Economic and social demands for coastal protection
Coastal Engineering 52 (2005) 819 – 840
www.elsevier.com/locate/coastaleng
Economic and social demands for coastal protection
P. Polome a,*, S. Marzetti b, A. van der Veen a
´
a
CTW–WEM, University of Twente, P.O. Box 217, 7500AE, The Netherlands
b
Department of Economics, University of Bologna, Italy
Available online 25 October 2005
Abstract
The purpose of this paper is to present methods and examples of economic valuation in the framework of cost–benefit
analysis of coastal defense schemes. We summarize the concepts of value in economics and their application to coastal erosion
defense. We describe the results of an original benefit transfer exercise on beach recreation, that is, whether and how values
known for some sites can be used to assess the value of some other sites. We present six original case studies on the valuation of
the benefits of coastal erosion defense; four of them focus on beach recreation in Italy, one focuses on the conservation of the
Venice heritage, and one on biodiversity in The Netherlands. The results of the case studies are illustrative of the diversity of
values for the many types of non-marketed assets that may be protected from sea erosion.
D 2005 Elsevier B.V. All rights reserved.
Keywords: Economic valuation; Coastal erosion; Non-market benefits; Benefit transfer
intended as illustrations of the variety of coastal
1. Introduction
defense benefits and their valuation. Section 2 pre-
The purpose of this paper is to present methods and sents the results of original studies on the valuation of
examples of economic valuation in the framework of recreational benefits of coastal defense for four Italian
cost–benefit analysis (CBA) of coastal defense beaches. These case studies should be fairly represen-
schemes. The paper is intended for a broad scientific tative of coastal defense schemes for Northern Med-
audience without prior knowledge of economics. The iterranean beaches. Section 3 presents the very special
introduction of the paper presents the principles of case of the defense of the Venice lagoon. Section 4 is
CBA, summarizes the main notions of economic radically different since it is about a small unused
value, the most well-known valuation methods and natural area in the Northern Sea. Section 5 introduces
the main potential costs and benefits of coastal the technique of benefit transfer that is whether and
defense schemes. The following three sections are how economic values known at some sites can be
used to infer in some way the value of an original
site. This technique, when it can be applied, is very
economical because an economic valuation study can
* Corresponding author.
be quite expensive. The results of an original benefit
E-mail address: polome@ecru.ucl.ac.be (P. Polome).
´
0378-3839/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.coastaleng.2005.09.009
820 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
transfer exercise are presented. We do not claim that ferent valuation methods. Classical typologies of
the valuation studies presented in the present paper are values adapted from Turner et al. (1992), and Bower
representative of all the possible valuation cases of and Turner (1998) are presented in Table 1.1. The
coastal defense; yet we trust that, as a result of this third column indicates the valuation methods that
paper, the reader will have a general idea of what can would be most suitable for estimating each value.
be done regarding cost–benefit analysis of defense This is not an indication that it has been estimated.
schemes and will have enough examples to draw An overview of the valuation methods is given in the
upon to build his own valuation exercise, be it a sequel.
transfer exercise or an original study. That is the We now turn to a brief introduction of the eco-
main purpose of our paper. nomic valuation methods. The necessary data are
CBA is a process intended to measure whether the generally too specific to exist in any publicly available
sum of all the positive impacts of a project outweighs database and it is often necessary to use surveys to
the sum of its negative impacts once they are converted collect the data or to resort to benefit transfer (Section
in a single unit, often money; for a thorough review of 5). The valuation methods are divided into bstated
CBA in the case of environmental changes, see Hanley preferencesQ and brevealed preferencesQ; a detailed
and Spash (1993). In this introduction, we will review description can be found in Haab and McConnell
briefly the economic notion of value, the valuation (2002). Revealed preferences methods rely on actions
methods, the types of value, and the types of asset that individuals have taken in the past; one can dis-
that can be found at the coastline. The economic con- tinguish between bdirectQ and bindirectQ revealed pre-
cept of value that is most often used in a CBA is the ferences methods. Direct methods refer to changes
Willingness to Pay (WTP) defined as the maximum that directly affect marketed goods. A typical example
amount of money a person is willing to exchange to in the case of coastal defense is the demand for hotel
acquire a (public or market) good or service. The nights at a specific coastal resort. Indirect methods
economic value does not refer to an exchange of refer to changes in the provision of a non-marketed
money or to a price; the goal is to convert bindividual good that can be valued indirectly through estimation
utilityQ into money to match it against monetary costs of the changes in the demand of an associated mar-
such as those of building a coastal defense scheme. The keted good. A good example in the context of coastal
WTP is used, and not market prices, because the defense is the recreation quality of a beach. Recreation
coastal defense scheme changes the supply of non- is not in itself sold in a market; however, to enjoy
marketed goods: a government provides the defense recreation at the beach, visitors have to travel there.
scheme, but cannot charge the consumers for it; CBA One can then estimate the demand for travel to the
addresses this issue by converting the change of well- beach and proceed as in a case of direct methods.
being into money, and compares it to the actual money Direct methods, or bmarket pricingQ as indicated in
that has been spent on providing the good. The con- Table 1.1 can be briefly described as follows (see Fig.
version should be based on individual preferences; that 1.1; see also Lipton et al., 1995). First, the demand
is the case in the present paper. That definition of schedule of the market good is estimated. The sche-
economic value makes clear that a broad class of dule can be estimated at individual level (the price is
benefits should be considered in CBA. Yet, economic the observed individual price) or at the market aggre-
value is not the only criterion for deciding on public gated level. The area defined by the horizontal price
projects; equity considerations, precautionary environ- line, the demand curve and the vertical axis is defined
mental standards, and regional economic constraints as the consumer surplus. The producer surplus is
can be seen as complements to CBA. defined similarly, but is often not estimated in practice
One purpose of this introduction is to make clear because supply is assumed completely inelastic (ver-
the diversity of value categories and assets at the tical schedule). Second, using the estimated demand
coast. The value of a coastal defense scheme is com- schedule, we forecast the change in demand caused by
posed of the sum of the values of the consequences of the change that we want to value (e.g. an eroded beach
that scheme on the seafront, avoiding double-count- versus a nourished beach); in Fig. 1.1, the demand
ing. Often different types of values will require dif- schedule shifts up. The change in value is the change
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 821
Table 1.1
Coastal defense values
Use generated values
Direct use values Consumptive: fishing; agriculture; transport; construction Market pricing (possibly adjusted)
and maintenance costs
Non-consumptive: recreation Travel cost; stated preferences
Indirect use values Flood control; storm protection; sedimentation; Market pricing; hedonic pricing;
habitat loss reduction; landscape; human health stated preferences
Non-use and option generated values
Option values Insurance value of preserving options for use Stated preferences
Quasi-option values Value of increased information in the future (biodiversity) Stated preferences
Existence and bequest values Knowing that a species or system is conserved; passing on Stated preferences
natural/heritage assets intact to future generations;
moral resource/non-human rights
in consumer surplus, in most cases a good approxima- is marginal, the supply of the additional nights has a
tion to the WTP for the change. zero (or very low) marginal cost. If the change is not
The complete procedure of estimation of the sup- marginal, for example if hotels have to be built to
ply and demand schedules, and forecasting their accommodate the additional nights, then costs have to
change, is often a complex task, especially if there be taken into account and the demand and supply
exist goods that are substitute or complement to the schedules should be estimated.
market good of interest. Things may be simpler if Indirect revealed preferences methods are used for
the change can be said to be marginal. In that case, goods for which there is normally no observable
the price of a market good is sometimes equivalent to demand but there is a complementary or substitute
the marginal social benefit of a unit of that good; as an market good. The travel cost method is concerned
approximation, and if the market can be said to be with changes in the quality of a recreational site.
competitive, the social benefit of a project that The value of that site is estimated on the basis of
increases marginally the output of such a good can the demand for travel to that site, travel being the
be taken as the product of price and quantity. For market good complement to the recreational site.
example, regarding the increase in the number of Hedonic pricing captures the WTP associated with
hotel nights caused by a (small) beach protection variations in property values that result from the pre-
scheme, the marginal social benefit can be said to sence or absence of specific environmental attributes.
be equal to the number of additional nights times Stated preferences methods are used for changes in
the price of the rooms on the ground that if the change non marketed goods such as landscape, natural or
Fig. 1.1. Consumer and producer surplus.
822 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
cultural heritage that have no complementary or sub- Enhancement effects include: increased output of
stitute market good. In that case, one can only resort the seafront (e.g. creation of recreational fishing
to directly asking individuals (in a survey) how much opportunities); water quality changes (eutrophication,
they are willing to pay to obtain that change (or to red tides); conflicts among different types of recrea-
avoid it). The precise way to ask that question is the tion users of beach areas.
subject of much debate and has given rise in practice Preservation effects refer to natural areas. The
to several methods. The contingent valuation (CV) is benefits stemming from the preservation of a natural
the most developed stated preferences method and is ecosystem are generally recreational use and non-use.
very well documented, see e.g. Bateman and Willis An in-depth case is described in Goodman et al.
(1999). Several examples are presented in details in (1996). Offshore sand and gravel mining (e.g. to
Sections 2, 3 and 4 of this paper. find the sand for beach nourishment) may affect fish-
We now turn to the question of what types of assets eries and habitats.
might be affected by a coastal defense scheme. We Indirect economic effects are bsecond roundQ
present here a summarized list; for a more detailed effects, e.g. constructions in hazardous areas in rela-
list, see Bower and Turner (1998), the bYellow Man- tion to coastal storms that are built because of the
ualQ of Penning-Rowsell et al. (1992) and Polome ´ protection granted by the defense scheme (resulting
(2002). possibly in a stronger scheme being necessary in the
Mitigation effects of coastal defense include the future; see Cordes and Yezer, 1998).
following categories: reduce damage to or prevent
destruction of coastal properties and cultural and heri-
tage assets from coastal storms and eroding shorelines; 2. Case studies on the use value of Italian beaches
reduce salinity intrusion; reduce sedimentation; restore
or preserve habitats or recreational opportunities (e.g. In this section, we present the most significant
sand beach). results of four case-studies at Italian beaches. For
Buildings damage can be valued in two ways. the complete results, see Marzetti (2003, D28/A).
Erosion can cause complete loss of the building Two small surveys were administered at the beach
(sinking); the literature (Mendelsohn and Neumann, of Ostia near Rome (100 interviews on the beach,
1999) suggests estimating the discounted value of summer 2002) and on Pellestrina Island in the Lagoon
the building from the current time until the expected of Venice (80 residents and 75 beach visitors, July
sinking time, allowing for market adjustment of the 2002), respectively. Two larger surveys were adminis-
building price (zero at the time of sinking). That tered at Lido di Dante near the town of Ravenna (an
produces in fact a lower bound on the value since on-site sample of 600 interviews, August 2002) and at
Trieste (a sample of 600 residents, November 2002).1
the change is non-marginal (from the point of view
of the individual house owner); a more appropriate The purpose of the surveys was to value informal
measure is the WTP to prevent the loss, which may beach recreation (a non-marketable good); the value
be difficult to measure due to the emotional nature of of the daily beach use was estimated per individual
the good. An upper bound may be the discounted visitor. The methodology that was chosen is a version
value of the building not allowing for market adjust- of the CV method implementing the Value Of Enjoy-
ment of the building price. The rationale would be ment (VOE) as described in the Yellow Manual of
that from a welfare point of view, what matters is
that the people who would lose their house to the sea
must find a replacement, that is, a house not threa- 1
For the Lido di Dante survey, the tourists’ characteristics may
tened by the sea, for which the market does not change depending on the months of the tourist season, since that site
is mainly visited by foreigners and Italians from different regions.
adjust the price. Instead of a complete loss, erosion
The other sites are visited by residents or people who live nearby,
may only increase the probability of temporary
who generally visit the beach from May to September, but also in
flooding; the literature (see Dorfman et al., 1996) autumn–winter. Therefore, the results of the Lido di Dante survey
suggests valuing that loss of welfare through hedonic likely describe only the preferences of the tourists present on the
pricing. beach at the time of survey.
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 823
Penning-Rowsell et al. (1992). The valuation question
has an open ended format: respondents are asked to
state the value of enjoyment at the seafront in different
scenarios. Alternative formats of CV (such as those
implementing the WTP format for example) require
the specification of a payment vehicle (such as a tax,
entry fee or voluntary donation), while this is not
required for the VOE version. At the Lido di Dante
beach, Trieste (Barcola) seafront and Pellestrina
beach, which are beaches with no admittance fee, at
the time of the surveys any form of payment would
have been unpopular, therefore the VOE format was
found preferable for beach visitors and residents.
Beach access is not free of charge on most of the
beach at Ostia, but the VOE format was nevertheless
applied to compare the results with those of the other
Italian sites.
In CV surveys with the VOE format, each user is
asked to estimate the value he/she attributes to the
enjoyment obtained from a visit to the beach in dif-
ferent scenarios. At the heart of the CV approach is the
questionnaire, presenting plausible scenarios in which
the valuations can be made. To make those valuation Composition 1. Simulation of the Barcola seafront after the beach
exercises easier, the respondents are shown visual expansion.
support such as pictures representing the various sce-
narios. For example, the visitors to a certain beach can about o million. The beach uses in the status quo
17
be shown pictures of the beach in its current state and and in the expansion scenarios were evaluated in two
pictures of what the same beach would look like if seasons: spring/summer and autumn/winter. In the
erosion was allowed to take place. The basic VOE Pellestrina survey only the value of the status quo
questionnaires used for the Italian case studies are (an already completely artificial beach as shown in
those published in Penning-Rowsell et al. (Appen- Picture 1) is estimated.
In the Lido di Dante questionnaire, beach use is
dices 4.2(a) and (b)): the Standard site user question-
naire and the Standard resident questionnaire. The valued in three scenarios: status quo, hypothetical
questionnaires were adapted to the Italian case studies erosion and hypothetical expansion. Pictures 2 and 3,
by asking the beach use value not only in spring/ and Compositions 2–5 were presented to respondents.
summer but also in autumn/winter. The Lido di Dante beach is divided into two parts: the
Since each of the four sites has distinctive char- developed and semi-developed area (where sunbathing
acteristics, different questionnaires were used. The buildings are on the beach — mainly in the developed
main characteristic of the Trieste (Barcola) question- part), and the undeveloped or natural area. These two
beach areas were photographed in their current state at
naire is the valuation of the beach use in two scenarios
(status quo and a hypothetical artificial beach expan- the survey time. Picture 2 describes the status quo of
sion). The Barcola seafront is defended from the sea the developed and semi-developed area, while Com-
by an artificial wall that protects the road and pedes- positions 2 and 3 describe the same area in the
trian paths, and there is currently a very narrow pebble hypothetical situations of erosion and artificial expan-
beach. Composition 1 was presented to respondents, sion, respectively. Picture 3, instead, describes the
describing the project of building two artificial bea- status quo of the natural area, while Compositions 4
ches at the Barcola seafront, each 400 m long and 40 and 5 describe this area in the hypothetical situations
of erosion and expansion, respectively.
m wide. The total cost of the project was estimated at
824 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
while the situation of erosion is shown in Picture 5;
both pictures were presented to the respondents.
Italian nationals were interviewed in Trieste, Pel-
lestrina Island and Ostia, while in Lido di Dante, an
international tourist site, foreign visitors were also
interviewed. Most respondents favor the artificial pro-
tection of beaches from erosion. Composite inter-
vention (groynes, nourishment and submerged
breakwaters) and pure nourishment are the most pre-
ferred kinds of defense structures (see Marzetti et al.,
2003). Regarding the time spent on the beach in the
present state, in spring/summer the daily beach use of
Italian beaches is generally intense: in Lido di Dante
people stay about 5 h per day on average, 2.4 in
Trieste, 4 in Ostia, and 4 (day visitors) and 3.2 (resi-
dents) in Pellestrina. In autumn/winter however, the
time spent on the beach is about 1 h. The mean
number of days spent on Italian beaches in spring/
summer is fairly high: Lido di Dante about 12.4 days
(tourists), 23 (day visitors) and 47 (residents); Trieste
(residents) 15 days; Ostia (residents and day visitors)
89; and Pellestrina 70 days (residents) and 46 (day
visitors). The number of visit days in autumn/winter is
smaller than in spring/summer. In spring/summer a
number of respondents visit the beach more than once
per day.
Picture 1. Pellestrina Island beach. The individual value of the beach recreational use
changes according to the site characteristics. Table 2.1
In the Ostia survey, the status quo – already artifi- shows the mean daily use values of the four Italian
cially protected – and the situation of erosion are beaches according to the beach characteristics, scena-
valued; the status quo is described in Picture 4, rios, seasons, and population groups. Extreme values
Picture 2. Lido di Dante developed beach in its present state.
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 825
Picture 3. Lido di Dante undeveloped beach in its present state.
were excluded. Regarding the beach characteristics almost the same value by respondents, much higher
in the present state, the developed and semi-developed than the Barcola seafront in Trieste (very small gravel
areas of the Lido di Dante beach (Picture 2) have a beach), and Pellestrina (completely artificial, made of
lower value than the undeveloped (natural and unpro- dark sand, Picture 1).
tected, see Picture 3) area, probably because the latter Table 2.1 also shows considerable variations in the
is a natural beach with dunes; very rare in the region daily use value in each scenario status quo (present
(Marzetti and Zanuttigh, 2003). The undeveloped state), erosion and expansion, as indicated above. The
beach of Lido di Dante has a higher value than the eroded beach value is lower than the current state
undeveloped beach of Ostia (artificially expanded and beach value in Lido di Dante and Ostia (Compositions
less attractive). The developed Lido di Dante and 2 and 4, and Picture 5). The lowest mean use value for
Ostia beaches (very wide and long, with light sand, an eroded beach is elicited at Ostia. The estimated
and artificially protected; Pictures 2 and 4) are given value of the hypothetical artificially protected beach is
Composition 2. Lido di Dante developed beach in a hypothetical situation of erosion.
826 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
Composition 3. Lido di Dante developed beach in a hypothetical situation of expansion.
higher than the status quo value: in Lido di Dante the visitors is higher in autumn/winter, but the mean
mean use value of the protected beach (Compositions number of days and the daily mean hours are
3 and 5) is 2.5% higher than the status quo value, lower in autumn/winter. The values of the Lido di
while in Trieste it is 58.8% higher (Composition 1). Dante and Pellestrina beaches are much higher in
This divergence may be explained by the difference in spring/summer than in autumn/winter. Not only did
beach expansion with respect to the status quo. the respondents who visit the beach in autumn/winter
Considering the mean use value according to the state lower values (in summer they stay on the beach
different seasons, as shown in Table 2.1, the value of on average much longer than in winter), but the
the Barcola seafront in Trieste is slightly higher in majority of respondents do not visit the beach in
autumn/winter than in spring/summer; the number of winter. In particular, as regards the Lido di Dante
Composition 4. Lido di Dante undeveloped beach in a hypothetical situation of erosion.
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 827
Composition 5. Lido di Dante undeveloped beach in the hypothetical situations of expansion.
beach, the mean use values in autumn/winter have while in autumn/winter it was elicited from residents.
been computed for the whole sample (people who do This may be due to the fact that in spring/summer the
not visit the beach in autumn/winter have a zero tourists who travel to Lido di Dante on holiday value
value for the daily beach use) and for people who beach recreational activities highly; while the resi-
visit the beach in autumn/winter only. In spring/ dents in spring/summer suffer a loss of enjoyment
summer the main activities are sunbathing, relaxing due to congestion, and attribute a greater value to
and swimming, while in autumn/winter the majority daily beach use in autumn/winter because there is
of respondents only walk. no congestion. On Pellestrina Island, the residents’
Finally, considering population groups Table 2.1 average estimated value for the beach was higher
shows that at Lido di Dante, the highest mean use than for day visitors. The daily use value also changes
value in spring/summer was elicited from tourists, considerably according to nationalities. At Lido di
Picture 4. Ostia beach in the current state.
828 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
Picture 5. Ostia beach in an eroded state.
Dante, foreign visitors (except Dutch respondents) pret the valuation question conditionally on being at
gave higher use values than Italian visitors. or near the beach. Also, the visitors’ trip usually has
The VOE is intended to measure the value of the multiple destinations, and in practice it is not always
recreational activities on a specific beach or destina- possible to establish the share of this cost for one only
tion; it should be interpreted as the cost of the most destination. Consequently, the CV method with VOE
comparable activity. It is likely that respondents inter- format cannot be used to assess the influence of the
Table 2.1
Beach use values in Euros per person per day
Mean value Spring/summer Autumn/winter
Status quo Eroded Protected Status quo Expanded
Lido di Dante 27.67 13.26 28.37 4.10*
North1 (developed) 25.41 11.47 27.43 16.38**
North2 (semi-developed) 27.21 9.94 26.35 17.60**
South (undeveloped) 32.44 21.49 33.39 19.62**
Residents 10.25 9.33 23.14 27.89*
Day visitors 23.21 10.76 24.91 4.32*
Tourists 32.28 15.51 31.53 3.25*
Nationals 26.45 12.49 17.99
German 30.93 16.45 28.65
French 30.00 14.04 33.36
Swiss 53.33 28.70 36.38
Dutch 22.50 5.50 25.00
Other nationalities 39.33 14.08 31.73
Trieste (residents) 5.24 8.32 5.25* 6.45
Ostia 17.91 2.05
Developed area 23.28 2.47
Undeveloped area 6.21 1.15
Pellestrina 9.23 3.54*
Residents 9.69 5.01*
Non-residents 8.72 2.11*
* Indicates the whole sample; ** indicates people who visit the beach in autumn/winter only.
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 829
travel cost on the elicited beach use value. Respon- may take the nature of extreme flooding events. The
dents who do not like the eroded or artificially pro- coastal defense program of Venice consists of differ-
tected beaches have the option of going to an ent kinds of interventions: (i) defense and rebalance of
alternative beach. In the hypothetical erosion situa- the morphological and hydrodynamic system of the
tion, 16.4% of respondents would stop visiting the lagoon, (ii) defense of the buildings, (iii) elevation of
Lido di Dante beach, and 29.1% would visit it less or floors and pavements, (iv) protection of the natural
much less often, while as regards the Ostia beach 36% barriers of Pellestrina and Lido islands from sea ero-
of respondents would stop visiting the beach and 39% sion by the building of artificial beaches protected by
would visit less often. In the situation of expansion, low crested structures, and (v) the temporary closure
only a few respondents would reduce the number of of the three inlets with mobile floodgates – the famous
visits (4.8% in Lido di Dante and 4.5% in Trieste) and MO.S.E. – built inside the lagoon across each inlets.
would go to another beach. The amount of public funds involved is considerable.
Computation of the aggregate use values of the In particular, the Italian Government has allocated
considered beaches meets the difficulty of measuring about o million in 15 years (more than o million
65 4
the number of day visitors. No official data about the per year) for the implementation of MO.S.E as from
total number of visits per year to these beaches exist; 2005. Because public funds are scarce, the implemen-
only data about tourists are available from local tation of a coastal defense project competes with that
records. Nevertheless, if the sample is representative, of other projects. Therefore, not only does the use
using the CV survey, an estimate of the number of day value of Venice have to be included in the CBA, but
visitors on the beach can be made. For example, at also its option value and non-use values.
Lido di Dante, the CV survey shows that 44.8% of the A CV survey was administered to assess the future
respondents are day visitors and they visit the beach use and non-use values of Venice and its lagoon.
on average just under 23 days per year; using the VOE Depending on the relevant population, different
estimates from Table 2.1, it can then be shown that the kinds of surveys can be administered. Given the avail-
estimated total loss of enjoyment due to beach erosion able funds and because Venice is visited by 10 million
at Lido di Dante is more than o million per year
3 people of all nationalities per year, an on-site sam-
(Zanuttigh et al., 2005). Trieste, on the other hand, is pling of visitors (tourists and day visitors in the most
only visited by (about 235,000) residents; the beach crowded streets of Venice, national and foreign, aged
expansion is important, and the aggregate annual 18 or over) was chosen. The main aims of the survey
value of the beach change has been estimated about are: (i) to assess the amounts that the respondents are
o million per year.
15 willing to pay to maintain or improve the existing
quality level of Venice as cultural heritage; (ii) to
investigate the donation and non-donation motives
of the WTP; (iii) to collect information about the
3. The Venice case study
social characteristics of the respondents, and type
This section illustrates the valuation of the coastal and frequency of visits to Venice.
defense of a cultural and historical heritage site, the The questionnaire was drawn up considering the
city of Venice, with a focus on option and non-use specific characteristics of the site, and the kind of
values (see Table 1.1). The aim is to estimate the survey chosen; a detailed version of it can be found
willingness to pay (WTP) for the defense of Venice in Marzetti (2003, D28/B-I). In particular, for the
from sea flooding by means of a CV survey. In this value elicitation questions, the bmodified double-
section of the paper the main results are presented; for referendumQ format was used (double dichotomous
the complete results see Marzetti (2003, D28/B-I). choice plus open-ended questions; see Shechter et
In 1987, Venice and its lagoon were designated al., 1998). The payment vehicle was one donation
World Heritage Site by the UNESCO. The town, with per year. Respondents were first reminded that there
its architectural and historical characteristics, requires are many other worthy causes to contribute to, and
rational management and protection because it is presented with the high water defense program of
affected by floods and high water phenomena which Venice (they were shown Composition 6 below);
830 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
followed by visiting museums. A large proportion of
respondents (93%) are in favor of the implementation
of the protection program; of those against the project,
just over 3% were Italians and 6% non-Italians.
In answering the value elicitation questions, 71.1%
of the respondents stated that they would be willing to
pay at least o to cover the cost of the flood and
1
coastal defense program (77.7% of the Italians and
69% of the foreigners) and 40.9% would be willing to
pay more than o Considering the whole sample,
1.
respondents indicate values from 0 to 100; the mean
WTP for the defense of Venice is o4.85 per year
(standard deviation 11.16). The day visitors’ mean
WTP is o 3.95, while the tourists’ mean WTP is
5.56 (Marzetti and Lamberti, 2003). As shown in
o
Fig. 3.1, the mean WTP differs widely according to
nationality: French and German respondents have the
smallest mean values, while US and Italian respon-
dents the greatest mean values.
In addition, 64.4% of the people claiming to be
willing to pay at least o to cover the cost of the
1
Venice defense program are 100% sure that they
would indeed pay the stated amount if actually asked
to; 1.3% of the respondents claim to be very uncertain.
The mean subjective probability to pay is 0.88. Taking
the probability of paying into account, the expected
Composition 6. Venice Lagoon — The MO.S.E.
WTP is o 4.39 (standard deviation 10.41). Considering
then they were asked (i) whether they were willing to only those respondents who are certain to pay (368
pay o per year to a non profit agency for that
1 people), the mean WTP is o 7.81 (median 5.00 and
program; if the reply was yes, (ii) they were asked standard deviation 13.18). We highlight that, because
whether they were willing to pay more; if the reply Venice is a UNESCO World Heritage Site, the aggre-
was again yes, (iii) the maximum WTP was asked. gation level is the entire world (King, 1995); we
Given the hypothetical nature of the CV survey sce- cannot estimate the aggregate value of Venice ascribed
nario, the elicited WTP could be different from the true to option value and non-use values, but only the
WTP, therefore respondents were also asked how con- aggregate WTP of tourists and day visitors in Venice.
fident they were, on a scale from 1 to 100, that they Therefore, because Venice is visited by about 10 mil-
would really donate the elicited amount (Champ et al., lion people per year, the WTP of tourists and day
1997). Before administering the main survey, a pilot visitors in Venice for option price and non-use values
survey was administered to test the questionnaire. could be more than o40 million per year.
The sample consists of 1000 face-to-face inter- The respondents who were willing to pay at least
views of 10–15 min each; 24.2% of interviewees o for the cost of the program were also asked their
1
were Italians and 75.8% non-Italians (European and donation motives. Most of them were willing to pay
non-European). A high percentage of the non-Italians for preserving Venice for future generations, just over
were from Germany, Great Britain and the USA. 17% for visiting the city in the future, and 10.5% just
There were 55.7% of tourists and 44.3% of day for knowing that Venice exists. People who would not
visitors. 58.4% of the respondents revealed their donate for the protection program (289 respondents)
annual household income. The respondents’ main were also asked their motives. About 38% of these
recreational activity is walking around the streets, respondents think that paying for this project is the
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 831
State’s duty; just over 18% said that the protection is Access is forbidden to Normerven and its location
not their problem because they do not live in Venice; behind the dyke makes it invisible to all except those
and just under 12% thought that the money should be who are specifically searching it. Therefore, the site
spent on some other projects. has virtually no recreation or tourist value. It has no
value as a protective device either because it is so
small comparatively with the dyke it is set against; at
best it may reduce the maintenance cost of the dyke
4. The Normerven case study
but in such a small scale that it can be considered
Using a CV survey, we value a small restored negligible. There are, however, the classical non-use
marine natural area called Normerven in the Nether- motives for value: altruism, care for future genera-
lands. Normerven was formerly a natural mudflat tions, duty towards the environment. . .
set along the dyke protecting the Netherlands from In a face-to-face CV survey, the respondents were
the Waddenzee (a huge shallow lagoon). Because of presented (in their home) with hypothetical scenarios
human action, Normerven was reduced to a thin of valuation in which the site would be replicated at
band of land just in front of the dyke, but was various locations along the coast of the South Wad-
later restored to a state comparable to the historical denzee region. The respondents were told that the
one. The restoration was achieved by filling up the government of the Province (the relevant authority
area formerly occupied by the mudflat and defend- for that kind of project) intended to build from one
ing it from sea erosion by constructing two low to ten new sites similar to Normerven. After a descrip-
crested structures, one facing south and the other tion of Normerven and of the project in details, the
west, while the east was closed by the dyke. The respondents were shown three pairs of cards in
structures were just low enough to be overtopped sequence. Each card represented an alternative future
on a few winter tides but not more often; this had described in terms of two characteristics: the cost of
the purpose of maintaining suitable conditions for the project and the number of sites that would be built.
seabirds nesting. The restored area appears to be The so-called cost of the project is not in fact
stable since 1995 and has seen a spectacular related to the actual cost of building the new sites; it
increase in the number of nesting pairs of birds, is a hypothetical amount that varies among respon-
reaching for some species 2% to 3% of the Wad- dents. The purpose is to observe how the respondents
denzee population (based on computations from react to the bcostQ they are shown. For that reason, the
RIKZ, 1999). bcostQ is called a bbidQ in the current context. The
Fig. 3.1. Mean WTP according to nationalities.
832 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
exact location of the sites was shown on a map. The indicates that the respondent prefers the do-something
number of nesting pairs of birds that could be alternative to the do-nothing one. Empty cells are
expected was also stated, in absolute values and in empty by design.
relative terms with respect to the total for the whole As expected, in most cases the frequency of Yes
Waddenzee. In each pair of cards, one of the alter- decreases when the bid increases. It was expected that
natives was always the do-nothing option, that is, the frequency would increase when the number of
Normerven is not replicated; that costs zero since sites increased, but that turned out to be true only
maintenance of Normerven is negligible. The respon- from 1 to 3 sites. From 3 to 5 sites the frequency is
dents knew in advance that they would be shown three roughly stable, and then decreases sharply for 10 sites.
alternative futures, but they were not told which char- In other words, the marginal utility of an additional
acteristics they would have. For each pair of cards, the site is actually zero after the third site and negative
respondents were asked to indicate their preferred after the fifth site. The reason for that behavior may be
alternative. The payment vehicle was the real estate that the new sites are competing with other uses and
tax, paid by every household in the Netherlands, non uses of the coastline. Extra sites are not bother
because it is the only one on which the government things equalQ because they occupy space, thus the
of the province has a substantial influence. respondent’s WTP for an extra site can actually be
The sample was selected randomly from the census negative because his WTP includes the disutility of
file of the North region of the North-Holland pro- some lost space or increased nuisance. For example,
vince, where Normerven is located. The survey was some respondents stated that one of the sites would
administered sequentially in rounds of about 100 reduce the usage of a local sea port by partially
questionnaires (see Hanemann and Kanninen, 1999, blocking its entrance (each site location had in fact
for a survey of sequential administration). After each been planned with engineers and marine biologists).
round, a quick analysis of the answers made it possi- Too many birds may also generate a series of nui-
ble to update the bids if needed. Only one bid update sances. This feeling of competing usages or that there
occurred, between the 2nd and 3rd rounds. Exactly is already enough nature or birds in the region, is the
600 questionnaires were completed, out of which second motive (a little under 20%) for a No answer,
some 73 are excluded for this analysis. The two after the cost of the alternative (42.4%).
most typical reasons for exclusion are that the inter- Since the respondents had three valuation choices,
viewer made some mistakes in the alternatives that the most flexible model to represent their choices is
had to be shown to the respondents and that the the trivariate probit. It can be shown that with our
respondent chose not to answer (an option that he data, this model is observationally equivalent to a
was explicitly given). Since the remaining 527 obser- random effect panel data probit model in which the
vations have each three valuation choices, there are means are not equal to each other’s in the three
1581 lines of data. Table 4.1 shows the proportion of choices. The formulation of the model is described
Yes answer for each pair (bid, site). A bYesQ answer in Greene (1993). The estimation results confirm that
the larger the bid, the less likely is a Yes answer. An
increase in the number of sites corresponds to an
Table 4.1
increase in the probability of a bYesQ answer for low
Relative frequencies of yes
numbers of sites (1 to 3) but to a decrease for large
Bid # Extra sites # Observations
numbers (5 to 10). Respondents tend to answer bYesQ
1 3 5 10 more often when they are members of environmental
6 0.73 0.77 349 organizations, when they work part or full time, when
12 0.61 0.71 0.48 89 they spend a large part of their leisure outdoors, when
18 0.64 0.63 0.74 219
they think that there are many threats to the environ-
24 0.45 0.54 0.51 0.36 100
ment and when they think that many aspects of the
40 0.59 0.56 0.49 0.28 252
environment should be bhelped.Q A larger proportion
50 0.50 0.45 0.40 86
80 0.50 0.39 0.32 288 of bYesQ answers occurred in the first valuation ques-
150 0.39 0.22 198 tion than the next two. A similar phenomenon occurs
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 833
Fig. 4.1. Median WTP function.
in double-bounded CV; several explanations are pos- of each additional site decreases as the total number of
sible, see Hanemann and Kanninen (1999) for an sites increases. For a detailed version of these results,
overview of that discussion. see Polome et al. (2003).
´
The estimated model is a Random Utility Model. It
is compatible with economic theory and can be used
to extract a welfare measure in a manner similar to 5. Benefit transfer
that of Hanemann (1984). The relevant welfare mea-
sure in this case is the WTP because the survey This section presents an example of benefit transfer
depicts a situation in which the respondents do not for coastal defense. The technique of benefit transfer
own the additional natural areas and may (collec- is intended to assess whether and how economic
tively) decide whether to acquire them or not. The values known at some sites can be used to infer the
median of the WTP is the amount such that the value at an original site, called the study site. Ideally,
probability of a Yes answer is .5. It is a more robust one would like to estimate a transfer function for each
statistic than the expected WTP because it is less type of benefit present at a coastal defense site (Table
sensitive to the tails of the statistical distribution 1.1); that is, for each type of benefit, a function
chosen for estimation. The main results of the estima- linking the value to socio-economic and physical
tion are shown in Fig. 4.1. The results shown corre- characteristics of the study site. However, for most
spond to the most conservative scenario; they types of benefit there are only a few studies or none at
constitute a lower bound. all. The only exception is a composite of several
The value of the original Normerven site can be recreational activities at the coast, called binformal
extrapolated as shown in Fig. 4.1. It is apparent that it beach recreationQ in some references. A transfer func-
is this first site that generates most value. From there, tion for that category of benefit is estimated in this
the WTP follows a quadratic curve that culminates at section. That is the same category of benefit as the one
studied in the Italian case studies of Section 2.2 A
3 new sites and then starts decreasing (5 new sites are
still worth more than one). As discussed already after figure is also presented describing the probability that
Table 4.1, one should not be surprised of this phe- the transferred benefit fall within bounds of the value
nomenon: the additional sites are competing with that would have been estimated with a new study.
other uses in terms of space, thus respondents may Such a figure lets the users of the transfer function
consider that there are too many bird areas similar to decide what level of risk they are willing to take or
Normerven. whether they prefer instead to undertake a new study.
This result has a direct bearing on benefit transfer The data come from three sources. The first one is
(see the next section), namely that the value of two a library search of published and unpublished papers,
identical sites may differ accordingly with the order in
which they are provided. If the conclusions of this 2
Because of time constraints however, the results of those studies
chapter can be generalized, then the (marginal) value could not be included in the benefit transfer exercise.
834 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
including reports. It is important not to restrict the count and the effect on the estimation of the individual
search to published papers; otherwise a selection bias visitor’s value is unclear.
could appear. The second source of data comes from The data set has 106 observations, but only 38
unpublished British results collected by Professor C. different sites. Some sites have been observed during
Green (Flood Hazard Research Centre, Middlesex more than one year, and for some sites there were
University). Those data are scarce regarding the hypothetical behavior questions such as bhow much
description of each site being valued and the socio- would you value this beach if it was erodedQ (the
economic characteristics of the local or visiting actual phrasing of the question is unknown for most
populations. Furthermore the value concept used in studies). Only three countries provided data: the UK,
those data is the Value of Enjoyment (VOE) devel- with 79 observations, the US with 22, and the Nether-
oped by Penning-Rowsell et al. (1992) instead of the lands with 5.
more standard WTP. VOE is to be seen as an In our data set, there is information on three cate-
average of the prices of experiences similar to a gories of variables. A first category, X, is the site
visit to the beach; WTP is the maximum amount a characteristics, containing two variables: site type
person would pay to visit the beach or to preserve it, and site quality. Sites are classified in 3 types: Coastal
depending on what the researcher intends to esti- resort (101 observations), Beach (5) and Dune (2). A
mate. The third source of data comes from studies site can have three bquality levelsQ: current state (64
by the US National Oceanic and Atmospheric observations), eroded (20) or defended (24). This
Administration collected by Professor W. M. Hane- measure of quality is very coarse. bCurrentQ refers to
mann (University of California at Berkeley). Those the beach as it is at the moment of the study; as far as
data are also scarce regarding the physical descrip- we can say on the basis of the present data set, this is
tion of the beach and the socio-economic character- in fact a wide range of qualities. It merely denotes a
istics of the visitors, but they are based on more coastal site that is enjoyable under normal conditions.
conventional value concepts. bErodedQ indicates a quality, usually hypothetical, in
In none of the data sources used in this exercise which only a narrow range of the beach remains in
substitute sites were completely taken into account. place, if any. bDefendedQ indicates that a coastal
This is a drawback (see Herriges and Kling, 1999) and defense scheme, also usually hypothetical, is imple-
it indicates that the estimated values in each study are mented that partially modifies the aspect of the beach
to be taken as upper bounds because the loss of the and may enlarge it. We do not have information about
site corresponds to the value of the site. If substitutes the exact scheme that was used at each site; it is likely
are taken into account, the loss of the site corresponds that nourishment was the main defense, possible
to the difference of values with the next best site. accompanied at some sites by groynes or boulders.
Another shortcoming of benefit transfer in this case This is only a guess from the information that we
relates to the number of visits to the beach. All the have: a coastal site is usually the object of a study
available values are per visit to the beach. To estimate when it is already somewhat eroded; the defense
the value of the beach itself, it is still necessary to scheme aims at restoring it to previous levels.
know the total amount of visitors to the beach and A second category of variables, Y, is the socio-
their number of visits. That information was not avai- economic variables. The only variable here is repre-
lable and is generally difficult to acquire. An estima- sented by means of 4 categories of respondents: the
tion of the prospective visitors, who would appear local visitors (16 observations), the non-local visitors
following an improvement of the beach, was also among which those who stay a single day (15) and
absent. However most surveys are concerned about those who stay longer (15), and those observations for
preserving the beach in its current state; hence pros- which this distinction is not made (60).
pective visitors are not an issue. An additional pro- The last category of variables, Z, relates to the
blem is the on-site sample bias. That bias is due to the study itself. A first variable in this category is the
fact that when visitors are randomly selected on-site year the study took place, ranging from 1975 to 1995,
on a beach, the frequent visitors are over-sampled (see with most studies in the early nineties. A second
Shaw, 1988). This will bias upward the estimate of the variable is the concept of value that has been used:
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 835
VOE in 78 cases, WTP for use in 13 cases and example, the site is eroded). Model (2) is called a
consumer surplus in 15 cases. panel data model: for each site, there can be more than
The value itself is expressed per visit per person in one observation. The main formal difference with
Brouwer’s model (1) is that the intercept term a i is
Euro of 2001, adjusted by the consumer retail price
index of the relevant countries up to 2001 and then now specific to each site (it is indexed by i). The
converted to Euro using the average rate for 2001. The interesting feature of the site-specific intercept term of
average of the values (across all sites and all qualities) the panel data model (2) is to account for all the
is nearly 17, with standard deviation around 14, mini- differences in values across sites not accounted for
mum 1, maximum nearly 92. Table 5.1 compares the by the regressors. For example, although income is
data used in this report with the three other known not observed, the effect of the average visitors’
references in which a value for transfer is suggested. income on site i estimated value is captured by the
intercept term a i . Thus the OLS bias problem is
To formalise the analysis, we start with the proto-
type linear benefit transfer function from Brouwer avoided.
(2000): When the goal of the study is to estimate the
marginal effect on the measure of value (V) of a
Vi ¼ a þ bXi þ cYi þ dZi þ ei ; ð1Þ change in some characteristic of the beach, the panel
data model (2) is always to be preferred because it
where a, b, c, d are parameters to estimate, V is the
avoids the biases caused by the missing regressors.
value per site per visit for a given policy, X, Y and Z
However, more regressors can be included in model
have been defined above and i indexes the studies.
(1) than in model (2) because all the variables that do
Because we have no data on several variables that
not change over the year are captured by the indivi-
could explain the value, such as beach width and
dual specific constants a i of the panel data model (2)
length or respondents’ income, Ordinary Least
and have therefore to be excluded from that model.
Squares (OLS) estimation of the coefficients a, b, c,
For example, the country where the study took place
d of Brouwer’s model (1) is generally biased and
is a variable in model (1) but not in the panel data
inconsistent. This is a standard result about OLS:
model (2) because the site-specific intercepts repre-
missing regressors lead to bias on the coefficient
sent not only the country but also the region and any
estimates unless there is no correlation between the
variable which has no variation within one site. If we
missing variables and the included ones (an unlikely
would try to insert dummies representing the country
event). However, since in the current dataset, there is
in the panel data model (2), there would be linear
often more than one observation for a single site, an
dependence between them and the site-specific inter-
alternative benefit transfer function can be written as:
cepts a i and that would preclude estimation. There-
fore, when the goal of the study is to predict the value
Vit ¼ ai þ bXit þ cYit þ dZit þ eit ; ð2Þ
of one site given a series of characteristics, Brouwer’s
model (1) should be estimated using OLS. This is
where Vit is the value for site i in circumstance t. A
because biases in the estimated coefficients are not
circumstance can refer to a different point in time (a
important for prediction. Of course, the estimated
different year), or to some hypothetical situation (for
Table 5.1
Average value per visit to a beach (Euro of 2001)
Source Country Current state Eroded Defended Value concept
Average of available data UK 17.7 9.1 20.6 VOE
US 23.1 – – WTP for use or
consumer surplus
Penning-Rowsell et al. (1992) bgenericQ beach UK 15.6 8.2 18.7 VOE
US National Oceanic and Atmospheric Administration US 13.9 – – WTP for use
(informal communication, 1995) bstandardQ beach
Loomis and Crespi (1999) US 22.4 – – WTP for use
836 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
Table 5.2 Consumer Surplus, in the second one (Table 5.3) it
Panel data model bWTPQ is WTP, the dummy indicating the WTP for use.
Variable Coefficient P-value The first thing to remark from these tables is that
they are quite similar with the exception of the inter-
Intercept 19.38 0.002
T 0.22 0.49 cept term. The intercept changes because of the two
different dummies (WTP or CS), this is reasonable
Category of visitor (default is bunspecifiedQ)
because these dummies indicate a change in the aver-
Day 4.70 0.22
age value of the site (the default is different), and
Local 1.55 0.69
hence of the intercept. The coefficients of the regres-
Stay 4.12 0.29
sors change little; this indicates that the effect of these
Concept of value (default is VOE or CS)
variables on the value is similar whatever the concept
À15.67
WTP 0
of value that is used. The effect of time (T) is statis-
tically negligible; that is, the value of sites for coastal
Quality of the site (default is bcurrentQ)
informal recreation has not changed noticeably
À8.37
Eroded 0
Defended 3.30 0.02 between 1975 and 1995 (in real terms since the
value is expressed in Euros of 2001). The effect of
coefficients of model (1) have no interpretation since the type of respondents (Local residents, Day visitors,
they are biased. Below the estimates of both models Stay visitors or Unspecified) is not statistically sig-
are presented and we show that the panel data model nificant either. The quality of the site (Current,
(2) is not as good a tool as model (1) when it comes to Defended, Eroded) is unquestionably very significant.
predict the value of one site. Finally, the high significance of the concept of
value used (VOE, WTP for use, Consumer Surplus)
5.1. Marginal effect: panel data results is worrisome. It is acceptable that different concepts
of value yield different values, but the problem is that
The date (T) of the study is a cardinal variable and different valuation methods and designs have been
is inserted in the regressions as a natural trend starting used for the different concepts. Therefore, we cannot
in 1975 (normalized to 1). The 4 categories of visitors tell whether the differences in value are genuine or are
(local residents, day visitors, stay visitors and unspe- led by the valuation method that has been used. If it is
cified type) are represented using three dichotomous the former, we would still have to decide which con-
variables (Local, Day, Stay), with the omitted category cept of value is more appropriate. If it is the latter,
(the default) being the unspecified type. The 3 cate- then benefit transfer of informal beach recreation is
gories of quality of the site (eroded, current quality, partially flawed since different methods lead to differ-
defended) are represented using two dichotomous
Table 5.3
variables (Eroded, Defended); the omitted category
Panel data model bCSQ
is the current quality. Finally, the concepts of value
Variable Coefficient P-value
(VOE, WTP for use, Consumer Surplus) have been
Intercept 10.22 0.08
represented by 2 dichotomous variables (WTP, CS),
T 0.22 0.48
the omitted category being VOE. It turns out that the
sum of WTP and CS is a vector of zeros and ones Category of visitor (default is bunspecifiedQ)
identical to the sum of certain site-specific constants; Day 6.26 0.11
this is a direct consequence of the fact that in most Local 3.12 0.42
Stay 5.67 0.14
cases the value of a site has been estimated using a
single concept of value. Therefore, one of these 2
Concept of value (default is VOE or WTP)
variables had to be removed to enable estimation CS 15.90 0
(otherwise, perfect collinearity impedes estimation),
but since the decision to remove is arbitrary, we pre- Quality of the site (default is bcurrentQ)
À8.32
Eroded 0
sent the 2 sets of results: in the first one (Table 5.2) the
Defended 3.30 0.01
variable removed is CS, the dummy indicating the
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 837
Table 5.4 the value concept variable is capturing some of the
OLS estimate of model (1) site or socio-economic characteristics because VOE is
Variables Coefficient Number only used in the UK and CS is mostly used in the US.
of cases
The average value is around 16 (of 2001) for the UK
À9.35
Intercept and 22 for the US sites.
T (1975 = 1, each year is one) 1.87
Country of study (default is UK) 79
5.2. Predicting values: ordinary least squares results
US 23.56 22
NL 1.39 5
As stated above, because of missing regressors, the
Type of site (default is bcoastal resortQ; 99
that is, an bequippedQ beach) OLS estimator of the coefficients of model (1) is
À10.94
Beach 5
generally biased and inconsistent. It is therefore not
À10.47
Dune 2
worth trying to correct for other possible estimation
Type of visitor (default is bunspecifiedQ) 60
problems. Yet, when the goal of estimation is predic-
À7.82
Day 15
tion (that is, transfer), bias in the estimated coeffi-
À9.78
Local 16
À8.00
Stay 15 cients is of no importance. Whether the coefficients
Concept of value (default is VOE) 78
are biased or not, the OLS estimator minimizes the
À22.66
WTP 13
prediction error by construction.
À12.44
CS 15
The estimation results are described in Table 5.4.
bQualityQ of the beach 64
As explained above, there are more variables in model
(default is bcurrentQ state)
À9.27
Eroded 20 (1) than in the panel data model (2), but the OLS
Unspecified defense 2.95 14
estimator is inconsistent and therefore the estimates of
À1.47
Defended by nourishment 4
the coefficients are not meaningful. For that reason,
Defended by nourishment plus groynes 3.13 4
their significance is not shown. The overall fit
(adjusted R 2) of the model is about 40%, but it is
ent values for the same beach. For certain situations, unclear whether it can be taken as a good measure of
the researcher was imposed the method (e.g. in the fit in the current context.
UK, only a specific CV format was admissible for
claims of funding to the former Ministry of Agricul- 5.3. Benefit transfer
ture, Fisheries and Food), but in general, this suggests
a lack of standards in applying valuation methods to To transfer values to an entirely new site, that is,
beach recreation. On the other hand, it is possible that to predict the value of the new site, one would
Fig. 5.1. Benefit transfer cumulative distribution of prediction errors.
838 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
simply substitute the new site’s characteristics and we have preferred the simplest models. More details
the estimates of Table 5.4 into model (1). To estimate can be found in Polome (2002).
´
the size of the error that one could commit by
proceeding in this way, we have devised the follow-
ing exercise. For each site, we ran the panel data and 6. Conclusions
OLS regressions without that site’s observation(s)
and predict its value. Then, to measure the gain of The contributions presented in this paper have
precision obtained by carrying a new study, we shown the important diversity of coastal values –
compared the predicted value(s) with the one(s) from informal enjoyment of a beach to heritage and
obtained from the original study(ies). The measure nonuse values – and have provided examples and
of prediction error is the proportion of deviation illustrations of estimation of these values. The focus
from the value reported for the site in absolute has been on the valuation of non market benefits.
term. In Fig. 5.1 below, the line referring to the In Section 2, the Penning-Rowsell et al. (1992)
baverageQ (triangles) represents the average-value value of enjoyment methodology has been adapted
prediction, that is, the value of one site is set equal for the valuation of four Italian beaches. These sur-
to the average value of all the other sites, regardless veys have shown mean values for informal recreation
of the sites characteristics. on a beach in its current state from o to o per
5 28
Fig. 5.1 reports the proportion (on the vertical visit. This is therefore of the same order of magnitude
axis) of predictions that falls below the error level as the US and UK beaches, even though there are
indicated on the horizontal axis. For example, the large variations across beaches, and some respondents
proportion of errors no larger than 40% in transfer- sometimes express very large values. The Italian sur-
ring a value is about 70% for OLS and 55% when veys have also shown that coastal visitors are sensitive
the prediction is the average of the values of the to the protection of coastal sites from erosion and
other sites. In other words, when transferring value flooding and that they are generally in favor of
using model (1) estimated by OLS (Table 5.4), there defense projects. The value of enjoyment may also
is a 70% chance of making an error of 40% or less, vary considerably accordingly with the season
and (approximately) a 90% chance of making an error (spring/summer or autumn/winter) or the type of visi-
of 100% or less. Fig. 5.1 is a truncation of the tor (resident or tourist).
complete plot since there is a non-zero probability The contingent valuation methodology can be used
of making an error larger than 200% (about 5% to value other, very different, types of coastal assets.
chance with OLS). In this paper, two cases were illustrated. The Venice
We say that model A predicts better than model B case study in Section 3 indicated values of an order
when the cumulative distribution of prediction errors between o and o per year per visitor for protecting
4 5
of model A is above that of model B. In that sense as Venice and its lagoon from erosion and recurrent
can be seen in Fig. 5.1, the panel data model with flooding using a complex defense scheme. A large
WTP dummy (Table 5.2) performs worse than the proportion of respondents expressed their certainty to
simple average of values. That does not undermine pay the amount elicited if actually asked, though some
the qualities of the panel data model (2) as an did not. This confirms the usefulness of a dcertainty
unbiased estimation of regression coefficients, but questionT after the valuation questions in order to
for prediction (that is, transfer) purposes, the best estimate the expected mean donation.
model is model (1) estimated by OLS. That is not to In the Normerven contingent valuation survey in
say that better estimators cannot be found, but we Section 4, we estimated a value function for increas-
would have to resort to more sophisticated econo- ing the number of seabird nesting areas in a Dutch
metric estimators (e.g. Tobit models to take into coastal province. It has been shown that even if the
account that values cannot be negative). Given the first area could have a rather large (purely non-use)
limitations of the data set on the one hand and, on the value (close to o per year for 10 years), subsequent
20
other hand, the fact that the transfer function should areas have strongly decreasing values. One implica-
be easily usable by non-economists field practitioners, tion of this result for the transfer of benefit is that
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 839
replication of a defense scheme may not lead to a References
replication of its value.
In Section 5, a compilation of the existing (pub- Bateman, I.J., Willis, K.G., 1999. Valuing Environmental Prefer-
ences: Theory and Practice of the Contingent Valuation Method
lished and unpublished) evidence on coastal protec-
in the US, EU and Developing Countries. Oxford University
tion benefit estimates has made clear that it does not
Press, Oxford.
seem possible to estimate a function that would Bower, Blair T., Turner, R. Kerry, 1998. Characterising and analysis
provide the total value of a coastal defense scheme benefits from integrated coastal management (ICM). Ocean and
for any single site. Instead, previous studies have Coastal Management 38, 41 – 66.
Brouwer, R., 2000. Environmental value transfer: state of the art and
concentrated on specific types of benefits. In parti-
future prospects. Ecological Economics 32 (1), 137 – 152.
cular, it appears that informal beach recreation has
Champ, P.A., Bishop, R.C., Brown, T.C., McCollun, D.W., 1997.
been studied more than any other type of activity at Using donation mechanisms to value nonuse benefits from
the beach, but even in this case most of the esti- public good. Journal of Environmental Economics and Manage-
mated values come from either the US or the UK. ment 33, 156 – 162.
Cordes, Joseph J., Yezer, Anthony M.J., 1998. In harm’s way: does
The reader should be cautious when using the results
federal spending on beach enhancement and protection induce
of the benefit transfer function presented in Section
excessive development in coastal areas? Land Economics 74
2 outside those two countries. Even within these two (1), 128 – 145.
countries, the probability that the transferred value is Dorfman, J.H., Keeler, A.G., Kriesel, W., 1996. Valuing risk-redu-
within 50% of the value from an original study is cing interventions with hedonic models: the case of erosion
protection. Journal of Agricultural and Resource Economics
only 75% (see Fig. 5.1). The average value of
21 (1), 109 – 119.
informal recreation on a beach in its current state
Goodman, S.L., Seabrooke, W., Daniel, H.M., Jaffry, S.A., James,
in these two countries, all methodologies together, is H., 1996. Results of a contingent valuation study of non-use
around o (of 2001) per visit. This value is within
20 values of coastal resources. MAFF.
the bounds of the Italian case studies presented in Greene, W.H., 1993. Econometric Analysis, 2nd edition. Prentice-
Hall, Englewood Cliffs.
Section 2.
Haab, T., McConnell, K.E., 2002. 1999. Valuing environmental and
natural resources: the econometrics of non-market valuation.
Edward Elgar, Cheltenham, UK.
Acknowledgements Hanemann, W.M., 1984. Welfare evaluations in contingent valua-
tion experiments with discrete responses. American Journal of
Agricultural Economics 66, 332 – 341.
EU support through RTD project DELOS, con-
Hanemann, W.M., Kanninen, B., 1999. The statistical analysis of
tract EVK3-CT-2000-00041, is gratefully acknow-
discrete-response CV data. In: Bateman, I., Willis, K. (Eds.),
ledged. Thanks are due Prof. C. Green and Prof. Valuing Environmental Preferences. Oxford University Press.
W. M. Hanemann for communicating data for the Hanley, N., Splash, C.L., 1993. Cost–Benefit Analysis and the
benefit transfer section. For the Italian case studies, Environment. Edward Elgar, Aldershot.
Herriges, J.A., Kling, C.L. (Eds.), Valuing Recreation and the
thanks are due to A. Lamberti for the engineering
Environment. Edward Elgar, Cheltenham.
aspects of the defense projects, the City Council of
King, O.H., 1995. Estimating the value of marine resources: a
Ravenna, the City Council of Trieste and the Con- marine recreation case. Ocean and Coastal Management 27
sorzio Venezia Nuova for their support and the (1–2), 129 – 141.
material provided, A. Cazzola and R. Filippini for Lipton, D.W., Wellman, K., Sheifer, I.C., Weiher, R.L., 1995.
Economic Evaluation of Natural Resources — A Handbook
the questionnaire formatting, F. Galassi for her work
for Coastal Policymakers: Silver Spring, Md.: U.S. Dept. of
with Excel and SPSS, B. Zanuttigh for the Lido di
Commerce, National Oceanic and Atmospheric Administration,
Dante pictures and Venice composition, C. Barbanti Coastal Ocean Office. Resource available at http://www.mdsg.
for the interviews in Pellestrina Island and collabora- umd.edu/Extension/valuation/handbook.htm.
tion on the Venice composition, and L. Franco and P. Loomis, J., Crespi, J., 1999. Estimated effects of climate change on
selected outdoor activities in the United States. In: Mendelsohn,
Scaloni for the interviews in Ostia. For the Dutch
Robert, Neumann, James-E. (Eds.), The Impact of Climate
case study, thanks are due Mindert de Vries (Delft
Change on the United States Economy. Cambridge University
Hydraulics) and Jentsje van der Meer (Infram) for Press, Cambridge, pp. xi,331.
the biological and engineering aspects of the defense Marzetti Dall’Aste Brandolini, S. 2003. D28/A and D28/B–I of the
project. DELOS final report D28 bEconomic and Social Valuation about
840 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
European Coastal Sites.Q Available on the DELOS webpage Polome, P. 2002. DELOS WP 4.1 Final Report (Benefit Transfer).
´
www.delos.unibo.it/. Available on the DELOS webpage www.delos.unibo.it/.
Marzetti Dall’Aste Brandolini, S., Lamberti, A., 2003. Economic Polome, P., van der Veen, A., Geurts, P.A.T.M., 2003. Contingent
´
and social valuation of the defence system of Venice and its valuation of a restored coastal natural area. Assessment of
lagoon (Italy). In: Ozhan, E. (Ed.), Proceedings of the Sixth Option Value and Non-Use Values — The Case-Studies of
International Conference on the Mediterranean Coastal Envi- Venice (Italy) and Normerven (The Netherlands), Report D28/
ronment, MEDCOAST 03, 7–11 October 2003, pp. 307 – 318. B-II of the DELOS Project.
Marzetti Dall’Aste Brandolini, Zanuttigh, B., 2003. Economic and RIKZ (National Institute for Coastal and Marine Management),
Social Valuation of Beach protection in Lido di Dante (Italy). In: 1999. Waddenzee Quality Status Rapport. Rapport RIKZ/
Ozhan, E. (Ed.), Proceedings of the Sixth International Con- 2000.008 (ISSN 0927-3980).
ference on the Mediterranean Coastal Environment, MED- Shaw, D., 1988. On-site samples’ regressions. Journal of Econo-
COAST 03, 7–11 October 2003. 319 – 330. metrics 37, 211 – 223.
Marzetti Dall’Aste Brandolini, S., Franco, L., Lamberti, A., Zanut- Shechter, M., Reiser, B., Zaitsev, N., 1998. Measuring passive use
tigh, B., 2003. bPreferences about Different Kinds of value. Environmental and Resources Economics 12, 457 – 478.
Low Crested Structures and Beach Materials: the Italian Case- Turner, R.K., Bateman, I., Brooke, J.S., 1992. Valuing the benefits
studies of Lido di Dante, Ostia and Pellestrina Island,Q D28/C of of coastal defence: a case study of the Adelburgh sea-defence
the DELOS project. Available on the DELOS webpage scheme. In: Coker, A., Richards, C. (Eds.), Valuing the Envi-
www.delos.unibo.it/. ronment: Economic Approaches to Environmental Valuation. J.
Mendelsohn, R., Neumann, J.E. (Eds.), 1999. The Impact of Cli- Wiley and Sons, New York.
mate Change on the United States Economy. Cambridge Uni- Zanuttigh, B., Martinelli, L., Lamberti, A., Moschella, P., Hawkins,
versity Press, Cambridge. S., Marzetti, S., Ceccherelli, V.U., 2005. Environmental Design
Penning-Rowsell,, Green, C.H., Thompson, P.M., Coker, A.M., of Coastal Defence In Lido di Dante, Italy, Coastal Engineering,
Tunstall, S.M., Richards, C., Parker, D.J., 1992. The Economics Special Issue on DELOS.
of Coastal Management: A Manual of Benefit Assessment
Techniques (Yellow Manual). Belhaven Press, London.
www.elsevier.com/locate/coastaleng
Economic and social demands for coastal protection
P. Polome a,*, S. Marzetti b, A. van der Veen a
´
a
CTW–WEM, University of Twente, P.O. Box 217, 7500AE, The Netherlands
b
Department of Economics, University of Bologna, Italy
Available online 25 October 2005
Abstract
The purpose of this paper is to present methods and examples of economic valuation in the framework of cost–benefit
analysis of coastal defense schemes. We summarize the concepts of value in economics and their application to coastal erosion
defense. We describe the results of an original benefit transfer exercise on beach recreation, that is, whether and how values
known for some sites can be used to assess the value of some other sites. We present six original case studies on the valuation of
the benefits of coastal erosion defense; four of them focus on beach recreation in Italy, one focuses on the conservation of the
Venice heritage, and one on biodiversity in The Netherlands. The results of the case studies are illustrative of the diversity of
values for the many types of non-marketed assets that may be protected from sea erosion.
D 2005 Elsevier B.V. All rights reserved.
Keywords: Economic valuation; Coastal erosion; Non-market benefits; Benefit transfer
intended as illustrations of the variety of coastal
1. Introduction
defense benefits and their valuation. Section 2 pre-
The purpose of this paper is to present methods and sents the results of original studies on the valuation of
examples of economic valuation in the framework of recreational benefits of coastal defense for four Italian
cost–benefit analysis (CBA) of coastal defense beaches. These case studies should be fairly represen-
schemes. The paper is intended for a broad scientific tative of coastal defense schemes for Northern Med-
audience without prior knowledge of economics. The iterranean beaches. Section 3 presents the very special
introduction of the paper presents the principles of case of the defense of the Venice lagoon. Section 4 is
CBA, summarizes the main notions of economic radically different since it is about a small unused
value, the most well-known valuation methods and natural area in the Northern Sea. Section 5 introduces
the main potential costs and benefits of coastal the technique of benefit transfer that is whether and
defense schemes. The following three sections are how economic values known at some sites can be
used to infer in some way the value of an original
site. This technique, when it can be applied, is very
economical because an economic valuation study can
* Corresponding author.
be quite expensive. The results of an original benefit
E-mail address: polome@ecru.ucl.ac.be (P. Polome).
´
0378-3839/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.coastaleng.2005.09.009
820 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
transfer exercise are presented. We do not claim that ferent valuation methods. Classical typologies of
the valuation studies presented in the present paper are values adapted from Turner et al. (1992), and Bower
representative of all the possible valuation cases of and Turner (1998) are presented in Table 1.1. The
coastal defense; yet we trust that, as a result of this third column indicates the valuation methods that
paper, the reader will have a general idea of what can would be most suitable for estimating each value.
be done regarding cost–benefit analysis of defense This is not an indication that it has been estimated.
schemes and will have enough examples to draw An overview of the valuation methods is given in the
upon to build his own valuation exercise, be it a sequel.
transfer exercise or an original study. That is the We now turn to a brief introduction of the eco-
main purpose of our paper. nomic valuation methods. The necessary data are
CBA is a process intended to measure whether the generally too specific to exist in any publicly available
sum of all the positive impacts of a project outweighs database and it is often necessary to use surveys to
the sum of its negative impacts once they are converted collect the data or to resort to benefit transfer (Section
in a single unit, often money; for a thorough review of 5). The valuation methods are divided into bstated
CBA in the case of environmental changes, see Hanley preferencesQ and brevealed preferencesQ; a detailed
and Spash (1993). In this introduction, we will review description can be found in Haab and McConnell
briefly the economic notion of value, the valuation (2002). Revealed preferences methods rely on actions
methods, the types of value, and the types of asset that individuals have taken in the past; one can dis-
that can be found at the coastline. The economic con- tinguish between bdirectQ and bindirectQ revealed pre-
cept of value that is most often used in a CBA is the ferences methods. Direct methods refer to changes
Willingness to Pay (WTP) defined as the maximum that directly affect marketed goods. A typical example
amount of money a person is willing to exchange to in the case of coastal defense is the demand for hotel
acquire a (public or market) good or service. The nights at a specific coastal resort. Indirect methods
economic value does not refer to an exchange of refer to changes in the provision of a non-marketed
money or to a price; the goal is to convert bindividual good that can be valued indirectly through estimation
utilityQ into money to match it against monetary costs of the changes in the demand of an associated mar-
such as those of building a coastal defense scheme. The keted good. A good example in the context of coastal
WTP is used, and not market prices, because the defense is the recreation quality of a beach. Recreation
coastal defense scheme changes the supply of non- is not in itself sold in a market; however, to enjoy
marketed goods: a government provides the defense recreation at the beach, visitors have to travel there.
scheme, but cannot charge the consumers for it; CBA One can then estimate the demand for travel to the
addresses this issue by converting the change of well- beach and proceed as in a case of direct methods.
being into money, and compares it to the actual money Direct methods, or bmarket pricingQ as indicated in
that has been spent on providing the good. The con- Table 1.1 can be briefly described as follows (see Fig.
version should be based on individual preferences; that 1.1; see also Lipton et al., 1995). First, the demand
is the case in the present paper. That definition of schedule of the market good is estimated. The sche-
economic value makes clear that a broad class of dule can be estimated at individual level (the price is
benefits should be considered in CBA. Yet, economic the observed individual price) or at the market aggre-
value is not the only criterion for deciding on public gated level. The area defined by the horizontal price
projects; equity considerations, precautionary environ- line, the demand curve and the vertical axis is defined
mental standards, and regional economic constraints as the consumer surplus. The producer surplus is
can be seen as complements to CBA. defined similarly, but is often not estimated in practice
One purpose of this introduction is to make clear because supply is assumed completely inelastic (ver-
the diversity of value categories and assets at the tical schedule). Second, using the estimated demand
coast. The value of a coastal defense scheme is com- schedule, we forecast the change in demand caused by
posed of the sum of the values of the consequences of the change that we want to value (e.g. an eroded beach
that scheme on the seafront, avoiding double-count- versus a nourished beach); in Fig. 1.1, the demand
ing. Often different types of values will require dif- schedule shifts up. The change in value is the change
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 821
Table 1.1
Coastal defense values
Use generated values
Direct use values Consumptive: fishing; agriculture; transport; construction Market pricing (possibly adjusted)
and maintenance costs
Non-consumptive: recreation Travel cost; stated preferences
Indirect use values Flood control; storm protection; sedimentation; Market pricing; hedonic pricing;
habitat loss reduction; landscape; human health stated preferences
Non-use and option generated values
Option values Insurance value of preserving options for use Stated preferences
Quasi-option values Value of increased information in the future (biodiversity) Stated preferences
Existence and bequest values Knowing that a species or system is conserved; passing on Stated preferences
natural/heritage assets intact to future generations;
moral resource/non-human rights
in consumer surplus, in most cases a good approxima- is marginal, the supply of the additional nights has a
tion to the WTP for the change. zero (or very low) marginal cost. If the change is not
The complete procedure of estimation of the sup- marginal, for example if hotels have to be built to
ply and demand schedules, and forecasting their accommodate the additional nights, then costs have to
change, is often a complex task, especially if there be taken into account and the demand and supply
exist goods that are substitute or complement to the schedules should be estimated.
market good of interest. Things may be simpler if Indirect revealed preferences methods are used for
the change can be said to be marginal. In that case, goods for which there is normally no observable
the price of a market good is sometimes equivalent to demand but there is a complementary or substitute
the marginal social benefit of a unit of that good; as an market good. The travel cost method is concerned
approximation, and if the market can be said to be with changes in the quality of a recreational site.
competitive, the social benefit of a project that The value of that site is estimated on the basis of
increases marginally the output of such a good can the demand for travel to that site, travel being the
be taken as the product of price and quantity. For market good complement to the recreational site.
example, regarding the increase in the number of Hedonic pricing captures the WTP associated with
hotel nights caused by a (small) beach protection variations in property values that result from the pre-
scheme, the marginal social benefit can be said to sence or absence of specific environmental attributes.
be equal to the number of additional nights times Stated preferences methods are used for changes in
the price of the rooms on the ground that if the change non marketed goods such as landscape, natural or
Fig. 1.1. Consumer and producer surplus.
822 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
cultural heritage that have no complementary or sub- Enhancement effects include: increased output of
stitute market good. In that case, one can only resort the seafront (e.g. creation of recreational fishing
to directly asking individuals (in a survey) how much opportunities); water quality changes (eutrophication,
they are willing to pay to obtain that change (or to red tides); conflicts among different types of recrea-
avoid it). The precise way to ask that question is the tion users of beach areas.
subject of much debate and has given rise in practice Preservation effects refer to natural areas. The
to several methods. The contingent valuation (CV) is benefits stemming from the preservation of a natural
the most developed stated preferences method and is ecosystem are generally recreational use and non-use.
very well documented, see e.g. Bateman and Willis An in-depth case is described in Goodman et al.
(1999). Several examples are presented in details in (1996). Offshore sand and gravel mining (e.g. to
Sections 2, 3 and 4 of this paper. find the sand for beach nourishment) may affect fish-
We now turn to the question of what types of assets eries and habitats.
might be affected by a coastal defense scheme. We Indirect economic effects are bsecond roundQ
present here a summarized list; for a more detailed effects, e.g. constructions in hazardous areas in rela-
list, see Bower and Turner (1998), the bYellow Man- tion to coastal storms that are built because of the
ualQ of Penning-Rowsell et al. (1992) and Polome ´ protection granted by the defense scheme (resulting
(2002). possibly in a stronger scheme being necessary in the
Mitigation effects of coastal defense include the future; see Cordes and Yezer, 1998).
following categories: reduce damage to or prevent
destruction of coastal properties and cultural and heri-
tage assets from coastal storms and eroding shorelines; 2. Case studies on the use value of Italian beaches
reduce salinity intrusion; reduce sedimentation; restore
or preserve habitats or recreational opportunities (e.g. In this section, we present the most significant
sand beach). results of four case-studies at Italian beaches. For
Buildings damage can be valued in two ways. the complete results, see Marzetti (2003, D28/A).
Erosion can cause complete loss of the building Two small surveys were administered at the beach
(sinking); the literature (Mendelsohn and Neumann, of Ostia near Rome (100 interviews on the beach,
1999) suggests estimating the discounted value of summer 2002) and on Pellestrina Island in the Lagoon
the building from the current time until the expected of Venice (80 residents and 75 beach visitors, July
sinking time, allowing for market adjustment of the 2002), respectively. Two larger surveys were adminis-
building price (zero at the time of sinking). That tered at Lido di Dante near the town of Ravenna (an
produces in fact a lower bound on the value since on-site sample of 600 interviews, August 2002) and at
Trieste (a sample of 600 residents, November 2002).1
the change is non-marginal (from the point of view
of the individual house owner); a more appropriate The purpose of the surveys was to value informal
measure is the WTP to prevent the loss, which may beach recreation (a non-marketable good); the value
be difficult to measure due to the emotional nature of of the daily beach use was estimated per individual
the good. An upper bound may be the discounted visitor. The methodology that was chosen is a version
value of the building not allowing for market adjust- of the CV method implementing the Value Of Enjoy-
ment of the building price. The rationale would be ment (VOE) as described in the Yellow Manual of
that from a welfare point of view, what matters is
that the people who would lose their house to the sea
must find a replacement, that is, a house not threa- 1
For the Lido di Dante survey, the tourists’ characteristics may
tened by the sea, for which the market does not change depending on the months of the tourist season, since that site
is mainly visited by foreigners and Italians from different regions.
adjust the price. Instead of a complete loss, erosion
The other sites are visited by residents or people who live nearby,
may only increase the probability of temporary
who generally visit the beach from May to September, but also in
flooding; the literature (see Dorfman et al., 1996) autumn–winter. Therefore, the results of the Lido di Dante survey
suggests valuing that loss of welfare through hedonic likely describe only the preferences of the tourists present on the
pricing. beach at the time of survey.
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 823
Penning-Rowsell et al. (1992). The valuation question
has an open ended format: respondents are asked to
state the value of enjoyment at the seafront in different
scenarios. Alternative formats of CV (such as those
implementing the WTP format for example) require
the specification of a payment vehicle (such as a tax,
entry fee or voluntary donation), while this is not
required for the VOE version. At the Lido di Dante
beach, Trieste (Barcola) seafront and Pellestrina
beach, which are beaches with no admittance fee, at
the time of the surveys any form of payment would
have been unpopular, therefore the VOE format was
found preferable for beach visitors and residents.
Beach access is not free of charge on most of the
beach at Ostia, but the VOE format was nevertheless
applied to compare the results with those of the other
Italian sites.
In CV surveys with the VOE format, each user is
asked to estimate the value he/she attributes to the
enjoyment obtained from a visit to the beach in dif-
ferent scenarios. At the heart of the CV approach is the
questionnaire, presenting plausible scenarios in which
the valuations can be made. To make those valuation Composition 1. Simulation of the Barcola seafront after the beach
exercises easier, the respondents are shown visual expansion.
support such as pictures representing the various sce-
narios. For example, the visitors to a certain beach can about o million. The beach uses in the status quo
17
be shown pictures of the beach in its current state and and in the expansion scenarios were evaluated in two
pictures of what the same beach would look like if seasons: spring/summer and autumn/winter. In the
erosion was allowed to take place. The basic VOE Pellestrina survey only the value of the status quo
questionnaires used for the Italian case studies are (an already completely artificial beach as shown in
those published in Penning-Rowsell et al. (Appen- Picture 1) is estimated.
In the Lido di Dante questionnaire, beach use is
dices 4.2(a) and (b)): the Standard site user question-
naire and the Standard resident questionnaire. The valued in three scenarios: status quo, hypothetical
questionnaires were adapted to the Italian case studies erosion and hypothetical expansion. Pictures 2 and 3,
by asking the beach use value not only in spring/ and Compositions 2–5 were presented to respondents.
summer but also in autumn/winter. The Lido di Dante beach is divided into two parts: the
Since each of the four sites has distinctive char- developed and semi-developed area (where sunbathing
acteristics, different questionnaires were used. The buildings are on the beach — mainly in the developed
main characteristic of the Trieste (Barcola) question- part), and the undeveloped or natural area. These two
beach areas were photographed in their current state at
naire is the valuation of the beach use in two scenarios
(status quo and a hypothetical artificial beach expan- the survey time. Picture 2 describes the status quo of
sion). The Barcola seafront is defended from the sea the developed and semi-developed area, while Com-
by an artificial wall that protects the road and pedes- positions 2 and 3 describe the same area in the
trian paths, and there is currently a very narrow pebble hypothetical situations of erosion and artificial expan-
beach. Composition 1 was presented to respondents, sion, respectively. Picture 3, instead, describes the
describing the project of building two artificial bea- status quo of the natural area, while Compositions 4
ches at the Barcola seafront, each 400 m long and 40 and 5 describe this area in the hypothetical situations
of erosion and expansion, respectively.
m wide. The total cost of the project was estimated at
824 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
while the situation of erosion is shown in Picture 5;
both pictures were presented to the respondents.
Italian nationals were interviewed in Trieste, Pel-
lestrina Island and Ostia, while in Lido di Dante, an
international tourist site, foreign visitors were also
interviewed. Most respondents favor the artificial pro-
tection of beaches from erosion. Composite inter-
vention (groynes, nourishment and submerged
breakwaters) and pure nourishment are the most pre-
ferred kinds of defense structures (see Marzetti et al.,
2003). Regarding the time spent on the beach in the
present state, in spring/summer the daily beach use of
Italian beaches is generally intense: in Lido di Dante
people stay about 5 h per day on average, 2.4 in
Trieste, 4 in Ostia, and 4 (day visitors) and 3.2 (resi-
dents) in Pellestrina. In autumn/winter however, the
time spent on the beach is about 1 h. The mean
number of days spent on Italian beaches in spring/
summer is fairly high: Lido di Dante about 12.4 days
(tourists), 23 (day visitors) and 47 (residents); Trieste
(residents) 15 days; Ostia (residents and day visitors)
89; and Pellestrina 70 days (residents) and 46 (day
visitors). The number of visit days in autumn/winter is
smaller than in spring/summer. In spring/summer a
number of respondents visit the beach more than once
per day.
Picture 1. Pellestrina Island beach. The individual value of the beach recreational use
changes according to the site characteristics. Table 2.1
In the Ostia survey, the status quo – already artifi- shows the mean daily use values of the four Italian
cially protected – and the situation of erosion are beaches according to the beach characteristics, scena-
valued; the status quo is described in Picture 4, rios, seasons, and population groups. Extreme values
Picture 2. Lido di Dante developed beach in its present state.
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 825
Picture 3. Lido di Dante undeveloped beach in its present state.
were excluded. Regarding the beach characteristics almost the same value by respondents, much higher
in the present state, the developed and semi-developed than the Barcola seafront in Trieste (very small gravel
areas of the Lido di Dante beach (Picture 2) have a beach), and Pellestrina (completely artificial, made of
lower value than the undeveloped (natural and unpro- dark sand, Picture 1).
tected, see Picture 3) area, probably because the latter Table 2.1 also shows considerable variations in the
is a natural beach with dunes; very rare in the region daily use value in each scenario status quo (present
(Marzetti and Zanuttigh, 2003). The undeveloped state), erosion and expansion, as indicated above. The
beach of Lido di Dante has a higher value than the eroded beach value is lower than the current state
undeveloped beach of Ostia (artificially expanded and beach value in Lido di Dante and Ostia (Compositions
less attractive). The developed Lido di Dante and 2 and 4, and Picture 5). The lowest mean use value for
Ostia beaches (very wide and long, with light sand, an eroded beach is elicited at Ostia. The estimated
and artificially protected; Pictures 2 and 4) are given value of the hypothetical artificially protected beach is
Composition 2. Lido di Dante developed beach in a hypothetical situation of erosion.
826 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
Composition 3. Lido di Dante developed beach in a hypothetical situation of expansion.
higher than the status quo value: in Lido di Dante the visitors is higher in autumn/winter, but the mean
mean use value of the protected beach (Compositions number of days and the daily mean hours are
3 and 5) is 2.5% higher than the status quo value, lower in autumn/winter. The values of the Lido di
while in Trieste it is 58.8% higher (Composition 1). Dante and Pellestrina beaches are much higher in
This divergence may be explained by the difference in spring/summer than in autumn/winter. Not only did
beach expansion with respect to the status quo. the respondents who visit the beach in autumn/winter
Considering the mean use value according to the state lower values (in summer they stay on the beach
different seasons, as shown in Table 2.1, the value of on average much longer than in winter), but the
the Barcola seafront in Trieste is slightly higher in majority of respondents do not visit the beach in
autumn/winter than in spring/summer; the number of winter. In particular, as regards the Lido di Dante
Composition 4. Lido di Dante undeveloped beach in a hypothetical situation of erosion.
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 827
Composition 5. Lido di Dante undeveloped beach in the hypothetical situations of expansion.
beach, the mean use values in autumn/winter have while in autumn/winter it was elicited from residents.
been computed for the whole sample (people who do This may be due to the fact that in spring/summer the
not visit the beach in autumn/winter have a zero tourists who travel to Lido di Dante on holiday value
value for the daily beach use) and for people who beach recreational activities highly; while the resi-
visit the beach in autumn/winter only. In spring/ dents in spring/summer suffer a loss of enjoyment
summer the main activities are sunbathing, relaxing due to congestion, and attribute a greater value to
and swimming, while in autumn/winter the majority daily beach use in autumn/winter because there is
of respondents only walk. no congestion. On Pellestrina Island, the residents’
Finally, considering population groups Table 2.1 average estimated value for the beach was higher
shows that at Lido di Dante, the highest mean use than for day visitors. The daily use value also changes
value in spring/summer was elicited from tourists, considerably according to nationalities. At Lido di
Picture 4. Ostia beach in the current state.
828 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
Picture 5. Ostia beach in an eroded state.
Dante, foreign visitors (except Dutch respondents) pret the valuation question conditionally on being at
gave higher use values than Italian visitors. or near the beach. Also, the visitors’ trip usually has
The VOE is intended to measure the value of the multiple destinations, and in practice it is not always
recreational activities on a specific beach or destina- possible to establish the share of this cost for one only
tion; it should be interpreted as the cost of the most destination. Consequently, the CV method with VOE
comparable activity. It is likely that respondents inter- format cannot be used to assess the influence of the
Table 2.1
Beach use values in Euros per person per day
Mean value Spring/summer Autumn/winter
Status quo Eroded Protected Status quo Expanded
Lido di Dante 27.67 13.26 28.37 4.10*
North1 (developed) 25.41 11.47 27.43 16.38**
North2 (semi-developed) 27.21 9.94 26.35 17.60**
South (undeveloped) 32.44 21.49 33.39 19.62**
Residents 10.25 9.33 23.14 27.89*
Day visitors 23.21 10.76 24.91 4.32*
Tourists 32.28 15.51 31.53 3.25*
Nationals 26.45 12.49 17.99
German 30.93 16.45 28.65
French 30.00 14.04 33.36
Swiss 53.33 28.70 36.38
Dutch 22.50 5.50 25.00
Other nationalities 39.33 14.08 31.73
Trieste (residents) 5.24 8.32 5.25* 6.45
Ostia 17.91 2.05
Developed area 23.28 2.47
Undeveloped area 6.21 1.15
Pellestrina 9.23 3.54*
Residents 9.69 5.01*
Non-residents 8.72 2.11*
* Indicates the whole sample; ** indicates people who visit the beach in autumn/winter only.
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 829
travel cost on the elicited beach use value. Respon- may take the nature of extreme flooding events. The
dents who do not like the eroded or artificially pro- coastal defense program of Venice consists of differ-
tected beaches have the option of going to an ent kinds of interventions: (i) defense and rebalance of
alternative beach. In the hypothetical erosion situa- the morphological and hydrodynamic system of the
tion, 16.4% of respondents would stop visiting the lagoon, (ii) defense of the buildings, (iii) elevation of
Lido di Dante beach, and 29.1% would visit it less or floors and pavements, (iv) protection of the natural
much less often, while as regards the Ostia beach 36% barriers of Pellestrina and Lido islands from sea ero-
of respondents would stop visiting the beach and 39% sion by the building of artificial beaches protected by
would visit less often. In the situation of expansion, low crested structures, and (v) the temporary closure
only a few respondents would reduce the number of of the three inlets with mobile floodgates – the famous
visits (4.8% in Lido di Dante and 4.5% in Trieste) and MO.S.E. – built inside the lagoon across each inlets.
would go to another beach. The amount of public funds involved is considerable.
Computation of the aggregate use values of the In particular, the Italian Government has allocated
considered beaches meets the difficulty of measuring about o million in 15 years (more than o million
65 4
the number of day visitors. No official data about the per year) for the implementation of MO.S.E as from
total number of visits per year to these beaches exist; 2005. Because public funds are scarce, the implemen-
only data about tourists are available from local tation of a coastal defense project competes with that
records. Nevertheless, if the sample is representative, of other projects. Therefore, not only does the use
using the CV survey, an estimate of the number of day value of Venice have to be included in the CBA, but
visitors on the beach can be made. For example, at also its option value and non-use values.
Lido di Dante, the CV survey shows that 44.8% of the A CV survey was administered to assess the future
respondents are day visitors and they visit the beach use and non-use values of Venice and its lagoon.
on average just under 23 days per year; using the VOE Depending on the relevant population, different
estimates from Table 2.1, it can then be shown that the kinds of surveys can be administered. Given the avail-
estimated total loss of enjoyment due to beach erosion able funds and because Venice is visited by 10 million
at Lido di Dante is more than o million per year
3 people of all nationalities per year, an on-site sam-
(Zanuttigh et al., 2005). Trieste, on the other hand, is pling of visitors (tourists and day visitors in the most
only visited by (about 235,000) residents; the beach crowded streets of Venice, national and foreign, aged
expansion is important, and the aggregate annual 18 or over) was chosen. The main aims of the survey
value of the beach change has been estimated about are: (i) to assess the amounts that the respondents are
o million per year.
15 willing to pay to maintain or improve the existing
quality level of Venice as cultural heritage; (ii) to
investigate the donation and non-donation motives
of the WTP; (iii) to collect information about the
3. The Venice case study
social characteristics of the respondents, and type
This section illustrates the valuation of the coastal and frequency of visits to Venice.
defense of a cultural and historical heritage site, the The questionnaire was drawn up considering the
city of Venice, with a focus on option and non-use specific characteristics of the site, and the kind of
values (see Table 1.1). The aim is to estimate the survey chosen; a detailed version of it can be found
willingness to pay (WTP) for the defense of Venice in Marzetti (2003, D28/B-I). In particular, for the
from sea flooding by means of a CV survey. In this value elicitation questions, the bmodified double-
section of the paper the main results are presented; for referendumQ format was used (double dichotomous
the complete results see Marzetti (2003, D28/B-I). choice plus open-ended questions; see Shechter et
In 1987, Venice and its lagoon were designated al., 1998). The payment vehicle was one donation
World Heritage Site by the UNESCO. The town, with per year. Respondents were first reminded that there
its architectural and historical characteristics, requires are many other worthy causes to contribute to, and
rational management and protection because it is presented with the high water defense program of
affected by floods and high water phenomena which Venice (they were shown Composition 6 below);
830 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
followed by visiting museums. A large proportion of
respondents (93%) are in favor of the implementation
of the protection program; of those against the project,
just over 3% were Italians and 6% non-Italians.
In answering the value elicitation questions, 71.1%
of the respondents stated that they would be willing to
pay at least o to cover the cost of the flood and
1
coastal defense program (77.7% of the Italians and
69% of the foreigners) and 40.9% would be willing to
pay more than o Considering the whole sample,
1.
respondents indicate values from 0 to 100; the mean
WTP for the defense of Venice is o4.85 per year
(standard deviation 11.16). The day visitors’ mean
WTP is o 3.95, while the tourists’ mean WTP is
5.56 (Marzetti and Lamberti, 2003). As shown in
o
Fig. 3.1, the mean WTP differs widely according to
nationality: French and German respondents have the
smallest mean values, while US and Italian respon-
dents the greatest mean values.
In addition, 64.4% of the people claiming to be
willing to pay at least o to cover the cost of the
1
Venice defense program are 100% sure that they
would indeed pay the stated amount if actually asked
to; 1.3% of the respondents claim to be very uncertain.
The mean subjective probability to pay is 0.88. Taking
the probability of paying into account, the expected
Composition 6. Venice Lagoon — The MO.S.E.
WTP is o 4.39 (standard deviation 10.41). Considering
then they were asked (i) whether they were willing to only those respondents who are certain to pay (368
pay o per year to a non profit agency for that
1 people), the mean WTP is o 7.81 (median 5.00 and
program; if the reply was yes, (ii) they were asked standard deviation 13.18). We highlight that, because
whether they were willing to pay more; if the reply Venice is a UNESCO World Heritage Site, the aggre-
was again yes, (iii) the maximum WTP was asked. gation level is the entire world (King, 1995); we
Given the hypothetical nature of the CV survey sce- cannot estimate the aggregate value of Venice ascribed
nario, the elicited WTP could be different from the true to option value and non-use values, but only the
WTP, therefore respondents were also asked how con- aggregate WTP of tourists and day visitors in Venice.
fident they were, on a scale from 1 to 100, that they Therefore, because Venice is visited by about 10 mil-
would really donate the elicited amount (Champ et al., lion people per year, the WTP of tourists and day
1997). Before administering the main survey, a pilot visitors in Venice for option price and non-use values
survey was administered to test the questionnaire. could be more than o40 million per year.
The sample consists of 1000 face-to-face inter- The respondents who were willing to pay at least
views of 10–15 min each; 24.2% of interviewees o for the cost of the program were also asked their
1
were Italians and 75.8% non-Italians (European and donation motives. Most of them were willing to pay
non-European). A high percentage of the non-Italians for preserving Venice for future generations, just over
were from Germany, Great Britain and the USA. 17% for visiting the city in the future, and 10.5% just
There were 55.7% of tourists and 44.3% of day for knowing that Venice exists. People who would not
visitors. 58.4% of the respondents revealed their donate for the protection program (289 respondents)
annual household income. The respondents’ main were also asked their motives. About 38% of these
recreational activity is walking around the streets, respondents think that paying for this project is the
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 831
State’s duty; just over 18% said that the protection is Access is forbidden to Normerven and its location
not their problem because they do not live in Venice; behind the dyke makes it invisible to all except those
and just under 12% thought that the money should be who are specifically searching it. Therefore, the site
spent on some other projects. has virtually no recreation or tourist value. It has no
value as a protective device either because it is so
small comparatively with the dyke it is set against; at
best it may reduce the maintenance cost of the dyke
4. The Normerven case study
but in such a small scale that it can be considered
Using a CV survey, we value a small restored negligible. There are, however, the classical non-use
marine natural area called Normerven in the Nether- motives for value: altruism, care for future genera-
lands. Normerven was formerly a natural mudflat tions, duty towards the environment. . .
set along the dyke protecting the Netherlands from In a face-to-face CV survey, the respondents were
the Waddenzee (a huge shallow lagoon). Because of presented (in their home) with hypothetical scenarios
human action, Normerven was reduced to a thin of valuation in which the site would be replicated at
band of land just in front of the dyke, but was various locations along the coast of the South Wad-
later restored to a state comparable to the historical denzee region. The respondents were told that the
one. The restoration was achieved by filling up the government of the Province (the relevant authority
area formerly occupied by the mudflat and defend- for that kind of project) intended to build from one
ing it from sea erosion by constructing two low to ten new sites similar to Normerven. After a descrip-
crested structures, one facing south and the other tion of Normerven and of the project in details, the
west, while the east was closed by the dyke. The respondents were shown three pairs of cards in
structures were just low enough to be overtopped sequence. Each card represented an alternative future
on a few winter tides but not more often; this had described in terms of two characteristics: the cost of
the purpose of maintaining suitable conditions for the project and the number of sites that would be built.
seabirds nesting. The restored area appears to be The so-called cost of the project is not in fact
stable since 1995 and has seen a spectacular related to the actual cost of building the new sites; it
increase in the number of nesting pairs of birds, is a hypothetical amount that varies among respon-
reaching for some species 2% to 3% of the Wad- dents. The purpose is to observe how the respondents
denzee population (based on computations from react to the bcostQ they are shown. For that reason, the
RIKZ, 1999). bcostQ is called a bbidQ in the current context. The
Fig. 3.1. Mean WTP according to nationalities.
832 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
exact location of the sites was shown on a map. The indicates that the respondent prefers the do-something
number of nesting pairs of birds that could be alternative to the do-nothing one. Empty cells are
expected was also stated, in absolute values and in empty by design.
relative terms with respect to the total for the whole As expected, in most cases the frequency of Yes
Waddenzee. In each pair of cards, one of the alter- decreases when the bid increases. It was expected that
natives was always the do-nothing option, that is, the frequency would increase when the number of
Normerven is not replicated; that costs zero since sites increased, but that turned out to be true only
maintenance of Normerven is negligible. The respon- from 1 to 3 sites. From 3 to 5 sites the frequency is
dents knew in advance that they would be shown three roughly stable, and then decreases sharply for 10 sites.
alternative futures, but they were not told which char- In other words, the marginal utility of an additional
acteristics they would have. For each pair of cards, the site is actually zero after the third site and negative
respondents were asked to indicate their preferred after the fifth site. The reason for that behavior may be
alternative. The payment vehicle was the real estate that the new sites are competing with other uses and
tax, paid by every household in the Netherlands, non uses of the coastline. Extra sites are not bother
because it is the only one on which the government things equalQ because they occupy space, thus the
of the province has a substantial influence. respondent’s WTP for an extra site can actually be
The sample was selected randomly from the census negative because his WTP includes the disutility of
file of the North region of the North-Holland pro- some lost space or increased nuisance. For example,
vince, where Normerven is located. The survey was some respondents stated that one of the sites would
administered sequentially in rounds of about 100 reduce the usage of a local sea port by partially
questionnaires (see Hanemann and Kanninen, 1999, blocking its entrance (each site location had in fact
for a survey of sequential administration). After each been planned with engineers and marine biologists).
round, a quick analysis of the answers made it possi- Too many birds may also generate a series of nui-
ble to update the bids if needed. Only one bid update sances. This feeling of competing usages or that there
occurred, between the 2nd and 3rd rounds. Exactly is already enough nature or birds in the region, is the
600 questionnaires were completed, out of which second motive (a little under 20%) for a No answer,
some 73 are excluded for this analysis. The two after the cost of the alternative (42.4%).
most typical reasons for exclusion are that the inter- Since the respondents had three valuation choices,
viewer made some mistakes in the alternatives that the most flexible model to represent their choices is
had to be shown to the respondents and that the the trivariate probit. It can be shown that with our
respondent chose not to answer (an option that he data, this model is observationally equivalent to a
was explicitly given). Since the remaining 527 obser- random effect panel data probit model in which the
vations have each three valuation choices, there are means are not equal to each other’s in the three
1581 lines of data. Table 4.1 shows the proportion of choices. The formulation of the model is described
Yes answer for each pair (bid, site). A bYesQ answer in Greene (1993). The estimation results confirm that
the larger the bid, the less likely is a Yes answer. An
increase in the number of sites corresponds to an
Table 4.1
increase in the probability of a bYesQ answer for low
Relative frequencies of yes
numbers of sites (1 to 3) but to a decrease for large
Bid # Extra sites # Observations
numbers (5 to 10). Respondents tend to answer bYesQ
1 3 5 10 more often when they are members of environmental
6 0.73 0.77 349 organizations, when they work part or full time, when
12 0.61 0.71 0.48 89 they spend a large part of their leisure outdoors, when
18 0.64 0.63 0.74 219
they think that there are many threats to the environ-
24 0.45 0.54 0.51 0.36 100
ment and when they think that many aspects of the
40 0.59 0.56 0.49 0.28 252
environment should be bhelped.Q A larger proportion
50 0.50 0.45 0.40 86
80 0.50 0.39 0.32 288 of bYesQ answers occurred in the first valuation ques-
150 0.39 0.22 198 tion than the next two. A similar phenomenon occurs
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 833
Fig. 4.1. Median WTP function.
in double-bounded CV; several explanations are pos- of each additional site decreases as the total number of
sible, see Hanemann and Kanninen (1999) for an sites increases. For a detailed version of these results,
overview of that discussion. see Polome et al. (2003).
´
The estimated model is a Random Utility Model. It
is compatible with economic theory and can be used
to extract a welfare measure in a manner similar to 5. Benefit transfer
that of Hanemann (1984). The relevant welfare mea-
sure in this case is the WTP because the survey This section presents an example of benefit transfer
depicts a situation in which the respondents do not for coastal defense. The technique of benefit transfer
own the additional natural areas and may (collec- is intended to assess whether and how economic
tively) decide whether to acquire them or not. The values known at some sites can be used to infer the
median of the WTP is the amount such that the value at an original site, called the study site. Ideally,
probability of a Yes answer is .5. It is a more robust one would like to estimate a transfer function for each
statistic than the expected WTP because it is less type of benefit present at a coastal defense site (Table
sensitive to the tails of the statistical distribution 1.1); that is, for each type of benefit, a function
chosen for estimation. The main results of the estima- linking the value to socio-economic and physical
tion are shown in Fig. 4.1. The results shown corre- characteristics of the study site. However, for most
spond to the most conservative scenario; they types of benefit there are only a few studies or none at
constitute a lower bound. all. The only exception is a composite of several
The value of the original Normerven site can be recreational activities at the coast, called binformal
extrapolated as shown in Fig. 4.1. It is apparent that it beach recreationQ in some references. A transfer func-
is this first site that generates most value. From there, tion for that category of benefit is estimated in this
the WTP follows a quadratic curve that culminates at section. That is the same category of benefit as the one
studied in the Italian case studies of Section 2.2 A
3 new sites and then starts decreasing (5 new sites are
still worth more than one). As discussed already after figure is also presented describing the probability that
Table 4.1, one should not be surprised of this phe- the transferred benefit fall within bounds of the value
nomenon: the additional sites are competing with that would have been estimated with a new study.
other uses in terms of space, thus respondents may Such a figure lets the users of the transfer function
consider that there are too many bird areas similar to decide what level of risk they are willing to take or
Normerven. whether they prefer instead to undertake a new study.
This result has a direct bearing on benefit transfer The data come from three sources. The first one is
(see the next section), namely that the value of two a library search of published and unpublished papers,
identical sites may differ accordingly with the order in
which they are provided. If the conclusions of this 2
Because of time constraints however, the results of those studies
chapter can be generalized, then the (marginal) value could not be included in the benefit transfer exercise.
834 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
including reports. It is important not to restrict the count and the effect on the estimation of the individual
search to published papers; otherwise a selection bias visitor’s value is unclear.
could appear. The second source of data comes from The data set has 106 observations, but only 38
unpublished British results collected by Professor C. different sites. Some sites have been observed during
Green (Flood Hazard Research Centre, Middlesex more than one year, and for some sites there were
University). Those data are scarce regarding the hypothetical behavior questions such as bhow much
description of each site being valued and the socio- would you value this beach if it was erodedQ (the
economic characteristics of the local or visiting actual phrasing of the question is unknown for most
populations. Furthermore the value concept used in studies). Only three countries provided data: the UK,
those data is the Value of Enjoyment (VOE) devel- with 79 observations, the US with 22, and the Nether-
oped by Penning-Rowsell et al. (1992) instead of the lands with 5.
more standard WTP. VOE is to be seen as an In our data set, there is information on three cate-
average of the prices of experiences similar to a gories of variables. A first category, X, is the site
visit to the beach; WTP is the maximum amount a characteristics, containing two variables: site type
person would pay to visit the beach or to preserve it, and site quality. Sites are classified in 3 types: Coastal
depending on what the researcher intends to esti- resort (101 observations), Beach (5) and Dune (2). A
mate. The third source of data comes from studies site can have three bquality levelsQ: current state (64
by the US National Oceanic and Atmospheric observations), eroded (20) or defended (24). This
Administration collected by Professor W. M. Hane- measure of quality is very coarse. bCurrentQ refers to
mann (University of California at Berkeley). Those the beach as it is at the moment of the study; as far as
data are also scarce regarding the physical descrip- we can say on the basis of the present data set, this is
tion of the beach and the socio-economic character- in fact a wide range of qualities. It merely denotes a
istics of the visitors, but they are based on more coastal site that is enjoyable under normal conditions.
conventional value concepts. bErodedQ indicates a quality, usually hypothetical, in
In none of the data sources used in this exercise which only a narrow range of the beach remains in
substitute sites were completely taken into account. place, if any. bDefendedQ indicates that a coastal
This is a drawback (see Herriges and Kling, 1999) and defense scheme, also usually hypothetical, is imple-
it indicates that the estimated values in each study are mented that partially modifies the aspect of the beach
to be taken as upper bounds because the loss of the and may enlarge it. We do not have information about
site corresponds to the value of the site. If substitutes the exact scheme that was used at each site; it is likely
are taken into account, the loss of the site corresponds that nourishment was the main defense, possible
to the difference of values with the next best site. accompanied at some sites by groynes or boulders.
Another shortcoming of benefit transfer in this case This is only a guess from the information that we
relates to the number of visits to the beach. All the have: a coastal site is usually the object of a study
available values are per visit to the beach. To estimate when it is already somewhat eroded; the defense
the value of the beach itself, it is still necessary to scheme aims at restoring it to previous levels.
know the total amount of visitors to the beach and A second category of variables, Y, is the socio-
their number of visits. That information was not avai- economic variables. The only variable here is repre-
lable and is generally difficult to acquire. An estima- sented by means of 4 categories of respondents: the
tion of the prospective visitors, who would appear local visitors (16 observations), the non-local visitors
following an improvement of the beach, was also among which those who stay a single day (15) and
absent. However most surveys are concerned about those who stay longer (15), and those observations for
preserving the beach in its current state; hence pros- which this distinction is not made (60).
pective visitors are not an issue. An additional pro- The last category of variables, Z, relates to the
blem is the on-site sample bias. That bias is due to the study itself. A first variable in this category is the
fact that when visitors are randomly selected on-site year the study took place, ranging from 1975 to 1995,
on a beach, the frequent visitors are over-sampled (see with most studies in the early nineties. A second
Shaw, 1988). This will bias upward the estimate of the variable is the concept of value that has been used:
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 835
VOE in 78 cases, WTP for use in 13 cases and example, the site is eroded). Model (2) is called a
consumer surplus in 15 cases. panel data model: for each site, there can be more than
The value itself is expressed per visit per person in one observation. The main formal difference with
Brouwer’s model (1) is that the intercept term a i is
Euro of 2001, adjusted by the consumer retail price
index of the relevant countries up to 2001 and then now specific to each site (it is indexed by i). The
converted to Euro using the average rate for 2001. The interesting feature of the site-specific intercept term of
average of the values (across all sites and all qualities) the panel data model (2) is to account for all the
is nearly 17, with standard deviation around 14, mini- differences in values across sites not accounted for
mum 1, maximum nearly 92. Table 5.1 compares the by the regressors. For example, although income is
data used in this report with the three other known not observed, the effect of the average visitors’
references in which a value for transfer is suggested. income on site i estimated value is captured by the
intercept term a i . Thus the OLS bias problem is
To formalise the analysis, we start with the proto-
type linear benefit transfer function from Brouwer avoided.
(2000): When the goal of the study is to estimate the
marginal effect on the measure of value (V) of a
Vi ¼ a þ bXi þ cYi þ dZi þ ei ; ð1Þ change in some characteristic of the beach, the panel
data model (2) is always to be preferred because it
where a, b, c, d are parameters to estimate, V is the
avoids the biases caused by the missing regressors.
value per site per visit for a given policy, X, Y and Z
However, more regressors can be included in model
have been defined above and i indexes the studies.
(1) than in model (2) because all the variables that do
Because we have no data on several variables that
not change over the year are captured by the indivi-
could explain the value, such as beach width and
dual specific constants a i of the panel data model (2)
length or respondents’ income, Ordinary Least
and have therefore to be excluded from that model.
Squares (OLS) estimation of the coefficients a, b, c,
For example, the country where the study took place
d of Brouwer’s model (1) is generally biased and
is a variable in model (1) but not in the panel data
inconsistent. This is a standard result about OLS:
model (2) because the site-specific intercepts repre-
missing regressors lead to bias on the coefficient
sent not only the country but also the region and any
estimates unless there is no correlation between the
variable which has no variation within one site. If we
missing variables and the included ones (an unlikely
would try to insert dummies representing the country
event). However, since in the current dataset, there is
in the panel data model (2), there would be linear
often more than one observation for a single site, an
dependence between them and the site-specific inter-
alternative benefit transfer function can be written as:
cepts a i and that would preclude estimation. There-
fore, when the goal of the study is to predict the value
Vit ¼ ai þ bXit þ cYit þ dZit þ eit ; ð2Þ
of one site given a series of characteristics, Brouwer’s
model (1) should be estimated using OLS. This is
where Vit is the value for site i in circumstance t. A
because biases in the estimated coefficients are not
circumstance can refer to a different point in time (a
important for prediction. Of course, the estimated
different year), or to some hypothetical situation (for
Table 5.1
Average value per visit to a beach (Euro of 2001)
Source Country Current state Eroded Defended Value concept
Average of available data UK 17.7 9.1 20.6 VOE
US 23.1 – – WTP for use or
consumer surplus
Penning-Rowsell et al. (1992) bgenericQ beach UK 15.6 8.2 18.7 VOE
US National Oceanic and Atmospheric Administration US 13.9 – – WTP for use
(informal communication, 1995) bstandardQ beach
Loomis and Crespi (1999) US 22.4 – – WTP for use
836 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
Table 5.2 Consumer Surplus, in the second one (Table 5.3) it
Panel data model bWTPQ is WTP, the dummy indicating the WTP for use.
Variable Coefficient P-value The first thing to remark from these tables is that
they are quite similar with the exception of the inter-
Intercept 19.38 0.002
T 0.22 0.49 cept term. The intercept changes because of the two
different dummies (WTP or CS), this is reasonable
Category of visitor (default is bunspecifiedQ)
because these dummies indicate a change in the aver-
Day 4.70 0.22
age value of the site (the default is different), and
Local 1.55 0.69
hence of the intercept. The coefficients of the regres-
Stay 4.12 0.29
sors change little; this indicates that the effect of these
Concept of value (default is VOE or CS)
variables on the value is similar whatever the concept
À15.67
WTP 0
of value that is used. The effect of time (T) is statis-
tically negligible; that is, the value of sites for coastal
Quality of the site (default is bcurrentQ)
informal recreation has not changed noticeably
À8.37
Eroded 0
Defended 3.30 0.02 between 1975 and 1995 (in real terms since the
value is expressed in Euros of 2001). The effect of
coefficients of model (1) have no interpretation since the type of respondents (Local residents, Day visitors,
they are biased. Below the estimates of both models Stay visitors or Unspecified) is not statistically sig-
are presented and we show that the panel data model nificant either. The quality of the site (Current,
(2) is not as good a tool as model (1) when it comes to Defended, Eroded) is unquestionably very significant.
predict the value of one site. Finally, the high significance of the concept of
value used (VOE, WTP for use, Consumer Surplus)
5.1. Marginal effect: panel data results is worrisome. It is acceptable that different concepts
of value yield different values, but the problem is that
The date (T) of the study is a cardinal variable and different valuation methods and designs have been
is inserted in the regressions as a natural trend starting used for the different concepts. Therefore, we cannot
in 1975 (normalized to 1). The 4 categories of visitors tell whether the differences in value are genuine or are
(local residents, day visitors, stay visitors and unspe- led by the valuation method that has been used. If it is
cified type) are represented using three dichotomous the former, we would still have to decide which con-
variables (Local, Day, Stay), with the omitted category cept of value is more appropriate. If it is the latter,
(the default) being the unspecified type. The 3 cate- then benefit transfer of informal beach recreation is
gories of quality of the site (eroded, current quality, partially flawed since different methods lead to differ-
defended) are represented using two dichotomous
Table 5.3
variables (Eroded, Defended); the omitted category
Panel data model bCSQ
is the current quality. Finally, the concepts of value
Variable Coefficient P-value
(VOE, WTP for use, Consumer Surplus) have been
Intercept 10.22 0.08
represented by 2 dichotomous variables (WTP, CS),
T 0.22 0.48
the omitted category being VOE. It turns out that the
sum of WTP and CS is a vector of zeros and ones Category of visitor (default is bunspecifiedQ)
identical to the sum of certain site-specific constants; Day 6.26 0.11
this is a direct consequence of the fact that in most Local 3.12 0.42
Stay 5.67 0.14
cases the value of a site has been estimated using a
single concept of value. Therefore, one of these 2
Concept of value (default is VOE or WTP)
variables had to be removed to enable estimation CS 15.90 0
(otherwise, perfect collinearity impedes estimation),
but since the decision to remove is arbitrary, we pre- Quality of the site (default is bcurrentQ)
À8.32
Eroded 0
sent the 2 sets of results: in the first one (Table 5.2) the
Defended 3.30 0.01
variable removed is CS, the dummy indicating the
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 837
Table 5.4 the value concept variable is capturing some of the
OLS estimate of model (1) site or socio-economic characteristics because VOE is
Variables Coefficient Number only used in the UK and CS is mostly used in the US.
of cases
The average value is around 16 (of 2001) for the UK
À9.35
Intercept and 22 for the US sites.
T (1975 = 1, each year is one) 1.87
Country of study (default is UK) 79
5.2. Predicting values: ordinary least squares results
US 23.56 22
NL 1.39 5
As stated above, because of missing regressors, the
Type of site (default is bcoastal resortQ; 99
that is, an bequippedQ beach) OLS estimator of the coefficients of model (1) is
À10.94
Beach 5
generally biased and inconsistent. It is therefore not
À10.47
Dune 2
worth trying to correct for other possible estimation
Type of visitor (default is bunspecifiedQ) 60
problems. Yet, when the goal of estimation is predic-
À7.82
Day 15
tion (that is, transfer), bias in the estimated coeffi-
À9.78
Local 16
À8.00
Stay 15 cients is of no importance. Whether the coefficients
Concept of value (default is VOE) 78
are biased or not, the OLS estimator minimizes the
À22.66
WTP 13
prediction error by construction.
À12.44
CS 15
The estimation results are described in Table 5.4.
bQualityQ of the beach 64
As explained above, there are more variables in model
(default is bcurrentQ state)
À9.27
Eroded 20 (1) than in the panel data model (2), but the OLS
Unspecified defense 2.95 14
estimator is inconsistent and therefore the estimates of
À1.47
Defended by nourishment 4
the coefficients are not meaningful. For that reason,
Defended by nourishment plus groynes 3.13 4
their significance is not shown. The overall fit
(adjusted R 2) of the model is about 40%, but it is
ent values for the same beach. For certain situations, unclear whether it can be taken as a good measure of
the researcher was imposed the method (e.g. in the fit in the current context.
UK, only a specific CV format was admissible for
claims of funding to the former Ministry of Agricul- 5.3. Benefit transfer
ture, Fisheries and Food), but in general, this suggests
a lack of standards in applying valuation methods to To transfer values to an entirely new site, that is,
beach recreation. On the other hand, it is possible that to predict the value of the new site, one would
Fig. 5.1. Benefit transfer cumulative distribution of prediction errors.
838 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
simply substitute the new site’s characteristics and we have preferred the simplest models. More details
the estimates of Table 5.4 into model (1). To estimate can be found in Polome (2002).
´
the size of the error that one could commit by
proceeding in this way, we have devised the follow-
ing exercise. For each site, we ran the panel data and 6. Conclusions
OLS regressions without that site’s observation(s)
and predict its value. Then, to measure the gain of The contributions presented in this paper have
precision obtained by carrying a new study, we shown the important diversity of coastal values –
compared the predicted value(s) with the one(s) from informal enjoyment of a beach to heritage and
obtained from the original study(ies). The measure nonuse values – and have provided examples and
of prediction error is the proportion of deviation illustrations of estimation of these values. The focus
from the value reported for the site in absolute has been on the valuation of non market benefits.
term. In Fig. 5.1 below, the line referring to the In Section 2, the Penning-Rowsell et al. (1992)
baverageQ (triangles) represents the average-value value of enjoyment methodology has been adapted
prediction, that is, the value of one site is set equal for the valuation of four Italian beaches. These sur-
to the average value of all the other sites, regardless veys have shown mean values for informal recreation
of the sites characteristics. on a beach in its current state from o to o per
5 28
Fig. 5.1 reports the proportion (on the vertical visit. This is therefore of the same order of magnitude
axis) of predictions that falls below the error level as the US and UK beaches, even though there are
indicated on the horizontal axis. For example, the large variations across beaches, and some respondents
proportion of errors no larger than 40% in transfer- sometimes express very large values. The Italian sur-
ring a value is about 70% for OLS and 55% when veys have also shown that coastal visitors are sensitive
the prediction is the average of the values of the to the protection of coastal sites from erosion and
other sites. In other words, when transferring value flooding and that they are generally in favor of
using model (1) estimated by OLS (Table 5.4), there defense projects. The value of enjoyment may also
is a 70% chance of making an error of 40% or less, vary considerably accordingly with the season
and (approximately) a 90% chance of making an error (spring/summer or autumn/winter) or the type of visi-
of 100% or less. Fig. 5.1 is a truncation of the tor (resident or tourist).
complete plot since there is a non-zero probability The contingent valuation methodology can be used
of making an error larger than 200% (about 5% to value other, very different, types of coastal assets.
chance with OLS). In this paper, two cases were illustrated. The Venice
We say that model A predicts better than model B case study in Section 3 indicated values of an order
when the cumulative distribution of prediction errors between o and o per year per visitor for protecting
4 5
of model A is above that of model B. In that sense as Venice and its lagoon from erosion and recurrent
can be seen in Fig. 5.1, the panel data model with flooding using a complex defense scheme. A large
WTP dummy (Table 5.2) performs worse than the proportion of respondents expressed their certainty to
simple average of values. That does not undermine pay the amount elicited if actually asked, though some
the qualities of the panel data model (2) as an did not. This confirms the usefulness of a dcertainty
unbiased estimation of regression coefficients, but questionT after the valuation questions in order to
for prediction (that is, transfer) purposes, the best estimate the expected mean donation.
model is model (1) estimated by OLS. That is not to In the Normerven contingent valuation survey in
say that better estimators cannot be found, but we Section 4, we estimated a value function for increas-
would have to resort to more sophisticated econo- ing the number of seabird nesting areas in a Dutch
metric estimators (e.g. Tobit models to take into coastal province. It has been shown that even if the
account that values cannot be negative). Given the first area could have a rather large (purely non-use)
limitations of the data set on the one hand and, on the value (close to o per year for 10 years), subsequent
20
other hand, the fact that the transfer function should areas have strongly decreasing values. One implica-
be easily usable by non-economists field practitioners, tion of this result for the transfer of benefit is that
P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´ 839
replication of a defense scheme may not lead to a References
replication of its value.
In Section 5, a compilation of the existing (pub- Bateman, I.J., Willis, K.G., 1999. Valuing Environmental Prefer-
ences: Theory and Practice of the Contingent Valuation Method
lished and unpublished) evidence on coastal protec-
in the US, EU and Developing Countries. Oxford University
tion benefit estimates has made clear that it does not
Press, Oxford.
seem possible to estimate a function that would Bower, Blair T., Turner, R. Kerry, 1998. Characterising and analysis
provide the total value of a coastal defense scheme benefits from integrated coastal management (ICM). Ocean and
for any single site. Instead, previous studies have Coastal Management 38, 41 – 66.
Brouwer, R., 2000. Environmental value transfer: state of the art and
concentrated on specific types of benefits. In parti-
future prospects. Ecological Economics 32 (1), 137 – 152.
cular, it appears that informal beach recreation has
Champ, P.A., Bishop, R.C., Brown, T.C., McCollun, D.W., 1997.
been studied more than any other type of activity at Using donation mechanisms to value nonuse benefits from
the beach, but even in this case most of the esti- public good. Journal of Environmental Economics and Manage-
mated values come from either the US or the UK. ment 33, 156 – 162.
Cordes, Joseph J., Yezer, Anthony M.J., 1998. In harm’s way: does
The reader should be cautious when using the results
federal spending on beach enhancement and protection induce
of the benefit transfer function presented in Section
excessive development in coastal areas? Land Economics 74
2 outside those two countries. Even within these two (1), 128 – 145.
countries, the probability that the transferred value is Dorfman, J.H., Keeler, A.G., Kriesel, W., 1996. Valuing risk-redu-
within 50% of the value from an original study is cing interventions with hedonic models: the case of erosion
protection. Journal of Agricultural and Resource Economics
only 75% (see Fig. 5.1). The average value of
21 (1), 109 – 119.
informal recreation on a beach in its current state
Goodman, S.L., Seabrooke, W., Daniel, H.M., Jaffry, S.A., James,
in these two countries, all methodologies together, is H., 1996. Results of a contingent valuation study of non-use
around o (of 2001) per visit. This value is within
20 values of coastal resources. MAFF.
the bounds of the Italian case studies presented in Greene, W.H., 1993. Econometric Analysis, 2nd edition. Prentice-
Hall, Englewood Cliffs.
Section 2.
Haab, T., McConnell, K.E., 2002. 1999. Valuing environmental and
natural resources: the econometrics of non-market valuation.
Edward Elgar, Cheltenham, UK.
Acknowledgements Hanemann, W.M., 1984. Welfare evaluations in contingent valua-
tion experiments with discrete responses. American Journal of
Agricultural Economics 66, 332 – 341.
EU support through RTD project DELOS, con-
Hanemann, W.M., Kanninen, B., 1999. The statistical analysis of
tract EVK3-CT-2000-00041, is gratefully acknow-
discrete-response CV data. In: Bateman, I., Willis, K. (Eds.),
ledged. Thanks are due Prof. C. Green and Prof. Valuing Environmental Preferences. Oxford University Press.
W. M. Hanemann for communicating data for the Hanley, N., Splash, C.L., 1993. Cost–Benefit Analysis and the
benefit transfer section. For the Italian case studies, Environment. Edward Elgar, Aldershot.
Herriges, J.A., Kling, C.L. (Eds.), Valuing Recreation and the
thanks are due to A. Lamberti for the engineering
Environment. Edward Elgar, Cheltenham.
aspects of the defense projects, the City Council of
King, O.H., 1995. Estimating the value of marine resources: a
Ravenna, the City Council of Trieste and the Con- marine recreation case. Ocean and Coastal Management 27
sorzio Venezia Nuova for their support and the (1–2), 129 – 141.
material provided, A. Cazzola and R. Filippini for Lipton, D.W., Wellman, K., Sheifer, I.C., Weiher, R.L., 1995.
Economic Evaluation of Natural Resources — A Handbook
the questionnaire formatting, F. Galassi for her work
for Coastal Policymakers: Silver Spring, Md.: U.S. Dept. of
with Excel and SPSS, B. Zanuttigh for the Lido di
Commerce, National Oceanic and Atmospheric Administration,
Dante pictures and Venice composition, C. Barbanti Coastal Ocean Office. Resource available at http://www.mdsg.
for the interviews in Pellestrina Island and collabora- umd.edu/Extension/valuation/handbook.htm.
tion on the Venice composition, and L. Franco and P. Loomis, J., Crespi, J., 1999. Estimated effects of climate change on
selected outdoor activities in the United States. In: Mendelsohn,
Scaloni for the interviews in Ostia. For the Dutch
Robert, Neumann, James-E. (Eds.), The Impact of Climate
case study, thanks are due Mindert de Vries (Delft
Change on the United States Economy. Cambridge University
Hydraulics) and Jentsje van der Meer (Infram) for Press, Cambridge, pp. xi,331.
the biological and engineering aspects of the defense Marzetti Dall’Aste Brandolini, S. 2003. D28/A and D28/B–I of the
project. DELOS final report D28 bEconomic and Social Valuation about
840 P. Polome et al. / Coastal Engineering 52 (2005) 819–840
´
European Coastal Sites.Q Available on the DELOS webpage Polome, P. 2002. DELOS WP 4.1 Final Report (Benefit Transfer).
´
www.delos.unibo.it/. Available on the DELOS webpage www.delos.unibo.it/.
Marzetti Dall’Aste Brandolini, S., Lamberti, A., 2003. Economic Polome, P., van der Veen, A., Geurts, P.A.T.M., 2003. Contingent
´
and social valuation of the defence system of Venice and its valuation of a restored coastal natural area. Assessment of
lagoon (Italy). In: Ozhan, E. (Ed.), Proceedings of the Sixth Option Value and Non-Use Values — The Case-Studies of
International Conference on the Mediterranean Coastal Envi- Venice (Italy) and Normerven (The Netherlands), Report D28/
ronment, MEDCOAST 03, 7–11 October 2003, pp. 307 – 318. B-II of the DELOS Project.
Marzetti Dall’Aste Brandolini, Zanuttigh, B., 2003. Economic and RIKZ (National Institute for Coastal and Marine Management),
Social Valuation of Beach protection in Lido di Dante (Italy). In: 1999. Waddenzee Quality Status Rapport. Rapport RIKZ/
Ozhan, E. (Ed.), Proceedings of the Sixth International Con- 2000.008 (ISSN 0927-3980).
ference on the Mediterranean Coastal Environment, MED- Shaw, D., 1988. On-site samples’ regressions. Journal of Econo-
COAST 03, 7–11 October 2003. 319 – 330. metrics 37, 211 – 223.
Marzetti Dall’Aste Brandolini, S., Franco, L., Lamberti, A., Zanut- Shechter, M., Reiser, B., Zaitsev, N., 1998. Measuring passive use
tigh, B., 2003. bPreferences about Different Kinds of value. Environmental and Resources Economics 12, 457 – 478.
Low Crested Structures and Beach Materials: the Italian Case- Turner, R.K., Bateman, I., Brooke, J.S., 1992. Valuing the benefits
studies of Lido di Dante, Ostia and Pellestrina Island,Q D28/C of of coastal defence: a case study of the Adelburgh sea-defence
the DELOS project. Available on the DELOS webpage scheme. In: Coker, A., Richards, C. (Eds.), Valuing the Envi-
www.delos.unibo.it/. ronment: Economic Approaches to Environmental Valuation. J.
Mendelsohn, R., Neumann, J.E. (Eds.), 1999. The Impact of Cli- Wiley and Sons, New York.
mate Change on the United States Economy. Cambridge Uni- Zanuttigh, B., Martinelli, L., Lamberti, A., Moschella, P., Hawkins,
versity Press, Cambridge. S., Marzetti, S., Ceccherelli, V.U., 2005. Environmental Design
Penning-Rowsell,, Green, C.H., Thompson, P.M., Coker, A.M., of Coastal Defence In Lido di Dante, Italy, Coastal Engineering,
Tunstall, S.M., Richards, C., Parker, D.J., 1992. The Economics Special Issue on DELOS.
of Coastal Management: A Manual of Benefit Assessment
Techniques (Yellow Manual). Belhaven Press, London.