Article
pubs.acs.org/est
Freshwater Availability and Water Fetching Distance Affect Child
Health in Sub-Saharan Africa
Amy J. Pickering†,§,* and Jennifer Davis‡,§,*
†
Emmett Interdisciplinary Program in Environment and Resources, School of Earth Sciences, Stanford University, Stanford, California
94305
‡
Environment and Water Studies, Civil and Environmental Engineering, Stanford University, Stanford, California 94305
§
Woods Institute for the Environment, Stanford University, Stanford, California 94305
S Supporting Information
*
ABSTRACT: Currently, more than two-thirds of the population in Africa must leave their home to
fetch water for drinking and domestic use. The time burden of water fetching has been suggested to
influence the volume of water collected by households as well as time spent on income generating
activities and child care. However, little is known about the potential health benefits of reducing water
fetching distances. Data from almost 200 000 Demographic and Health Surveys carried out in
26 countries were used to assess the relationship between household walk time to water source and
child health outcomes. To estimate the causal effect of decreased water fetching time on health,
geographic variation in freshwater availability was employed as an instrumental variable for one-way
walk time to water source in a two-stage regression model. Time spent walking to a household’s main
water source was found to be a significant determinant of under-five child health. A 15-min decrease in
one-way walk time to water source is associated with a 41% average relative reduction in diarrhea
prevalence, improved anthropometric indicators of child nutritional status, and a 11% relative
reduction in under-five child mortality. These results suggest that reducing the time cost of fetching
water should be a priority for water infrastructure investments in Africa.
■
INTRODUCTION
Currently, 44% of the world’s population must leave their
homes to fetch the water they need for drinking and other
domestic uses. Most of the households using such “nonnetworked” water supplies are located in sub-Saharan Africa
and southern Asia.1 Women and children are known to be the
main water carriers in low-income countries, often spending
more than one hour per water collection trip and making
multiple trips per day.2,3 Although the large global time burden
of water fetching is well known, the majority of water-related
health impact research has focused on water quality,4,5 leaving
the relationship between time spent water fetching and health
understudied.6
The time cost and physical burden associated with water
fetching translates into reduced volumes of water accessed by
households using non-networked sources. Previous research has
found an inverse association between volume of water used and
walk time to source; in particular, households whose water
sources are located more than 30 min away often collect less
water than is believed necessary for basic needs.7 The proximity
of water available to a household has also been demonstrated to
correlate with the frequency of hygiene behavior. For example,
mothers in Burkina Faso with piped water supplies in their
yards were three times more likely to perform regular hand
washing as compared to mothers using wells or public
standpipes outside their yards.8 Households in East Africa
with access to piped water on their plot have been found to use
© 2012 American Chemical Society
twice the volume of water for personal hygiene as compared to
those without on-plot access to piped water.9
Previous research on water access and health has explored
the impact of households gaining access to on-plot piped water
connections, but little is known regarding the extent to which
water fetching affects child health.10 One community-specific
study in Ethiopia found that installation of village taps reduced
time spent fetching water and child mortality, yet child
malnutrition increased.11 A recent systematic review of studies
investigating the relationship between distance from the home
to water source and diarrheal disease identified only six
studies, four of which did not adjust for possible confounding
variables. The review authors were unable to calculate a
quantitative relationship between distance and diarrheal risk
and concluded that more research is needed on this topic.6,12
To address this knowledge gap, this work investigates the
association between access to water and under-five child
morbidity and mortality. For the purpose of this analysis, water
access is defined by reported one-way walk time to a
household’s main drinking water source. Child health outcomes
include prevalence of diarrhea, cough, and fever; anthropometrics (height and weight); as well as childhood mortality.
Geographic variation in freshwater availability is employed as
Received:
Revised:
Accepted:
Published:
2391
September 10, 2011
December 28, 2011
January 12, 2012
January 12, 2012
dx.doi.org/10.1021/es203177v | Environ. Sci. Technol. 2012, 46, 2391−2397
Environmental Science & Technology
Article
interview. The DHS survey does not provide respondents with
a specific definition for diarrhea. Height/length and weight
were also measured for under-five children and then converted
to height-for-age z-scores (HAZ), weight-for-age z-scores
(WAZ), and weight-for-height z scores (WHZ). Z-scores for
each child were generated using World Health Organization
(WHO) child growth standards as a reference, and indicate the
number of standard deviations a child is above or below the
normal value given his/her age.14
Respondents in the DHS survey were asked how many
minutes it takes to walk one way to their households’ main
drinking water source. A value of 0 min was assigned for households with a water source located on their premises. Households
reported the type of source from which they collected drinking
water; these source water types were grouped into the following
categories: piped, well, spring, surface, rainwater, vendor, and
other.
Freshwater Availability. The freshwater availability data
set used in this analysis was generated by the University of New
Hampshire (UNH)/Global Runoff Data Centre (GRDC)
Composite runoff Fields (V1.0) climate driven water balance
model. 15 The model provides estimates of freshwater
availability (i.e., runoff) at the 0.5° latitude by 0.5° longitude
grid-based level. The area of each grid cell is approximately
55 km by 55 km (precise area varies with proximity to the
equator). The water balance model takes into account precipitation, air temperature, soil type, and land cover to estimate
a long-term average of total freshwater availability (mm/year);
it is calibrated using river gauge discharge data. Figure 1 shows
this gridded freshwater data set overlaid with GPS coordinates
of household clusters included in the DHS data set.
Estimating Strategy. This analysis employs an instrumental variable (IV) approach, a model estimation strategy
designed to estimate causal effects with nonexperimental
data.13 An important concern when analyzing such data is
that the independent variable of interest (water access) shares a
common cause with the outcome of interest (child health), and
the observed association is a result of confounding. An example
of a common cause that could confound the association
between water access and child health is the wealth of the
household. This issue can be addressed by identifying a source
of exogenous variation in distance to water source that is not
caused by endogenous variables such as household income.
In sub-Saharan Africa, where water delivery infrastructure is
limited, it seems reasonable to consider geographic differences
in freshwater availability as a source of exogenous variation in
household distance to water source. The IV modeling approach
employs a first-stage equation to model walk time to water
source as a function of freshwater runoff available to each
household. The predicted values of walk times are then used in
a second-stage equation modeling health outcomes as a
function of water collection time at the household level.
Model Specification. A two-stage model was estimated as
follows:
an instrumental variable for walk time to water source. The
instrumental variable approach is one strategy for estimating
causal effects within nonexperimental settings; it is used here to
estimate the effect of water access on child health outcomes.13
All data are drawn from sub-Saharan Africa, where 84% of the
population lack access to piped water into the dwelling or
yard.1
■
METHODS
Household Survey Data. Child health outcomes, water
access, and other household-level data were obtained from the
most recent Demographic and Health Survey (DHS) data set
from all countries in sub-Saharan Africa that included a walktime to water source variable, child anthropometrics, child
diarrhea prevalence, and GPS coordinates of household
clusters (Table 1). Surveys included in this analysis were
Table 1. Household Survey Observations (n = 199 067) from
26 Countries in Sub-Saharan Africaa
walk time (min)
country
N
Burkina Faso
10 645
Benin
5349
Congo
8992
Central African Rep.
2816
Cote d’Ivoire
1992
Cameroon
8125
Ethiopia
9861
Ghana
2992
Guinea
6366
Kenya
6079
Liberia
5799
Lesotho
3697
Madagascar
12 448
Mali
13 097
Malawi
10 929
Nigeria
28 647
Niger
4798
Namibia
5168
Rwanda
8649
Sierra Leone
5631
Senegal
10 944
Swaziland
2812
Tanzania
3215
Uganda
8369
Zambia
6401
Zimbabwe
5246
total
199 067
year
2003
2001
2007
1994−1995
1998−1999
2004
2005
2007
2005
2008
2007
2004
2009
2001
2004
2008
2006
2006−2007
2005
2008
2005
2007
1999
2006
2007
2005−2006
mean SD med
20
16
31
21
13
20
46
18
15
28
11
26
17
12
32
19
26
14
33
20
19
21
34
63
18
23
24
26
22
33
24
22
29
60
25
19
43
12
30
41
19
33
29
37
28
31
25
45
34
57
62
21
35
37
10
10
20
12
5
10
30
10
10
15
7
15
10
6
20
10
15
2
25
10
7
10
15
45
10
10
10
% on
plot
21
37
9
7
22
14
7
12
12
27
8
10
11
29
1
24
3
46
3
12
40
35
18
4
17
37
18
a
Number of observations are shown from each country, as well as
year(s) of survey, mean one-way walk time to water source in minutes
(min), standard deviation of mean walk time (SD), median walk time
(med), and percent of households with water source located on the
premises (% on plot).
TIMEi = β0 + β1Freshwateri + ΣkβkSourceik + β3Agei
conducted between the years of 1994 and 2009. The data set
consists of all living children under the age of 5 years at the
time of interview, as well as deceased children who would
have been less than 5 years old if still alive at the time of
interview (N = 199 067).
Questions in the DHS survey record whether each child
under the age of five in a participating household has been ill
with fever, cough, or diarrhea in the two weeks prior to
+ β4Genderi + β5Urbani + εi
^
Health i = α 0 + α1TIMEi + ΣkβkSourceik + α3Agei
+ α4Genderi + α5Urbani + εi
2392
dx.doi.org/10.1021/es203177v | Environ. Sci. Technol. 2012, 46, 2391−2397
Environmental Science & Technology
Article
Figure 1. Overlay of geo-referenced household clusters and gridded freshwater runoff in sub-Saharan Africa. Geo-referenced points of household
clusters provided by Demographic and Health Surveys. Gridded freshwater runoff (mm/yr) generated by UNH-GRDC Composite Runoff Fields
V1.0.
higher microbial contamination that can cause diarrheal illness
as compared to piped water or deep groundwater.16 To address
this issue, all models control for household water source type as
a proxy for water quality; surface water is the reference category
in all models.
One subgroup analysis was performed to compare the estimated
effect of water access on child health among households with and
without access to sanitation. The sample was stratified into two
groups: (1) households reporting use of any type of sanitation
facility (improved or unimproved); and (2) households practicing
open defecation.
To increase precision, all models include the gender and age
of each child (in months), and whether the household is
located in an urban or rural area. DHS-generated population
weights were used to obtain representative results. Standard
errors were adjusted to account for within-cluster correlation at
the geo-referenced cluster level. STATA 11 (StataCorp LP,
College Station, TX) was used for all data analysis.
where TIMEi is the one-way walk time to the main drinking
water source reported by household i and Freshwateri is the
quantity of freshwater available in the immediate area of
household i (Figure 1). ΣkSourceik is a series of dummy
variables specifying the type of water source that household i
̂ Ei)
accesses. The predicted values of walk time to source (TIM
generated by the first stage were used in the second-stage
health outcome model. Healthi outcomes modeled include
2-week prevalence of diarrhea, fever, and cough; WAZ, HAZ,
and WHZ; and child mortality. Binary health outcomes
(diarrhea, cough, fever, mortality) were estimated with probit
models, whereas continuous outcomes (WAZ, HAZ, WHZ)
were estimated using ordinary least-squares linear regression.
The parameter of interest in all health outcome models is α1,
the estimated average effect of a one-minute change in walk
time to water source for household i.
The two-stage estimation strategy was employed because
TIMEi is likely to be correlated with εi. The control variables
ΣkSourceik were included to address the possibility that
freshwater availability may affect the type of water source
accessed by households, which in turn may affect the quality
(and safety) of water. For example, areas with abundant water
resources may have a high prevalence of shallow wells or
surface water sources. Such sources generally deliver water with
■
RESULTS
Household Characteristics. Respondents include mothers
with a mean age of 29 (SD 7) and a mean 3.4 years of
education (SD 4). Approximately one-quarter of the total childlevel observations were collected from urban households, and
2393
dx.doi.org/10.1021/es203177v | Environ. Sci. Technol. 2012, 46, 2391−2397
Environmental Science & Technology
Article
the rest from households classified as rural. Almost one-fifth
(19%) of included households had access to electricity; 31%
owned a bicycle, 7% owned a fridge, and 60% had an earth floor
in their home. Access to sanitation facilities was varied: 7% used
flush toilets, 12% used ventilated improved pit (VIP) latrines,
43% used non-VIP latrines, 36% had no facility, and 1% used
other types of facilities (e.g., bucket latrine).
One quarter (25%) of households identified a piped source
(private or public tap) as their principal water source. Another
48% reported using protected or unprotected wells, whereas
10% used springs and 16% used surface water. Fewer than 1%
of surveyed households reported using rainwater (0.6%),
vendors (0.7%), bottled water (0.2%), or an “other” (0.6%)
type of principal water source. The mean one-way walk time to
a household’s main water source was 23 min (SD 37, median
10 min); 18% (N = 34 948) of households reported a water
source located on the premises.
Child Health. A total of 17% of children were reported to
have experienced diarrhea in the two weeks prior to interview.
Respondents reported fever for 27% of children and cough for
25% in the previous two weeks. The mean WAZ was −1.04
(N = 110 455, SD = 1.32), with 22% of children classified as
underweight by World Health Organization (WHO) standards
(WAZ < −2.0). The mean HAZ for children was −1.56 (N =
115 316, SD = 1.8), with 41% classified as stunted (HAZ <
−2.0). A total of 11% of children measured were identified as
Wasted (WHZ < −2.0), with a mean WHZ of −0.21 (N =
110455, SD = 1.49). Under-five child mortality was 9.9%.
First-Stage Model. Average freshwater runoff was 344
mm/yr (median 178, SD = 470, N = 191 333) and mean
monthly precipitation was 113 mm (median 95, SD = 103, N =
196 711). The results of the first-stage regression of one-way
walk time as a function of freshwater runoff and other control
variables are presented in Table S1 of the Supporting
Information (SI). As necessary for employing the instrumental
variable estimation approach, freshwater runoff is strongly
correlated to water fetching time in the expected direction
(t-statistic = −53.2, p < 0.001): as runoff increases, walk time
to source decreases. Diagnostic tests indicate that walk time is
sufficiently identified by freshwater availability (see SI for
details).17−20
Second-Stage Model. The average marginal effects (ME)
of walk time to water source on child health outcomes are
presented in Table 2. A 15-min decrease in one-way walk time
to water source is associated with a 7.1 percentage point
reduction in diarrhea (2-week prevalence), equivalent to a 42%
relative reduction from the mean prevalence in the study
population. The same 15-min decrease in one-way walk time is
also associated with a 4.5 percentage point reduction (17%
relative reduction) in fever, and with a 3.9 percentage point
reduction (15% relative reduction) in cough (all symptoms
refer to 2-week prevalence). The same decrease in one-way
walk time is associated with an increase of 0.6 in WAZ
(improved weight-for-age), an increase of 0.3 in HAZ (reduced
stunting) and a 0.5 increase in WHZ (reduced wasting). Underfive child mortality is lower by 1.1%age points (11% relative
reduction) with a 15-min decrease in walk time (Figure 2).
The estimated effect of walk time to source on child health
was much greater among households with access to sanitation
facilities as compared to those households practicing open
defecation (Table S2 of the SI).
Table 2. Average Marginal Effect (ME) on Child Health
Outcomes of Reduced One-Way Walk Time to Water
Sourcea
health
outcome
ME (1-min
decrease)
SE
N
Δ associated with 15-min
decrease (95% CI)
diarrheab
feverb
coughb
WAZ
HAZ
WHZ
mortalityb
−0.0047c
−0.0030c
−0.0026c
0.0371c
0.0176c
0.0337c
−0.0008c
0.0003
0.0004
0.0004
0.0032
0.0025
0.0031
0.0002
160 734
160 539
160 579
102 013
106 573
102 013
183 267
−7.1 (6.1−8.0) pp
−4.5 (3.2−5.8) pp
−3.9 (2.6−5.2) pp
+0.6 (0.5−0.7)
+0.3 (0.2−0.3)
+0.5 (0.4−0.6)
−1.1 (0.5−1.7) pp
a
Walk time to water source is instrumented by freshwater availability
(runoff); all models include water source type, urban or rural region,
child gender, and child age in months. SE is standard error of the
marginal effect; pp stands for percentage points; WAZ is weight-forage; HAZ is height-for-age; WHZ is weight-for-height. bProbit model.
c
P < 0.001.
■
DISCUSSION
This analysis of nationally representative data from 26 countries
in sub-Saharan Africa identifies walk time to water source as an
important determinant of child health. Even a 5-min decrease in
walk time to source is associated with an average 14% relative
reduction in 2-week diarrhea and an average increase of 0.2 in
the WAZ score for an under-5 child (Figure 2). A 15-min
decrease in one-way walk time to water source is associated
with an average relative reduction in risk of diarrhea by 41%,
which is comparable to the range of diarrhea risk reduction
found in meta-analyses of water quality, handwashing, and
sanitation interventions.21 These findings suggest that investments that reduce the time costs of water fetching time may
provide health benefits on par with those achieved by water
disinfection and hygiene promotion programs. Further research
should evaluate the health impact of technologies and strategies
that reduce water fetching time for households, such as carts,
bicycles, flexible hoses, and self-supply options such as
rainwater collection systems and hand-dug wells.22,23
Walk time to source was found to have a greater impact on
child health among households with access to sanitation (Table S2
of the SI). This result is consistent with research by Esrey et al.
(1996) in sub-Saharan Africa, in which health benefits associated
with gaining access to water supply at the household were
only observed when improved sanitation was also present.24
One possible explanation for this finding is that close proximity
to a water source may allow for more hygienic use of a sanitation facility (e.g., pour flushing, cleaning, postdefecation handwashing), which results in significant health benefits. Evidence
for such complementary health effect of latrine access and
increased water usage has been documented previously in
Lesotho.25
Several causal mechanisms underlying the observed relationship between time to water source and child health are plausible
(Figure S1 of the SI). Further walk times could restrict the
volume of water collected by households, thus impeding the use
of water for sanitation services and hygiene behaviors, such
as handwashing, that are known to reduce diarrheal and
respiratory illness.26,27 Longer storage time of drinking water
may offer increased opportunities for microbial contamination
though repeated extraction with dirty utensils and hands.28
Mothers that must spend significant time fetching water may
have less time to care for their children in ways that affect
health, such as seeking health care services when a child is ill.29
2394
dx.doi.org/10.1021/es203177v | Environ. Sci. Technol. 2012, 46, 2391−2397
Environmental Science & Technology
Article
Figure 2. Predicted probability of child health outcomes versus one-way walk time to water source. (A) Predicted probability of acute illness (2-week
period prevalence of diarrhea, fever, and cough) and mortality versus one-way walk time to water source (reported in minutes). Error bars display the
standard error of the predicted probability. (B) Predicted anthropometric Z-scores versus one-way walk time to water source. WAZ is weight-for-age,
HAZ is height-for-age, and WHZ is weight-for-height. Standard errors of the predicted Z-scores are displayed as error bars.
A distant water source may also preclude households from
using water for on-plot gardening, which can improve family
nutrition, or from using water for additional productive activities
(e.g., brick-making) that can augment income.30 Finally, time
spent collecting water reduces the time available to engage in
other income generating activities that can contribute to
household consumption. Additional research is needed to assess
the relative contribution of each of these causal pathways to child
health.
A fundamental limitation of this nonexperimental analysis is
the inability to directly assess whether freshwater runoff affects
child health only through its influence on domestic water
supply access. To explore this issue, several robustness checks
were carried out. For example, freshwater runoff in regions with
high precipitation levels may affect health through wealth
earned by differential income generating opportunities, such as
agriculture. In order to assess household wealth as a potential
confounding variable, all models were re-estimated after
including a household wealth index variable.31 Controlling for
this wealth index did not reduce the magnitude or significance
of any estimate. Similarly, estimates were not significantly
changed after controlling for educational attainment of the
respondent, nor of household participation in agriculture
(Table S3 of the SI).
For the subset of households from which dietary data were
collected (N = 76 038), a dietary diversity index (DDI) was
created as a proxy for nutritional intake of children and
included in re-estimated models of illness and anthropometrics
(see SI for details).32 With the exception of fever, all
associations between water access and health were preserved
in models that included the DDI, although the magnitudes of
the effect estimates on the anthropometric outcomes were
reduced (Table S3, column D of the SI). This reduction in effect
size would be expected if convenient access to water allows
households to irrigate, grow, and consume nutrient-rich foods.
Notably, access to irrigation increases the diversity of crops that
can be grown and has been shown to improve nutritional intake
among households in sub-Saharan Africa.33
There is evidence that precipitation and other climatic factors
can affect transmission of diarrheal pathogens.34 To control for
potential seasonality effects on health that might be related to
runoff, IV models were re-estimated after including the mean
monthly precipitation (mm/month) during the month in which
a survey was conducted.35 Including monthly precipitation did
not change the association between water access and diarrhea,
WAZ, HAZ, or WHZ; however, walk time associations with
fever and cough were no longer significant (Table S3 of the SI).
The magnitude of effects presented here must be considered
alongside the possibility for both upward and downward bias.
Walk time to source was drawn from self-reported data; the
accuracy of the results could thus be affected if people
systematically over or underestimate walk time. The association
between freshwater availability and walk time may also be
affected by the lack of temporal freshwater availability data. In
addition, the results of the presented alternative model
specifications do not eliminate the possibility that freshwater
availability is associated with child health outcomes through
pathways other than walk time to water source. At the same
time, the associations found between walking farther to water
source and higher probability of diarrhea, lower anthropometric
Z-scores, and increased risk of mortality are robust even after
controlling for wealth, seasonality, educational attainment, and
agricultural employment (Table S3 of the SI).
It should be noted that the relationship between time to
water source and volume of water collected may well be
nonlinear; however, there are limited recent data on this
association. A review conducted in 1987 of studies on water use
and time costs of collection found the following: (1) water
volumes used by households decrease sharply when the travel
time to source exceeds 5 min; (2) volumes remain fairly
constant during a collection time range of 5−30 min; and (3)
volumes decrease further when collection time exceeds 30 min.7
This pattern implies that there could be differentially large
health benefits for households with access to a water source in
the home or yard. In this analysis, when households with their
own water source are excluded, the magnitude of the effect of
walk time to source on health outcomes decreases only slightly
2395
dx.doi.org/10.1021/es203177v | Environ. Sci. Technol. 2012, 46, 2391−2397
Environmental Science & Technology
■
(Table S3, column A of the SI). This result, however, may be
affected by a shift in the first-stage regression, as households
with their own source tend to live in regions with relatively
lower runoff than those walking short distances to their water
source (see SI for additional discussion).
One implication of these findings is that water infrastructure
improvements that do not deliver water close to the home may
be unlikely to have major impacts on infectious disease among
children under five. This is particularly concerning for subSaharan Africa, where gains in access to improved water supply
over the past two decades have almost exclusively taken the
form of non-networked (off-plot) services. During the period
1990−2008, access to piped water into the dwelling, lot or yard
decreased in urban areas in sub-Saharan Africa (from 43% to
35%) and remained near constant (at 4−5%) in rural areas.1
The WHO/UNICEF Joint Monitoring Program (JMP) does
not consider proximity to water source when monitoring global
progress toward Target 10 of the Millennium Development
Goals, “to reduce by half the proportion of people without
access to safe drinking water and sanitation by 2015.”36
Although it is unofficially recognized that a water source should
be located within 1 km of the home, only water source type is
actually used by JMP to assess access to safe water (e.g., piped
water, borehole, or pump).36,37 This work suggests that walk
time to water source should be given more emphasis as an
essential indicator for monitoring global access to safe water.
Finally, when considered alongside predicted decreases in
freshwater availability for sub-Saharan Africa,38 these findings
suggest that the morbidity and mortality burden of water
fetching on African children will likely increase over the next
several decades. Recent research indicates that, by the year
2050, more than 1 billion people will live in areas with serious
water shortage, defined as having access to less than 100 L per
person per day.39 Wealthy countries can mitigate this water
security threat by building water delivery infrastructure and
treatment systems, but low-income countries struggle to make
these types of investments.40
■
REFERENCES
(1) WHO-UNICEF, Progress on Sanitation and Drinking-Water:
2010 Update. Joint Monitoring Programme for Water Supply and
Sanitation 2010.
(2) Sorenson, S. B.; Morssink, C.; Campos, P. A., Safe access to safe
water in low income countries: Water fetching in current times. Soc.
Sci. Med. 2011.
(3) Montgomery, M. A.; Elimelech, M., Water and sanitation in
developing countries: Including health in the equation. Environ. Sci.
Technol. 2007.
(4) Hunter, P. R. Household water treatment in developing
countries: Comparing different intervention types using metaregression. Environ. Sci. Technol. 2009.
(5) Schmidt, W. P.; Cairncross, S. Household water treatment in
poor populations: Is there enough evidence for scaling up now?
Environ. Sci. Technol. 2009, 43 (4), 986−992.
(6) Wang, X.; Hunter, P. R. A systematic review and meta-analysis of
the association between self-reported diarrheal disease and distance
from home to water source. Am. J. Trop. Med. Hyg. 2010, 83 (3), 582.
(7) Cairncross, S. The benefits of water supply. Dev. World Water
1987, 11.
(8) Curtis, V.; Kanki, B.; Mertens, T.; Traore, E.; Diallo, I.; Tall, F.;
Cousens, S. Potties, pits and pipes: Explaining hygiene behaviour in
Burkina Faso. Soc. Sci. Med. 1995, 41 (3), 383−393.
(9) Tumwine, J. K.; Katui-Katua, M.; Munguti, K. K. Drawers of
Water II: 30 Years of Change in Domestic Use & Enviromental Health in
East Africa. International Institute for Environment and Development:
London/Edinburgh, 2002.
(10) Zwane, A. P.; Kremer, M. What works in fighting diarrheal
diseases in developing countries? A critical review. World Bank Res.
Obs. 2007, 22 (1), 1−24.
(11) Gibson, M. A.; Mace, R. An energy-saving development
initiative increases birth rate and childhood malnutrition in rural
Ethiopia. PLoS Med. 2006, 3 (4), e87.
(12) Subaiya, S.; Cairncross, S. Response to Wang and Hunter: A
systematic review and meta-analysis of the association between selfreported diarrheal disease and distance from home to water source.
Am. J. Trop. Med. Hyg. 2011, 84 (3), 504.
(13) Angrist, J.; Krueger, A. B. Instrumental variables and the search
for identification: From supply and demand to natural experiments.
J. Econ. Perspect. 2001, 15 (4, Fall), 69−85.
(14) Onis, M. WHO Multicenter Growth Reference Study Group.
WHO Child Growth Standards based on length/height, weight and
age. Acta Paediatrica Suppl. 2006, 95 (450), 76−85.
(15) Fekete, B. M.; Vörösmarty, C. J.; Grabs, W. High-resolution
fields of global runoff combining observed river discharge and
simulated water balances. Global Biogeochem. Cycles 2002, 16 (3),
1042.
(16) Wright, J.; Gundry, S.; Conroy, R. Household drinking water in
developing countries: a systematic review of microbiological
contamination between source and point-of-use. Trop. Med. Int.
Health 2004, 9 (1), 106.
(17) Staiger, D.; Stock, J. H. Instrumental variables regression with
weak instruments. Econometrica 1997, 65, 557−586.
(18) Stock, J. H.; Yogo, M. Testing for weak instruments in linear IV
regression. In Identification and Inference for Econometric Models: Essays
in Honor of Thomas Rothenberg; Andrews, D. W. K., Stock, J. H., Eds.;
2005, 80−108.
(19) Angrist, J. D.; Pischke, J. S. Mostly Harmless Econometrics: An
Empiricist’s Companion; Princeton University Press: Princeton, NJ,
2008.
(20) Kleibergen, F.; Paap, R. Generalized reduced rank tests using the
singular value decomposition. J. Econometrics 2006, 133 (1),
97−126.
(21) Fewtrell, L.; Kaufmann, R. B.; Kay, D.; Enanoria, W.; Haller, L.;
Colford, J. M. Water, sanitation, and hygiene interventions to reduce
diarrhoea in less developed countries: a systematic review and metaanalysis. Lancet Infect. Dis. 2005, 5 (1), 42−52.
ASSOCIATED CONTENT
S Supporting Information
*
See Supporting Information for additional details on methods,
further discussion of results, Tables S1−S3, and Figure S1. This
information is available free of charge via the Internet at http://
pubs.acs.org/
■
Article
AUTHOR INFORMATION
Corresponding Author
*Phone: (510) 410-2666; e-mail: [email protected]
(A.J.P.). Phone: (650) 725-9170; e-mail: jennadavis@stanford.
edu(J.D.).
■
ACKNOWLEDGMENTS
A.J.P. was funded by the Stanford Interdisciplinary Graduate
Fellowship while conducting this analysis. We thank Grant
Miller for extensive advice on analysis methods, Teizeen
Mohamedali for excellent research assistance, Christine Moe for
analysis suggestions, and Ben Arnold and Ayse Ercumen for
helpful comments on an earlier draft. We also acknowledge
three anonymous reviewers for suggestions to improve the
manuscript.
2396
dx.doi.org/10.1021/es203177v | Environ. Sci. Technol. 2012, 46, 2391−2397
Environmental Science & Technology
Article
(22) Maiga, H.; Maiga, B.; Sutton, S. Self-supply in Mali. Waterlines
2006, 25 (1), 13−14.
(23) Mwenge Kahinda, J.; Taigbenu, A. E.; Boroto, J. R. Domestic
rainwater harvesting to improve water supply in rural South Africa.
Phys. Chem. Earth, Parts A/B/C 2007, 32 (15−18), 1050−1057.
(24) Esrey, S. A.; Water, Waste; Well-Being, A Multicountry Study.
Am. J. Epidemiol. 1996, 143 (6), 608−623.
(25) Esrey, S. A.; Habicht, J. P.; Casella, G. The complementary
effect of latrines and increased water usage on the growth of infants in
rural Lesotho. Am. J. Epidemiol. 1992, 135 (6), 659−666.
(26) Aiello, A. E.; Coulborn, R. M.; Perez, V.; Larson, E. L. Effect of
hand hygiene on infectious disease risk in the community setting:
A meta-analysis. Am. J. Public Health 2008, 98 (8), 1372−1381.
(27) Green, S. T.; Small, M. J.; Casman, E. A. Determinants of
national diarrheal disease burden. Environ. Sci. Technol. 2009, 43 (4),
993−999.
(28) Trevett, A. F.; Carter, R. C.; Tyrrel, S. F. Water quality
deterioration: A study of household drinking water quality in rural
Honduras. Int. J. Environ. Health Res. 2004, 14 (4), 273−283.
(29) Miller, G.; Urdinola, B. P. Cyclicality, mortality, and the value of
time: The case of coffee price fluctuations and child survival in
Colombia. J. Pol. Econ. 2010, 118 (1), 113−155.
(30) Moriarty, P.; Butterworth, J.; Martin, A.; Morris, M.; Nicol, A.;
Cousins, T. In Productive Use of Domestic Water Supplies: How Water
Supplies Can Play a Wider Role in Livelihood Improvement and Poverty
Reduction; IRC: Austin, 2003.
(31) Rutstein, S. O.; Johnson, K. The DHS Wealth Index (DHS
Comparative Reports No. 6); ORC Macro: Calverton, Md, 2004.
(32) Arimond, M.; Ruel, M. T. Dietary diversity is associated with
child nutritional status: Evidence from 11 demographic and health
surveys. J. Nutr. 2004, 134 (10), 2579−2585.
(33) Burney, J.; Woltering, L.; Burke, M.; Naylor, R.; Pasternak, D.
Solar-powered drip irrigation enhances food security in the Sudano,
̀
Proc. Natl. Acad. Sci. 2010, 107 (5), 1848.
Ä iSahel.
(34) Levy, K.; Hubbard, A. E.; Eisenberg, J. N. S. Seasonality of
rotavirus disease in the tropics: A systematic review and meta-analysis.
Int. J. Epidemiol. 2009, 38 (6), 1487.
(35) Hijmans, R. J.; Cameron, S. E.; Parra, J. L.; Jones, P. G.; Jarvis,
A. Very high resolution interpolated climate surfaces for global land
areas. Int. J. Climatol. 2005, 25 (15), 1965−1978.
(36) UNDP Indicators for monitoring the Millenium Development
Goals: Definition, Rationale, Concepts and Sources; United Nations
Publications: New York, 2003; Vol. ST/ESA/STAT/SER.F/95.
(37) Hunter, P. R.; MacDonald, A. M.; Carter, R. C. Water supply
and health. PLoS Med. 2011, 7 (11), e1000361.
(38) de Wit, M.; Stankiewicz, J. Changes in surface water supply
across Africa with predicted climate change. Science 2006, 311 (5769),
1917−1921.
(39) McDonald, R. I.; Green, P.; Balk, D.; Fekete, B. M.; Revenga, C.;
Todd, M.; Montgomery, M. Urban growth, climate change, and
freshwater availability. Proc. Natl. Acad. Sci. 2011, 108 (15), 6312−
6317.
(40) Vorosmarty, C. J.; McIntyre, P. B.; Gessner, M. O.; Dudgeon,
D.; Prusevich, A.; Green, P.; Glidden, S.; Bunn, S. E.; Sullivan, C. A.;
Liermann, C. R.; Davies, P. M. Global threats to human water security
and river biodiversity. Nature 2010, 467 (7315), 555−561.
2397
dx.doi.org/10.1021/es203177v | Environ. Sci. Technol. 2012, 46, 2391−2397