Decision and WTP for Weight Reduction

Willingness to Pay for Relocating Mobile Phone Base
Stations
Hsing-Chun Lin
Department of Applied Economics, National Chiayi University, Chiayi, 600, Taiwan
E-mail: [email protected]
Yih-Ming Lin
Department of Applied Economics, National Chiayi University, Chiayi, 600, Taiwan
E-mail: [email protected]
Paper presented in the joint conference of the Economy and Environment Program for
Southeast Asia (EEPSEA) and the East Asian Association on Environment and
Resource Economics (EAAERE), Hanoi, Vietnam, May 17-20, 2011
Abstract
The purpose of this paper is to estimate individual willingness to pay (WTP) for
relocateing mobile phone base stations (MPBSs) away from the vicinity of residences
and investigate factors that affect individual decision. The CV method is employed to
estimate the individual’s WTP and investigate the factors that affect such WTP. A
sample of 396 respondents was randomly selected and the survey was conducted by
face-to-face interviews. Results of the decision to whether to pay for removing
MPBSs show that the offered price is an important factor, as we as the cellular phone
monthly expenditure, the level of risk perception, and individual habits with regard to
engaging in exercise. The empirical results show that the mean and median of
estimated WTP were NT$5,657 and NT$5,864, respectively. Profile analysis suggests
that female individuals with a high level of education, a high level of risk perception,
high mobile phone expenditure is willing to pay the greatest amount to have the
MPBSs relocated.
Keywords: Contingent valuation method; Dichotomous choice model; Utility
different approach
JEL Classification: I31, Q51

Corresponding author. Assistant Professor, Department of Applied Economics, National Chiayi
University. Chiayi, 600, Taiwan. E-mail: [email protected]. Tel: 886-5-2732860. Fax:
886-5-2732853. Any remaining errors are ours and ours alone.
1. Introduction
Using a mobile phone has become an essential part of daily life in the past decade,
and has resulted in the number of mobile phones in use having dramatically increased
worldwide. It is reported that there were over four billion connections globally in
2008, and six billion is the number of phones predicted to be in use by the middle of
2012 (BBC, 2010). In Taiwan, the number of mobile connections has increased from
less than one million in 1996 to almost 27 million at the end of 2010. In addition, the
penetration of mobile phones has also risen from 4.51% in 1996 to 120.19% in 2010.
In order to provide a better mobile connection service to satisfy customer demand, the
number of mobile phone base stations (hereinafter MPBSs) has also increased in past
years. Despite the customers’ high demand for better mobile phone coverage in areas
where they live or work, most of them do not favor having MPBSs installed in their
neighborhood. Concerns associated with MPBSs have received increasing attention
and have recently become important public issues. Taiwan is in particular a very
densely populated island in which a population estimated to exceed 23 million lives
with 27,000 MPBSs on a island covering 36,000 square kilometers. That means that,
on average, there are three MPBSs for every four square kilometers. In particular, the
MPBSs are mostly located in densely populated residential, commercial and industrial
areas. The density of MPBSs is much higher in urban areas.
The MPBSs have become the most recent NIMBY1 facilities in Taiwan. The main
public concern related to MPBSs has to do with the fears of potential health hazards
from the radiation and/or electromagnetic fields that these devices produce. The
health risk effects associated with MPBSs have been a concern for the past decade
NIMBY or Nimby is an acronym for “Not In My Back Yard,” which is used more generally to
describe people who advocate some proposal, but oppose implementing it in such a way that would
require sacrifice on their part.
1
1
(Blettner et al., 2009; Scharie, 2010). It has been shown that living around the MPBSs
possibly increases the potential risk of certain symptoms and/or diseases, such as
headaches, impaired concentration, irritability, cancers, the stimulation of mast cells
that produce histamine, sleeping problems, etc. (Johansson, 2006; Santini et al., 2003;
Hunter et al., 2006; Danker-Hopfe et al., 2010). Even though the research results are
divergent, 2 the public surveys indicate high levels of public concern due to the
potential health hazards from the electromagnetic fields that the MPBSs emit.
Furthermore, the increasing media attention regarding the potential health risks
associated with the development of certain diseases related to MPBSs has caused such
fears to spread which has had a negative impact on property values or house prices. In
addition, the installation of MPBSs has also had negative impacts on house prices in
residential neighborhoods (Bond, 2005; Bond and Beamish, 2007; Bond and Wang,
2005; Bond, 2007). Bond (2007) points out that the installation of cellular phone
towers has a statistically significant effect on the prices of properties located near a
tower. The prices of properties generally decreased by more than 2% after a cellular
phone tower was built in Florida.
In addition, the noise they continuously make and the safety of the MPBSs3 are the
other reasons why MPBSs are so unpopular. The NIMBY phenomenon in regard to
MPBSs has become a serious social problem and an important public issue (Barnes,
1999; Szmigielski and Sobiczewska, 2000). In the past decade, there have been many
protest rallies and/or related events over the installation of MPBSs occurring in
On March 10, 2011, we used the keywords “cellular phone base station” to search for related medical
papers in PubMed. We found that there are more than 40 articles published between 2005 and 2010.
However, the research results related to the health risk effects of MPBSs are ambiguous. Some research
results show that the health effect of living around the MPBSs is not significant. (Elliott et al., 2010;
Eltiti et al., 2007)
3
From public surveys on the safety of MPBSs, it can be seen that residents living near MPBSs worry
about the possibility that the MPBSs could explode like electricity transformer boxes.
2
2
Taiwan. Most people who participate in the protest rallies and/or events believe that
the installation of MPBSs in the neighborhood could decrease the quality of the
environment and increase the chances of their developing certain diseases. So far,
there have been about 1,500 protest events and/or rallies occurring each year in
Taiwan in the past three years and the total number of protest events and/or rallies has
exceeded 10,000. In the meantime, the government has actively propagated how safe
the MPBSs are in the media. The phenomenon of NIMBY related to MPBSs has
occurred worldwide, such as in the U.S, Germany, Australia, New Zealand,
Bangladesh, and Taiwan (Campos, 2004; Bond, 2005; Bond and Beamish, 2007;
Bond and Wang, 2005; Bond, 2007; Scharie, 2010; Blettner et al., 2009)
Environmental quality and reduced health hazard improvements due to the removal
of MPBSs have no market price, which makes the benefits of such improvements in
environmental quality and reduced health hazards difficult to quantify. An approach
to the valuation of benefits which has been widely applied in evaluations of
environmental programs and health care is the contingent valuation (hereinafter CV)
method. The CV method is a well-established research approach which uses survey
techniques to elicit consumers’ valuations of non-market goods (Hanemann, 1984;
Hanemann et al., 1991; Mitchell and Carson, 1989; Haab and McConnell, 2002). The
benefits of environmental quality and reduced health hazard improvements have been
widely evaluated based on the concept of willingness to pay (hereinafter WTP) since
they are consistent with the principles of welfare economic theory and cost benefit
analysis. Based on the utility maximization principle, the benefit to a customer of a
service is defined as the maximum amount that the individual is willing to pay. The
sum of each individual’s WTP can be considered to be the benefit to society of the
intervention. Among various approaches to measure the WTP, the CV method is an
3
appropriate tool for evaluating a conceptually correct and complete measure of WTP.
In this study, we have employed the CV method and conducted a survey of the
general public to solicit the respondents’ WTP.
The CV method obtains evaluation data on non-market goods by asking a
respondent about his or her WTP. It elicits answers from the respondents to find out
what they would be willing to pay for non-market goods or services under certain
hypothetical market scenarios. In the absence of markets, we present consumers with
hypothetical market scenarios in which they have the opportunity to buy the good in
question. Previous applications of the CV method have often been limited to
measuring the benefit of public goods, such as air and water quality improvements
(Alberini et al., 1997; Cummings et al., 1986; Dickie and Gerking, 1991; Shaw et al.,
1999). Recently, the CV method has been widely applied in evaluating the value of
health (Fu et al., 1999; Johannesson et al., 1991). Many health economics researchers
have applied the CV method in their studies aimed at determining consumers’ WTP
for health care (Dickie and Gerking, 1991; Donaldson et al., 1998; Johannesson and
Johansson, 1997) and for assessing the benefits of some preventive care actions or
practices (Donaldson et al., 1995; Johannesson et al., 1991; Ryan, 1996, 1997;
Zethraeus, 1998; Phillips et al., 1997; Liu et al., 2009). It should be noticed that the
WTP measures may vary substantially and be affected by the design and data
collection method used in the survey. There is a consensus in the literature that a
face-to-face interview is the preferred method for obtaining reliable answers because
it allows for flexibility in controlling the flow and amount of information to be
presented (Olsen and Smith, 2001).
Based on a random utility framework, a utility difference model of Hanemann’s
welfare evaluation model (Hanemann, 1984) is proposed to elicit individual WTP. In
4
this study, we apply a binary contingent valuation question, in which each individual
or respondent is asked if he or she would like to pay an offered price or not. By
diversifying the offered price in different sub-sample questionnaires, the relationship
between the offered price and the proportion of respondents who are willing to pay
can be derived. A dichotomous choice model, such as a probit or logit model with
contingent valuation data is applied. Since the estimation technique underlying this
approach is consistent with a utility framework and widely applied, the dichotomous
choice approach to CV is employed in this study. To the best of our knowledge, no
research in the economics literature has so far estimated the WTP for preventing the
installation of MPBSs or investigated the factors that may affect the benefits from
such prevention using the CV approach.
The remainder of this paper is organized as follows. In Section 2, a simple utility
difference model which uses a probit model to investigate the determinants of WTP
for removing MPBSs is presented. Section 3 presents the survey data and the
variables employed in the empirical model specifications. In Section 4, we illustrate a
parametric estimate of empirical WTP values for removing MPBSs as well as three
non-parametric estimates of the WTP. The profile analyses of the predicted WTP are
also provided. Finally, Section 5 provides some concluding remarks.
2.
The Hanemann Utility Difference Model
In order to provide accurate estimates of the benefits from preventing the installation
of MPBSs in residential neighborhoods, the Hanemann utility difference model
(Hanemann, 1984) is employed in our analysis. A random utility framework is applied
to derive the WTP of an individual based on a dichotomous choice using contingent
valuation data. Following Hanemann’s model, the individual faces a choice as to
whether or not to pay the offered price, say NT$P, in order to prevent MPBSs from
5
being installed in the neighborhood of the respondent’s residence. If he/she decides to
pay the price (D = 1), then he/she will enjoy the effect of their removal but with an
NT$P reduction in income, and thus the individual’s utility can be expressed as:
U1  U  D  1, Y  P, S   V  D  1, Y  P, S   u1 ,
(1)
where D represents the decision variable as to whether or not to pay to prevent the
installation of MPBSs in the immediate vicinity, Y denotes the individual’s income, P
is the price offered for the remedy, S is a vector of other exogenous variables that may
affect the individual’s preferences, and u1 is the error term with zero mean. Similarly,
the utility of not paying can be expressed as:
U0  U  D  0, Y , S   V  D  0, Y , S   u0 .
(2)
Similarily, u0 is also the error term. Based on the utility maximization decision rule,
the difference in utility between those two scenarios, U1 – U0, will affect the
individual’s decision. An individual would like to pay $P (D = 1) if U1 > U0, and will
not pay otherwise (D = 0), if U1 < U0. That is, the probability of choosing to pay can
be expressed as:
Prob  D  1  Prob U1  U 0 
 Prob V  D  1, Y  P, S   u1  V  D  0, Y , S   u0  .
(3)
Rearranging equation (3), we have
Prob  D  1  Prob  u0  u1   V   F (V ) ,
and
Prob  D  0  1  Prob  D  1 ,
(4)
where V  V  D  1, Y  P, S   V  D  0, Y , S  is the difference in utility between
paying and not paying for preventing the installation of MPBSs;   u0  u1 , and
F (.) is the distribution function of  . Assuming the error terms are independently
6
and identically distributed as a normal distribution, equation (4) can be estimated
using the probit model.
In empirical estimation, the utility functional form, V(.), is usually specified by the
linear form or log-linear form, which can be expressed by
V  D  1, Y  P, S   1  1 (Y  P)   1S
,
or
V  D  1, Y  P, S   1  1 ln(Y  P)   1S
.
Similarly,
V  D  0, Y , S   0  0Y   0 S
,
or
V  D  0, Y , S   0  0 ln Y   0 S .
Therefore, the utility difference functional form can be expressed by
V  (1  0 )  (1  0 )Y  1P  ( 1   0 ) S ,
(5)
or
V  (1   0 )  1 ln(Y  P)  0 ln Y  ( 1   0 )S .
(6)
In Hanemann (1984) and Bowker and Stoll (1988), 1 and  0 are assumed to be
identical, and so equations (5) and (6) become
V  (1  0 )  1P  ( 1   0 ) S ,
(5’)
and
V  (1   0 )  1 ln(1 
P
)  ( 1   0 ) S .
Y
(6’)
Based on the above model, the respondent’s WTP is determined by
Pr ob(U1  U 2 )  Fu1 u2 (V )  0.5.
In the probit model, where Fu1 u0 (.) is the standard normal distribution function,
7
Fu1 u0 (0)  0.5 . Therefore, the WTP can be derived as:
WTP 
(1   0 )  ( 1   0 )Y  ( 1   0 ) S
1
(7)
.
or
WTP  Y  Y
o
1
exp[
(1   0 )  ( 1   0 ) S
1
],
(8)
In equation (5), the utility difference functional form can be rewritten as
V  (1  0 )  1P  ( 1   0 ) S ,
(9)
if the marginal utility of income is constant. Thus, the WTP can be derived as follows
WTP 
(1   0 )  ( 1   0 ) S
1
.
(10)
3. The Survey and Sample Data
The data used in this empirical study are collected during the period from June 1 to
June 30, 2006 in Chiayi city, Taiwan. Chiayi City is a moderately sized city in Taiwan,
in which there are about 372,000 inhabitants living in an area of 60 km2. In 2006,
there were about 600 MPBSs in Chiayi city, which means that, on average, there were
10 MPBSs per km2 in Chiayi city.4 The sample respondents were randomly selected
and the survey was conducted on the basis of face-to-face interviews which is the
preferred method for obtaining reliable answers. There were 576 respondents actually
living in Chiayi city surveyed and all of these respondents were adults, and desired to
see the MPBS towers removed. The respondents were asked to answer a list of
questions regarding their preferences, risk perceptions about living around MPBSs
and their WTP to have all MPBSs removed from their residential neighborhoods.
4
In most cases, there is more than just one MPBS on the roof of a building.
8
Socio-demographic data and other information such as that related to the usage of
mobile phones, as well as that concerning the respondent’s health condition, were also
collected for each respondent. Of the 576 questionnaires returned, 180 could not be
used for the analysis of WTP because of various omissions, or the lack of economic
means to pay the price. A sample of 396 observations which met the survey criteria
was thus employed.
In order to elicit the respondent’s WTP to have all MPBSs removed from his/her
residential neighborhood, the CV method is employed in the study. During the
interview, the respondent was first asked if he/she would really like to remove the
MPBSs from his/her residential neighborhood. For those respondents who answered
positively, they were then asked a follow-up question:
“We are going to ask you a hypothetical question. Assuming that the quality of
your mobile phone communication remains unchanged, are you willing to pay
NT$ XXX per year to remove all mobile phone base stations from your
residential neighborhood?
Note: Please note that the money will be taken from your own disposable income
and hence will decrease your private consumption.
(Note that the amount NT$XXX is the price offered by the interviewer).
In this survey, 10 different prices were randomly provided in the questionnaire
ranging from NT$100 ($3.08) to NT$7,000 ($215.38)(see Table 1). 5 During the
face-to-face interview, one price was randomly offered and the respondent was asked
if he/she would be willing to pay that offered price annually to have all MPBSs
5
The exchange rate between the New Taiwan dollar (NT$) and the US dollar in June 2006 fluctuated
between US$1 = NT$32.05 and US$1 = NT$32.73 and averaged about US$1 = NT$32.5.
9
removed from the residential neighborhood. Table 1 illustrates the distribution of
responses to those 10 offered prices. Among the 396 respondents, the proportion of
those saying yes to offer price NT$100 ($3.08) was higher than 90%, whereas only
39% responded positively to the offered price of NT$7,000($215.38). In general, the
percentage of those saying yes to an offered price decreased as the offered price went
up, which showed that this distribution satisfied the properties of a demand function.
[insert Table 1 here]
There are several types of explanatory variables employed in our empirical study,
including social demographic variables, risk perception variables, attitudes about
health variables, mobile phone usage condition variables, and the offered price based
on the WTP question, etc. The descriptive statistics and variable definitions are
reported in Table 2. The socio-demographic variables include the respondent’s gender
(SEX), age (AGE), years of schooling (EDU), annual income (INCOME), marital
status (MARRIED) and employment status (JOB). Table 2 indicates that half of the
sampled respondents are male and the respondent’s mean age is about 40 years old.
The respondents have an average of 13 years of schooling and the individual’s
average annual income is about NT$466,000 (US$14,338). More than 70% of the
respondents are married and 93% are employed.
[insert Table 2 here]
The risk perception variable represents the respondent’s perception regarding the
information on the health risks associated with living around MPBSs. Respondents
were queried on their risk information and awareness of 6 kinds of diseases or
symptoms.6 A score of one point was awarded to each question with a positive
response. A risk perception index (RPI), ranging from 0 to 6, was constructed by
6
The 6 kinds of diseases and/or symptoms include cancer, insomnia, miscarriage, headaches, tinnitus,
and giving birth to a malformed baby. Those diseases and/or symptoms are often shown in the
documents of protest events and/or rallies related to the MPBSs.
10
summing the scores for the 6 questions for each respondent. Table 2 indicates that the
average RPI was 3.48 with a standard deviation of 2.53 for the current sample. We
employed SPORT (if the respondent was in the habit of engaging in exercise) as the
proxy variable for the respondent’s attitude with regard to his/her health condition.
Since regular exercise is costly for an adult but good for health, a respondent who can
maintain the habit of engaging in exercise means that the respondent is much more
careful about his/her health condition. Table 2 shows that 59% of the sample
respondents engaged in some form of habitual exercise.
The mobile phone usage condition variables included CELLYEAR (years of using
a mobile phone) and CELLBILL (monthly mobile phone expenditure). As shown in
Table 2, the respondent had on average used a mobile phone for 6.55 years in the
sample. Furthermore, we found that the average respondent’s monthly mobile phone
expenditure was about NT$811 (US$24.95). The last explanatory variable employed
in this empirical study was the price asked in the WTP question. The average offered
price was about NT$2,041 (US$62.08), with the distribution of offered prices being
illustrated in Table 1.
4. The Benefits of Removing MPBSs
4.1 The Empirical Model Specification
There are five different model specifications estimated in this empirical study.
Models I and III are derived from equations (5) and (6), respectively, while Models II
and IV are derived from equations (5’) and (6’), respectively. Model II is Model I
without the INCOME term which is the empirical model when the marginal utility of
income remains unchanged in the two different decision situations (D=0 or D=1).
Similarly, Model IV is derived when the coefficients of LOG(INCOME) ( 1 and  0 )
11
in equation (6) are assumed to be the same in the two different decision situations
(D=0 or D=1). In Hanemann (1984), Model II and Model IV are developed.
Furthermore, in addition to Models II and IV, Model V is also adopted for the
empirical estimation by Bowker and Stoll (1988). A sample with 396 respondents is
employed in this empirical study. Table 3 presents the maximum likelihood estimation
results of the probit model for five different model specifications.
In Table 3, it is shown that the respondent’s decision is negatively correlated with
the prices offered (PRICE), as expected. The effect of price on the respondent’s
decision is highly significant at the 1% significance level in Models I, II and V. The
signs of the coefficients related to price and income in Models III and IV are also
consistent with theoretical predictions. It should be noted that those respondents who
have a habit of engaging in exercise (SPORT), have a high level of risk perception
(RPI), and with higher mobile phone expenditure (CELLBILL) are more likely to pay
the price for removing all MPBSs from their residential neighborhoods. We also find
that none of the socio-demographic variables are significant among the five models.
In general, the estimation results are robust and consistent, which means the empirical
results are more reliable. Regardless of the model specification used, the estimation
results are consistent with our expectations among the five different model
specifications.
[insert Table 3 here]
Furthermore, the empirical results indicate that the correct prediction percentages
of Model I and Model II are higher than in other models, with more than 78% of the
predictions being correct. Similarly, Models I and II have higher McFadden R2 values
than Models III and IV, which means that the model specifications of equations (5)
and (5’) are better than those of equations (6) and (6’) in this empirical study. In
addition, we find that the coefficient of INCOME is not significant in Model I, which
12
means that 1 and  0 in equation (5) are the same, which is consistent with
Hanemann’s (1984) model. Based on the AIC and SIC indicated in Table 3, it is
shown that Model II has the lowest values for both indices, which means that Model
II is supposed to be the best fitting model among the five model specifications. The
WTP estimated value can be obtained using the Model II estimates.
4.2 Willingness to Pay Analysis
The probit estimates of Model II are employed in this WTP analysis. Using the
probit estimates of Model II in Table 3, we can obtain the respondent’s predicted WTP
for removing MPBSs. As shown in Table 4, the average WTP for relocating MPBSs
away from the residential neighborhood is estimated to be NT$5,657 ($174.06) with a
standard deviation of NT$2,279 ($70.12). The median of the sample WTP is
NT$5,864 ($180.43) and the interquartile range is NT$3,250 ($100). The predicted
WTP seems to be distributed uniformly across the range.
[insert Table 4 here]
Furthermore, since the estimates of the parameters can be inconsistent if the
distribution specification is incorrect and the empirical results are sensitive to the
choice of probability function, a non-parametric approach is also employed to
calculate the mean and median WTP which has been proposed by Kriström (1990) in
which the specification of the distribution function is not critical. The proportion of
acceptance at different offered bids is used to construct a survival function. Based on
Kriström (1990), if the proportion of yes answers is greater at the higher of two
consecutive bids, the number of the yes answers for the two consecutive bids is added
and the proportions are calculated for the two bids combined. This process continues
until a non-increasing sequence of proportions is accomplished. A survival function
between these points is constructed using a linear interpolation approach. In
13
estimating the mean WTP, we assume that the highest WTP is NT$10,000 ($307.69)
(Kriström, 1990). Based on the distribution of responses illustrated in Table 2, the
empirical survival function in this study is depicted in Figure 1, which shows the
relationship between the offered price asked in the WTP question and the proportion
accepting the offered price. The non-parametric estimated mean can be calculated as
the area bounded by the survival function and the median can be obtained by a linear
interpolation approach. The mean and the median of WTP obtained by using the
non-parametric method are about NT$5,080 ($156.31) and NT$5,910 ($181.85),
respectively. The Turnbull expected distribution-free WTP estimates are also obtained.
The Turnbull lower bound and upper bound estimates of the expected WTP are about
NT$4,098 ($126.09) and NT$6,062 ($186.52), respectively (Haab and McConnell,
2002). We find that the means are smaller than the medians based on both the
parametric and non-parametric approaches. Furthermore, the parametric mean WTP,
the non-parametric mean WTP and the Turnbull upper bound estimated mean WTP
are quite close, and the estimate of the parametric mean WTP lies in between the two
non-parametric estimates.
[insert Figure 1 here]
4.3 Profile Analysis for Estimated WTP
Different combinations of respondent features can be estimated in order to
investigate the effects on the predicted WTP for removing the MPBSs. The results of
the WTP for some particular profiles are analyzed for the purpose of illustration. For
simplicity, we use three explanatory variables as control variables, in which two
variables, RPI and CELLBILL, are included in each profile to be combined with
another selected explanatory variable such as AGE, EDU, JOB and so on. The
estimated results are summarized in Table 5, in which the predicted WTP is measured
14
at mean values for the explanatory variables, except for the three control variables in
each profile.
Based on the results shown on Table 5, a high risk perception (RPI=5) respondent
is willing to pay NT$2,677 ($82.37) more than a low risk perception respondent. A
high monthly mobile phone expenditure (CELLBILL= 2000) respondent is willing to
pay NT$1,842 ($56.68) more than a respondent with low monthly mobile phone
expenditure (CELLBILL=300). Furthermore, the profile analysis also shows that the
age of a respondent does not have a great impact on his/her WTP to remove MPBSs.
We estimate that a female respondent would be willing to pay NT$1,143 ($35.17)
more than a male respondent, which means that a female would pay more attention
and worry more about her health hazards and environmental quality than a male. Even
though the estimated coefficient for the year of schooling (EDU) is not statistically
significant in maximum likelihood estimation results of probit regressions in Table 3,
Table 5 indicates that the year of schooling (EDU) is still an important factor affecting
the WTP. We can find that a respondent who graduated from college (EDU = 16), is
willing to pay NT$1,321 (40.65) more than a respondent who only graduated from
junior high school (EDU=9). It is also found that the conditions related to a
respondent’s job status as well as marital status do not seem to play an important role
in determining his/her WTP. Among the socio-demographic variables, we can not
investigate the effect of representative respondent income (INCOME) on WTP from
Table 5 since we employ the empirical results of Model II to estimate the respondent’s
WTP. Furthermore, having a habit of engaging in exercise or not (SPORT) is also an
important determinant of a representative respondent, whereas the impact of the
number of years the mobile phone has been used (CELLYEAR) is not so significant.
[insert Table 5 here]
The predicted WTPs are positive in all cases. On average, a high level of risk
15
perception (RPI=5) together with a high monthly mobile phone expenditure
(CELLBILL=2000) representative respondent has a relatively high WTP to have the
MPBSs removed, whereas a low level of risk perception (RPI=1) with a low monthly
mobile phone expenditure (CELLBILL =300) representative respondent has a
relatively low WTP in each case. Among the other explanatory variables, SPORT
(INCOME = 1), SEX (SEX = 0) and EDU (EDU=16) have the greatest WTP. In other
words, when marketing anti-radiation or anti-electromagnetic field devices, therapies
or medicine, the results suggest that one should perhaps target the segment of female
consumers with high education, high mobile phone expenditure, who have a habit of
engaging in exercise, and who have a high level of risk perception regarding the
MPBSs.
4.4 Policy Implications
Since the phenomenon of NIMBY related to MPBSs is gaining increasing attention
and has become an important public issue, many public policies related to the
installation of MPBSs have been suggested, such as requiring stricter regulations
about the installation of MPBSs, requiring the agreement of all residents living around
the MPBSs, and subsidizing the Telecom companies to develop more advanced
equipment to replace the MPBSs, etc. The estimated WTP could be considered as an
indicator of how cautious the sample respondents are perceived to be about their
environmental quality and health condition. Therefore, one policy implication is that
resources need to be directed toward both prevention and to managing the installation
of the MPBSs. In other words, the results regarding the WTP to have the MPBSs
removed can provide policy-makers with some useful information to help them to
determine whether to implement such policies and to understand which of these
policies related to MPBSs can receive greater support and who is much more
supportive of them.
16
5. Concluding Remarks
The study investigates the factors that may affect an individual’s decision with
regard to the installation of MPBSs, and develops a means of estimating WTP to have
MPBSs removed from residential neighborhoods. To the best of our knowledge, this
is the first study that attempts to evaluate the benefits of removing MPBSs by
adopting the CV approach in the economics literature. The sample respondents were
randomly selected and the data were collected based on face-to-face interviews in
Chiayi city, Taiwan during 2006. A total of 396 sample observations are used for the
WTP analysis.
The results of the decision as to whether to pay to have the MPBSs removed show
that the offered price is an important factor. As is expected, the higher the price that a
respondent has to pay, the less likely the respondent will be willing to pay.
Furthermore, the frequency of using a cellular phone, the level of risk perception, and
the individual’s habit of engaging in exercise are all found to be significant factors
that contribute to the respondent’s decision regarding the MPBSs. The mean and
median of the WTP to remove MPBSs from residential neighborhoods are estimated
to be NT$5,657 ($174.06) and NT$5,864 ($180.43), respectively, among the sample
respondents. However, the risk perception effect seems to dominate the WTP. Among
the explanatory variables, SPORT and CELLBILL have the greatest impact on WTP.
The results of the profile analysis suggest that the individuals with a high level of
education, high risk perception, high mobile phone expenditure, and who have the
habit of engaging in exercise are most likely to be willing to pay the greatest amount
to have MPBSs removed from the neighborhood.
17
This study is the first attempt based on the contingent valuation method to
investigate the respondent’s WTP to have MPBSs removed from residential
neighborhoods. The data employed in this empirical study are unique but not without
some limitations. The survey was conducted in a moderately sized city in Taiwan.
Therefore, the results should be interpreted with caution when the conclusions and
implications are generalized and applied to a larger population in a broader context.
18
REFERENCES
Alberini, A., M. Cropper, T. Fu, A. Krupnick, J. Liu, D. Shaw, and W. Harrington
(1997), “Valuing Health Effects of Air Pollution in Developing Countries: The
Case of Taiwan,” Journal of Environmental Economics and Management, 34(2),
107-126.
Barnes, J. R. (1999), “Cellular Phones: Are They Safe?” Professional Safety, 44 (12),
20-23.
BBC (2010) “Over 5 Billion Mobile Phone Connections Worldwide,” http://www.bbc.
co.uk/news/10569081.
Blettner, M., B. Schlehofer, J. Brecknekamp, B. Kowall, S. Schmiedel, U. Reis, J.
Potthoff, J. Schuz, G. Berg-Beckhoff (2009), “Mobile Phone Base Stations and
Adverse Health Effects: Phase 1 of a Population-based, Cross-sectional Study
in Germany,” Occupational and Environmental Medicine, 66(2), 118-123.
Bond, S. (2007), “The Effect of Distance to Cell Phone Towers on House Prices in
Florida,” Appraisal Journal, 75(4), 362-370.
Bond, S.G. (2007), “Cell Phone Tower Proximity Impacts on House Prices: A New
Zealand Case Study,” Pacific Rim Property Research Journal, 13(1), 63-91.
Bond, S.G., Beamish, K. (2005), “Cellular Phone Towers: Perceived Impact on
Residents and Property Values,” Pacific Rim Property Research Journal, 11(2),
158-177.
Bond, S. and Wang, K. (2005), “The Impact of Cell Phone Towers on House Prices in
Residential Neighborhoods,” Appraisal Journal﹐73(3), 256-277.
Bowker, J. and J. Stoll (1988), “Use of Dichotomous Choice Nonmarket Methods to
Value the Whooping Crane Resource,” American Journal of Agricultural
Economics, 70(2), 372-381.
19
Campos, J. (2004), “The Epidemic Spreads: Cellular Base Station Antennas,” Juniper
Berry Magazine, http://www.junipercivic.com/juniperberryarticle.asp?nid=209.
Cummings, R, D. Brookshire and W. Schulze eds. (1986), Valuing Environmental
Goods: A State of the Art Assessment of the Contingent Valuation Method, NJ:
Rowman & Allanheld.
Danker-Hopfe, H., H. Dorn, C. Bornkessel, and C. Sauter (2010), “Do Mobile Phone
Base Stations Affect Sleep of Residents? Results from an Experimental
Double-blind Sham-control Field Study,” American Journal of Human Biology,
22(5), 613-618.
Dickie, M. and S. Gerking (1991), “Willingness to Pay for Ozone Control: Inferences
from the Demand for Medical Care,” Journal of Environmental Economics and
Management, 21, 1-16.
Donaldson, C., P. Shackley, M. Abdalla and Z. Miedzybrodzka (1995), “Willingness
to Pay for Antenatal Carrier Screening,” Cystic Fibrosis, 4, 439-452.
Donaldson, C., A. Jones, T. Mapp and J. Olson (1998), “Limited Dependent Variables
in Willingness to Pay Studies: Applications in Health Care,” Applied Economics,
30, 667-677.
Elliott, P., M. Toledano, J. Bennett, L. Beale, K. de Hoogh, N. Best, and D. Briggs
(2010), “Mobile Phone Base Stations and Early Childhood Cancers:
Case-control Study,” British Medical Journal, BMJ2010; 340:c3077doi10.
1136/bmj.c3077.
Eltiti, S., D. Wallace, A. Ridgewell, K. Zougkou, R. Russo, F. Sepulveda, D.
Mirshekar-Syahkal, P. Rasor, R. Deeble, and E. Fox (2007) “Does Short-Term
Exposure to Mobile Phone Base Station Signals Increase Symptoms in
Individuals Who Report Sensitivity to Electromagnetic Fields? A Double-Blind
Randomized Provocation Study,” Environmental Health Perspectives, 115,
20
1603-1608.
Fu, T.-T., J.-T. Liu and J. Hammitt (1999), “Consumer Willingness to Pay for
Low-Pesticide Produce in Taiwan,” Journal of Agricultural Economics (U.K.),
Vol. 50, No. 2, pp. 202-233.
Haab, T., and K. McConnell (2002), Valuing Environmental and Natural Resources:
The Econometrics of Non-market Valuation, Edward Elgar Publishers, UK.
Hanemann, W. (1984), “Welfare Evaluation in Contingent Valuation Experiments
with Discrete Responses,” American Journal of Agricultural Economics, 66(3),
332-341.
Hanemann, W.M., J. Loomis and B. Kanninen (1991), “Statistical Efficiency of
Double-Bounded Dichotomous Choice Contingent Valuation,” American
Journal of Agricultural Economics, 73, 1255-1263.
Hutter, H., H. Moshammer, P. Wallner, and M. Kundi (2006), “Subjective Symptoms,
Sleeping Problems, and Cognitive Performance in Subjects Living near MPBSs,”
Occupational and Environmental Medicine, 63(5), 307-313.
Johansson, O. (2006), “Electrohypersensitivity: State-of-the-Art of a Functional
Impairment,” Electromagnetic Biology Medicine, 25, 245-255.
Johannesson, M., B. Jönsson and L. Borgquist (1991), “Willingness to Pay for
Antihypertensive Therapy – Results of a Swedish Pilot Study,” Journal of
Health Economics, 10, 461-474.
Johannesson, M., P.-O. Johansson, B. Kriström, L. Borgquist and B. Jönsson (1993),
“Willingness to Pay for Lipid Lowering: A Health Production Function
Approach,” Applied Economics, 25, 1023-1031.
Johannesson, M. and P.-O. Johansson (1997), “Quality of Life and the WTP for an
Increased Life Expectancy at an Advanced Age,” Journal of Public Economics,
65, 219-228.
Kriström, B. (1990), “A Non-Parametric Approach to the Estimation of Welfare
21
Measures in Discrete Response Valuation Studies,” Land Economics, 66,
135-139.
Liu, J., M. Tsou and J. Hammitt (2009), “Willingness to Pay for Weight-Control
Treatment,” Health Policy, 91, 211-218.
Mitchell, R. and R. Carson (1989), Using Surveys to Value Public Goods: The
Contingent Valuation Method, Washington, D.C.: Resources for the Future.
Olsen, J. and R. Smith (2001), “Theory versus Practice: A Review of ‘Willingness-To
Pay’ in Health and Health Care,” Health Economics, 10, 39-52.
Phillips, K.A., R.K. Homan, H.S. Luft, P.H. Hiatt, K.R. Olson, T.E. Kearney and S.E.
Heard (1997), “Willingness to Pay for Poison Control Centers,” Journal of
Health Economics, 16, 343-357.
Ryan, M. (1996), “Using Willingness to Pay to Assess the Benefits of Assisted
Reproductive Techniques,” Health Economics, 5, 543-558.
Ryan, M. (1997), “Should Government Fund Assisted Reproductive Techniques? A
Study Using Willingness to Pay,” Applied Economics, 29, 841-849.
Santini, R., P. Santini, P. Le Ruz, J. Danze and M. Seigne (2003), “Survey Study of
People Living in the Vicinity of Cellular Phone Base Stations,” Electromagnetic
Biology and Medicine, 22(1), 41-49.
Shaw, D., Y. Chien and Y. Lin (1999), “An Alternative Approach to Combining
Revealed and Stated Preference Data: Valuing Water Quality of the River
System in Taipei,” Environmental Economics and Policy Studies, 2(2), 97-112.
Szmigielski, S and E. Sobiczewska (2000), “Cellular Phone Systems and Human
Health – Problems with Risk Perception and Communication,” Environmental
Management and Health, 11(4), 352-368.
Zethraeus, N. (1998), “Willingness to Pay for Hormone Replacement Therapy,”
Health Economics, 7, 31-38.
22
Table 1. Distribution of responses to WTP question
Offer Prices
(NT$)
Number
Percent
100
33
8.33%
200
37
9.34%
400
48
12.12%
600
48
12.12%
1000
42
10.61%
1500
43
10.86%
2000
32
8.08%
3000
34
8.70%
5000
38
9.60%
7000
41
10.35%
Total
396
100%
23
Yes
No
30
(90.91%)
31
(83.78%)
44
(91.67%)
38
(79.17%)
32
(76.19%)
33
(76.74%)
19
(59.38%)
25
(73.53%)
24
(63.16%)
16
(39.02%)
292
(73.74%)
3
(9.09%)
6
(16.21%)
4
(8.33%)
10
(20.83%)
10
(23.81%)
10
(23.24%)
13
(40.62%)
9
(26.47%)
14
(36.84%)
25
(60.98%)
104
(26.24%)
Table 2. Definitions of Variables and Sample Characteristics
Variable Name
Description
Sample Mean
(S.D.)
Socio-demographics Variables
SEX
Dummy, (male = 1, female = 0)
AGE
Age of respondent
EDU
Years of schooling
INCOME
MARRIED
JOB
Annual income (NT$)
Dummy (married =1, otherwise= 0)
Dummy, (employed = 1, otherwise= 0 )
0.50
(0.50)
39.62
(11.92)
12.85
(2.70)
466,060.61
(457,670.09)
0.707
(0.456)
0.93
(0.25)
Risk Perception Variable
RPI
Risk perception index
from 0 to 6.
3.48
(2.52)
Attitude about Health Variable
SPORT
Dummy (=1 if respondent has the habit of
exercise, 0 otherwise )
0.593
(0.492)
Mobile Phone Use Condition Variables
CELLYEAR
Years of using mobile phone
CELLBILL
Monthly mobile phone expenditure (NT$)
6.55
(3.05)
811.42
(841.33)
Offered Price
PRICE
WTP question offered price (NT$)
Sample size
2040.91
(2198.42)
396
24
Table 3. Probit estimation results
Variable
Model I
Model II
Names
-0.479
-0.442
Constant
(0.719)
(0.715)
Model III
Model IV
Model V
3.347
(1.687)**
-0.146
(0.690)
1.599
(1.530)
-0.203
(0.167)
0.002
(0.010)
0.043
(0.029)
0.143
(0.226)
0.040
(0.334)
-0.203
(0.166)
0.001
(0.010)
0.038
(0.029)
0.124
(0.225)
0.063
(0.330)
-0.174
(0.164)
-0.002
(0.010)
0.042
(0.029)
0.158
(0.224)
0.038
(0.327)
-0.206
(0.163)
-0.005
(0.010)
0.029
(0.028)
0.090
(0.221)
-0.011
(0.317)
-0.224
(0.167)
0.001
(0.010)
0.035
(0.029)
0.124
(0.228)
0.096
(0.333)
RPI
0.133
(0.031)***
0.136
(0.031)***
0.129
(0.030)***
0.131
(0.030)***
0.142
(0.031)***
SPORT
0.362
(0.160)**
0.364
(0.160)**
0.395
(0.157)**
0.383
(0.156)**
0.391
(0.161)**
CELLYEAR
0.028
(0.027)
0.024
(0.026)
0.018
(0.026)
0.007
(0.025)
0.025
(0.027)
CELLBILL
0.236E-3
(0.105E-3)**
SEX
AGE
EDU
MARRIED
JOB
INCOME
0.220E-3
0.232E-3
(0.105E-3)** (0.103E-3)**
-0.191E-5
(0.215E-5)
-30.501
(6.861)***
Log(INCOME)
Log(1-PRICE/I
NCOME)
PRICE
0.190E-3
0.223E-3
(0.102E-3)** (0.107E-3)***
0.016
(0.120)
22.701
(5.834)***
-0.203E-3
-0.203E-3
(0.329E-4)*** (0.328E-4)***
-0.388
(0.063)***
Log(PRICE)
Log(INCOMEPRICE)
Model χ2
Correct
predictions (%)
McFadden R2
AIC
SIC
30.195
(6.793)***
76.692***
75.866***
57.606***
52.242***
77.410***
78.535
78.282
75.758
75.253
77.273
0.168
1.019
451.107
0.166
1.015
445.955
0.126
1.067
470.193
0.114
1.075
469.575
0.169
1.017
450.389
Sample size
396
396
396
396
396
Note: The numbers in parentheses are the standard errors of the estimates. ***, **, and *
denote significance at the 1%, 5%, and 10% confidence levels, respectively.
25
Table 4. Willingness-to-pay for relocating MPBSs
Willingness to Pay
Mean
(S.D.)
NT$
5,657
(2,279)
Percentile
10 th
25th
Median
75th
90th
2,462
4,025
5,864
7,275
8,441
26
Table 5. Predicted WTP for respondents of some particular profiles
Profile
Variables
RPI=1
RPI=5
CELLBILL=300 CELLBILL=2000
CELLBILL=300 CELLBILL=2000
Socio-demographics Variables
AGE = 25
3,366.24
AGE = 60
3,539.46
5,208.84
5,382.05
6,043.28
6,216.49
7,885.88
8,059.09
5,850.89
4,707.93
,
4,553.07
6,685.33
5,542.37
8,527.93
7,384.97
5,873.68
5,387.52
6,708.13
7,230.11
8,550.72
SEX=0
SEX=1
4,008.29
2,865.34
EDU=9
EDU=16
2,710.48
4,031.09
JOB = 0
JOB = 1
3,146.70
3,458.65
4,989.29
5,301.25
5,823.73
6,135.69
7,666.33
7,978.28
MARRIED=0
3,005.97
4,848.56
5,683.00
7,525.60
MARRIED=1
3,615.31
5,457.90
6,292.35
8,134.94
5,566.87
6,528.60
7,409.46
8,371.19
Mobile Phone Use Condition Variables
CELLYEAR=2
CELLYEAR=10
2,889.83
3,851.56
4,732.43
5,694.16
Attitude about Health Variable
SPORT=0
2,374.28
4,216.87
5,051.31
6,893.91
SPORT=1
4,164.77
6,007.36
6,841.80
8,684.40
a. Mean values were used for all variables except RPI, CELLBILL and another
explanatory variable to estimate individual WTP.
27
1
0.9
proportion of acceptance
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
1000
2000
3000
4000
5000
6000
7000
8000
offered price
Figure 1. The empirical survival function
28
9000
10000
11000