t . .l. n ltlvurovxflllu r . 4 {H.479}... dun}?! r .Ffl , r... 114.}! f in) .6. :1. ct? 5h!“ Susi ting . , u: . v 5 JL . o ‘ Ir... ‘0'- Ilv’filt ’v Ir£“ .1 L; D! )fv , titty: i i‘51311utiklilciivifl‘ . .79 It)... or t .1, E... 2 Bug .vte... . :3. {11.30 ft!!! 7.3“... 3S.9-.1oa.i-I§ I...) at)!» (53"...11". I.?J..0?!.l 14."! . .i.‘l!7¢.?v 13%.}!!‘11‘ w «55% 1 VI » h A .‘l .1 In} .1 . .I'.. l I V IIIIPIA 1‘! D . ‘ x l. ,‘ 1 - MI} J» ‘ I 1 1‘.‘ f o. , a y . . . . ..?,v.... ...: .... ...v.._.. . lr Lin .LE.‘I , . 3 a.» '3. 1| ”v.2? . . Q . 6 -Y s ‘ i .2 JW~::.\J\ n..».v..€....n.. 01153-3. u . b..?.. .351»... u. .......n 2?. ...}‘.....3Lh:vaw—: .415. 4.1.115" §Jmn‘.rb1}€u.i. 3H7 4 A. 2 $ 1 i ‘ . (I ‘9. 3.0.5.04. 1. .J mambws um" ’llllfill”\l\\\\\ um Michigan State University . This is to certify that the dissertation entitled RURAL HOUSEHOLDS' ENERGY CONSUMPTION IN CENTRAL JAVA. INDONESIA presented by Boen Huchtar Purnama has been accepted towards fulfillment of the requirements for Doctoral Forestry degree in Date M ILLIWD MS U is an Affirmative Action/Equal Opportunity Institution 0- 12771 PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or betore one due. f——————_———__——l DATE DUE DATE DUE DATE DUE " "T ‘4 i "f t,“ .9 Q (~1an JUYJ l F t__l F33 MSU Is An Attirmdive ActioNEquel Opportunity Institution RURAL HOUSEHOLDS’ ENERGY CONSUMPTION IN CENTRAL JAVA, INDONESIA by Boen Muchtar Purnama A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Forestry 1990 ’ “A ‘ I; .- ABSTRACT RURAL HOUSEHOLDS ENERGY CONSUMPTION IN CENTRAL JAVA, INDONESIA BY Boen M. Purnama Reforestation in Indonesia, to some extent, has been negated by the reliance of rural households on fuelwood. The question as to whether or not household demand for fuelwood can be altered is relevant. This study analyzes factors that may affect a household’s decision in choosing a particular type of energy, as well as the factors that may influence the levels of energy used. In addressing those interrelated questions, Tobit model is employed. Tobit model jointly estimates the amount and the probability of a household in choosing a particular type of energy for cooking based on household income, price of fuelwood divided by kerosene price, family size, land ownership, and education level of the head of family. The study uses data from Central Java region and shows that household income, family size and education significantly affect the amount and the probability of households in choosing fuelwood and kerosene. The effect of household income on the probability of using fuelwood and Boen Muchtar Purnama the amounts are negative and positive, respectively. The effect of family size is positive for fuelwood and negative for kerosene. Low education level of the head of household, positively associated with the usage of agricultural wastes and fuelwood with significant coefficients. In case of fuelwood consumption, the levels of use is 2.16 tons or 2.59 cubic meter per household per year or 0.49 cubic meters per capita per year. Increasing a household’s income by 10 percent reduces fuelwood use by 12.3 kilograms or 0.0012 tons per household per year. This would lead to decreasing total fuelwood consumption for the region by 0.30 million cubic meters or eliminating fuelwood use by approximately 117,000 households (3 percent of fuelwood using households). These results, indirectly suggest that reducing the deforestation rate by encouraging a household to use less fuelwood and use other types of fuel, theoretically, is a feasible endeavor. To achieve this objective, however, various programs such as job provision, education, and family planning have to be emphasized in the future. ACKNOWLEDGEMENTS I wish to express my sincere appreciation to Dr. Larry A. Leefers who as my major professor and the chairman of my guidance committee provided me with continued support, encouragement, and valuable suggestions. I am particularly indebted for his encouragement that led to the decision to work on this research topic and for his suggestion that useful findings can even be derived from the 'old’ survey data. My sincere appreciation is extended to my thesis committee members: Dr. Robert S. Manthy, Professor of Forestry; Dr. Stanley Thompson, Professor of Agricultural Economics; and Dr. Jeffrey R. Vincent, Assistant Professor of Forestry for their encouragement and valuable suggestions. To Dr. Nana Supriana, Director of Forest Products Research Institute (FPRI), Bogor, Indonesia, who allowed me to use the rural energy survey data (1983), I thank him sincerely. My thanks are also extended to my colleagues at FPRI, especially Mr. Buharman and Mr. Prahasto. I am also indebted to the government of Indonesia for financial support through the MUCIA-GPT II Project and for the opportunity given to study in the United States. ii Ms. Ruth Dukesbury has made a significant contribution in this work through her ability to make the unreadable become readable. For her patience in editing my work, I am deeply grateful. Throughout the course of this academic endeavor, my wife Wita, has been unstinting in her love and encouragement to finish. Her patience and sacrifice of her own valuable time has made this work possible. My deepest thanks are to her. Last but not least, my warm gratitude is to my son, Arry, who has been involved in this venture without his consent. iii TABLE OF CONTENTS LISTOFTABLES ......OOOOOOOOOOOOOOOOO OOOOOOOOOOOOOO LIST OF FIGURES ... ................................. I. INTRODUCTION ....................... ........... Background and justification ..... ...... ..... Objectives of study ......... .. .............. Scope of study ................... ........... Relation to other studies ..... .............. Organization of dissertation ... ............. II. REVIEW OF PREVIOUS RURAL ENERGY STUDIES ........ Rural energy studies ........................ Early studies ............................. Recent energy studies ....... . ............. Comparison of recent studies approaches.... Factors affecting energy consumption ........ Household income ........... . .............. Tastes and preferences ............... ..... Family size .. ............................. Energy availability ............. .......... Fuel prices ....... .................. . ..... Energy consumption and rural development Summary ..................................... III. RURAL ENERGY CONSUMPTION IN CENTRAL JAVA ....... Study location .............................. Energy use levels ........................... Household characteristics and energy use Energy consumption and rural development Summary ...... ............................... IV. METHODS ............................ ........... Decision to use a particular energy ......... The amounts of energy use ................... Model specification ......................... Decision to use a particular type of energy and its level ........... Total energy consumption .................. Expected coefficient signs of explanatory variables on probability and consumption ....... . ................ iv 77 V. FACTORS AFFECTING ENERGY CONSUMPTION .......... Variables used in the present study ......... Results of Tobit estimations ........ ....... Signs and statistical significance ....... Household income ....... . ............. Age of head of family ...... .......... Education level .. .................... Size of family ....................... Land ownership .. ..... ...... .......... Energy price ... ......... . ............ Region ......... ........ .. ............ The probability of using a particular type of energy ....................... The amounts of energy use ................ Total energy consumption ................... VI. CONCLUSIONS ................................... APPENDIXES ........................................ 1. The FPRI Energy Survey 1983 ............... 2. The probit ................... ..... . ....... 3. Comparison between probit models with and without district dummy variables 4. The result of the probit estimations ...... 5. Correlation matrix ........................ REFERENCES ........................................ 82 84 85 85 88 89 9O 92 94 95 112 Table 1. 10. 11. 12. 13. 14. LIST OF TABLES Fuelwood consumption per household in East Java province in 1973 ......... Rural fuelwood consumption in Indonesia... Summary of the variables used in the previous rural energy studies and their significances .................... Districts, counties, villages selected ............................... Household responses by energy type and district ........................... Average daily energy use per household by energy type and district for all samples ........................ Average daily energy use per household by energy type and district for samples using only one type of energy ........... Average daily energy use per household by energy type and district for samples who use one or more specific types of energy ........................ Daily household energy use by energy type and monthly income class ......... Daily household energy use by energy type and family size ................... Daily household energy use by energy type and education level ............... Daily household energy use by energy type and land ownership ................ Household characteristics by energy user categories ........................ Variables, variable names, and its expected sign of coefficients .......... vi 39 49 50 51 56 59 63 80 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. vii Variables summary .... ........ .... ........ The results of estimation using Tobit technique ........ ......... ...... ....... Probability of using energy F(Z) based on the probit and Tobit results ....... Mean values of the variables for samples using fuelwood and for all of the samples ........................... Decomposition of Tobit model for agricultural waste use ................ Decomposition of the Tobit model for fuelwood use ........................... Decomposition of Tobit model for kerosene use .......................... Effect of change in income on the expected use level E(Y), probability F(Z), and expected level conditional upon being above limit E(Y ) for agricultural wastes .................... Effects of change in explanatory variables on the expected use levels E(Y), probability of using F(Z), and expected level conditional upon being above the limit E(Y ) for fuelwood ............................... Effect of change in explanatory variable on the expected use level E(Y), probability F(Z), and expected level conditional upon being above the limit E(Y ) for kerosene ........... Elasticity of income calculated at the mean for all samples for agricultural wastes ................................. Elasticities calculated at the mean for all samples for fuelwood . .......... Elasticities calculated at the mean for all samples for kerosene ........... 83 86 95 98 100 101 105 107 110 111 28. 29. 30. 31. viii Comparison between survey results and Tobit on the proportion of households using a particular type of energy and their levels ....................... 114 The proportion of rural households by energy type in Central Java (p*N) ..... 116 Annual household energy consumption in rural areas of Central Java ........ 116 Total fuelwood consumption when household income increase by 1, S and 10 percent 00...... ....... .0... ..... O ..... 118 LIST OE FIGURES Figure 1. Map of Indonesia ... ...... ................ 2. The locations of the rural energy survey 1983 ......OOOOOOOOOOOO......OOOOOOOO0.0 3. Comparison of the logit and probit cumulative disributions ................. 4. The XiB distribution for all samples ...... ix 17 42 69 99 I. INTRODUCTION Bee round nd at f out one Biomass energy, fuelwood in particular, has long been used as a primary energy source among rural households and the urban poor (Susastro, 1983; Arnold and Jongma, 1978: Eckholm, 1975). Although fuelwood is primarily used for cooking purposes, in the mountainous areas fuelwood is also used by households as a heating source (Eckholm et al., 1984; Hadi, 1982). In addition to fuelwood, crop residues and cow dung are the other forms of biomass energy which have become increasingly important as either substitutes or supplements for fuelwood, especially in the areas where wood is becoming more scarce (Hadi et al., 1979; Eckholm et al., 1984). The awareness of the importance of fuelwood as a source of energy in developing countries has been stimulated by the work of Brick Eckholm (1975) who first introduced the notion of the ’other energy crisis: fuelwood'. Prior to that time, the rural energy problem was largely ignored and overshadowed by cheap oil prices and overly optimistic hope for industrialization in most developing countries (Hosier, 1985b). Many challenges facing other developing countries are shared by Indonesia, a country which comprises more than 1 13,000 islands, of which 6,000 are inhabited (Donner, 1987). The land surface of Indonesia covers an area of 1,999,443 square kilometers. The population of Indonesia in 1983 was 160 million, and ranked as the third largest population in Asia after the People’s Republic of China and India. However, the population growth rate of 2.3 percent per annum (Biro Pusat Statistik, 1984) was higher when compared to China and India having rates of 1.5 and 2.1 percent, respectively (Donner, 1987). Although Indonesia has a relatively abundant land surface, almost 62 percent of its population lives in Java, an island which covers only seven percent of the total Indonesian land area. As in other developing countries, fuelwood is an important source of energy in the rural areas of Indonesia. Nearly 80 percent of its total population in 1978 resided in rural areas (Hadi et al., 1979). Large numbers of rural households use fuelwood and other biomass substances as their primary energy source (Susastro, 1983; Haeruman, 1977; Sumarna and Sudiono, 1973). For example, about 75 percent of the rural households in West Java use fuelwood for cooking (Haeruman, 1977). Estimates of average annual per capita consumption of fuelwood in rural areas of Indonesia, based on previous studies, shows a wide variation from one region to another. The reported levels ranged from 0.36 to more than 2.00 cubic meter per capita per year (Soesastro, 1983). Relatively high per capita fuelwood consumption levels have increased the awareness of the possibility of fuelwood shortages. Hadi et al.(1979), by comparing fuelwood production capability and fuelwood consumption levels, noted the presence of fuelwood shortages in some areas of Java. In fact, fuelwood shortages were not new to Java. For instance, as noted by Hamzah, in 1923 fuelwood shortages had occurred in the Wonosobo district as the demand for fuelwood by the tobacco drying industries had exceeded supply capability (Hadi, 1982). Forest resource depletion in Java was mostly caused by an expansion of agricultural land resulting from demographic and economic development (Donner, 1987). An increasing population, which still largely relies on agricultural practices, has caused the need for more land to be used in food production. However, utilizing more land for growing food crops is usually at the expense of a natural forest cover. A limited land resource in Java on one hand and the growing labor force in rural areas on the other hand have caused an increase of the landless farmer population. This situation in combination with the lack of employment opportunities in non-agricultural sectors has put more pressure on forest lands. Related to the lack of income earning opportunities, two other factors causing deforestation were identified by Fattah (Hadi, 1982). The first is illegal timber harvesting by people cutting forest trees for construction timber and subsequently selling it as their means of generating income. The second is excessive cutting by rural people to fulfill their fuelwood needs. In the Jogyakarta region, however, people also collect fuelwood from forest lands and market it for their income (Dick, 1930). Impacts of population pressure on Java’s environment are significant. By 1981, estimated devastation to either forest or agricultural lands in Java was 1.184 million hectares (Biro Pusat Statistik, 1984). Although successful planting programs have been reported from Central and East Java, successful replanting programs were often negated by increases in deforested lands resulting from population growth (Donner, 1987). In other words, the rate of reforestation oftentimes was below the rate of deforestation. Consequently, growing needs for food and fuel among rural people, which had been reflected by the amount of devastated land, have certainly burdened the government in its efforts at replanting these devastated lands. Energy is a basic need, and it is a part of the government’s responsibility to make energy available for the people. In compliance with this task, the Indonesian government has implemented a subsidy on kerosene (Susastro, 1983). The reasons behind the kerosene subsidy are two- fold. First, the government is attempting to stop further deforestation by encouraging rural people to use more kerosene and hence reduce fuelwood use. Second, this subsidy is based on the equity consideration, in the sense that rural people and urban poor are expected to be able to better afford this more convenient and higher quality energy if kerosene prices are kept lower (Gillis, 1980). The subsidy policy has been criticized as short- sighted, because in the future it may promote the reliance on kerosene, a non-renewable energy resource, while at the same time it discourages the use of a renewable energy resource, fuelwood (Dick, 1980; Hadi, 1982; Donner, 1987). Further, the policy has been ineffective with respect to the income distribution. For example, only 20 percent of the kerosene was consumed by the poorest 40 percent of the population (Gillis, 1980). In other words, eventhough the kerosene subsidy helps the poor it benefits the relatively well-off even more. Therefore, since large numbers of rural households still rely on fuelwood, the pressure on forest and other vegetation still remains present. Despite the kerosene subsidy debate, the Indonesian government has implemented a nationwide energy program which is comprised of three sub-programs, -— energy production, diversification and conservation (Anonymous, 1979). In relation to rural energy, the energy production sub-program is focused on the effort to increase fuelwood production through the expansion of fuelwood plantations and the introduction of fast growing tree species suitable for fuelwood such as the giant ipil-ipil (Leucaena leucocephala) and kaliandra (Calliandra callothvrsus). The energy diversification sub-program, was designed to widen energy alternatives in order to reduce the reliance upon conventional energy sources. Introduction of biogas and solar energy to rural communities is implemented under this energy diversification program. Finally, the conservation sub-program is focused on the effort to increase efficiency in using energy. A more efficient stove such as the Singer stove has been introduced and promoted in rural communities (Anonymous, 1979). However, the effectiveness of these policy measures depends largely on the rural household as its ultimate target. The information on characteristics of rural households in association with energy use is certainly needed to evaluate the success of the energy policies. Unfortunately, little information of this kind is available. Objectives of study To facilitate a better understanding regarding rural energy use, in particular use for cooking purposes, this study is designed to achieve two objectives. The first objective is to provide a description of rural energy consumption. The type of energy used for cooking and the consumption levels for each particular fuel are presented. Household characteristics with regard to energy usage are analysed to provide information about the opportunities for interfuel substitution. The second objective is to analyze the relationship between households’ energy consumption and their economic and demographic characteristics and to examine the policy implications of this information. The direction as well as a magnitude of the effects of household characteristics on the decision to select a particular type of energy and the level of energy use are of interest. Implications of these characteristics are examined by estimating elasticities. A Tobit model is employed to accomplish this objective. Scope of study Although fuelwood crises commonly occur throughout Indonesia, especially in densely populated areas such as Java, the survey data are available only on rural energy use in the Central Java Region. The survey, from which data for this study originated, was conducted by the Forest Products Research Institute (FPRI) in 1983. Hence, this analysis is confined to that region. The second limitation of the present study is that, despite the fact that fuel availability is important in analyzing rural energy, the paucity of available data has focused this study on the consumption aspects of rural energy. Finally, the study is focused on analyzing consumption of energy for cooking purposes. As previously mentioned, the reliance on fuelwood has created a serious environmental problem (Nasendi, 1978; Hadi, 1982; and Donner, 1987). In fact, energy consumed by rural households is largely used for cooking purposes (Susastro, 1983) for which fuelwood is the primary energy source. Thus, analyzing energy use for cooking, fuelwood in particular, may also provide useful information regarding environmental problems associated with rural household energy consumption. Relation to other studies Few studies on rural energy in Indonesia have been undertaken. Sumarna and Sudiono (1973) studied the fuelwood consumption by three sectors (i.e. household, industry, and railroad company) in the Province of East Java. Most rural energy studies following the research of Sumarna and Sudiono focused on finding the consumption levels of particular types of energy. Studies by Nasendi (1978) and Dwiprabowo et al. (1980) were among the few that tried to analyze relationships between various socio-economic factors and energy consumption levels. Hadi (1982) also analyzed the effects of socio-economic factors on consumption levels. However, unlike Nasendi and Dwiprabowo et al., she proceeded by first, predicting the probability of households in selecting a particular fuel. This probability was analysed using the logit technique. Then, she estimated levels of energy use by relating them to socio-economic factors. The level of a particular type of energy use was estimated using the ordinary least squares (OLS). To do so, she restricted the estimation to subsets of the samples for each type of energy. Tobit, a model that links the level of household fuelwood use and its economic and demographic characteristics, is utilized in this study. This model is also known as a limited dependent variable or a censored sample technique (Judge et al., 1985). It allows all energy users to be incorporated in the model instead of only subsets of energy users. The Tobit formulation is based on the cumulative normal function. Organization of dissertation Rural energy consumption, particularly in Java, has turned into a crisis which has negative impacts on the environment. Deforestation as a result of the growing population needing food and fuel for their subsistence livelihood is reflected by increasingly huge, denuded land areas. Eventhough this energy crisis has caused the government to considerably increase spending for land conservation and reforestation programs, nevertheless, the presence of an energy crisis, always results in rural people and the poor being worse-off. Subsistence households may adjust to fuelwood shortages by using less wood, switching to less convenient forms of energy (ie., agricultural wastes) or using up more of their precious time to collect wood. 10 It is clear that rural energy, in this case rural energy consumption, needs to be analyzed not only to increase the understanding of problems associated with rural energy, but also to provide information that may be useful as a basis for a sound energy policy. Previous studies on fuelwood and other types of rural energy consumption in Indonesia are described to provide insight concerning the present status of rural energy usage. Factors likely to affect energy consumption are discussed. This is followed by a chapter which includes a descriptive analysis of the survey data. Energy users are defined and their characteristics are also identified. Categorization of rural households into energy user groups with discernible characteristics facilitates understanding of the decision-making process within households. Moreover, this categorization also facilitates an understanding of changes in energy-use patterns associated with changes of household status from subsistence living to the more modern, market-oriented households. The presence of a decision-making process within a household in selecting a particular energy, is used as a basis for developing a probability model. The discussion of the Tobit model, which is employed to estimate the probability and amount of energy use, is presented. Variables used and expected coefficient signs of explanatory variables are discussed. 11 The study results are then presented. Tobit results are discussed and various factors that significantly affect the estimation are identified. Finally, the conclusions and possible implications of this study on rural energy policy are presented. II. REVIEW OF PREVIOUS RURAL ENERGY STUDIES This chapter has three sections. The first section focuses on the methods employed in past studies. It begins with explanations regarding earlier studies which were mostly concerned with finding data on consumption levels. The weaknesses of these approaches are presented. Then, a discussion is presented of several energy studies which collected information beyond the consumption level figures. The discussion in this first section leads to the method used in the present study, which is discussed in Chapter IV. Variables that likely affect household energy use, especially the variables used in the past studies, are discussed in the succeeding section. The discussion in this section is used as a basis in selecting variables for the present study. Finally, the section discussing the changes in the patterns of rural energy use is presented. This section provides insights concerning the possible interfuel substitution as households move from subsistence to more modern life. Rural energy studies d'es As previously mentioned, forest land in Java, a densely populated island, has been a victim of population pressure for a long time. The strain on the sustainability of forest 12 13 resources has been caused by local people in meeting their needs for food, fodder, and fuelwood through illegal cutting and land encroachment. Reasonably, most of the early rural energy studies in Indonesia were conducted by foresters, primarily in relation to the efforts of the Forest Agency to protect forest resources. Those energy studies were intended to provide data on the levels of fuelwood consumption, and then by comparing them with the supply capability, the Forest Agency could decide whether or not it was necessary to designate a parcel of forest land for fuelwood production. Soedarmo was among the earliest who studied fuelwood consumption in 1955 (Hadi, 1982). He estimated fuelwood use by household sector based on the assumption that fuelwood use per capita per year was 0.5 cubic meter. Although overall supply and demand were in balance, fuelwood shortages occurred in several districts. As a result of his estimation, the Forest Agency implemented fuelwood plantation programs. A more in-depth study on fuelwood consumption was carried out by Sumarna and Sudiono (1973) in East Java province. One among their primary objectives was to examine the patterns and levels of fuelwood consumption in the household sector. Stratified random sampling was used by Sumarna and Sudiono in selecting household samples. The stratum were urban and rural areas. Surabaya, the capital city of the 14 province, was treated as a special stratum. In addition, two districts and three districts were chosen at random to represent urban areas and rural areas, respectively. From each selected district, villages were randomly chosen. In total, 50 villages were selected for the entire province. The sampling intensity of the survey was 0.02 percent covering as many as 1,250 households as survey respondents for the whole province. Among others, findings were: fuelwood is still a dominant energy source in rural areas ninety six percent of the total fuelwood consumed in East Java was used by households and levels of fuelwood consumption per household per year for the entire province was 2.51 cubic meters (Table 1). The per capita consumption level was calculated from the average per household energy use level divided by the average family size. By using the average family size of 4.9 persons, calculated average consumption per capita per year for the entire province was 0.51 cubic meters. Several studies on fuelwood consumption have been undertaken in Indonesia. The summary of their estimation regarding fuelwood consumption levels are presented in Table 2. A map of location of those studies is presented in Figure 1. Those studies generally covered very limited geographic areas, such as a watershed or a river basin area (RBA), although, estimations for the entire country had also been made. 15 Table 1. Fuelwood consumption per household in East Java province in 1973 Stratum Consumption level (cubic meter) Surabaya municipal 0.13 Urban areas 0.56 Rural areas 2.80 Province 2.51 Source: Sumarna, K. and J.Sudiono (1973). FPRI Report No. 8. Previous studies did provide some indication regarding the importance of fuelwood as a source of rural energy. However, those studies were characterized by lack of sufficient information and by inconsistencies. As a result, it was difficult to draw useful conclusions which might have provided reliable information for policy purposes (Susastro, 1983). Several factors may lead to variations in estimated fuelwood use levels. The first problem is related to how the actual energy consumption is measured. In most previous studies, estimated energy use was obtained through interviewing, rather than as a result of a direct observation (Hadi, 1982). Household respondents were asked to indicate the amounts of fuelwood or kerosene they normally used and then, the interviewer weighed and recorded the indicated amounts to provide use estimates. 16 Table 2. Rural Fuelwood Consumption in Indonesia (in cubic meter per capita per year) 1 Location Year Consumption Source levels Indonesia 1956 0.50 LPHH (1970) 1970 0.72 FAO estimates; Silitonga (1974) 1975 0.86 Chandrasekharan (1977) 1976 0.84 Atje (1979) Java 1976 0.79 Atje (1979) 1978 1.00 Hadi et al. (1979) East Java 1971 0.51 Sumarna and Sudiono (1973) 1978 1.27 Hadi et al. (1979) Cent.Java 1978 0.64 Hadi et al. (1979) 8010 RB2 1969 0.74 LPHH (1969) Solo RB 1975 0.36 Wiersum (1976) Solo RB 1976 1.13 Wangsadijaya et at.,(l979) West Java 1977 2.08 Haeruman et al.(1977) 1978 0.43 Hadi et al. (1977) Citanduy RB 1977 2.22 Nasendi (1978) Citarum RB 1979 2.53 Rusydi et al.(1979) Outside Java 1976 0.96 Atje (1979) Bali 1978 1.06 Hadi et a1. (1979) Aceh:urban 1978 0.27 Dwipra owo et a1. (1980)— rural 0.73 lwkowt Cited from Susastro (1983), p. 5. RB is an abbreviation for River Basin. Forest Products Research Institute (FPRI) Report No. 155, 1980, pp 25-32. 17 Asma_ .uuccoa aouu vuusov0:— 0&8: . . - 4 a q 1&3 .mz . . . . 9...... JJ t 0..o \ ‘HQ’MWVS.|:0 0.0 ‘ ‘ U g r P! u .0. :o:or \. nMMHHHHHHHHUW D. AHUHUUV eea zeszw A 0 00.0. 0. a. m f «.353 b. u: C.“ .8333. l.d...§d. D lib b eon once. a. .uuku u‘! 700.0 039.030 9 Am in”... ..Qéfl. .Mr .. s I." 005.33 0090.10. 10.0999/0 '0'0-5'0 0’ 9 «v .1. . 3254...; 0...... W 0 eon eeaezo ‘ one“- o'ee . ‘0 a D L b eon 2.2» 2:: .0 S mmzooz— no .2: D ._ ousuum 18 In general, previous studies did not seriously consider variations of calorific value among different types of fuels (Hadi, 1982; Susastro, 1983). These variations are even more important when considering agricultural wastes, which normally have lower heat content than fuelwood and are increasingly used by rural households. If both fuelwood and agricultural wastes are simply added together into one energy category, this might account for those discrepancies. Finally, most past studies were only concerned with finding fuelwood consumption levels and did not analyze the effects of socio-economic variables on consumption (Hadi, 1982). As a result, one could not really explain what factor or combination of factors accounted for any differences in the consumption levels. Recent energygstudieg Efforts to counter those weaknesess have been undertaken in more recent studies. For example, most recent studies have been based on the observation and direct measurement of the actual energy used instead of interviewing respondents (Susastro, 1983; Hadi, 1982). The presence of agricultural wastes has been treated by separating them into one type of energy source. Nonetheless, most of the studies were still concerned with finding consumption level figures. The exceptions are studies by Nasendi (1978), Dwiprabowo et al. (1980) and Hadi (1982); these researchers tried to analyze the effects of socio-economic factors on rural energy 19 consumption. The remainder of this section presents an explanation of their studies and a comparison of their approaches for analyzing rural energy. Nasendi (1978) studied rural energy consumption in the Citanduy River Basin of West Java with a special emphasis on fuelwood. His objectives were to estimate demand elasticity and to develop fuelwood demand projections for the region. The sampling technique employed was stratified random sampling. Three strata were determined based on population density per hectare of dry land (ie., low, medium, and high densities). The low density area was an area with less than 45 persons per hectare of dry land. Medium density was an area with the number of persons per hectare ranging from 45 to 70 persons. Finally, the high density area was an area with more than 70 persons per hectare of dry land. Two villages for each stratum were selected at random. Each village represented rural and urban areas, respectively. Twenty households from each village were randomly selected resulting in total respondents of 120 households for the entire Citanduy River Basin. Amounts of energy used by individual households such as fuelwood, charcoal, and kerosene were directly observed and measured. Economic and demographic characteristics of households, such as family income, education, family size and average age of husband and wife that might have affected energy consumption, were recorded. Relationships between the 20 quantity energy consumed per household and economic and demographic factors were specified as follows: Qij = a + lej + CXZj + dX3j + eX4j + Yij where: Qij = energy consumed per year by household j, in kilogram or litre; i=1 for fuelwood, i=2 for charcoal, i=3 for kerosene le = family income of household j ij = family size of household j X3j = husband's level of education in household j X4j = average age of husband and wife of household j Yij = error terms For each population density, Nasendi further classified population into low, medium, and high income classes. The low income class was for households with family income per year Rupiahs (Rps). 499,000 or less. The medium income class was comprised of those having income between Rps. 500,000 and Rps. 999,000 per year. Finally, high income households had income per year of Rps. 900,000 or more (Nasendi, 1978). To analyze the effects of socio-economic variables on energy use, separate regressions were estimated for each population density, location (urban and rural), and income class. The regression equation for estimating the effects of socio-economic variables on energy consumption for the 21 entire population, regardless of the differences on population density, location, and income class, was also derived. These separate estimations were intended to analyze the effects of economic and demographic variables on energy use levels for the following conditions: 1. energy use levels given the differences in income levels and population density, 2. energy use levels given the differences in income levels, 3. energy use levels given the differences in population density, and 4. energy use levels regardless of differences in income and population density. Several study results are notable. Income levels were only important in explaining charcoal and kerosene consumption. They were less important in explaining fuelwood consumption in all locations, except for fuelwood use by low income households in low population density areas. Fuelwood consumption for low income households in low density areas was negatively affected by income levels. The effect of income on kerosene consumption was positive and significant for low income families in low and medium density areas. Family size was a primary factor affecting the amounts of energy used for all type of energy for households with low income levels in low density areas as well as for high 22 income households in high density areas. The effect of family size was also true in both urban and rural areas. Tastes and preferences which were represented by the variables of the household head’s education level and the average age of husband and wife were less important in explaining energy consumption. Separate estimations based on income class show that family size significantly affected energy consumption for all types of energy in all income classes. However, within high income class households, education level of family head and average age of husband and wife were more important than family size in explaining energy use. The family income only significantly affected kerosene use in low as well as medium income classes. The results of estimations based on location categories indicated that family income, family size, and education level significantly affected fuelwood use in low population density areas. The family size variable was more important than the income variable since it affected fuelwood use in most locations. Estimation of energy use for the entire region, regardless of the differences in population density, income class and location, shows that family size was the primary factor affecting the use of all types of energy, fuelwood, charcoal, and kerosene. The income variable was only important in explaining charcoal and kerosene consumption. 23 Tastes and preferences which were represented by education level of family head and the average age of husband and wife only affected kerosene use. Another study conducted by Dwiprabowo et al. (1980), examined fuelwood and other fuel consumption by households and industries in the Aceh province of Sumatera. The objectives of their study were to estimate the levels of fuelwood use by households and industries, and to analyze the relationship between household fuelwood consumption levels and various factors that may affect them. In selecting household samples the authors utilized urban and rural areas as their survey stratum. Banda Aceh, the capital city of the province was chosen as reflecting urban, whereas Aceh Besar district represented a rural area. Counties in each district, as well as the villages in each selected county, were randomly chosen. Fuelwood consumption was specified by using a multiple regression technique. Various variables that were perceived to affect fuelwood consumption levels were included in the model which was specified as follows: Q = a +bX1 + ch + dX3 + eX4 + fX5 + U where: Q = amount of fuelwood consumed ( kilograms per capita per day) X1 = household income (Rps. per month) X2 = family size (N) fuelwood price (Rps per kilogram) X 0.) ll 24 X4 = education levels - quantity of substitute fuels consumed (litres >< U1 I per capita per day) U error terms Two estimations were made, one for urban and rural areas, respectively. Study results indicate that fuelwood consumption in urban as well as rural areas was significantly and positively related to income and size of family. This implies that an increase in either income or family size caused household fuelwood consumption to increase. Fuelwood price significantly affected fuelwood consumption, both in urban and rural areas, but with a negative coefficient. This implies that increases in fuelwood prices would reduce fuelwood consumption. Education level only significantly affected fuelwood consumption in urban areas, but was not important in explaining fuelwood use in rural areas. Hadi (1982), in her study in rural areas of West Java, set two study objectives namely to describe the patterns of household energy consumption and to analyze the effects of socio-economic factors on the choice of fuel types and on the consumption levels of chosen fuels. Considering that not all household samples used any one particular fuel, Hadi employed the logit model to predict the probability of a household choosing one type of fuel such as fuelwood, agricultural wastes and kerosene. The probability of 25 choosing a particular fuel was specified as a function of the location of a household and its socio-economic characteristics. The logit model was specified as follows: Pi = Prob (Y1 = l) i = 1, ...... ,n household samples Y1 = dependent binary variables; = 1, if taking a particular fuel = 0, otherwise Pi = the probability of Y taking the value 1 Xi = vector of independent variables affecting Pi b = vector of estimate parameters The model restriction is: 0 < Pi < 1 To further analyze rural energy use, she linked the amounts of particular fuels used by households to their socio-economic characteristics. The model employed was follows: Y = a + bixij + Yj where: Y = level of use of particular fuel Xij = vector of explanatory variables - error terms r0>H3m 0003 m can .m J. .m .N .H 0030.03 an .38 3088 G 332 >033 >805 :03 on... no mcowunooa 05. .~ 9.53m 42 25.. Sam ..., 28 e . ...—.30 :31... 1 Hanson m m 95.5798» 13:... J a ’ ... a. . 0000x0258 J omwunacm O .oifi a): .035U \s 9252mm 0:328 M O cnmcoanxum 20—. Dem 210-. 43 Table 4. Districtsi counties, and villages selected — District County Village Bantul Kretek Pr.tritis Donotirto Tirtohargo Pajangan Sendangsari Triwidadi Banyumas Kebasen Kebasen Madirancan Cindaga Patikraja Patikraja Notog Blora Ngawen Ngawen Srigading Cepu Cepu Jipang Ngloram Pemalang Rd.dongka1 Rd.dongkal Kalimas Gongseng Petarukan Petarukan Loning Purworejo Loano Loano Jetis Maron Banyuurip Banyuurip Kebasen Temanggung Ngadirejo Banjarsari Mendari Katekan Pringsurat Klepu Ngipik l Central Java province: Banyumas, Blora, Pemalang Purworejo and Temanggung. Yogyakarta province : Bantul district. 44 Based on survey data, the type of energy used for cooking by a household in the rural area of Central Java Region varied from agricultural wastes to kerosene. Agricultural wastes, henceforth are called wastes, are comprised of a wide range of substances such as residues from agricultural crops (e.g., rice stalks, rice husks, peanut shells, soybean shells, corn stalks, corn cobs, corn husks, and cassava stalks). Wastes such as coconut husks, coconut shell, coconut leaves and twigs were also commonly used especially in Banyumas and Purworejo districts. A household might use one type of energy, but they might also use a combination of energy types. Based on the types of energy used, household samples are categorized into five groups as follows: 1. Agricultural wastes users (AGWA) consist of households using wastes entirely as their energy sources for cooking, 2. Mixed biomass energy users (FWAW) are households who use fuelwood in combination with wastes, 3. Fuelwood users (WOOD) are households using entirely fuelwood for their cooking purposes, 4. Mixed energy users (KEBI) are households using a combination of fuelwood and kerosene, and 5. Kerosene users (KERO) are households using only kerosene for their cooking energy. 45 The survey data shows that biomass energy, wastes or fuelwood, still plays an indispensable role as a source of energy in rural areas. Biomass energy is used by nearly 90 percent of household samples in the region (Table 5). The percentage of households using biomass energy varied from one district to another. It ranged from 76 percent in Pemalang to more than 97 percent in Bantul district. Fuelwood is an important biomass energy source in most of the sample districts, followed by wastes. In three districts (Blora, Pemalang, and Temanggung) fuelwood was the only type of biomass energy used by household samples. On the other hand, in three other districts (Bantul, Banyumas, and Purworejo) the percentage of households using wastes, especially in combination with fuelwood (FWAW) was higher. Differences on the relative contribution of a particular type of energy in each district may be associated with its relative availability. Although kerosene distribution has penetrated rural areas of Java (Susastro, 1983), its role as an energy source for rural people, in particular for cooking, is still limited. Susastro’s observation is also supported by this survey data which shows that less than 20 percent of household samples in Central Java Region used commercial energy. No single household sample using kerosene (KERO) was recorded in Bantul district. The highest percentage of households using kerosene was in Pemalang (16 percent). 46 . 8.9.3 39.330... :33 05 3 meow: .3 3359.3 96 335:0th E 8.53.: 88.83 88.3 88.3 8~.oov 85.38 80.83 8m.3 up .. om mm 83 03 mp we 5.33. 88.83 8.23 87.: 8903 8m.03 88.8 88.3 mmw w e F: F: o 8 953553 8983 SN: 8~.S 80....3 89.: 8.25 8~.o~v a. ~ ~ . a. m 3 3 0.3.322 88.85 8.203 8.25 8~oC 8~ot 80.8 80.8 mm? 8 o no no 0 o 9.395.. 88.83 8~.$ 85.: 8203 8703 88.8 88.8 o: Z N 03 03 o o 238 88.85 88.8 85.5 8..”83 8m.mmv 88.03 88.05 0: o N n: be PM om .353 80.83. 8va 8.23 8m.oov 80.~3 80.38 8.23 «N— o . m w: an S m mag-Scam Bonanza: ago» 803 :5: (30¢ .33 88. 8.8. I 3983 3:33 a 327.5 use on? >965 >9 momcoomot 30:33: .m use.» 47 Pemalang district, according to Biro Pusat Statistik (1987), is located in the zone which is categorized as a developed area. Therefore, high kerosene usage in Pemalang might be due to a better distribution channel for kerosene relative to channels in other districts. The average daily energy use by energy type per household is presented in Tables 6, 7, and 8. There are three different calculations on the levels of energy usage. The first calculation is based on the energy level used when all household samples are included, regardless of energy type actually used (Table 6). The second calculation is based only on household samples using one particular type of energy. For example, average fuelwood per household is calculated based on household samples using entirely fuelwood (Table 7). The last calculation includes household samples using one or more specific types of energy. For example, the average fuelwood use per household is calculated based on household samples who used either entirely fuelwood, or used fuelwood in combination with other energy types, either wastes or kerosene (Tables 8). The series of tables clearly indicate that the levels of energy use will increase if non-users are excluded from calculation. For example, in Table 6, average daily fuelwood use per household is 5.35 kilograms, and when non- users are dropped, the fuelwood use level is increased to 6.87 kilograms (Table 7). 48 Fuelwood use by households using a combination of energy, for instance fuelwood with agricultural wastes, is lower than households using entirely fuelwood. However, the amount of reduction may depend on the amount of fuelwood substitute used, ceteris paribus. For example, when a household used only fuelwood, the amount of wood use per day per household was 6.87 kilograms (Table 7). But when a household used fuelwood in combination with another energy type, the amount of fuelwood used was somewhat less, 6.18 kilograms (Table 8). Daily fuelwood consumption per household among samples who used only fuelwood for all regions was 6.87 kilograms. The levels of consumption among districts ranged from 5.14 kilograms in Pemalang to 8.52 kilograms in Temanggung (Table 7). Kerosene consumption per household per day for kerosene user (KERO) was 0.89 litres. Average consumption ranged from the lowest daily use of 0.77 to 1.17 litres in Blora and Banyumas districts, respectively (Table 7). Standard errors of the mean energy use levels, both fuelwood and kerosene, were consistently high either at the region or district levels. These high standard errors indicated that the daily use of a particular energy among households varied considerably. The variation of energy use levels might be closely related to the variability of economic and demographic characteristics of households. For example, considerable variations on household size across 49 Table 6. Average daily energy use per household by energy type and district for all samples — Energy type District N Ag.wastes Fuelwood Kerosene Banyumas 124 0.59 6.64 0.10 (1.57); (3.90) (0.33) Bantul 119 1.36 4.55 0.01 (2.56) (3.76) (0.06) Blora 119 — 5.31 0.08 (3.87) (0.26) Pemalang 122 - 4.29 0.21 (3.42) (0.26) Purworejo 125 4.06 3.46 0.02 (2.95) (3.56) (0.12) Temanggung 123 - 7.78 0.08 (4.73) (0.25) Region 732 1.04 5.35 0.08 (2.29) (4.15) (0.28) w» IN The levels of wastes and fuelwood used are in kilograms (kg), kerosene is in litres. Figures in parentheses are standard errors. 50 Table 7. Average daily energy use per household by energy type and district for samples using only one type of energy- Energy type District Ag.wastes Fuelwood Kerosene N L 3 N L N L Banyumas 3 6.41 78 8.22 9 1.17 (5.96) (3.13) (0.31) Bantul 20 5.94 66 6.56 - - (3.83) (3.59) Blora - - 106 5.90 11 0.77 (3.97) (0.38) Pemalang - - 93 5.14 20 0.92 (2.93) (0.48) Purworejo 25 5.86 5 6.12 2 0.94 (4.07) (0.17) Temanggung - - 111 8.52 8 0.86 (4.34) (0.38) Region 48 5.76 459 6.87 50 0.89 (3.10) (3.83) (0.36) #1 The levels of ag.wastes and fuelwood used are in kg., kerosene is in litres. % N = number of samples, L = levels. Figures in parentheses are standard errors. 51 Table 8. Average daily energy use per household by energy type and district for samples who use one or more specific types of energy — Energy type District Ag. wasteg Fuelwood Kerosene N L — N L N L Banyumas 37 2.17 112 7.35 12 1.01 (2.39); (3.40) (0.48) Bantul 51 3.17 99 5.58 2 0.44 (3.09) (3.41) (0.18) Blora - - 108 5.84 13 0.77 (3.65) (0.33) Pemalang - - 102 5.13 29 0.87 (3.11) (0.60) Purworejo 118 4.13 98 4.41 4 0.57 (2.86) (3.46) (0.45) Temanggung - - 115 8.32 12 0.77 (4.41) (0.36) Region 206 3.70 634 6.18 72 0.80 (2.97) (3.84) (0.45) 3* WHN in kg., kerosene used is in litres. = levels. N: number of samples, Figures in parentheses are standard errors. L The levels of ag.wastes and fuelwood used are 52 household samples might lead to the high variation of the daily energy use level per household as shown by high standard errors. Another factor contributing to the variation in the daily use of a particular energy may be the relative availability of respective types of energy to households. For instance, a household living near a forest or other fuelwood source may tend to use more fuelwood than a household far from a source, ceteris paribus. Unfortunately, instead of recording a distance to fuelwood source, survey recorded the distance to any energy source (Appendix 1, p. 144). Therefore, specific analysis using this variable can not be pursued. Household characteristics and energy use The variability on daily energy use levels by households as shown in the previous section encourages further exploration of the effects of some economic and demographic variables on energy use. In fact, those variables are likely to be interrelated in influencing the type of energy used as well as the levels of energy consumed by households. However, the following presentation emphasizes the effects of individual variables, such as income, family size, education and land ownership, on energy use . 53 Although income is not the only variable affecting household energy use, it is an important economic variable that is often used to measure the relative well-being of a household. Family income is expected to affect the level and type of energy use. To identify the income effect on energy use, household samples were separated into three monthly income categories. The levels of energy use by energy type and income class are presented in Table 9. Household fuelwood use levels were significantly different across income classes. The use of fuelwood tends to increase as family income increases. The plausible explanation for this phenomenon is that households tend to purchase more food as their income increases which in turn, requires more fuelwood for cooking. Although fuelwood use levels increased as household income increased, the number of households using fuelwood tended to decrease as household income increased. Similarly, the number of households using wastes decreased as household income increased. The difference across income classes in the amounts of kerosene and wastes used were not statistically significant. The insignificance among kerosene users might have been due to the small number of samples in the low-income brackets using kerosene. Yet, there was a tendency towards the number of kerosene users increasing as household income increased. This implies, that as household income 54 Table 9. Average daily household energy use by energy type and monthly income class 1 Income class (in thousand Rps.)Z Energy type < 25 25 - 75 > 75 Waste? ”5 N - 118 77 10 levels (kg) 3.780 3.557 4.270 Fuelwood ** N 310 261 61 levels (kg) 5.648 6.625 7.088 Kerosene ns N 4 12 57 levels (litre) 0.503 0.537 1.145 L) IN 3 Household may use a combination of energy, therefore one household may be counted more than once. Rps. stand for Rupiahs, the Indonesian currency. Exchange rate in 1983: 1 US $ = Rps. 450. N = number of samples. ns Non-significant at 95 percent level. Significant at 99 percent level. 55 increases, households tend to shift from using biomass energy to more modern types of energy such as kerosene. Family size significantly affected household consumption of fuelwood and wastes, but not kerosene use (Table 10). The relatively small sample of households using kerosene might have contributed to this insignificance. Increasing use of fuelwood and wastes along with increases in family size, ceteris paribus, might be due to the fact that a household with a larger family has more labor available for collecting fuelwood and wastes than a smaller size household would have. Among biomass energy users, household biomass energy use increases as household size becomes larger, but at decreasing rate. For example, the per capita daily fuelwood use for small, medium, and large families was 1.028, 0.727, and 0.473 kilograms, respectively. Education of the head of family is significantly related to kerosene use, but not to biomass energy use (Table 11). It might be possible that when fuelwood and other biomass energy are relatively available, rural households would more likely use those energy sources, regardless of their educational background. The largest number of households using biomass energy were those having low formal education or none at all. On the other hand, among kerosene users the number of households with higher education level (ie., middle school Table 10. 56 Average daily household energy use by energy type and family size — Energy type Family size 5 < 5 5 - 7 >7 Wasteg ** N — 103 79 23 level (kg) 3.186 4.289 4.158 Fuelwood** N 223 297 111 level (kg) 4.741 6.582 8.104 Kerosene ns N 22 41 10 level (litre) 0.555 1.210 1.135 1:» Household might used a combination of energy, hence it might be counted more than once . IN Average family size for each family size groups were 3.1, 5.9, and 8.8, respectively. ; N = number of samples Significant at 95 percent level 5 Non-significant at 95 percent level 57 graduates or higher) was larger than households with lower educational backgrounds. This implies that, the higher their education is, the less likely that rural households will use fuelwood. Land ownership or size of holdings significantly affected fuelwood and kerosene used by households. 0n the other hand, waste usage was not influenced by land ownership (Table 12). A possible explanation of this, was the fact, that agricultural wastes are residues from agricultural products which are often perceived as ’unuseful’ substances and therefore, they are largely left in the field. They, in turn, become a free source of energy for people having little or no land. In other words, land ownership is not a necessary condition for having access to waste energy sources. Fuelwood use levels tend to increase as land owned by a household increases. This might be related to the fact that increases in land owned mean more fuelwood is available, and hence more fuelwood is consumed. With respect to kerosene usage, there may be rural households who own land and experience a surplus of agricultural products that could be traded for cash; or other households which possess land, but are engaged in off-farm income-generating activities (e.g., teachers and government employees). For this segment of the rural population, who may no longer be living in a subsistence condition, commercial energy such as kerosene is an affordable energy alternative and hence they would most likely use kerosene for their cooking energy. Table 11. 58 Average daily household energy use by energy type and education level — Energy type Education level uneducated middle high and elemen school school or tary school college Waste? ns N — 188 16 1 level (kg) 3.808 2.741 3.000 Fuelwood ns N 570 42 20 level (kg) 6.160 6.642 6.228 Kerosene ** N 13 42 17 level (litre) 0.503 0.934 0.821 W’ Household might use a combination of energy hence it might be counted more than once. Education refers to the education of head of family. 2 ns Non-significant at 95 percent level. — N = number of samples. Significant at 95 percent level. Table 12. 59 Average daily household energy use by energy type and land ownership — Energy type Land ownership (hectare) 0.25 0.25 - 0.50 0.50 Waste? “5 N - 93 43 65 level (kg) 3.500 4.030 3.911 Fuelwood ** N 265 121 229 level (kg) 5.060 6.463 7.406 Kerosene ** N 32 5 34 level (litre) 0.708 0.666 0.983 %' Household might be use a combination of energy, so it might be counted more than once . ns Non significant at 95 percent level. 2 N = number of samples. Significant at 95 percent level. 60 Energy consumption and rural development This section focuses on the linkage between energy use patterns and status of households in the transition from subsistence to more market-oriented activities. The reason for examining this linkage is to identify dynamic elements influencing rural energy consumption patterns. As previously explained, Hosier (1985b) classified households into groups that reflected a transition from subsistence households toward more market-oriented households. Then, he identified the type of energy used by each household category. Hosier concluded that as a rural household moves toward becoming more a modern household, it tended to shift from using fuelwood and wastes to using kerosene. i In this study, the approach is reversed, in the sense that first, households are categorized into energy user group based on the type of energy they used. Then, household characteristics which would most likely distinguish one energy user group to another are identified. The presumption regarding this approach is that the categorization based on energy used starts with household behavior rather than life-style categories. Reasonably, there may be various household characteristics that can be used as a basis to distinguish one energy user group over 5 See explanation in the chapter of previous studies (p. 36). 61 another. This section is intended to identify the characteristics of households that most likely relate to specific energy-user groups. Compared to the type of categorization used by Hosier (1985b), categorizing of households into user groups in this study is convenient and logical. Since the type of energy used by a household can be observed, there is not any problem in fitting a certain household into a specific category. The major criticism on the life-style categorization system used by Hosier is the difficulty in fitting households having continuous characteristics into discrete categorizations. In fact, numerous households simply not conform into any category. For example, what kind of category would be appropriate for a head of household working as a wage earner who simultaneously has a tract of land providing him with additional income? Would he be a wage earner or cash-crop farmer? Although Hosier developed a criterion to solve this problem, the criterion itself contained arbitrary elements. As noted previously in this chapter, household samples can be categorized into five energy user categories: agricultural waste users (AGWA), waste and fuelwood users (FWAW), fuelwood users (WOOD), biomass and kerosene users (KEBI), and kerosene users (KERO). Additionally, we hypothesize that there must be some characteristics associated with each energy-user group. 62 Some variables characterizing those user groups are presented in Table 13. Analysis of variance shows that several household characteristics such as income, education, and household size are statistically very significant across user groups. Household income and education of the head of household show a consistent pattern, hence they can be easily interpreted. For instance, households with high income levels can be expected not to belong to the AGWA group, because they are more likely to use fuelwood or kerosene. Education levels of household heads also showed similar patterns. The higher the household head’s education was, the less the likelihood that this corresponding household would use fuelwood. On the other hand, even though household size is significant, there is no consistent pattern shown by this variable. The plausible explanation for the lack of pattern associated with family size may be that some variables were correlated with household size in explaining household behavior toward the choice of energy. Another possible explanation is that, while household size certainly influenced the level of energy use, some other variables might be more important in explaining energy choice than the household size variable. The effect of the land ownership variable on energy use categorization was not significant. However, among households using biomass energy (AGWA, FWAW, WOOD, and KEBI) User group N 63 Table 13. Household characteristics by energy user category ‘ Variables i Family Head of fam. Land Family income** Education** own. Size** (RPS-/m0) (Ha) AGWA 48 25.244 1.79 0.474 4.18 FWAW 153 30.326 1.86 0.556 4.80 WOOD 459 47.531 1.81 0.536 5.54 KEBI 22 118.968 2.68 0.736 6.68 KERO 50 143.970 3.20 0.533 5.04 ; Income refers to monthly household income in ** 1000 Rps. Education refers to education of family’s head; Education is categorized into 1= no education, 2=elementary graduate, 3= middle school graduate, 4= high school graduate, and 5 = college and university graduate. Land refers to household land ownership in hectares. Size refers to size of family. Significant at 99 percent level. 64 there was a pattern of land ownership associated with type of energy user. On average AGWA users had less land and KEBI users had the largest amount. Unlike others, KERO users were largely comprised of wage earners. Thus, they possessed land, it was not necessarily used as a source of fuelwood. Instead, it was more of a source for additional income. Although some variables do differ across user groups, more information is still needed in examining this categorization. Information such as types of occupation, sources of income and distance to fuelwood sources are important. Summary Biomass energy, fuelwood in particular is still a dominant source of energy in rural areas of Central Java. Although kerosene is increasingly being used in rural areas, kerosene use for cooking purposes is still limited. Agricultural wastes play a significant role in the districts of Bantul, Banyumas, especially in Purworejo. High average daily energy use among rural households tended to associate with high income class group. There is also an indication, that less respondents among the highest income class group used agricultural wastes. 65 Family size tends to positively affect levels of household daily energy used. The larger the size of family is, the higher the levels of energy use would be. However, the amount of use per capita declines as family size become larger. Households with the head of family having lowest education level tend to rely on biomass energy (ie., agricultural wastes and fuelwood) for their cooking purposes. Average daily fuelwood consumption tends to positively related to land holding size. The amount is higher for households having larger tract of land. This tendency is not clear on agricultural wastes and kerosene use. High variation in energy consumption exists both in per household and in per capita levels. Thus, it is crucial to pursue analyses beyond average consumption levels in order to identify the factors affecting those variations. The classification of household samples into five energy-user groups provides more information concerning the likelihood of household response to energy options. Income and education provide consistent patterns across energy user groups. These results strongly support the importance of including the decision to use or not to use a particular type of energy among households in modelling rural energy consumptions patterns. IV. HBTHODB In the traditional approach to demand analysis, utility is assumed to be derived from consuming a bundle of goods. In this sense, goods are perceived as direct objects of utility. Yet, this approach ignores intrinsic properties that goods may have (Lancaster, 1966). In fact, characteristics of goods provide very useful information for consumers in choosing particular goods. Lancaster further argues that utility is derived from the properties or characteristics of the goods, not from goods themselves. He defines consumption as a process in which a combination of goods is transformed into output which is viewed as a bundle of characteristics. This bundle of characteristics is called a commodity (Becker, 1965). In other words, goods are inputs and commodities are outputs that are produced by combining various inputs.v This argument implies that commodities, not goods, actually enter directly into the utility function. The formulation of energy use in this study is in line with this notion of commodities, in the sense, that heat for cooking, not fuelwood, actually is consumed by a household. Therefore, fuelwood and other energy sources are inputs that together with other inputs (e.g., stoves) produce heat as a commodity. While other energy sources are substitutes for fuelwood, a stove is a complementary good. This formulation 66 67 is also used by Hardie and Hassan (1986) and Skog (1986) in their energy studies in the United States. Given the notion of heat as a commodity, we can formulate energy consumption decisions as containing two interconnected and continuous processes. First, given the availability of various types of fuel that can be used to produce heat, an individual household will decide to choose a particular energy among the alternatives. After a particular energy is chosen, the second step taken by a household is to determine the amounts of energy to use. Decision to use a particular energy The model of a qualitative choice has a general form as follows (Pindyck and Rubinfeld, 1981): 'U H. II F(X’B) = F(Zi) (1) Pi = probability to choose 1 type energy X = vector of explanatory variables B = vector of coefficients 21 = index which is determined by explanatory variable X F = cumulative probability function Two commonly used alternative probability functions are the normal and the logistic functions. If we assumed the cumulative probability is normally distributed, the probitl specification is appropriate. Assuming that the probability l The probit specification is explained in Appendix 2. 68 is distributed in the logistic results in the logit specification. One difference between the two distributions is that logit has a flatter tail (Pindyck and Rubinfeld, 1985) as shown in Figure 3. However, both probit and logit yield similar results, therefore the choice between them usually is based on a practical reason as Aldrich and Nelson (1984) noted: The logistic and normal curves are so similar as to yield essentially identical results. In practice, they yield estimated choice probabilities that differ by less than 0.02 and which can be distinguished, in the sense of statistical significance, only with very large samples. The choice between them, therefore, revolves around practical concerns such as the availability of computer programs and personal preference and experience (p. 34). The amount of energy use In this study the amount of a particular fuel used by an individual household for cooking is estimated by relating the level of use to the household’s economic and demographic characteristics. The estimation of the amount of a particular energy used by a household is modelled by assuming weak separability of a household’s utility from energy use. That is, an equation to predict amount of energy used is formed without referring to prices of non- energy products used (Skog, 1986). In analyzing factors affecting fuelwood use levels, Hadi (1982) in her study on rural energy in West Java employed ordinary least square (OLS) techniques to estimate 69 113' F(Z) —-—- probit model I -———'. .1031: model ’ L 2 Figure 3. Comparison of logit and probit cumulative distributions (Adopted from Pindyck and Rubinfeld, 1981) 70 fuelwood consumption among households using fuelwood for their cooking. To solve the problem of zero responses, Hadi excluded all non-users in estimating fuel- wood consumption. A seperate estimation for the probability and the ammount of a particular energy used, as in two-step approach, could only be done if in fact these equation are statistically independent. Yet, if equations are interdependent, the 01.5 tech- nique will yield a biased estimate (Hardie and Hassan, 1986; Tobin, 1958). Hardie and Hassan further show that the expected value of the random com- ponent in estimating the amount of energy used using OLS is as follows: E(Qi|di=l)=012}\i (2) where: e, = conditional expected error, 0.2 . intercquation variance, k, = f(.)/F(.) The terms f(.) and F(.) are the density function and the cumulative function of the standard normal random variable, respectively. The conditional mean of 9, would equal zero if 0,; equalled zero. That is, if both equations (estimating choice and amount) are independent. To solve the problem, Hardie and Hassan (1986) assumed that interdependence existed and introduced an instrumental variable A, in their equation for estimating the amount of energy used. Then, they tested the signifi- cance of the coefficient for A, 71 Survey data for the present study indicate that not all household samples used fuelwood for their cooking energy. However, the characteristics of all individual households were recorded. This situation, wherein some observations on the dependent variable corresponding to known sets of independent variables are not observable, is known as a censored sample (Judge et al., 1985). However, instead of employing approach used by Hardie and Hassan (1986), the Tobit model, which is designed to overcome a censored sample, is employed in the present study. Like probit, the Tobit technique allows one to have zero observations. The difference is, instead of qualitative binary responses, a dependent variable in the Tobit technique is the amount of energy used. But, unlike the probit, Tobit simultaneously estimates both the amount of energy use and the probability. A generalized Tobit model (Judge et al., 1985) can be written as follows: Y1 = XiB + ei if Yi > 0 = 0 otherwise (3) Y1 = amount of fuelwood burned (kg per day), Xi = vector of explanatory variables, B = coefficient vector, and ei = disturbance terms. 72 Assuming that 8 out of T observations are zero, the regression function can be written as : E(YiIXi, Yi >0) = Xi’B + E(ei|Yi > 0) i = 1, ...... T-S (4) If the disturbances, ei, are independent and normally 2 distributed N(0,0’) then: E(ei|Yi > 0) = E(ei|ei > -Xi’B) =CTf(Zi)/F(Zi) (5) where: Zi = Xi’B/d f(.) and F (.) = the density function and the cumulative density function of a standard normal random variable, respectively. Both f(.) and F(.) are evaluated at the argument (Judge et al., 1985).; Hence, the regression function can be written as: E(Yi|Xi, Yi > 0) = Xi’B +0'f(zin(zi) (14) i = l,....T—S The second term on the right hand side of equation (6) is an expected value of random component as in equation (2). If this second term is omitted during estimation, a problem is 2 Both the probit and Tobit are based on cumulative normal distribution. As Figure 3 shows, the slope will depend on the value of XiB selected in evaluating the impact of change. 73 created. The estimator of B is biased and inconsistent, using either the entire sample or the subsample of complete observation (Judge et.al, 1985). Thus, the potential bias is eliminated in the Tobit by jointly estimating the choice and the amount. To solve this problem, a maximum likelihood procedure will provide consistent and asymptotically normal parameter estimates for censored samples when the disturbance terms are normally distributed (Amemiya, 1973). Model specification Based on the discussion of previous studies and data available from the 1983 Rural Energy Survey the following variables are used as explanatory variables: household income (INCOME), age of the head of family (AGE), education of the head of family (EDUC), family size (SIZE), and size of land owned by household (LAND). Distance to a fuelwood source is a variable reflecting fuelwood scarcity (Hosier, 1985b). Unfortunately, the distance variable (DIST) available from the survey did not reflect the distance to a fuelwood source. Instead, it reflected the distance to any particular energy source used by a respective respondent. Since this variable could not be used to test the hypothesis that distance to the wood source may affect household fuelwood use, the DIST variable had to be dropped. Available fuelwood price data is used to reflect its relative scarcity. 74 Fuelwood price (PRIWOO) and kerosene price (PRIKER) are used as explanatory variables in this study. Price data are based on the recollections of household respondents regardless of whether or not the households did in fact use a particular energy. In the case of fuelwood, specific information on which households actually purchased fuelwood is not available. Therefore, fuelwood price is used as relative scarcity proxy for all the household samples. Qggision to use a particular type of energy and its level A Tobit model was developed to estimate the amount of energy use as well as the probability of using a particular type of energy. Both Tobit and the probit are specified as follows: Y1 = c + alxl + azxz + a3X3 + a4X4 + a5x5 + a6ED +%:91Di (15) where: for Tobit: Yi = amount of energy i used (1 = agricultural wastes, fuelwood, kerosene), = 0, otherwise, for probit: Y1 = 1, if a respective household use energy i, = 0, otherwise. for both: c = constant, X1 = income (in Rupiahs per month), 75 X2 = ratio of fuelwood price/kerosene price (in litres/kilograms), X3 = age of the head of family (in years), X4 = family size, X5 = land ownership (in hectares per household). ED = dummy variable for education, coded 1 if education of the head of family is elementary or less, and 0 if otherwise, Di = dummy variables for six districts where RBA RBT RBL RPM RPW RTM Estimation for wastes, fuelwood and kerosene) is done separately. :1, ll H each if if if if if if Banyumas: Bantul: Blora: Pemalang; Purworejo; Temanggung. (i = 1,... otherwise otherwise otherwise otherwise otherwise type of energy (agricultural 'I5)l This separation is based on the assumption that an individual household usually faces two alternatives energies at a time. This sense is supported by the results of descriptive analysis (Chapter III). For example, households using agricultural wastes are likely to perceive fuelwood as a substitute for wastes, whereas among wealthier households using fuelwood, kerosene is a potential substitute. Use of single equation model such as Tobit for a particular energy 76 explicitly incorporates the interdependence between energy and its close substitute. Tptal energy consumption Total fuel consumption of energy i (01) for the entire region is calculated based on the number of households using fuelwood in an entire region multiplied by average fuelwood burned per household (Hardie and Hassan, 1986). The equation is as follows: 01 = Pi * N * Qi (16) where: 01 = total daily energy consumption for the entire region (tons), pi = proportion of households using fuelwood, N = total households for the entire region, qi = the amounts of daily fuelwood used per household (in kilograms), and H. II type of energy. Proportion of households using fuelwood, pi, and the quantity of fuelwood used by a household per day, qi, are estimated based on the Tobit results. If one were pursuing a two-step approach similar Hadi's, probit could be used to estimate Pi and OLS could be used to estimate 9i. This is not, however, the preferred approach. 77 Expected coefficient signs of explanatopy variables on probability and consumption Household income is expected to be negatively related to fuelwood consumption, because fuelwood for cooking is mostly perceived as an inferior good. So, increasing household income will broaden energy alternatives available to the respective household. Increasing purchasing power enables them to purchase a more convenient form of energy such as kerosene. Hence, it is expected that the use of fuelwood will decrease as income increases because households may shift from fuelwood to using kerosene. This also implies that the probability of households using fuelwood decreases as household incomes increase. Demand analyses generally include the price of the good in question and the price of its substitute. In most developing countries, fuelwood is often freely collected and seldom passes through the marketplace (de Montalembert and Clement, 1983: Hosier, 1985b; Wardle and Palmieri, 1981). Hence, the market price for fuelwood, which is often available in the official statistics publication, only provides a tentative indication of the real cost. Cost of fuelwood is different for different people. For most rural households, fuelwood price refers to their own time and labor devoted to collecting wood. For the fuelwood trader, the fuelwood cost might include labor costs, costs of equipment to harvest it, as well as transportation and 78 storage costs (Wardle and Palmieri, 1981). In fact, the market price simply does not accurately reflect all those costs. For people who never pay cash for fuelwood, price is most likely not an important determinant in fuelwood consumption. On the other hand, for people who buy fuelwood from the market or local trader, the price of fuelwood is a crucial factor (Wardle and Palmieri, 1981). Unfortunately, in this present study, information regarding whether given households purchase or collect fuelwood is not available. Therefore, the price of fuelwood which was based on respondents' recollection is used to reflect a relative scarcity of fuelwood regardless of how households actually obtained it. Energy studies in West Java by Hadi (1982) and Hosier (1985b) in rural Kenya did not include price of wood as an indicator of scarcity. Hosier developed three surrogates for wood price namely time spent in collecting wood, distance to wood sources, and an indices for scarcity. None of those surrogates for price were statistically significant in affecting fuelwood consumption. Dwiprabowo et al. (1980) in their study in Aceh province of Sumatera included fuelwood price in their estimation and found that fuelwood price affected fuelwood consumption significantly. In this study price of fuelwood is expected to negatively affect the probability as well as the level of fuelwood use. On the other hand, the effect of kerosene 79 price is expected to be positive on the probability of choosing fuelwood. The amounts of fuelwood used is a priori assumed to be independent of the price of kerosene. The effect of family size on fuelwood consumption is expected to be positive. Increasing numbers of family members will certainly increase the amount of food the family consumes. Probability of using fuelwood, ceteris paribus, is expected to be positively influenced by family size. As noted in the third chapter, households with larger family sizes are likely to use biomass energy due to the relative availability of labor needed to collect fuelwood. The effect of family size on fuelwood use is also expected to be positive. Education is likely to affect the type of occupation of the head of family and to some extent it may correlate to family income. For example, the study by Susastro (1983) indicated that among the fuelwood purchasers in rural areas were notably teachers and other income earners. The majority of these individuals have higher education levels compared to those in average rural households. The higher the education of the head of the family is, the less likely that the respecting household will use fuelwood. In other words, the effect of education on both the probability and fuelwood consumption is expected to be negative. Land ownership is expected to positively affect both the probability and level of fuelwood consumption. The reason is clear: land possession guarantees availability of 80 fuelwood for household use. The importance of the home garden as a source of fuel and food in Java, and particularly in Central Java, has been extensively discussed, for example by Stoler (1978). The effect of districts, both in their magnitude and the signs of coefficient are not all clear. The uncertainly is due to the fact that districts are created merely as administrative units. Thus, they do not necessarily reflect the ecological stratification which may be more important in influencing fuelwood consumption. In fact, ecological variability within districts can be very substantial. For example, Pemalang district covers ecologically distinct lands ranging from lowland coastal to semi-highland areas. The district also consists of villages in different developmental stages, which to a lesser degree, influence fuelwood consumption. Hadi (1982) found that fuelwood consumption in more modern villages is less than that in traditional villages. Expected signs of coefficient of variables used in this study are presented in Table 14. 81 Table 14. Variables, variable names and expected coefficient signs Variable Variable name Expected sign 1 1. Household INCOME - income 2. Fuelwood PRIWOO - price 3. Kerosene PRIKER + price 4. Education EDUC - of the family head 5. Family size SIZE + 6. Land LAND + ownership 7. District RBA= Banyumas ? RBT= Bantul RBL= Blora RPM= Pemalang RPW= Purworejo ll-l Signs +, -, and ? represent a positive, negative, and an ambiguous effect, respectively. V. FACTORS AFFECTING RURAL ENERGY CONSUMPTION The explanation of factors affecting energy consumption is presented in three sections. It begins with a brief explanation of the variables used in the estimations. The subsequent section focuses on the explanation and discussion of various factors affecting levels of a particular type of fuel used and the probability (Tobit model results). Then, the estimation of energy use for the entire region is presented. Variables used ip the present study Tobit analysis is used to estimate the relationship between the amount of a particular type of fuel used and a set of explanatory variables. The variables are comprised of economic and demographic characteristics of household samples. Table 15 provides a list of explanatory variables which are used in the estimations in the next two sections. 82 83 Table 15. Variables summary Acronym Variable (unit) INCOME Household monthly income (Rps. 1,000). AGE Age of the head of family (years). SIZE Family size (number of members). LAND Household land ownership (hectares). ELES Dummy variable, coded 1 if the education of household head is elementary school or less, and 0 if otherwise. PRIRAT The ratio of fuelwood price and kerosene price (litres/kilograms). RBA Dummy variable for district, coded 1 if Banyumas, and 0 otherwise. RBL Dummy variable for district, coded 1 if Blora, and 0 otherwise. RBT Dummy variable for district, coded 1 if Bantul, and 0 otherwise. RPM Dummy variable for district, coded 1 if Pemalang, and 0 otherwise. RPW Dummy variable for district, coded 1 if Purworejo, and 0 otherwise. 84 Results of Tobit estimations A prior comparison of several estimations using the Chi-square test indicates that the inclusion of district as a dummy variable, in aggregate, is significantly different from that of the estimations without dummy variables for districts (Appendix 3). This implies that the district is crucial in estimating either the probability or the amounts of energy used. As a result, dummy variables for district are included in estimations. However, in the Tobit for estimating levels of agricultural wastes usage, districts variables are dropped altogether due to the failure of the model to achieve convergence after 20 iterations.l The education background of the head of family is categorized differently than the categorization used in Chapter III (p. 61). In this chapter, education levels are classified into two categories, elementary school or less and middle school or more. Model estimations using three education categories did not provide any significant effect of education on the probability of using a particular type of energy. This may be due to the limited sample size for education levels at or above the middle school level. Further, this aggregation scheme yields reasonable outcomes when employed in both Tobit and the probit estimations. l A computer printout is available, but the result is not amenable to interpretation. 85 Finally, the selection of final estimates is also based on the coherence of the coefficient signs to their expected signs and their statistical significances. The validation steps taken here follow general validation tests suggested by Kaplan (1964) which include the test of correspondence, coherence, and pragmatism. That is, how study results parallel the real world, relate to the body of knowledge, and are useful in the sense of their workability. The results of Tobit analysis appear in Table 16 (The probit results are presented in Appendix 4). Discussion of Tobit results begins with an explanation of the estimated sign and statistical significance of each coefficient. This is followed by the magnitude of the effects of the explanatory variables on the probability of using fuels and on levels of use. Siqn§:and statistical significance Household income Household income in Tobit models for agricultural wastes and fuelwood, as expected, provides a negative sign. The coefficient of income variable, reflecting the effects of household income on the probability of choosing agricultural wastes and fuelwood and the amounts of use, are statistically significant. The results imply that household income did alter the pattern of energy use and the probability of using biomass energy. Table 16. The results of estimation using Tobit technique Type of Energy Variable Wastes Fuelwood Kerosene **1 CONSTANT -8.l442 — 1.4334 -o.7529 INCOME -o.370E-01** -0.107E-01** 0.434E-02** (0.953E-02)3 (0.0034) (0.113E-02) PRIRAT 15.1992* 11.5729** (7.0216) (3.4772) AGE 0.876E-01** -0.498E-02 0.200E-02 (0.229E-01) (0.134E-01) (0.613E-02) SIZE 0.891E-03 0.5083** - 0.1146** (0.1546) (0.847E-01) (0.0416) LAND 1.0863 0.3083 - o.2391** (0.3924) (0.2193) (0.0828) ELES 1.5160 s 2.9074** - 1.5741** (1.0302) (0.5092) (0.1920) RBA —1.3897** - 0.2976 (0.5435) (0.2280) RBL -2.8649** 0.0463 (0.5550) (0.2230) RBT -3.4234** -1.1645** (0.5561) (0.3305) RPM -3.2334** -o.1106 (0.5500) (0.2305) RPW -3.9387** -0.1878 (0.6274) (0.3163) L.likelihood= -859.97 —1894.80 -155.67 1 ** , * and $ are significant at 1, 5 and 10 percent, respectively. Figures in parentheses are standard deviations. 87 The negative sign shows the tendency of decreasing the probability of a household using fuelwood as the household's income becomes higher. In other words, a household with a higher income level may likely shift from using fuelwood to using kerosene (or other more convenient forms of energy). However, among households which decide to use fuelwood, increasing income may increase food consumption and thus, would require the use of more energy for cooking. Increasing a household’s income also decreases the household’s probability choosing agricultural wastes. They may switch to either fuelwood or kerosene. Which energy form they may actually use depends on how much their income increases, ceteris paribus. The descriptive analysis presented in the previous chapter (Table 13) indicates that differences in the average monthly income are statistically significant for the users of agricultural wastes (AGWA), fuelwood and wastes (FWAW), fuelwood (WOOD), fuelwood and kerosene (KEBI), and kerosene (KERO) . Conversely, the coefficients of the household income variable in kerosene results were positive with significant coefficients. Therefore, higher income encourages a household to use kerosene and probably other convenient forms of energy as well. These findings are also supported by the descriptive analysis presented in Chapter III. The descriptive analysis shows that a household in the 88 highest income level category did belong to the kerosene user group (KERO). On the other hand, household samples using entirely agricultural wastes (AGWA) did belong to the group with the lowest average income level. Age of head of family In most cases, the variable (AGE) is not significant in estimating either the probability of a household using a particular type of energy or the amounts of a particular fuel used. This is consistent with the findings of Nasendi (1978) and Hadi (1982) that the inclusion of the average age of husband and wife, and age of husband did not provide a significant effect on fuelwood consumption. The age variable, however, significantly affected the level of wastes used with a positive sign. This implies, ceteris paribus, that a household with an older family head is more likely to use more agricultural wastes. The older family head is likely to have a smaller number of family members (Appendix 5) which means less labor is available. As a result, this encourages them to use the energy source which is relatively more available near the house such as wastes. In fact, a home garden (see the discussion on land ownership) is also an important sources for agricultural wastes such as branches, twigs, and leaves which are common source of energy. 89 Education level Low education levels (ELES) positively affected both the probability of using agricultural wastes and fuelwood and their amounts. On the other hand, the ELES variable is negatively and significantly influences both the probability of using kerosene and the amounts of kerosene used. The results imply that, given the availability of energy, a household with a higher education level has a lower probability for using fuelwood and wastes than those households with lower education levels. There are several possible explanations for this phenomenon. First, education may widen one's opportunity to seek jobs other than farm work, such as teaching or other salary earning type occupations. Having cash money at hand, in turn, seems to broaden one’s access to commercial energy, especially kerosene. Second, in addition to the occupation type, Burk noted the importance of a "reference group" in affecting expenditure patterns (Hadi, 1982). A peer group, for example, may also influence the life style of a member of a corresponding group, including the energy type they use. Third, it is likely that people with higher levels of education have more access to various kinds of information. For example, access to farm inputs, capital markets and new technologies may likely benefit them more than would be the case for those with little or very limited amounts of information. As a result, they may be 90 able to improve their land productivity and produce a surplus that could be traded for cash. They even have the possibility to getting non-farm jobs to supplement their income. Thus, in relation to the categorization by Hosier (1985a), they may more appropriately be categorized as cash-surplus farmers, rather than subsistence farmers. The evidence shows that cash-surplus farmers use less fuelwood as compared to subsistence farmers. The descriptive analysis in Chapter III also showed that the average education of a household using entirely kerosene (KERO) was higher than that of a household using agricultural wastes (AGWA). In fact, there is a gradual increase in terms of the average education of household head when moving from the users of the least convenient form of energy to the most convenient energy forms. §i§e of family The family size variable is not significant in Tobit models estimating agricultural waste usage. As mentioned previously, wastes might be relatively available and easier to collect than fuelwood, hence its collection does not require too much labor. Family size yields a statistically significant coefficient for Tobit fuelwood models with positive signs. This sign and its significance implies that both the probability of using fuelwood and the amounts used is higher as family size gets larger. 91 A possible explanation for this tendency is the fact that increasing the number of household members, especially children, may mean additional labor for the family. The importance of children among peasant families was studied by Nag et al. (1978) in a Javanese village in the Jogyakarta area. They found that there was a tendency among households to adopt the reproductive strategy which calls for "having as many children as they can afford and find useful". This strategy is based on the commonly held perception among the peasants that children are a source of capital and labor. Children also mean security, because children are expected to take care of their parents when parents reach old age. Nag et al. (1978) also found that the average time spent in collecting fuelwood among children of age six to fourteen years was longer in Java compared to the time spent in Nepalese villages. Larger families have more labor available for various activities including collecting wood. The family size variable is negative and significant in the Tobit model for kerosene. Larger families among kerosene-using households result in additional income sources and the ability to afford commercial energy, kerosene. Alternatively, a larger family has more labor available for gathering wood, hence discouraging households to use kerosene. As a result, the amounts of kerosene used is lower as family size gets larger. 92 Land ownershipg Land ownership, measured in hectares, is positive but not significant in affecting the amount of agricultural wastes and fuelwood. Variables other than land holdings may be more important in explaining the amount of wastes and fuelwood used. On the other hand, the quantity of kerosene is negatively and significantly influenced by land holdings. Increasing land ownership likely increases income and may encourage households to use more kerosene as this commercial fuel become more affordable. Land has a crucial role within Javanese society. Land is not only perceived as a productive means, but it is also a status symbol for the owner. In the case of fuelwood production, land possession especially in the form of a home garden is a significant contribution as a fuelwood source (Stoler, 1978). Unfortunately, in this present study there is no information regarding the type of land use by each household sample. As previously mentioned, a surplus of agricultural products may result from increasing land holdings. These products can be traded for cash leading to increases in household income. As a household’s income increases, its expenditures for food stuffs are also expected to increase therefore requiring increased quantities of fuelwood. However, above certain income levels, households may consider using their excess income for switching 93 completely to commercial energy and dropping fuelwood entirely as an energy source for cooking. En r rice The coefficients for the price ratio are positive and significant in both Tobit models for fuelwood and kerosene. For kerosene, this implies that increasing the price ratio increases amount of kerosene used as expected. The positive and significant price ratio for estimating fuelwood levels is different than expected. A possible explanation is that fuelwood users are comprised of both households which purchase fuelwood and households which collect fuelwood. As long as a fuelwood source is relatively available, households collecting fuelwood might use more than households purchasing it. For the latter group, the amount of fuelwood used is most likely limited by the price of fuelwood. As a result, when both groups are included in the calculation of the average amount of fuelwood used per household, the average figure may be biased upward. Unfortunately, this argument cannot be verified further as the information that would permit a categorization of fuelwood users into purchasers and collectors is not available. In spite of how one measures fuelwood scarcity, increasing wood scarcity is most likely to reduce the probability of households using fuelwood. However, 94 whether or not the corresponding household will actually consume kerosene may depend on other factors such as income and energy availability. Region Regional effects are different for each type of energy. All districts included as dummy variables give negative and significant coefficient in the Tobit model for fuelwood use. This reflects a substantial variation in fuelwood use levels across the sample districts: all are lower than use levels in the district of Temanggung. The levels of fuelwood used in Temanggung district is calculated by setting all district dummy variables equal to zero. Tobit results for kerosene also gives negative and significant coefficient for Bantul district. The probability of using a particular type of energy As previously mentioned, the Tobit model provides both estimations on the probability and the amount of energy use. The probit model, on the other hand, only provides an estimate of the probability of using a particular type of energy. To see a consistency between the two models, the results of probit and Tobit are used as a basis in predicting the probability. The estimations are calculated at the mean for all samples, the results are presented in Table 17. 95 Table 17. Probability of using energy F(Z) based on the probit and Tobit results Energy type Model Ag. wastes Fuelwood Kerosene probit 0.25 0.92 0.02 Tobit 0.25 0.89 0.02 Both the probit and Tobit provide consistent results in estimating the probability of using a particular type of energy. Those models yield a similar probability for agricultural wastes and kerosene. In estimating the probability of using fuelwood, the probit gave somewhat higher value the than Tobit model. However, both results show that the probability of using fuelwood are high. Since the Tobit model provides more information than the probit, Tobit results are used in the subsequent analysis. The amountgygf energy use In order to describe in detail the effect of a household’s characteristics on energy use levels, the Tobit results for each type of energy are examined. The probit and Tobit as previously mentioned are based on the cumulative standard normal distribution function, hence the marginal change is greatest near the point where the probability to use fuelwood is 0.5 and the change gradually decreases as we move away from 0.5 in either 96 direction. Therefore, the interpretation of the probit and Tobit are usually evaluated at different values of XiB. The Tobit results for each type of energy are evaluated at mean values for all samples and for samples using a particular type of fuel in question. The interpretation of the Tobit follows the procedure proposed by McDonald and Moffitt (1980). The expected value for level of fuelwood use for all samples, E(Y), can be calculated based on the expected level of use conditional upon being above the limit, E(Y*), times the probability of being above the limit, F(Z), or: E(Y) = F(Z)* E(y*) (17) They further decomposed the effect of a change in the kth, variable of X, on Y into two parts: (1) the change in Y of those above the limit, weighted by the probability of being above the limit, and (2) the change in the probability of being above the limit weighted by the expected value of Y if above (McDonald and Moffitt, 1980). The total effect is as follows: dE(Y)/ka = F(Z)[dE(Y*)/ka] + E(Y*)[dF(Z)/dxk] (18) As mentioned previously, the Tobit models are evaluated at the means for all samples and means for samples using a particular type of fuel. However, the computation of XiB for fuelwood based on the variable mean for samples using fuelwood and the mean for all samples (Table 18) yield 97 values of XiB equal 6.23569 and 5.93024, respectively. Since those figures are almost the same, for comparison purposes the Tobit results are interpreted at two values: (1) XiB = 6.2357 calculated on the mean values for households using fuelwood, and (2) XiB = - 0.500 which is chosen arbitrarily to represent the lower and negative values. The distribution of XiB for fuelwood is presented in Figure 3. Results of the Tobit decomposition for agricultural wastes, fuelwood and kerosene are presented in Tables 19, 20, and 21, respectively. As previously mentioned, the marginal change of the probability near 0.5 is higher than at the point distant from 0.5. For example, at the mean for household samples using fuelwood with F(Z) = 0.9333, the marginal change in the probability (dF(Z)/ka) is less than at XiB = -0.500 with F(Z)=0.4528 (Table 20). Specifically, the values of the marginal change in the probability are 0.031 and 0.095, respectively. This implies that the change in one explanatory variable, for example household income, may 98 Table 18. Mean values of the variables for samples using fuelwood and for all of the samples. Mean for Variable samples using all samples fuelwood(N=634) (N=732) INCOME 45.29302 50.71264 PRIRAT 0.09982 0.10037 AGE 49.38013 49.63115 SIZE 5.39274 5.28962 LAND 0.55395 0.54740 99 Number 0 * 122121 .. 120: 7 113 ' a: . - - _ 100"“ —— , 2 92 l i L . 80 (.._,,_.,._..-__...,._._. .. 3 v ’ ‘ I ............ . ____________ 1 1‘ ‘ l L l| ‘3 ‘ 60 -- 53 1 ; i _ , 1 ' --~----~-—--~——-—-< ’ 1“ “i 43 1 ‘ 1| * 39 . _, 4o-—-—-—- 1 (l y ”-1 i i l 1 i l W ‘ 20 ~-—- ‘ 1 3 1 | (+5 ----------- l i I l 7 l I l l ‘ 0 '05 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5105 XB ‘ all urnplu " “mole. unit-g luoluood Figure 4. X18 distributions for all samples 100 Table 19. Decomposition of Tobit model for agricultural waste use Evaluated at Mean for samples using wastes x18: -2.7059 Mean for all samples XiB=-3.7923 Expected waste used (kilogram) E(Y) Probability of using wastes: F(Z) Expected wastes used conditional upon being above limit (gilogram): E(Y ) Change due to a a change in variable Xk : dE(Y)/ka dF(§)/dxk dE(Y )/ka Change due to a change in waste use:* F(Z).dE(Y )/ka Change due to a change in probability of using wastes E(Y ).dF(Z)/dxk 0.8585 0.2546 3.3718 0.2546 Bk 0.0560 Bk 0.2588 Bk 0.0658 Bk (26 %) 0.1888 Bk (74 %) 1.1692 0.3192 3.6629 0.3192 Bk 0.0624 Bk 0.2842 Bk 0.0907 Bk (28 2) 0.0907 Bk (72 %) 101 Table 20. Decompositions of Tobit model for fuelwood use Evaluated at Mean for Mean for all samples households the value of using fuelwood XiB=-0.5000 XiB= 5.0237 XiB=6.2357 Expected fuelwood used (kilograms): E(Y) 1.4122 5.2464 6.3510 Probability of using fuelwood (in proportion): F(Z) 0.4528 0.8868 0.9333 Expected fuelwood used conditional on being above limit (kilograms): E(Y ) 3.1222 5.9195 6.8057 Changes due to a change in variable Xk : dE(Y)/dxk 0.4528 Bk 0.8868 Bk 0.9333 Bk dF(Z)/dxk 0.0955 Bk 0.0461 Bk 0.0310 Bk dE(Y )/ka 0.3409 Bk 0.6924 Bk 0.7740 Bk Change due to a change in fuelwood use (kilograms): F(Z) dE(Y*)/dxk 0.1541 Bk 0.6140 Bk 0.7223 Bk (34%) (69%) (77%) Change due to a change in probability of using fuelwood: E(Y*) dF(Z)/dxk 0.2987 Bk 0.2727 Bk 0.2110 Bk (66%) (31%) (23%) 102 Table 21. Decomposition of Tobit model for kerosene use Evaluated at Means for Mean for sample all samples using kerosene XiB=-1.6206 XiB=-1.0540 Expected kerosene used (in litre): E(Y) 0.0077 0.0387 Probability of using kerosene (in proportion): F(Z) 0.0250 0.1021 Expected kerosene use conditional upon being above limit (in litre): E(Y*) 0.3084 0.3793 Changes due to a change in variable Xk : dE(Y)/dxk 0.0250 Bk 0.1021 Bk dF(§)/dxk 0.0706 Bk 0.2139 Bk dE(Y )/ka 0.1303 Bk 0.2052 Bk Change due to a change in kerosene use: F(Z). dE(Y )/ka 0.0032 Bk 0.0209 Bk (13%) (20.5%) Change due to a change in probability of using kerosene: E(Y ).dF(Z)/ka 0.0218 Bk 0.0811 Bk (87%) (79.5%) 103 likely to encourage households in the border (near point 0.50) to switch from fuelwood and use other type of fuel for their cooking energy. Decomposition of Tobit models for agricultural wastes and kerosene (Tables 19 and 21) show that values of E(Y), F(Z), and E(Y*) are consistently greater when evaluated at the mean for samples using either wastes or kerosene than at the mean for all samples. Those values of XiB for samples using either wastes and kerosene are located closer to the 0.50 than all samples. The total response to a change in one explanatory variable can be separated into the response due to increasing the level of a particular fuel used by households and to increasing the proportion of households using that particular type of fuel. The relative contribution of each part in the total response will depend on the value of XiB selected. In this study, decomposition evaluated at the mean of households using fuelwood shows that 77 percent of the total response is due to increasing the amount of fuelwood used, whereas the rest is due to the increasing proportion of households using fuelwood. Conversely, at the point XiB = -0.500, the larger part of the total response (66 percent) is caused by the increase in the probability of households using fuelwood. 104 In general, this implies that a marginal change in one explanatory variable, ceteris paribus, will lead largely to a change in the amount of use among households already using a particular type of fuel. Whereas, the total response to change in one variable among non-using households, will be influenced more by changes in the probability of using a particular fuel. It is important to examine the effect of a change in an individual variable on the expected amount of energy used, E(Y), the probability, F(Z), and the expected amount of energy used conditional upon being above the limit, E(Y*). To evaluate the effect, each individual variable which gives a significant coefficient, is assumed to increase by 1, 5 and 10 percent, holding other variables constant at their means. The results are presented in Tables 22, 23, and 24. Elasticities are presented in Tables 25, 26 and 27. Increasing some explanatory variables by 1 and 5 percent give negligible effects both on the probability and the expected amount of energy use. Income increases by 5 percent reduces the daily household amount of wastes and fuelwood used by 0.012 and 0.029 kilograms, respectively when evaluated at the mean for all samples. The reduction on fuelwood use level, for example, comes forth from an 0.2 percent decrease in the expected probability of using fuelwood and 0.020 kilograms decrement in the amount of fuelwood used among households who actually used fuelwood. 105 Table 22. Effect of change in income on the expected use level E(Y), probability F(Z), and expected level conditional upon being above the limit E(Y ) for agricultural wastes Calculated at mean for all samples 1% 5% 10% E(Y) - 0.0030 - 0.0124 - 0.0376 F(Z) - 0.0031 - 0.0063 - 0.0126 E(Y*) - 0.0054 - 0.0075 - 0.0497 106 Table 23. Effects of change in explanatory variable on the expected use level E(Y), probability of using F(Z), and expected level conditional upon being above the limit E(Y ), for fuelwood Calculated at mean for all samples 1% 5% 10% INCOME SIZE PRIRAT INCOME SIZE PRIRAT INCOME SIZE PRIRAT Change in E(Y) in kilograms: - 0.0033 - 0.0287 - 0.0202 0.1178 0.0127 0.0678 Change in F(Z) - 0.0000 - 0.0019 - 0.0005 0.0057 0.0004 0.0035 Change in E(Y*) in kilograms: - 0.0037 - 0.0197 - 0.0193 0.0939 0.0103 0.0527 0.0411 0.2218 0.1048 0.0076 0.0110 0.0057 0.0337 0.1742 0.1017 107 Table 24. Effects of change in explanatory variable on the expected use level E(Y), probability F(Z), and expected level conditignal upon being above the limit E(Y ), for kerosene Calculated at mean for all samples 1% 5% 10% Change in E(Y) in litres: INCOME 0.0000 0.0002 0.0004 PRIRAT 0.0004 0.0016 0.0037 SIZE - 0.0004 - 0.0008 - 0.0066 LAND - 0.0001 - 0.0001 - 0.0020 Change in F(Z): INCOME 0.0000 0.0010 0.0056 PRIRAT 0.0000 0.0044 0.0094 SIZE - 0.0002 - 0.0016 - 0.0197 LAND - 0.0000 - 0.0006 - 0.0062 Change in E(Y*) in litres: INCOME 0.0000 0.0056 0.0148 PRIRAT 0.0080 0.0148 0.0224 SIZE - 0.0085 - 0.0140 - 0.0149 LAND - 0.0029 - 0.0052 - 0.0068 108 From the above explanations several observations concerning rural household behavior toward energy usage using Tobit model are notable. The effects of 1 and 5 percent change in one explanatory variable using Tobit results yield a small change in the probability to use or not to use a particular energy. The effects are somewhat larger if we increase explanatory variable by 10 percent. This implies that a substantial change in most explanatory variables is needed to influence the probability. The change on the probability to use or not to use a particular type of energy is greater near mid-point in the cumulative normal distribution function. Therefore, for households close to this point, a small change in one variable may be sufficient to affect their decision to choose one energy over another. Conversely, for those far away from that point, a substantial change in an explanatory variable is needed to change their decision. Subsistence households may need a considerable increase in their income, ceteris paribus, before switching to commercial energy. 0n the other hand wealthier rural households may shift from using wood to kerosene with from a small increase in their income. The Tobit model indicates that among households using a particular energy, a change in one variable is associated with a change in the amount of energy used and the probability of use. Elasticities are presented in Tables 25, 26 and 27. 109 Table 25. Elasticity of income calculated at the mean for all samples for agricultural wastes Elasticity E(Y) - 0.5565 F(Z) - 0.4127 E(Y*) - 0.1440 110 Table 26. Elasticities calculated at the mean for all samples for fuelwood Variable Elasticity INCOME E(Y) - 0.0917 F(Z) - 0.0282 E(Y*) - 0.0635 SIZE E(Y) 0.4545 F(Z) 0.1398 E(y*) 0.3145 PRIRAT E(Y) 0.2579 F(Z) 0.0793 E(Y*) 0.1785 111 Table 27. Elasticities calculated at the mean for all samples for kerosene Variable Elasticity INCOME E(Y) 0.7146 F(Z) 0.6215 E(Y*) 0.0930 PRIRAT E(Y) 3.7720 F(Z) 3.2812 E(Y*) 0.4909 SIZE E(Y) 1.9682 F(Z) 1.7119 E(Y*) 0.2561 LAND E(Y) 0.4250 F(Z) 0.3697 E(Y*) 0.0533 112 Total energy consumption The calculation of the total energy use for each type of fuel by household for the entire region is based on the equation (16) described in Chapter IV. For example, the total daily energy use for entire region, Qi, is computed as: Qi = Pi * N * Qi (19) where: Qi = annual use of energy i (i=1, 2, and 3, each for wastes, fuelwood, and kerosene), pi = proportion of households using energy 1, N = number of households, and qi = the amount of energy i used by household. The proportion of households using fuelwood, p, and the amount of daily fuelwood use by a household, q, are computed based on Tobit results. The population of households in the region, N, is based on the estimates released by the Central Bureau of Statistics (Biro Pusat Statistik, 1984). The results of computation along with the respective figures from the survey are presented in Table 28. The proportion of household using a particular type of energy based on Tobit estimations and the survey show that the proportion of households using agricultural wastes, fuelwood and kerosene are underestimated. Estimation of proportion of households using agricultural wastes yields a substantially low figure compared to the survey results. 113 This underestimation is likely due to the exclusion of regional dummy variables in Tobit equation for agricultural wastes. The underestimation in Tobit for wastes and kerosene might have resulted from too few samples for household using those types of energy as observed in the survey. Waste users for example, had not even been observed in three districts (ie., Blora, Pemalang, and Temanggung). No single household using kerosene for cooking was observed in the district of Bantul, while the total number of households using kerosene for cooking in the entire region accounted for less than 10 percent of the total household samples. Since the estimate for agricultural wastes yields very low proportion, prediction of the total wastes use for the entire region is not undertaken. Total number of households in the rural area of Central Java region was approximately 4,744,042 households in 1983 (Biro Pusat Statistik, 1987). The proportion of households using each type of fuel based on the survey and Tobit results are presented in Table 29. The total energy consumption by rural households for the entire region based on the Tobit models in comparison with survey results is presented in Table 30. Annual consumption for each type of fuel by household is calculated based on the level of daily energy use both from the survey results and Tobit estimates. Table 28. 114 Comparison between survey results and Tobit on the proportion of households using a particular type of energy and their levels Proportion Levelsl Energy tYPe/ - users survey Tobit survey Tobité Wastes 0.274 0.027 3.70 3.37 (2.97)—3- Fuelwood 0.866 0.856 6.18 5.91 (3.84) Kerosene 0.098 0.075 0.80 0.31 (0.45) Mr Fuelwood and wastes levels are in kilograms, kerosene is in litres. IN Ibo Expected amount conditional being above the limit E(Y ). Figures in parentheses are standard deviations. 115 Annual household energy used is based on the assumption that daily energy use recorded by the survey reflects an ordinary daily use throughout the year. A similar assumption was used by Sumarna and Sudiono (1973) in estimating the annual energy use by household in East Java province. This assumption, however, ignores the fact that seasonality may affect the level of energy used per year by households. Thus, the estimation results should be carefully used. Susastro (1983) and Komarudin (1989) noted that efficiency of a traditional stove commonly used in rural areas of Indonesia, including Central Java region is relatively low. Stove efficiency ranges from 4 to 16 percent. In the present study, no adjustment is made for efficiency. Thus, energy consumed by sampled households represents a fuel as an input to produce heat. The level of fuelwood used per household per year based on survey results and Tobit analysis are 2.71 and 2.59 cubic meters, respectively.; By considering that average size of household based on the survey is 5.28 persons per family, fuelwood use per capita were 0.51 and 0.49 cubic meters per year, respectively. 2 Conversion factor used is: 1 ton solid wood = 1.2 cubic meter roundwood (Sumarna and Sudiono, 1973 p. 17). 116 Table 29. The proportion of rural household by energy type in Central Java region (p*N) Energy type Survey Tobit Fuelwood 4,108,340 4,060,900 Kerosene 464,916 355,803 Table 30. Annual households energy consumption in rural areas of Central Java Survey Tobit Energy type Per Total Per Total household household Fuelwood(tons) 2.25 9,243,765 2.16 8,771,544 Kerosene 0.29 134,826 0.11 39,138 (in 1000 litres) 117 In the case of fuelwood consumption, the Tobit results indicate that increasing household income by 1, 5 and 10 percent reduces the amount of fuelwood per household by 0.0037, 0.0197 and 0.0337 kilograms calculated at the mean for all samples (Table 23). On a per year basis, the average reduction are 1.35, 7.19 and 12.3 kilograms per household or 0.001, 0.007 and 0.012 tons, respectively. These reductions are relatively small. Clearly, increasing income by 1 and 5 percent will not significantly reduce the total fuelwood consumption for the entire region. For example, annual fuelwood consumption with and without increasing income is approximately 10.5 million cubic meters. Increasing income by 10 percent, however, decreases region’s fuelwood consumption about 0.30 million cubic meters (Table 31). This would be equivalent to eliminating fuelwood consumption for approximately 117,005 households. 118 Table 31. Total fuelwood consumption if household income increase by 1, 5 and 10 percent Total fuelwood use Income constant Income increase 1% 5% 10% In tons: 8,771,544 8,767,483 8,743,118 8,519,009 In cubic meters: 10,525,853 10,520,980 10,491,741 10,222,811 VI. CONCLUSIONS AND IMPLICATIONS This study is aimed at addressing two objectives. The first is to present a description of rural household energy consumption. This includes the identification of type of energy, amount of use, and characteristics of energy users. The second objective is to analyze the relationship between household’s economic and demographic characteristics and energy use. The effect of a household's characteristics on both the probability of using a particular type of energy and its level of use are of study interest. Study conclusions and implications are presented in this chapter. Conclusions Biomass energy, especially fuelwood and wastes, still play a major role as an energy source for cooking purposes in rural areas of the Central Java region. In 1983, approximately 93 percent of rural households sampled in this region used biomass energy; nearly 86 percent used fuelwood. As many as 6.5 percent of sampled households used only wastes for their cooking energy. However, the contribution of wastes as a source of energy, either as a sole energy source or in combination with other types of energy, was substantial especially in the village samples of the Bantul and Purworejo districts. In those districts, the proportion of household samples using wastes as a sole energy source or 119 120 in combination with other types of energy was 42.8 percent and 93.6 percent, respectively. Fuelwood was the only biomass energy used by rural households in the Blora, Pemalang, and Temanggung districts. These differences indicate the importance of including the district as a variable in estimating rural energy use. Categorization of household samples into five energy user groups (AGWA, FWAW, WOOD, KEBI, and KERO) reveals the presence of interfuel substitution. The energy options facing rural households ranged from the most inconvenient (ie., wastes) to the most convenient form of energy (ie., kerosene). In fact, household's economic and demographic characteristics relate to the type of energy they used as a primary energy source for cooking. In other words, a decision-making process within a household exists for deciding to use or not to use a particular type of energy. Descriptive analysis of household samples (Chapter III) shows that several characteristics such as household income, education level of family head, household size, and land ownership influenced either type of energy selected by a household or its level of usage. In the case of fuelwood consumption, increasing household income increased the amount of fuelwood used by a household. However, the number of households using fuelwood decreased as household income increased. 121 Further analysis revealed factors linking household economic and demographic characteristics to energy-user categories and indicated that family income, education of the head of family, and family size were significantly different among energy user groups. This implies that a household may change its preference toward a particular type of energy as its economic and demographic characteristics change. As people become more market-oriented they tend to choose more convenient types of energy and use lesser amounts of biomass energy such as fuelwood and wastes. A Tobit model linking the household choice of a particular type of energy and the amount to its economic and demographic characteristics was utilized. Tobit results, compared to the probit results, provide consistent estimates for household’s probability of using a particular type of energy. Tobit results show that household income and household size altered both the probability of a household in choosing fuelwood and its amount. Dummy variables for education of the head of the household (ELES) yielded significant coefficients with negative and positive signs for kerosene and biomass energy, respectively. A household with the head of family having less education (ELES) has a higher probability of choosing fuelwood than a household head with a higher education background. The effects of family size are positive for fuelwood and negative for kerosene, both are with significant coefficients. As family size gets larger and more labor is 122 available, households tend to use more fuelwood and less kerosene. This implies the importance of population variable on rural energy consumption. The Tobit decomposition following McDonald and Moffitt (1980) provides important results. The total response of a change in one explanatory variable can be broken down into two parts. The first is a response due to the change in the amount of energy used. The second is a response due to the change in proportion of households using fuelwood. Evaluation at the mean for all samples indicates that contributions to the change in the amount of fuelwood use and the change in a number of households using fuelwood are 77 percent and 23 percent, respectively. On the other hand, among non-users (XiB = -0.500), the contribution of each part is reversed, 34 percent and 66 percent, respectively. These results suggest that theoretically it is possible to alter one’s decision to use or not to use fuelwood by developing policies which change levels explanatory variables. The response of each household to a change in explanatory variable is different depending on the status of household’s economic and demographic characteristics. For example, one important rural energy policy question may relate to reducing the total fuelwood consumption in order to reduce the deforestation rate. Household income is an important variable that is statistically significant in reducing the probability of selecting fuelwood as well as the level of fuelwood used. 123 The effect of increasing levels of explanatory variables on the probability of using a particular type of fuel and the amount are different for each type of fuel. However, increasing household income by 1 and 5 percent resulted in little reduction of the probability of using all types of energy as well as on their amounts. As shown by Tobit results, the probability of using fuelwood is barely affected by a marginal change in income with 1 and 5 percent increases. The effects on the other types of energy are also negligible. The Tobit results indicate that the amount of fuelwood use, among fuelwood users, decreases as much as 0.0037 and 0.0197 kilograms as household income increases by 1 and 5 percent, respectively. Or, it equals 0.001 and 0.007 tons per household per year. In terms of total fuelwood use for the entire region, these increase in income by 5 percent or less may not significantly affect the total fuelwood consumption for the entire region, which remains at approximately 10.5 million cubic meters. If income increases further, for example by 10 percent, the total fuelwood consumption for entire region decreases by 0.30 million cubic meter or equivalent to eliminating fuelwood consumption by more than 117,000 households. 124 ram Several limitations in this study are notable. First, this study focuses on demand side. A paucity of data on rural energy availability is a major constraint in providing more complete picture of rural energy status. Estimation of demand with single equation model is very restrictive as it assumes that the decision to choose a particular type of fuel among energy alternatives is independently determined. In other words, this model does not permit the error term eij to be correlated. This study is based on the survey data collected by Forest Products Research Institute (FPRI) of Indonesia in 1983. Although various data are available, some important information were not properly recorded. Heat content of energy used in rural areas was not measured during the survey, hence the conversion of energy use from a physical unit (kilograms) into BTU could not be done. As a result, energy input in this study is measured in kilograms for fuelwood and agricultural wastes and litres for kerosene. Important variables, such as distance to fuelwood sources which may be a better proxy than price in reflecting fuelwood scarcity were not available. Lastly, the survey only covered one period of time, hence the seasonality is not reflected by the survey data. 125 This study is confined to Central Java. Although the magnitude will certainly be different from region to region, however, the implications may be applicable to other region as well. Implications The study results indicate that both the probability and the level of a particular energy used are influenced by several explanatory variables. However, as shown by the effect of marginal change in explanatory variables, the impact is negligible in practical terms. In other words, to significantly affect overall rural energy consumption a substantial change in some explanatory variable such as income, education, and family size are crucial. Thus, if the government wants to be more successful in halting deforestation, the national economic development plan currently implemented should focus on providing more job opportunities and hence income-earning opportunities, particularly for rural people, as higher income is equated with less fuelwood use. The nationwide education program that is currently being carried out will also positively contribute to reducing the deforestation rate as suggested by the Tobit results. Opening up the opportunity for people to obtain better education may result in decreasing demand for fuelwood, as higher education levels lower the probability of using fuelwood. 126 The family planning program plays a positive and significant role in altering rural energy consumption, as variable household size is significantly affect both the probability and the amount of energy use. Decreasing population which in turn is reflected by decreasing the number of household relying on fuelwood will substantially affect the total fuelwood consumption. These solutions through increasing real income, education and family planning are long-term programs. In the mean time, fuelwood is still a basic energy necessity for most rural households, particularly in the Central Java region. Fuelwood use level per household per year based on the Tobit estimation is 0.49 cubic meters per capita per year. In other words, there must be a short-term solution for meeting basic energy needs. Implementing fuelwood production programs on private lands will be restricted by the fact that available land resources in Java, the most densely populated island, are becoming increasingly more scarce as the competition for land for different purposes such as housing and industrial sites becomes more intense. Agroforestry which integrates multiple production on a given tract of land (Wiersum, 1979) is one alternative land use practice that may be appropriate for the conditions in Central Java. This approach is appropriate since fuelwood is a basic need along with food and shelter. Solution to fuelwood problem, hence must be considered from broader perspectives (Dewees, 1989). 127 As implied by the low efficiency level of traditional stoves (Susastro, 1983; Komarudin, 1989), the possibility of reducing rural household’s consumption of biomass energy, fuelwood in particular, through increasing its efficiency is worth pursuing. The study by Argawal (1988) noted, however, that the introduction of more efficient stoves such as Singer and Lorena in developing countries was not successful in the past. Nonetheless, increasing the efficiency of stoves should be considered in dealing with rural energy use. Fuelwood for the major part of rural households is still a freely gathered energy source. Time devoted to collecting fuelwood might be a more appropriate indicator of the relative scarcity of fuelwood. The use of price ratio for all respondents in this study yielded a coefficient sign different than expected for fuelwood. This positive and significant coefficient of the price ratio might be caused by the aggregation of purchasers with non-purchasers. The household production model as suggested by Laarman (1987) and which is widely used in the agricultural sector for evaluating subsistence farmer behavior (Singh et al., 1986) may likely provide a richer explanation of rural household behavior related to fuelwood consumption. The household production model formally integrates the variable of time devoted to production within a household-firm in its structure. In future surveys, fuelwood collection time should be included. 128 Finally, further studies on demand as well as on supply are important to undertake. A repeated survey to obtain a series data is worth pursuing as it will permit evaluation of the dynamic aspects of rural energy consumption. Appendix 1. FPRI ENERGY SURVEY 1983 The present study is based on data collected through a rural energy survey in Central Java Region conducted by the Forest Products Research Institute (FPRI), Indonesia in 1983. A description of the survey follows. Background and objectives Fuelwood is still a primary source of rural energy in Java. The levels of fuelwood consumption per capita ranges from 0.5 cubic meters to more than 2 cubic meters per capita per year (Hadi et al., 1979; and Sumarna and Sudiono, 1973). Several studies have reported that the growing demand for fuelwood in some areas, especially on the island of Java, are beyond supply capability. These fuelwood shortages have been associated with forest resources depletion. Lack of reliable data and information regarding rural energy patterns is a hindrance to development of a sound rural energy program. To fulfill the needs for more information, a study on fuelwood consumption by rural households in Java was conducted. There were three survey objectives. First was to estimate the levels of fuelwood and other types of energy consumed by rural household. The second was to analyze the effects of economic and demographic factors on energy 129 130 consumption. The final objective was to identify the relative contribution of each type of energy in meeting rural energy demands. Survey implementations Field work was carried out in January and February of 1983. As many as 24 field workers were trained prior to the survey execution. In the actual field work, enumerators were supervised by the field coordinators. Each district was controlled by one field coordinator. Survey locations Surveys were conducted in Central Java Region. This region is comprised of two provinces, Central Java and Yogyakarta. The survey covered six districts, namely Banyumas, Blora, Pemalang, Purworejo, Temanggung, and Bantul in the province of Jogyakarta. Two counties in each district were randomly selected, and in each selected county three villages were selected at random. The selection of districts, counties and villages was done prior to the field work. In each village, 25 households were randomly selected to be survey respondents. Therefore, the total number of households selected in this survey for the entire region was 750 households. The selection of households was undertaken in the field based on the household list provided by the 131 provincial village offices. Households were selected at random. Samples results selected is presented in Table 4 page 43. Energy use measurements Energy use levels both for cooking and lighting by household samples were recorded through direct measurement during three consecutive days of visiting. From the three visits, surveyors recorded two observations, each observation represented the energy use per day by each respective household. The calculated daily energy consumption level was the average of these two observations. Questionnaires used in the survey are presented in the questionnaire section. Energy consumed was recorded separately for each type of energy use (e.g., energy for cooking and type of fuel). The measurement unit used for fuelwood and other biomass energies was a unit of weight (in kilograms) and for kerosene it was a unit of volume (in litres). Fuelwood and biomass energy consumed were weighed and recorded separately. This separation was based on the evidence that the heat content of a fuelwood species, on the average, is higher than that of other biomass energies such as agricultural wastes (Direktorat Jenderal Ketenagaan, 1982). Heat content of fuelwood and other biomass energies depends among other factors on the wood species and its 132 moisture content. For a given wood species, the drier the wood is, the higher the heat content. Unfortunately, the moisture content of wood and biomass energy used by households was not measured during the survey. In order to reduce variation of calorific values among wood and other biomass substances, respondents were asked to use ’ready to burn’ fuelwood. By this, it is meant that the varieties of fuelwood used by households were similar in the degree of their air-dry condition which decreased the variation in calorific value between different wood species. Economic and demographic factors Important economic factors in a demand study are income and price of goods. In this survey income was defined as the annual total income received by a household. It consists of income earned by the household head and by other household members for a common use. Regularly received monetary additions from other family members or relatives who did not live in the same household were added to the household’s total income. Income from agricultural production was calculated based on revenues received minus costs required in production activities. Productivity of agricultural land was estimated based on data published by The Biro Pusat Statistik (Central Bureau of Statistics). Cost items were based on respondents’ recollections. 133 Prices of fuelwood and kerosene recorded were based on price levels prevailing in the market place in each corresponding village. Price information was mostly obtained from respondents, fuelwood traders and kerosene retailers. Both income and prices were measured in Rupiah (Rps), the Indonesian currency. Recorded demographic factors included information regarding the head of a household such as age, education, and occupation: further information about a household included that of household composition, size, education of family members and land ownership. Educational background was recorded in categories such as; (1) an uneducated, (2) graduated from an elementary school, (3) graduated from a junior high school, (4) graduated from a senior high school, and (5) graduated from a college or university. Survey limitations To derive a useful reference from a sample using statistical techniques, reliable data are needed. Unfortunately, not all questionnaires collected had accurate information. Hence, to select reliable questionnaires, data screening processes were carried out. As a result of data screening, the number of questionnaires considered as reliable was 732 out of 750 Questionnaires. Only 18 questionnaires were dropped for 134 various reasons; some were incomplete, inconsistent, or not answered (Buharman, 1988). Being ’incomplete’ meant important data was not recorded. For example, only one of two observations which were supposed to have been reported was recorded. ’Inconsistency’ meant data recorded was vague and enumerator could not provide a reasonable explanation. Non-response was simply the result of failure to interview a potential respondent. For example, a household sample, for some reason or other, was not available for interviewing during the survey period. Data on distance to fuelwood sources was actually intended to be used as a proxy for a relative scarcity of fuelwood, since the largest part of the fuelwood is freely collected. However, the distance to a fuelwood source had been largely misinterpreted by the field workers due to unclear explanations during survey training. Instead of recording the distance to fuelwood source, enumerators had recorded the distance to the particular energy source (biomass, kerosene) which the respective household happened to be using. As a result, the available data on distance does not reflect the data to a fuelwood source in particular. As a result, the distance variable cannot be used to test the hypothesis that fuelwood use will decrease as the distance to wood sources increases. 135 Available data summaries contain other information concerning household samples such as household income, price of energy, age and education of the household head, family size, and land possession. These data summaries are used in this present energy study. Variables summary The mean and standard deviations of economic factors such as income and energy prices and demographic factors such as household size, land ownership, and age of the head of household are presented in Table A1. The distribution of household samples based on the education of the head of family is presented in Table A2. Distance to fuelwood and agricultural waste sources is presented in Table A3. 136 Table A1. Summary of variables for the region (N = 732) Variable Unit Mean Std.Dev. Household (1000Rps/mo) 50.71 66.89 income Family size (N) 5.28 2.16 Land (hectares) 0.55 0.81 ownership Age of head (years) 48.93 12.34 of household Fuelwood (Rps/kg) 13.05 3.61 price Kerosene (Rps/litres) 131.08 9.24 price Table A2. Households samples by education (N=732) Education level N % Without a formal 190 25.9 education Elementary 429 58.6 school Middle school 80 10.9 High school 31 4.2 College or 2 0.4 university 137 Table A3. Distance to energy source Energy type Mean Sd Max. Min. Agricultural wastes 0.17 0.27 1.20 0.00 Fuelwood 0.75 1.68 17.5 0.00 138 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DP. 01 : RESPONDENT’S IDENTITY Village : ........... Enumerator: .......... County :........... Date : ........... District : ........... No. Description Family’s head Husband/wife 1. N a m e 2. S e x 3. Age (years) 4. Education 5. Primary job 6. Secondary job 7. No. of family member 8. Land ownership (Ha) a. Dryfield b. Ricefield c. Homegarden d. Other garden c. Others Note: 139 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DP. 02 : HOUSEHOLD MEMBERS Village : ........... Enumerator: ........ County : ........... Date ........ District : O O O O O C O O O O O No. N a m e Status Sex Age Educ. Occupation Note: 140 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DP. 03 : FUELWOOD SOURCE Village : ................ Enumerator: ............ County : ................ Date : ............ District : ................ No. Description Energy Note 1 2 3 1. Type of energy 2. Source of energy 3. Frequency of collection 4. Amount per collection 5. Location of source 6 Distance to source 7. Frequency of purchasing 8. Amount per purchasing 9. Price per unit (Rps) 141 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DP. 04 and 05: FUELWOOD CONSUMPTION Village ................ Enumerator: ............ County : ............... Date ° ............ District : ............... No. of Description Measurement Note observation Amount of stock .......... kg Before using .......... kg Amount added (+) .......... kg Amount taken (-) .......... kg After using .. ... ... kg Amount used .......... kg Cooking frequency between two visits Number of people eating between two visits: ....... 142 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DP. 06 AND 07 :KEROSENE CONSUMPTION Village ° ............ Enumerator: ............ County :... ............ Date : ............ District :............... No. of Type of Description Measurement Notes observation usage Cooking Stock amount . ........ kg Before using .......... kg Amount added .......... kg Amount taken .......... kg After using ............ kg Amount used .......... kg Frequency of cooking between two visits : ...... Number of people eating between two visits : ...... Lighting Before using ... ....... kg After using .......... kg Amount used .......... kg 143 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DP. 09 : INCOME END EXPENDITURE Village : ......... .. Enumerator: ........... County : ........... Date : ........... District : ........... Activities Amount Note (RPS-) Expend. 1. Consumption a. Food b. Clothes/housing c. Education d. Health care c. Others 2. Production costs (e.g. farm inputs) 3. Savings Total Income 1. Agricultural a. Rice b. Dryland c. Homegarden d. fish pond 2. Trader 3. Government employees/ army 4. Home industry 5. Other sources Total 144 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DP. 10 : FUELWOOD TRADER Village : ..... ......... Enumerator : ............. County :..... ......... Date :.... ......... District : .............. No. Description Note 1. Trader’s name 2. Address/location 3. Amount purchased 4. Price per unit 5. Frequency of purchasing 6. Source of fuelwood 7. Amount sold per day 8. Selling price 9. Type of consumer *) 10. Consumer origin **) Notes *) Household/home industry/rural industry **) Within village or outside vilage 145 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DS. 01 : LAND USE PATTERN Village : ............. Enumerator : ............ County : ............. Date : ............ District : O O O O O O O O O O O O O No. Land use Area Percentage Note (ha) (35) 1. Rice field a. irrigated b. non-irrigated 2. Dry-field 3. Homegarden 4. Estate 5. Forest land 6. Bare land 7. Lake/swamp 8. Fish ponds 9. Others Total 146 AGENCY FOR FORESTRY RESEARCH AND DEVELOPMENT FORM DS. 02 : TYPE OF OCCUPATION Village :..... ...... ... Enumerator : ........... County :. ............. Date : ........... District : .............. No. Description Number % Note 1. Farmers (owners) 2. Farm workers 3. Laborers 4. Private employee 5. Govt. employees/Army 6. Traders 7. Others Total Appendix 2. The probit pecisipn to use fuelwood Economic theory explains that consumption decisions among individual households are driven by a rational behavior assumption where individuals maximize utility subject to a budget constraint. Total utility gained by an individual is derived from a bundle of commodities that are actually chosen from the set of alternative bundles of commodities available. In explaining how utility is derived from consuming energy, first, it must be assumed that energy utility is weakly separable from the utility of other goods. Being weakly separable means that the demand for heating is independent of prices and quantities of non-heating items (Deaton and Muellbauer, 1980). Following Skog (1986), if the utility from heating for cooking and from non-heating products are written as U1(q1) and U2(q2), respectively, then total utility (U) can be written as: U = U (U1(Q1). U2(qz)) (l) where: U1(q1) = utility from heating for cooking and, U2(q2) utility from non-heating items. From the previous description of the energy user in the Central Java region it was clear that an individual household faces a range of alternative fuels from the least convenient form of energy ( e.g., agricultural wastes) to a 147 148 more convenient form of energy (e.g., kerosene). Different users will make decisions based on individual utility. By holding the assumption that an individual behaves rationally a household may choose one particular fuel or a combination of fuels that provides more utility. In other words, a decision to choose a particular type of energy contains an element of probability. To explain the probability of choosing a particular type of fuel, such as fuelwood, first it is assumed that the utility gained by a household from consuming this energy contains both fixed and random components (Skog, 1986). To simplify, let us assume that a household faces the use of two types of energy, fuelwood and non-fuelwood. The utility obtained from using fuelwood by household n is expressed as: U1n = fi1 + 81n (2) Similarly, the utility derived from consuming non-fuelwood by household n, is written as: UZn = 02 + ezn (3) where: 01 and 02 = representative consumer utility, and eln and e2n random components. 01 and fig, which are representative consumer utilities, can 149 be expressed in the following utility relation (Hardie and Hassan, 1986): Uin = ui (X'B), i = 1 or 2 (4) where, X is a vector of economic and demographic characteristics of a household. B is a vector of coefficients expressing relationship between utility and the associated characteristics. As a rational actor, a given household n will choose fuelwood if: Uln > U2n (5) or 61 + eln > 6'2 + 92n- (6) This inequality can be rewritten as follows: 51 - 02 > ’eln + ezn (7) Since eln and e2n are random variables, the probability that an individual household n will choose fuelwood may be written as: Pln = Prob ( Ul - U2 > "eln + e2n ) (8) To derive a probability function, equation (8) is rewritten as follows: Pln = Prob (—U1 -—U2 > -e1n + e2n ) = F(X’B) (9) Probability of household n choosing fuelwood, and 150 F = the cumulative distribution of (-e1n + €2n)- To develop an empirical model based on equation (8), an assumption about the distribution of (-e1n + e2n) must be made. In this study, we assume that both enl and en; are normally distributed, therefore (-e1n + 92n) also has a normal distribution and hence a probit function is developed. Skog (1986) noted a necessary condition for eln and e2n to be normally distributed in that all relevant variables must be included to represent terms U1 and U2 and a proper functional form must be chosen. Since we assume that eln and 92n are normally distributed, then, Pln (or to be more general P1) in equation (9) is represented by the following cumulative normal distribution: p . [___‘__.<-~2’2>d.\- (11) where: Pi = probability of a household to select energy i (i= 1, 2; 1 = fuelwood, 2 = non-fuelwood) Parameters of X’B are estimated by a maximum likelihood procedure. A micro computer version of Time Series Processor (TSP) which provides probit estimation used in the present study (Lilien, 1987). For Tobit estimation, TSP mainframe version 4.1 is used in this study. Appendix 3. Comparison between Probit Models with and without district dummy variables Variable Model I Model II C -16.3617 -12.6910 INCOME -0.0078** -0.0064** AGE -0.0028 -0.0031 EDUC -0.4320** -0.4449** SIZE 0.2116** 0.2124** LAND 0.1928* 0.2082** PRIWOO -0.0202 0.0324* PRIKER 0.1416** 0.1061** RBA 0.2132 - RBT -0.4985* - RBL 0.2652 - RPM 0.3667 - RPW -0.5881 - Log likelihood -213.4222 -222.4163 x2 = 2[—213.42- (-222.42)] = 17.98 2 _ X (5’ 01) — 15.09 2 _ X (5’ 05) — 11.07 Fail to reject that Region = 0 151 152 Appendix 2. (continued) Variable Model IV Model V C -9.7823 -14.1231 INCOME -0.0053** -0.0069** AGE -0.0033 -0.0032 EDUC - - SIZE 0.2094** 0.2124** LAND 0.1921* 0.1898* PRIWOO 0.0336* -0.1679 PRIKER 0.0967** 0.1412** PRIRAT - - NOED -2.1485 -2.5328 ELEM -2.5519 -2.9770 MIDE -3.5304 -3.9916 HIGH -3.2876 -3.6l69 RBA - 0.0870 RBT - ~0.6757** RBL ' - 0.2288 RPM - 0.3897 RPW - -0.6454 Log likelihood -215.8683 -205.0763 x2 = 21.57 x2 table (0.1)= 15.09 ll 0 Fail to reject that Region 153 Appendix 2. (continued) Variable Model VI Model VII C 2.7527 4.7197 INCOME -0.0048** -0.0061** AGE -0.0034 -0.0037 EDUC - - SIZE 0.2048** 0.2009** LAND 0.1783* 0.1596* PRIWOO - - PRIKER - - PRIRAT 2.1485 -6.4157 NOED -2.0080 -2.2380 ELEM -2.3511 -2.6618 MIDE -3.4480 -3.7398 HIGH -3.2169 -3.4465 RBA - -0.3838 RBT - -0.9309 RBL - -0.3893 RPM - -0.3125 RPW - -1.4312 Log likelihood -226.1167 -209.7044 x2 = 32.84 x2 table (0.1) = 15.09 Fail to reject that Region = 0 Appendix 2. (continued) 154 Variable Model VIII Model IX C 1.4298 -0.3l91 INCOME -0.0063** -0.0049** AGE -0.0032 -0.0033 EDUC - - SIZE 0.1869** 0.1929** LAND 0.1633* 0.1780* PRIWOO - - PRIKER - - PRIRAT -6.7945* 2.1652 NOED - - ELEM 0.7791** 0.4648** MIDE -0.3803 -0.2975 HIGH - - RBA -0.3783 - RBT -0.8358** - RBL -0.3205 - RPM -0.1960 - RPW -1.4064** - Log likelihood -213.5377 -229.1264 x2 = 31.16 x2 table (0.1) = 15.09 Fail to reject that Region 0 155 Appendix 2. (continued) Variable Model X Model XI C -12.1272** -17.4977 INCOME -0.0054** -0.0072** AGE -0.0030 -0.0028 EDUC - - SIZE 0.1948** 0.1978** LAND 0.1910* 0.1938* PRIWOO 0.0354* -0.0188 PRIKER 0.0902** 0.1404** PRIRAT - - NOED - - ELEM 0.7693** 0.6490* MIDE -0.2964 -0.4525 HIGH - - RBA - 0.0854 RBT - -0.5831** RBL - 0.3003 RBM - 0.4939 RPW - -0.6172 Log likelihood -219.1249 -208.8885 x2 = 20.46 x2 Table (0.01) = 15.09 Fail to reject that Region Appendix 4. The results of the probit estimations Type of Energy Variable Wastes Fuelwood Kerosene CONSTANT -4.9870** 1.2801* -0.8310 INCOME -0.0027'5 1 -0.0062** 0.0061** (0.0031) (0.0013)2 (0.0015) PRIRAT —6.7425 16.8107** (3.8776) (4.7115) AGE -0.0027 -0.0033 -0.0004 (0.0070) (0.0024) (0.0049) SIZE -0.0388 0.1841** -0.1803** (0.0492) (0.0393) (0.0578) LAND 0.6376** 0.1512 -0.3576** (0.1358) (0.0875) (0.1180) ELES 0.9607** 1.0458** -2.0669** (0.2673) (0.1731) (0.2310) RBA 3.6942** -0.3008 -0.5673* (1.5533) (0.2906) (0.3225) RBL -0.0705 -0.2836 -0.0278 (2.2476) (0.3092) (0.3251) RBT 4.1972** -0.7665** -1.5039** (1.5567) (0.2684) (0.4240) RPM -0.l618 -0.2151 0.0478 (2.1860) (0.3101) (0.3100) RPW 5.9537** -1.3773 —0.0975 (1.5660) (0.3104) (0.4201) Loglgkelihood= -162.29 -214.41 —103.57 x 24.56 84.85 45.26 ** I IH percent level, respectively. IN 156 and $ are significant at 1, 5 and 10 Values in parentheses are standard deviations. Appendix 5. Correlation matrix X1 X2 x3 x4 x5 X6 X7 X1 1 -0.035 0.383 0.460 0.425 0.171 -0.052 X2 1 -0.109 -0.085 0.036 -0.005 -0.053 X3 1 0.099 0.104 0.094 -0.159 X4 1 0.182 0.089 0.062 x5 1 -0.014 0.005 X5 1 -0.255 X7 1 X1 = households income X2 = age of the head of household X3 = education X4 = family size X5 = land ownership X5 = fuelwood price X7 = kerosene price 157 REFERENCES Anonymous. 1983. Energi dan Pemerataan (Energy and Equality). Direktorat Jendral Ketenagaan, Dept. Pertambangan dan Energy. Jakarta. Aldrich, J.H. and F.D. Nelson. 1984. Linear probability, logit, and probit models. (Quantitative applications in the social sciences Series.) Beverly Hills, CA: Sage Publications. 95 p. Argawal, B. 1988. "Why stoves are resisted". Unasylva 35 (140): 22-26. Arnold, J.E.M. and J. Jongma. 1978. "Fuelwood and Charcoal in developing countries". Unasylva 29 (118): 2-9. Bajracharya, D. 1983. "Fuel, Food or Forest? Dilemmas in a Nepali Village". World Development 11(12): 1057-1074. Barnes, C., J. Ensminger and P. O’Keefe. 1984. Wood Energy and Households: Perspective on Rural Kenya. Stockholm: Beijer Institute. lp: R. Hosier (1985b). Becker, G. S. 1965. "A Theory of The Allocation of Time". The Economic Journal 75 (229): 493-517. Bohi, D. R. 1981. Analyzing Demand Behavior. A Study of Energy Elastilities. Baltimore, MD: The John Hopkins University Press. 177 p. Bryant, R.R. 1986. "The 0.8. Residential Demand for Wood". The Energy Journal 7(7): 138-139. Biro Pusat Statistik. 1976. Pengeluaran Konsumsi. Survey Sosial Ekonomi Nasional. VUS 78-26. Jakarta, Indonesia. . 1984. Statistik Tahunan Indonesia. No. 03300.8513. Jakarta, Indonesia. Buharman. 1988. Personal communications. de Montalembert, M.R. and J. Clement. 1983. Fuelwood Supplies in the Developing Countries. FAO Forestry paper No. 42. FAO, Rome. 158 159 Deaton, A. and J. Muellbauer. 1980. Economic and Consumer Behavior. New York, NY: Cambridge University Press. 450 p. Dewees, P.A. 1989. "The woodfuel crisis, reconsidered: observations on the dynamic of abundance and scarcity". World Development 17(8):1159-1172. Dick, H. 1980. "The Oil Price Subsidy, Deforestation, and Equity". Bulletin pf Indonesian Economics Studies XVI (3): 32-60. Direktorat Jenderal Ketenagaan. 1982. Hasil Lokakarya Energy Pedesaan. Direktorat Jendral Ketenagaan, Dept. Pertambangan dan Energi, Jakarta. Donner, W. 1987. Land Use and Environment in Indonesia. Honolulu, Hawaii: University of Hawaii Press. 368 p. Dwiprabowo, H.; Roliadi, H. and Buharman. 1980. Pola Konsumsi Kayu Bakar san Bahan Bakar Lain Oleh Rumah Tangga dan Industri di Daerah Istimewa Aceh. Laporan BPHH NO. 155: 25-32. Eckholm, E.P. 1975. The other energy crisis: Fuelwood. World Watch Paper no. 1. Washington, D.C. 22 p. Eckholm, E.P. , G. Foley, G. Barnard, and L. Timberlake. 1984. Fuelwood: The energy crisis that won’t go away. Washington, DC: International Institute for environment and development. 107 p. FAO. 1977. Fourth World Food Survey. FAO Food and Nutrition ser. 10. lg : J.G. Laarman and M.K. Wohlgenant (1984). Foley, G. 1985. "Wood fuel and conventional fuel demand in the developing world". Ambio 15(4-5): 253-258. 1986. "The economic of fuelwood substitution". Unasylva 38 (151) : 12-20. and vanBuren, A. 1980. " Substitute for Wood". Unasylva 32 (130): 1-24. Gillis, M. 1980. Energy demand in Indonesia: Projection and Policies. Development discussion paper no. 92. Cambridge, MA: Harvard Institute for International Development. 160 Hadi, S.S.: Buharman: Hartoyo: and B. Purnama. 1979. Fuelwood and agricultural residues use in Indonesia. (in Indonesia). Working paper presented at the Energy Workshop. Indonesia National Committee, World Energy Conference (KNI-WEC). Jakarta, Indonesia. Hadi, S.S. 1982. A quantitative analysis of household energy consumption in rural West Java: with particular emphasis on socio-economic influence. Unpublished PhD dissertation, University of Wisconsin, Madison. 112 p. Hardie, I.W. and A.A. Hassan. 1986. An Econometric Analysis of Residential Demand for Fuelwood in the United States, 1980-1981. Forest Science 32(4): 1001 - 1015. Hosier, R. 1985a. "Household Energy Consumption in rural Kenya". Ambio 15 (4-5) : 225 - 227. 1985b. Energy Use in rural Kenya: Household Demand and Rural Transformation. Stckholm, Swedia: The Beijer Institute. 142 p. Hosier, R. and J. Dowd. 1987. "Household Fuel Choice in Zimbabwe: An emperical Test of Energy Ladder Hypothesis". Resources and Energy 9(4): 347-361. Hugart, D. 1979. Prospects for Traditional and Non- conventional Energy Source in Developing Countries. Staff Working Paper No. 346. Washington, DC: World Bank. 132 p. Judge, G.G. ; W.E. Griffiths : R.C. Hill; H. Lutkepohl and T.C. Lee. 1985. The theory and practice of econometrics. New York, NY: John Wiley and Sons. 1019 p. Kaplan, A. 1964. The Conduct of Inquiry. Scranton, PA: Chandler Publishing Company. 428 p. Komarudin, A. 1989. "Energy for Rural Development in Indonesia". Indonesian Journal of Tropical Agricultural 1(1):23-31. Laarman, J.G. 1987. Household Demand for Fuelwood. lg: M. Kallio, D.P. Dykstra and C.S. Binkley (Eds.), The Global Forest Sector: An Analytical Perspective. Chichester, U.K.: John Wiley and Sons. pp:335-367. 161 Larmaan, J.G. and Wohlgenant. 1984. "Fuelwood Consumption: A Cross-Country Comparison. Forest Science, 30(2) 383-392. Lancaster, K.J. 1966. "A New Approach to Consumer Theory". Journal Of Eglitical Economics (JPE), 74 (2): 132-157. Lipfert, F.W. and J.L. Dungan. 1983. "Residential Firewood Use in the United States". Science Vol. 219 (March): 1425-1426. Makhijani, A. and A. Poole. 1975. Energy and Agriculture in the Third World. Cambridge, MA: Ballinger Publishing Co. 160 p. McDonald, J.F. and R.A. Moffitt. 1980. "The uses of Tobit Analysis". The Review of Economics and Statistics 62 (2): 318-321. Nasendi, B.D. 1978. Analisa Konsumsi Sumber Daya Energy Pedesaan khususnya Kayu Bakar di Daerah Aliran Sungai Citanduy, Jawa Barat. Tesis Magister Sains. Sekolah Pasca Sarjana IPB, Bogor. Openshaw, K. 1978. "Woodfuel- A Time for Reassessment". Natutal Resources Forum (3):35-51. IQ: R. Hosier. 1985b. Pindyck, R.S. and D.L. Rubinfeld. 1981. Econometrics Models and Economic Forecasts. Second Edition. New York, NY: McGraw-Hill Book Co. 630 p. Singh, I.; L. Squire.; J. Strauss. 1986. Agricultural Households Models. Baltimore, MA: The John Hopkins University Press. 325 p. Skog, K.E. 1986. Factor Influencing County Level Household Fuelwood Use. Unpublished PhD Dissertation. Michigan State University. East Lansing, MI. Susastro, H. 1983. Rural Energy Survey in West Java. 13: Hadi Susastro et al. (eds.), Jakarta: Yayasan Proklamasi. Center for Strategic and International Studies (CSIS). pp: 237-283. Sumarna, K. and J. Sudiono. 1973. Fuelwood Consumption by Households, Indistry, and Railway Company in East Java (In Indonesia). FPRI Report no.8, Bogor, Indonesia. 162 Tobin, J. 1958. "Estimation of Relationship for Limited Dependent Variables". Econometrica (26): 24- 26. Wardle, P. and M. Palmieri. 1981. "What does Fuelwood really costs?" Unasylva 33(131): 20-23. Wiersum, K.F. 1981. Outline of the Agroforestry concept. 1p: VieWpoints on Agroforestry. K.F. Wiersum (ed.). Agricultural University Wageningen, Netherlands. 185 p. "11111111111111