Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Transportation Research Part D 9 (2004) 341356

www.elsevier.com/locate/trd

Combining socio-acoustic and contingent valuation surveys


to value noise reduction
Thomas Bue Bjrner

*,1

AKF, Institute of Local Government Studies, Nyropsgade 37, DK-1602 Copenhagen V, Denmark

Abstract
A combination of a socio-acoustic survey on self-reported noise annoyance and a contingent valuation
questionnaire is used to estimate willingness to pay for noise reduction for urban residents living in
Copenhagen. It is found that the annoyance level has a signicant eect on the stated WTP. Expected WTP
per dB reduction is subsequently calculated by combining WTP for each annoyance level with the estimated
dose-response function for the relationship between noise exposure and annoyance. It is found that the
expected WTP for a one dB noise reduction is increasing with the noise level from e.g. 2 EUR at 55 dB to 10
EUR at 75 dB.
 2004 Elsevier Ltd. All rights reserved.
Keywords: Noise exposure and noise annoyance; Willingness to pay; Contingent valuation

1. Introduction
Dierent valuation methods have been used to estimate the external costs associated with road
trac. For trac noise the hedonic pricing method (HPM) is the most widely used approach, but a
number of studies have also relied on the contingent valuation (CV) method and other stated
preference methods, see e.g. Navrud (2002) and Bateman et al. (2001) for recent surveys of the
literature.
The main strength of the HPM is that it relies on actual behaviour observed in the housing
market, while the CV method relies on respondents statements about what they would pay for a
*

Tel.: +45-33-11-03-00; fax: +45-33-15-28-75.


E-mail address: tbb@dors.dk (T. Bue Bjrner).
1
Present address: Danish Economic Council, Adelgade 13, DK-1304 Copenhagen K, Denmark. Tel.: +45-33-13-5128; fax: +45-33-32-90-29.
1361-9209/$ - see front matter  2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.trd.2004.05.002

342

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

hypothetical reduction in noise. However, there are also some disadvantages associated with the
use of the HPM with respect to trac noise reductions. For example, trac noise is positively
correlated with other disturbances from trac, so it is likely that the HPM provides an upwardly
biased estimate of the value of noise reduction alone. In addition, according to theory (e.g.
Brookshire et al., 1982) the house price dierential calculated from the HPM only provides an
upper bound on the willingness to pay (WTP) for noise reductions. 2 These limitations of the
HPM suggest that it is worthwhile to explore further the potential of stated preference methods
for producing monetary valuations of the benet of noise reduction.
Of course there are also a number of methodological problems associated with the CV method.
In general, the method may produce too high an estimate of WTP, because of its hypothetical
nature. However, recent studies suggest that the bias is limited, when CV responses are interpreted
correctly (Vossler et al., 2003; Champ and Bishop, 2001). CV studies on noise reduction also face
the challenge of explaining reductions in noise levels in a way that is scientically correct and also
understandable to the respondent. Early CV studies used scenarios with specied reductions in
noise levels, typically a 50% reduction (Pommerehne, 1988). However, with this scenario it is
dicult to check whether the respondents understand what the reduction in noise would mean to
them (Navrud, 2002).
Recently, CV researchers have adopted methods developed by noise researchers, who for
several years have analysed the impact of noise on the experienced (self-reported) annoyance. This
line of research has developed into standardised methods for asking questions in socio-acoustic
surveys about level of annoyance using a ve-level annoyance scale (not at all annoyed, slightly
annoyed, moderately annoyed, very annoyed and extremely annoyed) specied in International
Organization for Standardization (2003). There also exists a large body of scientic evidence on
the exposureresponse relationship between noise and level of annoyance (measured as the
probability of being annoyed at a particular annoyance level). Recent examples include Klboe
(2003) and Miedema and Oudshoorn (2001).
The questionnaire applied in this study combines the socio-acoustic survey tradition with CV
questions on the WTP for removing the noise annoyance, following previous attempts by Lambert et al. (2001) and Navrud (2000). In the following we extend the method used in these previous
studies by linking the estimated WTP for removal of the noise annoyance with exposure
annoyance relationships. This allows us to calculate expected WTP per dB reduction at dierent
noise levels.
The results suggest that the combined socio-acoustic CV survey can provide reasonable estimates of the value of noise reduction. Thus, the expected annoyance level has a signicant
impact on the stated WTP, and the WTP also increases with the level of income. When WTP
results are combined with the exposureannoyance relationship we nd that the expected annual
WTP per dB reduction is increasing with the noise level from e.g. 2 EUR at 55 dB to 10 EUR at
75 dB.

That is the value that can be derived from the so-called rst stage of the hedonic pricing method. Earlier dierent
approaches were used to obtain the correct individual WTP in a second stage of the HPM, but these second-stage
methods have generally been abandoned due to methodological weaknesses, see e.g. Haab and McConnell (2002).

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

343

2. Survey instrument and data collection


The questionnaire consisted of four sections. In the rst a few simple demographic questions
were used as warm-up. The second section started with questions about the level of annoyance
from road trac noise. Subsequently, additional questions were asked about other disturbances
from trac (vibrations, safety, air pollution, dirt and dust) and noise annoyance from other noise
sources. These questions were included in order to remind the respondents that there are a number
of dierent disturbances associated with trac. In the valuation scenario, however, it was emphasised that these other disturbances would not be aected by the noise reduction. The third
section contained the hypothetical scenario and the WTP questions (see below). In the nal
section standard questions on the socioeconomic characteristics of the respondent and his/her
dwelling were included.
In the payment scenario the respondent was asked to consider the following:
. . . envisage that a new, eective, noise-reducing surface is laid on the street where you
live. The noise-reducing surface lowers the noise level so much that you are no longer
bothered, disturbed or annoyed by the noise. For instance, loud noises in connection
with loud acceleration and braking will be caught by the new surface made out of
noise-absorbing material. Other nuisances from trac such as air pollution, dust, dirt
or the risk of trac accidents are not aected. . .
The scenario further specied that the new surface would be more expensive to lay and
maintain than the standard road surface, and that part of these costs would have to be paid by the
households beneting from the noise reduction. The payment method was described as a xed
annual fee per household earmarked for the extra cost of removing the noise annoyance in the
street where the respondent lives. A standard budget reminder was included. To elicit the WTP an
open-ended question format was used. 3 In the past the dichotomous choice format has generally
been preferred, because it is thought to be easier to answer and because it is incentive compatible
in theory. However, the dichotomous choice format has also been suspected of providing upward
biased estimates of WTP due to yeah-saying. 4
With respect to the scenario, it is worth noting that Danish road administrators have been
testing a noise-reducing surface on a road section in Copenhagen and several participants in initial
focus group interviews had heard about this test. Thus, the idea of a technical solution that
reduces noise without inuencing other externalities is already familiar to at least part of the
general public in Copenhagen.
3

The open-ended WTP question followed after a payment card, included to ease the cognitive task of the respondent.
Similar combinations of payment card with an open-ended follow up have been applied by e.g. Dubourg et al. (1997)
and Ready et al. (2001). An English translation of the questionnaire can be found in Bjrner et al. (2003).
4
From a more pragmatic viewpoint it was necessary to obtain information on the WTP to eliminate noise annoyance
for each of the ve annoyance levels in order to make the results correspond to the exposureresponse functions. The
dichotomous choice format generally requires larger samples than open-ended format, and it would have been dicult
to obtain suciently large sample sizes for a dichotomous choice format for the high annoyance categories. Thus, only
7% of the respondents in the sample stated that they were extremely annoyed by road trac noise, even though
respondents exposed to high levels of trac noise were heavily over-sampled.

344

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

A sample of 2200 respondents from the Municipality of Copenhagen was selected. Respondents
living in areas with relatively high trac levels were oversampled in order to obtain a reasonable
number of respondents exposed to medium and high noise levels. Respondents living in privately
owned dwellings were also over-sampled in order to obtain a sucient number of responses from
owners to allow comparisons between owners and renters.
The questionnaire was mailed at the beginning of November 2002, with two reminders to
improve the response rate. In total 1149 questionnaires were returned, an overall response rate of
53% (not counting 30 questionnaires which were undeliverable due to changed or unknown addresses). Neither noise level nor type of ownership for the dwelling had an impact on the response
rate at a 5% signicance level.
Data on noise exposure were obtained from the Environmental Protection Agency of
Copenhagen. They calculated the noise exposure using the Nordic Prediction Method for trac
noise, where noise is described as a function of trac levels, speed and share of heavy trac. The
calculated noise at street level was subsequently combined with GIS data to calculate the distance
between the noise source and each dwelling. Noise exposure in dB is measured as LAeq24 . 5

3. Estimating the relationship between noise exposure and annoyance level


The valuation scenario is closely linked to the level of annoyance experienced by the respondent. To apply the WTP results it is therefore important to determine the exposureresponse
relationship between noise and annoyance.
The questions on noise annoyance used the new international verbal rating scale described in a
recent specication from International Organization for Standardization (2003). The annoyance
scale measures increasing level of annoyance in ve ordered categories (not at all annoyed, slightly
annoyed, moderately annoyed, very annoyed and extremely annoyed). To model the level of
annoyance one can apply an ordered probit or logit model, also sometimes denoted ordinal probit
or logit models. These models use data at the individual level, which makes it possible to control
for socioeconomic characteristics of the respondents, including characteristics of the dwelling. A
description of the ordered logit model focussing on noise and noise annoyance is provided by
Klboe (2003).
Dierent estimations of the ordered probit model have been carried out with noise as an
explanatory variable in combination with dierent sets of socioeconomic variables. Estimation
results are shown in Tables 1 and 2. 6 In Table 1 results are shown when noise enters the index
function in the simple linear way. In Table 2 an additional term with noise squared has been
added as explanatory variable.
Focusing rst on Table 1 it appears that the level of noise is signicant (at the 1% level) with the
expected positive sign. It also appears that the coecient on noise is very stable across the

LAeq24 is a single gure that reects the distribution of sound throughout the day using so-called A-weighted
decibels, which approximates the sensitivity of the human ear by weighting more heavily medium and high frequencies.
6
Estimation was carried out with LIMDEP7.

Table 1
Ordered probit for noise annoyance levels (noise entering linear)
Parameter
INTERCEPT
dB
AGE
AGE AGE
FEMALE
CHILDREN
CAR
LHHINCOM
OWNER
HOUSE
TERMOWIN
REDUHEAR
HOMEDAY
VERYSEN
SOMESEN
BEDRNOIS
Mu(1)
Mu(2)
Mu(3)
N
Log L
Pseudo R2

)4.921
0.063
0.064
)0.00069
0.111
)0.077
)0.064
0.006
)0.061
0.064
0.100
)0.104
0.079
0.603
0.341
0.259
1.224
1.924
2.568
1073
)1429.8
0.086

Model 2
Std. err.


0.585
0.004
0.013
0.00013
0.067
0.081
0.078
0.054
0.072
0.140
0.070
0.121
0.084
0.112
0.077
0.072
0.051
0.062
0.075

Parameter

Model 3
Std. err.

Parameter

Model 4
Std. err.

Parameter

Std. err.



)3.193
0.062

0.245
0.004

1.112
1.782
2.408

0.046
0.057
0.069

)4.421
0.062
0.066
)0.0007
0.116
)0.037
)0.052
)0.003
)0.048
0.005
0.071
)0.026
0.101

0.572
0.004
0.012
0.0001
0.067
0.081
0.077
0.053
0.071
0.141
0.070
0.119
0.083

)4.385
0.063
0.059
)0.0006

0.327
0.004
0.010
0.0001

1.188
1.873
2.511

0.049
0.060
0.074

1.136
1.813
2.447

0.047
0.057
0.070



1073
1451.9
0.072

1142
1560.1
0.071

1142
1579.8
0.059

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

Model 1

Note: Signicance levels at 1%, 5% and 10% are indicated by three, two and one asterisk(s), respectively.

345

346

Table 2
Ordered probit for noise annoyance levels (noise entering linear and squared)

INTERCEPT
dB
dB dB
AGE
AGE AGE
FEMALE
CHILDREN
CAR
LHHINCOM
OWNER
HOUSE
TERMOWIN
REDUHEAR
HOMEDAY
VERYSEN
SOMESEN
BEDRNOIS
Mu(1)
Mu(2)
Mu(3)
N
Log L
Pseudo R2

Model 2

Model 3

Model 4

Parameter

Std. err.

Parameter

Std. err.

Parameter

Std. err.

Parameter

Std. err.

1.386
)0.148
0.0018
0.062
)0.00069
0.122
)0.077
)0.081
)0.002
)0.053
0.073
0.119
)0.101
0.072
0.642
0.337
0.252
1.235
1.950
2.611

1.313
0.041
0.0003
0.013
0.00013
0.068
0.081
0.078
0.054
0.072
0.138
0.070
0.123
0.084
0.113
0.077
0.073
0.052
0.063
0.078

1.580
)0.140
0.0017
0.064
)0.00072
0.128
)0.038
)0.068
)0.012
)0.039
0.012
0.087
)0.020
0.094

1.262
0.040
0.0003
0.012
0.00013
0.067
0.080
0.077
0.053
0.071
0.139
0.070
0.120
0.083

1.366
)0.131
0.0016
0.057
)0.00063

1.162
0.038
0.0003
0.010
0.00010

2.473
)0.131
0.0016

1.121
0.038
0.0003

1.120
1.801
2.442

0.046
0.057
0.071

1073
1418.4
0.093

0.050
0.061
0.075

1.198
1.896
2.549
1073
1441.4
0.079

0.047
0.058
0.072

1.145
1.832
2.481
1142
1549.9
0.077

Note: Signicance levels at 1% 5% and 10% are indicated by three, two and one asterisk(s), respectively.

1142
1569.5
0.065

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

Model 1

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

347

dierent models, i.e. inclusion/omission of socioeconomic variables does not aect the size of the
positive coecient on noise. 7 A similar result is noted in Klboe (2003).
Few of the socioeconomic variables in the rst model in Table 1 are signicant, but age is an
exception (when entering linear and squared). Given the coecients in model 1 the impact of age
on noise annoyance increases up to the age of 46 after which it declines. Socioeconomic variables
not signicant (at 5% level) include gender (FEMALE), presence of children in the household
(CHILDREN), car ownership (CAR), log of household income (LHHINCOM). Respondents
who own their dwelling (OWNER) as opposed to respondents living in rented or cooperative
housing do not have signicantly dierent levels of annoyance. The same applies for respondents
with reduced hearing and respondents living in dwellings with double/triple glass thermal windows (as opposed to single glass or ordinary double glass windows). Respondents indicating that
they are very or somewhat sensitive to noise (VERYSEN and SOMESEN) as well as respondents
indicating that their bedroom is facing a street with trac noise (BEDRNOISE) have signicantly
higher levels of annoyance. However, one could argue that the self-reported sensitivity variables
as well as BEDRNOISE can hardly be regarded as true independent variables, because respondents stating that they are not annoyed are likely to state later in the questionnaire in order to
demonstrate consistency that they are not noise sensitive. Thus, statements about noise annoyance
and noise sensitivity may be viewed as two sides of the same coin. In model 2 the self-reported
sensitivity variables as well as the variable indicating that the respondents bedroom is facing a
street with trac noise are omitted. Omitting these variables does not have any impact on the size
or signicance of the other variables.
In Table 2 a squared term for noise is included in the model, which is also signicant (1% level).
In this table the coecient on the linear noise term becomes negative. The eect of noise on
annoyance level is negative until a cut-o level and subsequently increasing. The cut-o level is 41
dB, which is probably below the noise levels relevant for determination of exposureresponse
relationships for noise. The other conclusions related to Table 1 still apply to Table 2.
As already noted, age is the only socioeconomic variable that had a statistically signicant
impact on noise annoyance. The estimates suggest that annoyance is highest (other things being
equal) at the age of 46. To illustrate the impact of age the predicted probabilities of being (at least)
very annoyed and (at least) slightly annoyed are illustrated in Fig. 1 for a person aged 46 and a
person aged 23 (same probabilities for a person being 70 years old). The probabilities of the
person at the age of 23 (or 70) are shown as broken lines. The predictions are based on the
estimates from model 3 in Table 2 (where noise squared was included as explanatory variable,
too).
It appears from Fig. 1 that the impact of age is substantial. As an example the probability of
being very (or extremely) annoyed at a noise level of 65 dB is 0.21 for a person aged 46, while it is
only 0.12 for a 23 or 70 year old.
Even though age appears to be an important factor in the determination of noise levels it may
be dicult to include this information in predictions and project appraisals, because the age
distribution of the aected individuals may not be observed by the planner and in any case may
change over time as neighbourhoods change character. Therefore, exposureresponse relationship

This also applies to the marginal eects of noise.

348

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356


1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

Slightly+ (age 46)


Very+(age 46)
Slightly+ (age 23 or 70)
Very+ (age 23 or 70

40

45

50

55

60

65

70

75

dB

Fig. 1. Annoyance level and age.

1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0

Slightly
Moderately
Very
Extremely

40

45

50

55
60
dB (LAeq24)

65

70

75

Fig. 2. Actual and predicted annoyance levels.

has also been calculated when noise (and noise squared) are included as explanatory variables
alone without any socioeconomic characteristics (i.e. model 4 in Tables 1 and 2). 8
In Fig. 2 the predicted probabilities from the models in which noise enters linearly and as a
quadratic term (without taking age eects into account) are compared with the observed share of
annoyed respondents in dierent noise intervals. In the gure the solid lines represent the predicted probabilities from the quadratic model, the broken lines are from the model with noise
entering only linearly, while the individual marks are the observed shares in dierent intervals.
It appears from Fig. 2 that there are some dierences in the predicted probabilities. Both
models appear to give reasonable predictions of the observed shares of annoyance levels, though it
seems that the quadratic model (i.e. the solid lines) is slightly better at predicting the observed

In the current sample the age distributions are rather similar in the dierent noise intervals, indicating that it is not
important at least not in this case to take age dierences into account when performing predictions of aggregate
annoyance levels.

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356


100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

349

No response
Protest Zero
Legitim Zero

el

ry

m
tre
Ex

od
M

Ve

er
at

ly
ht
Sl

ig

N
ot

Positive

Fig. 3. WTP response by annoyance levels.

average shares. As the squared term was also signicant, this also implies that this model should
be preferred.
Miedema (2001) surveys the relationship between socioeconomic variables and annoyance
levels. Consistent with the above ndings he reports that age is signicant, with annoyance being
lower for young and old respondents than those in between, while there does not appear to be any
signicant dierence related to gender. However, in contradiction to the results in Tables 1 and 2,
he also reports that persons who own their own dwelling are more annoyed than persons who rent
their dwelling.

4. Analysis of WTP responses


Of the 1149 questionnaires returned 594 (52%) stated a positive WTP for removal of the noise
annoyance. With respect to the other respondents debrieng questions were included in order to
identify protest zero bids from genuine zero WTP. Genuine zero bids were obtained in 275 (24%)
cases, protest zero bids in 202 (18%) cases, while item non-response was obtained in 78 (7%) cases.
The most frequently used explanation for a legitimate zero bid, was that the respondent stated
that s/he was not suciently annoyed by the noise. With respect to protest bids a number of
dierent explanations were given (general objection against paying more tax, car drivers should
pay etc.). 9
The distribution of the four response categories according to annoyance levels is illustrated in
Fig. 3. As one would expect the share of positive bids increases with annoyance level, while the
share of genuine zero bids decreases. Closer inspection of the gure reveals that the share of
positive bids reaches its peak for the moderately annoyed, while it is lower for the very and extremely annoyed. However, one should take into account that the numbers of respondents in
these two categories are relatively small (132 and 83 individuals) and the dierences among the
shares of positive bids in the three highest annoyance categories are not signicantly dierent at
the 5% level according to a t-test.

A high share of zero bids (legitimate and protest) has also been found in previous CV studies on noise reduction, see
e.g., Lambert et al. (2001) and Vainio (1995, 2001).

350

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

It is also interesting to observe that the shares of protest bids appear roughly constant across the
annoyance levels (and are not, in fact, statistically dierent at the 5% level according to a t-test). A
priori there was concern that respondents exposed to high noise levels would be more sceptical
about the valuation scenario, because they might nd it unlikely that a noise-reducing surface
could eliminate the annoyance. However, Fig. 3 suggests that this was not a major problem.
Results from WTP estimations are reported in Table 3. In the model 1 only dummy variables
for each annoyance level are included as explanatory variables. These estimates correspond to the
simple mean WTP by annoyance level (in DKK per year). All estimates of the mean WTPs are
signicantly dierent from 0 at the 1% level. The 95% condence intervals for the WTP of each
annoyance level are also included. Though the condence intervals of some of the adjacent
annoyance intervals in model 1 do overlap, it also appears that there are signicant dierences in
the WTP between most non-neighbour annoyance levels. For example the estimated mean WTP
of the very annoyed is not signicantly dierent from the WTP of the moderately annoyed, but it
is signicantly higher than the WTP of the slightly and the not annoyed.
In models 2 and 3 in Table 3 dierent sets of socioeconomic characteristics of the respondents
are included. As expected the WTP increases with income. 10 Stated WTP is signicantly higher
when there are children in the household (CHILDREN), while the insignicant parameter on
SINGLE suggests that the WTP of one-adult households is not signicantly dierent from the
WTP of two-adult households (base case). The parameters on the dummy variables for high and
medium education levels suggest that the WTP increases with education (base case is short/no
education). With respect to the other variables neither age, living in a house (as opposed to
apartment), car ownership, membership of environmental organisations nor employment in the
public sector appeared to have a signicant eect on WTP. Finally, it is worth noting that the
WTP of respondents owning their dwelling was not signicantly dierent from respondents
renting. Estimates of the external costs of noise using hedonic pricing methods must necessarily
rely on the preferences of individuals living in privately-owned dwellings, 11 so to apply the results
of the hedonic studies in social cost-benet analyses planners must make the assumption that the
preferences of owners are similar to the preferences of renters. The insignicant parameter on the
owner dummy variable supports this assumption.
In preliminary regressions noise level was also included as an explanatory variable, but noise is
not signicant when included along with the annoyance level dummies. If the annoyance level
dummies are omitted then noise becomes signicant with the expected positive sign, but the
goodness of t statistics suggest that the model with annoyance level dummies included provides a
better description of the data.
The parameters estimated in models 2 and 3 in Table 3 can be used to calculate WTP at different annoyance levels for the average respondent. Such a calculated WTP is in principle superior
to the estimates of WTP from model 1 in Table 3, because it cleans out the (potential) dierences
in WTP that derive from dierences in the distribution of the socioeconomic characteristics

10
The estimated parameters on household income in models 2 and 3 (Table 3) correspond to an income elasticity of
0.72 and 0.78 at the average levels of income and WTP.
11
Assuming of course that there is rent control. If there is no rent control hedonic methods can also be applied with
the rent as endogenous variable (as it has been done in some cases).

Table 3
WTP estimation
Model 1

N
R2

Model 3

Std. err.

95% condence
interval

Parameter

Std. err.

331.3
628.1

86.5
81.4

161.8
468.6

500.9
787.6

)631.3
)381.5

203.1
179.1

1470.1

168.6

1139.6

1800.5

441.8

1918.2
2683.3

551.2
611.3

837.9
1485.2

2998.5
3881.4

Parameter

Std. err.
317.7
295.7

)1029.4
)732.4

)233.2
)30.5

)981.2
)692.4

233.7

)16.2

899.8

123.1

880.4
1725.8

472.8
619.4

)46.3
511.8

1807.1
2939.9

526.9
)308.9
2.5
724.0

283.3
197.2
0.5
243.5

)28.5
)695.5
1.5
246.6
)64.1

252.2

867
0.0661

95% condence
interval

161.3

829
0.1363

95% condence
interval
)1603.8
)1272.0

)358.6
)112.9

286.7

)438.9

685.0

619.1
1457.9

542.3
577.2

)443.8
326.6

1682.0
2589.3

1082.2
77.7
3.6
1201.3

509.0
)319.4
2.3
717.9

254.1
184.9
0.6
250.0

10.9
)681.9
1.2
227.8

1007.0
43.1
3.5
1208.0

568.4

215.0

162.1

)102.8

532.8

3.6
146.4
404.4
147.1
161.0
)213.7

4.0
153.3
698.1
144.8
161.5
238.8

)4.2
)154.1
)963.9
)136.7
)155.5
)681.7

11.3
446.9
1772.6
430.9
477.4
254.3

177.7

214.7

)243.2

598.5

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

Not annoyed
Slightly
annoyed
Moderately
annoyed
Very annoyed
Extremely
annoyed
Children
Female
Income 100a
High
education
Medium
education
Age
Single
House
Owner
Car
Environmental
member
Public sector

Model 2

Parameter

829
0.1425

Note: Signicance levels at 1%, 5% and 10% are indicated by three, two and one asterisk(s), respectively. Standard errors corrected for heteroskedasticity using the method of White (1978).
a
Means that the parameter is multiplied by 100.

351

352

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

between respondents with dierent annoyance levels. However, in this case the WTP at dierent
annoyance levels of the average respondent is very close to the WTP estimates given in model 1 in
Table 3.
The positive mean WTP for respondents not being annoyed by trac noise may perhaps be
seen as counterintuitive, but it should be recalled from Fig. 3 that less than one fourth of the notannoyed stated a positive WTP. The not-annoyed providing positive bids motivated this by
indicating that they would reduce noise due to concern for other members of their household or
by indicating a more altruistic motive for their positive WTP (e.g. reduce annoyance for other
inhabitants of the street or simply support a good cause). Although such expressions of altruism
may be a legitimate reason for stating a positive WTP, it has been debated whether it is reasonable
to use altruistically motivated WTP in social cost-benet analyses. As an example Milgrom (1993)
argues that there is a potential Pareto improvement ifand only ifthe project passes the benetcost test with the altruistic values excluded, because counting one persons WTP for anothers
happiness in a benet-cost calculation amounts to double (or triple or . . .) counting of the beneciarys benets. However, Johansson-Stenman (1998) and McConnell (1997) argue that this
applies only to some forms of altruism (i.e. pure altruism).
In Table 4 the WTP for noise reduction is compared with results obtained by Lambert et al.
(2001) and Navrud (2000), who also use scenarios with elimination of annoyance in the payment
scenario, combined with questions on annoyance levels. Lambert et al. conducted face-to-face
interviews with 331 dierent households living along highways and motorways in the RhoneAlpes Region in France. Open-ended WTP was elicited using a payment card. They found signicant dierences in the WTP for the dierent annoyance levels. It appears that similar levels of
WTP are obtained in the present study and in Lambert et al. for low annoyance levels, while the
WTPs obtained in this study are substantially larger than found by Lambert et al. at the higher
annoyance levels.
The study by Navrud (2000) involved interviews with 600 respondents in Oslo and Ullensaker
in Norway. Both dichotomous choice (DC) and open-ended (OE) formats were used to elicit the
WTP. For Oslo the estimated overall WTP for reducing annoyance was between 165-275 EUR
when the dichotomous format was used and 120 EUR when the open-ended format was used. The
overall WTP for Oslo corresponds well with the overall level of WTP at 135 EUR found in this
article. Both studies rely on a sample of respondents living in a large city. However, Navrud did
not nd any signicant dierences in the WTP for dierent annoyance levels.
The studies by Vainio (1995, 2001) were also carried out using respondents in an urban settingHelsinki in 1993and used a scenario which comes close to the elimination of noise
annoyance scenario. 12 For respondents exposed to at least 55 dB he obtains a mean WTP at 605
FIM per year corresponding to 72 EUR in 1993 prices or 81 EUR in 2001 (using increments in the
Finnish consumer prices based on Vainio (2001)), i.e. lower, but in the same range as the overall
mean obtained in this study. 13
12

Vainios scenario used the wording that trac volume would diminish considerably (the street would become a
residential street).
13
It is dicult to make strict comparisons of the overall WTP for removing annoyance, because average exposure
levels in the dierent samples may be dierent. The same applies with respect to socioeconomic characteristics such as
income.

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

353

Table 4
Comparing WTP per year per household from dierent studies
Not annoyed

Slightly
annoyed

Moderately
annoyed

Very annoyed

Extremely
annoyed

All

This study
WTP in EURa
Lambert et al.
(2001) WTP in
EUR
Navrud (2000)
WTP in EUR

45 [22; 67]

85 [63;106]

198 [153; 242]

257 [113; 404]

361 [200; 522]

135

47

61

78

101

130

73

Vainio
(1995, 2001)

120
(OE-WTP)
165275
(DC-WTP)
81 EUR in
2001

95% condence interval from estimation given in brackets. 1 EUR 7.43 DKK as on October 2002.

Summarising, it appears that the WTP bids obtained here seem reasonable. Thus, the WTP
increases with the level of annoyance as well as with the level of income of the respondents and the
estimated WTP for elimination of the noise annoyance is in the same range as found in previous
CV studies using similar scenarios.

5. Combining exposureannoyance and valuation responses


The mean WTP for each annoyance level can be combined with the exposureannoyance
relationship from Section 3 to calculate the expected WTP for a given reduction in noise. This is
described in Eq. (1) for a reduction in noise exposure from dBX to dBY for a given number of
households:
EWTP

5
X



X
Y
GdB
 GdB
 WTPk
k
k

k1
X
is the (expected) number of households
where k indicates annoyance level (from 1 to 5), GdB
k
being annoyed at level k at the dBX noise level, while WTPk is the mean WTP for each annoyance
level.
In Fig. 4 Eq. (1) has been applied to calculate the EWTP for one household of a one dB noise
reduction (at dierent initial noise levels).
The calculation of the EWTP is carried out using both exposureannoyance relationships
illustrated in Fig. 2, i.e. with noise entering both linearly and quadratically and with noise entering
the function only linearly. In addition, the EWTP is also calculated when the WTP of the notannoyed is given as either 331 DKK (about 44 EUR) or assumed to be 0 DKK/EUR instead.
In Fig. 4 the expected mean WTP illustrated in the rst (black) bar is based on the exposure
annoyance relationship, where noise also enters squared and where it is assumed that WTP of the
not-annoyed is equal to 331 DKK (about 44 EUR). In this base case the expected mean WTP per

354

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356


12.00
10.00
EUR

8.00
6.00
4.00
2.00
0.00

45

50

Base Case

55

60
65
70
75
Noise (dB)
NA -WTP = 0 Linear Linear + NA-WTP=0

Fig. 4. Expected WTP per dB reduction (EUR/house hold/year).

dB reduction exceeds 10 EUR per year per household at an initial noise level of 75 dB. However,
the mean WTP per dB decreases with reductions in initial noise level. At 60 dB the value of a dB
reduction is just below 4 EUR. At 45 dB the value is below 1 EUR.
The second bar in Fig. 4 is calculated subject to the assumption that the mean WTP for the notannoyed is equal to zero. It appears that the expected WTP is slightly higher for all initial dB
levels as compared with the base case. It may perhaps seem counterintuitive that a reduction in
one of the estimated WTP values (by annoyance level) yields an increase in the EWTP for a dB
reduction. However, Eq. (1) is basically the dierence in two weighted means of the WTP by
annoyance levels, where the probabilities of being annoyed at dierent exposure levels are the
weights. When noise is reduced the probability of not being annoyed is increased, so WTP for not
being annoyed attains a higher weight. It is the dierence in the mean WTP between the
annoyance levels (as opposed to the overall level of the mean WTP for the dierent annoyance
levels) that generates expected gains, when annoyance probabilities are altered.
The two last bars illustrate the mean WTP per dB reduction when exposureannoyance relationships are based on the model in which noise entered only linearly. In this case the expected
mean WTP per dB reduction is more constant across initial noise levels, i.e. higher at low initial
noise levels, but lower at higher initial noise levels (as compared with the base case). Thus, it
appears that the choice of functional form of the exposureannoyance relationship is not trivial
when combining the WTP for elimination of noise annoyance for dierent annoyance levels.
The results in Fig. 4 can be compared with results from a hedonic pricing study for Copenhagen
carried out as supporting evidence. The hedonic pricing study is described in more detail in
Bjrner et al. (2003), but the main result is that the price of dwellings is reduced by 0.5% per dB.
Converting the percentage reduction in dwelling prices into annualised values yields an average
value of noise reduction per dB (per dwelling per year) in the range of 12 to 20 EUR. This is
generally higher than the values from the contingent valuation study (Fig. 4), but this corresponds
with the theoretical observation by Brookshire et al. (1982) that house price dierentials calculated from the HPM provide an upper bound for the WTP for noise reduction (assuming both are
measured without bias). In addition, it also seems likely that the value of noise reduction obtained
from HPM may be upward biased as the house prices may also capture other external eects from
trac than just noise.

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

355

6. Conclusion
In this study, the tradition of noise researchers of using a so-called socio-acoustic survey to
obtain information on (self-reported) noise annoyance levels has been linked with traditional
contingent valuation questions on the WTP for avoiding the noise annoyance. Results of the
study suggest that there is a potential for combined socio-acoustic contingent valuation surveys.
First of all, it appears that plausible responses to the WTP questions were oered by the
respondents. Thus, the WTP for avoiding annoyance increased with the annoyance level of the
respondents and most of the not-annoyed respondents gave (genuine) zero bids. WTP also increased with the level of household income and with the presence of children in the household. As
an extension of the previous studies using the same approach, the WTP for avoiding noise
annoyance was subsequently combined with exposureannoyance relationships in order to calculate the expected WTP per dB reduction. The size of the expected WTP per dB reduction appears reasonable as compared with results from a hedonic pricing study carried out in the same
area.
However, it also appears that it is important for the estimate of the expected WTP per dB
reduction to obtain the correct functional form for the exposureannoyance relationship. When
using the preferred exposureannoyance relationship, the value of noise reduction depends on the
initial level of noise. At 55 dB the expected WTP per dB reduction was about 2 EUR per year,
while it was about 10 EUR at 75 dB.
Most studies on the value of noise reduction are based on property value data, because rent
control makes it unreasonable to apply data on rents. In order to apply these results in social costbenet analyses it is therefore necessary to assume that the preferences of owners and renters are
the same. Results of this study suggest that this assumption is reasonable as ownership did not
have signicant impact on neither annoyance nor WTP.
Acknowledgements
The research was supported by the Danish Environmental Research Programme. The article is
based on a research report co-authored by Jacob Kronbak and Thomas Lundhede (Bjrner et al.,
2003). The author is indebted to St
ale Navrud for several helpful suggestions. The author is also
grateful for valuable comments from Cliord S. Russell, Christian Hjorth-Andersen, Hans
Bendtsen and Lars Ellebjerg Larsen. Of course all errors and omissions are the sole responsibility
of the author.
References
Bateman, I.J., Day, B., Lake, I., Lovett, A.A., 2001. The Eect of Road Trac on Residential Property Values: A
Literature Review and Hedonic Pricing Study. Report (January 2001) to the Scottish Oce, Development
Department, Edinburgh, School of Environmental Sciences, University of East Anglia, Norwich.
Bjrner, T.B., Kronbak, J., Lundhede, T., 2003. Valuation of noise reductioncomparing results from hedonic pricing
and contingent valuation. SM publication no. 51 (research report), AKF Forlaget, Copenhagen.
Brookshire, D.S., Thayer, M.A., Schulze, W.D., Darge, R.C., 1982. Valuing public goods: a comparison of survey and
hedonic approaches. American Economic Review 72, 165177.

356

T. Bue Bjrner / Transportation Research Part D 9 (2004) 341356

Champ, P.A., Bishop, R.C., 2001. Donation payment mechanisms and contingent valuation: an empirical study of
hypothetical bias. Environmental and Resource Economics 19, 383402.
Dubourg, W.R., Jones-Lee, M.W., Loomes, G., 1997. Imprecise preferences and survey design in contingent valuation.
Economica 64, 681702.
Haab, T.C., McConnell, K.E., 2002. Valuing Environmental and Natural Resources. The Econometrics of Non-market
Valuation. Edward Elgar, Cheltenham.
International Organization for Standardization, 2003. Assessment of noise annoyance by means of social and socioacoustic surveys. ISO Technical Standard 155666: 2003. International Organization for Standardization, Geneva.
Johansson-Stenman, O., 1998. The importance of ethics in environmental economics with a focus on existence values.
Environmental and Resource Economics 11, 429442.
Klboe, R., 2003. Noise annoyance, stress and health eects. The Journal of Aviation Environment Research
Supplement Volume 7, 4851.
Lambert, J., Poisson, F., Champlovier, P., 2001. Valuing benets of a road trac noise abatement programme: a
contingent valuation study. INRETS-LTE, Bron, France. Paper presented at the 17th International Congress on
Acoustics, Rome.
McConnell, K.E., 1997. Does altruism undermine existence value? Journal of Environmental Economics and
Management 32, 2237.
Miedema, H.M.E., 2001. Noise & Health: How Does Noise Aect Us? Internoise 2001 Proceedings, The Hague.
Miedema, H.M.E., Oudshoorn, C.G.M., 2001. Annoyance from transportation noise: relationships with exposure
metrics DNL and DENL and their condence intervals. Environmental Health Perspectives 109, 409416.
Milgrom, P., 1993. Is sympathy an economic value? Philosophy, economics and the contingent valuation method. In:
Hausman, J.A. (Ed.), Contingent Valuation: A Critical Assessment. Elsevier, Amsterdam.
Navrud, S., 2000. Economic benets of a program to reduce transportation and community noiseA contingent
valuation survey. In: Proceedings of Internoise 2000, Nice.
Navrud, S., 2002. The State-of-the-Art on Economic Valuation of Noise. Department of Economics and Social
Sciences. Agricultural University of Norway.
Pommerehne, W.W., 1988. Measuring Environmental Benets: A Comparison of Hedonic Technique and Contingent
Valuation. In: Bos, D., Rose, D.M., Seidl, C. (Eds.), Welfare and Eciency in Public Economics. Springer, Berlin.
Ready, R.C., Navrud, S., Dubourg, W.R., 2001. How do respondents with uncertain willingness to pay answer
contingent valuation questions? Land Economics 77, 315326.
Vainio, M., 1995. Trac Noise and Air Pollution. Valuation of Externalities with Hedonic Price and Contingent
Valuation Methods. PhD dissertation. Helsinki School of Economics and Business Administration, Acta
Universitatis Oeconomicae Helsingiensis.
Vainio, M., 2001. Comparison of Hedonic Prices and Contingent Valuation Methods in Urban Trac Noise Context.
Paper presented at the 2001 International Congress and Exhibition on Noise Control Engineering, The Hague.
Vossler, C.A., Kerkvliet, J., Polasky, S., Gainutdinova, O., 2003. Externally validating contingent valuation: an openspace survey and referendum in Corvallis, Oregon. Journal of Economic Behavior and Organization 51, 261277.
White, H., 1978. A heteroskedasticity consistent covariance matrix and a direct test for heteroskedasticity.
Econometrica 46, 817838.

You might also like