Sarah Hinkfuss, ‘10

Harvard University, Department of Economics

This paper investigates the extent to which the informal water market equitably distributes water to urban households. Stated differently, given a government monopoly that falls short of its purpose, can the “invisible hand” equitably distribute a precious resource? Development economists commonly find evidence of ethnic discrimination—here defined as horizontal inequity—and regressive pricing—here defined as vertical inequity—in free market settings, while contradictory neoclassical analysis suggests that mutual self-interest will result in a naturally equitable marketplace. The paper probes this under-researched question by constructing a household water demand model with household-level data gathered during a field survey in July and August 2009 of 265 households in the Emir Ali neighborhood of Ayn al-Basha, Jordan. The paper uses the household water demand model to analyze the equity characteristics of the local water market using OLS and IV methods. The results suggest that the informal market is non-discriminatory but exhibits regressive pricing, which implies that the informal market falls short of a truly equitable distribution.

Introduction

“You have to assure equality for the neighborhood to accept stress. Then the challenge of water storage, call it, creates a unity of purpose.”

-Former Jordanian Minister of Water and Irrigation, 10 August 2009

In an interview, Munther Haddadin, the former Jordanian Minister of Water and Irrigation, summarized the implications of his country’s water scarcity on the optimal distribution of the resource at the household level: “Local water stress can only be met with a unity of purpose, which requires equity” (Interview K, 2009). Local distribution of water in Jordan occurs through not only the formal sector—the public water utility—but also through informal entrepreneurs who sell water from tanker trucks to households. Equity in the formal market is ensured by 91% access rates and a progressive tariff structure (World Development Indicators, 2005). In the tanker market, however, the level of equity is unknown.

There are two types of equity: horizontal equity and vertical equity. Horizontal equity concerns the similar treatment of all similar households and the dissimilar treatment of all dissimilar households. The designation of a similar versus a dissimilar household should only be made on the basis of willingness to pay (WTP) differences among households and not on issues of race, ethnicity, or religion. I call horizontal inequity discrimination. Vertical equity occurs when pricing is progressive, like progressive taxation (Mohamed, Kremer, and Kumar, 2009).

This paper investigates the extent to which the informal water market improves the horizontal and vertical equity of water access and distribution. Stated differently, given a government monopoly that falls short of its purpose, can the “invisible hand” equitably distribute a precious resource? Using the Emir Ali neighborhood of Ayn al-Basha, Jordan as a case study, I find mixed results: the informal market is non-discriminatory, but exhibits regressive pricing.

In brief, the findings presented in this paper are based on empirical, qualitative, and quantitative case study research. The basis of the study is a randomized survey of households I conducted in Ayn al-Basha during August 2009. The survey concerned household demographics, water use, and water satisfaction. I also conducted in-depth interviews of key informants at the local and national levels, including those in the health care sector, the local water authority, the national water office, and private water firms. Despite detailed study of the local context of water service provision, the case study of Ayn al-Basha is necessarily a snapshot in time. My work remains, however, the first step into an under-researched field.

This paper presents the first econometric foray into understanding how the informal market distributes water to urban households. As climate change accelerates the aridity of already water scarce countries—namely those in the Middle East—establishing mechanisms that equitably distribute this precious resource will become matters of domestic political stability and global security. Higher water scarcity will likely mobilize the informal market to play a greater role in resource distribution. Development economists and natural resource practitioners must understand the informal market and find ways to improve its performance. The inter-disciplinary methods mobilized in this paper can be applied to other informal markets in the region.

The household water market and the invisible hand

The formal sector in Jordan inadequately distributes water due to the country’s severe water scarcity (FAO, 2009). In response to its extremely low levels of water, the Jordanian government rations water and residents store it in tanks. Local entrepreneurs capture residual household water demand by selling spring water directly to households from water tanker trucks. The water tanker market exhibits the three characteristics of the market laid out by Adam Smith: the interaction of supply and demand, competition, and self-interest (Smith, 1759). The tanker market is the informal market—not just a private market—because the tanker drivers’ operation depends on the Water Authority Jordan (WAJ) not regulating the tanker prices. The tanker market operates without meaningful institutional oversight or regulation (Interview N, 2009; Odeh, 2009; Sass and Brubach, 2008).

Household demand in the informal water market, let alone the level of equity in the market, is currently poorly understood. Two competing hypotheses exist on the ability of this special water market to equitably distribute the resource. On the one hand, neoclassical economic analysis suggests that the tanker market results in equity gains because it exists as a free market response to the inefficiencies of the public market. Both the suppliers and consumers voluntarily enter the market so even if there is inequity, it must improve welfare based on the assumption that households are rational economic actors. On the other hand, development economic analysis suggests that households have no choice but to enter the informal market to meet household water needs. As a result, these households exhibit extremely inelastic demand for water and must accept the price set by the suppliers. The tankers maximize profit by favorably distributing to households in their family network, which creates inequity in the forms of discriminatory distribution and a regressive pricing scheme.

Local context

Ayn al-Basha is a city in the Balqa governorate, adjacent to Amman, the capital city of Jordan. Emir Ali is a small neighborhood of about two square kilometers and 1,383 households (Dorsch Consult, 2007). This area is ideal to study the informal water market for four reasons:

  1. Emir Ali, with its exclusive Muslim majority and level of water scarcity, is representative of other areas in the region (Department of Statistics, Jordan, 2006; Sass and Brubach, 2008). Hence, this case study may be meaningfully generalized to the greater region.
  2. The local tanker water market is robust because household water demand is greater than piped water supply due to the use of sub-par materials and low-skilled public workers.
  3. Both public workers and households in the area are familiar with foreign water professionals who work in coordination with international organizations.
  4. The neighborhood exhibits high levels of ethnic diversity because it is adjacent to the Baqa’a Palestinian refugee camp. This enables the exploration of discrimination issues.

Literature review

The empirical literature confirms that household water demand obeys the law of demand. First, water demand increases as household size increases, but less than proportionally due to household economies of scale (Strand and Walker, 2005; Whittington and Bal Kumar, 2005). Second, numerous studies establish that household demand for primary sources of water is inelastic, which is consistent with demand for a necessary good (Ruijs, Zimmerman and Van den Berg, 2008; Strand and Walker, 2005).

In addition to demand elasticity, many authors consider income elasticity. Kanyoka, Farolfi, and Mordet (2008) analyze demand and income elasticity by considering how demand elasticity of a household changes as income changes. Their model uses a choice approach in surveying to determine household preferences and WTP for water services in rural South Africa. The households surveyed had inconsistent water access. The authors find that households with lower incomes have higher price elasticity, which negatively impacts their WTP, as compared to more well off households. Strand and Walker (2005) find income elasticities below 0.1.

While there is some disagreement on the particular level of income elasticity, there is near universal agreement that household demand for a primary source of water exhibits positive income elasticity. Hajisyprou, Koundouri, and Pashardes (2002) establish the theoretical basis of water as a normal good, and Le Blanc (2008) empirically demonstrates this finding. He finds that household expenditures on water increase with income, but less than linearly. In South Asia, Le Blanc finds that households spend, on average, 2% of their income on water consumption. The share of income increases to 3.5% in Latin America and Eastern Europe. No comparable review papers consider the Middle East.

The preponderance of the literature on household demand for water from the informal market is based on studies in peri-urban areas where the informal market is the sole supplier of water to these households. In such situations, the household water demand exhibits the same qualities as the demand for water from the formal sector. This is not representative of Ayn al-Basha, however, because households use both the piped water and tanker water supply.  When households are presented with water from the informal sector to supplement or substitute water from the formal sector, such as in my study area, then demand may be different. In fact, Strand and Walker (2005) note an enormous effect of the water’s source on the household’s level of demand elasticity. Additionally, numerous papers by Dale Whittington find that the mix of water supply sources available to households changes the optimal household consumption of the various sources (Whittington and Bal Kumar, 2005).

Only one paper constructs household demand in the informal market based on observed behavior. Many other authors, such as Whittington, use the household cost of coping mechanisms and contingent valuation surveys to estimate household water demand (Whittington, et al., 1990; Whittington and Bal Kumar, 2005; Whittington, Lauria and Mu, 1991; and Whittington, Mu, and Roche, 1990). Iskandarani’s (2001) seminal paper is notable not just because it applies econometric analysis to a new market, but also because she does so in Jordan.

Iskandarani (2001) models household demand for the tanker market based on a household survey of 200 households in Jordan conducted in the summer of 1999. She first models the discrete choice of households to consume tanker water based on the price of the tanker water, the amount of piped water that the household receives, their monthly income, their water storage capacity, and their level of satisfaction with piped water supply. Iskandarani does not explain, however, why households have a choice at all over what type of water to consume and if the quantity of piped water is set exogenously. Iskandarani finds that household demand for tanker water is price inelastic and that when the full price of water is considered, households spend more than 3% of income on water than is currently assumed.

Overview of the paper

The paper proceeds in four sections. Section II presents the theoretical model of household water demand. The assumptions of the model are based on the dominant assumptions in the literature and my assessments from fieldwork. The model predicts both horizontal and vertical inequity. Section III describes the data mobilized to evaluate the theoretical model. Section IV uses the data to empirically investigate the informal water supply in Emir Ali neighborhood. I construct household demand for tanker water using instrumental variables, as suggested in the water demand literature. I find no evidence of horizontal inequity and some evidence of vertical inequity. Finally, Section V concludes.

As demonstrated in the literature review, very few microeconomics papers convincingly construct household demand for water from observed household behavior, let alone demand for a good in the informal market. This paper advances the natural resource and development economics literature by addressing this challenge.

Theoretical model

“If you don’t have (money), then consume less…and some people, they wash the trees.”

Answer of a water official in response to a question about varying water use among local households, 11 August 2009

This chapter presents an analytical framework for determining the equity of the private water market operating in tandem with the public market. The framework models and evaluates horizontal and vertical equity in the tanker market. To do this, I first construct a theoretical model of household water demand. From there, I examine both whether the household characteristics that influence water demand are congruent with horizontal equity and whether the dominant pricing structure exhibited by the demand curve is consistent with vertical equity.

I build household demand for tanker water based on fundamental water needs. Households require a certain quantity of tanker water for basic survival and will pay all of their income, short of income for other necessary goods like food and shelter, to receive this quantity. Household consumption at all quantities less than or equal to the survival quantity will take place irrespective of the price of the tanker water. Once the basic quantity is secured, however, household willingness to pay (WTP) for water is a function of its price, and households will consume less water as the price increases. At low quantities of household water consumption, a small decrease in price induces a large increase in water consumed, while at high quantities of household water consumption the household already has sufficient water and needs a large decrease in the price to warrant purchasing any additional quantities. The household demand curve for tanker water that exhibits these characteristics is shown in Figure 2a.

The supply relationship is similarly constructed. Tanker drivers exhibit a degree of power in setting prices, which allows them to better recover costs and increase profits. In essence, this means the more a given household consumes, the higher price the tanker driver will charge the household for the water. The tanker supply curve exhibits monopolistic competition.

The interaction of the demand and supply curves creates the tanker water market, as in Figure 2b. The optimal allocation of water, MM, occurs at the intersection of the curves.

Figure 2a
Figure 2a
Figure 2b
Figure 2b

One would expect all households of similar income levels to share the same equilibrium point. However, my interviewees suggested that at each income level, Jordanians, or those with influential family networks, are preferentially served by the market relative to Palestinians, or those without influential family networks. Jordanians are native to the country, and Palestinians are considered immigrants, even if a Palestinian was born in Jordan. Jordanian families dominate government positions and are considered more powerful through their ties to the King. Although both ethnic groups may access the market, ethnicity results in two market demand curves and water allocations rather than one, as shown in Figure 2c. If true, this violates horizontal equity. This conclusion is expected because tanker water is distributed within family networks, so tanker drivers set lower prices to their family than to other consumers. Given that families are all of one ethnicity, the model predicts the tanker market is horizontally inequitable because the price the tanker driver sets for water is determined by ethnicity.

Additionally, the shape of the demand curve presented predicts that the tanker market will be vertically inequitable. As in Figure 2d, the quantity of water consumed by poor households is closer to their minimum survival level, where demand is near completely inelastic. Therefore, price-setting tanker drivers may improve their profit margin by exploiting this high inelasticity and setting higher prices for poor consumers than for wealthy ones.

Figure 2c
Figure 2c
Figure 2d
Figure 2d

The appendix derives a model of household water demand for to support these predictions.

Theoretical applications of the model on equity

The theoretical model of the tanker market forms the basis for predicting the extent to which the tanker market improves equity in the water market. An analysis of horizontal and vertical equity in the water market surveys the household distribution of access to fairly priced, high-quality, reliable, satisfactory water. The model captures household access by the quantity received and the price paid at the household’s optimal demand. The demand defines the household’s willingness to pay for the water, or the value of the tanker water to the water.

Horizontal equity

Evidenced by the role of C in the theoretical model, the tanker market is horizontally inequitable because household ethnicity mediates household access to tanker water. The discrimination effect is demonstrated in the optimal quantity demanded, the elasticity of demand, and the income elasticity. Holding all else constant, Equation (15) in the Appendix predicts that Palestinian households will consume an amount less of tanker water than Jordanian households.

(16)

Likewise, holding all else constant, Equation (16) predicts that Palestinian households will have less elastic demand than Jordanian households across all quantities.

(17)

Equations (16) and (17) demonstrate that Palestinian households derive less utility from a given quantity of tanker water consumed than their Jordanian counterparts. Finally, holding all else constant, Equation (15) in the Appendix predicts that the income elasticity of Palestinian households will be half that of the Jordanian households. Lower income elasticity means that Palestinian households across different income groups will have less variation in tanker water consumption patterns than Jordanian households and be charged more at every quantity consumed than their identical Jordanian counterparts. None of the other variables in the theoretical model pose a threat to the horizontal equity criteria.

Figure 2c illustrates the horizontal inequity of the model. The figure modifies the depiction of the household demand curve to represent only the price, φ, along the y-axis, rather than φC as in the other figures. C impacts the shape of the household demand curve. The graph shows that for equivalent levels of consumption, Jordanian households always pay less per cubic meter than Palestinian households. The difference in demand between the ethnicities peaks at a consumption quantity past the minimum level of water required by the household. Additionally, as the Jordanian demand curve is always greater than the Palestinian demand curve, the Palestinian households have less elastic demand, as the model predicts.

Vertical equity

The theoretical model predicts that the tanker market is vertically inequitable. As explored in the horizontal equity section, the theoretical market model predicts that Palestinian households will pay more per cubic meter because the tanker drivers will respond to the differential WTP by charging the Palestinian households more to exploit the associated efficiency gains. The additional price that the Palestinian households must pay to consume the same amount of water as their otherwise equal Jordanian counterparts is double the price that the Jordanian households pay.

(18)

The pricing structure is most regressive when the level of piped water consumed by the household does not meet the household’s Mmin for two reasons. First, poor households are more inelastic in the tanker market than in the piped water market, given the elasticity of piped water supply.  Second, usually a small increase in price means that the tanker driver will lose a considerable amount of quantity supplied, however households have extremely high tolerance for price increases in their survival surplus portion of the demand curve.  Therefore, tanker drivers may face an incentive toincrease prices for households that consume less water, which in most cases are the poorest households.

Limitations of model and theoretical analysis

The theoretical model is the best approximation for actual behavior in the tanker market in Ayn al-Basha because the assumptions are based on observation and interviews and not solely on existing models. However, no model can capture the diversity of choices facing households and their resulting behaviors. For example, the model does not account for family time investments in consuming tanker water, such as boiling and freezing water, and the quality and reliability of tanker water may impact the vertical equity of the market based on which households disproportionately bear the costs of low quality and unreliable water. Additionally, wasta is not as simplistic as a binary variable that represents household ethnicity. There are enormous ranges of influence demonstrated by households. Future iterations of the model should relax the assumption of wasta as a binary variable.

Data

“Studies are needed to focus on understanding the nature of household demand for water and should attempt to express it in terms of household demand functions.” –An appeal for new research by Munther Haddadin, the Former Minister of Water and Irrigation in his 2006 book, Water Resources in Jordan

The quality of the data available to a study determines the basic value of an empirical investigation. My data comes from three sources. First, the Water Authority Jordan (WAJ) central statistics office in Amman provided the household-level piped water data. Second, I used Geographic Information Systems (GIS) to interpret all of the household-level geographic relationships from the maps of Emir Ali neighborhood in Ayn al-Basha. Engicon produced the maps in coordination with Dorsch Consult. Third, all other household-level data comes from a randomized household-level survey that I conducted in Ayn al-Basha during 2009. Each set of data offers its own advantages and limitations that in concert uniquely qualify this paper’s novel investigation of equity in the informal market.

Methodology

To collect the household level information, I conducted a randomized household survey of 295 households (ME 1.8%) over three weeks during July and August 2009, in the Prince Ali neighborhood of Ayn al-Basha. Pre-field work, I randomly chose the households to be surveyed using a uniquely identifying houseplot number. The surveys followed a script and were conducted in Arabic with a translator unless the respondent elected to speak in English.

The final sample is only 265 households. Ten households declined to participate in the survey, I was unable to interview seven households, and an additional thirteen surveys had to be excluded due to failure to follow the survey script. The respondents at households who declined to participate did so only when I was with a male translator. I speculate that this was due to the cultural apprehension about speaking with non-familial males. These families were scattered throughout the study area, so I expect their exclusion to create no bias in the survey. At the seven households where I was unable to find someone to interview, the house itself was either under construction or there were only family members under the age of 18 at home. The exclusion of the households under construction has little bearing on the distribution of households included in the survey because these houseplots do not yet represent families demanding tanker water until they live in their homes.

On the other hand, the second class of families who I was unable to interview due to no adult being present are likely the very poor families in which both parents (and any grandparents) select to work outside of the home and do not come home for lunch time (as is customary). As I only surveyed during normal working hours (8AM to 4PM), these very poor households are thus excluded. Consequently, my equity analysis will likely overestimate the true equity effects because the poorest of the poor have been excluded. Finally, the households who were excluded due to not following the household script were randomly distributed throughout the area and more often than not occurred on the first day of a translator’s time with me.

Variables for the empirical investigation

The primary variables used in the analysis are the dependent and independent variables. The key dependent variable is the tri-monthly consumption of tanker water. I consider all households in the survey area—those that consume tanker water and those that do not consume tanker water. The tri-monthly consumption is the total volume of a household’s ground level and rooftop tankers, multiplied by the probability that the tanker water comes daily, multiplied by 90 for the number of days in three months. If a household does not consume tanker water then the family has a zero probability of consuming tanker water on a given day, and the household’s tri-monthly consumption of tanker water is 0. I use a three-month time period because this is the period over which piped water billing occurs.

The key independent variable is the per cubic meter cost of tanker water. As in many other utility markets, no distinction can be made between the average and marginal cost of the tanker water due to the structure of this question in the survey. The question asked respondents: “How much do you pay per cubic meter for water from tankers?” The price included in the regressions, thereby, is the average price and the marginal price is indeterminate.

The empirical investigation of equity relies on the accuracy and precision of two household-level characteristics: income and ethnicity. Regarding household ethnicity, this was gathered carefully and incidentally. As a foreigner, it was inappropriate and politically incorrect in all circumstances to ask any individual his or her ethnicity. Many locals regard this to be an improper question to ask each other as well, although at times my translators encountered resistance when the interviewee refused to continue unless the translator said to be of the interviewee’s ethnicity. Household ethnicity was recorded only when one of the following three situations occurred: the interviewee volunteered his or her ethnicity during the course of the survey; the interviewee gave his or her surname and the surname is known to be of one ethnicity or another; or the interviewee spoke in the dialect consistent with one ethnicity or another.

Empirical investigation

“Water didn’t come this week and pressure was bad last week, so we haven’t bathed….”

Household 252 describing her satisfaction with the quantity of water supplied, 19 July 2009

This chapter empirically tests the predictions of horizontal and vertical inequity in the tanker market from the model in Section II with the data described in Section III. The objective of this section is three-fold: first, to confirm empirically the assumptions of the theoretical model; second, to construct household demand for tanker water; and third, to use demand to test the predictions of horizontal and vertical inequity from the theoretical analysis.

I begin by confirming two assumptions underlying the theoretical model: one, the tanker water is a uniform good, or a good that is identical across all units sold; and two, all households have equal access to the tanker market. Next, I construct household demand for tanker water. Due to simultaneous determination of price in demand and supply models, I reduce the threat of reverse causality in the demand function by constructing the tanker water price that the household faces using household characteristics, or instrumental variables (“IV”), that determine the price and do not determine the quantity demanded by the household. I use the constructed low price variable in the demand equation in place of the low price variable, which robustly isolates household demand—the effect of price on the quantity demanded. This method considers the actual price paid by the household.

I use the predicted demand to determine the level of equity in the tanker market. First, I reject the predictions of the theoretical model by finding evidence of horizontal equity. I find the ethnicity variables insignificant in the demand equation, which excludes discrimination as a determinant of tanker water consumption. This result is robust to the demand specification. On the other hand, I confirm the theoretical model’s prediction of vertical inequity. I find that the poor pay more for tanker water given their household income than the rich, which is evidence of a regressive pricing structure. This result is robust with regard to the mechanisms households use to cope with imperfect water supply, or the correlates of consumption.

Methodology

Constructing demand

Demand for a normal good, such as water, is a function of three considerations: the price; household income—their budget constraint; and household preferences. I mobilize the empirical implications of the model in Chapter 2 to specify the demand relationship among these factors.

The naïve model is the simplest version of household demand for tanker water. This model simply regresses the quantity of water consumed on the price of water, controlling for the household’s budget constraint and household preferences:

(1)

where Income is a vector of a household’s relevant budget constraint variables, Preferences is a vector of a household’s relevant preferences variables, and e is the common measurement error. Elasticity of demand is defined as

(2)

The results of the naïve model are threatened by simultaneity bias, which results from reverse causality between variables. Assuming that water is a normal good, reverse causality could arise if the price impacts the quantity consumed and the quantity consumed impacts the price. This creates bias in the estimation of demand elasticity that may lead to spurious results.

I disentangle the effect of price in the demand equation from the determination of price in the supply equation by creating the price variable anew. I extract from price what is meaningful to households but not meaningful in determining the price in the supply equation. The OLS method accomplishes this by separating the one supply curve into two so that demand can be determined as households differentially react to the two supply curves. I separate the supply of the high priced tanker water, Price_High, from the supply of the low priced tanker water, Price_Low. Both variables are binary, where Price_Low takes on the value of 1 if the tanker water cost is equal to or less than the median price. This method addresses concerns of reverse causality because although the quantity may determine the price, an absolute measure, it does not determine whether a household faces a high price or a low price, a relative measure.

The OLS model improves Equation (1) by replacing the general price variable with Price_Low. I also add to the model the interaction terms suggested by the theoretical model, specifically the interaction of price and income and the interaction among preference variables.

(3)

where all variables are defined as above. The elasticity of demand is defined as

(4)

Although this model reduces the threat of simultaneity, it does not eliminate the concern because the Price_Low and Price_High dummies, though relative values, are still endogenously determined in the market. The Dichotomous Price model improves the OLS model by exogenously identifying the price variable using instruments to eliminate the reverse causality concerns. I use four instrumental variables:

  • House Years—the number of years the household has lived in the house. The longer the family has lived in the same house, the more likely they are to be established in the community and have stronger social connections with local business owners, including tanker drivers. These households face the low tanker price;
  • Prob Piped—the probability that the piped water comes to the household over the course of a representative day. The lower the probability that water will come, the greater the desperation of the household, and the higher the price that the household will accept. Note that the probability of piped water supply is not directly correlated with the quantity of piped water that comes during a particular supply period;
  • Reader—whether or not the household knows their water meter reader. Knowing the water meter reader indicates how well connected to the local community is a household. It also demonstrates the respondent’s attention to the individuals that handle the household’s water; and
  • Unskilled—whether or not the household’s working members hold unskilled jobs. Unskilled workers are less likely to be in positions of power within the community and thus have less pull with the tanker drivers for favors. The families of unskilled workers, therefore, likely pay higher prices for tanker water.

Each of these instruments consistently and efficiently estimates the endogenous parameter, Price_Low. I confirm by testing the relevance and exogeneity of each instrumental variable. The instruments are relevant because each is strongly correlated with the tanker price, or the covariance of the instrumental variable and the endogenous independent variable, Price_Low, is not zero. The practical test for relevance is the first stage F-statistic, which must be above 10. An instrument is valid if it is not correlated with the quantity of tanker water that the household consumes or if the covariance of the instrumental variable and e is zero. The test for exogeneity is the Hansen J-statistic of the overidentifying restriction test, which is the F-statistic computed using the homoskedasticity-only c2s distribution with m-k degrees of freedom.

The Dichotomous Price model is as follows:

First stage:

(5)

 

Second stage:

(6)

where all variables are defined as above. As the Price_Low variable is binary, I estimate this model using the probit linear regression procedure. The elasticity of demand in the Dichotomous Price is defined as in the OLS model shown in Equation (4).

Finally, the IV with Continuous Price model further improves on the Dichotomous Price model by using as the independent variable the particular price that households face. I use the same exogenous and relevant instrumental variables as in the Dichotomous Price model, and the first and second stage regressions are also the same except that the dependent variable in the first stage regression is now Price. The IV with Continuous Price model is as follows:

First stage:

(7)

Second stage:

(8)

where all variables are defined as above. The elasticity of demand is defined as in Equation (4).

The estimation of tanker water demand may be biased if there are any omitted variables. Possible omitted variables include the altitude of the household’s houseplot and the quantity of water received from other water sources. First, a local Jordanian mid-level worker in the water sector suggested that altitude directs tanker water consumption in Ayn al-Basha because the households at highest altitude receive insufficient quantities of water from the piped supply. These households must rely on tanker water to meet their water needs (Interview A, 2009). I tested this effect by including the altitude in the demand regressions. I determined houseplot altitude by interpolating from the altitude isoclines at 50-meter intervals in the GIS data of Emir Ali from Engicon (Engicon, 2009). I found altitude to be insignificant and non-informative in the demand regressions. Therefore, the omission of altitude does not bias the results.

Second, the quantity of water received from other water sources may affect the quantity of tanker water demanded if a household receives large amounts of water outside of the tanker and piped water sources. In this case, the quantity of water received by the household from the third source of water should be controlled as the quantity of piped water is controlled. My survey includes household level data on bottled and cooler water consumption. I included this water in the demand regressions and found bottled and cooler water consumption to be insignificant and non-informative. However, many households also consume spring water. The survey did not ask households how much spring water they consume. If the omission of household spring water creates omitted variable bias, then this would overestimate household demand for tanker water because consuming spring water leads a household to consume less tanker water and be less willing to pay more for tanker water. The threat of this bias is minimal because only a small fraction of households in the area also consumed spring water.

Testing equity

I use the IV with Continuous Price model, the most robust model, to test the levels of horizontal and vertical equity in the informal market. First, I test horizontal equity by including ethnicity in the demand function, as given by the theoretical model. The specification becomes

First stage:

(9)

Second stage:

(10)

The null hypothesis is β16 = β17 = 0 and the alternate hypothesis is β16 = β17 < 0.

There is ethnic discrimination in the tanker market if I reject the null hypothesis, and there is discrimination against households of Palestinian ethnicity if I reject the null in favor of the alternate hypothesis.

Additionally, I check to see if the result is robust to alternate specifications of demand by determining the relationship between ethnicity and the quantity of tanker water consumed at its most basic linear relationship with all possible interaction terms, as given below

(11)

The null hypothesis is that β6 = β7 = β8 = β9 = β10 = β11 = 0, and the alternate hypothesis is that β6 = β7 = β8 = β9 = β10 = β11 < 0. There is ethnic discrimination if I reject the null hypothesis, and there is discrimination against Palestinian households if I reject the null in favor of the alternate.

I test vertical equity by analyzing the distribution of costs for tanker water across the households on the basis of income. First, I confirm that tanker water is a normal good by analyzing the income elasticity in the IV with Continuous Price model. Income elasticity is

(12)

I expect the elasticity of income to be positive.

Second, I analyze the variable role of tanker water consumption in a household’s budget constraint. I determine who pays the highest prices for water, absolutely and relatively. Additionally, I discuss the role of coping mechanisms in clarifying the level of vertical equity.

Demand results

The informal market plays a significant role in providing households access to water. The average quarterly total water supply is 55.9 cubic meters. Forty-six percent of households use tanker water. On average, tanker water constitutes 8.4% of household water supply.

Uniformity of tanker water

I test the assumption from the theoretical model that tanker water is a uniform good. If tanker water is not uniform, then the quantity demanded should be qualified by the quality of the water, as suggested by Olmstead (2010) and the World Bank (1995).

Based on my field research, I predict that there is only one variety of tanker water. Though the water comes from different springs, the tanker drivers consider the quality of the water at these different springs equivalent because they are all equally monitored by the government and the tanker drivers switch among the various springs when filling up the tanks depending on the daily quota of water to be sold at the springs.

The alternate hypothesis is that there are multiple varieties of tanker water. If there were more than one variety, I would expect the better quality water to be more expensive. To test the alternate hypothesis, I run an OLS regression with heteroskedastically-robust standard errors of the per unit price of tanker water on the household’s satisfaction with tanker water, controlling for the household’s satisfaction with piped water. Table 3A in the Appendix presents the results.  All tables may be found in the Appendix online. Regression 1 includes the most relevant satisfaction variables to the perceived “quality” of the water. Regression 2 includes additional satisfaction variables. Finally, Regression 3 includes only the aggregate satisfaction variables, the Satisfaction Index variables, for tanker and piped water.

As the table shows, the only significant predictor of the per cubic meter cost of tanker water is satisfaction with the quantity of tanker water received. The significant variable describes the quantity dimension of the uniform good and not the quality dimension, so it does not indicate that there is more than one quality of tanker water. The significance of the Satisfaction Index in Regression 3 suggests that the price paid influences the perceived quality of the water, but Regression 4 confirms that this significance comes wholly from the satisfaction with the quantity of the tanker water supplied as the price per cubic meter increases. I conclude that the tanker water in the study area is a uniform good.

Predicting tanker use

In order to construct household demand for tanker water, all observed households must be on the same demand curve. The theoretical model demonstrates that this occurs as long as two conditions are met. First, no household that uses tanker water to meet its minimum survival threshold should be considered in the data set. Since such a household would have an exorbitantly high willingness to pay for tanker water, this behavior would not accurately represent the normal tanker market. Therefore, all households must meet their minimum water needs through the piped water network. This insures that the portion of the demand curve of all households that does not increase utility—since it is below the survival level—is piped water. Second, WTP characteristics should be the only demographic household characteristics that explain whether or not a household uses tanker water. If a particular household characteristic excludes a household from the tanker market, then household demand will be biased. As long as these two conditions are met the consumption and price data from all households will construct one demand curve, after controlling for WTP characteristics.

The first condition is met if households receive more piped water than the minimum quantity required for survival. Chapter 2 adopts the assumption from Gleick (1996) that humans require at least 20L of water per day for basic survival, or 1.8 cubic meters of water per person over a three-month period. Empirically, I find that the average level of piped water consumption per person among households is 9.175 cubic meters for three months, with a range extending from a minimum of 1.88 cubic meters to a maximum of 67.38 cubic meters. This demonstrates that all households included in the survey meet their minimum water needs through piped water and confirms the applicability of the theoretical model in Chapter 2 to the situation of household water demand in Emir Ali.

I test the second condition by determining which characteristics differentiate families that consume tanker water from those that do not. To determine which household characteristics, if any, are significantly related to the use of water tankers, I regress the Tanker Use dummy on all household characteristics. The Tanker Use dummy is 1 if the household uses tanker water. Table 4B presents the household characteristics that are expected to influence water demand. None of the characteristics are significantly related to Tanker Use, indicating that there is no demand-based reason why some households use tanker water and others do not and it must instead be related to the particular water supply conditions of a household.

Table 4C considers whether any other household demographic characteristics are correlated with the Tanker Use dummy, including the household’s ethnicity. The only significant coefficient is whether the respondent is male or female. Regression 1 suggests that when the sex of the survey respondent is male, then the household is 15.0% less likely to use tanker water. This is likely because males advocate better for the household to the local water ministry to ensure that the household’s water needs are met through the piped water. Ethnicity is included in each regression, though it is never significant. This invalidates the theoretical model assumption that Palestinian families are more likely to purchase water from tankers as a result of under-fulfilled water needs from the piped water network.

The only significant difference between families that use tanker water and those that do not use tanker water is whether or not the piped water network provides for the water needs of the household. There are no demographic household characteristics that explain the distribution of household tanker water use, including ethnicity. All observed households share the same demand curve because both conditions are upheld. The empirical household demand construction may proceed by the assumptions of the theoretical model.

Specifying demand

Given the wide array of available budget constraint and preference variables available in the data to measure household demand for tanker water, I use the naïve model to determine which have the most explanatory power. The best model is

(13)

OLS model results

The unexpected results on the direction of the demand elasticity in the naïve model confirm the concerns about the threat of simultaneity. As feared, the simultaneity creates bias in my estimation of demand elasticity that spuriously suggests positive demand elasticity where negative demand elasticity is expected. To address the simultaneity, I disentangle the effect of price in the demand equation from its determination in the supply equation. I use Specification (3) with the specific budget constraint and preferences variables given in Specification (9).

Table 4D presents the results of the OLS model. In the naïve model (not presented) the demand elasticity is positive, but when the model is re-specified to restrict the relationship between the price and quantity to the effect of the price on quantity, the demand elasticity is negative. The regression coefficient indicates that charging a household a low price increases water consumption by 6.5 cubic meters than if the household were charged a high price.

In Regression 1, the only significant variable is the interaction term between per capita income and the tanker water price. The term suggests that given a certain level of per capita income, households that face a low tanker price consume less than households that face a high tanker price. The magnitude of the coefficient on this interaction term is much smaller than the price variable coefficient, preserving the negative demand relationship. Moreover, the magnitude of the interaction term shows that poorer households are less responsive to changes in price than are wealthier households, as predicted by the theoretical model.

Regression 2 is the complete demand specification. The price of the tanker water, captured by the Price_Low dummy, has the greatest effect on the amount of tanker water that the household consumes. This regression demonstrates that the households in Emir Ali demand tanker water to supplement the piped water supplied but will dampen consumption if the price is not right. Figure 4A in the appendix shows the predicted values of the natural log of tanker water consumption per household against the Price_High dummy. As discussed, the demand curve shown is negative because as the price increases, the quantity demanded decreases.

Dichotomous Price model results

The OLS model likely does not correct all of the reverse causality in the specification of demand because the variables that separate supply, Price_Low and Price_High are still determined within a system with reverse causality. The Dichotomous Price model improves the OLS model because it instruments for the values of Price_High and Price_Low with the WTP factors that most likely determine whether a household faces a high or a low tanker price. I expect this model to reduce the threat of omitted variable bias because the interaction terms consider the range of omitted variables that help determine tanker price.

I confirm the relevancy of the four instrumental variables individually, and the relative exogeneity of the variables jointly. The F statistic is 11.055, greater than the desired 10, with a p-value of 0.026. The J statistic is 2.18 with a p-value of 0.536, suggesting exogeneity.

The results from the second stage regression are presented in Table 4E. Regression 1 and 2 are identical except that Regression 2 requires the price variable to be a dummy, whereas the price variable in Regression 1 is a continuous variable over the range 0 to 1, as predicted by the instrumental variables. I present the constructed demand relationships from the two regressions in Figures 4B and 4C in the Appendix. Regression 1 is superior because it captures more of the variance in the instrumental variables than the price specification in Regression 2.

All variables lose significance in both regressions except the number of individuals in the household, the interaction term between the household’s per capita income, and the interaction term between price and household income. In this model, the loss of significance of the independent variable suggests that the interaction term between the income and price is more important than the role of the price alone in determining the quantity of water consumed by households. Regression 1 suggests that as a household moves from the high price to the low price, at a certain level of per capita income, the household’s demand for tanker water increases by 0.559 cubic meters. The regressions confirm the negative demand relationship and suggest the household’s tanker water consumption is highly responsive to increases in tanker price.

The demand elasticity suggested in the Dichotomous Price model should be interpreted as the maximum household responsiveness to changes in price because it reflects the effect on consumption of households moving from the mean high price to the mean low price, which is a much greater price difference than any household actually faces in the market.

IV with Continuous Price model results

I improve the Dichotomous Price model by mobilizing the full variation in tanker price faced by households to determine household tanker water consumption. I use the same instruments as in the Dichotomous Price model for the same reasons. I confirm the relevance and relative exogeneity of the four instruments using as indicators the second-stage F-statistic and J-statistic. The F statistic is 14.195, well above the desired 10, confirming relevance. The J statistic is 2.794, which rejects the null, and suggests that the instruments are jointly valid. This model most successfully constructs household tanker water demand because it best explains the variation in the independent variable, the price. Additionally, this model largely confirms the predictions of the theoretical model.

Table 4F presents the results and Figure 4D illustrates the constructed demand relationship. There are five significant determinants of household tanker water demand. When the price increases by one JD, household consumption of tanker water decreases by about 3.5 cubic meters. This is a considerable effect because it is more tanker water than the average household consumes in one month. This result is consistent with the traditional model of household demand and with the predictions of the theoretical model in Section II.

Additionally, with each additional family member, the household demands about 1 additional cubic meter per month, or 1000 liters of water per month. As there are returns to scale in household water use in terms of cooking, cleaning, gardening and other household activities, it is consistent with the existing literature that a marginal family member should only require the household to reconfigure household water demand slightly above the level of water required for daily survival (i.e. 600 liters per month). The theoretical model predicted the positive relationship between demand and the number of family members.

All of the variables are consistent with the predictions of the theoretical model, except for income. The income results will be discussed in Section 4.3.2. in the context of vertical equity. To confirm that the elasticity differential is not caused by a difference in the quantity of piped water available to households, Regression 2 adds an interaction term between the per capita income and the quantity of piped water consumed. This interaction term is insignificant.

I cannot conclude whether or not the theoretical model’s prediction that household demand for tanker water decreases in the quantity of piped water consumed is true. In the first regression the household’s quantity of tanker water consumed is increasing in the quantity of piped water that the household consumes, indicating a positive correlation between household consumption and demand. However, this effect is likely spurious because the independent effect loses significance when I add the interaction term of per capita income and quantity of piped water consumed. This suggests that there is some positive correlation between higher income households and households who receive substantial quantities of water through the piped network. To further explore the possibility of negative feedbacks, I include in Regression 3 a new binary variable to capture consumption of piped water that is 1 when the household consumes less than the average amount of piped water in the area. Controlling for piped water consumption, all of the coefficients on the variables interacted with piped water consumption lose significance. However, the new binary variable has a negative coefficient, which is consistent with the interpretation that when households receive small quantities of piped water, tanker water consumption is lower, and vice versa.

The final determining factor warranting special attention in the household demand function is the tanker satisfaction index. In Regression 1, the level of household satisfaction with the tanker water has the most positive and significant effect on the quantity of water consumed by the household. This is consistent with the role of the reliability and quality variables in the theoretical model. As predicted by the theoretical model, the interaction term between the quantities of piped water consumed and tanker satisfaction is negative, albeit insignificant.

The IV with Continuous Price model is the optimal model of household demand for tanker water given the available data, however it is far from a complete or perfect specification for two main reasons. First, the demand regression has a very small number of observations because the data includes few households with complete characteristics. The number of observations considered is only 44. Second, the constructed model of demand fails to predict nearly half of the variation in the quantity demanded. This is shown by the R2 value of 50%, which means that the model explains only about 50% of the variation in household demand. However, it is worth noting that thisR2 is on the upper edge of those in other water demand models in the literature.

Part of the remaining 50% of unexplained variance may be due in part to errors in variable bias. There may be measurement error in some of the variables, particularly in two variables critical to the demand regression, income and the number of people in the household. I discussed and for the most part dismissed in Chapter 3 the possible internal validity threats to the measurement of household income. I concluded that the levels of household income are relatively consistent. However, the absolute level of household income—the external validity of the data—is still questionable. Some households probably underestimate their income due to fear of tax repercussions or jealousy, while others likely overestimate their income out of pride. The net result does not wash away this concern of errors in variable bias, but magnifies it.

In conclusion, I find that the elasticity of demand at the average price and average quantity consumed is -1.5863. This suggests that when the price increases by 1%, the quantity of tanker water demanded decreases by 1.59%, holding all else constant. The elasticity likely changes throughout the year based on changing local weather conditions and the complementary availability of water. This elasticity is higher than the average elasticity of demand in the formal water market as demonstrated in the literature. Consistent with the theoretical model, the elasticity is increasing in the relative price of the water to the household and decreasing in the quantity consumed. However, the elasticity is not dependent on household income or the quantity of piped water consumed, as predicted.

Equity results: Horizontal equity

Horizontal equity occurs when all similar individuals are treated the same—an absence of discrimination—and households are only differentiated by their willingness to pay (WTP) for a good. I consider the level of horizontal equity in the informal water market by analyzing to what extent ethnicity influences water consumption patterns. Table 4G presents the results of household demand that is modified to include the ethnicity terms as set forth by the theoretical model. Regression 1 includes a dummy for Palestinians households, while the subsequent regressions include both the dummy and interactions between the ethnicity term and household income, as recommended by the theoretical model. Each of the instrumental variables regressions maintained the relevance and relative exogeneity of the instruments.

Across all of the regressions presented, the household’s ethnicity has no significant effect on the quantity of tanker water consumed.  It is noteworthy that the direction of the ethnicity dummy suggests that Palestinian households consume less tanker water than their non-Palestinian neighbors, however this effect is insignificant. Even if the results were significant, the coefficient in Regression 1 shows that magnitude of the effect would be trivial. On average, Palestinian households demand about 1.5 cubic meters less of tanker water than Jordanian households, which is half of the effect of a 1JD change in price and even smaller than any of the income effects. Additionally, including the ethnicity variable in the demand specification decreases the magnitude of the price coefficient by a cubic meter and the magnitude of the coefficient on the natural log of the number of people in the household by one-half cubic meter.

I check the robustness of these results by creating a new specification of demand. Rather than assuming the demand specification motivated by the theoretical model, as given by the IV with Continuous Price model, I assume a strictly linear relationship among the classic demand determinants: price, family size, quantity of piped water supplied, income, ethnicity, and all interactions with ethnicity. I vary the identification of the price variables as well, considering the identity, square, and natural log. Changing the specification allows me to see if the horizontal equity evaluation is hindered by the former specification. Tables 4H and 4I present the results: across all specifications of tanker water demand, ethnicity has no significant effect.

Therefore, I accept my null hypothesis that there is no discrimination on the basis of ethnicity in the tanker market in Ayn al-Basha. This finding rejects three components of the theoretical model. First, the ethnicity variable is insignificant in the demand equation. Second, the interaction term is also insignificant. Third and finally, the aggregate demand is unaffected by the relative distribution of household ethnicity in Ayn al-Basha.

Equity results: Vertical equity

Vertical equity occurs when individuals face a neutral or progressive pricing scheme, similar to the structure of progressive taxes with a net transfer from rich to poor. Under progressive pricing, the poor pay a lower proportion of income for water than the wealthy.

First, I analyze the elasticity of income to determine whether or not tanker water is a normal good, as predicted by the theoretical model. Household consumption of normal goods increases as household income increases. I use the IV with Continuous Price model to test for normalcy of tanker water but substitute the natural log of the quantity variable for the variable’s identity, so income elasticity is simply the coefficient of the natural log of the income variable. Tanker water may be one of three types of goods depending on the value of the coefficient. A positive coefficient less than 1 suggests that tanker water is a normal good because the quantity demanded increases as income increases, but less than linearly. A positive coefficient greater than 1 suggests that tanker water is a luxury good because the quantity demanded increases more than linearly as income increases. Finally, a negative coefficient suggests that tanker water is an inferior good because the quantity demanded decreases as income increases.

I find that the elasticity of income is negative, but extremely insignificant. This suggests that tanker water may be an inferior good. This is inconsistent with the theoretical model, albeit inconclusively, which predicted the tanker water to be a normal good. In fact, the preferences specification model, Stone-Geary preferences, does not allow goods to be inferior. I can resolve this empirical finding with the theoretical model by conceiving of qT as a number somewhere between 0 and -1 for tanker water, and not 0 as assumed in Chapter 2. The possibility that q is -1 precludes households who can avoid using the tanker water from participating in the market. Therefore, tanker drivers only enhance equity in the water market to create inequity in the water supplied by taking advantage of the ready market by selling suboptimal products.

Moreover, the finding of tanker water as an inferior good is consistent with reports that the piped water supply outlines the local power structure or that wealthy and powerful households receive more consistent water supply than poor households (Interview M, 2009). There is a chance, however, that the possibility of inferiority is spurious because the water supplied to the rich households in the area is likely underreported in the government data.

The possible inferiority of tanker water has incredible implications—unforeseen in the theoretical model—on the very existence of vertical equity in the tanker market. If tanker water is an inferior good, then households who can choose to consume water from sources other than tankers do. This suggests that the very existence of the tanker market serves some social good of enhancing equity because households with no other options for water supply turn to this market. Additionally, the possibility that tanker water is inferior flies in the face of the common misconception among foreign water workers that only the rich use tankers. One interviewee, for instance, when asked who uses tankers responded “those who pay” (Interview P, 2009).

Regardless of the income elasticity, many households use tanker water to fulfill their water demand, making vertical equity an important consideration in either case. If the good is normal, then vertical equity must be assessed to analyze the distribution of costs for this desirable and necessary good across the households on the basis of income. Alternatively, if the good is inferior, then the question of vertical equity is even more important because regressive pricing would be a double whammy against poor households: consuming an inferior product—perhaps due to questionable quality of the good—and paying more for the good.

In vertical equity terms, the question is whether the high-income households pay a smaller fraction of their income for tanker water than do low-income households, as predicted by the theoretical model. Given that the first stage least-squares regressions in the IV with Continuous Price model explains little variation in the price, I am unable to construct a reliable model to predict the price of water that households face. Therefore, I rely on price point comparisons to analyze the progressivity of the tanker market.

To uncover the particular relationship among household income and prices, I divide the households into five income brackets. The average monthly income of the first bracket is 100JD, of the second is 250JD, of the third is 400JD, of the fourth is 600JD, and of the highest is 1000JD. Table 4J summarizes average and total prices paid by the households in the five income brackets, both absolutely and relatively—the absolute level weighted by the household’s income.

According to these values, it is clear that the system is absolutely neutral, but relatively regressive. The poorest households pay the lowest price, though not significantly different from the average prices paid by the wealthiest households (T = 0.5227). However, the poorest households pay the highest price relative to their income (Bracket 1 to 2, T = 2.21257; Bracket 1 to 3, T = 2.2409; Bracket 1 to 4, T = 2.5478; Bracket 1 to 5, T = 2.3185).

The percentage of income used by households on tanker water ranges between 1% and 23% of the household income. These numbers seem to be absurdly high. However, a 2008 survey of household water use by a German consulting firm found that households spend upwards of 6.3% of their income to receive enough water to cover only 100 liters per capita per day (Sass and Brubach, 2008). The high envelope for the percentage of income required suggests that the income estimation has consistent downward bias. Regardless of the specific value, however, the poorest households pay the highest price relative to their income.

There is vertical inequity evidenced by this regressive relative pricing scheme. The regressive scheme most likely results from quantity-restricted demand. Based on interviews with households and corroborated by this data, the amount of water that households receive from tanker drivers is contingent on the size of their rooftop tanks. Yet, the amount of water that households pay for from tanker drivers is based on the size of the tanker driver’s tank truck. Not all households have an equal total volume of water tanks or total number of water tanks because household storage tanks are expensive (Gerlach and Franceys, 2009). Tank size increases with household income (T = 2.42), and so the gap between quantity received and the quantity that should be received given the total amount paid is larger in poor households than it is in rich households (corroborated by field observations of Interview W, 2009; and Odeh, 2009). In other words, poor and rich households pay similar amounts, but poor households receive less tanker water due to the limitation of smaller household rooftop tanks.

The role of household water tank volume in contributing to vertical inequity suggests the issue may also be described as a technological problem that reinforces the local cycle of poverty.

The household water storage tank is just one of the many additional costs of consuming water in the tanker market that comprise the total price—the effective price—of water. Though the coping methods are not statistically evaluated, they are worth commenting on briefly because they create further vertical inequity in the informal market. In response to suboptimal water quality, households cited boiling water, freezing water and both boiling and freezing water. About 30% of the households have filters for the water from the water storage tanks—piped water and tanker water—and an additional 26% of households also use bottled water for drinking and sometimes cooking. Filters range in price from 100 to 1000 JD for the initial investment and maintenance costs an average of 14JD every 4 months. Bottled water costs an average of a shocking 0.280 JD per liter, which is equivalent to 280 JD per cubic meter.

In addition to purchasing household water storage tanks, households also respond to suboptimal water quantity conditions by adjusting water consumption. One respondent shared her experience in the quote featured at the beginning of this chapter. Other households shared that they make more frequent trips to the mosque when water supply is low or seek donations from local water stores. Moreover, households with more time than money may travel to local public springs to fill up containers with water. While these coping behaviors, and many others, likely contribute to the inequity, insufficient data exists to determine their magnitude.

Conclusion

“The regular type of water management in Jordan is crisis management.”

Leading Engineer, who has worked in the area for a foreign consulting firm for more than fifteen years, 17 August 2009

Through its role mediating access to goods, the informal sector has increasingly substantial and far-reaching implications on household welfare. This paper aims to redirect the attention of economic practitioners in the developing world from the formal sector to the informal sector. I construct household demand for water from the informal market, and use the demand curve as the basis to explore the informal market’s performance. The scarcity of econometric work in this area is due in part to the a lack of data. I embraced this difficulty as a challenge, not a limitation. I came to understand the market by carefully mapping the interactions of households in my study area with water from the formal and informal sectors through observation, interviews, and primary source data collection. This paper offers a glimpse into the elusive informal market, and the theoretical and empirical methodologies herein developed may serve as a basis for future investigations of the informal market in other areas.

In Section I, I provide motivation for the paper by asking to what extent the informal market, given the failure of the formal sector, can equitably distribute a precious natural resource. I find that the informal market in the Emir Ali neighborhood of Ayn al-Basha, Jordan is an equitable distributor so long as a household’s payment for water is tied to consumption alone, not the consumption of others who buy from the same supplier. Stated differently, for the most part the tanker market equitably improves access to and distribution of water. However, when tanker drivers quote an inflated price to one household to compensate for a lower price offered to another, the informal market exhibits vertical inequity, and the poor suffer.

Section II develops a theoretical model of household demand for water from the informal sector as the residual demand not met by the piped water market. I assume that tanker firms are monopolistically competitive and that tanker drivers use their price-setting power to both discriminate against Palestinian households and charge more to poor households. To empirically test these predictions, I conducted a randomized household survey of 275 households in the Emir Ali neighborhood, as described in Section III.

The empirical investigation in Section IV confirms two antecedents to the empirical application of the theoretical model; constructs household water demand using three specifications that independently address simultaneity in supply and demand; and finally determines the level of equity in the market by exploring discrimination and regressive pricing. I use an IV with Continuous Price model to estimate that the household elasticity of demand is -1.56, or that when the price per cubic meter of tanker water increases by 1%, the quantity of tanker water demanded decreases by 1.59%, holding all else constant. The other significant predictors of household water demand are as predicted by the theoretical model: income, the number of people in the household, the quantity of piped water consumed, and interaction terms among these variables. I identify and dismiss possible omitted variables, and conclude that the only other threat of bias in the estimation of demand may be errors-in-variables bias.

Third, I use the IV with Continuous Price model to estimate the extent to which there is horizontal and vertical inequity in the tanker market. I find no evidence to corroborate the theoretical model’s prediction of horizontal inequity in the informal water market because I find that ethnicity has no significant influence on water consumption patterns. Additionally, I am unable to validate the theoretical model’s assumption that tanker water is a normal good. The possible inferiority of tanker water suggests that the informal market may be more important to the water allocation of poor households than wealthy households. This possibility raises the stakes of vertical inequity in the tanker market because the consumers in question are the most vulnerable water users in the area. I find that the poor households pay a greater proportion of their income for less water than wealthy households pay and receive. Two confounding factors likely create the vertical inequity: on the supply side, tanker drivers provide water at lower prices to households in their family network; and on the demand side, poor households have household storage tanks of less volume than wealthy households.

As water scarcity continues to increase in coming years, the formal supply will continue to fall short of household water demand, thus expanding the size of the informal tanker water market. The larger the tanker market, the more important the consequences of the vertical inequity in the market become. Intervening today may preempt a crisis tomorrow.

References

“Secondary Data DATABASE.” Department of Statistics, Jordan. 6 September 2008. http://www.dos.gov.jo/home_e.htm.

“Comprehensive Subscriber Survey of Balqa Governorate.” Dorsch Consult. April 2007.

Emir Ali Pilot Area Piped Water Network. Amman, Jordan: Engicon, 2009.

Emir Ali Pilot Area Houseplots. Amman, Jordan: Engicon, 2009.

Gerlach, Esther, and Richard Franceys. “Regulating Water Services for the Poor: The Case of Amman.” Geoforum 30, no. 3 (2009): 431-441.

Gleick, Peter. “Basic Water Requirements for Human Activities: Meeting Basic Needs.” Water International 21 (1996): 83-92.

Hajisyprou, S., P. Koundouri, and P. Pahardes. “Household Demand and Welfare: Implications of Water Pricing in Cyprus.” Environment and Development Economics 7 (2002): 659-685.

Iskandarani, M. “Water Market Participation and Effective Water Prices in Jordan: Globalization and Water Resources Management: The Changing Value of Water.” Paper presented at the University of Dundee International Specialty Conference, August 6-8, 2001.

Kanyoka, P., S. Farolfi, and S. Morardet. “Household’s Preferences and Willingness to Pay for Multiple Use Water Services in Rural Areas of South Africa: An Analysis Based on Le Blanc, David. “A Framework for Analyzing Tariffs and Subsidies in Water Provision to Urban Households in Developing Countries.” Working paper, DESA, 2008.

Mohamed, Ahmed Shawky, Alexander Kremer, and Manish Kumar. “Assessing the Efficiency and Equity of Water Subsidies: Spending Less for Better Services.” In Water in the Arab World, edited by N. Vijay Jagannathan, Ahmed Shawky Mohamed, and Alexander Kremer. Washington DC: World Bank, 2009.

Odeh, Nancy. “Towards improved partnerships in the water sector in the Middle EasT: A Case Study of Jordan.” PhD diss. Massachusetts Institute of Technology, 2009.

Olmstead, S. “The Economics of Water Quality.” Forthcoming in Review of Environmental Economics and Policy, 2010.

Ruijs, A., A. Zimmerman, and M. Van den Berg. “Demand and Distributional Effects of Water Pricing Policies.” Ecological Economics 66 (2008): 506-516.

Sass, Jan, and Katrin Brubach. “Water Management Program in the Middle Governorates—Zarqa, Balqa and Madaba.” GFA Consulting Group, in association with Engicon, 2008. Amman, Jordan.

Smith, Adam. The Theory of Moral Sentiments, edited by A.L. Macfie and D.D. Raphael. Oxford: Oxford University Press, 1976.

Strand, J., and I. Walker. “Water Markets and Demand in Central American Cities.” Environment and Development Economics 10 (2005): 313-335.

United Nations Food and Agriculture Organization. http://www.fao.org/nr/water/aquastat/main/index.stm.

Whittington, D., D. Lauria, and X. Mu. “A Study of Water Vending and Willingness to Pay for Water in Onitsha, Nigeria.” World Development 19, no. 2-3 (1991): 179-198.

Whittington, D., and K. Bal Kumar. “Coping with unreliable public water supplies: Averting expenditures by households in Kathmandu, Nepal.” Water Resources Research 41 (2005): W02012.

Whittington, Dale, Xinming Mu, and Robert Roche. “Calculating the Value of Time Spent Collecting Water: Some Estimates for Ukunda, Kenya.” World Development 18, no. 2 (1990): 269-280.

“Monitoring Environmental Progress: Report on a Work in Progress.” World Bank (1995).

“World Development Indicators Database.” World Bank. 2005. http://devdata.worldbank.org/wdi2005/Cover.htm.

Interviews

Interview A. Conducted by the researcher with a local Jordanian mid-level water worker. 10 August 2009.

Interview K. Conducted by the researcher with a former Minister of Water and Irrigation. 10 August 2009.

Interview M. Conducted by the researcher with a Western contractor who has worked in the water sector for more than 40 years. 1 July 2009.

Interview N. Conducted by the researcher with a Jordanian water engineer. 7 July 2009.

Interview P. Conducted by the researcher with a representative from a European consulting firm who conducted a qualitative household level survey on water use in 2007. 14 September 2009.

Interview Q. Conducted by the researcher with one of the senior managers at WAJ Balqa. 7 July 2009.

Interview W. Conducted by the researcher with the leading engineer at a foreign consulting firm that has worked in the area for more than fifteen years. 17 August 2009.

 

 

 

Comments:

1 COMMENT

  1. Nice work Sarah, but your conclusion (“The results suggest that the informal market is non-discriminatory but exhibits regressive pricing, which implies that the informal market falls short of a truly equitable distribution.”) cannot ever be satisfied. Prices that vary by income (were that possible) would THEN be discriminatory at the same time as they are “progressive.”

    Instead, use the rule of one instrument per objective, i.e., sell water at a market clearing (efficient) price and take care of equity via income transfers.

LEAVE A REPLY