There is no specific database from which to extract the urban development patterns for Uganda and several data sets point to different urbanisation patterns for the country. About 15 percent of Uganda’s population is categorized as urban and is projected to double by 2035 (MLHUD, 2013; UBOS, 2013) . This pattern of growth is associated with an increasing number of informal settlements across the country. Estimates by the MLHUD (2008) and Uganda Human Settlements Network give a total of 2.45 million to 3.2 million slum residents in Uganda.

This paper is based on a survey conducted between December 2013 and March 2014 in slums, referred to in the paper as low income settlements in the urban areas of Kampala, Jinja, Arua and Mbarara (Figure 1). It is such developments in these urban centres that have, in part, contributed to the development of the new urban policy for the country. The slum profile (an inventory of slum characteristics) of each urban area is given in Table 1. It is these slums that have been the target of several funding agencies promoting energy improvements among the urban poor across the country. It is also recognized that these urban centres hold the largest number

of households residing in informal settlements across the country. The results of this paper also fit closely into the UN-Habitat Medium Term Strategic and Institutional Plan (2008-2013), which calls for taking a step further to create the necessary conditions to stabilize the growth of slums and set the agenda for the subsequent reduction in and reversal of the number of slum dwellers.

Data collection was anchored at household level and a questionnaire was the main data collection instrument. Questionnaires were administered to a randomly selected set of 574 households in selected low-income settlements in the four urban areas depending on their willingness to participate in the survey. Specific attention was paid to residential energy utilization, indicators of energy utilization and adaptation to energy insecurity. The estimated slum population in the four selected urban areas is 1,104,322 and given that the average household size in urban areas in Uganda is 3.9 people, according to the Uganda Bureau of Statistics, the assumed number of households in the selected low income settlements was 283,460. It was from this population that the sample size was determined using guidelines given by Ric Coe (1996) and Equation (1) given by Krejcie and Morgan (1970) .

where:

S = Required sample size.

X² = The table value of chi-square for 1 degree of freedom at the desired confidence level (0.10 = 2.71; 0.05 = 3.841; 0.01 = 6.64; and 0.001 = 10.83).

N = The population size.

P = The population proportion (assumed to be 0.50 since this would provide the maximum sample size).

d = The degree of accuracy expressed as a proportion (0.05).

Descriptive statistics were generated to describe respondents’ characteristics and energy utilization patterns in the low income settlements. Ordered logistic regression was applied to explore the determinants of energy utilisation and how these shape adaptation among households. To use logistic regression, the response must only have two outcomes, i.e. yes/no, 0/1. The aim of this method is to fit a regression equation to the data (assuming two explanatory variables) where α is the intercept, β₁ and β₂ are the parameters for the explanatory variables x_i, p is the proportion for the event of interest and In denotes the natural logarithm function. The term on the left hand side of the equation is known as the logit link function of p. According to the Cook, Dixon, Duckworth, Kaiser, Koehler, Meeker, & Stephenson (2000) is of the form:

The explanatory variables can be quantitative (discrete or continuous) or qualitative (factors).The data was further studied with factor analysis in mind to assess relative adaptation of residents to energy insecurity in the low income settlements. The underlying assumption of factor analysis given by Hair, Anderson, Tatham, & Black (1998) & Kaiser (1960) were confirmed before it was utilised. The ultimate aim was to identify the observed underlying structure in the data matrix of variables against which residents adapt to energy insecurity. All the basic assumptions, routines and procedures given by Pallant (2011) were followed in the Statistical Package for Social Scientists (SPSS V.19) to run the logistic regression and factor analysis.

Energy demand patterns of urban households, especially the poor, largely revolve around household energy end-uses such as cooking and lighting as well as energy services for home-based commercial and productive activities. It was found that charcoal (34.9%) was the dominant energy source for cooking followed by kerosene (25.0%) and fuel wood (20.6%) (Figure 2). These results are consistent with MEMD & UBOS (2014); Menya et al. (2013); Ngaira & Omwayi (2012) and WFP et al. (2009) observations that charcoal is the most popular and predominant source of energy used in urban settings; where, for example, between 200,000 and 230,000 tonnes was reported for the Kampala City (Knöpfle, 2004) .

Wood biomass will likely remain the dominant household energy source for cooking and heating for several decades in Uganda due to low accessibility to alternative energy sources (Gore, 2008) . This is an old observation given by Morgan & Moss (1985) who noted that wood and charcoal are often regarded as cheap alternatives to electricity, for domestic cooking and small industrial purposes, even in high temperature processes. Preference for wood and charcoal may be derived from tradition, the taste of food cooked in this way, reliability of supply and ease of transport and storage. Use of wood requires just three stones or extremely cheap locally made stoves. Often wood is much better suited to large-scale cooking, as for large families, organisations, or for special occasions. Also, traditional round-bottomed cooking pots can be used with wood fires, while modern stoves require the additional expense of flat-bottomed, usually metal, saucepans. At a broad level, the results speak to Coelho & Goldemberg (2013) who observed that poor urban households were found to be using different fuels and their energy demand for cooking and heating is usually met from a mix of sources, including modern and relatively clean sources, such as LPG and kerosene to traditional biomass.

By their spatial location in well-defined urban centres in the country, one would have assumed that the electricity infrastructure in the studied areas would be elaborate and available, but the selected settlements have insufficient electricity infrastructure and so residents use alternative sources of energy such as charcoal, firewood, paraffin/candle. This is not any different from the results of the 2002 Population and Housing Census and the UNHS 2005/06 which indicated that 1 out of every 2 households in urban areas in Uganda used paraffin as their main source of energy for lighting. The percentage of households using electricity for cooking in the studied areas is still relatively low at 4.3%. This again is consistent with Knöpfle (2004) observations that charcoal is per energetic content (MJ) on the average even less expensive than firewood, which is often considered as the source of energy for the poor. In the kitchens of Kampala there are still mostly inefficient iron stoves used and even if people can afford a more efficient energy source for cooking―like LPG or electricity―they still use charcoal. The importance of charcoal among households can be gauged by its nickname, “black gold”, as it is referred to by some traders in Kampala City (UNDP, 2013) . Furthermore electricity needs a high initial investment for buying an electric cooker/oven, and has the disadvantage of insecure power supply in many areas. Therefore electricity, in spite of still being the focus of national energy policies in most of the countries, is generally used for lighting, household appliances and home-based income generating activities.

When all the logistic regression routines in SPSS suggested by Cook et al. (2000) were followed, all the predictor variables given in Table 2 were entered in Equation (1) to explain utilization of only three selected energy types: firewood, charcoal and electricity. Of the sixteen factors, only three were statistically significant. Results in Table 3 indicate that there is a significant power for variables household size (p = 0.045), and share of adults in the household (p = 0.003) in explaining fire wood consumption in the low income settlements. The model explained 4.3% (Nagelkerke R²) of the variance in fuelwood utilization and correctly classified 62.3% of the cases. The same process was applied for charcoal utilization and none of the factors entered into the model were significant. For electricity utilization, the model indicates that there is a significant power for variables gender (p = 0.045), and share of adults in the household (p = 0.069). The model explained 24.4% (Nagelkerke R²) of the variance in electricity utilization and correctly classified 77% of the cases. These results corroborate Egeru (2014) who noted that household size increased firewood demand due to the increased demands for cooking in a household with a large number of people.

Figure 3 illustrates the energy problems experienced in the low income settlements. What is most apparent in the figure in that energy insecurity in the low income settlements coalesce around four major components: high retail prices/tariffs for energy, electricity shortages, wood shortages and waste/inefficient use. The majority of Ugandans who depend on charcoal and firewood for cooking have to brace themselves for higher prices of charcoal. It is possible that the price of charcoal is a reflection of excessive demand for charcoal, especially in urban areas, while other alternatives like hydro-electricity power are too costly and unreliable for the average Ugandan to afford. Despite government’s commitment to bolster supply of electricity across the country, the availability of reliable and cheap electricity remains a major challenge for low income settlements. The access to electricity in the selected settlements is constrained by high power tariffs and inadequate transmissions and distribution network. It has been reported elsewhere that Ugandans pay the highest power bills in East Africa. Even at a time when government was offering a substantial amount of money in subsidies to the power sector, a unit

^**Significant at 5% level; ^***Significant at 10% level.

of electricity for domestic consumption was the equivalent of $0.154 per unit in Uganda but $0.098 per unit in Kenya and $0.082 per unit in Tanzania (Wesonga, 2014).

Notwithstanding the country’s commendable progress on reforms and private investment, Uganda has suffered chronic power shortages which have resulted into uneven distributions, chronic rationing, frequent load shedding during the day and night and yet demand continues to grow at an annual rate of about 7 percent. The instability of electricity supply is further exemplified in frequent “load shedding”, with up to 12 hour daily cycles in Kampala (Lee, 2013) . Some of the electricity disruptions may seem simple and localised but taking longer to address them is a big problem. Uganda has a regular supply of electricity of approximately 305 MW, still 75 MW short of the 380 MW peak demand (Mawejje, Munyamnonera, & Bategeka, 2012) . From time to time, customers especially in low income settlements have to bear with fluctuations in electricity supply and power surges from complete blackouts, rolling blackouts, to brownouts (voltage drops), which are strong elements of Uganda’s energy supply. Although a number of dams have come to the grid, with an addition of several megawatts, the ever growing demand for electricity together with inadequacies in the whole electricity transmission and distribution chain mean that the full benefits of these investments will not be fully realised in low income settlements.

Households do not remain passive under conditions of energy insecurity but adapt themselves in various ways, depending on the socio-economic position of individual households. Households undertake a wide range of actions and conscious effort to stay afloat and maintain their livelihoods. To describe the adaptive mechanisms used by households in low income settlements, the data collected was studied with a factor analysis together with a principal Axis Factor (PAF) with a Varimax (orthogonal) rotation of 18 questions on adaptation to energy insecurity was conducted. The ultimate aim was to screen variables, identify underlying aspects common to many variables and cluster them into homogenous sets. An examination of the Kaiser-Meyer Olkin measure of sampling adequacy suggested that the sample was suitable for factor analysis (KMO = 0.947) for the four urban centres. KMO values of 0.947 derived in this study suggest that sufficient correlations among the variables existed to warrant factor analysis.

In this paper, three factors had eigen-values over 1 and were chosen for the final analysis for all the four towns. The cumulative variance explained by the factors was 75% for all the towns. A Principal Components Analysis (PCA) with a Varimax (orthogonal) rotation of 14 initial questions from the questionnaire was used to extract factors. When loadings less than 0.5 were excluded for all towns, the analysis yielded a three-factor solution that household members felt are adaptation measures that they use to adjust to energy insecurity. These would be qualified as three domains; where household members take personal responsibility and self-discipline themselves to adjust. The results of an orthogonal rotation of the solutions are shown in Table 4.

The ability of households to optimize creates significant heterogeneity in type of activities that they engage in. Table 4 shows thirteen items that clustered on factor 1. All the factors are clustered around adjustments in cooking practices and energy substitution. As energy source becomes scarce, unreliable and expensive, households have adapted their food preparatory practices and cooking styles by omitting or substituting some essential energy-demanding foods in order to save cooking energy. There are several foods whose preparation has changed given the availability and cost of energy but the most classic example in the low income settlements are dry beans and bananas. Typical cooking practices today involve pre-soaking hard grains and dry beans in water

several hours before cooking to soften them; thereby requiring less time and energy during preparation. To respond to high fuel prices and scarcities, the practice of simmering matooke and ugali/kawunga (steamed bananas and posho) gently has reduced gradually and there is preference of boiled bananas fingers (katogo) to the traditionally steamed and mashed banana type.

Although charcoal was a constant in most households in the low income settlements it was observed that when charcoal availability reduces and its prices increase, there was a movement down the energy ladder to cheaper energy sources such as firewood and crop residues. Even for those households that depend on electricity for lighting, there is a tendency to switch to cheaper energy once they experience price increases and load shedding. There is a shift from the use of electricity to the use of kerosene lamps and candles (known locally as tadooba), and further down to wick and wax candles during the electricity load shedding phases. Firewood sources are not within easy reach either and especially for low income settlements in Arua town have increased tremendously over the years. This can be compared with MWE (2007) which noted that on average the distance travelled to collect firewood had increased from 0.06 km in 1992 to more than 1 km in 2007. Recent estimates by Egeru et al. (2014), NPA (2010) and Agea et al. (2010) indicated that the distance travelled had increased to 2 ± 7 km and 8 - 12 km and respectively. The increasing distance to traditional sources of energy places extra demands on women and children in the low income settlements. These results corroborate well with GIZ (2011) which found energy substitution behaviour involving the use of kerosene, candles, and solar power among households in Nepal.

Other adaptation measures adopted within this cluster include the widespread adoption of energy conservation measure such as energy saving bulbs, switch off unnecessary unusable lights in rooms and security lights if and where available, abandonment of traditional cooking practices and widespread adoption of preparation of foods that require less energy. Other households have reduced the consumption of energy in a variety of ways including; 1) reducing the number of cooked meals substantially; 2) increasingly paying attention non-home based food consumption and increasing attention being paid to the purchase of cooked food; 3) limiting the frequency of use of household appliances such as flat irons, electric kettles, radios and televisions where these are available; 4) preparing one meal in a day or large quantities to preserve energy for the next day; 5) abandoning boiling water for domestic purposes; and 6) shifting to less energy demanding foods. In extreme cases, especially with high electricity tariffs, residents have gone ahead to illegally connect, tamper with electricity supply metres and steal electricity, a practice that has been reinforced by the high number of unlicensed wiremen with the direct connivance of UMEME personnel (ERA, 2011) . Theft is also significant in slums and areas with informal settlements, whose inhabitants often do not meet the legal requirements to become regular customers of the electricity company (World Bank, 2010) . To delve into the magnitude of this problem including who is involved, what incentives exist for people to obtain illegal connections and the losses associated with illegal connections were not the main concerns of this paper but Mwaura (2012), ERA (2011) and several media reports of deaths from electrocutions indicate that the practice is rampant, especially in low income settlements. In general this is also consistent with Cooke et al. (2008) that households alter their behaviour in the presence of sufficient scarcity in ways that are least costly to them.

Three items clustered onto factor 2; referred to here as self-generation and fuel substitution including the use of generators. It was difficult to ascertain how widespread the practice of self-generation was among households in low income settlements, but generators were apparent among small-scale enterprises in the settlements especially retail shops and entertainment hotspots. Today, probably as many Ugandans are ‘electrified’ through the use of lead-acid batteries, small diesel and petrol generators, and photovoltaic systems as are connected to the national grid. However, current prices of generator sets may be higher that most households would be in a position to afford. The use of solar panels was most prevalent in Arua town and this could be explained by the lack of an electricity grid network in the region. Data from the World Bank (2007b) indicates that running a generator is between 2 and 6 times more expensive than obtaining electricity from the public grid (World Bank, 2007b) and the use of generators in slum settlements should be regarded as a short term measure.

Finally, one item loaded onto factor 3. This is related to households making certain that they adjust their sleeping patterns (habits and lifestyles) in order to cope to several instabilities in electricity supply. Two responses were reported here; while the stress and routine changes disrupted routines and sleep schedules/cycles for some people especially for families with working couples and those with children, for others, power outages provided a much needed push to go to bed just a little bit earlier. Detailed studies are required here but as Schocker (2012) noted; people are hostages to the daylight schedule and when the power goes out, there is a disruption in zeitgebers “time givers,” or those items like light exposure, exercise, bathing, social interaction and eating that the body uses to figure out where it is in a 24-hour period.

The critical household factors in low income settlements including household size, share of adults in the household and gender greatly influence household energy utilization. The energy lifestyle of people in low income settlements is built on scarcity, uncertainty, and inadequacy. There is a need to get energy supply and prices right for urban areas and low income settlements in particular. The continued increases in the costs of energy can potentially erode livelihoods for people in low income settlements. Charcoal use is a constant and dominates the energy source across all households studied and it is important to understand the urban charcoal market in greater detail; while reforming the electricity sector to take care of the urban poor. It is also important that a continuous supply and use of this resource needs to be understood clearly from its health, environmental and broader climate change implications. The continued rural extraction of wood for charcoal is responsible for encroachment into major fragile ecosystems and forested areas across the wider Ugandan landscape especially in the wood and charcoal producing zones.

The diversity of adaptation strategies is indicative of the ability and capacity of household members to cope with adversity. These strategies must be understood in depth in order to build resilient livelihoods in low income settlements. Low income settlements may not usually realise the extent to which they depend on various forms of energy to get through each day but there is a need for a strategy that provides uninterrupted supply of energy, promotes energy conservation, and minimises total energy costs that affect low income settlement across the country.