valuation of marine ecosystem threshold effects: theory and practice in relation to algal bloom in...
TRANSCRIPT
Valuation of Marine Ecosystem Threshold Effects: Theory and Practice in relation to
Algal Bloom in the North Sea
Alberto Longo1, Tim Taylor
2*, Marta Petrucci
2, Anil Markandya
2,3 Walter Hecq
4 and
Veronique Choquette4
1 Queen’s University, Belfast, UK
2 Department of Economics and International Development, University of Bath, UK
3 FEEM, Italy
4 ULB, Belgium
*corresponding author: [email protected]
Abstract
Threshold effects in marine ecosystems present particular issues in relation to valuation
and inclusion in policy making. This paper presents a theoretical basis for the inclusion of
threshold effects in economic analysis, focusing on the particular case of algal blooms.
Choice experiment techniques are then used to derive estimates of values placed on algal
bloom in the North Sea coast of Belgium.
1. Introduction
The potential for threshold effects in marine ecosystems has been the subject of much
research in recent years. The ongoing EC-funded THRESHOLDS project, of which this
paper is part, attempts to identify and value the impacts caused by these effects.
This paper presents a theoretical basis for the inclusion of threshold effects in economic
analysis of marine policy issues. The case of algal bloom in Belgium is then used to show
the application of this methodology to an applied case. To value the impacts of different
nutrient levels (and associated algal blooms), a primary valuation study is conducted.
Choice experiments are employed to elicit values for the duration and extent of the blooms
and for congestion on the beaches on the North Sea coast of Belgium.
2. Integrating Threshold Effects into Economic Analysis of Marine Systems
From a theoretical perspective, thresholds create impacts that require special treatment
when valuing externalities. If a particular activity creates an environmental pressure such
that it changes the ecological regime, there will be a discontinuity in the valuation function
(Arrow et al, 1995). Figure 2.1 shows how these effects arise when there is a
discontinuity:
i. In the activity-environmental pressure relationship;
ii. In the pressure-ambient state relationship, and;
iii. In the ambient state-valuation relationship1
The figure below shows multiple thresholds, i.e. in the extreme case where there are
thresholds in all the relationships shown above. For a more detailed account of the
construction, see Markandya et al (2005).
In Figure 2.1, the inter-relationship between the threshold effects is central to the
valuation. If, as shown by the bold lines in the top left quadrant, there is a step-type
change in the ambient environment resulting from a change in pressure, and this does not
then deteriorate further with increased pressure, then such a step change dominates the
other changes in terms of the valuation and would be the only threshold-effect of interest
to the policy maker. In reality, multiple step-functions may exist or there may be other
complexities in moving from one state of nature to another (for example, points of no
return or hysteresis). For example, a sigmoid type pressure-state relationship would mean
that other thresholds would also take effect – and this is shown by the curved line in the
top left quadrant.
The framework also allows for the treatment of two key elements that arise when
considering thresholds in marine ecosystems – those of uncertainty and hysteresis.
Figure 2.1: Valuation in the presence of multiple discontinuities
2.1 Treatment of Uncertainty
A number of uncertainties exist in the analysis. For example, we may be uncertain of the
exact level of the threshold of nutrient flow (or type of nutrient) that may lead to the
growth of algal blooms. The marine ecosystem is a complex system, with a number of
exogenous factors - such as climate - that influence the state of nature at any one time.
1 With three possibilities, there are 7 combinations of discontinuities to consider. Discontinuous functions
and valuation functions are shown in bold.
Pressure
Activity
Level X
X0
Ambient State
Total Damages €
Uncertainty can be considered through introducing stochastic elements into the assessment
framework developed in this section. Introducing probabilities would lead to the
consideration of expected values of different activity levels and associated environmental
impacts. Figure 2.2 illustrates this case. If, for example, one is uncertain as to the pressure
that will lead to a change in the ambient state – but one thinks it lies between certain
pressure levels then one may define the pollution loading as a distribution function rather
than a deterministic value. The joint probability distribution of the outcome of interest (for
example, willingness to pay for the abatement of impact for a particular activity) can then
be quantified using Monte Carlo simulation of the distributions involved. Bayesian belief
network models may be used to update such probability distributions as new observations
become available, demonstrating the value of information for reducing uncertainty. Of
course, complexity increases as additional uncertainties are added.
Figure 2.2: Valuation in the presence of uncertainty
2.2 Hysteresis and Discounting
Marine ecosystems are often characterised by a degree of hysteresis – that is, there may be
significant time lags between removing the pressure and a return to the original state of
nature. An example of this in the case of algal bloom may be the case where blooms lead
to effects on the actions of tourists, who may choose to go elsewhere instead of revisiting
the site on their next beach trip, or where algal blooms influences food web dynamics.
The rate of time preference is normally reflected using discounting. Hence, we would take
the net present value of the project inputs and outputs across time. Hysteresis may also
increase the difference between compensating and equivalent variation welfare measures
of nutrient abatement policies on either side of the threshold. For example, a study of
willingness to pay to avoid future nutrient loading should produce lower values than
willingness to pay to reduce existing nutrient loads once the threshold has been crossed if
respondents are aware of hysteresis in the system but still willing to deal with the problem.
Pressure
Activity
Level X
X0
Ambient
State
Total Damage €
P1
P2
P0
The use of discounting in this context also throws up some interesting questions in terms
of policy applications that may mirror the debate in the climate change literature on the
discounting of longer-term impacts. It implies a limited “policy-relevance” of restoring
ecosystems to their previous state in the case of a threshold that exhibits hysteresis. This
may lead some to suggest that such longer term benefits of ecosystem restoration should
be considered differently to normal project impacts – e.g. through use of a different
discount rate or a declining discount rate over time to promote sustainability (see, for
example, HM Treasury Green Book which proposes this for climate change projects).
It should be noted that the framework presented in this paper is essentially static. It is clear
from analysis of the scientific literature that a more dynamic framework may be needed,
and here we deal with dynamic aspects in an ad hoc manner so that they are not ignored.
Further work may include the development of a dynamic model to link emissions to
impacts.
3. Valuing Algal Blooms
The link between algal blooms and nutrient loading is well established in the scientific
literature. Within the context of the current Thresholds project, three major coastal areas
will be subjected to valuation exercises on the costs of algal blooms. We now present the
case of algal blooms, using the framework presented above to highlight the particularities
of assessing the costs of increased nutrient load after a certain level of nutrient loading is
exceeded.
Activity-Pressure relationship
The linkage between economic activity and environmental pressure in this context is a
complex one. The causes of nitrogen and phosphorous entering the watersheds and being
discharged to the coastal area are multiple and diffuse. Table 3.1 summarises activities
that may lead to such nutrient discharge.
Table 3.1: Increased Nutrient Loads and Potential Thresholds
Cause of increased nutrient load Potential threshold
Agricultural use of fertiliser Exceeding maximum take-up of nitrogen by soil
Sewage discharge Exceeding maximum capacity of sewage
treatment plants (STP's) may lead to steeper
increases in nutrient discharge
Surface erosion Extreme weather and surface erosion may cause
pulses of particle bound nutrients
Pressure-Environmental state relationship
The linkage between nutrient loads and the environmental state exhibits a large threshold
effect in that algal blooms arise with increased levels of nutrients, given specific climatic
and other conditions. A more complex and realistic example (from freshwater) is that
while total phosphorous exhibits a linear relationship with total phytoplankton biomass
(wet weight or ChlA), the fraction of cyanobacteria relative to algal biomass exhibits a
threshold. Whether cyanobacteria are toxic or not depend on a number of ecological
factors such as inter-species competition for light and nutrients and is an uncertainty
regarding the health impacts of the threshold.
Environmental state-valuation relationship
The linkage between environmental state and valuation depends critically on the uses of
the coastal zone. Table 3.2 summarises different uses of the coastal zone and highlights
where thresholds may exist for algal blooms in this context.
Table 3.2: Potential thresholds in state-valuation relationship
Use of coastal zone Potential thresholds
Ecosystem service –
Biodiversity
Algal blooms restrict sunlight and impact on viability of existing
ecosystem.
Fisheries Certain types of algal blooms are harmful and may lead to fishing activity
being banned. In addition algal blooms may have an impact on the quantity
and quality of fish stocks, given that fish stocks are related to water
quality.
Swimming Algal blooms may discourage bathing. Toxic algal blooms may also lead
to the banning of swimming in certain areas.
Walking The perception of algal blooms may fall at a certain level of algal bloom
coverage or at a certain distance from the bloom/sea. Odour may also arise
from greater densities of algal bloom.
Amenity Changes in perception of algal bloom may affect house prices.
Note: harmful algal blooms (HAB's) are not necessarily toxic.
The combination of the potential thresholds results in a relationship between the levels of
nutrient that could take the form shown in Figure 3.1. These illustrate the situation where
the threshold levels vary by impact, resulting in a step function for the nutrient loading-
damage relationship. Of course, the actual values may vary and be even more complex
than shown here, but the purpose of this presentation is to show how the relationship
would be estimated and put together.
Figure 3.1: Theoretical Damage Function for Algal Blooms
Integrated Framework
Figure 3.2 brings together the different elements that go into linking emissions of N,P, K
to values of algal blooms. Here, a linear relationship is shown between emissions of NPK
and the level of algal bloom – which is consistent with the literature. One or more of the
nutrients may be a limiting factor, but this is not considered here. Until the level of algae
reaches E2 there is no effect on behaviour – this corresponds to Z1 above. As the level of
activity expands, this has an impact on emissions which leads to an impact on the level of
algae as shown in the top left quadrant. The damages associated with the activity level is
shown in the bottom right quadrant.
Nutrient
Loadings
Damages €
Z1 Z2 Z3 Z4
X1
X1+X2+X3
X1+X2+X3+
X4
X1+X2+X3+
X4+X5
Figure 3.2: Integrated framework for Algal Bloom
Cost-benefit analysis of changes in nutrient flow and valuation of thresholds
A consequence of the above analysis is that cost-benefit analysis of changes in nutrient
flow need to consider such threshold effects. This is very different from the standard
environmental economics approach, which assumes linearity in the impacts and may
assign a unit value to a pollutant.
To focus on the valuation of algal blooms in the context of a multi-attribute environmental
problem, a combination of approaches will probably be required. A recent study for the
EC combined market and non-market approaches to value the impacts of algal blooms,
using CVM to estimate the impacts on tourists and market based approaches to estimate
the impacts on the fisheries industry (Scatasta 2004). Among other things, this study
shows that thresholds also need to be considered in the valuation of ecosystem services.
To a certain extent these have been in terms of changing water quality indices and
recreational uses (see, for example, FWR, 1996; Markandya and Tamborra, forthcoming)
but the linkage to activity and pressure is often poorly defined. Further work as part of the
EC-funded Thresholds project attempts to develop these linkages for marine eco-systems,
including a contingent valuation of algal blooms in three areas: Mallorca, the North Sea
and the Black Sea. The next section of this paper presents a case study of the causes of
eutrophication and a choice experiment for the valuation of algal blooms off the North Sea
coast of Belgium.
Activity
Emissions
Level of
Algae
Damages (€)
E1 E2
4. Valuation of Algal Bloom: Case study of the North Sea Coast of Belgium
The North Sea is an arm of the Atlantic Ocean that washes the shores of North-Western
Europe, from Great Britain eastwards to Norway. It covers an area of nearly 575 000 km2,
and drains several of Europe’s major rivers, including the Thames, the Elbe, the Rhine and
the Scheldt. Its Southern bight is understood to be the area comprised between the English
Channel to the South and, to the North, the area of thermal stratification of the water
column in summertime. While the Northern part of the North Sea has oceanic properties,
the Southern bight is largely influenced by water inflows from the English Channel and
the continent’s major rivers. In that context, the Belgian part of the North Sea – also called
Belgian coastal zone (BCZ) – may be considered a good example of the hydrological and
anthropogenic dynamics that underpin the ecological state of the Southern bight of the
North Sea.
The BCZ is a small area of approximately 3,500 km2, spread along Belgium’s 66 km of
coast. It is characterized by shallow waters (between 20 and 30 m deep) and a tidal motion
that induces strong along-shore currents and ensures a complete mixing of the water
column. Water composition in the BCZ results from two main hydrological dynamics:
currents coming in from the English Channel eastwards, which bring into the BCZ water
inflows from the Atlantic Ocean as well as the Seine and Somme rivers; and freshwater
coming in westwards from the Scheldt, Yser and Rhine estuaries, heavily impacted by
anthropogenic emissions. The extent to which the Atlantic water and river plumes reach
the BCZ varies from year to year, depending on wind conditions, among other factors.
From a political standpoint, the BCZ lies under the jurisdiction of the Belgian federal
government, whose DG Environment is responsible for managing the marine environment.
However, Belgian environmental policies concerning the North Sea fall under a strong
regional cooperative approach, spearheaded by the 1992 OSPAR Convention for the
Protection of the Marine Environment of the North-East Atlantic. This Convention is an
international agreement signed by 15 countries and the European Union, and which
imposes onto its Member States a series of dispositions related to the reporting, prevention
and elimination of marine pollution, as well as the management of human activities that
may impact on the quality of the marine environment.
At the European level, the BCZ watershed is regulated by the 2000 Water Framework
Directive (WFD) of the European Commission, which requires Member States to take
action so that all inland and coastal waters may reach a satisfactory status by the year
2015. The 1991 Nitrate Directive, which aims at restricting water pollution by nitrate
emissions from agriculture, is also an important piece of environmental legislation
affecting the management and conditions of the BCZ watershed. Finally, at the national
level, the BCZ is protected by a series of federal laws and orders, adopted by independent
decisions of the Belgian authorities and/or in application of European directives and
international commitments. Among these laws and orders, important ones are the 1999
Law for the Protection of the Marine Environment and the 2003 Royal Order establishing
the authorization procedure to undertake economic activities in the BCZ. Nevertheless, it
is important to note that, although the BCZ itself is a federal resource, its inland watershed
and the economic activities undertaken therein largely fall under the jurisdiction of (sub-
national) regional authorities. The Belgian coastline, for instance, belongs to the province
of West Flanders, hence is managed by the government of the Flemish region.
From an economic standpoint, the Belgian coast and BCZ is an area of significant activity.
Merchandise traffic to and from the ports of Antwerp, Zeebrugge, Ghent and Oostende,
among others, result in over 400,000 ships crossing the BCZ every year. It is estimated
that 152 million tons of merchandise are processed in the port of Antwerp alone, making it
the fourth largest port in the world. Fishing is also an important activity in the BCZ,
though its dynamism has lately been hampered by the imposition of quotas following deep
concerns over overfishing throughout the Southern bight of the North Sea. Nevertheless, it
is estimated that about 120 fishing boats are still active in the BCZ, with an approximate
yearly catch of 23,800 tons of fish and shrimps. As for aquaculture, it is not yet
extensively developed in the area, but a project led by the Port of Oostende to implement
offshore mussel farms has been approved in late 2005. Besides these, the industrial
activities undertaken within the BCZ are few. There is some ongoing sand extraction, and
an offshore wind farm of about 60 units is expected to be built within the next few years.
On the recreational side, however, the Belgian coastline is home to many activities. There
are a total of 14 seaside towns along Belgium’s 66 km of coast, and they are all readily
accessible by road as well as by public transport. The national railway company (SNCB)
has 5 coastal train stations linking the coast to the country’s main cities, and the Flemish
public transport society (De Lijn) has a tram line linking all 14 seaside towns together. An
extensive network of trails also allows easy access to most of the coastline for bikers and
hikers.
Among the 14 seaside towns of the Belgian coast, the most popular ones are generally
considered to be Knokke-Heist, Blankenberge, Oostende, Nieuwpoort and De Panne. In
order to provide a general idea of the type of recreational activities being developed along
the Belgian coast, as well as the level of difference between the various coastal towns,
Table 4.1 summarizes the main recreation-related characteristics of the coastal towns with
train stations. It shows that most of them offer a variety of water-related activities (such as
sailing, surfing, angling, etc.), as well as non water-related ones (such as horseback riding,
tennis, golf, etc.). That, of course, comes in addition to all the recreational activities that
require no specific infrastructure, such as sea-bathing or walking along the beach. The
Belgian coast being highly urbanized, every coastal town also has a plethora of restaurants
and cafes from which the view of the coast can be enjoyed.
Tourists constitute a significant share of all beach users on the Belgian coast. However,
given local climatic conditions, the flux of tourists at the Belgian coast is very uneven
throughout the year, the busiest season occurring between May and September, with a
peak during the school holidays of July and August. Smaller peaks are also observed
during other holiday periods, such as the Easter holidays in spring.
Table 4.1: Recreational Characteristics of the 5 main towns of the Belgian Coastline
Eutrophication and Belgian coast
The BCZ hydrological signature includes significant water inflows from five of Europe’s
major rivers : the Seine and Somme, whose waters are brought in from the English
Channel through the Straight of Dover; and the Yser, Scheldt and (to a lesser extent)
Rhine, whose plumes are transported westwards by water currents. These rivers are highly
impacted by human activities and bring into the BCZ nutrient-enriched waters with a large
excess of nitrogen over phosphorous and silica. The quantity and spatial distribution of
these excess nutrients throughout the BCZ varies from year to year – but it generally
results in a eutrophication event in the spring of every year. The spring-time
eutrophication event observed in the BCZ is a bloom of Phaeocystis globosa, colony-
forming algae that, in the presence of excess nitrogen and low silica, comes to dominate
other species of phytoplankton. The duration, timing and magnitude of the Phaeocystis
blooms is highly variable from year to year – but we can generally expect them to occur
between March and May, and to affect most or all of the Belgian coast. The main
observable symptom of Phaeocystis blooms is the creation of a slimy matrix which results
in the deposition of thick layers of foam on the beach. This foam, like the Phaeocystis
globosa itself, has not been associated with any form of toxicity to humans, neither
directly, nor through the ingestion of fish of shellfish that has been in contact with the
bloom. However, though they are non-toxic, Phaeocystis blooms are considered to be
harmful algal blooms (HAB), because they do constitute a visual nuisance, release an
unpleasant smell and may stain the clothes.
Drivers of Eutrophication
The bloom-causing nutrient enrichment of the BCZ waters results largely from the
emission of nitrogen and phosphorous compounds by human activities taking place on the
continent, throughout the Scheldt and Yser watersheds (and, to a lesser extent, the Rhine,
Seine and Somme watersheds).
In general, the sources of nutrient emissions into the hydrological network can be divided
into two categories: point sources and diffuse sources. The first refer to the direct emission
into streams and rivers of treated or non-treated wastewater from urban discharges, as well
as water-borne emissions from industrial facilities. As for the latter, they refer to
agricultural activities that use nitrogen and phosphorous compounds as fertilizers or that
release these compounds in the shape of animal manure. In both cases, excess nutrients -
those that are not absorbed by the crops - will be transported by rainfall, going either to
the streams and rivers, or percolating the subsoil to reach the underground aquifers. Either
way, numerous ecological processes intervene between the release of the nitrogen and
phosphorous compounds by the agricultural facilities and their arrival into the
hydrological network, as a result of which there is no simple link between the quantity of
nutrients released by agricultural activities and the quantity that reaches the rivers.
Figure 4.1 presents the estimated level of nitrogen and phosphorous input into the North
Sea in Belgium, for each of the three main sectors involved (agriculture, industry and
households), for the years 1985, 1995 and 2000. It shows that the nitrogen input (in tons)
largely exceeds that of phosphorous, and that the agricultural sector is the main culprit of
nitrogen emissions. We can also see that the input of phosphorous has declined
significantly (by almost 60%) between 1985 and 2000, while that of nitrogen fell
importantly in the industrial sector, but went up in agriculture, resulting in a modest
overall reduction (about 20%).
Figure 4.1: Evolution of nitrogen and phosphorous input into the North Sea in
Belgium, 1985-2000
As part of the OSPAR Convention, Belgium has committed to reduce its input of nutrients
into the North Sea to 50% of their 1985 levels. Thus, Figure 4.1 reveals that this objective
has been reached for phosphorous, but that improvement is still needed for the reduction
of nitrogen emissions.
The main impacts of the Phaeocystis blooms that result from the nutrient enrichment
of the BCZ is the creation of a slimy matrix in BCZ waters and the deposition of an
odorous foam on the beaches of the Belgian coast. There has been no reported
episode of Phaeocystis–related toxicity in the area, and no other reported impact on
human health. Similarly, no episodes of fish mortality have been observed,
hydrological conditions being dynamic enough to prevent oxygen depletion in the
water column. Nevertheless, that does not mean that Phaeocystis blooms are without
impacts.
Table 4.2 presents a series of possible impacts of the bloom together with a subjective
appraisal of the level of importance.
Table 4.2: Possible impacts of Phaeocystis blooms in the BCZ
In this paper we focus mainly on the impact on recreational activities – as the main
expected impact of the Phaeocystis blooms is the impediment or degradation of the quality
of recreational activities due to the presence of foam on the beach. However, previous
studies suggest that the socio-economic value of this impact is likely to be low. For
example, a study by Persoone et al (1994) based on a summertime survey of 1200
respondents in 8 Belgian coastal towns revealed that an algal bloom would lead only 6.5%
of the respondents to cancel their trip to the beach. Another survey of coastal residents and
visitors, conducted by Rousseau et al (2003), also indicates that only 10% of the
respondents consider the accumulation of foam on the beach as a major problem. Such a
low perception of the problem may perhaps arise from the fact that the algal bloom occurs
before the high season begins for recreational activities.
The impact of Phaeocystis blooms on commercial fishing activities has also been studied
by Rousseau et.al (2003), and was also found to be nearly negligible. Indeed, a survey of
21 fishermen working on 16 ships in the BCZ indicates that fishermen are generally
familiar with the algal blooms, and some of them report net clogging and smaller shrimp
catch as nuisances occurring during the blooms. They do not, however, consider these
nuisances as possible causes for income losses.
Finally, the impacts of Phaeocystis blooms on biological resources and ecosystem
structures are still poorly known.
To analyse the impacts of the Phaeocystis blooms on recreational activities in the BCZ we
employ conjoint choice analysis methods. A brief description of the methodology is
presented in the next section, before the questionnaire developed is described and initial
findings from analysis of the surveys are presented.
Conjoint Choice Analysis
A useful tool to assess the monetary valuation of thresholds effects is provided by the
conjoint choice analysis technique.2 Such analyses are also known in the literature as
conjoint choice experiments or choice experiments. In a typical conjoint choice survey,
respondents are shown various alternative representations of a good, which are described
by a set of attributes, and are asked to choose the most preferred (Hanley et al., 2001). The
alternatives differ from one another in the levels taken by two or more of the attributes.
This approach has the advantage of simulating real market situations, where consumers
face two or more goods characterized by similar attributes, but different levels of these
attributes, and are asked to choose whether to buy one of the goods or none of them.3
To motivate the statistical analysis of the responses to conjoint choice experiment
questions, it is assumed that the choice between the alternatives is driven by the
2 Choice experiments and other stated-preference (SP) techniques have recently emerged as a complement to
revealed-preference (RP) techniques. While RP evaluate economic agents’ behaviours in real markets, SP
involve choice responses evoked in hypothetical markets. The interest in hypothetical behaviour in
economics arises from different reasons, such as the necessity to investigate economic agents’ preferences
for new policies that might be implemented, for the development of a new product or good, or for evaluating
goods that are not traded in real economic markets. All these examples make it clear that it is not possible to
estimate agents’ preferences using revealed preferences.
Choice experiments analysis allows a great deal of flexibility because researchers can explore how a change
in the hypothetical scenario influences people’s responses, and compare the current scenario with many
hypothetical alternatives. This is particular helpful for informing policy decisions before the policy itself has
been decided upon.
Usually revealed preference data from regular marketplaces (such as the labour and the housing market)
contain information about actual market equilibria for the behaviour of interest, and can be used to infer
short-term departures from the current equilibria. In contrast, stated-preference data like responses to choice
experiments questions are especially rich in attribute trade-off information. Therefore, stated-preference data
are useful in estimating future changes in agents’ behaviour (Louviere et al., 2000).
In a single choice experiment exercise researchers learn only which alternative is the most preferred, but the
result of the exercise does not tell anything about the preferences for the options that have not been chosen.
A single choice experiment exercise does not offer a complete preference ordering. Therefore, if researchers
want to know a complete ordering of preferences it is necessary either to ask a respondent to do many choice
exercises, or to survey more respondents varying the levels of the attributes. 3A simple example of the application of the conjoint choice methodology can be described by studying the
choice of a ‘car’. When we choose to buy a car we compare the ‘levels’ taken by the ‘attributes’ that
describe them. A car can be described by several attributes: make, number of doors, price, engine, etc. The
levels for the attribute ‘make’ can be: Ford, FIAT, Renault, etc. The levels for the attributes ‘number of
doors’ can be: 3 or 5. And so on for the other attributes.
respondent’s underlying utility. The respondent’s indirect utility is broken down into two
components. The first component is deterministic, and is a function of the attributes of
alternatives, characteristics of the individuals, and a set of unknown parameters, while the
second component is an error term. Formally (see Alberini et al, forthcoming),
(1) ijijij VV ε+= ),( βx
where the subscript i denotes the respondent, the subscript j denotes the alternative, x is
the vector of attributes that vary across alternatives (or across alternatives and
individuals), and ε is an error term that captures individual- and alternative-specific factors that influence utility, but are not observable to the researcher. Equation (1) describes the
random utility model (RUM).
Respondents are shown two or more representations of the good. The alternative they
select is the one that gives them the highest utility. Because the observed outcome of each
choice task is the selection of one out of K alternatives, the appropriate econometric model
is a discrete choice model expressing the probability that alternative k is chosen. Formally,
the probability that respondent i chooses alternative k is given by:
(2) kjVVVVVVV(Vk jkKkkk ≠∀>=>>>= )Pr(),...,,Prchosen) is Pr( 21 ,
If the error terms ε in (1) are independent and identically distributed and follow a standard type I extreme value distribution, it can be shown (Train, 2003) that the probability that
respondent i picks alternative k out of K alternatives is:
(3)
∑=
=K
j
ij
ikk
1
)exp(
)exp()Pr(
βw
βw
Where wij is the vector of all attributes of alternative j. Equation (3) is the contribution to
the likelihood in a conditional logit model. The full log likelihood function of the
conditional logit model is
(4) ∑∑= =
⋅=n
i
K
k
ik kiyL1 1
) choosesPr(loglog ,
where yik is a binary indicator that takes on a value of 1 if the respondent selects
alternative k, and 0 otherwise, and Pr(i chooses k) is equal to Pr(k) in equation (3).
For large samples and assuming that the model is correctly specified, the maximum
likelihood estimates β are normally distributed around the true vector of parameters β,
and the asymptotic variance-covariance matrix, Ω, is the inverse of the Fisher information matrix.
Once model (4) is estimated, the rate of trade-off between any two attributes is the ratio of
their respective β coefficients. The marginal value of attribute l is computed as the negative of the coefficient on that attribute, divided by the coefficient on the price or cost
variable:
(5)
2ˆ
ˆ
β
β llMP −= .
The willingness to pay for a commodity is computed as:
(6) 2ˆ
ˆ
β
βx iiWTP −= ,
where x is the vector of attributes describing the commodity assigned to individual i. It
should be kept in mind that a proper WTP can only be computed if the choice set for at
least some of the choice sets faced by the individuals contains the “status quo” (in which
no commodity is acquired, and the cost is zero).
The conditional logit model described by equations (1)-(6) is easily amended to allow for
heterogeneity among the respondents. Specifically, one can form interaction terms
between individual characteristics, such as age, gender, education, etc., and all or some of
the attributes, and enter these interactions in the indirect utility function.
Whether or not interaction terms are included, implicit in the conditional logit model is the
assumption of Independence of Irrelevant Alternatives (IIA), which states that the ratio of
the odds of choosing any two alternatives depends only on the attributes of the alternatives
being compared, and is not affected by the attributes of other alternatives. Formally,
(7) )exp(
)exp(
)exp()exp(
)exp()exp(
)Pr(
)Pr(
βw
βw
βwβw
βwβw
ih
ik
j ijih
j ijik
h
k==
∑∑
.
An implication of the IIA is that, as shown in equation (7), adding another alternative, or
changing the characteristics of a third alternative, does not affect the relative odds between
alternative k and h. IIA generally imposes restrictive substitution patterns among the
alternatives. A change in the attributes of one alternative, therefore, changes the
probabilities of the other alternatives proportionately to satisfy the conditional logit’s
requirement that the ratio of these probabilities remains the same (Train, 1999). This
implies that the conditional logit is not well suited for alternatives that individuals
perceive as close substitutes of one another.4 Researchers are thus advised to test for
violations of this assumption using the appropriate Hausman test, and to consider models
that relax it, such as the multinomial probit and mixed logit models.
Questionnaire Design
A survey instrument based on the conjoint choice analysis technique was designed for
implementation in three regions of Europe, namely the North Sea coast of Belgium,
Mallorca and the Black Sea coast of Bulgaria. The questionnaires implemented were
4 An example of a situation where the IIA would not be plausible is the blue bus/red bus example due to
McFadden (1974). Consider commuters initially choosing between two modes of transportation, car and red
bus. Suppose that a consumer chooses between the car and bus with equal probability, 0.5, so that the ratio
in equation (16) is one. Now suppose a third mode, blue bus, is added. Assuming bus commuters do not care
about the colour of the bus, consumers will choose between these with equal probability. But then IIA
implies that the probability of each mode is 1/3; therefore, the fraction of commuters taking a car would fall
from 1/2 to 1/3, a result that is not very realistic. While this example is admittedly extreme (in practice, one
would group the blue bus and red bus into the same category), it indicates that the IIA property can impose
unwanted restrictions in the conditional logit model (Wooldridge, 2002). [To grasp how the introduction of a
new alternative might alter the odds of choosing for two existing alternatives, consider the following
example based on the presidential elections. Suppose than an individual prefers candidate A over candidate
B. When asked to indicate his preference, clearly, this individual would vote for candidate A. Suppose now
that a third candidate, C, is added that appears to have a stronger chance to win than A, and is preferred to B,
but less preferred than A. Strategic considerations would lead the individual to pick C, even he is the
second-best candidate, in an effort to avoid the victory of the least preferred candidate, B.]
designed so as to facilitate investigation of the potential for transfer of the results of
surveys on algal bloom valuation. Here we present solely the results of the Belgian survey.
The survey tool was pre-tested using focus groups. The sample population was taken to be
the Belgian users of beach resorts. The questionnaire was structured as follows. First, the
respondents were asked about their use of beaches, including general length of stay at the
resort, what they did at the beach, who traveled with them and the importance they placed
on various attributes in choosing which of the beach resorts to visit (including
environmental attributes). The respondents were then informed about the algal bloom
issue in the Belgian coast area and asked whether they had experienced a bloom episode
and whether the bloom could affect their activities at the beach.
The questionnaire then turned to issues relating to policy options to reduce algal bloom
(the major visible impact of which was foam in the Belgian case). A policy scenario was
presented, with the payment vehicle being an addition to water charges. Respondents were
asked the level of their current water bill.
A series of four choice sets was then presented to each respondent. In total there were 28
choice sets used (in 7 different sets). An example is given Figure 4.2.
Figure 4.2: Choice Set Example
Characteristics Project A Project B No project
Quantity of foam Small quantities of foam on the beach
Foam only on water surface
Big quantities of foam on the beach
Duration of the foam (between April and June)
8 weeks 4 weeks 8 weeks
Congestion
High
The nearest person (or group of people) is less than 3 m away from you
Medium
The nearest person (or group of people) is between 3 and 20 m
away from you
Medium
The nearest person (or group of people) is between 3 and 20 m
away from you
Additional charge for 2007
36€ 108€ 0€
Four attributes were used: the quantity of the algal bloom, the duration of the bloom, the
level of congestion at the site and the level of additional water charge. The quantity of
foam and congestion descriptions was aided by the use of photographs shown in Box 1.
Box 1: Attribute descriptions and photographs
Quantity of Foam
Low quantity
There is foam on the surface of the
water.
Middle quantity
There is some foam deposit on the
beach (small isolated patches)
High quantity
There is a lot of foam deposit on the
beach (large continuous stretches)
Congestion
Low congestion
The nearest person (or group of people) is
more than 20 m away from you
Mid congestion
The nearest person (or group of people) is
between 3 and 20 m away from you
High congestion
The nearest person (or group of people) is
less than 3 m away from you
Follow-up questions followed, to assess the difficulty of the exercise for the respondents
and what attributes the respondents focused on. In the case of the status quo option being
chosen in all cases, a question was posed as to why this was the case.
Finally, socioeconomic characteristics of the respondent were collected.
The next section presents preliminary results of the analysis of the questionnaires.
Results
Descriptive statistics
Our first order of business is to explore the characteristics of the sample of the Belgian
residents interviewed. Table 4.3 reports descriptive statistics of our sample. Our sample is
composed by a slight majority of males, is well educated, and has an average before-tax
income of €30,474. About 62.1% of respondents have a job, almost 19% are unemployed,
11.6% are retired, about 4% are students, less than 3% is homemaker and only 12.1% is
member of an environmental organization. The average number of adults and children
within a household is 1.9 and 1.04 respectively.
Table 4.3: Descriptive statistics
Observations Mean
Standard
Deviation Min. Max.
Age 548 44.05 15.55 18 84
Male (dummy) 550 0.56 0.50 0 1
Gross personal income 448 30474 23011 5000 80000
No education (dummy) 548 0.00 0.00
No elementary education
(dummy) 548 0.01 0.07
Elementary education (dummy) 548 0.11 0.31
Basic education (dummy) 548 0.28 0.45
High school (dummy) 548 0.18 0.39
College (dummy) 548 0.19 0.39
University (dummy) 548 0.19 0.40
Phd (dummy) 547 0.04 0.20
Full time worker (dummy) 549 0.36 0.48
Part time worker (dummy) 549 0.13 0.34
Self employed (dummy) 549 0.12 0.33
Homemaker (dummy) 549 0.03 0.17
Unemployed (dummy) 549 0.19 0.39
Retired (dummy) 549 0.12 0.32
Student (dummy) 549 0.04 0.20
Adults in the household 546 1.91 0.78 0 13
Children in the household 543 1.05 1.10 0 4
Annual water bill 311 337.76 179.69 58 1000
Environmental group member
(dummy) 545 0.12 0.33
Table 4.4 and Table 4.5 summarize the attitudes of our respondents towards recreational
activities at the beach. About 61% of the respondents stay overnight at the beach, and 78%
of them goes often to the beach. This is confirmed by the high number of beach trips taken
by our respondents in a year: on average our respondents go almost for one month to the
beach every year, and only 7% of these trips last less than a day. Respondents travel an
average of 1.5 hours to reach the beach and they usually go with their family (38%),
partner (31%), or friends (24%). Among the factors affecting the recreational experience
at the beach, our respondents deem the cleanliness of the beach as a very important
element, while opinions seem quite spread about congestion at the beach: more than half
of the respondents do not feel that congestion is an important characteristic to consider
when choosing among different beaches. The last two entries of Table 4.4 concern algal
bloom and they show that 63% of our respondents have previously seen algal blooms at
the North Sea and a similar percentage suggest that algal bloom may spoil their activities
at the beach.
Table 4.4: Beach recreation attitudes
Observations Mean Standard
Deviation Min. Max.
Staying overnight (dummy) 551 0.61 0.49
Goes often to the beach (dummy) 550 0.78 0.42
Number of visits 550 28.39 48.76 1 365
Time spent at the beach…
Half day (dummy) 549 0.07 0.25
One day (dummy) 549 0.40 0.49
2-3days (dummy) 549 0.26 0.44
>3 days (dummy) 549 0.27 0.45
Time travel to the beach (min) 548 91.68 52.48 0 240
Who do you go to the beach with…
with family (dummy) 549 0.38 0.49
with friends (dummy) 549 0.24 0.43
with partner (dummy) 549 0.32 0.47
on my own (dummy) 549 0.16 0.36
beach cleanliness is not important
(dummy) 545 0.05 0.22
beach cleanliness is important
(dummy) 545 0.34 0.48
beach cleanliness is very important
(dummy) 545 0.60 0.49
crowd is not important (dummy) 548 0.53 0.50
crowd is important (dummy) 548 0.35 0.48
crowd is very important (dummy) 548 0.11 0.32
Experienced algal bloom (dummy) 554 0.63 0.49
Algal bloom affect activities
(dummy) 554 0.62 0.49
Table 4.5 reports the activities carried out by our respondents at the beach. The results
indicate that most people go walking (90%), and quite surprisingly 38% of them go
swimming. Other popular activities are sunbathing, cycling and other beach sports carried
out by 35%, 32% and 16% of respondents respectively.
B. Conjoint Choice Experiments
Our conjoint choice experiments present respondents with four choice questions, each
described by two hypothetical scenarios plus the status quo. Before reporting the results of
these choice questions, it is necessary to describe the coding of the levels of the attributes
used in the analysis. The attribute for water quality, the quantity of foam caused by algal
blooms present in the water, was described to our respondents as low, medium or high and
accompanied by visual descriptions.
Table 4.5: Activities carried out at the beach
Observations Mean Standard
Deviation
any kind of walking (dummy) 555 0.90 0.30
swimming (dummy) 554 0.38 0.49
sunbathing (dummy) 554 0.35 0.48
cycling (dummy) 552 0.32 0.47
beach sports (dummy) 554 0.16 0.37
running (dummy) 554 0.11 0.32
diving (dummy) 554 0.08 0.27
petangue (dummy) 554 0.08 0.27
skating (dummy) 554 0.06 0.24
wind sports (dummy) 554 0.06 0.24
fishing from land (dummy) 554 0.03 0.16
fishing from boat (dummy) 554 0.01 0.11
horseback riding (dummy) 554 0.01 0.07
other activity (dummy) 554 0.14 0.34
The attribute for the duration of the algal bloom is measured in weeks and takes on values
of 2, 4 and 8 weeks, 8 being the status quo. The attribute for the congestion at the beach,
described as ‘the nearest person is situated’ within 3 metres, between 3 and 20 metres and
beyond 20 metres, is coded as 1.5, 11.5 and 30, meaning that the closest person is located
at 1.5, 11.5 and 30 metres respectively.
When we analyzed the data, we first ran mixed logit models, but found no evidence of
random parameters. We therefore report in Table 4.6 the results from two specifications of
the conditional logit model, the first one presenting results with only the attributes, and the
second one with the attributes and interaction terms to investigate whether attitudinal
differences among respondents and attributes interaction affect willingness to pay.5
The first specification, Model 1, is our basic model that uses only the attributes of the
choice experiments as independent variables. We use effects codes dummies for the
attributes quantity of foam and congestion (Louviere et al, 2000). For the quantity of
foam, we create a dummy low foam which is equal to 1 when the attribute takes the low
quantity level, 0 when it takes the middle quantity level, and -1 when it takes the high
quantity level of foam. A second dummy, middle foam is equal to 1 when the attribute
takes the middle quantity level, 0 when it takes the low quantity level and -1 when it takes
the high quantity level. The dummy omitted is the one for the high quantity level of foam.
We used dummy variables for the quantity of foam because it would have been difficult to
translate into quantitative variables the qualitative information on the quantity of foam.6
The use of dummy variables would also help us identifying thresholds effects, in case the
coefficients on the dummies present coefficients significantly different from one another.
5 The two specifications of the conditional logit model do not use Alternative Specific Constants for the
hypothetical programs, given that we could describe the status quo option with the levels of the attributes
(see Holmes et al, 2003).
6 In future studies, supported by the scientific input that will be available in the THRESHOLDS IP towards
the end of the project, we aim to find a more precise relationship between human activities and the quantity
of algal bloom. At this moment, we feel that the assumptions we should make to use a continuous variable
for the quantity of foam would be too strong.
For similar reasons and in a similar way we created effect codes dummies for the attribute
of congestion. Low congestion takes a value of 1 if the nearest person is more than 20 m
away, 0 if the nearest person is between 3 and 20 meters away, and -1 if the nearest person
is less than 3 meters away. Middle congestion takes a value of 1 if the nearest person is
between 3 and 20 meters away, 0 if the nearest person is more than 20 m away, and -1 if
the nearest person is less than 3 meters away. The reference dummy is the one for the high
level of congestion.
Table 6.6: Conditional Logit Model results
Model 1 Model 2
Coeff. t-ratio Coeff. t-ratio
cost -0.0055 -6.12 -0.0181 -9.00
duration -0.0608 -5.09 -0.0230 -1.75
low congestion 0.0374 0.80 -0.0298 -0.60
medium congestion -0.1401 -3.67 -0.0418 -1.02
low foam 0.0908 2.68 -0.3320 -4.13
medium foam 0.0465 1.39 0.3974 5.03
(low foam)*(duration) 0.0783 5.61
(medium foam)*(duration) -0.0714 -4.91
(number of children)*(cost) 0.0014 2.10
(full time worker)*(cost) 0.0056 3.79
(goes often to the beach)*(cost) 0.0084 4.63
observations 2115 2075
Log likelihood function -2281.64 -2281.64
The variables duration and cost are continuous variables that measure the effects of the
number of weeks of foam presence and the cost of the policy to the respondent
respectively.
Model 1 shows that, ceteris paribus, our respondents shy away from more expensive
projects, and dislike a long presence of foam along the coast. Both coefficients of cost and
duration are, in fact, negative and significant. The marginal price of one week of presence
of algal blooms is equal to €10.98; that is respondents are on average willing to pay almost
€11 to reduce by one week the duration of the algal bloom. When we look at the effect of
the quantity of algal bloom, respondents do prefer projects that offer a low quantity of
foam, being the coefficient of the dummy low foam positive and significant. Also the
coefficient of the dummy middle foam is positive and significant, but its effect is about
half the effect of the dummy low foam: our respondents are willing to pay €16.39 to have
a program that guarantees a low level of foam, and are willing to pay only €8.40 for a
program that entails a middle level of foam. As we are using effects codes dummies, to
calculate the effect of the high foam level we sum the coefficients of low foam and middle
foam and change the sign: our respondents are therefore willing to accept a compensation
of €24.79 for accepting the presence of a high quantity of foam.
When we look at the effects of congestion, our results are less straightforward. Our
respondents are not particularly bothered or appealed by a low presence of people, being
the coefficient of low congestion not significant, seem to be concerned by a middle level
presence of people at the beach, being the coefficient of the middle congestion dummy
negative and significant, and seem to like when a high level of congestion is present.7
These latter results seem counterintuitive, but if read together with the information on
table 6.4, where about half of the sample says that the crowd at the beach is not important,
lead us to conclude that probably this attribute was not considered as important as the
other attributes in the choice sets.
To further explore this result and to investigate heterogeneity among respondents we run
Model 2 with interaction terms. This model aims to investigate how socio-economic
characteristics of the respondents affect WTP and to see whether people considered any
relationships between the two attributes for algal bloom. A LR test shows that Model 2
outperforms Model 1 (LR test=83.25). The results still show that respondents shy away
from project that, ceteris paribus, are more expensive and entail longer periods of algal
blooms. The interaction terms with the socio-economic characteristics of the respondents
show that our sample is internally valid. The interaction term between the dummy variable
for beach users and the cost of the policy is positive and significant, suggesting that beach
users have a higher WTP than non-users. To further explore whether people with higher
income have higher WTP, we add the interaction term between a dummy for full time
workers and the cost of the program. The sign of the coefficient of this variable is positive
and significant, indicating that full time workers are WTP €0.31 more than those that are
not working full time. Our results also show that households with children are on average
willing to pay more than those with no children, suggesting that our respondents do care
for future generations.
The results for the quantity of foam should now be interpreted by looking at the
coefficients of the dummy variables low foam, middle foam and at the interaction terms
between these two dummies and the continuous variable of the duration of the foam. The
coefficients show how for short periods of algal blooms, respondents are willing to accept
a medium to high quantity of foam, while when the duration of the bloom increases,
people prefer low quantity of foam. To explain, only when the algal bloom lasts for 6
weeks of more, our respondents are willing to pay €1.72 and €5.88 to avoid a middle and
a high quantity of foam level respectively and are willing to pay €7.61 to guarantee a low
level of foam. The results of the simulation of the WTP for the three levels of foam
quantity for blooms lasting up to 16 weeks calculated from Model 2 are reported in Figure
6.3. An interesting feature of the figure is that our respondents always bear a negative
WTP for the worst case scenario of foam quantity, and their WTP is quite inelastic for
high foam quantities with respect to the duration of the foam.
The dynamics of the results suggest a complex interaction between duration, level of
bloom and the willingness to pay. Users appear to be willing to pay to ensure a medium
level of foam for a duration of up to 6 weeks. This may seem counter intuitive, but there
may be positive externalities associated with the foam. For instance, some beach users
have been seen to play with the foam deposits on the beach, which may be particularly the
case in early April and May when the temperatures are not so high as to generate the large
7 The effect of a high level of congestion is calculated by summing the coefficients of the two dummies for
low and medium congestion and changing the sign. Hence the effect of high congestion is positive.
odour impacts that may be expected from the bloom deposits on the beach. This
interaction requires further investigation – and will be explored in both the Bulgarian and
Mallorcan studies under the EC Thresholds project. However, after 6 weeks it can be seen
that there is significant willingness to pay to avoid foam, or to guarantee low foam in the
case in the Belgian coastal zone.
The results of Model 2 also illustrate now that the two dummies for the level of congestion
are now both not significant, suggesting that people seem not to be affected by the
presence of people at the beach in Belgium, and are therefore not willing to pay anything
for a policy that would lead to any change in the number of people visiting the Belgian
coast. These results are also consistent with debriefing questions at the end of the choice
questions when we asked respondents to rank the importance of each attribute. Table 6.7
shows that congestion ranks indeed as the least important attribute.
Table 6.7 Ranking of the attributes (1=most important; 4=least important) Observations mean st. dev.
Quantity of foam 504 1.81 0.87
Cost of the policy 504 2.41 1.25
Duration of algal bloom 504 2.44 0.86
Crowding at the beach 504 3.33 0.88
Comparison to previous studies
Previous studies on the valuation of algal blooms includes Stoltel et al (2003), who
examined the valuation of “harmful algal blooms” in European marine waters. This study
Figure 6.3 WTP for different foam levels
-60
-40
-20
0
20
40
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
duration of foam in weeks
Euro
WTP low foam
WTP middle foam
WTP high foam
combined market approaches with contingent valuation and compared the effects of HABs
across Europe, focusing on Spain, Italy, Ireland, The Netherlands, France and Finland.
Here we focus on the valuation using CVM, which included the impacts of shellfish
toxicity. A pooled analysis showed willingness to pay varied between 23 and 41 euros per
person per year – which is higher than the values reported above. This is to be expected as
the timing of the algal bloom event is different in Belgium, and the impacts of the algal
bloom in the Belgian case are different – Stoltel et al (2003) considered impacts on health
as well as recreation. Hence, our findings are broadly in line with the previous valuation
literature on algal blooms.
5. Conclusions
The consideration of threshold effects in policy analysis of marine systems requires the
use of specific economic tools. This paper has first presented a framework for the use of
such tools in the marine policy case. It takes the example of algal bloom, which is an
important issue in coastal ecosystems. Bloom events may have impacts on a range of
activities, including industry and recreation. Drivers include agriculture, industry and
waste water treatment – all of which lead to emissions of nutrients to the coastal system.
Of particular interest in this paper is the impact of algal bloom on recreational users. This
paper shows the use of choice experiments to identify the impacts on willingness to pay to
avoid the bloom for both quantity of the bloom and the duration of the bloom event.
Congestion at the tourist site is also considered. In terms of the extent of bloom, Belgian
beach users are willing to pay €16.39 to have a program that guarantees a low level of
foam, and are only €8.40 for a program that entails a middle level of foam. In terms of a
high quantity of foam, beach users are willing to accept a compensation of €24.79.
Duration and willingness to pay has a complex dynamic relationship. However, when the
algal bloom lasts for 6 weeks of more, beach users are willing to pay €1.72 and €5.88 to
avoid a middle and a high quantity of foam level respectively and are willing to pay €7.61
to guarantee a low level of foam
The cost of actions to mitigate N,P, K emissions should be compared with the benefits of
reductions in algal bloom levels. The mitigation costs are subject to ongoing research in
the current Thresholds project, with agriculture and waste water treatment being
considered.
Further work is planned on testing the possibility for benefit transfer between different
sites and types of bloom - between case study sites in Belgium, Bulgaria and Mallorca.
This will yield insights into how people perceive algal blooms in different environments.
The case of Phaeocystis blooms investigated in this paper differs quite radically in terms
of timing from that in the Mediterranean and Black Sea cases, being in the Spring rather
than Summer may explain the relatively low levels of willingness to pay reported.
6. References
Alberini, A., A. Longo and M. Veronesi (forthcoming), “Basic Statistical Models for
Conjoint Choice Experiments,” in Valuing Environmental Amenities using Choice
Experiments: A Common Sense Guide to Theory and Practice, B. Kanninen (ed.),
Springer.
Arrow, K. et a, 1995. Economic growth, carrying capacity and the environment, Science,
268, 520-521.
ECOHARM, 2005. The Socioeconomic Impact of Harmful Algal Blooms on Marine
Coastal Waters. Report prepared for European Commission under contract EVK3-CT-
2001-80003,.
Goett, A., Hudson, K., Train, K., 2000.Customers Choice Among Energy Suppliers: The
Willingess-to-Pay for Service Attributes, Energy Journal.
Hanley, N., Mourato, S., Wright, R.E., 2001. Choice Modelling Approaches: A Superior
Alternative for Environmental Valuation? Journal of Economic Surveys, 15, 435-462.
HM Treasury, 2006. Green Book. Available at: http://greenbook.treasury.gov.uk/.
Holmes, T.P. and W.L. Adamowicz (2003) Attribute-Based Methods, in Champ, Patricia
A.; Boyle, K.J.; Brown (eds.) A Primer on Nonmarket Valuation, Kluwer Academic
Publishers, Dordrecht.
Louviere, J.J., Hensher, D.A., Swait, J.D., 2000. Stated Choice Methods: Analysis and
Applications. Cambridge University Press: Cambridge.
Markandya, A., Tamborra, M., 2006. Green Accounting in Europe: A Comparative Study.
Edward Elgar: Cheltenham.
Markandya, A., Taylor, T., Longo, A., Barton D., 2005. Marine Ecosystem Thresholds –
Towards an integrated theoretical framework for valuation. In E. Ozhan (ed) Proceedings
of the Seventh International Conference on the Mediterranean Coastal Environment,
MEDCOAST 05, 25-29 October 2005, Kusadasi, Turkey, MEDCOAST, Middle East
Technical University, Ankara, Turkey, 2005; Volume 1 pp127-138.
McFadden, 1974. The Measurement of Urban Travel Demand, Journal of Public
Economics, 3, 303-328.
Persoone et al (1994) Definiering en toepassing van ecologische criteria en economische
indicatoren voor de efectsstudie en kostenbepaling van diverse types van verontreiniging
in the Noordzee, Jaarverslag 2 Impulsprogramma Zeewetenschappen 1992-1997.
Rousseau, V. et al (2003) Identification of Belgian Maritime Zones Affected by
Eutrophication (IZEUT), Scientific Support Plan for a Sustainable Development Policy
(SPSDI), Sustainable Management of the North Sea, Brussels, 65 p.
Scatasta, S., 2004. The Role of Respondents’ Experience in Contingent Valuation
Analysis: The Case of Harmful Algal Blooms and European Union Coastal Tourism. 13th
Annual EAERE Conference, 25-28 June 2004 Budapest, Hungary.
Smith, M.D., Crowder, L. B., 2005. Valuing Ecosystem Services with Fishery Rents: A
Lumped-Parameter Approach to Hypoxia in the Neuse River Estuary. FEEM Nota di
Lavoro 115. Available from www.feem.it
Stoltel, W., Scatasta, S., Granéli, E.1, Weikard, H.P., and van Ierland, E. (2003)
“ECOHARM: The Socio-economic Impact of Harmful Algal Blooms in European Marine
Waters”, European Commission: Brussels (EVK3-CT-2001-80003)
Taylor, T., Markandya, A, Walton H., 2005. Review of Literature on Valuation of Coastal
Ecosystem Services. Report prepared for the European Commission – Deliverable D.1.1.3
of the THRESHOLDS project.
Train, K.E., 2003. Discrete Choice Methods with Simulation. Cambridge University Press:
Cambridge.
Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. John
Wiley and Sons: Chichester.
Acknowledgements
The authors would like to acknowledge the contributions of David Barton (NIVA), Antoni
Riera, Dolores Garcia and Cati Torres (all UIB) in the formulation of the questionnaire.
The assistance of Mateo Cordier is gratefully acknowledged in the compilation of the
dataset from the survey. All errors are, of course, our own.
The European Union (Contract No. 003933) is gratefully acknowledged financial support
in this study, which forms part of the Thresholds Integrated Project.