Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

WATER RESOURCES RESEARCH, VOL. 42, W12407, doi:10.

1029/2005WR004833, 2006

Water shortages and implied water quality:


A contingent valuation study
Margarita Genius1 and Konstantinos P. Tsagarakis1
Received 22 December 2005; revised 19 August 2006; accepted 25 August 2006; published 14 December 2006.
[1] This paper analyses the extent to which households in an urban area are willing to pay
to ensure a fully reliable water supply when the latter induces changes in drinking water
quality. The water supply system in the city of Heraklion, Greece, is characterized by
periodic water rationing, which is more pronounced in the summer months. The
generalized use of cisterns and even water tanks helps residents cope with quantity
shortages but has a negative effect on the quality of the water reaching their taps. The
results of our contingent valuation show that respondents not affected by shortages and
already drinking tap water have a smaller willingness to pay, while positive perceptions on
quality have a positive effect.
Citation: Genius, M., and K. P. Tsagarakis (2006), Water shortages and implied water quality: A contingent valuation study,
Water Resour. Res., 42, W12407, doi:10.1029/2005WR004833.

1. Introduction claims that the quality is good. The main explanation for
[2] The provision of a reliable water supply network this phenomenon is that the water offered by the water
constitutes one of the main objectives of municipal water company (at the source) is not of the same quality as the
authorities who face the problem of matching a seasonal water arriving in consumers’ taps as water supply disrup-
dependent water demand with a supply that is affected by tions introduce soil in the water supply and subsequently
weather conditions. Whenever the water management policy water sits in the water tanks. The situation is very different
entails the avoidance of water shortages, the water capacity in other Greek cities which are not exposed to periodical
needs to be sufficiently developed so as to satisfy demand in rationing. In a recent report about the quality of drinking
periods of high demand or periods of drought. According to water in selected European cities, almost all survey respond-
Griffin and Mjelde [2000] such a policy might not be optimal ents in Athens claimed to drinking tap water [Scientific
if the costs of ensuring full reliability are high and a policy of Committee on Toxicity, Ecotoxicity and the Environment,
periodic shortfalls might be superior. 2003].
[3] The case of Heraklion constitutes an example where [4] The objective of the present study is to analyze
periodic water supply interruptions have come to be an whether residential water users in Heraklion would be
integral part of the everyday life of the city dwellers. Water willing to pay higher water bills to ensure a continuous
supply cutoffs are very frequent and last longer in the flow of water when the implications of the latter are
summer months, when demand is high due to the tourist twofold. On the one hand, the probability of running out
season but they are not uncommon during the rest of the of water can, in principle, be reduced to zero while on the
year. When considering the costs necessary to expand the other the water quality will not be deteriorated between
water supply to a fully reliable level one should take into source (water company) and the consumer’s tap. Our results
account as well the private costs to dwellers of any invest- show that households would indeed welcome an uninter-
ments they make to face the shortages. In Heraklion most rupted water service, moreover, willingness to pay is
households have cisterns or even water tanks (a cistern is a predicted to be higher for those households suffering water
prefabricated tank of a standard size made of galvanized shortages and resorting to bottled water. Negative experi-
metal or plastic placed on the roofs or in the basement of a ence with water appearance and/or smell plays an important
building, while water tanks are made from reinforced role both to explain the decision to drink tap water and the
concrete normally placed underground) that allow them to amounts people are willing to pay and favorable water
smooth their water consumption in the face of possible quality perceptions have a positive effect on WTP.
service interruptions, therefore supply disruptions effectu-
ated by the municipal authority do not always materialize in 2. Water Supply in Heraklion City
water shortages for the households. However, their occur- [5] Heraklion City is the fourth largest in Greece with a
rence has an immediate effect on the consumption of tap population of 140,000 during the winter and 160,000 during
water as drinking water since many households are reluctant the summer. The city is located in the northern part of the
to drink tap water and this despite the water managers’ island of Crete. Water supply and wastewater treatment is
managed by the Municipal Enterprise of Water Supply and
1
Department of Economics, University of Crete, Rethymno, Crete, Sewerage (MEWSS) of Heraklion which is owned by the
Greece. municipality and serves 46,300 customers.
[6] Water supply is achieved through a potable supply
Copyright 2006 by the American Geophysical Union.
0043-1397/06/2005WR004833 network that covers almost 100% of households, including

W12407 1 of 8
W12407 GENIUS AND TSAGARAKIS: WTP FOR QUALITY-INDUCED IMPROVEMENTS W12407

some areas outside the municipality’s borders. Water comes (12.8 mg/L), NO 2 (0 mg/L), Mg
+2
(38 mg/L), pH varied
from about 56 wells and springs in the wider area. The main from 7.2 to 8.
source of water is from wells, which are drilled from 100 to [9] Although arsenic concentrations in the water supply
400 m deep, where water is not influenced by surface are of great concern in both developing and developed
runoffs. Also, there are no heavy industries in the area that countries [Nickson et al., 2005; Shaw et al., 2005], the
could influence the underground water quality. These sour- analysis conducted by Heraklion’s MEWSS showed that it
ces are selectively used throughout the year as water quality was not detectable. However, higher concentrations of
may significantly vary especially when underground sour- Chloride (Cl) and Sodium (Na+) were reported in some
ces are overexploited. The various waters are mixed and cases. Chloride concentration in two of the distribution
chlorinated in seven points before entering the distribution sources was above 250 mg/L, which is the maximum
network. Network losses are estimated at 35% and illegal indicated value from the 83/98/EC directive. These values
connections losses are estimated at 3% making a total of ranged from 195 to 398 mg/L and 42.6 to 270 mg/L for the
38%. Water consumption to domestic, industry, hotels and first and the second source respectively. The first distribu-
other uses (commercial shops, cafe, restaurants, schools, tion source also exceeded the recommended 200 mg/L
hospitals, etc.) is 72.8%, 1.3%, 1.6% and 24.3% respectively. concentration of Sodium which was reported in the 98.5 –
Average water consumption per inhabitant is estimated at 230 mg/L range. This is due to the fact that the aquifer is
153 L/d. close to the sea and wells inevitably drop the groundwater
[7] Water supply interruptions result from both the fact level causing sea intrusion that increases the salinity of the
that water demand exceeds water supply, especially during groundwater. This is the reason for observing high concen-
the summer months and that the municipality’s storage trations of Cl and Na+.
tanks are not sufficient to cover demand during peak [10] The above figures indicate the absence of known
periods. As a response to the present water deficit, most health risks coming from drinking tap water. This is further
households have at least a 1 m3 cistern or larger built water supported by the fact that the local press and the media have
tanks where water is stored to provide water in times when not reported such concerns either.
supply interruptions are in effect. These tanks’ water ca- [11] In spite of the seemingly satisfactory quality indica-
pacity allows to cover at least a day’s demand. Water supply tors, water reaching households’ taps is often looked upon
may be continuous for some days during the winter, and with suspicion since its quality is altered during the distri-
switch to one day supply – one day lack of supply during the bution process as a consequence of supply interruptions.
summer. Therefore the maximum period without water Because of water network construction deficiencies, when
supply can be 24 hrs. During the current year the construc- supply is on, water exits from the pipes to the soil. When
tion of a dam (30 Million m3 total capacity) at Aposelemis water supply stops, pipes are emptied. This makes some
ephemeral river (20 km east of Heraklion city) has started water come back into the pipes sweeping soil which sedi-
which will provide potable water for the wider area includ- ments. This soil is carried to the households’ water tanks
ing the city of Heraklion. when the next supply starts. Then soil settles down in the
tank and any turbulence of incoming or outgoing water
3. Water Quality and the Effect of Water keeps particles in suspension, leading it to the consumers’
Supply Deficiencies tap. While the water company tries to keep water supply and
interruptions evenly distributed to all regions of Heraklion,
[8] Water quality standards are given by the 83/98/EC all households are not affected in the same way since houses
Drinking Water Directive [European Union, 1998] which located at higher altitudes of the city will be the first that
has been adopted by the Greek legislation. The municipality will experience interruptions in the water supply compared
has made available to us data related to water analyses from to those in lower altitudes.
7 different distribution sources where 124 samples per year
are analyzed. According to this data the parametric values
set by the Part A and B of the first annex of the directive 4. Valuing an Improvement in Water
were not exceeded. More specifically, all elements mea- Supply Services
sured (F, Fe+2, Cu+2, Mn+2, Zn+2, Pb+2, Cr+3, Cd+2, Ni+2) [12] Tap water is used for many different purposes in
were found in concentrations of approximately 10– 40 times today’s households’ activities and the spectrum of its uses
less than the maximum allowable or not detectable (Cd+2, widens with the rising of living standards. Tap water is used
Ni+2, pesticides) or zero values (TC, FC, fecal Streptococci, for bathing, washing, watering plants, drinking, watering
Pseudomonas). The concentration of trihalomethanes lawns, filling up pools and a myriad of other activities that
(THMs) in drinking water has been studied in various cities expand as a country’s development level increases. When
of Greece [Golfinopoulos, 2000; Golfinopoulos and facing water shortages, households are going to make
Nikolaou, 2005] and was found to be less than 56 mg/L choices as to what water needs or wants are going to be
(100 is the maximum allowable) for the city of Heraklion, satisfied first and it is conceivable that they do not value in
which is among the lowest observed levels in many cities of the same way water used for drinking as water used for
the developed world [Kampioti and Stephanou, 2002]. watering plants. In addition, the alternative uses of water
Residual chlorine was 0.05 – 0.4 mg/L. Considerably lower imply different demands on the characteristics of water,
than the indicated upper bounds set by Annex 1, Part C of water quality is important when it comes to drinking water
the directive, were the concentrations found for the follow- but it is less important when it comes to other uses. Since
ing parameters (maximum values reported in parenthesis): the value households assign to an improvement of the water
Special conductivity (1700 mS/cm), SO2 4 (110 mg/L), NO3

supply depends on the level of the current service, it is
important to know how the latter is related to the current
2 of 8
W12407 GENIUS AND TSAGARAKIS: WTP FOR QUALITY-INDUCED IMPROVEMENTS W12407

uses households make and also how they perceive the ces for drinking water, quality perceptions and socioeco-
different characteristics of water currently available to them. nomic characteristics have an impact on the value
[13] One approach followed in the literature is the avert- households place on the amelioration of the water supply
ing behavior approach whereas households’ incur in certain system.
expenditures in order to cope with deficiencies in the water
supply. For instance, Pattanayak et al. [2005] consider five 5. Elicitation Format, Survey Design,
types of averting expenditures related to the collection, and Data Description
pumping, treating, storing and purchasing of water in
Katmandu, Nepal, and find that these costs are lower than [16] An important element of a contingent valuation
the estimated WTP for improved services that they derived study is the elicitation format. It has long been recognized
in the same study. Um et al. [2002] use the averting in the contingent valuation literature that the adopted
behavior method augmented by a perception measure to elicitation format could affect the estimates of mean will-
derive estimates of willingness to pay for safe tap water in ingness to pay [see Cameron and Quiggin, 1994; Alberini et
Korea. Indeed they claim that it is perceived pollution al., 1997]. While the single bound format has been identi-
levels and not actual pollution levels that explain averting fied as being incentive compatible, under certain assump-
behavior and therefore they use households’ perceptions to tions, and is the format proposed in the guidelines set forth
explain whether averting actions are undertaken. by the NOAA panel [Arrow et al., 1993], it requires larger
[14] Another approach is the application of the contingent sample sizes than the double bound format to achieve the
valuation method which has been extensively used for same efficiency levels. In effect, the introduction of a
valuing nonmarket resources. The contingent valuation follow-up question conveys more information about the
method is a technique used to elicit values for nonmarket WTP distribution while it does not substantially increase
goods based on the survey responses to hypothetic envi- the cost of conducting the survey. While the double bound
ronmental scenarios whereas the respondents are asked to has become the most widely used elicitation method, several
state their willingness to pay for a change in the environ- studies have reported anomalies introduced by its iterative
mental good. Several studies have been conducted in nature which put into question the reliability of the
developing and developed countries to estimate the WTP responses to the second question. DeShazo [2002] considers
for improvements in drinking water quality and/or reliabil- asymmetric elicitation effects in the double bound format
ity. In developed countries, studies related to water quality depending on whether the elicitation sequence is ascending
have focused more on health risk reduction scenarios [see (second bid is higher than first) or descending (second bid is
Carson and Mitchell, 2006] while studies dealing with lower than first). Borrowing arguments from prospect
water supply reliability have often focused on willingness theory he shows that responses to ascending sequences
to pay to reduce the frequency and/or duration of shortages. are subject to framing effects whereas respondents who
For instance, Howe et al. [1994] analyze WTP for higher accept the first bid form a reference point and view the
levels of reliability (and WTA for lower levels of reliability) second bid as a loss. Therefore framing effects result in a
in three Colorado towns and find that WTP increases with higher number of negative responses to the second question.
the level of reliability offered. In addition, WTP increases as Since responses to descending sequences are not subject to
one moves from the high-reliability town to the low- framing effects, he recommends that follow-up questions
reliability towns, however the aggregate WTP does not should be restricted to ‘‘those that monotonically decrease
compensate for the higher costs of a more reliable water in value’’ [DeShazo, 2002] or else use the single bound
system while only in the higher-reliability town aggregate format.
WTA could be covered by the savings derived from a less [17] Given the above considerations, we adopt DeShazo’s
reliable system. Hensher et al. [2005] conduct choice recommendation and ask a follow-up question only to those
experiments between different levels of shortages at differ- respondents who had negative answers to the first bid.
ent prices for the water bill and report that both frequency [18] Data were collected through face-to-face interviews
and length of water interruptions are important to customers in the summer of 2004. At the beginning of the interview
while their WTP decreases with each successive water the respondent was informed that he/she had been randomly
interruption. Barakat and Chamberlin, Inc. [1994] consider selected to participate in a research conducted by University
the total elimination of specific types of future shortages and of Crete and the National Agricultural Research Foundation.
report that WTP increases with increasing duration and It was stated that the objective of the research was to
frequency of shortages while duration seems to be a more investigate consumers’ evaluation of potable water quality
important determinant of WTP. In developing countries coming from the potable water network. It was also men-
the major focus has been on inferring willingness to pay tioned that any answers would be kept confidential, as there
for connections to the public water network or for access was no identity data required. Before starting the interview
to public sources of water, [see, e.g., McPhail, 1994; the respondent was asked to find the last available water bill
Whittington et al., 1990]. and have it in hand. This was necessary in order to recover
[15] In the present study, the proposed improvement in information such as water consumption and water bill paid
the water supply consists in the provision of a continuous by the household. The first part of the interview included
water flow which, as explained above, affects both the questions to learn about the respondent’s water saving
quantity of water available to the consumers and the quality activities, the frequency of water shortages and the severity
of the water running through their taps. We use a variant of of its effects, the perceptions about water quality and water
the double bound contingent valuation model while we drinking behavior. A brief explanation about the nature and
focus on whether experience with shortages, different sour- reasons for water interruptions and its effects on water
quality was presented followed by a possible program that
3 of 8
W12407 GENIUS AND TSAGARAKIS: WTP FOR QUALITY-INDUCED IMPROVEMENTS W12407

Table 1. Description of Variables


Variable Name Description

Age respondent’s age


Female 1: female, 0: male
Numinf number of infants (below 2 years old) in the household
Famsize total number of household members
Garden 1: garden, 0: no garden
Sqmhse square meters of house
Educ1 1: high school not completed, 0: otherwise
Educ2 1: high school completed but not beyond, 0: otherwise
Inc1 1: monthly family income below or equal to ¼1500,
C 0: otherwise
Inc2 1: ¼1500
C < monthly family income C ¼3000, 0: otherwise
Herak number of years living in Heraklion
Tank 1: water tank or cistern present, 0: otherwise
Filt 1: some filter used, 0: otherwise
Bottle 1: drink bottled water, 0: otherwise
Bill quarterly total water bill per person in ¼
C
Cook 1: use tap water to cook most of the time, 0: otherwise
Watcolor 1: noticed blurred tap water color many times or always, 0: otherwise
Watsmell 1: noticed chlorine smell in tap water many times or always, 0: otherwise
Percqual1 1: perceived residential water quality is good or very good, 0: otherwise
Percqual2 1: perceived residential water quality is average, 0: otherwise
Watshopba 1: if water shortage problem is very or extremely important, 0: otherwise
Watqualpba 1: if water quality problem is very or extremely important, 0: otherwise
Savein 1: if take actions to save water inside the house, 0: otherwise
Nowat 1: has run out of water sometime in the past year, 0: otherwise
Once 1: has run out of water just once in the past year, 0: otherwise
2 to 6 1: has run out of water 2 to 6 times in the past year, 0: otherwise
7 to 12 1: has run out of water 7 to 12 times in the past year, 0: otherwise
>12 1: has run out of water more than 12 times in the past year, 0: otherwise
Shoeff 1: shortage effects on family are important or even serious, 0: otherwise
Drink 1: drink tap water, 0: otherwise
a
This question refers to a general context where respondents also evaluated problems of crime, unemployment, pollution,
traffic, etc.

could avoid them. The proposed program comprised the resulting increase in the bill was than if the bid had the form
investment in new sources of groundwater and the improve- of a fixed Euro amount per cubic meter.
ment of the pipe network system to reduce leakages. [21] Several pilot studies were implemented to determine
[19] The respondents were told that the cost of imple- the offered bids and to polish the questionnaire. Two
menting such a program would have to be covered by all versions of the questionnaire were randomly assigned to
residents connected to the water network in the same way as different respondents and the corresponding initial bids
connection to the sewerage network is being currently were 0.20 and 0.40 respectively. If the response to the first
charged, i.e., through a percentage increase in the water bid was yes there was no further question but if the answer
bill. They were also told that their answers, though confi- was no a second lower bid was offered (0.10 and 0.20
dential, could play an advisory role as to the acceptability respectively). Respondents who answered no-no to both
and economic feasibility of the program. The willingness to bids were asked what was the maximum amount they were
pay question was posed as follows: ‘‘Assume that it is willing to pay and those reporting a zero value were asked a
possible to ensure a continuous (24 hour) water supply, so follow-up question to determine whether they were real
that the water quality between the source (municipal au- zeros.
thority) and your tap remains stable, would you be willing [22] A sample of 440 households was selected and a
to pay an extra x% of your current bills for this to happen?’’ maximum of three trials was set for those households that
The last available bill was used for illustration purposes to could not be contacted on the first visit. In total 294
show respondents what the extra x% meant in absolute questionnaires were completed since 65 refused to partici-
terms, but it was made clear to respondents that the pate and 81 kept postponing the appointment over the set
percentage increase would affect all ensuing bills. limit or were not found in their houses in any visit. From the
[20] We opted for bids expressed as percentages of the 294 questionnaires some were disregarded since respond-
bill for different reasons. On the one hand, stated prices ents either failed to respond to some crucial question, there
should bear some resemblance to what respondents’ actu- were errors or the debriefing question identified them as real
ally experience [see Hensher et al., 2005] where increases in zeros (9 respondents) leaving us with 281 questionnaires.
the water bill are expressed as a percentage of the annual Table 1 describes the variables used in the analysis
estimated bill, Schläpfer and Schmitt [2005] where bids are which can be grouped according to (1) socioeconomic
expressed as a percentage increase in taxes, second most characteristics (respondent’s age, gender, family composi-
additional fees charged by the water company are calculated tion, income, education level, number of years residing in
as percentages of the bill and finally, we felt that it was Heraklion), (2) water use (garden, square meters of house,
easier for the interviewer to explain and compute what the bill per person in the household, whether actions are taken
4 of 8
W12407 GENIUS AND TSAGARAKIS: WTP FOR QUALITY-INDUCED IMPROVEMENTS W12407

to save water usage in the house (savein); information was Table 2. Descriptive Statistics: Means and Standard Deviationsa
also collected on whether tap water, bottled water or other
water is used for drinking and whether tap water is used for Tap Drinkers: Nondrinkers: All:
n = 42 n = 239 n = 281
cooking, (3) water quantity characteristics (whether the
household has suffered water shortages in the past year Age 42.38 (17.04) 42.70 (15.4) 42.64 (15.62)
(nowat) and whether he feels shortages have important Female 0.52 (0.51) 0.58 (0.50) 0.56 (0.50)
Numinf 0.05 (0.22) 0.15 (0.38) 0.13 (0.36)
effects on family (shoeff), presence of a water tank (tank)), Famsize 2.86 (1.46) 3.04 (1.25) 3.01 (1.28)
(4) water quality characteristics (problems with color Garden 0.19 (0.40) 0.27 (0.44) 0.26 (0.44)
(watcolor), with smell (watsmell), presence of filter), and Sqmhse 85.31 (24.62) 94.64 (28.59) 93.25 (28.20)
(5) respondent’s perceptions about quality (percqual1, Educ1 0.21 (0.42) 0.27 (0.44) 0.26 (0.44)
percqual2; respondents were asked to rate the quality of Educ2 0.26 (0.45) 0.26 (0.44) 0.26 (0.44)
Inc1 0.33 (0.48) 0.36 (0.48) 0.36 (0.48)
potable water in Heraklion from very good to very bad on a Inc2 0.60 (0.50) 0.54 (0.50) 0.54 (0.50)
5 point scale). Herak 27.79 (20.44) 26.56 (19.16) 26.74(19.33)
[23] Out of 281 respondents only 42 (around 15%) Tank 0.98 (0.15) 0.97 (0.16) 0.97 (0.17)
reported drinking tap water, always or always except on Filt 0.98 (0.15) 0.16 (0.36) 0.28 (0.45)
Bottle — 0.85 (0.36) 0.72 (0.45)
rare occasions, 72% buy bottled water, 13% get water from Bill 13.7 (7.34) 12.37 (5.70) 12.57 (5.98)
third sources, namely, 8% use water from a spring or well Cook 1 0.69 (0.47) 0.73 (0.44)
and 5% get drinking water from another municipality. Some Watcolor 0.10 (0.30) 0.37 (0.48) 0.33 (0.47)
73% of the included households use tap water for cooking. Watsmell 0.14 (0.30) 0.66 (0.48) 0.58 (0.49)
[24] Descriptive statistics for all variables considered are Percqual1 0.07 (0.26) 0.02 (0.140 0.03 (0.17)
Percqual2 0.64 (0.49) 0.36 (0.480 0.41 (0.49)
reported in Table 2, both for the total sample and for the Watshopb 0.88 (0.33) 0.91 (0.29) 0.90 (0.30)
subgroups of tap water drinkers and non drinkers. It is worth Watqualpb 0.79 (0.41) 0.95 (0.22) 0.93 (0.26)
noting that 97% of the interviewed households have a water Savein 0.48 (0.51) 0.55 (0.60) 0.54 (0.59)
tank or cistern which provides them with some protection Nowat 0.57 (0.50) 0.41 (0.49) 0.44 (0.50)
Once 0.19 (0.40) 0.07 (0.27) 0.09 (0.29)
against water shortages. As far as socioeconomic variables 2 to 6 0.26 (0.45) 0.12 (0.32) 0.14 (0.35)
are concerned it is worth noting that the percentage of 7 to 12 0 0.07 (0.26) 0.06 (0.24)
households with infants is smaller for tap drinkers while >12 0.12 (0.33) 0.15 (0.36) 0.15 (0.35)
their education level seems to be higher. It is also worth Shoeff 0.31 (0.47) 0.51 (0.50) 0.48 (0.50)
pointing out that the percentage of households who report a
Standard deviations are given in parentheses.
problems with the water appearance and/or smell is much
smaller for drinkers. This could be because (1) sensitivity to
color and smell might not be homogenous or (2) 98% of
households who drink tap water use a filtering device, equation (1). The underlying willingness to pay is a latent
mostly a filter attached to the tap or water tank with some variable that depends linearly on a set of regressors and an
households even boiling the water before they drink it error term given by su with mean zero and variance s2
(while only 16% of nondrinkers use filters) or, finally, while what is observed is the YES/NO answer to the
(3) we suspect that the quality of the supplied water is not different bids. Some of the possible determinants of the
the same in all neighborhoods due to different groundwater dependent variable are related to socioeconomic character-
sources and different ages of the pipe network system. istics, experience with shortages, water quality and also to
These facts could explain as well why 71% of respondent’s drinking behavior.
who drink tap water perceive the quality as being average or
very good while this number falls to 38% for the rest. We WTPi ¼ Xi0 b þ dDi þ sui i ¼ 1; . . . ; n ð1Þ
also note the percentage of tap drinkers who have suffered a
water shortage is higher than that of the rest while they seem where WTPi is respondent’s i underlying willingness to pay
to be less negatively affected by the shortages. Because we for the improvement, Xi is a vector of explanatory variables
do not have data on the duration of each shortage and the for respondent i that affect willingness to pay, Di takes the
characteristics of water tanks, we cannot conclude that the value 1 if the household uses tap water as drinking water
types of shortages incurred by the households are uniform and 0 otherwise, and ui  N(0,1).
for all respondents. For instance, some households have [26] Equation (1) can be readily estimated by maximum
underground cement water tanks and a small water cistern likelihood using the observed responses, the presented bids
on the roof and then water is pumped from the tank to the and the normality assumption to compute the three under-
cistern. Since electricity shortages are not uncommon in the lying probabilities: P(yes), P(no,yes) and P(no,no). How-
island one could have the paradoxical situation of having a ever, the estimation of equation (1) is complicated by the
full tank but being unable to pump the water so that it is presence of the dummy variable Di. The tap drinking
distributed to all faucets. Other households just have a dummy variable separates respondents into two treatment
cistern on the roof and their ability to store water is more groups: tap water drinkers and non drinkers. As is the case
limited. in the treatment effects literature [see, e.g., Greene, 2003]
care has to be taken since respondents have not been
6. Econometric Specification randomly assigned but have self-selected themselves into
[25] The econometric model to analyze the responses to the two treatments and these endogenous treatment effects
the present CVM survey is the censored regression model could lead to selectivity bias. The latter might arise in the
proposed by Cameron and James [1987] and given by present context when unobserved factors that affect the

5 of 8
W12407 GENIUS AND TSAGARAKIS: WTP FOR QUALITY-INDUCED IMPROVEMENTS W12407

probability of drinking tap water, affect the willingness to the number of household members and the number of liters
pay as well. For instance if it were the case that tap drinkers during a two month period to arrive at the final cost. The
are more ‘‘conformist’’ and less sensitive to quality than drinking water consumption is assumed to be the same for
other respondents they would display a lower willingness to all respondents and household members and was estimated
pay whether or not they actually drank tap water. One using the daily drinking water requirements per person
possible solution to the endogenous treatment effects prob- reported by FAO [World Health Organization, 2000].
lem is to jointly estimate model (1) and a model that Although the specific liter amount is arbitrary, we should
explains the choice of source for drinking water while note that it will only affect the magnitude of the cost
allowing for correlation of the errors of both equations. parameter estimate but not its significance.
This is the approach we undertake in the present analysis. [30] We further make the assumption that both errors
[27] The model explaining the decision to drink tap (u and e) follow a bivariate normal distribution,
water is based on the random utility model (RUM) [see    
McFadden, 1974], whereas a consumer facing J alternatives 0 1 r
ðui ; ei Þ  BVN ; :
0 r 1
will choose that alternative that maximizes his/her utility
given a budget constraint. The RUM specification adopted
below is closely related to that of Shaw et al. [2005] who [31] If r is equal to zero there is no selectivity bias
model choice of drinking water source but in a context problem and equation (1) can be estimated independently
where health risks are present due to arsenic contamination. of equation (5) otherwise the two equations need to be
In the present case we have two alternatives corresponding estimated jointly and the log likelihood in this case is given
to drinking tap water or not drinking it, where the second by,
option represents a costly alternative to the first. Therefore,  
along with individual characteristics, the cost of both alter- X ti1 þ Xi0 b
ln Lðb; d; g; sÞ ¼ ln F Zi0 g; ; r
natives will be an important determinant of drinking Di ¼0
s
behavior. Denoting by UDi and UNDi the derived utility Ii1 ¼1
X  
from drinking tap water and not drinking tap water respec- ti2  Xi0 b
tively, we have, þ ln F Zi0 g; ;r
Di ¼0
s
Ii1 ¼0
Ii2 ¼0
UDi ¼ aD þ bðYi  CDi Þ þ Wi0 d D þ Di  
ð2Þ X ti1  Xi0 b
UNDi ¼ aND þ bðYi  CNDi Þ þ Wi0 d ND þ NDi þ ln F Zi0 g; ;r
Di ¼0
s
Ii1 ¼0
where Yi is income of i, CDi is the cost of tap water as faced Ii2 ¼1
by i, CNDi is the cost of the alternative as faced by i, Wi is  
ti2  Xi0 b
characteristics of respondent i that do not change across  ln F Zi0 g; ;r
s
alternatives, and uDi and uNDi are components of utility not   0  
X t i1 þ Xi b þ d
observed by the researcher. þ 0
ln F Zi g; ;r
[28] The consumer will choose tap water as a drinking Di ¼1
s
Ii1 ¼1
source if   0  
X
0 ti2  Xi b þ d
þ ln F Zi g; ; r
Ui ¼ UDi  UNDi > 0 ð3Þ Di ¼1
s
Ii1 ¼0
Ii2 ¼0
or    
X ti1  Xi0 b þ d
þ ln F Zi0 g; ; r
Ui* ¼ ðaD  aND Þ þ b ðCNDi  CDi Þ þ Wi0 ðd D  d ND Þ þ ei Di ¼1
s
Ii1 ¼0
¼ Zi0 g þ ei > 0 ð4Þ Ii2 ¼1
 
 
ti2  Xi0 b þ d
 ln F Zi0 g; ; r ð6Þ
Therefore the resulting model for choice of drinking water s
can be written as
  where F is the cdf of the standard bivariate normal, ti1 and
Di ¼ I Zi0 g þ ei i ¼ 1; . . . ; n ð5Þ
ti2 correspond to the bids offered (ti1 > ti2), and Ii1 and Ii2
where I(.) is an indicator function that takes the value 1 if take the value 1 if the answer is yes to the first, second bid
the argument inside the brackets is positive and 0 otherwise, respectively and zero otherwise.
Zi is a vector of explanatory variables that include
individual characteristics and choice attributes such as cost, 7. Results
and ei  N(0,1).
[32] Estimation results from (6) reveal that the null
[29] In order to estimate equation (5) we have to make
hypothesis that r is equal to zero cannot be rejected at
some assumptions about the cost of the two alternatives.
any usual significance level (5%, 10%) using the likelihood
The cost of tap water is assumed to be zero while for the
ratio test, implying that the equations for drinking and
cost of the alternative source we use the average price per
willingness to pay can be estimated separately, and these
liter of bottled water in local supermarkets for bottle
are the results reported in Table 3.
drinkers, and the local lowest price per liter of bottled water
[33] As far as the decision to drink tap water is concerned,
for the rest. The aforementioned prices were multiplied by
the estimated parameters have the expected signs. More

6 of 8
W12407 GENIUS AND TSAGARAKIS: WTP FOR QUALITY-INDUCED IMPROVEMENTS W12407

Table 3. Estimation Results (t-Statistics in Parentheses) (from carrying bottles to boiling water and so on). The
quadratic term for age implies that willingness to pay
Variable Drink Equation WTP Equation increases with age up to a certain point (50 years) and then
Constant 1.227 (2.414) a
0.022 (0.140) it decreases which could be explained by the reproductive
Cost 0.010 (2.006)a cycle of the family (presence of young children in the
Filter 2.670 (5.842)a household) and a turning point in expected future earnings.
Herak 0.018 (2.633)a [35] As far as the variables related to drinking behavior
Female 0.068 (2.372)a
Savein 0.048 (1.857)b and water quality are concerned, our findings point out that
Sqmhse 0.002 (2.902)a those households who already drink tap water place less
Bill 0.009 (2.536)a value on the proposed improvement than bottle drinkers
Watshopb 0.122 (2.669)a (who bid the highest values) and drinkers from third
Garden 0.101 (2.811)a
Educ2 0.099 (3.012)a sources. More specifically the difference in the percentage
Inc1 0.168 (2.177)a bids for bottle drinkers and tap drinkers is around 20%. It
Inc2 0.156 (2.215)a could be the case that tap water drinkers regard the current
Bottle 0.087 (1.922)b level of water quality as satisfactory and place a smaller
Famsize 0.047 (2.059)a
Watcolor 0.568 (1.304) 0.130 (3.589)a
value in the incremental quality brought about by the
Watsmell 0.716 (2.083)a proposed plan. Bottle drinkers incur higher costs in order
Age 0.022 (3.228)a to acquire drinking water and therefore place a higher value
Agesquared/100 0.022 (3.195)a on the proposed change. It should be mentioned at this point
Numinf 0.114 (2.314)a that the variable for filter was not significant in the WTP
Percqual1 0.110 (1.673)b
Drink 0.114 (2.277)a equation. Those households who report problems with the
Nowat 0.060 (1.917)b appearance of water (watcolor) are on average less willing
s 0.168 (10.137)a to pay for the proposed plan and this could be due to the fact
Log likelihood 84.71 191.128 that they are skeptical that the high-quality scenario will
a
Significance at the 5% level. also apply to them or is attainable. On the other hand, those
b
Significance at the 10% level. respondents who perceive water quality as being good or
very good are willing to pay more on average than those
who perceive the current quality from average to bad. This
specifically, the results show that households who use a result could seem counterintuitive but in fact, quality
filter are more likely to drink tap water, households who perceptions reported by respondents refer to the water
have reported problems with smell or color are less likely to supplied by the municipal company which might differ
drink tap water (although watcolor is not significant at the from water reaching their taps for the reasons mentioned
10% level) and those who have resided in Heraklion for a in the introduction. In addition, perceptions about water
longer period of time are less likely to drink tap water. quality could be based on a wider information set than the
Possible explanations for the last finding could be related to respondents’ individual experience from his/her tap.
the fact that long-term residents are more exposed to [36] Estimates of mean willingness to pay are given in
different channels of information and ‘‘neighborhood talk’’ Table 4, on average respondents are willing to pay an
compared to newcomers and also that for the latter, tap additional 42.6% of their 3-month bill for the proposed plan
water might have been the main source of drinking water in ¼ 13.8). The mean willingness to pay was also estimated
(C
their previous location of residence. As far as the cost using the first bid only, namely the single bound model, and it
variable is concerned, it has a positive sign as expected was found to be higher and around 46% or ¼ C 15.1. Taking
and is significant at the 5% level. Therefore an increase in into consideration that this sum is not a once-and-for-all
the cost of bottled water would lead to an increase in the payment but would be levied on all future bills, we can
probability of drinking tap water. Because it is possible that conclude that households in the sample value the proposed
the use of a filter is endogenous in the drinking decision we plan quite positively. Although the size of the sample is small,
used the Rivers and Vuong [1988] two-step procedure to test local authorities can find this figure useful in their water
for the exogeneity of filter and we failed to reject the null management activities once they compute the cost of exploit-
hypothesis of exogeneity. ing additional sources of groundwater. Raising water fees has
[34] For the willingness to pay equation the signs for often encountered strong opposition in Greek municipalities
income are the ones expected from economic theory, and fear of retaliation in the following elections affects the
namely low income people are less willing to pay for an timing and occurrence of such measures.
improved water supply (the highest income group is the
baseline category). Households being adversely affected by 8. Conclusions
water interruptions (nowat) are willing to pay more
[37] In situations where deficiencies in the quantity di-
although it is significant at the 10% level but not at the
mension of the supply of water affect the quality dimension
5%. The opposite is true for those households who engage
in water saving activities (savein) and those who are heavier
users of tap water as reflected by the sign of garden, famsize Table 4. Mean Willingness to Pay and Confidence Intervals
and bill per person. Female respondents are on average
willing to pay more than their male counterparts and this WTP Estimates All Drinkers
could be explained by the more active role played by ¼13.9) 26.7% (C
Mean: Percent of bill (amount in Euros) 42.6% (C ¼8.8)
females in ensuring the provision of water for the household 95% Confidence interval [39.0%, 46.2%] [20.3%, 33.0%]

7 of 8
W12407 GENIUS AND TSAGARAKIS: WTP FOR QUALITY-INDUCED IMPROVEMENTS W12407

as well, perceptions of quality are important determinants of DeShazo, J. R. (2002), Designing transactions without framing effects in
willingness to pay to avoid shortages. In the case of iterative question formats, J. Environ. Econ. Manage., 43, 360 – 385.
European Union (1998), Council directive 98/83/EC of 3 November 1998
Heraklion, where most of the residents install tanks to cope on the quality of water intended for human consumption, Off. J. Eur.
with quantity shortages, drinking behavior is greatly affected Commun., L330, 32.
by shortages. Our study finds that both experience with Golfinopoulos, S. K. (2000), The occurrence of trihalomethanes in the
shortages and drinking water from sources other than the tap drinking water in Greece, Chemosphere, 41, 1761 – 1767.
Golfinopoulos, S. K., and A. D. Nikolaou (2005), Survey of disinfection
are important determinants of willingness to pay, with by-products in drinking water in Athens, Greece, Desalination, 176,
average willingness to pay being much higher for house- 13 – 24.
holds who rely on bottled water. The actual amount we find Greene, W. H. (2003), Econometric Analysis, Prentice-Hall, Upper Saddle
for the average willingness to pay is way below the River, N. J.
expenses households incur in order to buy bottled water. Griffin, R. C., and J. W. Mjelde (2000), Valuing water supply reliability,
Am. J. Agric. Econ., 82, 414 – 426.
An averting behavior model might predict a much higher Hensher, D., N. Shore, and K. Train (2005), Households’ willingness to pay
mean WTP than the one we find but it might be debatable for water services attributes, Environ. Resour. Econ., 32, 509 – 531.
whether expenses in bottled water can be considered as Howe, C. W., M. G. Smith, L. Bennett, C. M. Brendecke, J. E. Flack, R. M.
averting expenditures in the present case. Some important Hamm, R. Mann, L. Rozaklis, and K. Wunderlich (1994), The value
of water supply reliability in urban water systems, J. Environ. Econ.
issues have not been thoroughly investigated in the present Manage., 26, 19 – 30.
study. For instance, the process quality perceptions’ forma- Kampioti, A. A., and E. G. Stephanou (2002), The impact of bromide on
tion has not been addressed and the use of filters has not the formation of neutral and acidic disinfection by-products (DBPs) in
been analyzed in detail. Issues of information, or the lack Mediterranean chlorinated drinking water, Water Res., 36, 2596 – 2606.
of it, rise in both cases. Are households aware of what McFadden, D. (1974), Conditional logit analysis of qualitative choice be-
havior, in Frontiers in Econometrics, edited by P. Zarembka, pp. 105 –
their filtering devices do? Do they keep an adequate filter 142, Elsevier, New York.
maintenance schedule? What are their sources of informa- McPhail, A. (1994), Why don’t households connect to the piped water
tion about water quality? Local authorities could play an system? Observations from Tunis, Tunisia, Land Econ., 70, 189 – 196.
important role in providing information to residents about Nickson, R. T., J. M. McArthur, B. Shrestha, T. O. Kyaw-Myint, and
D. Lowry (2005), Arsenic and other drinking water quality issues,
the quality of residential water and affect their perceptions. Muzaffargarh District, Pakistan, Appl. Geochem., 20, 55 – 68.
Pattanayak, S. K., J. C. Yang, D. Whittington, and B. Kumar (2005),
[38] Acknowledgments. The first author acknowledges partial sup- Coping with unreliable water supplies: Averting expenditures by house-
port from a Marie Curie Transfer of Knowledge Fellowship of the European holds in Katmandu, Water Resour. Res., 41, W02012, doi:10.1029/
Community’s Sixth Framework Programme under contract MTKD-CT- 2003WR002443.
014288. The authors are grateful to G. Dialynas and M. Koumakis, Rivers, D., and Q. H. Vuong (1988), Limited information estimation and
employees of the MEWSS of Heraklion, for providing the data related to exogeneity tests for simultaneous probit models, J. Econ., 39, 347 – 366.
the water supply, its quality, and quantity. Also, thanks are owed to M. Schläpfer, F., and M. Schmitt (2005), Choices about public goods: The role
Manioudaki, E. Papadoyannakis, E. Skoula, K. Panagiotakis, and C. of information shortcuts, paper presented at the 14th Annual Conference
Kontaki for helping with data collection. Last but not least, we would like of the Association of Environmental and Resource Economists, Bremen,
to thank the reviewers and the Associate Editor for their invaluable Germany.
comments that have helped improve the paper. The usual disclaimer Scientific Committee on Toxicity, Ecotoxicity and the Environment (2003),
applies.
Opinion of the CSTEE on ‘‘Studies Concerning the Quality of Drinking
Water in Selected European Cities’’, Rep. C2/VR/csteeop/12062003/D
References (03), Brussels.
Alberini, A., B. J. Kanninen, and R. T. Carson (1997), Modeling response Shaw, W. D., M. Walker, and M. Benson (2005), Treating and drinking well
incentive effects in dichotomous choice contingent valuation data, Land water in the presence of health risks from arsenic contamination: Results
Econ., 73, 309 – 324. from a U. S. hot spot, Risk Anal., 25, 1531 – 1543.
Arrow, K., R. Solow, P. R. Portney, E. E. Leamer, R. Radner, and Um, M. J., S. J. Kwak, and T. Y. Kim (2002), Estimating willingness to pay
H. Schuman (1993), Report of the NOAA Panel on Contingent Valua- for improved drinking water quality using averting behavior method with
tion, NOAA, Silver Spring, Md. (Available at http://www.darp.noaa.gov/ perception measure, Environ. Resour. Econ., 21, 287 – 302.
library/pdf/cvblue.pdf.) Whittington, D. A., J. Briscoe, X. Mu, and W. Barron (1990), Estimating
Barakat and Chamberlin, Inc. (1994), The value of water supply reliability: the willingness to pay for water services in developing countries: A case
Results of a contingent valuation survey of residential customers, Calif. study of the use of contingent valuation surveys in southern Haiti, Econ.
Urban Water Agencies, Sacramento, Calif. Dev. Cult. Change, 38, 293 – 311.
Cameron, T. A., and M. D. James (1987), Efficient estimation methods for World Health Organization (2000), Bottled drinking water, Fact Sheet 256,
closed-ended contingent valuation surveys, Rev. Econ. Stat., 69, 269 – Geneva, Switzerland.
276.
Cameron, T. A., and J. Quiggin (1994), Estimation using contingent valua-
tion data from a ‘‘dichotomous choice with follow-up’’ questionnaire, 

J. Environ. Econ. Manage., 27, 218 – 234. M. Genius and K. P. Tsagarakis, Department of Economics, University
Carson, R. T., and R. Mitchell (2006), Public preferences toward environ- of Crete, University Campus, 74100 Rethymno, Crete, Greece. (genius@
mental risks: The case of Trihalomethanes, in Handbook on Contingent econ.soc.uoc.gr)
Valuation, edited by A. Alberini and J. R. Kahn, pp. 394 – 420, Edward
Elgar, Northhampton, Mass.

8 of 8

You might also like