energies Article Cooking Fuel Usage in Sub-Saharan Urban Households Ting Meng 1 , Wojciech J. Florkowski 2,* , Daniel B. Sarpong 3, Manjeet Chinnan 2 and Anna V. A. Resurreccion 2 1 Academy of Global Food Economics and Policy, Beijing Food Safety and Strategy Research Base, College of Economics and Management, China Agricultural University, Beijing 100083, China; tmeng@cau.edu.cn 2 Department of Food Science and Technology, University of Georgia, Griffin, GA 30223-1797, USA; manjeet.chinnan@gmail.com (M.C.); annaresurreccion@gmail.com (A.V.A.R.) 3 Department of Agricultural Economics and Agribusiness, University of Ghana, Legon, Accra P.O. Box LG 25, Ghana; dbsarpong@ug.edu.gh * Correspondence: wojciech@uga.edu; Tel.: +1-770-228-7233 Abstract: This study models the frequency use of wood, charcoal, liquid gas, electricity, and kerosene in urban households in Ghana and supplements the literature on cooking fuel choices. The modeling is based on survey data collected in several major Ghanaian cities. Survey results indicate that charcoal and liquid gas are frequently used in meal preparation, while the frequency use of firewood, kerosene, and electricity is limited. Frequency use is estimated using the ordered probit technique. Five cooking fuel use equations identify income, socio-demographic characteristics, and location of urban residents as influencing the frequency use. Statistically significant effects measure probability changes in each of the four fuel categories. Income and education increase the probability of often or very often of using liquid gas or electricity to cook. The effect of being employed by the government is similar but less consistent. Age, household size, and marital status are linked to frequency use,   but differently affect specific fuels. As the number of children or adults increases in a household, Citation: Meng, T.; Florkowski, W.J.; so does the probability of using firewood or charcoal, but this also increases the probability that Sarpong, D.B.; Chinnan, M.; such households never use liquid gas or electricity for cooking. Regional differences indicate Tamale Resurreccion, A.V.A. Cooking Fuel residents heavily rely on wood and charcoal, and infrequently use liquid gas or electricity. Multiple Usage in Sub-Saharan Urban cooking fuel use behavior may reflect risk aversion to fuel shortages. Increasing incomes and Households. Energies 2021, 14, 4629. improving education will drive the probability of an increased use of cleaner cooking fuels and https://doi.org/10.3390/en14154629 decreased use of fuel mixes, benefiting meal preparers’ health and the environment. Academic Editors: Xiaolin Wang and Keywords: cooking fuel choice; survey; ordered probit; probability change; income; location keyword Firoz Alam Received: 21 June 2021 Accepted: 27 July 2021 1. Introduction Published: 30 July 2021 Daily cooking is indispensable for sustaining millions of people and in many countries Publisher’s Note: MDPI stays neutral cooking fuels represent a major portion of utilized energy. Cooking with modern fuels with regard to jurisdictional claims in allows labor and natural resources to be reallocated from fuel collection and production published maps and institutional affil- towards income-generating purposes [1,2], while significantly reducing time spent on iations. cooking. In sub-Saharan Africa, 80% of the population cooks with solid fuels (wood, charcoal, or coal) as compared to 58% in China and 71% in India [3]. However, in 2000, 25% of the population in that part of Africa had access to electricity [4], increasing to 44% in 2017 [5]. The fast-expanding access to electricity obscures the prevalent use of charcoal in Africa [6]. The number of people depending on solid fuels is expected to substantially Copyright: © 2021 by the authors. increase in sub-Saharan Africa [7]. After Nigeria, Ghana is the second leading charcoal Licensee MDPI, Basel, Switzerland. This article is an open access article producer [8]. Two common fuels, kerosene and gas (primarily liquid gas), are used by only distributed under the terms and 7–10% [3,9] and 4% of the sub-Saharan population, respectively [3], but in Ghana the use conditions of the Creative Commons of liquid gas has quickly increased, especially in urban areas [10]. Attribution (CC BY) license (https:// Household members, especially women, spend a large amount of time on fuel col- creativecommons.org/licenses/by/ lection and cooking using traditional fuels [2,11]. Moreover, high levels of indoor air 4.0/). pollution from burning biomass for cooking pose a health risk [12] and have been listed Energies 2021, 14, 4629. https://doi.org/10.3390/en14154629 https://www.mdpi.com/journal/energies Energies 2021, 14, 4629 2 of 18 among the top 10 risks damaging human health [13], especially for children. Indoor air pollution from biomass burning is responsible for 4.2 child deaths per 1000 population due to pneumonia [3,14]. About 3.5 million people died prematurely in 2010 due to indoor air pollution resulting from firewood, charcoal, and other biomass combustion [15]. In Ghana, women are the primary meal preparers and they bear a lot of the impact from cooking fuel emissions [16]. Biomass use in cooking appears to result in environmental problems as well. Soil and land degradation are related to the usage of biomass, particularly around densely populated areas. In the dry savannah zone of northern Ghana, charcoal production exacerbates deforestation [17,18]. The average annual deforestation in Ghana was estimated at 0.3% between 2000 and 2015 [19]. Biomass burning contributes to black carbon [20,21] and greenhouse gas (GHG) emissions [3,22]. In 2015, Ghana’s population was exposed to toxic particulate matter P2.5 at a level exceeding World Health Organization standards [19]. Both carbon and gas emissions are aggravated by inferior cooking equipment [16] despite progress in that area [23,24]. The multiple negative effects of traditional biomass fuel use suggest that it is desirable to accelerate the shift from various types of biomass energy to modern cooking fuels. Mauritius, Republic of South Africa, and Ghana were among the leading countries achieving considerable progress in increasing access to modern energy services for their citizens [25]. In Ghana, electricity demand is growing 6–7% annually [16] and electricity use for cooking increased in urban areas between 2005 and 2013 [26]. Promotion of modern cooking fuels may not induce the complete abandonment of traditional fuel usage. In fact, a growing body of evidence indicates that using multiple cooking fuels simultaneously is a common phenomenon among households in developing countries (for example, [6]). A predominant cooking fuel combination in Guatemala is firewood and liquid petroleum gas, used by 26% of urban households, while 52% of rural households in Vietnam often use wood complemented by straw, and 34% of households in rural South Africa use both firewood and kerosene for cooking [1]. The effect of factors determining energy choice decisions was quantified in two separate studies conducted in the Republic of South Africa and Burkina Faso [27,28]. The widespread use of multiple fuel types in the developing world was reflected in previous studies’ focus on the household’s main cooking fuel choice without capturing the behavior of households’ multiple fuel frequency use. The objective of the current study is on cooking fuel usage by modeling the frequency of five alternative fuel types. The focus on frequency of use of different cooking fuels is a novel and supplements previous studies on the choice of cooking fuel filling the gap in empirical research. The study uses data obtained from a survey conducted in urban areas of Ghana. During the survey, households shared details about frequency of the most common fuels used for cooking the main meal of the day, i.e., wood, charcoal, liquid gas, electricity, and kerosene. Cooking fuel use cannot be underestimated because it affects the daily life of virtually all households, but previous research focuses on fuel use by other sectors. The choice of cooking fuel in Ghana and sub-Saharan Africa, by tradition, relates to the choice of the dish prepared for a particular meal because it apparently affects the meal’s taste. A household makes a deliberate choice of fuel and selects the one that is most appropriate for preparation of a particular type of food or dish. This study contributes to research on household behavior regarding cooking fuel use and examines fuel usage frequency instead of solely exploring household decisions of cooking fuel choice. Knowledge of frequency supplies detailed and generally missing information about household behavior in cooking fuel use. Additionally, by examining multiple fuel use in cooking in various areas of the country, results of the study recognize the relevance of subtle regional differences in cooking fuel use and possible implications for both human and environmental health. Governments, policy-makers, and NGOs can use results to improve the assessment of different cooking fuel usage in other countries in West Africa. Furthermore, insights gained about household constraints in using modern cooking fuels help to modify energy strategies. Such policies may be more effective by recognizing that, ultimately, the urban Energies 2021, 14, 4629 3 of 18 household transition from traditional biomass cooking fuels to a mix of cooking fuels is conditioned by user, household, and location-specific factors. The remainder of the paper includes the description of the data and estimation ap- proach in the section Materials and Methods and is followed by the section Results. The latter section includes the description of the summary survey results and the empirical estimation results describe the equations modeling the frequency of using each cooking fuel. Next, the section “Discussion” elaborates on the statistically significant factors influencing the frequency of cooking fuel use, while the last section presents conclusions. 2. Cooking Fuel Use in the Literature Earlier studies have established that socio-demographic factors, situational factors, and psychological factors affect household cooking fuel usage, especially in developing countries. Socio-demographic factors, including age, gender, education, employment status, household type, and location have a significant influence on cooking fuel usage. Education, household wealth and socio-economic status are well recognized predictors of adoption and sustained use of clean modern fuels [29]. Sharma (2018) [30] lists the determinants of household cooking energy demand as literacy status, landholding size and household size. Cleaner energy is more likely to be used in households where the head of the family has a higher level of education and income, whereas larger households and households with a higher proportion of dependent members, and an older person or female as head of the family, are less likely to use cleaner energy [31]. Lower education levels of the household and private house ownership are linked to the consumption of fuels such as firewood and kerosene [32]. Moreover, residence in urban areas positively affects choice of liquefied petroleum-gas as compared to traditional fuels [33]. Current situational factors such as presence and availability of cleaner and modern energy alternatives, and associated financial supports increase the adoption of cleaner energy. The presence of alternative fuels that are less polluting and more convenient increases the adoption of clean energy [30]. Access to renewable energy leads to a preferred selection of modern, clean fuels [32]. Availability of electricity positively affects choice of liquefied petroleum-gas as compared to traditional fuels [33]. Moreover, the lack of modern energy technology options and the lack of financing and other support decrease the use of clean energy [34]. Finally, psychological factors (e.g., values, motivation, vulnerability) also influence cooking fuel adoption among households. The lack of motivation and the pressure for switching over to cleaner facilities has impeded the transition towards clean cooking equipment [34]. Perception of benefits of “easy cooking” and “saving time and money” promotes the adoption of biogas [35]. Moreover, attitude toward technology and awareness of the health risks of using traditional cooking fuel change energy adoption behavior [36]. The choice of one cooking fuel does not completely eliminate the dependence on the use of other cooking fuels, resulting the use of multiple fuels [37]. Consequently, the important question is what factors influence the frequency of a given cooking fuel, which is the focus of the current study. 3. Materials and Methods Household behavior regarding cooking fuel use in sub-Saharan Africa differs from fuel use in developed countries. Cultural preferences and lifestyles condition the choice of cooking fuel in daily main meal preparation. It is not unusual for a household to own several cooking stoves, each using a different type of fuel. The frequency of a cooking method and the concomitant cooking fuel choice is driven by the frequency of preparing a specific type of a dish. Additionally, disruptions in the supply of modern fuel types encourage having an alternative cooking stove using other fuels to assure food preparation. Multigenerational households where more than one person participates in cooking may have multiple cooking stoves and fuels meeting preferences and skills of various preparers. Thus, the frequency of cooking fuel use suggests the empirical approach that permits Energies 2021, 14, 4629 4 of 18 modeling the frequency of fuel use. The estimation of cooking fuel choice requires data that are not readily available and the current study applies data gathered through a household survey. 3.1. Data The study uses survey data collected in three large cities in Ghana (Accra, Takoradi, and Tamale) in 2011. The surveyed cities are located in the northern and southern parts of the country and in two different ecological zones, i.e., Takoradi and Accra are located in the coastal zone of southern Ghana, and Tamale, capital of the Northern Region, is located in the dry savannah zone. The share of urban population in Ghana exceeds 50% [38]. Cost limitations and logistics as well as the concentrated use of cooking fuel restricted the scope of the survey to major urban areas. The drafted structured questionnaire consisted of sections collecting information needed for the study. Questions probed respondents about their usage frequency of the most common cooking fuels, i.e., wood, charcoal, liquid gas, electricity, and kerosene. The reported usage frequency of each fuel was measured on a scale from one to four with an increasing number indicating its more frequent use (i.e., 1 = not used at all, 2 = seldom, 3 = often, 4 = used very often). A separate set of questions also probed commonly sought (in national census surveys) personal information and household characteristics such as age, gender, occupation, income, and household composition. Trained enumerators using a pre-tested questionnaire conducted the face-to-face interviews. Among residents of Tamale, enumerators involved in the national statistical panel data collection, who were fluent in Dagbani, a regional language, conducted the interviews. Among residents of Takoradi and Accra, the majority speaks English and the interviews were mostly conducted in that language. Neither language nor the formulation of questions in the survey instrument posed difficulties to respondents during the pilot stage of the survey. A typical interview lasted about 35–45 min. Following the collection of the questionnaires, the information was entered into the dataset. A total of 1010 households were participating in the survey. However, the number of fully completed questionnaires was less and varied with regard to the questions about the five cooking fuels. As a result, the estimation of each equation of cooking fuel use frequency differed and ranged from 1006 in the case of charcoal to 940 for the use of kerosene. 3.2. Estimation Approach The study applies the ordered probit regression model to investigate how the socio- demographic factors affect an urban household’s usage frequency of certain cooking fuels. Social science research commonly uses ordinal numbers to measure and quantify phenomena transformed into variables. The ordered probit model is common in empirical studies applying the ordered categorical variable. In the basic framework of the regression model (Equation (1)) Y* is the latent variable behind the fuel use frequency, X denotes the selected explanatory variable vector, B is the coefficient vector, and e is the error term, which is assumed to follow normal distribution: Y∗ = XB + e (1) where Y* is the latent variable depicting the frequency of using a specific cooking fuel, X is the vector of the explanatory variables, including in this study socioeconomic, demographic, and location features, B is the vector of the coefficients to be estimated, and e is a vector of error terms. The relation between the latent variable Y* and the dependent variable Y is defined in Equation (2): Y = i, i f cuti−1 < Y∗ < cuti, where i = 1, 2, 3, 4 (2) when the latent variable is between particular cut points [39] and i indicates one of the four categories of using the cooking fuel associated with the use of the ordered probit, i.e., not Energies 2021, 14, 4629 5 of 18 at all, seldom, often, and very often (see also the description in the Section 3.1 Data). The dependent variable is equal to a certain ordinal level, where Cut’s are parameters to be estimated assuming Cuti−1 < Cuti (because of convenience in the estimated model, Cut0 and Cut4 are used to denote negative infinite and infinite) [39]. The probability of fuel usage frequencies can be expressed as the difference between two Cumulative Distribution Functions (CDFs) of normal distribution (Equation (3)): Prob(Y = i) = Prob(cuti−1 < Y∗ < cuti) = Prob(cuti−1 − XB < e < cuti − XB) (3) = F(cuti − XB)− F(cuti−1 − XB) where F() is the CDFs of normal distribution, and Cut’s and B are coefficients to be estimated. In each equation, the likelihood function of the empirical model (Equation (4)) is the product of all possible probabilities with the indicator variable d as corresponding power, and N is the total sample size, Likelihood = ∏ ∏ Prob(Y = i)d(Y=i) where j = 1, 2, . . . , N, d = 1 i f Y = i; d = 0, otherwise (4) j i In Equation (4), j indicates the jth respondent, i indicates the ith category of fuel usage frequency referring to a specific cooking fuel, and d = 1 only if a certain category is selected. For example, if there are n1, n2, n3, and n4 respondents who choose the use of liquid gas with the frequency “not at all”, “seldom”, “often”, and “very often”, then the likelihood function for the spe[cific case in this study is: [ Likelihood = Prob(Y = 1)]n1[Prob(Y = 2)]n2 Prob(Y = 3)]n3[Prob(Y = 4)]n4. (5) The ordered probit coefficients lack direct interpretation that generates practical knowl- edge until the coefficients are converted into probability changes. In particular, probability change measures an increase or a decrease of using one of the five cooking fuels with the frequency captured by the four-step scale. The conversion allows measuring specific probability changes in response to changes in the explanatory variables. The explanatory variables are shaped by the unobservable preferences reflected in identifiable socio-demographic characteristics. Thus, both consumer theory and previous studies about cooking fuel choices guide the explanatory variable selection in the current study. Evidence shows that selected socioeconomic and demographic factors significantly affect household cooking fuel usage. Several studies closely link income growth with changing household fuels [28,40,41]. Other studies indicate that households of respondents with high educational attainment levels tend to use modern fuels which offer significant time savings, especially for women [1]. Demographic characteristics are also relevant because the household energy type selection can be captured by its size and composi- tion [1,28]. Finally, given that Ghana encompasses several different ecozones, the location of a household in a particular urban area is explicitly included by adding a place indicator. 4. Results 4.1. Survey Results The usage frequencies regarding each cooking fuel type are displayed on Figure 1. Regarding traditional biomass fuels, only about 20.9% of the surveyed households still use wood, and among the users of wood for fuel, one half of them seldom use wood only. Wood generates smoke and, compared to other cooking fuels considered in the study, is inconvenient to handle. About 86.6% of the surveyed households use charcoal seldom, often, or very often. The popularity of charcoal among urban sub-Saharan African households has been estimated at above 80% [42]. Charcoal is characterized by high energy content and clean burning as compared to wood. The use of charcoal as the primary fuel for cooking among urban households in Ghana has been estimated at about 53% [43]. In Tamale, charcoal has been reported as the main cooking fuel for about 50% of the residents Energies 2021, 14, x FOR PEER REVIEW 6 of 19 inconvenient to handle. About 86.6% of the surveyed households use charcoal seldom, often, or very often. The popularity of charcoal among urban sub-Saharan African house- Energies 2021, 14, 4629 holds has been estimated at above 80% [42]. Charcoal is characterized by high energ6yo cf o18n- tent and clean burning as compared to wood. The use of charcoal as the primary fuel for cooking among urban households in Ghana has been estimated at about 53% [43]. In Ta- male, charcoal has been reported as the main cooking fuel for about 50% of the residents anandd4 455%%i ninG GreraetaetrerA Acccrcar,a[, 4[44]4.].T Thheec ucurrrernent ts tsutuddyya alslosor erpepoortrstst htheer erlealtaitvieveu uses,ea, nandda baobuout t 60% of respondents indicate they use charcoal for cooking their main meal of the day often 60% of respondents indicate they use charcoal for cooking their main meal of the day often or very often (Table 1). or very often (Table 1). FiFgiugurere1 .1T. Thheeu useseo fofif vfievefu feuleslsin inc ocokoiknigngth tehem maianinm meaelaol fotfh tehed adyayb ybyfo fuorufrr ferqeuqeunecnycyc actaetgeogroireisesin in uurbrabnanh ohuousesheohlodlds sin inG Ghahnaan,ai,n i%n%(n (n= =9 5965)6.).S Soouurcrec:e:S Suummmmaaryryr erespspoonnsessesf rforommt htheeh hoouuseshehooldlds usurvrveye.y. TaTbalbele1 .1D. Desecsrcirpitpivtievest sattaistitsictiscos fovf avrairaibalbelseisn icnlculdueddedin inth tehee mempipriicriaclaml modoedle. l. VariabVlea rNiaabmlee Name VariabVlea DriaebsclreipDteioscnr/iUpntiiotns /oUf nMitesaosfuMreemaesnutr ement MMeaenan StSdtdD Devev Dependent variables 1 FDreepqeunednenctyv oafr icaobles 1 oking How often do you use wood to cook your main meal of the day? 1 = not at all; Frequency of cooking with wood How often do you use wood to cook your main meal of the 1.365 0.807 with wood 2 = seldom; 3 = often; 4 = very often day? 1 = not at all; 2 = seldom; 3 = often; 4 = very often 1.365 0.807 Frequency of cooking How often do yoHuo uwseo cfhteanrcdooaly toou cuosoek cyhoaurrc omalatino cmoeoakl yoof uthrem daaiyn?m 1e =a lnFreq o oft at 2.779 1.037 withu cehnacrycooafl cooking waitlhl; c2h =a rsceoldalom; 3t h= eofdtaeyn?; 41 == vneorty aotfatelln; 2 = seldom; 3 = often; 4 = very often 2.779 1.037 Frequency of cooking How often do yoHuo uwseo lfitqeunidd ogayso utou csoeolkiq yuoidurg masation cmooeakl yoofu thr em daaFr in y?m 1e =a lnooft 2.848 1.345 wietqhu lieqnuciydo gfacso oking waitth allilq; u2 i=d sgealdsomt;h 3e =d oafyt?en1; =4 n= ovtearty aolfl;te2n= seldom; 3 = often; 4 = very often 2.848 1.345 FFrreeqquueennccyy ooff ccooookkiinngg wHithowel eocfttreinci tdyo yoHuo uwseo eftlenctrdiocityyo utou csoeoekl eycoturirc imtyation cmooeakl yoof uthr em daainy?m 1e =a lnooft the day? 1 = not at all; 2 = seldom; 3 = often; 4 = very often 1.117.5175 0.505.5353 with electricity at all; 2 = seldom; 3 = often; 4 = very often FFrreeqquueennccyy ooff ccooookkiinngg wHithowke oroftseenn edo yoHuo uwseo kfteernosdeoney otou cuosoekk yeorousre mneatino cmoeoakl yoof uthrem daaiyn?m 1e =a lnooft at 1.216.0260 0.604.6949 with kerosene all; 2 = seldom; 3t h= eofdtaeyn?; 41 == vneorty aotfatelln; 2 = seldom; 3 = often; 4 = very often Independent variables Independent variables Demographic factors Married Demographic facto=r1s if a respondent is married 0.753 0.431 MAgaerried =1 if a respondenAt cistu malaarrgieedin years 39.02.27253 10.06.54631 AChgiel dren Actual age in yeaNrsu mber of household members between 13–18 years old 03.998.2322 11.200.6556 CAhdiuldltren Number of houseNhuomldb mereomfbheoruss beehtowldeemn e1m3–b1e8r sybeaetrws oeledn 19–60 years old 2.008.7983 1.715.2105 AElduelrt Number of houseNhuomldb mereomfbheoruss beehtowldeemn e1m9–b6e0r sy6e1aryse oalrds old or older 0.125.3087 0.510.7551 Elder Number of houseShocoilod-e mcoenmombiecrfsa 6ct1o rysears old or older 0.153 0.505 Income Socio-economic facHtoorus sehold income in the month preceding the survey/in 646.070 785.081 Income Household incomGeh iann athiaen mcoednitsh preceding the survey/in Ghanaian cedis 646.070 785.081 Self employed =1 if a respondent is self-employed 0.642 0.480 Government or civil employed =1 if a respondent is gov/civil employee 0.243 0.429 =1 if a respondent has a secondary education (including Secondary Education Senior high/GCE O-A level, Vocational school, Technical 0.382 0.486 school, Teacher training) Energies 2021, 14, 4629 7 of 18 Table 1. Cont. Variable Name Variable Description/Units of Measurement Mean Std Dev =1 if a respondent has a college education (including College Education Senior high/GCE O-A level, Vocational school, Technical 0.134 0.340 school, Teacher training, University, or postgraduate) Location Tamale =1 if a household is in Tamale 0.186 0.389 Takoradi =1 if a household is in Takoradi 0.208 0.406 1 See the distribution of frequency use in Table 2 Table 2. The use of five fuels in cooking the main meal of the day by four frequency categories in urban households in Ghana, in%. Fuel Not at All Seldom Often Very Often Tamale Wood (n = 179) 54.75 17.88 9.50 17.88 Charcoal (n = 185) 8.65 20.00 23.78 47.57 Liquid gas (n = 177) 61.02 3.95 6.21 28.81 Electricity (n = 172) 98.26 0 1.16 0.58 Kerosene (n = 172) 89.53 6.98 0.58 2.91 Takoradi Wood (n = 207) 86.47 8.21 2.90 2.42 Charcoal (n = 210) 13.81 33.81 17.62 34.76 Liquid gas (n = 209) 25.84 7.18 7.18 59.81 Electricity (n = 207) 89.86 4.83 3.38 1.93 Kerosene (n = 206) 80.58 14.56 2.91 1.94 Accra Wood (n = 570) 86.32 8.77 3.33 1.58 Charcoal (n = 608) 14.80 28.29 32.57 24.34 Liquid gas (n = 603) 22.89 4.15 13.93 59.04 Electricity (n = 562) 84.34 9.79 3.74 2.14 Kerosene (n = 562) 81.14 11.39 4.8 2.67 About two thirds of the respondents, 63.9%, use liquid gas often or very often for cooking the main meal of the day (Table 1). This share is much higher than the national average, which indicates 9.5% of all households used liquid gas for cooking in 2008. The expansion of liquid gas use among urban households may reflect the transition to modern cooking fuels as part of Ghana’s energy policy and reflects the government subsidy program (despite periodic shortages in the liquid gas supply). According to the survey results, only a small number of households cook with kerosene or electricity, despite how easy and safe it is to use either cooking fuel (though relative prices favor kerosene use over electricity). Electricity is used by about 11.2% and kerosene by 17.3% of respondents at least occasionally, and the shares correspond to 14% for electricity and 20% for petroleum in the country’s total annual energy consumption. Among responding users of either of the two fuel types, only 1.7% and 2.6% report using kerosene or electricity, respectively, very often to cook the main meal of the day. The survey results are summed by displaying the definition, measurement units, and descriptive statistics (the mean and standard deviation) of factors used as variables in the empirical analysis (Table 2). In the sample, 60.6% of survey participants are from Accra, 20.8% from Takoradi, and the remaining 18.6% from Tamale. The average respondent is 39.2 years old. Among surveyed households, three out of four are households of married respondents. Married households are more likely to cook meals because cooking is expected as a part of division of tasks in a household. In addition, the typical household has one teenage household member (between 13 and 18 years old), two adult members (between 19 and 60 years old), and 0.15 elder members (more than 61 years old). The mean household income in the month preceding the survey is 646.1 cedi (USD 1 = 1.49989 Energies 2021, 14, 4629 8 of 18 Ghanaian cedi, 1 January 2011 [34]). In terms of employment status among respondents, 64.2% report being self-employed, 24.3% are government or civil employees, and 11.5% are either not employed, students, or retired. In terms of educational attainment level, 38.2% of respondents have a secondary education and 13.4% have a college education. In addition to the statistics summary for the entire sample, consumer profiles linked to each cooking fuel are provided in Table 3. Because the use of multiple cooking fuels is a common phenomenon among households and the focus on the frequency of use in the current study, the described consumer profiles associated with the use of each cooking fuel refer to the frequency categories “often” or “very often”. The frequent use of firewood in cooking characterizes households that tend to be larger, having more children and adults, including elderly members (Table 3). Additionally, frequent users of firewood seldom have a college education, their households report lower incomes, and are more likely to be located in Tamale. Table 3. Consumer profiles based on the survey results associated with the frequency use of each cooking fuel “often” or “very often”. Characteristic Wood Charcoal Liquid Gas Electricity Kerosene Demographic Factors Married 0.79 0.71 0.79 0.84 0.76 Age 43.05 39.34 39.18 43.69 42.70 Children 1.51 1.03 0.90 1.22 1.14 Adult 2.74 2.17 2.05 2.61 2.50 Elder 0.40 0.13 0.13 1.15 0.19 Socio-economic factors Income 617.07 432.93 845.99 1116.86 896.47 Self employed 0.65 0.74 0.57 0.60 0.69 Government employed 0.22 0.10 0.34 0.29 0.20 Secondary Education 0.38 0.31 0.45 0.43 0.37 College Education 0.03 0.03 0.20 0.13 0.06 Location Tamale 0.41 0.29 0.11 0.20 0.22 Takoradi 0.08 0.23 0.23 0.08 0.09 N 100 316 553 86 94 Households reported using charcoal have the lowest average income and are likely lo- cated in Tamale or Takoradi as compared to households using other cooking fuels (Table 3). Respondents from those households are likely self-employee, seldom include government employees, or have college education. In addition, respondents using liquid gas “often” or “very often” are, on average, the youngest and have the smallest number of adults and children as compared to the frequent users of other cooking fuels. The respondents from such households are more likely to be college-educated and employed by the government as compared to households using often other types of cooking fuels. Households using electricity “often” or “very often” are among those with the highest incomes, their average age is the highest, and the largest number of the elderly. Respon- dents from those households tend to be well educated, and many of them have a secondary or college education and many are employed by the government (Table 3). Households using kerosene to cook “often” or “very often” have incomes higher than users of all other fuels except electricity. Additionally, those households are relatively large, tend to have older members, and include a sizable share of Tamale respondents. Those respondents likely represent the self-employed and those with a secondary education (Table 3). 4.2. Empirical Model Estimation Results Estimation results of modeling the frequency use of the five cooking fuels are reported in Table 4. The system of equations performs well, as indicated by the likelihood test value significant at p < 0.0001. The need for practical interpretation of results required Energies 2021, 14, 4629 9 of 18 that the estimated ordered-probit coefficient be converted into a change in probability in the specific fuel frequency use in response to a unit change in an explanatory variable. Figure 2a–e show the probability changes for each frequency category, i.e., not using the specific cooking fuel at all, using it seldom, often, or very often. Because the changes in probability are of primary interest in learning about the factors relevant to various cooking fuel use frequency, the description of the results focuses only on the marginal effects (or effects in the case of binary explanatory variables) that are statistically significant. Table 4. Ordered probit estimation results of the cooking fuel use frequency in urban households of Ghana. Variable Name Wood Charcoal Liquid Gas Electricity Kerosenen = 956 n = 1003 n = 989 n = 941 n = 940 Demographic factors Married 0.1895 −0.0260 0.2432 ** 0.3993 *** −0.1523(0.1244) (0.0831) (0.0952) (0.1470) (0.1113) Age 0.0042 0.0013 −0.0013 0.0117 ** 0.0094 *(0.0050) (0.0036) (0.0041) (0.0058) (0.0048) Children 0.1188 *** 0.0458 −0.0356 −0.1171 ** 0.0090(0.0396) (0.0306) (0.0351) (0.0591) (0.0410) Adult 0.1782 *** 0.0434 ** −0.0406 0.0609 0.0900 ***(0.0279) (0.0218) (0.0265) (0.0384) (0.0280) Elder 0.3274 *** −0.0312 −0.0876 −0.3248 ** −0.0701(0.0900) (0.0685) (0.0811) (0.1637) (0.1039) Socio−economic factors Income(100 cedi) −0.0179 ** −0.01352 *** 0.07495 *** 0.0126 * 0.01137 *(0.0082) (0.0050) (0.0111) (0.0072) (0.0066) Employ_self −0.1225 −0.1662 0.0022 −0.6144 *** −0.2352(0.1414) (0.1140) (0.1288) (0.1704) (0.1441) Employ_gov −0.4879 *** −0.4936 *** 0.5376 *** −0.3169 * −0.0797(0.1999) (0.1335) (0.1571) (0.1921) (0.1749) Educ_sec −0.3897 *** −0.2873 *** 0.5085 *** 0.4430 *** −0.1294(0.1196) (0.0816) (0.0927) (0.1426) (0.1127) Educ_col −0.9952 *** −0.7488 *** 0.5666 *** 0.8895 *** −0.4550 **(0.2760) (0.1306) (0.1652) (0.1897) (0.1962) Location Tamale 0.7463 *** 0.3226 *** −0.6749 *** −0.9639 *** −0.3922 ***(0.1292) (0.1028) (0.1229) (0.2555) (0.1522) Takoradi −0.1196 0.0343 0.0666 −0.1854 −0.0315(0.1435) (0.0930) (0.1081) (0.1482) (0.1262) Parameters Cut1 1.4452 −1.5130 0.1049 1.8263 1.1277(0.2704) (0.2010) (0.2250) (0.3134) (0.2644) Cut2 2.0867 −0.5343 0.2783 2.3704 1.7650(0.2761) (0.1983) (0.2253) (0.3187) (0.2686) Cut3 2.5563 0.2612 0.6420 2.8728 2.185371(0.2841) (0.1970) (0.2258) (0.2873) (0.2739) Log Likelihood −532.60 −1265.23 −914.25 −391.37 −559.06 Likelihood test 260.20 167.19 316.52 118.00 43.89 (p-value) with DF = 12 (<0.001) (<0.001) (<0.001) (<0.001) (<0.001) Note: Standard errors are in parentheses. * indicates the significance at 10% level. ** indicates the significance at 5% level. *** indicates the significance at 1% level. 4.2.1. Wood The marginal effects of using wood as a cooking fuel (Figure 2a) show that eight factors significantly influenced wood use frequency. Each additional teenage child in the household lowered the probability of never using wood by 2.7%. Such an addition increased the use of firewood, but with decreasing effect. One more teenager raised the infrequent use of firewood by 1.7%, but the effect was negligible for more frequent use categories. A similar pattern of probability changes on the use of wood for cooking was Energies 2021, 14, 4629 10 of 18 Energies 2021, 14, x FOR PEER REVIEW 9 of 19 established when the number of adults aged 19–60 and elderly (61 years old or older) increases. Adding one person in each age group increased the “seldom” use of wood by 2.4% antden4d.3 %to, hreasvpee cotlidveerly m(Femiguberers2, aa)n. d include a sizable share of Tamale respondents. Those Incrreesapsoinngdiennctosm liekerlayis reesptrheesepnrto tbhaeb sileitlyf-eomf “pnleovyeerdu asnindg t”howsoeo wdiitnh cao soekcionngdaanryd elodwucearstion (Ta- the probblaeb 3il)i.t y of using it with any of the other three frequency categories (Figure 2, panel a), but the induced change is very small. Only a substantial income increase could be expe4c.t2e.d Etmoplioriwcaelr Mthoedeul sEestoimf fiatrieown oRoesdu.ltBs eing a government employee increased the probability of never using wood by 9.4% and lowered the frequency probability associated Estimation results of modeling the frequency use of the five cooking fuels are re- with the “seldom”, “often”, and “very often” category by 5.8%, 2.2%, and 1.4%, respectively. ported in Table 4. The system of equations performs well, as indicated by the likelihood The effect of education as compared to those without formal or only elementary education consistetnetslty vwalausea ssisgonciiafitceadnwt aitth pt