World Development Vol. 36, No. 10, pp.

1916–1937, 2008
! 2008 Elsevier Ltd. All rights reserved
0305-750X/$ - see front matter
www.elsevier.com/locate/worlddev
doi:10.1016/j.worlddev.2008.02.010

Slum Real Estate: The Low-Quality High-Price

Puzzle in Nairobi’s Slum Rental Market and
its Implications for Theory and Practice
SUMILA GULYANI
World Bank, Washington, DC, US

and

DEBABRATA TALUKDAR *
State University of New York at Buffalo, US
Summary. — This study of 1755 households in Nairobi’s slums challenges the conventional belief
that slums offer low-quality low-cost shelter to a population that cannot afford better standards. In
Nairobi, slums provide low-quality but high-cost shelter. Although slum residents pay millions of
dollars in rents annually, and better quality units command higher rents, very little is being re-in-
vested to upgrade quality. To resolve the challenge that the Nairobi puzzle poses for theory and
practice, we develop a new analytical framework for understanding quality of living conditions.
Improving conditions in Nairobi’s slums requires, we argue, two simultaneous interventions: alter-
ation of the tenure mix to enhance owner occupancy and infrastructure investment.
! 2008 Elsevier Ltd. All rights reserved.

Key words — informal housing markets, tenure, infrastructure, slum upgrading, quality, Kenya,
Africa

1. INTRODUCTION Recovery Strategy (Government of Kenya,

In a report titled Challenge of Slums, the Uni-
ted Nations-HABITAT (2003) estimated that,
in 2001, about 870 million people were living
in urban slums in developing countries. If cur-
rent trends continue unchecked, the report ar-
gues, the number of slum residents will grow
to 1.43 billion by 2020. World leaders and
development agencies are again, after a signifi-
cant hiatus, displaying their concern about the
issue, and slums appear to be back on the core
development agenda. At UN-sponsored sum-
mits in 2000 and 2002, world leaders agreed
to a set of millennium development goals which
include a commitment to significantly improve
the lives of at least 100 million slum dwellers
by 2020 (UN-HABITAT, 2003). In a nation
such as Kenya, this commitment now appears
in national development plans and is high-
lighted as a key task in the National Economic
1916
2003).
The commitment to improve the lives of slum
dwellers is well intentioned and important, but
the task of achieving this goal is fraught with
problems (Gulyani & Bassett, 2007; World
* We are grateful to Natasha Iskander, Ellen Bassett,
Peter Marcuse, Jacqueline Klopp, and four anonymous
reviewers for excellent comments; Jaime Biderman,
Geoffrey Bergen, Colin Bruce, Makhtar Diop, and
Matthew Glasser for strongly supporting this study;
James Mutero for leading the fieldwork in Nairobi;
Alyssa Boyer and Cuz Potter for research assistance;
Adam Samuel for editing tips; and the World Bank and
Norwegian Trust Fund for financing this work. The
views expressed here are neither attributable to nor
necessarily shared by the World Bank. Final revision
accepted: February 22, 2008.
 SLUM REAL ESTATE
Bank, 2006). First, a generation of earlier ef-
forts, starting in the 1970s, to upgrade urban
slums has, at best, been only partially success-
ful. Second, there is little information on and
understanding of urban poverty in general,
and the situation in slums in particular—there
are few large-scale or representative studies that
examine the current lives of slum residents,
their living conditions, or the nature of poverty
that they face. Consequently, in most develop-
ing countries, there are no reliable estimates
even of basic indicators—such as the number
of people residing in these slums and the pro-
portion of them that are poor. The UN-HABI-
TAT’s figures cited above are also the subject
of much contention and debate, and the data
problem is particularly acute in Sub-Saharan
Africa. This type of ambiguity regarding scale
combined with the lack of systematic studies
documenting conditions inside informal settle-
ments, means that our understanding of the
nature of slums and their dynamics is extremely
limited. Such a partial understanding may well
have been a key reason for the limited success
of previous programs, and is highly likely to
confound the design, implementation, and suc-
cess of the forthcoming generation of slum-
improvement efforts as well.
As a step toward addressing this gap in our
knowledge base, we conducted a large-scale,
and representative study of 1,755 households
residing in Nairobi’s slums. 1 The focus was
on understanding who lives there, what they
do, and how they live. In this paper, we use
data from our population-weighted stratified
random sample to examine one aspect of life
in the slums—quality of housing and living
conditions, and its relationship to the price res-
idents pay for their accommodation. Specifi-
cally, we focus on the following questions:
how inadequately housed and under-served are
slum residents in Nairobi? How much do they
pay for their housing and what factors explain
variation in their rents?
The literature on housing in developing
countries valorizes slums as an affordable op-
tion for the low-income (see Gulyani & Bassett,
2007 for a recent review). It argues that
although slums are sub-par from a technocratic
planning perspective and unpleasant by middle-
class standards, they play a crucial role by
proving shelter at low-cost to the urban poor.
Further, if given even implicit recognition by
government—for instance, through a morato-
rium on evictions—these settlements consoli-
date and housing improves over time.
1917
We find that Nairobi’s slums constitute a
serious challenge to the aforesaid conventional
wisdom. In Nairobi, the slums indeed cater to
low-income residents and offer low-quality liv-
ing conditions. The surprise is that the resi-
dents, most of whom are very poor, are
paying a large sum of money in rents each
year—in 2004 they paid, on average, US$126
(Ksh 9840) per household, amounting to a
cumulative sum of about US$31 million (Ksh
2.35 billion). Despite such large payments and
a significant history of implicit recognition by
government, however, housing in the slums
has not improved and consolidated. In other
words, Nairobi’s slums offer low-quality but
high-cost housing, and, even worse, they seem
to be stuck in this sub-optimal equilibrium.
By identifying this puzzle and providing a rare
insight into the scale and nature of Nairobi’s
massive informal rental market, this paper
makes an empirical contribution.
Further, to resolve the challenge that the
Nairobi case poses for both theory and prac-
tice, we develop a new framework for analyzing
and improving ‘‘quality of living conditions.’’
Specifically, we argue that quality of living con-
ditions must be understood along four dimen-
sions—tenure, infrastructure, the unit itself,
and the neighborhood—and that these factors
interact with each other to determine the out-
come. Using this framework, we show why
Nairobi’s slums are stuck in a low-quality
high-cost trap, and suggest how one might
intervene to improve the situation. The frame-
work also helps us understand, for example,
why a slum upgrading program that focuses
on improving infrastructure (rather than, say,
changing the tenure status) may be the right ap-
proach in certain contexts, but is likely to fail in
Nairobi’s slums. At a broader level, this frame-
work offers a step toward a more complete the-
ory of living conditions and their dynamics—
because it facilitates a multi-dimensional under-
standing of quality, highlights interactions
among variables and over time, and helps ex-
plain variation among contexts.
This paper is structured as follows. Section 2
briefly reviews the research methodology, pov-
erty incidence in the slums, and demographics
of slum households. Section 3 focuses on liv-
ing conditions—it presents the analytical
framework and uses it to assess quality in Nai-
robi’s slums. Section 4 examines the rental
market and factors influencing rents. Section
5 returns to the framework, and Section 6 con-
cludes.
1918 WORLD DEVELOPMENT
2. METHODOLOGY, THE DATA, AND
THE TYPICAL SLUM HOUSEHOLD
This study is part of a larger World Bank-ini-
tiated research project on slums in Africa. 2
This paper is based on data gathered, in Febru-
ary, and March 2004, from an in-depth survey
of 1,755 households residing in Nairobi’s
slums. A population-weighted stratified ran-
dom sample of households was created as fol-
lows. For census purposes, Kenya’s Central
Bureau of Statistics (CBS) has divided Nairobi
into about 4,700 enumeration areas (EAs), of
which 1,263 as categorized as ‘‘EA5,’’ or
‘‘informal settlements;’’ these are all character-
ized by poor quality sub-standard housing and
poor infrastructure. For our purpose, all 1,263
EAs are ‘‘slums’’ (see Section 3 for definition).
From these 1,263 slum EAs, 88 were randomly
selected, and lists of their residents were up-
dated by CBS through a field-based re-listing.
About 20 households were then selected ran-
domly, from the updated lists, in each of the
88 EAs (see Appendix 1 for a map of surveyed
areas), and all households were assigned a
weight adjusted to reflect their probability of
selection. In the results presented in this paper,
all household-level data are weighted but indi-
vidual-level data are not.
(a) Slum residents account for at least 30% of the
population
In the 1999 national census, Nairobi’s popu-
lation was 2.14 million, and slums accounted
for 640,000 people or about 30% of the city’s
population. It is worth noting that ‘‘30%’’
seems to be a conservative estimate and may re-
flect, in part, CBS’s use of a relatively stringent
definition of slums. 3 We use this number—
640,000 slum residents in 1999—as establishing
a ‘‘floor’’ or minimum number of slum resi-
dents in the city. By the time of the survey in
2004, the slum population is estimated to have
grown from 640,000 to 810,000. 4 The sam-
pling, and results of this study, therefore, per-
tain to the total population of the 1263 EAs
in 2004, or an estimated 810,000 people.
(b) Disaggregating ‘‘poor’’ and ‘‘non-poor’’
households in slums
We use a discrete poverty measure—an
expenditure-based poverty line—to disaggre-
gate our sample into poor and non-poor house-
holds. Here, the poverty line is defined as a
monthly expenditure of Ksh 3,174 (US$42)
per adult equivalent, excluding rent; (its inclu-
sion would make some renters seem artificially
richer than home owners). 5 This measure is
based on the 1997 urban poverty threshold as
defined by government, and adjusted for infla-
tion during the period 1997–2004. 6 Using this
poverty line, 73% of the households are ‘‘poor’’
and 27% are ‘‘non-poor’’—that is, poverty inci-
dence in the slums is very high.
(c) Key demographics, and the typical slum
household
The average slum household has three mem-
bers (Table 6 in Appendix 1). By comparison,
mean household size is 3.2 for Nairobi and
3.4 for urban Kenya (Census, 1999). The rela-
tively small household size in slums is partly
attributable to the high proportion of single-
person households—they account for almost a
third (32%) of all households. The average
household head is 35 years old, and most are
men; female-headed households account for
18% of all households. As locally believed,
there are more male than female residents; the
ratio, however, is 55:45 belying the notion that
slums are inhabited largely by males.
3. ASSESSING QUALITY: THE LIVING
CONDITIONS DIAMOND
Although slums universally invoke images of
squalor, not all are equally bad. Slums vary—
some have better access to water, others offer
better quality housing units, and some are
physically safer. While the vast literature on
slums cumulatively demonstrates that there is
diversity, it offers few conceptual models that
can comfortably accommodate and explain
such a variation in quality of living conditions
among settlements. This is partly due to the
fact that there are few large-scale empirical
studies that systematically document quality
and variations in it, let alone explain why and
under what conditions quality improves.
As a step toward a broader conceptual model
for documenting, understanding, and compara-
tively analyzing living conditions among
settlements, we develop a new framework. Ana-
lytically, we propose that living conditions—in
any settlement, including slums—can be under-
stood as a combination of four types of factors:
(i) tenure, (ii) infrastructure, (iii) unit quality,
and (iv) the neighborhood’s condition and
 SLUM REAL ESTATE
location. These four factors interact with, and
influence each other, and they collectively deter-
mine the overall quality of living conditions in a
given settlement. We represent this framework
graphically as a diamond with four vertices,
and, for ease of reference, call it the living con-
ditions diamond (Figure 1). The literature elu-
cidates the importance of each of the four
components individually, but we aim to extend
current conceptualizations by creating a com-
posite and by proposing that the four aspects
are interdependent. To illustrate the logic be-
hind the framework, and its relationship to
existing literature, we use the tenure compo-
nent. 7
We make three main propositions regarding
tenure. First, the type and security of ten-
ure—for example, whether a resident is tenant
or owner and whether there is an imminent
threat of displacement—is in itself an impor-
tant aspect of quality of living conditions. It
is, therefore, a component of the framework.
Second, tenure and its perceived security are re-
lated to the three other components: unit qual-
ity; infrastructure access; and the
neighborhood’s condition and location. Third,
the direction of influence between tenure and
any other factor can run either way—for exam-
ple, enhanced tenure security can facilitate
Figure 1. Living conditions diamond.
1919
investment in the unit, or vice versa. The links
between tenure and each of the other three
components—as theorized in the literature,
and in this framework—are discussed in turn
below.
The relationship between tenure and the unit
is much studied and theorized. Many research-
ers have argued that residents of slums, despite
their apparent poverty, will invest to improve
their housing, but to do so they need to have
some security of tenure (e.g., De Soto, 2000;
Turner, 1972). There continue to be raging de-
bates, however, regarding how tenure security
can be enhanced, and whether it can ‘‘follow’’
rather than ‘‘lead’’ investment. For De Soto
(2000), land titling is the right approach to con-
ferring tenure security and unleashing ‘‘dead
capital’’; in this view, legal tenure is a precursor
to housing investment and also leads to broad-
er welfare gains for residents. By contrast,
other analysts have documented cases where
the sequence was reversed—in Amman, Jor-
dan, and Voi, Kenya, residents invested in
making their homes more permanent as a strat-
egy for securing their tenure (Bassett, 2007;
Razzaz, 1993). In our framework, both are va-
lid: tenure and unit quality are related, and a
change in either one can lead to a change in
the other.
1920 WORLD DEVELOPMENT
Tenure is also related to infrastructure access.
Because government agencies see slums as tem-
porary or illegal, they are reluctant to invest in
extending public services—water supply, elec-
tricity, drainage, sewerage, garbage removal,
and street lighting—to them (UN-HABITAT,
2003). If so, a first step to improving services,
and living conditions would be to deem settle-
ments legal. This logic resonates with De Soto’s
(2000), and supports his ‘‘title first’’ approach;
titles can, theoretically, encourage not only pri-
vate investment in units but also public invest-
ment in infrastructure. In direct contrast,
infrastructure proponents argue that the infra-
structure can lead or even replace efforts to for-
malize tenure in slums because infrastructure
investment signals ‘‘acceptance’’ or implicit rec-
ognition of informal settlements by government
(Gulyani & Bassett, 2007; World Bank, 2002).
Infrastructure investments reduce the perceived
threat of demolition and offer de facto tenure
security which, in turn, encourages residents
to invest in their units. In our framework, ten-
ure and infrastructure interact, and, in contrast
to debates in the literature, we posit that either
one can lead to improvements in the other. In
other words, the framework allows for the con-
sideration of multiple directions of causality
and different entry points—the ‘‘correct’’ solu-
tion, we contend, is likely to vary by context.
Finally, tenure is related to a neighborhood’s
condition and its location. Neighborhoods with
better conditions—public amenities (e.g.,
parks, schools), safety, layouts that permit ac-
cess—are likely to be associated with greater
tenure security. The location matters as well—
for instance, slums on hazardous sites (and, of-
ten, those in prime downtown locations) are
more likely to be slated for demolition, even
by governments that otherwise support slum-
improvements.
Overall, the framework contributes to exist-
ing literature in three ways. First, it creates a
composite that offers a more comprehensive
and multi-dimensional approach to document-
ing, and understanding quality. Documenting
indicators related to the four dimensions of
the Living Conditions framework can together
provide a good overview—a snapshot at one
moment in time—of conditions in a settlement.
These can also be converted to a numerical in-
dex, by assigning weights to various indicators.
Second, it emphasizes interactions among the
variables that are usually treated or studied in
isolation, and allows for a dynamic understand-
ing of living conditions. A change in one factor
can influence changes in the other three, and
conditions change over time. Therefore, if data
are collected at periodic intervals and plotted
using the Diamond, the relationship between
various factors can be explored and changes
in quality analyzed over an extended time-
frame. By drawing attention to the dynam-
ics—between variables and over time—the
framework helps emphasize that settlements
can follow different development trajectories,
and that these can shift with time (e.g., ‘‘slums
of hope’’ can become ‘‘slums of despair,’’ and
vice versa).
Third, it provides a conceptual model that
can not only accommodate but also help ex-
plain variation in quality of living conditions
among settlements. It allows us to move be-
yond the simplistic notion that slums are homo-
geneously poor in quality, and facilitates
comparative analyses that can reveal why they
differ. It also moves us away from seeking ‘‘uni-
form’’ or globally applicable definitions of
slums—of the kind the UN-HABITAT (2003)
suggests we establish—and toward a more con-
text-specific understanding of them. 8
In the next sub-sections, we discuss how the
framework can be used to define slums, and,
then, use it to assess quality of living conditions
in Nairobi’s slums. We show that each of the
four dimensions is in itself a composite of sev-
eral factors and requires the use of multiple
indicators. Our selection of indicators is illus-
trative rather than prescriptive or comprehen-
sive. We return to the framework in Section 5
to demonstrate how it serves as an analytical
tool and provide an understanding that is dif-
ferent from current conceptualizations.
(a) Defining and documenting slums using the
framework
Slums score poorly on one or more of the four
dimensions of the Living Conditions frame-
work. In fact, in most parts of the developing
world, slums have some combination of the fol-
lowing four characteristics in early stages of
their development: (i) informal or illegal land
tenure; (ii) physical infrastructure and ser-
vices—such as water supply, electricity, drain-
age, sanitation, and street lighting—that are
all highly inadequate; (iii) housing units built
with poor-quality or impermanent construc-
tion materials, usually in violation of various
standards; and (iv) neighborhood conditions
and location that are poor, and often also fail
to meet legally specified regulations and plan-
 SLUM REAL ESTATE
ning standards. If defined very broadly, the
term ‘‘slum’’ can include settlements that have
any of the above characteristics.
In this paper, we use a narrow definition of
slums, focusing on those that score poorly on
all four dimensions of the framework—tenure,
infrastructure, unit quality, and neighborhood
conditions. In the slums included in our study,
housing and infrastructure are visibly below
minimum standards and land tenure is informal
or illegal. This definition excludes, for instance,
housing that is built to standard but is deemed
informal because it lacks official paperwork or
violates land use and zoning regulations. De-
fined as such, Nairobi’s slums are a subset of
informal settlements (since all slums are infor-
mal, we use the terms interchangeably in this
paper). In the following sub-sections, we exam-
ine each of the four dimensions of the Diamond
in turn, to create a multi-dimensional snapshot
of living conditions in Nairobi’s slums.
(b) Tenure in Nairobi’s slums: tenuous, short,
and disproportionately rental
In Nairobi’s slums, we find that the tenure
mix is unusually skewed toward tenancy—
92% of the households are rent-paying tenants
and 8% are owner-occupiers. Only 6% claim
they own both their house and their land, while
2% say they own the structure but not the land.
Within this small group of owner occupants,
60% rent out at least one room and are ‘‘resi-
dent landlords.’’ Resident landlords thus ac-
count for 5% of the slum population,
indicating that the vast majority are ‘‘absentee
landlords’’ who do not live in the slums. Local
experts emphasize that the technically correct
term is ‘‘structure owner’’ rather than ‘‘land-
lord’’ because almost none of them own the
land (the terms landlord and structure owner
are used interchangeably in this paper).
Tenancy contracts in slums are based almost
entirely on verbal agreements—only 3.6% of
the renters have a formal written tenancy agree-
ment. Their tenancy is, hence, fragile in legal
terms. It is also short in duration—we find that
slum residents have lived in their current home
for an average of five years and in their current
settlement for about nine years. In other words,
the turnover rate is high. About 9% report that
they have been evicted at least once. Neverthe-
less, about half the residents feel that they have
secure tenure. The latter, combined with the
fact that tenants move frequently, suggests that
there is a fairly well-established rental ‘‘mar-
1921
ket’’ or system, with enforceable rules of the
game.
The high tenancy rate in the slums reflects a
larger problem facing the city as a whole—
82% of the city’s residents are tenants (Census,
1999), indicating that households across differ-
ent income classes have difficulty owning prop-
erty in Nairobi. 9 This means, for example, that
programs to regularize tenure or create owner-
ship for slum residents must consider potential/
latent demand for such assets from better-off
citizens. In this context, slum upgrading pro-
grams that create assets—such as housing with
legal title—for slum residents that the better-off
lack are likely to be the subject to gentrifica-
tion, unless they are explicitly scaled and de-
signed to counter such effects.
(c) Housing size and quality—crowded and
constructed with impermanent materials
Crowding, as measured by persons per room,
is far worse in the slums than in Nairobi city
and the nation as a whole. There are 2.6 per-
sons per room in Nairobi’s slums as compared
to 1.7 for Nairobi, and 1.55 for Kenya. Fur-
ther, almost the entire stock of slum housing
is constructed with poor-quality and imperma-
nent building materials. For instance, only 12%
of the units have permanent external walls—
constructed with stone, brick, or block. This
quality indicator compares unfavorably with
the Nairobi average of 56%, and also with the
national average of 26% (Census, 1999). The
latter is worth noting, given that housing qual-
ity in rural areas is known to be very poor.
(d) Infrastructure access and service delivery:
worse than anticipated
There is almost universal agreement—be-
tween policy makers, academics, city residents,
development agencies, and other experts—that
Nairobi’s slums have inadequate infrastructure.
At the same time, aggregate data on infrastruc-
ture access from the National Census of 1999
indicate that the city’s residents (and, by exten-
sion, its slum residents) are far better-off than
those in other parts of the nation. This raises
the question: just how inadequately served are
Nairobi’s slum residents?
To provide a systematic and empirical over-
view, we plot ‘‘access’’ indicators for six types
of infrastructure as an ‘‘infrastructure services
polygon’’ (see Figure 2). The polygon shows
that access to public infrastructure in the slums
1922 WORLD DEVELOPMENT
Figure 2. Infrastructure services polygon. Note: Y: yard tap, H: in-house water connection, P: public sewer, S: soak pit.
is appalling—less than a quarter of the polygon
is shaded, demonstrating that less than quarter
of the slum population has access to any of the
six services examined. This falls far short of the
normative goal of universal access to these ba-
sic services, reflected as fully (100%) shaded
polygon. The polygon also reveals variation
among services—it shows, for instance, that
about 22% have electricity connections but only
0.9% benefit from public garbage collection ser-
vice. We discuss some of the service-specific
findings from Table 1 to provide an insight into
(a) inequality in access in the slums relative to
the city; and (b) the types of (informal) options
the unserved are forced to rely upon.
(i) Water: cash-and-carry service, low use levels,
and high-prices
Previous studies have argued that the urban
poor, especially those residing in slums, are
poorly served by public water utilities. In Nai-
robi’s slums only about 19% of the households
are directly connected to the public water util-
ity—4% have private piped connections, and
15% have yard taps (Table 1). By comparison,
the aggregate city-wide water connection rate
is estimated to be 71–72% (Collignon & Vézina,
2000). In other words, there is a gap of over
50% points in utility connection rates between
the city and its slums.
The majority—64%—of slum residents,
therefore, rely on kiosks and buy water by the
bucket (or 20-l ‘‘jerrycans’’). About 5% rely
on ground or other natural sources, 2% on ven-
dors, and 1% on neighbors. Given the reliance
on sources such as kiosks, the unit cost of water
borne by slum residents is high, and their use le-
vel is low. Specifically, water use averages
about 23 liters per capita per day (lcd), and
the median value is 20 lcd. Slum residents
pay, on average, Ksh130/m3 (US$1.73/m3) for
their water, and the median cost is KSh100/
m3 (US$1.33/m3). Based on data from other
studies on water use and prices in Kenya (see
Gulyani, Talukdar, & Kariuki, 2005; Thomp-
son et al., 2000), we can categorically conclude
that Nairobi’s slum residents are seriously un-
der-served.
(ii) Reading by kerosene lamps: few electricity
connections
Only one in five homes in Nairobi’s slums
(22%) is connected to electricity (Table 1). Con-
nections rates are significantly higher among
non-poor households (31%) compared to the
poor (18%). Among the 78% who are uncon-
nected, almost all rely on kerosene as their pri-
mary lighting fuel. Use of electricity as a
lighting fuel—an indicator of its relative ‘‘ac-
cess, availability, and affordability’’—is far
lower in Nairobi’s slums compared to the city
and to urban Kenya, but still higher than in
rural Kenya.
(iii) Toilets: shared facilities with high loading
levels
Only about one-quarter of the households
have access to a private toilet facility. The
majority (68%) of slum residents rely on shared
toilet facilities, and loading factors are high. On
average, 19 households or 71 people depend on
one shared toilet. About 6% of the households
are even worse off than those sharing facili-
ties—they have no access to toilets and have
 SLUM REAL ESTATE 1923

Table 1. Infrastructure and services

All Poor Non-poor
N Value N Value N Value
Water use and unit cost
Mean per capita water use, liters per day*** 1750 23.4 1277 21 473 29.8
—median use, liters per day 20.0 20 20.0
Mean unit water cost, USD/m3** 1750 1.7 1277 1.7 473 1.8
—median cost, USD/m3 1.3 1.3 2.0
Primary sources of water*** (%)
Private piped 67 4 47 4 18 4
Yard tap 267 15 185 14 82 17
Kiosk 1127 64 838 65 289 61
Others (incl. missing) 294 17 212 17 84 18
Total 1755 100 1282 100 473 100
Connections to electricity(%) 1755 1282 473
Households connected ** 22 18 31
HHs aware of informal connections 44 50 44
Primary source of home lighting** (%) 1755 1282 473
Electricity 22 19 30
Kerosene 77 80 70
Solar/others 1 1 1
Street lighting(%) 1755 1282 473
HHs who say there is street lighting 16 15 15
Toilet facilities—Type (%): 1755 1282 473
No facility/flying toilets 6 6 6
Individual toilet (VIP/ordinary pit/WC) 25 24 26
Shared/public toilet 68 68 67
Average no. HHs sharing a toilet facility (no.) 1615 19.1 1180 19.4 435 18.5
Average no. people sharing toilet facility* (no.) 1613 71.3 1181 73.8 432 64.7
Excreta disposal system (incl. sewerage)* (%) 1755 1282 473
Formal NCC connection to public sewer 12 11 14
Informal connection to public sewer 17 18 14
Septic tank/soak pit 1 0 2
Pit latrine 64 64 63
Garbage disposal system (%) 1755 1282 473
Dumping in own neighborhood 78 79 76
Burning/burying in own compound 10 10 9
Organized private collection system 11 10 13
City collection system 1 1 1
Gray water disposal/drainage (%) 1755 1282 473
Pour into the drain 71 71 70
Pour into the road, or pavement 19 19 17
Pour into pit latrine 1 1 1
HHs with a drain outside their house 58 56 61
HHs whose drain works most of the time 25 25 26
Note: Significance of difference between poor and non-poor indicated by asterisks: ** = 5%, * = 10%.
Figures may not add to 100% due to rounding.

to use open areas or ‘‘flying toilets’’ (i.e., plastic
bags that are tied up and flung away).
(iv) Neighborhood-level sanitation infrastructure
and services: a health crisis?
Access to various neighborhood-level ser-
vices is no better, and the state of sanitation-re-
lated services is the cause for alarm. First, solid
waste disposal services barely exist. Less than
one in a hundred households (0.9%) is served
by a publicly-provided garbage collection sys-
tem; consequently, most households dump or
burn waste in their immediate vicinity. Second,
about 12% of households say they are
1924 WORLD DEVELOPMENT
connected to a public sewer, and 1% have a
soak pit; the majority rely on pit latrines that
contaminate ground water if they are unlined
or when they overflow. Third, only 25% have
a working drain outside their house; stagnant
water is a health hazard and, during the rainy
season, flooding of homes is common.
(e) Neighborhood and location: unsafe from
crime in their own neighborhood
For brevity, we report just one neighbor-
hood-level indicator—safety. 10 As many as
63% of households report that they feel unsafe
inside their slum and 27% report that someone
in their household actually experienced a crim-
inal incident over the previous 12 months.
Among those who were victims, the majority
report that the incident occurred inside their
own settlement. The average number of inci-
dents experienced by a victimized household
in the previous year was 1.67, with an average
of 1.2 occurring inside the slum and 0.46 else-
where.
(f) Conclusions on quality
In sum, the quality of living conditions is
unambiguously bad, with appalling indicators
on each of the four dimensions—tenure, infra-
structure, unit quality, and neighborhood—of
the living conditions diamond (see Figure 1).
In the existing literature on slums, it is widely
accepted that such low quality and standards
help keep housing affordable (e.g., Gilbert,
2003; Gilbert, Mabin, McCarthy, & Watson,
1997). Intuitively, quality and price should be
correlated—the better the standard of living
conditions, the higher the market price or per-
ceived value of a unit or neighborhood. In the
next section, we test whether poor quality is
indeed keeping rents low and affordable in
Nairobi’s slums, and also examine which as-
pects of quality are correlated with higher
rents.
4. THE SLUM RENTAL MARKET
Several analysts have highlighted the impor-
tance of rental housing in providing shelter
for the poor (e.g., Gilbert, 2003; Kumar,
1996; Miraftab, 1997). Some even argue that
the expressed demand is for low-cost rental
accommodation and strongly oppose the
emphasis, in development planning, on home-
ownership schemes (Hansen, 1997). Yet others
protest the stereo-typical portrayals of land-
lords as villains, arguing that there is wide-
spread prevalence of petty landlordism that
generates only small profits while facilitating
consolidation in self-help settlements (e.g., Dat-
ta, 1995; O’Connor, 1983). In Gaborone, Bots-
wana, for example, women outnumbered men
as landlords among the urban poor in self-help
housing areas and they earned relatively low
rents (Datta, 1995). Similarly a small-scale
study of two low-income settlements in South
Africa found that ‘‘few landlords make any
money at all . . . [and] the poor quality of
accommodation keeps rents low . . .’’ (Gilbert
et al., 1997). Based on a review of such analy-
ses, the UN-HABITAT (2003) concluded that
‘‘aid programs for rental tenure remain a ne-
glected element of international assistance,
and knowledge about informal landlords and
tenants and the kinds of programs that might
benefit them are rare.’’
By contrast, among local experts, and policy
makers, 11 there is a raging debate over whether
the rents are too high, and whether the Nairobi
slum rental market, allegedly dominated by
politically well-connected landlords, is helping
or hurting the poor. Based on analyses pre-
sented in this section, we will argue that the ren-
tal market in Nairobi’s slums is not serving the
poor well, and that it is highly desirable to
reach a better outcome. Although the case
poses a major challenge for proponents of ren-
tal housing, 12 it also contributes to this litera-
ture by revealing conditions under which
rental housing can hurt rather than help the
poor.
In this section, we take a closer look at rents
and behavior of the rental market. Ascertaining
whether ‘‘rents are too high’’ is, however, not
easy; it requires an examination from two
opposite perspectives—from the demand-side
to assess affordability, and from the supply side
to assess returns on investment. We start by
examining rent levels and affordability, cash
flow in the rental market, and factors influenc-
ing rents. On the supply side, we examine rates
of return and review what we know about land-
lords.
(a) Rents are not highly affordable and operate
within a narrow band
At first glance, the slums appear to offer
accommodation at a wide variety of prices.
As Table 2 shows, the maximum reported
 SLUM REAL ESTATE 1925

Table 2. Rents in different areas (administrative non-poor (Figure 4). The non-poor do spend
divisions) of Nairobi more on rent—but, surprisingly, only about
N Mean Median Min. Max. SD 20% more—than the poor. On average, the
non-poor pay a monthly rent of Ksh 913
Dagoretti 295 1090 900 300 3230 570
(US$12) or about 11% of their income, while
Westlands 57 1080 1000 400 3000 590
the poor pay Ksh 753 (US$10) accounting for
Pumwani 111 890 700 300 3000 590
Central 189 860 700 260 2600 400
about 12% of their income; the median rent is
Makadara 170 780 750 400 2000 200
Ksh 750 (US$10), and Ksh 700 (US$9) for
Embakasi 167 670 600 200 3200 330 the two, respectively (Table 3).
Kasarani 182 640 500 250 2000 320 The relatively small difference in rental
Kibera 450 620 500 300 2000 260 expenditures between the poor and non-poor
reflects in our view, lack of choice. First, as a
Total 1621 790 700 200 3230 440 comparison of Figures 3 and 4—cumulative
plots of rents and incomes—graphically re-
veals, there is a significant gap in the income
monthly rent is Ksh 3,230 (US$43.1) which is curves for poor versus non-poor households,
about 16 times higher than the minimum rent but the profile of rents paid by the two groups
of Ksh 200 (US$2.7). It also shows, as one is virtually identical. Second, there is a large
might expect, that rental values differ among difference in their absolute expenditures in
the eight administrative divisions of the city. other categories such as food, water, and trans-
Rents are, on average, highest in the slums lo- port. Compared to the poor, the non-poor
cated in Westlands and Dagoretti, and lowest spend about twice as much on food per person
in those located in Kibera and Kasarani. A clo- per day, 1.5 times more on water, and three
ser examination suggests, however, that major- times more on transport (Table 3). The diver-
ity of the rents fall within a narrow range, and, gence in transport expenditure patterns is par-
as noted earlier by Amis (1984), the typical rent ticularly striking. The median absolute
is anything but low or easily affordable. expenditure on transport is zero for poor
To assess affordability, we looked at income, households—that is, half of all poor house-
and expenditure levels of residents (Table 3). holds spend nothing on transport, and opt to
We know that 73% of slum households fall be- walk; they appear to be coping by cutting back
low the absolute poverty line, defined as an on a ‘‘luxury’’ such as transport. In other
expenditure of Ksh 3,174 (US$42) per adult words, the poor spend far less than the non-
equivalent per month, excluding rent. With re- poor when they have discretion; with respect
spect to income, households reported an aver- to housing, the poor seem not to have discre-
age monthly per capita income of Ksh 3,705 tion—the market does not offer cheaper, or
(US$49). The median income is Ksh 3,000 more affordable units. This appears to be a
(US$40)—that is, half of the slum households key reason why rental expenditures differ little
earn less than US$40 per capita per month. between the poor and non-poor. Overall, this
As one would expect, poor and non-poor discussion suggests that even the cheapest slum
households differ significantly—compared to housing, although less expensive than other op-
the poor, average income and expenditure tions in the city (e.g., Huchzermeyer, 2006), 14
among the non-poor is twice as high (see Table is probably not cheap enough for its poor resi-
3 and Figure 3). dents.
Regarding rents, four points are worth high-
lighting. First, rent constitutes a major expense (b) The multi-million dollar slum rental market
for slum households—second only to food—
and it accounts for about 12% of monthly Rents constitute a significant expense for
household income on average. 13 Second, the these relatively poor people individually, but
typical rent is significant in absolute terms— how much is this real estate worth cumula-
the average monthly rent is Ksh 790 (US$11), tively? In 2004, slum residents paid at least
and the median rent is Ksh 700 (US$9). Third, US$31 million (Ksh 2.3 billion) in rents. This
rents move within a narrow band overall. Apart figure is estimated using an average rent of
from the outliers on either end, the vast major- Ksh 790 per month per household, a tenancy
ity (80%) pay rents between Ksh 400 and 1,200 rate of 92%, and a slum population of
(US$5.3–16.0). Fourth, there is only a small dif- 810,000. This is a very large sum—it exceeds,
ference between the rents paid by the poor and for example, Nairobi city’s annual budget for
 1926
Table 3. Incomes and expenditures
All HH Poor HH Non-poor HH

WORLD DEVELOPMENT
N Mean Median N Mean Median N Mean Median
Per capita income in Ksh/month ** 1,400 3,705 3,000 1,011 2,776 2,444 389 6,023 5,500
Per capita expenditure on basics, typical month (Ksh) ** 1,754 2,493 2,000 1,281 1,874 1,750 473 4,121 4,000
Per capita spending on food per day (Ksh) *** 1,755 43.00 40.00 1,282 34.31 31.00 473 66.04 60.00
Per capita spending on transportation per day (Ksh) *** 1,754 9.68 4.00 1,281 5.98 0.00 473 19.44 15.00
Per capita daily expenditure on water (Ksh) *** 1,754 2.95 2.40 1,281 2.57 2.00 473 3.95 3.00
Rent per month (Ksh) ** 1,601 797 700 1,173 753 700 428 913 750
Household monthly spending on refuse collection (Ksh) 161 51 40 117 51 40 44 51 40
Household spending per month on electricity (Ksh) ** 362 286 250 218 295 275 144 274 200
Note: (i) Statistical significance of difference between poor and non-poor indicated by asterisks. *** = 1%; ** = 5%, and * = 10%.
(ii) The relevant exchange rate is US$1 = Ksh75.
 SLUM REAL ESTATE 1927

Figure 3. Cumulative plot of per capita household income. Note: households are categorized as ‘‘poor’’ or ‘‘non-poor’’
based on a ‘‘0,1’’ classification using an expenditure-based (not income-based) poverty threshold.

Figure 4. Cumulative plot of typical monthly rent.

investment as well as operations and mainte-
nance. 15 If viewed as repayments on a com-
mercial housing development loan, this cash
flow would be sufficient to leverage, and pay
off a 20-year loan of about US$358 million, at
an interest rate of 6%. Unfortunately, very little
of this vast cash flow is currently being invested
or leveraged to improve living conditions in the
slums; this is evident from the previous section
on quality.
(c) What drives rents in Nairobi’s informal
housing market?
We find that rents vary not only between
slums located in different parts of the city but
also within each settlement. To identify vari-
ables that drive rents in slums, we conducted
a multivariate regression analysis with
‘‘monthly rent’’ as the dependent variable. The
complete results of this hedonic-type analysis
1928 WORLD DEVELOPMENT

are presented in Table 4 and the main findings
are discussed below.
The results show that this model—comprised
of independent variables corresponding to the
four dimensions of the living conditions dia-
mond—explains 50.5% of the observed varia-
tions in rents. Given that the regression
analysis is based entirely on cross-sectional
data, the explanatory power of this rent regres-
sion model is quite high. By using a step-wise
regression approach, we examined the relative
explanatory power of the four categories of
variables. Variables related to the unit size
and quality have the greatest explanatory
power—they explain about 32% of the varia-
tion in rents. Infrastructure variables explain
an additional 8% of the variation, neighbor-
hood conditions and location account for
about 10% and tenure-related variables explain
about 0.5% of the total variation in rents. In the
discussion below, we take a statistical signifi-
cance of 10% or less as the threshold value
for determining whether a variable is systemat-
ically correlated with rent.
Characteristics of the rented unit strongly
influence rent. A unit’s size, in terms of number
of rooms, is strongly and positively related to
rent. Housing quality also matters—rents are
higher for units that received a higher quality
rating from enumerators, and also for housing
constructed with permanent building materials.
Units with external walls of stone, brick, or
block have higher rents than those with walls
of wood, mud, or tin. Similarly, units with ce-
ment or wooden floors have higher rents than
those with mud floors.
Unit-level infrastructure access is also corre-
lated with rent. Units with an electricity con-
nection, and those with an in-house or yard
connection for water, have higher rents than
units without these amenities. Similarly, rents
are higher for units that provide ‘‘reasonable’’

Table 4. Regression for monthly rent (N = 1601)

Coefficient Std. Err. P-value
(1) Unit size and quality
Number of rooms for household 314.270*** 64.096 0.000
Assessor’s estimate of building quality 77.894*** 29.560 0.010
Permanent walls 305.829*** 67.993 0.000
Concrete roof !78.396 95.140 0.412
Permanent floor 115.384*** 21.733 0.000
(2) Infrastructure: unit and neighborhood-level
Has electricity 241.176*** 26.683 0.000
Has piped water 73.353** 35.664 0.043
Reasonable access to toilet 33.523* 19.704 0.092
Has drain outside !18.418 16.080 0.255
Internal roads unpaved !16.021 39.745 0.688
Street lights exist 42.600 71.856 0.555
(3) Neighborhood conditions and location
Settlement has been improved !91.958 68.597 0.184
Public school present in neighborhood 48.946* 26.482 0.068
Considers area safe !15.872 15.190 0.299
Neighborhood location (ii) (ii) (ii)
(4) Tenure
Perceives tenure to be secure !20.448 21.295 0.340
Length of stay in unit (in years) !2.689 1.895 0.159
Constant 150.388 99.620 0.135
Adj. R2 = 0.505
Notes: (i) The regression model is estimated using the sampling weights for households. Statistical significance of
coefficients is indicated by asterisks: *** = 1%, ** = 5%, and * = 10%.
(ii) The results for the 24 ‘‘location’’ dummy variables are not shown here to conserve space. Out of the 24, 15 show a
statistically significant difference in rents as compared to the location selected as ‘‘base’’ case (viz., Mathare).
 SLUM REAL ESTATE
access to a toilet facility, defined as those where
the renter shares a toilet facility with less than
ten households (as noted earlier, the average
loading is 19 households per facility).
In terms of neighborhood-level infrastruc-
ture, we find that drains, street lights, and
paved internal roads do not affect rents system-
atically. This is not to suggest that such infra-
structure has no value in terms of rent.
Rather, as our additional analyses suggest, this
is mostly because infrastructure facilities vary
more across locations than within a settlement,
and because of multicollinearity with the ‘‘pub-
lic school’’ variable. This means that any sys-
tematic effects of neighborhood-level
infrastructure are being mostly captured by
the ‘‘location’’ and ‘‘public school’’ variables.
Among other ‘‘observed’’ neighborhood
characteristics, or features for which we con-
trolled, schools matter—rents are higher in set-
tlements with a public school. This may be
related to the fact that, in 2003, the government
introduced free primary education in public
schools; if so, landlords are benefiting from
government investment in education. We find
that residents’ perception regarding the relative
safety of their neighborhood (in terms of crime)
is not correlated in any significant way to the
rents that they pay.
Location of a slum directly influences rent.
Of the 23 location dummies, 15 are statistically
significant—compared to Mathare (the ‘‘base,’’
or ‘‘comparator’’ case), rents are systematically
higher in five locations and lower in the remain-
ing ten. The results strongly underscore that
even after controlling for several observable
characteristics of the house, infrastructure,
and neighborhood, the location imposes a
‘‘premium’’ or a ‘‘discount’’ in the home rental
markets across Nairobi’s slums.
Finally, we find no systematic relationship
between rent and the renter’s perception of ten-
ure security. Nor do we find a relationship be-
tween rents and the length of a household’s
stay in their unit. The former does not suggest
that tenure security has no rental value. Rather,
as in the case of neighborhood-level infrastruc-
ture such as drainage, tenure security also var-
ies more across locations than within a
settlement. The ‘‘location’’ dummies appear to
pick up the systematic effects of tenure security
on rent.
In sum, we find a formal (but distorted) real
estate market in the midst of informality. Rents
in Nairobi’s slums increase with number of
rooms and are higher for units that have per-
1929
manent walls and floors, as well as those with
connections to electricity or piped water. They
are higher in neighborhoods that have a public
school, and the sub-division in which a unit is
located imposes a premium or discount on the
rental value. Overall, as in formal real estate
markets, rents in Nairobi’s slums depend on
or vary with a unit’s size, location, construction
quality, and access to infrastructure.
(d) Supply side issues: extraordinary returns and
entry barriers?
Previous studies suggest that the business of
being a slum landlord in Nairobi is highly prof-
itable. In his study of landlordism in Nairobi’s
slum of Kibera, Amis (1984) estimated that
‘‘[In the case of] a 10-room structure, the
[undiscounted] annual capital return was 131
percent. Thus after only nine months the land-
lord’s income is pure profit, since the mainte-
nance, and running costs are more or less
non-existent.’’ In a recent rapid appraisal com-
missioned by UN-HABITAT, Syagga, Mitul-
lah, and Karirah-Gitau (2002) also concluded
that in Kibera—with an average monthly rent
per room of Ksh 1,300—it takes only 9.75
months for an investor to recoup the Ksh
12,686 investment entailed in constructing a
room. Other informal settlements have lower
rents but returns are still excellent—annual re-
turns range from 60% to 80% and payback
periods average 16 months.
We find that rents are lower than those re-
ported by Syagga et al. (2002). Our data show
that rents in Kibera average about Ksh 620
(US$8) and they are lower than the average
rent in the city’s slums (Ksh 790 or US$11) (Ta-
ble 2). If so, then the pay back period for a Ksh
12,686 investment for a single room would be
20.4 months rather than 9.75 months—still
high, but not quite as high as the rapid apprai-
sal suggests. It is, therefore, worth moving be-
yond rapid appraisals and toward more
rigorous assessments of the costs and benefits
of being a slumlord in Nairobi; our data
provide a crucial building block for such anal-
yses.
High returns are often indicative of high en-
try barriers. Entry into the business of being a
slumlord in Nairobi entails, at a minimum, con-
struction or purchase of a unit. And this,
according to existing studies (Amis, 1984; Syag-
ga et al., 2002), requires (i) political connec-
tions; (ii) payment of significant (illegal) fees
to get permission to build or approve a unit;
1930 WORLD DEVELOPMENT

and (iii) willingness to bear the risk of demoli-
tion and loss of capital. What is the profile of
people who have successfully overcome such
barriers to own a shack?
(e) Who owns the shacks? Owner-occupiers and
landlords
In his study of Kibera’s landlords, Amis
(1984) noted that there is an informal system
for allocating building rights and that it is an
important source of political patronage for pol-
iticians and the local administration (i.e., gov-
ernment employees). He also found that some
landlords owned multiple units (e.g., four of
the largest landlords owned more than 100
units each), and that many of the large land-
lords were from the administration. Syagga
et al. (2002) came to similar conclusions: much
of the squatter land in Kibera has been ac-
quired or allocated by politicians and govern-
ment employees with enough influence to
ensure that they are not displaced. In their sam-
ple of 120 landlords, 41% were government offi-
cers, 16% were politicians, and 42% were other
absentee owners. Only a handful of the struc-
tures belonged to people who lived in the slums.
This is unlike the situation in Mathare and
Pumwani, the report notes, where a large num-
ber of investors are residents who ‘‘lived at a le-
vel fairly similar to their tenants and
demonstrated a keen interest in maintaining
the community and improving it’’ (Syagga
et al., 2002).
We find a somewhat larger proportion of
both owner-occupiers (8% of residents), and
resident landlords (5% of residents). The 90 res-
ident landlords in our sample report an average
of 5.9 rent-paying tenants (Table 5). Disaggre-
gated by welfare, resident landlords who are
poor report an average of 4.6 rent-paying ten-
ants as compared to 8.4 tenants among non-
poor landlords. When asked to estimate the
sale value of their home, owner-occupiers who
are poor placed the average value at Ksh
239,203 (US$3189), as compared to Ksh
723,977 (US$9653) among the non-poor. Their
own estimates can provide an alternative basis
for understanding decisions by slum investors
regarding whether to build or not, and whether
to sell or rent; they can also be used to under-
stand what kind of deal the landlords may or
may not be willing to accept, if the government
were to come in with proposals for slum
upgrading. More research is required to im-
prove our understanding of both resident and
absentee landlords, their incentives, and what
it would take to have them invest in improving
living conditions in the slums.
5. SOLVING THE HIGH-COST
LOW-QUALITY PUZZLE WITH THE
FRAMEWORK
The above discussion suggests that Nairobi’s
slums are stuck in a high-cost, low-quality trap.
The tenants are paying a lot for very poor qual-

Table 5. Owner-occupiers and resident landlords

All Poor Non-poor
N Percent (%) N Percent (%) N Percent (%)
Feel tenure is secure 150 62.2 105 62.3 45 61.8
Believe its easy to sell or buy property in neighborhood* 150 33.7 105 38.2 45 23.3
Aware of a property sale in vicinity in last 12 months* 150 33.9 105 38.0 45 24.1
Mean expected value of property if sold (Ksh)** 150 398,400 105 239,200 45 724,000
Type of ownership document 150 105 45
–None 36.6 37.2 35.0
–Temporary occupation license 19.2 22.7 11.0
–Freehold title 11.2 8.1 18.6
–Certificate of title (long-term lease from city 7.8 7.8 7.6
council/govt.)
–Letter from the chief (provincial administration) 4.8 6.2 1.5
–Others including share certificate 20.4 17.9 26.3
Rent out rooms (i.e., are resident landlords) 150 60.0 105 57.5 45 65.9
Mean number of rent-paying tenants 150 5.9 4.6 8.4
Note: Statistical significance of difference between poor and non-poor indicated by asterisks: *** = 1%, ** = 5%, and
* = 10%.
 SLUM REAL ESTATE
ity housing. Despite the large cash flow in the
system, and the fact that better quality units
command higher rents, landlords are not rein-
vesting to significantly improve the housing
stock. It is easy to vilify the landlords and they
often are—by local media, and by local and
international experts. In our view, however,
the quality problem cannot be blamed on the
landlords alone. Even if the landlords were,
for instance, nicer, or better regulated, or sub-
ject to somewhat more competition, this would
not, we argue, lead to large gains in quality of
living conditions for current slum residents.
(a) The framework as a diagnostic and analytical
tool
As the living conditions diamond visually
shows, the quality problem is not limited to
the unit alone. Infrastructure services and
neighborhood conditions—two of the other
dimensions of the diamond—have very poor
quality indicators as well, and these too have
not been improving over time. This stubborn
lack of improvement and consolidation in Nai-
robi’s slums stands in contrast to the experience
in slums of many other cities, such as Dakar
and Rio de Janeiro, where conditions have im-
proved steadily over time.
We argue that the current trap is the result of
an interaction of factors along different dimen-
sions of the living conditions diamond. Together
these factors have created severe disincentives
for action and investment for all three of the
key stakeholders in Nairobi’s slums—landlords,
tenants, and government. A key piece of the
problem, and by extension the solution, lies in
two variables associated with tenure. But it is
not the usual suspects—illegal land tenure or
distortions in the rental market. Rather, in Nai-
robi, the problem is that the tenure mix is highly
unusual—there are too many tenants and too
few owner-occupiers and resident landlords—
and the duration of stay of an average resident
is relatively short. To offer one comparison, in
Dakar’s slums the proportion of tenants to own-
er occupants is 29:71 as compared to 92:8 in Nai-
robi. 16 Further, the average resident in Dakar’s
slums has stayed in the same house for 19 years,
as compared to an average stay of five years in
Nairobi’s slums. The combination of excessive
rental in the tenure mix and a high turnover rate
among slum residents has weakened the incen-
tives for landlords and tenants to invest in
improving their unit, the infrastructure, and/or
their neighborhood.
1931
What are the (dis)incentives for improving the
units? Tenants are not only poor but also mo-
bile, and these factors appear to reduce their
ability and willingness to invest in their units.
If tenants believed that could stay longer, or if
they had rights vis-à-vis their landlords, they
may have invested in the units even though they
do not own them. What about the landlords? Is-
sues pertaining to the competitiveness of the
rental market aside, the fact that majority of
the landlords are ‘‘absentee’’—in that they do
not live on site—means that they neither suffer
from the appalling living conditions nor gain
any of the prestige associated with owning and
living in a good-quality and well-functioning
home. This eliminates a major incentive for
investment in their units. For instance, if the
roof was constantly leaking during the rainy
season, a resident landlord would have a greater
incentive to replace it than a non-resident owner
who neither experiences the misery it causes, nor
sees it, nor feels any social connection to or
pressure from the tenant to fix the problem.
There are, however, financial incentives—if
landlords invest in improving the quality of
their units they can get higher rents. Why is this
incentive not working? One possible reason is
that investing heavily in units may attract un-
wanted attention from the authorities. Land-
lords are likely to be averse to such attention
for fear that the government will assert its rights
over the land, and directly or indirectly, cut into
their profitable business (e.g., by extracting
kick-backs, trying to regulate it, tax it, or,
worse, declare it illegal and re-possess the land).
Not investing in the units allows them to stay
‘‘under the radar’’ or relatively invisible.
What about incentives for improving infra-
structure or other neighborhood conditions?
Landlords and tenants could, for instance,
potentially lobby government for such
improvements. It seems, however, that the ab-
sence of government in service provision has
created profitable infrastructure businesses for
landlords and eroded their incentives for seek-
ing improvements. Local experts suggest—and
our data and other studies support their obser-
vation—that landlords own and control not
just the housing units but also many of the yard
taps, water kiosks, in-house connections for
electricity and water, and many of the toilet
facilities (including ‘‘public’’ pay-per-use toi-
lets), and all these are known to be good busi-
nesses. 17 If so, landlords are unlikely to want,
let alone ask, government to improve services
in the slums; such interventions would directly
1932 WORLD DEVELOPMENT
reduce the scale and scope of their infrastruc-
ture-related businesses. In other words, land-
lords are strongly likely to prefer—and work
to maintain—the status quo.
Tenants do have incentives to seek infrastruc-
ture and neighborhood improvements but, thus
far, wide-spread tenant action has not material-
ized. Even though they move their home often,
tenants tend to stay in the same settlement
longer—on average about nine years—and they
are the ones who suffer most directly from the
horrible neighborhood conditions and inade-
quate access to infrastructure. They can poten-
tially organize to improve their neighborhood
in ways that do not entail heavy cash invest-
ment—for example, by organizing around gar-
bage collection or community policing, and by
using their power to vote to pressure govern-
ment into improving services. Certain slum
communities are indeed organizing and mobi-
lizing with assistance from NGOs, but such ef-
forts are yet to go to scale.
In improving infrastructure and neighbor-
hood conditions, however, the key actor is the
government. Infrastructure and neighborhood
amenities have public good characteristics
meaning that, even in conditions highly condu-
cive to private investment, public investment is
required to achieve optimal levels of service.
These improvements require more than tacit ap-
proval—public utilities and infrastructure agen-
cies need approval to connect slum residents
and also need significant budget to make this
happen. Unfortunately, the government has
not invested—there have been no large-scale,
let alone city-wide, government-supported pro-
grams to improve public infrastructure in the
slums. This is, in part, attributable to the com-
plex political economy in the slums. Given the
high proportion of tenants and the politically
well-connected landlords, it is hard to design
win–win solutions and any intervention carries
a high risk of serious opposition from potential
losers; this creates a major disincentive for gov-
ernment to act. In sum, the rare tenure mix,
combined with the lack of investment and effort
by all three of the main stakeholders that is, the
government, landlords, and tenants, has to-
gether resulted in the appalling conditions in
the slums. So, what should be done?
(b) Using the framework to propose an
alternative solution
In contrast to the comprehensive thinking
that characterized early slum upgrading pro-
jects, the current tendency of practitioners
and academics is to identify a key entry
point—usually, either tenure or infrastruc-
ture—and suggest that action on their selected
factor alone is sufficient to generate wider ben-
efits, and improvements (De Soto, 2000; World
Bank, 2002). Grappling with the Nairobi case
revealed, however, that approaches which favor
one entry point over others reflect a partial
understanding of slums, and, thereby, offer
incomplete solutions: these solutions might
work in certain conditions but will fail in other
contexts.
In Nairobi’s slums, neither the titling-based
approach recommended by De Soto (2000),
nor supports for rental housing advocated by
Gilbert (2003), nor the infrastructure upgrading
approach championed by the World Bank is
likely to work independently. Titling-based ten-
ure legalization is likely to confer benefits on
structure owners and not tenants, unless explic-
itly treated as a land reform exercise. Support-
ing structure owners, as providers of rental
housing, will not work without altering the
structure of the market and eliminating other
disincentives. Investments in infrastructure are
likely to translate into higher rents—as shown
by the rent analysis—and, thereby, displace
many of the current slum residents, not least
because the majority of them fall below the
poverty line.
The government’s goal should be to break
the high-cost low-quality trap across Nairobi’s
slums. Unlike previous small-scale or pilot pro-
jects, this effort needs to be systematic and city-
wide. And the government needs to act on two
facets of the diamond—tenure and infrastruc-
ture—simultaneously. By enhancing owner
occupancy in the tenure mix and simulta-
neously upgrading infrastructure, the govern-
ment can create a basis for private
investments in units and facilitate neighbor-
hood-level improvements. In fact, the new
infrastructure could be explicitly designed to
create additional neighborhood-level bene-
fits—for example, while paving paths and
building drains, the settlement layout can be al-
tered to facilitate access by fire trucks. In other
words, the dual-entry approach, if well de-
signed, can catalyze improvements on all four
fronts—it can set in motion a longer-term pro-
cess of consolidation of the housing stock and
neighborhoods. Using these principles, several
design options are possible.
The government could, for instance, offer a
slum development package that combines
 SLUM REAL ESTATE
infrastructure investment with legitimization of
tenure. Residents would select infrastructure
investments based on their priorities and neigh-
borhood conditions, subject to a budget con-
straint. Government would legally recognize
tenure and transfer title to individuals and/or
the community (as in the Voi case). This offer
would be open to all slums, but would initially
be implemented only in those settlements where
tenants and landlords reach agreement on a
formula for sharing the costs and benefits of
these changes—including the terms under
which landlords agree to divest some of their
units to current tenants.
An example of such a ‘‘deal’’ is the one bro-
kered by Pamoja Trust, an NGO, in Nairobi’s
Huruma settlement. This agreement—between
landlords, tenants, and Nairobi City Coun-
cil—required landlords to give up control over
some of their units in exchange for a formal ti-
tle and investment in improving the retained
units (Personal interview with Jane Weru,
Director, Pamoja Trust, September 20, 2006;
also see Baliga, 2005). For instance, a landlord
with 12 units agreed to retain three and give up
the remaining nine, which were then legally
transferred to the tenants. Subsequently, the
NGO also facilitated construction of new/im-
proved units for the 70 participating house-
holds, using a combination of grant funds and
residents’ savings. In our view, a key advantage
of this solution is that it results in a better mix
of owner-occupiers, tenants, and landlords on
the site; it is a small-scale land reform project.
Overall, we now know that the value of main-
taining status quo is a flow of at least US$31
million per year to landlords and those they
need to unofficially pay—none of this translates
either into improved living conditions for ten-
ants, or into official revenues or taxes for gov-
ernment. The government, as the official land
owner in majority of the city’s slums (Syagga
et al., 2002), can convene a meeting with the
two other stakeholders (the structure owners
and the tenants) to decide whether and how this
pie can be carved up in a fairer manner. Theo-
retically, if it were rich and generous, the gov-
ernment could also offer to just buy out the
landlords with an appropriate compensation
package. Yet another alternative would be for
the government to try and mandate a solution,
but the complex political economy may not al-
low for such an approach to work. On our part,
we favor a negotiation between the three stake-
holders, with the government using a combina-
tion of carrots and sticks to reach a solution.
1933
6. CONCLUSIONS, AND IMPLICATIONS
FOR THEORY, AND PRACTICE
In summary, this study finds that the vast
majority (92%) of Nairobi’s slum residents do
not own their homes. They are tenants, and
they are crammed—2.6 persons per room—in
poor-quality housing structures built with
impermanent materials. Their access to basic
infrastructure is appalling. For instance, only
22% of slum households have an electricity con-
nection and barely 19% have access to a supply
of piped water, in the form of either an in-
house water connection or a yard tap.
Rents in the slums are significant, despite the
informal tenure status, the poor quality of
housing, and inadequate infrastructure. In
2004, slum households paid ‘‘structure owners’’
an average rent of about US$11 per month, a
sum that amounts to an extraordinary US$31
million for the year. Unfortunately, this large
cash flow does not translate into investments
that improve the quality of living conditions
for the residents, nor does any fraction of it ac-
crue formally to the official land owner, that is,
the government. This informal rental market,
albeit distorted, appears to follow many of the
rules that apply in ‘‘formal’’ real estate mar-
kets—rents vary with house size, quality, infra-
structure access, and location.
Although previous studies of Nairobi’s slums
have documented some of the above features,
researchers and practitioners have struggled
with their interpretation of the findings. All
seem to agree that the situation in Nairobi’s
slums is bad, but there are major disagree-
ments, for example, on whether the rents are
too high and need to be regulated, whether
the landlords are villains or not, and whether
this rental housing market is helping or hurting
the poor. Not surprisingly, then, there is no
agreement on what to do about the ‘‘terrible
living conditions’’ in these slums. From an aca-
demic perspective, many researchers have
acknowledged that Nairobi does not fit well
with their broader conceptual framework and
argument, but have left the case hanging as
an outlier or unresolved puzzle.
We argue that Nairobi’s slums should not be
dismissed as an outlier or unique case. First, we
speculate that Nairobi is probably not entirely
unique—there are possibly many other cities
in which sub-standard slum accommodation
comes at a high-price, and yet others that have
slums with a high proportion of renters (Mum-
bai may be an example of both). Identification
1934 WORLD DEVELOPMENT
and analyses of such cases and rental markets
are important areas for future research. Sec-
ond, the Nairobi case and other alleged outliers
should not be dismissed precisely because they
bring into question the very core of our current
understanding of slums. As difficult cases, they
force us to revise or expand the conceptual
frameworks that we use to analyze and under-
stand slums. We offer the living conditions dia-
mond as an analytical framework, and a step
toward a more complete theory of the dynamics
of living conditions in settlements. This frame-
work allows us to understand, for example,
why Nairobi is a difficult but not counter-intu-
itive case, and how its ‘‘slums of despair’’ might
be converted into ‘‘slums of hope.’’
What are the implications of this study for
improving living conditions in Nairobi’s slums?
NOTES
1. For an excellent large-sample study of population
and health in Nairobi’s slums, see APHRC (2002). For a
profile of Nairobi, see UN-HABITAT (2006).
2. Results are reported in a series of papers, including
World Bank (2002, 2004, 2006), Gulyani and Talukdar
(2007), and Gulyani and Bassett (2007).
3. Some studies estimate that slum residents account
for 55% of Nairobi’s population (e.g., Matrix Consul-
tants/USAID 1993 cited in Alder, 1995). Possible
explanations for the divergence include (i) use of
different boundaries for Nairobi—some focus on the
metropolitan area and include slums abutting the city’s
boundaries, while CBS excludes slums outside the
administrative boundary, and (ii) CBS may have under-
estimated the number of EAs that are slums—it may
have mis-categorized some of the EAs either accidentally
or by using too narrow a definition (also see Section 3).
Additional research is required to resolve this issue.
4. During 1989–99, Nairobi’s annual population
growth rate was 4.85%; we use this rate to estimate that
the slum population grew to 810,000 by 2004.
5. Adult equivalents are calculated as follows: children
of age 0–4 are 0.24 adult equivalents, children of age 5–
14 are 0.65, and individuals of age 15 or more are 1.0 (or
‘‘adult’’) (see World Bank, 2006 for details).
6. For information on Government of Kenya’s poverty
line, see World Bank (2003). We used the Consumer
Price Index to adjust the 1997 poverty line for inflation.
In Nairobi, the standard prescription of in situ
upgrading of infrastructure will not work.
Usual approaches to legalizing tenure or sup-
porting rental markets are likely to benefit
structure owners rather than tenants. Using
the living conditions diamond, we argue that
instead of choosing either infrastructure or ten-
ure as a single entry point, the government
needs to act on both simultaneously. The gov-
ernment can deploy infrastructure investment-
cum-tenure legitimization as incentives, and
use them explicitly to enhance owner occu-
pancy in slums. The government’s goal should
be to break the low-quality high-cost trap in
slum housing, and one way to get there is to
bring both structure owners and tenants to
the table, and negotiate a deal.
7. For a more detailed theoretical discussion of the
framework, see Gulyani and Bassett (2008).
8. Although it acknowledges that slums are diverse,
UN-HABITAT (2003) argued for a quantitative, oper-
ational, and common definition of slums. It defines
‘‘slum as an area that combines, to various extents, . . .:
(i) inadequate access to safe water; (ii) inadequate access
to sanitation and other infrastructure; (iii) poor struc-
tural quality of housing; (iv) overcrowding; and (v)
insecure residential status;’’ it also proposes thresholds
for each (Table 1.2, p. 12).
9. The wider land market in the city is not working
well. For discussions on the complexity and constraints
in the land market, see, for example, Government of
Kenya (2007), Southall (2005), and Klopp (2000).
10. Our definition of neighborhood ‘‘conditions’’ is
broader than safety; it also includes amenities (such as
parks, open spaces, community halls, schools), housing
density, and settlement layout and access (e.g., whether
fire engines, and ambulances can enter). Further, ‘‘loca-
tion’’ refers not only to the relative access that a
neighborhood provides to jobs, schools, and markets,
but also the environmental acceptability or precarious-
ness of its physical site.
11. During our research (2003–07), we have benefited
from discussions with numerous Kenyan experts, includ-
ing James Mutero (consultant), Peter N’gau (Nairobi
University), Paul Syagga (Kenya Planning Association),
Jane Weru (Pamoja Trust), Julius Malombe (Ministry of
Housing), and Mukami Kariuki (World Bank).
 SLUM REAL ESTATE
12. Gilbert (2003, p. 239) himself admitted that Nai-
robi is a challenge: ‘‘Renting in Nairobi currently
represents a worst case scenario, in terms of living
conditions offered to tenants; it is a classic example of
what rental housing should not be like. It should be
noted, however, that these rental units provide affordable
housing to a considerable share of the labor force’’
[emphasis added].
13. Amis (1984) found that rents accounted for 14% of
income, constituted a major burden for most slum
residents, and were squeezing other basic expenditures,
including food. He denounced the assumption that an
expenditure of 20% of income is affordable, noting that
‘‘to suggest that such individuals are not paying
enoughon housing, as are the international agencies, is
little short of criminal.’’
14. According to Huchzermeyer (2006) ‘‘Within the
current housing market in Kenyan cities, an affordable
housing alternative to wattle and daub slum rooming
does not exist. . . . The next best accommodation . . .
REFERENCES
Alder, G. (1995). Tackling poverty in Nairobi’s informal
settlements: Developing an institutional strategy.
Environment and Urbanization, 7(2), 85–107.
Amis, P. (1984). Squatters or tenants: The commercial-
ization of unauthorized housing in Nairobi. World
Development, 12(1), 87–96.
APHRC (African Population and Health Research
Center), (2002). Population and health dynamics is
Nairobi’s informal settlements: Report of the Nairobi
cross-sectional slum survey 2000. Nairobi: APHRC.
Baliga, N. (2005). Community-based upgrading and
tenure regularization: Insights from Nairobi, Kenya.
Unpublished masters thesis. Urban Planning,
Columbia University.
Bassett, E. M. (2007). The persistence of the commons:
Upgrading and tenure decision-making in Voi,
Kenya. African Studies Quarterly, 9(3).
CBS (Central Bureau of Statistics). (2003). Geographic
dimensions of well-being in Kenya: Where are the
poor? CBS, Ministry of Planning and National
Development. Nairobi: Regal Press Kenya.
‘‘Census 1999.’’ (2002). The 1999 population and housing
census: Popular report. Nairobi: CBS.
Collignon, B., & Vézina, M. (2000). Independent water
and sanitation providers in African cities: Full report
of a ten-country study. Washington DC: The World
Bank.
Daily Nation (The). 25 October, 2006. Preparation of
LASDAP: Local authority revenue and resource
envelope. Public notice issued by Town Clerk, City
Council of Nairobi.
Datta, K. (1995). Strategies for urban survival? Women
landlords in Gaborone, Botswana. Habitat Interna-
tional, 19(1), 1–12.
1935
though still unauthorized, is in extremely dense multi-
storey rooming districts. However, rents for single room
. . . with shared toilets . . . are at least three times those
for slums.’’ Huchzermeyer reports rents for wattle, and
daub units at Ksh 400–500, and for a single room in a
Huruma multi-storey at Ksh 1,800–2,000.
15. Nairobi city’s budget for 2006–07 was estimated at
Ksh 5.44 billion, with Ksh 1.83 billion for investment,
operations, and maintenance, and the rest for salaries
and allowances (The Daily Nation, October 25, 2006).
16. Data drawn from a parallel survey of 1960 house-
holds in Dakar’s slums (see World Bank, 2004).
17. Among slum residents with electricity, just over
half (53.8%) pay their landlord for it, and not the utility
(World Bank, 2006). Landlords also serve as intermedi-
aries between the water utility and users; they collect
payments from users and decide what price to charge
(Water & Sanitation Program, 2007).
De Soto, H. (2000). The mystery of capital: Why
capitalism triumphs in the West and fails everywhere
else. New York: Basic Books.
Gilbert, A. (2003). Rental housing: An essential option for
the urban poor in developing countries. Nairobi: UN-
HABITAT.
Gilbert, A., Mabin, A., McCarthy, M., & Watson, V.
(1997). Low-income rental housing: Are South Afri-
can cities different? Environment and Urbanization,
9(1), 133–148.
Government of Kenya. (2003). Economic recovery strat-
egy for wealth and employment creation. Ministry of
Planning and National Development.
Government of Kenya. (2007). National land policy.
Kenya Ministry of Lands. May.
Gulyani, S., & Bassett, E. (2008). The living conditions
diamond: A framework for assessing quality. Working
Paper. AFTU1. Washington, DC: The World Bank.
Gulyani, S., & Talukdar, D. (2007). Inside informality:
Poverty, micro-enterprises and living conditions in
Nairobi’s slums. Working Paper. AFTU1. Washing-
ton, DC: The World Bank.
Gulyani, S., & Bassett, E. (2007). Retrieving the baby
from the bathwater: Slum upgrading in Sub-Saharan
Africa. Environment and Planning C: Government and
Policy, 25, 486–515.
Gulyani, S., Talukdar, D., & Kariuki, R. M. (2005).
Universal (non)service? Water markets, household
demand and the poor in urban Kenya. Urban
Studies, 42(8), July.
Hansen, K. (1997). Keeping house in Lusaka. New York:
Columbia University Press.
Huchzermeyer, M. (2006). Slum upgrading initiatives in
Kenya within the basic services and wider housing
1936 WORLD DEVELOPMENT
market: A housing rights concern. In Paper presented
at ISA conference, Durban, South Africa, July.
Klopp, J. (2000). Pilfering the public: The problem of
land grabbing in contemporary Kenya. Africa Today,
47(1), 7–26.
Kumar, S. (1996). Landlordism in third world urban
low-income settlements: A case for further research.
Urban Studies, 33(4–5), 753–782, May.
Miraftab, F. (1997). Revisiting informal-sector home
ownership: The relevance of household composition
for housing options of the poor. International Journal
of Urban and Regional Research, 21(2), June.
O’Connor, A. (1983). The African city. London: Hutch-
inson.
Razzaz, O. (1993). Examining property rights and
investment in informal settlements: The case of
Jordan. Land Economics, 69, 341–355.
Salon, D., & Gulyani, S. (2008). Travel ‘‘choices’’ of
slum residents in Nairobi, Kenya. In Paper presented
at transport research board meetings, Washington DC,
January.
Southall, R. (2005). The Ndungu report: Land and graft
in Kenya. Review of African Political Economy,
32(103), 142–151, March.
Syagga, P., Mitullah, W., & Karirah-Gitau, S. (2002). A
rapid economic appraisal of rents in slums and
informal settlements. Prepared for Government of
Kenya and UN-HABITAT Collaborative Nairobi
Slum Upgrading Initiative.
Thompson, J., Porras, I. T., Wood, E., Tumwine, J. K.,
Mujwahuzi, M. R., Katui-Katua, M., et al. (2000).
Waiting at the tap: Changes in urban water use in
East Africa over three decades. Environment &
Urbanization, 12(2), 37–52, October.
Figure 5. Nairobi City depicting the sampled slum neighborhoods. Source: Salon and Gulyani (2008).
Turner, J. (1972). Housing as a verb. In J. F. C. Turner,
& R. Fichter (Eds.), Freedom to build (pp. 148–175).
New York: Macmillan.
UN-HABITAT. (2003). The challenge of slums, Global
report on human settlements 2003. Nairobi: WSP.
UN-HABITAT. (2006). Profile of Nairobi, Kenya, rapid
urban sector profiling for sustainability DVD. Nai-
robi: UN-HABITAT.
Water and Sanitation Program. (2007). Citizen’s report
card on urban water, sanitation and solid waste
services in Kenya: Summary of results from Nairobi.
Nairobi: WSP.
World Bank. (2002). Slum upgrading in Africa: Ten
country assessment reports. Africa Upgrading Team
(Gulyani, S., Farvacque-Vitkovic, C., Debomy, S.,
Connors, G., Casalis, A., Carroll, A., Banes, C.).
Working Papers. Africa Urban and Water. Wash-
ington, DC: The World Bank.
World Bank. (2003). Kenya: A policy agenda to restore
growth. Report No. 25840-KE. Washington DC: The
World Bank.
World Bank, (2004). Upgrading of low-income settle-
ments in Africa. Consultants’ report, TF No. 024943,
Africa Urban and Water. Washington, DC: The
World Bank.
World Bank. (2006). Kenya: Inside informality: Poverty,
jobs, housing and services in Nairobi’s slums. Report
No. 36347-KE. Washington DC: The World Bank.
APPENDIX
See Figure 5 and Table 6.
 Table 6. Demographics, household size, and composition
All Poor Non-poor
N Percent of total (%) N Percent of total (%) N Percent of total (%)
Households 1755 1282 473

SLUM REAL ESTATE

Household size 2.97 3.38 1.88
Single-person households 560 31.9 261 20.4 299 63.2
Female-headed households 310 17.7 244 19.0 66 14.0
Mean age of household head (yrs) 34.8 34.8 34.6
Median age:household head (yrs) 32.0 32.0 32.0
Total number of individuals 5256 100.0 4345 100.0 911 100.0
Gender ratio, male:female 2899:2357 55:45 2320:2025 53:47 579:332 64:36

Available online at www.sciencedirect.com

1937