'', in Environmental Valuation: New Perspectives Eds K G Willis,
J T Corkindale (CABI,Wallingford, Oxon) pp 118 ^ 143
ß Crown Copyright
Contingent valuation design and data treatment 53
Conditions of use. This article may be downloaded from the E&P website for personal research
by members of subscribing organisations. This PDF may not be placed on any website (or other
online distribution system) without permission of the publisher.