Spotting the invisible crisis: early warning indicators in urban slums of Nairobi, Kenya

By Lilly Schofield, Shukri F Mohamed, Elizabeth Wambui Kimani-Murage, Frederick Murunga Wekesah, Blessing Mberu and Thaddaeus Egondi, Catherine Kyobutungi and Remare Ettarh

Lilly Schofield has worked in many countries in Africa and Asia as a nutrition consultant. Until June 2012 she was the Evaluation and Research Support Advisor for Concern Worldwide Kenya. She has recently joined Save the Children UK as a Nutrition Advisor.

Shukri F Mohamed works at the African Population and Health Research Centre (APHRC) as a Research Officer. She coordinated, managed and led the activities of the Indicator Development for Surveillance of Urban Emergencies (IDSUE) project, described here. She led the project analysis write-up and APHRC’s input into this article.

Elizabeth Wambui Kimani-Murage is an Associate Research Scientist at APHRC and a Wellcome Trust Fellow. She has about 10 years of experience of research in environmental health and public health issues in sub-Saharan Africa. She was the project manager in the project overseeing the operations of the project, including data collection and management.

Frederick Murunga Wekesah is a Research Officer at the APHRC with research interests in epidemiology, MCH and HIV in the context of urbanisation. Frederick trained the data collection team, supervised data collection and participated in data management, analysis and the actual writing of the report.

Blessing Mberu works on policy relevant research on migration, urbanisation, adolescent reproductive behaviour and poverty in sub-Saharan Africa under the Urbanisation and Wellbeing Research Programme at APHRC.

Thaddaeus Egondi is a data analyst working under Health Challenges and Systems programme at APHRC. He was the data analyst for the IDSUE project responsible for data management and indicator generation.

Catherine Kyobutungi heads the Health Challenges and Systems Research Programme at ADHC Her research interests are in the epidemiology of non-communicable diseases in the African region and in health systems strengthening.

Remare Ettarh was associate research scientist at APHRC between 2009 and 2012.

This project was funded by the US Office of Foreign Disaster Assistance (US-OFDA). The authors also acknowledge funding for the NUHDSS from the Bill and Melinda Gates Foundation. Elizabeth Kimani-Murage is a Wellcome Trust Fellow. The authors extend thanks to the IDSUE field team.

This article shares experiences and findings of operational research in Kenya¹ to identify indicators that can help detect the ‘tipping point’ from chronic need to crisis in vulnerable urban populations.

Despite the recognised risk and growing international commitment to address urban crises, urban environments and actors working in them are still often plagued with a dearth of information. Until recently, urban areas were often excluded from national and sub-national surveys under the assumption they would skew data and obscure negative trends in rural areas. Even when they are included, data are rarely disaggregated between wealthier urban neighbourhoods and slums, leading to a homogenisation that hides the true situation in both areas. Furthermore, high levels of food insecurity and unstable livelihoods that typify urban poverty are often thought of in terms of chronic development challenges. While sustainable improvements in these indicators for many of the world’s urban poor will certainly require long-term investment in development and poverty alleviation by government, civil society, non-governmental organisations (NGOs) and other key stakeholders, there is growing recognition that urban areas also face significant risk for both rapid and slow onset disasters.

The key characteristic of urban vulnerability is its multi-dimensional nature and the fact that both chronic and acute vulnerabilities co-exist and overlap in the same populations, with hazards coming from both the local environment (fire, flood, localised violence, etc.) and global forces (price shocks, for example). These vulnerabilities also interact to exacerbate the effects of one another. Because of this, even moderate shocks can precipitate a crisis among many of the urban poor. Understanding the tipping points at which vulnerability transforms into acute need is therefore critical.

Accumulated rubbish in Korogocho slums, Nairobi, Kenya

The urban poor are exposed to multiple interconnected vulnerabilities and risks from the environment in which they live and the livelihoods which they pursue. Most slum dwellers lack secure tenure to their land and housing, making them vulnerable to sudden eviction and destruction or loss of property. The overcrowding and non-durable nature of most slum housing coupled with inadequate sanitation and water services means that disease transmission is high. Health service provision is limited also affecting health status of the population.

Urban centres are characterised by a highly monetised economy meaning that even the extreme poor must access all or almost all of their basic needs through the market; this means access to an income is essential for household well-being. Wage labour is a centrepiece of urban economies as the informal economy is a key employer of the urban poor. The urban poor can also face higher per unit costs for key goods and services and limited access to these services compared to their wealthier urban counterparts because of the inability to buy in bulk/larger quantities. Slum dwellers in Nairobi pay 11 times² as much for water as residents in better off neighbourhoods because, while non-slum dwellers pay monthly bills to the City Council, slum residents obtain water in piecemeal from exploitative vendors within the community; Dhaka slum dwellers pay 25 times as much³. Many urban poor rely on low quality, casual, insecure and low paying jobs, some of which are also subject to seasonal slow-downs. A review of employment data from five countries found that employment rates were not largely different between urban poor and non-poor but uncertainty of employment was higher among the poor. The uncertainty of livelihoods coupled with need for cash to meet basic needs also translated into greater adoption of dangerous and undesirable livelihoods such as commercial sex work, crime, scavenging and working on garbage dumps, and child labour. These livelihood strategies expose those who practice them to greater risk of ill health and physical harm, reinforcing and deepening the poor health of the community.

On top of these mutually reinforcing poverty domains come additional shocks or stressors. These can be both external, such as a broad economic downturn, conflict, drought, floods or extreme weather in food producing regions, or increases in global food prices. They can also be more localised, such as escalating gang violence, localised water shortages, or an infectious disease outbreak. These shocks interact with the underlying vulnerabilities limiting household’s ability to meet their basic needs. When the combination of shocks and underlying vulnerabilities overwhelm the coping capacity of the community, the tipping point is reached and excess morbidity and mortality result.

Prediction of this tipping point is the key focus of a project on Indicators of Urban Emergency, nested on the Nairobi Urban Health and Demographic Surveillance System (NUHDSS). Specifically, the project has focused on identifying key indicators that can be routinely monitored to detect when an urban poor community is approaching the tipping point. This article describes the preliminary steps employed to identify and define these indicators and lessons learned. To inform indicator development, the project involved a number of studies and analyses that are outlined below.

Life in Korogocho slums,
Nairobi, Kenya

Study methods

Phase 1

The first phase of the study involved a qualitative study and retrospective data analysis.

Qualitative study

A qualitative study was conducted in two slum settlements in Nairobi Kenya, Korogocho and Viwandani, where APHRC runs a longitudinal health and demographic surveillance system, the NUHDSS, with about 71,000 individuals living in about 28,500 households in these two areas. Both slums are located on the outskirts of Nairobi city about 10 km from the city centre. The two communities are informal settlements characterised by poor housing, lack of clean water, poor sanitation, unemployment, poverty and overcrowding. Viwandani slum is located very close to the city’s industrial area and it is home to many low income young people who are predominantly male and have migrated from rural areas to work in the industries. As a result, a high proportion of married men in Viwandani do not live with their spouses who have been left behind to farm and take care of the children. Korogocho, by contrast, is a more established slum settlement with a high proportion of married men living with their spouses and children.

The target study population was residents of the two slum communities. The key eligibility criterion for participating was based on residency as defined by the NUHDSS. A resident must be registered in the NUHDSS with a minimum stay of four months. The NUHDSS captures particulars of location of household and its members (composition). Participation in the project was voluntary and all participants who consented to participate confirmed this by signing a pre-written consent form.

The main aim of this study was to capture the perspective of the urban poor and their own understanding of humanitarian crisis. Both focus group discussions (FGDs) and key informant interviews (KIIs) were conducted. The focus groups were conducted among four distinct groups: younger men (15-24 years), older men (25+ years), female household heads, and married women and unmarried girls (15+ years) with participants selected using purposive sampling. There were 10 FGDs with each group consisting of 8-10 people and 12 KIIs. The KIIs were conducted among the community leaders including teachers, religious leaders, community based organisation (CBO) leaders, women group leaders, youth group representatives, village leaders, administrative leaders and health professionals. All data collected were recorded, transcribed, translated, coded and analysed in MAX QDA using the constant comparison method.

Retrospective analysis

At the same time as the qualitative study, the study team conducted a retrospective analysis of data from the NUHDSS and nested studies collected from the two slum communities over a four year period (2007 to 2010). Though not collected for the purpose of exploring urban humanitarian crisis, the data from NUHDSS and nested studies includes vital information on household size and dynamics, health, nutrition, crime and household assets. Additionally, the period of time studied (early 2007 to late 2010) encompassed a period of political upheaval and price increases that severely affected Kenya’s urban slum dwellers. Following the disputed 2007 elections, violence erupted across many areas of the country and the slums became the main areas of unrest in urban centres. This period of upheaval was followed closely by global food price increases that saw the price of staple food rise, severely affecting poor urban dwellers ability to meet their basic needs. The aim of this analysis was to explore trends in the demographic data to identify how key demographic and health variables reflected the changing situation on the ground and if any of these variables were early warning signs for the substantial increases in negative coping and food insecurity that were seen. Bivariate, graphical and regression analysis was used to compare the level of variables between the different time periods.

Phase 2

Phase 1 generated an initial list of indicators, identified by combining and comparing results from the qualitative and retrospective analysis and informed by a review of relevant literature. The second phase of the study involved prospective monitoring of these indicators on a quarterly basis, beginning initially in the same two slums and later expanding in a phased manner across multiple informal settlements of Nairobi and Kisumu. The indicator tool was continually refined and revised throughout this phase in an iterative manner as new data and learning was added from each round and study site. In the initial three rounds, 500 households were randomly sampled from each study site. In round four, the sample size was reduced to 400 per site after post- hoc power calculation on previous rounds showed that this sample size had adequate power.

The full results of these first phases are available elsewhere⁴. Below is a summary of some of the key findings and implications of these finding for future work on urban emergency monitoring.

Results

Qualitative study

Respondents were first asked to describe the normal or every-day conditions of their lives and then to think of crises times and describe what made those periods crises and how they coped with such times.

Normal situation

The ‘normal’ non-crisis situation described by respondents was still one characterised by extreme poverty, food insecurity, high level of insecurity and tenuous livelihoods. Jobs were hard to come by, mostly casual and irregular. To get and/or maintain a job, men often bribe while women told of how they have to give their bodies to men.

"If your husband works at the construction sites, when it rains nobody gets out his cement and so at that time you don’t get anything…" (FGD, Married women, Korogocho)

"If you don’t know anyone, stay in the house…..The people who suffer a lot are young women because for a young woman, you have to undress to get a job…So the job goes to the boys who have the money because they will give it (bribe)" (FGD, Unmarried girls, Viwandani)  

Even in non-crisis times, many households faced conditions of food insecurity and often reduced meal size and frequency to get by (Table 1).

…And I can tell you there are many people here who sleep on porridge only. You find that they drank porridge in the morning, never had anything at lunch time and then in the evening they make the same porridge. (FGD, Older Men, Korogocho) 

Crisis times

Given that slum residents typically live ‘on the edge’, a small disturbance of the normal situation tends to lead to an acute crisis. When asked to describe times of crisis, respondents identified a wide variety of crisis types. The most commonly mentioned was the post-election violence that racked Nairobi’s slums in 2008. However other crisis mentioned were economic (food price crisis in 2008/2009), social (increased drunkenness and drug abuse in youth), health related (recurrent cholera outbreaks) and security (inter-ethnic fighting in 1997 and 1999). Increased insecurity was mentioned as both a warning sign of crisis and as a key defining feature of crisis periods.

Food security, both in terms of availability and access, was negatively impacted in crisis periods. A security crisis caused many traders and shop keepers to pull out of the slums or close their shops temporarily. Simultaneously, household purchasing power was eroded due to worsening employment conditions and increasing prices

Coping strategies

Table 1 summarises the coping strategies employed by the urban poor in both normal and crisis situations. It shows that there was not a distinct set of ‘crisis coping strategies’ but rather the same coping mechanisms were used in crisis and non-crisis times, though they may be used more often or to a more marked degree during a crisis. Food related coping strategies (reducing meals size, frequency, switching to cheaper foods, scavenging food from dumpsites) were common and used in both crisis and non-crisis times.

Retrospective data analysis

The final composite dataset for the retrospective data analysis contained 18,371 records from 4,286 households over a four year period from Jan 2007 to October 2010. Three time periods were defined based on when post-election violence occurred, and when maize prices (the main stable food) peaked and began to fall.

The pre-emergency period was defined as January 2007-December 2007.

The emergency period ran from January 2008-June 2009. This covered both the post-election violence period and the period when food price inflation was rampant, with the maize price peak occurring in May 2009.

Post emergency was defined as July 2009-October 2010.

A number of key variables were selected to assess the effects of the post-election violence and subsequent food price increases on urban slum dwellers. The variables were grouped into four main domains, including food security, personal security, water and sanitation and health outcomes⁵. Household food security was measured using a food security composite index⁶ and Table 2 presents these results. During the emergency and post-emergency periods, a significantly higher proportion of households reported that they would use extra money if this were available to make dietary changes to acquire increases in quantity or mineral/vitamin content of foods. Using the food security index, more households were classified in the poorest category during the emergency (34.2%) and post-emergency (35.1%) phases than pre-emergency (31.6%), though the increase was not significant. This trend may reflect a long term poverty impact of the emergency, where households employed negative coping strategies to protect consumption in the emergency period that compromised long term well-being and thus were not able to recover once prices reduced. Data on coping strategies are not collected in these datasets so it was not possible to test this hypothesis.

HH: Household

The household income and expenditure patterns, wealth status, general living conditions and dwelling structure features were also examined. As highlighted in Table 3, in the emergency period, slum households ranked themselves lower on a scale of 1 (very poor) to 10 (very rich) compared to pre-emergency and post-emergency phases. Both household average monthly incomes and expenditures significantly increased in the emergency and post-emergency phases compared to pre-emergency. This may be at least partly explained by increased remittances and donations in time of need/crisis. Absolute expenditures on different goods varied across the three time-periods depending on the type of good, for example, energy expenditures were highest during the emergency period, while water expenditures were highest in the post emergency period and health expenditures highest in the pre-emergency period.

The results on households general living conditions and dwelling structure characteristics (as measured by the dwelling index), showed that a similar proportion of households (43%) were in the poorest category during the pre-emergency and emergency periods, however, the proportion increased slightly (47%) in the post-emergency phase.

Prospective indicator monitoring

Results from the retrospective and qualitative investigations were combined with information from the existing literature to develop an initial list of indicators to test prospectively.

The first four rounds of data collection covering one year from April 2011 to April 2012 were collected at 3-month intervals (Table 4). The aggregate figures showed stable levels for most key indicators and largely agreed with the qualitative results depicting a chronically poor but stable state of food and livelihood insecurity. There was heavy reliance on negative coping strategies with over 70% of households in all rounds using at least one negative coping behaviour in the last month.

These figures masked local level variation however. When each site was analysed separately for changes in food security and employment, a more complex picture emerged. The percentage of severely food insecure in both sites varied from round to round (Figure 1) though the variation occurred in different directions, i.e. rounds with higher food insecurity in Korogocho corresponded with rounds of lower food insecurity in Viwandani and vice versa. The fluctuation may also be reflecting seasonal variations in food security, though why these fluctuations would be opposite in the two study sites remains a question for further study.

* Composite price for standard basket of basic commodities including maize flour, sukuma wiki (kale), water, cooking fat, cooking fuel (paraffin).

A similar pattern holds when employment of main breadwinners is considered (Figure 2). Breadwinners in Viwandani worked more hours on average in round 3 compared to rounds 1 and 2 and a lower percentage was unemployed (1.3% in round 3 compared with 2.5% and 3.9% in Rounds 1 and 2). Korogocho breadwinners faced differing conditions; they worked fewer hours and a higher percentage had no work at all in round 3 compared with the preceding two rounds. Round 4 saw a fall in unemployment in Korogocho and a slight increase in Viwandani.

Given the reliance of slum dwellers in Korogocho and Viwandani on market purchased food, the fluctuations in breadwinner employment may be the underlying cause, at least partially, of the fluctuations in food security. A further analysis of what distinguishes the employment patterns in these two slums is required to understand why employment opportunities should improve in one when they worsen in another. One hypothesis is that while breadwinners in both areas are predominately engaged in casual labour (44% and 54% in Viwandani and Korogocho respectively), Viwandani’s location adjacent to the industrial quarter means that more of these casual labourers are working in the formal sector as daily labourers at factories and construction sites. However, Korogocho residents’ casual labour is more often tied to the informal waste management industries that derive from the main city dump to which it is adjacent.

Lessons learned

The preliminary findings of this project highlight several key lessons learned and some of the unique challenges of monitoring and identifying crisis in urban poor environments.

There are high levels of chronic vulnerability and poverty in these populations and therefore a thin line between normal and crises situations. Some experts insist that lives in such slum settings are in perpetual emergency. Slum dwellers often described the same types of coping strategies employed to meet everyday challenges and shortfalls as those used during crises. Even in periods of relative stability, a large proportion of slum dwellers suffer from food insecurity and utilise negative coping strategies such as removing a child from school, reducing food intake, stealing, sending family members away or begging to make ends meet.

The change in food insecurity indicators over time in the two study sites showed significant conflicting patterns with one area improving at the same time the other worsened. Local conditions can vary significantly between slums in the same urban area and worsening conditions in one slum can accompany improvements in another slum. In this study, this may have been partly explained by the different locations and associated employment opportunities. This raises the important question of what level of disaggregation is necessary and appropriate when monitoring urban poor populations. The quantitative data presented here highlight the risk of aggregating different slums together and masking changes at a slum level. Further investigation is needed to understand what is driving the changes in food and livelihood security in these slums and why they are responding differently to the same city level conditions.

A key question that arises is whether we can define common characteristics of urban poor communities that dictate how they will be affected by municipal level changes in prices, food availability etc. Can we identify zones within larger urban areas that while not geographically continuous, behave in the same way to external and internal stresses or shocks? Or is each slum community different? The implications of this issue are important for any early warning work in urban areas because it determines at what level indicators must be monitored to detect crisis and what information can be extrapolated from one slum community to another.

Way forward

The on-going work on urban early warning indicators in Kenya is resulting in expansion to more sites both within Nairobi and other urban centres in Kenya. This will allow for testing of indicators and validation across a wider range of contexts. It is also facilitating further exploration of the spatial heterogeneity of urban slums to address questions around what factors drive the inter-slum differences as discussed above.

Further prospective monitoring of the draft indicators and revision is also required to distinguish those that are sensitive to change early in a developing crisis.

Feedback from key stakeholders after the first year results led to expansion of the data collection to include detailed income and expenditure information. The aim of this information is to develop a ‘survival basket’ analysis to define a minimum level of consumption needed to meet basic needs and subsequently validate the draft indicators against this threshold to identify those that show strong agreement.

Further work will involve development of a refined indicator list for surveillance of urban emergencies and development of thresholds/cut-offs/decision algorithms for indicators to define an urban emergency.

In sum, through the development and preliminary testing of an indicator framework, several key lessons have been learned and baseline values established for the indicators. Further development and use of this tool will allow for better understanding of the spatial variations within urban poor communities in Nairobi and other urban areas in Kenya.

For more information, contact Elizabeth Kimani, email: ekimani@aphrc.org, and Lilly Schofield, email: lilly.schofield@gmail.com

¹Indicator Development for Surveillance of Urban Emergencies Research Report for Year 1. Concern Worldwide and APHRC.

²Beall, J, and Fox, S (2006). Urban Poverty and Development in the 21st Century: Towards an Inclusive and Sustainable World, 2006, Oxfam GB, Development Studies Institute, London School of Economics

³Beall et al (2006). See footnote 2

⁴See footnote 1

⁵There were no specific livelihood measures included in the DSS data hence no livelihood domain included in analysis.

⁶The components that were used for computing the food security index included the following questions: how often do you purchase the following staple foods (maize meal, githeri and kales); how many meals were served to the household members during the last two days; during the last seven days, for how many days were the following foods served in a main meal eaten by the household (chapati, meat and bread); whether the household had enough money to buy food, whether the children slept without eating in the past 30 days, and whether the adults slept without eating in the past 30 days. Two questions on sanitation were also included.