ELSEVIER Energy Economics 19 (1997) 465 475

Energy Economics

Improved stoves in Sub-Saharan Africa: the case of the Sudan

E i m a n O. Ze in -E labd in

Franklin and Marshall College, Economics Department, P.O. Box 3003, Lancaster, 17604, USA

Abstract

The paper estimates charcoal demand and supply elasticities to determine rebound effects from improved stoves in the Sudan. These are increases in fuel consumption resulting from gains in real income upon the use of more efficient appliances, and from downward price adjustments associated with the reduction in fuel requirements. The findings are that: (1) charcoal markets are characterized by low elasticities; and (2) 42% of fuel savings are lost, mostly as a result of large price adjustments since low elasticities place more of the burden of market adjustment on prices than on quantities. Nonetheless, price-related effects may be small because of increases in charcoal prices as a result of accelerated deforestation. © 1997 Elsevier Science B.V.

JEL classification: Q41

Keywords: Improved stoves; Deforestation; Fuelwood

1. Introduct ion

Biomass provides over 70% of total energy supplies in Sub-Saharan Africa. This pressure on wood resources is one of the main causes of forest depletion in the region. Three to four million hectares of land were annually cleared in the 1980s, resulting in decline of the main sources of foreign exchange receipts, namely agriculture and tourism, due to soil erosion and habitat loss. One of the measures undertaken to combat deforestation in Sub-Saharan Africa is improving the effi- ciency of fuelwood (firewood, charcoal) utilization through the introduction of improved cooking stoves in households, where over 80% of all fuelwood is consumed.

0140-9883/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved P I I S 0 1 4 0 - 9 8 8 3 ( 9 7 ) 01025-0

466 E.O. Zein-Elabdin / Energy Economics 19 (1997) 465-475

The ability of improved stoves to engender a significant and lasting reduction in fuelwood consumption is jeopardized by the presence of second-order, or rebound, effects, in the form of further consumption of fuelwood. This consumption may result from gains in real income generated by the use of more efficient appliances (income effects), or from downward adjustments in fuelwood prices following the initial reduction in fuel requirements (price effects))

This paper examines the effectiveness of improved stoves in slowing the pace of forest depletion in Sub-Saharan Africa by evaluating the rebound effects associated with them in a case study of the Sudan. The study involves econometric estimation of some demand and supply parameters in the market for charcoal in the capital city of Khartoum, as well as analysis of the role different elasticities play in the generation of rebound effects. The paper argues that in theory price effects may be substantial because fuelwood markets are characterized by relatively low elastici- ties that restrict quantity adjustments, thereby resulting in large price movements. However, in reality price effects may be rather small because the downward price adjustments following the introduction of improved stoves may be offset by upward pressures as a result of the acceleration of deforestation and subsequent shortages of fuelwood.

This work has been motivated by the dearth of research in this area. Even though improved stoves were introduced almost half a century ago, and have since spread widely, their impact on forest depletion has not been adequately explored. This is due to narrowness of the database and lack of understanding of the market for traditional fuels. As a result, evaluation of improved stove programs has been limited to aspects of technical design and diffusion strategies, leaving out the important issue of their effect on the demand for fuelwood.

Only two attempts have so far been made at gauging the behavioral response to improved stoves. Using various economic parameters from a range of countries, Jones (1988, p. 262) concluded that leakages in fuel savings from improved stoves could amount to 60% of designed capacity. Previously, on the basis of a general equilibrium, input-output model in which household income elasticities were only implicitly included, Murck et al. (1985, p. 1231) estimated that such leakages reduced the efficiency of improved stoves in the Sudan by 48%.

The present study uses the parameters from two demand and supply equations to obtain the second-order effects of improved stoves, thereby using fewer data points. This also allows explicit analysis of the role-played by the different elastici- ties in determining the effectiveness of improved stoves, and therefore the results lend themselves to clear policy recommendations. The results have important implications for countries and organizations involved in efforts to arrest deforesta- tion, and for those concerned with biomass energy in Africa.

The following section outlines the scope and method of analysis. Section 3 reviews the econometric results. Section 4 calculates the rebound effects of improved stoves. Finally, Section 5 discusses the implications of the findings.

1The idea of rebound effects has long been debated in the context of modem energy. Khazzoum (1980), held that the use of more efficient appliances would have the same effect as would a drop in the price of fuel.

E.O. Zein-Elabdin / Energy Economics 19 (1997) 465-475 467

2. Scope of analysis, method, and data

Demand and supply elasticities for charcoal were estimated for the city of Khartoum, and the rebound effects of improved stoves were calculated on the basis of the estimated parameters. Although the Sudan has been selected here partly as a matter of data availability, this country displays most of Sub-Saharan Africa's economic and environmental ailments. Per capita income is estimated at less than $300 per year. Modern fuels provide less than 20% of total energy supplies. Annual deforestation stands at 504000 h, the highest in the region (World Resources Institute, 1991, p. 292). The Sudan also has been the site of a number of biomass energy initiatives, including the major USAID effort in Africa involving the introduction of improved stoves. Consequently, the Sudan provides a reasonable case for generalizing findings to Sub-Saharan Africa.

The analysis is limited to charcoal for a number of reasons. It is used by the majority of urban households. Unlike firewood, it is fully commercialized, thereby allowing better observation of market relations. Most improved stoves introduced in Sub-Saharan Africa are designed to burn charcoal. Moreover, the environmental damage associated with charcoal use is much greater vis-h-vis firewood. The latter is mostly obtained by rural households, gathering branches and twigs, while charcoal is produced by clearing vast tracts of land to supply large urban popula- tions, therefore imposing intensified pressure on local ecosystems.

The research method is econometric estimation of a single household demand equation and a supply function, using time series data. The two equations are estimated independently of one another. Simultaneous solution was precluded by the fact that the two equations are derived from separate data sets. a Both equations are in double-logarithmic form to allow direct interpretation of the estimated coefficients as elasticity measures.

The demand equation is derived from a continuous utility function, and depicts consumption as a function of price and income. In linear form the demand equation is

In Qdt = In A + a In Pet + 6 In Yt + / x (1)

where Qdt is household consumption of charcoal at time t, A is a constant, Pet is the price of charcoal, Yt is income, both at time t, and /z is an error term. The parameters o~ and 6 are the price and income elasticities respectively.

The supply equation shows quantities supplied to Khartoum as a function of charcoal prices. After taking logarithms, supply can be expressed as

In Qst = In D + 4) in Pet + /X (2)

2Th i s is a typical problem in the analysis of fuelwood markets. Consumption data are collected by energy departments, while supply information is handled by forestry personnel.

468 E.O. Zein-Elabdin / Energy Economics 19 (1997) 465-475

where Qst indicates quantities of charcoal supplied at time t, D is a constant, and 4~ is the supply elasticity)

The hypothesis is that the market for charcoal is characterized by relatively low elasticities. The literature on traditional fuels uniformly indicates income elastici- ties below unity. As will be discussed later, previous studies of fuelwood in Sub-Saharan Africa frequently show charcoal to be an inferior fuel. Price elasticity is expected to be relatively low owing to the lack of close substitutes. The use of modern fuels is prohibited by a significant price differential. On the supply end, various rigidities, such as transportation bottlenecks, result in a small supply elasticity.

The implication of these low elasticities for the charcoal market is that quantity movements would be limited, meaning that the bulk of market adjustments must be borne by prices. These small quantity responses are the reason why the introduc- tion of improved stoves is likely to be accompanied by sizable price adjustments. In other words, it is expected that price-related rebound effects from improved stoves will be relatively large.

Rebound effects are calculated using the derived elasticity estimates, as well as household budget shares. The supply elasticity is used to infer the approximate size of the price adjustments mentioned above. This procedure is a simple attempt to circumvent data shortages that often hinder the analysis of important issues in many African countries. Comparison of these findings with those of previous studies below will show that the procedure yields satisfactory results while tremen- dously reducing data requirements. Ordinary least squares were used. Lack of data ruled out a more appropriate technique such as instrumental variables estimation. Tables 1 and 2 give the data used for the study.

3. Estimation results

Table 3 shows the results of estimation (Zein-Elabdin, 1993). Overall, a good fit was obtained. All coefficients are significant. The demand and supply equations have their expected signs. Household consumption of charcoal exhibits the stan- dard characteristics of a demand relationship, being an inverse function of price. Charcoal is a normal good, consumption varying directly with income. The quanti- ties of charcoal supplied are directly determined by price.

All elasticities are below unity, consistent with the hypothesis stated earlier. The price coefficient (-0.55) reflects that quantity demanded responds less than proportionately to price changes. A 10% rise in the price of charcoal will, ceteris paribus, reduce household consumption by 5.5%. This is only the second estimate of price elasticity for traditional fuels in Sub-Saharan Africa. The price elasticity of

3To test for identification the supply equation was also estimated with charcoal transportation cost as an independent variable since it is not a shift variable in the demand equation.

E.O. Zein-Elabdin / Energy Economics 19 (1997) 465 475

Table 1 Khartoum households data

469

Year Charcoal consumption Charcoal price Income c (TOE) a (Ls/sack) b (Ls million)

1960 1591.60 0.64 38.00 1961 1943.28 0.64 48.00 1962 2345.40 0.64 49.00 1963 2922.48 0.65 55.00 1964 3713.75 0.67 62.00 1965 4898.05 0.68 66.00 1966 6260.30 0.70 83.00 1967 8385.30 0.70 95.00 1968 11 437.20 0.70 101.00 1969 16474.90 0.72 114.00 1970 23 340.00 0.75 133.00 1971 32257.60 0.80 159.00 1972 34244.37 0.85 216.00 1973 36959.04 0.90 324.00 1974 38 852.10 1.00 400.00 1975 42215.80 1.30 500.00 1976 45 664.45 1.50 632.00 1977 49747.62 1.75 807.00 1978 52 717.50 2.00 1004.00 1979 55 044.00 2.40 1181.00 1980 58 249.10 3.00 1442.00 1981 60 686.20 3.00 1819.00 1982 62 641.00 3.50 2331.00 1983 64 946.00 5.00 2609.00 1984 67 157.00 6.00 3247.00 1985 70381.00 9.00 7383.00 1986 73 759.00 10.00 9564.00 1987 77 300.00 35.00 12490.00 1988 99 778.00 91.67 22 069.00 1989 103 666.00 80.00 44 491.00 1990 108564.00 93.00 80000.00

Sources: Sudan National Energy Administration (1987, 1991) and International Monetary Fund (1970, 74, 77, 79, 82, 83, 85, 90). aTons of oil equivalent. b Ls is Sudanese pounds. The average mass of a sack of charcoal is 36 k. CNominal GDP is used as a proxy for total household income. Deflation raised multicolinearity sharply.

demand for fuelwood in Ethiopia was previously estimated at - 0.37 (Kidane, 1991, p. 133), a level certainly comparable with the present estimate.

Income elasticity is 0.87, implying that household income exerts a stronger effect on charcoal consumption than do prices. Although in size it is consistent with the literature, the sign of the income elasticity is contrary to earlier evidence showing charcoal to be an inferior fuel in Sub-Saharan Africa. Previous studies indicate income elasticities of -0 .59 and -0 .23 for the Sudan and Kenya respectively

470

Table 2 Charcoal supply data

E.O. Zein-Elabdin / Energy Economics 19 (1997) 465-475

Year Quantity (tons) Price (L/sack)

1972 172 270 0.80 1973 175 808 1.25 1974 181506 1.30 1975 201 640 1.50 1976 208418 1.50 1977 218 395 2.00 1978 227 213 2.50 1979 237 785 3.00 1980 247 488 3.75

Source: Mukhtar (1982).

Table 3 Estimation results

Equation Parameters a R 2 DW

b ce 6 Demand 4.90 - 0.55 0.87 0.95 0.697

(0.697) ( - 0.18) (0.13)

d Supply 12.10 0.025 0.90 1.432

(0.02) (0.03)

Source: Zein-Elabdin (1993). ab and d are the coefficients for the constant term in the demand and the supply equations respectively,

is own-price elasticity, 6 is income elasticity, and 4~ is supply elasticity. Standard errors are in parentheses. Results are corrected for first-order serial correlation, as indicated by the Durbin-Watson (DW) statistic.

(Whi tney et al., 1985, p. 12; Hughes -Cromwick , 1985, p. 277). F u e l w o o d income elast ici ty in E th iop ia was previously es t ima ted at - 0 . 2 5 (Kidane, 1991, p. 133). On

the contrary , F e r n a n d e z (1980), p. 69, found that income elast ici ty for non-commer- cial fuels in u rban areas in the Sudan was 0.51.

It is difficult to draw s t rong conclusions f rom these figures, but since fue lwood is genera l ly cons ide red an infer ior fuel, a posi t ive income elast ici ty may ref lect a worsening of real income levels in Kha r toum. Hughes -Cromwick ' s s tudy of Kenya seems to suppor t this conclusion. She found that charcoal income elast ici ty dec reased as income levels rose, turning negat ive at the u p p e r m o s t s trata.

The supply side of the m a r k e t for charcoal indicates that suppl ies r e spond posi t ively to pr ice f luctuat ions. This is the first t ime fue lwood supply elast ici ty has been e s t ima ted for Sub-Saha ran Africa. The low es t imate (0.25) suggests that

E.O. Zein-Elabdin / Energy Economics 19 (1997) 465-475 471

relatively large changes in the price of charcoal are required to produce an effect on the quantities supplied. This may be attributed to the fact that charcoal supplies are not readily expandable. Production technology (the earth kiln) does not allow operation in the rainy season, storage facilities are rudimentary, and transportation is constrained by shortages of fuel and spare parts.

On the whole, these results indicate a limited response of both households and firms to changes in charcoal prices, suggesting that the burden of adjustment in this market falls on prices. Consequently, improved stoves will probably generate large price effects that will weaken their ability to cut fuelwood consumption drastically.

4. The rebound effects of improved stoves

The rationale behind improved stoves is that the fuel efficiency of traditional stoves can be raised or even multiplied by simple modifications in design. This will in turn reduce overall fuelwood consumption, and consequently forest destruction. A wide range of improved stove designs can be found. Reported fuel savings vary from 10% to 60%, although accuracy of measurement is still an issue. Stoves also vary widely in cost. 4

At any rate, the change in fuel consumption caused by improved stoves has typically been measured as the change in fuel efficiency multiplied by the number of stoves in use. This merely yields the first-order effects, i.e. the exogenous shift in demand due to technological improvement. However, by definition, improved stoves are designed to precipitate changes in household behavior in relation to energy use; therefore, their long-term impact depends on these higher-order effects, which, if sufficiently large, could undermine this strategy.

As mentioned previously, the rebound effects of improved stoves are manifested in income- and price-induced purchases of fuelwood. The income effect reflects reallocation of savings resulting from lower fuel requirements toward further fuelwood consumption. This is perhaps best exemplified by households that were forced to cut back wood consumption owing to the high cost. Clearly, the higher the income elasticity of demand for fuelwood, the larger will be the second-order purchases, and, in turn, the larger will be the losses in fuel savings induced by real income increases. The price effect is triggered by downward adjustment in fuel- wood prices subsequent to the initial reduction in consumption. Again, the magni- tude of price-induced purchases is directly determined by price elasticity of demand, as well as indirectly by supply elasticity.

To obtain the income effect (dCi), the fuel efficiency improvement from new stoves (dC) is multiplied by the share of charcoal in the household budget (s). This yields an estimate of the gain in real income to the household from the use of more

4The cost of improved stoves is an important element in households' decision to adopt them. Stoves are generally subsidized, ranging in cost from $3 to $20. The cost of modern stoves has little bearing on the decision to adopt improved stoves because of the substantial price differential between the two.

472 E.O. Zein-Elabdin / Energy Economics 19 (1997) 465-475

efficient stoves. Multiplying this increase in real income by income elasticity (6) gives the percentage increase in household consumption of charcoal. For instance, if a stove doubles fuel efficiency, and charcoal absorbs 20% of the household budget, that household will save about 10% of total expenditures, assuming fixed prices. Using the estimated income elasticity of 0.87, the increase in charcoal consumption will be 8.7% 5

The price effect (dCp) is obtained as the outcome of charcoal price adjustments multiplied by the household response to them. In a more generous data situation it would have been more appropriate to solve simultaneously the demand and supply equations to derive the required parameters. In the absence of such data, the following formula is used:

dCp = a( log 4))dC (3)

where dCp is the percentage loss of fuel savings resulting from the drop in charcoal prices. This drop is approximated by multiplying the logarithm of the supply elasticity (~b) by the initial reduction in charcoal consumption (dC). The logarithm of the supply elasticity is used since the latter exerts an inverse effect on price movements. 6 The price adjustment is multiplied by the price elasticity of demand ( a ) to obtain the price effect. The total rebound effects (dC T) can be expressed as

d C T = ( 6 s + a log ~b)dC (4)

To be sure, this is a drastically simplified procedure. Nevertheless, the results are comparable with those obtained by previous authors, without similar demands on data. The only hazard here is that total rebound effects may be a multiplicate of the income and price effects as opposed to being their sum, in which case the outcome would be even smaller than calculated here. The present results, then, may serve as an upper bound to the estimate of the second-order effects of improved stoves.

On the basis of the parameters estimated here, about 41.7% of the designed fuel efficiency gains from improved stoves may be lost to purchases of fuel. This is based on a budget share of 10%, prevailing in the Sudan in the 1980s (Jones et al., 1989, p. 95). A stove designed to cut fuelwood consumption by 30% will in reality achieve a reduction of 17.49%. As expected, the bulk of these second-order effects comes from price adjustments (amounting to 33% of fuel savings), whereas the gains in real income absorb only about 8.7%.

5This calculation of the income effect is based on Jones' equation:

d C i = 6 s d C

where d C i represents the percentage of fuel savings lost as a result of income-induced purchases of charcoal. 6Eq. (4) is derived by the author to obtain an estimate for the price effect. Logarithmic derivation approximates the negative but complex relationship between supply elasticity and price adjustments to change in demand more accurately than simply taking an inverse or a root.

E.O. Zein-Elabdin / Energy Economics 19 (1997) 465-475 473

The main factor behind large price effects is the low supply elasticity, which will cause prices to drop substantially in response to reductions in demand, thereby generating further purchases of fuelwood. This is consistent with the earlier argument that small quantity responses leading to large price adjustment will reduce the effectiveness of improved stoves.

These results are substantially close to those obtained previously for the Sudan, where the loss of fuel savings was estimated by Murck et al. (1985) at 48%. However, the results indicate fuel losses considerably lower than the 60% calcu- lated by Jones (1988). This is partly explained by the fact that Jones used an income elasticity of 1% and a budget share of 20%, and allowed for leakages in the form of household purchases of traditionally fueled manufactured goods. Jones also assumed a relatively high supply elasticity of unity. The results, however, are consistent with his findings in terms of the relative sizes of the income and price effects (his results were 20% and 40% respectively).

Although the price effects of improved stoves are substantial, they may not be fully realized. The reason is that rising fuelwood prices may offset, at least partially, the downward price adjustment that is at the heart of the price effect. The cost of fuelwood transportation has been rising steadily in recent years as a result of higher fuel costs and receding forest areas under the pressure of deforestation. This rise in fuelwood prices can be expected to be relatively large precisely because quantity movements are such that prices must pick up the burden of adjustment.

Accordingly, the drop in fuelwood prices following the introduction of improved stoves may be considerably smaller than calculated here, resulting in a lower price effect. Establishing this conclusion naturally requires more empirical evidence on the relationship between deforestation and fuelwood prices, which is a much needed direction for future research, especially in savannah regions as current research is heavily focused on rain forests.

5. Policy implications

The findings indicate that improved stoves would be most effective in places with low demand elasticities, especially if income elasticity is negative, since it will encourage households to move up the fuel ladder. For example, the previously reviewed parameters suggest that the rebound effects from improved stoves will be smaller in Ethiopia and Kenya than in the Sudan, although not knowing the supply elasticity for the former two makes this a rather tenuous conclusion.

A small budget share implies low real income gains, and hence would result in relatively small leakages. One might therefore expect that improved stoves will be less effective in the West African Sahel region, where fuelwood expenditures amount to one-quarter of household income. Assuming similar elasticities to those estimated here, this budget share suggests that rebound effects may reduce the effectiveness of improved stoves in West African countries by over one-half.

The results also suggest that supply elasticity must be raised in order to minimize the leakages in fuel savings from improved stoves. To put it in counterintuitive

474 E. 0. Zem-Elabdin /Energy Economics 19 (1997) 465-475

terms, in order to raise the effectiveness of improved stoves in slowing deforesta- tion, the systems by which fuelwood is delivered to households ought to be improved. This prescription, however, may prove hazardous unless it is feasible to raise supply elasticity without simultaneously lowering costs of production, thereby shifting the supply curve outward, and as a result, increasing fuelwood consump- tion. In other words, the efficacy of improved stoves would be greatly enhanced by a number of complementary measures to influence fuelwood market responses.

To indicate the potential of improved stoves for reducing forest depletion in the Sudan, fuel savings may be translated into forest area, although this exercise should generally be supplemented by qualitative evidence on the source of wood. It is estimated that about 239 200 h of land are annually cleared to provide Khartoum with charcoal (Muhammad Ali, 1984, p. 27). Linear extrapolation suggests that if improved stoves with one-third efficiency improvement are introduced, taking the rebound effects into account, about .41 836 h of land may be saved annually. However, this assumes an implausible 100% stove adoption ratio. The most successful improved stove program to date (in China) has only reached about half of the households in areas where stoves have been distributed. Still, assuming that one-third of households use improved stoves, about 12551 h could be saved annually.

The potential for forest savings will be greater if fuelwood prices are directly related to the rate of deforestation, and the price-related rebound effects can safely be ignored. In that case, the area saved may go up to 19 655, representing about 4% of annual deforestation in the Sudan. This is not an insignificant reduction, particularly given the low cost of implementation of improved stove programs.

Given that reforestation efforts in the Sudan are negligible, and given the high cost of modem energy imports, improved stoves represent an opportunity worth pursuing. However, at the same time, the size of the rebound effects implies that stoves with modest efficiency improvement may not be justified as a measure to contain deforestation, unless, perhaps, the small fuel savings are balanced by an aggressive diffusion campaign.

On a more general level, the relative size of the demand elasticities estimated here suggests that measures to influence fuelwood consumption will probably be more effective if aimed at household income than at fuelwood prices. Moreover, the low price elasticity indicates that raising fuelwood prices, as economists often prescribe, will have a modest effect on deforestation caused by the use of biomass energy.

Acknowledgements

My thanks to Professor Milton Russell for his help throughout this research project. Also many thanks to Janet Knoedler for her valuable comments on earlier drafts of the paper.

E.O. Zein-Elabdin / Energy Economics 19 (1997) 465-475 475

References

Fernandez, J.C., 1980, Household energy use in non-OPEC developing countries, Rand Corporation, Santa Monica, CA, prepared for the US Department of Energy.

Hughes-Cromwick, E.L., 1985, Nairobi households and their energy use: an economic analysis of consumption patterns, Energy Economics, (October), 265-278.

International Monetary Fund, 1970, 74, 77, 79, 82, 83, 85, 90, International financial statistics, (International Monetary Fund, Washington, DC), several issues.

Jones, D., 1988, Some simple economics of improved cookstove programs in developing countries, Resources and Energy 10, 247-264.

Jones, M. et al., 1989, Energy efficient stoves in East Africa: an assessment of the Kenya ceramic JIKO (stove) program, USAID and Oak Ridge National Laboratory, Oak Ridge, Tennessee, report of the Office of Energy and Technology, USAID, in cooperation with Regional Economic Development Services Office of East and Southern Africa.

Khazzoum, J.D., 1980, Economic implications of mandated efficiency standards for household appli- ances, Energy Journal (October), 21-40.

Kidane, A., 1991, Demand for energy in rural and urban centers of Ethiopia: an econometric analysis, Energy Economics 13(2) (April), 130-134.

Muhammad Ali, T.E.H., 1984, Some aspects of alternative energy: the supply of fuelwood in Khartoum province, Petroleum and Minerals 26-28 (in Arabic).

Mukhtar, M.E., 1982, Trends of production and consumption of woodfuel in Khartoum province and suggested measures for satisfying future needs, M.Sc. Thesis, University of Khartoum.

Murck, B.W., C.M. Dufournaud and J.B.R. Whitney, 1985, Simulation of a policy aimed at the reduction of wood use in the Sudan, Environment and Planning A 17, 1231-1242.

Pindyck, R.S. and D.L. Rubinfeld, 1981, Econometric models and economic forecasts, 2nd ed. McGraw- Hill, New York.

Sudan National Energy Administration, 1987, 1991, Sudan energy handbook, Ministry of Energy and Mining, Khartoum.

Whitney, J.B.R., C.M. Dufournaud and B.W. Murck, 1985, Energy use and the environment in the Sudan. Project Ecoville Working Paper No. 22, (University of Toronto, Toronto).

World Resources Institute, 1991, World Resources 1990-91 (Basic Books, New York). Zein-Elabdin, E.O., 1993, The impact of improved stoves on the demand for fuel wood, and its relation

to deforestation in sub-Saharan Africa, Ph.D. Disertation, University of Tennessee, Knoxville, TN.