WPS5388 Policy Research Working Paper 5388 Poverty, Living Conditions, and Infrastructure Access A Comparison of Slums in Dakar, Johannesburg, and Nairobi Sumila Gulyani Debabrata Talukdar Darby Jack The World Bank Africa Region Sustainable Development Division July 2010 Policy Research Working Paper 5388 Abstract In this paper the authors compare indicators of living conditions are not as bad as in Nairobi. These development, infrastructure, and living conditions in the findings suggest that reduction in income poverty slums of Dakar, Nairobi, and Johannesburg using data and improvements in human development do not from 2004 World Bank surveys. Contrary to the notion automatically translate into improved infrastructure that most African cities face similar slum problems, they access or living conditions. find that slums in the three cities differ dramatically from Since not all slum residents are poor, living conditions each other on nearly every indicator examined. also vary within slums depending on poverty status. Particularly striking is the weak correlation of measures Compared to their non-poor neighbors, the poorest of income and human capital with infrastructure access residents of Nairobi or Dakar are less likely to use water and quality of living conditions. For example, residents (although connection rates are similar) or have access of Dakar's slums have low levels of education and high to basic infrastructure (such as electricity or a mobile levels of poverty but fairly decent living conditions. phone). Neighborhood location is also a powerful By contrast, most of Nairobi's slum residents have jobs explanatory variable for electricity and water connections, and comparatively high levels of education, but living even after controlling for household characteristics and conditions are but extremely bad. And in Johannesburg, poverty. Finally, tenants are less likely than homeowners education and unemployment levels are high, but to have water and electricity connections. This paper--a product of the Sustainable Development Division, Africa Region--is part of a larger effort in the department to improve the global knowledge base on African infrastructure as part of the Africa Infrastructure Country Diagnostic. The paper has benefited from funding, at different stages, from the Norwegian Trust Fund, Public Private Infrastructure Advisory Facility (PPIAF), Water and Sanitation Program-Africa (WSP-Africa), the Africa Region of the World Bank, and the multi-donor supported Africa Infrastructure Country Diagnostic (AICD). Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at sgulyani@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Poverty, Living Conditions, and Infrastructure Access A Comparison of Slums in Dakar, Johannesburg, and Nairobi Sumila Gulyani, Debabrata Talukdar, and Darby Jack Acknowledgments The paper has benefited from funding, at different stages, from the Norwegian Trust Fund, Public Private Infrastructure Advisory Facility (PPIAF), Water and Sanitation Program-Africa (WSP-Africa), the Africa Region of the World Bank, and the multi-donor supported Africa Infrastructure Country Diagnostic (AICD). The authors would like to thank Vivien Foster for valuable comments and advice on this study, Candy Chang of Red Antenna for excellent graphics, Steven Kennedy and his team for a thorough edit, and Aijaz Ahmad for helping with administration in many ways. Contents 1 Introduction 1 Data 2 2 Comparing cities 4 The development diamond: a multidimensional view of poverty in the slums 4 Assessing quality: the living conditions diamond 7 The infrastructure polygon 10 3 Inequality within slums: comparing poor and nonpoor households 15 The development diamond 15 The living conditions diamond: dwelling quality and tenure 16 The infrastructure polygon: The poor are worse off 17 4 Spatial heterogeneity in access to services 20 Electricity connections: Poverty, tenancy, and location matter 21 Piped water: tenancy and location matter, but poverty does not 22 5 Conclusions 23 References 24 Tablular appendix 25 Appendix A Spatial heterogeneity in service access in Johannesburg 48 Electricity connections 48 Piped water connections 50 Appendix B Supply-side vs. demand-side determinants of infrastructure access 52 The heuristic computational approach 52 Application of the approach to our data for the three cities 53 iii 1 Introduction Today more than half the world's population lives in urban areas. The transition from a rural world to an urban one has been led by the rapid, dramatic, and difficult urbanization in the developing world. The shift has been difficult, not least because cities in the developing world have neither been able to plan for nor to keep pace with this transformation. Consequently, more and more urban residents live in unplanned, squalid settlements that lack access even to basic services such as piped water, sanitation, drainage, and electricity. The situation is deemed so alarming that one analyst describes Earth as a "planet of slums" (Davis 2006). Despite the magnitude of urbanization, and despite widespread agreement on its significance, however, we lack answers to fundamental questions about the lives of urban residents--especially slum residents--in the developing world. This gap is particularly pronounced in Africa, which is both the most rapidly urbanizing and the poorest region of the world (UNFPA 2007). In this study, we empirically examine development and living conditions in slums. What levels of infrastructure services do slum residents tolerate? How do these levels vary, within and across cities? What explains these variations? Drawing on random samples of slum residents in Dakar, Johannesburg, and Nairobi, this paper explores these questions and offers some novel, provocative, and policy-relevant empirical evidence. Our main findings are as follows: The cities in our sample exhibit heterogeneity--slums in the three cities differ dramatically from each other on nearly every development indicator that we examined. This finding belies the notion that most African cities face a more or less similar slum problem. By extension, it also challenges the idea that one approach to--or template for--the upgrading of slums can work in all African cities. In comparing the three cities, it is striking that measures of income and human capital correlate poorly with infrastructure access and quality of living conditions. In Dakar, slum residents have fairly decent living conditions, even though they have low levels of educational attainment and high levels of income poverty. By contrast, in Nairobi slum's living conditions are appalling although most slum residents have jobs and a comparatively high level of education. Although far from conclusive, this evidence suggests that reduction in income poverty and improvements in human development do not automatically translate into improved infrastructure access and quality of living conditions for the urban poor. Future research will want to focus on the factors that explain the variation--and facilitate improvements--in quality of living conditions. Within each city, and even within a given slum, we again find heterogeneity. First, we were encouraged to find that not all slum households are poor. Second, slum residents have heterogeneous living conditions, depending on their welfare/poverty status and on the neighborhood in which they reside. In comparing poor and nonpoor slum residents within a given city, we find the poor often lagging behind their nonpoor neighbors in access to some basic infrastructure and they endure worse living conditions. For instance, in both Nairobi and Dakar, the poor are systematically less likely to have POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS either an electricity connection or a mobile phone. Water-connection rates are similar for the two groups, but water-use is lower among poor households. We were encouraged to find no statistically significant difference in their educational attainment levels. Spatially, we find considerable variation across slum neighborhoods within a city. In Nairobi and Dakar, and for both electricity and water connections, neighborhood location is a powerful explanatory variable, even after controlling for household characteristics and poverty. This suggests that supply-side factors play an important role in determining infrastructure access. Finally, tenancy is a key variable in this issue of access to certain services--tenants are systematically less likely to have water and electricity connections as compared with homeowners. A tenant might be less willing or able to pay. The utilities might be less willing (or able) to connect. It might be a combination of the two. The current paper also contributes to a methodological framework for a comparative analysis of slum residents' experiences. We show how radar graphs of thematically grouped variables--the development diamond, the infrastructure polygon, and the living conditions diamond--provide a useful way to aggregate data and to examine differences within and across cities. To our knowledge, this is the first comparative study of African cities that draws on multisectoral random samples of slum residents (indeed, the three data sets that we analyze appear to be the only high-quality multisectoral household survey data available for urban slums in Africa). The paper proceeds as follows. Section 2 presents cross-city comparisons that are organized around the development diamond, the infrastructure polygon, and the living conditions diamond. Section 3 uses this framework to compare poor and nonpoor households. The fourth section examines spatial heterogeneity and what it implies about supply and demand of key infrastructure, while Section 5 offers conclusions. Data In February­March 2004, a World Bank-led research team conducted household surveys in the slums of Nairobi and Dakar. The questionnaires in the two cities were almost identical to allow for an apples-to- apples comparative analysis and to establish a base for comparative studies including other cities. Although the sampling methodology differed somewhat for the two cities, both sought to produce a population-weighted stratified random sample of slums. The study covered 1,755 households in Nairobi and 1,960 households in Dakar. In Nairobi the 1,755 households were selected from 88 Enumeration Areas (EAs). For census purposes, Kenya's Central Statistics Bureau (CBS) has divided Nairobi into about 4,700 EAs, of which 1,263 are categorized as EA5, or slums, which are characterized by substandard housing and poor infrastructure. The 88 EAs in our sample were selected randomly from the subset of 1,263 EA5s and weighted by population. Because the household lists were not up-to-date, a complete relisting was conducted in each EA selected for study, and the sample households were chosen randomly from the revised lists. CBS collaborated with the World Bank's team in designing the sampling frame of this study and also carried out the field-based relisting of households in the 88 EAs. Overall, the Nairobi data set is a 2 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS population-weighted stratified random sample, and it is representative of the 1,263 EAs that CBS categorized as slums. The Dakar survey covered 1,960 households selected randomly from a random sample of 99 EAs, from a universe of 2,074 EAs in the city, stratified by level of past slum-upgrading efforts. The household sample was selected from a 2002 listing of households, and stratification was carried out based on both official and unofficial information. In Johannesburg, a different World Bank research team led surveys of 5,100 households over six weeks in August and September 2001. The survey formed part of a larger effort to monitor service delivery to the poor and focused on what the study team termed "poorly served" areas. Using data from a 1996 citywide sample, Bank staff identified 921 EAs that qualified as poorly served.1 These EAs comprise 125,409 households, or about 24 percent of the city's population. The 5,100 surveyed households are representative of this underserved quartile and are drawn from 253 of the 921 poorly served EAs. For purposes of comparison with the slums of Nairobi and Dakar, however, we analyze only a subset comprising 1,618 households living in informal housing, mostly shacks. 1 Cut-off points for inclusion in the set of poorly served areas is as follows: 5 percent of the households with less than basic water and sanitation, 10 percent for electricity, and 20 percent for housing quality. These cut-off points were selected to give the maximum feasible population of EAs to be surveyed. 3 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS 2 Comparing cities This section focuses on variations across the three cities on key development indicators. We start by briefly summarizing the differences in demographic profiles across the slum residents in the three cities. The mean household size (9.6) in Dakar is strikingly larger than in Nairobi (3.0) and Johannesburg (3.7) (appendix table 1). A relatively high percentage of single-person households exist in the slums of Nairobi (28 percent) as compared to Dakar (1.7 percent) and Johannesburg (9.2 percent). Although the proportion of males to females is about 50:50 in the slums of Dakar and Johannesburg, in Nairobi it is about 55:45. Household heads in Nairobi's slums were much younger (with a mean age of about 35 years) than those in Dakar (with a mean age of about 52 years); for Johannesburg, corresponding data are not available. The development diamond: a multidimensional view of poverty in the slums We map poverty in the slums of the three cities using a four-dimensional framework that includes monetary welfare, employment, education, and living conditions. These factors are graphically represented as four vertices of the development diamond (figure 2.1), which is discussed in more detail in Gulyani and Talukdar (2007, 2008); we use it here to provide a snapshot of different aspects of poverty in the slums. We select absolute indicators for each of the four aspects and suggest that the (normative) goal should be to achieve a 100 percent score along each axis. Thus, a fully shaded diamond would indicate a city that has eliminated deprivation. For this comparative analysis of slums in the three cities, we plot the proportion of (i) households living above the absolute (expenditure-based) poverty line as defined by the government; (ii) adults employed in jobs or microenterprises; (iii) adults with primary education; and (iv) households with living conditions that meet at least a minimal quality threshold, defined as a unit with permanent walls, water, and electricity.2 We posit that the dimensions of the development diamond interact over time to determine the welfare of a given community or household. For instance, improvements in education can improve access to jobs, and this can, in turn, help lower the incidence of income poverty. With more income, residents can invest in improving living conditions and education levels. Slums in all three cities fall far short of the 100 percent target along each of the four dimensions-- poverty incidence is high, unemployment is a serious problem, education is anything but universal, and living conditions are poor. What is striking, however, is the variation among the cities--slum residents in 2 These indicators establish only a basic or minimum development diamond and are meant to be illustrative. Conceptually, the famework can be modified for a given context and the indicators can be more sophisticated. For example, the indicators can be changed to include both "access" and "quality" aspects, and the thresholds can be set at higher levels. For instance, the measure for education could be set at "proportion completing high school" rather than primary or secondary education and/or modified to record the proportion who can successfully pass certain types of tests. 4 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Nairobi are much better educated; those in Dakar enjoy better living conditions overall, and those in Johannesburg have both the greatest access to regular jobs and the most unemployment. This means that the key problems, and thus high-priority actions, differ rather dramatically by city. Figure 2.1 Development diamonds for three cities As figure 2.1 illustrates, the four factors are weakly correlated in each of the three cities. These asymmetries suggest that reduction in income poverty is in itself insufficient to ensure better living conditions for slum residents, although it may help to meet this goal. Similarly, improved access to education alone is not sufficient to ensure access to jobs. These implications are both commonsensical and well supported by research, yet governments and development practitioners tend to favor sector- specific initiatives. Such initiatives rarely reflect the multiple dimensions of welfare implicit in the development diamond. The development diamond offers a composite picture of the multifaceted and 5 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS interactive nature of the challenge of development in a given context. Over time, this composite picture helps us come to a more nuanced understanding of the relationship among the four dimensions. In the rest of this section we briefly discuss these four dimensions. Monetary welfare Income and expenditure underpin a household's economic life, and yet they are difficult to measure. Respondents may gain little from revealing their real income, or they may not remember and aggregate their numerous transactions. With these concerns in mind, the surveys in Dakar and Nairobi employed three approaches to estimate monetary welfare:3 Assessment of household expenditures above or below the poverty line. The enumerator calculated a household-specific poverty line based on the age and number of household members, and the respondent answered whether expenditure was above or below that threshold. Total household expenditures adjusted for family size (measured in adult equivalents) Total household income adjusted for family size (also measured in adult equivalents) The first of these measures was judged to be most reliable and is used as the primary measure of poverty throughout this report. In Nairobi, the analysis took the 1999 poverty line as defined by the Government of Kenya and adjusted it for inflation to calculate the poverty threshold for 2004. Using this expenditure-based poverty line--defined as an expenditure of Ksh 3,174 (US$42) per adult equivalent per month, excluding rent--about 72 percent of the slum households are poor and 28 percent are nonpoor. In Dakar, the 2004 official poverty line was 27,705 F CFA (US$53) per adult equivalent, including rent. The discrete wealth indicator suggests that about 18 percent of the individuals have expenditures above the poverty line, while about 82 percent fall below the poverty line. For Johannesburg, the survey did not have information on income; so Figure 1 uses an estimated value of about 50 percent of households under poverty line. Education Educational attainment varies markedly across the three cities (appendix tables 2 and 3). In Dakar, about a third of adults have a primary-level education, and fewer than 10 percent completed high school. Corresponding figures are much higher in Nairobi (79 percent have finished primary school; 31 percent high school) and Johannesburg (70 percent have completed primary; 44 percent high school). It is worth noting the contrast between Dakar, which has low educational attainment and fairly good infrastructure, and Nairobi and Johannesburg, with their relatively high levels of educational attainment and poor infrastructure. This suggests that while human capital can certainly affect access to jobs and poverty levels, it appears to be neither necessary nor sufficient for slum households to attain decent living conditions or access to basic infrastructure services such as water and electricity. 3 Factor analysis based on household assets was also carried out, but slum households exhibited too little variation in assets for meaningful results. 6 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Employment Unemployment is widespread among Johannesburg's slum residents, where every other adult is unemployed. That compares to an unemployment level of 26 percent in Nairobi and only 6 percent in Dakar. Within households the story is much the same: a typical household in Johannesburg slums will have one-third of adult members unemployed, compared to one-fifth in Nairobi and only one-twentieth in Dakar (appendix tables 4 and 5). At the same time, Johannesburg also has the highest proportion of regularly employed adults (28 percent) compared with 25 percent in Nairobi and only 8 percent in Dakar. Self-employment or work in household-owned microenterprises is uncommon in Johannesburg but prevalent in Nairobi and Dakar, where about one in five report that they work in their own household microenterprise. Nairobi's slum residents report the highest level (24 percent) of casual employment, followed by Johannesburg (10 percent) and Dakar (8 percent). Overall, among the three cities, slum residents in Johannesburg are the most likely to be either unemployed or regularly employed and the least likely to be working in a household microenterprise. Living conditions We developed two proxy indicators to arrive at an overview of living conditions and infrastructure access across slum areas in the three cities--both indicators are composites. The first indicator establishes the percentage of households with access to piped water and electricity, and the second determines the percentage of those with piped water, electricity, and permanent external walls. On either indicator, Dakar leads by a sizable margin--76 percent of Dakar's slum households have both piped water and electricity compared with about 31 percent in Johannesburg and only about 7 percent in Nairobi. Another way to compare living conditions among the three cities is to note that only about 3 percent of Dakar's slum households have neither piped water nor electricity, but the corresponding number jumps to about 44 percent for Johannesburg and to 66 percent for Nairobi. In terms of the second indicator, about 74 percent of Dakar's slum households have access to piped water and electricity and reside in houses constructed with permanent external walls. This is true of only 3 percent of the slum households in Nairobi. (Housing quality indicators are not available for Johannesburg.) Less than 1 percent of Dakar slum households have none of three--piped water, electricity, permanent external walls. For Nairobi, by way of contrast, the corresponding number jumps to about 63 percent. Clearly, while the study areas in all three cities can be called slums, those in Dakar offer greater access to critical infrastructure than those in either Johannesburg or Nairobi. The conditions are particularly bad in Nairobi. In the next section, we move from coarse indicators of basic or decent living conditions to a more comprehensive framework for assessing and understanding quality of living conditions. Assessing quality: the living conditions diamond Although slums typically evoke images of squalor, some slums offer decent living conditions. Some have good access to water; others offer better-quality housing units, and some are physically safe. To 7 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS document and compare living conditions across cities, we use a new tool--termed "the living conditions diamond"--developed in Gulyani and Talukdar (2008). According to this framework (represented graphically as a diamond in figure 2.2), living conditions vary in four ways: tenure, infrastructure, unit quality, and neighborhood condition and location. These four factors interact with and influence each other. Figure 2.2 Living conditions diamonds for three cities In the rest of this section, we use the living conditions diamond to determine the quality of slum settlements in the three cities. Given data constraints regarding neighborhood conditions, we focus our discussion on the other three factors. The discussion below shows that each of the three dimensions is in 8 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS itself a composite of several factors and requires use of multiple indicators. Our selection of indicators is illustrative rather than prescriptive. In graphing the living conditions diamond, we have simplified even further. For simplicity and ease of illustration, we have opted to plot only one proxy indicator or an index for each of the four dimensions. We use the following indicators in figure 2.2: the percentage of households with permanent walls (which is a proxy for unit quality); percentage who own their home (a proxy for tenure); percentage who feel safe in their neighborhood (a proxy for neighborhood conditions); and a composite infrastructure index (average access or connection rate across eight different infrastructure services). Condition of housing units: space and quality The mean household size in Dakar (9.6) is about three times larger than those in Nairobi (3.0) and Johannesburg (3.7). At the same time, the mean number of rooms per household is also significantly higher in Dakar (4.1) than in Nairobi (1.2) and Johannesburg (1.8). Consequently, crowding levels--the average number of persons per room--are comparable across the three slums: 2.8 in Dakar, 2.6 in Nairobi, and 2.3 in Johannesburg (appendix table 6). The condition of the housing stock is much better in Dakar than in Nairobi. For example, about 96 percent of the houses in Dakar's slums--but only 12 percent of those in Nairobi--have external walls constructed with permanent materials (brick/stone/concrete blocks).4 Similarly, about 32 percent of households in Nairobi's slums have dirt floors; the corresponding number in Dakar is only 10 percent. Tenure mix and turnover rate The tenancy mix--or, the ratio of owner-occupiers to tenants--is highly skewed in Nairobi, where only 8 percent of households own their own houses. In Dakar, about 74 percent are owner-occupiers and 26 percent tenants. Johannesburg has the highest owner-occupancy rate--89 percent; the remaining 11 percent do not own their house, but only 4 percent report that they pay rent for their accommodations. The turnover rate among slum residents is high in Nairobi and low in Dakar. Median stays in the current house and slum settlement are three years and six years (average of five and nine years) respectively for Nairobi households, compared to 17 and 20 years (average of 19 and 21 years) respectively for Dakar households. Although the mean and median length of stay cannot be computed for Johannesburg's slum residents, 23 percent of the residents report they have stayed in their current home for more than 10 years, while 33 percent have stayed for fewer than 5 years. Neighborhood safety and facilities Assessing the condition of a neighborhood and its location requires the use of multiple indicators. Here we use three proxy indicators: whether slum residents feel safe in their neighborhood, and whether or not their neighborhood has health and education facilities, that is, clinics and schools. (We have these data for Nairobi and Dakar, but not for Johannesburg.) 4 Data on physical condition of housing are not available for Johannesburg. 9 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS The majority of slum residents in both cities report that they feel unsafe in their own settlement. Specifically, when asked if they feel safe in their settlement, 48 percent of Dakar's slum residents and 37 percent of those in Nairobi answered in the affirmative. With respect to health infrastructure, as many as 89 percent of slum households in both Nairobi and Dakar report they have health-service facilities in their settlements. In terms of education infrastructure, we find that schools in general, and private schools in particular, are much more widely prevalent in the slums of Nairobi than in Dakar. About 88 percent of slum households in Nairobi say that they have public or private school facilities in their settlements compared with about 75 percent in Dakar. Breaking this down by type of school, about 52 percent of slum households in Nairobi say that they have public school facilities in their settlements compared with about 59 percent in Dakar. For private schools, the corresponding values are 85 percent in Nairobi and 58 percent in Dakar. Infrastructure Basic infrastructure services are an essential component of quality of living conditions. We examined eight services in detail: water, electricity, toilets, sewage disposal, drainage, garbage collection, public transport, and telephone services. Each of these services is discussed in the next section. (For several indicators, we restrict our comparison to Nairobi and Dakar because comparable data were unavailable for Johannesburg.) The infrastructure polygon By definition, slums in all three cities are poorly served. There are variations, however, among services within a given city and large differences in access among the three cities. In this section, we look more closely at access to different services, specifically focusing on the following questions: What is the infrastructure status of the slums? Are service indicators better in certain sectors? Which cities are doing better in serving their slum areas and in what types of services? Figure 2.3 clearly shows that none of the three cities offers anything approaching universal coverage to its slum residents. In relative terms, households in Dakar's slums are systematically and significantly better off than those in either Johannesburg or Nairobi, on almost all key infrastructure services. In terms of water and electricity connection rates, Dakar leads Johannesburg by a factor of two and Nairobi by a factor of four. Regarding households with both piped water and electricity connections, Dakar's slum households still have significant advantages--76 percent compared to 31percent for Johannesburg and only 7 percent for Nairobi. Figure 3 also allows for a comparison across services; it shows, for instance, that in Nairobi about 22 percent have electricity connections but only 10 percent benefit from garbage collection services. 10 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Figure 2.3 Infrastructure polygons for three cities In the discussion below we highlight some of the service-specific findings, presented in appendix tables 7.1­7.5, to provide an insight, first, into the types of informal options the unserved are forced to rely on, and, second, some of the implications of poor access/service. Water connection rates and use Access to private, piped water connections--in-house or a yard tap--is remarkably high in Dakar (84 percent), lower in Johannesburg (38 percent), and quite low (19 percent) in Nairobi. In Dakar, the vast majority of slum residents use private piped connections as their primary source of water. In Johannesburg, communal taps are the most common source, with 42 percent of slum households relying 11 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS on them. In Nairobi's slums, given the low connection rate, water kiosks are the most prevalent primary source--64 percent of slum households buy water by the bucket (or 20-liter Jerri can) from water kiosks. Most Johannesburg slum households are found to use a single source of water for their daily needs, while about 10 percent of Nairobi's slum households and 13 percent of those in Dakar depend on multiple sources. Further, when we analyzed those households using multiple water sources, the percentage with private piped water connections is 0 percent for Nairobi and about 68 percent for Dakar. That 68 percent for Dakar amounts to 9.5 percent of all households with private piped-water connections; presumably these households have to rely on alternate backup water sources either because of uncertainty of the private piped water supply or because they cannot afford to use piped water for activities such as washing and cleaning. Despite the dramatic differences in the primary water sources available in Nairobi and Dakar, the per capita water use is low in both cities in absolute terms--it hovers around 20 liters per capita per day (lcd)--and it is only slightly higher in Dakar as compared to Nairobi. Specifically, median and mean values of per capita water consumption for Nairobi's slum households are 20 lcd and 23 lcd respectively, and the corresponding values for Dakar's slum households are 22 lcd and 28 lcd. The low level of use in Dakar lends further credence to the hypothesis that water prices may be high for slum residents; this is an area for further research. (We lack daily water-use data for Johannesburg.) Electricity connection rates and street lighting At 82 percent, electricity connection rates are high in the Dakar slums; the figures are lower in Johannesburg (39 percent) and lowest in Nairobi (22 percent). Not surprisingly, 77 percent of the slum residents in Nairobi use kerosene/paraffin as an alternate source of home lighting. This figure is 91 percent in Johannesburg; in Dakar 5 percent of slum residents use kerosene/paraffin. In Dakar's slums, 59 percent of the households report street lights in their neighborhoods compared with only about 14 percent of households so reporting in Nairobi. (We do not have comparable data for Johannesburg.) These findings suggest that electricity supply infrastructure--transmission and distribution networks--is more established in Dakar's slums than in Nairobi's. In Section 4 we offer further analysis at the residential neighborhood level to see how household electricity connections vary across the geographic divisions within the slum settlements in each city. It allows us to gain insights into the extent to which electricity access in a city is affected by demand-side (affordability) factors versus supply-side (presence of trunk infrastructure) issues. Telephone: mobile phones rather than landlines Dakar's slum residents also have superior access to telecommunication services, as compared with Nairobi and Johannesburg. About 29 percent of Dakar's slum households have working landline telephones, but this service is not really an option in the other two cities--only 0.2 percent of slum residents in Nairobi and 2.1 percent of those in Johannesburg have landline telephones. The big communications story, however, lies in the expanding use of mobile phones. By a significant margin in the three cities, access to mobile phones outstrips that to landlines. Here again, Dakar's slum residents enjoy better access--50 percent of households in Dakar have a mobile phone compared to about 20 percent in Nairobi and 22 percent in Johannesburg. 12 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Toilets and sewage disposal services Dakar's slum households have better access to toilet facilities, with the mean number of households sharing a toilet facility at about 1.4 compared with 19.1 and 16.2 for Nairobi and Johannesburg respectively. Even after accounting for the mean size of households (which is much higher in Dakar, at 9.6 percent, than in Nairobi and Johannesburg--3.0 and 3.7 respectively), we find that on average about 13 people use one toilet in Dakar compared with 57:1 in Nairobi and 60:1 in Johannesburg. With 60 percent of households relying on public toilets, these are the most commonly used facility in Nairobi's slums. About 6 percent of Nairobi slum households report they have no access to toilet facilities and therefore resort to the bush or a wrap-and-throw option infamously described as "flying toilets." A septic tank or a VIP (ventilated improved pit) latrine is the most widely used form of sewage disposal in Dakar's slums, used by about 80 percent of households. In contrast, the most widely used disposal system in Nairobi's slum is an ordinary pit latrine--64 percent of the households rely on them as their disposal system. Comparable information is not available for Johannesburg. Solid waste/garbage disposal services Access to solid-waste removal services is almost universal in Johannesburg, reasonably high in Dakar, and in Nairobi almost nonexistent. Specifically, about 90 percent of Johannesburg's slum households have some form of an organized garbage-disposal system, compared with 76 percent in Dakar and only 12 percent in Nairobi. Although we do not have data on specific collection systems for Johannesburg, we do for Nairobi and Dakar. Interestingly, while only 12 percent of Nairobi households have an organized collection system, private collection is a central part of that system, accounting for 11 out of those 12 percent. In contrast, for Dakar, 70 out of the total 76 percent of slum households with access to organized collection systems in fact depend on city/municipal collection systems. As for those households without any access to an organized garbage disposal system, the predominant method is "dumping in the neighborhood" in both Nairobi and Dakar. Graywater disposal services As we do not have data for graywater disposal services for Johannesburg, we will focus only on Nairobi and Dakar. The most prevalent method for graywater disposal in Nairobi's slums is "pouring into the drain" used by about 71 percent of households. In contrast, the most prevalent method in Dakar's slums (used by about 62 percent of households) is "pouring on the road or pavement." Different graywater disposal practices is not surprising given that about 58 percent of Nairobi slum households report some kind of drain outside their house; the figure is only about 7 percent of Dakar households. The lack of drainage infrastructure in Dakar seriously compromises the potential gains from improvements in other services. Many neighborhoods, especially those in low-lying areas, flood extensively. This adversely affects both health and living conditions. Construction of drains is not enough. They need to work. Of those households with drains outside their homes, only about 44 percent of slum residents in Nairobi report that such drains work properly most of the time, compared with about 75 percent in Dakar. Although Nairobi's slums certainly have a more extensive drainage network compared with Dakar, most of this network is not maintained well 13 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS enough to be in working condition--the result is that only 25 percent of Nairobi's slum residents have a working drain outside their home (the comparable figure for Dakar is 5 percent). Access roads and internal paths About 98 percent of slum households in Nairobi say that they have internal roads in their settlements compared to about 35 percent in Dakar. As for access roads to the settlements, the corresponding values are 98 percent in Nairobi and 56 percent in Dakar. We do not have corresponding data for Johannesburg. The findings seem to suggest that Nairobi's slums have better internal and access road infrastructure compared with those in Dakar. Because drains are often built along with roads, Nairobi's relatively better road infrastructure in slums may explain why they also have a significant drainage network. Transportation Only Johannesburg seems to offer public transit services used by slum residents. In Nairobi and Dakar, most slum residents walk. For the slum residents in all the three cities, walking to school and work is an important mode of transport. Walking is the primary mode of transport for majority (55 percent) of Nairobi residents; only about 20 percent use motorized public transportation. In Dakar, walking is the most common form of transport: the primary transportation mode is walking for about 40 percent of the residents, while about 16 percent use public transportation. In Johannesburg, vehicles (motorcycles and microbuses) are slightly more common forms of transportation than walking. Interestingly, almost all slum households in Johannesburg have some members who use motorcycles for transportation and 23 percent of slum households have members who use trains, which are not available at all in the other two cities. Overall, use of motorized public transportation is low and this seems to reflect a combination of constraints--financial constraints at the household level and infrastructure or service constraints at the settlement level. Expenditure allocation for rent and basic services Slum residents in Nairobi and Dakar reported their expenses on rent and basic services (appendix table 8). As a percentage of basic household expenditure, both Nairobi and Dakar households are found to have comparable expenditures on rent (about 12 percent and 15 percent respectively) and water (about 2 percent and 3 percent respectively). But the expenditure levels for electricity (2.5 percent) and refuse collection (0.3 percent) for Dakar households are about half of those for Nairobi slum residents--4 percent for electricity and 0.8 percent for refuse collection. 14 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS 3 Inequality within slums: comparing poor and nonpoor households In this section, we examine residents' access to infrastructure and services among households according to their welfare level. We looked for systematic differences in access between households below and above the poverty line. This approach allows us to see if slums are uniformly bad or if some residents face conditions that are worse than those of their neighbors. We use respective national governments' expenditure-based poverty lines, adjusted for household size, to categorize slum households as either poor or nonpoor. Because no data on income and poverty level for households are available in our Johannesburg survey, we restrict our discussion here to Nairobi and Dakar. As described above and in section 1, poverty was assessed for each household (for which size and age were also determined) according to household-specific poverty lines calculated in the field. Respondents were asked if their total monthly expenditures were above or below the computed amount. Those who responded that they spent less than the poverty level were considered poor. The development diamond Poverty rates are high, but not everyone is poor As noted earlier, slum settlements in both Nairobi and Dakar have high household poverty rates--72 percent and 82 percent respectively--based on respective national poverty threshold levels. But, strikingly, not all slum residents are poor. About 1 in 4 or 5 of the slum households in the two cities are above the poverty line. Poor households are larger. In both cities poor slum households are likelier to report having more household members living together than the nonpoor, and they are also less likely to be single-person households (appendix table 9). Although the proportion of female-headed households is similar among poor and nonpoor slum households in both cities, in Dakar the heads of poor households are found to be about five years older (53 versus 48 years) than those of nonpoor households. Education: the poor are less likely to have secondary and postsecondary education With respect to education level among individual adults living in the slum settlements, those from poor households are systematically less likely to have secondary and postsecondary level education in both cities (appendix table 10).5 The same finding holds when we look at the highest level of education within households (appendix table 11). 5 We also tested whether there was any systematic difference between poor and nonpoor households in terms of access to basic public health services, using BCG immunization rates as a proxy indicator. Encouragingly, we find that almost all (about 97-98 percent) slum children of 0-14 years of age are immunized against BCG in both cities, and there is relative little difference in the immunization rates among children from poor and nonpoor households. 15 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Employment: a welfare gap in Nairobi but not in Dakar Regarding the employment status of adults in the two cities, interestingly, Dakar's poor and nonpoor are not very different. In Nairobi's slums, however, poor adults are less likely to be employed than their nonpoor neighbors. The same finding holds when we look at employment within households. Taken together with education levels, this suggests that while adults in nonpoor slum households are likely to have better educations than those in poor households in both cities, only in Nairobi does this systematically correlate with their chances of being employed (appendix tables 12­13). Decent living conditions: the poor lag behind the nonpoor in Nairobi but not in Dakar As noted in our discussion of cross-city analysis, we created a summary indicator for tabulating access to basic or decent living conditions--the percentage of households that have access to both piped water and electricity and are constructed with permanent external walls. We then examined whether this indicator differs systematically among poor and nonpoor households. In Dakar's slums, there is no statistically significant difference between poor and nonpoor households on this composite proxy indicator. By contrast, in Nairobi's slums, poor households are systematically (statistically significant at 1 percent level) worse off compared with their nonpoor counterparts. In Nairobi (but not in Dakar), living conditions--defined in terms of access to basic utility services and housing--vary systematically with the relative economic or welfare level of a slum household. Whether such variation in utility services is purely due to demand-side/affordability differences or due to neighborhood-level supply bottlenecks is something that we further explore in the next section (Section 4). In the next two subsections, we examine quality of living conditions--and, especially, infrastructure access--in more detail. The living conditions diamond: dwelling quality and tenure Housing units: the poor are more cramped and, in Nairobi, their house quality is worse As earlier noted, while the overall conditions of housing units remain dismal in both cities, we do find systematic differences in the condition of housing units occupied by poor and nonpoor households (appendix tables 14.1­14.2). For example, poor households are likely to live in more crowded conditions with the mean number of persons per room in Nairobi being 3.0 for poor households and 1.6 for nonpoor households; the corresponding figures in Dakar are 2.9 and 2.3 respectively. Regarding materials (permanent versus less permanent) used for the walls, roofs, and floors of the housing units, there is not much of a difference between poor and nonpoor households in Dakar's slums. In Nairobi's slums, however, the nonpoor are more likely to live in housing units built with permanent materials than their poor neighbors are. Tenure: more a city-level problem rather than a poverty issue? Poor and nonpoor households in Nairobi's slums have similar levels of land/structure ownership and tenancy rates. The turnover rate for the two groups does not differ much either--both poor and nonpoor households in Nairobi's slums have lived in their slums, on average, for similar lengths of time. 16 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS In Dakar, at first glance and somewhat erroneously, the poor seem to have better tenure indicators than the nonpoor. For instance, the poor are more likely (73.5 percent versus 59.9 percent) to own land/structure and are less likely (23.7 percent versus 36.4 percent) to be tenants than nonpoor households. Further, the poor have resided in their slum settlement for about five years longer than the nonpoor. These indicators are influenced, however, by the national government's definition of poverty-- because rent is included in the government's expenditure-based poverty threshold, rent-paying tenants are more likely to rise above the poverty line than homeowners are. Additional analyses are required, therefore, to understand the links between poverty and tenure for the case of Dakar. Among those who rent, the poor are more likely to have no agreement (at times suggesting a squatter status) and are less likely to have formal written agreement than their nonpoor counterparts in both cities. Finally, we do not find a significant difference between poor and nonpoor households in either city regarding perceptions of house tenure security. At a broader level, it is important to test whether tenure influences (or is related to) some of the outcomes that we have been examining. We find that two sets of variables do exhibit statistically significant correlations with tenure status (that is, difference between tenant and nontenant households has a p value <0.05). First, data from Nairobi, Dakar, and Johannesburg show that homeowners have larger families (household size) and live in larger homes. Second, tenure affects infrastructure connections. When present, the effect is large in magnitude, but strikingly the sign of the effect varies across cities. In Nairobi and Dakar, owners are more likely to have electricity and water connections; in Johannesburg the opposite is true. The relationship between tenure and the likelihood of having basic infrastructure-- particularly water and electricity connections--is further explored in Section 4. Neighborhood safety and facilities: mo welfare gap We do not find significant differences between poor and nonpoor households in either city regarding their perceptions of neighborhood safety as well as the actual incidence of crimes (personal and property). With respect to school facilities, cross-tabulations suggest that nonpoor households are more likely to have public and private schools in their neighborhood. When controlling for other factors, using logistics regression analyses, the results are different. We find that the likelihood of a household having public or private school in its vicinity depends not on the household's poverty level but on the neighborhood; this is true in both cities. The story regarding availability of health clinics is similar. A regression analysis shows that the likelihood of a household having a nearby health clinic, in either city, depends not on the poverty level of the household but on the neighborhood. The infrastructure polygon: the poor are worse off Now we turn to the fourth dimension of the living conditions diamond--infrastructure. In this section we examine six different infrastructure services and find that, compared to the nonpoor, the poor are systematically worse off (appendix tables 15.1­15.5). Although this makes intuitive sense in general, it is remarkable that these differences manifest themselves even among slum residents, who are already deprived relative to other residents. We also find that, compared with the nonpoor, poor slum households 17 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS in both cities allocate a larger share of their household budget for rent and services such as water, electricity, and refuse collection (appendix table 16). Water: the nonpoor use more and pay higher unit prices The distribution of primary water sources is similar between poor and nonpoor households in both cities. This suggests that nonpoor households have the same water-supply options and constraints as poor households. Nevertheless, nonpoor households consume much more water in both cities (by as much as 30­40 percent). In Nairobi, nonpoor households are also found to have slightly higher per-unit cost for water than the poor households. In Section 4 (below), we conduct further analyses at the residential neighborhood level to gain insights into how water connection rates in a city are affected by demand side (affordability) and supply side (presence of trunk infrastructure) factors. Electricity: connection rates are higher among the nonpoor Better-off households have better access to home electricity connections in both cities. When it comes to neighborhood street lighting, however, we do not see a statistically significant difference between nonpoor and poor households in either city. In Section 4, along with water connections, we also examine electricity connections rates at the neighborhood level to assess the extent to which electricity connection rates are affected by demand- and supply-side factors. Telecom: access to (mobile) phones is higher among the nonpoor The relative economic clout of nonpoor households does translate to better access to cell phone services in both cities. Specifically, about 28 percent of nonpoor households have cell phone service compared to 16 percent of poor households in Nairobi's slums. The corresponding numbers in Dakar's slums are 59 percent and 47 percent. Land-based phone services are practically nonexistent in Nairobi's slums, so the welfare gap cannot be tested. Dakar's slums do have landlines and, compared with the poor, the nonpoor are more likely to have landline phones. Sanitation: the nonpoor are better off As for sanitation infrastructure services, while the overall situation is dismal, nonpoor households appear to have marginally better (statistically not significant for Nairobi but significant for Dakar) services compared to the poor households in both cities. For example, in both cities, nonpoor households are more likely to have individual VIP or flush toilets as well as septic tank/soak pit for toilet sewage disposal. Specifically, about 25 percent of nonpoor households have such types of toilets compared to 22 percent of poor households in Nairobi. For Dakar, the corresponding numbers are 60 percent and 51 percent respectively. The positive differential in favor of the nonpoor households is more pronounced for garbage disposal services, especially in Dakar's slums. In Dakar, about 84 percent of nonpoor households have organized garbage disposal services compared to 75 percent of poor households. In Nairobi, the corresponding numbers are 15 percent and 11 percent, respectively. 18 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Roads: nonpoor report better access roads Nonpoor households in both cities are slightly more likely to report that they have reasonable access roads in their neighborhood. With respect to the availability of internal road facilities within their neighborhoods, the nonpoor, again, have a slight advantage in Nairobi, but the situation is reversed in Dakar. Motorized public transport? As for the primary mode of transport to school and work, in Nairobi's slums, individuals from poor households are more likely to walk than those from nonpoor households. In Dakar, however, there is no welfare gap--individuals from poor or nonpoor households are about equally likely to walk to school and work. 19 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS 4 Spatial heterogeneity in access to services To what extent does infrastructure access vary among and within settlements in a city? What is the relative role played by supply and demand? To what extent does tenure status and its security influence access to services? Answers to such questions can help us better understand why infrastructure access or hook-up rates are low and, thereby, help us design tailored solutions for improving overall coverage. They can provide insights on whether the priority is to invest in additional trunk infrastructure, in making the service more affordable, or in dealing with insecurity of tenure. In this section we take a closer look at differences in access to services across space--spatial heterogeneity--within each city. This serves two purposes. First, it provides an insight into the extent and nature of spatial heterogeneity, or inequality in services, within cities. Second, it provides some clues as to why some households lack services. The reasoning is simple: If none of the households surveyed within a given administrative district use a given service, then that service is probably unavailable. For example, an area that reports no households with piped water likely lacks the requisite trunk infrastructure, while a neighborhood with a small percentage of households that enjoy piped water probably has the trunk infrastructure. In the latter case we might suspect that affordability, not access, was the main barrier to household-level access. The analysis proceeds as follows. We examine two categories of infrastructure for Nairobi and Dakar: home electricity and piped water connections. (We did not have comparable data for Johannesburg.)6 To divide the cities spatially, we relied on divisions and enumeration areas (EAs) used to design the sampling frame. Our analysis thus covers 8 divisions and 88 constituent EAs for Nairobi, and for Dakar 10 divisions and 99 constituent EAs. For each category and for each city, we first tabulate coverage rates by division and by EA for the two selected services. This tabulation shows how the coverage varies across the divisions in each city and within each division across its constituent EAs. Next, we investigate what might explain spatial variations in access. In this context, the variations have two possible sources: supply-side factors and demand-side factors. Supply-side factors refer to differences in access arising from the relative extent of supply trunk lines available across and within the divisions in which they reside. Demand-side factors refer to differences arising from factors such as a household's welfare status (poverty level) and tenancy status (which can in turn affect a household's ability or willingness to pay for infrastructure). By running logistic regressions at the household level, we investigate what roles supply- and demand- side factors play in explaining the observed variations in coverage. A household's access to electricity or water is modeled as a function of indicator variables for poverty-level (poor versus nonpoor), tenancy status (tenant versus owner), and location (at the division level) of the household. Statistically significant coefficients for household poverty level and/or tenancy status in such analysis will show that demand-side factors help to explain the observed variations in access across the divisions in the given city. Similarly, statistically significant coefficients for division indicator variables will indicate that significant variation 6 However, we do an analogous type of analysis for Johannesburg's underserved areas (see appendix A). 20 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS in access rates across the divisions in a given city can be attributed to supply-side factors. The extent of such roles can be inferred from the estimated odds ratios of the corresponding coefficients, which show the ceteris paribus effects of the household's poverty level, tenancy status, and residential location on the probability that the household has access to a given infrastructure service. 7 Electricity connections: poverty, tenancy, and location matter Appendix tables 17 and 18 summarize how the electricity coverage varies across the divisions in each city and within each division across its constituent EAs in Nairobi and Dakar. All divisions in both cities report at least some households with electricity. This suggests that--at least at a very coarse spatial scale--the trunk infrastructure exists in all divisions. The coverage levels do, however, vary considerably across divisions. For instance, in Nairobi, the minimum division level coverage is 8 percent (Embakasi) and the maximum level is 47 percent (Dagoretti). The corresponding figures for Dakar are 27 percent (Nolivet/Barak) and 93 percent (Diokoul). At the EA level, the story appears to be quite contrasting between the two cities. While both cities report EAs that completely lack access, in Nairobi we find that 75 percent (6 out of 8) of the divisions have at least one or more EAs that completely lack access to electricity and about 25 percent (21 out of 88) of all EAs have no access. But only 10 percent of the divisions in Dakar have at least one or more EAs that completely lack access, and just about 1 percent of all EAs have no access. Although Dakar exhibits a significantly higher access level for electricity throughout its slum settlements, the access level is much lower in Nairobi, and that too is concentrated in about 75 percent of its slum settlements. Appendix tables 19 and 20 present the results from the household-level logistic regression results that shed insight into the extent to which tenancy, demand, and supply help to explain the observed spatial differences in coverage levels in the two cities. The results show (at 10 percent significance) that in each city the likelihood of a household's access to electricity is affected by all three factors. That is, households' access to electricity is affected by their poverty level, their residential location, and also their tenancy status (with tenants being worse off relative to homeowners). Specifically, the likelihood of a poor household having access to electricity is reduced to only 57 percent and 54 percent of that of a nonpoor household in Nairobi and Dakar respectively. Similarly, the likelihood of a tenant household having access to electricity is reduced to only 37 percent and 40 percent of that of a non-tenant household in Nairobi and Dakar respectively. As the results also show, several locations have odds ratios that are statistically significant in each city. This indicates that, after controlling for both tenancy and poverty/affordability, residential location also matters greatly in determining access--a likely reflection of differences in supply-side bottlenecks. For example, the likelihood that a resident of Kibera in Nairobi will have an electricity connection is only 33 percent of that of a slum resident in Nairobi's Central division. Similarly, the likelihood of households in Dakar's Dalifort neighborhood of having access to electricity is only 31 percent of those residing in the city's Arafat settlement. 7 We also use a heuristic computational approach developed in the study by Foster and Araujo (2004) to investigate the relative roles of demand and supply side factors in explaining the variation in coverage rates across divisions in each city for electricity and piped water. The approach and results from that analysis are discussed in appendix B. 21 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Piped water: tenancy and location matter, but poverty does not Appendix tables 21 and 22 summarize the spatial variation in piped water coverage in the two cities. As with electricity, all divisions in both cities report at least some households with private piped water connections, indicating that the trunk infrastructure extends to all the divisions. The variation in coverage levels across divisions within Nairobi is higher (s.d.: 22 percent) than in Dakar (s.d.: 16 percent). For Nairobi, the minimum division level coverage is 8 percent (Embakasi) and the maximum level is 38 percent (Westlands). The corresponding figures for Dakar are 59 percent (Dalifort) and full coverage at 100 percent (Nolivet/Barak). As with electricity, the access to piped water connections also differs significantly in the two cities. In Nairobi, we find that 75 percent (6 out of 8) of the divisions have at least one or more EAs that completely lack private piped water connections and about 31 percent (27 out of 88) of all EAs have no access. In stark contrast, none of the 10 divisions in Dakar have an EA that completely lacks access to piped water connections. So, for both electricity and piped water, we find that Nairobi exhibits a significantly lower access level than Dakar all through its slum settlements, and its spatial access is concentrated in about 70­75 percent of its slum settlements. Appendix tables 23 and 24 present the results from the household level logistic regression results. In contrast to our findings for electricity, the results show (at 10 percent significance level or less) that poverty level does not play much of a role in a household's access to a piped water connection in either city. The tenancy status matters in Dakar but not in Nairobi. Specifically, the likelihood of a tenant household having access to piped water is reduced to about half (46 percent) that of a nontenant household in Dakar. As with electricity, the results again show that, after controlling for demand side factors, residential location also matters in determining access to piped water. Overall, the results indicate that a household's access to piped water is affected essentially only by supply side factors in Nairobi, but both tenancy status and supply side factors play a role in Dakar. Unlike the case of electricity, water connections are not influenced by a household's poverty level in either city. 22 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS 5 Conclusions The preceding analysis shows that much can be learned from a comparative study of African cities. Here we underscore what we consider to be three salient conclusions. A first key finding of our work is that, within the admittedly small sample of cities that we examined, neither human capital nor income appears to fully explain the observed variation in infrastructure access. This finding, if borne out by additional work, presents a substantial puzzle--and challenges the idea that reduction in income poverty is sufficient to improve the lives of slum residents. A second key insight from this work is that the challenges facing slum residents vary considerably from neighborhood to neighborhood, and still more from city to city. This heterogeneity raises an important set of social science questions: How do variations in poverty and human capital relate to variations in urban infrastructure? What institutions mediate this relationship? What lessons do better- served places offer to places still struggling to meet basic needs? The remarkable variation in the data also suggests an important lesson for policymakers: Interventions must be adapted to local conditions. The third lesson is primarily methodological. The current paper shows how a multisectoral approach reveals insights that sector-specific studies would almost certainly miss. Efforts to understand urban poverty must recognize multiple dimensions. We show how visual presentations of carefully chosen indicators--grouped into what we term the development diamond, the living conditions diamond, and the infrastructure polygon--can provide a simple yet powerful tool for comparative analysis. Demographic data from Africa (and from around the world) show that future generations will live mostly in cities. But under what living conditions? This remains unclear. If we are to chart a course to a prosperous, sustainable urban future, then we need to understand clearly where we are today, and the play of forces that got us here. The research described here points to a rich set of issues for additional research. Priorities include: Better data. The power of comparative analysis proceeds from good data drawn from many different cities. Furthermore, scholars require longitudinal (panel) data in order to draw conclusions regarding urban outcomes and their underlying causes. Data that sheds light specifically on informal sources of income will be particularly useful. Better theory. One line of investigation that seems ripe is the institutional context of improvements in slum conditions. How do the key institutional players and the rules of the game vary from city to city? Are there important success stories that can help ground a theory of the institutional antecedents for systematic improvements in slum conditions? Better analytical frameworks. These will help us arrive at better interpretations of multisectoral and multidimensional data on living conditions and poverty, to translate such data into well-tailored urban interventions, and to assist in communicating the justification for selected interventions. 23 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS References Foster, V., and M. C. Araujo. 2004. "Does infrastructure reform work for the poor? A case study from Guatemala." Policy Research Working Paper 3185. World Bank, Washington, D.C. Davis, M. 2006. Planet of Slums. London and New York: Verso. Gulyani, S., and D. Talukdar. 2007. "Inside Informality: The Links between Poverty, Micro-enterprises, and Living Conditions in Nairobi's Slums," Working Paper, Africa Urban and Water 1. World Bank, Washington, D.C. ------. 2008. "Slum Real Estate: The Low-Quality High-Price Puzzle in Nairobi's Slum Rental Market and Its Implications for Theory and Practice," World Development 36(10). UNFPA. 2007. State of world population 2007. New York: United Nations. 24 Tablular appendix As a general note to the tables in this annex, all household-level data analysis is based on sampling weights used in the survey; individual level data analysis is not. Table 1. Demographics, household size, and composition Nairobi Dakar Johannesburg N % of total N % of total N % of total Households 1,755 100 1,960 100% 1,618 31.7% Household size 2.97 9.58 3.69 Single-person households 482 27.6 34 1.7% 148 9.2 % Female-headed households 304 17.3 400 22.0% Mean age of household head (yrs) 34.8 52.4 Median age: household head (yrs) 32.0 52 Age profile N N Age 0-4 825 15.7 2,315 12.3 Age 5­14 (school age children) 976 18.6 4,361 23.1 Age 15+ ("adults") 3455 65.7 12,213 64.7 Age 0­10 1481 24.4 Age 11­20 1075 17.7 Age 21­40 2474 40.7 Age 41­65 971 16.0 > 65 79 1.3 Total no. of individuals 5256 100.0% 18,889 100.0 6080 100.0% Gender ratio, male: female Ratio (N) Ratio (%) Ratio (N) Ratio (%) Ratio (N) Ratio (%) All individuals 2899: 2357 55: 45 9470:9419 50:50 3019: 3061 50:50 Table 2. Educational attainment across individuals Individuals 15 years of age or older Dakar Johannesburg (slum) Nairobi N % N % N % None 786 20.6 7,721 64 937 26.4 Primary 1,849 48.4 3,296 27.3 933 26.3 Secondary 1,162 30.4 820 6.8 1,000 28 Postsecondary (including vocational/technical) 22 0.6 220 1.8 564 15.9 Don't know/other -- - - - 119 3.4 Total 3819 100 12057 100 3434 100 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 3. Within households Attainment of the most educated member of the household; individuals 15 years of age or more Nairobi Dakar Johannesburg N % N % N % None 223 12.7 466 24.7 239 14.1 Primary 833 46.8 848 43.3 337 21.4 Secondary 686 39.8 472 23.6 574 36.1 Postsecondary (including vocational/technical) 13 0.7 174 8.4 436 28.4 Total 1755 100 1960 100 1586 100 Table 4. Employment status across individual adults Individuals 15 years of age or more) Nairobi Dakar Johannesburg N Percent N Percent N Percent Unemployed 1,008 26.3 690 5.7 1,813 51.0 Self-employed (nonfarm) 726 19.0 2,782 23.1 5 0.1 Self-employed (farm) 0 0.0 0 0.0 10 0.3 Regular employee 954 24.9 944 7.8 1,007 28.3 Casual employee 913 23.9 1,018 8.4 354 10.0 Nonpaid family worker1 14 0.4 3,043 25.2 114 3.2 Student/apprentice 172 4.5 2,052 17.0 27 0.8 Others2/Don't know 40 1.1 1,532 12.7 223 6.3 Total 3,827 100 12,061 100 3,553 100 1 Includes housewife/homemaker 2 Includes pensioner/retired persons and disabled persons Table 5. Employment status within households Nairobi Dakar Johannesburg N Percent N Percent N Percent Self-employed 1,755 19.5 1,763 18.94 1,591 0.20 Casual employed 1,755 27.8 1,763 6.53 1,591 7.95 Regular employed 1,755 27.1 1,763 6.31 1,591 21.24 Unemployed 1,755 21.3 1,763 4.60 1,591 34.10 26 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 6. Housing unit and tenure conditions: neighborhood safety Nairobi Dakar Johannesburg Indicator N Value N Value N Value HOUSING UNIT CONDITION Mean household size 1,755 3.0 1,960 9.6 1,618 3.7 Mean number of rooms per household 1,755 1.2 1,956 4.1 1,609 1.8 Mean number of persons per room 1,755 2.6 1,956 2.8 1,609 2.3 Housing Structure--External Walls: 1,755 1,951 Brick/stone/block 12.0 89.3 Concrete (permanent) 6.6 Mud/wood or mud/cement 37.2 Mud only 0.5 Wood only 6.0 3.0 Tin or corrugated iron sheet (zinc) 44.6 0.5 All others 0.2 Housing Structure--Material for Floor: 1,755 1,950 Earth/clay 31.9 9.5 Tiles 13.0 Cement 67.9 76.4 All others 0.2 1.1 TENURE CONDITIONS Home/land ownership (% of HHs): 1,755 1,960 1,618 Own both land and structure 6.0 57.6 19.5 Own the structure but not the land 2.1 13.7 68.9 Tenants 91.9 25.8 11.3 Others 0.0 2.9 0.3 Type of tenancy agreement (% of HHs): 1,613 918 Written formal tenancy agreement 3.5 5.9 Verbal agreement 96.0 49.9 No agreement (squatter) 0.5 44.2 % of households who feel they have secure tenure 1,755 50.6 1,803 71.3 Mean time in years lived in the current house 1,755 4.9 1,954 18.7 Median time in years lived in the current house 1,755 3.0 1,954 17.0 1,618 5 to 9 Mean time in years lived in the settlement 1,755 8.8 1,953 20.8 Median time in years lived in the settlement 1,755 6.0 1,953 20.0 CRIMINAL LAW & ORDER % of households which has been a victim of crime in the past year 1,755 27.3 1,955 19.1 1,618 27.9 % of households who feel safe in their settlements 1,755 36.5 1,955 51.9 27 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 7-1. Infrastructure and services Indicator Nairobi Dakar Johannesburg N Value N Value N Value Water use and unit cost Mean per capita water consumption in litres per day 1,750 23.4 1464 28.2 Median value: per capita water consumption in liters per day 20.0 21.9 Mean unit water cost in Ksh/Jerri can for those without connections 1,719 2.60 Median value: unit water cost in Ksh/Jerri can 2.00 Mean unit water cost in Ksh/Jerri can for those with connections 31 1.86 Median value: unit water cost in Ksh/Jerri can 1.38 Mean unit water cost in xx/liter for those without connections 813 0.10 Median value: unit water cost in xx/liter for those without connections 0 Mean unit water cost in xx/liter can for those with connections 1007 1.1 Median value: unit water cost in xx/liter can for those with connections 0.56 Primary sources of water (%) Piped in house/on-site: formal connection 67 4 1345 69 6 0 Piped in house/on-site: informal connection 36 2 Yard tap 240 15 274 15 589 36 Water kiosk 1,129 64 Water vendors 33 2 Private/public well 7 0 Neighbor 17 1 57 3 149 10 Street/communal taps/standpipes 708 42 Ground tanks next to house 44 3 Borehole/rainwater/well 0 0 Groundwater/natural sources 87 5 Other 6 0 86 7 More than one source 15 1 11 1 No single primary water source 167 9 260 12 TOTAL 1,755 100 1,960 100 1,618 100 28 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 7-2. Infrastructure and services (cont'd) Indicator Nairobi Dakar Johannesburg N Value N Value N Value Connections to electricity (% of households) Households with their homes connected to electricity 1,755 21.6 1960 81.9 1618 38.6 Primary source of home lighting (% of households)1 : 1,755 1958 1618 Electricity 22.0 82.1 38.6 Kerosene/paraffin 77.1 4.8 90.8 Bougie 12.9 Battery 17.0 Candle 85.1 Solar/other 0.9 0.2 Street lighting (%) Households who say street light services exist in their settlements 1,755 14.3 1953 58.7 Telephone access (%) Land telephone with service 1,755 0.2 1960 28.8 1618 2.1 Cell/mobile phone with service 1,755 19.7 1960 49.4 1618 22.2 Note: For Johannesburg, the data across households are not mutually exclusive, as in that survey each household was asked to give multiple (as opposed to only the "primary") sources used. 29 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 7-3. Infrastructure and services Indicator Nairobi Dakar Johannesburg N Value N Value N Value Toilet Facilities Type--% of households who use: 1,755 1,957 1,618 No facility/flying toilets 5.9 Neighbor's facilities 2.7 Individual latrine with tank 38.4 8.9 Individual ordinary pit latrine 1.5 33.1 Individual VIP latrine 0.4 33.0 1.7 Flush toilet/water closet 22.9 19.8 16.2 Traditional latrine 3.2 Chemical toilet 36.8 Public/shared latrine 59.4 2.2 Public/shared VIP latrine 8.4 Other 1.6 0.7 3.4 Average number of households sharing a toilet facility 1,615 19.1 1,658 1.4 1,329 16.2 Excreta disposal system (incl. sewerage) (%) 1,755 1,960 Formal connection public sewer 11.8 7.7 Informal connection to public sewer 17.2 Septic tank/soak pit 0.8 58.5 VIP latrine with tank 22.2 Pit latrine 63.9 Others 8.9 NA/don't know 6.3 2.7 30 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 7-4. Infrastructure and services (cont'd) Indicator Nairobi Dakar Johannesburg N Value N Value N Value Garbage disposal system (%) 1,755 1,958 1,618 No organized system 87.7 23.7 9.5 Dumping in own neighborhood 78.1 22.9 Burning/burying in own compound 9.6 0.8 Organized system 12.3 76.3 90.5 Private collection system 10.9 3.9 City/Municipal collection system 0.9 69.3 Others 0.5 3.1 Graywater disposal/drainage % of households with the following graywater disposal system: 1,755 1,960 Pour into the drain 70.7 6.5 Pour into the road or pavement 18.7 62.2 Pour into pit latrine 1.0 In a septic tank 2.5 Puisard 6.7 In an old well 1.5 Caniveaux 3.5 Other 9.6 17.1 % of households with a drain outside their house 1,755 57.6 1,960 6.7 % of households whose drain works properly most of the time 1,011 43.7 154 75.3 31 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 7-5. Infrastructure and services (cont'd) Indicator Nairobi Dakar Johannesburg N Value N Value N Value Health services % of 0­14 year old immunized against BCG (tuberculosis)1 1,801 97.7 6,384 98.9 % of households who say health services exist in their settlements 1,755 88.8 1,960 88.9 % of households who say quality of health services is at least "satisfactory" 1,618 73.7 School facilities % of households who say public school services exist in their settlements 1,755 51.8 1,960 59.3 % of households who say private school services exist in their settlements 1,755 85.2 1,960 58.1 % of households whose children need to travel 30 min. or less to school 1,388 69.8 Roads % of households who say internal road systems exist within their settlements 1,755 97.6 1,960 35.0 % of households who say access road services to their settlements exist 1,755 98.3 1,960 55.5 Mode of transportation to school and work by individuals (%) 1, 2 By foot 2,680 55.0 5,949 39.5 84.0 Bike 64 1.3 36 0.2 0.7 Motorbike 99.8 Private car 6 0.1 240 1.6 1.6 Shared taxi / gypsy taxi 3 0.1 100 0.7 Taxi 2 0.0 36 0.2 0.5 Microbus/matatu 962 19.8 89 0.6 93.0 Bus regular 16 0.3 2,230 14.8 7.5 Train 21.3 Other 58 1.2 261 1.7 0.4 NA 1,033 21.2 5,575 37.0 No response 47 1.0 1 0.0 Do not know 542 3.6 TOTAL 4,871 100 15,059 100 1,618 Selective overall infrastructure access % of households who have neither piped water or electricity 1,755 66.0 1,960 2.8 1,618 44.3 % of households who have both piped water and electricity 1,755 6.5 1,960 75.7 1,618 30.7 % of households who have no piped water or electricity or permanent wall 1,755 62.8 1,951 0.5 % of households who have piped water and electricity and permanent wall 1,755 3.4 1,951 74.3 1 Based on individual level survey data that is not weighted since the available sampling weight information is at the household unit level. 2For Johannesburg, the data is collected at the household level and are not mutually exclusive, as in that survey each household was asked to give multiple modes (instead of the primary mode) used by household members. 32 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 8. Share of major expenditures in total monthly income and expenses Percent Nairobi Dakar N Mean Median N Mean Median As percentage of basic HH expenditure in a typical month Rent 1,601 17.0 14.6 474 82.5 12.0 Food 1,754 58.8 52.9 1,832 n.a. n.a. Transport 1,753 11.4 6.7 1,427 n.a. n.a. Water (All households (HH)) 1,753 4.4 3.6 n.a. n.a. Water (HH with connection) 1,380 9.6 2.2 Water (HH without connection) 937 3.1 0.0 Electricity 362 4.7 4.0 1,480 36.5 2.5 Refuse collection 161 1.0 0.8 64 8.4 0.3 As percentage of monthly HH income Rent 1,282 11.7 10.0 329 22.4 15.8 Food 1,400 41.2 37.5 1,171 n.a. n.a. Transport 1,399 7.7 4.3 935 n.a. n.a. Water (All HHs) 1,400 3.1 2.6 n.a. n.a. Water (HH with connection) 899 5.2 3.2 Water (HH without connection) 616 0.9 0.0 Electricity 253 3.5 2.9 944 7.7 3.3 Refuse collection 129 0.7 0.6 37 0.6 0.4 33 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 9. Demographics, household size, and composition Nairobi Dakar Poor Nonpoor Poor Nonpoor Percent N Percent of total N Percent of total N of total N Percent of total Households 1,282 72.5 473 27.5 1615 82.4 345 17.6 Household size a, d 3.38 1.88 10.06 7.25 Single-person households a 199 15.2 283 60.1 13 0.8 21 6.1 Female-headed households c 238 18.5 66 14.0 333 22.0 67 22.0 Mean age of household head (yrs) d 34.8 34.6 53.2 48.1 Median age: household head (yrs) 32.0 32.0 53 47 Age profile N Percent N Percent N Percent N Percent Age 0­4 717 16.5 108 11.9 1,968 12.0 347 13.8 Age 5­14 (school age children) 877 20.2 99 10.9 3,799 23.2 562 22.3 Age 15+ ("adults") 2,751 63.3 704 77.3 10,601 64.8 1,612 63.9 Total no. of individuals 4,345 100.0 911 100.0 16,368 100.0 2,521 100.0 Ratio Ratio Gender ratio, male: female Ratio (N) Ratio (%) Ratio (N) Ratio (%) (N) (%) Ratio (N) Ratio (%) 8,203: All individuals 2,320: 2,025 53: 47 579: 332 64: 36 8,165 50:50 1,267: 1,254 50:50 Note: Statistical significance for difference in the corresponding indicator value between "poor" and "nonpoor" households: (1) In Nairobi: "a"-- 1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. Table 10. Education level across individual adults Individuals 15 years of age or older Nairobi Dakar Poor Nonpoor Poor Nonpoor Education level N % N % N % N % Less than primary education 669 22.6 117 13.6 6,814 65.2 907 56.7 Completed primary education 1,494 50.5 355 41.2 2,809 26.9 487 30.5 Some/completed secondary education 790 26.7 372 43.2 664 6.4 156 9.8 Postsecondary training 5 0.2 17 2.0 171 1.6 49 3.1 Total 2,958 100 861 100 10,458 100 1,599 100 34 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 11. Highest education level within households Nairobi Dakar Poor Nonpoor Poor Nonpoor Education level N % N % N % N % Less than primary education 164 12.8 59 12.3 379 24.5 87 25.6 Completed primary education 628 48.2 205 43.3 725 44.9 123 35.2 Some/completed secondary education 486 38.7 200 42.5 376 22.7 96 28.0 Postsecondary training 4 0.3 9 1.9 135 7.9 39 11.2 Total 1,282 100 473 100 1,615 100 345 100 Note: Difference between P-NP household is significant at 1% level for both cities. Table 12. Employment status across individual adults (>15 years of age) Nairobi Dakar Poor Nonpoor Poor Nonpoor N Percent N Percent N Percent N Percent Unemployed 873 29.4 135 15.7 592 5.7 98 6.1 Self-employed (non-farm) 544 18.3 182 21.1 2,418 23.1 364 22.7 Regular employee 646 21.8 308 35.8 798 7.6 146 9.1 Casual employee 716 24.1 197 22.9 878 8.4 140 8.7 Nonpaid family worker1 13 0.4 1 0.1 2,611 25.0 432 27.0 Student/apprentice 144 4.9 28 3.3 1,807 17.3 245 15.3 Others2/Don't know 30 1.0 10 1.2 1,354 13.0 178 11.1 Total 2,966 100 861 100 10,458 100 1,603 100 1 Includes housewife/homemaker 2 Includes pensioner/retired persons and disabled persons Table 13. Employment status within households Nairobi Dakar Poor Nonpoor Poor Nonpoor N Percent N Percent N Percent N Percent Self employed a 1,282 18.6 473 22.0 1,442 18.5 321 20.8 Casual employed 1,282 27.8 473 27.9 1,442 6.2 321 8.1 Regular employed a, d 1,282 23.4 473 36.8 1,442 5.8 321 8.8 Unemployed a 1,282 25.4 473 10.5 1,442 4.7 321 4.2 Note: Statistical significance for difference in the corresponding indicator value between "poor" and "nonpoor" households: (1) In Nairobi: "a"-- 1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. 35 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 14-1. Housing unit and tenure conditions: neighborhood safety Nairobi Dakar Poor Nonpoor Poor Nonpoor Indicator N Value N Value N Value N Value HOUSING UNIT CONDITION Mean household size a, d 1,282 3.38 473 1.88 1,615 10.1 345 7.3 Mean number of rooms per household d 1,282 1.2 473 1.2 1,613 4.1 343 3.7 Mean number of persons per room a, d 1,282 3.0 473 1.6 1,613 2.9 343 2.3 Housing structure--material for external walls: c, d 1,282 473 1,607 344 Brick/stone/block 9.3 19.0 91.4 79.1 Concrete (permanent) 4.6 16.1 Mud/wood or mud/cement 40.2 29.1 Mud only 0.4 1.0 Wood only 6.3 5.2 2.9 3.8 Tin or corrugated iron sheet (zinc) 44.0 46.5 0.7 0.0 All others 0.2 0.2 Housing Structure- Material for Floor: d 1,282 473 1,606 344 Earth/clay 33.7 27.3 10.5 4.8 Tiles 10.3 25.8 Cement 66.1 72.7 78.4 66.7 All others 0.2 0.0 0.8 2.7 Note: Statistical significance for difference in the corresponding indicator value between poor and nonpoor households: (1) In Nairobi: "a"--1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. 36 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 14-2. Housing unit and tenure conditions; neighborhood safety (cont'd) Nairobi Dakar Poor Nonpoor Poor Nonpoor Indicator N Value N Value N Value N Value TENURE CONDITIONS Home/land ownership composition (% of HHs): d 1,282 473 1,615 345 Own both land and structure 5.8 6.8 59.1 49.9 Own the structure but not the land 2.1 2.0 14.4 10.0 Tenants 92.1 91.2 23.7 36.4 Others 0.0 0.0 2.8 3.7 Type of tenancy agreement (% of HHs): e 1,182 431 720 198 Written formal tenancy agreement 2.9 5.2 4.7 10.2 Verbal agreement 96.5 94.4 49.0 53.3 No agreement (squatter) 0.6 0.5 46.2 36.5 % of households who feel they have secure tenure 1,282 49.6 473 53.5 1,478 71.3 325 71.2 Mean time in years lived in the current house: b, d 1,282 5.1 473 4.4 1,610 19.6 344 14.7 Median time in years lived in the current house 1,282 3.0 473 3.0 1,610 18.0 344 10.0 Mean time in years lived in the settlement: d 1,282 8.6 473 9.1 1,609 21.4 344 17.7 Median time in years lived in the settlement 1,282 6.0 473 6.0 1,609 20.0 344 15.0 CRIMINAL LAW & ORDER % of households which has been a victim of crime in the past year 1,282 26.1 473 30.5 1,610 19.3 345 17.9 % of households who feel safe in their 1,282 37.7 473 33.1 1,614 50.9 341 56.8 settlements: f Note: Statistical significance for difference in the corresponding indicator value between "poor" and "nonpoor" households: (1) In Nairobi: "a"-- 1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. 37 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 15-1. Infrastructure and services Nairobi Dakar Poor Nonpoor Poor Nonpoor Indicator N Value N Value N Value N Value Water use and unit cost Mean per capita water consumption in liters per day a, d 1,277 21.0 473 29.8 1,215 26.8 249 34.9 Median value: per capita water consumption in L/day 20.0 20.0 20.9 25.0 Mean unit water cost in Ksh/Jerri can for those without connections b 1,257 2.55 462 2.74 Median value: unit water cost in Ksh/Jerri can 2.00 3.00 Mean unit water cost in Ksh/Jerri can for those with connections 20 1.77 11 2.03 Median value: unit water cost in Ksh/Jerri can 1.25 1.48 Mean unit water cost in XOF/liter for those without connections 665 0.10 148 0.11 Median value: unit water cost in XOF/liter for those without connections 0.02 0.55 Mean unit water cost in XOF/liter can for those with connections 827 1.1 180 1.1 Median value: unit water cost in XOF/liter can for those with connections 0.55 0.67 Primary sources of water (%) Private piped 48 4 19 4 1102 68 243 70 Piped in house/on-site: informal connection Yard tap 164 14 76 17 238 16 36 11 Water kiosk 835 65 294 61 Water vendors 26 2 7 2 Private/public well 6 0 1 0 Neighbor 15 1 2 0 46 3 11 3 Street/communal taps/standpipes Ground tanks next to house Borehole/rainwater/well Groundwater/natural sources 49 4 38 8 Other 6 0 0 0 More than one source 14 1 1 0 11 1 0 0 No single primary water source 131 10 36 8 206 11 54 15 TOTAL 1,282 100 473 100 1,615 100 345 100 Note: Statistical significance for difference in the corresponding indicator value between "poor" and "nonpoor" households: (1) In Nairobi: "a"-- 1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. 38 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 15-2. Infrastructure and services (cont'd) Nairobi Dakar Poor Nonpoor Poor Nonpoor Indicator N Value N Value N Value N Value Connections to electricity % of households with their homes connected to electricity: a, e 1,282 18.2 473 30.5 1,615 80.9 345 86.4 Primary source of home lighting (% of households) : a 1,282 473 1,613 345 Electricity 19.1 29.6 81.2 86.1 Kerosene/paraffin 80.0 69.5 5.2 3.0 Bougie 13.3 10.9 Solar/other 0.9 0.9 0.3 0.0 Street lighting (%) Households who say street light services exist in their settlements 1,282 14.4 473 13.9 1,608 57.9 345 62.7 Telephone access (%) 1,282 473 1,615 345 Land telephone with service d 0.2 0.2 26.9 38.0 Cell/mobile phone with service a, d 16.4 28.4 47.3 59.3 Note: Statistical significance for difference in the corresponding indicator value between poor and nonpoor households: (1) In Nairobi: "a"--1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. 39 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 15-3. Infrastructure and services (cont'd) Nairobi Dakar Poor Nonpoor Poor Nonpoor Indicator N Value N Value N Value N Value Toilet facilities Type--% of households who use: 1,282 473 1,612 345 No facility/flying toilets 5.8 6.1 Neighbor's facilities 2.81 2.37 Individual latrine with tank 39.63 32.48 Individual ordinary pit latrine 1.8 0.5 Individual VIP latrine 0.4 0.4 32.70 34.29 Flush toilet/water closet 22.1 24.9 18.51 26.15 Traditional latrine 3.30 2.66 Public/shared latrine 60.5 56.4 2.27 1.63 Public/shared VIP latrine 7.7 10.4 Other 1.6 1.3 0.79 0.43 Average number of households sharing a toilet facility e 1,180 19.4 435 18.5 1,398 1.4 260 1.8 Excreta disposal system (incl. sewage) (%) c, e 1,282 473 1,615 345 Formal connection public sewer 11.0 14.0 7.7 7.7 Informal connection to public sewer 18.3 14.2 Septic tank/soak pit 0.4 1.6 56.8 66.7 VIP Latrine with tank 23.0 18.1 Pit latrine 64.2 63.2 Others 9.4 6.2 NA/Don't know 6.1 6.9 3.1 1.2 Note: Statistical significance for difference in the corresponding indicator value between "poor" and "nonpoor" households: (1) In Nairobi: "a"-- 1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. 40 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 15-4. Infrastructure and services (cont'd) Nairobi Dakar Poor Nonpoor Poor Nonpoor Indicator N Value N Value N Value N Value Garbage disposal system (% of households) d 1,282 473 1,613 345 No organized system 88.8 84.9 25.4 15.7 Dumping in own neighborhood 79.0 75.7 24.7 14.5 Burning/burying in own compound 9.8 9.2 0.7 1.2 Organized system 11.2 15.1 74.6 84.3 Private collection system 10.1 13.1 67.3 79.2 City/Municipal collection system 0.7 1.4 4.0 3.3 Others 0.4 0.6 3.3 1.8 Graywater disposal/drainage % of households with the following "graywater" disposal system: d 1,282 473 1,615 345 Pour into the drain 71.0 70.0 6.50 6.62 Pour into the road or pavement 19.4 16.8 63.03 57.90 Pour into pit latrine 1.0 0.7 In a septic tank 1.86 5.55 Puisard 5.98 10.11 In an old well 1.33 2.15 Caniveaux 3.61 3.18 Other 8.5 12.5 17.68 14.49 % of households with drain outside house 1,282 56.4 473 60.5 1,615 6.4 345 8.4 % of households whose drain works most of the time 1,282 25.0 473 25.7 122 73.8 32 80.9 Note: Statistical significance for difference in the corresponding indicator value between "poor" and "nonpoor" households: (1) In Nairobi: "a"-- 1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. 41 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 15-5. Infrastructure and services (cont'd) Percent Nairobi Dakar Poor Nonpoor Poor Nonpoor Indicator N Value N Value N Value N Value Health services % of 0­14-year-olds immunized against BCG (tuberculosis)1 1,559 97.8 207 96.6 5,509 99.0 875 98.3 % of households who say health services exist in their settlements b, f 1,282 87.4 473 92.3 1,615 89.6 345 85.5 School facilities % of HHs who say public school services exist in their settlements f 1,282 50.9 473 54.4 1,615 59.7 345 57.6 % of HHs who say private school services exist in their settlements a 1,282 84.0 473 88.4 1,615 57.9 345 59.3 Road facilities % of HHs who say internal road systems exist within their settlements f 1,282 97.4 473 98.0 1,615 36.2 345 29.0 % of HHs who say access road services to their settlements exist c 1,282 98.0 473 98.9 1,615 54.5 345 60.2 Primary mode of transportation to school and work by individuals (%) By foot 2,259 58.1 421 42.9 5,111 39.1 838 42.2 Bike 47 1.2 17 1.7 29 0.2 7 0.4 Private car 1 0.0 5 0.5 188 1.4 52 2.6 Shared taxi / Gypsy taxi 3 0.1 0 0.0 84 0.6 16 0.8 Taxi 0 0.0 2 0.2 20 0.2 16 0.8 Microbus/matatu 595 15.3 367 37.4 80 0.6 9 0.5 Bus regular 13 0.3 3 0.3 1,921 14.7 309 15.6 Other 40 1.0 18 1.8 227 1.7 34 1.7 NA 896 23.0 137 14.0 4,933 37.7 0 0.0 No response/ Do not Know 36 0.9 11 1.1 480 3.8 705 35.5 TOTAL 3,890 100 981 100 13,073 100 1,986 100 Selective overall infrastructure access % of households who has neither piped water or electricity a 1,282 68.6 473 59.1 1,615 2.7 345 2.9 % of households who has both piped water and electricity a 1,282 4.8 473 10.9 1,615 74.9 345 79.5 % of households who has no piped water or electricity or permanent wall a 1,282 65.2 473 56.3 1,615 0.3 345 1.1 % of households who has piped water and electricity and permanent wall a 1,282 2.0 473 7.1 1,615 73.7 345 76.9 Note: Statistical significance for difference in the corresponding indicator value between "poor" and "nonpoor" households: (1) In Nairobi: "a"-- 1% level; "b"--5% level; "c"--10% level; (2) In Dakar: "d"--1% level; "e"--5% level; "f"--10% level. 1 Based on individual level survey data that is not weighted since the available sampling weight information is at the household unit level. 42 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 16. Share of major expenditures in total monthly income and expenses Percent Nairobi Dakar Poor Nonpoor Poor Nonpoor N Mean Median N Mean Median N Mean Median N Mean Median As percentage of basic HH expenditure in a typical month Rent 1,173 17.0 15.0 428 16.9 13.9 359 101.3 12.5 115 23.9 10.0 Food 1,281 60.3 54.5 473 54.8 48.0 n.a. n.a. n.a. n.a. n.a. n.a. Transport 1,280 9.7 0.0 473 15.8 13.3 n.a. n.a. n.a. n.a. n.a. n.a. Water (all HHs) 1,280 4.8 4.0 473 3.4 2.6 n.a. n.a. n.a. n.a. n.a. n.a. Water (HH with connection) 1117 10.8 2.3 263 4.7 2.1 Water (HH without connection) 755 3.8 0.0 182 0.1 0.0 Electricity 218 5.4 4.2 144 3.7 3.1 1,200 43.9 2.8 280 4.5 1.7 Refuse collection 117 1.0 0.8 44 1.0 0.6 54 9.9 0.3 10 0.2 0.2 As percentage of monthly HH income Rent 927 11.9 10.6 355 11.3 10.0 241 23.2 16.7 88 20.4 12.5 Food 1,011 43.1 40.0 389 36.5 35.0 n.a. n.a. n.a. n.a. n.a. n.a. Transport 1,010 6.6 0.0 389 10.3 10.0 n.a. n.a. n.a. n.a. n.a. n.a. Water (All HHs) 1,011 3.4 3.0 389 2.3 1.8 Water (HH with connection) 714 5.5 3.3 185 4.3 2.7 Water (HH without connection) 488 1.0 0.0 128 0.4 0.0 Electricity 151 4.0 3.3 102 2.7 2.5 746 8.1 3.6 198 6.2 2.5 Refuse Collection 94 0.8 0.6 35 0.6 0.4 31 0.7 0.5 6 0.4 0.3 Table 17. Nairobi home electricity: coverage insights Coverage (percentage of households with access) # of EAs Across constituent EAs with no Division Number of EAs electricity Number of HHs Across divisions min max med mean sd 1. Central 10 0 198 28 5 90 18 28 28 2. Makadara 9 1 175 10 0 21 10 10 6 3. Kasarani 10 2 198 11 0 30 10 11 9 4. Embakasi 9 5 179 8 0 35 0 8 12 5. Pumwani 6 1 120 35 0 85 30 35 33 6. Westlands 3 0 60 27 10 60 10 27 29 7. Dagoretti 17 2 331 47 0 100 40 47 32 8. Kibera 24 10 494 12 0 85 8 13 19 ALL 88 21 1,755 22 0 100 11 22 26 43 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 18. Dakar home electricity: coverage insights Coverage (percentage of households with access) # of EAs Across constituent EAs with no Division Number of EAs electricity Number of HHs Across divisions min max med mean sd Nolivet/Barak 4 1 40 27 0 73 17 27 32 Arafat 11 0 200 87 65 100 89 87 11 Hann 12 0 250 86 75 96 88 86 8 Dalifort 4 0 60 70 60 75 72 70 7 Santiaba 4 0 80 81 70 100 76 81 13 Taïf 10 0 200 79 65 90 83 79 9 Guinaw 16 0 400 85 73 95 88 85 9 Médina 16 0 300 80 60 100 80 80 12 Wakhinane 9 0 180 83 75 95 80 83 7 Diokoul 13 0 250 93 75 100 95 93 8 ALL 99 1 1960 82 0 100 85 82 16 Table 19. Nairobi home electricity: logistic regression results for household access Variable Odds-ratio Standard error p-value Poor 0.574 0.096 0.001 Tenant 0.371 0.111 0.001 Locations 1. Central (base) 2. Makadara 0.305 0.132 0.008 3. Kasarani 0.325 0.141 0.011 4. Embakasi 0.219 0.134 0.015 5. Pumwani 1.497 0.986 0.542 6. Westlands 0.769 0.594 0.735 7. Dagoretti 2.032 0.953 0.134 8. Kibera 0.330 0.161 0.026 N 1,755 U2 0.14 44 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 20. Dakar home electricity: logistic regression results for household access Variable Odds-Ratio Standard error p-value Poor 0.544 0.104 0.002 Tenant 0.399 0.061 0.000 Locations 1. Nolivet/Barak 0.052 0.043 0.001 2. Arafat (base) 3. Hann 0.815 0.280 0.553 4. Dalifort 0.313 0.090 0.000 5. Santiaba 0.418 0.175 0.041 6. Taïf 0.505 0.160 0.034 7. Guinaw 0.709 0.234 0.300 8. Médina 0.475 0.144 0.016 9. Wakhinane 0.670 0.210 0.204 10. Diokoul 1.355 0.546 0.454 N 1,960 U2 0.07 Table 21. Nairobi piped drinking water: coverage insights # of EAs Coverage (percentage of households with access) with no Across constituent EAs piped water Division Number of EAs Number of HHs Across divisions min max med mean sd 1. Central 10 4 198 24 0 75 23 24 25 2. Makadara 9 3 175 13 0 72 5 13 23 3. Kasarani 10 0 198 23 5 70 18 23 21 4. Embakasi 9 5 179 8 0 45 0 8 15 5. Pumwani 6 0 120 24 5 85 13 24 31 6. Westlands 3 1 60 38 0 85 30 38 43 7. Dagoretti 17 7 331 21 0 70 20 21 23 8. Kibera 24 7 494 11 0 76 5 11 17 ALL 88 27 1755 19 0 85 10 18 22 45 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 22. Dakar piped drinking water: coverage insights # of EAs Coverage (percentage of households with access) with no Across constituent EAs piped Division Number of EAs water Number of HHs Across divisions min max med mean sd Nolivet/Barak 4 0 40 100 100 100 100 100 0 Arafat 11 0 200 93 80 100 95 93 6 Hann 12 0 250 91 75 100 93 91 7 Dalifort 4 0 60 59 18 93 63 59 31 Santiaba 4 0 80 99 96 100 100 99 2 Taïf 10 0 200 93 75 100 95 93 9 Guinaw 16 0 400 93 75 100 95 93 7 Médina 16 0 300 95 80 100 96 95 6 Wakhinane 9 0 180 64 5 90 75 64 27 Diokoul 13 0 250 97 90 100 100 97 4 ALL 99 0 1960 90 5 100 95 90 16 Table 23. Nairobi piped water: logistic regression results for household access Variable Odds-ratio Standard error p-value Poor 0.803 0.137 0.202 Tenant 0.782 0.235 0.417 Locations 1. Central (base) 2. Makadara 1.083 0.964 0.929 3. Kasarani 1.294 0.725 0.647 4. Embakasi 0.239 0.176 0.055 5. Pumwani 1.271 0.998 0.761 6. Westlands 3.406 3.234 0.200 7. Dagoretti 1.007 0.546 0.990 8. Kibera 0.530 0.296 0.259 N 1,755 U2 0.04 46 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table 24. Dakar piped water: logistic regression results for household Variable Odds-ratio Standard error p-value Poor 0.715 0.156 0.129 Tenant 0.460 0.072 0.000 Locations 1. Nolivet/Barak* 2. Arafat (base) 3. Hann 0.686 0.244 0.293 4. Dalifort 0.113 0.061 0.000 5. Santiaba 4.772 4.152 0.076 6. Taïf 0.859 0.396 0.744 7. Guinaw 1.067 0.411 0.866 8. Médina 1.172 0.485 0.702 9. Wakhinane 0.118 0.053 0.000 10. Diokoul 1.765 0.730 0.173 N 1,920 U2 0.15 * Dropped because all the 40 sample households in this location have access to piped drinking water. 47 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Appendix A Spatial heterogeneity in service access in Johannesburg This section examines differences in access across space within Johannesburg. As noted earlier in Section 4 in the paper, we discuss this analysis for Johannesburg separately from that of Nairobi and Dakar because of lack of comparable underlying data in two ways. First, unlike for Nairobi and Dakar, the data for Johannesburg are not exclusively for slum settlement but for selected underserved urban areas. Second, we do not have a household poverty-level measure for Johannesburg, as we do for Nairobi and Dakar. We still wanted to glean some insights, however, from the spatial heterogeneity in infrastructure access within Johannesburg based on an analogous type of analysis. Accordingly, here we rely on the 11 divisions and 253 constituent enumeration areas (EAs) that were used to design the sampling frame for Johannesburg underserved areas. We examine the same two categories of infrastructure: home electricity and piped-water connections as we did for Nairobi and Dakar. As for the logistic regressions at the household level to investigate whether the supply and demand side factors do play roles in explaining the observed variations in coverage levels of these two types of infrastructure, we use whether a household resides in a shack-type dwelling or not as the relative measure of household poverty level in absence of direct data on poverty level. In all other respects, the analysis here follows the same approach as we used for Nairobi and Dakar. The results are discussed below. Electricity connections Table A1a summarizes how the electricity coverage varies across the 11 divisions in the city and within each division across its constituent EAs. All the divisions report at least some households with electricity suggesting the presence of trunk infrastructure in all divisions. However, the coverage levels do vary considerably across divisions with a mean of 71 percent and s.d. of 39. The minimum level coverage is 49 percent (Johannesburg South) and the maximum level is 94 percent (Sandton/Rosebank). We also find that 64 percent (7 out of 11) of the divisions have at least one or more EAs that completely lack access to electricity and about 16 percent (41 out of 253) of all EAs have no access. Table A1b presents the results from the household-level logistic regression results that shed insight into the extent of roles by demand and supply side factors in explaining the observed spatial differences in electricity coverage levels within the city. The results show (at 10 percent significance level or less) that tenancy status does not play much of a role in terms of a household's access to electricity. However, in terms of whether the household lives in a shack or not, it does appear to affect the probability of access. Specifically, the likelihood of a shack household having access to electricity is only about 14 percent of that of a nonshack household. We also find that several location odds ratios are statistically significant. That indicates that, after controlling for demand side factors, residential location also matters greatly in determining access--a likely reflection of differences in supply side bottlenecks across locations. For example, the likelihood of households residing in Roodepoort of having access to electricity is only 10 percent of those residing in Alexandra division. 48 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table A1a. Johannesburg home electricity: coverage insights Coverage (percentage of households with access) # of EAs Across constituent EAs Number with no Number Across Division of EAs electricity of HHs divisions min max med mean sd Alexandra 35 0 738 80 30 100 88 80 20 Central region 17 0 332 87 20 100 100 87 23 Diepkloof / meadowlands 27 8 545 57 0 100 90 57 46 Diepsloot 16 4 321 67 0 100 95 67 46 Doornkop / soweto 45 12 941 55 0 100 85 55 46 Ennerdale / orange farm 58 10 1,158 75 0 100 95 75 40 Johannesburg south 4 1 81 49 0 95 51 49 39 Midrand / ivory park 31 0 568 87 7 100 95 87 24 Northcliff 3 1 59 67 0 100 100 67 58 Roodepoort 11 5 216 55 0 100 100 55 52 Sandton / rosebank 6 0 139 94 80 100 100 94 9 All 253 41 5098 71 0 100 95 71 39 Table A1b. Johannesburg home electricity: logistic regression results for household access Variable Odds ratio Standard error p-value Shack 0.142 0.029 0.000 Tenant 1.312 0.216 0.101 Locations 1. Alexandra ("base") 2. Central region 1.155 0.628 0.791 3. Diepkloof / Meadowlands 0.199 0.112 0.005 4. Diepsloot 0.694 0.626 0.686 5. Doornkop / Soweto 0.474 0.222 0.113 6. Ennerdale / Orange Farm 1.270 0.550 0.581 7. Johannesburg south 0.216 0.233 0.157 8. Midrand / Ivory Park 1.426 0.737 0.493 9. Northcliff 0.078 0.085 0.021 10. Roodepoort 0.103 0.083 0.005 11. Sandton / Rosebank 2.706 1.378 0.053 N 5098 U2 0.22 49 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Piped water connections Table A2a shows the spatial variation of the piped water coverage. As with electricity, we find that all the divisions report at least some households with piped water access suggesting the presence of trunk infrastructure in all divisions. The coverage levels do vary considerably across divisions with a mean of 75 percent and s.d. of 37. The minimum level coverage is 53 percent (Diepkloof/Meadowlands) and the maximum level is the full 100 percent coverage (Sandton/Rosebank). We also find that 55 percent (6 out of 11) of the divisions have at least one or more EAs that completely lack access to piped water and about 9 percent (23 out of 253) of all EAs have no access. Table A2a. Johannesburg piped drinking water: coverage insights Coverage (percentage of households with access) # of EAs Number with no Number Across constituent EAs Across Division of EAs piped water of HHs divisions min max med mean sd Alexandra 35 0 738 93 45 100 100 93 16 Central region 17 0 332 94 5 100 100 94 23 Diepkloof / Meadowlands 27 5 545 53 0 100 48 53 44 Diepsloot 16 4 321 63 0 100 100 63 47 Doornkop / Soweto 45 4 941 67 0 100 85 67 38 Ennerdale / Orange Farm 58 6 1,158 67 0 100 95 67 40 Johannesburg South 4 1 81 69 0 100 88 69 47 Midrand / Ivory Park 31 0 568 94 21 100 100 94 17 Northcliff 3 0 59 68 5 100 100 68 55 Roodepoort 11 3 216 64 0 100 100 64 50 Sandton / Rosebank 6 0 139 100 100 100 100 100 0 All 253 23 5098 75 0 100 100 75 37 Table A2b presents the results from the household level logistic regression results for piped water access. As in the case of electricity, we again find that (at 10 percent significance level or less) tenancy status does not play much of a role in terms of a household's access to piped water, but whether the household lives in a shack or not does. Specifically, the likelihood of a shack household having access to piped water is only about 16 percent of that of a non-shack household. We also find that, after controlling for demand side factors, residential location also matters greatly in determining access. For example, the likelihood of households residing in Roodepoort of having access to electricity is only 7 percent of those residing in Alexandra division. 50 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table A2b. Johannesburg piped drinking water: logistic regression results for household access Variable Odds ratio Standard error p-value Shack 0.157 0.040 0.000 Tenant 1.100 0.238 0.660 Locations 1. Alexandra (base) 2. Central Region 0.524 0.535 0.528 3. Diepkloof / Meadowlands 0.064 0.033 0.000 4. Diepsloot 0.174 0.158 0.057 5. Doornkop / Soweto 0.328 0.183 0.048 6. Ennerdale / Orange Farm 0.194 0.125 0.012 7. Johannesburg South 0.131 0.193 0.170 8. Midrand / Ivory Park 1.269 0.944 0.749 9. Northcliff 0.035 0.027 0.000 10. Roodepoort 0.073 0.076 0.013 11. Sandton / Rosebank* N 4,959 U2 0.23 * Dropped because all the 139 sample households in this location have access to piped drinking water. 51 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Appendix B Supply-side vs. demand-side determinants of infrastructure access This appendix first discusses a heuristic computational approach developed in the study by Foster and Arauju (2004) to investigate to what extent demand and supply side factors might explain any observed spatial variations in access across neighborhoods (EAs) within a city. We present and discuss results based on this approach for Nairobi, Dakar and Johannesburg slum/underserved settlements. The heuristic computational approach Denote by C the percentage coverage or connection rate of a service in the population. This is the number of households using the service divided by the total number of households (with appropriate survey-based household weights). Next define the access rate (A) as the number of households living in communities or primary sampling units where service is available divided by the total number of households. Finally, denote by U the take-up or hook-up rate that is the number of households actually using the service (i.e., connected to the network) divided by the number of households living in communities where service is available. The coverage rate is the product of the access and take-up rates (C=AxU). The share of the population not served by the network is 1-C. The objective is to assess whether the unserved population is not served due to a demand-side problem (the service is available, but not taken up by the households, probably because it is not affordable, but perhaps also because it is of low quality) or a supply-side problem (the service is simply not available). F&A define the pure demand-side gap (PDSG) as: PDSG A C A (1 U ) (1) This definition implies that when there is access in the areas where the households live, if a household does not take-up the service, it is symptomatic of a demand issue. Thus, lack of demand is responsible for all of the difference between the neighborhood access rate and the actual coverage rate. Next, the authors define the supply-side gap as follows: SSG (1 C ) PDSG (1 A U ) A (1 U ) 1 A (2) In other words, the supply gap is the difference between the neighborhood access rate and the coverage rate. Said differently, the sum of the pure demand-side gap, the supply-side gap, and the coverage rate is equal to one: PDSG SSG C 1 (3) However, in areas that are not covered by the network, and are responsible for the supply gap above, it is likely that even if supply were available, some households would not take up the service due to affordability issues. If one assumes that the take-up rate in non-served areas would be similar to the take- up rate in areas where there is service now, the additional coverage that we would obtain by providing 52 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS access to these areas would be equal to the supply-side gap times the take-up rate where there is access. This is defined as the pure supply-side gap: PSSG SSG U (1 A) U (4) The difference between the pure supply-side gap and the supply-side gap can then be deemed to represent a combined demand and supply-side gap, since first there is no access to the service, and second even if there were access, some households would not be connected. F&A defined this as the mixed demand and supply-side gap, defined as follows: MDSSG SSG (1 U ) (5) Given the above definitions, the proportion of the deficit in coverage that is attributed to demand-side factors is defined as the ratio of the pure demand-side gap to the unserved population. The proportion of deficit attributable to supply-side factors is the ratio of the pure supply-side gap divided by the unserved population. Finally, the proportion of deficit attributable to both demand and supply-side factors is the ratio of the mixed demand and supply-side gap divided by the unserved population. The sum of the three proportions is equal to one. Application of the approach to our data for the three cities We applied the aforesaid approach to our data for the three cities with respect to the two types of infrastructure: electricity and piped water. The results for electricity are presented in Tables B1a-c and for piped water in Tables B2a-c. A household is deemed to have access to piped water or electricity if the household lives in an EA (which is the primary sampling unit of the survey to which the household belongs) where at least one household has access. We discuss here the averages across the 3 cities for each of the two selected infrastructures. All averages are provided both with sample population weights and without weights. Since the results are quite similar, we discuss the results with respect non-weighted averages for the 3 cities. (See tables B1a-c and tables B2a-c.) Electricity The results suggest that access at the neighborhood (EA) level is fairly widespread in all the three cities: 80, 98 and 82 percent for Nairobi, Dakar and Johannesburg respectively. However, take-up rates are especially lower in Nairobi at 27 percent compared to 78 percent and 86 percent in Dakar and Johannesburg respectively. This means that the coverage rate for electricity in Nairobi is only 22 percent compared to 77 percent and 70 percent in Dakar and Johannesburg respectively. The proportion of the deficit in coverage attributable to demand-side factors is 76 percent for Nairobi compared to 97 percent and 52 percent in Dakar and Johannesburg respectively. On the other hand, the proportion of the deficit in coverage that is attributable only to supply-side factors is 6 percent for Nairobi compared to 1 percent and 42 percent in Dakar and Johannesburg respectively. The combined demand and supply-side problems account for 18 percent in Nairobi, 2 percent in Dakar and 5 percent in Johannesburg. Clearly, these results suggest that demand-side factors may be larger than supply-side factors in explaining lack of electricity infrastructure coverage in the 3 cities. 53 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Piped water As with electricity, the results for piped water suggest that access at the neighborhood (EA) level is fairly widespread in all the 3 cities: 71 percent, 100 percent and 90 percent for Nairobi, Dakar and Johannesburg respectively. However, again take-up rates are especially lower in Nairobi at 29 percent compared to 88 percent and 84 percent in Dakar and Johannesburg respectively. This means that the coverage rate for piped water in Nairobi is only 20 percent compared to 88 percent and 76 percent in Dakar and Johannesburg respectively. The proportion of the deficit in coverage attributable to demand- side factors is 64 percent for Nairobi compared to 100 percent and 68 percent in Dakar and Johannesburg respectively. On the other hand, the proportion of the deficit in coverage that is attributable only to supply-side factors is 12 percent for Nairobi compared to 0 percent and 27 percent in Dakar and Johannesburg respectively. The combined demand and supply-side problems account for 25 percent in Nairobi, 0 percent in Dakar and 6 percent in Johannesburg. Again, as with electricity, clearly these results based on the heuristic approach suggest that demand-side factors may be larger than supply-side factors in explaining lack of piped water infrastructure coverage in the three cities. Table B1a. Nairobi home electricity Percent attributable to demand-side Mixed demand and supply- Take-up rate given access attributable to both supply attributable to supply-side and demand-side factors Pure demand-side gap Pure supply-side gap Unserved population Proportion of deficit Proportion of deficit Proportion of deficit Supply-side gap Coverage rate Access rate factors only factors only side gap Division Central 100 28 28 72 72 0 0 0 100 0 0 Makadara 89 11 10 90 79 11 1 10 88 1 11 Kasarani 80 14 11 89 69 20 3 17 78 3 19 Embakasi 44 18 8 92 36 56 10 46 40 11 50 Pumwani 83 42 35 65 48 17 7 10 74 11 15 Westlands 100 27 27 73 73 0 0 0 100 0 0 Dagoretti 88 53 47 53 41 12 6 5 78 12 10 Kibera 58 21 12 88 46 42 9 33 53 10 38 Simple average 80 27 22 78 58 20 4 15 76 6 18 Weighted average 76 27 22 78 54 24 6 19 71 7 22 54 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table B1b. Dakar home electricity Percent attributable to demand-side Mixed demand and supply- Take-up rate given access attributable to both supply attributable to supply-side and demand-side factors Pure demand-side gap Pure supply-side gap Unserved population Proportion of deficit Proportion of deficit Proportion of deficit Supply-side gap Coverage rate Access rate factors only factors only side gap Division Nolivet/Barak 75 36 27 73 48 25 9 16 66 12 22 Arafat 100 87 87 13 13 0 0 0 100 0 0 Hann 100 86 86 14 14 0 0 0 100 0 0 Dalifort 100 70 70 30 30 0 0 0 100 0 0 Santiaba 100 81 81 19 19 0 0 0 100 0 0 Taïf 100 79 79 21 21 0 0 0 100 0 0 Guinaw 100 85 85 15 15 0 0 0 100 0 0 Médina 100 80 80 20 20 0 0 0 100 0 0 Wakhinane 100 83 83 17 17 0 0 0 100 0 0 Diokoul 100 93 93 7 7 0 0 0 100 0 0 Simple average 98 78 77 23 20 3 1 2 97 1 2 Weighted average 99 82 82 17 16 1 0 0 98 0 0 55 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table B1c. Johannesburg home electricity Percent attributable to demand- Pure demand-side gap attributable to supply- Pure supply-side gap Unserved population supply and demand- Proportion of deficit Proportion of deficit Proportion of deficit Mixed demand and Take-up rate given attributable to both side factors only side factors only Supply-side gap supply-side gap Coverage rate Access rate side factors Division access Alexandra 100 80 80 20 20 0 0 0 100 0 0 Central Region 100 87 87 13 13 0 0 0 100 0 0 Diepkloof / Meadowlands 70 81 57 43 13 30 24 6 30 57 13 Diepsloot 75 89 67 33 8 25 22 3 24 67 8 Doornkop / Soweto 73 75 55 45 18 27 20 7 40 45 15 Ennerdale / Orange Farm 83 90 75 25 8 17 15 2 32 61 7 Johannesburg South 75 65 49 51 26 25 16 9 51 32 17 Midrand / Ivory Park 100 87 87 13 13 0 0 0 100 0 0 Northcliff 67 100 67 33 0 33 33 0 0 100 0 Roodepoort 55 100 55 45 0 45 45 0 0 100 0 Sandton / Rosebank 100 94 94 6 6 0 0 0 100 0 0 Simple average 82 86 70 30 11 18 16 2 52 42 5 Weighted average 84 84 71 29 13 16 13 3 55 38 7 Table B2a. Nairobi piped water Percent attributable to demand- Pure demand-side gap attributable to supply- Pure supply-side gap Unserved population supply and demand- Proportion of deficit Proportion of deficit Proportion of deficit Mixed demand and Take-up rate given attributable to both side factors only side factors only Supply-side gap supply-side gap Coverage rate Access rate side factors Division access Central 60 40 24 76 36 40 16 24 47 21 32 Makadara 67 20 13 87 54 33 7 27 62 7 31 Kasarani 100 23 23 77 77 0 0 0 100 0 0 Embakasi 44 18 8 92 36 56 10 46 40 11 50 Pumwani 100 24 24 76 76 0 0 0 100 0 0 Westlands 67 57 38 62 29 33 19 14 46 31 23 Dagoretti 59 36 21 79 38 41 15 26 48 19 34 Kibera 71 16 11 89 60 29 5 25 67 5 28 Simple average 71 29 20 80 51 29 9 20 64 12 25 Weighted average 69 26 17 83 52 31 8 22 63 10 27 56 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS Table B2b. Dakar piped water Percent attributable to demand- Pure demand-side gap attributable to supply- Pure supply-side gap Unserved population supply and demand- Proportion of deficit Proportion of deficit Proportion of deficit Mixed demand and Take-up rate given attributable to both side factors only side factors only Supply-side gap supply-side gap Coverage rate Access rate side factors Division Nolivet/Barak 100 access 100 100 0 0 0 0 0 -- -- -- Arafat 100 93 93 7 7 0 0 0 100 0 0 Hann 100 91 91 9 9 0 0 0 100 0 0 Dalifort 100 59 59 41 41 0 0 0 100 0 0 Santiaba 100 99 99 1 1 0 0 0 100 0 0 Taïf 100 93 93 7 7 0 0 0 100 0 0 Guinaw 100 93 93 7 7 0 0 0 100 0 0 Médina 100 95 95 5 5 0 0 0 100 0 0 Wakhinane 100 64 64 36 36 0 0 0 100 0 0 Diokoul 100 97 97 3 3 0 0 0 100 0 0 Simple average 100 88 88 12 12 0 0 0 100 0 0 Weighted average 99 89 89 10 10 0 0 0 97 0 0 Table B2c. Johannesburg piped water Percent attributable to demand- Pure demand-side gap attributable to supply- Pure supply-side gap Unserved population supply and demand- Proportion of deficit Proportion of deficit Proportion of deficit Mixed demand and Take-up rate given attributable to both side factors only side factors only Supply-side gap supply-side gap Coverage rate Access rate side factors access Division Alexandra 100 93 93 7 7 0 0 0 100 0 0 Central Region 100 94 94 6 6 0 0 0 100 0 0 Diepkloof / Meadowlands 81 65 53 47 28 19 12 7 60 26 14 Diepsloot 75 84 63 37 12 25 21 4 32 57 11 Doornkop / Soweto 91 74 67 33 24 9 7 2 73 20 7 Ennerdale / Orange Farm 90 74 67 33 23 10 7 3 70 22 8 Johannesburg South 75 92 69 31 6 25 23 2 19 74 6 Midrand / Ivory Park 100 94 94 6 6 0 0 0 100 0 0 Northcliff 100 68 68 32 32 0 0 0 100 0 0 Roodepoort 73 88 64 36 9 27 24 3 25 66 9 Sandton / Rosebank 100 100 100 0 0 0 0 0 -- -- -- Simple average 90 84 76 24 14 10 9 2 68 27 6 Weighted average 91 82 75 25 16 9 7 2 72 19 6 57 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS About AICD This study is a product of the Africa Infrastructure Country Diagnostic (AICD), a project designed to expand the world's knowledge of physical infrastructure in Africa. AICD will provide a baseline against which future improvements in infrastructure services can be measured, making it possible to monitor the results achieved from donor support. It should also provide a better empirical foundation for prioritizing investments and designing policy reforms in Africa's infrastructure sectors. AICD is based on an unprecedented effort to collect detailed economic and technical data on African infrastructure. The project has produced a series of reports (such as this one) on public expenditure, spending needs, and sector performance in each of the main infrastructure sectors--energy, information and communication technologies, irrigation, transport, and water and sanitation. Africa's Infrastructure--A Time for Transformation, published by the World Bank in November 2009, synthesizes the most significant findings of those reports. AICD was commissioned by the Infrastructure Consortium for Africa after the 2005 G-8 summit at Gleneagles, which recognized the importance of scaling up donor finance for infrastructure in support of Africa's development. The first phase of AICD focused on 24 countries that together account for 85 percent of the gross domestic product, population, and infrastructure aid flows of Sub- Saharan Africa. The countries are: Benin, Burkina Faso, Cape Verde, Cameroon, Chad, Côte d'Ivoire, the Democratic Republic of Congo, Ethiopia, Ghana, Kenya, Lesotho, Madagascar, Malawi, Mozambique, Namibia, Niger, Nigeria, Rwanda, Senegal, South Africa, Sudan, Tanzania, Uganda, and Zambia. Under a second phase of the project, coverage is expanding to include as many other African countries as possible. Consistent with the genesis of the project, the main focus is on the 48 countries south of the Sahara that face the most severe infrastructure challenges. Some components of the study also cover North African countries so as to provide a broader point of reference. 58 POVERTY AND INFRASTRUCTURE ACCESS IN SLUMS The World Bank is implementing AICD with the guidance of a steering committee that represents the African Union, the New Partnership for Africa's Development (NEPAD), Africa's regional economic communities, the African Development Bank, the Development Bank of Southern Africa, and major infrastructure donors. Financing for AICD is provided by a multidonor trust fund to which the main contributors are the U.K.'s Department for International Development, the Public Private Infrastructure Advisory Facility, Agence Française de Développement, the European Commission, and Germany's KfW Entwicklungsbank. The Sub-Saharan Africa Transport Policy Program and the Water and Sanitation Program provided technical support on data collection and analysis pertaining to their respective sectors. A group of distinguished peer reviewers from policy-making and academic circles in Africa and beyond reviewed all of the major outputs of the study to ensure the technical quality of the work. The data underlying AICD's reports, as well as the reports themselves, are available to the public through an interactive Web site, www.infrastructureafrica.org, that allows users to download customized data reports and perform various simulations. Inquiries concerning the availability of data sets should be directed to the editors at the World Bank in Washington, DC. 59