World Development 103 (2018) 297–310

Contents lists available at ScienceDirect

World Development
journal homepage: www.elsevier.com/locate/worlddev

Are we underestimating urban poverty?

Paula Lucci a, Tanvi Bhatkal b,⇑, Amina Khan a
a
Overseas Development Institute, United Kingdom
b
Overseas Development Institute and University of Cambridge, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: Data collection methods and poverty measures have not caught up with the reality of an increasingly
Accepted 11 October 2017 urbanised world; as a result, urban poverty may be underestimated. This has important implications
Available online 29 November 2017 for targeting interventions and allocating resources in the 2030 Agenda for Sustainable Development.
Several problems affect the measurement of urban poverty: definitions of ‘slum’ settlements vary widely,
Keywords: data collection may undercount slum populations, insufficient data disaggregation may conceal intra-city
Urban poverty disparities, and common indicators and assumptions may be ill-suited to assessing both income and mul-
Slums
tidimensional poverty in urban contexts. However, not enough is known about the extent to which these
Data
Urbanisation
issues affect the resulting estimates. This paper contributes to the existing literature by illustrating the
Measurement scale of the bias associated with common practices in measuring urban poverty at different stages of
Global South the production of poverty estimates. The analysis draws on selected examples in the literature alongside
new analysis of data from Demographic and Health Surveys and Household Income and Expenditure
Surveys. The article also provides recommendations on how to address each of these problems to
improve urban poverty measurement.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction (Gibson, 2015). It is therefore unsurprising that these tools are in

With urbanisation currently accelerating in many countries, it is
becoming increasingly important to raise the profile and improve
our understanding of deprivation in urban contexts.
Over the last two decades, there has been a growing discussion
about the ‘urbanisation of poverty’. Ravallion, Chen, and Sangraula
(2008) made the first decomposition of the international dollar-a-
day poverty estimates by rural and urban location. They high-
lighted that although urbanisation plays a positive role in overall
poverty reduction, the urban share of poverty is rising. Further,
the UN Human Settlements Programme (UN-Habitat) – the only
source of internationally comparable data on slum dwellers – esti-
mates that 881 million people or 30% of developing countries’
urban populations live in slums (UN-Habitat, 2014), and that this
could rise to 3 billion or 60% by 2050 (UN DESA, 2013, 2014). This
rise in the urban share of poverty notwithstanding, data collection
methods and poverty measures have not caught up with the reality
of an increasingly urbanised world. In fact, household surveys – the
main instruments to collect data on poverty – have not changed
much in 30–40 years, when the focus was mostly on rural poverty
⇑ Corresponding author.
E-mail addresses: p.lucci@odi.org (P. Lucci), tb563@cam.ac.uk (T. Bhatkal),
a.khan@odi.org (A. Khan).
many aspects inadequate to account for living standards in an
era of increasing urbanisation.
Many argue that urban poverty is commonly underestimated
(for example, Mitlin & Satterthwaite, 2013; Parnell, 2005; Sabry,
2010; Tacoli, 2007; Thanh, Anh, & Phuong, 2013). Poor urban pop-
ulations, such as those living in informal settlements, are often
undercounted, and the indicators used to measure basic depriva-
tions are not providing policy-makers with the information they
need to formulate and implement policy to tackle urban depriva-
tions (Lucci & Bhatkal, 2014).
Despite well-known measurement problems (Carr-Hill, 2013;
Mitlin & Satterthwaite, 2013), not enough is known about their
scale. Using selected examples from the literature and new analy-
sis of Demographic and Health Survey and Household and Income
Expenditure data, this paper contributes to the existing literature
by estimating the scale of the bias in urban poverty measurement
at different stages of the production of poverty estimates.
While discussions about data may appear very technical, they
are also inherently political and have important implications for
interventions. If current estimates underestimate deprivation in
urban contexts, then governments’ and donors’ priorities and
resource allocations may neglect pockets of deprivation in cities.
A focus on the persistence of deprivations and intra-city inequali-
ties in urban areas is salient in the context of increasing calls for

https://doi.org/10.1016/j.worlddev.2017.10.022
0305-750X/Ó 2017 Elsevier Ltd. All rights reserved.
298 P. Lucci et al. / World Development 103 (2018) 297–310
data-driven policy and progress tracking in international develop-
ment discourse and governmental attention on ‘smart’ cities.
The discussions in this article are relevant to ongoing interna-
tional debates about implementing and monitoring the new Sus-
tainable Development Goals (SDGs), particularly Target 11.1 on
‘ensuring access for all to adequate, safe and affordable housing
and basic services and upgrade slums’, as well as the New Urban
Agenda (the outcome of the 2016 Habitat III conference).1 Our anal-
ysis also speaks to the ‘Leave No One Behind’ agenda, which posits
that progress on the SDGs should include the hard to reach; this
includes marginalised urban communities, such as slum dwellers.
This paper is structured as follows: Section 2 discusses how
estimates of the number of slum dwellers, arguably a high propor-
tion of the urban poor, vary according to the definitions used. Sec-
tion 3 sets out some of the problems with the data that are
currently collected, particularly the undercounting of slum dwell-
ers and the lack of disaggregated data beyond urban averages. Sec-
tions 4 and 5 discuss how commonly used indicators and
assumptions can underestimate monetary and multidimensional
poverty, respectively, in urban contexts. Section 6 concludes by
summarising our findings and providing recommendations to
address the gaps that have been identified.
2. Problems related to definitions of ‘slums’
It is hard to discuss urban poverty without focusing on slums, as
they often include most poor people in cities in the developing
world. The term ‘slum’ has been used to cover a range of deficien-
cies in housing and basic services, and different organisations –
even within a country – often use varying definitions. In this sec-
tion, we show how this variation makes it difficult to measure
the number of people living in these areas.
UN-Habitat has developed a cross-nationally applicable defini-
tion. A set of people living under the same roof in urban areas that
lack one or more of the following are defined to be living in slums
or informal settlements (UN Habitat, 2016):
 access to improved water services which includes piped con-
nection to house or plot; public stand pipe serving no more than
5 households; protected spring; rain water collection; bottle
water; bore hole; and protected dug well
 access to improved sanitation services which includes direct
connection to public sewer; proper flush latrine; pit latrine with
slab; ventilated improved pit latrine; and direct connection to
septic tank
 sufficient living space, with fewer than four people per habit-
able room
 structural quality or durability of housing that is built in a non-
hazardous location and has a permanent and adequate struc-
ture able to protect inhabitants from extreme climatic
conditions
 security of tenure that enables them to live with security, peace
and dignity – this is included in the definition but not in slum
measurement due to insufficient data (UN-Habitat, 2016).
The main advantage of this definition is that it allows for inter-
national comparisons, which are of interest to donors and multilat-
1
Habitat III refers to a global summit, formally the UN Conference on Housing and
Sustainable Urban Development, held in Quito, Ecuador, on 17–20 October 2016. The
UN has called the conference, the third in a series that began in 1976, to ‘reinvigorate’
the global political commitment to the sustainable development of towns, cities and
other human settlements, both rural and urban. The product of that reinvigoration,
along with pledges and new obligations, is referred to as the New Urban Agenda. That
agenda will set a new global strategy around urbanisation for the next two decades
(Citiscope, 2015).
eral organisations as they consider resource allocation across
countries. This was the definition used to monitor the ‘slum’ target
in the Millennium Development Goals (MDG 7, Target 11) and now
for SDGs (Goal 11, Target 11.1) (IAEG, 2016). It also provides a com-
prehensive picture of housing and basic service deprivations in
urban settings, as a household lacking even one of the conditions
described above would classify as a slum.
However, this definition, while comprehensive, is not suffi-
ciently nuanced to distinguish different types of housing depriva-
tions needed to inform policymaking. For example, it does not
include density criteria for a settlement, whereas a high density
of households is a characteristic commonly associated with infor-
mal settlements. This means that, under the UN definition, a singu-
lar household living in a precarious building in the inner city would
qualify as a slum household.
In principle, this would be addressed if the total number of slum
households could incorporate density criteria and the type of
building structure (e.g. to distinguish between a slum settlement
or inner city tenement). Further disaggregation by the number of
deprivations that a household experiences would also provide a
more complete picture of the depth of deprivations. In fact, it is
common for some of these deficiencies to be experienced jointly,
which can make addressing standalone issues (such as drinking
water without sanitation) somewhat problematic.2
Finally, urban deprivations across regional and country contexts
are often distinct, and global generalisations can be unhelpful
(Gilbert, 2013). Thus, while the UN-Habitat definition and numbers
have been used to monitor global voluntary commitments such as
the MDGs and now the SDGs, countries usually deploy their own
definitions of slum settlements in their own planning.
Take the example of India: In 2001, the country identified slums
in the Census for the first time. Slums were identified at the neigh-
bourhood rather than household level, and defined as areas satisfy-
ing any of the following three criteria: (1) all areas in a town or city
notified as ‘slum’ by state or local governments and union territory
administration under any Act, including a ‘Slum Act’; (2) all areas
recognised as ‘slum’ by state or local government and union terri-
tory administration that may not have been formally notified as
‘slum’; or (3) a compact area with a population of at least 300 peo-
ple or around 60–70 households of poorly built and congested ten-
ements in an unhygienic environment, usually without adequate
infrastructure and proper sanitary and drinking water facilities.3
The first two criteria are based on administrative designations
and require official recognition of neighbourhoods as slums by
local or state governments. However, these definitions are inher-
ently arbitrary, as these governments employ different criteria
(MHUPA, 2010) due to varying incentives (for instance, including
slums on official lists has resource implications as municipal
authorities are meant to provide notified and recognised slums
with basic services). The third criterion identifies areas as slum set-
tlements based on measurable attributes, using similar conditions
to the UN-Habitat definition but with the inclusion of a density cri-
terion. These slums are often inhabited by newer migrants, and
generally have poorer access to basic facilities as authorities have
no obligations on provision (IIPS & Macro International, 2007).
However, the definition does not define the attributes it specifies
(e.g. what constitutes an ‘unhygienic environment’) (Patel,
Koizumi, & Crooks, 2014; Risbud, 2010).
2
One other criticism of the slum definition used for the MDG target (and now for
the SDGs) is that it overlaps with the water and sanitation targets (Gilbert, 2014).
3
This definition was amended following from the Pronab Sen Committee
(Government of India, 2010) to reduce the density requirement to 20 households,
making it less restrictive. This definition is used by the National Sample Survey
Organisation in household surveys; however, the 2011 national Census used the 2001
definition.
 P. Lucci et al. / World Development 103 (2018) 297–310
90
80
70
60
50
40
30
20
10 21 50 46 63 35 41 55 20 48
0
Delhi Meerut Kolkata Indore
Census (2001 definion) UN Habitat definion (no tenure)
Fig. 1. Share of population classified as ‘slum’ in Indian cities, by definition, 2005–2006. Source: Authors’ calculation based on IIPS and Macro International (2007).
The neighbourhood approach identifies slum settlements more
clearly. However, it has its own limitations. One criticism refers to
the threshold used: the criterion of a minimum of 60–70 clustered
households fails to recognise smaller slum settlements. These are
usually more recently formed and house particularly disadvan-
taged households (Patel et al., 2014). This is one reason that this
criterion was amended to 20 households following recommenda-
tions by the Pronab Sen Committee (Government of India, 2010).
In fact, given the particular vulnerabilities smaller, recently
formed, settlements may face, it may be useful to consider them
as a separate category (i.e. differentiating settlements by their size
and/or year of formation).
In addition, this approach ignores differences in housing depri-
vation within a neighbourhood, which could be significant. As a
result, better-off households would be classified as slum house-
holds if the majority of their neighbours had inadequate infrastruc-
ture (Patel et al., 2014). To some extent, this is an inevitable
consequence of a neighbourhood-based approach, but there could
be ways around it. For example, the deprivations experienced by
households within slum settlements should be distinguishable.
The way a slum is defined can have a huge impact on estimates
of the number of slum dwellers. For instance, Fig. 1 presents our
calculations of slum estimates using the UN-Habitat and the
national Census definitions for eight Indian cities for which the
Demographic and Health Survey (DHS) (2005–2006) provides rep-
resentative data. In half of these cities, we find that the proportion
of households estimated to be living in slums using UN-Habitat’s
definition4–6 is more than two times higher – and up to four times
higher – than that measured using the definition from the national
Census.
Further, in the two cities that have tenure security data avail-
able (which is included in UN-Habitat’s conceptual definition but
not in its measurement, as there is typically insufficient data) this
share could be even higher, although the difference is not statisti-
cally significant. This suggests tenure insecurity may not overlap
fully with the other deprivations.
There are a number of possible explanations for the large differ-
ence between estimates. First, the Census definition includes a set-
tlement density criterion, whereas UN-Habitat does not, so the
latter definition would include households in deprived smaller
neighbourhoods or inner city tenements. That said, the Census def-
4
The DHS data identified those households in the survey sample that were
designated as slums as per the 2001 Census. We compare these to those recognised as
slum households using the UN-Habitat definition based on available data on housing
conditions and access to basic services.
5
The DHS 2005–06 did not ask how many households shared a water source so we
consider all shared sources as improved.
6 7
Following from WHO accepted standards, we exclude children under 12 months,
and count those between the ages of one and 10 years as half a person.
299
57 64 76 36 55 17 66 19 74
Mumbai Nagpur Hyderabad Chennai
UN Habitat definion (including tenure)
inition could misclassify better-off households in slums as slum
households. Further, the way adequate sanitation is defined is
likely to vary across the two, with UN-Habitat having a more ambi-
tious definition. Finally, the Census definition is more subjective
and requires recognition from government, which could give them
an incentive to undercount slums (Agarwal, 2011).
3. Problems with the underlying data
This section highlights two problems with the underlying data
used to generate urban poverty estimates. First, we describe how
undercounting of slum dwellers can happen for several practical
and political reasons, and, more fundamentally, because of unrep-
resentative sampling frames. Secondly, we discuss how the lack of
granular data on slums and reliance on urban averages conceals
intra- and inter-city disparities. In both cases, we provide new
analysis to estimate the scale of the problem.
3.1. Undercounting
Household surveys provide the data used to estimate income
and multidimensional poverty in urban areas, and the number of
slum dwellers (Census data are also used for the latter). Yet these
sources can undercount marginalised urban populations like those
living in informal settlements (Mitlin & Satterthwaite, 2013). The
homeless, those living in their workplace and mobile groups like
seasonal and temporary migrant workers, who often live in infor-
mal settlements, are often missed from the data (Carr-Hill, 2013;
Deshingkar, 2006; Sabry, 2010).
3.1.1. Populations missing for practical and political reasons
There are practical reasons why household surveys may under-
estimate the poverty rates in slum areas. For instance, certain areas
may be included in sampling frames but in practice may be missed
or not thoroughly covered by enumerators because they appear
hostile and unsafe or are hard-to-reach – for example, places
where water is dirty, defecation is out in the open, sewers are
uncovered or have reached capacity, and sanitation and hygiene
are low.7
In addition, a large number of the urban poor often live on the
outskirts of the city, particularly around megacities, and may be
counted as ‘‘rural” (Sabry, 2010). This concern is salient in the con-
text of urban development in megacities of the global South that
has displaced poorer populations to the outskirts, either as a result
Skype interview with Erica Hagen, Ground Truth Initiative/Map Kibera, December
2015.
300 P. Lucci et al. / World Development 103 (2018) 297–310
of market forces (Roy, 2011; Yntiso, 2008) or resettlement due to
urban redevelopment policies (Desai, 2012; Roy, 2005).
Political incentives also sometimes render populations in slum
settlements invisible in the data, particularly when politicians
and other government officials frame definitions of slums in line
with their political motivations. Thus, Yelling (1986: 1) declares
that a slum is ‘‘a term in the discourse of politics rather than scien-
ce”. If public land or land owned by politicians is occupied by slum
dwellers under tacit informal arrangements that do not appear in
any official documents, rising land values or disputed land owner-
ship can motivate politicians to use slum dwellers’ invisibility to
evict or resettle them forcefully and without compensation.8
Of course, the picture is nuanced and in some instances govern-
ments have the opposite motivation. For instance, governments
seeking slum dwellers’ political support may have an incentive to
make them visible and address their needs,9 while local govern-
ments appealing to higher levels of governments for resources may
face incentives to over-count if resource allocation is linked to the
number of slum dwellers being reported. Moreover, changing polit-
ical priorities may lead to policy inconsistencies over time and a
changing use of politicised slum numbers (Ghertner, 2010; Sheikh
& Banda, 2015).
Slum dwellers too may have competing motivations. They may
insist on being counted so they can then exert pressure on govern-
ments to respond to their needs, as demonstrated by the many
enumerations carried out by slum federations (Patel, Baptist, &
D’Cruz, 2012). On the other hand, they may wish to be invisible
if they fear eviction.
3.1.2. Populations missing because of unrepresentative sampling
frames
Survey sampling frames are based on Census data, and there-
fore will replicate any biases in them (Carr-Hill, 2013; Lucci &
Bhatkal, 2014). Although Censuses are meant to enumerate the
entire population of a country by design, they too may leave out
certain populations for the same reasons described above in the
case of household surveys (Carr-Hill, 2013).
More fundamentally, Census data are collected only every 10
years; this means that, in places marked by rapid urbanisation
where the population of informal settlements is changing, Census
data can very quickly become out of date. For example, a World
Bank poverty assessment conducted for Cambodia in 2006 voiced
the concern that informal settlements had grown since the sam-
pling frame was constructed in 1999 and therefore the poverty fig-
ures might understate their scale (World Bank Cambodia, 2006 in
Mitlin & Satterthwaite, 2013). This problem can be more acute in
some countries, particularly conflict-ridden ones, where Censuses
have not been carried out for over 10 years. For instance, the latest
Census in Pakistan will be conducted in 2017, after a gap of 19
years.
3.1.3. The scale of undercounting: selected examples
By definition, it is difficult to quantify the scale of undercount-
ing, and most sources and estimates produced at international and
national levels likely face this limitation to some extent. Using esti-
mates as in Carr-Hill (2013), we find that, if one in every 10 slum
dwellers or one in every five slum dwellers is uncounted in global
slum population figures, the number of slum dwellers being left
out of Censuses and sampling frames of household surveys ranges
8
Personal interview with Arif Hasan, Chairperson, Orangi Pilot Project-Research &
Training Institute, Karachi, November 2013.
9
Skype interview with Erica Hagen, Ground Truth Initiative/Map Kibera, December
2015.
from 88 to 176 million people.10 This is a sizeable number which
underscores how large the problem of undercounting may be
globally.
At a micro level, contestations over the official and unofficial
numbers of people who live in Kibera in Kenya illustrates the
potential for undercounting. Kibera consists of 13 villages; it is 5
km southwest of the centre of Nairobi and covers about 2.5 km2.
By some population estimates, it is ‘the largest slum in Africa’;
by others it is second to Soweto in South Africa (Marras, 2009).
Yet available data on its population are quite unreliable and highly
contested: we simply do not know its ‘true’ population is, or the
rate at which it is growing.
Recent population estimates range from about 170,070 (KNBS,
2009) to 270,000 (MKP, 2009). In 2008, the Map Kibera Project
(MKP) team conducted a door-to-door census of the population in
Kianda (one of the 13 villages) alongside a mapping of its physical
features. Extrapolating on the basis of Kianda’s population
(15,219),11 the MKP estimated Kibera’s total population to be
between 235,000 and 270,000 (Marras, 2009; MKP, 2009). In 2009,
the KEYOBS RESPOND project12 used satellite images of built struc-
tures in Kibera to estimate population per structure and reported that
the number ranged between 199,959 and 205,108 (Maron, 2010).
Both the MKP and the KEYOBS RESPOND project have published their
methodology, and their estimates are perceived as credible (Maron,
2010). We contrast Census numbers with those from these two
sources; if we regard the latter as realistic estimates of Kibera’s popu-
lation, then the 2009 Census (170,070) undercounted the population
of Kibera by 30,000–100,000 people or by 18–59% (see Table 1).
3.2. Lack of granularity
Another problem in using household surveys to estimate urban
poverty is that sample sizes are often too small. This means break-
downs are available only for broad geographical areas, such as
rural/urban areas or regions (Lucci & Bhatkal, 2014), and not for
cities, let alone slum areas. This means, there is no information
about development outcomes and access to services among mar-
ginalised groups within cities or about differences between cities.
We show that this lack of granularity matters by using DHS data
to estimate the difference in deprivation levels within cities. In par-
ticular, we compute the Multidimensional Poverty Index (MPI)13
for five Indian cities that are projected to be among the top 30 largest
urban agglomerations globally in 2030 (UN DESA, 2014) and for
which city level data are available. This exercise demonstrates that
the poverty headcount ratio for slum households ranges from
between nearly double to four times the ratio for non-slum house-
holds (Fig. 2). This illustrates how city averages conceal differences
within cities, and that these gaps vary considerably between cities.
10
We updated Carr-Hill’s (2013) estimates of slum dwellers being left out of
Censuses and out of sampling frames of household surveys by using latest UN-Habitat
figures (2014).
11
Given that the type, dimension and distribution of the buildings observed in
Kianda are typical of all of Kibera, the MKP team estimated the population of the
entire slum by multiplying the population density found in Kianda (95,120 people per
km2) by the area of Kibera (2.3–2.5 km2), while factoring in an estimated error of 7%
(Marras, 2009; MKP, 2009).
12
The RESPOND Humanitarian Global Mapping Services Project was part of the
Copernicus programme of the European Commission (EC). It aimed to use geographic
information to improve the work of the European and international humanitarian
community. KEYOBS, a Belgian enterprise, was one of the project’s partners.
13
This analysis draws on city-level data from the latest Demographic and Health
Surveys. The MPI considers outcomes on three dimensions: health, education and
living standards. A household is considered poor if it is deprived in one third or more
of the weighted indicators considered (details on indicators in Table A1 in Annex 1).
Note that these shares are likely to be underestimates as some of the indicators
included in the MPI may not always be appropriate for urban areas (discussed in
Section 5).
 P. Lucci et al. / World Development 103 (2018) 297–310 301

Table 1
Contestations over Kibera’s population.

Type of instrument to collect data Source and type of source Year Slum population estimates Differences between Differences between
official and alternative official and alternative
sources sources in percentage
Kenya Population and Housing KNBS, Government 2009 170,070 N/A N/A
Census
Satellite image analysis of built KEYOBS RESPOND, Alternative 2009 199,959–205, 108 30,000–35,000 18–21%
structures in Kibera
Door-to-door Census; mapping MKP, Alternative 2008 235,000–270,000 65,000–100,000 38–59%
of physical features of Kianda village

Sources: Maron (2010), Marras (2009), Marx, Stoker, and Suri (2013), MKP (2009).

Non-slum Slum
35%
30%
30%
Share of populaon (%)

25%
25% 23%

20% 18%
16%
15%
10%
10% 8% 8%
7% 7%

5%

0%
Delhi Kolkata Mumbai Hyderabad Chennai

Fig. 2. MPI headcount ratio by type of residence for selected cities, % of population (2005-06). Note: More details included in Table A3 in Annex 1. UN-Habitat’s definition is
used to identify slums. Note that all the issues discussed in this paper (undercounting, including an outdated Census sampling frame, and the appropriateness of indicators –
including those within the living standards dimension of the MPI, which we discuss in Section 5) also apply to these numbers. Source: Authors’ calculation based on IIPS and
Macro International (2007).

4. Problems with monetary poverty measures in urban contexts
In addition to data coverage, the assumptions and methods
used to calculate poverty also impact estimates. In this section,
we focus on some problems related to calculating income/
expenditure-based poverty in urban contexts.
Poverty levels are typically estimated by counting the number
of people who fall below a common threshold (or poverty line).
Drawing on household surveys, governments set a minimum level
of welfare based on income or expenditure, below which individu-
als are classified as poor.14 The most commonly used method to
establish this minimum poverty line is to calculate the cost of a ‘min-
imum food basket’ needed to achieve a minimum calorific intake
(around 2100 calories). Typically, the food basket is constructed
based on the consumption of a reference group of low-income
households with some allowances made for expenditure on non-
food items. Often, two poverty lines are calculated: a lower one
focused on the non-food spending of those households whose total
spending equals the food poverty line and an upper poverty line
which uses a higher non-food allowance calculated from the food
budget share of households whose food spending exactly meets the
food poverty line (Gibson, 2007). Price data needed to calculate the
costs of the minimum consumption basket are gathered through
either the same survey or the Consumer Price Index (CPI).15 These
poverty lines are used to monitor poverty trends, target resources
14
Often the consumption or income an individual needs is adjusted based on their
age and household size.
15
In addition, the World Bank sets an international poverty line based on the
national poverty lines of 15 of the poorest countries. This was updated in 2015 to
$1.90 in 2011 Purchasing Power Parity (PPP) terms – to make comparisons across
countries. These are used to monitor trends for international commitments to reduce
extreme poverty and by donors and international organisations to monitor progress
across countries and identify areas of highest need.
where poverty is highest and evaluate the impacts of interventions
or of external shocks. If the underlying assumptions are more
attuned to rural contexts, then the estimates produced for urban
areas could be misleading.
To compute a poverty line, it is necessary first to identify a min-
imum basket of goods and services that separate the poor and the
non-poor and then to price that basket. Here, we focus on how the
methods used to estimate income poverty may misrepresent urban
poverty in each of these stages. First, we discuss how food and non-
food allowances are accounted for, arguing that this may misrepre-
sent consumption patterns of the urban poor. Second, we consider
how prices for a minimum consumption basket are derived, show-
ing that this may not take into consideration differences in the cost
of living between urban and rural areas, and cities of different
sizes.
4.1. The composition of food and non-food allowances in the poverty
line
4.1.1. Outdated questionnaires do not reflect current food consumption
habits
Urbanisation can alter diets and eating habits, and this may not
be properly captured by questionnaires gathering the information
on food expenditure needed to estimate poverty. As Gibson (2015)
points out, many Household Consumption Expenditure Surveys
(HCES) record food consumption by providing long lists of ingredi-
ents such as rice, wheat and flour and asking how many units are
consumed and how much money is spent at the household level,
assuming households eat communally. While this made sense in
the early 1990s when the majority of the poor population was still
rural and likely to eat meals together, it is less appropriate for the
urban poor, who are more likely to eat independently of other
302 P. Lucci et al. / World Development 103 (2018) 297–310
household members, either buying street food or purchasing meals
to eat at home.
Consequently, there are two related problems. The first is that
survey questionnaires have greater detail for food ingredients than
purchased meals. For instance, Dupriez, Smith, and Troubat (2014)
show in their analysis of HCES interview questionnaires that the
average number of groups in the food list is 110 but an average
of just three of these are for meals and other forms of food eaten
away from home. Along similar lines, Bangladesh’s Income and
Expenditure Survey in 2010 shows 129 categories for ingredients
compared to 13 for dining out (authors’ analysis of Bangladesh
Bureau of Statistics, 2011). Moreover, information on quantities
is often missing, rendering it impossible to translate this informa-
tion to calories consumed (Gibson, 2015).
The second problem is that this information is often recorded at
the household level rather than for each person. The latter would
be more appropriate for urban contexts in which household mem-
bers may eat outside the house. Reinforcing this point, Beegle, de
Weerdt, Friedman, and Gibson (2012)’s survey experiment in Tan-
zania shows that urban households report 29% lower consumption
if surveyed with a household-level diary rather than a personal
one, while there was no difference for rural areas. The headcount
poverty ratio was over 10 percentage points lower if a personal
diary rather than a household one (with frequent visits) was used.
The general implication is that levels of consumption in urban
contexts are likely to be higher than currently reported, and that
current methods of accounting for food allowances may overesti-
mate urban poverty. A further argument to this point is that in
urban areas with greater access to a variety of food, food baskets
can be changed to minimise expenditure without compromising
calorific intake. Thus, seasonal factors inherent in rural food con-
sumption may be much weaker in urban areas. However, since
all poverty estimates use fixed food baskets, this factor is ignored.
Urban poverty would perhaps be lower if this were considered.
4.1.2. Non-food allowances, central to urban poverty, are inadequately
covered
Another area where income poverty measures may misrepre-
sent urban poverty is in their treatment of non-food allowances.
While the monitoring of food purchases is a core part of all surveys,
questionnaires vary in their inclusion of other important consump-
tion items, such as housing, health, education, energy and water
(Chandy, 2013). Evidence suggests many such items are likely to
be more expensive for urban dwellers, such that excluding them
may underestimate urban poverty (Sabry, 2010).
In urban contexts, meeting non-food needs, such as housing
(either rented or self-built) and other basic services (e.g. paying
for safe water and toilets) can be very costly. Housing in particular
can represent a high share of the urban poor’s budgets but it is dif-
ficult to measure, particularly for owner occupiers (e.g. the infor-
mation needed to measure it is often missing). In fact, in some
cases, because surveys do a poor job of accounting for the con-
sumption of housing services, it is dropped from analyses
(Gibson, 2015). It is more straightforward to account for the costs
facing tenants, which are of course much higher in urban areas,
particularly large cities.
Similarly, poor households often pay higher rates to purchase
water from informal providers. For instance, according to Citizen
Report Card (CRC) research in Kenya, urban households that
obtained their water from a kiosk paid between two to five times
more per unit of water than those that received water through
the network (Twaweza, 2010; WSP, 2007). Sometimes the situa-
tion is even worse, for instance in Nairobi’s Kibera slum, where
households paid up to 30 times more per unit compared with what
middle- and higher-income residents paid (ibid.).
Transport can also represent a substantial share of the urban
poor’s budgets. In Harare, the urban poor spend more than a quar-
ter of their disposable income on transport; in Kampala, they
spend almost half (World Bank & IMF, 2013). Other costs, such as
health care and medicines, school fees and school materials and
fuel, can also be high in some contexts. Of course, rural households
face many of these costs, but because of the concentration of peo-
ple and high demand in cities, particularly large ones, these tend to
be higher in urban areas.
Further, a study by Chibuye (2011) calculated the costs of food
and non-food needs in a series of cities in Zambia and compared it
with the expenditure used in the official figures in 2006. Her calcu-
lations for a poverty line vary from the official figures in that she
includes housing and more generous, realistic allowances for fuel,
soap, electricity and water. Her results showed non-food needs in
Lusaka that were 10 times higher (K996,100 or approximately
$1.50 a day for a household of six) than that in official figures
(K88,709, approximately $0.13 per day); much of this increase
owed to housing needs. Her estimates also show the differences
in expenditures between larger and smaller cities.
Our own analysis of Household Income and Expenditure Survey
for Bangladesh 2010 (Bangladesh Bureau of Statistics, 2011) also
illustrates how urban poor households, particularly in metropoli-
tan areas, spend a higher proportion of their total expenditure on
housing, electricity and transportation (Table 2). For example, poor
households in metropolitan areas spend on average a share of 10%,
9% and 15% of their total spending on rents, electricity and trans-
port fares compared to 2%, 3% and 8% nationwide.
Given that poverty lines do not account properly for non-food
costs, Mitlin and Satterthwaite (2013) argue that the allowances
made for non-food items are simply too small, meaning the pov-
erty line is set too low. A common way of calculating the cost of
non-food items is estimating how much those on/or close to the
poverty lines spend on these items. However, there are questions
as to whether taking the poor as the reference group for non-
food spending is appropriate, since they are unlikely to have the
resources to afford essential non-food items. Sabry (2009) consid-
ers what the non-food allowance estimated by the World Bank
would buy in Greater Cairo, Egypt and finds that it would not be
enough to cover even very inadequate housing, education and
transport. In the case of India, Chandrasekhar and Montgomery
(2010) conclude that the urban poverty line would require signifi-
cant upward adjustment to account for costs of adequate housing.
4.2. Accounting for price differentials
The methods used to account for differences in the cost of living
may also bias the measurement of urban poverty. Prices are
needed to place a monetary value on the food basket and for
non-food needs where allowances for the latter are made. Even
alternative methods to calculate a poverty line that require less
information, such as the Food Energy Intake method,16 depend on
price information to adjust for differences in nominal expenditures
between different areas. In addition, a price index is needed to calcu-
late the change over time in the cost of reaching a poverty line
(Gibson, 2007).
There are, of course, variations in costs and prices across differ-
ent locations, and therefore spatial price deflators are needed. As
information on the composition of a minimum consumption bas-
ket is based on a reference group of low-income households, if dif-
ferences in costs of living between areas are ignored then
16
The food energy intake method is used when price information is unavailable, and
involves plotting expenditure (or income) per capita against food consumption (in
calories per person per day) to identify the expenditure threshold at which a
household acquires enough food (Haughton & Khandker, 2009).
 P. Lucci et al. / World Development 103 (2018) 297–310
Table 2
Share of total household expenditure on selected non-food expenses in Bangladesh (2010).
Area Rent
* *
Poor Average
Rural 0% 1%
Urban 1% 6%
Metropolitan 10% 21%
Total 2% 5%
Notes:
*
The poverty identification is based on the ’upper poverty line’ which provides for a non-food allowance equal to the average non-food expenditure of households whose
food consumption was equal to the food poverty line. In Bangladesh in 2010, this was separately produced for rural, urban and metropolitan areas of each division.
**
Transport fares includes household expenditure on bus, rickshaw/van, taxi/tempo/mishuk, boat/launch and train fares.
Source: authors’ calculation based on Bangladesh Bureau of Statistics (2011).
households from areas where prices are high, typically larger cities,
are less likely to be included in the reference group (Gibson, 2007).
Yet poverty lines do not always reflect these differences in
prices. From an analysis of poverty lines in 53 countries, Mitlin
and Satterthwaite (2013) found that 12 countries did not make
any differentiation based on location. In seven it was not feasible
to ascertain the approach taken to deal with price differentials;
nine differentiated between urban and rural only; and about half
of them (25) followed a more complex differentiation of prices
(e.g. between urban centres of different sizes).
Furthermore, differences between prices may be based on dif-
ferentials in food prices, whereas spatial differences for non-food
prices may be higher (Hentschel & Lanjouw, 1996). There is also
a question about the availability of data for different items – for
example most of the data on non-food items focus on goods and
services found in shops or markets, with little information on
transport, health care, water bills, housing and schooling expendi-
tures, which are all likely to be higher in urban areas, particularly
in larger cities. The CPI is based on average consumption, so there
is also a question of whether it represents a typical poor person’s
expenditure.
Price information can be sourced from expenditure surveys
when these include information on quantities in addition to total
expenditure, but most commonly a price index like the CPI is used.
It is rare for the CPI to detail geographical differences, even in
developed countries like the UK or New Zealand, let alone in devel-
oping countries, where price information is more incomplete
(Gibson, 2007). Jolliffe (2006) shows how adjusting poverty mea-
sures in the US to account for cost-of-living differences between
metro and non-metro areas reverses the common finding that pov-
erty is higher in non-metro areas. In fact, the author finds that
metro poverty is greater than non-metro poverty in terms of
prevalence, depth and severity over the entire period under study
(1991–2002). The author argues that this has implications for the
allocation of social assistance funds, which have tended to focus
on non-metro areas.
It is important to highlight that in some cases the CPI draws
more heavily on information in urban areas or big cities in different
regions; in these instances, there may be an urban bias in the
prices used (Chandy, 2013). In cases where poverty lines are con-
structed using prices with an urban bias, the price information is
less likely to underestimate costs in urban areas (and to overstate
them in rural ones). Further, the extent to which the price informa-
tion draws on data from bigger and smaller urban areas could
impact poverty numbers in larger and smaller cities.
The World Bank’s international poverty line also often uses the
CPI at a national level to adjust changing prices over time to a
specific base year. These adjustments use urban and rural relative
prices only in the case of large countries like China and India; for
the others, national-level CPIs are used (World Bank, 2015). Fur-
ther, the price information used for conversions of poverty lines
into a PPP US dollar, the International Comparison Program (ICP)
303
Electricity Transport fares**
Poor Average Poor* Average
1% 1% 7% 5%
5% 3% 7% 4%
9% 4% 15% 6%
3% 2% 8% 5%
data, is also likely to have an urban bias for some big countries like
China (Ravallion, 2014). In fact, for the 2005 round of PPPs (but not
the most recent round), Chen and Ravallion (2008) used adjust-
ments to address urban bias. This adjustment reduced poverty in
China by nearly half in 2005, from 26.4% to 15.6%.
Adjustments also occur at the national level, as in Latin Ameri-
can countries, where urban bias is likely. The Socio-Economic Data-
base for Latin America and the Caribbean (SEDLAC) database scales
up all rural incomes by 15% to account for urban–rural price differ-
ences (SEDLAC (Socio-Economic Database for Latin America, n.d.).
Views on the extent of urban bias in the more recent 2011 ICP
round differ, with some (World Bank, 2015) suggesting it was
much better at capturing rural prices than the 2005 round, partic-
ularly with respect to China, India and Indonesia (Jolliffe & Beer
Prydz, 2015).
Ultimately, to make more conclusive points about the impacts
of the price information used in urban poverty measurement, fur-
ther work would be needed to consider in more detail the informa-
tion used in price surveys in different countries and the extent to
which it represents sub-national heterogeneity. But the main point
remains: how urban (including differences between large and
smaller cities) and rural prices are treated and whether or not they
are adjusted can make a big difference to the resulting estimates.
A World Bank assessment for Ethiopia (2005, cited in Mitlin &
Satterthwaite, 2013) provides examples of how poverty estimates
can vary if locally specific poverty lines are used. While the govern-
ment estimated urban poverty at 33% in 1995 and 37% in 1999, the
use of specific poverty lines that account for price differentials in
different locations and household substitution of food bundles in
response to price changes puts these numbers at 32% and 46%.
The use of an upper urban poverty line puts these figures at 47%
and 70%. In this case, more detailed assessments of food and
non-food costs of living in different areas resulted in an increase
in the numbers of urban residents estimated as having consump-
tion below the poverty line.
In short, monetary poverty measures can misrepresent urban
poverty, but the direction of the bias is less clear. While the way
food allowances are accounted for may overstate urban poverty,
the incomplete accounting of non-food allowances is likely to
underestimate it. Further, more granular price information would
be needed to account for costs of living in different areas, particu-
larly between larger and smaller cities.
5. Problems with indicators in multidimensional measures in
urban contexts
Since the 1990s, the idea that poverty is multidimensional has
become widely recognised, with the UN and its agencies introduc-
ing wider interpretations of poverty. In principle, such multidi-
mensional measures of poverty should address some of the
shortcomings of monetary poverty indicators in urban contexts,
particularly limitations relating to capturing non-food needs by
304 P. Lucci et al. / World Development 103 (2018) 297–310

measuring outcomes in areas like housing, education and health as the cost of accessing water can sometimes constrain poor urban
directly. Yet some indicators used in common multidimensional dwellers.
measures may not be the most appropriate for dense urban con- For instance, a CRC exercise in Kenya in 2007 found that more
texts. To a large extent, the reason for the choice of these indicators than a third (36%) of poor households relied on water kiosks for
is the availability of comparable cross-national data. their water and typically made four to six trips a day to fetch water
In this section, we focus on indicators of access to basic infras- which, even when nearby, consumed a considerable amount of
tructure services and decent housing, which are salient depriva- time (Twaweza, 2010; WSP, 2007). In Nairobi, households spent
tions in urban contexts and constitute UN-Habitat’s definition of on average 54 min going to the kiosk in normal times, and more
a ‘slum’. Of course, several other dimensions of poverty – such as than twice that in times of water scarcity, while in Kisumu the sit-
education, health, and decent work (as the urban poor generally uation was worse: households relying on water kiosks spent
suffer from poor working conditions in the informal sector; Chen almost two hours (112 min) collecting water every day, and more
2005) – are also relevant and, crucially, are highly unequally dis- than three hours in times of scarcity (ibid.).
tributed within and between urban areas, but this lies beyond Similarly, the frequency of service poses a challenge. A survey of
the scope of the present paper.17 In any event, indicators of depriva- slums in Mumbai found that while only 4.5% of households needed
tions in dimensions such as health and education are less likely to to travel or wait 30 min or more to get water from a standpipe, an
require differentiation in urban and rural contexts. additional 17% could access the standpipes for less than two hours
a day (Bag, Seth, & Gupta, 2016).
Further, as discussed, accessing basic services such as water is
often expensive but the existing indicator does not consider afford-
5.1. Access to basic infrastructure services
ability. For instance, research across four cities – Harare (Zim-
babwe), Windhoek (Namibia), Blantyre (Malawi) and Dar es
Access to safe drinking water and adequate sanitation is often
Salaam (Tanzania) found that accessing 50 litres per person per
measured as the proportion of people that have access to ‘im-
day of municipal water can cost households between 11% and the-
proved’ facilities as defined by the WHO/UNICEF Joint Monitoring
oretically over 100% of their income (Mitlin & Walnycki, 2016).
Programme (JMP). Based on the JMP categorisation, the following
In the case of sanitation, the JMP identifies as improved sanita-
are considered as improved drinking water source: piped water
tion a flush toilet to piped sewer systems, septic tank or pit latrine;
into dwelling or yard/plot; public tap or standpipe; tubewell or
ventilated improved pit latrine; pit latrine with slab and compost-
borehole; protected dug well; protected spring; rainwater and bot-
ing toilet (WHO/UNICEF, 2015). Unlike drinking water, the JMP
tled water if water for non-drinking purposes is from an improved
definition of ‘improved’ access excludes facilities shared by two
source (UNICEF and World Health Organization, 2015).
or more households. This definition varied from that used by UN-
This means shared sources – such as a public tap – are consid-
Habitat to identify slum households, which does not impose
ered to be ‘improved’, which neglects the prevalence of higher
restrictions and considers shared facilities as ‘improved’ (UN-
demand and overcrowding in urban areas, particularly in dense
Habitat, 2016). Having said this, UN-Habitat previously only
settlements, where a public tap may be shared with hundreds
included those that were private or ‘shared by a reasonable num-
(Tacoli, McGranahan, & Satterthwaite, 2015). As a result, household
ber of households’ (UN-Habitat, 2003).
members may have to queue to access drinking water, spending
While, in principle, private facilities should be the standard to
considerable periods of time waiting for and carrying water, result-
aspire to, experts argue that in the short term, it is an unrealistic
ing in unsatisfactory access and inadequate supply.
expectation. Moreover, there are good examples of hygienic shared
Yet the majority of urban households that access improved
toilets (Mitlin, 2015). Community-driven sanitation solutions in
facilities have shared access. Of the 139 countries considered to
dense urban settlements provide an opportunity to come up with
have met the MDG target for access to water between 1990 and
the most appropriate design and technology, although they pose
2015, in over a quarter of them, less than half of the population
considerable challenges in implementation (McGranahan and
accessed piped water to their premises (WHO and UNICEF, 2015).
Mitlin, 2016).
For instance, in Nigeria, only 3% of the urban population had a
Currently, shared facilities are classified as ‘unimproved’ as they
water connection on their premises, although 81% was considered
are shared, suggesting these definitions need to be revisited (ibid.;
to have access to improved water.
WHO and UNICEF, 2015). Ideally, data and criteria would distin-
There has been some recent progress in measurement recognis-
guish between shared facilities that are hygienic and adequate
ing this problem. In 2008, the JMP introduced a four-rung ladder to
and those that are not, but definitions currently used are not
track progress on these indicators in a more refined manner. In the
nuanced enough to allow for this type of distinction.
case of water, the JMP identifies ‘piped water on premises’ and
Other facilities considered ‘improved’ may not necessarily be
‘other improved drinking water sources’ as two categories of
adequate for good health in dense urban settlements. For example,
improved water sources (UNICEF and World Health Organization,
toilets that connect to a septic tank or pit latrines may work well in
2015). In addition, it specifies that an ‘improved’ source ‘‘by the
many rural contexts or low-density urban areas where they can be
nature of its construction and when properly used, adequately pro-
emptied regularly and safely or there is space to build additional
tects the source from outside contamination”. UN-Habitat uses a
pits. However, they may be poor alternatives in dense urban con-
similar definition in its criteria to identify slum households, but
texts (Graham & Polizzotto, 2013). For instance, pit latrines in
has now proposed considering water from a public tap as improved
urban areas fill up quickly, and maintenance can be difficult
only if it is shared by five or fewer households (UN-Habitat, 2016).
(Sutcliffe & Bannister, 2014). In many slum settlements, when they
Nonetheless, neither of these revised definitions consider
are not regularly re-sludged, storage can fill up, rendering them
whether and the extent to which people can access water in prac-
unusable – and harming health outcomes. The most effective sys-
tice. The amount of water and times at which it is available as well
tem for collecting and disposing of waste in high-density urban
contexts is a sewer system, with wastewater treatment
17
Another area which is important for poverty measurement that is beyond the (Satterthwaite, 2014). Local, decentralised sanitation solutions
scope of this paper is subjective well-being and perceptions of poverty (Ravallion,
2012; Pradhan and Ravallion, 2000; Kingdon & Knight, 2006). Multidimensional
have worked well in some circumstances to address environmental
poverty indicators that incorporate subjective well-being offer to consider and health challenges, but these have largely taken place on a small
prioritise dimensions of well-being that poor people value.
 P. Lucci et al. / World Development 103 (2018) 297–310
scale and require further research to be measured and classified as
‘improved’ sources.
Fig. 3 illustrates the difference that varying the sanitation indi-
cators can make to the measures of access to sanitation in urban
areas. We consider four different categories: flush latrines to piped
sewer systems (not shared); the current JMP definition; flush latri-
nes to piped sewer systems allowing for shared facilities; and mod-
ified JMP (allowing shared facilities with up to five households).
Differences in access can be sizeable depending on the criteria
used. For instance, in Kolkata in India 47% of people were identified
as having improved sanitation facilities as per the JMP definition,
but only 18% had individual flush latrines to a piped sewer system
in 2005-06. In Kinshasa, these shares were 20% and 0% in 2014. It is
worth mentioning that, with sanitation, there is a distinction
between access and use, with the latter influenced by behavioural
and attitudinal barriers (the DHS asks households what type of toi-
let facility household members usually use).
Compared with the current JMP definition, if we consider ‘lim-
ited sharing’ as improved, a significantly higher proportion of
households are counted as using adequate sanitation facilities.
However, when excluding pit latrines and septic tanks – which
are often inappropriate for urban settings – from the measure of
‘improved’ sanitation, the proportion of households with improved
facilities is lower in most cities. Therefore, in some regards, the JMP
definition seems too narrow for the urban context (i.e. by not
allowing any sharing of facilities); on the other hand, in some
respects it is too wide (i.e. by including pit latrines as improved).
While the net effect would vary based on local circumstances, indi-
cators should be more nuanced since these differences have varied
policy implications.
5.2. Access to decent housing
Housing indicators used to assess well-being do not always con-
sider the range of deprivations city dwellers experience. The MPI
considers a household deprived in housing if they had dirt, sand
or dung floors (the restrictive choice owed to the limited data that
were available to construct a cross-national measure).
However, this may underestimate poverty by failing to consider
precarious multi-storey dwellings, common in slum settlements
(Satterthwaite, 2014). There may be concerns regarding the quality
of the walls and roofs, which may not protect people from adverse
environmental conditions.
100%
Share of populaon (%)
75%
50%
25%
0%
Delhi (2006) Kolkata (2006) Mumbai (2006)
JMP definion (not shared)
Flush or pour flush to piped sewer system (not shared)
Fig. 3. Share of population with improved sanitation in selected cities (%). Note: *Modified JMP includes improved facilities shared by up to five households as ‘improved’.
Sources: Authors’ calculation based on IIPS and Macro International (2007), Philippine Statistics Authority and ICF International (2014), Ministère du Plan et Suivi de la Mise
en oeuvre de la Révolution de la Modernité, Ministère de la Santé Publique de République Démocratique du Congo, et MEASURE DHS, ICF International (2014).
305
In addition, dwellings may not have sufficient living space for
all members. In addition to overcrowding, insufficient space some-
times limits households’ options for basic services and housing
improvements. For instance, Bag et al. (2016) find that small plot
sizes of 25 square yards or less in ‘relocated colonies’ in Delhi mean
households with poor provision of sewerage systems are unable to
construct their own sanitation facilities, which forces them to use
paid public toilets.
In the five Indian cities in our analysis, the majority of house-
holds have improved floors, and even when incorporating durabil-
ity of walls and roofs, this share is largely unchanged in most cities.
However, when considering overcrowding, this share drops signif-
icantly: the difference ranges from 11 percentage points in Hyder-
abad to up to 33 percentage points in Mumbai (authors’
calculations based on IIPS & Macro International, 2007).
5.3. Multidimensional poverty index (MPI)
So far, we have shown that some parameters of the multidimen-
sional indicators used to measure access to basic services and hous-
ing do not adequately consider pertinent concerns in urban areas.
Improved water facilities include shared facilities (in informal set-
tlements, they could be shared by hundreds); improved sanitation
can include types of facilities that compromise good health in urban
areas (e.g. facilities connected to a septic tank or pit latrines) and
exclude others that may be shared but may be fit for purpose; and
housing indicators based on flooring materials can underestimate
overcrowding, among other housing deprivations.
As an illustrative example, we assess the extent to which
adjusting the indicators on water, sanitation and housing under
the living standards component of the MPI (while preserving its
health and education components) affects measures of multidi-
mensional poverty in seven cities. We considered a household
deprived if:
 The household does not have access to piped drinking water on
premises.
 The household does not have flush or flush pour latrine to piped
sewer system.
 The household does not have a ‘finished’ roof, walls or floors or
it has four or more people sharing a sleeping room (see Annex 1
for more details on the MPI and its methodology, and how the
modifications made compare with the MPI indicators).
Hyderabad Chennai (2006) Manila (2013) Kinshasa (2014)
(2006)
Modified JMP (shared up to 5 hhs)
Flush latrines to piped sewer system (shared up to 5 hhs)
306
45%
40%
35%
Share of populaon (%)
30%
25%
20%
15%
10%
5%
0%
Delhi (2006) Kolkata (2006) Mumbai (2006) Hyderabad Chennai (2006) Manila (2013)
Fig. 4. Headcount ratio for multidimensional poverty in selected cities (%). Sources: Authors’ calculation based on IIPS and Macro International (2007), Philippine Statistics
Authority and ICF International (2014), Ministère du Plan et Suivi de la Mise en oeuvre de la Révolution de la Modernité, Ministère de la Santé Publique de République
Démocratique du Congo, et MEASURE DHS, ICF International (2014). Note: More details on the methodology are included in Annex 1.
Even amending only selected indicators from the ‘living stan-
dards’ dimension in urban areas leads to considerable changes in
the MPI (Fig. 4). The difference in the poverty headcount ratio
(the share of people identified as poor) between the global MPI
and the modified MPI is 5 percentage points or more in six out of
seven cities. As an example, in Delhi this would amount to over
one million people.
There are also differences in the average intensity of poverty
among the poor, or the share of weighted indicators in which the
poor are deprived, of up to 3 percentage points (in Meerut) (see
Table A2 in Annex 1 for further results).
6. Conclusion
As the pace of urbanisation accelerates, it is increasingly impor-
tant that the way we monitor poverty trends is suitable for urban
contexts. Some aspects of data collection and poverty measures at
national and international level were devised at a time when pov-
erty was mostly rural, and now require adjustments.
In this paper, we examined different ways in which current
measures may underestimate urban poverty. We have considered
problems with varying definitions of slums and issues related to
the data (e.g. undercounting of slum populations and lack of
detailed geographical disaggregation beyond urban averages). We
have also discussed how particular assumptions and indicators
used in monetary and multidimensional measures of poverty
may be inappropriate for urban contexts, and proposed some indi-
cators that might be more suitable. We conclude by providing a
series of suggestions for improvements with regard to each of
the problems identified in this article.
6.1. Addressing problems with slum definitions
Our analysis demonstrates how national and global level esti-
mates on the size and proportion of slum dwellers can vary signif-
icantly with the definitions used.
It is worth considering whether it is still appropriate to club
together one or a few households with housing or basic service
deprivations and larger squatter settlements within the same def-
inition, as the UN-Habitat definition does. Adding more nuanced
breakdowns that differentiate between types of housing depriva-
P. Lucci et al. / World Development 103 (2018) 297–310
Original Modified
Kinshasa
(2006) (2014)
tions (e.g. squatter settlement versus inner city precarious build-
ing) could address this issue (this would require agreement on
settlement density/housing structure criteria). There is also scope
to improve the transparency of definitions and specific thresholds
used for different criteria. At present, it is often difficult to find this
information by reading metadata published online by data
producers.
6.2. Addressing problems with data
6.2.1. Undercounting in data collection
Many of the biases inherent in undercounting slum populations
can be corrected by improving data collection. NSOs and interna-
tional organisations (e.g. those coordinating MICS, DHS, LSMS,
etc.) all have a role to play in this regard.
As in the case of definitions, it would be useful to increase trans-
parency over the methods used to construct sampling frames.
Metadata could state clearly the extent to which methods used
could undercount certain populations, such as slum dwellers. This
way, independent sources can verify and replicate these estimates
on their own. Cross-checking and using different sources of data
produced by different actors (e.g. global numbers, national num-
bers, enumerations by civil society) could also help highlight issues
relating to missing slum populations.
For instance, many slum dwellers (organised as federations)
have recently carried out enumerations that can complement and
verify the data collected through conventional methods in addition
to filling data gaps. Federations use this rich information to nego-
tiate with local government, influence resource flows and develop-
ment priorities and make poor communities vocal and visible
(Know Your City, 2016).
In some instances, it has also been possible to compare
community-generated data across countries and to collect such
data using Global Positioning System (GPS) technology (Beukes,
2014). GPS-based data have location and time components, so
clearly identifiable points with precise GPS coordinates allow slum
dwellers to be ‘proactive in defining their own spaces and in put-
ting themselves on the maps’ (ibid: 2), especially when govern-
ments’ city maps fail to register their neighbourhoods’ (ibid.).
In addition, more countries, particularly those where urbanisa-
tion is taking place at a fast pace, could consider conducting
 P. Lucci et al. / World Development 103 (2018) 297–310
slum-specific Censuses and surveys. For example, in Bangladesh,
slum-specific Censuses are being carried out in between two Pop-
ulation and Housing Census rounds. If a slum-specific Census is
carried out at the mid-point between two Population and Housing
Census rounds, the data are likely to better reflect conditions in
slums, and to provide consistent time series data to compare
changes over time. In addition, census listings can be updated
using aerial photography; as some countries like Chile do (INE,
2009). The UN Population Fund (UNFPA), which is coordinating
the 2020 round of the Population and Housing Census, is ideally
placed to spearhead this exercise and to mobilise countries in
doing so. Finally, it would also help if Population and Housing Cen-
sus data generally included slum settlements as another geo-
graphic unit, making these estimates easily available with every
round (e.g. through NSO portals).
Slum-specific surveys can use standardised instruments of the
DHS or MICS, but with sampling frames designed especially for
slums. For instance, the African Population and Health Research
Centre (APHRC) designed a survey for Nairobi’s slums in 2000,
the results of which could be compared with Kenya’s 1998 DHS
(APHRC, 2012). The Kenyan National Bureau of Statistics (KNBS)
conducted a MICS exclusively for Mombasa’s slums in 2009. Such
surveys are exceptions, however, not the rule (Carr-Hill, 2013).
Big data, in particular mobile phone call detail records (CDRs),
can also be used to capture high-frequency data on poor mobile
populations that move in and out of slums. CDRs have been used
to analyse mobility and migration patterns, and to predict poverty
headcount rates. In Rwanda, for example, Blumenstock (2012) esti-
mated internal migration patterns using a dataset containing
mobile phone CDRs for 1.5 million people between 2005 and
2009 and from calling about 900 subscribers to collect anonymised
demographic data. Similarly, Wesolowski and Eagle (2009) used
CDRs for 6 million mobile phone subscribers in Nairobi, Kenya,
between June 2008 and 2009 to examine the dynamics of the Kib-
era slum – how long people stay there and where people move to,
as well as where they work. Soto, Frias-Martinez, Virseda, and
Frias-Martinez (2011) use call data records to predict poverty in
a Latin American city with about 500,000 citizens, and compare
their findings with official estimates.
6.2.2. Lack of granularity
The lack of granular survey data means we have information
only for urban averages, and do not know about the extent of
intra- and inter-city disparities. Household surveys often have
small sample sizes that do not allow for disaggregation of data
beyond urban averages.
Conducting slum-specific surveys or ensuring that household
surveys have large enough samples sizes to provide results for
slum settlements can help with this issue. Trying and testing
new methods and new technologies to capture slum data more
easily and frequently (in periods where survey data are unavail-
able) could also be useful.
6.3. Addressing problems with monetary poverty measures
Commonly used indicators like the income poverty headcount
ratio can also underestimate urban poverty. In particular, the lack
of accounting for higher non-food spending and relative costs of
living in urban areas can result in urban poverty lines that are
too low.
6.3.1. Outdated questionnaires do not reflect current food consumption
habits
One problem with monetary measures is that questionnaires do
not reflect changing habits (e.g. an increasing tendency for house-
hold members to eat separately and away from home). Household
307
consumption and expenditure surveys should include questions
for meals out and make sure these include quantities consumed
to calculate caloric intake. Ideally, food expenditure should be
recorded by each adult rather than by a household head, and using
diaries as this method is perceived to be more accurate. Of course,
this is less feasible in countries with high rates of illiteracy, and
accuracy has been shown to diminish over time as respondents
develop fatigue (Beegle et al., 2012).
6.3.2. Non-food allowances, central to urban poverty, are inadequately
covered
Another central problem of monetary measures is that the data
needed to account for non-food allowances, key elements of low-
income households’ budgets in urban areas, are often not consis-
tently collected. Expenditure surveys should include comprehen-
sive questions on non-food needs, including housing, transport,
water, sanitation, schooling and health care. It is also worth consid-
ering a review of the most common methods to account for non-
food allowances in urban contexts, as the most popular ones are
likely to set the urban poverty line too low.
6.3.3. Differences in costs of living are insufficiently accounted for
Finally, as the importance of comparing poverty incidence and
standards of living across areas and provinces is likely to increase,
further differentiation of the cost of living between different areas
(large cities versus smaller ones) will be needed too. There is also
scope here to consider whether new technologies can help produce
more frequent and granular data on prices in different areas.
6.4. Addressing problems with multidimensional measures
6.4.1. Access to basic services: Accounting for dense urban contexts
with high demand
Sources that are adequate in rural settings may not be appropri-
ate in urban areas. In this context, for instance, only having data on
distance from or time taken to walk to a drinking water source is
not enough; information is needed on how many people share a
source (and/or waiting times) in order to account for overcrowding
in dense slum settlements.
There is also scope to be more nuanced about the definition of
access to improved water, building on the ‘ladder’ approach set
out by the UNICEF/WHO JMP. For instance, in addition to a
share/unshared distinction, there could be another category of data
readily available for limited sharing (i.e. by ‘up to five households’).
This is in line with what is currently proposed by the slums targets
in the SDGs (UNSD, 2016). Although the number of surveys includ-
ing a question on sharing services has increased over the years,
data on the extent of sharing are rare – only 85 out of over 400 sur-
veys in 2015 lent themselves to distinguishing the number of
households sharing facilities (UNICEF and WHO, 2015).
In the case of access to sanitation facilities, there is scope to
revisit the categories used and what constitutes ‘improved’
access. Although the JMP classifies all shared facilities as unim-
proved, shared sanitation is on the first step of the sanitation
ladder and as such, should be reported individually (UNICEF
and WHO, 2015). Some have suggested that a threshold of five
or more households, as in the case of water facilities, known
as ‘limited sharing’, should be included in the ‘improved’ cate-
gory for sanitation (ibid.). Others argue that even limited sharing
has negative impacts on health and should not be considered
‘improved’. Note that limited sharing would still consider com-
munity toilet blocks – which, for instance, work well in Mum-
bai’s dense slums settlements as a short-term solution – as
unimproved facilities (Mitlin, 2015).
In addition to issues of overcrowding, indicators used must
pay attention to the implications of service quality. As argued
308 P. Lucci et al. / World Development 103 (2018) 297–310

in Section 5, a pit latrine or septic tank may be sufficient in rural Acknowledgements

areas but not in densely populated urban settlements where
they may fill up rapidly and communities may find it difficult
to get them emptied or to build new pits. In fact, the absence
of a piped sewerage network in urban areas raises concerns
about the adequacy of sanitation services given adverse health
risks caused by poor faecal waste management (Mitlin, 2015).
This and ongoing discussions on shared facilities and how to
classify hygienic toilet blocks shared with more than five house-
holds warrant further discussions relating to the JMP criteria of
improved sanitation. Other aspects, such as the quality of the
service (e.g. the quality of the water, the hygienic condition of
toilets or whether access is frequently available), its affordability
and the number of hours in which households can access water,
and indeed sanitation, are also important.
6.4.2. Access to housing: Collecting information beyond the quality of
the flooring
In the case of housing too, more consistent data collection on
quality is needed – beyond the material of the flooring of house-
holds (e.g. on materials of roofs and walls) to include information
about the extent of overcrowding and tenure security, which fewer
household surveys collect. Again, this will require strengthening
data collection efforts in these areas.
In short, improvements in data collection are urgently needed.
This will help governments and campaigners better understand
the consequences of urbanisation and push for policies that can
improve poor city dwellers’ lives. Of course, data alone will not
be enough to achieve these changes, but they will be a step in
the right direction.
We are grateful to two anonymous reviewers for their com-
ments which helped improve and clarify this paper. Sincere thanks
to Steven Commins, Christopher Hoy, Claire Melamed, Gregory
Pierce, Emma Samman, Pronab Sen, Suman Seth, and Elizabeth Stu-
art for their useful comments to an earlier version of this paper.
The William and Flora Hewlett Foundation and the Department
for International Development (DFID) provided financial support
to undertake this research. The usual disclaimers apply.
Funding
This article is based on research that was supported by the Wil-
liam and Flora Hewlett Foundation and the Department for Inter-
national Development (DFID). The sponsors have not been
involved in any stage of study design; collection, analysis, or inter-
pretation of data; drafting of the article; or decision to submit the
article for publication.
Annex 1: Indicators and weights used in calculating the
multidimensional poverty Index, and selected results
It is widely acknowledged that poverty is not about just income
or consumption but is a much wider concept. In this context,
recent efforts have attempted to measure deprivation in a way that
accounts better for broader aspects of well-being. The MPI (intro-
duced by OPHI and UNDP, and published annually by the Human
Development Report Office) is one such an initiative. This

Table A1
Weights and indicators used in original and modified version.

Dimension Weight A household is considered deprived if. . .

Original indicator Modification
Health 1/6 Nutrition:* Any adult or child for whom there is nutritional –
information is malnourished
1/6 Child mortality:** Any child has died in the household within the past –
five years
Education 1/6 Years of schooling: No member older than 10 years has completed 5 –
years of schooling
1/6 School attendance: Any school-age child is not attending school up to –
the age they would complete Class 8
Living standards 1/18 Cooking fuel: The household cooks with dung, wood or charcoal –
1/18 Sanitation***: The household’s sanitation facility is not improved, or it The household does not have flush or flush pour latrine to
is improved but shared with other households piped sewer system, or it shares its facilities with other
households
1/18 Water:**** The household does not have access to an improved The household does not have piped drinking water to premises
drinking water source or it is a 30-min walk or more from home, or plot/yard
round trip
1/18 Electricity: The household has no electricity –
1/18 Housing: The household has a dirt, sand or dung floor The household does not have a ‘finished’ roof, walls or floors, or
it has 4 or more people per sleeping room
1/18 Assets: The household does not own more than one radio, TV, –
telephone, bike, motorbike or refrigerator and does not own a car or
truck

Note: While we have amended access indicators such that they consider urban-specific deprivations, data are unavailable on some indicators (e.g. number of hours of water
access).
*
Adults are considered malnourished if their Body Mass Index is below 18.5 m/kg2. Children are considered malnourished if their z-score of weight-for-age is below -2
standard deviations from the median of the reference population
**
The DHS does not collect data on whether a child has died in the past five years; instead, it asks women if they have a son or daughter who has died without stipulating a
time period.
***
A household is considered to have access to improved sanitation if it has some type of flush toilet or latrine, or ventilated improved pit or composting toilet, provided they
are not shared.
****
A household has access to clean drinking water if the water source is any of the following types: piped water, public tap, borehole or pump, protected well, protected
spring or rainwater, and is within 30 minutes’ walk (round trip).
 P. Lucci et al. / World Development 103 (2018) 297–310 309

Table A2
MPI and modified MPI results for cities.

City Original Modified Difference (% pts)

H A M0 H A M0 H A
Delhi (2006) 15% 42% 0.063 21% 44% 0.093 6% 2%
Kolkata (2006) 17% 46% 0.076 27% 46% 0.122 10% 0%
Mumbai (2006) 15% 41% 0.060 20% 43% 0.085 5% 2%
Hyderabad (2006) 20% 41% 0.080 24% 44% 0.108 5% 4%
Chennai (2006) 14% 40% 0.055 24% 41% 0.102 11% 2%
Manila (2013) 1% 33% 0.003 4% 54% 0.023 3% 21%
Kinshasa (2014) 24% 43% 0.103 41% 43% 0.176 17% 0%

Note: H refers to the headcount ratio; A is the intensity of poverty among the poor or average the share of weighted indicators in which the poor are deprived; and M0 is the
modified headcount ratio or MPI, calculated as the product of H and A.

Table A3
MPI results for cities by type of residence.

City H A M0
Non-slum Slum Non-slum Slum Non-slum Slum
Delhi (2006) 7% 23% 35% 44% 0.025 0.102
Kolkata (2006) 7% 30% 39% 48% 0.028 0.143
Mumbai (2006) 10% 18% 38% 41% 0.037 0.073
Hyderabad (2006) 8% 25% 36% 41% 0.031 0.105
Chennai (2006) 8% 16% 37% 40% 0.029 0.064

Note: H refers to the headcount ratio; A is the intensity of poverty among the poor or the average share of weighted indicators in which the poor are deprived; and M0 is the
modified headcount ratio or MPI calculated as the product of H and A.

internationally-comparable index includes deprivations in three
dimensions (Table A1).
1. health – measured through child mortality and nutrition
2. education – measured by years of schooling and enrolment
3. living standards – using water, sanitation, electricity, cooking
fuel, floor, assets
The three dimensions are equally weighted, as are all indicators
within each dimension. The methodology used identifies house-
holds deprived in each of these indicators, and then classifies
households as poor if they suffer deprivations across a third or
more of the weighted indicators. The MPI is then the product of
the proportion of population that is poor (H or headcount ratio)
and the average share of weighted indicators that poor people
are deprived on (A or intensity of poverty).
Table A1 shows the different weights and indicators used in the
MPI. Section 5 discusses how the estimated poverty rate as mea-
sured by the MPI would vary if the indicators used to assess depri-
vation included urban-specific areas of concern. In particular,
modifications were introduced in the living standards component.
The weights and original measure follow from the methodology
used to calculate the global MPI as discussed in Alkire and Robles
(2015). To retain comparability we have only modified existing
measures instead of adding new indicators. Owing to a lack of data
on quality and cost of basic services, these factors have not been
accounted for.
References
Agarwal, S. (2011). The state of urban health in India: Comparing the poorest
quartile to the rest of the urban population in selected states and cities.
Environment and Urbanization, 23(1), 13–28.
Alkire, S., & Robles, G. (2015). Multidimensional Poverty Index – 2015.
Methodological notes and results.
APHRC (African Population and Health Research Centre) (2012). Nairobi cross-
sectional slums survey 2012. Nairobi: APHRC.
Bag, S., Seth, S., & Gupta, A. (2016). A comparative study of living conditions in slums of
three metro cities in India. Leeds University Business School Working Paper No.
16–07. Delhi: Centre for Development Economics.
Bangladesh Bureau of Statistics (2011). Household income and expenditure survey
2010. Dhaka: Ministry of Planning, Statistics Division.
Beegle, K., de Weerdt, J., Friedman, J., & Gibson, J. (2012). Methods of household
consumption measurement through surveys: Experimental results from
Tanzania. Journal of Development Economics, 98(1), 3–18.
Beukes, A. (2014). Know your city: Community profiling of informal settlements.
Briefing, London: IIED.
Blumenstock, J. E. (2012). Inferring patterns of internal migration from mobile
phone call records: Evidence from Rwanda. Information Technology for
Development, 18(2), 107–125.
Carr-Hill, R. A. (2013). Missing millions and measuring development progress.
World Development, 46, 30–44.
Chandrasekhar, S., & Montgomery, M. R. (2010). Broadening poverty definitions in
India: Basic needs in urban housing Human Settlements Working Paper. London:
IIED.
Chandy, L. (2013). Counting the poor. Methods, problems and solutions behind the
$1.25 a day global poverty estimates. Research Paper. Washington, DC: Brookings
Institute and Development Initiatives.
Chen, M. (2005). Rethinking the informal economy linkages with the formal economy
and the formal regulatory environment. UNU Wider, Research Paper No. 2005/10.
Chen, S., & Ravallion, M. (2008). China is poorer than we thought, but no less successful
in the fight against poverty. Policy Research Working Paper 4621. Washington,
DC: World Bank.
Chibuye, M. (2011). Interrogating poverty lines: The case of Zambia Human
settlements working paper. London: IIIED.
Citiscope (2015). What is Habitat III? Accessed 05.03.2016 http://citiscope.org/
habitatIII/explainer/2015/06/what-habitat-iii.
Desai, R. (2012). Governing the urban poor: Riverfront development, slum
resettlement and the politics of inclusion in Ahmedabad. Economic and
Political Weekly, 47(2), 49–56.
Deshingkar, P. (2006). Internal migration, poverty and development in Asia. Sussex
and London: Institute of Development Studies and Overseas Development
Institute.
Dupriez, O., Smith, L., & Troubat, N. (2014). Assessment of the reliability and relevance
of the food data collected in national household consumption and expenditure
surveys. Rome and Washington, DC: FAO, IHSN and World Bank.
Ghertner, D. A. (2010). Calculating without numbers: Aesthetic governmentality in
Delhi’s slums. Economy and Society, 39(2), 185–217.
Gibson, J. (2007). A guide to using prices in poverty analysis. Department of
Economics, Hamilton: University of Waikato.
Gibson, J. (2015). Poverty measurement: We know less than policy makers realize.
Working paper in economics 8/15. Hamilton: University of Waikato.
Gilbert, A. (2013). How to help, and how not to help, the poor in the megacities of
the South. City, 17(5), 628–635.
Gilbert, A. (2014). Shelter and the Millennium Development Goals. Development
Progress Blog. http://www.developmentprogress.org/blog/2014/01/20/shelter-
and-millennium-development-goals.
Government of India (2010). Report of the Committee on Slum Statistics/Census. New
Delhi: Ministry of Housing and Urban Poverty Alleviation, National Buildings
Organisation.
310 P. Lucci et al. / World Development 103 (2018) 297–310
Graham, J. P., & Polizzotto, M. L. (2013). Pit latrines and their impacts on
groundwater quality: A systematic review. Environmental Health Perspectives,
121(5), 121–130.
Haughton, J., & Khandker, S. R. (2009). Handbook on poverty and inequality.
Washington, DC: World Bank.
Hentschel, J., & Lanjouw, P. (1996). Constructing an indicator of consumption for the
analysis of poverty. World Bank living standard measurement study 124,
Washington, DC: World Bank.
IAEG (Inter-agency Expert Group on the SDGs) (2016). Metadata for the proposed
global Indicators (as of 18 January 2016).
IIPS (International Institute for Population Sciences) and Macro International
(2007). National family health survey (NFHS-3), 2005–06: India. Mumbai: IIPS.
INE (Instituto Nacional de Estatísticas) (2009). Marco maestro de viviendas. Santiago:
INE.
Jolliffe, D. (2006). Poverty, prices and place: How sensitive is the spatial distribution
of poverty to cost-of-living adjustments? Economic Inquiry, 44(2), 296–310.
Jolliffe, D., & Beer Prydz, E. (2015). Global poverty goals and prices: How purchasing
power parity matters. Policy Research Working Paper 7256. Washington, DC:
World Bank.
Kingdon, G. G., & Knight, J. (2006). Subjective well-being poverty vs. income poverty
and capabilities poverty? The Journal of Development Studies, 42(7), 1199–1224.
KNBS (Kenya National Bureau of Statistics) (2009). 2009 population and housing
census. Nairobi: KNBS.
Know Your City (2016). http://knowyourcity.info/map.php#/app/ui/about3.
Accessed 05.01.2016.
Lucci, P., & Bhatkal, T. (2014). Monitoring progress on urban poverty: Are indicators fit
for purpose? London: ODI.
Maron, M. (2010). Kibera’s Census: Population, politics, precision. Map Kibera Trust.
Nairobi: Map Kibera Trust.
Marras, S. (2009). Kibera: Mapping the unmapped. Milan: University of Milano
Bicocca.
Marx, B., Stoker, T., & Suri, T. (2013). Economics of slums in the developing world.
Journal of Economic Perspectives, 27(4), 187–210.
McGranahan, G., & Mitlin, D. (2016). Learning from sustained success: How
community-driven initiatives to improve urban sanitation can meet the
challenges (Vol. 87) World Development, pp. 307–317.
MHUPA (Ministry of Housing and Urban Poverty Alleviation) (2010). Report on
committee of slum statistics/census. New Delhi, India: Ministry of Housing and
Urban Poverty Alleviation, Government of India.
Mitlin, D. (2015). Building towards a future in which urban sanitation leaves no one
behind Environment and Urbanisation Brief 32. London: IIED.
Mitlin, D., & Satterthwaite, D. (2013). Urban poverty in the Global South: Scale and
nature. London: Routledge.
Mitlin, D., Walnycki, A. (2016). Why is water still unaffordable for sub-Saharan
Africa’s urban poor? IIED Briefing, March 2016. http://pubs.iied.org/17353IIED/.
MKP (Map Kibera Project) (2009). http://mapkiberaproject.yolasite.com/maps-and-
statistics.php. Accessed 01.01.2016.
Ministère du Plan et Suivi de la Mise en œuvre de la Révolution de la Modernité
(MPSMRM), Ministère de la Santé Publique (MSP), and ICF International (2014).
Enquête Démographique et de Santé en République Démocratique du Congo 2013–
2014. Rockville, Maryland, USA: MPSMRM, MSP and ICF International.
Parnell, S. (2005). Constructing a developmental nation-the challenge of including
the poor in the post-apartheid city. Transformation: Critical Perspectives on
Southern Africa, 58(1), 20–44.
Patel, S., Baptist, C., & D’Cruz, C. (2012). Knowledge is power – informal
communities assert their right to the city through SDI and community-led
enumerations. Environment and Urbanization, 2012(24), 13–26. https://doi.org/
10.1177/0956247812438366.
Patel, A., Koizumi, N., & Crooks, A. (2014). Measuring slum severity in Mumbai and
Kolkata: A household-based approach. Habitat International, 300–306.
Philippine Statistics Authority (PSA) and ICF International (2014). Philippines
National Demographic and Health Survey 2013. Manila, Philippines, and
Rockville, Maryland, USA: PSA and ICF International.
Pradhan, M., & Ravallion, M. (2000). Measuring poverty using qualitative
perceptions of consumption adequacy. Review of Economics and Statistics, 82
(3), 462–471.
Ravallion, M. (2012). Poor, or just feeling poor? On using subjective data in measuring
poverty. http://elibrary.worldbank.org/doi/abs/10.1596/1813-9450-5968.
Ravallion, M. (2014). India’s puzzling new PPP. Blog, Washington, DC: CGD.
Ravallion, M., Chen, S., & Sangraula, P. (2008). New evidence on the urbanization of
global poverty. Background Paper for the World Development Report. Washington,
DC: World Bank.
Risbud, N. (2010). Typology of slums and land tenure in Indian cities. In National
workshop on land tenure issues in slum-free city planning. Ahmedabad: Center for
Urban Equity, CEPT University.
Roy, A. (2005). Urban informality: Toward an epistemology of planning. Journal of
the American Planning Association, 71(2), 147–158.
Roy, A. (2011). The blockade of the world-class city: Dialectical images of indian
urbanism. In: Worlding cities (pp. 259–278). Oxford, UK: Wiley-Blackwell.
Sabry, S. (2009). Poverty lines in greater Cairo: Under-estimating and misrepresenting
poverty Poverty reduction in urban areas series working paper 21. London: IIED.
Sabry, S. (2010). How poverty is underestimated in Greater Cairo, Egypt.
Environment and Urbanisation, 22(2), 523–541.
Satterthwaite, D. (2014). MDG experience regarding improved drinking water,
sanitation and slums and the lessons for implementation of the post 2015 agenda.
London: IIED.
SEDLAC (Socio-Economic Database for Latin America and the Caribbean) (n.d).
Socio-economic database for Latin America and the Caribbean. Accessed on March
15th 2016.
Sheikh, S., & Banda, S. (2015). Surveying slums. Economic & Political Weekly, 50(22),
73.
Soto, V., Frias-Martinez, V., Virseda, J., & Frias-Martinez, E (2011). Prediction of
socioeconomic levels using cell phone records. In: Proceedings of the 19th
international conference on user modeling, adaption, and personalization. UMAP 11
(pp. 377–388). Berlin, Heidelberg: Springer-Verlag.
Sutcliffe, M., & Bannister, S. (2014). Africa. In United Cities and Local Governments
(Ed.), Third global report on local democracy and decentralisation: Basic services for
all in an urbanising world. Abingdon and New York: Routledge.
Tacoli, C. (2007). Poverty, inequality and the underestimation of rural-urban
linkages. Development, 50(2), 90–95.
Tacoli, C., McGranahan, G., & Satterthwaite, D. (2015). Urbanisation, rural–urban
migration and urban poverty. Working Paper. London: IIED. http://pubs.iied.org/
pdfs/10725IIED.pdf.
Thanh, H.-X., Anh, T. T., & Phuong, D. T. T. (2013). Urban poverty in Vietnam: A view
from complementary assessments. IIED Human Settlements Working Paper
Series, No. 40. London: IIED.
Twaweza (2010). It’s our water too! Bringing greater equity in access to water in Kenya
Policy brief 09/2010. Nairobi: Uwazi.
UN DESA (UN Department of Economics and Social Affairs) (2013). World economic
and social survey: Sustainable development challenges. New York: UN DESA.
UN DESA (UN Department of Economics and Social Affairs) (2014). World
urbanization prospects The 2014 revision. New York: UN DESA.
UN-Habitat (UN Human Settlements Programme) (2003). The challenge of slums –
global report on human settlements 2003. Nairobi: UN-Habitat.
UN-Habitat (UN Human Settlements Programme) (2014). The global urban indicators
database 2014. Nairobi: UN-Habitat.
UN-Habitat (UN Human Settlements Programme) (2016). SDG goal 11 monitoring
framework. Nairobi: UN-Habitat.
UNICEF (UN Children’s Fund) and WHO (World Health Organization) (2015).
Progress on sanitation and drinking water: 2015 update and MDG assessment. New
York and Geneva: UNICEF and WHO.
UNSD (UN Statistics Division) (2016). UN millennium development goals’ indicators.
New York: UNSD.
Water and Sanitation Program (WSP) – Africa (2007). Citizen report card for water in
urban areas. Nairobi: WSP.
Wesolowski, A. P., & Eagle, N. (2009). Inferring human dynamics in slums using mobile
phone data. Santa Fe, NM: Santa Fe Institute.
World Bank (2015). Purchasing power parities and the real size of world economies: A
comprehensive report of the 2011 International Comparison Program. Washington,
DC: World Bank.
World Bank and IMF (International Monetary Fund) (2013). Global monitoring report
2013: Rural-urban dynamics and the Millennium Development Goals. Washington,
DC: World Bank and IMF.
Yelling, J. (1986). Slums and slum clearance in Victorian London. London: Allen and
Unwin.
Yntiso, G. (2008). Urban development and displacement in Addis Ababa: The impact
of resettlement projects on low-income households. Eastern Africa Social Science
Research Review, 24(2), 53–77.