Technological innovation and dispersion: Environmental benefits and the adoption of improved biomass cookstoves in Tigrai, northern Ethiopia

Energy Economics 67 (2017) 337–345

Contents lists available at ScienceDirect

Energy Economics journal homepage: www.elsevier.com/locate/eneeco

Technological innovation and dispersion: Environmental benefits and the adoption of improved biomass cookstoves in Tigrai, northern Ethiopia Zenebe Gebreegziabher a,⁎, G. Cornelis van Kooten b, Daan P. van Soest c a b c

Department of Economics, Mekelle University, Adi-haqui Campus, P.O. Box 451, Mekelle, Tigrai, Ethiopia Department of Economics, University of Victoria, Canada Department of Economics, Tilburg University, The Netherlands

a r t i c l e

i n f o

Article history: Received 12 October 2014 Received in revised form 26 August 2017 Accepted 28 August 2017 Available online 07 September 2017 JEL classification: Q12 Q16 Q24

a b s t r a c t This paper empirically analyzes adoption and fuel savings efficiency of improved biomass cookstove technology using survey data from a cross-section of 200 farm households from the highlands of Tigrai, northern Ethiopia. Results indicate that these farm households are willing to adopt improved biomass cookstove innovations if this leads to economic savings. Moreover, results suggest significant positive environmental externalities. On a per household basis, we found that adopters collect about 70 kg less wood and about 20 kg less dung each month. The adoption of improved biomass cookstoves reduces harvest pressure on local forest stands: assuming an average of 120 metric tons of biomass per ha, we find the potential reduction in deforestation amounts to some 1400 ha per year – an important saving. Further, the reduced use of dung as a fuel has a positive impact on soil productivity in agriculture. © 2017 Elsevier B.V. All rights reserved.

Keywords: Improved cookstoves and fuel-savings Economic development Land degradation Technological adoption

1. Introduction Land degradation and deforestation are global environmental concerns. Land degradation affects some two-thirds of the world's agricultural land, resulting in reduced agricultural productivity (UNDP, 2002), along with income loss and food insecurity that threaten livelihoods (Nkonya et al., 2011). It is particularly a vexing problem in developing countries because it leads to a poverty trap. Poverty is one of the ultimate causes of land degradation that, in turn, exacerbates poverty by reducing the quality of the most important resource available for economic development. For example, despite the considerable progress that has been made in reducing poverty in some parts of the world over the past couple of decades, and notably in East Asia (IFAD, 2010), about 11% of the world's population is estimated to live below the international poverty line of $1.90 per person per day (World Bank, 2016). Further, the majority of the world's least well off are rural, young, poorly educated and mostly engaged in the agricultural sector, with over half of the world's extremely poor located in Sub-Saharan Africa. Among other factors, declining agricultural productivity resulting from land ⁎ Corresponding author. E-mail address: [email protected] (Z. Gebreegziabher).

http://dx.doi.org/10.1016/j.eneco.2017.08.030 0140-9883/© 2017 Elsevier B.V. All rights reserved.

degradation and the deterioration of other natural resources has contributed to this rural poverty. In most developing countries, inefficient exploitation of the land reduces the amount of resource rents that can be collected, while lowering available future rents as land resources are degraded over time (van Kooten and Bulte, 2000). Consequently, increasing poverty combined with a lack of property rights to land causes peasants to invest too little in land improvements. A cycle of land degradation occurs because, as forests are depleted, people turn to grasses, crop residues and livestock dung for fuel, which degrades the land further (Pearce and Warford, 1993, p. 25). This is certainly true in countries where deforestation is a major problem, causing many peasants to switch from fuelwood to crop residues and dung for cooking and heating purposes, thereby damaging the agricultural productivity of cropland (World Bank, 2006). With dwindling forest resources, Arnold et al. (2006) indicate that in rural areas people first switch to less preferred woody alternatives (e.g., fuelwood from less-preferred tree species) and then to other biomass fuels like crop residues, animal dung and even noxious weeds. Amacher et al. (2004) also argue that, as fuelwood scarcity increases, dung and crop residues eventually dominate, with adverse consequences for agriculture. Perhaps not surprisingly then, Newcombe

338

Z. Gebreegziabher et al. / Energy Economics 67 (2017) 337–345

(1989) estimated that burning of 7.9 million metric tons of dung per year in Ethiopia was associated with a reduction of agricultural productivity of 6 to 9% of the country's gross national product from lost nutrients associated with manure. Ethiopia is the second most populous country in Africa with a population of over 90 million, and a per capita annual income of $470 (UN, 2015). It experienced double-digit, broad-based growth over the past decade (i.e., real GDP grew by an average 11% annually), compared to a regional average of about 5% (World Bank, 2014). Only about 12% of the country's surface, or 13.0 million ha of 110.4 million ha of total land, is forested (WBISPP, 2004, 2005; Moges et al., 2010), compared to some 40% about 100 years ago (EFAP, 1994). Recently, Ethiopia adopted a new forest definition to better capture its dry and moist lowland vegetation sources. The new definition specifies forests as “land spanning at least 0.5 ha covered by trees or bamboo, attaining a height of at least 2 m and a canopy cover of at least 20% or trees with the potential to reach these thresholds in situ in due course” (MEFCC, 2016).1 Standing timber amounts to some 285 million m3, or about 22 m3 per ha. Total biomass in forest ecosystems amounts to 573 million metric tons (t), or 44 t per ha. In 2005, 108.879 million m3 of timber was harvested for fuelwood (24.5% more than in 2000) along with 2.982 million m3 for industrial roundwood (an increase of 21.3% compared to that of 2000), all of which were consumed domestically; harvests of timber for other uses were insignificant in comparison (FAO, 2006). Between 1990 and 2010, the average annual rate of deforestation was nearly 1% (FAO, 2010), one of the highest rates in the world. Local deforestation and forest degradation also contribute to greenhouse gas emissions (Angelsen and Brockhaus, 2009; IPCC, 2007). Some further argue that biomass fuel burning with inefficient stoves could even contribute to global warming more than stoves using fossil fuel (Sagar and Kartha, 2007). Consequently, there is renewed interest in improving cookstoves as a potential solution to these important problems. A review of relevant literature identifies the following outstanding issues. First, land degradation is a major concern that undermines the livelihoods of rural people in many developing countries, including Ethiopia. The substitution of dung and crop residues for fuelwood is a major factor leading to land degradation. Second, the use of inefficient cookstoves also contributes to deforestation and de-vegetation.2 Third, whereas a substantial and growing literature characterizes the important features of adopting agricultural production technologies (Feder et al., 1985; Dadi et al., 2004; Barrett et al., 2004), knowledge about the characteristics of stove adoption is sparse. Fourth, empirical evidence regarding the effectiveness or fuel savings of the improved cookstoves (ICS), particularly biomass stoves and the potential for further interventions to promote these stoves, is scanty and mixed. Fifth, some argue that efforts to promote stoves have shown success only in towns, where the stoves are seen as saving money, but not in rural areas, where they are merely saving time or biomass. However, one would expect that economic savings—not only in terms of money but also in terms of time or biomass—would provide sufficient incentives for the adoption of ICS. Hence, it is of interest to investigate whether it matters that the stoves are merely saving time or biomass and not saving money in terms of their successful promotion. Last, although outside the scope of this study, there also appears to be a relatively new and growing interest on the need for ICS because of the health impacts of indoor air pollution from the use of biomass fuels (Larson and Rosen, 2002; Sinton et al., 2004; Hutton et al., 2006; Balietti and Datta, 2017). The Ethiopian government has embarked on a two-pronged strategy to stem deforestation and the degradation of agricultural lands: tree planting (afforestation) and dissemination of fuel-efficient cookstoves. 1 WBISPP (2004, 2005) and Moges et al. (2010) employ the FAO definition of forests. The new definition has increased the forested area from 12% to 15.5%. 2 The term de-vegetation refers to the removal of vegetation and exposure of bare soil throughout one growing season, or more (Lund, 2012).

More recently, the Federal Government of Ethiopia declared its intention to distribute 9.4 million ICS of various sorts within five years as a key part of the country's Climate Resilient Green Economy (CRGE) strategy. However, the effectiveness or fuel savings efficiency at the field level of these ICS has not been systematically evaluated. The purpose of the current study is to examine the potential of the second strategy, namely, the dissemination of fuel-efficient cookstoves. To do so, we use a dataset that covers 200 households from the highlands of Tigrai, northern Ethiopia. The specific objectives of the study are to (1) determine what characterizes ICS adoption; and (2) estimate the impact of disseminating ICS in redressing the fuel problem and land degradation. Key questions of interest are: how effective is the improved cookstove promoted as a remedy to the fuel problem and land degradation? What characterizes stove adoption? Should it matter that the stoves are merely saving time or biomass but not money? In doing so, our aims are both to determine the propensity to adopt new stoves and to isolate how adoption of improved stoves changes behavior (including, e.g., the frequency with which households prepare hot dishes and the number of cattle they might keep). Because we cannot a priori exclude the possibility of a rebound effect—that use of more efficient stoves actually increases fuel demand rather than reducing it, we need to pay particular attention to the consequences for actual use of the improved stove.3 In the next section, we review biomass cooking fuels, stoves and environmental degradation. Our conceptual and empirical model of the stove adoption process and its impacts is provided in Section 3, while the data, survey design and context are discussed in Section 4. Section 5 presents empirical results. The conclusions follow in Section 6. 2. Biomass cooking fuels, stoves and environmental degradation: review Environmental degradation manifested in the form of land degradation and deforestation presents a formidable challenge to many developing countries resulting in declining agricultural productivity which, in turn, exacerbates poverty (Johnson et al., 1997). In this regard, land degradation is serious, affecting some two-thirds of the world's agricultural land, resulting in a decline in agricultural productivity (UNDP, 2002). It results in income loss and food insecurity, thus threatening livelihoods (Nkonya et al., 2011). Of particular concern in this regard is the soil fertility (nutrient) depletion aspect of land degradation. The burning of dung and crop residues for fuel purposes—which were previously sources of soil humus and fertility—has resulted in a progressive decline in land quality and agricultural productivity. In the context of Ethiopia, this has increased farmers' vulnerability to shocks, food insecurity and poverty (Amsalu, 2006). Scherr (2000) identifies soil erosion, nutrient depletion, de-vegetation, loss of biodiversity, soil compaction, acidification, and watershed degradation as common problems of land degradation in the densely-populated marginal-developing regions. Campfires and low-efficiency stoves that rely on fuelwood are the traditional means for cooking and baking in most developing countries. Wood, dung and crop wastes are typically used with efficiencies not exceeding 12 to 15% (Sullivan and Barnes, 2006; Gebreegziabher, 2007; Gebreegziabher et al., 2012). It also appears that the use of inefficient stoves contributed to deforestation and de-vegetation. Although there are debates on the contribution of firewood usage to deforestation (including arguments that firewood collection might cause forest degradation but not deforestation), the literature suggests that the major 3 The new technology might increase fuelwood consumption because, as Wirl (2000) notes, people care only about the services a stove provides (e.g., heat generated) and not about fuelwood use itself. A more efficient stove implies that the same heat is obtained from less wood, so the price of heating services declines and thus households increase their demand for cooking services. This can result in more fuelwood use if the percentage change in services demanded is larger than the percentage change in fuel efficiency associated with the new stove.

Z. Gebreegziabher et al. / Energy Economics 67 (2017) 337–345

drivers of deforestation vary by context and continent (Arnold et al., 2006; Boucher et al., 2011), with firewood collection believed to be the major driver of forest degradation in sub-Saharan Africa (Geist and Lambin, 2002). This is especially true in Ethiopia where traditional biomass constitutes over 95% of domestic energy consumption with fuelwood accounting for about 90% (Gebreegziabher and van Kooten, 2013). Some argue further that, particularly in urban areas, poorer people are paying higher prices for useable energy than more well off consumers because of the inefficiency of the traditional biomass cookstoves and kerosene lamps, and that inefficient cookstoves involve a financial burden leading to inequity (Barnes et al., 2004). Cooke et al. (2008) argue that there is little economic investigation of the impact of using an improved biomass cookstove on actual household fuelwood demand. Improved cookstoves (ICS) have been regarded as technological alternatives to alleviate or mitigate the fuelwood crisis (for a crisis is what it is), particularly since the oil price shocks of the 1970s (Barnes et al., 1993). The greater energy efficiency of improved stoves can reduce fuel shortages and pressure on local forests. However, knowledge about the characteristics of ICS adoption is sparse. Moreover, empirical evidence on the effectiveness or fuel saving efficiency of the ICS, particularly biomass stoves, is scanty (Bluffstone, 1989). For example, Amacher et al. (1996) argue that ICSs significantly decrease fuelwood consumption and can be an important curb on deforestation. In contrast, using data from three villages (one remote and two with good market access) in Jiangxi province in China, Chen et al. (2006) found possession of ICS does not affect fuelwood consumption in the more remote villages but increases fuelwood consumption in villages with good market access. This runs counter to arguments that seek to promote improved stoves. Arnold et al. (2006) argue that efforts to promote stoves have shown success only in towns where the stoves are seen as saving money, but not in rural areas where they are merely saving time or biomass. More importantly, perhaps, there is a relatively new and growing interest for improved stoves because they help reduce the health impacts of indoor air pollution caused by the use of biomass fuels (Larson and Rosen, 2002; Sinton et al., 2004; Hutton et al., 2006; Balietti and Datta, 2017). Agarwal (1983) summarizes the factors likely to affect the diffusion of wood-stoves. These are (1) technical aspects related to the method of wood-stove design and development; (2) private financial aspects of investing in an improved wood-stove; (3) infrastructural aspects, such as extension and credit provision; (4) cultural aspects including attitude to change; and (5) social structure. Social structure here refers to inequalities in social status and the unequal nature of the power balance between different classes, casts or genders. In their synthesis of woodfuels, livelihoods and policy interventions, Arnold et al. (2006) argue that the fuelwood discourse or crisis has shown a classic pattern of thesis and antithesis over the last few decades. The use of fuelwood in developing countries is apparently not growing at the rates assumed in the past. Nonetheless, the researchers acknowledge that the complex reality in developing countries could seldom be captured in such clear narratives. For example, it might not be the case for Ethiopia, supporting the need for location or country specific studies. They argue that responses to fuelwood shortage are largely conditioned by the household's capacity to access resources such as labor, land and money, as well as other factors such as access to common pool resources and the availability and price of substitute fuels. 3. Household adoption of improved cookstoves: conceptual and empirical framework Improvements in energy efficiency, especially in relation to the cookstove technology, can affect the utility of our representative household in various ways – via reductions in the time spent collecting biomass for fuel use, via changes in the cooking pattern, et cetera. To

339

establish how the adoption of an ICS is expected to affect household welfare, we postulate the following household utility function: U i ¼ U ðci ; cf i ; tscwi ; tscdi ; ani ; zi Þ;

ð1Þ

where ci denotes household i's consumption during the period under consideration, cfi is the frequency with which the household cooks (number of times per month), tscwi is the time spent by the household collecting woody biomass for fuel purposes, tscdi is the time spent collecting dung for fuel, and ani is the number of farm animals the household owns. Cooking frequency is likely to be driven by the dietary habit (types of dishes a household usually consumes) and the availability of and ability to purchase different food items, such as grains.4 A household's use of time certainly involves fuel collection, cooking, childcare, cleaning, different types of income generating activities, and consumption (leisure). However, emphasis has been placed on time spent collecting fuel—particularly wood and dung—as the interest here is to assess how the adoption of an ICS affects household welfare. Finally, zi is a vector of household characteristics that includes the number of household members, household income, and so on. Note that the partial derivatives of this function (i.e., the marginal utilities) are not necessarily restricted to be positive. For example, increased consumption levels presumably increase household utility, whereas increased time spent collecting fuelwood reduces it. Given a household's utility function in Eq. (1), whether or not the household decides to adopt the new cooking stove depends on how the use of this new stove affects its utility. Random utility theory postulates that, for a given cost of adopting the new stove (the ICS), a household is more likely to adopt the new technology if the associated change in utility is positive (Manski, 2001). If E denotes the efficiency of the cooking technology employed, ΔE is the discrete change in energy efficiency associated with adopting the new technology; further, using index j to enumerate all arguments of household i's utility function (xji), the household is more likely to i adopt the new technology the larger is ΔU i ¼ ∑i ∂U ∂x ji

dxji dE

i ΔE. Here, ∂U is ∂x ji

the marginal utility of household i resulting from a change in argument dx

xj, and dEji is the total impact of the improved cooking technology on utility argument j. This total impact is the sum of the direct impact of the ICS   ∂x on argument j ∂Eji , plus the sum of all the indirect effects of all other variables in household i's utility function that are affected by the ICS   ∂x z on argument j ∑z ∂xzj ∂x . For example, if the ICS allows for spending ∂E less time collecting fuelwood, more time is available for other activities. to denote the probability of adoption by household i, we So, using PAdopt i   dx1i have P Adopt ¼ PðAdopt ΔE; dxdE2i ΔE; … where x1i = tscdi , i ¼ 1Þ ¼ f i dE x2i = ani, etc., and where the regression coefficients capture the marginal utilities associated with the variables of interest. Our identification strategy for the adoption process is as follows. We first estimate

dxji dE

¼

∂xji ∂E

∂x

z þ ∑z ∂xzj ∂x – the total impact of improved cook ∂E

stove adoption on those arguments of the utility function that are likely to be affected by the change in technology. Under the assumption that households adopting the ICS are not systematically different from those who did not adopt, we estimate the direct and indirect impacts of the ICS by comparing the consumption and time allocation decisions of households with the ICS to those without. Next, we estimate the relative importance of these impacts for adoption by regressing the adoption dummy variable, Adopti, on the vector of predicted changes d^ x Δ^x j ¼ dEj ΔE: To allow for household heterogeneity, we also include household characteristics like income, family size, et cetera, in the adoption regression equation. 4 It could be that they buy grains from the market and in the end cook it at home. However, it has to be noted that buying prepared/cooked meals is mostly not done by the target community.

340

Z. Gebreegziabher et al. / Energy Economics 67 (2017) 337–345

We proceed as follows. First, we test the validity of the key assumption that all households are drawn from essentially the same distribution by testing whether adopters and non-adopters differ systematically with respect to any of the key characteristics for which we have information. We obtain support for our assumption if we are unable to reject the hypothesis that there are no systematic differences in the household characteristics of adopters and non-adopters. Next, we esti  dx ∂x ∂x z ΔE for all i, noting that the impact mate Δx j ¼ dEj ΔE ¼ ∂Eji þ ∑z ∂xzj ∂x ∂E of ICS on one variable is not fully independent of the impact of ICS on another. Thus, the system of partial impacts is estimated using a three-stage least squares (3SLS) regression model, with all truly exogenous variables (e.g., location, household characteristics) as instrumental variables (IVs).5 Note that the 3SLS combines two-stage least squares (2SLS) with seemingly unrelated regressions (SUR), implying system estimation (Zellner and Theil, 1962). The key assumption here is that there are no systematic differences between households that have adopted the new stove and those that have not; they are drawn from essentially the same distribution.6 On the basis of these 3SLS regression models, we ∂^ x estimate ∂Eji by calculating ^x j ðAdopt ¼ 1Þ−^x j ðAdopt ¼ 0Þ. Third, the adoption decision is regressed on all partial technology impacts as well as a vector of key household characteristics. Specifically, the probability of household i adopting the ICS (i.e., Adopt = 1 as opposed to Adopt = 0) is determined as follows:
P i ðAdopt ¼ 1Þ ¼ f Δxji ; yi ; si ; li ; with Δxji as in above;

ð2Þ

where yi is exogenous income of household, si denotes household size (number of household members), and li denotes other household characteristics including location or agro-ecology conditions of the village where the household resides (upper or middle highlands vis-à-vis lower highlands) (Barrett et al., 2004). Our two-step procedure considerably mitigates the endogeneity problem of determining a household's propensity to adopt a new stove, as well as the main factors affecting that propensity, and the household-specific benefits the stove is expected to provide, especially given the households of both samples are drawn from the same distribution. If the households in the two samples do not differ systematically with respect to essential household characteristics, we can infer that all households are potential adopters of new stoves. 4. Data, survey design and context 4.1. Data and survey design Our data are from a survey of 200 households in Tigrai region, northern Ethiopia, collected during the year 2000. Different brands or versions of ICS have been promoted in the region (see below). This was the only dataset involving observations on improved biomass cookstoves that we could find in Ethiopia particularly as it pertains to the adoption of the three-stove Tigrai variant cookstove technology (see Gebreegziabher et al., 2005). Two-stage sampling was used to select the sample households. Four households were chosen at random from each of 50 tabias—the smallest administrative unit in the region— selected at random from a total of 600 tabias in the province. The (local) tabia administration is responsible for maintaining household lists from which the households targeted for interview were chosen. A survey method was employed in this research for data collection. A questionnaire that can generate the desired data was designed and pre5 We opted for IV-system (i.e., 3SLS) estimation instead of propensity score matching (PSM) because, if we use PSM, we would be matching with the IVs which should be avoided (see Wooldridge, 2009). 6 Statistically speaking, in order for the two sub-samples to be comparable and to carry on the necessary computations/estimations, the two sub-samples must be drawn from the same population or distribution. Therefore, this key assumption is guaranteeing or saying that the two sub-samples are essentially drawn from the same distribution.

tested at field level for verification and further modifications. The survey instrument was translated into the local language (Tigrigna) and administered to the participating households using two enumerators who were trained for the data collection. Both quantitative and qualitative data were collected on the household's collection and consumption of various biomass fuel types; time spent collecting various biomass fuel types; food cooking/baking frequency of household; demographic characteristics of the household including age, gender and literacy level of household head, and household size. Responses on the amount of the different biomass fuel types consumed by the household were collected in local units. Every local unit was measured for each household to facilitate conversion into metric units and minimize error.7 Also obtained from the survey were family resource endowments including total land ‘leased’, cultivated area, number of trees, livestock holdings, and type of stove used by the household. Also available from the survey are village level factors, including agro-ecological conditions or altitude range, and distance travelled (time spent) to collect different fuels. Data on cooking/baking frequency of a household was weighted for respective end-use share in the total household fuel consumption, using Annex Table A.2. Table 1 provides summary statistics of the variables considered in the empirical analysis. 4.2. Study context Research and development efforts to improve cook stove technologies began in Ethiopia in the 1980s with the World Bank Energy Sector Assessment (World Bank, 1984). Besides identifying short- to long-term options for alleviating the fuel problem/crisis in the country, the assessment also carried out kitchen-lab investigations of fuel-saving efficiency of various stoves. Stove types considered in the investigation and the test results are provided in Annex Table A.1. As it was quite apparent that injera baking – the production of a pancake like type of bread, typical to Ethiopia – accounts for about half of all the household fuel consumption, injera cookers received priority. The Tigrai type8 stove was found to be twice as efficient as open fire tripods with up to 25% fuel savings (World Bank, 1984). Two years later, a program of massive diffusion of efficient cooking stoves was designed, with the intention of dissemination of the Tigrai type stoves with some improvement or modifications. The intention was to induce all rural households to adopt this improved stove within 20 years, with the aim of reducing overall energy consumption over this period by a factor two (ENEC and CESEN, 1986a). This massive stove diffusion program was also envisaged to cover essentially all urban households using traditional fuels within the specified time period. Continuing improvement in stove designs was also expected in the years beyond the planning horizon, that is, beyond 2005, in order to allow a further decrease in the demand for traditional fuels of about 50%. The importance of an efficient extension service was recognized to support the diffusion of the efficient stoves.9 However, these stoves had no chimney, which is detrimental to family health as cooking areas filled with smoke. Hence, a second generation stove arose as the partially clay-enclosed stove was subsequently improved upon by the introduction of a stove that included a chimney and an even lower grate height10 and that was entirely enclosed 7 The enumerators or data collectors were provided with weighing balances upon departure for field work. The respondents are kindly requested to display a sample of the quantity of wood and/or dung collected at a time in local units. Then, measurement of the actual quantities (in metric units) of the local units of wood and dung displayed is taken using a weighing balance. 8 The Tigrai type stove was an indigenous innovation by the local people to the growing fuel scarcity and high fuel prices in the area (ENEC and CESEN, 1986b). 9 The extension service was regarded essential for providing assistance in the use of the stoves in new households and monitor the activities to determine the degree of appropriateness and utilization as well as the physical conditions of the stove (ENEC and CESEN, 1986a). 10 Refers to the height of the wall or insulation, i.e., from the hearth to the mogogo or baking plate.

Z. Gebreegziabher et al. / Energy Economics 67 (2017) 337–345 Table 1 Summary statistics of variables considered in the analysis (n = 200). Variable

Mean

Std. dev. Min. Max

Family size Adult males in household Adult females in household Age of household head Education of head of household (years of schooling) Sex of head Female Male Exogenous income (ETB/month) Number of cattle Land area (hectares) Cooking frequency (monthly) Wood price/shadow (ETB/h) Dung price/shadow (ETB/h) Wood consumption (kg/month) (adopters) Dung consumption (kg/month) (adopters) Wood consumption (kg/month) (nonadopters) Dung consumption (kg/month) (nonadopters) Kerosene consumption (lit/month) Households involved in tree planting Households not involved in tree planting Total no. of trees per household (all trees) Time collecting wood and/or dung (minutes/month) Use wood from own trees (=1; 0 otherwise) Yes (% of households involved) No Use improved stove (=1; or, 0 otherwise) Yes (% of households involved) No Location or agro-ecology (% of households involved) Upper highland Middle highland Lower highland (reference)

5 1 1 48 0.92

2 1 1 13 1.47

1 0 1 23 0

12 5 4 85 7

341

since improved stoves are constructed with the supervision of an extension/home agent, quality is assumed to be guaranteed. Hutton et al. (2006) argued that some households might continue to use their traditional stove for certain tasks. But, in our case, adoption requires that households dismantle their old stoves, so that one type of stove is in use at any given time. Moreover, there is generally inadequate room in a household's living quarters for more than one stove. 5. Estimation results

21% 79% 0.35 4 0.834 53.0 1.483 0.266 117.9 79.2 210.1 96.0 1.75 93% 7% 74 1850.4

2.86 3 0.496 19.7 7.285 0.849 86.3 91.2 100.8 96.1 6.89

0 0 0 12.7 0 0 0 0 0 0 0.11

25 14 2.5 210.3 18.376 3.618 420.0 508.3 1107.2 628.5 97.68

172 2032.6

0 0

1834 11,817.5

In this section we present estimation results. Before proceeding, it is necessary to check whether we can reject the hypothesis that the households in the two samples (those who have and those who have not adopted an improved stove) are drawn from the same distribution. We employ Mann-Whitney U tests to validate and check whether the two samples differ in terms of some key characteristics.12,13 The results do not allow us to reject the hypothesis that the two samples do not differ with respect to any of the individual household characteristics, suggesting that data or observations are drawn from the same population. Although this is not a definitive proof, it does give credence to our claim that it is the household-specific combination of characteristics that determines whether a household adopts a new stove. 5.1. The first-stage regression results

18 82 40.5 59.5

18 50 32

Note: ETB is the Ethiopian currency, Ethiopian Birr; $1 USD = 8.3279 ETB during the survey year.

(RTPC, 1998). The new technology consists of a baking oven, a stove for heating water and sauces, and a grain-roasting compartment. Thus, with little additional effort, the three stove Tigrai variant yielded more fuel savings, and it is this variant of stove that we examine in our empirical analysis. Dissemination of improved stoves in Tigrai also started before 1991. However, it was in the post-1991 period that it was more strengthened. For instance, a total of 77,563 new-technology ICS (i.e., the three stove Tigrai variant) were disseminated or built in rural Tigrai during the years 1991/92–1996/97 (BoANR, 1997).11 However, despite all these attempts, it is not clear whether the ICSs disseminated actually have the desired quality in terms of fuel savings particularly at the field level and whether there is a scope for further improvements. Moreover, it is not yet clear what factors determine the adoption of these fuel saving stoves. Therefore, there is a clear need for careful and rigorous empirical investigation. As is clear from Annex Table A.2, the fact that the vast majority of biomass fuel is used for baking and cooking as opposed to lighting and heating justifies the R&D effort on injera cookers (stoves). It is also worth noting that in the diffusion process peasants do not purchase these stoves but instead each person is taught how to build an improved stove based on information or advice provided by an extension agent. Advice/instruction about improved stove construction involved standard stove dimensions including grate height, gate area, etc. Moreover, 11 This job of improved stove dissemination was held under the Rural Women's Desk in the Bureau of Agriculture and Natural Resources (BoANR). The responsibility for technical design and stove development was under the Rural Technology Promotion Center (RTPC) within the BoANR (Gebretsadik et al., 1997).

We investigate the total impact of ICS on the various arguments in the household's utility function,

dx j dE

¼

∂x j ∂E

∂x

z þ ∑z ∂xzj ∂x . To do so, we first ∂E

estimate xj(.) for the subset of farmers who have adopted the ICS, and also for the subset of farmers who have not yet adopted it. We then esd^ x timate dEj by x^ j ðAdopt ¼ 1Þ−x^ j ðAdopt ¼ 0Þ. We do not impose any restriction that slope and/or intercept coefficients have to be identical across the two samples. We found that technology affects four key variables – time spent collecting dung, time spent collecting fuelwood, number of livestock, and cooking frequency. Time collecting dung and time collecting fuelwood are unlikely to be independent, and similar cross-relationships are likely to hold for other arguments in the utility   ∂x function ∂xzj ≠0 . This gives rise to endogeneity and simultaneity considerations that we address as follows. First, we conducted a Hausman test for endogeneity (see Verbeek, 2008), which caused us to reject the null hypothesis in favor of the alternative – that endogeneity is a concern. Because the model is overidentified, we also carried over identification tests (also known as Sargan J-tests). We cannot reject the null hypotheses that the instruments are uncorrelated with the error term in all cases suggesting the instruments that we use are valid.14 As a result, the four equations were estimated using three-stage least square (3SLS), that is, running 2SLS (two-stage least squares) or IV regression of the four equations as a system with the following instrumental variables: location (middle and upper highlands), family size and composition (i.e., number of adult females), land size, and the dummy variable ‘use wood from own trees’. 12 The Mann-Whitney U test is a non-parametric test. We use nonparametric tests' because our variables of interest are not normally distributed (Siegel and Castellan, 1988). First, all observations in each of the adopters and non-adopters sub-samples are ranked from highest to lowest with respect to the variable of interest. The two series are then merged. Upon going down the merged series, if the observations roughly alternate between the two sub-samples, then the Mann-Whiney U test fails to reject the assumption that observations are drawn from the same population. Alternatively, if observations from the top of the list all come from one sub-sample while those at the bottom from the other, the test rejects the null hypothesis of a common population (Harnett, 1982). 13 For completeness we also provide the test results (p-values) of the two-sided MannWhitney U tests with respect to whether the two samples differ in terms of these key characteristics in Annex Table A.3. 14 We carried out over identification tests, especially in relation to the equations (models) time spent collecting wood and time spent collecting dung and Χ2(1) = 0.000, yielding p = 0.995 in both models.

342

Z. Gebreegziabher et al. / Energy Economics 67 (2017) 337–345

Table 2 3SLS regression results for households that have adopted (n = 81) and have not adopted (n = 119) improved stove. Explanatory variable

Exogenous income (ETB/month) Family size (number) Family size squared Number of adult females (number) Land size (tsmdi)b Number of cattle (number of heads) Time collecting wood and/or dung (min/month) Wood from own trees (=1; else 0) Middle highland (=1; 0 otherwise) Upper highland (=1; 0 otherwise) Constant R

2

Adopter

Non-adopter

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Cooking frequency

Number of cattle

Time collecting wood

Time collecting dung

Cooking frequency

Number of cattle

Time collecting wood

Time collecting dung

0.045a (0.028) 8.931** (3.820) −0.764** (0.314)

0.009*** (0.004)

−1.794 (2.979) 105.793 (106.906)

1.271 (1.136) −19.004 (45.156)

0.032a (0.020) 8.645** (3.646) −0.585* (0.332)

0.007*** (0.003)

0.199 (1.923) 159.500* (89.110)

−0.033 (0.734) 13.760 (23.283)

260.180 (333.536) −680.773 (494.877)

224.481** (98.594) −64.241 (257.873) 6.338 (129.570)

1.479*** (0.508)

640.906** (294.599) −57.725 (391.651)

160.493** (73.494) −19.381 (158.236) 16.139 (90.968)

−0.508 (0.574) −0.725 (0.740) 1.531** (0.733) 0.2223

707.369 (498.363) 160.026 (434.613) −480.649 (550.395) −123.845 (685.972) 0.1527

1.898*** (0.632)

−0.004 (0.003)

29.066*** (9.601) 0.0603

−0.005** (0.003)

0.483 (0.570) −0.509 (0.731) 0.876 (0.752) 0.2723

−251.022 (488.509) −934.825** (443.828) −1115.262** (567.849) 1838.313*** (685.459) 0.0974

229.598** (113.641) 83.074 (164.182) −200.398 (175.073) 0.1959

34.255*** (10.473) 0.0710

209.414* (116.893) 144.842 (151.095) −120.192 (211.326) 0.0723

Notes: Standard errors in parentheses: ***, ** and * indicate significance at the 1%, 5% and 10% levels (or better), respectively. a p ≤ 0.11. b tsmdi is local unit for land area: 1 tsmdi = 0.25 ha.

The 3SLS estimator addresses endogeneity, is consistent and, in general, asymptotically more efficient than 2SLS (Kennedy, 2008). The regression results are provided in Table 2. Most of the variables in the two cooking frequency equations (columns 1 and 5 of Table 2) are statistically significant at the 10% level or better. All households cook more often the larger household income, the larger the family (albeit at a decreasing rate), and the less time they allocate to fuel collection, which is probably indicative of a readily available (nearby) source of fuel. Moreover, adopters cook less frequently than non-adopters, probably because of the convenience provided by the improved stove to combine baking and other cooking activities.15 Only exogenous income and the area of land ‘controlled’ by the household are found to be statistically significant variables explaining cattle ownership in both the non-adopter and adopter equations (columns 2 and 6 in Table 2). As expected, both variables contribute positively to the number of cattle a household will own. It would appear that households own cattle as a form of wealth, especially because private landownership is not permitted. Somewhat surprisingly, the household's location or agro-ecology is not found to affect the number of cattle it keeps.16 The most important factors explaining the amount of time allocated to collect wood in the non-adopters equation (column 7, Table 2) are family size (larger families need to collect more wood) and the number of adult females in the household (as adult females traditionally engage in fuel collection). Location or agro-ecology was found unimportant. In contrast, neither family size nor number of adult females is statistically significant in the adopters equation (column 3), but households located 15 In addition, perhaps the fuel savings attribute of the ICS might also enable adopting households to cook/bake more injera at a time with limited available amount of fuel(s) which could in turn possibly reduce their cooking frequency, hence, rendering the negative correlation. 16 Results are in reference/relation to lower highland. Many studies have used dummy variables to capture the effect of location or village agro-ecological conditions (see, e.g., Pender et al., 2006).

in the middle or upper highlands spend less time on wood collection, with coefficients for these location dummy variables statistically significant at the 5% level or better. Families that have adopted the improved stove spend less time collecting wood as such stoves are more efficient in their use of wood. Interestingly, exogenous income, land area, and whether or not one uses wood from trees located on the homestead are not found to be important determinants of time spent collecting fuelwood. The number of adult females and the household's location (in the middle highlands) provide a statistically significant explanation of household time spent on dung collection (columns 4 and 8, Table 2). However, unlike the case for fuelwood, number of adult females is statistically significant in the adopters as well as non-adopters equation. And only the middle highlands dummy variable is statistically significant, but then in both equations. Surprisingly, while dung collection time would be expected to be inversely related to the number of cattle owned by the household, it turned out statistically insignificant and positive (both in the non-adopter and adopter equations). Exogenous income has the expected positive sign, but is only statistically significant in the adopters' equation. Also family size and the size of the land area are found to be statistically insignificant determinants of time spent collecting dung. We now proceed by using the results of Table 2 to determine savings from adopting the new stove technology. The predicted savings—when using an improved stove—in cooking frequency, time spent collecting fuelwood, time spent collecting dung, and livestock numbers are provided in Table 3. As indicated above, these are obtained by multiplying the differences in the estimated coefficients with the householdspecific values of the explanatory variables. The difference in intercepts   ∂x provides us with an estimate of ∂Ej ΔE , and the other arguments can   ∂x z be used to calculate ∂xzj ∂x ΔE . ∂E We find that the use of an improved stove is correlated with lower cooking frequencies, less time spent collecting fuel (both wood and dung), and increased cattle ownership. As indicated by the negative

Z. Gebreegziabher et al. / Energy Economics 67 (2017) 337–345

343

Table 3 Predicted time and other savings from adopting an improved stove, standard error (in parentheses) and t-tests of difference from zero. Item

Cooking frequency

Number of cattle

Time collecting wood

Time collecting dung

Fuelwood (kg/month)a

Dung (kg/month)a

Predicted savings ðΔ^xi Þ

4.697 (0.708) 6.63

−0.599 (0.142) −4.22

472.665 (66.171) 7.14

40.840 (18.950) 2.15

68.278 (22.575) 3.02

19.899 (11.371) 1.75

t-Values

Note: a Predicted savings obtained from estimated derived demand functions.

savings in Table 3, household's cattle holding increases by about 0.6 animals on average probably because cattle are kept as a store of wealth (i.e., asset) and as a status symbol (Gryseels, 1988; Mogues, 2006; AACCSA, 2015). The latter result also indicates that, at the margin, any change that affects the value of cattle to the household affects livestock holdings. Thus, even though cattle might be kept because they also provide dung for cooking and now less dung is required per household, each family will strive to increase their cattle holding. The savings in terms of woody biomass and dung are also provided in Table 3. The estimated savings are obtained by comparing the predicted demands for adopters and non-adopters of the improved stove technology. The results are interesting because they suggest that adoption of the new stove reduces harvesting pressure on local forest stands. Not only do the times spent collecting dung and wood go down, but, as a result, less wood and dung are used for cooking purposes. On a per household basis, we calculate adopters collect 68.278 kg less wood each month and about 19.899 kg less dung each month. These results were found to be significant at the 1% and 10% levels, respectively. However, results also suggest that adoption of an improved stove could have mixed environmental benefits. Grazing pressure on communal lands is likely to go up, as the number of cattle increases by an average of 0.6 per household. 5.2. The adoption of improved cooking stoves in Tigrai We can now determine the factors that probably affect the adoption decision (Eq. (2)). Apart from the predicted savings in cooking frequency, cattle holding, and the amount of time allocated to collect fuelwood or dung, we hypothesize that the decision to adopt an improved cooking stove also depends on other household characteristics, including household income, size, and location. Other adoption studies have focused on socio-economic factors characterizing adoption such as the literacy level or education of head, attitude, etc. However, it is important to note that unlike other adoption studies the focus of this study is on the role of economic savings in inducing adoption. The argument in here is that it is the utility gains, i.e., expected savings, in terms of cooking frequency, time spent collecting fuels, and the number of cattle that ultimately matter whether or not the household adopts the stove. Availability of credit was also considered unimportant as the innovation/technology in question is self-made and does not involve any cash outlay. The results of the probit regression are presented in Table 4. The results are revealing. The savings in cooking frequency, time spent collecting wood, and cattle number are all statistically significant factors explaining adoption. Cattle or livestock in the study area are regarded as store house of value. It could be that gains in fuel savings are invested in cattle. The saving in time collecting dung is not found to be an important factor in the adoption decision, even though one would expect time spent collecting dung to decline as a result of adopting the new stove (perhaps because of low energy value and low magnitude of saving in case of dung). We also find that, having controlled for the impact of household characteristics on the households' savings, their direct impact on the decision to build a new stove is negligible. Only households located in the upper highlands and in the middle highlands are found to be less likely to adopt new stoves.

6. Discussion Improved biomass cookstoves have been regarded as technological alternatives to alleviate or mitigate the fuelwood problem (crisis), particularly since the oil price shocks of the 1970s, as their greater energy efficiency can overcome fuel shortages. However, knowledge about the characteristics of stove adoption is sparse. By and large, empirical evidence regarding the effectiveness or fuel saving efficiency of these improved stoves, particularly biomass cookstoves, is scanty and mixed. Using a dataset on cross-section of 200 farm households from the highlands of Tigrai region, northern Ethiopia, we investigated the role of disseminating improved stoves in redressing the fuel problem and its implications for the environment, as well as the factors that affected the stove adoption decision. Key questions involved are: How effective are the improved biomass cookstoves being promoted as a remedy to the fuel problem and land degradation? Should it matter that the stoves are merely saving time or biomass and not money? What characterizes stove adoption? The results indicate that peasants in Tigrai region, Ethiopia, are willing to adopt new technologies if these result in economic savings. Barnes et al. (1993) argue that improved stove programs have been most successful when targeted to specific areas where fuelwood prices or collection time are high. In this study, we found that the adoption of a more energy efficient or improved cookstove is proportional to the economic savings in fuel collection time, cooking frequency, and cattle required for everyday purposes. Our research also suggests that there may be a significant positive environmental externality in terms of slowing down the degradation of agricultural and forested lands. Unlike the findings of Arnold et al. (2006), our results indicate that it should not matter whether stoves merely save time or biomass. In fact, results suggest that expected savings, in terms of cooking frequency and time spent collecting wood, would also provide sufficient incentives to induce stove adoption decision. We find that the use of an improved biomass cookstove is correlated with lower cooking frequencies, less time spent collecting fuel (both wood and dung), and increased cattle Table 4 Probit regression of the adoption of an improved cooking stove in Tigrai, Ethiopia (n = 200). Source: Authors' calculations. Explanatory variable

Estimated coefficienta

Standard error

Saving in cooking frequency Saving in cattle numbers Saving in time collecting fuelwood Saving in time collecting dung Household income Household income squared Family size Middle highlands (=1; otherwise 0) Upper highland (=1; otherwise 0) Constant LR χ2(9) Pseudo R2 Predicted prob. of adopting ICS

0.0160* 1.5606** 0.0009* 0.0057 0.0167** −0.00002b −0.3482 −1.2333*** −1.8988** −1.3884 17.59** 0.0652 0.2884

0.0092 0.7587 0.0005 0.0040 0.0076 0.00001 0.2188 0.4395 0.7852 0.9657

(at the mean)

Notes: a ***, ** and * indicate statistically significant at 1%, 5% and 10% level (or better), respectively. b p-Value = 0.108.

344

Z. Gebreegziabher et al. / Energy Economics 67 (2017) 337–345

ownership. The latter result is suggestive of the fact that cattle in the study area are kept as a store of wealth and as a status symbol. In traditional societies, like the community in question, irrespective of where the savings come from, it is kept in the form of cattle or livestock. Based on our findings, improved biomass cookstoves appear to reduce environmental degradation in three ways: (1) By switching to an improved stove as opposed to the traditional one, less dung is collected as fuel so more manure is available to benefit the soil. (2) Adoption of improved stoves results in less wood used as fuel, ceteris paribus, thus reducing deforestation pressure. As a result, more wood is available for others (non-adopters and adopters), which implies less dung and crop residues will be used for fuel. (3) Finally, through its effect on time savings, cookstove adoption results in less time spent collecting fuelwood and dung, and in less time spent cooking. If labor markets also function fairly well, this would mean that more time is available for off-farm work, leading to less time spent in agricultural and forestry activities. That, in turn, would imply reduced pressure on forests and agricultural land. However, grazing pressure on communal lands is likely to go up, as the number of cattle per household increases. The importance of improved biomass cookstoves can be determined from our results. For example, consider a setting/region where there are some 700,000 rural households (CSA, 2008). Given that a household will adopt a new stove with probability 0.2884 (see Table 4), this implies that some 201,000 households are likely to adopt the more efficient stove technology. Given that each adopting household collects on average 68.3 kg less fuelwood and 19.9 kg less dung per month (see Table 3), total potential savings could amount to 165,461 t of fuelwood and 48,209 t of dung per year. Clearly, these estimates are based on potential, which assumes that all dung not collected would have been used in agriculture, and that all fuelwood savings come from forested areas. Nonetheless, in terms of wood alone, assuming an average of 120 t of biomass per ha (much higher than the current average), the potential reduction in deforestation amounts to nearly 1379 ha per year, which is not an inconsequential saving. Furthermore, assuming that there are almost 1 million ha of cropland in the setting/region in question (Gebreegziabher et al., 2011), the dung saving translates into about 0.05 of an additional ton of organic matter per hectare per year, again a substantial benefit. Last but not least, the current study investigated the adoption and fuel-savings efficiency of one set of improvements to the biomass cookstove. However, there exist many types or variants of improved biomass cookstoves in Ethiopia and Tigrai province in particular. Therefore, there is a need for further research as well regarding the fuel-savings efficiency of the other improved biomass cookstoves. Acknowledgements The authors thank the Agricultural Economics and Rural Policy Group at Wageningen University, the Netherlands, and the REPA Research Group at the University of Victoria, Canada, for research support. We also thank two anonymous referees for their thoughtful comments and constructive suggestions on an earlier version of the paper. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.eneco.2017.08.030. References AACCSA (Addis Ababa Chamber of Commerce and Sectoral Associations), 2015. Value Chain Study on Meat Processing Industry in Ethiopia, “Strengthening the Private Sector in Ethiopia” Project Finance by the Danish Embassy in Ethiopia. AACCSA, Addis Ababa, Ethiopia (http://addischamber.com/wp-content/uploads/2017/01/ValueChain-study-on-Meat-Processing.pdf (retrieved May 2017)). Agarwal, B., 1983. Diffusion of rural innovations: some analytical issues and the case of wood-burning stoves. World Dev. 11, 359–376.

Amacher, G.S., Hyde, W.F., Kanel, K.R., 1996. Household fuelwood demand and supply in Nepal's Tarai and Mid-Hills: choice between cash outlays and labor opportunity. World Dev. 24, 1725–1736. Amacher, G.S., Ersado, L., Hyde, W.F., Osario, A., 2004. Tree planting in Tigray, Ethiopia: the importance of human disease and water microdams. Agric. Syst. 60, 211–225. Amsalu, A., 2006. Caring for the Land: Best Practices in Soil and Water Conservation in Beressa Watershed, Highlands of Ethiopia. (PhD thesis). Wageningen University, The Netherlands. Angelsen, A., Brockhaus, M., 2009. Realising REDD +: National Strategy and Policy Options. Center for International Forestry Research, Bogor. Arnold, J.E.M., Köhlin, G., Persson, R., 2006. Woodfuels, livelihoods, and policy interventions: changing perspectives. World Dev. 34, 596–611. Balietti, A., Datta, S., 2017. The Impact of Indoor Solid Fuel Use on the Stunting of Indian Children. Center for International Development, Harvard Kennedy School of Government, mimeo. Barnes, D.F., Openshaw, K., Smith, K.R., van der Plas, R., 1993. The design and diffusion of improved cooking stoves. World Bank Res. Obs. 8, 119–141. Barnes, D.F., Krutilla, K., Hyde, W., 2004. The Urban Household Energy Transition: Energy, Poverty, and the Environment in the Developing World. Energy Sector Management Assistance Program (ESMAP), World Bank, Washington, DC http://www.esmap.org/ esmap/sites/esmap.org/files/Rpt_UrbanEnergyTransition.pdf, Accessed date: 21 May 2012. Barrett, C.B., Moser, C.M., McHugh, O.V., Barison, J., 2004. Better technology, better plots, or better farmers? Identifying changes in productivity and risk among Malagasy rice farmers. Am. J. Agric. Econ. 86, 869–888. Bluffstone, R., 1989. Improved cookstove promotion and the proper public role: a case study of Salyan, Nepal. A Report for the National Cooperative Business Association and US Agency for International Development, Kathmandu, Nepal. BoANR (Bureau of Agriculture and Natural Resources), 1997. Agricultural Development and Natural Resource Activities 1978/79–1996/97. BoANR, National Regional State of Tigrai, Mekelle, Ethiopia. Boucher, D., Elias, P., Lininger, K., May-Tobin, C., Roquemore, S., Saxon, E., 2011. The Root of the Problem - What's Driving Tropical Deforestation Today? Union of Concerned Scientists (UCS), Cambridge, MA Chen, L., Heerink, N., van den Berg, M., 2006. Energy consumption in rural China: a household model for three villages in Jiangxi Province. Ecol. Econ. 58, 407–420. Cooke, P., Köhlin, G., Hyde, W.F., 2008. Fuelwood, forests and community management—evidence from household studies. Environ. Dev. Econ. 13, 103–135. CSA (Central Statistical Authority), 2008. Summary Statistical Report of the 2007 Population and Housing Census. CSA, Addis Ababa, Ethiopia. Dadi, L., Burton, M., Ozanne, A., 2004. Duration analysis of technological adoption in Ethiopian agriculture. J. Agric. Econ. 55, 613–631. EFAP (Ethiopian Forestry Action Plan), 1994. Vol II. The Challenges to Development. EFAP Secretariat, Ministry of Natural Resources Development and Environmental Protection, Addis Ababa, Ethiopia. ENEC and CESEN (Ethiopian National Energy Committee and Centro Studio Energia), 1986a. Co-operation Agreement in the Energy Sector: Main Report. Ministry of Mines and Energy and CESEN-Ansaldo/Finmeccanica Group, Addis Ababa, Ethiopia. ENEC and CESEN (Ethiopian National Energy Committee and Centro Studio Energia), 1986b. Cooperation Agreement in the Energy Sector: The Rural/Urban Household Energy Survey Part 1; Design, Organization and Basic Results. Ministry of Mines and Energy and CESEN-Ansaldo/Finmeccanica Group, Addis Ababa, Ethiopia. FAO (Food and Agriculture Organization of the United Nations), 2006. http:// www.fao.org/forestry/foris/webview/forestry2/index.jsp?siteId=6833&sitetreeId= 32006&langId=1&geoId=0, Accessed date: 1 June 2006. FAO (Food and Agriculture Organization of the United Nations), 2010. Global Forest Resources Assessment 2010 - Country Report Ethiopia. Food and Agriculture Organisation (FAO), Rome, Italy www.fao.org/forestry/fra/fra2010/en/, Accessed date: 15 September 2010. Feder, G., Just, R.E., Zilberman, D., 1985. Adoption of agricultural innovations in developing countries: a survey. Econ. Dev. Cult. Chang. 33, 255–298. Gebreegziabher, Z., 2007. Household Fuel Consumption and Resource Use in Rural-Urban Ethiopia. (PhD Thesis). Wageningen University, The Netherlands. Gebreegziabher, Z., van Kooten, G.C., 2013. Does community and household tree planting imply increased use of wood for fuel? Evidence from Ethiopia. Forest Policy Econ. 34, 30–40. Gebreegziabher, Z., Gebremedhin, B., Mekonnen, A., 2011. Institutions, sustainable land use, and consumer welfare: the case of forest and grazing lands in northern Ethiopia. Environ. Dev. Econ. 17, 21–40. Gebreegziabher, Z., Mekonnen, A., Kassie, M., Köhlin, G., 2012. Urban energy transition and technology adoption: the case of Tigrai, northern Ethiopia. Energy Econ. 34 (2), 410–418. Gebreegziabher, Z., van Kooten, G.C., van Soest, D.P., 2005. Land degradation in Ethiopia: what do stoves have to do with it? University of Victoria Working Paper 2005–16 Gebretsadik, G., Tesfay, A., Abreha, T., Hassen, E., 1997. Tigray Rural Household Energy Development: Issues and Strategies. BoANR, National Regional State of Tigray, Mekelle, Ethiopia. Geist, H.J., Lambin, E.L., 2002. Proximate causes and underlying driving forces of tropical deforestation. Bioscience 52 (2), 143–150. Gryseels, Guido, 1988. Role of Livestock on Mixed Smallholder Farms in the Ethiopian Highlands. A Case Study From the Baso and Worena Wereda Near Debre Berhan. (Dissertation). Agricultural University, Wageningen, The Netherlands. Harnett, D.L., 1982. Statistical Methods. Adison-Wesley, Reading, Massachusetts. Hutton, G., Rehfuess, E., Tediosi, F., Weiss, S., 2006. Evaluation of the Costs and Benefits of Household Energy and Health Interventions at Global and Regional Levels. World Health Organization (WHO), Switzerland.

Z. Gebreegziabher et al. / Energy Economics 67 (2017) 337–345 IFAD (International Fund for Agricultural Development), 2010. Rural Poverty Report 2011: Overview. IFAD, Rome http://www.ifad.org/rpr2011/report/e/overview.pdf, Accessed date: 21 December 2010. IPCC (Intergovernmental Panel on Climate Change), 2007. Climate Change 2007: Synthesis Report, IPCC Fourth Assessment Report. IPCC, Geneva, Switzerland. Johnson, D.L., Ambrose, S.H., Bassett, T.J., Bowen, M.L., Crummey, D.E., Isaacson, J.S., Johnson, D.N., Lamb, P., Saul, M., Winter-Nelson, A.E., 1997. Meanings of environmental terms. J. Environ. Qual. 26, 581–589. Kennedy, P., 2008. A Guide to Econometrics. 6th edition. Wiley-Blackwell, Malden, MA. Larson, B.A., Rosen, S., 2002. Understanding household demand for indoor air pollution control in developing countries. Soc. Sci. Med. 55, 571–584. Lund, H.G., 2012. Definitions for degradation of forests and devegetation of other vegetation types. Review of Published Definitions of Degraded Forestland, Devegetation, and Related Terms and Antonyms. Chapter 2. Prepared for the IPCC NGGIP-LULUCF Programme. Manski, Ch.F., 2001. Daniel McFadden and the econometric analysis of discrete choice. Scand. J. Econ. 103 (2), 217–229. MEFCC (Ministry of Environment, Forest and Climate Change), 2016. Ethiopia's forest reference level submission to the UNFCCC. http://redd.unfccc.int/files/2016_submission_frel_ethiopia.pdf, Accessed date: May 2016. Moges, Y., Eshetu, Z., Nune, S., 2010. Ethiopian Forest Resources: Current Status and Future Management Options in View of Access to Carbon Finances (Unpublished). Mogues, Tewodaj, 2006. Shocks Livestock Asset Dynamics and Social Capital in Ethiopia DSGD Discussion Paper No 38. IFPRI, Washington DC. Newcombe, K., 1989. An economic justification for rural afforestation: the case of Ethiopia. In: Schramm, G., Warford, J.J. (Eds.), Environmental Management and Economic Development. Johns Hopkins University Press, Baltimore, US. Nkonya, E., Gerber, N., von Braun, J., De Pinto, A., 2011. Economics of Land Degradation: The Costs of Action Versus Inaction. IFPRI Issue Brief 68. Pearce, D.W., Warford, J.J., 1993. World Without End. Oxford University Press, Oxford. Pender, J., Place, F., Ehui, S. (Eds.), 2006. Strategies for Sustainable Land Management in the East African Highlands. International Food Policy Research Institute (IFPRI), Washington, DC. RTPC (Rural Technology Promotion Center), 1998. Biomass Stove (Mogogo Eton) Test Report. RTPC, Mekelle, Ethiopia. Sagar, A.D., Kartha, S., 2007. Bioenergy and sustainable development? Annu. Rev. Environ. Resour. 32 (1), 131–167.

345

Scherr, S.J., 2000. A downward spiral? Research evidence on the relationship between poverty and natural resource degradation. Food Policy 25 (4), 479–498. Siegel, S., Castellan Jr., N.J., 1988. Nonparametric Statistics for the Behavioral Sciences. 2nd edition. McGraw-Hill, New York, p. 1988. Sinton, J.E., Smith, K.R., Peabody, J.W., Yaping, L., Xiliang, Z., Edwards, R., Quan, G., 2004. An assessment of programs to promote improved household stoves in China. Energy Sustain. Dev. 8, 33–52. Sullivan, K., Barnes, D., 2006. Energy policies and multitopic household surveys: guidelines for questionnaire design in living standards measurement studies. Energy and Mining Sector Board Paper No. 17. World Bank, Washington DC. UN, 2015. The changing face of Ethiopia, Africa Renewal Magazine, August 2015. http:// www.un.org/africarenewal/magazine/august-2015/changing-face-ethiopia (accessed August 2017). UNDP (United Nations Development Program), 2002. Preventing Land Degradation Sustaining Livelihoods: Experiences from GEF and UNDP Projects. UNDP, New York. van Kooten, G.C., Bulte, E.H., 2000. The Economics of Nature. Blackwell, Oxford. Verbeek, M., 2008. A Guide to Modern Econometrics. 3rd edition. John Wiley & Sons Ltd., West Sussex. WBISPP, 2005. A National Strategy Plan for the Biomass Sector. MoARD, Addis Ababa, Ethiopia. WBISPP (Woody Biomass Inventory and Strategic Planning Project), 2004. Forest Resources of Ethiopia. MoARD, Addis Ababa, Ethiopia. Wirl, F., 2000. Lessons from utility conservation programs. Energy J. 21, 87–108. Wooldridge, J.M., 2009. Should Instrumental Variables Be Used as Matching Variables? Department of Economics, Michigan State University, East Lansing, MI World Bank, 1984. Issues and options in the energy sector. Report No. 4741-ET. The World Bank, Washington, DC. World Bank, 2006. Rural Energy Strategy for Bangladesh. The World Bank, Washington DC. World Bank, 2014. Ethiopia overview. http://www.worldbank.org/en/country/ethiopia/ overview (retrieved 22 July 2014). World Bank, 2016. Poverty and Shared Prosperity 2016: Taking on Inequality. The World Bank, Washington DC. Zellner, A., Theil, H., 1962. Three-stage least squares: simultaneous estimation of simultaneous equations. Econometrica 30 (1), 54–78.