Willingness to Pay for Reduced Visual Disamenities from Off-Shore Wind

Farms in Denmark

Jacob Ladenburg
*
& Alex Dubgaard
**

*
Ph.D.-student, Environmental Economics and Rural Development Division, Institute of Food and Resource Economics, The Royal
Veterinary and Agricultural University, Rolighedsvej 25, 1958 Frederiksberg C, Denmark, phone: +0045 3528 2273, jala@kvl.dk

**
Associate Professor and Research Director, Environmental Economics and Rural Development Division, Institute of Food and
Resource Economics, The Royal Veterinary and Agricultural University, Rolighedsvej 25, 1958 Frederiksberg C, Denmark, phone:
+0045 3528 2280, adu@kvl.dk


Abstract
Expansion of the off-shore wind power plays a significant role in the energy policies of many EU
countries. However, off-shore wind farms create visual disamenities. These disamenities can be
reduced by locating wind farms at larger distances from the coast and accepting higher costs per
kWh produced. In this paper willingness to pay for reducing the visual disamenities from future off-
shore wind farms is elicited using the economic valuation method Choice Experiments. The
valuation scenario comprises the location of 720 off-shore wind turbines (equivalent to 3600 MW)
in farms at distances equal to: 12 km, 18 km or 50 km from the shore, relative to an 8 km baseline.
Using a fixed effect logit model average willingness to pay amounts were estimated as: 46, 96 and
122 Euros/household/year for having the farms located at 12, 18 and 50 km from the coast as
opposed to 8 km. The results also reveal that WTP deviates significantly depending on the age of
respondents and their experiences with off-shore wind farms.

Keywords: Visual Disamenities, off-shore wind farms, choice experiments, fixed effect logit,
willingness to pay.

1 Introduction
During the last few years off-shore wind power generation has grown significantly and several
countries have great plans for future expansion. Wind power is a clean technology but it is not
without negative environmental impacts. The most significant externalities are visual disamenities
and noise nuisances. Off-shore location of wind turbines eliminates noise nuisances but visual
disamenities prevail. The visual disamenities can be reduced and even eliminated by moving
wind farms to larger distances from the coast. However, the costs per kWh produced increase as the
distance is augmented. Hence, the social planner is confronted with a trade-off between minimizing
the disamenities, on one hand, and accepting higher costs of power generation on the other. Welfare
economic appraisal of alternative wind farm locations requires monetisation of the un-priced
externalities.

Using the economic valuation method Choice Experiments the study presented in this paper has
estimated willingness to pay (WTP) for reducing visual disamenities from future off-shore wind
farms in Denmark. To the best of the authors knowledge this is the first attempt to investigate
preferences and WTP for reducing the externalities from off-shore wind farms.

In Denmark wind power has had a prominent position in energy policies for the past couple of
decades and Denmark was one of the first countries to establish large off-shore wind farms. This
provides a good basis for investigating preferences regarding the externalities from these
constructions. The valuation scenario comprised the location alternatives 12 km, 18 km and 50 km
from the shore relative to an 8 km baseline. The effect of extending the distance was visualized
using computer simulations. There was a fairly high positive WTP for having future wind farms
located at greater distances from the shore. However, there were significant differences in WTP
between age groups with younger people showing virtually no WTP for having wind farms located
further away from the cost. The paper describes the results of the valuation scenarios with particular
emphasis on differences in attitudes and WTP between demographic groups. We start with an
overview of previous valuation studies in the field of power generation.
2 Previous Studies on External Costs of Electricity Generation
Assessing the cost of the external effects from energy production initially came to the centre of
attention in the beginning of the 1980s (Schuman & Cavanagh, 1982). The number of studies
increased during the 1990s as a consequence of European policy makers growing concern about
external costs (European Commission, 1995, European Commission, 1999).

The external effects of wind power generation have also been investigated and various methods
have been used to elicit the monetary values of the externalities. Non-preference based methods -
primarily Abatement Cost and Damage Cost techniques - are the main approaches used in the
studies conducted by the (European Commission, 1995, European Commission, 1999). These
techniques are also used in other studies such as (Friedrich & Voss, 1993) and (El-Kordy et. al ,
2002).

However, revealed preference methods have also been applied (Jordal-Jrgensen, 1995, Sterzinger
et. al , 2003) as well as stated preference methods (Alvarez-Farizo & Hanley, 2002, Bergmann et.
al , , Ek, 2002, Navrud, 2004). The revealed preference studies used the house price method to
examine the relationship between property prices and wind turbines in the adjacent area. These
investigations did not reveal any statistically significant effects on house prices from adjacent wind
turbines (Sterzinger et al., 2003).

In contrast, the results from the stated preference studies indicate that there are significant external
costs associated with wind turbines. The studies and their results are not directly comparable.
However, they jointly verify that people have preferences for reducing different types of external
effects from wind turbines such as visual disamenities, reductions in wild life etc. These studies also
show that benefits - such as pollution reduction and the employment effects - add to the acceptance
of wind turbines. The studies, the valuated attributes of the wind turbines, and the associated WTPs
are presented in Table 1.

Table 1: Previous stated preference studies on environmental costs of wind power
Study Method Capacity
specified in
scenario
Scenario/Attributes Significant
WTP
WTP
Euro/year/
household
Navrud
(2004)
Contingent
Valuation
1.5-6.7 TWh Substituting wind power by
improving existing
hydropower
+ 110-130

Ek (2002)

Choice
Experiments

-

Location of turbines:
-Mountainous
-On-shore
-Off-shore
Noise impacts
Size of turbine
Grouping of turbines:
-individual
<10
10-50


+
+
+
-
-

+
+
+


0
12
29
-
-

10
20
0

Alvarez-
Farizo &
Hanley
(2002)

Contingent
Rating &
Choice
Experiment

-

Protection of:
Cliffs
Habitat & Flora
Landscape


+
+
+


22
38
37

Bergmann et
al. (In Press)

Choice
Experiment

-

Landscape impacts
Wildlife impacts
Air pollution
Employment benefits

-
+
+
- (+)

12
6
20


The studies in table 1 have different focal points in the description of wind power development and
the associated external costs. Navrud (2004) takes a holistic approach in estimating the overall
external cost of wind power development in Norway. In the study the external costs of 1.5-6.7 TWh
of wind power development are estimated by identifying the WTP for substituting wind power with
hydro power by improving efficiency of existing hydro power facilities. The WTP is estimated to
110-130 Euro/year/household for eliminating the environmental cost associated with an increase in
Norwegian wind power capacity by 1.5-6.7 TWh respectively
1
. Ek (2002), Alvarez-Farizo and
Hanley (2002) and Bergmann et al. decompose the external costs of wind power development into
attributes representing potential impacts of wind power development. Ek (2002) finds that Swedish

1
To estimate the environmental cost of wind power the respondents were asked how much they were willing to pay for
increasing the renewable energy capacity by a more efficient use of existing water generating power facilities (no
increase in environmental impacts) as opposed to wind power development
households are willing to pay 29 Euro/year for locating wind turbines off-shore and 12 Euro/year
for locating them in the low land as opposed to locating them in the mountains. The environmental
costs of off-shore location and location in the low land are therefore perceived as smaller than
development in the mountains. The same survey also finds preferences for medium size wind farms
(<10 turbine) as opposed to individual turbines or large wind farms (10-50 turbines). The size of the
turbines and the noise levels did not have a significant impact on the choice by the respondents.

Alvarez-Farizo and Hanley (2002) analysed the environmental cost of wind power development in
an important natural heritage area in Spain. The results show that the environmental cost expressed
through WTP is 21-38 Euros/year/household. WTP was elicited for protection of the attributes of
the natural heritage such as cliffs, flora and fauna, and the landscape.

Bergmann et al. include a potential increase in new local, permanent jobs in the valuation scenario.
In addition to expressing positive preferences for new jobs, the respondents were willing to pay 20
Euros/year for reducing air pollution, 12 Euros for reducing landscape impacts and 6 euros for
reducing impacts on wild life.

The three studies above verify that wind turbine development is associated with environmental
costs and that the costs depend strongly on the location of the wind turbines. Thus, the
environmental costs of wind power development can be reduced by selecting appropriate sites for
this development.

Of the above mentioned investigations Ek (2002) is the only study addressing off-shore location of
wind turbines. It was established that the external costs of wind power generation can be reduced by
locating wind turbines off-shore. However, Ek's valuation scenarios did not consider alternative
locations of off-shore wind farms with respect to the distance from the shore. To the best of our
knowledge the present study of distance-dependent disamenities from Danish off-shore wind farms
is the first to address the distance aspect explicitly.
3 The Survey
The valuation survey was based on a sample consisting of 700 randomly drawn respondents from
the Danish population between 20 and 65 years. Addresses were drawn from the Danish Civil
Registration Systems database. The survey mode was a mailed questionnaire. Each respondent
received a cover letter and an 11 pages questionnaire. 375 questionnaires were returned. Of these 13
were not completed satisfactorily. This leaves an effective sample of 362 respondents equivalent
to an effective response rate of 51.7%, which can be considered as good (Bateman et. al , 2002). In
the questionnaire each respondent was asked to evaluate three choice set consisting of two
alternatives. For more information on the survey see (Ladenburg et. al , 2005).
3.1 The Scenario
The scenario setting up the valuation experiments was based on an off-shore wind power
development plan from 1996. In this plan it was stipulated that 35% of the Danish electricity
consumption should come from wind power by 2030 (Danish Energy Authority, 1996). 4000 MW
was expected to be developed off-shore, which is equivalent to nearly 75% of the total wind power
expansion planned in the period. Given the present off-shore capacity of about 400 MW the
scenario entails a 3600 MW off-shore expansion. Taking the rapid development in wind turbine
technology into account 5 MW turbines (100 m high and 120 m wing span) were used in the
valuation scenario to give a realistic description of future development. Consequently, the scenario
entails the establishment of 720 (720*5=3600 MW) turbines off-shore. To minimise potential value
biases in the survey it was emphasised that the location of future off-shore wind farms would be
chosen in such a way that the impact on biodiversity and landscapes would be minimised.

Different payment vehicles were tested in a focus group, i.e. a lump sum payment per household
versus a payment depending on electricity consumption. The focus group considered a consumption
dependent payment (higher price per kWh) as most fair. However, the focus group acknowledged
the difficulties respondents would have remembering the annual household electricity bill. The
payment mode was therefore defined as a uniform annual surcharge (lump sum) on all households
electricity bill. Respondents were urged to be absolutely sure that their household would in fact be
willing to pay the amount specified in the chosen alternative. In addition a cheap talk reminder was
included stating that in previous surveys respondents have had a tendency to overstate their WTP
such as recommended in the literature (Cummings & Taylor, 1999, List, 2002). Finally a budget
reminder asked the respondents to consider their budget limitations when making their choices
(Arrow et. al , 1993).

The attributes in the choice sets are presented in Table 2 below.

Table 2: List of attributes defining the visual externalities of off-shore wind farms
Attributes Description/levels
Distance from the shore 8 km, 12 km, 18 km and 50 km
Number of turbines per wind farm 49, 100 and 144
Number of off-shore wind farms in
Denmark
2

5, 7 and 14
Annual cost (Euro)/household/year 0, 12.5, 23, 40, 80 and 175
3.1.1 Distance
The distances in Table 2 were set to illustrate possible future locations of wind farms relative to the
shore. 8 km from the shore is considered to be the minimum distances for futures wind farms
(Gaarde, 2003). 12 and 18 km from the shore are perceived as realistic distances and 50 km is the
distance where a wind farm consisting of 5 MW turbines cannot be seen from the shore (Nielsen,
2003).
3.1.2 Number of Turbines per Wind Farm and Number of Wind Farms
The number of turbines (49, 100 and 144) represents possible wind farm sizes. 49 turbines per farm
is less than the number of turbines in the existing off-shore wind farms at Horns Rev (80 turbines)
and Nysted (72 turbines). 100 and 144 turbines pr wind farm must therefore be considered relatively
large wind farms, but within the expected range of future wind farms (Madsen, 2005). After
counselling with wind farm developers, ELSAM and ENERGY E2, it was decided to work with
quadratic shaped wind farms in the visualisations. The number of turbines per wind farms is
therefore 7
2
, 10
2
and 12
2
. In the scenario the total number of turbines must sum up to 720 turbines.
Accordingly, the number of turbines per wind farm and the total number of farms are perfectly
(negatively) correlated.

2
The number of wind farms is perfectly correlated with the number of turbines per wind farm. Consequently, this
variable was not included as an attribute in the design of the survey.
3.1.3 Payment Vehicle
The costs/prices were set between 0 and 175 Euros/year assumed to be paid over the electricity bill.
The plausibility of levels and payment mode was tested during the focus group interview.
3.1.4 Design of Attributes and Choice Sets
The number of possible combinations given the attributes and the attribute levels are 3*4*6 =72. It
was decided to implement a fractional design of 36 alternatives. It would have been possible to do a
smaller main effect design, which would increase the representation of each alternative in the data.
However, it was believed that the visual impacts associated with the size and distance attributes
could be casually correlated. Therefore, it was important to control for to possible interaction effects
in the elicited choice model. The alternatives were generated in the SAS system by using the
macros and the design efficiency recommendations in (Kuhfeld, 2004). The initially generated
alternatives were afterwards blocked in choice sets of two and combined in groups of three choice
sets. To minimise both the number of dominating alternatives and non causal alternatives the
swapping procedure presented in (Huber & Zwerina, 1996) was used to construct the final choice
sets. Based on discussions with the focus group and taking into account the policy frame of the
scenario describing the future wind power development in Denmark it was decided not to include
a status quo option in the generated choice sets.

The visual impacts of the generated alternatives were illustrated by a visualisation of each of the
alternatives. The visualisations were made by a specialised consulting company in computer based
visualisations, see appendix A. It must be emphasised that the generated illustrations represent a
view to the wind farms under almost perfectly clear weather conditions. On many days during a
year the view to off-shore wind farms will be more blurred due to changing weather conditions.
Consequently, the chosen presentation may have resulted in a tendency for respondents to overrate
the actual disamenities from off-shore wind farms.
4 The Econometric Model
The choices between alternative off-shore development plans in the choice sets are modelled in a
random utility framework. In the model the utility associated with an alternative is represented by a
systematic component and random component taking into account the unobserved utility of the
alternative.

U
na
= V
na
+
na

U
na
is the utility which respondent n associates with alternative a. V
na
is the systematic component,
which is a function of the characteristics/attributes of the off-shore development plan and the
characteristics of the respondents.
na
is the error component. Thus, both V and are alternative and
individual specific.

In a choice set consisting of two alternatives a and b, alternative a is chosen if and only if the utility
associated with alternative a is larger than the utility of alternative b, that is U
na
> U
nb
and vice
versa.

The probability that person n chooses alternative i from j can then be expressed as

P
ni
= Pr(V
na
+
na
> V
nb
+
nb
b a)

= Pr(
nb
-
na
< V
na
- V
nb
b a)

That is, the difference in the observed utility of alternative a and b is larger than the difference in
the unobserved utility of alternative b and a. The difference in the observed utility V
nab
between
two alternatives is defined as

V
nab
= x
nab


where x
nab
represents the difference in the characteristics between alternatives a and b (x
na
-x
nb
).

Defining t as the number of the choice set evaluated by the respondent (t = 1, 2 and 3 in the present
case) the difference in the observed utility is equal to:

V
nt
=x
nt


Assuming a linear in parameters conditional utility function the systematic component can be
expressed by (Maddala, 1983):

V
nt
= x
nt

+
Z
n


Where x
nj
is the difference in the attributes of the off-shore wind farms and z
n
is the characteristics
of the respondent
3
. In general only a limited number of the parameters entailed in Z
n
are observable
by the analyst and can be controlled for in the model. The unobserved individual effect is captured
by the error term. Given that t =1, 2 and 3 - and assuming that the individual fixed effect is time
invariant - and the error term is logistically distributed - the unobserved individual effects can be
controlled for in a fixed effect binary logit model as described by (Chamberlain, 1980).

The indirect utility function is now

U
ni
= x
nt

+
Z
n
+
n
+
nt

Where
n
is the individual effect and
nt
is the error term, which is logistically distributed

The above model can be estimated by maximising the likelihood conditional on the
n
.
(Chamberlain, 1980)

Following (Hsiao, 2003) the joint probability function for the observed sequence of choices of
individual i conditional of the sum of individual choices (sum of 1 and 0s) can be expressed as


In the numerator the joint probability depends on the sum of the differences in characteristics of the
x
nt
choice set (and the individual) each weighted by the choice indicator y
nt
(y
nt
= 1 if alternative a is
chosen or 0 if alternative b is chosen).

=
=
=

=
T
t
B D
T
t
njt nt
T
t
nt nt
nt n
n nj
d x
y x
y y
1 ~
1
1
] exp[
] exp[
) | Pr(

In the denominator

sums over the possible combinations of choice sequences

n nj
B D
~

3
In the model the Z
n
s are incorporated by interacting them with the x
ni
s
D
nj
= (d
nj1
,d
nj2
,,d
njT
| d
njt
= 1 or 0) conditional on . In s y
T
t
nt
=
=
1

=

T
t
njt nt
d x
1
] exp[ each j
combination is evaluated by the sum of the difference in characteristics x
nt
, each weighted by j
th

choice sequence choice indicator d
njt
(d
njt
= if alternative a is chosen or 0 if alternative b is chosen).
It is clear that individuals, where

0 or T, do not ad any information as the conditional

probability of the observed sequences is equal to 1, which is equal to the unobserved
n
= or -.
=
=
T
t
nt
y
1

Normalising () with the sequence D
n1
= (d
n11
.....d
n1T
) the conditional probability can be rewritten as

=
=
=
+

=
T
t
D B D
T
t
t n njt nt
T
t
t n nt nt
nt n
n n nj
d d x
d y x
y y
1
)
~
( 1
1
1
1
1
] ( exp[ 1
)] ( exp[
) | Pr(

which gives the following conditional log-likelihood function
Log L
*
=

= =

+
C n
T
t D B d
T
t
t n njt nt t n nt nt
n n j n
d d x d y x
1 )
~
( 1
1 1
1
) ( exp 1 log ) (

Where { }

= =
=
T
t
nt
T
t
nt
y T y n C
1 1
0 , |
From the conditional fixed effect logit model the parameters representing the characteristics of the
off-shore wind farms can be identified and the conditional utility function can be estimated. Let

V
nj
= P
P
+ T
T

be the conditional utility function, where
P
represents the marginal utility of income (the price) and

T
represents a vector of marginal utilities of the other attributes. Totally differentiating the
conditional utility function and setting equal to zero (dV/dx
nj
=0), the marginal rates of substitution
between the off-shore wind farm characteristics are derived.

dV =
P
dP +
T
dT

= 0 and rearranging


As cost per household is one of the characteristics the marginal rate of substitution can be
interpreted as the maximum amount the individual is willing to pay to achieve/avoid a change in
one of the other attributes.
P
T
dP
dT

=
5 Results
In the analysis of preferences for reducing visual disamenities three nested models are elicited. The
first model is an attribute only model (sample average model), which only includes variables
representing the attributes of the wind farms. In the second model it is tested if respondents, who
can see off-shore wind farms from either their house or summerhouse, have stronger preferences for
reducing the visual disamenities of off-shore wind farms (the NIMBY-effect). In the final model it
is tested if preferences for reducing the visual disamenities are age-dependent. The different models
are presented in Table 3.

Table 3: Fixed effect logit models
1 Model 2 Model 3 Model
Param
t-value
aram
t-value
aram
t-value
eter
Estimate
P eter
Estimate
P eter
Estimate
DIST12 0.45502 (3.13) 0.41846 (2.79) 0.56869 (3.44)
DIST18 0.95507 (7.15) 0.91577 (6.77) 1.10333 (7.35)
DIST50 1.20957 (7.63) 1.14810 (7.14) 1.32989 (7.32)
SIZEM -0.01271 (-0.12) -0.02186 (-0.20) -0.01002 (-0.09)
SIZEL -0.13827 (-1.31) -0.13311 (-1.25) -0.14156 (-1.31)
PRICE -0.00133 (-11.53) -0.00134 (-11.47) -0.00134 (-11.40)

_DIST

12
12

Rchi2 269.47 (6) 276.79 (9) 290.71 (12)

1.93420

2.07725 V (1.78) (1.86)
V_DIST18 2.83758 (1.96) 2.98136 (2.00)
V_DIST50 3.33982 (2.16) 3.52173 (2.22)
Age29_DIST -1.02725 (-2.50)
Age29_DIST18 -1.40698 (-3.59)
Age29_DIST50 -1.20133 (-3.09)

L
McFadden R
2
0.208 0.216 0.231

5.1 Model 1
f l contains only the variables representing the attributes of the alternatives. It can
s mentioned previously the wind farm size variables (number of turbines per farm) are perfectly
5.2 Model 2 - View Model
des (positive/neutral or negative) towards the existing off-shore wind

The irst mode
therefore be characterised as the average preferences of the sample. The distance to the shore
variables are dummy coded, referring to a base distance of 8 km from the shore. The coefficients
are all significant and have the expected signs. The price attribute is highly significant and negative
as expected.
DIST12
<
DIST18
<
DIST50
are significantly different from each other
4
. Hence, respondents
associate wind farms located at distances equal to 12, 18 and 50 km from the shore with higher
utility/lower disamenity costs than wind farms located at 8 km. Accordingly, respondents have
unambiguous preferences for reducing the visual externalities from off-shore wind farms.

A
(negatively) correlated with the attribute controlling for the total number of off-shore wind farms in
Denmark. Some respondents might prefer large wind farms so that the total number of wind farms
will be smaller. Other respondents might prefer smaller wind farms so that the visual externality per
wind farm is minimised, whereas the number of wind farms will be higher. The size variables
(SIZEM and SIZEL) are both negative, but not statistically significant. This indicates that overall
large size of wind farms is associated with a negative utility, but preferences among the respondents
are rather heterogeneous. Thus, it was not possible to identify any systematic differences between
the respondents with regard to whether they prefer small and a relatively large number of wind
farms as opposed to large and relatively few wind farms.
In Ladenburg (2006) the attitu
farms in Denmark are analysed using a probit model. The model revealed that attitudes depend on
whether respondents can see off-shore wind farms from their residence or summerhouse. In the
present sample 17 respondents (app. 5 % of the sample) stated that they can see off-shore wind

4

DIST18
=
DIST50
is rejected at 0.0503 level.
farms from either their residence or from their summerhouse. The preferences of these respondents
might be different from the rest of the respondents. More specifically it was tested in model 2 if

i i
DIST V u DIST u _ ><

where V_DIST represents the respondents who can see off-shore wind farms from their home or
summerhouse and i correspondents to one of the three distance dummies.

The estimated coefficients in Table 3 indicate that respondents, who can see off-shore wind farms
from their house or summer house, have strong preferences for reducing the visual disamenities
associated with off-shore wind farms
5
. Thus, the variables 0<
V_DIST12
<
V_DIST18
<
V_DIST50
.
V_DIST18

and
V_DIST50
are significant at the 0.05 level and
V_DIST12
at the 0.076 level, whereas the variables
are jointly significant at the 0.06 level. Since only 3*17 observations represent the three V_DIST
i

variables the significance levels are perceived as quite good. Still, given the relatively small number
of observation the estimated coefficients should be interpreted with some care.

In the overall sample the inclusion of the V_DIST
i
variables reduces the size of the coefficients of
the DIST
i
parameters with approximately 5-10%. In other words, the estimated WTP for mowing
future wind farms away from the coast is lower for respondents who cannot see off-shore wind
farms from their residence or summerhouse.

The question then is how to interpret the result that WTP seems to be higher for people having a
view to existing off-shore wind farms. As a possible explanation these respondents might be
concerned that future wind farms will be located in the same area as the existing ones. The affected
respondents would consequently have motivations for expressing strong preferences for mowing the
future wind farms to greater distances, so that they themselves would be less affected
6
. Another
explanation could be that the respondents in question have stronger preferences for the reducing
visual disamenities because they have more experience with the visual impacts from off-shore wind
farms. If this is the case the presence of off-shore wind farms seem to affect preference formation
notably in a negative way what the tolerance towards the externalities are concerned. Of course,
this would have significant policy implications in the sense that the expansion off-shore wind power
would generate negative feelings towards such a programme. More research could be needed to
uncover whether such mechanisms prevail.
5.3 Model 3- Age Model
In an analysis of attitudes towards wind power development in Denmark older peoples expressed
less positive feelings than younger respondents (Ladenburg, 2006). In model 3 it was tested if WTP
for reducing the visual disamenities of off-shore wind farms is age dependent. Various age and
distance interactions have been tested, such as interaction between ln(age) or age(linear) and the
distance parameter. Inclusion of the variable ln(age)*DIST
i
improves the model significantly
compared to the base model (no interactions). This indicates that the perception of the disamenities
of off-shore wind farms is a increasing function of the age of respondents the younger the

5
It would have been interesting to analyse if the preferences within the group were dependent on which of the existing
wind farms the respondent live next to as the visual impacts is quite different between the different wind farms
5
(see
SEAS 2000, ELSAM 2000). However, the number of observations is not large enough to support a differentiation of
the respondents.
6
However, the respondents have not exhibited dominant/lexicographical preferences for larger distances in the choice
sets.
respondent the less negative the perception of the visual impact. It turned out that a dummy variable
coding depicted this heterogeneity in preferences better. In the final model a dummy variable
(AGE29_DIST
i
), representing respondents younger than 30 years, was included. This model
specification is presented in row four in Table 3. The coefficients
AGE29_DISTi
are all negative and
significant indicating that respondents younger than 30 years have significantly weaker preferences
for reducing the visual disamenities from off-shore wind farms. In fact, the joint utility of reducing
the visual disamenities for respondents below 30 years (
DISTi
+
AGE29_DISTi
) is not significantly
different from 0. In other words, respondents younger than 30 years seem to be indifferent between
having the wind farms located at 8, 12 18 or 50 km
7
.

These results point towards a structural difference in respondents preferences around the age of
30
8
. From a policy point of view the observed relationship between the age of respondents and the
perception of the disamenities is interesting. It indicates that coming generations might be much
less sensitive to visual disamenities from off-shore wind farms than the present older generations. If
so, the disamenity costs will tend to decline over time, everything else equal. Again, this is an issue
which calls for further investigations.
5.4 Estimates of WTP
Based on the coefficients in Table 3 WTPs can be calculated for each of the models. In Figure 1 the
WTPs from model 3 are presented with model 1 (attribute model) as a benchmark.

-100,00
0,00
100,00
200,00
300,00
400,00
500,00
600,00
0 10 20 30 40 50 60
Distance from the coast (km)
W
T
P

h
o
u
s
e
h
o
l
d
/
y
e
a
r
WTP Attribute model
WTP Model 3, Age>29 years
WTP Model 3, V_DIST
WTP Model 3, Age<30 years

Figure 1: WTP for reducing the visual disamenities by mowing wind farms from 8 to 12, 18 and 50 km

In model 3 the WTPs for reducing the visual disamenities are a function of age and whether the
respondent can see off-shore wind farms from their home or summerhouse. Accordingly, the WTP
estimates are divided into three WTP functions each function representing one of the different
subgroups in the model. Starting with the (general) attribute model respondents are found to be

7
The interpretation is supported by eliciting a model for the respondents younger than 30 years only. In the model the
DIST12, DIST18 and DIST50 are all insignificant.
8
The data set do entail enough observations to verify if the cut of age is exactly 30 years or if preferences smoothly
change in the late twenties.
willing to pay 46, 96 and 122 Euros on average for having off-shore wind farms located at 12, 18
and 50 km respectively. Turning to the subgroups respondents who can see off-shore wind farms
from their home or summerhouse are willing to pay almost 4 times as much, specifically 264, 407
and 484 Euros for moving the future wind farms to 12, 18 and 50 km respectively. The WTP
function representing respondents who are older than 29 years (AGE>29 years) is quite identical to
the attribute model function. The WTPs are as explained a little larger. Respondents older than 29
years are willing to pay 57, 110 and 133 for reducing the visual disamenities by mowing the wind
farms to 12, 18 and 50 km respectively. The WTP function representing the respondents younger
than 30 years (AGE<30 years) is represented with WTP below zero for mowing the turbines to 12
and 18 km respectively, but as noted previously the joint utilities are not significantly different from
0. It would therefore be more correct to consider the WTP for reducing the visual disamenities of
respondents below 30 years to be zero.
5.5 Marginal WTP
In the survey, the reductions of visual disamenities are defined on a discrete scale (wind farms can
be located on 8, 12, 18 and 50 km). The decision criteria for determining the locations of future
wind farms are based on a more continuous scale. It would be relevant, therefore, to have the
preferences for reducing the visual disamenities of off-shore wind farm defined on a more useful
scale. Though not continuous, the observed WTPs exhibit certain levels of marginal properties
indicated by the flattening of the lines joining the WTP points in Figure 2.
The marginal properties of the preferences (WTP/km/household/year) can be explored by
estimating the difference in WTP between mowing the wind farms from 8 to 12, 12 to18 and 18 to
50 km from the coast divided with the difference in km. The constructed marginal WTPs are
presented in Figure 2 below

-20,00
-10,00
0,00
10,00
20,00
30,00
40,00
50,00
60,00
70,00
0 10 20 30 40 50 60
Distance from the coast (km)
W
T
P
/
k
m
/
h
o
u
s
e
h
o
l
d
/
y
e
a
r
WTP/kmAttribute model
WTP/kmModel 3, Age>29 years
WTP/kmModel 3, V_DIST
WTP/kmModel 3, Age<30 years

Figure 2: Marginal WTP for reducing the visual disamenities

The marginal WTPs in Figure 2 illustrate the policy relevant aspects of the estimated WTPs in
figure 1. In the distinct intervals defined by the discrete levels of the distance attribute the marginal
WTPs drop from 15 euro/km (8-12km) to less than 1 euros/km (18-50 km) in the attribute model. A
similar drop in the marginal WTP can also be observed in the marginal WTPs calculated for the
different subgroups. In this relation, it is interesting that the marginal WTP for moving wind farms
from 18 to 50 is less than 2.5 euro/km in the model for respondents who can see off-shore wind
farms from their residence or summer house. This indicates that regardless of the subgroup, the
benefits associated with mowing wind farms further out than 18 km are small.
6 Conclusion
Off-shore wind power generation is expanding almost world wide. In spite of the fact that wind
power is a clean technology off-shore wind farms can cause visual disamenities. The present study
shows that in general people in Denmark are willing to pay for reducing visual disamenities from
future off-shore wind farms. The valuation scenario comprised the construction of 3600 MW of off-
shore wind turbines in farms located at either 12 km, 18 km or 50 km from the shore relative to an 8
km baseline. The results show that the Danish population in general holds significant preferences
for reducing the visual disamenities from off-shore wind farms. However, preferences vary across
subgroups in particular respondents who can see off-shore wind farms from their residence or
summerhouse and respondents younger than 30 years. Using a fixed effect logit model it was found
that the respondents in general are willing to pay 46, 96 and 122 Euros/household/year for having
future wind farms located at 12, 18 and 50 km from the coast as opposed to 8 km. The respondents
who can see off-shore wind farms from their residence or summer house were found to have very
strong preferences for reducing the visual disamenities. These respondents were willing to pay 264,
407 and 484 Euro/household/year for the above mentioned distances. Interestingly, respondents
younger than 30 years did not have significant preferences for reducing the visual disamenities.
Their WTP for locating off-shore wind farms at larger distances from the coast were close to zero.

Overall these results indicate that there will be considerable social benefits associated with
diminishing the visual disamenities from future off-shore wind farms. However, the results also
indicate that the marginal benefits (benefit/km) from extending the distance are decreasing.
Especially, the marginal benefits from locating wind farms at larger distances form the coast than
18 km are found to be very small. Assuming that marginal costs of power generation are increasing
as function of distance to the coast this indicates that the socially optimal location of off-shore wind
farms is unlikely to go beyond 18 km from the shore.

Finally, the deviations in preferences (experience and age) are worth putting forward, as their policy
implications justify further research.

References

Alvarez-Farizo, B. & Hanley, N. (2002): Using conjoint analysis to quantify public preferences
over the environmental impacts of wind farms. An example from Spain, Energy Policy, vol. 30, no.
2, pp. 107-116.
Arrow, K., R. Solow, P. R. Portney, E. E. Leamer, R. Radner, & H. Schuman (1993): Report of
the NOAA Panel on Contingent Valuation The National Ocean and Atmospheric Association's
Damage Assessment and Restoration Program (DARP), Washington, USA.
Bateman, I., R. T. Carson, B. Day, M. Hanemann, N. Hanley, T. Hett, M. Jones-Lee, G.
Loomes, S. Mourato, E. zdemiroglu, D. W. Pearce, R. Sugden, & J. Swanson (2002):
Economic Valuation with Stated Preference Techniques - A Manual Edward Elgar, Cheltenham,
UK.
Bergmann, A., Hanley, N., & Wright, R. Valuing the attributes of renewable energy investments,
Energy Policy, vol. In Press, Corrected Proof.
Chamberlain, G. (1980): Analysis of Covariance with Qualitative Data, Review of Economic
Studies, vol. 47, pp. 225-238.
Cummings, R. G. & Taylor, L. O. (1999): Unbiased Value Estimates for Environmental Goods: A
Cheap Talk Design for the Contingent Valuation Method, American Economic Review, vol. 89, no.
3, pp. 649-665.
Danish Energy Authority (1996): Energy 21, The Danish Government's Action Plan for Energy
Danish Energy Authority, Ministry of Environment and Energy.
Ek, K. (2002): Valuing the Environmental Impacts of Wind Power, a Choice Experiments
Approach.
El-Kordy, M. N., Badr, M. A., Abed, K. A., & Ibrahim, S. M. A. (2002): Economical evaluation
of electricity generation considering externalities, Renewable Energy, vol. 25, no. 2, pp. 317-328.
European Commission (1995): ExternE: Externalities of Energy Luxembourg: Office for Official
Publicationsof the European Communities..
European Commission (1999): ExternE: Externalities of Energy Luxembourg: Office for Official
Publicationsof the European Communities..
Friedrich, R. & Voss, A. (1993): External costs of electricity generation, Energy Policy, vol. 21,
no. 2, pp. 114-122.
Gaarde, J. (0303): Personal comment on feasible distances from the shore of future off-shore wind
farms. 0303.
Ref Type: Personal Communication
Hsiao, C. (2003): Analysis of Panel Data Cambridge University Press.
Huber, J. & Zwerina, K. (1996): The Importance of Utility Balance in Efficient Choice Designs,
Journal of Marketing Research, vol. 33, pp. 307-317.
Jordal-Jrgensen, J. (1995): [In Danish] Samfundsmssig vrdi af vindkraft, visuelle effekter og
stj af vindmller- kvantificering og vrdistning [Welfare Economic Value of Wind Power, Visual
and Noise Effects of Wind Turbines- Quantification and Valuation] Institute of Local Goverments
Studies- Denmark.
Kuhfeld, W. (2004): Marketing Research Methods in SAS. Experimental Design, Choice,
Conjointm and Graphical Techniques, [online]. SAS Institute Inc.,
<URL:http://support.sas.com/techsup/technote/ts694.pdf >.
Ladenburg, J. (2006): Attitudes towards Wind Power Development in Denmark, The Royal
Veterinary and Agricultural University, Food and Resource Economic Institute, Working Paper No.
3.
Ladenburg, J., Dubgaard, A., Martinsen, L., & Tranberg, J. (2005): Economic Valuation of the
Visual Externalities of Off-Shore Wind Farms, Report from the Food and Resource Economic
Institute, The Royal Veterinary and Agricultural University, Copenhagen, Report no.179.
List, J. A. (2002): Do explicit Warnings Eliminate the Hypothetical Bias in Elicitation Procedures?
Evidence from Field Auctions for Sportscards, The American Economic Review, vol. 91, no. 3, pp.
1498-1507.
Maddala, G. S. (1983): Limited-Dependent and Qualitative Variables in Econometrics Cambridge
University Press, Cambridge, MA.
Madsen, B. T. (2005): Emerging Offshore Wind Power Markets- Conditions, Quantitied and
Timetable BTM, Consult.
Navrud, S. (2004): M LJ KOSTNADER AV VINDKRAFT I NORGE Sammendragsrapport til
SAMRAM-programmet.Norges Forskningsrd.Notat .Institutt for konomi og
Ressursforvaltning.Universitetet for Milj- og Biovitenskap (UMB).
I
Nielsen, P. (2003): Off-shore Wind Energy Projects - Feasibility Study Guidelines.
Schuman, M. & R. Cavanagh (1982): A Model Conservation and Electric Power PLan for the
Pacific Northwest, Appendix 2: Environmental Cost Seattle: NCAC.
Sterzinger, G., F. Beck, & D. Kostiuk (2003): The Effect of Wind Power Development on Local
Property Values Renewable Energy Policy Project.


Appendix A: Example of Choice Set

Distance: 8 km, No. Turbines: 144, Number of wind farms: 5 and Cost: 40



Distance: 18 km, No. Turbines: 49, Number of wind farms: 15 and Cost: 80