Determinants of Household Transition of Cooking Fuel in Energy-Rich Peripheries: Evidence from Mozambique

1. Introduction

The trajectory of human history has been fundamentally determined by energy access, transitions, and use [1,2]. That narrative has yet to fully unfold in most developing countries [3], including those with abundant energy resources, such as Mozambique. The limited access to clean cooking fuel and electricity affects a substantial majority of the population, a global concern addressed in the UN's 2030 Agenda for Sustainable Development Goals (SDG7) [4,5].

Access to energy should be guided by a transition perspective that minimizes detrimental health and environmental impacts and enhances productivity. A US$28 billion annual investment in reliable energy sources, systems, and infrastructure in Africa by 2030 is essential to meet current population growth and energy demand [6]. In 2018, the Mozambican Government launched the “Energia Para Todos—ProEnergia” (energy for all) initiative supported by international donors with an equivalent amount of US$148 million [7]. The initiative aimed only at expanding infrastructures and increasing household connections for electricity access, neglecting the low rates of clean cooking access. When analyzing human rights and access to modern energy, Hesselman [8] found that although international human rights law and protocols have extensively addressed the right to ‘energy services’ and gender perspective, they focused on the ‘right to electricity’, lacked details, and had limited monitoring practices. Gregory and Sovacool [9] showed that investment in energy systems and infrastructures is one of the key impediments to energy access. Additionally, comprehensive development planning [10] is fundamental in transitional geographies.

According to a CEAGE and Winrock International study [11], Mozambique's deforestation and forest degradation rate has risen considerably, with an annual loss of 137,300 hectares. Each year, fuelwood extraction, charcoal production, and use result in approximately 600,000 tCO2e, accounting for 7% of the country's emissions. The national forestry inventory report [12] found that the deforestation rate was at 220,000 hectares per year, and Nogueira Lisboa et al. [13] showed that clean-cutting charcoal accounted for roughly 9%, representing a 2% rise since 2016.

Mozambique possesses abundant natural resources, including renewable and non-renewable resources. However, more than 90 percent of its population relies on solid biomass for cooking [14,15], while 47 percent have no access to electricity [14,16]. Several studies have shown that the use of solid biomass is associated with emissions of harmful household air pollutants with health implications [17,18]. Moreover, the considerable time spent gathering firewood contributes to low productivity in numerous developing countries [19,20]. Studies reveal limited effectiveness and reach of clean biomass cooking initiatives, despite promotional efforts by many donor organizations [21,22]. Hutton et al. [23] suggest that to address the low adoption of solid biomass, the transition to clean and modern cooking fuels (CCE), such as natural gas or liquefied petroleum gas (LPG), is essential and preferable among households [24,25].

Following its 1975 independence, the Mozambican government enacted numerous energy policies and strategies [26,27,28,29,30,31]. The policies aimed to attract foreign direct investment in the energy sector, establishing energy regulatory bodies (CNELEC, INP, and ARENE)[32,33,34], an energy fund (FUNAE)[35], and state-owned energy companies (EDM, HCB, CARBOMOC, ENH, and Petromoc) [36,37,38,39,40]. The policies prompted substantial investment in the exploration and discovery of onshore and offshore gas resources [41,42]. On the other hand, the country depends heavily on imported petroleum products, including LPG for cooking, to meet its energy demands [14,43]. From 1958, Mozambique used to refine on average 14,500 barrels per day of petroleum products from SONAREP Refinery, a private Portuguese-owned refinery, located in Matola, Maputo province [44,45]. The geopolitical developments in the region and fears of sanctions against the South African regime [44] led to a significant reduction in the refinery's output during the early 1970s. Later in the 1970s, following independence, the FRELIMO government's nationalization efforts rendered it inoperative [45,46]. International oil market volatility might cause fuel prices and transport fares to rise. Historically, rising fuel prices in Mozambique and other developing nations of the global south have sparked violent protests and unrest among consumers [47,48].

Several studies have explored household energy use patterns and behaviors by analyzing factors that control the household energy transition from traditional to modern [49,50]. Earlier research focused solely on monetary factors, such as household income and fuel prices, leading to the "energy ladder" theory. This theory suggests that when a household's income and living standards rise, they often switch from traditional, polluting, and inefficient fuels to cleaner, modern, and more efficient alternatives [49,50,51].

The earlier theory failed to account for the prevalence of middle- and high-income households adopting various fuels daily, spanning both upper and lower levels of the energy ladder or transition reversal [52,53]. In some cases, low-income and rural households are unable to access modern and advanced fuels due to higher fuel and stove prices, fuel security concerns, and availability [53,54,55]. The theory was also criticized for being overly simplistic, with detractors claiming it neglected other significant factors related to fuel-staking behavior.

The new theory suggests that the factors controlling household energy transition are not limited to socioeconomic status; they also include household characteristics, demographics, and energy systems [52,56]. Van Der Kroon et al. [57] highlight external factors, such as biophysical, political, and institutional influences, as well as internal opportunities, as crucial determinants of a household's fuel choice. While both hypotheses are acceptable, the “energy stacking” framework provides a more complete and structured approach for evaluating how households behave during an energy transition and pinpointing policy priorities for fuel switching [57], especially in most developing countries, where multiple fuel stacking is evident.

Previous studies have examined the use of local energy for international trade and regional cooperation [58,59], energy justice and planning [10,60], the role of energy systems in transitions [61,62,63,64,65], and the adoption of biomass and improved cookstoves [66,67,68]; research in this field focusing on household choice of cooking energy transition in energy-rich peripheries is limited. Also, much of the research on cooking fuel access in developing countries has centered on either the adoption of imported LPG or issues of gender and health.

In their study, Kerimray et al. [69] observed that energy poverty and fuel stacking were less severe in energy-rich areas compared with other regions. The study highlighted key advantages, including improved energy infrastructure, lower energy costs, and higher earnings.

Energy resource-rich nations and the energy transition nexus have been a subject of recent literature, with particular emphasis on rents and institutional frameworks for low-carbon-intensity energy sources such as electricity [70,71,72]. Although some highlight the positive role of natural gas rents, none have addressed the issue of cooking fuels.

Consequently, the role of domestic natural gas in promoting the transition to clean and efficient cooking is being overlooked. As far as the authors are aware, no prior research has examined how the consumption of locally produced gas affects transitions to and access to CCE in Sub-Saharan Africa (SSA), particularly in Mozambique. Thus, this study aims to contribute to understanding the factors influencing households' and micro and small enterprises' (MSMEs) cooking fuel transitions, using data collected from two provinces (Inhambane and Maputo City) in Mozambique, where locally produced natural gas is a primary cooking fuel for some households.

The research examines households and micro, small, and medium enterprises (MSMEs) in both rural and urban geographies. A qualitative and quantitative analysis of the primary data is used to explain the probability of their choices in the absence of the primary cooking fuel.

The remainder of the paper is structured as follows. Section 2 presents the study’s materials and methodology, including a description of the study area, survey procedures and data, and the empirical models used for analysis. Section 3 presents the results and discussion of the study. While section 4 offers a summary of findings and the resulting policy implications.

2. Materials and Methods

2.1. Study Setting

The study was carried out between September and November 2024 in two provinces currently utilizing domestically produced natural gas as a primary cooking fuel (Figure 1): 1) Inhambane (Govuro, Inhassoro, and Vilanculos districts) with a combined population of 269,781, and 2) Maputo City (KaMpfumo, KaNlhamankulu, and KaMaxaquene districts) with 559,042 residents [73,74,75,76]. Geographically, they are located in the southern region of the country. In addition, all districts in the study sites are located in coastal regions, where 60% of the Mozambican population lives [77].

According to national statistics [78], 27% of households in Inhambane use biomass cookstoves, while only 7% use electric or gas stoves. In contrast, 82% of households in Maputo use biomass cookstoves, compared with 57% using electric or gas cookstoves. Electricity access is much higher in Maputo, at roughly 99% of households, than in Inhambane, where it's just 42% [14]. Furthermore, 93 % of Mozambican households spend an average of 1.5 to 7 hours per week obtaining cooking fuel and 5 to 15 minutes setting up their cookstoves. Maputo city has the lowest percentage, at 64%, while Inhambane is about 95% [79].

Like most other Mozambican spaces, the study area comprises urban, rural, and a transition area called “peri-urban” [80]. Urban and peri-urban spaces are characterized by rapid expansion through self-occupation and informal settlements as a result of ‘benefits of dysfunctionalities’ and poor urban planning [81,82,83]. The expansion of urban and peri-urban areas has outstripped the development of urban services [84]. This also holds true in sparsely populated rural spaces [80,84].

More than half of the economically active population works in the informal sector [79]. This type of vulnerable employment is most prevalent in rural areas (67%), compared with urban and peri-urban areas, which are at 30%. The national statistics showed that Inhambane province had an informal employment rate of 55%, contrasting sharply with Maputo, the capital city, which had the lowest rate at 16%.

2.2. Energy Resources

Around 43% of Mozambique's total area (34 million hectares) is covered by natural forests [85], which provide energy to local communities and the majority of Mozambican households. The forest and biomass resources amount to approximately 1.74 billion cubic meters [86], with an annual biomass production and consumption ranging from 13.8 to 15.8 million tons [87]. This usage accounts for more than 80% of the fuel households use for cooking.

On the other hand, Mozambique possesses recoverable resources of about 130 trillion cubic feet of natural gas and 158 million barrels of oil [88,89], with Inhambane province comprising 4 trillion cubic feet of gas and 11 million barrels of oil, discovered between 1961 and 2003 [89,90]. Current production of energy resources is about 138.8 million BTU of liquefied natural gas (LNG) each year, 1,000 barrels of condensate per day by Coral Sul FLNG [91], by Italian multinational Eni, 192.5 gigajoules (GJ) of natural gas; and 500 thousand barrels of condensate per year [92,93] by South African petrochemical company Sasol, from Inhambane’s Pande, Temane, and Inhassoro fields. The country uses 26.2 GJ (13.6 percent) of this energy, with only 3.1 GJ used for cooking.

Furthermore, Mozambique imports roughly 48,300 metric tons of liquefied petroleum gas (LPG) each year for cooking [14] via the Maputo, Beira, and Nacala ports, which host LPG storage facilities. In Inhambane, LPG cylinders come from the neighboring province of Sofala; thus, the average retail price is about 2.20 US$/kg, 39 percent pricier than that of Maputo and Beira cities at an average of 1.34 US$/kg [94] (at an exchange rate of 64 MZN = 1 US$).

Clean cooking access is estimated at 2.7 percent, with Maputo at 19.1 percent, the highest in the country, and Inhambane averaging 3.5 percent [95].

2.3. Energy Systems and Infrastructures

In contrast to electricity infrastructures and other services, which were inherited from the colonial era [96], city gas infrastructure and systems were planned and constructed from scratch in the post-independence era. The first pipeline system implemented in the country was a small-scale project by the Mozambican National Oil Company (ENH) in the 1990s, as part of its efforts to develop the Pande and Temane gas fields. In 1992, it commenced operations in three districts within Inhambane, near the Pande & Temane gas fields [97]. Natural gas is now used in homes and restaurants for cooking, in industry, and for electricity generation, and the network has expanded to over 670 km [97].

The Maputo pipeline originated from a 25-year petroleum production agreement (PPA) signed in 2000 by Sasol, ENH, Companhia Moçambicana De Hidrocarbenetos (CMH), and the government of Mozambique [98,99]. Among other projects, the agreement involved building an 865 km export ROMPCO pipeline from Temane (Inhassoro District), Mozambique, to Secunda, South Africa [99]. This project commenced in April 2002 and incorporated five gas off-take points located in Temane, Chigubu, Macarretane, Magude, and Ressano Garcia [100]. In 2005, a 100 km natural gas pipeline was extended from Ressano Garcia to Matola, an industrial district of Maputo [59,101]. At first, the gas was limited to industries, transportation, and power generation [101]. By the end of 2013, its reach extended 72 km, including Maputo City and the Marracuene District [59]. A pilot program providing natural gas to households in Maputo city launched in 2018 in the Aeroporto-A neighborhood (in KaNlhamankulu) [102].

As a result of further expansion, more than 4,000 households and MSMEs in Maputo and Inhambane are connected to the natural gas pipeline systems. The study area and its key underlying energy systems are shown in Figure 1.

2.4. Sample Size

The sample size was not predetermined for the study. The study sample comprised 434 households and MSEs (N=434) from two provinces: 66 (15.2%) in Govuro, 110 (25.3%) in Inhassoro, and 139 (32.0%) in Vilanculos (Inhambane Province); and 60 (13.8%) in Nlhamankulu, 25 (5.8%) in KaMpfumo, and 34 (7.8%) in KaMaxaquene (Maputo City).

2.5. Data Collection and Management

A multimethod approach was used to examine clean energy access and consumption patterns in both rural and urban settings, with an emphasis on locally produced natural gas resources. A literature review, in-person interviews, and a survey in the Inhambane and Maputo provinces were employed. Questionnaires and interviews were conducted from September through November of 2024 and in the Portuguese language. This method is beneficial because it allows for the collection of both qualitative and quantitative data, gives participants a voice, improves the validity of the findings, and enables comprehensive analysis of complex phenomena [103,104,105,106].

A questionnaire contained a mix of open and closed questions. This survey was designed to gather input regarding the factors that influence a respondent's choice of primary energy [107,108,109]; in particular: i. demographic) urbanicity and aggregate size; ii. socioeconomic) head of household’s income, education, primary and alternative energy spending; iii. energy systems) incidents or accidents related to energy use; iv. sociocultural) energy stacking; A description of the factors is shown in Table 1. The questionnaire covered a considerable range of subjects, and a portion of the survey data is not pertinent to this publication.

Data was collected using MS Forms online or, if unavailable, using paper questionnaires. Notes were taken to capture responses not addressed by the questionnaire. To conduct the statistical analysis, the data were cleaned, coded, and analyzed with the mlogit package [110] of R’s version 4.5.2 [111]. Coding was crucial for improving the analysis of the quantitative data. [106].

2.6. Analytical Techniques

In this paper, descriptive statistics, cross-tabulation, p-values, and multinomial logistic (MNL) models were employed. We employed descriptive statistics to outline the demographic and socioeconomic traits of the respondents. Additionally, the efficiency of energy systems and the respondent’s perception of energy services were evaluated.

Given that social and behavioral responses are modeled using discrete variables, choices are mutually exclusive; a common way to represent binary events is a dichotomous (dummy) variable [112]. And when dealing with dummy dependent variables, predicted values from linear regression or ordinary least squares (OLS) yield an uneven split between the two categories, with observations tending to fall toward the extremes [112]. OLS models for a dichotomous response pose additional problems of nonlinearity, nonadditivity, nonnormality, and heteroskedasticity [112,113]. As such, we employed logistic models in quantitative analysis. This method is more appropriate for our data [114,115]. In addition, given that our sample size exceeds 100 observations, the risk of conducting significance tests or maximum likelihood (ML) estimation is minimal [116].

Households' decisions regarding cooking fuel, specifically their choice of one type over alternatives like piped natural gas, can be modeled using the Random Utility Model (RUM) [109]. Discrete choice models can be derived from the assumption of households’ utility-maximizing behavior [117]. Consider household head $i$, confronted with fuel alternatives $j=1,2,\dots J$. When a household $i$ opts for fuel type $j$ out of set if alternatives $J$ (natural gas, electricity, LPG, and biomass), he or she derives a utility level of $U_{ij}$. The function is dependent on observed covariates $V_{ij},$ and unknown factors, ${ϵ}_{ij}$. The observed factor is the standard random utility model and can be written as:

$$V_{ij}={\alpha }_j+{{\beta }_{1}Pr}_{ij}+{\beta }_2{Inc}_{ij}+{\beta }_3{Exp}_{ij}+{\gamma }_j{X}_i$$

(1)

In equation 1, ${\alpha }_j$ denotes alternative-specific constant; ${Pr}_{ij}$, ${Inc}_{ij}$, and ${Exp}_{ij}$are the alternative and choice situation specific covariates for price, incidents and expenditure, respectively, with generic coefficients ${\beta }_1$, ${\beta }_2$ and ${\beta }_3$; and $X_i$ is the individual specific covariates with alternative specific coefficients ${\gamma }_j$. Therefore, the above function can be determined as:

$$U_{ij}=V_{ij}+{ϵ}_{ij}$$

(2)

Here, $U_{ij}$ represents utility of energy type $j$ for households’ $i$ choice, $V_{ij}$ is observable covariates and unknown parameters to be estimated, and ${\epsilon }_{ij}$ is a random deviation which contains all the unobserved determinants of the utility.

Depending on households view on each alternative, the alternative $l$ is therefore chosen if ${ϵ}_j<\left(V_{ij}-\right.\left.V_{ij}\right)+{ϵ}_{ij}\forall j\ne l$ and if its components are distributed independently, identically distributed (iid) extreme value, then the probability of choosing this alternative can be determined as follows:

$$\begin{gathered} P_{ij}={\displaystyle \frac{\mathit{exp}\left(V_{ij}\right)}{\sum _{j=1}^J \mathit{exp}\left(X_{ij}\right)}} \\ ={\displaystyle \frac{\exp \left({\alpha }_j+{{\beta }_{1}Pr}_{ij}+{\beta }_2{Inc}_{ij}+{\beta }_3{Exp}_{ij}+{\gamma }_j{X}_i\right)}{\sum _{j=1}^J exp\left({\alpha }_j+{{\beta }_{1}Pr}_{ij}+{\beta }_2{Inc}_{ij}+{\beta }_3{Exp}_{ij}+{\gamma }_j{X}_i\right)}} \end{gathered}$$

(3)

2.6. Description of Variables

We investigated the factors driving household fuel choice by examining economic and socio-demographic variables, both with and without consideration of the resource periphery effect. Price and consumption expenditures in monetary terms for energy types such as NG, LPG, electricity, and biomass constitute the economic variables, with natural gas serving as the reference for primary cooking energy. Participants were surveyed, for example, regarding their total household expenditure on natural gas and substitute fuels, denominated in Mozambique meticais. The pricing data for each area was obtained directly from the retail establishments operating in those locations. Any households for which the energy expenditure could not be clearly determined were removed from consideration. Table 1 presents descriptions of outcome and explanatory variables, including sociodemographic and technical-related factors such as gender, age, aggregate size, incidents, and energy stacking.

Some variables in the models are expressed in natural logarithms to achieve normality. Summary statistics of the variables are shown in Table 2.

2.7. Limitations

The study is limited by its focus on understanding fuel choice among households connected to locally produced gas in the northern Inhambane province and Maputo city. Data from households without piped natural gas, as well as from other regions, would have allowed broad cross-comparisons of choice behavior, clean cooking access, and other pertinent factors. While acknowledging the cited limitations, it nonetheless makes a valuable contribution to understanding fuel choice behavior and access to clean cooking in Mozambique and other gas-rich countries.

3. Results and Discussions

The empirical analysis in the study begins with a descriptive analysis, and the extent to which households were influenced by a combination of factors was examined using a chi-squared test, leading to a multinomial logistic regression analysis.

3.1. Descriptive Statistics

3.1.1. Income and Accessibility

Table 3 shows respondents' income groups across rural, peri-urban, and urban areas, with 50.1% in vulnerable and semi-vulnerable groups (i.e., informal businesses, unemployed, and entrepreneurs). The remaining were a non-vulnerable group with a fixed income. Our findings support other studies showing that Mozambique’s labor force is characterized by vulnerable sources of income or a higher level of self-employed workers [118,119].

Furthermore, 91% of respondents thought that fuel prices were accessible or reasonable. Some respondents suggested that the fuel prices, natural gas in particular, should be cheaper because it belongs to the community or is produced locally (“não há razão de ser caro, porque o gás vem da nossa terra”), indicated a male resident in the Vilanculos district. Our interview captured another contested factor: most households residing in Inhambane province pay additional monthly fixed charges or “taxa fixa”, consisting of 318 and 3,000 Meticais (~4.89 and 46.15 $US), imposed by the operator on domestic and commercial consumers, respectively. For some households, these charges often exceed the actual monthly energy expenditure. These charges enable the operator to offset revenue shortfalls [120], especially in an environment with heavily subsidized fuel prices and no price revisions for over twenty years.

3.1.2. Fuel Stacking

Among 434 respondents, 56% switched their fuels and stoves (see Table 3), indicating a high prevalence of cooking fuel stacking, with respondents using additional fuels alongside their primary source. The main fuels were biomass (76.2%), electricity (7%), and LPG (3%). Respondents used electricity only for heating and microwaving food; biomass and PLG fueled both their cooking and heating needs. Refusal by the operator to extend gas lines to additional kitchen areas among MSMEs was the reason to consider LPG as a substitute for city gas. While some faced problems with their gas stoves or payment issues, others had visitors who needed to have their plans adjusted.

A strong rural and peri-urban bias was evident in the survey; 94% of respondents who stacked fuel lived outside urban centers, while only 6% lived in urban areas. Intriguingly, about 80% of participants who perceived the gas price as cheaper or reasonable staked the most, compared with 5% who thought fuel prices were high or very expensive.

Across all spending levels, respondents stacked their fuels. A direct relationship was also observed between the number of aggregates and total expenditure. A parallel trend was observed between aggregate counts and biomass stacking levels.

3.1.2. Security Concerns

Table 3 also indicates that only 18% of the participants in Inhambane and Maputo have security concerns about using gas as their primary cooking fuel. Several participants in the Inhassoro, Govuro, and Vilanculos districts reported that they often experience gas leaks in their neighborhoods, but these have not caused major incidents. However, in Maputo and Beira, a number of fatal LPG gas cylinder explosions have been reported [121,122].

Gas incidents and accidents have also been a concern for many households and residents around developed and developing countries. Because of human error, lack of inspection, and equipment and infrastructure in India, Pakistan, Ghana, and Kenya, it resulted in numerous explosions and human loss [123,124,125,126], while corroded pipes and leaks have been the main causes of explosions, building destruction, injuries, and deaths in the United States, China, Canada, and Great Britain [127,128,129,130]. Regulatory failure has been cited as the main contributor to the gas infrastructure and equipment explosions in Pakistan, and there have been calls for proper regulatory reforms, public awareness, and education [125], and cylinder innovation in Ghana [131] to reduce accidents and promote gas as the best alternative cooking fuel. When analyzing the total number of gas accidents in China, Pang et al. [132] found that from 2012 to 2023, the number of accidents was about 380 to 610 per month. Compared with piped gas, LPG had 43% more explosion-related accidents, 24% more fires, and 70% more asphyxiation and poisoning, while the former had 144% more leakages [132].

3.2. Multinomial Logistic Regression Analysis

The multinomial logistic regression models were used to analyze household fuel choices in Mozambique's energy periphery. The analysis considered four fuel categories: Electricity, NG, LPG, and Biomass, with NG as the reference fuel category.

The models were fitted both with and without accounting for the location effect. Table 4 presents the result of the multinomial logit model without the resource periphery effect. The coefficients indicate the direction of the effects. The probability of a household opting for a specific cooking fuel is interpreted by estimating the predicted probability for all households.

3.2.1. Income

In terms of socio-economic factors, the results revealed that monthly income enhances the use of energy-mix and has a significantly positive effect on fuel choice for all fuels at 1% and 5% levels of significance, indicating that the predicted odds of choosing electricity as an alternative to NG increased by 0.32% with an increase in monthly income (see Table 4). Similarly, for LPG increase by 0.48%, and for biomass, by 0.24%. This suggests that even wealthier households are likely to switch fuels, thereby supporting the energy stacking hypothesis. The results for LPG are similar to those reported by Masera [136] and to studies supporting the energy ladder hypothesis. The research emphasizes the significance of a household's socio-economic status in its decision-making process for cooking energy, particularly within the framework of energy stacking. The energy mix selection, with a substantial proportion of cleaner fuels, remains significantly influenced by household income. The findings reveal that Mozambican households are not consistently progressing up the energy ladder. Instead, some are regressing or reverting to less clean fuel sources, suggesting a reversal of the energy transition.

A country's general economic circumstances can affect changes in energy-related behaviors, with contributing factors such as income, unemployment, and energy prices playing a role [57].

3.2.2. Demographics

The coefficients for demographics are statistically significant at the 5% level for electricity and LPG but are insignificant for biomass (see Table 4). The results showed that, compared to peri-urban and rural areas, the predicted odds of preferring energy other than natural gas are higher by 25.5 for electricity and by 10.3 for LPG in urban areas. A possible explanation is that during the early phases of urban expansion, and given the availability of abundant biomass fuel around cities, urban dwellers predominantly use it for cooking, with little reliance on other sources, such as electricity and LPG [133].

The findings presented in Table 5 demonstrate the same scenario of the relationship between the choice of cooking fuel and its influencing factors, including the energy-rich periphery effect, using the MNL. Additionally, it indicated that the odds of choosing biomass (statistically significant at the 5% level) as an alternative cooking fuel to natural gas were 96.2% lower than in non-energy-rich peripheries. Domestic household fuels produced in Mozambique are relatively lower in the surrounding producing regions, which makes it accessible to lower-income households. The majority of oil- and gas-producing and exporting countries have relatively lower domestic fuel prices, mainly due to subsidies [134,135]. Another explanation may be alternative-specific benefits, such as efficiency. A 46-year-old respondent from Inhassoro indicated that “natural gas is faster than biomass, and because it is extracted here, it’s much cheaper than alternatives.” Furthermore, other respondents linked the use of NG and increased productivity: “You do not need to waste time fetching cooking fuel; you just need to turn on the valve, and you are ready to go”, a female primary school teacher in Vilanculos highlighted.

3.2.3. Education

At a 1% significance level, we found that years of education had a positive effect on cooking energy preference. The odds of choosing electricity as an alternative fuel increased by 45.3% (coefficient of 0.4335), and the odds of choosing LPG increased by 85.7% (coefficient of 0.6188) with a 1-year increase in education, controlling for other predictors (see Table 4 and Table 5). Consequently, households with higher qualifications demonstrate a better understanding of the advantages of cleaner fuels and find it easier to adopt them than less-qualified households, since a lack of technological awareness deters acceptance. This finding corroborates the results on modern fuel preferences [136,137].

In urban settings, a positive and significant correlation is observed between the adoption of clean energy sources (electricity and LPG) and education level. One possible explanation is that people pursuing higher education face a greater opportunity cost with modern fuels, which offer significant time savings [133]. Additionally, individuals with higher levels of education are increasingly aware of the detrimental health consequences associated with cooking with wood and charcoal [52].

3.2.4. Expenditure

The results in Table 4 and Table 5 demonstrate a statistically significant positive correlation between expenditure and all fuels (significance level < 0.01). The findings reveal that households, regardless of their socio-economic status, do not tend to alter their consumption in response to fuel prices. These findings are substantiated by results from Nigeria and Mexico [52,142], which assert that energy goods are a basic necessity.

3.2.5. Infrastructures and Energy Systems

A model excluding demographic and socioeconomic factors was analyzed to understand the effect of alternative-specific variables on decision-making. We found that natural gas-related incidents have a significant positive impact on switching to all fuel alternatives (Table 6). Many respondents indicated that they switched their primary cooking fuel (natural gas) due to incidents involving gas leaks, low gas pressure, and prepayment system issues. A retired male respondent in Vilanculos said that “sometimes, due to low gas pressure during the night period, we are obliged either to cook around 3:00 PM, before everyone is back from work, or switch to other alternatives.” Furthermore, an informal businesswoman in a female-headed household in the Costa de Sol neighborhood in Maputo shared that “the prepayment system is the key reason for temporary disconnection from natural gas and switching to other alternatives”.

Challenges to the current energy infrastructure stem from rapid urban expansion, population growth, and increased energy demand, as the network was originally designed in the 1990s for a smaller consumer base. Additionally, gas transmission networks are susceptible to extreme natural disasters, third-party actions, and operational failures, which risk their integrity and safety [138,139,140]. Thus, to ensure the efficient distribution of natural gas to end-users, adding or upgrading compressor stations in certain areas can help maintain adequate pressure and flow rates throughout the pipeline system. Routine inspections and dynamic maintenance, intelligent monitoring, automation solutions [140,141], servicing of pipeline networks and related systems to ensure safety and sustained operational efficiency.

The factors of gender, household age, and size seem to play little or no role in shaping cooking fuel preferences both within and outside resource peripheries.

3.2.6. Diagnostic Tests

We tested the model for the Property of Independence from Irrelevant Alternatives (IIA) using the Hausman-McFadden IIA test [142]. We reject the null hypothesis of the choices being independent (see Table 7). Although the model exhibits the IIA property, households in this study opted for multiple cooking fuel choices, which may have violated the property or been unrealistic.

4. Conclusions and Policy Implications

This paper analyzed the determinants of households’ cooking energy decision-making in and out of energy resource-rich regions of Mozambique, employing a logistic regression model based on data from 434 households connected to domestically produced natural gas. In addition, the interviewees' remarks validate our empirical findings on coping mechanisms for accessing clean cooking fuel in developing countries. Based on our predictions, we can draw the following conclusions: First, we found that both supply and demand factors shape household fuel choices. On the demand side, variables such as income, spending, education, and geographic location, along with proximity to energy exploitation regions, show a significant correlation with decision-making. On the supply side, we've identified energy systems and related infrastructures as the key drivers of household energy selection. Second, price, gender, age, and household size are not significant determinants of decision-making.

There has been little prior knowledge of household energy transitions in Mozambique's gas-rich peripheries, and this paper makes a useful contribution to filling this knowledge gap. Mozambique can learn valuable policy lessons from nations that have successfully promoted the adoption of clean cooking. While formulating strategies and policies for the promotion of natural gas and other transition cooking fuels, it is recommended that elements of the supply and demand sides, such as socioeconomic, environmental, cultural, demographic, and infrastructural factors, be taken into consideration. For instance, the Kenyan Government [148] identified several key behavioral aspects and communication strategies to boost clean cooking adoption. Among others, it included branding; behavior change strategy; media advocacy to enhance public awareness and understanding of clean cooking; partnerships; special and dedicated events; and the involvement of private sector/industry players in the process. Central government and institutional commitment, subsidies, financial donor support, infrastructure readiness, effective policies, campaigns, and communication strategies, and public response were among the factors contributing to the successful implementation of clean cooking in many developing countries [143,144,145]. Concurrently, households shall be encouraged to transition to sustainable cooking fuels, moving from inefficient, traditional energy to modern, cleaner energy sources by addressing the factors influencing their energy choices. Successfully implementing this strategy could lead to greater public trust and broader adoption of modern energy alternatives in transport and industry. This would result in reduced reliance on forest biomass for cooking energy, better public health, increased productivity, and rapid progress towards the SDG-7 objectives.