Comparing Contingent Valuation and Averting Expenditure Estimates of the Costs of Irregular Water Supply

Ecological Economics 146 (2018) 250–264

Contents lists available at ScienceDirect

Ecological Economics journal homepage: www.elsevier.com/locate/ecolecon

Analysis

Comparing Contingent Valuation and Averting Expenditure Estimates of the Costs of Irregular Water Supply

MARK

Jennifer Orgill-Meyera,⁎, Marc Jeulandb, Jeff Albertc, Nathan Cutlerd a

Franklin and Marshall College, Department of Government, Public Health Program, United States Duke University, Sanford School of Public Policy, Duke Global Health Institute, United States c Thrive Water, United States d USAID, United States b

A R T I C L E I N F O

A B S T R A C T

Keywords: Nonmarket valuation Convergent validity Water quality Water reliability Utility water supply

We compare two methods—contingent valuation and averting expenditures—to measure the demand for improved water reliability in urban Jordan. Traditionally, averting expenditures (a revealed preference measure) have been considered a lower bound for demand relative to contingent valuation (a stated preference measure) estimates. We develop a theoretical model to show that this relationship critically depends on household perceptions. In our setting, this insight is important, because households appear to have relatively low confidence in both the reliability and quality of existing water supplies, even though water quality tests suggest that utility water is safe to drink from a microbial perspective. Averting expenditures, which reach 4% of monthly expenditures on average, include substantial purchases of non-network water sourced from water shops or tankers, as well as costs in terms of water collection time, storage and in-home treatment. In contrast, the contingent valuation responses, while correlated with coping costs, reveal low willingness to pay for increases in water reliability from the utility network. We attribute this departure from the traditional relationship between averting expenditures and contingent valuation to the lack of household confidence in the quality of utilityprovided water. Our study thus adds to previous evidence in the literature, which points to the importance of consumer perceptions in determining demand for environmental improvements.

1. Introduction

fueling a need for methods that would support allocation of this nonmarket resource, as well as investment to support it. Due to the general lack of market mechanisms to allocate water (and other similar environmental goods), the economic value of water is rarely directly observed. Thus, an economic perspective on the problem of managing complex water resource tradeoffs necessitates careful valuation work, to understand both the efficiency and distributional consequences of investments and institutional changes in the sector. Indeed, environmental economists have developed a sophisticated body of methods – both revealed and stated preference approaches – to measure the ex ante demand for such nonmarket environmental improvements. Stated preference (SP) methods include contingent valuation or choice experiments, and rely on survey responses from which the willingness to pay for specific changes can be derived. Revealed preference (RP) methods, on the other hand, examine individuals' existing choices in order to make inferences about the marginal benefits of similar improvements. Such RP

As global population increases and consumption of water continues to rise, concerns that humankind is entering a new age of global water scarcity are increasingly common (Postel, 1997; Vörösmarty et al., 2000). To some, rising water scarcity is uniquely worrisome because this resource is essential for myriad purposes – for drinking and critical domestic uses, as an input to food and industrial production processes, and for general human and ecological well-being – and yet is rarely allocated using mechanisms that effectively manage scarcity (Hanemann, 2005; Rijsberman, 2006; Whittington, 2016).1 Because of the essentialness of this resource, many argue that growing scarcity creates a zero sum game that will inevitably lead to widespread social destabilization and environmental damage. And though many populations already face water availability problems and still find ways to manage complicated tradeoffs between uses, increasing scarcity is

⁎

Corresponding author. E-mail addresses: [email protected] (J. Orgill-Meyer), [email protected] (M. Jeuland), jeff[email protected] (J. Albert). Besides essentialness, other features of water that challenge the use of markets for allocation include: high spatial and temporal variability; renewability and mobility; varying degree of non-rivalness (e.g., for non-rival recreation or spiritual aspects) and non-exclusivity; high fixed cost of transport and storage (which challenges reallocation across space and leads to natural monopoly); and pollutability. Others point out that people simply do not think about water as they do about other resources, see for example Whittington (2016). 1

http://dx.doi.org/10.1016/j.ecolecon.2017.10.016 Received 21 December 2016; Received in revised form 26 September 2017; Accepted 12 October 2017 0921-8009/ © 2017 Elsevier B.V. All rights reserved.

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

urban and rural environments. Households in many low-income settings compensate for unreliable and low-quality water supply by spending time and money seeking alternative sources or engaging in expensive private treatment (Whittington et al., 1990). Consequently, prior research has found that households gaining access to dependable water supply benefit from time savings, productivity, and positive changes in quality of life (Devoto et al., 2011). Galiani et al. (2009) show that expansions in the water supply network reduce household water expenditures by decreasing household reliance on more costly and distant water sources. The literature also provides evidence of health benefits and better economic outcomes from investments in piped water supply. For example, the expansion of piped water supply has been found to significantly reduce childhood diarrhea and child and infant mortality rates (Galiani et al., 2005; Gamper-Rabindran et al., 2010). These changes may stem from utilities' ability to efficiently treat water and subsequently provide high quality water through piped networks (Alsan and Goldin, 2015; Cutler and Miller, 2005). In contrast, epidemiological meta-analyses fail to find convincing evidence that water supply improvements (in the absence of complementary water quality improvements) deliver reliable health gains (Fewtrell et al., 2005; Waddington and Snilstveit, 2009). Our study carefully considers how particular constraints – household perceptions – affect demand for such water supply improvements, which may impede the success of interventions (Jeuland et al., 2015).3

approaches include the travel cost method (TCM), hedonic valuation and averting expenditure methods, among others. Given the range of valuation methods that exist, it is natural to wonder whether they generate consistent measures of environmental benefits, and under what conditions. In fact, comparisons of stated and revealed preference methods have a long history in environmental economics (Carson et al., 1996; Haener et al., 2001; Whitehead, 2005; Whitehead et al., 2010). This literature has primarily focused on the estimation of recreational demand in higher income countries, yielding many comparisons of stated preference and travel cost measures of willingness to pay. Generally, the literature has concluded that stated preference estimates have convergent validity with those obtained from the TCM, and that the latter therefore often provides a meaningful measure of WTP even if it only reflects use values. In this paper, we provide a different and much less common comparison that is perhaps more relevant to the problem of water scarcity and analysis of water allocation tradeoffs – that between averting expenditures (or coping costs) and contingent valuation. These methods are frequently used to estimate the economic benefits of improvements in water supply, but existing literature does not provide conclusive evidence on the nature of the relationship between them. Building on earlier insights in environmental economics (Freeman, 1979; Mäler, 1974), Wu and Huang (2001) provide a theoretical model suggesting that averting expenditures are a lower bound for WTP. Defensive expenditures in most settings cannot feasibly reduce the effects of inadequate environmental services to zero, i.e., they are not perfect substitutes for environmental improvements. Pattanayak et al. (2005) produce a similar result in an application to improved water supply in Kathmandu, Nepal, but Rosado et al. (2006) question whether contingent valuation and averting expenditures are measures arising from the same valuation process. This paper adds to this limited existing literature by assessing the extent to which these two valuation methods are related, demonstrating that subjective perceptions of quality provide a key reason why they may be different, and discussing their relevance for thinking about the value of water service improvements. We also provide new evidence on the economic burden of unreliable water supplies in a particularly water-poor country in the Middle East. This evidence is timely because one of the Jordanian government's major current objectives is to improve water security for its urban population. This goal is being supported by numerous policy reforms and changes in the water sector, including corporatization of municipal water utilities, reallocation of water to higher value users, investment in improved infrastructure, and development of expensive alternative water sources (Royal Commission for Water, 2009). All of these changes are occurring in a complicated political economy context that strongly constrains opportunities for reform through more rational pricing of water, due to widespread popular opposition to higher water bills (Haddadin, 2006).2 In Section 2, we discuss the current literature on the measurement of social benefits from improved water supply, paying particular attention to averting expenditure or coping cost methods. Section 3 presents a theoretical model that considers more carefully the issue of substitutability across water sources. We describe the study setting and our data in Section 4. In Section 5, we provide an overview of our empirical methods. We present results in Section 6 and conclude in Section 7.

2.2. Evidence of Benefits From Ex Ante Averting Expenditure Studies In the context of water supply and sanitation, averting expenditures, or coping costs, refer to monetized coping behaviors that households undertake when faced with intermittent water supply. Such behaviors can include purchases of alternative water sources, time spent seeking alternative sources of water (e.g., from water shops or tankers), and treatment costs to improve water quality. Multiple studies have documented the significance of such coping costs in settings with irregular or contaminated water supplies (Katuwal and Bohara, 2011; Pattanayak et al., 2005). Quantifying and monetizing these coping costs provides a potentially useful measure of the social benefits from improvements to municipal water networks, if such improvements eliminate the need for such behaviors. Environmental economists generally argue, however, that such measures represent a lower bound for benefits, since households may be willing to pay considerably more than what they can privately spend to improve water supplies, due to technological or institutional constraints (Freeman, 1979). 2.3. Contingent Valuation

There is a well-established public health and economic literature on the social benefits of investments in improved water services across

An alternative way to measure the ex ante social benefits from water supply investments is to directly elicit the stated willingness to pay (WTP) for such improvements, using stated preference methods such as contingent valuation (CV) (Carson, 2000). To elicit demand using CV, survey enumerators present to respondents a detailed scenario of a hypothetical change (e.g., a discrete improvement in water reliability) coupled with a payment mechanism that is well suited to the specific context and characteristics of the improvement. Price levels that households would need to pay (using a credible payment vehicle such as an increased tariff) for the improvement are then randomized across respondents, who indicate during the interview whether or not they would pay the specified price for it. The main critique of CV studies is that they suffer from hypothetical bias (Ajzen et al., 2004; Blumenschein et al., 1998; Murphy and

2 Urban water tariffs in Jordan are higher than they are in many Middle Eastern countries, but most utilities nonetheless do not fully recover costs (Jordanian Ministry of Water and Irrigation, 2013; Sommaripa, 2011).

3 It is important to note that we are defining the success of interventions from the perspective of benefits reaching households. There are other beneficiaries from water supply improvements, such as water utilities, which may reduce non-revenue water.

2. Background 2.1. The Benefits of Improved Water Supply: Empirical Evidence

251

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

max U = (GαWβ, L, Z; θ)

Stevens, 2004). Since households are not actually required to pay the price, they may provide a dishonest response, which is typically thought to bias WTP estimates upwards. Yet research has shown that careful survey design, the use of a realistic payment mechanism, and extensive enumerator training can minimize hypothetical bias (Blumenschein et al., 1998; Carson, 2000), and CV is now widely applied by researchers to assess the value of potential water supply improvements in many low and middle-income countries (Van Houtven et al., 2017; Whittington, 2010; Dutta et al., 2005; Whittington et al., 1991; Whittington et al., 2002; Briscoe et al., 1990; Casey et al., 2006; Goldblatt, 1999). Nonetheless, there are relatively few comparisons of CV and revealed preference measures in this domain. Only two papers have compared CV and averting expenditure measures, for example (Pattanayak et al., 2005; Rosado et al., 2006). Pattanayak et al. (2005) directly compared CV measures of demand for improved water supply with averting expenditure measures using data from Kathmandu, Nepal. Consistent with expectations from the environmental economics literature, they find that averting expenditure measures lie below stated WTP. Rosado et al. (2006) use a different approach; they estimate a bivariate probit model based on both CV and coping cost data, allowing for heteroskedasticity between and within measures. These authors find no evidence that CV and averting expenditure data arise from the same structure of preferences. In contrast to these papers, we show how household perceptions can influence measures of demand and reverse the conventional relationship whereby averting expenditure estimates are a lower bound for those derived using CV.

s. t. tG + pW + Z ≤ Y E+L+C≤T wE=Y G ≤ Gmax

(2)

The amount spent on network water, non-network water, and the numeraire commodity must be less than or equal to total income (Y), which is equal to the total time spent on employment (E) multiplied by the hourly wage (w). t represents the tariff on network water,4 and p is the price of non-network water. Further, total time spent on employment, leisure, and treating network water (C) is less than or equal to the total time endowment (T).5 We assume that time spent treating network water decreases as perceptions of network water (α) improve. We also assume that time spent treating network water increases as the total amount of network water consumed increases. Mathematically, C = h (α) and C = g(G) where:

∂C <0 ∂α ∂C >0 ∂G This formulation is important because it leads to imperfect substitutability between network and non-network water. As perceptions of network water improve, treatment costs decrease, causing households to substitute away from non-network water sources. However, as households consume more network water, treatment costs increase. Finally, the last constraint in the model illustrates that a household is constrained in its choice of G by network water reliability (Gmax). In other words, households cannot consume more network water than they receive from the network provider.

3. Theoretical Framework To help illustrate this idea, we start by developing a theoretical model that shows how households might respond to increases in water reliability, and to examine how perceptions of network water affect household water source decisions.

3.2. Household Responses to Increased Network Water Reliability

Our theoretical model offers a simplified representation of the willingness to pay for improvements in water supply, and takes as its starting point the idea that different sources of water may not be perfect substitutes for reasons that go beyond water availability. We thus begin by assuming that household utility is made up of total water quantity consumed, where G represents network water, and W represents nonnetwork water. We use L and Z to denote leisure and a numeraire commodity, respectively. Household preferences enter the utility function U through θ such that:

To explore how households respond to improvements in network water reliability, we examine two types of households: households for whom the last constraint does not bind (G < Gmax) and households for whom the last constraint does bind (G = Gmax). The first type of household's optimal choice of network water quantity (G*) is not constrained by network water reliability. For this type of household, a positive change in water reliability, Gmax, has no ∂U = 0). As a result, we expect a $0 stated effect on household utility ( ∂G max willingness to pay for improvements in water reliability. In this case, averting expenditures may exceed stated WTP as households may still be consuming non-network water. For the second type of household, without loss of generality, we assume that the other constraints bind and write the Lagrangian as follows:

U = (GαWβ, L, Z; θ)

 = U (GαWβ, L, Z; θ) + λ(w(T − L − C) − tGmax − pW − Z)

3.1. Theoretical Model Overview

(1)

(3)

As in Pattanayak et al. (2005), we employ duality to rewrite the maximization problem in terms of the expenditure function shown below, where U* represents the optimal level of utility.

We assume that UZ > 0 and UL > 0. Note that we have modeled utility to be Cobb-Douglas with respect to network and non-network water, to account for imperfect substitutability between these sources. In addition, we assume decreasing returns from these water sources (α + β < 1). Parameters α and β are functions of household perceptions of network water and non-network water, respectively. Such perceptions include opinions about water quality, taste and pressure, utility corruption, etc. Others have allowed water to enter the utility function indirectly through its effect on health (Pattanayak et al., 2005) or its use in the production of other commodities (such as gardening) (Abrahams et al., 2000). In our study site in Jordan, households use water for a variety of purposes, and so we model water purchases entering directly into the utility function to not limit our results to particular uses of water. The household maximization problem is:

ψ = w(C + L − T) + tGmax + pW + Z + μ(U∗ − U (GαWβ, L, Z; θ) ) (4) The household's problem is thus transformed from a utility maximization problem to an expenditure minimization problem, as is common in the environmental valuation literature (Freeman, 2003). We are interested in exploring the effects of a change in the quantity of

Water in our sample area is metered; households do not pay a flat fee. We assume that treatment of non-network water is negligible, which our data support. 4 5

252

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al. ∂2C

network water supply. Applying the envelope theorem, and noting that the optimal levels of C* and W* depend on Gmax, we observe the following:

a decreasing rate such that ∂G∂α is negative. As a result, we can conclude ∂G ∂α

> 0. In other words, as perceptions of network water improve, that households will consume more network water. This result is important, because as households consume more network water, they approach Gmax and move from the first case detailed above (WTP = 0) to the second case. An important caveat is that when perceptions of network and non-network water are equivalent (α = β), then perceptions do not affect the optimal choice of G* (this follows from Eq. (9)).

∂ψ ∂C ∂W =w +t+p ∂Gmax ∂Gmax ∂Gmax −μ

β ∂W ⎞ ⎤ ∂U ⎡ α β ⎛ α ∂G + G W ∂Gmax ⎢ W ∂Gmax ⎠ ⎥ ⎝ G ∂Gmax ⎣ ⎦ ⎜

⎟

(5)

The additional expenditure that households are willing to incur, or WTP, for an increase in water reliability is made up of (1) the savings from avoided monetary expenses for non-network water supply (the third term) plus (2) the additional utility that comes from consuming new levels of network and non-network water supply (the final term), net of (3) the time households must spend treating additional network water (the first term), and (4) the additional monetary expense spent on the greater quantity of network water (the second term). We expect stated willingness to pay (the sum of terms 1–4) to exceed averting expenditures, and therefore consumption of network water to increase, if and only if this entire quantity is positive, which will typically be the case when the household is consuming at Gmax (except for the very special case where G* = Gmax). It is important to note that our model only includes recurring or ongoing coping costs, or non-network water expenditures and treatment time costs. In reality, households may also invest in “sunk” coping costs such as large storage containers to store water. If these “sunk” costs are included in the definition of coping costs, averting expenditures may exceed stated WTP even if the terms on the right hand side of Eq. (5) are net positive, and even when household consumption of network water increases following the reliability improvement.

3.4. Implications of Theoretical Model Through this theoretical model, we have shown that there are certain cases where averting expenditures may exceed stated WTP, reversing the nature of the relationship that has been identified in previous literature (Mäler, 1974; Wu and Huang, 2001). Specifically, averting expenditures may exceed stated WTP when the optimal choice of network water quantity is not constrained by reliability, and if the additional expenses of network water and treatment costs exceed the savings on non-network water and direct utility benefits (see Eq. (5)). We also discuss how averting expenditures may exceed stated WTP if we include sunk averting expenditures in the latter term, for example large storage containers or expensive treatment technologies. Finally, we show the importance of network water perceptions; as perceptions of network water improve, a household's optimal quantity of network water increases. This increase in the optimal quantity of network water increases the probability of a positive stated WTP as G* approaches Gmax. 4. Study Setting and Data

3.3. Effect of Water Quality Perceptions on Choice of Network Water Quantity

Jordan is one of the most water poor countries in the world (Haddadin, 2006). In 2007, the country had a per capita annual water share of 166.4 m3, nearly 3.5 times below the average of other Middle Eastern countries, and 40 times below the world average (FAO, 2007). In recent years, population growth, urbanization, an influx of refugees, and extremely high water losses have all intensified this strain (Jaber and Mohsen, 2001; Haddadin, 2006; Royal Commission for Water, 2009). A general definition that covers both physical and administrative water losses is one pertaining to volumes that fail to produce revenues (i.e. non-revenue water) (Al-Ansari et al., 2013). Such water losses typically result from degraded and leaking pipes in the distribution system, broken meters, illegal water usage, poorly installed meters, and/or weak governance (Al-Ansari et al., 2014). These losses contribute to poor cost recovery, and to a lack of maintenance and further degradation of water delivery systems, both of which exacerbate water scarcity.6 Our sample includes households located in two governorates: Zarqa and the eastern part of Amman.7 We randomly sampled households living in select census blocks throughout the two sample areas. In total, we surveyed 3358 households, with 2258 residing in Zarqa, and 1100 in Amman.8 This survey generated the baseline data for the impact

As shown above, a determining factor for whether averting expenditures exceed stated WTP is the extent that a household's optimal choice of network water quantity is constrained by network water reliability. In this section, we illustrate how perceptions of water quality (α) may affect the optimal choice of network water quantity. In the original optimization problem, households are choosing both G and W such that the Lagrangian is:

 = U (GαWβ, L, Z; θ) + λ(w(T − L − C) − tG − pW − Z)

(6)

and:

∂ ∂C = αGα − 1Wβ − λ ⎛w + t⎞ ∂G ⎠ ⎝ ∂G

(7)

∂ = βWβ − 1Gα − λp ∂W

(8)

Setting both conditions (7) and (8) to 0 and rearranging terms, we can solve for G*:

αWp

G∗ =

(

)

∂C

β w ∂G + t

6 On average, Al-Ansari et al. (2014) estimated that water losses countrywide reached 75 million Jordanian Dinar (JD) in 1998 (equivalent to 154 million 2014 USD), and predicted that this figure would continue to grow as population increased. Consistent with this expectation, the volume of NRW country-wide has increased from 124 million m3 in 2000 to 183 million m3 in 2013 (MWI 2013). 7 We chose to survey households in Zarqa and adjoining parts of Amman as part of the baseline activities for an impact evaluation of a large planned investment in water and sewerage in this governorate. This US$278 investment was prioritized and planned by the Government of Jordan and funded by the Millennium Challenge Corporation (MCC). It aims to reduce poverty and stimulate economic growth by increasing the effective supply of water available to inhabitants. The increase in supply would come from improvements in the efficiency of water delivery, the extent of wastewater collection, and the capacity of wastewater treatment. More information on the Jordan Compact is available at: https:// www.mcc.gov/where-we-work/program/jordan-compact. 8 Note that this sample is not representative of the population, as it was constructed to

(9)

To examine how consumer perceptions affect the optimal choice of network water quantity, we take the derivative of G* with respect to α as shown below:

∂G = ∂α

(

(

)

2

)

∂C ∂ C Wp ⎡β w ∂G + t ⎤ − αWp ⎡β w ∂G∂α ⎤ ⎣ ⎦ ⎣ ⎦

⎡β ⎣

(

∂C w ∂G

)

2

+t ⎤ ⎦

(10)

∂C ∂G

> 0 so the first term in the numerator is positive. We Recall that assume that as network water perceptions improve, the amount of additional treatment needed for additional government water increases at 253

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

(fecal coliform and E. coli). We found little evidence of E. coli contamination problems in either stored network or stored shop water—only three of 91 samples of stored shop water had modest E. coli contamination (7, 28, and 54 CFU/100 mL respectively). Total coliforms were found in 10% of stored network water samples, but the counts were all low (< 100 CFU/100 mL in all but one case). In contrast, we detected total coliforms in over 70% of shop water with 29% of the samples having > 100 CFU/100 mL. These higher rates of contamination in shop water are likely because this water is stored for longer periods of time or because of poor storage container maintenance. These findings together with household perceptions of network water quality suggest that households may be overly confident in the quality of shop water, relative to network water. We observe five primary water-related coping costs in our data: Water purchases from non-network water expenditures, water collection times, water treatment costs, water storage costs, and repair costs for household plumbing, storage, or infrastructure. Table 2 contains a summary of how these coping costs were calculated from the data. In addition to coping costs, we also recorded information from the most recent water bills in households that had their water bill on hand (37% of households). These bills allowed us to observe the most recent water usage and water costs, and compare expenses for network water to coping costs. Lastly, the survey included a CV experiment. Households were offered a scenario in which they were asked whether they would vote in favor of an improvement in their current water supply reliability given a specified increase in their bill. Three levels of the reliability increase – an additional 24 h/week, 72 h/week, or an increase to continuous supply – were randomly assigned. The corresponding (also randomized) hypothetical bill increases were 1, 2, 4, 6, 9, 12, or 25 JD per month.11 The CV scenario was designed to measure the ex ante benefits of the infrastructure investment that would accrue to households, and thus targeted improvements to water supply only (and not water quality, which will largely remain unchanged). The randomized bill increases range from 0.2–5.5% of average monthly household expenditures. Households were then asked whether they would vote in favor of the change. After the vote was recorded, we asked households a number of debriefing questions to see why they responded the way they did, and to measure their level of certainty about their responses. The Appendix contains the full CV script.

Table 1 Descriptive statistics. N

Mean

SD

Min

Max

Demographics Household size Number of children under 5 Female head of household Number of disabled HH members Jordanian Zarqa resident

3359 3359 3359 3359 3359 3359

4.91 0.42 0.15 0.07 0.93 0.67

2.05 0.70

1 0 0 0 0 0

14 5 1 6 1 1

Socioeconomics Average number of years of adult education Homeowner Total expenditure Own washer Own computer Own air conditioner Own vehicle Took out a loan in the past year

3359 3359 3348 3359 3359 3359 3359 3359

10.6 0.73 451 0.98 0.45 0.20 0.45 0.20

3.45

0 0 80.5 0 0 0 0 0

17 1 868 1 1 1 1 1

3319 3359

0.71 5.12

3.22

0 0

1 10

3245 3281

0.19 2.83

1.17

0 1

1 5

Perceptions and attitudes Believe diarrhea can be prevented Perceived safety of WAJ water (0 = not at all, 10 = completely) Had complaints about WAJ water reliability Sanitation situation: 1 = excellent, 5 = very poor Reported experiencing water shortage

0.30

136

3343

0.23

0

1

Water sources Subscribed to WAJ water Subscribed to WAJ wastewater Use water from a borehole or a well Use water from water shops Use water from tankers Use other water source

3359 3245 3359 3359 3359 3359

0.97 0.79 0.03 0.32 0.04 0.02

0 0 0 0 0 0

1 1 1 1 1 1

Water and sanitation behaviors Treats water currently Keeps storage containers elevated Had soap on hand Has a private toilet Has a toilet with no sewer

3359 3177 3135 3359 3359

0.36 0.68 0.88 0.76 0.24

0 0 0 0 0

1 1 1 1 1

evaluation of the infrastructure investment described in footnote 7. It elicited information about demographics, socioeconomic factors, household members' health, water sources, water treatment and storage behaviors, and sanitation behaviors. Additionally, enumerators observed water bills for households subscribed to the Water Authority of Jordan (WAJ) and observed water meters whenever possible. The survey was pre-tested in 50 households in non-sample areas in Zarqa. Table 1 provides descriptive statistics for our sample. The average household has 5 members and owns its home. At the time of the survey, 97% of households were subscribed to WAJ for water services, but 38% of households supplemented their water supply with water from shops. Additionally, 36% of households also reported treating their water in some way.9 Most households were not confident in the safety of the water provided through WAJ.10 We collected water samples of stored network and shop water from a random subsample of 450 households for microbiological testing

5. Empirical Strategy 5.1. Coping Costs Non-network water expenditures were measured directly in the survey—as shown in Table 2. We summed the monthly total that respondents reported spending on each (of seven total) non-network water sources.12 Calculations of water collection, treatment, and storage costs require additional assumptions, primarily about the value of time and lifespan of equipment. Infrastructure costs were converted to monthly equivalents by amortizing them using a 5% discount rate.13 For water collection, the survey provides estimates of the time spent on each water collection trip. Based on frequencies reported in a follow up survey with the same households, we assume that households make 2.4 trips/week to collect water from each non-network source. We also value the time spent collecting water at half of the average hourly wage reported in the data. For water treatment costs, survey data provided household estimated monthly treatment costs from consumables, as

(footnote continued) inform an impact evaluation that will compare treated and non-treated areas (where treatment refers to improvement of water supply and sewer improvements). 9 4% of households report boiling water, 6% use simple filters (e.g. sieve through a cloth or net), and 25% use an advanced filter (e.g. ceramic filter or Aquaguard). Slightly > 1% of the sample uses reverse osmosis, chemical disinfectants, alum or other treatment methods. 10 We conducted a perception elicitation game in which respondents were asked to distribute 10 buttons across two piles—one indicating that network water was safe to drink and one indicating that it was not safe to drink. On average, respondents placed 5 buttons in the safe pile and 20% put no buttons in the safe pile. A number of other questions from our survey confirm that households are skeptical of network water quality.

11

1 USD = 0.71 Jordanian Dinar. The seven non-network water sources are boreholes, dug wells, water shops, water tankers, rainwater, bottled water, and other sources. 13 This is a common discount rate used for these types of infrastructure (Whittington et al., 2012). 12

254

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

Table 2 Averting expenditure calculations. Averting expenditure

Survey data used

Assumptions

Non-network water expenditures

• Estimates of monthly non-network water purchases (pi) for every non-network source (i)

None

Water treatment costs

• Estimated monthly treatment costs (c) • Estimated equipment costs (e)

• 5 year lifespan of equipment • 5% discount rate

Water storage costs

• Estimated monthly costs of cleaning storage containers (C) • Number of storage containers (S) • Estimated volume of storage container (v)

Repair costs

• Estimated yearly repair costs (r) • Estimated time spent on repairs (s)

• Price of storage container (P)—varying by size • Lifespan (l) of storage container—depending on size 5% discount rate • Value of time is average wage (w/2)

Water collection costs

• Estimates of water collection time (minutes) (ti) per trip for each non-network water source (i) • Trips per week (g)

Calculation 7

∑ pi i=1

• Value of time is average wage (w/2) • Four weeks/month

e ∗ 0.05∗1.055 ⎞ c+⎛ ⎞ ⎛ ⎝ 12 ⎠ ⎝ 1.055 − 1 ⎠ ⎜

C + ((S v ∗Pv )/12)

r + s∗

⎟

0.05∗1.05lv 1.05lv − 1

( ) w 2

12 7

∑ (ti/60)∗g∗ (w/2)∗4 i=1

then estimated from the log-likelihood function:

well as equipment costs. The equipment that households reported using most frequently was advanced filtration,14 which we assume to have a lifespan of 5 years (Whittington et al., 2007). The amortized monthly cost of equipment was added to the monthly cost of consumables to obtain the total water treatment cost. For the cost of storage equipment, we use market prices and varying lifespans measured in an informal survey of shops selling containers of various sizes and materials (e.g., plastic or concrete).15

N

l=

∑ Si Ti ln[1 − FWTP (t)] + Si (1 − T)i ln[FWTP (t) − FWTP (0)] 1

+ (1 − Si) ln[FWTP (0)]

(14)

where we assume the following logistic functional form for FWTP(t):

1 if t > 0 1 + eη − λt 1 FWTP (t) = if t = 0 1 + eη FWTP (t) = 0 if t < 0 FWTP (t) =

5.2. CV Estimates To calculate the mean WTP from the CV exercise, we utilize a spike model because of the very high number of refusals in the data, even at our lowest price offer (Kriström, 1997). The spike model assumes that the data follow a two-part distribution, with a mass of probability at a price of zero, followed by a continuous probability distribution at nonzero prices. To empirically estimate WTP using this model, we first define the following terms:

Si = 1 if WTP > 0 (0 otherwise) Ti = 1 if WTP > t (0 otherwise)

We can interpret η as the marginal utility of the reliability improvement, and λ as the marginal utility of income. The spike is the 1 probability that FWTP (t) = 1 + eη . Following Kriström (1997), and assuming that λ > 0, we calculate mean WTP to be:

ln[1 + eη] λ

5.3. Exploring the Relationship Between CV Estimates and Averting Expenditures Our final analysis aims to demonstrate how coping costs and CV estimates are related to one another. Using a logit model, we predict the probability of voting in favor of the change at an illustrative bill increase (t = 4) and reliability improvement (r = continuous) for all individuals in the sample.16 We then regress water-related coping costs on these predicted probabilities to assess the relationship between the two measures.

(12)

Accepti is a binary variable that indicates whether the respondent accepted the proposed price offer. Xi is a vector of individual characteristics, and t indicates the quarterly bill increase. The randomized reliability level (r) can take three values such that Z represents a 24hour increase, 72-hour increase, or continuous availability in water reliability. We estimate Eq. (12) for each reliability increase. We then use the relationship in Eq. (12) to predict the probability of accepting this near zero price offer for each individual in the sample. From these predictions, we create Si as follows:

Si = 1 if  Accept i > 0.5 Si = 0 if  Accept i ≤ 0.5

(16)

(11)

To estimate Si in our sample, we first estimate the probability of accepting the improvement among respondents that are given a zero or near-zero price offer (this was 1 JD in our data). Using a logit model, we estimate the following equation from the data:

Accept i = γ + δXi + ε if t = 1 and r = Z

(15)

 t = 4,r = continuous + εi Copei = ϕ + νAccept i

(17)

We bootstrap standard errors for ν using 1000 simulations. In Eq. (17), Copei refers to total averting expenditures. We also estimate Eq. (17) using both sunk and not-sunk coping costs as our dependent variables.17 The coefficient of interest, ν indicates the strength of association between the probability of accepting the 4 JD bill increase and the amount of water-related coping costs.

(13)

We additionally allowed Ti = 1 if respondents stated they were certain that they would accept a bill increase of t. The spike model is

16 We use these values to standardize the probabilities of accepting the WTP scenario across respondents. 17 We categorize storage and treatment equipment as “sunk” coping costs, because the household has already invested in these technologies and assuming no re-sale market, cannot recuperate these expenses. The coping costs that are not sunk are all of the continuing averting expenditures that the household makes, such as purchases of non-network water, collection costs, etc.

Advanced filtration was defined in our study as use of Aquaguard or a ceramic filter. Most of the households in our sample (92%) had very large outdoor water storage containers that we did not include in our coping cost calculations. We omit these containers because they are large one-time capital expenses that are often included in the price of the land or house itself. 14 15

255

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

Table 3 Coping costs summary.

Non-network water expenditures Treatment costs Storage costs Repair costs Collection costs Total water-related coping costs

N

Mean

SD

Min

Max

3359 3359 3359 3359 3359 3359

8.04 2.71 5.18 1.41 0.42 17.8

34.3 5.20 10.4 4.20 0.86 36.7

0 0 0 0 0 0

440 37.7 100 27.6 6.10 463

Percent of Total Expenditure

8

6

4

2

0

0-25%

25-50% Coping costs

2

50-75%

75-100%

Water Utility Bills

Fig. 2. Water-related coping costs and water utility bills as a percentage of total expenditures by expenditure quartile. 1

0

0-25%

25-50%

50-75%

Non-network water expenditures Treatment costs Collection time costs

Mean probability of accepting bill increase .1 0 .2 .3 .4 .5

Percent of Total Expenditure

3

75-100% Storage costs Repair costs

Fig. 1. Coping costs as a percentage of total expenditures, by expenditure quartile.

6. Results 6.1. Coping Behaviors and Costs

0

Table 3 presents a summary of the average of different categories of monthly coping costs. Non-network water expenditures represent the highest of these costs. For the average household in our survey, waterrelated coping costs are 18 JD, or 4% of average monthly expenditures.18 Fig. 1 shows the percentage of total expenditures that households spend on coping costs disaggregated by expenditure quartile.19 We observe that averting expenditures are a much higher percentage of total expenditure for poor households; those in the lowest quartile spend approximately 6% of their total expenditure budget on coping costs. While the main focus of this paper is on water-related coping costs, it is relevant to note that there are significant wastewater related coping costs in this area as well (see Tables 1A, 2A, and Fig. 1A in the Appendix).

5

10 15 Bill increase offered in WTP scenario

+24 hr/week water supply Continuous water supply

20

25

+72 hr/week water supply

Fig. 3. Probability of voting in favor of a reliability improvement given monthly bill increase (in JD).

6.3. Willingness to Pay for Improved Water Reliability Fig. 3 summarizes the CVM data, where the percent voting positively corresponds to those respondents who indicated they were very certain they would accept the price offer, and among households who passed a comprehension test indicating that they understood key elements of the scenario. Consequently, our results should be minimally affected by misunderstanding and hypothetical bias. We observe a downward sloping demand curve, with probability of accepting increased billing amounts decreasing in the level of the increase. We also observe that the probabilities of accepting the price offer increase with the level of reliability offered. Overall WTP derived from the CV experiment is fairly low; the probability of accepting continuous water supply is < 0.5 for even a small increase in the bill amount (1 JD). This is consistent with general opposition to water bill increases noted in previous work (Sommaripa, 2011), and motivates our use of the spike model to estimate average WTP, a procedure that yields a mean estimate of 4.2 JD/quarter, or 1.4 JD/month for continuous water supply (see Table 4). The mean monthly WTP is 0.55 JD/month and 1.34 JD/ month for a 24-h is 0.55 and a 72-h increase in water reliability, respectively. Survey responses indicate that most households who voted in favor of the bill increase did so because they suffer from a lack of reliability and think the improvement is worth the cost. Few households (2%) indicated that they thought the improvement would save them time and money. These responses may indicate that households are not fully

6.2. Coping Costs and Water Utility Bills The average monthly water utility bill (18 JD) is roughly equivalent to monthly coping costs in our sample. Fig. 2 shows that along every expenditure quartile, households spend roughly the same amount on averting expenditures as they do on water utility bill.20

18 Table 3A in the appendix contains descriptive multivariate regressions that relate household characteristics to coping costs. Important to note is the negative correlation between perceived network quality and coping costs. 19 We use predicted total expenditures to account for missing data and misreported data. The regression we use to generate these total expenditures is shown in Table 5A. 20 Given that only 37% of households had their water utility bill on hand, we used imputed water utility bills to create this figure.

256

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

summarizes the results of the regression analysis detailed in Eq. (17). The dependent variable in the first column is total water-related coping costs, and sunk and not-sunk coping costs are shown in the second and third columns, respectively. While we do not observe a statistically significant relationship between total water-related coping costs and the predicted probabilities of accepting the CV offer, we do observe statistically significant relationships between the two measures when we decompose coping costs into sunk and not-sunk costs (columns 2 and 3). Specifically, households that we predicted would accept the 4 JD bill increase had sunk coping costs that were roughly 5 JD lower than households that we did not predict would accept the bill increase. On the other hand, households that we predicted would accept the price increase had 16 JD higher non-sunk coping costs compared to households that did not accept the price increase.

Table 4 Mean quarterly WTP from spike models.

Spike Mean WTP Bootstrapped SE Mean WTP (USD) N

24-hour supply increase

72-hour supply increase

Continuous supply

0.67 1.65 (0.18) 1.17 1041

0.36 4.02 (0.24) 2.85 1187

0.29 4.21 (0.24) 2.99 1218

Predicted Probability of Accepting Price Offer

.6

.5

7. Conclusion and Discussion .4

This study considered the demand for improved water supply reliability in two urban settings in Jordan. Though we find a statistically significant positive relationship between stated WTP for improved reliability and non-sunk coping costs, we show that averting expenditures exceed stated WTP for greater reliability by a very large amount. On average, water-related coping costs make up 4% of monthly household expenditures, roughly equivalent to spending on monthly water utility bills. This relative burden, as a fraction of expenditure, is particularly significant for the poorest households. On average, we can interpret our regression results comparing stated WTP and coping costs to mean that a 16 JD in non-sunk coping costs would induce the average household to switch from rejecting an offer of continuous supply reliability at a price of 4 JD, to accepting it. In other words, households that had higher recurring coping costs such as non-network expenditures were more likely to be willing to pay for a reliability improvement in network water supply. In contrast, households with higher sunk coping costs (such as large capital investments in storage technologies) had lower stated WTP for water reliability improvements. This latter finding is consistent with the idea that such households are less affected by a constraint limiting their network consumption, given their investments in storage infrastructure. Our data also suggest that households are not confident in the quality of network water, and that they perceive that non-network water sources such as shop water provide safer drinking water, even though objective water quality measures do not appear to support this perception, at least by the time households consume that water. In the presence of such skewed perceptions, averting expenditures and WTP estimates of demand may diverge substantially, because households do not want additional network water, preferring instead to purchase expensive alternatives to it. Given the conventional wisdom in the literature that empirical measures of demand using averting expenditure methods provide a lower bound relative to the CVM, these findings were unexpected. Nonetheless, they are consistent with a model in which households consider different water sources to be imperfect substitutes due to differences in quality. Thus, while the reason for the divergence between measures appears to arise from the same underlying phenomenon related to substitutability that has been previously discussed in the literature, its policy implications are substantially different, as we discuss below. In addition, the correlation between stated WTP and coping costs suggests that the measures do reflect similar preferences. These findings have important implications for cost-benefit analyses of water supply improvements, which tend to use averting expenditures and CV measures interchangeably as measures of social benefits. The results from this study imply that the two measures, while related, may provide very different estimates, and should be interpreted carefully. Furthermore, we show that the generally assumed relationship—averting expenditures as a lower bound for CV—may not always hold, and that it likely depends on the degree of perceived

.3

.2

.1 0

100

200 300 Water Related Coping Costs

400

500

Fig. 4. Relationship between predicted probability of accepting CV scenario (4 JD monthly bill increase and continuous water supply) and coping costs.

Table 5 Relationship between stated WTP and coping costs.

Predicted probability of accepting 4 JD price increase continuous reliability offer R-squared Observations

(1) Total coping costs

(2) Sunk coping costs

(3) Not-sunk coping costs

10.8 (6.88)⁎

− 4.94⁎⁎⁎ (1.64)

16.2⁎⁎ (6.57)

0.001 3124

0.003 3124

0.002 3124

Bootstrapped standard errors in parentheses. ⁎⁎⁎ p < 0.01. ⁎⁎ p < 0.05. ⁎ p < 0.1.

internalizing or understanding the tradeoff between improved reliability and reductions in coping costs, or that they feel the savings will be limited because investments in storage and treatment equipment have been made, and/or network and non-network water are imperfect substitutes. The most common reason respondents cited for voting against the proposed improvement was the household budget constraint (50%), followed by water reliability not being a significant problem for the household (28%). Only 5% of nay-voters believed that the proposed scenario would not provide benefits—suggesting that protest voters made up only a small percentage of the nay-voters. Table 4A in the Appendix provides multivariate regressions that show correlations between household characteristics and WTP from the CV exercise.

6.4. Relationship Between Coping Costs and CV Estimates Fig. 4 presents the relationship between coping costs and the probability of voting in favor of an improvement to continuous reliability given a 4 JD increase in the monthly water bill. Table 5 257

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

commenced in 2014 that was meant only to improve reliability. Thus, our findings suggest that households may undervalue these new investments due to concerns over quality, and point to the need for further evaluation of the intervention's impacts on beneficiaries.

substitutability between water from different sources. In the survey, we explicitly told households that the quality of network water would not be affected. This was consistent with the relevant policy scenario being valued, which was a real-world water infrastructure investment Appendix A Table 1A Averting expenditure calculations for sanitation. Toilet infrastructure costs Time spent on trips to toilet

Pit emptying costs

• Estimated costs to replace toilet (r) • Connection fee to WAJ-wastewater (w)

• Average lifespan of toilet is 20 years • 5% discount rate • Estimated time (minutes) spent walking to toilet for households • Value of time is average wage that used shared toilets (t) (w/2) • 3 trips per day per household member. Household size = h • Cost of emptying pit (p) None • Frequency of emptying pit in months (m)

(r + w) ∗ 0.05(1.05)20 12 1.0520 − 1

t ∗ w ∗ ⎛ ⎞ ⎛ ⎞ 3h ⎝ 60 ⎠ ⎝ 2 ⎠

p∗ m 12

Table 2A Coping costs summary for sanitation.

Toilet infrastructure costs Time spent on trips to toilet Pit emptying costs Total sanitation related coping costs

N

Mean

SD

Min

Max

3359 3359 3359 3359

8.23 0.13 16.8 25.2

4.41 0.87 38.0 38.3

0 0 0 0

63.3 9.90 1230 1247

Table 3A Multivariate regressions of coping costs (CC).

Household size Head of household age Female head of HH Head of HH literate Monthly expenditures (JD) Index of asset ownership Enumerator rating of wealth (1 = very poor, 5 = rich) Resides in Zarqa Jordanian

Total waterrelated CC

Total water & wastewater CC

Total waterrelated CC

Total water & wastewater CC

0.12 (0.33) − 0.04 (0.06) − 1.38 (1.74) − 7.41⁎⁎ (3.71) 0.00 (0.01) 1.75 (1.21) 1.88⁎⁎ (0.79) 3.26⁎⁎ (1.44) − 2.39 (3.53)

1.05⁎⁎ (0.52) 0.06 (0.09) − 5.32⁎ (2.77) − 17.35⁎⁎⁎ (6.04) 0.01 (0.01) 0.34 (1.93) 4.28⁎⁎⁎ (1.24) − 0.15 (3.59) 1.28 (4.17)

−0.15 (0.34) −0.09 (0.07) −0.83 (1.74) −4.96 (3.70) 0.00 (0.01) 1.24 (1.24) 1.73⁎⁎ (0.79) 3.23⁎⁎ (1.39) −3.77 (3.70) 4.95⁎⁎⁎ (1.46) 1.67 (1.73) −10.29⁎⁎⁎ (3.18) 5.90⁎

− 0.05 (0.46) − 0.02 (0.08) − 2.96 (2.46) − 8.31 (5.73) 0.01 (0.01) − 0.01 (1.71) 2.37⁎⁎ (0.98) 3.37⁎ (2.03) − 2.86 (3.95) 5.35⁎⁎⁎ (1.60) 2.46 (2.22) − 62.57⁎⁎⁎ (4.10) − 6.09

Resident of current area for 8 years or more HH suffers from water shortage HH owns private toilet Connected to WAJ wastewater network

258

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

(3.12) 0.81 (2.86) −0.39⁎ (0.21) 2.03 (2.02) −1.38 (1.31) −3.86 (3.00) −2.44 (1.56) 25.87⁎⁎⁎ (8.35) (1.56) 3003 0.02

Filed complaint with WAJ in past year Perceived safety of WAJ water (0 = completely unsafe, 10 = completely safe) Household member had diarrhea in past 2 weeks

Rated sanitation situation as poor Received financial assistance from NAF Took out loan in the past year Constant

22.55⁎⁎⁎ (8.04)

36.04⁎⁎⁎ (11.48)

Observations R-squared

3124 0.01

3124 0.02

(4.36) 7.23 (6.32) − 0.12 (0.25) 2.81 (2.54) − 3.02⁎ (1.73) − 5.24 (3.80) − 2.43 (1.82) 90.92⁎⁎⁎ (10.38) (1.82) 3003 0.29

Robust standard errors in parentheses. ⁎⁎⁎ p < 0.01. ⁎⁎ p < 0.05. ⁎ p < 0.1.

Table 4A Logit regressions of WTP (CV Scenario)—only including respondents who understood scenario. (1)

Tariff offered in WTP scenario Increase reliability 72 h/week Increase reliability to continuous service

(2)

(3)

Accept WTP

Marginal effects

Accept WTP

Marginal effects

Accept WTP

Marginal effects

−0.11⁎⁎⁎ (0.01) 0.27⁎⁎ (0.13) 0.59⁎⁎⁎ (0.13)

− 0.02

−0.11⁎⁎⁎ (0.01) 0.32⁎⁎ (0.14) 0.59⁎⁎⁎ (0.14) 0.01 (0.03) −0.01⁎⁎⁎ (0.00) 0.14 (0.18) 0.34 (0.23) 0.00⁎⁎ (0.00) −0.15 (0.11) 0.03 (0.08) 0.68⁎⁎⁎ (0.13) −0.31 (0.24)

− 0.02

−0.12⁎⁎⁎ (0.01) 0.35⁎⁎ (0.14) 0.62⁎⁎⁎ (0.14) 0.01 (0.03) −0.01⁎⁎ (0.01) 0.25 (0.19) 0.48⁎ (0.25) 0.00⁎ (0.00) −0.10 (0.12) 0.05 (0.08) 0.61⁎⁎⁎ (0.15) −0.19 (0.26) −0.42⁎⁎⁎ (0.13) 0.86⁎⁎⁎ (0.13) −0.18 (0.22) 0.10 (0.23) −0.13 (0.21)

− 0.02

Household size fHead of household age Female head of HH Head of HH literate Monthly expenditures (JD) Index of asset ownership Enumerator rating of wealth (1 = very poor, 5 = rich) Resides in Zarqa Jordanian Resident of current area for 8 years or more HH suffers from water shortage HH owns private toilet Connected to WAJ wastewater network Filed complaint with WAJ in past year

259

0.05 0.11

0.06 0.11 0.00 − 0.00 0.03 0.06 0.00 − 0.03 0.01 0.12 − 0.06

0.06 0.11 0.00 − 0.00 0.04 0.08 0.00 − 0.02 0.01 0.11 − 0.03 − 0.07 0.15 − 0.03 0.02 − 0.02

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

Perceived safety of WAJ water (0 = completely unsafe, 10 = completely safe) Household member had diarrhea in past 2 weeks Rated sanitation situation as poor Received financial assistance from NAF Took out loan in the past year Constant Observations R-squared

−0.45⁎⁎⁎ (0.12) 1925 0.07

−1.43⁎⁎ (0.66) 1822 0.09

0.02 (0.02) −0.27 (0.18) −0.03 (0.12) −0.17 (0.42) 0.05 (0.14) −1.63⁎⁎ (0.71) 1757 0.12

0.00 − 0.05 − 0.00 − 0.03 0.01

1757 0.12

Robust standard errors in parentheses. ⁎⁎⁎ p < 0.01. ⁎⁎ p < 0.05. ⁎ p < 0.1.

Table 5A OLS regression predicting monthly expenditures. Reported total monthly expenditures Average adult education Own house Own washer Own computer Own air conditioner Own vehicle Took out loan in past year Number of household members with diarrhea in past 2 weeks Number of household members with other water related diseases Zarqa resident Jordanian Has children under age 5 Female head of household Constant Observations R-squared Robust standard errors in parentheses. ⁎⁎⁎ p < 0.01. ⁎⁎ p < 0.05. ⁎ p < 0.1.

Willingness to pay for hypothetical water supply reliability improvement

260

13.30⁎⁎⁎ (1.985) − 5.364 (11.73) 69.44⁎⁎⁎ (20.10) 79.30⁎⁎⁎ (12.39) 95.65⁎⁎⁎ (16.70) 130.2⁎⁎⁎ (10.75) 29.42⁎ (16.18) 1.606 (10.20) 6.749 (26.72) − 57.71⁎⁎⁎ (14.75) 58.76⁎⁎⁎ (15.54) − 17.02⁎⁎ (7.398) − 36.43⁎⁎⁎ (12.79) 121.2⁎⁎⁎ (27.69) 3261 0.16

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

261

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

262

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

network losses in Jordan. J. Water Resour. Prot. 6, 83–96. Alsan, Marcella, Goldin, Claudia, 2015. Watersheds in infant mortality: the role of effective water and sewerage infrastructure, 1880 to 1915. NBER (National Bureau of Economic Research), Cambridge, MA (September 5, 2017). http://www.nber.org/ papers/w21263.pdf. Blumenschein, Karen, et al., 1998. Hypothetical bias contingent valuation. South. Econ. J. 65 (1), 169–177. Briscoe, John, et al., 1990. Toward equitable and sustainable rural water supplies: a contingent valuation study in Brazil. World Bank Econ. Rev. 4, 115–134. http:// www.jstor.org/stable/3989925 (September 5, 2017). Carson, Richard T., 2000. Contingent valuation: a user's guide †. Environ. Sci. Technol. 34 (8), 1413–1418. Carson, Richard T., Flores, Nicholas E., Martin, Kerry M., Wright, Jennifer L., 1996. Contingent valuation and revealed preference methodologies: comparing the estimates for quasi-public goods. Land Econ. 72 (1), 80. http://www.jstor.org/stable/

References Abrahams, Nii Adote, Hubbell, Bryan J., Jordan, Jeffrey L., 2000. Joint production and averting expenditure measures of willingness to pay: do water expenditures really measure avoidance costs? Am. J. Agric. Econ. 82 (2), 427–437. http://dx.doi.org/10. 1111/0002-9092.00036. (September 5, 2017). Ajzen, Icek, Brown, Thomas C., Carvajal, Franklin, 2004. Explaining the discrepancy between intentions and actions: the case of hypothetical bias in contingent valuation. Personal. Soc. Psychol. Bull. 30 (9), 1108–1121. http://dx.doi.org/10.1177/ 0146167204264079. (September 5, 2017). Al-Ansari, Nadhir, Al-Oun, Salem, Hadad, Wafa, Knutsson, Sven, 2013. Water loss in Mafraq Governorate, Jordan. Nat. Sci. 5 (3), 333–340. http://dx.doi.org/10.4236/ns. 2013.53046. (September 5, 2017). Al-Ansari, Nadhir, Alibrahiem, N., Alsaman, M., Knutsson, Sven, 2014. Water supply

263

Ecological Economics 146 (2018) 250–264

J. Orgill-Meyer et al.

www.cambridge.org/core/product/identifier/S1068280500005761/type/journal_ article (September 7, 2017). Pattanayak, Subhrendu K., Yang, Jui-Chen, Whittington, Dale, Bal Kumar, K.C., 2005. Coping with unreliable public water supplies: averting expenditures by households in Kathmandu, Nepal. Water Resour. Res. 41 (2). http://dx.doi.org/10.1029/ 2003WR002443. (September 7, 2017). Postel, Sandra, 1997. In: Starke, Linda (Ed.), Last Oasis: Facing Water Scarcity - Sandra Postel - Google Books. W.W. Norton & Company, New York, London. https://books. google.com/books?hl=en&lr=&id=GFMfRVZH9G8C&oi=fnd&pg=PR7&dq= postel+1997&ots=JvWxtX0rR2&sig=M9X0v3r9QP0nyMZjnCRhUjxXCss#v= onepage&q=postel1997&f=false (September 7, 2017). Rijsberman, Frank R., 2006. Water Scarcity: Fact or Fiction? Agric. Water Manag. 80 (1–3), 5–22. http://linkinghub.elsevier.com/retrieve/pii/S0378377405002854 (September 7, 2017). Rosado, Marcia A., Cunha-E-Sá, Maria A., Ducla-Soares, Maria M., Nunes, Luis C., 2006. Combining averting behavior and contingent valuation data: an application to drinking water treatment in Brazil. Environ. Dev. Econ. 11 (6), 729. http://www. journals.cambridge.org/abstract_S1355770X0600324X (September 7, 2017). Royal Commission for Water, 2009. Water for Life: Jordan's Water Strategy 2008–2022. (Amman, Jordan). Sommaripa, Leo, 2011. Water Public Expenditure Perspective. Royal Commission for Water (Government of Jordan), Amman. Vörösmarty, Charles J., Green, Pamela, Salisbury, Joseph, Lammers, Richard B., 2000. Global water resources: vulnerability from climate change and population growth. Science 289 (5477). http://science.sciencemag.org/content/289/5477/284 (September 7, 2017). Waddington, Hugh, Snilstveit, Birte, 2009. Effectiveness and sustainability of water, sanitation, and hygiene interventions in combating diarrhoea. J. Dev. Eff. 1 (3), 295–335. http://dx.doi.org/10.1080/19439340903141175. (September 7, 2017). Whitehead, John C., 2005. Environmental risk and averting behavior: predictive validity of jointly estimated revealed and stated behavior data. Environ. Resour. Econ. 32 (3), 301–316. http://dx.doi.org/10.1007/s10640-005-4679-5. (September 7, 2017). Whitehead, John C., Phaneuf, Daniel J., Dumas, Christopher F., Herstine, Jim, Hill, Jeffery, Buerger, Bob, 2010. Convergent validity of revealed and stated recreation behavior with quality change: a comparison of multiple and single site demands. Environ. Resour. Econ. 45 (1), 91–112. http://dx.doi.org/10.1007/s10640-0099307-3. (September 7, 2017). Whittington, Dale, 2010. What have we learned from 20 years of stated preference research in less-developed countries? Ann. Rev. Resour. Econ. 2 (1), 209–236. http:// dx.doi.org/10.1146/annurev.resource.012809.103908. (September 7, 2017). Whittington, Dale, 2016. Ancient Instincts: Implications for Water Policy in the 21st Century. http://dx.doi.org/10.1142/S2382624X16710028. (September 7, 2017). Whittington, Dale, Mu, Xingming, Roche, Robert, 1990. Calculating the value of time spent collecting water: some estimates for Ukunda, Kenya. World Dev. 18 (2), 269–280. Whittington, Dale, Lauria, Donald T., Mu, Xinming, 1991. A study of water vending and willingness to pay for water in Onitsha, Nigeria. World Dev. 19 (2–3), 179–198. http://linkinghub.elsevier.com/retrieve/pii/0305750X9190254F (September 7, 2017). Whittington, Dale, Pattanayak, Subhrendu K., Yang, Jui-Chen, Bal Kumar, K.C., 2002. Household demand for improved piped water services: evidence from Kathmandu, Nepal. Water Policy 4 (6), 531–556. Whittington, Dale, Michael Hanemann, W., Sadoff, Claudia, Jeuland, Marc, 2007. The challenge of improving water and sanitation services in less developed countries. In: Foundations and Trends® in Microeconomics. 4(6–7). pp. 469–609. http://www. nowpublishers.com/article/Details/MIC-030 (September 7, 2017). Whittington, Dale, Jeuland, Marc, Barker, Kate, Yuen, Yvonne, 2012. Setting priorities, targeting subsidies among water, sanitation, and preventive health interventions in developing countries. World Dev. 40 (8), 1546–1568. http://linkinghub.elsevier. com/retrieve/pii/S0305750X12000411 (September 7, 2017). Wu, Pei-Ing, Huang, Chu-Li, 2001. Actual averting expenditure versus stated willingness to pay. Appl. Econ. 33 (2), 277–283. http://dx.doi.org/10.1080/00036840122947. (September 7, 2017).

3147159?origin=crossref (September 5, 2017). Casey, James F., Kahn, James R., Rivas, Alexandre, 2006. Willingness to pay for improved water service in Manaus, Amazonas, Brazil. Ecol. Econ. 58 (2), 365–372. http:// linkinghub.elsevier.com/retrieve/pii/S0921800905003228 (September 5, 2017). Cutler, David M., Miller, Grant, 2005. The role of public health improvements in health advances: the twentieth-century United States. Demography 42 (1), 1–22. Devoto, Florencia, et al., 2011. Happiness on Tap: Piped Water Adoption in Urban Morocco. Dutta, Venkatesh, Chander, Subhash, Srivastava, Leena, 2005. Public support for water supply improvements: empirical evidence from unplanned settlements of Delhi, India. J. Environ. Dev. 14 (4), 439–462. http://dx.doi.org/10.1177/ 1070496505281841. (September 5, 2017). FAO, 2007. Aquastat. Fewtrell, Lorna, et al., 2005. Water, sanitation, and hygiene interventions to reduce diarrhoea in less developed countries: a systematic review and meta-analysis. Lancet Infect. Dis. 5 (1), 42–52. http://linkinghub.elsevier.com/retrieve/pii/ S1473309904012538 (September 5, 2017). Freeman III, A. Myrick, 1979. Approaches to measuring public goods demands. Am. J. Agric. Econ. 61 (5), 915. http://dx.doi.org/10.2307/3180346. (September 5, 2017). Freeman, A. Myrick, 2003. The Measurement of Environmental and Resource Values, 2nd ed. Resources for the Future Press, Washington DC. Galiani, Sebastian, Gertler, Paul, Schargrodsky, Ernesto, 2005. Water for life: the impact of the privatization of water services on child mortality. J. Polit. Econ. 113 (1), 83–120. http://dx.doi.org/10.1086/426041. (September 5, 2017). Galiani, Sebastian, Gonzalez-Rozada, Martin, Schargrodskyj, Ernesto, 2009. Water expansions in shanty towns: health and savings. Economica 76 (304), 607–622. Gamper-Rabindran, Shanti, Khan, Shakeeb, Timmins, Christopher, 2010. The impact of piped water provision on infant mortality in Brazil: a quantile panel data approach. J. Dev. Econ. 92 (2), 188–200. http://linkinghub.elsevier.com/retrieve/pii/ S0304387809000212 (September 5, 2017). Goldblatt, Michael, 1999. Assessing the effective demand for improved water supplies in informal settlements: a willingness to pay survey in Vlakfontein and Finetown, Johannesburg. Geoforum 30 (1), 27–41. http://linkinghub.elsevier.com/retrieve/ pii/S0016718598000347 (September 5, 2017). Haddadin, Munther J., 2006. Water Resources in Jordan. Resources for the Future Press, Washington DC. Haener, M.K., Boxall, P.C., Adamowicz, W.L., 2001. Modeling recreation site choice: do hypothetical choices reflect actual behavior? Am. J. Agric. Econ. 83 (3), 629–642. http://dx.doi.org/10.1111/0002-9092.00183. (September 5, 2017). Hanemann, W. Michael, 2005. The Economic Conception of Water. Department of Agricultural and Resource Economics, UC Berkeley. Van Houtven, George L., Pattanayak, Subhrendu K., Usmani, Faraz, Yang, Jui-Chen, 2017. What are households willing to pay for improved water access? Results from a metaanalysis. Ecol. Econ. 136, 126–135. http://dx.doi.org/10.1016/j.ecolecon.2017.01. 023. Jaber, Jamal O., Mohsen, Mousa S., 2001. Evaluation of non-conventional water resources supply in Jordan. Desalination 136 (1–3), 83–92. http://linkinghub.elsevier. com/retrieve/pii/S0011916401001680 (September 7, 2017). Jeuland, Marc, et al., 2015. Do decentralized community treatment plants provide better water? Evidence from Andhra Pradesh. http://www.ssrn.com/abstract=2589196 (September 5, 2017). Jordanian Ministry of Water and Irrigation, 2013. Jordan Water Sector Facts and Figures 2013. http://www.mwi.gov.jo/sites/en-us/Documents/W in Fig.E FINAL E.pdf (September 7, 2017) (Amman). Katuwal, Hari, Bohara, Alok K., 2011. Coping with poor water supplies: empirical evidence from Kathmandu, Nepal. J. Water Health 9 (1). http://jwh.iwaponline.com/ content/9/1/143 (September 5, 2017). Kriström, Bengt, 1997. Spike models in contingent valuation. Am. J. Agric. Econ. 79 (3), 1013–1023. Mäler, Karl-Göran, 1974. Environmental Economics; A Theoretical Inquiry. Published for Resources for the Future by the Johns Hopkins University Press, Baltimore. Murphy, James J., Stevens, Thomas H., 2004. Contingent valuation, hypothetical bias, and experimental economics. J. Agric. Resour. Econ. Rev. 33 (2), 182–192. https://

264