Cooking energy choices in urban areas and its implications on poverty reduction

ABSTRACT

Developing nations face higher stakes in the race towards a cleaner energy future, where it's a matter of life, wealth, and basic human needs. Using Tanzania Panel Survey data, this study examined urban households' cooking energy choices and their impact on reducing poverty. Employing advanced statistical models, it found that traditional energy sources still dominate cooking methods, and households relying on them are more likely to be trapped in poverty. Clean energy sources, on the other hand, reduce poverty prevalence. Education and income diversification programs can facilitate an immediate shift towards clean energy and achieve Sustainable Development Goals. This study highlights the urgency of the situation and the need for decisive action towards a cleaner, equitable world for all.

KEYWORDS:

• Cooking energy
• Household poverty
• Choice modelling
• Tanzania
• Sustainable development
• Clean energy

Introduction

Poverty has been a persistent challenge in developing countries, particularly in Sub-Saharan Africa (SSA), where the majority of the world's poor population reside (Tonini et al. 2022). Tanzania is no exception to this problem, with more than 27.4% of the country's population living in extreme poverty conditions (Doggart et al. 2020). This issue has made it difficult for countries to achieve local and international development targets, including the Sustainable Development Goals 2030, such as affordable and clean energy (WBA 2020; Aziz, Barua, and Chowdhury 2022; Halkos and Gkampoura 2021).

Access to energy is a crucial component for development, as it is a determinant of livelihood status at the household level (Karmaker et al. 2022) and has backward effects on the household's economic status (Rahut et al. 2020). However, about 3.0 billion people worldwide have no access to clean cooking energy, while 1.3 billion people lack access to electricity, and 2.7 billion people depend on dirty energy sources such as firewood and biomass for cooking and heating, which have adverse effects on human health and the natural environment (Aziz, Barua, and Chowdhury 2022; Pangaribowo and Iskandar 2022).

Sustainable Development Goal 7 (SDG7) is focused on ensuring access to affordable, reliable, sustainable, and modern energy for all by 2030. This goal aims to increase the share of renewable energy in the global energy mix, improve energy efficiency, and provide universal access to electricity and clean cooking solutions. The lack of access to clean cooking facilities is a major problem in many developing countries, and it has significant impacts on SDG7. According to the International Energy Agency (IEA), around 2.6 billion people worldwide still lack access to clean cooking facilities. This means that these households primarily rely on traditional biomass, coal, or kerosene for their cooking needs. The use of these fuels for cooking can have significant health and environmental impacts, which in turn can impact SDG7 (IEA 2020).

The burning of solid fuels such as coal and wood releases harmful pollutants such as carbon monoxide, nitrogen oxides, and particulate matter, which can cause respiratory and other health problems. According to the World Health Organization (WHO), exposure to these pollutants from indoor cooking fires causes around 4 million premature deaths per year, mainly in developing countries (Awan, Bilgili, and Rahut 2022; IEA 2020). This not only has a significant impact on human health but also affects the economic development of these countries, as people who are sick cannot contribute fully to the economy. Furthermore, the use of solid fuels for cooking also contributes to deforestation and climate change. Deforestation leads to a loss of biodiversity, which in turn can impact the food security and livelihoods of people living in affected areas. The burning of solid fuels for cooking also emits greenhouse gases, which contribute to climate change.

To address this issue, SDG7 aims to provide universal access to clean cooking solutions by 2030. This can be achieved by promoting the use of clean fuels such as liquefied petroleum gas (LPG), biogas, and ethanol, as well as by promoting the use of clean cookstoves that are more efficient and emit fewer pollutants. Achieving this goal will have significant health, environmental, and economic benefits, as it will reduce the harmful impacts of indoor air pollution, reduce deforestation, and contribute to mitigating climate change. In addition to the lack of access to clean cooking facilities, the lack of access to electricity is another major problem that impacts SDG7. According to the IEA (2020), around 770 million people worldwide still lack access to electricity, mainly in sub-Saharan Africa and South Asia. Lack of access to electricity can limit people's ability to access education, healthcare, and other basic services, and can also limit economic development (IEA 2020; Awan, Bilgili, and Rahut 2022; WHO 2018).

Therefore, SDG7 is critical for ensuring access to affordable, reliable, sustainable, and modern energy for all, including universal access to electricity and clean cooking facilities. Achieving this goal will have significant health, environmental, and economic benefits, and will help to promote sustainable development and reduce poverty worldwide.

On the other hand, in Sub-Saharan Africa, woodfuel (firewood and charcoal) and farm residues are the most common forms of energy used for cooking and heating. Due to their low prices, they are widely used in small-scale businesses such as bakeries, brick-making, and drying agricultural products such as tobacco and tea (Aikins and McLachlan 2020). As a result, nearly 90% of the population in SSA depends on either firewood or charcoal for cooking (Aikins and McLachlan 2020). In Tanzania, charcoal is mostly used in urban areas, while firewood is most common in rural areas. More than 70% of households in urban areas of Tanzania depend on charcoal, and in 2020, Tanzania was ranked 7th in global charcoal production (Citizen 2021).

The World Health Organisation reports an average of 1.5 million premature deaths each year due to indoor air pollution from household solid fuel consumption (Garba and Bellingham 2021; Twumasi et al. 2021; Gitau et al. 2019). The choice and amount of energy used by households for lighting and cooking are rooted in poverty, according to the World Health Organization (Apergis, Polemis, and Soursou 2022). Access to cleaner energy reduces the prevalence of diseases as well as mortality rates among children and adults. It also reduces the time spent gathering firewood for both women and children (Sun et al. 2022; Rosenthal et al. 2018) moreover, the adoption of clean cooking energies among households is vital for achieving the Sustainable Development Goal 13 (Climate Action) (Paudel et al. 2021).

The government and non-governmental organisations have undertaken initiatives to reduce household poverty by promoting and increasing the use of clean energy in Tanzania. These efforts align with the Sustainable Development Goals targets of ending poverty globally, particularly ensuring access to electricity, and access to clean cooking technologies and fuels for all people by 2030 (Drescher and Janzen 2021). The global focus has always been to respond to climate change by promoting the use of renewable and low-carbon energy through increased spending and diversification (Twumasi et al. 2021; Gafa and Egbendewe 2021).

Empirical review

Numerous studies have been conducted on the drivers of energy choices without integrating their effects toward poverty alleviation/reduction. Although several studies have examined the determinants of energy choices, they fail to explore the relationship between energy choices and poverty reduction in developing countries (Mperejekumana et al. 2021; Pangaribowo and Iskandar 2022; Wassie, Rannestad, and Adaramola 2021; Twumasi et al. 2021; Shrestha et al. 2021; Zhu et al. 2022; Tonini et al. 2022; Mwaura, Okolobio, and Ahaibwe 2014).

For instance, Zhu et al. (2022) studied drivers of cooking energy choice among households in China and found that en ergy choice is driven by the education level of the household head, living conditions, as well as the accessibility and affordability of energy. Similarly, Pangaribowo and Iskandar (2022) employed a multivariate probit model to analyse determinants of household cooking energy choices by using Indonesia Family Life Survey-East data. The study revealed that education, sources of household income, and household size are significant determinants of energy choices.

However, studies have overlooked the use of Probit, Two-Stage Residual Inclusion, and Control Function Approach in studying the drivers of energy choice as well as its effect on poverty reduction among urban households in Tanzania. Furthermore, the studies have not analysed the different factors influencing the renewable and non-renewable cooking energy choice among urban households, which this study does (Mperejekumana et al. 2021).

Several studies have examined the relationship between poverty and household energy choices, but only in one direction (Halkos and Gkampoura 2021; Abdulganiyu 2019; Olang, Esteban, and Gasparatos 2018). However, there is a gap in the literature regarding the extent to which these energy choices contribute to poverty reduction in developing countries. To address this gap, instrumental variables models like Two-stage Residual Inclusion (2SRI) and Control Function Approach (CF) are needed to provide counterfactual information on the effects of cooking energy (Dendup and Arimura 2018; Dagnachew et al. 2019; Mperejekumana et al. 2021; Rose et al. 2022). While some studies have explored the determinants of energy choices and energy poverty (Awan, Bilgili, and Rahut 2022; Brown and Vera-Toscano 2021), the relationship between energy choices and poverty reduction in developing countries has not been sufficiently studied. Moreover, the use of instrumental variables models such as 2SRI and CF in exploring the effects of cooking energy on poverty reduction has not been adequately explored in the literature, therefore this study fills such empirical and methodological gaps.

Theoretical foundation

This study has utilised the energy ladder model, which provides a valuable framework for understanding household energy choices and transitions. According to the model, households progress from using traditional energy sources like firewood and crop residues to more transitional options such as charcoal and coal, and eventually to modern sources like electricity, LPG, and solar as their income level increases (Bensch, Jeuland, and Peters 2021; Pangaribowo and Iskandar 2022). However, it is important to note that the energy ladder model only accounts for income-related factors and disregards other non-income factors that may affect household energy decisions, which are captured in the household utility theory. Therefore, by employing all factors together the model becomes more holistic in understanding household energy choices, which helps in developing effective policies and initiatives that promote sustainable and equitable access to clean energy for all.

Additionally, this model highlights the importance of income as a key driver of energy choice, emphasising the need for poverty reduction strategies that can increase households’ access to cleaner energy options. (Figure 1)

Figure 1. Extended poverty-energy ladder. Source: Authors’ design (2022).

Household utility theory on cooking energy choice

Households’ decisions on the type of cooking energy they use are guided by economic theory and their rational thinking. According to the theory of residential energy demand, households consider their preferences and budgetary constraints when choosing the type of cooking energy that will provide them with the greatest perceived utility (Greve and Lay 2023; Kitole and Sesabo 2022).

The choice of cooking energy is influenced by the socioeconomic characteristics of each household, which vary from one household to another. This study considers nine exhaustive outcomes, including firewood, solar, LPG, biogas, kerosene, coal, charcoal, farm residues, and electricity, as the dependent variable for the household's primary source of cooking energy. For a household that faces a choice between these outcomes, the utility function of such a household can be expressed as follows: (1) $U_{tj}={\beta }^{\mathrm{\prime }}{x}_t+{\epsilon }_{tj}j=0,1,2,3,4$(1) When the household choice$j$ is made, the underlined assumption is $U_{tj}$ is at maximum among the given set of utilities. That is; $j$ is chosen when $U(alternativej)>U(alternativek),\gamma j\ne k.$

Data and methods

Non-experimental research design is important in this study because it allows for the exploration of data sets from the Tanzania Panel Survey of 2020/2021 collected by the National Bureau of Statistics (NBS). This approach is particularly useful in cases where experiments are not possible or ethical, such as studying the energy choices of households in real-life settings. In this study, the non-experimental research design helps to provide a realistic representation of the energy choices made by households in urban areas of Tanzania. By using data from the Tanzania Panel Survey, the study has access to a large and representative sample of households, which increases the external validity of the findings. This means that the results of the study can be applied to other similar settings in Tanzania and other Sub-Saharan African countries with similar energy choices.

Furthermore, the non-experimental research design helps to address some of the limitations of experimental research designs, such as the lack of control over extraneous variables and the inability to manipulate variables of interest. The study can control for some of the confounding variables that may influence the energy choices made by households, such as income, household size, and education levels.

Econometric model specification

This study has employed the Multinomial Logit Model [MNL] to analyse determinants of household cooking energy choices followed by Kapsalyamova et al. (2021) as well as Kitole and Sesabo (2022) which have described the basis of how MNL is powerful when having more than three choices. Therefore, The Multinomial Logit Model [MNL] can be specified as: (2) $log\left[{\displaystyle \frac{{\pi }_j(x_i)}{{\pi }_k(x_i)}}\right]={\alpha }_{0i}+{\beta }_{1j}{x}_{1i}+{\beta }_{2j}{x}_{2i}+\dots +{\beta }_{pj}{x}_{pi}$(2) Since $j$ ranges from $1$ to $k-1$ and $i$ from $1$to $n$ then the reduced form of the equation will be expressed as; (3) $log({\pi }_j(x_i))={\displaystyle \frac{exp({\alpha }_{0i}+{\beta }_{1j}{x}_{1i}+{\beta }_{2j}{x}_{2i}+\dots +{\beta }_{pj}{x}_{pi})}{1+{\sum }_{j=1}^{k-1}exp({\alpha }_{0i}+{\beta }_{1j}{x}_{1i}+{\beta }_{2j}{x}_{2i}+\dots +{\beta }_{pj}{x}_{pi})}}$(3) The maximum likelihood approach is used to estimate the specified MNL model in equation (2). This model is chosen based on its ability to use the cumulative distribution function of the logistic distribution and other important factors (Ishengoma and Igangula 2021; Kebede, Tolossa, and Tefera 2022; Mekonnen and Kohlin 2008). It is preferred over the Multinomial Probit (MNP) model, which assumes that residuals are normally distributed, as MNL assumes that residuals are identically and independently distributed. The use of MNP is not suitable for this study as it only considers choices up to three options (Kitole and Sesabo 2022).

To analyse the effects of cooking energy choices on poverty reduction in urban Tanzania, this paper employs the Instrumental Variable model. Endogeneity arises due to the correlation between household energy choices and the error term, which violates the Gaus-Markov theorem assumptions. The econometric methods of two-stage residual inclusion (2SRI) and instrumental variable (IV) are used to address this problem (Kitole, Lihawa, and Nsindagi 2022).

As suggested by Hausman (1978), 2SRI is an effective method for controlling endogeneity for both nonlinear and linear models. This method is a two-step procedure in which residuals are calculated in an estimated reduced form and then used as additional explanatory variables in the second-stage regression. Testing the null hypothesis of exogeneity of a subset of regressors in this model is equivalent to variables addition tests for the equality to zero of the coefficients of the first-stage residuals in the second-stage reduced equation. This approach is also known as the control function (CF) or two-stage residual inclusion with extensions.

Thus, endogeneity occurs during estimating equation 4 whereas the urban household poverty status is indicated by $y_i$ as the binary outcome variable; ${\beta }_1$is a $K\times 1$ vector and $X$ is $n\times k$ matrix of covariates while ${\mu }_i$ is error term is expressed as; (4) $y_i={\beta }_0+{\beta }_1{X}_i^{{}^{\mathrm{\prime }}}+{\mu }_i$(4) As the result of endogeneity, a correlation between the explanatory variable and error term is not zero $((E(X,\mu )\ne 0)$ thus we apply instrumental variables (Kitole, Lihawa, and Mkuna 2022) and therefore equation 5 is reduced to: (5) $y_{1i}^{\ast }=\beta y_{2i}+\gamma X_{1i}+{\mu }_i$(5) (6) $y_{2i}=X_{1i}{\alpha }_1+X_{2i}{\alpha }_2+v_i$(6) Therefore, the $y_{1i}^{\ast }$ in equation 5 (reduced form) is the outcome variable for the $i^{th}$observation (which is the urban household poverty status).

Moreover, to estimate the effects of the cooking energy choice on household poverty status the current study has employed the Probit model due to its ability in providing probabilistic estimates on the chances for each cooking energy choice to cause poverty among households in urban Tanzania.

Consider, (7) ${\epsilon }_i=N(0,{\delta }^2)$(7) (8) $Prob(Y_i=1)=F\left({\displaystyle \frac{{\beta }_0+{\beta }_1{X}_{1i}}{\delta }}\right)$(8) Whereas $F$ is the standard normal cumulative density function, thus, for the derivatives to obtain the marginal effects given $Y=1,$ equation 8 becomes; (9) ${\displaystyle \frac{\mathrm{\partial }Prob(Y_i=1)}{\mathrm{\partial }{X}_1}={\displaystyle \frac{\mathrm{\partial }F\left({\displaystyle \frac{{\beta }_0+{\beta }_1{X}_{1i}}{\delta }}\right)}{\mathrm{\partial }{X}_1}=f\left({\displaystyle \frac{{\beta }_0+{\beta }_1{X}_{1i}}{\delta }}\right){\beta }_1}}$(9) Therefore, (10) $y_i=\{\begin{array}{c}0if{y}_i^{\ast }\le 0\\ 1if{y}_i^{\ast }>0\end{array}$(10) Whereas $y_i=1$ represents urban households that live below the global poverty line index of US$ 1.9 while $y_i=0$ for non-poor households or households living above the poverty threshold. The calculation of household daily consumption (expenditure) for the comparison based on the poverty threshold has been done based on the World Bank poverty measurement (Kitole and Sesabo 2022)

Results

Table 1 provides descriptive results of the sample of households in urban areas in Tanzania. The Table 1 reveals important characteristics of the households, including the gender of household heads, education levels, employment status, land ownership, and agricultural participation. The results indicate that male-headed households are more prevalent, comprising 69.92% of the sample, while female-headed households make up the remaining 30.08%. Additionally, a significant portion of the households has no formal education, with 33% of the heads of households never having attended any formal education. Only 1.13% of household heads who attended secondary education did not complete their studies, while 6.76% completed secondary education, and 4.13% had more than secondary education.

Table 1. Summary of descriptive statistics.

Variable	Categories	Frequency	Percentage
Sex	Male	1,913	69.92%
	Female	823	30.08%
Total		2,736	100%
Level of education	No formal education	903	33.00%
	Some Primary	317	11.59%
	Completed Primary	1,187	43.38%
	Some Secondary	31	1.13%
	Completed Secondary	185	6.76%
	More than Secondary	113	4.13%
Total		2,736	100%
Employment Status	Employed	520	19.01%
	Self Employed	1,568	57.31%
	Paid Household Worker	9	0.33%
	Unpaid Household Worker	98	3.58%
	Unemployed	78	2.85%
	Retired	135	4.93%
	Never Worked	328	11.99%
Total		2,736	100%
Land Ownership	Don’t Own (No)	1,860	68.00%
	Own (Yes)	876	32.00%
Total		2,736	100%
Livestock Ownership	Don’t Own (No)	2,076	75.88%
	Own (Yes)	660	24.12%
Total		2,736	100%
Business Ownership	Don’t Own (No)	2,015	73.65%
	Own (Yes)	721	26.35%
Total		2,736	100%
Agriculture Participation	Don’t Participate (No)	1,572	57.46%
	Participate (Yes)	1,164	42.54%
Total		2,736	100%

Regarding employment, the majority of household heads were self-employed in various economic activities, comprising 57.31% of the sample. Meanwhile, 19.01% were employed in various public and private sector entities. Unemployed heads of households represented 2.85%, while 11.99% had never worked. Furthermore, the study found that land ownership is not widespread among households in the sample. Only 32% of households own land with a title deed, while the majority (68%) did not have any. Finally, 42.54% of households participate in agricultural activities. These descriptive results provide a useful foundation for exploring the relationship between household characteristics and energy choice.

Table 2 presents the findings on the distribution of the primary sources of cooking energy among the households in the study. The results indicate that charcoal is the most commonly used source of cooking energy, with 41.52% of households using it as their primary source. Firewood follows as the second most common source, with 19.663% of households using it, while solar ranks third with 18.90%. The shares of LPG and electricity as primary sources of cooking energy are relatively low, with 7.89% and 6.33% respectively. Farm residues and coal are the least used sources of cooking energy, accounting for only 0.07% and 0.11% respectively. These findings provide insights into the current patterns of energy use in households in urban areas of Tanzania and could inform policies and programs aimed at promoting sustainable energy use.

Table 2. Energy use among households.

Variable	Frequency	Percentage
Electricity	173	6.33%
Solar	517	18.90%
Industrial gas (LPG)	216	7.89%
Biogas	60	2.19%
Kerosene	91	3.33%
Coal	3	0.11%
Charcoal	1,136	41.52%
Firewood	538	19.66%
Farm residues	2	0.07%

Drivers of cooking energy choice among urban households

Table 3 displays the results of a Multinomial logit model analysis with charcoal as the base outcome due to its prevalence as the most frequently used cooking energy source among the nine options examined. The analysis found that larger household size was associated with an expected decline of 8.76% in the likelihood of households choosing electricity over charcoal. This result is consistent with previous research that suggests households with more active working family members tend to have higher incomes and are more likely to use cleaner cooking energy sources (Gyamfi et al. 2022; Ishengoma and Igangula 2021).

Table 3. Multinomial logit on household’s cooking energy choice.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Electricity	Solar	Industrial Gas (LPG)	Biogas	Kerosene	Coal	Firewood	Farm Residues
Household size	−0.0876** (0.0399)	0.0966 (0.120)	−0.242*** (0.0026)	−0.287*** (0.0779)	−0.436** (0.0761)	0.154 (0.444)	0.132*** (0.0249)	0.0908 (0.247)
Sex (Male)	0.0449 (0.197)	−1.088 (0.773)	0.138 (0.178)	−0.200 (0.333)	−0.670** (0.217)	1.072 (1.516)	0.289* (0.112)	−16.38 (4,565)
Age	0.0159** (0.0007)	−0.0268 (0.0238)	0.0129** (0.0010)	−0.00379 (0.0122)	0.00574 (0.00975)	−0.137 (0.0874)	0.0227*** (0.0045)	0.0315 (0.0596)
Marital status (Married)	−0.187 (0.124)	0.885** (0.442)	−0.245** (0.114)	0.257 (0.200)	−0.0802 (0.160)	−1.459 (1.131)	−0.128 (0.0914)	−0.682 (1.778)
Employed	0.0554** (0.0118)	0.123 (0.118)	0.134*** (0.0370)	0.121 (0.0652)	−0.0342 (0.0647)	−0.0689 (0.334)	0.0609** (0.0293)	−0.579 (0.858)
Formal education	0.0900*** (0.0656)	0.302 (0.200)	0.410*** (0.0553)	0.8492 (0.9937)	0.123 (0.0900)	−14.20 (585.9)	−0.0496 (0.0531)	−14.89 (1,670)
Farming activities	−0.619*** (0.204)	0.572** (0.546)	−0.269 (0.181)	−0.372 (0.326)	0.455*** (0.412)	2.593 (1.765)	0.682*** (0.111)	0.424*** (0.142)
Business activities	0.427** (0.200)	0.413 (0.591)	0.782*** (0.200)	−0.502 (0.328)	0.236 (0.245)	1.015 (1.443)	−0.424*** (0.142)	−1.396 (1.497)
Income	1.25007*** (0.08008)	−8.4800* (2.68007)	2.2007*** (0.71008)	2.25007*** (0.93008)	−2.48007 (1.72007)	−1.03005 (6.43006)	−8.92007** (2.75008)	−5.86008 (4.59007)
Land ownership	−0.0248 (0.218)	−0.0850 (0.574)	0.233 (0.192)	0.308 (0.337)	−0.993** (0.484)	0.0224 (1.698)	0.986*** (0.126)	0.530 (1.666)
Observations	2,734	2,734	2,734	2,734	2,734	2,734	2,734	2,734
LR ch2 (80)	1,048.55
Prob > Chi 2	0.0000
Pseudo R2	0.3579

Standard errors in parentheses.

***p < 0.01, **p < 0.05, *p < 0.1.

Moreover, household size was negatively and significantly associated with the use of other clean cooking energy sources, including industrial gas (LPG), biogas, and kerosene, by 24.2%, 28.7%, and 43.6%, respectively, at a significance level of 1%. However, household size was positively associated with the use of firewood by 13.2%. Several studies in developing countries have found similar results and suggested that households with more non-working members may struggle to afford or prioritise the use of cleaner cooking energy sources, as much of their income may be spent on food consumption (Wassie, Rannestad, and Adaramola 2021; Kebede, Tolossa, and Tefera 2022; Kapsalyamova et al. 2021; Ronzi et al. 2019).

The analysis further revealed that male-headed households were more likely to use firewood as a cooking energy source than female-headed households, relative to the base outcome. Previous studies suggest that women are typically responsible for household work, including cooking, in many developing countries, which may make them more aware of the negative health and environmental impacts of using dirty cooking energy sources and more likely to prioritise cleaner alternatives (Sedai, Nepal, and Jamasb 2021; Gitau et al. 2019; Qiu et al. 2022).

The age of the household head was found to significantly influence the use of industrial gas (LPG), electricity, and firewood. Specifically, a one-year increase in the age of the household head was associated with a 1% increase in the likelihood of households choosing electricity over charcoal and a 2.27% increase in the likelihood of households using firewood instead of charcoal. However, age alone cannot directly influence energy choices, as income levels vary by age and can strongly influence household energy consumption patterns (Jacques-Aviñó et al. 2022; Tonini et al. 2022; Mwaura, Okolobio, and Ahaibwe 2014; Karimu, Mensah, and Adu 2016; Rosenthal et al. 2018).

Additionally, the employment status of the household head was found to significantly influence the use of electricity, industrial gas, and firewood, with employed heads of household being more likely to choose these energy sources over charcoal. Education level and ownership of business enterprises were also found to be important factors in households’ energy choices, with formal education and business ownership increasing the likelihood of households choosing electricity and industrial gas (Pangaribowo and Iskandar 2022; Wassie, Rannestad, and Adaramola 2021; Zhu et al. 2022; Twumasi et al. 2021; Njiru and Letema 2018).

Moreover, the study conducted by Karmaker et al. (2022) and Pangaribowo and Iskandar (2022) suggested that farmers may be more inclined to switch to cleaner energy sources as they earn additional income from their farming activities. This finding is contradictory to the current study, which found that being a farmer was associated with a reduced interest in using electricity for cooking. The reasons behind this disparity may be attributed to the contextual differences in terms of income level, access to energy infrastructure, and sociocultural beliefs about traditional cooking methods.

Furthermore, the use of farm residues as a source of cooking energy is often associated with poverty, as it is considered the least expensive option for households (Pachauri et al. 2021; Dagnachew et al. 2020; Adeyemi and Adereleye 2016). This suggests that households in urban areas of Tanzania who engage in farming activities may be more likely to use farm residues as a cooking energy source due to their limited financial resources. Therefore, the significant relationship between farming activities and the use of farm residues for cooking may indicate the presence of poverty among households in these areas.

Effects of household cooking energy choices on poverty

Assessment on the effects of household cooking energy choices on poverty is based on estimations of equations 5 and 6 which employed the instrumental variable due to the endogenous relationship between the poverty and unobserved explanatory variables in the cooking energy choices. Thus, to solve such a problem the instrument used in this study was the distance from the homestead to market place after confirming the existence of endogeneity as then we reject the null hypothesis that there is no endogeneity at 5% (see the last row of Table 4). On the other hand, the endogeneity is confirmed due to the significance of the residuals of cooking energy choices at the two-stage residual inclusion (2SRI) model. On the other hand, the Pseudo R squared of 0.3579 indicates that the statistical model explains 35.7% of the variance in the dependent variable. In other words, the model is able to account for approximately one-third of the variation in the outcome variable based on the predictors included in the model.

Table 4. Marginal effects of factors explaining the effects of Cooking Energy Choices on Poverty Reduction in Tanzania.

Dependent variable is poverty status	Estimation methods
	Probit (1)	2SRI (2)	Control function approach (3)
Socioeconomic characteristics
Age	0.0934399** (0.011461)	0.1388563*** (0.040235)	0.1711303*** (0.036517)
Age squared	−0.0004188*** (0.00016)	0.0011169** (0.0005038)	0.001313** (0.0006038)
Sex (Male)	−0.0786953 (0.15652)	−0.1367008 (0.2825611)	−0.19438514 (0.2091031)
Income	0.1465603** (0.05831)	0.1574773*** (0.0146501)	0.1903533*** (0.0132720)
Years of schooling	−0.0933381** (0.001654)	−0.10282597* (0.0610443)	−0.1426193** (0.0617977)
Married	−0.0672995* (0.001586)	−0.0935041** (0.0137595)	0.1103841*** (0.02589123)
Employed	−0.184208* (0.078210)	−0.210074** (0.068342)	−0.250821*** (0.042902)
Cooking Energy Choices
Electricity	−0.1649084* (0.014430)	−0.1726841** (0.0657123)	−0.2135497*** (0.0222663)
Solar	−0.2193702** (0.094915)	−0.2428733** (0.090226)	−0.2763372*** (0.073291)
LPG	−0.1278926** (0.039011)	−0.1573496** (0.0435052)	−0.1866471** (0.0403722)
Biogas	−0.0302766*** (0.000093)	−0.090918** (0.0031315)	−0.1083577** (0.0034191)
Kerosene	−0.0059421*** (0.00141)	−0.0146029*** (0.0048019)	−0.038882*** (0.0065293)
Farm residues	0.0808349** (0.01125)	0.10374517** (0.0135535)	0.1500122** (0.0495906)
Firewood	0.189564** (0.0272601)	0.1975217** (0.034502)	0.2150202*** (0.013562)
Charcoal	0.242005** (0.064005)	0.273382*** (0.091162)	0.318853*** (0.088941)
Electricity residual		−0.1274317 (0.0912781)	−0.606732 (0.4730524)
Solar residual		−0.6388615** (0.2722084)	−0.5079913** (0.1992683)
LPG residual		−0.069315* (0.001832)	−0.1010511*** (0.0007881)
Biogas residual		−0.471921** (0.167672)	0.2409207*** (0. 1109457)
Kerosene residual		−0.015731* (0.006912)	−0.1870436*** (0.034591)
Farm residual		0.369280** (0.120326)	0.388021*** (0.106453)
Firewood residual		0.359192* (0.115721)	0.3047320** (0.100564)
Charcoal residual		0.2212467*** (0.0748135)	0.2092182*** (0.038106)
Electricity*residual			−0.1421110*** (0.0281522)
Solar*residual			0.2900782*** (0.0382171)
LPG*residual			0.2270104*** (0.002460)
Biogas*residual			−0.0891471*** (0.0057302)
Kerosene*residual			−0.0401612 (0.0004072)
Farm Residue*residual			0.0001191 (0.0000226)
Firewood*residual			−0.1610241*** (0.02101)
Charcoal*residual			0.2241621*** (0.0191132)
Sample size	2,734	2,734	2,734
Pseudo R squared	0.2785	0.2940	0.3107
Instrumented	Cooking energy Choices
Instrument	Distance from homestead to the market place
Wald test of exogeneity	Corr = 0 Chi-square = 8.95		Prob > chi2 = 0.0158

Standard errors in parentheses.

***p < 0.01, **p < 0.05, *p < 0.1.

Results on Table 4 reveal that having controlled effects of endogeneity and biases, the use of cooking energy sources such as electricity, solar, LPG and Kerosene are significantly associated with a decline in household poverty. This result has been observed from all three models (Probit, 2SRI and CF) of which the size of the coefficients have been observed to improve from one model to another/ This imply that model biases have been suppressed after controlling endogeneity and heterogeneity whereas heterogeneity is verified being controlled due to the significant of the cooking energy choices and residuals. However, the results of this study have been interpreted based on the Control Function (CF) Approach.

Specifically, findings show that choosing electricity as a source of cooking energy reduces the likelihood for household falling into poverty by 21.35%. Moreover, poverty prevalence across households that chose solar and LPG as their cooking energies were found to decline by 27.63% and 18.66% significantly. On the other hand, the use of farm residues, firewood and charcoal were found to increase poverty among households significantly at 15%, 21.50% and 31.88% respectively. These findings are supported by studies of Ashagidigbi et al. (2020) and Jacques-Aviñó et al. (2022) who argued that the use of dirty cooking energy sources has an impact not only on the health of those who are directly engaged in cooking or utilising the energy rather multiple effects to the extent of lowering livelihood of many households. Berkouwer and Joshua (2022) as well as Kapsalyamova et al. (2021) suggested that an increase in household costs of taking care the sick due to poor or dirty cooking energy sources prone most households in developing countries falling into a poverty trap.

Demand for renewable and Non-renewable energy among households

The findings of this study suggest that the demand for cooking energy is an important factor in explaining the quality of life that people live. To gain a deeper understanding of households’ demand for cooking energy choices, this study comparatively analysed households’ demand based on two major types of energies: renewable and non-renewable energy, as presented in Table 5.

Table 5. The Multinomial Logit on estimating the household demand for renewable and non-renewable energy among households.

Variable	Non-renewable energies								Pooled Non-Renewable Energy (n = 2219)	Pooled Renewable Energy (n = 517)
	Electricity dy/dx	Charcoal dy/dx	Firewood dy/dx	LPG dy/dx	Kerosene dy/dx	Biogas dy/dx	Coal dy/dx	Farm residue dy/dx
Location (Rural)	−0.0414*** (0.0012)	0.3270*** (0.1125)	0.2592** (0.0928)	−0.3327** (0.1047)	−0.1273 (0.0049)	0.0019 (0.1470)	0.0157 (0.1020)	0.0198** (0.003)	0.1309** (0.0429)	0.3612** (0.1154)
Sex (Male)	−0.3065 (0.2919)	−0.10404 (0.3406)	0.0566 (0.5349)	0.0107 (0.6101)	0.1013 (0.0801)	−0.2511 (0.2191)	0.1479 (0.2743)	0.1328 (0.0946)	0.1285*** (0.0029)	−0.0401** (0.0109)
Household size	−0.2727*** (0.0752)	0.1058* (0.0224)	0.0250 (0.0711)	0.0179 (0.0515)	0.1573 (0.1910)	0.0623 (0.2142)	0.1862 (0.1005)	0.1046 (0.2103)	0.1068* (0.0348)	−0.0402 (0.1503)
Age	−0.0335*** (0.0014)	−0.0459*** (0.0106)	−0.0217** (0.0104)	0.0023 (0.0005)	0.0531 (0.0027)	0.0291 (0.0807)	−0.0073 (0.0018)	0.3534 (0.5632)	0.0353* (0.0142)	−0.1350*** (0.0060)
Marital status	0.2441** (0.0943)	0.0919* (0.0228)	−0.1129** (0.0116)	0.1802*** (0.0021)	−0.0022 (0.0121)	−0.1762 (0.4069)	−0.1912 (0.2242)	0.0232 (0.0775)	0.0519 (0.0059)	0.0198** (0.0001)
Farming activities	−0.4416*** (0.0987)	0.3711*** (0.0198)	0.1674** (0.0350)	−0.1056* (0.0012)	0.1541* (0.0623)	0.02081 (0.0809)	0.0820 (0.1125)	0.1009 (0.5213)	0.2113*** (0.0009)	−0.1912** (0.0123)
Years of schooling	0.16481** (0.0112)	−0.2525** (0.1136)	−0.2371* (0.1076)	0.18956** (0.0018)	−0.0256 (0.0031)	0.0001 (0.0203)	0.0022 (0.0583)	−0.4482*** (0.0398)	0.0924* (0.0119)	0.2338*** (0.0301)
Income	0.1575*** (0.0571)	−0.1185*** (0.0054)	−0.2309** (0.0614)	0.1404*** (0.0000)	−0.1041** (0.0077)	0.0071** (0.0013)	−0.1266 (0.1485)	−0.5106*** (0.0509)	0.1691** (0.0553)	0.2803*** (0.0089)
Poverty status	−0.2515** (0.0815)	0.1502** (0.0315)	0.1479** (0.0214)	−0.1736* (0.0173)	0.0947** (0.0100)	−0.0613* (0.0119)	0.0569 (0.057)	0.7687*** (0.1170)	0.0980** (0.0096)	−0.1560** (0.0038)
Employment	0.2770*** (0.0794)	−0.2453*** (0.0780)	−0.2136** (0.0966)	0.1067** (0.0176)	0.0672** (0.0147)	0.0701 (0.1602)	0.0935 (0.1101)	−0.3601*** (0.0401)	0.0239 (0.0578)	0.0411* (0.0177)
Business activities	0.1726** (0.0264)	−0.2071** (0.0749)	−0.1773** (0.0415)	0.1835** (0.0031)	−0.3227** (0.0049)	0.1562 (0.3261)	0.0991 (0.1209)	−0.1460 (0.0182)	0.0115 (0.0298)	0.0934* (0.0316)
Observations	2219	2219	2219	2219	2219	2219	2219	2219	2219	517
LR ch2	689.55								478.784	62.055
Prob > Chi 2	0.0000								0.0000	0.000
Pseudo R2	0.3291								0.4104	0.3705

Standard errors in parentheses.

***p < 0.01, **p < 0.05, *p < 0.1.

The results on Table 5 show that living in a rural area increases the likelihood of a household consuming renewable energies higher than non-renewable energies at a proportional difference of 23.03% significantly compared to households living in urban areas. These findings are consistent with those of Dimoso and Kitole (2021) and Martey et al. (2022), who argue that renewable energies such as solar are more commonly used in rural areas than in urban areas in most developing countries.

The findings also show that male-headed households have different preferences towards energy types. While households headed by males have a positive preference for non-renewable energy at a significant level of 12.85%, the preference is negative and significant at 4.01% for renewable energy. This implies that any program that aims to change or improve household energy use should not ignore the role of males, despite women being the core of domestic activities in most households in developing countries. Studies by Mperejekumana et al. (2021) and Gafa and Egbendewe (2021) have different findings than the current study and suggest that women are key actors in most of the domestic energy use, and should be given high priority in enhancing community stewardness towards the use of clean energies.

Furthermore, the findings have shown that household engagement in farming activities increases the likelihood of non-renewable energy use significantly at 21.13%, while reducing the households’ preference for renewable energy use significantly at 19.12%. Additionally, engagement in business activities increases the household's preference for non-renewable energy insignificantly, while increasing that of renewable energy significantly at 9.34%. Therefore, income-generating activities play a greater role in determining households’ choice of energy. Studies by Qiu et al. (2022), Ishengoma and Igangula (2021), and Alem and Demeke (2020) argue that as household income increases, the preference for cleaner energies increases.

Discussion and conclusion

The study findings reveal that there is a need for deliberate efforts to improve households’ income through investment in income-generating activities or projects in both rural and urban areas. This could be achieved through government policies aimed at supporting small businesses, agriculture, and entrepreneurship. For instance, the government could provide tax incentives or grants to small businesses and entrepreneurs to enable them to set up businesses that will create employment opportunities and increase household incomes. Additionally, the government could invest in infrastructure that supports agriculture, such as irrigation systems, to improve yields and increase incomes for farmers. Such efforts will not only help to reduce poverty levels but also increase households’ preference towards clean energy.

Moreover, the study recommends that governments should promote the use of renewable energy sources such as solar, wind, and hydropower to reduce dependence on non-renewable sources of energy. The government could provide subsidies for households to purchase and install solar panels or other renewable energy systems. Additionally, the government could invest in renewable energy infrastructure such as wind turbines and hydropower plants, which will create employment opportunities and reduce the cost of renewable energy. Furthermore, the government could develop policies that encourage the use of clean energy in public institutions such as schools and hospitals, which will serve as a model for households to follow.

Finally, education and awareness campaigns could be launched to inform households about the benefits of using clean energy, the health risks associated with traditional energy sources, and the impact of energy choices on the environment. The campaigns could be carried out through community-based organisations, schools, and other public institutions. Additionally, the government could work with NGOs and the private sector to provide training and education on the use of renewable energy sources, such as solar panels and wind turbines. This will help to increase households’ knowledge and understanding of clean energy and encourage their adoption.

In conclusion, achieving sustainable development goals of clean energy requires a collaborative approach from stakeholders such as governments, NGOs, and the private sector. Therefore, there is a need to form partnerships and collaborations that leverage the resources, expertise, and knowledge of various stakeholders towards the common goal of clean energy. By adopting these recommendations, governments can promote the use of clean energy while also reducing poverty levels and improving the well-being of households.

Disclosure statement

No potential conflict of interest was reported by the author(s).