Drivers of Fuel Choice for Cooking among Uganda’s Households

This paper examines the factors that Uganda’s households consider when making fuel choices for cooking and investigates the key drivers of fuel choice. The study adopts a quantitative cross sectional research design. The dependent variable of the empirical model is a qualitative response variable which defines three mutually exclusive and highly differentiated discrete choices for cooking fuels, namely: the traditional fuel (firewood), the transitional fuel (charcoal), and the modern (LPG & electricity). Results from the study show that the most important drivers of fuel choice for cooking in Uganda’s households are: household income, age of household head, gender of household head, marital status, education levels of household head and location of a household. Results also reveal high dependency of firewood as cooking energy source among households in Uganda.


Introduction
Over 40% of the global population rely on biomass fuels for their cooking, heating and lighting needs [1]. The use of biomass fuels (mainly firewood and charcoal) combined with inefficient cook stoves is responsible for over 4 million deaths annually due to indoor air pollution [2]. This is even more significant in Sub Saharan Africa where over 720 million people are reported to have died due to indoor air pollution in 2016, the majority of which were women and children [3]. Other consequences of continued use of biomass fuels include deforestation, the withdrawal of children from school, women drudgery, soil erosion and im-The theory postulates that the consumer's or household's choice of goods is based on the characteristics or attributes they possess and are the basis of consumer preference or utility. These attributes are assumed to be the same across all households, however, the utility derived by the household is subjective and depends on the households' preferences. The information they possess regarding the attributes of the fuels combined with the economic and non-economic factors influences their decision making. The economic factors include availability of the fuel source, price of fuel, household income and expenditure, while the non-economic factors include socio-economic characteristics such as household size, age, gender, house ownership, type of dwelling, location of residence, distance to fuel source, and access to electricity [20]. In the spirit of [21], this study assumes that a household faces a choice among different fuel types, and chooses the option that maximizes its utility.
Assuming that choice set C includes all the possible fuel types available to a household, and assume that the utility derived depends on the choice made from this set, a household will select that choice that maximizes its utility, that is; where for any household i, a given level of utility will be associated with any alternative fuel choice j. The utility derived from any alternative fuel type depends on the attributes (Z) of the fuel type and the economic and non-economic factors affecting households' decision. The household decision to choose a particular choice (j) will depend on the fact that the utility got from this set is higher than the utility associated with another choice of fuel. The probability ij P that alternative j is chosen by household i is given by; ( ) Consequently if the i th household chooses fuel type j, then ij U is the highest utility obtainable from among the j possible alternatives.

Theoretical Literature
The theory of fuel choice is often based on the energy ladder hypothesis. Originally proposed by [22], the energy ladder hypothesis presents the view that households discontinue the use of biomass fuels as their income increases [1]. The energy ladder hypothesis depicts a linear three-stage switching process. The first stage involves a heavy reliance on spans of old biomass fuels, such as firewood, cow dung and straws, etc. while in the second stage household moves to "transition" fuels involving the use of kerosene, coal and charcoal, and in the third stage, they switch to the use of LPG, natural gas or electricity [23]. ( Figure  1) To some scholars (for instance, ( [24] [25] [26], the energy ladder hypothesis overlooks the significance of other factors when making energy choices and has been criticised on these grounds. Instead, the critics of the energy ladder argue Open Journal of Energy Efficiency Figure 1. Energy ladder. Source: [20]. that households would use multiple fuels knows as fuel stacking as their income increases instead of moving up the "energy ladder", which means that with an increase in income, traditional fuels are not completely replaced, but are rather used in combination with modern clean fuels.

Energy Stacking
The energy stacking model hypothesizes that households use a range of energy sources regardless of the income levels [27]. The difference among energy portfolios is reflected in the variety of energy sources and their corresponding proportions to total energy. As a result, climbing the energy ladder does not mean abandoning any energy completely [28]. The energy stacking hypothesis is an alternative hypothesis to energy ladder hypothesis for the better understanding of energy transition of households. The hypothesis asserts that rise in household income level will result to the shift to the use of modern energy sources such as electricity, which will take place in term of simultaneous use of both traditional and modern energy sources. For instance, poor households tend to use traditional energy such as biomass, thus an increase in income will necessitate the use of modern energy such as electricity, although the use of traditional energy will not be discontinued for some activities [29]. In Uganda, a large proportion of middle-income households who could in principle afford clean and convenient forms of fuels continue to rely fully or partly on traditional biomass fuels [7]. Several factors such as age, family size, level of education of household head, location, and marital status are important factors that determine household cooking fuel choice [29]. Therefore, income, although very important, is not the only determinant of household cooking fuel choice. Household fuel choice should be explained as a portfolio choice rather than as a ladder process [29].
Therefore, understanding households' fuel choice is considered under the general framework of new consumer theory [19] [30], which suggests that consumers derive utility not from a commodity but from the attributes embedded in a commodity. Information at households' disposal about the various fuels influences their decisions which are driven by households' economic and non-economic constraints. The economic factors may include availability and market price of fuel, household income and expenditure, while the non-economic factors may include socio-economic characteristics such as household size, age, gender, house ownership, type of dwelling, location of residence, distance to fuel source, and access to electricity [29].

Empirical Literature
Many studies have found that household fuel choice depends on numerous factors. [31] [32] used Multinomial Logit model to analyze household energy choices and discovered that household income, age of the household head, level of education of the head of the household, household size, the dwelling ownership, occupation of the household head, number of rooms, number of years the house was built, size of the resident, ratio of female in the household, had a positive relationship with the household fuel choice. On the other hand, other studies for instance [33]) and [34] found the same variables to have negative relationship with fire wood use, thereby encouraging the adoption or use of electricity and or gas. These inconclusive findings could be as a result of the studies having been carried out in different environment and using different data. This signifies that energy choice of households varies from one region to another and not all factors are equally important in determining energy choice in different areas and regions.
Furthermore, some studies, for instance [35] [36] [37] employed ordered logit/Probit models to examine the factors that influence household energy choice. variables such as; income, firewood price, education level of household head, share of dwelling with other people, urban household, access to Liquefied Petroleum Gas(LPG) were found to have a positive relationship with the probability of adopting cleaner energy. While other variables, such as; electricity price, price of kerosene, age of the household head, household size, gender (male) of the household head, and access to fire wood, have negative effect on the probability of the use of clean and efficient fuels. The major limitation of these studies is that they assume that the various household energy choice categories are in an ordered ranking manner which is not realistic.
Literature on household fuel choice conducted in both developed and developing countries show that there are limitations in a number of related studies, including the scope covered, model used, the variables included in the study among others. For instance, some studies on household energy choice (for instance [38] [39] [40]) focus only on electricity aspect of household energy consumption, thereby neglecting other aspects like consumption of fuel wood and other solid fuels, households' consumption of kerosene, fuel for transportation, as well as liquefied natural gas as a source of household energy consumption. In addition, some studies such as [41] [42] focus only on fuel wood analysis as a source of household energy neglecting other aspects like; kerosene, electricity gas and transportation. Thus from the literature reviewed, it is shown that not all factors have equal importance in determining the pattern and behavior of household energy consumption and that energy choices differ in different areas due to differences in socio-economic settings, environmental factors, and cultural factors as well as the average level of development in the area. Moreover, many authors have considered the use of a single energy source in their studies. This study fills the gap by considering multiple energy choices for household cooking.

Methodology
The study adopts a quantitative cross sectional research design. The paper uses secondary data, which have been obtained from [5]. Data Analysis was done using STATA statistical package, version 15.0. A multinomial logit model was used to estimate the underlying empirical model. We found this analytical model appropriate because it describes the behavior of consumers when they are faced with a variety goods with a common consumption objective. In addition, we base the choice of the model on its ability to perform better with discrete choice studies as confirmed by [43]. However, its goodness rests on the premise that the phenomena being studied must be highly differentiated by their individual attributes. This condition is fulfilled in our study since our empirical model examines choice between a set of mutually exclusive and highly differentiated fuels such as firewood, charcoal, kerosene, gas, and electricity.

Theoretical Model
The probability that a household chooses one type of fuel is restricted to lie between zero and one. The model assumes no reallocation in the alternative set and without changes in fuel prices or fuel attributes. The model also assumes that households make fuel choices that maximize their utility [43]. The model can be expressed as follows: where;

[ ]
Pr Y j = is the probability of choosing either firewood, kerosene, gas or electricity with firewood as the reference household fuel category; J is the number of fuels in the choice set; j = 0 is firewood; X i is a vector of the predictor (exogenous) factors/variables and β j is a vector of model parameters.
Re-writing Equation (1) gives; Re-arrangement Equation (4) using the odds ratio gives the following empirical model: In Equation (5), the quantity 1 is the odds ratio. The logit (log odds) is a linear function of the independent factors X i 's. Equation (3) allows for the interpretation of the logit elasticities for variables in the same way as in linear regressions. This equation expresses the odds ratio of selecting a fuel type with respect to the reference category. Differentiating Equation (3) we obtain the marginal effects (Greene, 2003).
The marginal effects measure the expected change in the probability of choosing one fuel alternative with respect to a unit change in an explanatory variable. For instance, the expected change in probability of choosing a particular fuel type with respect to a one-year change in age of household head. The model follows from the assumption that the random disturbance terms are independently and identically distributed [43]. In addition, [44] show that even if the number of alternatives is increased (from 2 to 3 to 4) the odds of choosing an alternative fuel remain unaffected. That is, the probability of choosing a particular fuel type is independent of presence of other fuel choices. A positive marginal effect implies an increase in the likelihood that a household will choose the alternative fuel. A negative marginal effect indicates that there is less likelihood that a household will change to alternative fuel.

Specification of the Empirical Model
Following Equation (3), the empirical model to be estimated is specified as: where: β's are coefficients of the equation, ε is the residual term INC = Household income levels; SEX = Sex of household head; EDUC = Level of education of household head; REG = Region where the household is located and Loc = Location of the household.

Variables Used in the Study and Their Measurement
Cooking fuel choice is the dependent variable. Various socio-economic variables such as income levels of the household, age of household head, education level of the household head, gender of household head, employment status of a household head, marital status of a household head as well as location of the household are investigated as the underlying predictor variables. The study variables, their measurement scales and units of measurement are shown in the following Table 1. Table 2 show that of the 15,654 total households, the majority 11,191 Open Journal of Energy Efficiency   The descriptive statistics in Table 4 further show that whereas urban households dominate the use of modern fuels as compared to their counterparts in rural areas, the rural households dominate the use of traditional fuels.     Table 5  with married, divorced and widowed household heads respectively. On the other hand, the relative log odds of transitional fuels for cooking compared to traditional fuels reduce by 0.822 and 0.877 with married and widowed household heads respectively. These results suggest that married, divorced and widowed household heads are more likely to reduce the use of modern fuels for cooking whereas married and widowed household heads are more likely to reduce the use of transitional fuels for cooking. Estimates in Table 5 show that the relative log odds of choosing modern fuel for cooking rather than traditional fuel will increase by 1.699, 1.346 and 3.000 with household heads who had attained higher secondary school, diploma and degree levels of education respectively. Likewise, the results show that the relative log odds of choosing transitional fuel for cooking rather than traditional fuel will increase by 0.

Marginal Effects after the Multinomial Logistic Regression
The marginal effects show the percentage change in the odds ratio attributable to a unit change in one of the variables as shown in Table 6. From Table 6, the marginal effects estimates show that an increase in the household's income by one percent reduces the probability of using traditional fuels by 5.4% while the probability of using transitional and modern fuel increases by 5.1% and 0.3% respectively. The results confirm the energy ladder hypothesis [22], which postulates that as household's income increases, households will move up the energy ladder shifting from the use of traditional fuel to modern fuels. However, the change is marginal which confirm the International Energy Agency's (IAE'S) standpoint that income growth does not automatically lead to a greater uptake of clean fuels for cooking [10]. Age of household head reduces the probability of using transitional fuels and modern fuels for cooking by 0.3% and 0.0% respectively, but it (age) increases the probability of using traditional fuels for cooking by 0.4%. We find this result making sense because as the household heads grow older, they tend to get inclined to traditional fuels as a matter of habit as compared to younger household heads. In many developing countries like Uganda, where savings are inadequate and investment in capital is relatively low, the older you get, the less income you have. As a result, older household heads tend to stick to traditional fuels which are cheap and readily available. Our results are supported by findings other scholars in related studies for instance, [30] and [32]. The probability of using modern fuels and traditional fuels for cooking reduces by 1.1% and 15.7% respectively if the household is male headed; but the probability of using transitional fuels for cooking increases by 16.9% if the if the household is male headed. These results suggest that male headed households have preference of traditional fuels as compared to transitional and modern fuels. We find this result surprising given the cultural values the males tend to attach to their food preparation methods most especially in developing countries like Uganda. Moving from a household head in paid employment to a household head in self-employment, the probability of using transitional fuels reduces by 2%. The probability of using transitional fuel in household headed by subsistence worker reduces by 21% and increases the probability of using traditional fuels by 20%. This is supported by the findings of [45] who found that occupation is much more important than income in the cooking fuel transition. Residing in urban areas increases the probability of using modern fuels by 1%, increases the probability of using transitional fuels by 20% and reduces the probability of using traditional fuels by 21%. The estimated coefficient for household head's education is positive and statistically significant at 5% for the probability of a household choice of transitional fuels as their main cooking fuels and statistically significant at 1% for the use of traditional fuel. A household head who has completed some primary has a probability of 0.6% of using transitional fuel and a probability of 0.2% of not using traditional fuel. A household head who has completed primary education has a probability of 0.7% of using transitional fuel and a probability of 0.5% of not using traditional fuel. A household head who has completed some secondary, lower secondary and higher secondary have a probability of 13%, 15% and 17% of using transitional fuel respectively and a probability of 10%, 12% and 16% of not using traditional fuel. A household head who has completed diploma education has a probability of 16% of using transitional fuel and a probability of 15% of not using traditional fuel. Household heads with a degree have a probability of 0.2% of using modern fuel and a probability of 21% of not using traditional fuel. This implies that an increase in household head's education is likely to influence the choice of modern and transitional fuels over traditional fuel. A possible explanation is that, increased level of education improves households' income, taste, knowledge of fuel attributes and preference modern clean fuels. A highly educated household head is likely to lack time to collect fuelwood because of the opportunity cost of the time and would rather purchase fuelwood alternatives which are cleaner but expensive ceteris paribus. This finding is consistent with the finds of [29].

Conclusion
The study presents the set of key factors that determine the choice of cooking fuels in households of Uganda. The study uses cross sectional data from [5]. The MNL analysis estimates show that the log odds of choosing modern as well as transitional fuels over the traditional fuel increase with household incomes, household having higher secondary, Diploma or Degree education levels and a household located in Urban. On the other hand, estimates show that the log odds of choosing modern as well as transitional fuels over the traditional fuel decrease with age of household head, household head being male headed, household head being in subsistence employment, household head being a married female household head being divorced and household head being widowed.
The study results thus show that the most important drivers of fuel choice for cooking in Uganda's households are: household income, age of household head, gender of household head, marital status of the household head, education level of the household head, employment status of the household head and location of a household. The paper further reveals high dependency of firewood as cooking energy source among households in Uganda.

Policy Implications
From the study, the following implications were drawn; 1) Many households continue to rely on traditional fuels, and this finding implies that there will be a decline in agricultural productivity through deforestation, loss of natural habitat for wildlife and not forgetting the impact on the health of women and children. To mitigate the negative consequences of heavily relying on firewood, policy makers should make modern fuels more accessible and affordable so that firewood and charcoal be used and harvested more sustainably. There is also need to sensitize communities on the role of environment and the danger they pause when they misuse it.
2) Government should provide incentives and enabling environment for private sector participation in supply of modern energy. Improvement of infrastructure in the countryside will encourage private sector investment and hence removal barriers to accessing of cleaner fuels.
3) Government and other sectors stakeholders should also encourage demand for modern energy through promotion of cleaner energy by making them more accessible and affordable.

Study Limitations and Areas for Further Research
This study uses cross sectional data from [5]. This data is a one-survey period data. A similar study can be extended to cover a panel of the UNHS data covering previous periods as well as the most recent period when the data is available. In this case, panel data analytical techniques may be employed, and the analysis involving comparison with the results of the one-survey period studies may reveal whether or not there has been social-economic transformation in the energy choice for cooking among Uganda's households. The quantitative nature of the study is also a limitation. The study may be improved if a similar qualitative study is carried out and field interviews conducted.