Content uploaded by Mercedes Sánchez
Author content
All content in this area was uploaded by Mercedes Sánchez on Jan 08, 2014
Content may be subject to copyright.
1
How much are people willing to pay for silence?
Jesús Barreiro*
Mercedes Sánchez*
Montserrat Viladrich-Grau+1
* Business Management Department
Universidad Pública de Navarra
31006 Pamplona, Spain
e-mail: mersan@unavarra.es
+Economics Department
Universidad Pública de Navarra
31006 Pamplona, Spain.
e-mail: mviladrich@unavarra.es
This is the preprint of an article whose final and definitive form has been published in the
Applied Economics @ 2005 copyright Taylor and Francis. Applied Economics is available
online at http://www.informaworld.com/ with the URL
http://www.informaworld.com/openurl?genre=article&issn=0003-
6846&volume=37&issue=11&spage1233
Acknowledgements: The authors gratefully acknowledge the help of Diego Azqueta, Pere
Riera, Roberto Penin, José Orden, Teresa Garcia Marco and Ana I. Sanjuán. The authors
would also like to extend their special thanks to Joseph Cooper (ERS-USDA) for his kind
permission to use his one and one-half bound routine. Remaining errors are the responsibility
of the authors. This paper is partially funded by a grant from the Plan de Formación y de
Investigación y Desarrollo del Departamento de Educación del Gobierno de Navarra (1998),
also Jesús Barreiro would like to acknowledge the funding received by CICYT grant number
SEC96-0648, and Montserrat Viladrich-Grau by CICYT grant number SEC2000-0838.
1 Corresponding author.
2
How much are people willing to pay for silence?
ABSTRACT
Despite its major importance in the urban environment, the problem of noise has received little
attention from environmental economists. In this paper we estimate the value of a noise
reduction program in a Spanish city. The chosen technique is contingent valuation with a one
and one-half bound question format. Through our estimations we have found, first of all, that
urban residents generally value noise negatively and, specifically, that a Spanish household is
willing to pay approximately 4 euros per year per dB reduced. A further finding is that there is
no scope sensitivity effect; which means that households display a willingness to pay different
amounts for two different degrees of noise reduction.
Keywords: Urban environment; noise; contingent valuation; one and a half bound question
format.
3
1. Introduction
Noise, which continues to be one of the main environmental problems facing Europe, is
gaining in importance as a result of the rise in noise levels that comes about with increasing
economic activity. Noise is something to which we are exposed throughout life. It is exposure
to levels above 40 dB(A) that begins to influence our well being, while levels above 60 dB(A)
are considered detrimental to our health.1 Excessive levels of noise have both physiological,
and psychological consequences. The physiological effects include, for example, hearing
impairment, disturbed sleep, high blood pressure, stomach ulcers and other digestive
disorders. Among the psychological effects we can also mention greater levels of anxiety,
irritability and nervousness; it also influences social behaviour and cognitive development
(Bolaños and Ochoa, 1990; Guski, 1989). High noise levels have a negative impact not only
on health but also on other areas of life, and therefore give rise to economic consequences.
Exposure to high levels of noise decreases the ability to concentrate, increases the likelihood
of errors of perception, interferes with communication, and causes difficulties in the learning
process among children (Grandjean and Gilgen, 1976).2 Other economic consequences are
losses in property value and increased health expenditure. Noise affects not only urban areas
and human health, but also the natural environment.
Major sources of noise are road, air, and rail traffic, together with industry and
recreational activities. Road traffic is responsible for 32% of the European Union (EU)
population being exposed to noise levels of over 55dB(A) (EEA, 1999), a problem that is
becoming more serious, despite corrective measures and improvements in car and truck
design.3 Greater economic activity has brought about this increase in the level of road noise
and the trend is expected to prevail in the future. A 30% increase in passenger car
transportation is expected for the period (1995- 2010), this figure rises to 50% when referred
1 Noise level is determined by measuring the intensity of sound pressure levels in decibels (dB). Decibels are measured on a logarithmic
scale, ranging from 0 (human audibility threshold) to 130 (pain threshold). For most purposes however, this scale is weighted by the
frequency sensitivities of the human ear, know as A-weighting. The range for everyday noise on this scale goes from 45 to 115 dB(A). To
describe the impact of noise on humans, the so-called Equivalent Sound Pressure Level (Leq) is calculated, that is, the mean value of sound
intensity over time expressed in decibels.
2 The WHO (1993) has also demonstrated that the capacity of language acquisition in certain populations such as young children can
sharply decrease in environments with high levels of noise.
3 The EU noise standards have been lowered from above 90 dB(A) for heavy lorries and 80 dB(A) from passenger cars in 1972 to 80 and 74
dB(A) in 1996, respectively. These reductions in standards have been made possible by widespread application of low noise technology,
resulting in decreases in engine and exhaust noise, for example. However, the European Environmental Agency (EEA) points out that
4
to freight transportation4. Second in importance is airport and air-traffic noise, 3 million
people are exposed to aircraft noise exceeding 55dB(A). Even though noise exposure at major
European airports is unlikely to see any increase before 2010 -due to the phasing out of noisier
aircraft- noise exposure from regional airports is expected to increase beyond that date.
Another source of concern is the progressive trend to shift freight transport from day to night-
time. The development of high-speed trains will also lead to increases in the noise level
generated by this means of transport. Sources of noise linked to industry and recreational
activities are difficult to compile because of the wide variety involved, however, there is
strong evidence of an increase in citizen complaints about street noise and locally concentrated
sources such as sports arenas and discotheques.
The European Commission has reacted to these trends and already in its 5th
Environmental Action Programme stated that “no person should be exposed to noise levels
which endanger health and quality of life.”5 To attain this goal and correct the above-
mentioned tendencies several options are available. First, technological and engineering
actions such as low-noise product development, insulation for sound-proofing homes and the
development of low-noise tyres and surfaces are ways of reducing noise emission. Second, the
planning of land use can also help by separating incompatible functions and establishing quiet
areas or noise-abatement zones. Additionally, educational and informative measures, such as
increasing the noise awareness of the population by providing information either on the
number of complaints made, or on the increase in noise resulting from high-speed driving,
could also help to lower noise levels. Finally, legal action would also be required, in other
words, it would be necessary to fix limits or approve guidelines, after agreeing on criteria for
measuring noise, minimum requirements for the acoustic properties of dwellings, and the
enforcement of such regulations.
Any such abatement measures would almost certainly be costly and their
implementation would only be justified if the economic benefits of noise reduction are a
matter of importance to European citizens. Noise, in economic terms, is a negative externality
and a public “bad,” however, it is one of the pollution problems that has attracted least
nowadays the dominant source of road noise is caused by the friction of tyres on the roads at speeds above 40 and 50 km/hour.
4 See the "Environment in the European Union at the Turn of the Century" as reported in the EIONET Noise Newsletter, European
Environmental Agency, no. 2, September 1999.
5
attention among environmental economists. There have been no comprehensive studies to
evaluate the social cost of the different types of urban noise and the benefits to be gained from
its reduction. Most studies focus on the evaluation of the social cost of airport noise. More
specifically, a large proportion of noise reduction benefit studies focus on measuring the loss
in property value associated with aircraft noise using the hedonic price method (HPM).
Collins and Evans (1994) and Yamaguchi (1996) applied the HPM to study the loss in
property values associated with aircraft traffic noise produced by the Manchester and London
airports. In the United States, Levesque (1994), and O’Byrne (1985) have also applied this
methodology to a study of the economic impact of aircraft noise. Additionally, HPM is also
the methodology most widely used to study the values associated with road traffic noise
reductions, (Soguel (1994) and Renew (1996)).
Relatively few contingent valuation (CV) studies have been conducted to assess the
value of city noise reduction. One of the major reasons for this may be the difficulty involved
in constructing a good CV survey able to provide an accurate description of a reduction in
noise that would be easily understandable for respondents. Many of these studies,
Saelensminde (1999) and Pommerehne (1988) among others, ask respondents to value
percentage reductions in noise levels without checking to make sure that people understand
what such a reduction would mean to them. Recently, however, several studies have explored
new ways of describing noise reductions that are easier for respondents to understand. Vainio
(1995, 2001) has applied the contingent valuation method (CVM) in order to estimate the
willingness to pay (WTP) for reducing traffic-noise to such a considerable extent that “the
street would become a residential street”. Navrud (2000) and Lambert et al. (2001) describe
noise reduction in terms of annoyance and elicit the WTP for a noise reduction program that
will eliminate some determined levels of annoyance.
The main goal of this paper is to advance in this direction by using the contingent
valuation method to estimate the economic value of a reduction in noise. More specifically, we
estimate the economic value of a reduction in the volume of noise tolerated by the inhabitants
of a medium-sized Spanish city. We describe the change in noise reduction level by referring
to the noise levels endured by respondents at different times of day and on different days of the
5 See Dobris Assessment, p.5, Ch.16.
6
week. Furthermore, in applying the CVM to estimate the economic value of a reduction in
noise, we use the one and one-half-bound format recently proposed by Cooper and Hanemann
(1995) and by Cooper et al. (2002). The one and one-half-bound format is less likely to result
in follow-up response bias than the double bound and triple bound alternatives. Furthermore,
this method still retains much of the efficiency of the double bound alternative and requires
less information than either the double bound or triple bound models. Because of its novelty,
we outline a brief description of the OOHB one and one-half-bound methodology in the next
section.
The contingent valuation method, however, has not been exempt of criticism. A
second important goal of this paper is to assess the relevance of a major contingent valuation
caveat, i.e., its lack of sensitivity to different levels of provision of a public good. Therefore,
we test for the presence of any scope sensitivity effect in our study. In their 1992 paper,
Kahneman and Knetsch criticise the contingent valuation method, by showing that in certain
cases people attach the same value to different goods. To avoid this critique we test for the
presence of any scope sensitivity effect. We carry out a double test in which we first measure
noise objectively and compare the sums that respondent are willing to pay for different degrees
of noise reduction. We repeat this exercise by measuring noise subjectively. In both cases, we
reject the existence of any scope sensitivity effect.
The next section presents and briefly analyses the one and one-half-bound
methodology. In the third section, we discuss the survey design, question format and model
construct, and explain the data collection process. In the fourth and fifth sections, we present
the definition of variables used in the estimation and the representative summary statistics,
respectively. Section six presents the results from the one and one-half-bound estimated
models. In section seven the results for the scope sensitivity test are presented and then
discussed. In the last section we summarise the major conclusions of the study.
2. Methodology: One and one-half-bound question format
The basic assumptions underlying the CVM are - as Kristöm pointed out in his 1990
paper and thesis- that individuals have an approximate idea of the amount of money that they
7
are willing to pay to acquire the good under evaluation, and that individuals will report the true
value, given that the survey has been designed optimally. The application of this
methodology, however, can give rise to several problems that may cause valuation biases. In
this paper we dedicate special attention to reducing question format bias by applying the
methodology proposed by Cooper and Hanemann (1995) and Cooper et al. (2002).
These authors propose an alternative question setting to the classical single bound and
double bound question formats: the one and one-half-bound format. Bishop and Heberlein
presented the single format in 1979 in which each respondent is quoted a single monetary
amount and asked if he or she is willing to pay that amount. Years later, Hanemann, Loomis
and Kanninen (1991) introduced the double bound method where respondents are presented
first with a price as in the single bound case, but, having once responded, are quoted a
different price and again asked if they are willing to pay that second amount. Though the
double bound format provides more efficient coefficient estimates than those facilitated by the
single bound method, this does not mean that it is totally error-free. Carson et al. (1992),
Cameron and Quiggin (1994), McFadden and Leonard (1993), and Kanninen (1995) give
different reasons to explain how the responses to the second bid may be inconsistent with the
responses to the first. The main one of these is that respondents switch from a market setting
for the first bid to a bargaining setting for the second bid, making it difficult to compare the
responses to the two bids. Cooper and Hanemann (1995), and Cooper et al. (2002) present a
solution to this problem by devising a multiple bound method that is free of response bias to
the follow-up bid. Specifically, they construct the so-called OOHB format, a specification that
should significantly reduce the risk of the survey moving into a bargaining setting when the
interviewer proposes a follow-up bid.
OOHB question format assumes that there is uncertainty about the cost of providing
the good to be valued. The interviewer only knows an interval of variation for this cost, which
may range from a lower to an upper bound, called BIDL and BIDU, respectively, (i.e. BIDL <
BIDU). Application of this methodology involves the following steps: first, before being asked
to answer the questions that elicit willingness to pay, the respondent is informed about both
the lower and upper bounds, referred to as the limits of the expected cost of the environmental
good. Next, the interviewer randomly chooses one of these two points as the initial value at
8
which to elicit the respondent’s willingness to pay. Then, if BIDU is chosen and the
respondent says NO, the respondent is asked if he is willing to pay BIDL. Similarly, if BIDL
were the first value asked and the respondent says YES, then the respondent would be asked if
he is willing to pay BIDU. In the other two cases the elicitation process stops when the first
price proposed is BIDU and the respondent says YES, and if the first price proposed is BIDL
and the answer is NO.
Thus, the elicitation process can result in six sets of answers. If the lower-end bid
(BIDL) is randomly drawn as the starting bid, then the possible response alternatives are: no,
yes-no and yes-yes. If the upper-end bid (BIDU) is randomly drawn as the starting bid, the
possible response paths are: yes, no-yes and no-no. So that, the one and one-half-bound log-
likelihood function can be written as:
ln ( ) *ln ( ) *ln ( , ) *ln ( )
*ln ( ) *ln ( , ) *ln ( )
Ld BIDL d BIDU BIDL d BIDU
d BIDU d BIDU BIDL d BIDL
inn ii
ny yn iii
yy yy i
iyy ii
ny ny iii
nn nn i
i
N
θππ π
ππ π
=+++
++
=
∑
1
where, i are individuals; jl
π
, is the probability of the jlth response, where j can take two
values, Yes or No, depending on whether the respondent is willing to pay the initial sum
quoted in the elicitation question. Similarly, l will take value, Yes or No, depending on the
answer to the follow up value presented in the elicitation question (that is, j=Y or N, and l=Y
or N); finally djl is the binary indicator variable.
The probability
π
jl is found to be the interval between two bids, e.g.
)()()( BIDLWTPBIDUWTPBIDUWTPBIDL
yn <Λ−≤Λ=<≤=
ππ
, where (.)Λis the
cumulative distribution function. In our case we choose the logistic distribution function
)( x
β
Λ to carry out our estimations. Since nnn
ππ
=, nyyn
ππ
= and
ππ
yy y
=, the likelihood
function can be simplified to:
[] [ ]
[]
ln ( ) *ln ( ) *ln ( , )
*ln ( )
Ld d BIDL d d BIDU BIDL
dd BIDU
ininn nn ii
yn iny yn ii
iyy iyyy i
i
N
θππ
π
=∨+∨ +
∨
=
∑
1
9
Note that the OOHB estimation method presents each respondent with two bids, the
lower and upper bids, or BIDL, and BIDU, respectively. Therefore, we could obtain two
estimated parameters if these two bid vectors were to be used in the estimation process, as are
obtained in the double bound. However, only one vector is used as a BID independent variable
in the estimation, as is the case in a single bound model. The elements from which this BID
variable is derived are a combination of the BIDL and BIDU vectors; whether it is the lower
or the upper bid value that is included in the BID variable to estimate the model depends on
the respondent’s reply. Thus, if the BIDL were to be drawn first and the respondent were to
say YES to this first bid, and NO to the follow-up bid, only the low bid value would be
included in the BID independent variable used in the estimation. Additionally, the
probability associated with that (Yes,No) answer would be represented by:
)()()( BIDLWTPBIDUWTPBIDUWTPBIDL
yn <Λ−≤Λ=<≤=
ππ
. However, if the
answer were YES to both the first and second bids, the bid value included in the BID
independent variable would be the upper bound and the probability associated with it would be
)(1)( BIDUWTPWTPBIDU
yy ≤Λ−=∞<≤=
ππ
. Likewise for a (No, No) response,
where the probability associated with the response can be represented by
)( BIDLWTP
nn ≤Λ=
π
.
3. Survey design
The city selected for this study, Pamplona, is located in the northern part of Spain,
between the Pyrenees and the Cantabrian Sea. It can be considered, with respect to noise, to be
average among Spanish cities of its size (approximately 300,000 inhabitants). The acoustic
map of the city, drawn up in 1997, shows that 59 percent of the measurements taken were
above 65 db(A), that is the upper limit recommended by the World Health Organisation
(WHO), while the 75 dB (A) level, that considered harmful by the WHO, was reached in only
9 per cent of the cases, (Arana and Garcia 1990; Arana 1997). In this study the average noise
level was 67.1 dB(A).
The survey was conducted on the basis of telephone interviews, which were held
during December 1999. The city was divided into 14 neighbourhoods and the interviews were
distributed among them according to their population. The survey content was structured in
10
three sections: i) description of the good being valued, ii) explanation of the circumstances
under which the good will be provided and formulation of the questions designed to elicit
respondents’ willingness to pay, and iii) features of the respondents' personal profile.
The goal of the first set of questions is to focus the respondent’s attention on the good
to be valued considering their daily relationship with noise. We started by formulating
questions that would help people recall the everyday noise levels to which they were exposed.
Among other questions, we asked: i) what type of noise was more disturbing for the
respondent; ii) when noise was more of a nuisance, during the day or at night; and iii) what
type of noise was more disturbing at each time of day. Through this range of questions we
expected respondents to recall the noise levels to which they were usually exposed and,
therefore, to be able to understand the noise reduction proposed. In this first section of the
survey, we also included questions that would help us to rank the importance of urban noise
for residents in Pamplona. Thus, for example, we asked respondents to rank noise against
other urban problems, such as security on the streets and in the neighbourhood, and garbage
collection. In this way, noise is placed in the wider context of urban problems, in order to
avoid part-whole bias.
In the second section of the survey, we presented the features of the provision of the
good, “noise reduction” and elicited the amount that the respondents would be willing to pay
for a particular degree of noise reduction. In other words, we explained how a reduction in the
noise level would be provided, what the baseline level of provision would be, who would
provide this reduction, how it would be provided, and the method of payment. Note that,
although most of us are familiar with the steps involved in building a public garden, we are
less aware of the possible ways of achieving a reduction in noise levels. We therefore
described the three measures that the city council apply if the noise reduction program were
approved. The first of these involves conducting a noise control campaign; the second,
developing a program of surveillance that would include fines for infringement; while the third
would require covering street traffic lanes with noise absorbing asphalt. Likewise, we asked
respondents to value each of these measures independently (from highly effective to non-
effective). This valuation serves a double purpose: first, it serves as an indicator of which of
these policy measures is seen to be more useful, and second, it forces the respondent to think
11
about each measure.
Once the attention of the respondent is focused on the noise reduction problem, we
used examples to describe the likely implications of these three measures in terms of noise
reduction. We pointed out that such measures would have implications for both day-time and
night-time noise, and therefore, our description included examples of both day and night
reductions. Once the degree of noise reduction had been successfully explained to the
respondent, we presented the questions concerning valuation. We pointed out that the
measures envisaged by the city council are costly and that respondents will have to contribute
to finance them if they are finally approved. We informed them that a research team from the
Universidad Pública de Navarra had estimated the cost of such policies, and presented each
respondent with an estimated interval for those costs. The extreme values of this interval
coincide with the upper and lower bids that would later be presented to the respondent in the
elicitation question. In this formulation the cost of the good in question was placed in a
framework of uncertainty. The interviewer claimed to be uncertain about the exact cost of the
good, but assured the respondent that she knew it to lie somewhere within the interval defined
by the extreme values BIDL and BIDU. As a method of payment we chose to present
increases on city taxes, since we found this to be the least troublesome, because other city
services, such as garbage collection, are paid through city taxes. We then asked about each
individual’s WTP.
The values of the BIDL and BIDU ranged between 3.12 € and 62.5 €. These values
were chosen after conducting several experimental open format surveys, where we asked for
the maximum willingness to pay. Our bid choice aimed to cover the central 95% of the
observed WTP distribution. Three intervals of variation for the lower and upper bids were
chosen: i) 3.12 € and 21.87 €; ii) 12.5 € and 43.75 €; and iii) 25 € and 62.50 €. The questions
were asked in pesetas the old Spanish currency, (1€ =166 pesetas). The sample was therefore
divided into three sub-samples. The total sample size is of 600 observations, distributed as
shown in Table 1. Note that in order to set up these values we did not consider the real cost of
the program and therefore these were not real cost estimates.
TABLE 1 ABOUT HERE
12
Half of each sub-sample is presented, as a first value, with a BIDL (for example, in the
case of the second interval (12.5 €)) while the other half is presented with the corresponding
BIDU (i.e. (43.75 €)). If the BIDL is drawn and the respondent is willing to pay the value
specified, we then ask if he is willing to pay the corresponding BIDU. If the respondent is not
willing to pay the amount corresponding to the lower bid, we then enquire what is the most he
is willing to pay for the reduction in the noise level. Though this last question is not necessary
when applying OOHB methodology, we asked the open-ended question to enable us to
perform additional estimations and consistency tests. Also, in the event that respondents were
not willing to pay any positive amount, we asked for the reasons for their attitude, in order to
isolate protest zeros.
A similar process was carried out when we chose BIDU as our starting value. If
respondents answered negatively, they were then asked if they were willing to pay the BIDL
amount, if the answer was still negative, they were asked the maximum they were willing to
pay. In this case, as before, if they were not willing to pay any positive amount, they were
asked to give the reasons for their attitude. Those who answered affirmatively to the suggested
upper bound, were also asked for the maximum they would be willing to pay. Finally, in the
third section of the questionnaire we asked for the respondents’ personal data, such as age,
gender, and income level. To obtain complementary information we also asked respondents if
they had invested any amount in insulation for their homes.
As we pointed out in the introduction, one of the main goals of this study was to test
for any possible scope effect. We distinguished between two possible degrees of reduction in
day-time noise levels. Half of the sample was asked to value option 1 and the other half to
value option 2. Under the first alternative, we requested the respondent to value the reduction
in noise-level that takes place between “a working day during working hours and the same
day at 9:30 p.m.” The other half of the sample was asked to value the reduction in noise-level
that takes place between “a working day during working hours and a Sunday morning.” In
order to isolate the effect of the bid from the effect of the noise reduction option, half of the
respondents for each bid were faced with one option and the remaining half with the other.
The night-time noise level was maintained constant for the whole sample.
13
4. Definition of the variables used in the estimation
Recall that, when applying the OOHB question format, the dependent variable takes 6
different values depending on which bid was drawn first and on the respondent’s answer path.
The six possible paths were mentioned in the methodology section, three correspond to the
lower bid: no, yes-no, yes-yes; and three to the upper bid: yes, no-yes, no-no.
The variable BID, which refers to the first price offered to the respondent, is used as an
independent variable in the OOHB estimation. As we said before, if the low bid was drawn
first, for example (12.5 €), and the respondent said YES to this, and NO to the follow-up bid
(i.e. 43.75 €) the bid value considered in the BID vector was 12.5 €. However, if the answer
was YES to the first offer and YES to the second, the bid value included in the bid vector was
43.75 €. If, on the other hand, the upper bid was drawn first (i.e. 43.75 €) then the BID
variable took the higher value 43.75 € if the answer to the first question was YES. Similarly,
the variable took the value, 12.5 €, if the answers were NO to the first question and YES to the
second.
Several sociological variables were also defined. Noise is perhaps the most
educationally related environmental externality. We have found several previous studies to
show that the higher the level of education the greater the annoyance experienced from noise.
To estimate the effect of educational level on a person's willingness to pay for a degree of
noise reduction, we introduced the variable UNIVER that takes a value of 1 if the respondent
has a university degree and 0 otherwise. We expected this variable to be positive and
significant. We also distinguished between respondents who had invested in insulation for
their homes and those who had not. A person who has already invested in defensive measures
is more likely to be sensitive to noise. Therefore, we defined the dummy variable ISOLA that
takes a value of 1 if they have invested in insulation and 0 otherwise.
In addition, we also expected respondents who attach a high value to the noise
reduction program proposed by the city council to show more willingness to pay for it than
other citizens. Thus, we defined the dummy variable VALPRO that takes a value of 1 if the
14
program proposed is highly valued by the respondent and 0 otherwise. Specifically, the
variable VALPRO takes a value of 1 if, in the evaluation question, the respondent attributed a
value of 7 or more to the noise reduction program, and 0 otherwise6.
But this is not the only survey question we can use to measure respondents’ sensitivity
to noise. In addition, we asked respondents to declare the level of nuisance suffered from
different types of noise. Specifically, we asked respondents, through three different questions,
to assign a number between 0 and 10 to the level of disturbance suffered from noise, i) in the
city, ii) in their particular neighbourhood and iii) at home. Using the responses to these three
questions we defined the dummy variable SENSI that takes value 1 when the respondent is
“highly sensitive” to the nuisance caused by noise and 0 otherwise. In order to define what
we mean by a “highly sensitive” individual, we classified respondents in two groups using an
iterative 2-means cluster analysis (Malhorta, 1993). This methodology calculates two average
cluster means - one for the highly sensitive and one for the insensitive individuals in the
sample- and assigns observations to one group or the other depending on the difference
between their response and that average. This procedure works as follows: first, we select an
initial cluster centre, in our case we chose the extreme responses 0 and 10. Next, each
observation was assigned to the cluster with the closest cluster centre.7 Once all the
observations had been assigned to a group, a new mean for each group was calculated. These
means are the two new cluster centres. Then the difference between the new cluster centres
and the original ones were computed. If this difference is below a given number, the process
stops, otherwise the second step is repeated with the new means and the cluster centres are
updated. The process is repeated reiteratively until a minimum change or a maximum iteration
number is reached. The cluster centres for our two groups are given in Table 2. Therefore,
SENSI takes a value of 1 for those individuals whose answers to the three selected questions
are closer to the answers for the more sensitive group, and a value of 0 otherwise.
TABLE 2 ABOUT HERE
Although other sociological variables, such as level of income, age, and gender were
6 Respondents were asked to give a value between 0 (minimum) and 10 (maximum).
15
considered, none of them resulted significant in the relevant regressions. The INCOME
variable presents the usual problems in this type of studies, because, a large proportion of
respondents failed to answer this question. We were not surprised, therefore, by the lack of
significance of this variable. We also tried using an occupational variable as a proxy for the
income variable. In the survey respondents were asked about their profession, and grouped on
this basis, taking into account their average income level but the results continued to be
insignificant. Neither gender, nor age was significant. For gender we did not have any a priori
expectation, however, we expected middle-aged people to value the absence of noise more
highly. Young people usually either enjoy noisy environments or are not greatly disturbed by
them. On the other hand, old people’s hearing capacity is diminished and this could be the
reason why they are less bothered by high noise levels. Additionally, we attempted to discover
the characteristics of families that were willing to pay larger amounts of money for a noise
reduction. We asked, for example, whether there was any family member with a chronic
illness or disease. We expected this type of families to be more sensitive to excessive noise
levels, but this factor proved non-significant.
As we said before, we were also interested in testing for any possible scope effect and
therefore we divided the sample into two groups. Each group was asked to value a different
degree of noise reduction, Alternative 1, or Alternative 2. We introduced the variable OPTION
that takes a value of 0 if the respondent values the reduction in the level of noise that takes
place “between a working day during working hours and the same day at 9:30 p.m.” (i.e.
Alternative 1) and a value of 1 if he values the difference between “working hours and Sunday
mornings” (i.e. Alternative 2). If there is no scope sensitivity effect we expect the option that
represents the greater noise reduction to be more highly valued.
5. Some descriptive statistics
As mentioned, the total sample size was 600, comprising 43.1% men and 56.8%
women. We also obtained a representative sample of educational levels: 24.2% of the
population had only primary studies, 17.7% had finished basic education (8 years), 24.4% had
finished high school, and approximately 24% of the population had undergone university
7 The distance was measured with the Euclidean distance.
16
studies (either at bachelor or masters level). Most of the respondents were flat owners,
specifically 87.5% of the total sample were owners and only 11.8% were tenants.
Respondents showed less tolerance of night-time noise than of daytime noise. 53.8%
of the population emerged as finding night-time noise more disturbing, compared to the 38.5%
that find day-time noise more disturbing. Trash trucks were signalled as the origin of the most
disturbing noise during the night. With respect to day-time noise, 33.3% of the population
reported being disturbed by traffic noise. When respondents were asked which type of traffic
noise they considered more disturbing, 87.3% of them mentioned motorcycle noise, even
though there are 15 times more cars than motorcycles in Pamplona. Another cause of noise
during the daytime are the activities of the city works department.
When compared with other problems in the city, noise reduction was not considered a
priority. Neighbourhood security, dirt and litter and dog excrement were considered more
important problems. Nevertheless over 50% of the total sample gave noise a score of 5 or
over when asked if it was an important issue in their neighbourhood, the average score being
5.78. There is also consensus among respondents in considering that high levels of noise are
dangerous for health, most consider stress to be the main problem caused by noise. In general
95.3% of the population are happy with their neighbourhood and enjoy living there.
However, 227 people, that is 37.8% of the sample, have had insulation work carried
out in their homes to combat cold and/or noise. In order to distinguish between these two
reasons for insulating we also asked what the main reason for the investment had been. For
21.9% of those who had made the investment, the main reason was excessive noise level, for
39.3% it was to insulate from the cold and 34.8% of the population said for both reasons
equally. The types of investment made were most often the installation of either primary or
secondary double glazing. In only 20.7% of cases did the cost of the investment exceed 1,803
€.
Finally, we should mention that there were 188 zeros (31.3% of the sample). We
8 The other problems mentioned score above 6 on this same question.
17
follow the recommendations of the National Oceanic and Atmospheric Administration report
by Arrow et al. (1993) and differentiate between real zeros and protest zeros. Therefore, zero
answers had a follow-up question in order to detect whether these zeros were real or protest-
based. People who do not consider noise a major problem and, thus, would not be willing to
pay for a program aimed at reducing noise would give real zeros. Reasons we interpret as real
zero are the following: i) “I do not notice noise”, ii) “I do not consider it important” and iii)
“I’m not interested in noise reduction.” On the other hand protest zeros were interpreted when
the reasons given were: i) “I already pay enough in taxes to the local council” and ii) “Noise
cannot be reduced with these policies”. These reasons account for 133 protest answers.
6. One and one-half-bound estimations
We obtained the OOHB estimations using Cooper et al. (2002) routine. To perform
these estimations we have assumed a logistic distribution for the WTP.9 The results obtained
are presented in Table 3. In particular we present two estimations first for the full sample -
with protest zeros- and second for the sample excluding protest zeros. Sample size excludes 8
observations where no answers were obtained for the dichotomous choice valuation questions.
We expected the coefficient of the BID variable to be negative and significant, and thus to
show that the higher the price quoted, the lower the proportion of affirmative answers. This
result emerges to be true in the two estimated equations.
Furthermore, when attempting to determine whether a reduction in the level of noise
will increase respondents' well-being, it is of primary importance to compute the welfare
measures. For calculating these measures we estimate a spike model for the expected WTP
function following Cooper et al. (2002). We assume that there is some positive probability
that an individual may be indifferent to a given noise reduction and therefore, in this event, his
willingness to pay would be zero. We model indifference as being equivalent to a probability
mass or spike at zero. The point estimated for WTP mean is calculated by integrating this
density function between 0 and ∞.
9 We also fitted a probit model to this data set that produced very similar results.
18
In the full sample case, the mean unrestricted willingness to pay in the one and one-
half-bound model is 21.86 € per year and 26.41 € per year when protest zeros are excluded.
The corresponding confidence intervals were computed using the jack-knife method, with
1,000 repetitions in each case. Using this methodology it is possible to obtain an
approximation of the distribution of the WTP from the original CV data set. First, for a
sample of size N a new data set is generated by drawing N observations, with replacement,
from the original data set. Each observation of the original data set is assigned a 1/N
probability of being drawn. For each new data set new logit coefficients are estimated using
the maximum likelihood estimation method. Those estimated coefficients are used to
calculate the expected WTP. And this sequence of events is repeated T times to form the
empirical distribution for the expected WTP. The standard deviations from this distribution
are calculated and also the confidence intervals. The results of these estimations are also
presented in Table 3. The reported WTP corresponds to the point estimates. Clearly, in our
case, the expected willingness to pay is positive indicating that a reduction in the level of noise
will improve the well-being of the inhabitants of Pamplona.
TABLE 3 ABOUT HERE
Not only the price or bid would have an effect on the dependent variable, however. As
we said before, we expect other characteristics of the population to have an impact on the
WTP for some noise reduction, therefore, we introduce several explanatory variables in our
regression equation. Henceforth, we will only report the estimated results without protest
zeros. These results are presented in Table 4. The effective sample size for the models with
socio-economic variables is thus 419 and, for the models including the SENSI variable, the
sample is further reduced to 404 observations due to some item non-response on one or more
of the segmentation variables. As expected, the variable VALPRO presents a significant and
positive parameter estimate in all regressions, showing that respondents who attach a high
value to the noise control program proposed in the survey also give a greater proportion of
affirmative answers for such a noise reduction program than respondents who do not value the
program as highly. Further, the respondents who assign a high value to the reduction
measures proposed in the survey do not necessary coincide with those who are most disturbed
by noise. However, we also expect respondents who are most disturbed by noise to be more
19
willing to pay for some noise reduction. Therefore, we introduce the variable SENSI to test
this hypothesis. Recall that SENSI takes a value of one for persons highly sensitive to noise
and zero otherwise and, as expected, the estimated coefficient is positive in all equations. Note
that, even with the introduction of SENSI, the variable VALPRO remains significant.
Accordingly, and as expected, both coefficients are positive, indicating that a higher valuation
of the program and a high level of sensitivity to noise led to greater willingness to pay.
As we have mentioned before, we expected more highly educated people to be more
annoyed by noise and, therefore, to be willing to pay more. The estimated coefficient
corresponding to the variable UNIVER is positive and significant in all regressions, revealing
that our hypothesis was correct, people with a university degree show more willingness to pay
for a decrease in the level of noise to which they are exposed. This is frequently found in this
type of studies: the higher a person's level of formal education, the lower his/her tolerance of
unwanted sound. People with higher levels of education are usually employed in intellectual
work or jobs that demand greater concentration; they are therefore more likely to be disturbed
by the presence of noise than people in other types of work. Therefore, these individuals
show more willingness to pay for a reduction in the noise level than others who are less highly
educated. Furthermore, the variable UNIVER presents a positive correlation with income level
and agents with higher income levels usually demand a higher quality of life, which includes
less unwanted noise.
TABLE 4 ABOUT HERE
The variable ISOLA is never significant, which shows that families that have already
invested some of their income in defensive measures are not willing to pay any more than
other citizens for additional degrees of noise reduction. There are two possible explanations
for this. Even though households that have invested in defensive measures demonstrated a
lower tolerance for noise than other members of their community, once the investment is in
place, their level of disturbance is the same or lower than that of their neighbours, and they
therefore have no reason to pay more. On the other hand, these families may also consider that
they have already invested enough money in noise reduction devices. Other variables were
considered in the regressions, such as age, gender or income level. Since the estimated
coefficients of these variables were never significant in these estimated regressions, it was
20
decided that they should be dropped.
7. Test for scope sensitivity
We introduced the variable OPTION in the estimated regressions equation to test for a
“scope sensitivity effect.” Scope or scale effects arise when respondents do not distinguish
differences in the quantity of the good under valuation. Therefore, in order to avoid critiques
such as those made by Kahneman and Knetsch (1992) and Desvousges et al. (1993),10 we
tested to see whether people attach different values to different degrees of noise reduction.
Under Alternative 1, respondents are asked to value in monetary terms a decrease in the level
of noise similar to the reduction that takes place “between the working hours of a working day
and 9:30 p.m. of the same day.” Under Alternative 2, respondents value the reduction in the
level of noise that takes place “between the working hours of a working day and Sunday
morning.” We tested for the presence of this scope effect using the OOHB estimation
procedures. The single bound results were very similar, for brevity, only the OOHB results are
presented here.
The estimated results, which were presented in Table 4, showed that the sign of the
variable OPTION is positive and significant, which means that respondents value Alternative
2 more highly. These results are also confirmed in the OOHB estimations presented in Table
5, where we introduced OPTION as the only explanatory variable. In other words, the
inhabitants of Pamplona place a higher value on the reduction in noise level that takes place on
Sunday morning. The significance of this parameter showed that there is no scope sensitivity
effect, that is, different degrees of reduction in the noise level are valued differently.
Furthermore, when we estimate a separate regression for each Alternative, we see that
Alternative 2 is valued more highly than Alternative 1. This confirms our finding that there is
no scope sensitivity effect, in as much as the reduction in the level of noise that takes place on
Sunday morning is more highly valued. However, we need to know if whether the OPTION
variable presents the correct sign, that is, whether this result (WTPalt2 > WTPalt1) is consistent
with the real noise reduction. This result would be consistent if Alternative 2 represents a
10 For example, this last author reported no difference of the willingness with respect to the number of waterfowl that could be saved by a
21
greater reduction in the level of noise than Alternative 1.
TABLE 5 ABOUT HERE
Though noise can be objectively measured in decibels dB(A), the level of noise
disturbance suffered by the population does not usually coincide with the objective measure of
noise. That is, two noises of the same measured intensity (i.e. same level of dB(A)) may give
rise to different levels of disturbance, depending on the physiological characteristics of the
individuals, the kind of activity they have to carry out, or on other environmental or personal
factors. We have therefore considered two measures of noise, one is “subjective”, while the
other is “objective.”
We obtain the “subjective” measurement of noise by means of the questionnaire. In
the survey, we asked respondents to rank three moments in the day according to the level of
noise endured: i) working hours during working days, ii) night hours (i.e. 9:30 p.m.) on
working days, and iii) Sunday morning hours. The question was asked twice, first we asked
them to rank the three situations from the noisiest to the least noisy, and later on in the survey
we asked the contrary, that is, to rank the alternatives from the least to the most noisy. The
91.0% of the population ranked “working hours on working days” as the noisiest period of the
day. The results were similar for the second question, the rank order was the same and 85.6%
of the sample considered that Sunday morning hours were the least noisy time of the week.
That is, the biggest reduction in the level of noise would take place between the level
perceived during “working hours on working days and that perceived in Sunday morning
hours.” If noise is considered as a “bad” we should, however, expect the option that
represents the greatest noise reduction to be the most highly valued. That is, the variable
OPTION should present a positive and significant parameter estimate. According to the
subjective perception of noise nuisance described above, the sign of the estimated coefficient
corresponding to the variable OPTION is appropriate. That is, we would expect willingness to
pay for Alternative 2 to be greater. Therefore, the results obtained in Tables 4 and 5 assure,
first, that there is no scope sensitivity effect and that the OPTION variable presents the correct
sign according to the subjective perception of those inhabitants of Pamplona who are willing
protection programme.
22
to pay more for a greater reduction in the level of noise.
For “objective” measures of noise, we use the noise levels measured for the city of
Pamplona. Observations were carried out in a wide variety of city locations. Most locations
were measured for noise only during working hours, but in five neighbourhoods, a continuous
(24 hour) measure of noise was recorded. This continuous measurement enables us to
compare the different levels of noise in the same location during 24 hour periods. The
continuous measurements were taken for the neighbourhoods of: Casco Viejo, La Chantrea,
San Jorge, San Juan, and Segundo Ensanche. From these measurements we have obtained the
average level of noise in these neighbourhoods in the three periods considered in our survey.
First, the Sunday morning noise level was calculated by averaging the level of noise measured
between 10 a.m. and 1 p.m. To obtain the level of noise at 9:30 p.m., we calculated the
average of noise between 9 and 11 p.m.11 An average of the levels of noise during working
hours was also calculated to obtain the noise level during working hours on working days.
These results left us with two types of neighbourhoods i) those where the level of noise is
higher during working days at 9:30 p.m. than on a Sunday morning, i.e. Casco Viejo and San
Jorge, and ii) those where the level of noise is higher on Sunday mornings than at 9:30 p.m. on
a working day, i.e. La Chantrea, San Juan and Segundo Ensanche. These measurements are
presented in Table 6.
TABLE 6 ABOUT HERE
We expected the OPTION variable to be positive and significant for the first type of
neighbourhood, and negative and significant for the second type. To test these expectations we
estimated two regression equations, one for Casco Viejo and San Jorge (henceforth in the
tables CV and SJo) and a second for La Chantrea, San Juan and Segundo Ensanche
(henceforth in the tables CH, 2EN and SJ). The estimated results obtained are presented in
Table 7.12 Due to the scarcity of data, we include in these estimations only those explanatory
variables that were significant in all the previous regression equations. These tables show that
those neighbourhoods in which there is a more marked reduction in the level of noise on
11 We had measures of the level of noise corresponding to the 9 p.m. to 10 p.m. period, but these exhibited very high variability.
23
Sunday morning (Casco Viejo and San Jorge) are consistently willing to pay more for this
reduction than for the reduction that takes place at 9:30 p.m. In other words, willingness to pay
in these neighbourhoods is consistently different for the two options presented.
Therefore, we also reject the existence of any scale or scope effect also when taking
into account the “objective” measure of noise for the neighbourhoods of Casco Viejo and San
Jorge, where respondents show more willingness to pay for a greater reduction of noise in
objective terms. However, we cannot entirely rule out the presence of a scope sensitivity
effect if we take into account the objective measures of noise in the quieter neighbourhoods of
La Chantrea, San Juan and Segundo Ensanche. In these cases the coefficient of the variable
OPTION is non-significant, showing that the inhabitants of these neighbourhoods are not
willing to pay more for a greater reduction in the level of noise. This result can be justified,
however, because the general level of noise in these neighbourhoods is lower and there is less
difference between the noise levels at 9:30 p.m. on a weekday and on Sundays mornings than
there is in the more noisy neighbourhoods, as can be observed in Table 6. In other words, in
the neighbourhoods of San Juan and Segundo Ensanche, in particular, it could prove difficult
to distinguish between the levels of noise at 9:30 p.m. and on Sunday mornings.
Summarising, most of the tests performed show people to be generally willing to pay
more for a greater decrease in noise level and therefore these results enable us to rule out the
presence of any scope sensitivity effect. Further investigation is needed, however, to explain
the differences encountered when comparing “subjective” and “objective” measures of
disturbance from noise. Individuals are known to differ in their perception of various types of
transport noise, for example, for the same noise level of 60 db(A), approximately 15% of
people claim to be highly annoyed when this comes from aircraft noise, 10% when it is a
question of road noise only 5% when it is produced by a railway.13 From our study it emerges
that sensitivity can also differ between different days of the week, and that other factors (apart
from source) may interfere in people's perception of noise as a nuisance. In particular, we
would need to find out what psychological or sociological factors intervene in the fact that the
decreases in noise levels that take place “between working hours on a working day and
Sunday morning” are almost always ranked as being greater than those that take place
12 Owing to their small sample size, these two groups failed to provide consistent results and therefore no OOHB models are presented.
24
“between working hours on a working day and 9:30 p.m. of the same day” even though this is
not the case. As an initial approximation, we suggest that the sensation of calm and relaxation
associated with Sunday morning must clearly have an influence on results.
Additionally, we observe that respondents in the noisier neighbourhoods were willing
to pay more for a noise reduction than respondents in quieter neighbourhoods. The
neighbourhoods where the level of noise is higher both on working days at 9:30 p.m. and on
Sunday mornings, show a willingness to pay of approximately 28.16 €, while in the less noisy
neighbourhoods (La Chantrea, Segundo Ensanche and San Juan) willingness to pay stands at
23.87 €. In other words, at first sight, respondents in noisier neighbourhoods seem to be
willing to pay more for a reduction in noise than respondents in quieter neighbourhoods, even
though, paradoxically, the neighbourhoods of San Jorge and Casco Viejo are two of the city’s
low-income neighbourhoods. Recall that we have not introduced income as an independent
variable in the regression equation because we had a large number of non-responses that
would have greatly reduced the number of observations.
TABLE 7 ABOUT HERE
8. Concluding remarks
We have applied contingent valuation methodology to estimate the economic value of
a non-market good, i.e., a reduction in the level of noise in a Northern Spanish city. The
results clearly indicate that households display a positive willingness to pay for a reduction in
noise, which in turn shows that such a reduction will improve the well-being of the inhabitants
of Pamplona. In particular, we have estimated a mean WTP that varies between
approximately 26 and 29 € per household, per year. Our estimated household willingness to
pay falls close to the values estimated in other European studies. However, it must be kept in
mind that many methodological differences exist among these studies, each of which uses at
least a different estimation technique, question format, and/or vehicle of payment.
Our household willingness to pay (without protest zeros) represents 0.27% of total
13 See European Environmental Agency, 1999.
25
annual income, which is not much lower than the 0.32% reported in Vainio (1995 and 2001)
in his CV study of noise reduction for the city of Helsinki. Furthermore, our household
willingness to pay per dB, per year falls close to the 4 € mark. The Vainio study, in the
meantime, reports an approximate WTP per dB, household and year that oscillate between 6
and 9 €, while Garrod et al. (2001) and Scarpa et al. (2001) report a estimated mean WTP for
traffic calming measures in three English towns of about 2- 4 € per dB, per household, per
year. Lambert et al. (2001) focusing on the Rhone-Alpes region reported an estimated WTP
of 7 €. Other studies, however, have estimated a higher WTP, see, for example, Navrud
(2000), who, in his Norwegian study, reports a willingness to pay of between 21 and 34 € per
dB, per household, per year. Clearly, the discrepancies in the findings of these studies can be
explained, not only by the different methodologies used, but also by factors relating to
different income levels, institutional settings, cultures and preferences.
We see a further positive contribution of our study in the fact that we have shown there
to be no scope sensitivity effect. We employed two alternative means, one, a measure of
“subjective” perception, the other, a measure of “objective” perception of noise annoyance.
Summarising, our results show that there is no scope sensitivity effect and that the tests
performed therefore show that most people are willing to pay more for larger decreases in the
level of disturbance from noise. Further investigation is needed, however, to sort out the
differences encountered when comparing “subjective” and “objective” measures of noise. Our
study shows that the decrease in noise that takes place “between working hours on a working
day and a Sunday morning” are almost always ranked as being greater than that which takes
place “between the working hours of a working day and 9:30 p.m. of the same day”, even
though this is not actually the case. This difference between “subjective” and “objective”
ranking seems to suggest that people are less annoyed by exposure to noise on non-working
days. If this conclusion were proved to be true, it would imply that policy-makers should focus
their efforts on reducing noise exposure on working days.
26
References
- ARANA, M. (1997). Estudio y Evaluación de la Problemática del Ruido en Pamplona
y Poblaciones Colindantes. Fase I. Mimeo.
- ARANA, M. and GARCÍA, A. (1990). Estudio del ruido ambiental en Pamplona.
Revista de Acústica 1 y 2.
- ARROW, KENNETH J., R. SOLOW, P. R. PORTNEY, E.E. LEAMER, R. RADNER
and H SCHUMAN (1993). Report of the National Oceanic and Atmospheric
Administration Panel on Contingent Valuation, Federal Register 58 (10), 4016-614.
- BOLAÑOS, F., and J.M. OCHOA (1990). Medida y control del ruido, Ed. Boixareu,
Barcelona.
- CAMERON, T.A. and J. QUIGGIN (1994). Estimation using contingent valuation
data from a Dichotomous Choice with Follow-up Questionnaire, Journal of
Environmental Economics and Management, 27(3), p.218-234.
- CARSON, R., R. MITCHEL, M. HANEMANN, R. KOPP, S. PRESSER and P.
RUUD (1992). A Contingent Valuation Study of Lost Passive Used Values Resulting
from the Exxon Valdez Oil Spill, Report to the Attorney General of Alaska.
NRDA1nc, La Jolla CA.
- COLLINS, A., and A. EVANS, (1994). Aircraft noise and residential property values:
an artificial neural network approach. Journal of the Transport Economics and Policy,
May, 144-52.
- COOPER, J.C. and M. HANEMANN, (1995). Referendum contingent valuation: how
many bounds are enough? Working Paper, ERS-USDA, Food Consumption and
Economics Division, July.
- COOPER, J.C.; HANEMAN, M. and SIGNORELLO, G. (2002). One and one-half
bound dichotomous choice contingent valuation, forthcoming 2002, Review of
Economics and Statistics.
- DESVOUSGES, W.H.; REED JOHNSON, F.; DUNFORD R.W.; HUDSON, S.P.;
WILSON, K.N. and KEVIN J.B. (1993). Measuring natural resource damages with
contingent valuation test of validity and reliability, in J. Hausman (ed.), Contingent
Valuation a Critical Assessment, North Holland, Amsterdam.
- EUROPEAN ENVIRONMENTAL AGENCY (1995). Europe’s Environment: The
Dobris Assessment. Eds. D. Stanners and P. Bourdeau. Copenhagen. Office for
Official Publications of the European Communities, Luxembourg and Elsevier Science
Ltd., Oxford, UK.
- EUROPEAN ENVIRONMENTAL AGENCY (1999). Environment in the European
Union at the turn of the Century, Copenhagen. Office for Official Publications of the
European Communities, Luxembourg and Elsevier Science Ltd., Oxford, UK.
27
- EUROPEAN ENVIRONMENTAL AGENCY (1999). EIONET Noise Newsletter,
European Environmental Agency, n 2, September.
- GARROD, G.D., SCARPA, R. AND K.G. WILLIS (2001). Estimating the benefits of
traffic calming in through routes: A choice experiment approach. Forthcoming in
Journal of Transport and Policy.
- GRANDJEAN, E., and A. GILGEN (1976). Environmental factors in urban planning.
Taylor and Francis Ltd., London.
- GUSKI, R. (1989). El ruido. Efecto de los sonidos no deseados, Ed. Herder,
Barcelona.
- HANEMANN, M., J. LOOMIS, and B. KANNINEN, (1991). Statistical efficiency of
double-bounded dichotomous choice contingent valuation. American Journal of
Agricultural Economics, 73 (4), 1255-1263.
- KAHNEMAN, D. and KNETSCH, J.L. (1992) Valuing public goods: the purchases of
moral satisfaction. Journal of Environmental Economics and Management, 22 (1), 57-
70.
- KANNINEN, B.J. (1995) Bias in discrete response contingent valuation. Journal of
environmental economics, 28 (1), 114-125.
- KRISTÖM, B. a) (1990) Economics of including recreation as a purpose of Easter
water projects. Journal of Farm Economics, 46, 1148-1157.
- KRISTÖM, B. b) (1990) Valuing environmental benefits using the contingent
valuation method. An econometric analysis. Ph.D. University of Umea.
- LAMBERT, J., F. POISSON AND P. CHAMPLOVIER (2001) Valuing benefits of a
road traffic noise abatement program: a contingent valuation study. INRETS-LTE,
Bron, France. Paper presented at the 17th International Congress on Acoustics, Rome,
September 2-7, 2001.
- LEVESQUE, T.J. (1994) Modelling the effects of airport noise on residential housing
markets: a case study of Winnipeg International Airport, Journal of Transport
Economics and Policy, 199-210.
- MCFADDEN, D. and G. LEONARD (1993) Issues in contingent valuation of
environmental goods: methodologies for data collection and analysis in J.A.
HAUSMAN (ed.) Contingent Valuation a Critical Assessment. North Holland,
Amsterdam.
- O’BYRNE, P.H. et al., (1985) Housing values, census estimates, disequilibrium and
the environmental costs of airport noise: a case study of Atlanta. Journal of
Environmental Economics and Management, 12, 169-178.
28
- NAVRUD, S. (2000) Economic benefits of a program to reduce transportation and
community noise. A contingent Valuation Study. In Proceeding of Internoise 2000,
Nice, Francia.
- POMMEREHNE, W.W. (1988) Measuring Environmental benefits: A comparison of
hedonic technique and contingent valuation, in Dieter Bos, D. M. Rose and C. Seidl
(eds.): Welfare and Efficiency in Public Economics, Springer; Berlin, Heidelberg, New
York.
- SAELENSMINDE, K. (1999) Stated choice valuation of urban traffic pollution.
Transportation Research Part D, 4; 13-27.
- SCARPA,R., WILLIS K.G. AND G.D. GARROD (2001) Estimating WTP for
effective enforcement of speed reduction from dichotomous-choice CV: The case of
rural trunk roads, Environmental and Resource Economics, 20(4); 281-304.
- RENEW, W.D. (1996) The relationship between traffic noise and house price.
Conference of the Australian Acoustical Society, Brisbane, 13-15th November, 1996.
- SOGUEL, N. (1994) Evaluation monetaire des atteinties a l’environment: une étude
hedoniste et contingente sûr l’impact des transports, Imprimerie de l’Evolve, SA
Neuchatel.
- VAINIO, M. (1995). Traffic noise and air pollution: valuation of externalities with the
hedonic price and contingent valuation methods, PhD thesis, School of Economic and
Business Administration, Helsinki.
- VAINIO, M. (2001). Comparison of the hedonic prices and contingent valuation
methods in urban traffic noise context, paper presented at the 2001 International
Congress and Exhibition on Noise Control Engineering, The Hague, The Netherlands,
August 27-30, 2001.
- WHO (1993). The Environmental Health Criteria Document on Community Noise,
report on the Task Force Meeting, Düsseldorf, Germany, 24-28 November, 1992.
World Health Organization, WHO Regional Office for Europe, Copenhagen.
- YAMAGUCHI (1996). Sustainable investment and resource use, UNESCO/Paternon,
Caraforth.
29
Table 1. Distribution of observations by initial price used in the one and one-half-bound
question
Cost Interval N % Bid drawn first N %
3.12 - 21.87 €
192
32.0
3.12 €
104
54.2
21.87 € 88 45.8
12.5 - 43.75 €
200
33.3
12.5 €
109
54.5
43.75 € 91 45.5
25.00 - 62.5 €
208
34,6
25 €
111
53.4
62.5 € 97 47.6
TOTAL 600 100 600
30
Table 2. Citizen Sensitivity to Noise
Noise disturbance suffered in
SENSI = 1
More noise
sensitive
SENSI=0
Less noise
sensitive
Neighbourhood***
House***
City***
7
5
7
3
3
5
Percentage of Sample 59% 41%
*** Significant at the 99% confidence interval
31
Table 3. OOHB Results
OOHB Estimation with protest zeros OOHB Estimation without protest
zeros
Coefficient T-Statistic Coefficient T-Statistic
Constant 1.03045 9.815*** 2.04931 13.880***
Bid -0.000216 -13.710*** -0.00029 -13.794***
Log-likelihood
N
-649.66608
592
-482.26
460
E(WTP)
21.86 €
26.41€
Confidence intervals for mean WTP
With protest zeros Without protest zeros
99% CI 20.24 - 23.23 € 25.33 - 27.32 €
95% CI 20.77 - 22.93 € 25.62 - 27-09 €
90% CI 20.90 - 22.75 € 25.75 - 26.97 €
*** Significant at the 99% confidence level
32
Table 4. OOHB estimations with socio-economic variables for the sample without protest
zeros
Variable Without option
and sensi
With option and
without sensi
Without option and
with sensi
With option
and sensi
Intercept
1.2972
(5.72)***
1.0930
(4.33)***
1.2040
(4.59)***
1.0333
(3.74)***
Price -0.3031
(14.56)***
-0.3054
(14.56)***
-.3101
(14.33)***
-0.3125
(14.32)***
Valpro .7442
(3.53)***
.7564
(3.57)***
.8176
(3.80)***
.8260
(3.83)***
Isola .1206
(.61)
.1532
(.76)
.1229
(.60)
.1519
(.74)
Univer .5802
(2.56)***
.5717
(2.55)***
.5546
(2.42)***
.5520
(2.39)***
Option - .4058
(2.10)**
- .3822
(1.92)*
Sensi - - .2161
(1.07)
.1815
(.89)
LogL
N obs.
-438,28
419
-436,06
419
-415,95
404
-414.09
404
Confidence Intervals
E(WTP) 28.42 € 28.78 € 28.79 € 29.03 €
99% 27.21-29.32 € 27.62-29.61 € 27.50-29.56 € 27.80-29.71€
95% 27.44-29.17 € 27.95-29.42 € 27.83-29.41€ 28.25-29.58 €
90% 27.62-29.06 € 28.08-29.34 € 27.98-29.35 € 28.38-29.52 €
T-statistics are in parentheses ***, **, * Significant at the 99% , 95% and 90% confidence level respectively.
33
Table 5. OOHB Estimation results with the option variable without protest zeros
Full Sample with option
variable
For alternative 1 For alternative 2
Coefficient T-Statistic Coefficient T-Statistic Coefficient T-Statistic
Constant
1.8896
11.68***
2.0301
10.38***
2.0674
10.27***
Bid -0.2910 -15.00*** -0.3180 -11.24*** -0.2633 -9.89***
Option 0.3388 1.85*
Log-
likelihood
N Obs.
-479.37
459
-245.36
225
-233.01
234
E(WTP)
26.97 €
26.33 €
26.50 €
99% CI 25.54 - 28.01 € 24.88 - 27.54 € 24.89 - 27.87 €
95% CI 25.93 - 27.74 € 25.34 - 27.26 € 25.36 - 27.52 €
90% CI 26.11 - 27.65 € 25.44 - 27.12 € 25.61 - 27.38 €
T-statistics are in parentheses . ***, **, * Significant at the 99% , 95% and 90% confidence level respectively..
34
Table 6. Noise Measurement by Neighbourhood
dB(A) in Leq.
Neighbourhood Working Hours Working Days at
09:30 p.m.
Sunday Morning
Casco Viejo
71,1
72,0
68,6
San Jorge
75,1
71,3
67,3
La Chantrea
60,0
51,6
59,3
San Juan
65,6
63,3
65,3
II Ensanche
69,6
64,6
66,3
35
Table7. OOHB estimations for neighbourhoods with continuous noise measurement
Variables Without Explanatory
Variables
With Explanatory
Variables
CV and SJo CH, 2EN,SJ CV and SJo CH, 2EN,SJ
Intercept 1.52
(3.83)***
1.43
(4.96)***
1.32
(1.97)**
0.73
(1.75)*
Price -0.33
(5.75)***
-0.25
(7.87)***
-0.35
(5.24)***
-0.25
(7.88)***
Option 1.32
(2.43)**
-.02
(.06)
1.27
(2.32)**
0.08
(.25)
Valpro .85
(1.36)
0.91
(2.57)**
Univer -.93
(1.73)*
0.04
(0.13)
N.Obs. 59 146 59 146
E(WTP) 28.16 € 23.87 € 28.95 € 27.30 €
Log-
likelihood
-63.11 -160.34 -60.69 -156.98
T-statistics are in parentheses . ***, **, * Significant at the 99% , 95% and 90% confidence level respectively.
36
Appendices
Valuation questions in the survey (Alternative 2; 3.12-21.87 € cost interval BIDU drawn first)
15) From the experience of the effect of similar programs in cities like Pamplona it is possible
to reduce both day and night-time noise by applying all the measures described earlier.
To give you an idea of what this reduction would mean, we can assure you that daytime noise
would be reduced from the working day level to that of a Sunday morning*. Regarding night
noise it would mean a reduction from the level on a Saturday night to that of a Monday night.
What overall score would you give such a program on a scale of 0 to 10? _______
As you know, this program would be costly and the local council would need to ask citizens to
pay for it by introducing a local tax increase. By taking data from similar programs, a research
team from the Universidad Pública de Navarra has estimated the cost to be somewhere
between 3.12 € and 21.87 € per household per year.
Would you be willing to pay 21.87 € per year in order to reduce the level of noise as described
before?
YES NO
[INTERVIEWER: if the interviewee has any doubts regarding the reduction in noise level,
please repeat:
DAYTIME: weekday morning to Sunday morning
NIGHTIME: Saturday evening to Monday Evening
If interviewee answers no then ask]
And what about 3.12 € per year?
YES NO
[Ask all]
What would be the most you would be willing to pay in order to reduce the level of noise in
Pamplona?
_______________Euros
[If the answer is 0 or less than 0 then ask]
What are your reasons for not wanting to pay for such a program?
* The alternative scenario used for the scope sensitivity test had this description “daytime noise would be reduced from the working day level
to that of a working day at 9:30 p.m.”