+ All Categories
Home > Documents > 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1...

2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1...

Date post: 03-Sep-2020
Category:
Upload: others
View: 2 times
Download: 0 times
Share this document with a friend
32
TERI University Ph.D. Thesis, 2006 Review of Literature 2.1 Introduction This chapter reviews the previous published studies and empirical techniques that have been used to assess customer’s preferences for water services in the context of welfare change. These studies either use ‘revealed preference’ or ‘stated preference’ methods to quantify change in welfare. Revealed preference methods are subdivided into two distinct categories – hedonic pricing and averting expenditure; while stated preference methods use any of the three approaches – open-ended contingent valuation, contingent ranking and discrete choice methods. Discrete choice methods are further subdivided into referendum contingent valuation and choice modelling. The published studies are reviewed keeping in mind this classification. The chapter concludes with the overall summary of literature review, which is intended to set the basis for defining objectives of the current research. 2.2 Revealed Preferences The correspondence of choice and revealed preference is seen as an empirical matter, and this is indeed how the correspondence is viewed in the classical demand theory (Sen 1997). Considerable research efforts have been made in public economics to estimate the demand and choice for public goods with revealed preference data. These efforts have sufficient theoretical support as they employ market information for complementary or surrogate goods such as land/housing markets to assess change in levels of services for valuing monetary impacts of such changes (Karl-Göran Mäler and Vincent 2005). Haripriya and Kathuria (2005), Yusuf and Koundouri (2005) use hedonic pricing techniques whereas Dasgupta and Dasgupta (2004), Kolstad (2000), Abdalla et al. (1992), Akerman et al. (1991), Harrington and Portney (1987), Courant and Porter (1981) use averting expenditure for quantifying welfare change in estimation of benefits of better water supply. 2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974), this technique uses the assumption that the price of a non-market good is related to price of the market good and its associated attributes. The technique is mainly used in property value studies, where the price of a particular unit of housing is expected to include the value of environmental amenities, such as good water quality. Various attributes and 2
Transcript
Page 1: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

TERI University Ph.D. Thesis, 2006

Review of Literature

2.1 Introduction

This chapter reviews the previous published studies and empirical techniques that

have been used to assess customer’s preferences for water services in the context of

welfare change. These studies either use ‘revealed preference’ or ‘stated preference’

methods to quantify change in welfare. Revealed preference methods are subdivided

into two distinct categories – hedonic pricing and averting expenditure; while stated

preference methods use any of the three approaches – open-ended contingent

valuation, contingent ranking and discrete choice methods. Discrete choice methods

are further subdivided into referendum contingent valuation and choice modelling.

The published studies are reviewed keeping in mind this classification. The chapter

concludes with the overall summary of literature review, which is intended to set the

basis for defining objectives of the current research.

2.2 Revealed Preferences

The correspondence of choice and revealed preference is seen as an empirical

matter, and this is indeed how the correspondence is viewed in the classical demand

theory (Sen 1997). Considerable research efforts have been made in public

economics to estimate the demand and choice for public goods with revealed

preference data. These efforts have sufficient theoretical support as they employ

market information for complementary or surrogate goods such as land/housing

markets to assess change in levels of services for valuing monetary impacts of such

changes (Karl-Göran Mäler and Vincent 2005). Haripriya and Kathuria (2005),

Yusuf and Koundouri (2005) use hedonic pricing techniques whereas Dasgupta and

Dasgupta (2004), Kolstad (2000), Abdalla et al. (1992), Akerman et al. (1991),

Harrington and Portney (1987), Courant and Porter (1981) use averting expenditure

for quantifying welfare change in estimation of benefits of better water supply.

2.2.1 Hedonic pricing techniques

Pioneered by Griliches (1971) and formalized by Rosen (1974), this technique uses

the assumption that the price of a non-market good is related to price of the market

good and its associated attributes. The technique is mainly used in property value

studies, where the price of a particular unit of housing is expected to include the

value of environmental amenities, such as good water quality. Various attributes and

2

Page 2: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 7

TERI University Ph.D. Thesis, 2006

their associated prices of the property reveals the implicit prices of water quality,

known as hedonic prices.

Haripriya and Kathuria (2005) estimated the effect of water availability on house

price and the willingness to pay for a water project that provided reliable water

supply in Chennai using hedonic price technique. The authors argued that quantity

and quality of water supply in a particular zone may affect the preferences of the

households to reside in that area. Using a questionnaire survey of 1416 households

in 13 localities in the city centre of Chennai, the results indicated that the aggregate

WTP for an increase in water supply situation was approximately Rs 2480 million

and for improved water quality the aggregate WTP was Rs 981 million in urban

settlements. The higher price was revealed for piped water and improved water

quality in their premises in comparison to the existing situation of fetching water

through different sources such as ground water, private tankers, metro tanker,

public taps etc. Similarly, Yusuf and Koundouri (2005) investigated how the

decision on house location (urban/rural) affected the household’s valuation of

water-related characteristics of the house through hedonic price analysis. The

empirical analysis used data from the Indonesian housing market and suggested

that households valued access to safe and improved domestic water sources in

selecting housing locations.

Not many case studies are available to value improved water supply through

hedonic pricing technique. The largest problem that hedonic analysis faces is the

requirement that the researcher be able to control all relevant attributes, which is a

difficult task. Even where relevant proxy markets exist, it may be difficult to provide

adequate information about households’ actual willingness to pay for the goods in

question. Other problems with hedonic property valuation include, but are not

limited to, the need for the individuals in question to be aware of the differences in

attributes between potential housing units and jobs, insufficient market data, and

unknown functional forms of the pricing equations (Mitchell and Carson 1989).

According to Bateman et al. (2002), “hedonic pricing technique estimates the value

of a marginal change, in the provision of the good in question. To value discrete

changes, which are often of interest, requires the full ‘two stage’ approach to hedonic

price estimation. This can be complex and resulting estimates can be very uncertain.

Estimates derived from the first stage only, i.e., the change in property price with

respect to small change in the variable affecting the property price, are, however,

thought to be upper bounds to the theoretically complete estimate if one assumes

that there is a variation on the extensive margin (for example, alternative house

size).” Also revealed preference techniques are limited to estimating use value

Page 3: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 8

TERI University Ph.D. Thesis, 2006

associated with goods that are reflected in actual markets, or confined to those

services where there is an associated market.

2.2.2 Averting expenditure

Averting expenditure (AE) analysis captures economic value from customers’ ability

to purchase goods in traditional economic markets that can be used to mitigate

adverse environmental and health impacts. The theoretical foundation is based on

estimation of customers’ ‘production function’ from their demand and consumption

of goods and services as an approximation of benefits stemming from reduced

environmental risks (Dasgupta and Dasgupta 2004, Kolstad 2000, Abdalla et al.

1992, Whittington et al. 1990c, Harrington and Portney 1987, Courant and Porter

1981).

Dasgupta and Dasgupta (2004) estimated WTP in infrastructurally disadvantaged

slums of Delhi using production function approach and found that the averting

expenditure was substantial to secure a safe and reliable water supply. The authors

attempted to provide measure of the households’ minimum willingness-to-accept

for avoiding damages from consuming contaminated water through indirectly

estimating monetary value of health losses borne by households due to diarrhoeal

sickness. The average predicted probability of observing diarrhoeal sickness in the

sampled household was found to be 0.44 whereas the estimated cost of illness for

poor representative household due to diarrhoeal diseases was found to be Rs

1094.31 per annum.

In a revealed preference study carried out in Ukanda, Kenya in 1990 by Whittington

et al. (1990c) it was observed that ‘time’ to collect water and ‘price’ were two major

decision factors responsible for households’ preference for improved water supply.

Choice of different water systems results in different time savings, the service levels

offered from them have to be traded off in terms of increased costs of connection

and the benefits derived from reduced time spent in collecting water. In this study

households placed higher values to time saved due to better water supply sources.

This was roughly equivalent to the wage rate for an unskilled labour, though Asian

Development Bank at that time assumed that time savings should be valued at 50%

of the market wage rate for unskilled labour in the local economy. The households

were already paying 8% of their income on securing water from the vendors. The

authors while using conditional multinomial logit model to study the preference of

water supply systems took two sets of explanatory variables – (i) source attributes

which affect households’ utility such as duration of collection and taste of water, and

Page 4: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 9

TERI University Ph.D. Thesis, 2006

(ii) households’ characteristics that reflect the preference variations such as price of

water and households’ income. This study gave one of the first instances of valuation

benefits from time savings from an improved water supply in a developing country

which was much higher than the commonly held belief.

Many of researchers believe that AE have lower bounds compared to compensating

and equivalent variation, and thus may result in underestimation of customers’

actual expenditure on getting a reliable water supply (Kolstad 2000, Roach 1990,

Shortle and Roach 1989, Courant and Porter 1981). This may be case even if the

value of time that is forgone due to averting behaviour is included in the analysis

(Akerman et al. 1991). This was observed through constructing model of households’

utility function to study their expenditure on averting pollution. Courant and Porter

(1981) considered the response of customers who wish to maximize their utility to

diminutive changes in the levels of water quality, assuming that a better water

quality is preferred mainly because it reduces the need of higher expenditure to

avoid risk of contamination. Harrington and Portney (1987) argue that AE can

produce welfare measures accurately stemming from water quality improvement

project provided information on market mechanisms and customers’ behaviour is

correctly elicited. Hanley et al. (1997) list conditions required for better AE

estimates, such as – (i) AE should not exhibit ‘joint production’ and changes in

expenditure to averting behaviour should be entirely due to changes in quality of

specific goods in question, (ii) increase in AE should also improve the quality of

specific goods, and (iii) the specific goods should directly enter customers’ utility

function, not the inputs required to produce them.

As stated earlier, AE technique is confronted with ‘joint products’ problem, wherein

total expenditure may correspond to spending not only on the specific goods in

question, but also on some other unrelated goods. For example, households may

consume bottled water for drinking purposes not only in an effort to protect them

from contamination, but also because they consider that bottled water tastes better.

Also AE includes customers’ long-lasting investments in durable goods and has large

sunk cost such that the full price of such goods cannot be equated as payments for a

change in environmental quality (Bartik 1988). According to Bateman et al. (2002),

WTP estimates from averting expenditure analysis will be upper bounds rather than

the estimates that may be required for policy purposes.

Page 5: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 10

TERI University Ph.D. Thesis, 2006

2.3 Stated preference method

Direct or stated preference method uses survey based approach to valuate

hypothetical changes in levels of service provision by asking customers how much

they would be willing to pay for improved service conditions vis-à-vis existing

service levels (Tietenberg 2005). It is also called contingent valuation (CV) because

the alternate scenario presented to the customers is contingent upon future service

levels enumerated to them. Alternatively, for service failure, they are asked how

much they would be willing to accept compensation for the changed service levels.

The first one is called willingness to pay (WTP) and the second one is referred to as

willingness to accept compensation (WTA). With a strong foundation in

constrained utility maximization theory (McFadden 1974, Johansson 1995), stated

preference technique has become the most widely used tool for estimating benefits

associated with providing non-marketed goods and services, especially in

environmental economics literature. Olsen and Smith (1999) agree that ‘WTP is the

most appropriate way to value benefits provided by imperfectly provided goods and

services in monetary terms’.

Stated preference method is an omnibus for a variety of approaches which use any

of these three techniques – open-ended contingent valuation, contingent ranking

and discrete choice methods. Discrete choice methods are further subdivided into

referendum contingent valuation and choice modelling. The family of stated

preference techniques is shown in the figure 2.1.

Stated preference methods

Open-ended contingent valuation

Conjoint ranking and paired comparison

Discrete choice methods

Referendum contingent valuation

Choice modelling

Figure 2.1 The family of stated preference techniques

Page 6: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 11

TERI University Ph.D. Thesis, 2006

All of these methods approximate the demand curve that would have existed, had

there been a market for valuation of a non-market good like ‘water quality’.

According to (Haab and McConell 2002), stated preference data generated from a

‘questionnaire based survey’ allows estimation of changes in the provision of specific

goods and can be evaluated in the similar way as actual choice and preference

observed in the market. This allows to directly estimate customer’s WTP for water

supply improvements, which is governed by current level of services and coping

mechanisms against unreliable supply, customers’ expectations, and other socio-

economic variables (Haider and Rashid 2002, Raje et al. 2002, Zerah 2001, Asthana

1997). These considerations are taken into designing the survey instruments i.e. the

questionnaire. Questionnaires are deliberated to imitate the behaviour of the

respondents in the marketplace, which are common in all these methods except the

manner in which valuation question is posed to the respondents in assessing the

demand and preference behaviour (Mitchell and Carson, 1989). It basically

confronts the respondents with a given monetary amount under a hypothetical

scenario, resulting into a response. The purpose of appropriate elicitation format in

the valuation of non-market commodities is to present the hypothetical scenario in a

plausible manner so that respondents can express their true preference for them.

This has evolved from the simple open-ended questions to more complex choice

experiments. As respondents’ WTP for an improvement water supply services is a

function of the projected changes in the level of the service, all factors which control

individual’s valuation of the specified change are included in the questionnaire.

According to Whittington et al. (1993), a particular method assumes that a

household’s WTP for an improved water supply depends on several factors such as:

(a) existing levels of service provision and sources of supply; (b) tariff rate and

billing; (c) perceptions of reliability, pressure and quality as water problems ; (d)

Price of other related goods and services; (e) socio-economic characteristics such as

age, gender, family size, education level, occupation, income, expenditures and

assets; (f) housing characteristics; (g) levels of water consumption; and (h) attitudes

of the household which may serve as proxies for taste.

2.3.1 Open-ended contingent valuation

This has evolved from the simple open-ended question of early studies where

respondents are asked to state their maximum WTP amounts as point estimates.

McPhail (1993) in 5 small Moroccan cities found that households were ready to pay

anywhere from 7 to 10 per cent of households’ expenditure for securing residential

water connections even though they were provided with free and reliable standpost

services. Though this finding was not validated by later studies. Altaf (1994)

Page 7: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 12

TERI University Ph.D. Thesis, 2006

described the results of a WTP survey of approximately 1000 households conducted

in Gujranwala, Pakistan to document household response to inadequate water and

sanitation services and to estimate demand for improved public services. The study

showed that demand-side information about household preferences and priorities,

which have traditionally been neglected, could provide important input for

designing public policy aiming at better societal welfare. However, in this study,

respondents were willing to pay 50% less than the price they were currently paying

because they were not convinced from the hypothetical scenario that would result in

savings in procuring water from other sources, mainly private wells. The response

was explained by the fact that private wells were better alternative supply than the

public water supply and so the respondents did not deem promise of an improved

water supply plausible. Raje et al. (2002) presented an approach to quantify the

satisfaction level of consumers relating to water supply services and described the

impact of various factors on WTP through binary logistic regression analysis. Open-

ended elicitation format was employed for getting point estimates of households’

WTP. The study was conducted in Mumbai, one of the mega cities in India, through

a sample of 1000 households selected from 8 zones. Nearly 50% of respondents

were ready to pay partially more than their current bill amounts. The authors

assumed that that consumers’ affordability, satisfaction about water supply services,

and their beliefs in service provider might influence positively on their WTP.

However, it was observed that with increasing level of satisfaction of the consumers,

the probability of WTP decreased. But the change in the probability value of higher

WTP due to change in satisfaction level was insignificant making this variable a

weak indicator of households’ WTP.

Recently, Casey et al (2006) conducted a CV survey of over 1600 residents in the

city of Manaus in Brazil to assess households’ willingness to pay for improved water

services collecting data on their current water needs, health concerns, and

socioeconomic characteristics. Six low-income communities in the eastern area of

Manaus were initially selected for conducting the survey and, households were

randomly selected for in-person survey within these communities. OLS regression

analysis indicated that residents were willing to pay between US$6.12 to US$8.67

per month for improved water services with household size, years living in the

house, employment status, monthly expenses for water outside the home, electricity

bill having positive impact on the WTP whereas income and age of the households

had negative impact on their WTP. The authors used four elicitation techniques to

calculate WTP: open-ended, open-ended with a pre-qualifying statement,

descending and ascending dichotomous choice bidding game. This was done to

Page 8: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 13

TERI University Ph.D. Thesis, 2006

observe for any differences across these four elicitation formats to see if one or any

of them is resulting in significantly different WTP estimates. The open-ended and

augmented open-ended elicitation formats resulted in nearly identical mean WTP.

The descending bid dichotomous choice format gave the highest WTP and the

highest variance around the mean. When used as the dummy variable, this

elicitation format had statistically significant positive sign.

In open-ended direct question formats, respondents express their WTP for

alternative scenario (point estimates), which represents respondents’ maximum

willingness to allocate from income for deriving the same level of utility. No hint is

given to the respondents about the expected charge or expected worth of the

alternative options. Therefore, when the specific goods in question are being

provided ‘free’ or at least cost to them, they may have the difficulty to actually

express true economic value for the services. Consequently, this format is associated

with a large number of non-responses and ‘protests zero bids’. The term ‘protest

zero bids’ refers to a situation where a respondent indicates that his WTP is zero, not

because he has no value for the specific goods, but because he either objects to some

aspect of the survey; or he believes that no monetary value can be placed on the

good in question; or he has very low faith in the service provider’s ability to supply

the stated goods. The method is also associated with ‘strategic overstatements’

wherein the respondents may increase their WTP estimates deliberately so that

goods can be provided to them (NOAA 1993).

2.3.2 Conjoint ranking and paired comparison

In contingent ranking, several alternative options with different attributes (having

specific cost attribute) are presented to the respondents where respondents are

requested to rank these alternatives according to their preferences and ability to pay

(Roe et al. 1996). Thus a rank-wise preference structure can be obtained depending

upon respondents’ first or second choice with cost tag assigned to each of the

choices (Foster and Mourato 2002). In water supply sector, this technique has been

found to be quite useful especially in low-income areas by Coates et al. (2001) for

‘encouraging exchange of ideas between a water supply organization and community

groups’. Recently, Yang et al. (2006) have used conjoint ranking to evaluate

customers’ demand for alternative water supply system in three coastal towns in

southwestern Sri Lanka. They compared household preferences for four water

supply attributes – price, quantity, safety, and reliability for both poor and non poor

settlements. The authors reported that households possessed heterogeneous

preference structure for quantity, safety, and service options, but not the supply

Page 9: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 14

TERI University Ph.D. Thesis, 2006

duration. It was also observed that households in poor settlements had

comparatively lower capacity to trade-off discretionary income for reliable services

than the non poor households.

Mycoo (1995) attempted to examine the major problems of water supply in urban

Trinidad analyzing factors influencing customers’ willingness to pay for water

service improvements in the early 1990s. The author conducted the survey in 1993

in the capital region of Trinidad using stratified cluster sampling dividing the

sampling into six smaller clusters using a sample size of 360 households. In this

study CVM, contingent ranking and revealed preference method were collectively

used to determine households’ WTP. Reliability, pressure, quality and price

variables were explained using payment card approach in the hypothetical choice.

Since the households were finding it difficult to attach a price to the various

attributes of water supply (reliability, pressure and quality), the contingent ranking

asked them to rank these attributed in the order of priority. Revealed preference

method was based on household’s production function approach. The regression

model indicated that household income, the price of water, number of service hours,

and housing and land tenure influenced positively on WTP amounts, whereas

expenditure on electricity was negatively correlated with increasing WTP.

In this study the application of conjoint ranking was only limited to ranking or

prioritizing the water supply attributes without actually estimating households’

WTP. Therefore, CV method was employed to estimate households’ WTP which

indicated that 80% of the sampled households were willing to pay twice the

currently charged price for a reliable supply. However, this was not validated by the

revealed preference estimates which resulted in lower bounds of value of a reliable

supply. This was explained by the fact that about 75% of the population had already

replicated a water storage and distribution system in their residences by purchasing

and installing tanks and water pumps. This study therefore, gives differing estimates

of customers’ value from revealed and stated preference techniques.

2.3.3 Discrete choice methods

The discrete choice methods for estimating WTP for improved water supply have

been widely employed in both developing and developed countries to determine

affordability and WTP for water and sanitation services (Whittington et al. 2002,

1993, 1992, 1991, 1990ab, Misra 1999, Griffin et al. 1995, Whittington and Swarna

1994, Altaf et al. 1993, McPhail 1993, Whittington and Lauria 1991, Briscoe et al.

1990). These methods have been further sub-divided into referendum contingent

Page 10: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 15

TERI University Ph.D. Thesis, 2006

valuation or choice modelling depending upon the manners in which valuation

questions are posed to the respondents.

2.3.3a Referendum contingent valuation

The referendum CVM format is based on bid levels posed in different manner to the

respondents in the questionnaire – such as single bounded, double bounded or

multi-bounded. This format was introduced by Hoehn and Randall (1987) based

from the single-bounded dichotomous choice (DC) format earlier employed by

Bishop and Heberlin in 1979. This format presents the respondents with the

description of the good, and a single bid, or monetary value, which he may choose to

accept (‘yes’ vote) or reject (‘no’ vote) reflecting whether or not he supports the

provision of a specific good at the stated cost (Cameron and Quiggin 1994). The

single bid value is varied across the sample in the survey. The respondents

hypothetically weigh against their utility in ‘baseline scenario’ to that of utility under

the ‘improved alternative scenario’. These methods are viewed as producing a

hypothetical situation that is quite similar to actual market situations or actual

voting situations; and hence familiar to most individuals. The NOAA Panel on the

CVM recommended the DC referendum format as an appropriate elicitation method

in majority of circumstances (Arrow et al. 1993). However, one drawback of the

traditional DC and referendum formats is that they require the researcher to make

well-built postulation about the statistical distribution of respondents’ stated prices,

so that correct indirect utility function can be constructed to get consistent welfare

estimates (Ready et al. 1995, Mitchell and Carson 1989). This aspect can be met

better with large sample size and careful bid design. Carson et al. (1986) proposed

‘take it or leave it with follow-up’ format to overcome some of the drawbacks of the

traditional approach. In this method, if a respondent aggress to accept the

alternative scenario at the initial bid price, he is asked a follow-up WTP question

that uses a little higher price. On the other hand, the respondent may decline to

undertake the alternative scenario at the initial bid, in which case the follow-up

WTP question uses a lower price. This is very similar to iterative bidding game

which works like true purchasing scene in the market when a buyer tries to bargain

such as those in auction process (Mitchell and Carson 1989).

Whittington et al. (1991) observed that households’ perception of water supplied

from the local authority was generally unsatisfactory in Onitsha, Nigeria and this

resulted in their low willingness to pay. People perceived water quality from those

provided by the tanker trucks and water vendors better than those provided by the

public system. In the absence of reliable supply, households relied upon private

Page 11: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 16

TERI University Ph.D. Thesis, 2006

water vendors. Annually, the payments made to the private vendors were more than

double the O & M cost of supplying water through a piped network. Whereas the

public utility was recovering 1100 US $ from supplying 1.5 MGD water, households

were spending about 28000 US $ to obtain 2.96 MGD from well-organized network

of private water vendors during the dry season. In their study, they found that

households were willing to pay ‘surprisingly high’ amount to secure a reliable

supply. The authors did not use rigorous sampling strategy due to the absence of

appropriate sampling frames; instead biding game approach was employed in a

rapid reconnaissance survey. The study was one of the first examples to show the

scale and magnitude of water vending activities in metropolitan areas of a

developing country. The authors conclude that, “the revenue potential of the public

authority is tremendous and that the population of Onitsha would be much better

served if the utility viewed itself as a regulated utility, not as an agency providing a

social service”.

Two Indian case studies in rural and urban setting can be cited which give different

policy outcomes. In one study in rural Kerala, Singh et al. (1993) used referendum

CVM to study the households’ preference for individual domestic connection with

better quality in comparison to shared yard tap. They observed that households

were willing to share the higher cost for domestic connection if ‘local credit’ facility

was available to them. Vaidya (1995) in a study from urban Baroda reported that 80

% of the households with individual connection were willing to pay 3 times higher

amounts than the current bill if better pressure and quality was ensured to them.

WTP declined with increasing income of the households – households with highest

income were willing to pay just 60% higher than the households with the lowest

income; thereby indicating ‘limited reach for cross subsidization’ across different

categories of the households. The Kerala study indicated that the ‘transferability of

findings from one site to another was likely to lead to erroneous predictions even if

site characteristics, service offered and community characteristics are apparently

alike’.

Asthana (1997) used a multinomial logit model to predict household’s choice of

water supply system (private pipe potable water; public standpost; handpump;

dugwell; surface water) in Bhopal. The in-house piped water supply was found to be

the most preferred alternative option. The perceptions of quality were important

determinants of the choice of water supply source for drinking purposes. The paper

concluded that rather than trying to provide a free or heavily subsidized minimum-

Page 12: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 17

TERI University Ph.D. Thesis, 2006

service to all, the policy makers need to consider an improved service to all and

higher level of service to those who are willing to pay higher amounts.

According to Goldblatt (1999) an improved understanding of customers’ demand

and willingness to pay is not only imperative for financial sustainability of improved

water services, but also for enhanced extension to poorer sections of the urban

society. The author debates the usual effect of providing subsidized water services

without actually studying their demand and consumption pattern resulting into

utility’s decreasing revenue base and its inability to expand to poorer households. As

the author notes “those already with access to water will be paying below-cost prices

while the poorest of the poor households may be denied formal access completely.”

In a CV survey carried in greater Johannesburg in two informal settlements of

Vlakfontein and Finetown to study the water use behaviour of the households and to

study their WTP for improved water supplies, the study employed two controlled

hypothetical bidding approaches whereby households indicated their maximum

WTP for improved water supply scenarios. In the first approach, ‘volumetric bids’

were used in terms of cost per 25 liters of water varied through the sample and the

final amount selected as the maximum WTP. The second bidding approach called as

‘monthly bid’ asked households their maximum monthly water bill as maximum

WTP assigning different low or high starting points selected randomly. This was

done first, to minimize the starting point bias widely cited in the literature, and

second, to later compare the volumetric WTP with the monthly bill based upon the

consumption details provided by the respondents. No parametric models were used;

instead the bids and consumption levels were used to calculate the WTP amounts.

According to the study, Utility’s monthly revenue can be estimated keeping in mind

three attributes – (i) total percentage of households who agree to connect at the

specified price levels, (ii) price of water, and (iii) monthly water consumption. The

results showed that 64% of the households would pay for a reliable piped water

supply if the total monthly bill were 5% of income or below it testifying the

conventional five per cent rule. Volumetric bids and monthly bids yielded similar

results.

Koss and Khawaja (2001) measured customers’ valuation of water supply reliability

using the referendum CV technique. Respondents were asked to indicate their WTP

in order to avoid the occurrence of water shortages of a given frequency and

severity. The authors employed a double bounded dichotomous choice biding game

strategy. The results indicated that customers were willing to pay as much as

$16.92/month to avoid a 50% water shortage occurring over 20 years.

Page 13: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 18

TERI University Ph.D. Thesis, 2006

Whittington et al. (2002) examined household’s demand for improved services in

Kathmandu, Nepal through a randomly selected sample of 1500 households. The

objective was to study the preference for water supply from a private operator or

from existing public water supplier. The valuation section had one CV scenario with

different market segments depending upon the existing supply situation.

Households were presented with one of the eight different prices. The data was

modeled econometrically through discrete choice binomial logit model with linear

utility function. The results indicated that about 70% of the households in 5

municipalities were willing to pay a fivefold increase in the current average water

bill in addition to supporting improved services provided by a private supplier. The

study observed that both poor and non-poor segments of the population were

willing to support privatization plan that would improve water supply.

In a study from rural Bangladesh, Water and Sanitation Program, South Asia (WSP

2002) observed that households place higher value on safe drinking water free from

arsenic in arsenic-affected areas. The demand increased with higher income, and

decreased with higher price quoted for piped water supply. The results also

indicated that the higher the awareness and concern for arsenic contamination, the

greater is the inclination to opt for domestic piped water connections. In particular,

the mean WTP for recurring costs for standposts was 46% higher than the actual

costs, whereas it was 40% higher for domestic connections.

Salman and Emad Al-Karablieh (2004) estimated the willingness of farmers to pay

for groundwater resources under different conditions of water supply regimes using

referendum format. The results showed that the water values in the region were

undervalued as farmers were willing to pay 2.5 times extra than the current price of

groundwater. This finding suggested that decision makers can impose a price level

for groundwater from US$ 0.14 to 0.35/m3 corresponding to farmers’ WTP without

having any impact on the cropping pattern or the planted area. Information

available on water supplies, areas under irrigation and market conditions were the

basis of the calculations. The findings also indicated that there was potential to

decrease water consumption and reallocate it in an optimal way to increase the net

agricultural income in the study area.

Ntengwe (2004) examined the role played by consumer awareness in their

willingness to pay for water supply in two cities in Zambia using referendum design.

Survey conducted in Kitwe and Lusaka revealed that level of awareness, WTP and

cost recovery all varied directly. Whereas awareness might increase consumers’

Page 14: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 19

TERI University Ph.D. Thesis, 2006

WTP assisting service provider’s cost recovery, the research also revealed that

factors such as ability to pay, affordability of bills, quality of water and of the service

provided, as well as good business-consumer relations were important factors

affecting a utility’s ability to recover its costs.

Hensher et al. (2004) assessed household’s WTP to avoid outages (defined in terms

of frequency, timing, and duration) and information (defined in terms of

notification of interruption and telephone response to queries) for water supply and

disposal, in Canberra, Australia. They used a series of conditional logit and mixed

logit models. They observed that people were willing to pay affirmatively higher

amounts to secure less frequently interrupted water supply.

Gloria Soto Montes de Oca et al (2005) estimated willingness to pay for maintained

and improved water services using CV method testing two scenarios in three

residential areas in Mexico city in a split sample on 1424 households, one to

maintain and another to improve the service conditions using probit regression

model. Both the scenarios described the threat of water shortages over the next

decade offering differential benefits to the respondents. The ‘maintenance scenario’

was designed to ensure that the current service levels would be maintained whereas

the ‘improvement scenario’ presented additional reliability improvements. The

elicitation format was typical referendum styled where households were presented

with single bid price selected randomly from ten values to be accepted or rejected.

The WTP question used the referendum format, which presented the respondent

with a single buying price, which was accepted or rejected. Water shortages, water

pressure and water quality were considered as three major attributes to estimate the

service quality standards. The mean differences between the two scenarios were

found to be significant indicating that households were willing to pay more for

improved scenario than the maintenance scenario if income distribution was

ignored. The authors argue that apprehensions on the ability of the poor to pay for

the improved services are justifiable, and reasonable water pricing needed to reflect

the income constraints of poor people further justifying the use of differentiated

water prices between rich and poor segments of the society.

About 10,000 consumers were surveyed to assess coping costs and preferences for

improved water supply services in the Greater Bangalore, where there is currently a

water deficit estimated at about 60% of the demand (WSP 2005). Coping strategies

include private bore wells, sumps and overhead tanks to manage irregular and

inadequate supplies; booster pumps to cope with low pressure; use of storage

Page 15: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 20

TERI University Ph.D. Thesis, 2006

mechanisms, like earthen pots, small tanks and buckets by low income/poor

households; purchase from vendors and purchase of bottled water; time spent in

queuing up for water collection; use of filter or boiled water etc. Most non-domestic

consumers have their own borewell or purchase water from vendors. The average

coping cost for households is about Rs 24/m3, varying from Rs 3/m3 for high-rise

flats to Rs 13/m3 for low-income group plotted houses, and Rs 7/m3 in slum areas.

According to the study, the expected O&M cost of the upcoming water supply and

sewerage extension project is estimated at Rs 16/m3 suggesting that it would

generate significant savings. A detailed econometric analysis of the survey data was

also carried out to estimate the consumers’ WTP for improved WSS services. Among

household categories, WTP is about Rs 16/m3 for high-rise flats, between Rs 10 and

12/ m3 for low-income group plotted houses and between Rs 7 and 10/m3 for slum

households.

In all the studies cited above one or the other from of referendum bidding game is

employed. It emerges from the review that researchers using bidding game

approach outnumber those using other elicitation formats. Following justification is

given in the literature in support of the bidding game elicitation format:

(i) This format is more familiar to most respondents as they do bargaining

for a marketed good so they emulate true purchasing scene (Bateman et

al. 2002).

(ii) It is less prone to strategic and hypothetical bias (Carson et al. 2000).

The NOAA Panel also favoured dichotomous choice iterative bidding

game over other formats as an appropriate elicitation method (Arrow et

al. 1993).

(iii) The process of bidding helps the respondents to understand the value of

the goods in terms of ‘utility’ better with times to respond (Hoehn and

Randall 1987) and thus it is possible to clearly bring out the respondent’s

maximum WTP with more precise range than that of open ended CV

questions (Cummings et al. 1986).

(iv) It gives more precise estimates of WTP because of lower standard

deviation around mean WTP. According to Chowdhury (1999) the

advantage of iterative bidding games, is that answers often have a lower

standard deviation around the mean as compared with single bid games.

He chose bidding game format to study WTP for improved water supply

in Dhaka’s urban poor settlements because it worked better than direct

and open-ended elicitation formats.

Page 16: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 21

TERI University Ph.D. Thesis, 2006

(v) The approach is more participatory in nature than the other elicitation

formats.

However, there might be a small problem of starting the bid with the specific

amount, i.e. respondents may get overly or unjustifiably influenced by the starting

bid amount. For example, if the first bid is excessively higher than the expected

WTP of the respondents, the last selected bid will be higher than the true WTP.

Similarly, if starting bid is much less than the expected WTP of the respondents, the

last selected bid will be lower than the true WTP (Roberts et al. 1985). This may be

taken care of by actually calculating the cost of the alternate scenario and reflecting

these costs either in ascending or descending order so that respondents’ stated

maximum WTP from bidding game may lie in the range of the actual costs

previously estimated.

2.3.3b Choice modelling

Choice experiments or choice modelling are multi-attribute stated preference

techniques that which also belong to the family of stated preference method (Ryan

and Wordsworth 2000, Hanley et al. 2001, Adamowicz et al. 1998). This theoretical

construct has been the main starting point for modelling choices in several

applications, such as: transportation (Ben-Akiva and Lerman 1985), marketing

(Louviere 1995, Louviere et al. 2000), energy supply (Goett et al. 2000) and

environmental economics (Boxall et al. 1996). A summary of environmental

applications is given in Hanley et al. (2000). However, there are relatively few

studies that have used choice experiment models to evaluate the service

performance of water companies and customer preferences for attributes of water

supply and quality (Haider and Rasid 2002, Willis et al. 2005, Jalan et al. 2005,

MacDonald et al. 2005). In choice modelling, levels of water service improvements

are valued by their ‘attributes’ and customers are asked to choose between two, or

more than two scenarios having specific attribute bundles (Bennett and Blamey

2001). Price remains the most important attribute and it differs through the

scenario. Having different price levels across the scenarios helps to estimate

marginal utility estimates calculated from choice models which represents social

welfare (Hanley et al. 2005, Train 2003, Hanley et al. 1998). Therefore, choice

modeling approaches allow a more direct route to the valuation of the attributes

(characteristics of the goods), and their marginal changes rather than complete

value of the goods in totality (Bateman et al. 2002). This would mean that the

method would be preferred than the contingent valuation in contexts where it is

necessary to value individual attributes of the goods in question.

Page 17: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 22

TERI University Ph.D. Thesis, 2006

Haider and Rashid (2002) employed discrete choice experiment using a simple

conditional logit model to investigate customer preferences for variations in two

attributes: water pressure and water taste both expressed qualitatively as ‘worse,

same, improved’ in relation to a percentage change in water prices in Thunder Bay,

Ontario. The respondents were asked to choose between different combinations of

water prices, water pressure, and water taste for three alternative sources of water

supply. The survey results suggested that pressure and taste were very important

variable for improved water supply irrespective of the water rates. The questionnaire

had two distinct component in this study: (a) questions relating to important

variables that were responsible for choice of alternate water supply such as water

use, daily consumption of potable water, adoption of water conservation measures,

quality and overall inconvenience (b) set of discrete choice experiments which

presented three alternate sources of water supply. Each of these alternatives

consisted of three attributes having three attribute levels. The households were

asked to choose one of the three options based upon their comparative evaluation of

attribute levels associated with each of the alternatives.

Realizing that very few (one or two percent) respondents had actually suffered any

water service failure over the preceding year, Willis et al. (2005) presented the

respondents with 14 attributes (called as service factors (SF) in the paper) which

changed across all customers in UK. The large number of SFs meant that the 14 SFs

had to be blocked into 5 groups of 3 or 4 SFs to avoid cognitive limitations on

respondents’ ability to simultaneously trade-off large numbers of SFs. The ‘main

block’ permitted customers to trade-off some principal factors in water supply and

quality with wastewater disposal factors; whilst the other blocks investigated

consumers’ marginal rates of substitution (MRS) between different water supply

and quality factors, and wastewater factors. The valuation encompassed use, option,

nonuse, and altruistic utility: that is, the benefit each customer directly derives from

consumption of the SF (use value); the individual’s assessment of the option value of

the SF; the benefit customer derives from knowing that the SF is being improved

even though as a customer may not have been directly affected by it (nonuse value);

and the utility the customer receives from knowing that other customers would

benefit from the SF improvement (altruistic value). The stated choice analysis

estimated the MRS for each SF, and MRS for money. The implicit monetary benefit

of changes in each SF level, or WTP was determined by the ratio of these MRS:

between SFs and money.

Page 18: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 23

TERI University Ph.D. Thesis, 2006

Jalan et al. (2005) estimated WTP for clean water from a multinomial logit model

from a sample of households in Delhi where a household had different choice

available including no purification option. They observed that on an average,

households were willing to pay Rs 25 per month extra for clean drinking water. The

WTP of the richest households were almost twice than that of the poorest

households. Further, increasing the levels of education from illiterate to

matriculation resulted in doubling of expected WTP. Therefore, effect of education

was similar to the wealth effects. The authors studied the impact of different source

of awareness with respect to wealth effects to find out how awareness was impacting

on households’ behaviour towards adopting clean and safe drinking water measures.

The three determinant sources of awareness were – episodes of diarrheal sickness,

exposure to medium of mass communication such as radio, tv, newspaper ads, and

formal schooling.

MacDonald et al. (2005) have used choice modelling approach to highlight the

importance of objective methodologies in standard setting as well as to evaluate

incremental changes in the customer service standards of urban water supply in

Adelaide. They have used ‘water bills’ and ‘frequency of future supply interruptions’

as two important attributes to estimate the implict price of better water supply that

is more reliable using multinomial logit (MNL) and random parameters logit (RPL)

models. The attributes were described in terms of duration of interruption,

frequency of anticipated interruption, communication about interruption (phone

call/knock on door), alternative supply (central location/bottled water) and price

changes. They noted that customers were willing to pay to reduce the frequency and

duration of water interruption depending on the number of interruptions that a

customer faces each year. The households were willing to pay higher amounts for a

water supply that was less frequently interrupted and hence more reliable. In

explaining this positive WTP of households, two variables had no influence on

increasing their WTP. They were – (i) provision of an alternate supply during an

interruption, and (ii) advance notice of interruption or service failure by the utility.

This employs that even if an alternate arrangement of supply were made during a

service failure, or households were told in advance about interruption, their WTP for

a reliable supply would remain same; thereby indicating higher value attached to a

reliable water supply system.

When using the stated preference techniques, the main choice is between using

referendum contingent valuation and choice modelling. According to Bateman et al.

(2002) not all choice modelling techniques are consistent with underlying welfare

Page 19: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 24

TERI University Ph.D. Thesis, 2006

theory; also convincing evidence has not come so far on their superiority over

contingent valuation. However, choice modelling may do better if information is

required on relative values for different attributes of an environmental good (Foster

and Mourato 2003). To study the effectiveness of EU new Water Framework

Directive, Hanley et al. (2005) used choice modelling using mixed logit model to

assess if valuation is sensitive to the bid levels used in bidding game elicitation. They

observed non-significant impact of changing price vector on estimating households’

preference and WTP.

In the stated preference literature cited above, there are basically two steps in

obtaining welfare measure: the estimation step in which one gets parameter

estimates, and the calculation step when the welfare measure is calculated given the

parameters. Discrete choice data is statistically fitted in a specified probability

function and then welfare measures corresponding to the mean and the median

WTP values are computed. However, such measures are sensitive to the distribution

assumption (Carson et al. 1994). WTP when derived from CV studies is random for

several reasons. Frequently, it is a function of a random component of preferences

which is known to the respondent but not to the researcher. When WTP is computed

from the CV survey, same estimated parameters are used, infusing another source of

randomness. Researchers wish to estimate a sample mean WTP because it is a

logical step in the expansion of benefits to the population. The error distribution can

be specified as some parametric form, and the model can be estimated by maximum

likelihood technique (Hanemann and Kanninen 1999). There are also other flexible

formulations and non-parametric methods which can provide alternative fitting of

the observed data dropping scope for some of the biases (Creel and Loomis 1997).

Three types of formulations are generally discussed in the literature to estimate the

welfare change form WTP studies: (i) Parametric forms (for e.g., Logit, Probit,

Weibull), (ii) Semi-parametric forms (for e.g., Ayer’s and Turnbull’s estimators,

Smoothed Maximum Score (SMS) estimators), and (iii) Non-parametric forms (e.g.,

MSCORE, NPREG). Non-parametric treatment of discrete choice CV deals with

estimating WTP from without the constraint of a given distribution for the

unobserved component of preferences (Haab and McConnel 1997, Creel and Loomis

1997, Carson et al. 1994, Kriström 1990, Cosslett 1983, Turnbull 1976). They allow

only limited or no explanation of the effects of covariates. Therefore, despite of

methodological simplicity, they do not permit integration of explanatory variables in

explaining household’s WTP (Kriström 1990, Carson et al. 1994). While the model is

distribution free, there are many cases when the distribution or the functional form

can have a substantial effect on the estimates of WTP (Salvador and Leandro 2001).

Page 20: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 25

TERI University Ph.D. Thesis, 2006

Similarly, even though semi-parametric formulations have the advantage of having

more ‘flexible assumption of the true distribution’ observed in the data and allow

inclusion of covariates including the bid levels into the specification, they are

encountered with methodological pitfalls such as failure to provide ‘exact inference

for the predictive probability conditional on a particular observation of the

covariates’ (An 2000, Lewbel and McFadden 1997, Li 1996), misspecification errors

when homoschedasticity and unimodality assumptions do not hold for the collected

data (Chen and Randall 1997, Horowitz 1993). When the pattern of responses is well

behaved, the estimates of WTP will not be especially sensitive to the choice of

distribution assumption for the unobserved random component of preferences, or

for the functional form of the preferences function. Simple parametric formulations

such as logit and probit estimates gives better ‘estimate of the variance of the error

term’ when customers are sure about their choice and preference structure (Li and

Mattson 1995).

Bohara et al. (2001) examined the performance of three alternative distributional

assumptions – normal, lognormal and Weibull in estimating WTP. The normal

distribution provided unbiased and relatively accurate estimates of WTP,

irrespective of the proportion of the population holding negative WTP. According to

the authors, when a relatively small proportion (about 2%) holds negative values,

the lognormal and Weibull give inferior estimates, but they are still reasonably

accurate. As the proportion of the population with negative WTP increases to 14%,

the Weibull, and especially log-normal, distributions give positively biased estimates

of WTP, and appear to have no statistical advantage over the normal model. When

the proportion of the population with negative WTP increases from 14 to 30%, the

use of the lognormal and Weibull become increasingly untenable. They both begin

to grossly overestimate WTP. According to the authors, the Weibull or lognormal

models never out-perform the normal model. They further extended the simulation

to investigate the performance of mixture models of the lognormal and Weibull

distribution. Although mixture modelling decreases bias and reduces errors

compared to the standard models, it fails to out-perform the standard use of the

normal distribution, especially when 30% hold negative WTP values.

2.4 Comparison of stated and revealed preference methods

Stated preference methods can be used to cover a wider range of attribute levels for

the proposed ‘quality’ or ‘quantity’ changes of a good whereas revealed preference

data fails to provide any information about attribute trade-offs (Swait et al. 1994).

Stated preference methods are commonly criticized because the households’

Page 21: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 26

TERI University Ph.D. Thesis, 2006

behavior they represent may not be actually observed due to several biases (Mitchell

and Carson 1989, Cummings et al. 1986) and their failure to incorporate real market

constraints (Louviere et al. 2000). However, these methods offer the only means for

estimating the value of goods that have no proxy or surrogate markets. There is a

strong case for using revealed preference techniques whenever the relevant

willingness to pay information can be inferred from households’ actual decisions

(Bateman et al. 2002) as decisions actually made in the markets are inferred as

consistent indicator of preference. However, stated preference techniques become

indispensable when the WTP information that is required cannot be inferred from

the markets. On the other hand, revealed preference approaches have to rely on

information from markets for proxy private goods, consumption of which is a

prerequisite for benefiting from the goods in question. Stated preference approaches

may therefore score better over revealed preference approaches. The scope of stated

preference techniques is wider in the sense that they can potentially value all

impacts, both use and non-use values (Bishop et al. 1997).

In literature, comparison of estimates derived from stated and revealed preference

methods are quite common, which is termed as ‘convergent validity test’. Several

researchers have questioned the convergent validity test as they were getting

different value estimates from stated and revealed preference methods (Urama and

Hodge 2006, Bishop 2003, Hanley et al. 2003b, Carson et al. 1996, Mycoo 1995).

Carson et al. (1996) examined the relationship between contingent and revealed

preferences values from 83 studies that estimated the same environmental

amenities. The ratios of contingent values to revealed preference values in these

studies averaged 0.89 with a 95 percent confidence interval of 0.81 – 0.96.

Additional statistical tests confirmed the robustness of this result and the validity of

the approaches involved in the comparisons (Bishop 2003). But some researchers

failed construct validity tests to arrive at similar revealed and stated preference

values (Diamond 1996, Mycoo 1995, Hausman 1993, Diamond and Hausman 1994,

McFadden 1994). The conclusions of Diamond and Hausman (1993, p.4) were

typical of the group: “we conclude that the CV method does not measure an

economic value that conforms with economic preference concept. Thus, we also

conclude that it is not appropriate to include CV measures of stated WTP in either

benefit-cost analysis or compensatory damage measurement”. According to Bishop

(2003), these arguments were not convincing (p. 553) and reason for construct

validity failure was rooted in content validity failures. By computing the indirect

costs incurred by the customers to compensate for the intermittent, irregular, and

unpredictable water supply in Delhi, Zérah (2001) contributed to the ongoing

Page 22: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 27

TERI University Ph.D. Thesis, 2006

debate on moving to demand-oriented service delivery that addressed the

customers’ needs and their ability to pay for a reliable service. However, no

modelling assessment or behaviour of customers relating to their WTP was done;

instead the implicit cost of water supply unreliability was inferred as their indirect

WTP which was expressed through revealed preference data. Davis (2004) engaged

4 different techniques (a household, “intercept” and telephone survey, and a

sequence of focus group discussions) to estimate households willingness to pay for

better water supply in Odessa, Ukraine. The author observed that a particular data

collection technique can have different outcome, in particular, “social desirability

and time pressures can bias the results obtained in demand-assessment research

and lead to substantively different conclusions and policy recommendations”.

Recently, Urama and Hodge (2006) studied the convergence of stated and revealed

preferences data through two separate experiments conducted in southeastern

Nigeria. They compared irrigators’ stated WTP to mitigate soil and water pollution

problems in their farms with the actual expenditure from revealed preference data

to mitigate the same pollution problems for a river basin restoration scheme. The

authors state that, “we found that the farmers’ stated and revealed preferences were

significantly correlated but yielded significantly different means at 5% level. While

econometric analysis reveals a systematic association between key socioeconomic

variables and farmers’ stated WTP, lending credence to the CVM within its

theoretical framework, analysis of the qualitative follow-ups reveals inconsistencies

in some farmers’ stated and revealed preferences that cannot be satisfactorily

explained on the basis of economic theory alone.” They attributed this behaviour to

farmers’ lack of awareness about the problem, information complexity, risk

aversion, and peer pressure which limit their decision making ability.

Despite prevailing disagreement on accuracy of revealed preference as a true welfare

measure, the estimation of WTP from stated preference technique gives

conceptually convincing yet conservative estimates of actual benefits and costs

derived from improvements in environmental quality. Even though these methods

suffer from a number of problems, their appeal is that they examine customers’

behaviour in actual economic markets related to the provision of goods in question,

as well as establish a hypothetical market for that good by the survey instruments.

The participatory feature of decision-making is clearer in stated preference

techniques where households’ attitudes, motivation, preference and WTP are

assessed which are otherwise omitted in revealed preference studies (Bateman et al.

2002).

Page 23: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 28

TERI University Ph.D. Thesis, 2006

Table 2.1 Some major case studies in water supply sector with goods being valued and signs of major explanatory variables Study area with goods being valued

Method of estimation Important variables and their sign with respect to preference and/or WTP

Reference

Ukanda, Kenya; Value of ‘time’ to collect water and households’ preference for water supply options amongst wells, kiosks, and vendors

Revealed preferences, and discrete choice modelling to study the households’ preference of different water systems through conditional multinomial logit model

Time savings (+), price of water charged by the vendor (–), taste (–), income, women (+), education (+)

Whittington et al. 1990c

Onitsha, Nigeria; Preference for securing a reliable water supply as against water secured from vendors and unreliable supply from the local authority

Rapid reconnaissance survey using bidding game, surveys could not be carried out as per strict contingent valuation approach due to absence of well defined sample frames

Quality of water (+), price (–)

Whittington et al. 1991

Trinidad; Customers’ willingness to pay for water service improvements

CVM, contingent ranking and revealed preference method using stratified cluster sampling and OLS regression. Reliability, pressure, quality and price vectors were explained using payment card approach in the hypothetical choice

Income (+), price of water (+), number of service hours (+), housing and land tenure (+), expenditure on electricity (−)

Mycoo M 1995

Vlakfontein and Finetown, Greater Johannesburg; Demand and willingness to pay for improved water supplies in informal settlements

CV study employed two hypothetical bidding processes where households indicated their WTP for improved water supply scenarios - Volumetric bids and monthly bids, distribution of bid amounts was used to estimate WTP

Utility’s monthly revenue can be estimated keeping in mind three attributes – (i) total percentage of households who agree to connect at the specified price levels, (ii) price of water, and (iii) monthly water consumption levels.

Goldblatt M 1999

Page 24: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 29

TERI University Ph.D. Thesis, 2006

Mexico city; Willingness to pay for maintained and improved water services

CV method testing two scenarios in three residential areas in a split sample, one to maintain and another to improve the service conditions using probit regression model

Income (+), bid amount (–), knowledge of the water bill, water bill (–), water quality (–), family members (+)

Gloria Soto Montes de Oca et al 2005

Manaus, Amazonas, Brazil; Willingness to pay for improved water services

CV experiment using OLS regression through four elicitation techniques: open-ended, open-ended with a pre-qualifying statement, descending and ascending dichotomous choice bidding game on a randomly selected households

Income (–), Age (–), household size (+), energy bill (+), water expenditure (+), owned house (+)

Casey et al 2006

Kanye in southern Botswana; Willingness to pay for private water connection

CV approach through iterative bidding game from randomly selected households

Household income (+), level of education (+), employment status of the head of the household (+) and level of consumers’ awareness (+)

Mbata J N 2006

Page 25: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 30

TERI University Ph.D. Thesis, 2006

Table 2.2 Summary of case studies using CVM in water supply sector with different

options and prices

Study area with number of options

Sequencing of scenarios and eliciting procedures

Reference(s)

Anambra State in Nigeria. Public taps and private connections.

1 – described public taps, asked to bid. 2 – described private connections, asked to bid on unlimited quantity of water for fixed rate.

Whittington et al. 1990b

Willingness to pay for Water in Onitsha, Nigeria

Rapid reconnaissance survey of private water vendors and households’ willingness to pay for improved level of services, survey could not follow strict protocols, sample frames could not be constructed appropriately, valuation through bidding game procedure

Whittington et al. 1991

Improved Sanitation Services in Kumasi, Ghana

Two improved scenarios – water closets and a latrine connected to the city sewer systems, on 1200 randomly selected households

Whittington et al. 1993

Kerala, India. CVM to determine demand for piped connections

Two samples. Households with connections asked if they were willing to pay higher tariffs. Others were asked willingness to pay for connection and monthly charges. Yes/no type answers for quoted bid levels

Griffin et al. 1995

Willingness to pay for four water supply options in small towns in Uganda

Describe each option, setting out the institutional environment and a fixed capital cost contribution for each option. Options are described simultaneously. Then respondents are asked referendum style questions for O&M contribution for all four options starting with the most expensive option.

Nostrand 1997

Jakarta, Indonesia. One water supply option to determine value of piped water option only.

1 – asked whether wanted a connection based on three types of charges. 2 – offered connection charge in installments plus monthly charge for ordinary water quality. 3 – offered connection charge in installments plus monthly charge for better quality water. 4 – offered real water authorities connection and monthly rates. All yes/no answers. 5 – repeated the above but with reverse ordering of all questions.

McGranahan et al. 1997

Page 26: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 31

TERI University Ph.D. Thesis, 2006

Small towns water supply in Uganda. Two options with two payment methods. Option 1a – Public taps by jerry can. 1b – Public taps fixed monthly charge. 2 – Private connection with various public taps prices.

1 – Each option was described and bids elicited for paying by jerry can 2- Bids elicited for monthly metered charge 3 – Respondents asked to choose between monthly charge for private connection or pay by jerry can at taps (various prices). No lowering or raising of bids

Whittington et al. 1998

Estimation of demand and willingness to pay for improved water supplies in two informal settlements in Johannesburg

Done through two different bidding games in which respondents indicate whether they would be willing to pay specified amounts differing across samples. The final amount arrived at via this process was referred to as the ‘volumetric bid’ or ‘monthly bid’

Goldblat 1999

Dar es Salaam, Tanzania. Two Water Supply options – Piped connection or public kiosks.

1 – both options described in detail. Respondents asked to select preferred option. Split sample bidding on chosen option only.

Wedgwood 2000

Mombasa, Kenya. Six water supply option to three different market segments

High starting point bids for assessing demand for each option

Njiru and Sansom 2000

Bushenyi, Uganda. Five Water Supply Options

1 – All five options described in detail, respondents chooses one without knowing exact price but knows relative prices. 2 – Various bidding techniques on preferred option. 3 – Respondents allowed to choose option (not obligatory), bidding takes place.

Wedgwood et al. 2001

Kathmandu, Nepal. One main CV scenario with different market segments depending on existing supply.

Respondents were asked specific amounts to connect to the water supply scheme. Eight different prices were used in the questionnaire varied across households, only one price per household, no bidding

Whittington et al. 2002

Willingness to pay for improved water

Administration of two scenarios through a split sampling procedure: The first scenario

Gloria et al. 2005

Page 27: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 32

TERI University Ph.D. Thesis, 2006

supplies in Mexico City for two supply options concerning risk of water shortfalls and quality

(maintenance scenario) was presented to the respondents that would avoid the risk of shortfall and ensure current service level would be maintained. The second scenario (improvement scenario) would avoid this risk of shortfall as well as improve the service conditions. The WTP question employed referendum format, which presented the respondent with a single buying price selected randomly from ten values, which was accepted or rejected.

2.5 Bias1 in WTP studies: construction and specification of the

contingent market

WTP responses from the stated preference techniques are going to remain

hypothetical and there is always a risk of bias, size and direction of which may not

be all so easy to determine (Bateman et al. 2002). Misleading answers by the

respondents introduce bias in the survey that compromises the validity and

consistency of the welfare estimates (Cummings et al. 1995, Diamond and Hausman

1994, Neill et al. 1994). The CVM has been criticized in the literature for context

effects, framing effects and embedding effects mainly resulting into various types of

biases such as strategic bias, hypothetical bias, compliance bias, information bias,

and scenario mis-specification bias. According to Whittington (2004), three main

ways may result in bias during the CV survey: (i) background information in the CV

scenario is poorly and inaccurately provided; (ii) the description of the hypothetical

market is not done in a plausible manner, and (iii) complex referendum elicitation

formats and split-sample experiments. In the literature, several ways have been

devised to reduce the bias and extent of estimation inconsistencies through survey

design and implementation. There are few CV studies on evidence of accuracy in

developing countries as compared to what exists in developed countries

(Whittington et al. 2002).

1 Biased responses are defined as those final WTP values that differ systematically from the ‘true’ values placed by respondents on the good in question.

Page 28: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 33

TERI University Ph.D. Thesis, 2006

Table 2.3 Frequently cited biases in CV with examples from case studies Bias-type Description References Strategic bias Respondent deliberately expresses his WTP

amount to influence the decision to provide the goods resulting in under-representation or over-representation of the true WTP. Therefore he gets maximum share of ‘consumer surplus’ by being ensured provision of the good at a less price than he would actually be willing to pay.

Johannesson 1996, Mitchell and Carson 1989

It depends upon how meaningful CV scenario is described to the respondents. Bias is substantially reduced ‘if careful attention is paid to making the valuation scenario realistic.’

Jones-Lee 1989, Milon 1989, Groves 1973, Clarke 1971, Groves and Ledyard 1977, Smith 1979, Rowe et al. 1980, Bohm 1972, Brookshire et al. 1976

Discrete questionnaires can reduce the incentive for strategic bias.

Carson 1991, Mitchell and Carson 1989

Overall, empirical evidence suggests only weak forms of strategic bias occur

Morrison et al. 1996

The WTP question based on referendum style, which presents the respondent with a single bid price to be accepted or rejected, minimizes strategic bias and it is incentive compatible.

Bateman et al. 2005, Carson et al. 2000, Hoehn and Randall 1987

Hypothetical bias

Attitude-behaviour difference – respondents systematically overestimate their WTP in many CV studies due to their ‘hypothetical nature’. Cited as biggest drawback in the literature resulting in significant different welfare estimates between an individual’s stated WTP and his true WTP. In general, revealed WTP actually in the market when goods are provided may be less than the committed WTP during the survey.

Cummings and Taylor 1999, Johannesson 1996, Diamond and Hausman 1993

“a single purchase of an unfamiliar commodity represents a guess as to what the commodity might be worth, rather than an evaluation based on experience”

Diamond and Hausman 1993

Respondents have no underlying information or preference upon which to base their decision concerning goods’ value to them. Therefore, hypothetical bias is highly likely because it is rational for respondents to overstate their WTP values

Harrison and Rutström 1999, Seip and Strand 1991, Duffield and Patterson 1992, Loomis et al. 1994, Neill 1995, Cummings et al. 1995,

Page 29: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 34

TERI University Ph.D. Thesis, 2006

Bias-type Description References in a hypothetical scenario Brown et al. 1996, Fox et

al. 1998 “Respondents must be reminded of his

budget constraint; otherwise, he may exaggerate the value that he places on the hypothetical good”

Cropper and Oates 1992

Non-commitment – intention and behaviour is cross-checked when the customers are actually provided with the goods and services. “Survey responses do not require the same level of financial commitment as a ‘real’ purchase done in the market”.

Kemp and Maxwell 1993, Bishop and Heberlein 1986

Some of the studies could not find any convincing evidence of significant hypothetical bias. Particularly, it is difficult to attribute discrepancies between responses to hypothetical and real valuation questions to hypothetical bias alone, as combination of factors may be at work.

Brookshire and Coursey 1987, Dickie et al. 1987, Smith and Mansfield 1997

“problems due to the hypothetical nature of CVM questions arise more frequently and are more serious when the choices are between goods with which people are not familiar”

Chowdhury 1999

Compliance bias

Respondent tries to fulfill his perceived expectations of the interviewer or the purpose of the study unconsciously or consciously, i.e. they give responses which they believe the interviewers are looking for.

Mitchell and Carson 1989

This bias can be reduced by taking care not to provide respondents with any additional instructions beyond those that are carefully prepared in advance and given to all.

The extent of these types of biases will depend in part on whether or not the method for eliciting individuals’ true preferences is “incentive-compatible”. A demand revelation method is incentive-compatible if it creates a situation where it is in the individual’s best interests to reveal his true preferences for the good, and he will not be tempted to engage in free-riding behaviour.

Griffin et al. 1995, Mitchell and Carson 1989

Information Similar to the hypothetical bias. The WTP Loomis et al. 2000

Page 30: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 35

TERI University Ph.D. Thesis, 2006

Bias-type Description References bias estimates will depend upon how much and

what type of information is presented to the respondents in the description of the hypothetical non-market scenario.

Reduced variations in the results of final bids with increasing information

Pearce and Morgan 1994

Elements of a hypothetical scenario such as the payment vehicle and the amount of information provided to the respondents influence WTP values, since respondents are not valuing an abstract amenity, but a policy package.

Arrow 1986, Kahneman 1986, Randall 1986

Effect of information bias is very less compared to the other forms of bias.

Mitchell and Carson 1989

Scenario misspecification bias

Such biases can be a problem if the respondent perceives some aspect of the contingent market incorrectly. This bias may stem from either theoretical or methodological problems.

Loomis et al. 1996

Theoretical misspecification bias occurs if the scenario described by the researcher is incorrect from a theoretical standpoint or is contrary to established facts. In such a case, although the respondent may understand the presented scenario perfectly, his indicated WTP values are not valid

Mitchell and Carson 1989

Methodological misspecification occurs when some aspect of the contingent market is inadequately described, and thus respondent perceives the scenario in a different way unintended by the researcher

Cooper 1997

Elements of the hypothetical scenario that may be misinterpreted include descriptions of risk, the payment vehicle, and the implied budget constraint

Mitchell and Carson 1989

Giving respondents more time to think about their responses leads to a lowering of bids

Whittington et al. 1993

Non-response bias

Sample non-response leads to inconsistent welfare estimates and a vague mean WTP presenting a vexing problem in the CV studies

Seung-Hoon Yoo and Hee-Jong Yang 2001, Schuman 1996, Whitehead et al. 1993, Mattsson and Li 1994, Whitehead 1994

Page 31: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 36

TERI University Ph.D. Thesis, 2006

2.6 Conclusion: synthesizing common thread

The review has seen that most applications in the literature involve creative

blending of theory with the application to an empirical framework in diverse

situations using either revealed or stated preference techniques. It emerges from the

review that site and purpose-specific situations are crucial in the choice of a

particular valuation technique. Even though revealed preference methods reflect

real choices and take into account various constraints on households’ decisions,

such as market imperfections, budgets and time, they fail to provide the value of

changes in the quality or quantity provided by the policy change in the provision of

the good (Louviere et al. 2000). The improvement in service conditions of an

environmental good can be better estimated directly from stated preference

techniques, which are estimated from assessing customers’ choice through

administering specially designed questionnaires (Hensher et al. 2005, Louviere et

al. 2000). These techniques have emerged as the most widely used methods to

estimate customer’s preferences for non-market goods that have no well-defined

market price or market is imperfect or incomplete2. Several researchers have

effectively implemented the outcome of stated preference studies to design

appropriate public policy to determine affordability and WTP affecting water service

improvements and the price levels to achieve them (Mbata 2006, Casey et al. 2006,

Gloria et al. 2005, Ntengwe 2004, Davis 2004, Whittington et al. 2002, 1990ab,

1991, Goldblatt 1999, Asthana 1997, Griffin et al. 1995, McPhail 1993, Altaf et al.

1993). More advanced and recent studies are those by MacDonald et al. (2005),

Willis et al. (2005), Ahmad et al. (2005) and Hanley et al. (2005). The common

thread of these studies was on valuation of water supply improvements in terms of

customer’s WTP. Overall, the available literature indicates that the amount that

customers are willing to pay for improved water services varies widely depending

upon the scope, context and extent of reliability improvements. While using stated

preference techniques, bidding game elicitation formats outscore than the other

formats.

In the literature, most studies conducted in developing countries treat water as a

homogenous good even though there is a clear preference for quality differentiated

supply. Households are willing to pay higher amounts for better water quality and

assured reliability. People have different preference structure for water supply

services and they differ in their willingness to pay depending upon their quality,

taste, perception and other socio-economic characteristics. The service attributes for

2 According to Bishop (2003), CV is by far the most widely applied nonmarket valuation technique the world over. He writes – “without stated preference methods, there would still be an environmental economics, but its usefulness would be much diminished”.

Page 32: 2 Review of Literature - INFLIBNETshodhganga.inflibnet.ac.in/bitstream/10603/6299/11/11...2.2.1 Hedonic pricing techniques Pioneered by Griliches (1971) and formalized by Rosen (1974),

Review of literature 37

TERI University Ph.D. Thesis, 2006

urban water supply being provided by either markets or government institutions are

‘heterogeneous in their degree of marketability’ because of heterogeneous

customers’ preference and socio-economic settings, and, therefore, a site-specific

approach is required to assess demand and to provide adequate supply.

It emerges from the literature review that customers’ preferences and choice of

water supply options in a diverse planning environment has immense potential for

improvement by understanding their demand and preference structures. There is a

need to fill this gap by studying preference heterogeneity for quality and reliability

improvements in a mega city in heterogonous planning environments in both

planned and unplanned settlements.


Recommended