WATER RESOURCES RESEARCH, VOL. 44, W08414, doi:10.

1029/2007WR006225, 2008
Click
Here
for
Full
Article
An analysis of nonpayment of water services
in Nuevo Laredo, Mexico
Ismael Aguilar-Benitez1 and Jean-Daniel Saphores2
Received 4 June 2007; revised 27 March 2008; accepted 23 May 2008; published 9 August 2008.
[1] In many developing and developed cities, nonpayment for water services undermines
the financial health of water utilities; in poorer areas, it also hinders the water utilities’
ability to serve poor households. However, nonpayment has not received much attention
in the literature so far. In this paper, we analyze nonpayment for a unique data set of
households from Nuevo Laredo, Mexico. Our fieldwork and basic statistics indicate that
nonpayment is very common and affects both poor and affluent households, so ability to
pay may only partly explain nonpayment. Using panel data models to control for
unobserved household characteristics (to avoid omitted variable bias) and after accounting
for monthly effects, we find that long-term delinquent households consume
3% more
water than do other families; this number is larger for households who are retired and
affluent, which further compounds the adverse consequences of nonpayment.
Citation: Aguilar-Benitez, I., and J.-D. Saphores (2008), An analysis of nonpayment of water services in Nuevo Laredo, Mexico,
Water Resour. Res., 44, W08414, doi:10.1029/2007WR006225.

1. Introduction delinquent households have no incentive to conserve water.

[2] Many urban water utilities, both in developed and in This reduces the amount of water available for ecosystems
developing countries, report high rates of nonpayment for and damages biodiversity.
water services. The situation is probably worse in econo- [4] A number of reasons have been proposed for explain-
mies in transition: in Tajikistan, for instance, nonpayment ing nonpayment for water services, including poor service
affects up to 90% of households [Fankhauser and Tepic, (i.e., poor water quality or periodic service interruption),
2007]. Nonpayment is also an issue in developed econo- inconvenient payment options, a sense of entitlement, a lack
mies. For example, in 2003 the water industry in the United of enforcement, and inability to pay.
Kingdom was facing an annual nonrecoverable debt of £164 [5] In Mexico, all of these reasons have been invoked.
million or the equivalent of 3% of household charges [Water Public distrust of water quality is reflected by the world’s
UK, 2005]. Latin America is not spared. The nonpayment highest per capita consumption of bottled water and the
rate for water utilities in Argentina ranges from 28% to second highest of carbonated drinks [Milenkovic, 2003;
91%, with an average of 37% [Urbiztondo, 2001]. In Mexico, Dickerson and Martinez, 2006]. As illustrated below for
31% of annual bills for urban water services remain unpaid Nuevo Laredo, payment for water services can also be time
on average [Comisión Nacional del Agua (CONAGUA), consuming and costly. Moreover, the 1917 Mexican Con-
2006]. stitution gives the state property rights for water, so access
[3] Widespread nonpayment for water services has seri- to potable water is seen as a constitutional right. As a result
ous economic, environmental, and public health impacts. but also for public health considerations, cutting off water
First, it undermines the financial health of water utilities by services from inhabited buildings is prohibited (Article 121
depriving them of much needed resources, which could of the Ley General de Salud) unless explicitly allowed by a
finance overdue maintenance. Second, it prevents water local law. This complicates the collection of overdue bills,
utilities from expanding their services to meet the demand which is weakly enforced partly for political reasons.
created by explosive urban growth; this hurts poor people Finally, opponents of water pricing argue that many house-
who have to buy their water from private water vendors. In holds are simply too poor to pay for water services. So far,
Nuevo Laredo, for example, water delivered by truck costs however, there has been not quantitative analysis of this
10 times more than municipal water (at least $1.8/m3 assertion.
compared to $1.86 for all of the first 16 m3, a flat rate for [6] In this context, our paper makes two contributions.
the first block). Third, uncollected funds could be used to First, it brings some quantitative evidence to the debate
expand water treatment and reduce discharges of raw about pricing water on the basis of a unique data set of
sewage in the environment, a serious health hazard. Finally, Nuevo Laredo households. A simple statistical analysis
nonpayment tends to promote inefficient water use since shows that nonpayment or delayed payment is almost as
prevalent among more affluent families as it is for poorer
1 ones. This result and circumstantial evidence collected
El Colegio de la Frontera Norte, Nuevo León, Mexico.
2
Civil and Environmental Engineering, University of California, Irvine, during our fieldwork suggest that payment behavior in
California, USA. Nuevo Laredo is to a large extent due to institutional
weaknesses, inconvenience, mistrust about the quality of
Copyright 2008 by the American Geophysical Union. service, or a sense of entitlement.
0043-1397/08/2007WR006225$09.00

W08414 1 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

[7] Second, in the absence of socioeconomic data, we They find that households in transition countries can usually
rely on panel data models to control for unobserved house- afford to pay their utility bills, which represent approxi-
hold characteristics. After accounting for monthly effects, mately 1.9% of their monthly income. Moreover, nonpay-
we find that Nuevo Laredo households who skip payments ment is much more pervasive for water than for other
consume on average 3% more water than those who pay on utilities, which suggests that affordability is not the main
time, which further compounds the burden of nonpayment. cause of nonpayment for water services. However, some
Overconsumption linked to nonpayment affects affluent low-income groups would find water difficult to afford (i.e.,
households at least as much as poorer ones; it is especially water would cost more than 5% of their income) if tariffs
high (around 9%) for more affluent, retired households. were increased too quickly to achieve cost recovery.
This is not surprising as households who don’t pay for water [12] In general, the water literature invokes several other
services have few incentives to conserve water. To our reasons for nonpayment. They include inadequate billing
knowledge, our paper is the first household-level panel and collection [Savedoff and Spiller, 1999]; a lack of
analysis of nonpayment for water services. payment options [Foundation for Water Research, 2003];
[8] This paper is organized as follows. In the next section, and lax enforcement caused by insufficient resources and
we review some published papers dealing with nonpayment institutional obstacles [Shirley and Menard, 2002].
for water services. In section 3, we present our study area [13] In addition, the belief that water is a human right and
and our data; we also discuss nonpayment in Nuevo Laredo. legal restrictions on enforcing payment are also likely to
In section 4, we introduce our modeling approach, and in play an important role in explaining high rates of nonpay-
section 5 we discuss our results. In section 6, we conclude ment. Gleick [1998] argues that access to basic water
and present suggestions for future work. Unless otherwise requirements is a human right supported implicitly by
indicated, monetary amounts are in U.S. dollars. international law and State practices. To buttress his argu-
ment, he lists several international agreements that implic-
2. Literature Review and Background itly support the idea of a human right for water.
[14] The belief in a right to water has led to specific laws
[9] For some time already, cost recovery has been iden- and legal obligations. In the United Kingdom, for example,
tified as a priority to achieve the goals of efficiency and the Water Industry Act of 1999 removed the right of water
adequate coverage of water services [e.g., see World Water utilities to disconnect household customers for nonpayment.
Council, 2003]. Many studies have analyzed the role of Prepayment devices, which ‘‘self-disconnect’’ if customers
water pricing [e.g., see Dinar and Subramanian, 1997; do not purchase enough credits, also became illegal (they
Hanemann, 1998; Dinar, 2000; Organization for Economic are still used in the gas and electricity sectors). This new law
Cooperative Development, 1999], as well as the willingness has led to a 17% increase of outstanding revenues between
and ability to pay for water services in developing countries fiscal year 1998 – 1999, just before the Water Industry Act
[e.g., see Whittington, 1998; Whittington et al., 1990, went into effect, and fiscal year 2003 –2004 [Water Services
1991, 1993]. Although a number of papers mention the Regulation Authority, 2003]. This increase cannot be attrib-
causes and consequences of nonpayment for water serv- uted to a reduction in enforcement because water companies
ices, surprisingly few authors focus on these issues. Rare have stepped up revenue collection efforts. It is also
exceptions include those of Booysen and Le [2001], noteworthy that it did not occur for other utilities such as
Urbiztondo [2001], Kayaga et al. [2003], and Fankhauser electricity or mobile phone service [Water UK, 2005].
and Tepic [2007]. [15] In Mexico, Article 121 of the Ley General de Salud
[10] A couple of these papers analyze sociodemographic (Health General Law, HGL) states that water service pro-
factors that may explain nonpayment; although they provide viders cannot disconnect households unless explicitly
valuable insights, they rely on regression analyses applied allowed by other applicable general ordinances passed by
to small samples. In their 2003 study of 11 towns in state or municipal authorities. In a 2004 letter to its
Uganda, Kayaga et al. [2003] find that women, better affiliates, the National Association of Water and Wastewater
educated consumers and higher-income households are Companies of Mexico (ANEAS) claimed that the HGL does
better at paying their water bills. Urbiztondo [2001] focuses not ban disconnection if it is formally mentioned in cus-
instead on 30 water utilities in Argentina. He investigates tomer contracts as a sanction for nonpayment. Nevertheless,
five variables to explain nonpayment for water services: an in practice few Mexican water utilities implement discon-
index of basic unsatisfied needs, service area population, nection for dealing with nonpayment.
population growth, percentage of households with metered [16] In the next section, we summarize our data collection
water services, and type of service. He finds that except for and fieldwork activities, before presenting the study area
the index of basic unsatisfied needs, these variables are and the data used in our quantitative analysis of nonpay-
statistically significant. ment for water services in Nuevo Laredo, Tamaulipas.
[11] Overall, the water literature gives a mixed picture of
nonpayment. In very poor areas, poverty seems to strongly
impact nonpayment, as shown by Booysen and Le’s [2001] 3. Study Area and Data
study of municipal services in South Africa. This is not the 3.1. Data Collection: Fieldwork
case in middle income or more affluent countries, however, [17] To collect available data and to understand the
as illustrated by Fankhauser and Tepic’s [2007] study of institutional context in Nuevo Laredo, we performed some
nonpayment among low-income consumers in transition extensive fieldwork. Between the end of October 2005 and
countries. Using household surveys, they analyze the af- August 2006, we visited Nuevo Laredo 10 times and
fordability (i.e., the share of monthly household income conducted approximately 20 formal interviews. Interview-
needed to pay utility bills) of electricity, heating, and water.
2 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

ees included current and former COMAPA officials, mem- cluding a 40% surcharge for sewage services for 27 m3 (the
bers of the Nuevo Laredo planning department, a former average water consumption for metered accounts in Nuevo
bridge commissioner, local personalities concerned with the Laredo) represents less than 3% of the estimated mean
water situation in various colonias, as well as several elected monthly household income based on data from Encuesta
officials. We also talked extensively about the water situa- Nacional de Empleo Urbano (INEGI). The same amount of
tion in Nuevo Laredo with staff members from the local water would cost slightly less in Piedras Negras, Coahuila,
office of El Colegio de la Frontera Norte. but approximately double in Tijuana, the most expensive
[18] In addition, we collected and analyzed various docu- city along the border. For reference, the benchmark for
ments provided by COMAPA (including water bills), Nuevo measuring affordability for water services ranges from
Laredo [2006], and an extensive consulting report from 2.5% of total household income in the United States to
Parsons [2002]. We also reviewed key water laws such as 5% of total household income for the Asian Development
the 1997 Mexican Water Law and its 2004 amendments, Bank [Fankhauser and Tepic, 2007].
documents from Diario Oficial [2003], and the 2006 [23] A 2002 diagnostic prepared for the North America
Tamaulipas Water Law, which abrogates a 1992 law mod- Development Bank reveals, however, several weaknesses
ified in 1995 and 2002. Finally, with the assistance of with water services in Nuevo Laredo. These weaknesses
Rosalı́a Chávez, a research assistant from El Colegio de le include an aging infrastructure that provides inadequate
Frontera Norte, we analyzed available water articles pub- water pressure [Parsons, 2002], high unaccounted-for water
lished in three local newspapers: El Mañana, La Tarde, and (public use, but also water losses and illegal connections),
the Laredo Morning Times (in Laredo). which represents a staggering 50% of water production
3.2. An Overview of the Study Area [CONAGUA, 2006], and nonpayment. Indeed, COMAPA
typically collects revenues for only 30% of the total water it
[19] Nuevo Laredo, Tamaulipas, is separated from Lar- produces [CONAGUA, 2006].
edo, Texas, by the Rio Grande, which also marks the U.S.- [24] According to COMAPA officials, nonpayment is
Mexico border. In 2000, there were fifty-five maquiladora pervasive throughout the city, not just in low-income areas
plants in Nuevo Laredo, which employed approximately 20 [El Mañana, 2006]. Some of the causes of nonpayment are
percent of the total formal labor force of Nuevo Laredo related to billing practices. COMAPA bills its customers
[Cañas et al., 2007]. Nuevo Laredo’s location at the monthly, one month after reading meters, and because of
northern terminus of the national railroad and of the Inter- distribution delays, many households only have 10 days to
American Highway makes it the top inland port in Latin pay. If a household does not pay on time, a delinquent
America. As a result, the Nuevo Laredo area has been notice and a disconnection warning are enclosed in its next
growing briskly (over 40% between 1990 and 2000 alone) bill, which arrives almost one month later. By contrast, an
for almost two decades, which has stressed water systems overdue electricity bill can result in service termination
on both sides of the border. According to the Mexican the day after the due date. Moreover, although COMAPA
census, Nuevo Laredo had approximately 311,000 inhab- charges 2.7% for late payment, this fee is typically
itants in 2000 and its metropolitan area exceeded 660,000 waived during campaigns organized to promote payment
inhabitants. (C. Canturosas, Nuevo Laredo Council member and
[20] Water services in Nuevo Laredo are provided by COMAPA Administrative Council member, personal com-
Comisión Municipal de Agua Potable y Alcantarillado munication, 17 February 2006).
(COMAPA), a public water utility. Most households in [25] Paying water bills in Nuevo Laredo can be costly and
Nuevo Laredo have access to water 24 hours a day. time consuming, which is likely to impact payment habits.
According to CONAGUA [2006], only 2% of Nuevo Laredo COMAPA has two downtown offices, but parking there is
households were without municipal water connection in difficult and access is impractical without a car. Water bills
2005, and 9% had no adequate sanitation services, which can also be paid at grocery stores for a fee of five Mexican
compares favorably with Mexico as a whole (UNICEF, n.d.). pesos (
$0.46), which is one fourth of the water and sewer
[21] At the end of 2000, COMAPA was managing 71,914 charge for first block water users. This could explain that
accounts, of which 92.7%, 4.4%, and 0.1% were for house- some households prefer bundling their payments. Five
holds, commercial, and industrial users respectively; 0.5% mobile payment units also serve impoverished neighbor-
of these accounts pertained to COMAPA employees (who hoods but they operate only during payment campaigns.
do not have to pay for water), and the last 2.3% was not [26] COMAPA has three strategies to deal with nonpay-
attributed. In 2000, households consumed 83.3% of billed ment. The first one consists in organizing radio and news-
water, and almost two thirds of them were metered. Metered paper campaigns that promote the payment of overdue
households faced an 11 block water schedule with a flat rate water bills by waiving late charges and offering discounts.
of 20.41 pesos for the first block and increasing marginal Our interviews with COMAPA staff indicate that these
rates thereafter. A noteworthy feature of this water schedule campaigns are typically implemented once a year. Unfortu-
is that marginal and average prices are equal for blocks two nately, COMAPA does not keep track of these events, nor
and above, which creates small jumps in water bills as does it record approved discounts, which may be granted
consumption moves from one block to the next. We should by Nuevo Laredo Council members or by water officials
also mention that retirees receive a 50% discount on their (E. Covarrubias, COMAPA’s Technical Department Man-
water bill. ager, personal communication, 31 October 2005).
[22] Compared to other cities along the border, water is [27] A second debt recovery strategy consists in installing
relatively affordable in Nuevo Laredo. The average monthly flow reducers on the water intake of chronically delinquent
bill for 27 m3 residential metered water (the average water households. As of December 2005, COMAPA had installed
consumption for metered accounts in Nuevo Laredo), in-
3 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

15,000 flow-reducing devices (J. Valdez, COMAPA General idencial’’. Retired households make up the rest, with 4.5%
Manager, personal communication, 21 April 2006). This for retired ‘‘Domestico’’ and 1.6% for retired ‘‘Residencial’’
approach avoids direct disconnection, whose legality is users. This data set includes both metered and unmetered
contested, but it is costly: a 12 inch flow-reducing device costs accounts and contains the best available information on
$25 to $30 to purchase and install. COMAPA charges $23 1999 – 2000 water consumption and payment for Nuevo
dollars per reconnection, but only for customers who owe at Laredo. It gives us a snapshot of nonpayment behavior in
least a thousand Mexican pesos; reconnection is free for the Nuevo Laredo.
others. High costs, the difficulty of getting enough devices, [32] However, this large data set includes unmetered
and insufficient staff explain the slow rate of installation of accounts and readings from broken meters, which are
flow-reducing devices (E. Solis, COMAPA Commercial unsuitable for our panel data models where we investigate
Manager, personal communication, 21 April 2006). the link between nonpayment and water use. From inter-
[28] The third debt recovery strategy is disconnection. views with COMAPA staff, we learned that a water con-
However, it is only used for long-term delinquent house- sumption of either 25 or 35 m3 (the amounts assumed for
holds; in 2004, for example, only 671 delinquent accounts nonmetered domestic accounts) is assumed in the absence
were disconnected. Disconnection is rarely used for at least of consumption data. Therefore, for our panel data models
two reasons: political opposition and a lack of funds to pay we did not analyze accounts whose water consumption for
for an intensive disconnection program. As a result, discon- three or more months was constant and equal to zero, 25 m3,
nection is not a credible threat and many households wait or 35 m3, or accounts with missing values.
for rebate campaigns to pay their overdue water bills. More [33] Since payment data in our database are unreliable,
generally, the lack of enforcement is also partly the result of we generated dummy variables to track skipped payments;
the widespread insecurity in Nuevo Laredo, as COMAPA for a given month, they indicate whether payment was
employees are afraid of inspecting private houses (E. Solis, skipped for just the previous bill, the previous two bills,
COMAPA Commercial Manager, personal communication, the last three to five bills, or at least the last six consecutive
23 April 2006). bills (see below for the exact definition of these dummy
3.3. Data variables). To have a full set of well defined dummy
variables, we did not use the first six months of 1999, and
[29] COMAPA provided the two years of household data we dropped the second part of 1999 to avoid possible
(from January of 1999 to December of 2000) analyzed in this fallouts from the June 1999 rate increase. As a result, we
study. We believe that COMAPA shared this unique data set have a full year of data without price changes (2000) to
with us because its managers are genuinely concerned about study the impact of nonpayment behavior on water con-
the water situation in Nuevo Laredo. Data include billing sumption. The final data set for our panel data models is a
route, billing date, monthly water consumption, and pay- balanced panel with 28,358 accounts (70.1% ‘‘Domestico,’’
ment. According to COMAPA, payment amounts in our 22.5% ‘‘Residencial,’’ 5.0% retired ‘‘Domestico,’’ and 2.5%
database are not reliable because they do not properly retired ‘‘Residencial’’ users) with 12 monthly observations
account for discounts on late bills. To circumvent this each.
problem, we ignore payment amounts and focus on whether [34] In addition, we collected monthly weather data but in
or not households made payments toward their water bills the models presented herein we use simple monthly dummy
using dummy variables. Unfortunately, COMAPA does not variables to capture weather effects and any other monthly
collect detailed socioeconomic data about its customer and all variation.
our contacts confirmed that this information is unavailable.
Because of cost, time, and the security situation in the study
area, it was not possible to conduct interviews of randomly 4. Modeling Methodology
selected households in Nuevo Laredo. [35] To assess the impact of nonpayment on water use,
[30 ] Apart from a miscellaneous group (‘‘others’’), we rely on panel data models. Our goal is to detect differ-
COMAPA classifies its customers in seven categories: five ences in water use depending on past nonpayment behavior
groups of households, as well as commercial and industrial after controlling for monthly effects (related to changing
users. Household groups include low-income users weather conditions, for example) and unobserved household
(‘‘Domestico’’), more affluent users (‘‘Residencial,’’ i.e., characteristics. These characteristics depend on factors such
households located in middle- and high-income areas), as household size, age composition, income, or habits; we
COMAPA employees, as well as retired ‘‘Domestico’’ and assume that they are constant during year 2000, which is
‘‘Residencial’’ users. This classification was originally important to obtain consistent estimates of the impact of
based on the size and location of houses. nonpayment on water use. Allowing for differences between
[31] To get an overview of household nonpayment be- household groups (‘‘Domestico,’’ ‘‘Residencial,’’ and the
havior, we set aside industrial and commercial users, as well corresponding groups of retirees), we write the log of water
as ‘‘others’’ and COMAPA employees. We also discarded consumption in period t for household i as a linear function
accounts with missing consumption or payment informa- of monthly dummy variables, payment behavior, unob-
tion, or account for which records started (new service) or served characteristics lumped into a household specific
stopped (accounts closed) between January of 1999 and constant, and an idiosyncratic error term (Model 1):
December of 2000 in order to get clear temporal informa-
tion about payment behavior. This gave us a large data set
 X
4

with 61,342 household accounts, which encompasses ap- ln qi;t ¼ b0 þ Ik ðiÞ b 1;k M1;t þ . . . þ b 11;k M11;t þ g 1;k D1;i;t
k¼1
proximately 92% of all household accounts. In our data set,
77.7% of accounts are ‘‘Domestico’’ and 16.2% are ‘‘Res- þ . . . þ g 4;k D4;i;t Þ þ ci þ ei;t ; ð1Þ

4 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

where i = 1. . .28358 indexes households; t = 1. . .12 indexes idiosyncratic error (ei,t) conditional on unobserved house-
the months of 2000; and k = 1, 2, 3, and 4 refers hold characteristics (ci) and on all explanatory variables for
respectively to ‘‘Domestico,’’ ‘‘Residencial,’’ retired ‘‘Do- all time periods is assumed to be zero [Wooldridge, 2002,
mestico,’’ and retired ‘‘Residencial’’ households. The p. 253]. This implies that all explanatory variables are
following also applies to equation (1): uncorrelated with the idiosyncratic error ei,t in all time
[36] 1. Value qi,t is the water consumption for household i periods, which makes sense here since we control in
in month t. equation (1) for specific household characteristics through
[37] 2. The indicator function Ik(i) equals 1 if household i ci and for seasonal effects through Mj,t. We also expect
belongs to group k and 0 otherwise. idiosyncratic shocks to have negligible effects on nonpay-
[38] 3. Mj,t, j 2 {1,. . .11} are monthly dummy variables. ment because the cost of water services is only a small
They equal 1 when the water consumption month t equals fraction of the income of the households in our database,
calendar month j, 1 when t is December (t = 12), and 0 which does not include very poor households since they do
otherwise. This definition allows expressing monthly varia- not receive municipal water services. It follows that the
tions from the conditional mean of the sample rather than fixed effect estimators (FE) is unbiased and consistent as the
from the excluded month [Baum, 2006]. We allow monthly number of households goes to infinity while the number of
variations to differ between groups because different types time periods is fixed [Wooldridge, 2000].
of households may be affected differently by seasonal [44] If there is no correlation between unobserved
variations (e.g., wealthier households may have larger household characteristics ci and our nonpayment and
gardens, for example). monthly dummies, a random effect (RE) model is a
[39] 4. Dn,i,t with n 2 {1,. . .,4} are mutually exclusive worthwhile alternative to a FE model; in fact, RE models
dummy variables that equal 1 if household i is delinquent in are more efficient if some technical conditions are satisfied
month t, and 0 otherwise. When n = 1, household i did not [Wooldridge, 2002, p. 260]. Low correlations between
make a payment for its last received water bill (not the one unobserved household characteristics and nonpayment sug-
for the current month) but made a payment for the one gest that nonpayment may result from the decision to
before. When n = 2, no payment was received for the last consume water freely while waiting for the next round
two received bills but a payment was made for the bill three of rebates put in place by COMAPA. We therefore also
months ago. Conversely, n = 3 corresponds to nonpayment estimate FE counterparts to our models.
for the last three to five water bills in a row, and n = 4 [45] To improve the reliability of our inferences, we
indicates consecutive nonpayment for six months or more. estimate our FE and RE models with both robust standard
[40] 5. Value ci represents unobserved characteristics for errors (using the Huber/White/sandwich estimator) and with
household i. It captures the impact of socioeconomic clustered errors, which provide a robust variance estimate
variables, such as income, habits, or their stock of appli- that accounts for possible correlation between the monthly
ances. It is assumed to be fixed over the period of our observations of a given household [Wooldridge, 2002,
analyses. We treat the cis as random draws from a distri- section 13.8.2].
bution with zero mean since a constant is included in [46] After estimating equation (1), we calculate Dn,k, the
equation (1). average percentage change in water consumption due to
[41] 6. Value ei,t is the idiosyncratic error term for nonpayment behavior n 2 {1,. . .,4} for household group
household i in month t. k 2 {1,. . .,4}, from
[42] In light of our hypothesis, we are therefore interested
in the sign and in the magnitude of the g i,k (i = 1. . .4, k = Dn;k ¼ 100ðeg n;k  1Þ: ð3Þ
1. . .4) coefficients. To test for differences in the impacts of
seasonality and nonpayment on different household groups,
we also estimate Model 2 defined by
5. Results


ln qi;t ¼ b0 þ b1 M1;t þ . . . þ b11 M11;t þ g1 D1;i;t þ . . . þ g4 D4;i;t 5.1. Overview of Payment Behavior
X
4 [47] Results of our nonpayment analysis of 61,342 house-
þ Ik ðiÞðb1;k M1;t þ . . . þ b11;k M11;t þ g1;k D1;i;t hold accounts are presented in Table 1 and in Figure 1. We
k¼2 partitioned household accounts according to their payment
þ . . . þ g4;k D4;i;t Þ þ ci þ ei;t : ð2Þ behavior, from those who always paid on time to those who
did not pay at least six months in a row (or even not at all).
In equation (2), bj for j = 1. . .11 is the pooled estimate of We focused on the year 2000 since longer-term nonpayment
how household water consumption in month j differs from indicators could not be constructed for the whole of 1999.
the conditional mean, whereas bj,k with k = 2,. . .4 represents [48] First, we see from Figure 1 that only 50% of Laredo
the corresponding difference between group k and ‘‘Domes- households on average pay their water bill on any given
tico’’ households. Likewise, gn = g n,1 for n = 1. . .4 is the month. This compares with approximately 79% of house-
estimate of the impact on water consumption for nonpay- holds who pay their electricity bill on time in Nuevo Laredo
ment patterns n (see the definition of n in the description of (A. Barrera, Comisión Federal de Electricidad, personal
Dn,i,t above); and gn,k for n = 1. . .4 and k = 1. . .4 is the communication, 31 May 2007); unlike for water, house-
corresponding difference between group k and ‘‘Domes- holds who do not pay their electric bill on time may be
tico’’ households. quickly disconnected. In fact, we see from Table 1 that only
[43] To perform standard statistical tests, we make the 12.6% of households made water bill payments every
strict exogeneity assumption: the expected value of the month in 2000; this percentage is slightly lower for

5 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

Table 1. Nonpayment Summary Statistics for 2000a

‘‘Domestico’’ ‘‘Residencial’’ Retired ‘‘Domestico’’ Retired ‘‘Residencial’’ All Households

Number of Number of Number of Number of Number of

Accounts Percent Accounts Percent Accounts Percent Accounts Percent Accounts Percent

Paid Each Month

5086 10.7 1645 16.5 689 25.2 307 31.2 7727 12.6

Skipped Payments for up to Two Consecutive Months

16877 35.4 3897 39.1 1190 43.6 436 44.3 22400 36.5

Skipped Payments for at Least Three and at Most Five Consecutive Months
10864 22.8 2105 21.1 475 17.4 143 14.5 13587 22.1

Skipped Payments for at Least Six Consecutive Months but Made at Least One Payment During the Year
8811 18.5 1490 15.0 275 10.1 79 8.0 10655 17.4

Did Not Make a Single Payment

6034 12.7 818 8.2 101 3.7 20 2.0 6973 11.4

Total Number of Accounts With Information About Water Use and Payments
47672 100 9955 100 2730 100 985 100 61342 100
a
Table 1 is based on the 61342 households for which payment information data were available during year 2000. These also include accounts on a flat
rate and accounts with broken meters.

‘‘Domestico’’ users, and over 2.5 times higher for retired [50] Even more financially damaging for COMAPA,
households (up to 31.2% for retired ‘‘Residencial’’ users). Table 1 reveals that 28.8% of households did not pay their
[49] Second, over 36% of households skipped payments water bills for at least 6 months in a row; of those, 17.4%
for one or two months in 2000 (and many did so repeatedly made at least one payment during the year but 11.4% simply
during the year); interestingly, a larger percentage of afflu- did not pay at all. The former may have simply been waiting
ent and retired households adopted this behavior compared for rebates and payment campaigns; unfortunately,
to ‘‘Domestico’’ users. Moreover, 22.1% of households COMAPA did not keep track of these campaigns. For the
waited even longer (from 3 to 5 months) to make payments. latter, although some families were likely genuinely unable
In that case, the percentage of ‘‘Domestico’’ users is slightly to pay, the high percentage of households skipping all
larger than for other categories, especially compared to payments, especially the 8% of more affluent ‘‘Residencial’’
retirees. One possible explanation is that households wanted users, suggests that a number of households are taking
to avoid the hassle and/or the fees associated with paying advantage of a lack of enforcement linked to ambiguities
water bills at non-COMAPA payment locations. in the law. Finally, we note that retired households are much

Figure 1. Monthly breakdown of households who made payments toward their water bill. Figure 1 is
based on the 61342 households for which payment information data were available during year 2000
(47672 ‘‘Domestico,’’ 9955 ‘‘Residencial,’’ 2730 retired ‘‘Domestico,’’ and 985 retired ‘‘Residencial’’
households). These 61342 households also include accounts on a flat rate and accounts with broken
meters. On average, only 50% of households made payments each month. According to Comisión
Municipal de Agua Potable y Alcantarillado, however, most households caught up on overdue amounts.
6 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

Table 2. Summary Statistics for Panel Data Seta less likely to skip payment for six months or more; we
‘‘Domestico’’ ‘‘Residencial’’ conjecture that these households are more risk averse than
younger households.
Standard Standard
Variable Mean Deviation Min Max Mean Deviation Min Max 5.2. Panel Data Models
Not Retired
5.2.1. Sample Characterization
Ln(qi,t) 3.239 0.800 0 6.818 3.347 0.785 0 6.875 [51] Basic statistics for the data set of 28,358 household
D1,i,t 0.165 0.371 0 1 0.159 0.365 0 1 accounts used to estimate our panel data models are presented
D2,i,t 0.070 0.256 0 1 0.064 0.245 0 1 in Table 2. We see that retired households tend to consume
D3,i,t 0.094 0.291 0 1 0.079 0.270 0 1 slightly more than working ones, and that ‘‘Residencial’’
D4,i,t 0.153 0.360 0 1 0.113 0.316 0 1
Mj,t 0.000 0.408 1 1 0.000 0.408 –1 1 users consume more water than ‘‘Domestico’’ users, except
when they are retired. In general, nonpayment is more
Retired prevalent among ‘‘Domestico’’ than ‘‘Residencial’’ house-
Ln(qi,t) 3.614 0.705 0 5.756 3.490 0.659 0 6.080 holds. Moreover, many households are either late exactly one
D1,i,t 0.141 0.348 0 1 0.133 0.340 0 1
D2,i,t 0.049 0.215 0 1 0.046 0.209 0 1 bill, or delinquent for six month or more.
D3,i,t 0.059 0.235 0 1 0.051 0.220 0 1 [52] Figure 2 gives the monthly percentage of households
D4,i,t 0.063 0.242 0 1 0.045 0.208 0 1 who did not pay at least their last six water bills. On average,
Mj,t 0 0.408 – 1 1 0.000 0.408 – 1 1 13.7% of households each month were in that situation in
a
N denotes the number of observations in a group (‘‘Domestico’’ or 2000. ‘‘Domestico’’ households have the highest average
‘‘Residencial,’’ retired or not). For not retired ‘‘Domestico,’’ N = 238,476, percentage of long-term delinquent users (15.3%), but the
and for retired ‘‘Domestico,’’ N = 16,884. For not retired ‘‘Residencial,’’ corresponding percentage of ‘‘Residencial’’ household is
N = 76,560, and for retired ‘‘Residencial,’’ N = 8,376. There are 12 high too (11.3%); these percentages are less than half as
observations for each account, so N/12 is the number of accounts in a
group. Ln(qi,t) is the logarithm of monthly water consumption. Dk,i,t, k = large for retired households (6.3% for ‘‘Domestico’’ and
1,. . .,4 and Mj,t, t = 1,. . .,12 are dummy variables. D1,i,t = 1 if household i 4.5% for ‘‘Residencial’’ users).
did not pay exactly its last bill in billing period t, and 0 otherwise; D2,i,t = 1 [53] Unfortunately, Figure 2 also shows an increase
if household i did not pay exactly its last two bills in billing period t, and 0 during 2000 in the number of long-term delinquents among
otherwise; D3,i,t = 1 if household i did not pay its last consecutive three to
five bills in billing period t, and 0 otherwise; and D4,i,t = 1 if household i did
‘‘Domestico’’ households, so the assumptions necessary to
not pay at least its last six bills in a row in billing period t and 0 otherwise. obtain consistent estimates of our model parameters are not
Moreover, Mj,t = 1 when t (the water consumption month) equals j, 1 fully verified for ‘‘Domestico’’ households. Fortunately,
when the consumption month is December, and 0 otherwise. Monthly other nonpayment categories do not exhibit any clear trend
dummies are centered with respect to December, the last billing month of during 2000 for any other household group, apart from
2000, to express variations from the conditional mean of the sample rather
than deviations from the excluded month [Baum, 2006]. some seasonal variations, which we explicitly account for in
our models.
5.2.2. Estimated Coefficients for Model 1
[54] Our models were estimated using Stata. Estimates of
the impact of nonpayment on water use for Model 1

Figure 2. Monthly delinquent households who made no payments for 6 months or more. Figure 2 is
based on the 28358 households in the data set used to estimate our panel data models. They include
19873 ‘‘Domestico,’’ 6380 ‘‘Residencial,’’ 1407 retired ‘‘Domestico,’’ and 698 retired ‘‘Residencial’’
households. On any given month of year 2000, an average of 13.7% of households had not paid their
water bills for at least six consecutive months.

7 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

Table 3. Estimated Nonpayment Coefficients for Model 1 and Impact on Water Usea
Model Parameter FE FE With Clustered Errors RE RE With Clustered Errors

‘‘Domestico’’ (238,476 Observations for 19,873 Households)

g 1,1 0.0034 0.0034 0.0016 0.0016
(0.0024) (0.0022) (0.0024) (0.0022)
g 2,1 0.0015 0.0015 0.0039 0.0039
(0.0037) (0.0036) (0.0036) (0.0036)
g 3,1 0.0081* 0.0081* 0.0029 0.0029
(0.0037) (0.0046) (0.0036) (0.0045)
g 4,1 0.0272** 0.0272** 0.0273** 0.0273**
(0.0047) (0.0067) (0.0044) (0.0063)

‘‘Residencial’’ (76,560 Observations for 6,380 Households)

g 1,2 0.0049 0.0049 0.0070 0.0070*
(0.0046) (0.0042) (0.0045) (0.0042)
g 2,2 0.0074 0.0074 0.0107 0.0107
(0.0071) (0.0068) (0.0070) (0.0068)
g 3,2 0.0078 0.0078 0.0128* 0.0128
(0.0073) (0.0088) (0.0071) (0.0085)
g 4,2 0.0322** 0.0322* 0.0417** 0.0417**
(0.0100) (0.0134) (0.0093) (0.0125)

Retired ‘‘Domestico’’ (16,884 Observations for 1,407 Households)

g 1,3 0.0059 0.0059 0.0220** 0.0220**
(0.0078) (0.0067) (0.0079) (0.0067)
g 2,3 0.0093 0.0093 0.0292* 0.0292*
(0.0127) (0.0118) (0.0125) (0.0117)
g 3,3 0.0108 0.0108 0.0352* 0.0352*
(0.0154) (0.0194) (0.0153) (0.0190)
g 4,3 0.0456* 0.0456* 0.0794** 0.0794**
(0.0217) (0.0279) (0.0212) (0.0266)

Retired ‘‘Residencial’’ (8,376 Observations for 698 Households)

g 1,4 0.0049 0.0049 0.0162 0.0162
(0.0127) (0.0112) (0.0123) (0.0109)
g 2,4 0.0259 0.0259 0.0415* 0.0415*
(0.0221) (0.0220) (0.0218) (0.0216)
g 3,4 0.0100 0.0100 0.0095 0.0095
(0.0227) (0.0244) (0.0223) (0.0235)
g 4,4 0.0874** 0.0874* 0.1161** 0.1161**
(0.0287) (0.0427) (0.0268) (0.0402)
Corr(ci, Xib) 0.0026 0.0026 (assumed 0) (assumed 0)

Impact on Water Use for FE Models Impact on Water Use for RE Models

D4,1 +2.8% +2.8%

D4,2 +3.3% +4.3%
D4,3 +4.7% +8.3%
D4,4 +9.1% +12.3%
a
Variables and parameters are defined in equation (1). For the fixed effects (FE) and random effects (RE) models without clustered errors, we report
robust standard errors. Double asterisks and single asterisks indicate p values 0.01 and 0.10, respectively. For impact of nonpayment on water use, see
equation (4).

(coefficients g i,j with i = 1. . .4 and j = 1. . .4) are reported in in a row is statistically significant but only for ‘‘Domestico’’
Table 3; to save space, we did not report the constant term households and then its magnitude is small, and so is its
and the monthly dummy coefficients. The F test for indi- impact on consumption (D3,1 = 0.8% from equation (4)). By
vidual effects is highly significant, so pooled OLS would contrast, g 4,j (j = 1. . .4) is statistically significant for all
have been inappropriate here. In addition, three quarters of household categories and the corresponding impact on
the variations in the logarithm of water consumption come water use ranges from D4,1 = 2.8% for ‘‘Domestico’’ to
from individual effects (this is given by the ratio of the D4,4 = 9.1% for retired ‘‘Residencial’’ households. As
estimated variance of ci divided by the sum of the estimated expected, the impact on water consumption of long-term
variances of ci and ei,t). nonpayment is larger for ‘‘Residencial’’ than for ‘‘Domes-
[55] Let us first focus on FE models. We see that the tico’’ households, and this impact is larger for retired
coefficients for short-term (one or two months) nonpay- households compared to working ones (retired people
ment, g 1,j and g 2,j with j = 1. . .4, are insignificant for all typically spend more time at home and they may garden
four categories of households, so people who occasionally more). We note that estimated standard errors from the FE
skip a payment may simply be avoiding the inconvenience model with clustered errors are larger for g 3,j but especially
of paying their bill (the time spent waiting in line and/or the for g 4,j (j = 1. . .4), which suggests that errors for each
fee charged at non-COMAPA location). g 3,j (j = 1. . .4) for household are correlated over time.
households who skipped payments between 3 and 5 months

8 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

Table 4. Estimated Nonpayment Difference Coefficients for Model 2a

Model
Parameter FE FE With Clustered Errors RE RE With Clustered Errors

‘‘Residencial’’ (76,560 Observations for 6,380 Households)

g1,2 0.0015 0.0015 0.0086* 0.0086*
(0.0052) (0.0048) (0.0051) (0.0047)

g2,2 0.0059 0.0059 0.0146* 0.0146*

(0.0080) (0.0077) (0.0078) (0.0076)
g3,2 0.0003 0.0003 0.0099 0.0099
(0.0082) (0.0099) (0.0079) (0.0096)
g4,2 0.0049 0.0049 0.0144 0.0144
(0.0110) (0.0150) (0.0103) (0.0140)

Retired ‘‘Domestico’’ (16,884 Observations for 1,407 Households)

g1,3 0.0025 0.0025 0.0236** 0.0236**
(0.0081) (0.0070) (0.0083) (0.0071)
g2,3 0.0077 0.0077 0.0331* 0.0331**
(0.0132) (0.0124) (0.0130) (0.0122)
g3,3 0.0027 0.0027 0.0323* 0.0323*
(0.0159) (0.0199) (0.0157) (0.0195)
g4,3 0.0184 0.0184 0.0521* 0.0521*
(0.0222) (0.0287) (0.0216) (0.0273)

Retired ‘‘Residencial’’ (8,376 Observations for 698 Households)

g1,4 0.0015 0.0015 0.0178 0.0178
(0.0129) (0.0114) (0.0126) (0.0111)
g2,4 0.0243 0.0243 0.0454* 0.0454*
(0.0225) (0.0223) (0.0220) (0.0219)
g3,4 0.0180 0.0180 0.0066 0.0066
(0.0230) (0.0249) (0.0226) (0.0240)
g4,4 0.0602* 0.0602 0.0888** 0.0888*
(0.0290) (0.0432) (0.0271) (0.0406)
Corr(ci, Xib) 0.0026 0.0026 (assumed 0) (assumed 0)
a
Variables and parameters are defined in equation (2). For the FE and RE models without clustered errors, we report robust
standard errors. Double asterisks and single asterisks indicate p values 0.01 and 0.10, respectively. For ‘‘Domestico’’
(238,476 observations for 19,873 households), see Table 3 since gn = g n,1 for n = 1. . .4.

[56] We also estimated RE models with robust and households who did not pay for six consecutive months or
clustered errors. With the exception of g 1,1 and g 1,3, RE more.
coefficients are within two standard deviations of FE [58] For completeness, we also present results for RE
coefficients, and they suggest a larger impact of nonpay- models. These models detect small differences for both
ment on water use, especially for retired households. How- short-term nonpayment for all categories and for long-term
ever, RE models imply differences in water use for all retired delinquents, but this may be caused by slightly biased
retired ‘‘Domestico’’ households who skip payments, and estimates.
also for retired ‘‘Residencial’’ households who skip pay- 5.2.4. Results in Context
ment for up to two months, which is difficult to explain. A [59] To interpret these results, it is useful to put them in
Hausman test of the correlation between ci and our other context. First, year 2000 saw an improvement of the
explanatory variables, Xi, was rejected at 1%, so our RE Mexican economy, which was rebounding from the 1994
estimates are likely biased. Our preferred model is therefore devaluation of the peso. The unemployment rate (which in
the FE model with clustered errors. Mexico is questionable for a number of reasons but still
5.2.3. Estimated Coefficients for Model 2 provides a useful trend) decreased throughout the year (see
[57] To see if there are significant differences between statistics available at www.oecd.org), and the economy of
household groups regarding the impact of nonpayment on border cities (such as Nuevo Laredo) picked up. Second, we
water use, we estimated Model (2). The ‘‘Domestico/ know from Table 1 that nonpayment affects both more
Residencial’’ classification may also be considered a coarse affluent and poorer households, although to a different
proxy for income. Results are shown in Table 4. For FE degree. We also saw above that Nuevo Laredo households
models, we see that the only (mildly) statistically significant are much better at paying on time their electricity bill
difference emerges for retired ‘‘Residencial’’ households (which is typically higher than their water bill) and that
who did not pay their water bills for at least six consecutive late payment penalties (including disconnection) are
months, so the impact of long-term nonpayment does not enforced for electricity. This suggests that, although they
differ statistically between household categories, except could afford to, many households simply do not pay their
possibly for retired ‘‘Residencial’’ customers. This is not water bill on time for a variety of reasons, including distrust
very surprising since the standard errors of g 4,3 and g 4,4 of water quality, inconvenience, high transaction costs, a
(long-term delinquent retired households) are relatively sense of entitlement, weak enforcement, or the perverse incen-
large, which is likely due to the fairly small number of retired tives of payment campaigns. Since delinquent households

9 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

face a lower effective water price, it is not surprising that household categories (see Table 1), average monthly water
many of them consume more water than households who pay consumption (from Table 2), and percentage of water
on time. Conversely, a larger percentage of ‘‘Domestico’’ than overconsumption (using results for FE models, at the
‘‘Residencial’’ households skipped water bill payments in bottom of Table 3). This is enough to serve approximately
2000, so affordability cannot be ruled out at least for a 1,400 Nuevo Laredo households who do not receive water
segment of the population. services from the city and who buy at least some of their
[60] To save space, we did not report the estimated water from expensive water vendors.
coefficients of the monthly dummy variables. We find that [65] Estimating the utility-wide impact of nonpayment on
all time dummy coefficients are statistically significant, with COMAPA’s financial situation is unfortunately much more
signs and magnitudes consistent with seasonal variations: as difficult; it would require a detailed accounting analysis that
expected, water consumption is lower during late fall, we are unable to perform for lack of data. Difficulties
winter, and early spring, and it is higher during the summer. abound. First, as indicated above, the payment records we
Interestingly, Model 2 confirms that estimates of the month- received from COMAPA are not reliable because they
ly dummy variables coefficients differ between household reflects neither the rebates that are routinely granted by
categories. This suggests systematic seasonal effects for various officials nor the discounts given to nonpayers
different household categories, which may be partly due during payment campaigns. We also do not know the cost
to differences in outdoor water use (wealthier households of these events, which involve advertising and staff time, or
are likely to have more land). the cost of sending reminders to many late payers. Second,
5.2.5. Robustness Checks we do not know what percentage of long-term delinquent
[61] We also estimated our models for different partitions households will eventually pay their overdue bills. Third,
of our data set on the basis of water consumption (results are irregular revenues may require COMAPA to borrow money
not shown). We found that our estimates of overconsump- from banks in order to pay its bills. Finally, our analysis
tion are not due to a few heavy water users. However, focuses on metered accounts, which make up just two thirds
contrasting results for households who consume only in of all household accounts. A simple calculation for year
blocks one to three (low water users) with households who 2000 illustrates, however, the order of magnitude of the
consume in higher blocks (five to eleven) shows that more potential impacts from nonpayment: the total amount owed
affluent, heavy water users are consuming more water when by the 6,973 households who did not make any payment
they regularly skip water bill payments. represents 11.9% of the total billing for the 61,342 house-
[62] Finally, we estimated models with different dummy hold accounts for which payment information was available
variables for nonpayment: we replaced the indicator vari- (see Table 1).
able for households who did not pay their last three to five [66] Future work should analyze nonpayment in cities
bills in a row with a dummy for people who did not pay that keep better track of payments in order to distinguish
exactly their last three bills and people who did not pay their households who simply end up paying late from those who
last four to five bills in a row. Estimated coefficients (not are chronically delinquent. To better understand the causes
shown here) are similar to the results discussed above. of nonpayment, it will be necessary to conduct household
surveys to collect socioeconomic and demographic infor-
6. Conclusions mation; this was unfortunately beyond the capabilities of
this project, especially given the precarious security situa-
[63] In this paper we have analyzed the extent of house- tion in Nuevo Laredo. Breaking the vicious cycle of
hold nonpayment of water services in Nuevo Laredo and its nonpayment is a necessary step to better manage limited
impact on water demand using panel data models. Our water resources, which are especially scarce at the U.S.-
statistical analysis for 1999– 2000 reveals that nonpayment Mexico border [e.g., see Booker et al., 2005].
is widespread among both affluent (‘‘Residencial’’) and
lower-income (‘‘Domestico’’) customers. Indeed, only [67] Acknowledgments. Helpful comments from Mireille Jacobson,
50% of households on average paid their water bills on Ivan Jeliazkov, Giuseppe Ragusa, Belem Vasquez, and participants in a
time in 2000, which compares to 79% for electricity where seminar at the University of California, San Diego, are gratefully acknowl-
penalties and disconnections are enforced. In addition, we edged. We are also grateful to the Associate Editor, Michael Moore, and
three anonymous referees for making a number of excellent suggestions
show that metered households who fail to pay their water that improved this paper significantly. As usual, all remaining errors are our
bills for six consecutive months or more tend to consume responsibility.
3% more water on average; this overconsumption can reach
9% for more affluent households. Although affordability References
may play a role, to a large extent nonpayment is also likely Baum, C. F. (2006), An Introduction to Modern Econometrics Using Stata,
the consequence of lax enforcement, institutional weak- Stata Press, College Station, Tex.
Booker, J. F., A. M. Michelsen, and F. A. Ward (2005), Economic impact of
nesses, and a sense of entitlement about water services. alternative policy responses to prolonged and severe drought in the Rio
To our knowledge, this is the first quantitative analysis of Grande Basin, Water Resour. Res., 41, W02026, doi:10.1029/
nonpayment using household-level data, which is surprising 2004WR003486.
because nonpayment is a serious issue for many water Booysen, F., and R. Le (2001), Non-payment of services: A problem of
ability-to-pay, S. Afr. J. Econ., 69(4), 475 – 697.
utilities around the world. Cañas, J., T. M. Fullerton Jr., and W. M. D. Smith (2007), Maquiladora
[64] Although the utility-wide impact of nonpayment on employment dynamics in Nuevo Laredo, Growth Change, 38(1), 23 – 38,
water use is fairly modest, it is not negligible in this arid doi:10.1111/j.1468-2257.2007.00351.x.
region: it represents approximately 35,800 m3 per month Comisión Nacional del Agua (CONAGUA) (2006), State of the drinking
based on a simple calculation that takes into account the water, sewage and sanitation subsector (in Spanish), report, Secr. de
Medio Ambiente y Recursos Nat., Mexico City.
percentage of metered households in each of our four
10 of 11
W08414 AGUILAR-BENITEZ AND SAPHORES: NONPAYMENT OF WATER SERVICES W08414

Diario Oficial (2003), Rules of operation for programs dealing with irriga- Savedoff, W. D., and P. Spiller (Eds.) (1999), Spilled Water: Institutional
tion, water and sanitation by the National Water Commission, and any Commitment in the Provision of Water Services, Inter-Am. Dev. Bank,
amendments that apply starting in the year 2003 (in Spanish), report, Washington, D. C.
Mexico City. Shirley, M. M., and C. Menard (2002), Cities awash: A synthesis of the
Dickerson, M., and C. Martinez (2006), Mexico hopes tax on soda will country cases, in Thirsting for Efficiency, edited by M. M. Shirley,
refill lost oil revenue, Los Angeles Times, Business News, 12 Dec. pp. 1 – 41. The World Bank, Washington, D. C.
Dinar, A. (2000), The Political Economy of Water Pricing Reform. Wa- Urbiztondo, S. (2001), Collecting payment for water and sewer services in
shington D. C.: The World Bank, Oxford Univ. Press, New York. Argentina (in Spanish), report, FIEL, Buenos Aires.
Dinar, A., and A. Subramanian (1997), Water pricing experiences, Tech. Water Services Regulation Authority (2003), Industry information on
Pap. 386, The World Bank, Washington, D. C. household debt levels and debt recovery costs, RD 32/03, 23 Aug.,
El Mañana (2006), If they seek rate increases for drinking water (in Birmingham, U. K.
Spanish), 01/21. Water UK (2005), Household debt in the water industry, briefing paper,
Fankhauser, S., and S. Tepic (2007), Can poor consumers pay for energy London.
and water?: An affordability analysis for transition countries, Energy Whittington, D. (1998), Administering contingent valuation surveys in de-
Policy, 35(2), 1038 – 1049, doi:10.1016/j.enpol.2006.02.003. veloping countries, World Dev., 26(1), 21 – 30, doi:10.1016/S0305-
Foundation for Water Research (2003), Researching, developing and testing 750X(97)00125-3.
of payment strategies for the lower income groups at four selected com- Whittington, D., J. Briscoe, X. Mu, and W. Barron (1990), Estimating the
munities in order to manage charges for water use, Rep. 1202/1/03, willingness to pay for water services in developing countries: A case
Johannesburg. study of the use of contingent valuation surveys in southern Haiti, Econ.
Gleick, P. (1998), The human right to water, Water Policy, 1(5), 487 – 503, Dev. Cult. Change, 38(2), 293 – 311, doi:10.1086/451794.
doi:10.1016/S1366-7017(99)00008-2. Whittington, D., D. Lauria, and X. Mu (1991), A study of water vending
Hanemann, W. M. (1998), Price and rate structures, in Urban Water Demand and willingness to pay for water in Onitsha, Nigeria, World Dev., 19(2 – 3),
Management and Planning, edited by D. D. Baunmann, J. J. Boland, and 179 – 198, doi:10.1016/0305-750X(91)90254-F.
W. M. Hanemann pp. 137 – 169, McGraw-Hill, New York. Whittington, D., D. Lauria, A. M. Wright, K. Choe, J. A. Hughes, and
Kayaga, S., J. Calvert, and K. Sansom (2003), Paying for water services: V. Swarna (1993), Household demand for improved sanitation services
Effects of household characteristics, Utilities Policy, 11, 123 – 132, in Kumasi, Ghana: A contingent valuation study, Water Resour. Res.,
doi:10.1016/S0957-1787(03)00034-1. 29(6), 1539 – 1560.
Milenkovic, Z. (2003), Bottled water sales get Latin fever, 14 April, Euro- Wooldridge, J. M. (2000), Introductory Econometrics: A Modern Ap-
monitor Int., London. proach, 2nd ed. South-Western Coll., Belmont, Calif.
Nuevo Laredo (2006), Executive annex number 10/06; Comprehensive plan Wooldridge, J. M. (2002), Econometric Analysis of Cross Section and
for drinking water, sewage and sanitation of the City of Nuevo Laredo (in Panel Data, 1st ed. MIT Press, Cambridge, Mass.
Spanish), report, Mexico. World Water Council (2003), Report of the World Panel on financing water
Organization for Economic Cooperative Development (1999), The Price of infrastructure: Financing water for all, Marseille, France.
Water: Trends in OECD Countries, Paris.
Parsons (2002), Report on the Nuevo Laredo, Tamaulipas, drinking water
system: Diagnosis, modeling and expansion (in Spanish), Int. Boundary 

and Water Comm., El Paso, Tex. I. Aguilar-Benitez, El Colegio de la Frontera Norte, Técnicos 277 Colonia
Tecnológico Monterrey, Nuevo León, México. (iaguilar@colef.mx.)
J.-D. Saphores (corresponding author), Civil and Environmental
Engineering, University of California, Irvine, CA 92697, USA. (saphores@
uci.edu.)

11 of 11