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: ecstjt@bath.ac.uk

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.