WPS8729


Policy Research Working Paper                            8729




         Why Do So Many Water Points Fail
                  in Tanzania?
         An Empirical Analysis of Contributing Factors

                                 George Joseph
                              Luis Alberto Andres
                              Gnanaraj Chellaraj
                        Jonathan Grabinsky Zabludovsky
                          Sophie Charlotte Emi Ayling
                                 Yi Rong Hoo




Water Global Practice
February 2019
Policy Research Working Paper 8729


  Abstract
 According to the 2015 Tanzania Water Point Mapping data,                           than those managed by private operators or water authority.
 about 29 percent of all water points are non-functional, out                       Factors that cannot be modified such as hydrogeological
 of which 20 percent failed within the first year. This paper                       factors play a major role in determining water points failure
 analyzes the various factors which impact water point failure                      during the first year after installation. However, manage-
 and measures the relative contributions of these determi-                          ment type as well as the type of pump and technology
 nants. The results indicate that water points managed by                           matter considerably more in the short and medium term.
 village committees had a much higher likelihood of failure




 This paper is a product of the Water Global Practice. It is part of a larger effort by the World Bank to provide open
 access to its research and make a contribution to development policy discussions around the world. Policy Research
 Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at
 gjoseph@worldbank.org




         The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
         issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
         names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
         of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
         its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                       Produced by the Research Support Team
                        Why Do So Many Water Points Fail in Tanzania?
                          An Empirical Analysis of Contributing Factors1



                            George Joseph, Luis Alberto Andres, Gnanaraj Chellaraj
                             Jonathan Grabinsky Zabludovsky, Sophie Charlotte Emi
                                                  Ayling, and Yi Rong Hoo




JEL Classification: D02, O18, Q25


Key Words: Water Points, Tanzania, Functionality




1
   George  Joseph:  World  Bank  Water  Global  Practice,  gjoseph@worldbank.org  [Corresponding  author]  .Luis  Alberto  Andres: 
Global  Water  Practice,  World  Bank,  landres@worldbank.org.  Gnanaraj  Chellaraj:  World  Bank  Water  Global  Practice, 
gchellaraj@worldbank.org.  Jonathan  Grabinsky  Zabludovsky:  World  Bank  Water  Global  Practice,  jgrabinsky@worldbank.org. 
Sophie  Charlotte  Emi  Ayling:  Global  Water  Practice,  World  Bank,  sayling@worldbank.org.  Yi  Rong  Ho:  Water  Global  Practice, 
World Bank, yhoo@worldbank.org.     
This work was made possible by the Swedish International Development Cooperation Agency and was a 
background paper to the WASH Poverty Diagnostics in Tanzania 
The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily 
represent the view of the World Bank, its executive directors, or the countries they represent. The findings, interpretations, and 
any remaining errors in this paper are entirely those of the authors. 
    1. Background

        Despite the significant economic progress that Tanzania has experienced over the years and
considerable investments in water supply infrastructure through both government and donor funding, a
significant proportion of its population remains without proper access to improved drinking water. While
one of the agreed Millennium Development Goals (MDGs) is to halve the proportion of people that do
not have access to water services by 2015, Tanzania only increased its access to improved drinking water
from 54 percent to 56 percent (JMP, 2015). The country now faces an even more difficult task of meeting
the Sustainable Development Goals (SDGs) to provide universal coverage of safe water by 2030.
        Despite its efforts through the two phases of the Water Sector Development Program (WSDP),
one persistent problem that has adversely affected the country’s effort in increasing access to improved
water services is the prevailing high levels of non-functionality or failures of its current water
infrastructures and in particular, water points. Technology constraints also play an important role (de
Bont et al., 2019). This situation is not unique to Tanzania. Water point failures have been documented by
a number of studies across the Sub-Saharan African region. In countries such as Nigeria (Andres et al.,
2018) and Ghana (Fisher et al., 2015), water points tend to fail frequently, particularly during the early
years after construction. The situation is similar in Mozambique (Jansz, 2011), Uganda (Nekesa &
Kulanyi, 2012) and Ethiopia (Alexander et al., 2015; Schweitzer et al., 2015). However, evidence
indicates that the problem of water point failures may be relatively more serious in Tanzania with some
estimates putting the figure as high as 44 percent (Banks & Furey, 2016)
        The importance of access to water supply cannot be overstated. Apart from meeting the human
necessity of water, it has immediate health impacts (Hyland and Russ, 2019) which in turn affects other
developmental outcomes including education attainment and even long run poverty outcomes (Zhang and
Xu, 2016; Mangyo, 2008; Alderman et al., 2001). Evidence indicates that lack of functioning
infrastructure will undermine economic growth (Agenor, 2010; Barbier, 2004). Moreover, evidence also
indicates that infrastructure—including water and sanitation—is likely to offset moderate macroeconomic
shortcomings during the initial stages of economic development (Moller and Wacker, 2017; Gibson and
Rioja, 2017; Amann et al., 2016).
        In this paper, we utilize information from the Tanzania Water Point Mapping (WPM) compiled in
November 2016 to identify the reasons for the failure of water points in the country. Using a variety of
statistical methods, we seek to understand the impact of the following factors on water points failure: (i)
the age, (ii) technology type; (iii) geographic patterns; (iv) hydrological characteristics; (v) type of
promoter; and (vi) day to day management type of these points. The results indicate that water points

                                                    2 
managed by village committees had a much higher likelihood of failure than those managed by private
operators or water authority. Factors that cannot be modified such as hydrogeological factors play a major
role in determining water points failure during the first year after installation. However, management
type as well as the type of pump and technology matter considerably more in the short and medium term.
These findings highlight the importance of utilizing information on the hydrological characteristics in the
design, construction and maintenance of water points. On the other hand, the technology of water points
is a more important factor in explaining failure over time, controlling for other factors.


2.        Water in the Tanzania Context

          Tanzania is the fourth most populous country in Sub-Saharan Africa with more than 55.6 million
people2. According to UN Water (2013), Tanzania is ‘economically water scarce’ and in 2012, its annual
renewable water resource was estimated to be around 89km3, about 2000m3 of per capita availability. This
reportedly fell to 1,952m3 in 2014, and is projected to decrease to 873m3 by 2035. In addition to its
rapidly expanding population, there are also other significant pressures on Tanzania’s water resources
stemming mostly from its economic activities (Miller and Doyle, 2014; Lein, 2004). As in most other
countries, agricultural sector for example, by far consumes a large portion of its withdrawn water
(roughly 90 percent of total water withdrawals (UN-Water, 2013)3.
          Despite its economic progress over the years, its water infrastructure has failed to keep pace with
population increase. As a result, it is unable to provide adequate service coverage to its population4. This
is reflected in the mere 2 percentage points increase in the improved water coverage between 1990 and
2015. In fact, in urban areas, due to population growth and migration, improved water coverage actually
fell from 92 percent in 1990 to 77 percent by 2015 (JMP, 2015). In rural areas on the other hand,
improved water coverage is estimated to be at 45 percent in 2015 (JMP, 2015).
          To address the shortfalls in water supply infrastructure and to improve water resource
management throughout the country, in the wake of the Millennium Development Goals (MDG) the
Tanzania government developed the Water Sector Development Program (WSDP) which is set to run

2    Data are from World Development Indicators (database). 2016. World Bank, Washington, DC.
http://databank.worldbank.org/data/home.aspx
3 Other than that, the country’s climate and hence rainfall pattern also plays a significant role in influencing the supply of water

resources. The country experiences a monsoon-type climate with annual rainfall ranging from a high of >1000mm in the coastal
and highland areas to a low of around 600mm in the dryer central and northern areas. In the latter, the dry season can last up to 7
months, and rivers tend to run dry (Basalirwa et al., 1999). The El Niño/La Niña South Oscillation (ENSO) phenomenon can also
result in substantial impacts on intra-seasonal rainfall variability and inter-annual rainfall variability is a key challenge resulting
in droughts and floods (White and Tourre, 2007; Nicholson & Kim, 1997). Projections of Tanzania’s future rainfall patterns
based on climate change vary widely, which results in a challenging environment for planning and investment (Basalirwa et al.,
1999). There is a general consensus that total rainfall is set to increase in parts of the country (Cioffi et al., 2016), although the
already dry central area is potentially likely to experience a decrease in annual precipitation levels. (Cioffi et al., 2016)
4 Reported in UNPD, 2014 from JMP 2015




                                                                  3 
between 2006 and 2025, with the support of development partners (Carlitz 2017). The WSDP comprises
of three smaller programs: 1) water resources management; 2) Rural Water Supply and Sanitation
Program (RWSSP); and 3) urban water supply and sewerage. At the time of the initiation of RWSSP,
water coverage in rural areas in Tanzania was estimated to be around 53 percent based on its 2002
Population and Housing Census (Ministry of Water, 2006)5. The RWSSP aimed at improving water
coverage to its rural population to 1) at least 65 percent by 2010; 2) 74 percent by mid-2015; and 3) at
least 90 percent by 20256. The program is also largely funded by the World Bank and other international
donors and is one of the biggest in the African region (Jimenez and Perez-Foguet, 2010; Gine and Perez-
Foguet, 2008; World Bank, 2008).
              During the first phase of the WSDP (2007-15) implementation arrangements were decentralized
to the district levels (Local Government Agencies- LGAs), with the intention of providing better technical
support to the rural communities. The WSDP adopted a Community Driven Development (CDD)
approach in the rural water sector, where the village community and its institution, the COWSO
(Community Owned Water Supply Organizations), were given full ownership of their rural water system
along with the associated management responsibility. Under this arrangement, the Ministry of Water and
Irrigation (MOWI) set policies and guidelines and provided technical support to the Local Government
Authorities (LGAs). The actual implementation of new water projects is facilitated by the LGAs under the
President’s Office of Regional and Local Government (PORALG).
              One of the more ambitious projects that went hand-in-hand with the country’s effort in improving
water access was the creation of its Water Point Mapping (WPM) database. With the assistance of Water
Aid which has previously done similar work in Malawi (Stoupy & Sudgen, 2003), the country
implemented the WPM with the goal of gathering geographical data on all improved water points in an
area, in addition to management, technical and demographic information. The information collected was
used to create a baseline of accurate, reliable and up to date information on all improved water points in
addition to serving as a benchmark for future monitoring efforts, improving decision making as well as
allocation of resources for rural water supply services (Welle, 2010).
              Through the implementation of the first phase of the WSDP and by extension, the RWSSP, an
additional 8,285 water points have been installed covering almost 1.9 million additional beneficiaries.
This has increased rural water coverage from 55 percent in 2007 – roughly at the start of the program – to
                                                            
5 The figures reported in JMP (2015) were lower than those reported in the 2002 Housing Census. While population growth
pressures could have influenced the figures, the difference is most likely due to the different definition of improved water points
that were adopted by the two databases. Under the WSDP, the WPM have since adopted the definition of improved water points
that is consistent with the internationally accepted definition by WHO/UNICEF (2000). The figures presented in JMP (2015)
would be based on this definition.
6 Goals that were set partly through the commitments of the National Strategy for Growth and Reduction of Poverty also known

as MKUKUTA and by the Millennium Development Goals (MDG) respectively. 


                                                                4 
 
about 57 percent in 2012. The number of people with improved water access throughout the country on
the other hand has also increased from 21.5 million in 2007 to 22.4 million in 2012. While progress has
been made, the MDG targets were not met. Meeting the targets will be difficult especially when the
country has failed to meet its 2010 target (65 percent of water coverage for its rural population).
              Part of the underwhelming success of WSDP’s first phase can be attributed to the pressures of
rapid population growth as discussed earlier. However, there were several other challenges. Although the
WSDP was successful in changing the institutional arrangements in the sector, systemic challenges in
terms of capacity, operational costs, data accountability and coordination put the program’s
implementation and sustainability at risk. At the village level for example, the village committees or the
COWSOs have limited technical and managerial capacity while at the central government level, the
ministry has struggled to provide timely support, guidelines and budget to its implementation partners
(LGAs). Additionally, the high operating costs of certain types of water pumps that were selected by the
participating communities resulted in approximately half of the population paying more than 5 percent of
their household income for water access7. The process of data collection and management through
MOWI’s ambitious project to map all the water points in the country was not                   fully    institutionalized.
Hence it is difficult to hold the district water engineers accountable for dealing with non-functionality.
Lastly, while coordination between MOWI and other participating parties such as PORALG and LGAs
have steadily improved, local level planners and decision makers rarely receive feedback on
implementation issues, thus adversely affecting policy reforms and resource allocations at the national
level. WaterAid (2009) provided several recommendations that could potentially improve the country’s
rural water access: instituting, monitoring and regulation of COWSOs, improvement of the WPM data8;
promoting more community participation; providing autonomy in managing their water infrastructure;
and the provision of technical support for COWSOs to conduct complex maintenance works.
              With around 21 million Tanzanians still lacking access to improved water, the country fell short
of reaching its Millennium Development Goal (MDG) targets to halve the proportion of people without
improved drinking water and sanitation in 1990, by 2015. The new Sustainable Development Goal (SDG)
of reaching universal access to safe water and sanitation by 2030, will prove even more challenging. This
goal includes not only providing water access to the population, but also ensuring a continuous supply of
sufficient, affordable and clean water for all. While the World Bank’s support has made headway in
improving access to rural water supply in the country through Rural Water Supply and Sanitation

                                                            
7 It is reported that some were paying close to 30 percent of the household income. See DFID (2016): “Measuring and

Maximizing Value for Money of Rural Water Supply (RWS) Investments in TZ”.
8 Better WPM data is essential to the success of Tanzania’s water targets. According to Verplanke and Georgiadou (2017) and

Water Aid (2009) the WPM database for Tanzania is riddled with errors and inaccuracies. 


                                                               5 
 
Program (RWSSP) there is still an urgent need to reconfigure the institutional framework and retool the
sustainability approach to achieve a lasting and high-quality delivery of water services in Tanzania.

3.            Literature Review


                             In recent years, statistics on water points functionality have been collected in a number of
countries. One of the more comprehensive database is Akvo (2015) developed by the the Akvo
Foundation. It is currently monitoring data for over 120,000 water points across 37 countries. Globally,
20 percent of the water points are not functional while an additional 10 percent are functional but with
problems9. Other studies indicate that in Sub-Saharan Africa, it is estimated that about 30 percent to 36
percent of water points are not functional (Baumann, 2009; RWSN, 2009) with countries such as Cote
d’Ivoire (65 percent) and Sierra Leone (65 percent) experiencing high rates of water points failures and
countries such as Madagascar (10 percent) and Benin (22 percent) experiencing lower levels (Table 1,
Annex)10.
              Another problem is that there is no widely agreed upon definition of functionality. This is an
important aspect of data accuracy, because the inaccuracy of the data collected – especially when the
definition of functionality is not clearly defined – could adversely affect programs that are designed based
on flawed data (Jimenez & Perez-Foguet, 2010). While most surveys use the binary distinction
(Cairncross et al., 1980) of working or not working, which at times is considered insufficient (Carter &
Ross, 2016), other surveys included the concept of partial functionality (Wilson et al., 2016). Initiatives
such as the Akvo Flow and the Water Point Data Exchange (WPDx)11 were created to not only collect
information and indicators on water points across different countries but also harmonize the data
definitions which are often collected and measured differently. A recent World Bank study which
analyzes a range of indicators from countries and development partners, including 20 national monitoring
systems and 20 monitoring frameworks from donors proposed a shortlist of indicators and associated
metrics as a global framework: (1) service levels (the characteristics of water that users receive); (2)
functionality (the physical condition and functioning of a supply system); and (3) upkeep (those factors,
including external backup support, that affect the performance of the service provider in its roles of
operation, maintenance, and administration).



                                                            
9Additionally, the Water Point Data Exchange (WPDx) was also launched in May 2015 with the aim of compiling and
harmonizing statistics of water points globally which are often collected and shared using unique approaches. Data from Akvo
Flow for example will also be shared on WPDx.
10
       
11   WPDx’s website: https://www.waterpointdata.org/


                                                                 6 
 
              Previous studies in the literature on water point functionality found several factors that contribute
to their sustainability in the long run. These include technical conditions such as construction,
maintenance and age (O`Keefe-O’Donovan, 2012), social conditions such as competency and autonomy
in managing the water points (Marks et al., 2014; Harvey & Reed, 2007), financial conditions such as
funds for maintenance (Foster, 2013; O`Keefe-O’Donovan, 2012) and tariffs (Sayre and Taraz, 2019), as
well as hydrological determinants such as changing water tables or groundwater productivity (Andres et
al., 2018; Fisher et al., 2015; Miller and Doyle, 2014; Harvey, 2004). Additionally, Baumann (2009)
suggests that these different conditions could be categorized as “soft” and “hard”. “Soft” conditions
include community ownership, a perceived need for the water point, and user skills, behaviors, norms,
and practices. “Hard” conditions include human resources, financial resources and suitable technologies.
              One of the primary factors that can be considered as a “hard” condition is the continuous
maintenance or management of water points (Koestler et al., 2010; Morgan, 1993; Mudege, 1993). While
the likelihood of water points becoming dysfunctional, increases with age (Banks & Furey, 2016; Foster
2013; O`Keefe-O’Donovan, 2012), it can be mitigated to a large extent by proper maintenance. In
Tanzania for example, it is estimated that 27 percent of water points became dysfunctional when they
reach year two (Banks & Furey, 2016). Furthermore, Gondwe and Rukiko (1987) documented the lack of
maintenance as a contributing factor to water points and schemes failure in Tanzania. Additionally,
maintenance itself is more important than technology in sustaining water points functionality (Mudege,
1993). Evidence indicates that lack of maintenance results in water points failure as in in the case of
Morocco in the early 1980s (Lynch, 1984). Meanwhile, for Mexico studies indicate a bias towards new
water infrastructure rather than maintaining existing ones (Gibson & Rioja, 2017; McNeill, 1985). The
reduction of post-installation abandonment would thus improve the functionality and performance of
these water points (Carter & Ross, 2016).
              The cost of maintaining and sustaining water systems has been also emphasized (Whittington et
al., 1989) along with the availability of spare parts when needed (Gill & Flachenburg, 2015; Koestler et
al., 2014;).12 Furthermore it is also potentially cheaper to maintain than to install new water points
(Agarwal et al., 2015)13. Maintenance activities could be sustained by user fees, thus contributing to the
functionality of the water points in the long run (Foster, 2013). For Tanzania, using regression and
Bayesian network (BN) analyses, Cronk and Bartram (2017) found that the functionality of water points
is better sustained when a steady flow of funds is available (i.e when fees are collected on a monthly
basis) for maintaining the water points rather than in response to a system breakdown.
                                                            
12In some cases, politics also play an important role in the maintenance of such infrastructure (Carlitz, 2017).
13Notwithstanding transportation costs, the cost of maintaining water points could be as cheap as $10 while the cost of installing
new ones could be more than $1000


                                                                7 
 
              While the role of maintenance is crucial for the sustainability of water points, the manner in
which the water points are maintained or managed is also important. This is the “soft” condition
suggested by Baumann (2009) although the evidence is not clear. Cronk and Bartram (2017) found that
water points managed by private operators in Tanzania and Nigeria were more functional than those
managed by the local communities. Meanwhile, with increasing ownership of the community
participation in waterpoints management, sustainability can be improved. Holm et al., (2016) for example
suggested that the sustainability of water points in Malawi could be improved if the communities were
given autonomy in deciding the type of water points to be installed. This is also supported by Gill (2014)
and Gill and Flachenburg (2015) who found that water points were better sustained through increased
ownership by communities, and by WaterAid (2009) which found that more autonomous community
groups such as Water User Groups and Water User Associations were more successful in sustaining water
points. While private operators remain a viable alternative for the management of waterpoints, there is a
risk of profiteering which will be detrimental to the expansion of coverage and sustainability of
waterpoints in underserved populations (WaterAid, 2009). Finally, community management of water
points can also be useful in terms of pooling maintenance and financial risks (Koehler et al., 2015)14.
              A number of studies also found that functionality of water points post installation, is affected by
technology or its type. This is a “hard” condition under Baumann’s definition. In Tanzania, Cronk and
Bartram (2017) found that Nira handpumps were more functional than Afridev and India Mark II
handpumps. Similarly, using data mining techniques, Chowdavarapu and Manikanandan (2016) showed
that the extraction and water point type significantly predicts functionality. In addition, there could be an
endogenous relationship regarding the observed high levels of non- functionality of water points. Certain
types of water pumps are more expensive to sustain than others, and this could be challenging for
communities which lacked the financial resources and hence are unable to replace or repair water pumps,
leaving them dysfunctional. Meanwhile, a survey of 512 water points in Kenya for example found that 44
percent of submersible pumps were not functioning while only 2 percent of handpumps were not (Goodall
et al., 2016). Similar results were found for Timor Leste (Willets, 2012).
              Finally, a number of studies have also found the importance of hydrological factors in explaining
water point functionality. For Tanzania, Fisher (2015) found a statistically significant relationship
between functionality and groundwater storage. Water points in areas of low storage is associated with
higher functionality. Similarly, using data from the 2015 Nigeria National Water and Sanitation Survey,
Andres et al. (2018) found that hydrogeological factors are significant in explaining water point

                                                            
14Additionally, Koehler et al. (2015) also suggested that the sustainability of rural water supply can be further improved by
taking advantage of advances in monitoring and payment technologies.


                                                                 8 
 
functionality. In fact, among water points that are one-year old, hydrogeological factors such as
groundwater productivity, groundwater storage and depth to groundwater have the greatest effect among
other factors on their functionality.


       4. Data and Methodology


       (a) Water Point Mapping Data
              This study uses the WPM data compiled in November 2016)15. The analysis made use of data on
40,917 water points out of a total of 83,295. This final figure was derived after we 1) restricted our
sample to those between 1 and 20 years old; 2) removed other water schemes, defined as a water point
with multiple taps; 3) identified other additional duplicates; and 4) dropped water points with incomplete
information16. Information pertaining to the settlement type, types of pump, types of promoters and water
source is found in the WPM dataset, in addition to the functionality status of the water points. Following
Andres et al., (2018) and Fisher (2015) a number of hydrological information were combined with the
WPM dataset which were obtained from the British Geological Survey (BGS) (MacDonald et al., 2012).
The hydrological database included groundwater productivity17, groundwater storage18 and depth to
groundwater19. Each data were presented in the form of a map with a 5km resolution grid and they were
geospatially matched to each water point available in the WPM dataset where possible. The water points
were matched such that only about one percent of water points in the dataset did not have at least one of
those measurements.
       (b) Methodology and Empirical Strategy
              First, Locally Weighted Scatterplot Smoothing (“Lowess Smoothing” of running-line least
squares) is employed to estimate a locally weighted non-parametric regression of ‘non-functionality’ (0,1)
on age in years (1<age<15) for the water points in our data. Lowess Smoothing Curves across the
different subgroups of the water point characteristics such as the types of pumps and their management
type were also estimated to analyze the association between age and the probability of failure.


                                                            
15
      The distribution of water points is shown in the Map of Tanzania (figure 17). 
16 As Verplanke and Georgiadou (2017) highlighted, errors in the dataset such as the double counting of water points are
common within the WPM data.  
17 The groundwater productivity map indicates the borehole yields that can reasonably be expected in different hydrogeological

units (MacDonald, 2012). Definitions can be found at:
http://www.bgs.ac.uk/research/groundwater/international/africangroundwater/mapsDownload.html
18 Groundwater storage was estimated by combining the saturated aquifer thickness and effective porosity of aquifers across

Africa. For each aquifer flow/storage type an effective porosity range was assigned based on a series of studies across Africa and
surrogates in other parts of the world (MacDonald et al., 2012).
19 Depth to groundwater was modelled using an empirical rules-based approach, where depth to groundwater was assigned

according to rainfall and aquifer type, as well as proximity to rivers (MacDonald et al., 2012).


                                                                9 
 
              Second, to investigate the factors contributing to the failure or non-functionality of water points
in our study, a series of logistic regression analyses were estimated using the following equation:
               NON-FUNCT = β0 + β1AGEx + β2AGESQ + β3URBAN + β4ZONE+ β5BODY + β6PROMOTER
              + β7PUMP + β8GRWPRODUCT+ β9GRWSTORAGE + β10DEPTH + β+ ε
              The dependent variable is the likelihood that the water point is non-functional (NON-FUNCT)
which is a binary variable where 0 denotes one that is functioning and 1 denotes non-functional.
Following O`Keefe-O’Donovan (2012) and Komives et al (2008) , the age (AGE and AGESQ) of water
points is expected to be positively associated with higher rates of failures. URBAN denotes household
living in the urban areas and there is no particular expectation regarding this variable. Additionally, we
also account for the different geographical zones (ZONES) which separates the water points into lake,
central, coastal, northern and southern highlands. These zones control for the unobserved characteristics
common to them. In addition, we also control for the managing body (BODY) of the water points. These
are village committee including COWSO, water authority board and government and private service
providers. Similarly, we also account for the promoters (PROMOTER) of the waterpoints who are either
private, donors or others. Additionally, we also control for the type of pump (PUMP) for each water
point. Finally, following Andres et al (2018) we include several variables that represent the
hydrogeological conditions of the terrain where the water point is located. These measures include
groundwater productivity (GRWPRODUCT), ground water storage (GRWSTORAGE) and depth to
groundwater (DEPTH). For the purposes of the analysis, these potential determinants of failure are
grouped together as follows: those that are one year old, between 2 and 4 years old, between 5 and 10
years old and those that are older than 11 years. Subsequently, using the results from our logistic
regression estimates, we also apply Shapley (1953) decomposition to analyze the relative contributions of
these factors to the probability of failure20. Shorrock’s (1984; 1982; 1980) application of Shapley
decomposition is used to study the contribution of different determinants of water point functionality
utilizing regression analysis.
               Let S be the aggregate indicator which measures water scheme outcomes, Xk , where k = 1,2,
….m be a set of factors contributing to the value of S. The following equation can be written:
                                                           S = f(X1, X2,….Xm),
              Where f(.) is an appropriate aggregation function. The goal of all decomposition techniques is to
attribute contributions, Ck to each of the factors, Xk, so that the value of S be equal to the sum of m
contributions to water scheme functionality outcomes. These X include the explanatory variables of

                                                            
20
   In studies of determinants of poverty for example, the Shapley decomposition is often used to evaluate the contribution of each
determining factor by analyzing the decomposition of the observed variation or R-squared value. 


                                                                             10 
 
functionality outcomes used in the linear probability regression. The relative shares of the variation in
water point failure of the explanatory variables such as age, hydrology, localities, zones, pump type,
promoter and management body are decomposed to explain the contribution of each to the likelihood of
failure in each of these three age groups.
    (c) Descriptive Statistics
        Table 2 presents the characteristics of water points by different geographical zones. 29 percent of
all water points are non-functional, and this varies by regions with the Northern zone and Southern
highlands zone having the lowest rates at 24 percent and 25 percent respectively while the coastal and
central zone was the highest at 33 percent each. 62 percent of the water points are situated in areas with
low-moderate groundwater productivity, while 20 percent are situated in moderately productive
groundwater areas and 14 percent in highly productive areas. The groundwater productivity of water
points also varied considerably by regions. Low-moderate groundwater productivity accounted for 89
percent of water points in the Central zone but only 48 percent in the Northern zone, with others in
between. This is not surprising as Tanzania’s central zone is typically drier and hotter than the rest of the
country,
           For groundwater storage, 45 percent of the water points in the dataset are situated in areas with
low groundwater storage while another 35 percent are situated in areas with low-medium groundwater
storage and additional 19 percent in high groundwater storage areas. Again, there is considerable variation
among the regions with only 30 percent of all water points in the Coastal zone having low groundwater
storage whereas the same is true for 81 percent of all water points in the Central zone. For depth to
groundwater, about 68 percent of water points are situated in areas with very shallow depth to the
presence of groundwater. Again there is considerable variation among the regions. 81 percent of the
water points in the Southern Highlands zone are in areas with very shallow depth compared to 64 percent
in the Central zone and only 46 percent in the Lake Zone.
        89 percent of all water points in the dataset are located in rural areas with small variations across
the various regions. The additional 11 percent are located in urban areas, but it varied from only 4 percent
of water points in the Northern zone to 24 percent in the Central zone. In terms of the ecological class, 58
percent of water points are situated in areas with shrub and herb vegetation and an additional 21 percent
are situated in forests and woodlands. Shrub and herbs areas accounted for 64 percent of water points in
the Lake zone while it was only 46 percent in the Central zone. Forest and woodland accounted for 33
percent in the Coastal zone but only 1 percent in the Central zone. Finally, desert and semi-desert account
for only 2 percent in the Coastal zone but 48 percent in the Central zone.
        For types of pumps, gravity water pump accounted for 43 percent of all water points while
manual hand pumps accounted for 29 percent and motorized pump 19 percent. However, there are

                                                     11 
 
considerable differences across the regions. 41 percent of the water points in the Lake Zone have manual
hand pumps compared to only 8 percent in the Northern Zone. Motorized pumps ranged from a low of 8
percent in the Lake Zone to a high of 50 percent in the Central zone while gravity water pump accounted
for 70 percent of water points in the Northern zone but only 20 percent in the Coastal zone.
           Among the promoters, 39 percent are public, and 38 percent are private for Tanzania. Donors
only accounted for 5 percent across the country and the variation among the zones is also very small. A
majority of the water points in the dataset is managed by village committees (79 percent) compared to
those managed by water authority boards (11 percent). Private sector service providers only accounted for
6 percent of the total water points. There is very little variation across the zones.
           Among the sources, nationally, gravity flow systems accounted for 46% of all systems for
Tanzania, while the pump piped system (groundwater based) accounted for 50%. The corresponding
figures for Central zone it was 23% and 77%, for Coastal zone it was 21% and 66%, for Lake 40% and
57%, for Northern zone, 73% and 23% and for Southern Highlands zone 61% and 37%. Surface water-
based pump piped system and unknown accounted for a very small proportion of all systems for all
regions.
           Finally, regarding extraction technology, gravity flow system accounted for 43%, other systems
accounted for 32% and surface water pump accounted for 17%. However, for the Central zone, surface
water pump accounted for 49%, for Coastal zone 37%, and for Lake 45%. For Northern zone it accounted
for only 10% and Southern Highlands 23%. Gravity flow system accounted for 70% of all extraction
technology in the Northern zone and 60% in the Southern Highland zone. In other regions, gravity flow
system accounted for less than 50% of all technology.
           Table 3 presents groundwater productivity by various types of pumps. More than half of the water
points have low to moderate groundwater productivity and this is the same across all types of water
pumps. Similarly, more than half of the water pumps in the dataset are situated in areas with very shallow
groundwater depth and this is the same for all the different types of pumps (Table 4). Finally, low and
low-medium groundwater storage accounted for most of the water points across the different categories of
pumps (Table 5).

5.         Results


     (a) Functionality Analysis
           The results of a locally weighted scatterplot smoothing (LWSS) (“Lowess Smoothing” of
running-line least squares) regression of “non-functionality” (0,1) on age in years (1<age<20) for water
points are presented in Figures 1–15. The Lowess curve presented in Figure 1 clearly indicates that in


                                                      12 
 
general, the probability of water points failures rises steadily with age (years). On average, 20 percent of
all water points fail within the first year and over 40 percent fail 20 years after installation.
        Around 20 percent of water points in the rural areas are likely to fail within the very first
year after installation, while 10 percent are likely to fail in the urban areas (Figure 2). While the
likelihood of failure increases with age, it also varies significantly between urban and rural areas. Initially
the probability of failure is higher in the rural than in the urban areas, but by the seventh year, they are
equal, and by 20 years after installation, it is higher in the urban areas (40 percent) and lower in the rural
areas (35 percent).
        Regionally, the likelihood of failure is higher in the central zone than in other parts of the
country (Figure 3). During the first year, the probability of water point failure is higher in the Central
zone at 40 percent while it is much lower in the other zones. However, 20 years after installation, all
except the Northern zone converge at a higher level (between 30 and 45 percent). On the other hand, the
probability of failure in the Northern zone is around 15 percent during the first year and increases to only
30 percent by 20 years after installation.
        The likelihood of failure rises sharply with age for the donor managed water points (Figure
4). During the first year, both public and private managed water points exhibit a higher likelihood of
failure at 15 percent and 23 percent respectively. This increases steadily to around 40 percent for both
groups twenty years after installation. However, the likelihood of donor managed water point failing is
near zero during the first year, but increases sharply to 40 percent by the 10th year (lower than public and
private managed water points) before declining to around 30 percent by the 20th year.
        In terms of source, systems-gravity flow, pump piped (ground water) and pump piped
(surface water) have similar likelihood of failure (around 20 percent) during the first year (Figure
5). However, while there is a sharp rise in the failure rates of pump piped groundwater system to 40
percent by the 10th year and 50 percent by the 20th year, the probability of failure for the other two
systems only increase to 30 percent by the 20th year.
        Water points managed by village committees had a significantly higher likelihood of failure
than those managed by water authority or private operators (Figure 6). The water points managed by
village committees have a likelihood of failure of more than 20 percent during the first year of operation
rising steadily to 40 percent by year 20. Water points managed by the private sector starts out with a
failure likelihood of less than 20 percent in the first year declining to less than 10 percent by the third
year, increasing sharply to around 30 percent by the 10th year before falling to below 20 percent by the
20th year after installation. Water points managed by Water Boards start out with a failure likelihood of
less than 10 percent in year one and thereafter increasing steadily to 20 percent, 20 years after installation.



                                                       13 
 
        Regarding extraction technology, the groundwater pump system shows a probability of
failure of around 10 percent during the first year, while both gravity flow and surface water system
show twice the rates of failure (20 percent) (Figure 7). However, the failure likelihood of the gravity
flow system drops slightly after first year before increasing steadily to 30 percent by 20 years after
installation. Meanwhile, the likelihood of failure of the surface water pump increases sharply after the
first year before levelling off at around 30 percent by the 12th year and thereafter remaining steady.
Finally, the failure rates of groundwater pump increase sharply from 10 percent during the first year to
slightly less than 40 percent by the 10th year, and thereafter declining slightly to around 35 percent by 20
years after installation.
        Motorized pumps have a consistently higher likelihood of failure over the twenty-year
period relative to other pumps (figure 8). Initially, motorized pumps have a 30 percent likelihood of
failure in year one, and this increases steadily to around 45 percent by 20 years after installation.
Meanwhile, manual hand pumps start out with a failure likelihood of less than 10 percent in the first year,
increasing sharply to over thirty five percent by the 10th year, and falling to around thirty percent by the
20th year. Finally, gravity water pumps start out with a 20 percent likelihood of failure in the first year,
increasing steadily to a 30 percent likelihood of failure by the 20th year.
        Water points with low-medium groundwater storage start out with a high 35 percent failure
rate in the first year, initially declining to 25 percent by the 10th year and thereafter increasing to 40
percent by the 20th year (Figure 9). Meanwhile water points with both low and medium storage depth
have a similar failure rate of around 10 percent in the first year, and while the failure rate for both
increase steadily, it diverges sharply, and is considerably higher for medium, at 45 percent relative to low
at around 30 percent by the 20th year.
        The likelihood of failure for water points situated at shallow, very shallow and shallow-
medium groundwater depth is around 18 percent in the first year and increases at about the same
rate until around the 15th year (Figure 10). Thereafter it diverges sharply with the likelihood of failure
for water points at shallow-medium areas at over 50 percent, for shallow areas, over 40 percent and for
very shallow areas around 35 percent by the 20th year after installation.
        Water pumps with different groundwater productivity exhibit almost exactly the same
trends over the 20- year period regarding the likelihood of failure (figure 11). However, while they
all start out with similar rates of failure - around 20 percent in the first year – highly productive water
points have over 50 percent failure rates by the 20th year compared to others which have only 40 percent
likelihood of failure.
        For the ecological zones, water points located in forest and woodland as well as shrub and
herb vegetation have a 20 percent likelihood of failure in the first year while those located in desert

                                                      14 
 
and semi desert areas have a likelihood of failure of around 12 percent (Figure 12). However, while
the likelihood of failure for water points located in forest and woodland declines sharply after the first
year, before increasing to slightly less than 20 percent by the 20th year, those located in areas with shrub
and vegetation show a steady increase in failure likelihood to 40 percent by the 20th year after installation.
Finally, the failure likelihood of those located in desert and semi-desert increases from slightly over 10
percent in the first year to slightly over 40 percent in the 20th year after installation.
        Figures 13-15 examine how the failure rates vary by the type of pump and the hydrological
factors such as groundwater depth, groundwater storage and groundwater productivity to better
understand how technology interacts with geography in affecting the non-functionality of water points.
The failure rates of water points by depth to groundwater categories varied by type of pump
technology (figure 13). For instance, manual hand pump failure rates increased sharply from less than 10
percent for all depth to groundwater categories to 35 percent by the 12th year. After that, for hand pumps
in both shallow and shallow-medium areas, the likelihood of failure continues to increase reaching over
40 percent by the 20th year after installation. However, for hand pumps in very shallow areas, the failure
rate declines to around 30 percent during the same time period. For motorized pumps, the failure rate is
around 30 percent during the first year, after which there is a decline for those in very shallow and
shallow medium areas. However, after the fifth year both increase sharply. Failure rates of motorized
pumps with shallow medium reaches 60 percent by the 20th year after installation. For motorized pumps
in very shallow areas it only reaches 35 percent. For shallow depth to groundwater it increases from 30
percent from year one to over 50 percent in year 20. Lastly, for gravity pumps, the failure rate at shallow
medium and very shallow areas starts around 20 percent during the first year but declines to around 15
percent by the fifth year. Thereafter, the failure rate at shallow-medium areas increases sharply to 40
percent after 20 years while the failure rate of gravity pumps at very shallow areas increases to 30 percent
for the same period. Finally, for gravity pumps at shallow depth to groundwater areas, rate of failure starts
at 11 percent increasing steadily over the years, reaching 40 percent after 20 years.
        The failure rates of water points by groundwater storage categories also varied by type of
pump (figure 14). The likelihood of manual hand pump failure is less than 10 percent in the first year for
all cases, but after the third year they all diverge. Those with medium groundwater storage experience a
sharp increase in failure likelihood reaching almost 45 percent by the 12th year before declining to 35
percent by the 20th year. Failure rates of manual hand pumps at both low and low-medium areas show
parallel increase until the 10th year after which those in low medium areas declines to less than 30 percent
while those in medium areas increases to nearly 40 percent. For motorized pumps, failure rate for those
with low medium groundwater storage was nearly 40 percent during the first year and for the others it was
20 percent. However, failure likelihood is u-shaped for the former over the years, declining to 25 percent

                                                        15 
 
by the 10th year before increasing to 40 percent by the 20th year. Failure rate for those with medium
groundwater storage initially declines reaching 12 percent by fifth year. Thereafter it increases sharply to
40 percent by the 20th year. For those with low groundwater storage, there is a sharp increase from 20
percent failure rate in the first year to about 40 percent by the 10th year and it stays the same until the 20th
year. Finally, for gravity pump, the failure likelihood for those with low groundwater storage increases
from 5 percent in the first year to 25 percent by the 20th year after installation. For those with medium
storage it increases from 12 percent in the first year to over 40 percent by the 20th year. For those with
low medium storage it is a u-shaped curve, similar to the motorized pumps and the failure rates by year
are also similar.
        Finally, the failure rates of water points by groundwater productivity also varied by the
type of pump (figure 15).        For manual handpumps those with low moderate, moderate and high
productivity start out with a failure rate of less than 10 percent in the first year while those with very low
productivity start out at 15 percent. However, while failure rates for those with very low groundwater
productivity increases to 25 percent by the 20th year after installation, those in other categories show a
sharp increase. Manual handpumps with low moderate groundwater productivity shows an increase in
failure rates to 30 percent by the seventh year and thereafter it remains the same until the 20th year of
installation. The failure rates for those with high groundwater productivity increase sharply to 40 percent
by the 12th year and thereafter it stabilizes at that rate. Finally, failure rates of those with moderate
productivity increase to around 40 percent by the 12th year and thereafter it declines reaching less than 30
percent by the 20th year. For motorized pumps, the failure rates for those with high groundwater
productivity exhibits a u-shaped pattern, starting at nearly 50 percent by the first year, declining to around
20 percent by the seventh year, and increasing to nearly 40 percent by the 20th year. Similarly, for
motorized pumps with moderate groundwater productivity, failure rates start at near 50 percent for the
first year after installation but increases sharply to nearly 75 percent by the 20th year. For very low
productivity the failure rates range between 35 and 50 percent during the entire period and do not exhibit
a clear pattern. Finally, for gravity pumps, all categories exhibit an upward trend. Failure rates for very
low is 5 percent in the first year, increasing to nearly 50 percent in the 20th year after installation. For
those with high groundwater productivity, failure rates increase from 5 percent in year 1 to nearly 70
percent by year 20. For low moderate productivity, failure rates start out 20 percent in the first year,
initially declining, but sharply increasing to near 30 percent by the 20th year. For those with moderate
productivity, failure rates start near 30 percent in year 1, declines slightly to 20 percent by the fifth year
before increasing again to near 30 percent by the 20th year.




                                                      16 
 
       (b) Empirical Results
              Table 6 presents the results of the logistic regression21 estimates for the determinants of water
point failures for the full sample and four sub-samples for water points divided into the following age
groups: one year old, two to four years old, five to ten years old and eleven years and older. Analyzing
these sub- samples based on age has implications for policy, particularly by providing insights into the
types of interventions necessary at various time points during the life of water points.
              In general, and as expected, controlling for other factors, all water points experience higher
failure rates with age and this is significant at the one percent level. The age squared variable indicates
that the effect of age on water point failures increases at a declining rate for all water points and for water
points between 5 and 10 years old. For water points between 2 and 4 years old however, the relationship
is opposite. Basically, the probability of failure decreases at an increasing rate by age. All in all, there is a
complex relationship that is observed between age and water point failures if broken down into different
age groups. On the other hand, urban and rural water points are equally likely to fail controlling for other
factors in general. However, urban water points are less likely to experience failure at two to four years
old but more likely to fail when they are over 11 years old and they are significant at the 1 percent level.
              Using water points from the Lake zone as the reference group, failure rates are higher among
those in the Central zone, in all cases except for water points between 2 and 4 years old, and all are
significant at the one percent level. Those over 11 years old have an insignificant impact. Water points in
the Coastal zone also experience higher rates of failures compared to those in the Lake zone. The effect is
also significant at the one percent level for all water points, those between 5 and 10 years old and those
over 11 years old. The results are similar for the water points located in the Northern zone. For the
Southern highlands zone the non-functionality is positive and significant at the one percent level for all
water points in Tanzania, and for those over 11 years of age, and insignificant for others.
              Using village committees as reference, those water points managed by a water authority or
government have lower failure rates for all age groups. The results are similar for the private sector. For
the other sector, it is negative and significant only for those water points that are one year old, while it has
an adverse impact for water points that are over 11 years old. Compared to the public group, other
promoters have a lower failure likelihood than either donors or the private sector. Regarding the types of
pumps, all kinds of pumps have a higher likelihood of failure for all age groups and they are significant at
the one percent level using gravity flow pump as reference.



                                                            
21
   We have also found that the variance inflation factor for the independent variables we have included in the estimation model to
be around 2.90 which is not a serious problem.  


                                                               17 
 
        For groundwater productivity, water points with moderate productivity have a higher failure
likelihood in the full sample, those between 5 and 10 years old and those over 11 years using very low
and low to moderate groundwater productivity water points as reference. On the other hand, water points
with very high and high groundwater productivity have lower likelihood of failure in the first year, and
higher likelihood at over 11 years. For groundwater storage, those with low medium groundwater storage
has failure rates that are higher for the full sample, year 1 and those between 2 and 4 years old but are
insignificant thereafter. For those with medium storage, they have a higher likelihood of failure for all
cases, years 5-10 and over 11 years old. This result supports the findings from Fisher (2015) that water
points with low groundwater storage are associated with higher functionality in general. There is clear
evidence showing that hydrogeological factors are important predictors of water point failures as in the
case of Andres et al. (2018) and Fisher (2015).
        Finally, regarding depth to groundwater, using very shallow as reference, shallow groundwater
was positive and significant at the one percent level for all groups suggesting higher failure rates, except
those which are one year old and under, which was insignificant. The result was similar for those for
which no values were available. Finally, shallow medium had higher failure rates for only those water
points which are 2-4 years of age and those between 5 and ten years.
    (c) Shapley Decomposition of Water Points’ Failure
        In this section, we use the results from logistic regression to calculate the contribution of various
factors to the failure of water points (Table 7). We group together water points that are one year old, two
to four years old, five to ten years old and over eleven years old to determine failure by age. The predicted
effects of age, hydrogeology, localities, zones, pump types, promoter and management body in explaining
the likelihood of water scheme failure in each of these three age groups, along with the Shapley values
(Table 7) for each component’s contribution to failure, is calculated (Figure 16).             The marginal
contribution of pump type is at 58 percent for all water points for all age groups while age only accounts
for 16 percent. The marginal contribution of hydrological factors accounts for 6 percent for all water
points. During the first year, the marginal contribution of hydrological factors is the highest at 53 percent,
thus accounting for more than half of all components. However, the contribution of components changes
considerably by years 2-4. The pump type contributes more than half of the total when compared to all
the other components, accounting for 54 percent and this increases sharply to 63 percent by year 5 to ten
years old, and 70 percent for those over eleven years. There is no specific pattern regarding other
components as the water points age although hydrological terrain still contribute ten percent of reasons
for non-functionality for water points aged over five years. Type of promoters, localities and age make
small contributions to failure. Type of management body is important for years 2-4, and thereafter its
contribution declines significantly.

                                                     18 
 
               6.            Conclusions and Policy Implications
              The 2015 Tanzania Water Point Mapping data enables us to identify the failure rates of water
points and schemes across their lifetimes as well as the factors contributing to these failures.22 29 percent
of all water points are non-functional, and many failed within the first year. The results indicate that
nearly 20 percent of water points are likely to fail in the first year of construction, while nearly 40 percent
are likely to fail in the long run (within 20 years).
              The empirical results from the logistic regressions confirm the evidence provided in the Tanzania
context as well as those found in the literature. Water points managed by village committees had a much
higher likelihood of failure than those managed by private operators or water authority. This result is
similar to the ones found by Cronk and Bartram (2017).  Empowerment of village committees to manage
and maintain the water points under their supervision needs to be emphasized as recommended by
WaterAid (2009). Alternatively, a movement for more water points to be managed by private operators is
also viable although there is a risk of profiteering which would be undesirable especially when the aim is
to increase water access among rural populations (WaterAid, 2009). Additionally, following similar
results from Foster (2013), it was found that water points technology also affects their functionality.
Motorized pumps are more likely to fail than gravity flow pumps. In fact, as shown in the Shapley
decomposition, the technology of water points accounts for about 58 percent of the total water point
failures. Hence, there is a need to use appropriate technologies that is compatible with the hydrological
conditions while also emphasizing the management and maintenance capability of community led user
associations. Evidence also supports the recommendations of Holms et al., (2017) that villages should be
given more autonomy regarding decision to install and maintain water pumps.
              The variables described above such as the manner of how water points are being managed as well
as their technology, can be considered “modifiable” as suggested by Fisher (2015). Proper interventions
are needed to address the shortcomings contributing to water point failures. On the other hand, this study
also highlights the impact of factors contributing to water point failures that are not “modifiable”. These
are basically the geographic and hydrological factors such as the different geographical zones and
groundwater characteristics. These findings are useful to future decision-making processes in designing,
constructing or managing the water points at locations with varying hydrogeological characteristics. In
fact, the Shapely decomposition shows that the hydrological factors are more important than those of
locality and geographic zone types. While interventions cannot be designed to wholly change these

                                                            
22
   Note that this model does not intend to make causal inferences. Absent a control group, as in the case of
experimental or quasi-experimental research designs, it is not possible to make any strict causal inferences about the
role that each of these factors play in explaining the probability of water point failure. Further experimental research
is needed to analyze the causal factors driving water point failure.  


                                                               19 
 
factors, with appropriate and accurate hydrological data they can be designed to mitigate or reduce the
adverse effects on water points. Finally, the lack of hydrological data is a major impediment to
understanding the water point failures especially during the first year after their installation.
        It is evident that water points managed by village committees have a much higher likelihood of
failure than those managed by other groups, confirming the evidence discussed in the Tanzania context.
During the first years of the water points installation, hydrogeological factors play a major role in
determining their functionality. However, the type of pump matters over time with evidence showing that
gravity pumps have a much higher likelihood of failure over the long term. Overall, there is a need to
improve the capacity at the village level to manage and maintain water pumps as suggested. This can be
done through appropriate technical support and financial transfers from the LGAs to the villages. Finally,
technology as well as hydrological characteristics which affect the water pumps should also be taken into
consideration in constructing future water points.




                                                      20 
 
                                           Bibliography
Agarwal, K., Sarda, R. and Gajwani, M. 2015. Pump it up: Data Mining for Tanzanian Water Crisis.
      Department of Information Systems Management, Carnegie Mellon University.
Agénor, P.R. 2010. A Theory of Infrastructure-Led Development. Journal of Economic Dynamics and
       Control, 34: 932-950.
AKVO. 2015. Over One Million Surveys, Collected with Akvo FLOW. [ONLINE] Available
     at: https://akvo.org/blog/over-one-million-surveys-collected-with-akvo-flow/. [Accessed 30 May
     2018].
Alderman, H., Behrman, J.R., Lavy, V. and Menon, R. 2001. “Child Health and School Enrollment: A
      Longitudinal Analysis.” Journal of Human Resources, 36: 185-205.
Alexander, K.T., Tesfaye, Y., Dreibelbis, R., Abaire, B. and Freeman, M.C. 2015. “Governance and
       Functionality of Community Water Schemes in Rural Ethiopia.” International journal of public
       health, 60: 977-986.
Amann, E., Baer, W., Trebat, T. and Lora, J.V. 2016. “Infrastructure and its Role in Brazil's Development
      Process.” The Quarterly Review of Economics and Finance, 62: 66-73.
Andres, L.A., Chellaraj, G., Dasgupta, B., Grabinsky, J. and Joseph, G. 2018. “Why are so Many Water
       Points in Nigeria Non-Functional? An Empirical Analysis of Contributing Factors.” Policy
       Research Working Paper Series No. 8388, World Bank, Washington DC.
Banks, B. and Furey, S. 2016. What’s Working, Where, and for How Long. A 2016 Water Point Update to
       the RWSN (2009) Statistics, RWSN
Barbier, E.B. and Chaudhry, A.M. 2014. “Urban Growth and Water.” Water Resources and Economics,
        6:.1-17.
Barbier, E.B. 2004. “Water and Economic Growth.” Economic Record, 80: 1-16.
Basalirwa, C.P.K., Odiyo, J.O., Mngodo, R.J. and Mpeta, E.J. 1999. “The Climatological Regions of
        Tanzania Based on the Rainfall Characteristics.” International Journal of Climatology, 19: 69-80.
Baumann, E., 2009. “May-day! May-Day! Our Handpumps are not Working!” Rural Water Supply
      Network, Perspectives, 1.
Carlitz, R.D. 2017. "Money Flows, Water Trickles: Understanding Patterns of Decentralized Water
         Provision in Tanzania." World Development 93: 16-30.
Carter, R.C. and Ross, I. 2016. “Beyond ‘Functionality’ of Handpump-Supplied Rural Water Services in
        Developing Countries.” Waterlines, 35: 94-110.
Cairncross, S., Carruthers, L., Curtis, D., Feachem, R., Bradley, D. and Baldwin, G. 1980. Evaluation for
        Village Water Supply Planning. Published for International Reference Centre for Community
        Water Supply by John Wiley & Sons Ltd. Baffins Lane, Chichester, West Sussex PO19 1UD.
Chowdavarapu, I. K. and Manikandan, V. D. 2016. Data Mining the Water Pumps: Determining the
      functionality of Water Pumps in Tanzania using SAS Enterprise Miner. SAS South Central User
      Group Forum 2016
Cioffi, F., Conticello, F. and Lall, U. 2016. “Projecting Changes in Tanzania Rainfall for the 21st
        Century.” International Journal of Climatology, 36: 4297-4314.


                                                   21 
 
Cleaver, F. 1991. “Maintenance of Rural Water Supplies in Zimbabwe.” Waterlines, 9: 23-26.
Cronk, R. and Bartram, J. 2017. “Factors Influencing Water System Functionality in Nigeria and
       Tanzania: A Regression and Bayesian Network Analysis.” Environmental Science & Technology,
       51: 11336-11345.
De Bont, C., Komakech, H.C. and Veldwisch, G. J. 2019. “Neither Modern nor Traditional: Farmer-Led
      Irrigation Development in Kilimanjaro Region, Tanzania.” World Development, 116: 15-27.
Department for International Development (DFID). 2016. Measuring and Maximizing Value for Money of
       Rural Water Supply (RWS) Investments in TZ, Department of International Development (DFID)
Duti, V. 2012. Tracking Functionality for Sustainability. In Annual Review Conference of the Community
        Water and Sanitation Agency, Kumasi, Ghana.
Fisher, M.B., Shields, K.F., Chan, T.U., Christenson, E., Cronk, R.D., Leker, H., Samani, D., Apoya, P.,
        Lutz, A. and Bartram, J. 2015. “Understanding Handpump Sustainability: Determinants of Rural
        Water Source Functionality in the Greater Afram Plains Region of Ghana.” Water Resources
        Research, 51: 8431-8449.
Foster, T. 2013. “Predictors of Sustainability for Community-Managed Handpumps in Sub-Saharan
        Africa: Evidence from Liberia, Sierra Leone, and Uganda.” Environmental Science &
        Technology, 47: 12037-12046.
Gibson, J. and Rioja, F. 2017. “Public Infrastructure Maintenance and the Distribution of Wealth.”
       Economic Inquiry, 55: 175-186.
Gill, L. 2014. Evaluation of the Short, Medium and Long-Term Sustainability of Concern’s WASH
        Program in the Kagera Region of North-West Tanzania. Department of Civil, Structural and
        Environmental Engineering, Trinity College Dublin.
Gill, L. and Flachenburg, F. 2015. Sustainability of Hand-Dug Wells in Tanzania: Results from a Post-
         Evaluation. In Briefing Paper, 38th WEDC International Conference, Loughborough University
Giné, R. and Pérez‐Foguet, A. 2008, “Sustainability Assessment of National Rural Water Supply
       Program in Tanzania.” Natural Resources Forum 32: 327-342.
Gondwe, E. and Rukiko, M.M. 1987. “Failures of a Water Supply Scheme in a Developing Country.”
      Water international, 12: 114-119.
Goodall, S., Hope, R. and Katilu, S. 2016. Operational, Financial and Institutional Considerations for
       Rural Water Services: Insights from Kyuso, in 39th WEDC International Conference, Kumasi,
       Ghana, 2016.
Harvey, P.A. and Reed, R.A. 2007 “Community-Managed Water Supplies in Africa: Sustainable or
       Dispensable?” Community Development Journal, 42: 365-378.
Harvey, P. A. 2004. Borehole Sustainability in Rural Africa: An Analysis of Routine Field Data, in
       Proceedings of the 2004 WEDC International Conference, pp.147-150, Water, Eng. and Dev.
       Center.
Hazelton, D.G. 2000. The Development of Effective Community Water Supply Systems Using Deep and
       Shallow Well Handpumps. CSIR-Division of Roads and Transport Technology.
Holm, R., Singini, W. and Gwayi, S. 2016. “Comparative Evaluation of the Cost of Water in Northern
       Malawi: from Rural Water Wells to Science Education.” Applied Economics, 48: 4573-4583.


                                                  22 
 
Hyland, M. and Russ, J. 2019. “Water as Destiny-The Long-Term Impacts of Drought in Sub-Saharan
       Africa.” World Development, 115: 30-45
Jansz, S. 2011. A Study into Rural Water Supply Sustainability in Niassa Province, Mozambique. Water
        Aid.
Jiménez, A. and Pérez-Foguet, A. 2010. “Challenges for Water Governance in Rural Water Supply:
       Lessons Learned from Tanzania.” International Journal of Water Resources Development, 26:
       235-248
Koehler, J., Thomson, P. and Hope< R. 2015. “Pump-Priming Payments for Sustainable Water Services
       in Rural Africa.” World Development, 74: 397-411.
Koestler, A.G., Kahorha, J. and Biteete, L. 2014. “Supply Chain Analysis of Handpumps and Spare Parts
        in Eastern Democratic Republic of Congo.” UNICEF, Norway.
Koestler, L., Koestler, A., Koestler, M.A. and Koestler, V. J. 2010. “Improving Sustainability Using
        Incentives for Operation and Maintenance: The Concept of Water-Person Years.” Waterlines, 29:
        147-162.
Lein, H. 2004. Managing the Water of Kilimanjaro: Water, Peasants, and Hydropower Development.”
       GeoJournal, 61:155-162.
Lynch, W. 1984. “Handpumps in Rural Morocco.” Waterlines, 3: 15-18.
MacDonald, A.M., Bonsor, H.C., Dochartaigh, B.E.O. and Taylor, R.G. 2012. “Quantitative Maps of
      Groundwater Resources in Africa.” Environmental Research Letters, 7.
Mangyo, E. 2008. “The Effect of Water Accessibility on Child Health in China.” Journal of Health
      Economics, 27: 1343-1356.
Marks, S.J., Komives, K. and Davis, J. 2014. “Community Participation and Water Supply Sustainability:
       Evidence from Handpump Projects in Rural Ghana.” Journal of Planning Education and
       Research, 34: 276-286.
Mcneill, D. 1985. “The Appraisal of Rural Water Supplies.” World Development, 13: 1175-1178.
McPherson, H.J. and McGarry, M.G. 1987. “User Participation and Implementation Strategies in Water
      and Sanitation Projects.” International Journal of Water Resources Development, 3: 23-30.
Miller, B.W. and Doyle, M.W. 2014. “Rangeland Management and Fluvial Geomorphology in Northern
        Tanzania.” Geomorphology, 214: 366-377.
Ministry of Water. 2006. National Rural Water Supply and Sanitation Program (NRWSSP). United
        Republic of Tanzania.
Moller, L.C. and Wacker, K.M. 2017. “Explaining Ethiopia’s Growth Acceleration—The Role of
        Infrastructure and Macroeconomic Policy.” World Development, 96: 198-215.
Morgan, P. 1993. “Maintenance, the Key to Handpump Survival.” Waterlines, 11: 2-4.
Mudege, N. 1993. “Handpump Maintenance in Zimbabwe.” Waterlines, 11: 9-12.
Mudgal, A.K. 1997. India Handpump Revolution: Challenge and Change. Swiss Centre for Development
       Cooperation in Technology and Management.
Nekesa, J. and Kulanyi, R. 2012. “District Hand Pump Mechanics Associations in Uganda for Improved
       Operation and Maintenance of Rural Water-Supply Systems.” Waterlines, 31:170-183.

                                                 23 
 
Nicholson, S.E. and Kim, J. 1997. ‘The Relationship of the El Niño–Southern Oscillation to African
       Rainfall.” International Journal of Climatology, 17: 117-135.
O'Keefe‐O'Donovan, R. 2012. Cost Sharing and the Costs of Not Sharing: Cooperation and Incentives in
       the Maintenance of Public Goods, Oxford University.
Rural Water Supply Network (RWSN). 2009, Handpump Data, Selected Countries in Sub-Saharan
       Africa, RWSN.
Sayre, S.S. and Taraz, V. 2019. “Groundwater Depletion in India: Social Losses from Costly Well
       Deepening.” Journal of Environmental Economics and Management, 93: 85-100.
Schweitzer, R., Ward, R. and Lockwood, H. 2015. Wash Sustainability Index Tool Assessment: Ethiopia.
       USAID.
Shapley, L.S. 1953. Additive and Non-Additive Set Functions. Princeton University.
Shorrocks, A.F. 1984. “Inequality Decomposition by Population Subgroups.” Econometrica: Journal of
       the Econometric Society, 52: 1369-1385.
Shorrocks, A.F., 1982. “Inequality Decomposition by Factor Components.” Econometrica: Journal of the
       Econometric Society, 50: 193-211.
Shorrocks, A.F., 1980. “The Class of Additively Decomposable Inequality Measures.” Econometrica:
       Journal of the Econometric Society, 48: 613-625.
Stoupy, O. and Sugden, S. 2003. Halving the Number of People without Access to Safe Water by 2015-A
        Malawian Perspective. Water Aid
UN-Water. 2013. United Republic of Tanzania; UN-Water Country Brief. UN-Water
Verplanke, J. and Georgiadou, Y. 2017. “Wicked Water Points: The Quest for an Error Free National
       Water Point Database.” ISPRS International Journal of Geo-Information, 6:.244.
WaterAid Tanzania. 2009. Management for Sustainability: Practical Lessons from Three Studies on the
      Management of Rural Water Supply Schemes WaterAid
White, W.B. and Tourre, Y.M. 2007. “A Delayed Action Oscillator Shared by the ENSO and QDO in the
       Indian Ocean.” Journal of Oceanography, 63: 223-241.
Whittington, D., Mujwahuzi, M.R., McMahon, G. and Choe, K. 1989. Willingness to Pay for Water in
       Newala District, Tanzania: Strategies for Cost Recovery. Wash Field Report, 246
Willetts, J.R., 2012. A Service Delivery Approach for Rural Water Supply in Timor-Leste: Institutional
        Options and Strategy. Institute for Sustainable Futures
Wilson, P., Bonsor, H.C., MacDonald, A.M., Whaley, L., Carter, R.C. and Casey, V., 2016 UPGRO
       Hidden Crisis Research Consortium: Unravelling Past Failures for Future Success in Rural
       Water Supply: Initial Project Approach for Assessing Rural Water Supply Functionality and
       Levels of Performance. UPGRO
Welle, K. 2010. Water Point Mapping–A Tool for Increasing Transparency and Accountability. In IRC
       Symposium.
World Bank, 2008. Implementation Completion and Results; Rep. IDA-36230, IDA-3623A on a Credit in
       the Amount of SDR 20.8 Million (US$26.00 Million Equivalent) to the United Republic of
       Tanzania for a Rural Water Supply and Sanitation Project. World Bank

                                                  24 
 
World Health Organization, WHO/UNICEF Joint Water Supply and Sanitation Monitoring Program
      (JMP), 2015. Progress on Sanitation and Drinking Water: 2015 Update and MDG Assessment.
      World Health Organization.
Zhang, J. and Xu, L.C. 2016. “The Long-Run Effects of Treated Water on Education: The Rural Drinking
        Water Program in China.” Journal of Development Economics, 122: 1-5.




                                                25 
 
                                                       ANNEX
Table 1. Summary of Water Points’ Non-Functionality Rates from Selected Literature Review

Region/Country               Year(s) of Survey         Study/Compilation               Percent Non-
                                                                                      Functional
Global                       2010s                     Akvo (2015)                    20
                             Late 1970s                Cairncross et al. (1980)       30
Developing countries         1990s                     Duti (2012)                    30–40


Asia
    India                    1974                      Mudgal (1997)                  75
    Thailand                 1980s                     McPherson and McGarry (1987)   50


Sub-Saharan Africa           2008                      Baumann (2009)                 30
                             2000s                     RWSN (2009)**                  36
Angola                       2000s                     RWSN (2009)**                  30
Benin                        2000s                     RWSN (2009)**                  22
Cote d'Ivoire                2000s                     RWSN (2009)**                  65
Ethiopia                     2010                      Schweitzer et al. (2015)       43
Ethiopia                     2000s                     RWSN (2009)**                  35
Greater Afram (Ghana)        2010s                     Fisher et al. (2015)           20.6*
Madagascar                   2000s                     RWSN (2009)**                  10
Malawi                       2010s                     Holm et al. (2015)             22*
                             2010s                     Holm et al. (2016)             39*
                             2000s                     RWSN (2009)**                  40
Mozambique                   2000s                     Jansz (2011)                   20
South Africa                 1990s                     Hazelton (2000)                50
Sierra Leone                 2000s                     RWSN (2009)**                  65
Zimbabwe                     2000s                     RWSN (2009)**                  30
                             1989                      Cleaver (1991)                 32


Notes:
*Studies of specific region or states within country
*Statistics compiled from multiple sources/studies




                                                          26 
 
Table 2. Breakdown of Independent Variables by Region for Water Points


                Variables    Frequency    %      Central     Coastal      Lake      Northern    Southern Highlands
                                                Zone ( %)   Zone ( %)   Zone ( %)   Zone ( %)       Zone ( %)


Functionality
    Functioning               29,230     71 %     67 %        67 %        70 %        76 %            75 %
    Non-Functioning           11,687     29 %     33 %        33 %        31 %        24 %            25 %
Groundwater Productivity
    No-Value-Available          546      1%       0%          5%          0%          1%               1%
    Very low                    618      2%       2%          0%          3%          1%               1%
    Low-Moderate              25,560     62 %     89 %        67 %        58 %        48 %            66 %
    Moderate                   8,049     20 %     0%          8%          28 %        29 %            20 %
    High                       5,667     14 %     9%          20 %        7%          22 %            11 %
    Very High                   66       0%       0%          0%          1%          0%               0%
    Null                        411      1%       0%          0%          3%          0%               1%
Groundwater Storage
    No-Value-Available          404      1%       0%          3%          0%          1%               1%
    Low                       18,380     45 %     81 %        30 %        45 %        34 %            54 %
    Low-Medium                14,156     35 %     16 %        13 %        44 %        61 %            28 %
    Medium                     7,691     19 %     3%          53 %        9%          5%              17 %
    Null                        286      1%       0%          0%          2%          0%               0%
Depth to Groundwater
    No-Value-Available          198      0%       0%          1%          0%          1%               1%
    Very-Shallow              27,888     68 %     64 %        80 %        46 %        72 %            81 %
    Shallow                   10,894     27 %     28 %        14 %        49 %        21 %            16 %
    Shallow-Medium             1,934     5%       8%          5%          5%          6%               2%
    Medium                       3       0%       0%          0%          0%          0%               0%
Settlement Type
    Rural                     36,291     89 %     76 %        86 %        92 %        96 %            86 %
    Urban                      4,626     11 %     24 %        14 %        8%          4%              14 %
Ecological Class
    Forest & Woodland          8,393     21 %     1%          33 %        7%          23 %            30 %
    Shrub & Herb Vege         23,733     58 %     46 %        60 %        64 %        49 %            61 %
    Desert & Semi-Desert       5,825     14 %     48 %        2%          20 %        19 %             2%
    Open Rock Vegetation         3       0%       0%          0%          0%          0%               0%
    Water                       389      1%       0%          0%          3%          0%               0%
    Missing                    2,574     6%       5%          6%          6%          9%               6%
Types of Pump
    Manual Hand Pump          11,718     29 %     23 %        33 %        41 %        8%              30 %
    Motorized Pump             7,713     19 %     50 %        37 %        8%          17 %             4%
    Gravity Water Pump        17,486     43 %     19 %        20 %        35 %        70 %            60 %
    Other                      4,000     10 %     8%          10 %        17 %        5%               7%


Promoters



                                                27 
 
    Public                                15,800   39 %    32 %     34 %   50 %   37 %   33 %
    Private                               15,370   38 %    50 %     28 %   30 %   47 %   42 %
    Donor                                 2,249    5%          2%   9%     4%     3%     7%
    Other                                 7,498    18 %    15 %     29 %   15 %   13 %   17 %
Managing Body
    Water Authority Board                 4,692    11 %        4%   3%     11 %   33 %   4%
    Private                               2,541    6%          5%   15 %   3%     8%     1%
    Village Committee                     32,523   79 %    90 %     78 %   81 %   59 %   93 %
    Other                                 1,161    3%          1%   4%     5%     1%     2%
Source
    Gravity Flow System                   18,627   46 %    23 %     21 %   40 %   73 %   61 %
Pump Piped System - Groundwater based     20,304   50 %    77 %     66 %   57 %   23 %   37 %
Pump Piped System - Surface water based   1,883    5%          0%   13 %   3%     3%     2%
    Unknown                                103     0%          0%   0%     0%     1%     0%
Extraction Technology
    Gravity Flow System                   17,492   43 %    19 %     20 %   35 %   70 %   60 %
    Groundwater Pump                      3,548    9%      13 %     12 %   11 %   3%     6%
    Surface water Pump                    6,948    17 %    49 %     37 %   45 %   10 %   23 %
    Other                                 12,929   32 %    20 %     31 %   9%     17 %   12 %
Total                                     40,917

                                                            

 

 

 

 

 

 

 

 

 

 

 

 




                                                          28 
 
Table 3. Type of Water Pumps and Groundwater Productivity (liters/second) for Water Points 20 Years Old or Younger
                            No Values
                            Available       Very Low       Low-Moder      Moderate         High      Very High       Null      Total
Manual Hand Pump                97            258             7,371         2,354          1,433         26          179      11,718
Motorized Pump                 359             75             5,357             370        1,514         6           32        7,713
Gravity Water Pump              53            173             10,233        4,759          2,135         24          109      17,486
Other                           37            112             2,599             566        585           10          91        4,000
Total                          546            618             25,560        8,049          5,667         66          411      40,917
Key: Very Low: <0.1; Low: 0.1-0.5; Low-Medium: 0.5-1; Medium: 1-5; High: 5-20; Very High: >20


    Table 4. Type of Water Pumps and Depth to Groundwater (mbgl) for Water Points 20 Years Old or Younger

                            No Values Available     Very Shallow       Shallow        Shallow-Medium      Medium        Total
    Manual Hand Pump                  45                   7,699        3,511              463                0        11,718
    Motorized Pump                    75                   5,247        1,885              506                0         7,713
    Gravity Water Pump                50               12,476           4,193              764                3        17,486
    Other                             28                   2,466        1,305              201                0         4,000
    Total                             198              27,888          10,894             1,934               3        40,917
    Key: Very-Shallow: 0-7; Shallow: 7-25; Shallow-Medium 25-50; Medium 50-100




    Table 5. Type of Water Pumps and Depth to Groundwater Storage (mm) for Water Points 20 Years Old or Younger

                                     No Values Available       Low      Low-Medium            Medium          Null          Total
    Manual Hand Pump                         74                5,625        3,009                2,890        120          11,718
    Motorized Pump                          244                3,111        1,728                2,621         9            7,713
    Gravity Water Pump                       52                7,839        8,149                1,377        69           17,486
    Other                                    34                1,805        1,270                  803        88            4,000
    Total                                   404               18,380        14,156               7,691        286          40,917
    Key: Low: <1000; Low-Medium: 1,000-10,000; Medium:10,000-25,000




                                                       29 
 
Figure 1. Failure Curve of Water Points in Tanzania
                             .7
                             .6
    Probability of failure
                             .5
                             .4
                             .3
                             .2
                             .1
                             0




Figures 2 - 5. Distribution and Failure Curve of Water Point in Tanzania: Across Settlement Type,
Zones, Promoters, Source and Extraction Technology




                                               30 
 
Figure 6 - 8. Distribution and Failure Curve of Water Point in Tanzania: Across Settlement Type,
Zones, Promoters, Source and Extraction Technology




                                              31 
 
Figures 9-11 Distribution and Failure Curve of Water Points in Tanzania by Hydrological Factors




                                               32 
 
Figure 12. Distribution and Failure Curve of Water Points in Tanzania by Ecological Zones

         Distribution and Failure curve of water points in Tanzania
                                                                                                  by Ecological Zones
                                                                               using definition of functionality based on status variable

                                                                                                                   Distribution of water points by in Tanzania
                                                               Failure curves of water points in Tanzania
                                                                         By Ecological Class
                              0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1
     Probability of failure




                                                                                                                                    6.3%
                                                                                                                                 1.0%
                                                                                                                                 0.0%
                                                                                                                                              20.5%
                                                                                                                            14.2%




                                                                                                                                      58.0%



                                                               0         5        10         15        20
                                                                         Age (in Years)


                                                                         Forest & Woodland                                Shrub & Herb Vegetation
                                                                         Desert & Semi-Desert
                                                                                                                                                                  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 



                                                                                                                    33 
 
Figures 13-15. Curves by Hydrological Factors and Pump Type




                                                               
 
 
 
 
 
 
 




                                             34 
 
Table 6. Regressions on Non-Functionality of Water Points Aged 1-20 (0=functional; 1= non-
functional)
                                                         (1)             (2)            (3)               (4)               (5)
                    VARIABLES                    all water points     1 yr old    2 to 4 yrs old   5 to 10 yrs old   over 11 yrs old
    Age                                              0.016***                       -0.258***          0.079***           -0.024
                                                       (0.002)                       (0.075)            (0.021)          (0.015)
    Age Squared                                      -0.000***                      0.044***          -0.004***         0.001***
                                                       (0.000)                       (0.012)            (0.001)          (0.000)
    Settlement Type
         Urban                                       -0.007            -0.053      -0.034***          -0.010            0.040***
                                                     (0.007)           (0.038)      (0.014)           (0.012)            (0.012)
        (Reference Group: Rural)
    Geographical Zone
        Central Zone                                0.044***          0.184***     -0.078***         0.045***             0.019
                                                     (0.009)           (0.027)      (0.024)           (0.015)            (0.015)
          Coastal Zone                              0.049***             0.027       0.014           0.050***           0.085***
                                                     (0.007)           (0.032)      (0.015)           (0.012)            (0.012)
          Northern Zone                             0.059***            -0.032       0.006           0.078***           0.074***
                                                     (0.007)           (0.027)      (0.016)           (0.012)            (0.013)
          Southern Highlands Zone                   0.015***             0.013     0.033***            0.007              0.018
                                                     (0.007)           (0.030)      (0.014)           (0.012)            (0.011)
       (Reference Group: Lake Zone)
    Managing Body
       Water Authority Board or Government          -0.136***         -0.471***    -0.034***         -0.154***         -0.131***
                                                      (0.009)          (0.106)      (0.016)            (0.014)          (0.015)
          Private                                   -0.095***           0.072      -0.174***         -0.029***         -0.067***
                                                      (0.010)          (0.056)      (0.023)            (0.015)          (0.021)
          Other                                        -0.016         -0.111***      -0.040             -0.048         0.057***
                                                      (0.013)          (0.046)      (0.026)            (0.025)          (0.022)
          (Reference Group: Village Committee)

    Promoter
        Private                                        0.001            0.017        0.015           -0.018***           0.004
                                                      (0.005)          (0.029)      (0.011)            (0.008)          (0.009)
          Donor                                       -0.012                       -0.120***            -0.007           0.010
                                                      (0.009)                       (0.027)            (0.016)          (0.014)
          Other                                     -0.071***           0.045      -0.046***         -0.106***         -0.078***
                                                      (0.007)          (0.025)      (0.014)            (0.014)          (0.011)
        (Reference Group: Public)
    Types of Pump
        Manual Hand Pump                            0.102***          -0.148***     0.132***         0.153***           0.074***
                                                     (0.006)            (0.036)      (0.012)          (0.009)            (0.009)
          Motorized Pump                            0.123***           0.070***     0.137***         0.145***           0.099***
                                                     (0.007)            (0.019)      (0.014)          (0.011)            (0.013)
          Other                                     0.336***          -0.119***     0.223***         0.346***           0.422***
                                                     (0.007)            (0.057)      (0.014)          (0.011)            (0.011)
          (Reference Group: Gravity Flow Pump)




                                                                35 
 
    Groundwater Productivity
         Moderate                                        0.044***           -0.027         0.019           0.065***            0.059***
                                                          (0.007)          (0.025)       (0.015)            (0.013)             (0.011)
         High & Very High                                  0.003         -0.176***        -0.010             -0.007            0.052***
                                                          (0.008)          (0.030)       (0.015)            (0.013)             (0.013)
         No Values Available                               0.018            -0.196         0.011             0.001             0.074***
                                                          (0.020)          (0.182)       (0.034)            (0.035)             (0.032)
         (Reference Group: Very Low, Low & Moderate)
    Groundwater Storage
         Low Medium                                      0.014***         0.287***      0.026***             -0.022              -0.016
                                                          (0.007)          (0.020)       (0.013)            (0.012)             (0.011)
         Medium                                          0.038***            0.052         0.005           0.039***            0.039***
                                                          (0.007)          (0.037)       (0.015)            (0.012)             (0.012)
         No Values Available                             0.090***        -1.687***         0.052              0.092               0.048
                                                          (0.025)          (0.207)       (0.041)            (0.054)             (0.037)
         (Reference Group: Low)
    Depth to Groundwater
         Shallow                                         0.035***           -0.037      0.038***           0.037***            0.048***
                                                          (0.005)          (0.019)       (0.010)            (0.008)             (0.008)
         Shallow Medium                                    0.009           -0.076       0.044***           0.037***               0.015
                                                          (0.010)          (0.048)       (0.019)            (0.017)             (0.016)
         No Values Available                               -0.057         1.744***        -0.016          -0.159***            0.162***
                                                          (0.034)          (0.184)       (0.048)            (0.069)             (0.072)
         (Reference Group: Very Shallow)                                       -
                    Observations                          40,917             2,311        6,875             15,503              16,227
                 Pseudo R-squared                          0.079             0.197         0.077             0.078               0.097
    Robust standard errors in parentheses
    Coefficients reported are average marginal effect/probabilities
    Coefficients missing are omitted because of collinearity, or because there were no observations for that particular age range
    The reference group for each categorical variable is displayed
    * p<0.05, ** p<0.01, *** p<0.001
    Full model Variance Inflation Factor is 2.90
    Notes on recodings:
    Groundwater productivity
    Very Low category is coded together with Low Moderate category due to low sample size
         Null and No Valuess Available are also coded together
    Groundwater storage
         Null and No Values Available are also coded together
    Depth to Groundwater
         Medium is coded with Shallow Medium since the category has only 3 observations




                                                                  36 
 
Table 7. Shapley Values and Groups
                               Hydrological                        Pump              Management
                        Age                   Localities   Zones          Promoter
                                  data                             Type                Body
    All Water Points    16 %       6%            0%        4%      58 %     1%          11 %
        1 year old       n/a      53 %           2%        15 %    12 %     3%          15 %
     2 to 4 years old   1%         8%            2%        5%      54 %     6%          21 %
    5 to 10 years old   3%        10 %           0%        6%      63 %     1%          12 %
    over 11 years old   3%        10 %           1%        6%      70 %     1%          7%


Figure 16. Shapley Decomposition of Water Points’ Failure in Tanzania




                                                                                              
 
 
 
 
 
 
 
 
 
 
 
 
 
 

                                                  37 
 
Figure 17: Geographic Distribution of Water Points in Tanzania by Functionality




                                              38