Electricity access and charcoal consumption among urban households in Zambia

ABSTRACT

This study uses a nationally representative dataset of urban households in Zambia to examine household cooking fuel choice patterns and to quantify the effect of access to electricity on household charcoal consumption. We find charcoal to be the most prevalent cooking fuel, for both households with and without electricity access. Proportionately more charcoal users reside in low income residential areas. Using a two-stage econometric estimation procedure that accounts for endogeneity of access to electricity, we find that on average, households with access to electricity consume 54% less charcoal than their counterparts without access. Further, our results indicate that charcoal consumption tends to increase with income, but this increase attenuates as income increases further. Other socio-demographic variables such as age, education and household size are also important in influencing charcoal consumption. We discuss implications for interventions aimed at promoting cleaner energy sources and efficient charcoal use for cooking among urban residents.

KEYWORDS:

• Fuel choice
• urban households
• electricity
• charcoal
• Zambia

1. Introduction

An increasing proportion of the world’s population lacks access to clean sources of energy, and the problem is particularly pervasive in most developing countries. In 2012, an estimated 1.3 billion people, representing about 20% of the global population, did not have access to electricity (Mottaleb et al. 2017), while 2.7 billion people, or nearly two-fifths of the world population, depend on biomass energy sources for cooking (International Energy Agency [IEA] 2010). This proportion is highest in sub-Saharan Africa (SSA), with over 90% of the rural population relying on fuel wood. Meanwhile, in urban areas of SSA, over half of all households use fuel wood, charcoal or wood waste as the main source of cooking fuel (IEA 2006). The rising proportion of SSA population relying on charcoal for cooking has been and continues to be a source of environmental and health concern.

A number of studies on charcoal use in SSA indicate that about 80% of urban households in the region depend on charcoal for cooking, and the demand for this energy source is likely to increase in the foreseeable future (Arnold et al. 2006; Tembo et al. 2015; Zulu & Richardson 2013). Among the factors driving charcoal demand in urban areas of SSA countries, including Zambia, are urbanisation, growing population density, insufficient and erratic supply of modern forms of energy such as electricity, coupled with rising electricity tariffs (Mwitwa & Makano 2012; Zulu & Richardson 2013). However, the increase in fuel wood use can have adverse environmental, socioeconomic, and health effects. Unsustainable harvest of fuel wood degrades local forests, leading to deforestation, damaged wildlife habitat, and poor watershed functioning (Heltberg 2001; Hofstad et al. 2009; Köhlin et al. 2011). Other environmental effects may include soil erosion due to loss of tree cover, leading to declining agricultural productivity (Arnold et al. 2006; Kambewa et al. 2007; Alem et al. 2010). Other studies have highlighted the negative health effects of burning charcoal for cooking fuel, including respiratory problems, acute health problems (Ellegǻrd & Egneus 1992).

The continued rise in fuel wood use in Zambia, particularly charcoal, to meet urban household energy demands is placing the country among those with the highest deforestation rates, with an annual rate of 250,000 to 300,000 hectares per year (Vinya et al. 2011; Kalinda et al. 2013), thus reducing the country’s potential to contribute to climate change mitigation. Hence, unsustainable charcoal production could contribute to climate change and exacerbate its impacts.

Zambia has formulated numerous climate change policies and strategies that have a bearing on household level energy choice and use. These include National Climate Change Response Strategy of 2010, National Policy on Climate Change of 2016, National Policy on Environment of 2007, the 2014 National Forestry Policy, and the National Energy Policy of 2008. These pieces of legislation directly and indirectly promote use of clean energy at household level, as well as sustainable harvest of forest-based resources for energy production. However, these policies have not been very successful in influencing household energy use due to implementation challenges – mostly related with human and financial resource constraints. The success of these policies not only depend on resource availability, but also a good understanding of household level characteristics governing the choice of cooking energy source as well as household level drivers of charcoal consumption.

Numerous studies have been conducted in SSA pertaining to households’ cooking fuel choice, for both rural and urban households, with the aim of understanding fuel choice behaviour and its determinants. For instance, Campbell et al. (2003) assessed domestic fuel choices in selected towns of Zimbabwe; Pundo & Fraser (2006) conducted a study in Kenya’s Kisumu district to determine factors affecting rural households’ cooking fuel choice, and determined that education, ownership of dwelling place and type of food mostly cooked were some of the factors influencing the choice of cooking fuel. Hiemstra-van der Horst & Hovorka (2008) examined the fuel transition theory using data from Ngamiland district of Botswana, and found no evidence of energy transition from traditional to modern energy forms as wealth improved at household level.

These studies provide a good background for and insights into understanding household fuel choice behaviour in SSA. However, most of the existing studies tend to focus on particular regions of a country, and are thus unable to make inferences beyond the study regions. Furthermore, although these studies provide evidence suggesting that households would transition from less clean to cleaner fuels as incomes rise and other socioeconomic factors improve, there is little known about how access to clean fuels and forms of energy such as electricity reduce consumption of less clean fuels such as charcoal. Measuring the effect of access to electricity on charcoal consumption will help guide energy and forest policies form realistic expectations regarding household energy transition from charcoal to electricity. The lack of such empirical analyses has led to perceptions that increasing electricity access can significantly reduce reliance on charcoal and help households transition toward electricity. To the best of our knowledge, no empirical study has analysed the effect of access to electricity on household consumption of high emission fuels such as charcoal using nationally representative data. An exception here could be Shackleton et al. (2017) who examine fuel use patterns among communities with access to electricity in Eastern Cape of South Africa, with a particular focus on fuel wood.

The contribution of this study rests on the two research gaps identified above. First, the study is among the first in Zambia, and in SSA to use a nationally representative household level dataset to analyse urban household cooking fuel choice behaviour. The use of nationally representative, rather than regional data as in previous studies (e.g. Campbell et al. 2003; Pundo & Fraser 2006; Hiemstra-van der Horst & Hovorka 2008; Shackleton et al. 2017), permits us to account for spatial differentials that exists across regions in a country in terms of availability of and access to various fuel sources. Secondly, the study provides one of the first systematic and statistically rigorous estimations of the effect of access to electricity on charcoal consumption. Further, unlike other studies that treat alternative energy sources as exogenous, we account for potential endogeneity of access to electricity. In this way, our analytical framework is more comprehensive and realistic compared to those that assume exogeneity. Against this backdrop, the objectives of this paper are twofold; 1) examine urban households’ choice of main cooking fuel; and 2) estimate the effect of access to electricity on charcoal consumption among urban households in Zambia.

2. Data

This study uses data from 10,024 urban households drawn from the nationally representative Living Conditions Monitoring Survey (LCMS) of 2010, conducted by the Central Statistical Office (CSO). The LCMS sample follows a two-stage cluster sampling procedure, where in the first stage, 1,000 standard enumeration areas (SEAs) were selected with each SEA having equal probability of being selected. The second stage involved selecting 20 households from each sampled SEA using systematic sampling. We also used charcoal price and consumption data from the Urban Consumption Survey (UCS) conducted by the Food Security Research Project (FSRP) in the year 2007.

The LCMS includes data from both rural and urban households, with the latter being the focus of this study. The local authorities in the country have classified urban residential areas into low-cost, medium-cost, and high-cost areas, and the LCMS collected data across all three residential classifications. The LCMS also collected data on household energy use, including charcoal and electricity, and for different tasks such as cooking, lighting, and heating, as well as information on whether or not a household had access to electricity. The survey also collected demographic and socioeconomic characteristics of households. Thus, the data permits analysis that links household charcoal consumption and access to electricity, as well as socioeconomic, and demographic characteristics.

To estimate the quantities of charcoal consumed, data on charcoal expenditures were converted into physical units of consumption. Since the LCMS did not collect price data, regional unit prices were obtained from the UCS of 2007 and adjusted to 2010 prices using consumer price index. The average unit, weighted by the frequency of unit purchases recorded in the survey, was used to estimate charcoal weighted average price (price/kilogram [kg]). We then divided charcoal expenditures by estimated unit prices to obtain physical quantities of charcoal consumed. However, the energy data in the LCMS have two major limitations. Firstly, the survey asked only for the main¹ cooking fuel for each household, the main cooking device, and whether the house was connected to electricity. Thus, identifying fuel-stacking patterns – the use of multiple fuel sources simultaneously as more fuel options become available – is difficult with the LCMS data. Secondly, the LCMS does not capture information of whether or not a household is located in an area that is connected to the national electricity grid. As a result, we do not observe electricity availability. Hence, we rely on a two-stage estimation procedure to control for electricity availability in an area, as described in the next section.

3. Estimation and identification procedure

Like all typical marketed consumption goods, the quantity of charcoal demanded and consumed by a household is mainly determined by the good’s own price, the price of substitutes, household income, and other household socioeconomic characteristics. In order to analyse the factors affecting household monthly charcoal consumption, one can model household monthly charcoal consumption as a function of price, income and a set of other socioeconomic characteristics, such as household size, and the age and education level of the household head. The model can be expressed as:(1) $C_i=X_i^{\mathrm{\prime }}\beta +{\epsilon }_i$(1) where C_i is monthly charcoal consumption for household i, $X_i^{\mathrm{\prime }}$ represents factors postulated to influence charcoal consumption for household i, $\beta$ is a vector of pararmeter estimates, and ${\epsilon }_i\sim N\left(0,1\right)$ is the error term. This model can be estimated using ordinary least squares (OLS) estimation. However, we suspect access to electricity to be endogenous, because access could be explained by factors such as whether or not a geographical area is connected to the national electricity grid, which is unobserved in our data. Thus, we implicitly assume that the effect of connection to the national grid is zero and therefore part of the error term. In reality this may not be true, since households located in areas that are not connected to the national electricity grid have no way of getting connection. Without accounting for this possibility, our estimate of electricity connection will be confounded by the unobserved factors lumped in the error term, leading to violation of the exogeneity assumption of OLS, and the result is biased and inconsistent estimate (Wooldridge 2010). It can easily be shown that the model can be expressed as:(2) $C_i=X_i^{\mathrm{\prime }}\beta +\sigma A_i+{\rho }_i{G}_i+{\epsilon }_i$(2) where A_i is i^{t h} household’s access to electricity, G_i is whether an area is connected to the national electricity grid (G), which besides influencing consumption, also influences A_i. Thus, by assuming that G is part of the error term (since we do not observe it in our data), this specification makes A_i endogenous, leading to biased and inconsistent estimates. To guard against this problem, one might employ a control function approach as suggested by Wooldridge (2010) and applied by studies with similar econometric problems such as Ngoma et al. (2016) and Mulenga et al. (2017). However, application of the control function approach requires that we have a valid instrumental variable (IV). Since we do not have a valid instrumental variable (IV) to instrument access to electricity, we employ a version of the IV estimator suggested by Wooldridge (2008) in which the predicted values from the first-stage estimation are used as an instrument.

The first stage for this estimator involves estimating the reduced form by regressing the suspected endogenous binary variable (electricity access) on all exogenous variables via probit regression. In the second stage, quantity of charcoal consumed is regressed on all exogenous variables and the predicted values of electricity access from the first stage probit are used as an instrument. This method is fully robust to misspecification of the first stage probit (Wooldridge 2008), and thus a better option in the absence of a valid IV, as is the case in this study.

Thus, our estimation procedure is as follows; in the first stage we estimate a probit expressed by the equation:(3) $A_i={\mathit{X}}_i^{\mathrm{\prime }}\beta +{\nu }_i$(3) By definition $\text{E}\left[\mathit{X}\nu \right]=0$and if we specify a linear projection of u_i on v_i, we have $u_i=\psi v_i+{\epsilon }_i$, and by definition $\text{E}\left[\nu \epsilon \right]=0$, thus $\text{E}\left[\mathit{X}\epsilon \right]=0$, since now both u_i and v_i are uncorrelated with X. Therefore, predicted values of $A_i\left({\hat{A}}_i\right)$ are independent of u_i and C_i, hence can be used to instrument A_i. The second stage involves regressing C_i on all X and ${\hat{A}}_i$ as an IV for A. The second stage model is therefore formulated as:(4) $C_i=X_i^{\mathrm{\prime }}\beta +{\gamma }_1{\hat{A}}_i+{\mu }_i$(4)

From (4) above we can consistently estimate ${\gamma }_1$ than if we use a control function without an instrument as in Tembo et al. (2015), or by OLS. Table 1 below presents a summary of descriptive statistics of variables used in the model.

Table 1. Summary statistics of regression variables.

Variable	Mean	Std. Dev	Min	Max
Age of head (years)	41.00	12.53	10.00	90.00
Male headed (1 = yes)	0.77	0.42	0.00	1.00
Head completed high school (1 = yes)	0.30	0.46	0.00	1.00
Household size	4.56	2.34	1.00	20.58
Charcoal consumption (Kg/month)	24.96	23.86	0.00	675.47
Electricity access (1 = yes)	0.47	0.50	0.00	1.00
Monthly expenditure (ZMW1,000)	1.685	2.36	0.14	62.30
Charcoal price (ZMW/Kg)	1.12	339.05	0.53	1.35

Source: Authors’ calculations from LCMS 2010 data, CSO 2012.

4. Results and discussion

4.1. Descriptive results

Before delving into causal effects of access to electricity on charcoal consumption, we first present the results of descriptive analysis, which provide a foundation for the econometric model. This section presents mainly the correlations between household fuel choices and charcoal consumption, without making any causal inferences.

4.1.1. Fuel switching

This section is concerned with urban household fuel choice and how it relates to the standard energy ladder model on fuel choice. The energy ladder is based on the assumption of fuel switching by a household, whereby cleaner fuels displace less clean fuels as household income increases and welfare improves. However, this is not always the case because households use cooking fuels in complex combinations. In urban Zambia, less than 1% of households use gas, kerosene, and other fuels (Tembo et al. 2015). Because of this, our analysis focuses on the main fuels used, namely charcoal, electricity and a mix of electricity and charcoal. Therefore, in this study, we define fuel-switching as the choice between charcoal (solid fuel) and electricity (modern non-solid energy) or a mix of electricity and charcoal. Based on the data, we categorise households into three groups namely: 1) charcoal user–households use charcoal as the main source of cooking fuel; 2) mixed fuel user–households use a combination of charcoal and electricity; and 3) electricity user–households use electricity as the main source of cooking energy.

By defining categories this way, we are better able to analyse the extent to which electricity displaces charcoal (solid fuel) in response to changes in wealth status. Displacement of charcoal, to a large measure, is required if electricity is to have an impact on combating problems associated with the use of traditional fuels. The share of households in each fuel use category is shown in Table 2. Close to half (47%) of households still rely on charcoal as the main cooking fuel. In second place is mixed fuel reported by 34% of sampled urban households, and in a distant third place with only 14% of the sampled households are electricity users. This result shows that charcoal remains a widely used cooking fuel in urban areas of Zambia. Note that the percentages do not sum to 100 since some households reported their main cooking fuel as ‘wood’ or ‘other’, which we do not consider in our analysis.

Table 2. Main cooking, Urban Zambia.

Main Cooking Fuel	% Households
Charcoal user	47
Mixed fuel user	34
Electricity user	14

Note: The percentages do not sum to 100 since they include households reporting their fuel as ‘wood’ or ‘other.’

Source: Authors’ calculations from LCMS 2010 data, CSO 2012.

We further examined fuel choice patterns by income level by categorising households into 10 income groups using expenditure as a measure of income (expenditure deciles). Figure 1 shows the share of households in each expenditure decile in urban areas that belong in the three exclusive fuel switching categories. Results of this analysis indicate that charcoal is predominant in the lower deciles, until mixed fuel displaces it in the 6^th decile, while electricity-only becomes more prevalent relative to charcoal after the 8^th decile. This result seems to suggest that charcoal remains an important source of domestic cooking energy, even as household incomes increase. It appears that switching will only begin to be prevalent when incomes rise beyond average (5^th decile).

Figure 1. Main Cooking Fuel among Urban Households in Zambia by Expenditure Decile. Source: Authors’ calculations from LCMS 2010 data, CSO 2012.

It can be gleaned from this analysis that relative to charcoal, the role of electricity for cooking fuel remains low in urban areas, even among relatively wealthy households. In addition, our results suggest that electricity mostly complements rather than displaces charcoal as can be seen from the high percentage of household using both electricity and charcoal as income rises. Generally, there is little fuel switching in urban Zambia, with only the upper income deciles exhibiting fuel switching patterns. This is consistent with previous studies in the region (e.g. Campbell et al. 2003; Hiemstra-van der Horst & Hovorka 2008) which suggest that household fuel decisions are influenced by other factors besides income.

4.1.2. Multiple fuel choice in Zambia

This section examines the distribution of urban households using different energy types for cooking, across residential areas and as a total.² Figure 2 presents a graphical summary of the share of urban households by type of cooking fuel and type of residential area. Overall, we note that charcoal-only as the main cooking fuel, relative to other sources, remains dominant. When analysed across residential areas, results indicate a correlation between type of residential area and cooking fuel choice patterns. Charcoal-only is the most common cooking fuel in low-cost areas, representing close to two-thirds of households in this category. Electricity-only represents only one tenth of the total households in the low-cost areas. Among the medium- and high-cost areas, charcoal-only was the least commonly-reported cooking fuel, representing less than a quarter of households in the medium costs, and less than a fifth (16%) in the high-cost areas.

Figure 2. Percent Urban Households by Cooking Fuel Type and Residential Area. Source: Authors’ calculations from LCMS 2010 data, CSO 2012.

Generally, our result, showing more households in low-cost areas using charcoal, followed by medium-cost, and lastly high-cost areas, is consistent with Chidumayo et al. (2002), although their analysis was based on Lusaka city only. Among the factors identified to be associated with high dependence on charcoal in low-cost residential areas were erratic electricity supply, lack of access to electricity and electric cooking devices, and household size (Chidumayo et al. 2002). It is noteworthy that even in the medium and high cost areas, a mix of electricity and charcoal is more common, indicating fuel stacking pattern rather than fuel-switching. Thus, charcoal remains a dominant cooking fuel across all the three residential categories.

We further examined the use of the three main types of cooking fuel among urban households that have access to electricity in order to assess the role of charcoal among these households (Figure 3). From the results, a combination of electricity and charcoal stands out, with two-thirds of the households using both electricity and charcoal as the main source of cooking fuel, further reinforcing earlier results of predominant fuel-stacking rather than switching. The predominance of fuel-stacking pattern among the wealthy and those with access to electricity perhaps suggests that other factors besides income and access to electricity influence households’ tendency to use multiple fuels concurrently.

Figure 3. Electrified Urban Households by Type of Cooking Fuel. Source: Authors’ calculations from LCMS 2010 data, CSO 2012.

Analysis of variance (ANOVA) of socioeconomic variables across fuel choices revealed some interesting patterns across the three fuel choice groups. Table 3 below presents a summary of ANOVA results. Generally, households that used electricity only or a mix of electricity and charcoal appear to have higher incomes, more educated household heads, and were proportionately more male-headed as compared to those that use charcoal-only. We further conducted a Tukey post-hoc test to evaluate the statistical differences in means of the socioeconomic variables across the three fuel choice categories. Test results indicate statistically significant differences in means across the three fuel choice groups for all variables except gender of head and access to electricity. Gender of household head and electricity access were only statistically significantly different between electricity-only and charcoal-only, as well as between electricity and charcoal mix and charcoal-only. Age of the head was highest for households that use a mix of fuels, followed by charcoal-only and lastly electricity-only. This result is somewhat surprising as one would expect older heads to more likely use charcoal-only given their preference for foods that are perceived to cook well on charcoal stoves.

Table 3. Analysis of variance of socioeconomic variables by cooking fuel choice.

Variable	Main Cooking Fuel
	Electricity-only	Electricity-charcoal	Charcoal-only
Age of head	38.85	42.00	40.95
Male head	0.80	0.80	0.74
Head completed high school	0.59	0.47	0.08
Household size	4.54	5.75	5.00
Electricity access	0.77	0.78	0.13
Household monthly expenditure (ZMW’000)	2.42	2.37	0.74

Source: Authors’ calculations from LCMS 2010 data, CSO 2012.

A comparison of education of the household head shows a statistically significant difference across the three groups, with households that use electricity-only having proportionately more heads who completed high school, followed by those using a mix of electricity and charcoal, and in distant third place are households that use charcoal-only. This result is consistent with others such as Mottaleb et al. (2017), Pundo & Fraser (2006) and Heltberg (2001) who report that education of the head and spouse influence household use of cleaner and /or modern cooking fuels. In terms of household size, households that use a mix of electricity and charcoal had significantly more members compared to the other two categories, perhaps because of the high energy demand to cook meals for a large family, hence supplementing electricity with charcoal. As expected, there was a significantly higher proportion of households with electricity access among those that use electricity or a combination of electricity and charcoal than among those that use charcoal only. A comparison of household monthly expenditure, which we use as a proxy for income, reveal marginal, yet statistically significant differences between households that use electricity-only and those using a mix of electricity and charcoal. However, the difference in expenditure is larger between those that use charcoal-only and each of the other two categories, with the former having lower expenditure.

To summarise the descriptive results, a substantial proportion of urban households regardless of income or access to electricity continued to rely fully or partially on charcoal. Our results indicate that urban households use multiple energy sources at different points of the energy ladder. From other studies (Chidumayo et al. 2002; WHO 2014), a number of factors are identified as causing this behaviour, and these include cultural influences, cost of electricity, and use of charcoal as backup fuel in case of electricity power failure. The cost of electricity in Zambia has been rising rapidly leading most households to rely on charcoal. For example, Zambia Electricity Supply Corporation (ZESCO) increased electricity tariffs for domestic consumers by 50% in May 2017 and a further 25% in September 2017 (Zambia Daily Mail Newspaper 11 May, 2017 edition). Household cooking fuel choices under such circumstances tend to be driven mainly by costs, especially when cooking certain types of food that require a long time to cook – for example, beans, or dried fish.

4.2. Econometrics results

This section focuses on explaining factors affecting household charcoal consumption, particularly, the effect of access to electricity. A priori, we suspected access to electricity to be endogenous, therefore we conducted a test of endogeneity, and test results indicate evidence of its presence. Thus, the use of a modified IV estimator (as presented in the procedure and identification section) to guard against biased and inconsistent parameter estimates is justified. For ease of result interpretation, the dependent variable (monthly charcoal consumption) was log transformed so that we can interpret the coefficients on independent variables as percentage change in the dependent variable due to a unit change in a particular independent variable. Expenditure (a proxy for income) and charcoal price were log transformed, thus permitting for interpretation of the coefficients associated with these two variables as income and own-price elasticities, respectively.

Table 4 presents the determinants of urban household monthly charcoal consumption. The results reveal that age and education level of the head are important determinants of charcoal consumption, with older and less educated household heads significantly associated with higher charcoal consumption. According to the results, households headed by older heads would tend to use more charcoal, on average, than their counterparts headed by relatively younger heads. This is perhaps because older people tend to prepare meals using traditional methods, which may not be possible with electric cooking devices such as electric stoves. These results correspond with Kapfudzaruwa et al. (2017) who use cross country analysis to explore adoption of improved cook stoves in Africa. They find older people prefer traditional cooking appliances, compared with the younger generation, mostly in urban settings, who are more willing to use modern cooking fuels and devices, rather than use traditional appliances that lack modern appeal. Education of the head was found to negatively influence charcoal consumption, which is not surprising as education is expected to make people more aware of the health and environmental hazards of excessive use of less clean fuels such as charcoal.

Table 4. Determinants of household charcoal consumption (Log Kg of charcoal used per month).

Variable	Coefficient
Age of the head	0.005
	(0.001)**
Gender of the head (1 = male)	−0.018
	(0.020)
Head completed high school (1 = yes)	−0.061
	(0.029)*
Household size (adult equivalent)	0.040
	(0.004)**
Total household expenditure (^aZMW ‘00)	0.548
	(0.045)**
Square of household expenditure	−0.058
	(0.007)**
Charcoal price per kg	0.001
	(0.002)
Access to electricity (1 = yes)	−0.543
	(0.108)**
Joint provincial dummy (9 provinces)	55.96**
N	8,295

^aHundred of Zambian kwacha. As at 2010, the exchange rate was USD1 to ZMW 4.

Notes: District-level-clustered standard errors in parentheses; **p < 0.01, *p < 0.05.

As expected, household size had a significant and positive effect on monthly charcoal consumption. In terms of magnitude, increasing household size by one more member is associated with increases in charcoal consumption by 25 percentage points. Given the high levels of energy required to prepare meals for a large family, larger households are more likely to use more charcoal (either on its own or in combination with electricity) to prepare meals, relative to smaller households. This finding, to a large extent, corroborate that of Huang (2015), who determines that larger household size led to increased electricity consumption in Taiwan. However, on per capita energy use basis, Huang (2015) finds that larger household size reduced electricity consumption.

Total monthly expenditure was used as a proxy for household income, and had a positive and significant coefficient, implying that charcoal consumption, by an average urban household in Zambia, increases with income, holding all else equal. Model estimates show that a percentage increase in household income would result in a 0.28 percentage points increase in monthly charcoal consumption. This finding corroborates evidence from similar studies that charcoal is not an inferior good, contrary to most literature and the energy ladder model, in particular. This result is consistent with Mekonnen & Köhlin (2009) who find that even at higher income levels, urban households still use traditional energy sources such as charcoal in Ethiopia, mainly because of preferences, taste, reliability of supply, and cooking and consumption habits. However, the square of expenditure had a negative coefficient, suggesting that although charcoal consumption increases with income, this increase is at a decreasing rate and tends to diminish as income continues to increase.

To assess the effect of access to electricity on charcoal consumption, we included electricity connection as one of the explanatory variables, equal to 1 if a household has access and 0 otherwise. As expected, results show that access to electricity has a negative and significant effect on charcoal consumption. However, what is noteworthy here is the magnitude of this reduction. Results indicate that households with access to electricity consume 54% less charcoal per month than their counterparts who do not have access, all else equal. This finding is consistent with others such as Tembo et al. (2015) and Guta (2014), albeit the magnitude of the effect being smaller (47%) in Tembo et al. (2015). The difference in magnitudes is possibly due to the differences in model specifications and identification, with Tembo et al. having implemented a control function approach with the predicted values of access to electricity as a regressor, instead of as an instrument, as suggested by Wooldridge (2008).

The failure by most studies to control for possible endogeneity of access to electricity implicitly assumes that access to electricity is as given, and thus could lead to seemingly higher effects of access to electricity on reducing charcoal consumption, resulting in unrealistically high targets of reducing charcoal consumption by interventions aimed at increasing electricity access. This result has important implications for energy policies in Zambia, as well as other SSA countries, as it demonstrates that improved availability and accessibility of cleaner alternative fuels could help households cut back on the use of biomass based and less clean fuel sources such as charcoal. This finding is consistent with those of other studies (e.g. Chidumayo 2002; Ouedraogo 2006). However, the fact that a majority of households with access to electricity still use charcoal to prepare meals is indicative that charcoal will remain an important cooking fuel for the foreseeable future. A similar finding is reported in Shackleton et al. (2017), where despite having access to electricity, households in Makana district of South Africa continued to use fuel wood for cooking. This could mean that relying on improving access to electricity in order to reduce charcoal consumption may not produce the desired results.

Household cooking fuel preferences are influenced by myriad of socio-cultural settings (Beyene & Koch 2013). For example, in Zambia, it is common for households with access to electricity to cook certain food types using charcoal, rather than electricity. This is because charcoal – in addition to being cheaper than electricity – - is believed to enhance the taste of certain food types such as beans and dried fish. Spatially, we find that being in a particular province has significant influence on charcoal consumption, as indicated by the significant coefficient of the joint provincial dummy variable. This is indicative that regional differences – which could include culture, food type – affect household cooking fuel choice behaviour.

5. Conclusions

The objective of this study was to examine urban households’ cooking fuel choice patterns and behaviour, with the aim of understanding the extent to which charcoal is widely used in urban areas. The study also aimed to determine the effect of access to electricity on household charcoal consumption, by controlling for potential endogeneity of access to electricity. Results show that charcoal-only and electricity-only are at the two extreme ends of the income continuum, with the charcoal-electricity mix lying between the two extremes. Further, it is interesting to note that within this pattern, there is an emerging picture of fuel-stacking behaviour (electricity-charcoal mix) as income increases, a result inconsistent with expectations of fuel switching theory. Our results, like those of other similar studies, show that higher incomes are associated with higher likelihood of electricity use, however, in most cases, this tends to lead to household fuel-stacking (a mix of electricity and charcoal), rather than fuel-switching (from charcoal to electricity). This finding is likely related to the high cost and unreliable supply of electricity provision in some areas.

In terms of determinants of charcoal consumption, results show that one of the most influential factors is access to electricity, with households who have access to electricity consuming 54% less charcoal than their counterparts (without access), all else equal. It follows from this that access to electricity promises to slow rising charcoal demand, and associated deforestation and/or forest degradation. However, the fact that a relatively high percentage of affluent households, including those with access to electricity still use charcoal in combination with electricity, implies that charcoal will remain an important energy source in the foreseeable future. It would therefore be helpful for stakeholders designing interventions aimed at reducing charcoal use and/or replacing it with cleaner energy to not be too optimistic and simplistic in their approach. Such interventions could have a higher chance at success by focusing more on promoting sustainable charcoal production and ramping up promotion for adoption of improved cook stoves – especially among the low- and middle-income households, who make up the majority of charcoal users. Improved cook stoves use charcoal more efficiently than the typical Zambian charcoal stove (Mbaula), hence these devices could help reduce charcoal consumption among low- and middle-income households. In this way, the effect of charcoal use on forest resources and the environment could be reduced, while meeting the energy demand of urban households.

Furthermore, there is a need for continued efforts to connect more urban households to the national electricity grid, while exploring and promoting alternative and cleaner energy sources such as liquefied petroleum gas (LPG) and renewable energy technologies. However, such efforts need to be cognisant that cleaner alternatives such as LPG have significant cost implications, such as up-front and on-going costs relative to charcoal and electricity. Also, LPG is perceived to be a danger by most Zambian households, hence its uptake remains low.

From a methodological point of view, our attempt to control for endogeneity of access to electricity seems to suggest that availability of alternative energy sources should be treated as endogenous and controlled for accordingly. This analytical approach can also be extended to studies that compare biomass-based fuels with clean fuels such as LPG by controlling for variability in supply and availability of LPG across space. Future research on cooking fuel choice patterns and behaviour should consider employing multi-year data and where possible employ panel data methods in order to understand how fuel choice patterns and fuel substitutions and/or switching vary across time and food types.

Acknowledgement

The authors gratefully acknowledge financial support from the United States Agency for International Development (USAID) and the Swedish International Development Agency (SIDA) missions in Lusaka, through their partner, the Indaba Agricultural Policy Research Institute (IAPRI). Our thanks also go to the Central Statistical Office for providing us with the data used in this study. This article is based on the working paper version published in the IAPRI working paper series. Any views expressed or remaining errors are solely the responsibility of the authors.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes

1 The LCMS asked about the main source of cooking fuel. Thus, if a household used multiple cooking fuel sources, only the one considered as the main source by the household was captured.

2 At the sample selection stage of the LCMS, the urban SEAs were classified as low cost, medium cost, and high cost areas according to local authority classification of residential areas.