Extractive Industries for Development Series #16 June 2010 54929 Expenditure of Low-Income Households on Energy Evidence from Africa and Asia Robert Bacon Soma Bhattacharya Masami Kojima World Bank Group's Oil, Gas, and Mining Policy Division Oil, Gas, Mining, and Chemicals Department A joint service of the World Bank and the International Finance Corporation The Oil, Gas, and Mining Policy Division series publishes reviews and analyses of sector experience from around the world as well as new findings from analytical work. It places particular emphasis on how the experience and knowledge gained relates to developing country policy makers, communities affected by extractive industries, extractive industry enterprises, and civil society organizations. We hope to see this series inform a wide range of interested parties on the opportunities as well as the risks presented by the sector. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors and should not be attributed in any manner to the World Bank or its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. Copyright ©2010 www.worldbank.org/ogmc (or /oil or /gas or /mining) www.ifc.org/ogmc (or /oil or /gas or /mining) Cover photos: Oil rig, hematite-banded ironstone, LNG tanker Extractive Industries for Development Series #16 June 2010 Expenditure of Low-Income Households on Energy Evidence from Africa and Asia Robert Bacon Soma Bhattacharya Masami Kojima World Bank | Oil, Gas, and Mining Policy Division Working Paper Contents Acknowledgments ix Abbreviations x ExecutiveSummary 1 1 Background 10 Links between Oil and Energy Prices Paid by Households 12 Measuring Household Expenditure on Energy 13 Energy Ladder and the Energy Portfolio 15 Energy Prices 15 2 PreviousStudies 17 Islamic Republic of Iran (1999) 17 Ghana (1999) 19 India (1999-2000) 19 ESMAP Multicountry Study (Various Survey Dates) 20 Guatemala (2000) 22 Mali (2000-01) 23 IMF Multicountry Study (Various Dates) 23 Republic of Yemen (2003) 25 Luanda, Angola (2005) 25 Gabon (2005) 27 Madagascar (2005) 28 Common Findings 28 3 MethodologyandFindings 31 Measuring Total Household Expenditure 32 Constructing Expenditure Groups 33 Calculating Average Expenditure Shares and Uptake Rates 34 Countries Analyzed and Total Household Expenditures 36 Extractive Industries for Development Series iii Shares of Expenditure on Energy, Food, and Transport 39 Uptake of Different Energy Sources 52 Shares of Expenditure on Energy, Food, and Transport by User Households 60 Main Energy Source for Cooking 65 Main Energy Source for Lighting 70 Stylized Energy Facts 73 Implications for Universal Price Subsidies for Petroleum Products 75 4 Conclusions 80 Appendixes A Survey Data Sets and Estimation of Total Household Expenditure 87 B Additional Results 101 References 117 Figures E.1 Monthly Rural Household Expenditure on Biomass 4 E.2 Monthly Urban Household Expenditure on Biomass 5 3.1 Creating Quintiles: Example from Ghana 35 3.2 Shares of Total Household Expenditure on Various Energy Sources 40 3.3 Shares of Total Household Expenditure on Modern Energy, Food, and Transport 40 3.4 Share of Total Household Energy Expenditure Spent on Modern Energy by the Bottom and Top Quintiles 51 3.5 Monthly Rural Household Expenditure on Biomass 52 3.6 Monthly Urban Household Expenditure on Biomass 53 3.7 Rural Household Expenditure on Kerosene 54 3.8 Urban Household Expenditure on Kerosene 54 3.9 Main Cooking Fuel Across All Households 66 3.10 Main Cooking Fuel in Rural and Urban Areas 67 3.11 Main Lighting Source for All Households 71 3.12 Main Lighting Source in Rural and Urban Areas 71 3.13 Simulated Flat-Rate Universal Price Subsidy: 78 3.14 Simulated Flat-Rate Universal Price Subsidy: E 78 iv Expenditure of Low-Income Households on Energy 4.1 Use of Biomass as Main Cooking Fuel by Households in the Top Two Quintiles 83 Tables E.1 Survey Countries 2 E.2 Shares of Household Expenditure on Various Energy Sources, Food, and Transport: All Households (%) 7 2.1 Shares of Total Household Expenditure on Various Energy Sources in the Islamic Republic of Iran, 1999 (%) 18 2.2 Shares of Total Household Expenditure on Petroleum Products in Ghana, 1999 (%) 19 2.3 Shares of Total Household Expenditure on Cooking Fuels in India, 1999­2000 (%) 20 2.4 Shares of Total Household Expenditure on Energy for Cooking and Lighting in Various Countries (%) 21 2.5 Shares of Total Household Expenditure on All Energy Sources in Guatemala, 2000 (%) 22 2.6 Shares of Total Household Expenditure on Various Energy Sources in Mali, 2000­01 (%) 23 2.7 Shares of Total Household Expenditure on Petroleum Products in Bolivia, 2000 (%) 24 2.8 Shares of Total Household Expenditure on Various Energy Sources in Jordan, 2002­03 (%) 24 2.9 Shares of Total Household Expenditure on Various Energy Sources in Sri Lanka, 1999 (%) 24 2.10 Shares of Total Household Expenditure on Various Energy Sources in the Republic of Yemen, 2003 (%) 26 2.11 Shares of Total Household Expenditure on Various Energy Sources and Public Transport in Luanda, Angola, 2005 (%) 27 2.12 Shares of Total Per Capita Expenditure on Petroleum Products in Gabon, 2005 (%) 27 2.13 Shares of Total Household Expenditure on Various Energy Sources in Madagascar, 2005 (%) 28 2.14 Patterns of Energy Use Based on Selected Surveys 30 3.1 Average Annual Per Capita Expenditure and Urbanization in Sample Countries 37 3.2 Average Monthly Total Household Expenditure by Quintile (2005 $ at PPP) 38 Extractive Industries for Development Series v 3.3 Shares of Rural Household Expenditure on Various Energy Sources, Food, and Transport: All Households (%) 41 3.4 Shares of Urban Household Expenditure on Various Energy Sources, Food, and Transport: All Households (%) 42 3.5 Shares of Rural Household Expenditure on Various Energy Sources, Food, and Transport by Quintile: All Households (%) 44 3.6 Shares of Urban Household Expenditure on Various Energy Sources, Food, and Transport by Quintile: All Households (%) 46 3.7 Shares of Expenditure on Various Energy Sources, Food, and Transport for Quintiles at Similar Total Household Expenditure Levels (%) 50 3.8 Percentage of Rural Households Consuming Various Energy Sources, Food, and Transport 56 3.9 Percentage of Urban Households Consuming Various Energy Sources, Food, and Transport 58 3.10 Shares of Rural Household Expenditure on Various Energy Sources, Food, and Transport, by Quintile: User Households (%) 61 3.11 Shares of Urban Household Expenditure on Various Energy Sources, Food, and Transport, by Quintile: User Households (%) 63 3.12 Main Energy Source for Cooking: Percentage of Rural/ Urban Households Using That Source 68 3.13 Main Energy Source for Lighting in India, Pakistan, Uganda, and Vietnam: Percentage of Rural/Urban Households Using That Source 72 3.14 Main Energy Source for Lighting in Cambodia and Kenya: Percentage of Rural/Urban Households Using That Source 73 3.15 Patterns of Energy Use Based on Selected Surveys 74 A.1 Percentage of Households Using Biomass in Kenya 95 A.2 Percentage of Households Using Electricity in Kenya 95 B.1 Total Household Expenditure by Quintile Including and Excluding Nonpurchased Food: Cambodia (2005 $ at PPP) 102 B.2 Total Household Expenditure by Quintile Including and Excluding Nonpurchased Food: Kenya (2005 $ at PPP) 102 B.3 Total Household Expenditure by Quintile Including and Excluding Nonpurchased Food: Uganda (2005 $ at PPP) 102 vi Expenditure of Low-Income Households on Energy B.4 Total Household Expenditure by Quintile Including and Excluding Nonpurchased Food: Vietnam (2005 $ at PPP) 103 B.5 Shares of Household Expenditure on Various Energy Sources, Food, and Transport: All Households (%) 103 B.6 Shares of Household Expenditure on Various Energy Sources, Food, and Transport in Cambodia: All Households (%) 104 B.7 Shares of Household Expenditure on Various Energy Sources, Food, and Transport in Kenya: All Households (%) 104 B.8 Shares of Household Expenditure on Various Energy Sources, Food, and Transport in Uganda: All Households (%) 105 B.9 Shares of Household Expenditure on Various Energy Sources, Food, and Transport in Vietnam: All Households (%) 105 B.10 Shares of Household Expenditure on Various Energy Sources, Food, and Transport, by Quintile: All Households (%) 106 B.11 Percentage of All Households Consuming Various Energy Sources, Food, and Transport 108 B.12 Percentage of Households Consuming Various Energy Sources, Food, and Transport in Cambodia 110 B.13 Percentage of Households Consuming Various Energy Sources, Food, and Transport in Kenya 110 B.14 Percentage of Households Consuming Energy, Food, and Transport in Uganda 111 B.15 Percentage of Households Consuming Various Energy Sources, Food, and Transport in Vietnam 111 B.16 Shares of Expenditure on Various Energy Sources, Food, and Transport: User Households (%) 112 B.17 Main Energy Source for Cooking: Percentage of All Households Using That Source 114 B.18 Main Energy Source for Lighting in India, Pakistan, Uganda, and Vietnam: Percentage of All Households Using That Source 115 B.19 Main Energy Source for Lighting in Cambodia and Kenya: Percentage of All Households Using That Source 116 Extractive Industries for Development Series vii Acknowledgments This paper was prepared by Robert Bacon, Soma Bhattacharya, and Masami Kojima of the Oil, Gas, and Mining Policy Division. The paper benefited from helpful comments provided by Andrew Burns of the Development Prospects Group, Manohar Shamar of the East Asia and the Pacific Region, Gabriel Demombynes and Awa Seck of the Africa Region, Dan Biller of the South Asia Region, and Voravate Tuntivate of the Energy Sector Management Assistance Program--all of the World Bank--and Javier Arze del Granado of the International Monetary Fund. The authors are grateful to the following World Bank staff for provid- ing survey data and other related materials: · Aphichoke Kotikula for Bangladesh · Chorching Goh for Cambodia · Rinku Murgai for India · Vivi Alatas, Hendratno Tuhiman, and Lina Marliani for Indonesia · Johan A. Mistiaen for Kenya · Nobuo Yoshida and Tomoyuki Sho for Pakistan · Xubei Luo for Thailand · Rachel Sebudde for Uganda · Phuong Minh Le and Carolyn Turk for Vietnam Nita Congress edited and laid out the document, and Esther Petrilli- Massey of the Oil, Gas, and Mining Policy Division oversaw its produc- tion. Extractive Industries for Development Series ix Abbreviations HIES Household Income and Expenditure Survey IMF International Monetary Fund LPG liquefied petroleum gas PDS public distribution system PPP purchasing power parity SUSENAS National Socio-Economic Survey All dollar amounts are U.S. dollars. x Expenditure of Low-Income Households on Energy ExecutiveSummary Patterns of household energy use and expenditure have been the subject of a large number of studies. Household expenditures on energy--par- ticularly, how much the poor spend--have policy implications for several reasons. First, policies to mitigate or cope with energy price shocks are increasingly focusing on targeted support to low-income households as a way of limiting the fiscal cost of such policies while offering protec- tion to the most vulnerable members of society. Second, for governments looking to reform energy price subsidies, the effects on household wel- fare--especially effects on poor households--of price increases resulting from subsidy reduction/removal is an important policy consideration. But subsidies for liquid fuels targeting the poor are difficult to design and implement effectively, because liquid fuels tend to be used more by the rich than by the poor, and are also easy to transport (and hence to divert to nonpoor users). For this reason, there is a growing recognition of the need to move away from price subsidies for liquid fuels to alterna- tive forms of targeted assistance to compensate the poor for the adverse effects of higher fuel prices. Third, in areas where many households have not yet begun using modern commercial energy regularly, the amount they can afford to pay for such energy services is a relevant question. Quantifying expenditures on different types of energy at varying income levels provides a basis for addressing these questions. A crucial aspect of this quantification is the ability to examine energy use patterns as a function of income and to identify the poor. Nation- wide household expenditure surveys provide one of the best measures of poverty. These surveys assign a weight to each household so the results can be scaled up to the total population of the country. This, combined with detailed data on all important expenditures, enables a fairly accurate division of households into different expenditure groups. Where nation- ally administered household surveys collect disaggregated data on energy use, patterns of energy consumption and expenditure can be analyzed by income. Using data obtained in nationally administered household expendi- ture surveys, this study investigates the shares of household expenditure Extractive Industries for Development Series 1 devoted to energy at different income levels for Bangladesh, Cambo- dia, India, Indonesia, Kenya, Pakistan, Thailand, Uganda, and Vietnam (table E.1). The paper also examines expenditures on motorized passen- ger transport and food, two items for which the price of oil is an impor- tant component of their cost structure and which are consequential in the budget of poor households. The income levels are based on per cap- ita expenditure, with the country's population divided into five groups containing the same number of people in each. The study investigates the following questions: · What proportion of households use various energy sources--elec- tricity, petroleum products (kerosene, gasoline, diesel, and liquefied petroleum gas), natural gas, and biomass--and transport at different income levels and in rural and urban areas? · What are the main energy sources used for cooking and lighting? · What proportion of household income (using total household expen- diture as a proxy) is spent on petroleum products, on modern sources of energy (petroleum products, electricity, and natural gas), and on energy generally? How does spending on energy compare to what households spend on food and motorized transport? · How do the proportions vary across income levels and does the effect of higher energy prices bear more heavily on low- or high-income groups? For all forms of modern energy except kerosene, this study found that the proportions of households using different sources of energy at TableE.1 SurveyCountries Annualpercapita Urbanization Country Surveyyear expenditurea (%ofpopulation) Bangladesh 2005 657 25 Cambodia 2003­04 1,013 15 India 2004­05 707 25 Indonesia 2005 801 45 Kenya 2005­06 1,295 20 Pakistan 2004­05 1,005 32 Thailand 2006 3,073 30 Uganda 2005­06 926 15 Vietnam 2006 1,071 27 Sources: National household expenditure surveys and authors' calculations. a. Expenditures are in 2005 dollars at purchasing power parity. 2 Expenditure of Low-Income Households on Energy similar income levels were generally higher in urban than in rural areas.1 There were notable exceptions. For every quintile in Thailand, a larger percentage of rural households used gasoline and diesel than their urban counterparts. The same pattern was observed in Bangladesh, India, and Kenya for three or more quintiles. Because of the availability of natural gas in urban areas, a greater proportion of rural households used lique- fied petroleum gas (LPG) than their urban counterparts in every quin- tile in Pakistan. And while kerosene tended to be used more by rural households than urban in most countries, for all but one quintile in Indonesia--where kerosene was heavily subsidized and not rationed at the time--a larger percentage of urban households used kerosene. The uptake rate of biomass was nearly universal among rural households in many countries, as well as in urban Bangladesh, Cambodia, and Uganda. More than half the households in every quintile used motorized transport in Bangladesh and Pakistan, and more than half the households in four out of five urban quintiles used transport in India and Indonesia. Six surveys asked about the household's main energy source for light- ing. As expected, households that had access to electricity used it as the primary source of lighting. Access to electricity was very high in Viet- nam and essentially universal in Thailand. It was very low in Kenya and Uganda, especially among rural households, and low in rural Bangladesh and Cambodia. Those households without access to electricity used mostly kerosene for lighting, but firewood was important for lower quin- tiles in Kenya, and batteries were an important third main energy source in Cambodia. The expenditure shares of biomass and kerosene generally declined with rising quintile, but expenditures themselves rose for most countries among the bottom two to three quintiles. For kerosene, it would be rea- sonable to take expenditure levels as a first approximation of quantities consumed, particularly in countries with essentially pan-territorial pric- ing. For biomass, the relationship between quantity and value could be weak because of large geographic and temporal price variation as well as the difficulties of estimating equivalent market values of freely obtained biomass. Figures E.1 (rural) and E.2 (urban) show expenditures on bio- mass for quintiles converted to 2005 dollars at purchasing power parity. 1 Rural areas in household surveys may include areas that are more peri-urban than rural. For example, "rural users" of natural gas in Pakistan were most likely peri-urban residents. Extractive Industries for Development Series 3 FigureE.1 MonthlyRuralHouseholdExpenditureonBiomass 20 % share of household expenditure 20 Quintile 5 18 18 16 16 Quintile 1 14 14 12 12 2005 $ 10 10 8 8 6 6 4 4 2 2 0 0 an a a a m a sh nd a di si ny nd di na de st la ne In bo Ke ga ki ai et la do am Pa Th Vi U ng In C Ba Source: Authors' calculations. Note: In Kenya, 39 percent of rural households assigned an imputed value of zero to non- purchased biomass, including 68 percent of the bottom quintile, 48 percent of the second quintile, and 40 percent of the third quintile. It is striking that only in Thailand did the value of biomass consumed decline from the bottom to the second quintile in rural areas; the value in urban areas declined only in Indonesia, Pakistan, and Thailand. These findings might suggest that the quantity of biomass consumed in low- income countries tends to increase with rising income before falling. A related interesting finding is the extent to which the rich in the sur- vey countries were using biomass as their primary cooking fuel. These results should be interpreted with caution because households use a portfolio or fuel-stacking approach to cooking as income rises, and some households using two or more sources of energy might not have found it easy to name their primary cooking fuel. Only in high-income countries do households use a gas, electricity, or some combination of both for cooking; in developing countries, cooking with biomass is widespread. Using biomass may be time consuming, not only in terms of getting the fire started but also because, if not purchased, the fuel must be col- lected. Also, traditional use of biomass creates considerable indoor air pollution which is injurious to health. Despite these disadvantages, poor 4 Expenditure of Low-Income Households on Energy FigureE.2 MonthlyUrbanHouseholdExpenditureonBiomass 16 16 % share of household expenditure Quintile 5 14 14 12 12 Quintile 1 10 10 8 8 2005 $ 6 6 4 4 2 2 0 0 an a a a m a sh nd a di si ny nd di na de st la ne In bo Ke ga ki ai et la do am Pa Th Vi U ng In C Ba Source: Authors' calculations. Note: In Kenya, 21 percent of the bottom urban quintile and 7 percent of the next two urban quintiles assigned an imputed value of zero to nonpurchased biomass. households continue to use biomass if it carries low or zero monetary cost and they face cash constraints. What is surprising is that many rich households also use biomass--more than 90 percent of the fourth quin- tiles in several countries. Households do not abandon biomass use alto- gether for a variety of reasons, which include cost, the fact that modern fuel supplies are not always reliable or are time consuming to acquire where they live, and because of cooking practices and cultural prefer- ences. The persistent use of biomass even in urban areas and even as per capita expenditure reached upwards of $800, valued in 2005 dollars at purchasing power parity, shows that steps to move households away from biomass and toward modern energy sources will need to address a variety of concerns and problems. This is particularly the case for LPG, which, apart from electricity, is the cleanest modern fuel option in rural areas for cooking and heating. There are economies of scale in LPG deliv- ery, which also requires good road infrastructure. If LPG is delivered only once every so many days, a backup cylinder (at $20­30 each) becomes Extractive Industries for Development Series 5 essential. Delivery may also not be regular and reliable in a low-volume market, particularly if it is remote. The cost of cylinder management rises with declining cylinder size, but large cylinders mean large refill pay- ments, a problem for households with irregular cash income flow. These challenges all too often exist against the backdrop of more readily avail- able and much cheaper biomass, which is also suited for cooking tradi- tional meals. For policy to shift household fuel use from traditional biomass to cleaner cooking fuels, it would make sense to examine first how the urban rich could be persuaded to make this shift, because they are most likely to be able to afford it and have ready access to the LPG service infrastructure. If there are distortions in the market--a lack of competi- tion, an inadequate regulatory framework for the industry, poor enforce- ment of regulations, or any combination of these factors resulting in high prices, low quality of service, or both--the national energy ministry should take the lead in addressing them. The principal problems may, however, lie outside the energy sector: port congestion and slow customs clearance incurring high demurrage charges, bad road infrastructure, or the cost of doing business discouraging investment in bottling plants. Identifying and addressing these issues would require the involvement of other government ministries and agencies. Expenditures on different forms of energy can be averaged across all households or across user households. From the policy perspective, the average across all households is more important because the question for the government is how a particular policy might affect the entire popu- lation or the poor as a whole. When averaged across all households in the country, the share of total household expenditures spent on modern energy varied between 2 percent and 10 percent. When biomass was included to make up total energy, the expenditure share rose consider- ably in several countries, ranging from 7 percent to 12 percent, demon- strating the importance of traditional use of biomass. The expenditure share of purchased food was markedly higher than that of modern energy, by 20-fold or greater in Bangladesh and Cambodia. The expendi- ture share of motorized transport was about the same order of magnitude as that of petroleum products, and was lowest in Cambodia and Vietnam at less than 1 percent of total household expenditures (table E.2). The analysis of energy expenditures by quintile group gave useful information on the relative importance of energy to the poorest house- holds in each country. In Bangladesh, Cambodia, India, and Uganda, the share of expenditure on energy was greatest for the bottom quintile; 6 Expenditure of Low-Income Households on Energy TableE.2 SharesofHouseholdExpenditureonVariousEnergy Sources,Food,andTransport:AllHouseholds(%) Bangladesh Cambodia Indonesia Pakistan Vietnam Thailand Uganda Kenya India Expenditureitem Kerosene 1.0 1.0 1.5 2.3 2.1 0.3 0.0 1.5 0.3 LPG ND 0.2 1.1 0.2 0.1 0.2 0.6 ND 2.6 Gasoline and diesel 0.1 ND 0.8 1.0 0.2 1.0 6.1 0.2 3.1 Petroleumproducts 1.1 1.2 3.4 3.8 2.5 1.6 6.7 1.7 5.9 Electricity 1.1 0.8 2.4 3.4 0.2 3.8 3.1 0.4 3.0 Natural gas 0.3 NA NA 0.0 NA 0.6 0.0 NA NA Modernenergy 2.5 2.0 5.8 7.2 2.7 6.0 9.8 2.0 9.0 Biomass 4.7 4.8 5.4 1.6 1.4 3.1 0.6 4.4 3.1 Totalenergy 7.3 6.8 12 8.8 4.1 9.0 10 6.5 12 Purchased food 49 52 47 54 36 42 35 29 39 Nonpurchased food 12 18 7.9 7.8 21 10 8.7 24 12 Totalfood 61 70 55 62 57 52 44 53 51 Transport 2.5 0.2 2.4 2.4 3.2 3.0 1.7 2.0 0.6 Source: Authors' calculations based on surveys described in appendix A. Note: NA = fuel not available; ND = no question was asked concerning the fuel. Nonpur- chased items, including cashfree biomass, are included. this was also true of urban households in Indonesia and Thailand. Only Kenya, Pakistan, and Vietnam showed no such pattern, with the top quintile having the highest energy expenditure share in both rural and urban areas in Pakistan and Vietnam. The share of expenditure on mod- ern energy rose with income in all the Asian countries (except the top quintiles in Indonesia). There was no evidence that the share of modern energy increased with quintile level in Kenya and Uganda, although Kenya could exhibit such a trend if nonzero values are assigned to non- purchased biomass. There was no consistent relationship between the share of petroleum products and income. In India, Indonesia, Pakistan, Thailand, and Viet- nam, the share increased in both rural and urban areas with income. In Bangladesh and Cambodia, the share declined with the quintile group; this was also true in Uganda for all but the top quintile. For the bottom 40 percent, the spending on petroleum products was concentrated on kerosene, with the exception of Thailand, Vietnam, and urban Pakistan. Extractive Industries for Development Series 7 In those countries where the expenditure shares were already high in the middle of the last decade, subsequent oil price increases in 2007 and 2008 might have hit the poor hard. Universal price subsidies for petroleum products are common, and a number of governments that had earlier eliminated price subsidies reintroduced them as oil prices soared to historic heights in 2008. A recent estimate suggests that global pretax petroleum product subsidies increased from about $60 billion in 2003 to $520 billion by mid-2008 (Coady and others 2010). Simulation of universal flat-rate price subsi- dies for petroleum products using the data in this study suggests that such subsidies would be regressive for LPG, gasoline, and diesel in all countries where data are available, and for kerosene in half the countries. The rate of excluding the poor is very high for LPG, gasoline, and diesel, but low for kerosene in six countries where that product is widely used. These findings would suggest that universal subsidies for LPG, gasoline, and diesel would not help the poor when considering direct effects on household expenditures, but that a kerosene price subsidy could be pro- poor under certain circumstances. However, because kerosene is a nearly perfect substitute for diesel, when the price of subsidized kerosene is lower than that of diesel, the former is almost universally diverted to the automotive sector, benefiting businesses and higher-income households and potentially making the kerosene subsidy regressive even in countries where an analysis of household energy use might suggest it would be progressive. Combining the results of the above simulation with an examination of cash expenditures on food would further argue against price subsidies for petroleum products. The expenditure share of purchased food consti- tuted one-third or more of total household expenditures for every urban quintile and as much as 60 percent for the bottom four urban quintiles in Cambodia and the bottom two in Bangladesh. In rural areas, cash expenditures on food in every quintile comprised 50 percent or more of total household expenditures in Cambodia and Indonesia and one-third or more in Bangladesh, India, Pakistan, and Vietnam. In India, Indone- sia, Kenya, Pakistan, Thailand, and Vietnam, the ratio of expenditures on purchased food to those on petroleum products declined monotoni- cally with income in both rural and urban areas. If a 10 percent increase in petroleum product prices were to lead to a 1 percent increase in food prices, the indirect effect on food prices would be larger than the direct effect of higher oil prices for all quintiles except in India, Thailand, Vietnam, and urban Kenya. Excluding Thailand, the bottom quintile in 8 Expenditure of Low-Income Households on Energy both rural and urban areas in the remaining eight countries would be hit harder by higher food prices caused by higher oil prices than by the higher oil prices themselves. If that is the case, the policy response to help the poor cope with higher transportation fuel prices might more productively focus on assistance, ideally through targeted cash transfers, for food purchase--and more generally the basket of goods the poor con- sume--than on subsidizing fuel prices. Extractive Industries for Development Series 9 Chapter 1 Background Many studies have examined patterns of energy use and expenditure on energy by households in developing countries. These studies have vari- ously aimed to · obtain a detailed picture on patterns of access to, and use of, modern forms of energy · understand traditional use of biomass and its associated health effects · evaluate technical and policy options that can facilitate the transition to cleaner use of household energy · estimate the likely impact of higher energy prices on household wel- fare · assess the progressiveness, or its lack, of energy subsidy schemes Answering these questions entails examining households by their per capita income levels. Nationwide household expenditure surveys--such as the National Sample Survey of India and the Living Standards Mea- surement Study developed by the World Bank's Development Econom- ics Research Group--provide one of the best measures of poverty in a given country. These surveys assign a weight to each household, so that the weighted households can be extrapolated to the total population of the country. This methodology, taken together with detailed data on all important household expenditures over one year, enables a fairly accu- rate separation of households into different groups based on per capita expenditures, which are taken as a proxy for per capita income. Where nationally administered household surveys collect disaggregated data on energy use, patterns of energy consumption and expenditure can be ana- lyzed by expenditure group. National household expenditure surveys do not provide the level of detail on energy use of specialized energy surveys. They do not, for example, normally ask how many units of electricity the household con- sumes, or the distance to the closest shop selling liquefied petroleum gas (LPG). In this sense, their utility in studying energy use is limited. However, the coverage of these surveys--in terms of both geography 10 Expenditure of Low-Income Households on Energy and the breadth of expenditure categories--makes them uniquely valu- able. Specialized energy surveys seldom cover the country in a way that enables their results to be scaled up to the entire population, nor do they collect detailed data on total household expenditures. Specialized surveys can study low-, middle-, and high-income households, but such clas- sification tends to be more qualitative than that possible from nationally administered household expenditure surveys. For this reason, despite their limitations, household expenditure surveys have been used in many studies to understand the relationship between energy use and house- hold poverty. Household expenditures on energy--particularly, how much the poor spend--have policy implications for several reasons. First, policies to mitigate or cope with energy price shocks are increasingly focusing on targeted support to low-income households as a way of limiting the fis- cal cost of such policies while offering protection to the most vulnerable members of society. Second, for governments looking to reform energy price subsidies, the effects on household welfare, especially poor house- holds, of price increases resulting from subsidy reduction/removal is an important policy consideration. But subsidies for liquid fuels targeting the poor are difficult to design and implement effectively, because liquid fuels tend to be used more by the rich than by the poor, and are also easy to transport (and hence to divert to nonpoor users). For this rea- son, there is a growing recognition of the need to move away from price subsidies for liquid fuels to alternative forms of targeted assistance to compensate the poor for the adverse effects of higher fuel prices. Third, in areas where many households have not yet begun using modern com- mercial energy regularly, the amount they can afford to pay for such energy services is a relevant question. Quantifying expenditures on differ- ent types of energy at varying income levels provides a basis for address- ing these questions. This paper investigates the share of household expenditure devoted to energy at different income levels for a number of developing countries in Asia and Africa for which detailed data are available. It presents evidence relevant to the debate on the impact of higher energy prices--as a result of either higher world prices of fossil fuels or price subsidy reduction or removal--on the poor. By establishing the shares of household expendi- ture spent on different forms of energy, an estimate can be made of how changes in energy prices may affect household welfare directly. Further, by comparing shares of expenditure on energy at different income levels within a country, the extent to which the lowest-income households are Extractive Industries for Development Series 11 affected relative to higher-income households can be seen. This informa- tion can help inform the policy debate on the need to provide extra sup- port to these lowest-income households. The paper also examines expenditures on public transport and food, since the price of oil is an important component of their cost structure and both are significant items in the budget of poor households. Specifi- cally, public transport may be used extensively by the urban poor, and food typically constitutes the largest share of total expenditure among the poor. LinksbetweenOilandEnergyPricesPaidby Households Energy prices are location specific, some more than others. The cost and availability of electricity from hydropower depend largely on hydrology. Oil, in contrast, is the most internationally traded energy commodity and prices depend primarily on the cost of transport from oilfields or major refining centers. World oil prices are the most transparent and the most publicized: information on oil prices is readily available daily. Retail prices of petroleum products closely track those on the world market in some developing countries. Others, however, do not have completely liberalized markets, and international prices are not fully passed onto consumers, as documented by Bacon and Kojima (2006) and Kojima (2009a and b). Instead, subsidies, either implicit or explicit, keep domes- tic prices below the international equivalents. Households are affected by a change in oil prices through a variety of mechanisms, some direct and some indirect: · Households purchase a number of petroleum products directly, and petroleum product prices rise in step with crude oil prices. The impact of these price increases on a household depends on the importance of various fuels in the household budget and on the extent to which international prices have been passed through to domestic. · Households purchase other sources of energy besides petroleum prod- ucts, including electricity. The international prices of natural gas and coal have also risen in recent years, in part because oil, gas, and coal are substitutes over the medium to long run in many applications. The prices of electricity generated by fossil fuel combustion are determined in part by fuel costswhich are directly affected by higher oil prices if the fuel is a petroleum product, and indirectly affected in terms of higher coal and gas prices. If price increases of other fuels are passed through to households, they pay more for electricity, coal, and gas. 12 Expenditure of Low-Income Households on Energy Where petroleum products such as kerosene compete with biomass-- typically in urban markets--prices of biomass may also rise in tandem. · Petroleum products are an input to the production of other goods, many of which are purchased by households. In particular, the costs of transportation are directly affected by the price of diesel and gasoline. Higher domestic petroleum product prices lead to higher transporta- tion costs. These in turn are both a direct cost to households that use public transport, and an indirect one affecting the prices of goods transported. For example, food prices are affected by higher oil prices not only because food needs to be transported to markets, but also because diesel fuel is used to operate irrigation pumps and tractors, and fossil fuels are used in fertilizer manufacture. These direct and indirect links mean that an increase in international oil prices can have a series of effects on the prices faced by households. For a complete quantification of the changes of all prices faced by house- holds due to oil price changes, a detailed breakdown of the direct and indirect cost components of every item of household expenditure would be needed. To carry out such a calculation, an input-output table is required. Such information is available for relatively few countries, and where it is available, the table is sometimes based on data that are several years out of date. Partial equilibrium analysis using an input-output table provides upper bounds on the impact of higher energy prices on house- holds, because coefficients in the model are fixed and substitution is not modeled. A general equilibrium model considers market interactions and allows for relative changes in prices and resource flows, which affect macroeconomic and distributional outcomes. Partial and general equilibrium analyses are resource intensive, particu- larly the latter. Although limited in scope, useful information can still be gathered from analyzing household surveys alone: namely, current energy use patterns and expenditures, and first-order estimates of direct (and some indirect) effects of changing energy prices. Importantly, as mentioned above, household expenditure surveys enable separation of households into different income groups, so energy use can be analyzed as a function of per capita expenditure level, a common measure of poverty. This paper focuses on studies that report results by expenditure quintile or decile. MeasuringHouseholdExpenditureonEnergy Household expenditure surveys enable calculation of the shares of household expenditure spent on different forms of energy. These shares Extractive Industries for Development Series 13 can also be calculated for groups of households, allowing comparisons between households with high or low total expenditures per capita, or between rural and urban households. Because very few countries carry out such surveys every year, the information from the most recent avail- able survey may be a few years out of date. Depending on the country, expenditures are provided for a number of energy sources. This paper defines petroleum energy products as kero- sene, LPG, gasoline, and diesel. Other sources of energy are electricity, natural gas, coal, and firewood and other forms of biomass. Low-income households in some countries also use candles for lighting and car batteries as a source of power for electric appliances. Firewood and other forms of biomass may include charcoal, wood, straw, and dung. With the exception of charcoal, households often col- lect, rather than purchase, biomass, especially in rural areas. Surveys reporting these as expenditures have imputed a value to each freely obtained fuel. In some cases, such as in the National Sample Survey of India, the values are solicited from respondents by enumerators, injecting a large element of subjective judgment. Where there is a well-established market for firewood, as in many peri-urban and urban areas, the imputed values are more likely to reflect the market value of firewood in the com- munity. Household expenditures on energy can usefully be aggregated into three categories: · Expenditure on petroleum products (sum of expenditures on kero- sene, LPG, gasoline, and diesel) · Expenditure on modern forms of energy (sum of expenditures on kerosene, LPG, gasoline, diesel, electricity, and natural gas) · Expenditure on energy (sum of expenditures on kerosene, LPG, gaso- line, diesel, electricity, natural gas, coal, and firewood and other forms of biomass) Shares of total expenditure on different energy sources can be com- puted for each household. The shares can be averaged across all house- holds--whether they use the energy source or not--or across only those households that use the particular energy source. For the purpose of policy formulation, averaging across all households is more important than focusing only on users of a particular form of energy. This paper thus focuses primarily on shares averaged across all households in each income group under consideration. 14 Expenditure of Low-Income Households on Energy EnergyLadderandtheEnergyPortfolio Many authors have analyzed the use and choice of energy sources at the household level. Early studies on the use of fuels for the basic needs of lighting, cooking, and heating centered on the concept of an energy "lad- der." Later studies suggested that a portfolio, or fuel-stacking, approach is more realistic. The differences between the two approaches were well described in a study by ESMAP (2003c, pp. 11­12): The energy ladder model envisions a three-stage fuel switching pro- cess. The first stage is marked by universal reliance on biomass. In the second stage households move to "transition" fuels such as kero- sene, coal and charcoal in response to higher incomes and factors such as deforestation and urbanization. In the third phase house- holds switch to LPG, natural gas, or electricity... Yet the ladder image is perhaps unfortunate because it appears to imply that a move up to a new fuel is simultaneously a move away from fuels used hitherto... Evidence from a growing number of countries is showing multiple fuel use to be fairly common... The new perspective on household energy choice sees it as a portfo- lio choice more than as a ladder. A striking example of multiple fuel use was seen in a study conducted in rural Mexico. Households were found to move to multiple-fuel cook- ing with rising income in a bidirectional process--meaning that separate, coexisting factors simultaneously pushed households away from biomass and pulled them back. Even when households had been using LPG for many years, they rarely abandoned fuelwood use. Fuelwood savings from using LPG ranged from 35 percent on average in one village to as little as zero percent. Expenditures on fuelwood were higher in some cases in households using both fuelwood and LPG than in fuelwood-only house- holds (Masera, Saatkamp, and Kammen 2000). EnergyPrices The budget share allocated to a particular form of energy depends on a number of factors that underpin the energy portfolio or energy ladder explanations of household behavior. For a given household, the four important determinants of energy use are availability, income, the price of energy, and the prices of possible substitutes. These together influence the Extractive Industries for Development Series 15 quantity purchased and the expenditure on an energy source. Changes in expenditure share over time for a group can be related to changes in prices, quantities, the share of households using the given form of energy, and total household income (using total household expenditure as a proxy). For those countries where household expenditure surveys are available for more than one date, changes in expenditure shares can be related to changes in these four factors through a decomposition analysis (Bacon, Bhattacharya, and Kojima 2009). For those surveys that provide information on quantities purchased as well as expenditure on each energy source, it is possible to derive mea- sures of the prices (unit values) paid for each form of energy. Where such information is available, it is almost always for a fuel and not for electric- ity. Calculating prices paid allows examination of the hypotheses that (1) poor households tend to pay more per unit for a given fuel than do rich households, and (2) with the exception of biomass, rural households pay more than urban because of higher energy delivery costs. The first hypothesis is based on the possibility that there are econo- mies of scale in purchasing that credit-constrained low-income house- holds are unable to exploit. In addition, where fuels require transport from the retailer to the home, lower-income households may not be able to visit the lowest-cost suppliers because of the associated costs of transportation. The second hypothesis rests on the possibility that rural households may face higher retail prices because of the extra transport costs of supplying fuels to more remote areas. However, for firewood, where the main source of supply is the countryside, urban households are likely to face higher prices than rural. Where grid electricity is available, whether and how much to use depends in part on the pricing scheme in effect. For countries using rising block tariffs with lower prices charged for the first block(s) of purchases, the average price paid will be lower the smaller the amount purchased. However, where there is a fixed charge for access (such as a meter fee), or where connection charges have been recorded for house- holds connected during the survey period, the average prices paid rise with decreasing amount of electricity purchased. It is possible that the average price paid by households is U-shapedinitially declining with volume and then increasing. 16 Expenditure of Low-Income Households on Energy Chapter 2 PreviousStudies The World Bank and the International Monetary Fund (IMF), among other institutions, have published a number of studies that include detailed information on household energy consumption in develop- ing countries. The purposes of these studies differ considerably; con- sequently, close comparisons of the countries covered by these are not possible. Some studies focused on the choice of fuel for cooking, thus excluding consideration of the use of petroleum products for transpor- tation. Other studies looked primarily at the implications of subsidies given for petroleum products, excluding electricity and biomass from the analysis. Their inconsistencies notwithstanding, these studies can inform the present investigation. This chapter considers only those studies that analyze expenditures on energy by income group. In nearly all household survey analyses, expenditures are used as a proxy for income. Limiting the comparability of the results is that not all of the studies provide full details on how the quintile or decile groups are calculated--that is, whether they are based on per household expenditure or per capita expenditure rankings, and whether the groups contain equal numbers of people or of households. Where the method used in the study was indicated, this is noted here. Virtually none of the reports clarify what total household expenditures include, nor how the purchase of large durables is treated. As explained in chapter 1 and appendix A, their inclusion could change results signifi- cantly. A brief review of key studies is provided below; space constraints prevent presentation of all potentially relevant results. In the following discussion, the lowest quintile or decile represents the poorest house- holds, and the highest quintile or decile the richest. IslamicRepublicofIran(1999) The World Bank's study on the Islamic Republic of Iran (2003) provided information on household expenditures on the principal sources of energy for rural and urban quintiles. The country's population was first Extractive Industries for Development Series 17 divided into rural and urban areas before calculating quintiles separately for each area. The study used a household expenditure survey conducted in 1999 when world oil prices were at a historic low. Even at that time, the government of the Islamic Republic of Iran was heavily subsidiz- ing petroleum products, natural gas, and electricity; the study was a contribution to the analysis of the effects of reducing or removing these subsidies. The expenditure shares reflect the very low prices charged for energy, but still provide information on the patterns of expenditure of higher- and lower-income, as well as rural and urban, households (table 2.1). Table2.1 SharesofTotalHouseholdExpenditureonVariousEnergy SourcesintheIslamicRepublicofIran,1999(%) Petroleumproducts Elec- Natural Quintile Kerosene Gasoline Diesel Total tricity gas Total 1 1.6 0.2 0.1 1.9 1.1 0.4 3.3 2 1.3 0.2 0.0 1.5 1.1 0.3 2.9 3 1.0 0.3 0.1 1.4 0.9 0.4 2.7 Rural 4 0.9 0.3 0.1 1.3 0.9 0.4 2.6 5 0.4 0.2 0.1 0.7 0.4 0.2 1.3 All 0.7 0.2 0.1 1.0 0.7 0.3 2.0 1 0.7 0.2 0.0 0.9 1.2 0.2 2.3 2 0.6 0.2 0.0 0.8 1.2 0.2 2.1 Urban 3 0.5 0.3 0.0 0.8 1.2 0.2 2.2 4 0.4 0.3 0.1 0.8 1.2 0.2 2.1 5 0.1 0.3 0.1 0.5 0.8 0.1 1.2 All 0.3 0.3 0.0 0.6 1.0 0.1 1.7 Source: World Bank 2003. There is little difference among quintiles in terms of expenditure shares for a given energy source, with the exception of kerosene, which declined steadily as the quintile level increased. The top quintile also spent a much lower expenditure share on electricity. Expenditure on diesel was low for all groups, while the share for gasoline was similar across groups and between areas. The share of total spending on energy was higher for each rural quintile compared to the corresponding urban quintile; this difference was largely accounted for by the much greater share of expenditure on kerosene. 18 Expenditure of Low-Income Households on Energy Ghana(1999) Coady and Newhouse (2006) analyzed the potential effects of remov- ing fuel price subsidies in Ghana. Their study was based on a household expenditure survey for 1999 and presented information on expenditure shares by quintile for petroleum products (kerosene, LPG, and gasoline). The quintiles were based on national expenditures per equivalent adult (table 2.2). Rural and urban households were not disaggregated, and infor- mation on other forms of energy was not given in the study. The share of expenditure on petroleum products fell as income rose, but increased for the highest quintile. This pattern was explained largely by a steadily falling share for kerosene and rising shares for LPG and gasoline. In particular, the expenditure on gasoline increased markedly at the top quintile. Expendi- ture on LPG remained small throughout the income range. Table2.2 SharesofTotalHouseholdExpenditureonPetroleum ProductsinGhana,1999(%) Quintile Kerosene LPG Gasoline Total 1 5.9 0.0 0.1 6.0 2 4.1 0.0 0.1 4.2 3 3.4 0.0 0.2 3.6 4 2.4 0.1 0.2 2.7 5 1.6 0.2 2.1 3.9 All 3.5 0.1 0.6 4.2 Source: Coady and Newhouse 2006. India(1999-2000) ESMAP (2003a) reported on household energy consumption in India, focusing on those fuels used primarily for cooking--in particular, on the choice between biomass in its various forms and cleaner sources such as kerosene, LPG, or electricity. Deciles were based on per capita total expenditures, each containing the same number of households, and were calculated separately for the samples of rural and urban households. Shares of total expenditure allocated to clean energy sources are shown in table 2.3. The total share of expenditure on the three sources of energy rose throughout the expenditure range for rural households, and for urban households up to the sixth decile. Both rural and urban households allo- cated declining shares of total expenditure to kerosene and increasing Extractive Industries for Development Series 19 Table2.3 SharesofTotalHouseholdExpenditureonCookingFuels inIndia,1999­2000(%) Rural Kero- Elec- Urban Kero- Elec- decile sene LPG tricity decile sene LPG tricity 1 1.3 0.0 0.5 1 2.0 0.5 2.2 2 1.2 0.0 0.7 2 2.3 1.1 2.7 3 1.1 0.0 0.8 3 2.1 1.5 3.1 4 1.1 0.1 0.9 4 2.2 1.8 3.3 5 1.1 0.1 1.0 5 2.0 2.2 3.4 6 1.1 0.1 1.2 6 1.8 2.4 3.5 7 1.1 0.2 1.3 7 1.5 2.5 3.5 8 1.1 0.3 1.5 8 1.4 2.3 3.4 9 1.0 0.5 1.6 9 0.9 2.2 3.5 10 0.9 0.8 1.8 10 0.4 1.7 3.5 Source: ESMAP 2003a. shares to electricity as incomes rose. Rural households increased the share of expenditure on LPG, but urban households increased the share only up to the seventh decile, above which this energy source's share declined. The share of expenditure on each fuel was higher in urban deciles throughout the income range. ESMAPMulticountryStudy(VariousSurvey Dates) ESMAP (2003c) brought together results from several household surveys taken at different dates, including the India survey mentioned above. Because its focus was the choice and use of energy sources for nontrans- portation household activities, gasoline and diesel were not included. The countries and survey dates were South Africa (1993­94), Nepal (1995­96), Brazil (1996­97), Vietnam (1997­98), Nicaragua (1998), Ghana (1998­99), India (1999­2000), and Guatemala (2000). In each country, quintiles were based on per capita expenditure rankings with equal numbers of people in each quintile. Rural and urban quintiles were calculated separately rather than from aggregated countrywide house- hold ranked data, as this latter method would result in large differences between rural and urban households in total expenditure at the same quintile level. Information was analyzed for expenditures and use of ker- osene, LPG, electricity, coal and charcoal, and firewood. Histograms, but not numerical information, were given for rural and urban quintiles for 20 Expenditure of Low-Income Households on Energy shares of total household expenditure on each energy source. These indi- cate that, except in Brazil where the use of kerosene was negligible, the share of expenditure on kerosene declined with total expenditure level for rural households in all countries. Among urban households in India and Vietnam, the share initially increased before decreasing at the higher quintiles. The shares of expenditure on electricity and LPG increased with total expenditure in all countries and areas except urban Brazil and urban Guatemala, where they declined. The share of expenditure on pur- chased firewood declined in all urban areas but increased at low incomes for rural households before declining at the highest quintiles. An indication of the overall importance of energy in household bud- gets is given by the aggregate shares of expenditure on purchased energy for cooking and lighting (table 2.4). In all cases, the share of expenditure on energy is higher for urban households than for rural, except in South Africa and in India when collected firewood is taken into account. In sev- eral countries, urban households allocated more than 5 percent of total expenditure to energy. This is notable, given that all the surveys were conducted before 2001, when energy prices were much lower than they have been since 2004. In addition, if expenditures on gasoline and diesel were included, the shares would be higher, especially in urban areas. Table2.4 SharesofTotalHouseholdExpenditureonEnergyfor CookingandLightinginVariousCountries(%) Country Rural Urban Purchasedenergy Brazil (1996­97) 3.2 3.4 Ghana (1998­99) 3.1 5.0 Guatemala (2000) 6.2 6.7 India (1999­2000) 4.1 7.5 Nepal (1995­96) 2.1 6.0 Nicaragua (1998) 2.5 4.8 South Africa (1993­94) 5.9 3.7 Vietnam (1997­98) 2.9 5.6 Allenergyincludingcollectedfirewood India 8.3 8.0 Nepal 2.4 6.2 Vietnam 4.8 5.9 Source: ESMAP 2003c. Extractive Industries for Development Series 21 Guatemala(2000) ESMAP (2003b) analyzed the choice of household fuel for cooking and lighting in 2000 in Guatemala, especially the potential for switching from biomass to cleaner fuels. Accordingly, it did not cover expenditure on transportation fuels (gasoline and diesel) but did include informa- tion on purchased and collected firewood (evaluated at imputed prices). Rural and urban quintiles were based on national population quintile per capita expenditure groupings, thus standardizing differences in the total expenditures between the two areas. Average expenditures for kerosene, LPG, electricity, charcoal, and firewood were given for rural and urban quintiles for those households using a fuel, but shares of total expendi- ture were not given (table 2.5). Table2.5 SharesofTotalHouseholdExpenditureonAllEnergy SourcesinGuatemala,2000(%) Rural Expendituresforenergy Urban Expendituresforenergy quintile All Casha quintile All Casha 1 10 3.6 1 9.8 5.2 2 8.9 3.5 2 11 8.1 3 7.9 4.4 3 9.2 7.8 4 7.4 5.1 4 6.9 6.4 5 5.4 4.5 5 4.3 4.2 All 8.2 4.0 All 5.7 5.1 Source: ESMAP 2003b. a. Cash expenditures exclude imputed expenditures on freely acquired forms of energy such as firewood. Rural energy expenditures, including for collected firewood, declined with total expenditure but increased when only cash outlays were included. For urban households, both total energy expenditures and cash expenditures declined from the second quintile. Urban households spent more on energy (cash-only basis) than rural households at the same expenditure quintile. The study's detailed expenditure tables indicated that LPG and firewood were the most important sources of energy at low income levels for both rural and urban households, and that electricity expenditure increased more rapidly at higher income levels. Kerosene was relatively unimportant at all income levels for both rural and urban households. 22 Expenditure of Low-Income Households on Energy Mali(2000-01) Energy expenditure patterns in Mali were examined by Kpodar (2006) as part of an analysis of the distributional effects of oil price changes. The study, based on a household survey carried out in 2000-01, constructed national quintiles based on total expenditures per equivalent adult. The study provided information on energy expenditures (including for elec- tricity and charcoal) but did not distinguish between rural and urban expenditure patterns. Firewood is an important source of energy, but information was not available on its use; expenditure on LPG was unim- portant and not recorded. Table 2.6 indicates that the share of expenditure on energy rose with income level and that the shares of expenditure on all energy sources, except kerosene, also increased with income level. Diesel was unimport- ant throughout the income range, while the shares of spending on gaso- line, electricity, and charcoal all rose strongly at the highest quintile. Table2.6 SharesofTotalHouseholdExpenditureonVariousEnergy SourcesinMali,2000­01(%) Petroleumproducts Elec- Char- Quintile Kerosene Gasoline Diesel Total tricity coal Total 1 2.0 0.6 0.0 2.7 0.0 0.1 2.8 2 1.5 0.7 0.0 2.2 0.1 0.1 2.4 3 1.5 0.7 0.0 2.2 0.2 0.3 2.6 4 1.3 1.1 0.1 2.4 0.5 0.5 3.4 5 0.9 2.0 0.1 3.0 1.5 0.9 5.3 All 1.5 1.0 0.0 2.5 0.4 0.4 3.3 Source: Kpodar 2006. Note: Details may not sum to totals because of rounding. IMFMulticountryStudy(VariousDates) Coady and others (2006), in an IMF study, brought together informa- tion on patterns of household expenditure on petroleum products and energy for quintile groups in five countries to analyze the magnitude and distribution of fuel subsidies. Information on Ghana and Mali, two of the countries included, was taken from the sources described above. The other countries studied were Bolivia, Jordan, and Sri Lanka. For the latter two countries, information on expenditure on electricity as well as on petroleum products was included. None of the reported results Extractive Industries for Development Series 23 distinguished between rural and urban households, nor were data on firewood or biomass included. Tables 2.7­2.9 present the results for the additional countries. In all three, the share of expenditure on total energy (petroleum fuels only in Table2.7 SharesofTotalHouseholdExpenditureonPetroleum ProductsinBolivia,2000(%) Quintile LPG Transportationfuels Total 1 2.6 0.0 2.6 2 2.3 0.0 2.4 3 2.1 0.3 2.4 4 1.7 0.5 2.2 5 1.1 2.5 3.6 All 1.5 1.6 3.1 Source: Coady and others 2006. Note: Details may not sum to totals because of rounding. Table2.8 SharesofTotalHouseholdExpenditureonVariousEnergy SourcesinJordan,2002­03(%) Petroleumproducts Elec- Quintile Kerosene LPG Gasoline Diesel Total tricity Total 1 1.0 1.8 0.9 0.3 4.0 3.1 7.1 2 0.7 1.3 1.5 0.4 3.9 2.3 6.1 3 0.6 1.2 2.1 0.4 4.3 2.1 6.3 4 0.3 0.7 3.4 0.9 5.3 1.8 7.1 5 0.3 0.7 3.4 0.9 5.3 1.8 7.1 All 0.6 1.2 2.0 0.5 4.3 2.3 6.6 Source: Coady and others 2006. Table2.9 SharesofTotalHouseholdExpenditureonVariousEnergy SourcesinSriLanka,1999(%) Petroleumproducts Elec- Quintile Kerosene LPG Transportationfuels Total tricity Total 1 1.8 0.1 0.0 1.9 0.8 2.7 2 1.2 0.2 0.2 1.6 1.1 2.6 3 0.9 0.2 0.3 1.4 1.1 2.6 4 0.8 0.6 0.5 1.9 1.5 3.4 5 0.4 1.4 1.6 3.4 2.1 5.4 All 1.0 0.5 0.5 2.0 1.3 3.3 Source: Coady and others 2006. 24 Expenditure of Low-Income Households on Energy Bolivia) first declined but then rose with the level of total per capita expen- diture. Kerosene was not recorded in Bolivia, but in Jordan and Sri Lanka the share of expenditure on this fuel declined with increasing quintile. In all three countries, the share of total expenditure on transportation fuels (gasoline and diesel) increased at higher income levels. The shares of expenditure for LPG and electricity fell in Jordan but rose in Sri Lanka. RepublicofYemen(2003) ESMAP (2005), to examine household use of energy in the Republic of Yemen, collected survey information in 2003 on the expenditure and use of all sources of energy except gasoline. Expenditure shares were calcu- lated for each energy source for both rural and urban deciles as defined in national household income ranges; as a result, equivalent rural and urban deciles have similar average total household expenditure levels. When the survey data were collected, the government of the Republic of Yemen was following a policy of heavy price subsidization for petro- leum products, especially for LPG. This subsidization was in part meant to encourage a shift away from the use of firewood, which was widely used by households across all income levels. Data were collected for purchased firewood and charcoal as well as for other forms of biomass. Diesel was widely used, particularly in rural areas for agricultural pur- poses and for captive power generation where there was no grid electric- ity or the electricity supply was unreliable. The expenditure shares based on all households in the deciles are shown in table 2.10. In both rural and urban households, the share of expenditure on all forms of energy (excluding gasoline) declined by income level. For rural households, the shares of expenditure on kerosene, LPG, and biomass all declined with income, while the shares of expenditure on diesel and electricity increased. In urban households, the shares of expenditure on all sources of energy except diesel decreased at higher income levels. At lower deciles in both rural and urban areas, the total expenditure on energy accounted for very large shares of all household expenditure; urban house- holds dedicated a larger share to energy than did the equivalent rural decile even though average total household expenditures were similar. Luanda,Angola(2005) A poverty and social impact analysis was carried out by the World Bank (2005) in the Luanda province of Angola. This analysis concerned the effects of phasing out fuel and electricity price subsidies on households at different income levels. A relatively small household survey collected Extractive Industries for Development Series 25 Table2.10 SharesofTotalHouseholdExpenditureonVarious EnergySourcesintheRepublicofYemen,2003(%) Decile Kerosene LPG Diesel Electricity Biomass Allenergya 1 4.4 3.8 0.2 0.6 3.1 15 2 2.6 3.1 0.2 0.4 2.1 10 3 2.0 3.2 0.8 0.7 1.5 12 4 2.3 3.2 0.9 1.0 1.7 12 5 1.4 3.0 0.9 0.8 0.9 9.7 Rural 6 1.1 2.7 0.4 1.2 1.4 8.1 7 1.0 2.7 1.1 1.3 0.9 10 8 0.9 2.2 1.1 0.9 1.1 9.2 9 0.5 2.0 1.6 1.4 0.7 10 10 0.3 1.3 1.9 1.1 0.7 7.1 All 1.1 2.3 1.2 1.0 1.1 9.4 1 2.1 4.6 0.1 5.0 2.4 18 2 1.3 2.8 0.0 5.9 1.6 13 3 0.8 3.4 0.0 4.6 1.8 12 4 0.4 1.9 0.1 4.0 0.6 7.8 5 0.6 2.8 0.0 4.1 1.2 10 Urban 6 0.4 2.5 0.4 4.8 1.2 10 7 0.6 2.0 0.1 4.7 1.1 9.0 8 0.3 1.9 0.3 3.8 1.0 8.6 9 0.1 1.4 0.3 3.3 0.5 6.2 10 0.1 1.0 0.4 2.2 0.4 4.5 All 0.3 1.7 0.3 3.4 0.8 7.1 Source: ESMAP 2005. a. "All energy" consists of kerosene, LPG, diesel, electricity, biomass, candles, batteries, and maintenance and repair of self-generating units. expenditure information on the main commercial sources of energy but excluded biomass. The survey was carried out for rural and urban house- holds in Luanda and therefore did not represent expenditure patterns for the whole country. Quintiles were based on per capita expenditure, but results for the expenditure analysis did not separate rural and urban household expenditure patterns. The results for aggregate quintiles are shown in table 2.11. The share of total expenditure allocated to energy was very high for the lowest quintile but declined markedly at higher quintiles. LPG was the most important item of energy expenditure at all income levels. The shares of total expenditure for LPG, kerosene, electricity, and coal/charcoal all 26 Expenditure of Low-Income Households on Energy Table2.11 SharesofTotalHouseholdExpenditureonVarious EnergySourcesandPublicTransportinLuanda,Angola,2005(%) Petroleumproducts Coal/ Public Quin- Kero- Gas- Die- Elec- char- Total trans- tile sene LPG oline sel Total tricity coal energy port 1 2.3 13 1.9 0.2 17 4.3 0.9 23 3.8 2 1.7 5.8 1.5 0.2 9.2 2.8 0.9 13 6.0 3 1.1 4.3 1.3 0.2 6.9 2.9 0.5 10 6.1 4 1.1 3.5 1.5 0.3 6.3 3.1 0.5 10 5.5 5 0.4 2.9 2.0 0.5 5.8 2.2 0.6 8.6 4.3 All 1.3 5.7 1.6 0.3 9.0 3.0 0.7 13 5.1 Source: World Bank 2005. declined at higher quintiles, while the shares for gasoline and diesel rose from the third quintile onwards. The share of expenditure for public trans- port was also substantial, although it was lowest for the top quintile. Gabon(2005) El Said and Leigh (2006) looked at the magnitude of fuel price subsidies in Gabon using evidence drawn from a 2005 household survey. Shares of per capita expenditures allocated to kerosene, LPG, and transportation fuels (gasoline and diesel), were given for national decile groups. No separate information on other forms of energy expenditure or rural versus urban patterns were reported. Table 2.12 gives the results of these calculations. Table2.12 SharesofTotalPerCapitaExpenditureonPetroleum ProductsinGabon,2005(%) Decile Kerosene LPG Transportationfuels Total 1 1.3 1.2 0.0 2.6 2 0.8 1.6 0.1 2.5 3 0.6 1.6 0.1 2.3 4 0.5 1.8 0.3 2.5 5 0.4 1.6 0.2 2.2 6 0.4 1.8 0.3 2.4 7 0.3 1.7 0.5 2.5 8 0.2 1.6 1.1 2.9 9 0.2 1.5 1.4 3.0 10 0.1 1.1 2.5 3.7 All 0.5 1.5 0.6 2.7 Source: El Said and Leigh 2006. Extractive Industries for Development Series 27 The share of expenditure on transportation fuels rose, and that on ker- osene declined, with expenditure level. The share of expenditure on LPG initially increased but then declined over the top four deciles. The total share of expenditure on petroleum products remained fairly constant up to the eighth decile, above which it increased. The total share of petro- leum products is much lower than in some other countries. Madagascar(2005) Andriamihaja and Vecchi (2007) analyzed the potential impact of an increase in petroleum prices on household living standards in Madagas- car, using information from a 2005 household expenditure survey. The quintiles were based on national per capita annual expenditures for kero- sene, gasoline, diesel, and electricity. Information on differences between rural and urban households was presented only in diagrammatic form. Table 2.13 presents the results. The share of total expenditure on energy declined with quintile group until the highest quintile was reached. This effect was dominated by a falling share for kerosene and a rising share for electricity. The share for diesel was negligible at all income levels, while that for gasoline was negligible for all quintiles except the top level. Table2.13 SharesofTotalHouseholdExpenditureonVarious EnergySourcesinMadagascar,2005(%) Petroleumproducts Quintile Kerosene Gasoline Diesel Total Electricity Total 1 3.2 0.0 0.2 3.4 0.1 3.5 2 2.3 0.0 0.1 2.4 0.1 2.5 3 2.0 0.0 0.1 2.0 0.2 2.3 4 1.6 0.0 0.1 1.7 0.5 2.2 5 1.0 0.3 0.1 1.4 1.2 2.6 All 1.9 0.1 0.1 2.1 0.5 2.6 Source: Andriamihaja and Vecchi 2007. CommonFindings The surveys reviewed above were carried out in different years--corre- sponding to different international energy prices--and in countries with different energy pricing policies and varying levels of energy subsidies. Despite these differences, certain common features emerge. Classifying 28 Expenditure of Low-Income Households on Energy countries around a series of stylized facts helps highlight these similari- ties and differences. These simple criteria are used as a starting point for description of patterns of energy use; where fuller information is avail- able, as in chapter 3, more nuanced patterns of use may be discerned. The main variables considered are described below. · Was the share of total expenditure on energy greater than 5 percent for the lowest quintile?1 · Was the share of expenditure on energy generally greater for urban households than for rural households at the same quintile level? · Did the share of expenditure on kerosene tend to fall with rising income? · Did the share of expenditure on electricity tend to rise with income? · Did the share of expenditure on LPG tend to rise with income? · Did the share of expenditure on gasoline tend to rise with income? · Was the share of diesel less than 0.5 percent at all income levels? · Did the share of expenditure on petroleum products first decline and then rise with income (U-shaped response)? The tabulation of findings is given in table 2.14. The table indicates considerable variation among the countries sur- veyed. In about half the countries, the share allocated to energy was more than 5 percent for the lowest-income groups, indicating its direct impor- tance in household budgets. In the majority of countries where informa- tion was available, the share of expenditure on energy was higher for urban households in the same quintiles, but this finding should be inter- preted in light of the fact that quintiles were defined separately for the rural and urban groups in some countries and hence were not necessarily at similar per capita income levels. A common finding in all countries for which information was available is that the share of expenditure on kero- sene declined at higher income levels. The share of electricity at different income levels did not show a common tendency to increase or decrease, but was country specific. The share of expenditure on LPG decreased in more countries than those in which it increased, but information on LPG consumption was not available for many countries. The share of gasoline rose in all countries for which information was available, while the share of expenditure on diesel tended to be less than 0.5 percent at all income levels. The combination of a falling share of kerosene and a rising share 1 In all surveys, except for that in the Republic of Yemen, the figure for total energy excludes any imputed expenditure for firewood or other forms of biomass. Extractive Industries for Development Series 29 Table2.14 PatternsofEnergyUseBasedonSelectedSurveys U-shapedresponse Electricityrises Gasolinerises Kerosenefalls Diesel<0.5% Urban>rural Share>5% LPGrises Countryandsurveyyear Angola, 2005 Y -- Y N N Y Y Y Bolivia (petroleum products only), 2000 N -- -- -- N Y* -- Y Gabon (petroleum products only), 2005 -- -- Y -- -- Y* -- Y Ghana (petroleum products only), 1999 Y -- Y -- Y Y -- Y Guatemala, 2000 Y Y -- -- -- -- -- -- Iran, Islamic Rep., 1999 N N Y N -- Y Y N India (cooking and lighting), 1999­2000 -- Y Y Y Y -- -- -- Jordan, 2002­03 Y -- Y N N Y N Y Madagascar, 2005 N -- Y Y -- Y Y N Mali, 2000­01 N -- Y Y -- Y Y Y Sri Lanka, 1999 N -- Y Y Y Y* -- Y Yemen, Rep., 2003 Y N Y N N -- N Y Source: Authors' calculations. Note: Y = yes; N = no; -- = not available; * = information provided relates to both gasoline and diesel, which were not separated. Studies that did not provide sufficient information to complete at least two columns of the table were excluded. for gasoline did produce a U-shaped pattern of expenditure on petroleum products in most countries for which data were available. 30 Expenditure of Low-Income Households on Energy Chapter 3 MethodologyandFindings This study aimed to answer the following questions concerning the use of, and expenditures on, energy sources by households, and their impor- tance relative to household expenditures on food and transport, two items that are affected by higher energy prices: · What proportion of household income is spent on petroleum prod- ucts, on modern sources of energy, and on energy generally? How does spending on energy compare to what households spend on food and transport? · How do proportions vary across income levels, and is the effect of higher energy prices expected to weigh more heavily on low- or high- income groups? · Are there important differences in the patterns of expenditure on energy, food, and transport between rural and urban households at similar income levels? · What proportion of households use various energy sources at different income levels and in rural and urban areas? · What are the main energy sources used for cooking and lighting? To answer these questions, the study analyzed a number of house- hold expenditure surveys that provide similar coverage on the pattern of expenditures on energy sources, food, and transport. Production-related expenditures--such as diesel used to power irrigation pumps and the cost of transporting produce to markets--were excluded from the analy- sis. The proportion of household income spent on energy can be used to quantify the potential vulnerability of households to higher energy prices. Because households purchase energy items directly, their effects on household welfare are called direct effects. The proportion of house- hold income spent on food and transport can give an indication of indi- rect effects of higher energy prices. An input-output table may suggest, for example, that a 10 percent rise in oil prices will lead to a 3 percent rise in transport costs and a 2 percent rise in food prices. Potential increases Extractive Industries for Development Series 31 in the prices of these items can then be calculated for a given increase in oil prices and the impact on households in different income groups esti- mated. Such estimations based on an input-output table give an upper bound on the effects of higher oil prices. To take substitution and other effects into account, a general equilibrium model needs to be used. This study, however, stops at the question given in the first bullet above. To allow comparisons across surveys conducted in different countries, this study used the same methodology to move from original household expenditure data to an analysis of energy and other expenditure patterns. Notably, as described in the following sections, the study created a stan- dardized procedure for · defining the measure of total household expenditure to be used; · arranging the data in expenditure groups; · calculating average household expenditures on various energy sources by all households and average uptake rates (the percentage of house- holds using the energy source), and on food and transport. MeasuringTotalHouseholdExpenditure To determine the potential vulnerability of households to higher energy prices, expenditures on different sources of energy are compared to total household expenditure. The use of household expenditure rather than household income was determined by data availability, since many sur- veys provide evidence only on the former. The use of total household expenditure as a reference against which expenditure on fuels and electricity can be compared raises two further measurement problems. First, some recorded expenditure is actually based on consumption of nonpurchased goods and services. Households acquire certain goods without payment (for example, food and biomass), but household surveys generally collect such data and impute prices to the quantities acquired, thus increasing measured expenditure. In most cases, the dominant categories of imputed expenditure are food and housing. For biomass sources of energy, which can be collected rather than purchased, most surveys impute a value to the collected product. All these imputed expenditures are included in the present study in total household expenditure for those countries where the surveys have provided such information unless indicated otherwise. It is common for imputed values to be entered by the enumerators on the raw survey data sheets. The imputed values are intended to approximate local market values, but it is difficult to know how accurately the respondents and 32 Expenditure of Low-Income Households on Energy enumerators estimated the market values of nonpurchased goods and services. In the extreme case, no local market may exist--for example, all firewood in a community may be collected and not purchased. For this reason, the values of nonpurchased energy are expected to have the greatest uncertainties. The second issue in using total household expenditure as a proxy for disposable income is the purchase of high-cost durables in the year of the survey. Expensive durables that have a life of many years are likely to be financed out of savings. Including them in total household expen- diture--to represent the income level that determines the purchases of goods regularly acquired--may misrepresent household income. House- holds that purchase such a durable good in the year of the survey will appear to have a higher income than is normally the case and can be placed in a higher-income group than their long-term behavior would merit. They would be seen to be allocating a smaller share of total expen- diture on energy and other nondurable items than would have occurred had the survey been taken in a different year. To deal with this problem, expenditure on major durables is excluded from total household expen- ditures in this study. However, the classification used by household surveys in different countries does not permit a standardized definition of major durables; such items thus vary from country to country (see appendix A). ConstructingExpenditureGroups Because the primary purpose of this study is to quantify differences in expenditure patterns on energy, food, and transport across income levels, the method of grouping households by expenditure is critical. Poverty levels are conventionally analyzed by per capita income rather than household income, because the same household income with different household sizes clearly does not correspond to the same welfare level. Accordingly, households are divided into expenditure groups on a per capita basis, even though expenditure patterns are analyzed at a house- hold level. Expenditure surveys are carried out using households as the basic data unit with information included about the number of house- hold members, and these are used to construct such a distribution. The method of grouping used in this study is to rank households by their associated per capita total expenditure. The ranked households are then assigned to quintiles in order of increasing per capita expenditure so that each quintile contains an equal number of people rather than of Extractive Industries for Development Series 33 households, corresponding to 20 percent of the total surveyed popula- tion (figure 3.1a). Where this paper discusses consumption patterns for energy and other items at different income levels, the income levels refer to the quintiles computed on the basis of average per capita expendi- tures. In assigning households to quintiles, they are weighted by factors that ensure the sample represents the characteristics of the population as identified in a national census. These weights are coded in the responses of the household surveys. Since household size varies across the income distribution, the number of households per quintile will not be constant. More specifically, because poor families tend to be larger than rich, there are more households in upper quintiles using this approach. Where sepa- rate analysis for rural and urban households is reported, households from the nationally based quintiles were allocated to a rural or urban group. Hence the lowest-income rural quintile consists of those rural house- holds that are within the lowest national quintile; the lowest-income urban households are similarly derived (figure 3.1b). Note that rural areas in household surveys might have included areas more precisely considered peri-urban than rural. For example, rural users of natural gas in Pakistan were most likely peri-urban residents. This allocation method makes the per capita expenditure ranges at the same quintile for rural and urban groups similar, although not identi- cal. In contrast, the alternative method of dividing the population into rural and urban areas first and then creating quintiles in each region can make per capita expenditure in the same quintile group markedly differ- ent between rural and urban, introducing an additional complication in interpreting the results. CalculatingAverageExpenditureSharesand UptakeRates To calculate the average expenditure share of a particular item, this study averages the expenditure share for each household in the quintile by number of households--weighted for this purpose--in the quintile. This average is more important for policy purposes than an average taken across only those households that consume the particular good. When considering the impact of higher or lower prices, the government's first concern is the welfare of all households in a given group; a particularly important question is how the poor as a whole may be affected. For this reason, unless indicated otherwise, group averages are obtained by aver- 34 Expenditure of Low-Income Households on Energy Figure3.1 CreatingQuintiles:ExamplefromGhana a. Number of people by quintile Millions 18 16.4 16.4 16.4 16.4 16.4 16 15.2 14.5 13.5 14 12 10.7 10.2 10 8 6.2 5.7 6 4 2.9 1.2 1.9 2 0 Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 a. Number of households by quintile Thousands 4.4 5 4.0 4.1 3.7 4 3.1 3.3 3.3 3.3 2.7 2.7 3 1.7 2 1.3 1 0.4 0.6 0.2 0 Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Rural Urban Total Source: Authors' calculations. aging across all households, regardless of whether they consumed the good in question. There are two ways of computing the average share of total expendi- ture on an item. The first is the average for the group of each household's ratio of expenditure on the item to total expenditure (the democratic bud- get share); the second is the ratio of the average expenditure on the item of the group to its average total expenditure (the plutocratic budget share). This study uses the former method, because the variable of interest is usually the share of expenditure for households rather than the share of the group as a whole. Extractive Industries for Development Series 35 By identifying which households report zero expenditure on a par- ticular energy source, it is possible to define an uptake rate. Taking the ratio of the number of households using an energy source to the total number of households in that same group, the degree of uptake can be derived. This procedure was used to compute uptake rates for all forms of energy except biomass and electricity in Kenya, LPG in Thailand, and electricity in Uganda, as explained below and in appendix A. Taking the average expenditure on an energy source only for the subset of users within a group, a different share of expenditure on the energy source can be derived. The average share of expenditure by users will be higher than the average share for all households, and this difference will be larger the lower the uptake rate for the group. CountriesAnalyzedandTotalHousehold Expenditures The study focuses on a group of recent household expenditure surveys that provide extensive information on expenditures for various forms of energy. These surveys are for Bangladesh (2005), Cambodia (2003­04), India (2004­05), Indonesia (2005), Kenya (2005­06), Pakistan (2004­ 05), Thailand (2006), Uganda (2005­06), and Vietnam (2006). Details on the surveys are given in appendix A. These surveys, all conducted in 2003­06 and, with the exception of Thailand, for countries at fairly similar levels of per capita income, provide a basis for analyzing energy expenditure patterns under reasonably comparable external environ- ments. The surveys are mainly from Asian countries, where the wealth of material available permitted selection for almost identical time periods. The two African countries were included to provide a wider comparison from among readily available and comparable surveys. The original survey data were reported in current local currency prices. To provide a comparative evaluation of the results, average per capita expenditures are given in 2005 dollars valued at purchasing power parity (PPP), as shown in table 3.1. The table also provides compara- tive information on levels of average annual per capita expenditure from national accounts and the rate of urbanization in each country. The countries surveyed had fairly similar consumption levels valued at PPP, except Thailand, where the level was more than double that in Kenya, the next highest case. The expenditure surveys in some countries yield a different value for average consumer expenditure than that based on the national accounts because of the inclusion or omission of certain items in 36 Expenditure of Low-Income Households on Energy Table3.1 AverageAnnualPerCapitaExpenditureandUrbanization inSampleCountries Annualpercapitaexpenditure %ofhouse- Survey Fromnational holdslivingin Country year accounts Fromsurvey urbanareas Bangladesh 2005 706 657 25 Cambodiaa 2003­04 849 1,013 15 Cambodia b 2003­04 849 900 15 India 2004­05 1,170 707 27 Indonesia 2005 1,931 801 44 Kenya a 2005­06 938 1,295 25 Kenyab 2005­06 938 1,127 25 Pakistan 2004­05 1,479 1,005 32 Thailand 2006 3,773 3,073 32 Ugandaa 2005­06 572 926 17 Uganda b 2005­06 572 764 17 Vietnam a 2006 1,160 1,071 27 Vietnamb 2006 1,160 986 27 Source: National accounts data and exchange rates from World Bank 2009. Note: Expenditures are in 2005 dollars at PPP. a. Including nonpurchased food. b. Excluding nonpurchased food. the surveys; Bascand, Cope, and Ramsay (2006) note that such a dispar- ity is not unusual. Household expenditure surveys showed that urban households comprised less than half of all households in every country. They constituted 30 percent of households when averaged across all households in the study countries, ranging from 15 percent in Cambodia to 44 percent in Indonesia. The grouping of households into quintiles based on per capita expen- ditures, and the subsequent separation into rural and urban quintiles drawn from these national quintiles, also produces a distribution of total household expenditure by quintile. Mean monthly household expendi- ture, converted to 2005 dollars at PPP, for the different quintile groups are given in table 3.2. For Cambodia, Kenya, Uganda, and Vietnam, the value of nonpurchased food at imputed prices was sufficiently large to shift households from one quintile to another if it were excluded. For consistency, the tables in the main text are based on the inclusion of imputed food expenditure. The differences in total expenditure by quintile for these countries excluding the value of nonpurchased food Extractive Industries for Development Series 37 Table3.2 AverageMonthlyTotalHouseholdExpenditurebyQuintile (2005$atPPP) Bangladesh Cambodia Indonesia Pakistan Vietnam Thailand Uganda Quintile Kenya India 1 120 140 115 124 139 265 316 118 179 2 160 187 149 159 229 334 430 181 239 3 197 228 178 181 295 381 565 240 288 Rural 4 266 291 215 223 391 445 755 311 377 5 449 757 335 351 675 623 1,406 558 568 All 223 310 184 186 343 405 637 280 304 1 119 140 122 138 126 295 351 127 199 2 165 199 172 174 202 352 489 201 260 Urban 3 199 256 202 219 276 418 644 245 325 4 261 337 252 275 395 518 853 331 422 5 567 993 471 452 993 904 1,588 784 749 All 355 650 372 318 729 636 1,142 586 556 Source: Authors' calculations. are given in tables B.1­B.4 in appendix B. The survey in Vietnam did not include a question on imputed rent, so total household expenditures for that country are lower relative to the measure used in the other countries. Table 3.2 indicates that the households surveyed had an average urban monthly household expenditure (including nonpurchased food) of $550 or higher in Cambodia, Kenya, Pakistan, Thailand, Uganda, and Viet- nam; the average urban expenditure was lower than $400 in Bangladesh, India, and Indonesia. Thailand had the highest average rural expenditure (including nonpurchased food), followed by Pakistan, Kenya, Cambodia, and Vietnam. Excluding nonpurchased food reduced the average rural expenditure in Cambodia, Kenya, Uganda, and Vietnam by a substantial fraction, leaving Thailand and Pakistan distinctly higher than the other countries surveyed. In every country, the average expenditure for all urban households was higher than for rural. In Cambodia, India, Indonesia, Pakistan, Thai- land, Uganda, and Vietnam, the average expenditure of a given urban quintile was higher than that of its corresponding rural quintile. In Bangladesh, the first four rural and urban quintiles were similar to each other, with only the top urban quintile being well above the top rural 38 Expenditure of Low-Income Households on Energy quintile. Excluding the value of nonpurchased food reduced the average urban expenditure in Cambodia, Kenya, Uganda, and Vietnam by a small fraction. In Kenya, when the value of nonpurchased food was included, the bottom three rural quintiles had a higher average household expen- diture than the corresponding urban quintiles; this trend could not be discerned when nonpurchased food was excluded. SharesofExpenditureonEnergy,Food,and Transport Shares of total energy and of the individual sources of energy in the household budget are the starting point of this analysis. A detailed description of the fuels covered in the different questionnaires is given in appendix A, which also explains the grouping of fuels into the categories used here. Coverage varies from country to country; notably, questions were not asked about expenditure on LPG in Bangladesh and Uganda, and on gasoline and diesel in Cambodia. These omissions are likely to result in underestimation of expenditure on energy by higher-income households in those countries. Figure 3.2 shows expenditure shares for electricity, oil and gas (petro- leum products and natural gas), and biomass averaged across all house- holds. The expenditure share, inclusive of imputed values of nonpur- chased fuels, was highest for biomass in India, Cambodia, Bangladesh, and Uganda. As mentioned earlier, values of nonpurchased biomass fuels are likely to contain large uncertainties, and these findings should be interpreted with caution. In particular, 30 percent of all Kenyan households surveyed assigned a value of zero to nonpurchased biomass, thereby reducing expenditures on biomass and increasing the share of modern energy in total energy. The share for oil and gas was highest in Thailand, Vietnam, Indonesia, and India; because natural gas is not avail- able to households in these countries, the share represents that for petro- leum products. The expenditure share for electricity was highest only in Pakistan. Expenditures on transport were comparable to or lower than those in modern energy; as expected, expenditures on food were markedly higher (figure 3.3). Even in Thailand, which had the smallest difference between expenditures on modern energy and food, households on aver- age spent 3.5 times more on food than on modern energy. The difference widens to a factor of 4.4 when the imputed value of nonpurchased food is included. Extractive Industries for Development Series 39 Figure3.2 SharesofTotalHouseholdExpenditureonVarious EnergySources % of household expenditure 8 7 6.7 6 5.9 5.4 5 4.7 4.8 4.4 3.8 3.8 4 3.4 3.4 3.1 3.1 3.0 3.1 3 2.4 2.5 2.1 2 1.4 1.6 1.7 1.1 1.2 1.4 1 0.8 0.6 0.2 0.4 0 a h a a an nd m a di ya es si di nd na In st la ne en bo ad ga ki ai et do K am gl Pa Th Vi U an In C B Electricity Oil and gas Biomass Source: Authors' calculations. Figure3.3 SharesofTotalHouseholdExpenditureonModern Energy,Food,andTransport % of household expenditure 60 52 54 50 49 47 42 39 40 36 35 29 30 20 9.8 9.0 10 5.8 7.2 6.0 2.8 2.5 2.0 0.2 2.4 2.4 2.7 3.2 3.0 1.7 2.0 2.0 0.6 0 a sh a a an nd m a di ya si di nd de na In st la ne en bo ga ki ai a et do K am gl Pa Th Vi U an In C B Modern energy Transport Purchased food Source: Authors' calculations 40 Expenditure of Low-Income Households on Energy The shares of total expenditure devoted to the individual sources of energy, to subgroups of energy sources, and to total energy are presented for rural households in table 3.3 and for urban households in table 3.4. The tables also show the shares of total household expenditure spent on food (purchased and nonpurchased) and transport. They indicate that, in all countries and in both rural and urban areas, energy accounted for 4 percent to 14 percent of total household expenditure, ranging from 6 percent in urban Uganda and Kenya to 14 percent in urban Vietnam, and from 4 percent in rural Kenya to 12 percent in rural India. The Cam- bodian survey did not ask for information on expenditure on gasoline or diesel; thus, the extent to which households actually purchased these fuels as a total share of spending on energy was understated. In Kenya, 30 percent of households assigned an imputed value of zero to non- purchased biomass, again understating total energy spending. Since all Table3.3 SharesofRuralHouseholdExpenditureonVariousEnergy Sources,Food,andTransport:AllHouseholds(%) Bangladesh Cambodia Indonesia Pakistan Vietnam Thailand Uganda Kenya India Expenditureitem Kerosene 1.1 1.1 1.7 2.2 1.9 0.4 0.0 1.5 0.3 LPG ND 0.1 0.5 0.1 0.1 0.2 0.6 ND 1.9 Gasoline and diesel 0.1 ND 0.5 0.8 0.1 0.7 6.3 0.2 2.6 Petroleumproducts 1.2 1.2 2.7 3.3 2.1 1.3 7.0 1.7 4.7 Electricity 0.7 0.5 2.0 3.0 0.0 3.4 2.9 0.2 2.7 Natural gas 0.0 NA NA 0.0 NA 0.1 0.0 NA NA Modernenergy 2.0 1.7 4.7 6.3 2.2 4.8 9.9 1.9 7.4 Biomass 5.3 5.0 6.8 2.5 1.3 4.0 0.8 4.6 3.8 Totalenergy 7.3 6.8 12 8.8 3.5 8.8 11 6.5 11 Purchased food 49 52 49 55 35 42 35 27 38 Nonpurchased food 15 20 11 11 25 14 11 29 16 Totalfood 64 72 60 67 60 56 46 56 54 Transport 2.4 0.2 2.4 1.8 2.6 3.0 1.3 1.8 0.6 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel. Nonpur- chased items, including cashfree biomass, are included. In Vietnam, "gasoline and diesel" included only gasoline and lubricants; biomass included coal, as well as charcoal, firewood, and other forms of biomass. In Thailand, "natural gas" refers to compressed natural gas used as an automotive fuel. In Kenya, 39 percent of all rural households and 4 percent of urban households surveyed assigned an imputed value of zero to nonpurchased biomass. Extractive Industries for Development Series 41 Table3.4 SharesofUrbanHouseholdExpenditureonVarious EnergySources,Food,andTransport:AllHouseholds(%) Bangladesh Cambodia Indonesia Pakistan Vietnam Thailand Uganda Kenya India Expenditureitem Kerosene 0.6 0.5 1.2 2.4 2.7 0.1 0.0 1.1 0.3 LPG ND 0.8 2.5 0.3 0.4 0.3 0.5 ND 4.6 Gasoline and diesel 0.1 ND 1.5 1.3 0.5 1.7 5.6 0.4 4.3 Petroleumproducts 0.7 1.3 5.2 4.3 3.5 2.1 6.1 1.5 9.1 Electricity 2.2 2.8 3.6 4.0 0.7 4.8 3.5 1.1 4.0 Natural gas 1.0 NA NA 0.0 NA 1.6 0.0 NA NA Modernenergy 4.0 4.1 8.8 8.3 4.2 8.4 9.6 2.7 13 Biomass 3.1 3.2 1.6 0.5 1.8 1.1 0.2 3.6 1.2 Totalenergy 7.1 7.3 11 8.8 6.0 9.6 9.8 6.3 14 Purchased food 51 53 41 53 39 42 35 37 41 Nonpurchased food 3.9 6.6 0.8 3.2 6.9 2.0 4.5 6.5 2.6 Totalfood 55 60 42 56 46 44 39 44 44 Transport 2.8 0.1 2.5 3.1 4.9 3.1 2.5 2.5 0.5 Source: Authors' calculations. Note: See notes for table 3.3. surveys were carried out prior to 2007, these shares do not reflect high international oil prices in 2007­08. Results averaged across all house- holds are shown in table B.5. In Bangladesh, Cambodia, India, Indonesia, Pakistan, Thailand, and Uganda, the shares of expenditure on energy were similar between the rural and urban groups as a whole. In Kenya and Vietnam, the shares were substantially higher for urban households. Tables B.6­B.9 show that, when the value of nonpurchased food was excluded in Cambo- dia, Kenya, Uganda, and Vietnam, the share of expenditure on energy increased for both rural and urban households--substantially in the case of the former, indicating the importance of nonpurchased food. Modern energy--electricity, petroleum products, and natural gas--is more convenient and generates less indoor air pollution than firewood or other solid fuels. However, because modern energy is typically more expensive, it tends to be used less by lower-income households as well as by those in rural areas where biomass is more likely to be free or much cheaper. The household surveys analyzed in this study indicate that, with 42 Expenditure of Low-Income Households on Energy the exception of Uganda, modern energy comprised at least 50 percent of total energy expenditure for urban households on average. The share of modern energy in total energy consumed was substantially smaller in rural areas, and was less than 30 percent on average in Bangladesh and Cambodia. In Thailand, which had the highest household income level in the sample, modern energy's share of total energy was 98 percent in urban areas and 90 percent in rural. The share of petroleum products in total energy expenditure was smaller, but still greater than 50 percent in urban areas in India, Indo- nesia, Kenya, Thailand, and Vietnam. The proportion of total energy accounted for by petroleum fuels in rural areas was notably lower than in urban areas for India and Vietnam. The share of total household expenditure on electricity was higher in urban than in rural areas in all countries; with the exceptions of Bangla- desh and Uganda, the share of petroleum products was also greater in urban areas. Particularly in Thailand, but also in India, Indonesia, and Vietnam, the share of petroleum products in urban areas was significant, indicating the potential vulnerability of these households to large oil price increases. Among fuels, biomass was the most important in rural areas, with the exception of Kenya where kerosene's share of total was greater. This find- ing, which does not reflect the relative amounts of kerosene and biomass used by Kenyan households, is primarily due to nonpurchased biomass being given zero imputed values by many households. In Thailand, biomass was unimportant as an energy source in both rural and urban areas. LPG was not important for rural households except in Vietnam, but its consumption was substantial in urban areas in India and Vietnam. Kerosene was more important in rural than urban areas except in Kenya and was essentially not consumed by households in Thailand. Gasoline and diesel were more important in urban than in rural areas in all coun- tries and were of very considerable importance in Thailand, where they accounted for more than 5 percent of total expenditure in both rural and urban areas. The greater use of natural gas (where available) in urban areas reflects the fact that natural gas is usually not supplied to rural households. In Thailand, natural gas refers to compressed natural gas for use in vehicles, and its use at the time of the survey was limited. The expenditure share on all food, both purchased and nonpurchased, varied from 46 percent in Thailand to 72 percent in Cambodia for rural households, and from 39 percent in Thailand to 60 percent in Cambo- dia for urban households. The smallest expenditure share on purchased Extractive Industries for Development Series 43 food was 27 percent in rural Uganda; the highest share was 55 percent in rural Indonesia. The expenditure shares on purchased food were identi- cal across rural and urban households in Pakistan and Thailand, and very close in Bangladesh, Cambodia, and Indonesia. Purchased food com- prised at least 90 percent of the total value of food consumed by urban households in all countries except Cambodia and Uganda. In rural areas, purchased food comprised as little as 48 percent of total food consump- tion in Uganda, followed by 58 percent in Kenya, and 70 percent in Viet- nam. The highest share was 82 percent in India and Indonesia. Further insights into variations in the consumption of fuel, food, and transport can be obtained by examining expenditures by quintile. Expenditure shares for rural quintiles are shown in table 3.5 and for urban quintiles in table 3.6; shares for all households are shown in table B.10. In rural areas, the total energy share fell with rising quintile in Ban- gladesh, Cambodia, India, and Uganda; increased in Kenya, Pakistan, Thailand, and Vietnam; and increased through the bottom four quin- tiles in Indonesia. In urban areas, the share of spending on total energy decreased with rising quintile in Bangladesh, Cambodia, India, Indone- sia, and Uganda. It fell from the second to the fifth quintile in Kenya. In Table3.5 SharesofRuralHouseholdExpenditureonVariousEnergy Sources,Food,andTransportbyQuintile:AllHouseholds(%) Petro- Gaso- Nat- leum Energy Food Quin- Kero- line& Elec- ural Bio- prod- Trans- tile sene LPG diesel tricity gas mass ucts Modern Total P NP Total port 1 1.5 ND 0.0 0.3 0.0 6.2 1.5 1.8 8.0 56 13 70 1.8 Bangladesh 2 1.3 ND 0.0 0.4 0.0 5.8 1.3 1.8 7.6 51 17 68 2.2 3 1.1 ND 0.0 0.8 0.0 5.5 1.2 2.0 7.5 49 16 65 2.5 4 0.9 ND 0.1 1.1 0.1 4.7 1.0 2.2 6.9 45 15 60 2.9 5 0.6 ND 0.4 1.2 0.1 3.5 1.0 2.4 5.9 39 12 51 2.9 1 1.6 0.0 ND 0.1 NA 6.6 1.6 1.7 8.4 49 28 77 0.2 Cambodia 2 1.3 0.0 ND 0.1 NA 5.8 1.3 1.5 7.3 51 25 76 0.2 3 1.1 0.0 ND 0.3 NA 5.2 1.2 1.5 6.8 54 21 75 0.2 4 0.9 0.1 ND 0.6 NA 4.6 1.0 1.6 6.2 56 16 72 0.2 5 0.5 0.4 ND 1.3 NA 2.9 0.8 2.1 5.0 50 9.4 59 0.1 1 2.1 0.0 0.0 1.2 NA 8.8 2.2 3.4 13 55 8.8 63 1.5 2 1.9 0.1 0.1 1.7 NA 7.9 2.1 3.8 12 52 10 63 1.9 India 3 1.7 0.4 0.3 2.1 NA 6.9 2.4 4.4 12 49 12 61 2.4 4 1.4 0.9 0.7 2.4 NA 5.6 3.1 5.5 11 45 12 58 2.9 5 0.9 1.9 2.1 2.8 NA 3.1 4.9 7.8 11 41 8.9 50 3.8 44 Expenditure of Low-Income Households on Energy Petro- Gaso- Nat- leum Energy Food Quin- Kero- line& Elec- ural Bio- prod- Trans- tile sene LPG diesel tricity gas mass ucts Modern Total P NP Total port 1 2.1 0.0 0.2 2.4 0.0 3.7 2.4 4.8 8.5 55 15 70 1.2 Indonesia 2 2.3 0.0 0.6 2.9 0.0 2.9 3.1 5.9 8.8 56 11 68 1.7 3 2.3 0.1 0.9 3.4 0.0 2.2 3.5 6.9 9.1 56 11 66 2.2 4 2.4 0.1 1.3 3.5 0.0 1.6 4.1 7.6 9.2 56 8.4 64 2.2 5 1.8 0.3 1.6 3.0 0.0 1.0 4.3 7.3 8.3 52 8.7 61 2.2 1 2.1 0.0 0.0 0.0 NA 0.9 2.1 2.1 3.0 39 31 70 1.5 2 2.1 0.0 0.0 0.0 NA 1.3 2.1 2.1 3.4 37 29 66 2.2 Kenyaa 3 2.0 0.0 0.0 0.0 NA 1.3 2.1 2.1 3.4 35 27 61 2.5 4 1.8 0.0 0.2 0.1 NA 1.6 2.0 2.0 3.7 33 23 56 3.0 5 1.5 0.3 0.5 0.1 NA 1.5 2.3 2.4 3.9 30 16 46 3.8 1 0.5 0.1 0.1 3.2 0.1 4.6 0.7 4.0 8.6 48 12 59 2.7 2 0.5 0.1 0.3 3.2 0.1 4.4 0.9 4.2 8.6 44 14 58 2.9 Pakistan 3 0.4 0.2 0.4 3.3 0.1 4.1 1.0 4.4 8.6 42 15 57 3.0 4 0.4 0.3 0.9 3.5 0.1 3.5 1.6 5.3 8.8 39 16 55 3.1 5 0.3 0.5 1.7 3.6 0.1 3.3 2.5 6.3 9.6 35 15 50 3.4 1 0.0 0.3 4.6 3.2 0.0 1.6 5.0 8.1 9.8 37 17 55 0.8 2 0.0 0.7 5.8 3.1 0.0 1.1 6.4 9.6 11 37 13 50 1.0 Thailand 3 0.0 0.8 6.5 2.9 0.0 0.7 7.3 10 11 36 10 46 1.4 4 0.0 0.8 7.2 2.8 0.0 0.4 8.0 11 11 34 7 41 1.6 5 0.0 0.5 8.3 2.4 0.0 0.1 8.8 11 11 27 4 32 1.6 1 1.8 ND 0.0 0.1 NA 6.6 1.8 1.9 8.4 25 35 61 0.9 2 1.7 ND 0.0 0.1 NA 5.4 1.7 1.8 7.2 25 35 60 1.4 Uganda 3 1.5 ND 0.1 0.1 NA 4.5 1.6 1.8 6.3 26 33 59 1.7 4 1.4 ND 0.3 0.1 NA 3.9 1.6 1.8 5.7 27 29 56 2.2 5 1.2 ND 0.6 0.5 NA 2.7 1.8 2.2 4.9 30 17 47 2.5 1 0.4 0.2 1.1 2.4 NA 5.5 1.8 4.2 9.6 37 26 63 0.6 2 0.3 0.8 2.0 2.7 NA 4.6 3.0 5.7 10 38 20 58 0.6 Vietnam 3 0.3 1.8 2.8 2.7 NA 3.6 4.9 7.6 11 40 14 54 0.7 4 0.2 3.5 3.6 2.8 NA 2.6 7.4 10 13 39 9.8 49 0.6 5 0.2 4.3 4.4 2.8 NA 1.6 8.8 12 13 37 5.1 42 0.5 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; P = pur- chased; NP = nonpurchased. a. Nearly 40 percent of all rural households, and as much as 68 percent of the bottom quin- tile, assigned a value of zero to nonpurchased biomass (see appendix A). Extractive Industries for Development Series 45 Table3.6 SharesofUrbanHouseholdExpenditureonVariousEnergy Sources,Food,andTransportbyQuintile:AllHouseholds(%) Petro- Gaso- Nat- leum Energy Food Quin- Kero- line& Elec- ural Bio- prod- Trans- tile sene LPG diesel tricity gas mass ucts Modern Total P NP Total port 1 1.1 ND 0.0 1.1 0.1 5.9 1.1 2.3 8.3 62 6.5 68 1.7 Bangladesh 2 0.9 ND 0.0 1.6 0.3 5.0 0.9 2.8 7.7 60 6.6 66 2.0 3 0.7 ND 0.0 2.0 0.6 4.3 0.8 3.4 7.8 57 5.3 62 2.4 4 0.6 ND 0.0 2.6 0.9 3.3 0.6 4.2 7.5 55 4.0 59 2.8 5 0.3 ND 0.3 2.5 1.6 1.5 0.6 4.8 6.3 42 2.0 44 3.3 1 1.9 0.0 ND 0.3 NA 7.1 1.9 2.1 9.2 61 16 77 0.0 Cambodia 2 1.3 0.0 ND 1.1 NA 5.9 1.4 2.5 8.4 58 17 75 0.0 3 0.9 0.2 ND 1.4 NA 5.2 1.1 2.5 7.7 60 12 72 0.1 4 0.6 0.4 ND 2.3 NA 4.2 1.0 3.3 7.5 60 7.7 68 0.2 5 0.1 1.3 ND 3.9 NA 1.5 1.3 5.2 6.8 47 2.0 49 0.1 1 2.1 0.2 0.0 2.3 NA 5.7 2.3 4.6 11 51 0.9 52 1.6 2 1.9 0.7 0.1 2.6 NA 5.3 2.6 5.2 11 53 1.2 54 1.3 India 3 1.8 1.5 0.1 3.0 NA 3.9 3.5 6.5 11 50 1.3 51 1.8 4 1.8 2.6 0.5 3.4 NA 2.2 4.9 8.3 11 47 1.0 48 2.3 5 0.8 2.9 2.2 3.8 NA 0.5 6.0 9.8 10 37 0.5 37 2.9 1 3.5 0.0 0.1 3.5 0.0 2.1 3.6 7.2 9.4 59 5.1 64 1.2 Indonesia 2 3.4 0.0 0.5 4.0 0.0 1.1 4.1 8.1 9.2 58 4.0 62 2.0 3 3.1 0.1 0.8 3.9 0.0 0.7 4.3 8.2 8.9 56 5.0 61 2.8 4 2.7 0.2 1.2 4.1 0.0 0.3 4.5 8.7 8.9 55 3.1 58 3.4 5 1.6 0.6 1.8 3.9 0.0 0.1 4.4 8.4 8.5 48 2.0 50 3.7 1 3.1 0.0 0.0 0.0 NA 2.6 3.1 3.1 5.8 54 12 66 1.2 2 3.4 0.0 0.0 0.6 NA 4.3 3.4 4.0 8.3 48 10 58 2.0 Kenyaa 3 3.7 0.0 0.0 0.3 NA 3.2 3.8 4.1 7.3 48 7.6 56 3.3 4 3.1 0.0 0.0 0.6 NA 2.5 3.2 3.8 6.3 45 7.0 52 4.3 5 2.2 0.6 0.8 0.8 NA 1.0 3.6 4.4 5.4 35 6.2 41 5.7 1 0.1 0.1 0.2 4.7 1.2 3.2 0.4 6.4 9.6 51 3.0 54 2.3 2 0.2 0.2 0.3 5.0 1.4 2.3 0.6 7.0 9.3 49 2.8 52 2.6 Pakistan 3 0.1 0.3 0.4 4.9 1.5 1.7 0.8 7.2 8.9 46 2.6 49 2.8 4 0.1 0.3 0.8 5.0 1.7 1.1 1.3 8.0 9.1 44 2.1 46 3.4 5 0.0 0.3 3.2 4.7 1.6 0.3 3.5 9.8 10 36 1.4 38 3.2 1 0.0 0.5 4.7 3.8 0.0 1.3 5.2 9.0 10 41 12 53 0.8 2 0.0 0.8 5.2 3.8 0.0 0.7 5.9 9.7 10 41 8.6 49 1.3 Thailand 3 0.0 0.8 5.1 3.8 0.0 0.3 5.9 9.7 10 39 6.6 46 1.8 4 0.0 0.6 5.0 3.6 0.0 0.1 5.6 9.2 9.3 38 4.7 43 2.8 5 0.0 0.3 6.2 3.3 0.0 0.0 6.5 9.8 9.8 31 2.6 33 2.9 46 Expenditure of Low-Income Households on Energy Petro- Gaso- Nat- leum Energy Food Quin- Kero- line& Elec- ural Bio- prod- Trans- tile sene LPG diesel tricity gas mass ucts Modern Total P NP Total port 1 2.5 ND 0.0 0.0 NA 7.4 2.5 2.5 9.8 40 17 57 1.0 2 1.7 ND 0.0 0.2 NA 6.7 1.7 2.0 8.7 38 15 53 1.7 Uganda 3 1.8 ND 0.0 0.3 NA 5.6 1.9 2.2 7.8 41 10 51 1.5 4 1.3 ND 0.0 0.8 NA 4.7 1.3 2.1 6.8 42 6.5 49 2.3 5 0.9 ND 0.6 1.5 NA 2.5 1.5 3.0 5.5 35 5.2 41 2.7 1 0.6 1.2 1.6 3.3 NA 3.6 3.3 6.6 10 47 13 60 0.3 2 0.7 2.4 2.2 3.5 NA 3.1 5.3 8.8 12 50 6.2 56 0.6 Vietnam 3 0.6 4.2 3.4 3.7 NA 2.1 8.2 12 14 47 4.5 51 0.4 4 0.3 5.1 4.4 3.9 NA 1.3 9.9 14 15 43 2.1 45 0.5 5 0.1 5.0 5.0 4.2 NA 0.5 10 14 15 37 0.9 38 0.5 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; P = pur- chased; NP = nonpurchased. a. Twenty-one percent of the bottom quintile assigned a value of zero to nonpurchased biomass (see appendix A). Pakistan, it fell for the bottom three quintiles but rose at higher income levels, while in Vietnam it rose throughout the quintile range. The share of total expenditure on energy was highest for the lowest-income groups in seven countries in urban areas and in four countries in rural. The share of spending on electricity in rural areas increased with quin- tile level in Bangladesh, Cambodia, India, Kenya, Pakistan, Uganda, and Vietnam. The share in urban areas increased with rising quintile in Cam- bodia, India, Kenya, Uganda, and Vietnam, and through the bottom four quintiles in Bangladesh; no marked trend was observed in Indonesia or Pakistan. In Thailand, the share decreased in both rural and urban areas at higher income levels. Because urban household incomes are higher than rural for the same quintile groups, these results suggest that, as income increases, the share of expenditure on electricity also increases, until households reach the upper portion of the income distribution, at which point the share tends to decline. The pattern of expenditure share for kerosene was consistent across all countries--the share fell at higher quintile levels except in rural Indo- nesia and urban Kenya, Pakistan, and Vietnam. By contrast, the share of expenditure on LPG increased with quintile level except for urban Thailand, the top quintile in rural Thailand, and urban Vietnam. In Kenya, LPG consumption use was negligible; in Bangladesh and Uganda, data were not available. The share of expenditure on gasoline and diesel Extractive Industries for Development Series 47 increased with quintile level in both rural and urban areas in all coun- tries. The total expenditure share on petroleum products--character- ized by a declining share for kerosene and increasing shares for LPG and automotive fuels--increased with rising quintile in both rural and urban areas in Pakistan and Vietnam, in rural Indonesia and Thailand, and in urban India. It declined in both rural and urban Bangladesh and rural Cambodia. The share of expenditure on biomass, although higher in rural than in urban areas at each quintile level, decreased with rising quintile in all countries in both rural and urban areas, except in Kenya where this pattern was not observed because many households assigned a value of zero to freely collected biomass. The share of total food expenditure declined monotonically (steadily) with rising quintile in every country for both rural and urban households except urban India. The expenditure share of purchased food declined monotonically with rising quintile in all countries except Cambodia, Uganda, and Vietnam, and in rural Indonesia and urban India. The bot- tom two quintiles spent at least half of their total expenditure on pur- chased food in Bangladesh, India, and Indonesia, and in urban Cambo- dia. The expenditure share of purchased food in urban areas was higher than that in rural areas in 40 of the 45 quintiles in this study. Averaged across the 45 quintiles, the expenditure shares were 42 percent in rural areas and 48 percent in urban. This finding suggests that higher oil prices would have serious adverse indirect effects in terms of higher food prices on the poor, especially the urban poor. If a 10 percent increase in petroleum product prices were to lead to a 1 percent increase in food prices, the indirect effect on food prices would be larger than the direct effect of higher oil prices for all quintiles except in India, Thailand, Vietnam, and urban Kenya. The indirect effect would fall most heavily on the poor. Excluding Thailand, the bottom quintile in both rural and urban areas in the remaining eight countries would be hit harder by rising food prices caused by higher oil prices than the increase in oil prices themselves. If a 10 percent increase in petroleum product prices were to result in a 2 percent increase in food prices, the indirect effect through higher food prices alone would be greater than the direct effects of higher oil prices everywhere except the top one to three quin- tiles in Thailand and Vietnam. As expected, the imputed value of nonpurchased food was smaller for urban than rural households in every quintile across the nine countries. The expenditure share of nonpurchased food declined monotonically with rising quintile in Kenya, Thailand, Uganda, and Vietnam, and in 48 Expenditure of Low-Income Households on Energy rural Cambodia and urban Pakistan. The value of nonpurchased food exceeded that of purchased food in every quintile except in rural Uganda. The expenditure share of transport generally increased with rising quintile, except in Cambodia and Vietnam, the two countries with the lowest expenditure shares. The expenditure share was higher among urban households than among rural for every quintile only in Indonesia and Thailand. Expenditures on transport exceeded those on petroleum products for both rural and urban households in Bangladesh, Kenya, Pakistan, and Uganda. The average total household expenditures for urban quintiles were often not equal to those for their rural quintile counterparts, and a com- parison of energy consumption at similar income levels cannot be made by simply pairing equivalent quintile groups in urban and rural areas. However, there are a few cases where the mean expenditures for differ- ent quintiles were close, thereby permitting such a comparison. These include the following pairs of quintiles: Bangladesh urban quintile 1 (U1) and rural quintile 1 (R1), U2 and R2, U3 and R3, and U4 and R4; India U2 and R3; Indonesia U3 and R4; Kenya U4 and R4; and Uganda U3 and R3. Monthly household expenditures were within 2 percent of each other for Bangladesh, Kenya, and Indonesia; within 3 percent for Uganda (rural higher than urban); and within 4 percent in India (rural higher than urban). The expenditure shares for these pairings are displayed in table 3.7. Apart from Kenya and Uganda--where the inclusion of the value of nonpurchased food increased total rural expenditures by a large amount, thus reducing energy shares--the total share of expenditure on energy was similar between rural and urban quintiles at the same total house- hold expenditure level. However, in all cases, the shares of electricity and of modern energy were higher in the urban quintiles. Except for Kenya and Uganda, the share of expenditure on biomass was higher in the rural quintile. The share of expenditure on automotive fuels (gasoline and diesel) was higher in the rural households. This finding of higher rural expenditure on gasoline at the same total expenditure level was also identified by Bacon, Bhattacharya, and Kojima (2009) in their study of changing expenditure patterns in Indonesia and Pakistan. The share of expenditure on kerosene was higher in the urban quintile than in the corresponding rural quintile in India, Indonesia, Kenya, and Uganda; in Bangladesh, where the largest number of pairwise comparisons was avail- able, the share was consistently larger in rural areas. Extractive Industries for Development Series 49 Table3.7 SharesofExpenditureonVariousEnergySources,Food, andTransportforQuintilesatSimilarTotalHouseholdExpenditure Levels(%) Petro- Quintile Gaso- Nat- leum Energy Food Kero- line& Elec- ural Bio- prod- Trans- sene LPG diesel tricity gas mass ucts Modern Total P NP Total port R1 1.5 ND 0.0 0.3 0.0 6.2 1.5 1.8 8.0 56 13 70 1.8 U1 1.1 ND 0.0 1.1 0.1 5.9 1.1 2.3 8.3 62 6.5 68 1.7 R2 1.3 ND 0.0 0.4 0.0 5.8 1.3 1.8 7.6 51 17 68 2.2 Bangladesh U2 0.9 ND 0.0 1.6 0.3 5.0 0.9 2.8 7.7 60 6.6 66 2.0 R3 1.1 ND 0.0 0.8 0.0 5.5 1.2 2.0 7.5 49 16 65 2.5 U3 0.7 ND 0.0 2.0 0.6 4.3 0.8 3.4 7.8 57 5.3 62 2.4 R4 0.9 ND 0.1 1.1 0.1 4.7 1.0 2.2 6.9 45 15 60 2.9 U4 0.6 ND 0.0 2.6 0.9 3.3 0.6 4.2 7.5 55 4.0 59 2.8 R3 1.7 0.4 0.3 2.1 NA 6.9 2.4 4.4 12 49 12 61 2.4 India U2 1.9 0.7 0.1 2.6 NA 5.3 2.6 5.2 11 53 1.2 54 1.3 Indonesia R4 2.4 0.1 1.3 3.5 0.0 1.6 4.1 7.6 9.2 56 8.4 64 2.2 U3 3.1 0.1 0.8 3.9 0.0 0.7 4.3 8.2 8.9 56 5.0 61 2.8 R4 1.8 0.0 0.2 0.1 NA 1.6 2.0 2.0 3.7 33 23 56 3.0 Kenya U4 3.1 0.0 0.0 0.6 NA 2.5 3.2 3.8 6.3 45 7.0 52 4.3 R3 1.5 ND 0.1 0.1 NA 4.5 1.6 1.8 6.3 26 33 59 1.7 Uganda U3 1.8 ND 0.0 0.3 NA 5.6 1.9 2.2 7.8 41 10 51 1.5 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; P = pur- chased; NP = nonpurchased. As expected, the expenditure share of purchased food was higher and that of nonpurchased food was lower in urban areas in every case. The expenditure share of transport was comparable, except in India where it was lower for urban than rural households, and in Indonesia and Kenya where the reverse was found. The variation in shares of energy expenditures by quintile group indicates how the use of modern energy changes with rising income. Figure 3.4 shows the ratio of expenditure on modern energy to expendi- ture on all forms of energy for the highest and lowest quintile groups. In urban areas, the dependence on modern energy for the bottom quintiles was low in Bangladesh, Cambodia, India, and Uganda. The share was higher for the top quintile but remained relatively low in Uganda where 50 Expenditure of Low-Income Households on Energy Figure3.4 ShareofTotalHouseholdEnergyExpenditureSpenton ModernEnergybytheBottomandTopQuintiles % of total energy 99 99 100 100 94 96 96 85 85 84 83 80 79 80 74 77 67 67 67 6764 62 59 60 57 52 43 45 45 43 38 40 37 32 28 28 27 24 25 27 20 0 a h a a an nd m a di ya es si di nd na In st la ne en bo ad ga ki ai et do K am gl Pa Th Vi U an In C B Rural bottom quintile Rural top quintile Urban bottom quintile Urban top quintile Source: Authors' calculations. households still obtained a sizable portion of their energy from biomass; the share was not much higher in Bangladesh than in Uganda. In rural areas, except in Kenya (where 68 percent of the bottom rural quintile and 21 percent of the bottom urban quintile set imputed expenditures on biomass equal to zero), dependence on modern energy was even lower for the bottom quintile, and although much higher for the top quintile, was considerably less than in urban areas. The expenditure shares for biomass and kerosene generally declined with rising quintile, but expenditures themselves could still rise. For kerosene, it would be reasonable to take expenditure levels as a first approximation for quantities consumed, particularly in countries with essentially pan-territorial (nationally uniform) pricing. For biomass, the relationship between quantity and value could be weak because of large geographical and temporal price variation. Figures 3.5 and 3.6 show monthly expenditure on biomass for rural and urban quintiles, respec- tively, together with the expenditure share of biomass. Similarly, fig- ures 3.7 and 3.8 show expenditures on kerosene. It is striking that only in Thailand did the value of biomass consumed decline from the bottom to the second quintile in rural areas; the value in urban areas declined only in Indonesia, Pakistan, and Thailand. Extractive Industries for Development Series 51 Figure3.5 MonthlyRuralHouseholdExpenditureonBiomass 20 % share of household expenditure 20 Quintile 5 18 18 16 16 Quintile 1 14 14 12 12 2005 $ 10 10 8 8 6 6 4 4 2 2 0 0 an a a a m a sh nd a di si ny nd di na de st la ne In bo Ke ga ki ai et la do am Pa Th Vi U ng In C Ba Source: Authors' calculations. Note: In Kenya, 39 percent of rural households assigned an imputed value of zero to non- purchased biomass, including 68 percent of the bottom quintile, 48 percent of the second quintile, and 40 percent of the third quintile. These findings might suggest that the quantity of biomass consumed in low-income countries tends to increase with rising income before fall- ing. Similarly, the value of kerosene consumed generally rose with rising income among the bottom two quintiles in most countries. The evidence of rising consumption with income is stronger for kerosene than for bio- mass, because the relationship between quantity and value is more robust for kerosene. UptakeofDifferentEnergySources The pattern of uptake of different sources of energy is important in understanding differences in expenditure share by income group and by location, and forms a valuable supplement to information on average group expenditure patterns. In this study, a household was assumed to be consuming a good if it reported a positive expenditure on that good. There are four exceptions, explained in appendix A: 52 Expenditure of Low-Income Households on Energy Figure3.6 MonthlyUrbanHouseholdExpenditureonBiomass 16 16 % share of household expenditure Quintile 5 14 14 12 12 Quintile 1 10 10 8 8 2005 $ 6 6 4 4 2 2 0 0 an a a a m a sh nd a di si ny nd di na de st la ne In bo Ke ga ki ai et la do am Pa Th Vi U ng In C Ba Source: Authors' calculations. Note: In Kenya, 21 percent of the bottom urban quintile and 7 percent of the next two urban quintiles assigned an imputed value of zero to nonpurchased biomass. · LPG in Thailand, where the numbers presented are for those owning one or more LPG cookstoves · Biomass in Kenya, where the responses to a series of questions on dif- ferent forms of biomass were used to determine whether the house- hold was using biomass · Electricity in Kenya, where the responses to a series of questions on sources of electricity were used to determine whether the household was using electricity · Electricity in Uganda, where the higher of two percentages--those reporting positive expenditures or those reporting electricity as the primary source of lighting--for each rural and urban quintile was used to determine whether the household was using electricity and all other numbers calculated accordingly Uptake does not imply around-the-clock availability. Power outages in some of the survey countries are frequent. Irregular LPG supply is one reason many households continue to use biomass. What uptake does Extractive Industries for Development Series 53 Figure3.7 RuralHouseholdExpenditureonKerosene 8 8 % share of household expenditure 7 7 Quintile 5 6 6 5 5 2005 $ 4 4 Quintile 1 3 3 2 2 1 1 0 0 a a a a m sh an nd a di si ny nd di na de st la ne In bo Ke ga ki ai et la do am Pa Th Vi U ng In C Ba Source: Authors' calculations. Figure3.8 UrbanHouseholdExpenditureonKerosene 12 12 % share of household expenditure Quintile 5 10 10 Quintile 1 8 8 2005 $ 6 6 4 4 2 2 0 0 m a a a a h an nd a di si ny nd s di na de st la ne In bo Ke ga ki ai et la do am Pa Th Vi U ng In C Ba Source: Authors' calculations. 54 Expenditure of Low-Income Households on Energy indicate is that a particular energy source was sufficiently available for the household to have used it in the recent past. Table 3.8 gives uptake rates for rural quintiles and table 3.9 for urban quintiles; table B.11 gives uptake rates averaged across all households in each quintile. Tables B.12­B.15 give the equivalent values for Cambodia, Kenya, Uganda, and Vietnam when the value of nonpurchased food is excluded (which affects both total household expenditure and assign- ment of households to quintile groups). With the exception of Thailand, where access to electricity was virtu- ally universal, the uptake rate for electricity increased with rising quin- tile. This trend was particularly pronounced in urban Cambodia. In Indonesia, Pakistan, and Vietnam, the urban uptake rate was very high for all but the bottom quintile. In Kenya and Uganda, the uptake rate was almost zero for the lowest quintile even in urban areas; only within the top two (Uganda) or three (Kenya) urban quintiles did more than 20 percent of households use electricity. The uptake rate of electricity in rural areas was considerably lower in all countries. The exceptions to this were Thailand and Vietnam, where there was essentially no difference between rural and urban areas and where the uptake rate was already very high. In Uganda, less than 5 percent of rural households on average were using electricity, and only for the highest quintile did the rate rise above 10 percent. In Kenya, less than 7 percent of the total rural popula- tion used electricity. The uptake of LPG increased with quintile level in both rural and urban areas in those countries where this information was available. As an urban fuel, LPG was widely used in India (60 percent of households), Vietnam (70 percent), and Thailand (79 percent) but was little used in Indonesia, Kenya, and Pakistan, where the fuel of choice in urban areas was natural gas. In Thailand and Vietnam, the uptake rate for LPG in rural areas as a whole was substantial; elsewhere, only the top two rural quintiles in India, Indonesia, and Pakistan and the top quintile in Cam- bodia had uptake rates above 10 percent. The expenditure share of kerosene tends to fall with income, and its uptake rate depends critically on whether the rate of electrification is high--in the absence of electricity, households almost universally rely on kerosene for lighting--and how its price compares with that of LPG or, where piped gas is available, natural gas. An earlier study compared the use of kerosene by households in Indonesia and Pakistan and related the consumption patterns to prices paid per unit of useful energy (Bacon, Bhattacharya, and Kojima 2009). Because many governments subsidize Extractive Industries for Development Series 55 Table3.8 PercentageofRuralHouseholdsConsumingVarious EnergySources,Food,andTransport Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 93 ND 0.7 8.6 0.1 99 93 95 100 93 69 2 91 ND 0.5 16 0.6 100 91 94 100 94 78 Bangladesh 3 90 ND 0.4 29 0.9 100 91 96 100 92 83 4 90 ND 1.3 44 1.4 100 90 97 100 91 88 5 85 ND 5.3 56 4.7 99 87 97 100 92 89 All 90 ND 1.4 29 1.3 99 91 96 100 93 80 1 89 0.2 ND 1.2 NA 93 89 90 99 83 4.6 2 88 0.5 ND 3.0 NA 93 89 90 100 85 5.3 Cambodia 3 87 1.0 ND 6.0 NA 94 88 90 100 82 6.1 4 81 3.1 ND 13 NA 94 83 89 100 77 7.7 5 61 15 ND 33 NA 90 74 89 100 61 11 All 82 3.6 ND 11 NA 93 85 90 100 78 6.8 1 95 0.4 0.5 32 NA 98 95 99 99 46 54 2 95 2.1 1.9 46 NA 98 96 100 100 53 70 3 92 6.5 4.7 56 NA 97 95 100 100 59 77 India 4 86 17 11 69 NA 92 94 100 100 62 81 5 68 48 29 85 NA 75 92 99 100 58 82 All 89 12 7.4 54 NA 94 95 99 100 55 72 1 88 0.1 5.3 67 0.2 87 88 98 100 85 43 2 89 0.3 13 80 0.1 76 90 98 100 82 50 Indonesia 3 91 1.4 19 87 0.2 70 92 100 99 77 55 4 91 3.6 29 88 0.2 58 94 100 100 71 54 5 84 11 39 86 0.4 49 91 96 99 70 53 All 89 2.2 18 80 0.2 71 91 99 100 79 50 1 75 0.0 0.1 0.7 NA 99 75 75 98 94 30 2 87 0.1 0.2 1.2 NA 99 87 87 100 98 45 Kenyaa 3 90 0.1 0.6 3.5 NA 98 90 90 99 97 54 4 89 0.7 1.8 8.1 NA 98 90 90 100 98 65 5 87 5.6 4.7 21 NA 93 89 90 99 97 75 All 86 1.2 1.4 6.7 NA 97 86 87 99 97 54 1 45 2.5 7.5 66 2.3 95 51 97 100 63 87 2 44 5.2 11 71 2.7 96 53 98 100 72 91 Pakistan 3 44 7.7 11 76 2.1 97 54 99 100 74 92 4 38 10 20 81 4.3 93 56 99 100 76 91 5 31 18 29 87 5.8 91 58 99 100 77 91 All 41 8.5 15 76 3.4 95 55 98 100 73 91 56 Expenditure of Low-Income Households on Energy Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 1.1 44 66 97 0.0 69 68 99 99 94 19 2 0.8 67 78 99 0.0 60 81 100 99 91 25 Thailandb 3 0.9 80 83 99 0.0 46 86 100 99 88 30 4 0.6 86 85 99 0.0 28 89 100 99 78 32 5 0.3 86 88 99 0.1 14 91 100 100 62 33 All 0.8 72 79 99 0.0 46 82 100 99 84 27 1 88 ND 0.0 1.1 NA 96 88 88 99 97 13 2 95 ND 1.1 1.8 NA 97 95 95 100 97 23 Ugandac 3 97 ND 1.1 2.9 NA 97 97 97 99 95 32 4 96 ND 3.0 3.8 NA 97 97 97 99 93 39 5 91 ND 6.2 15 NA 89 91 95 99 80 48 All 94 ND 2.3 4.9 NA 95 94 95 99 92 33 1 45 2.8 26 89 NA 95 60 99 100 94 40 2 45 8.7 46 95 NA 95 70 100 100 93 48 Vietnam 3 43 23 58 96 NA 89 80 99 100 89 53 4 38 47 71 98 NA 76 89 100 100 84 55 5 31 71 78 99 NA 57 95 100 100 74 57 All 41 26 53 95 NA 85 78 99 100 88 50 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; P = pur- chased; NP = nonpurchased. a. For biomass and electricity, the percentages are for those households that reported consuming them. b. For LPG, the percentages are for those households that reported owning one or more LPG cooking stoves. c. For the top quintile for electricity, the percentage is set equal to that reporting electricity as the primary lighting source. kerosene as a social fuel, the relative end-user prices of kerosene and LPG do not necessarily follow price trends on the world market. Consistent with the foregoing general observations, the uptake rate of kerosene varied markedly from country to country. With the exception of Thailand, rural uptake was higher than urban, and indeed was about 90 percent in Bangladesh, India, Indonesia, Kenya, and Uganda. Urban uptake was also substantial in Indonesia, Kenya, and Uganda, but was about 50 percent in Bangladesh and India, much lower in Vietnam, very low in urban Pakistan (where the fuel of choice was natural gas), and essentially nonexistent in Thailand. Everywhere, apart from rural areas in Kenya and Uganda, uptake rates fell at higher quintile groups in both rural and urban areas. Extractive Industries for Development Series 57 Table3.9 PercentageofUrbanHouseholdsConsumingVarious EnergySources,Food,andTransport Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 81 ND 0.0 34 2.9 99 81 90 100 78 69 2 72 ND 0.1 50 9.0 97 72 90 100 72 76 Bangladesh 3 66 ND 0.4 59 16 97 66 89 100 66 81 4 57 ND 0.3 71 21 97 57 84 100 55 85 5 40 ND 5.0 81 50 96 44 87 100 48 93 All 55 ND 2.2 68 29 97 57 87 100 58 85 1 94 0.0 ND 5.5 NA 96 94 96 99 67 5.1 2 82 1.1 ND 16 NA 96 83 94 99 72 6.3 Cambodia 3 66 5.6 ND 34 NA 95 71 93 100 58 11 4 51 12 ND 51 NA 94 63 93 100 50 10 5 15 57 ND 90 NA 79 70 98 100 21 17 All 39 33 ND 64 NA 87 71 96 100 39 13 1 81 2.4 0.5 51 NA 81 82 94 97 11 41 2 88 9.8 1.0 70 NA 87 92 98 100 12 55 3 81 22 2.6 80 NA 75 91 99 100 11 67 India 4 69 45 10 89 NA 52 93 99 100 10 75 5 34 76 40 96 NA 18 93 99 100 9 77 All 50 59 26 90 NA 36 92 99 100 10 74 1 93 0.7 2.6 82 1.1 54 93 99 99 65 44 2 95 0.0 12 96 0.5 36 96 100 100 50 58 Indonesia 3 95 2.4 20 96 0.2 23 96 100 99 53 66 4 95 6.6 33 99 0.3 13 98 100 100 39 68 5 78 27 49 97 1.5 4.1 96 99 100 29 75 All 88 13 33 96 0.8 16 97 100 100 40 68 1 82 0.0 0.0 0.5 NA 86 82 82 99 74 24 2 87 0.0 0.0 16 NA 92 87 87 99 77 39 Kenyaa 3 86 0.4 0.1 23 NA 83 86 87 100 82 53 4 86 1.5 0.4 39 NA 86 86 89 100 85 67 5 80 22 9.5 66 NA 61 91 93 100 91 80 All 83 14 5.8 52 NA 70 89 91 100 88 71 1 16 3.5 8.1 90 35 65 25 99 100 26 82 2 13 4.6 11 96 46 51 25 100 100 22 83 Pakistan 3 13 7.1 12 96 53 44 27 99 100 23 89 4 9.8 7.9 20 97 64 32 33 100 100 20 92 5 4.2 7.9 48 99 82 12 55 100 100 15 88 All 8.7 7.0 29 97 66 30 40 100 100 19 88 58 Expenditure of Low-Income Households on Energy Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 0.9 61 65 98 0.0 54 68 99 99 82 18 2 0.3 78 76 99 0.0 39 80 100 100 80 28 Thailandb 3 0.4 86 72 100 0.0 21 80 100 99 68 36 4 0.3 84 68 99 0.0 8.5 77 100 100 55 51 5 0.1 76 72 100 0.0 2.5 77 100 100 37 54 All 0.2 79 71 100 0.0 11 77 100 100 51 48 1 96 ND 0.0 0.0 NA 95 96 96 99 72 16 2 95 ND 0.0 5.1 NA 98 95 96 99 68 31 Ugandac 3 90 ND 1.6 5.0 NA 97 90 92 100 62 32 4 82 ND 0.5 21 NA 96 82 92 99 47 43 5 69 ND 6.5 48 NA 82 72 91 99 44 51 All 76 ND 4.3 34 NA 87 78 92 99 48 47 1 42 11 33 89 NA 83 67 96 98 73 26 2 36 28 48 96 NA 74 74 100 100 56 39 Vietnam 3 31 49 67 99 NA 58 88 99 100 54 40 4 20 67 77 99 NA 42 90 100 100 41 43 5 12 90 88 100 NA 22 98 100 100 31 47 All 19 71 77 99 NA 38 91 100 100 40 44 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; P = pur- chased; NP = nonpurchased. a. For biomass and electricity, the percentages are for those households that reported consuming them. b. For LPG, the percentages are for those households that reported owning one or more LPG cooking stoves. c. For all quintiles but the second, the percentages are for those households reporting elec- tricity as the main lighting source. The uptake of gasoline and diesel was modest in most countries with the exception of Thailand and Vietnam. In all cases, the average urban uptake rate was higher than the rural rate; as expected, the uptake rate increased at higher quintiles in every country. The uptake of biomass also varied considerably, with Bangladesh, Cambodia, and Uganda exhibiting almost universal uptake in both rural and urban households, while the urban uptake was 70 percent in Kenya and below 40 percent in the other countries. However, in all countries rural uptake of biomass was high: the lowest rate, found in Thailand, was 46 percent. In India, Indonesia, Pakistan, Thailand, and Vietnam, the urban uptake of biomass declined strongly at higher quintiles. Similarly, Extractive Industries for Development Series 59 in rural India, Indonesia, Thailand, and Vietnam, uptake of biomass declined at higher quintiles. Natural gas was available to households only in urban Pakistan and, to a limited extent, urban Bangladesh. The uptake rate was very low in rural areas in both countries because, as in the rest of the world, natural gas is typically not available in such areas. Two-thirds of urban households in Pakistan used natural gas. Uptake increased steadily at higher quintiles among urban households in both countries. In Thailand, compressed natural gas as an automotive fuel was available, but at the time of the sur- vey was used by few households. As expected, virtually all households consumed purchased food; the handful that did not relied entirely on nonpurchased food. As expected, rural uptake of nonpurchased food was high, with the exception of India and the better-off in Cambodia and Thailand. The percentage of urban households consuming nonpurchased food was surprisingly high in many countries. The only group whose percentage was in single digits was the top quintile in India. The percentage of households reporting positive expenditure on trans- port also showed wide variation across countries. Cambodia had by far the lowest uptake rate in both rural and urban areas. Rural Pakistan, sur- prisingly, had the highest uptake rate. With the exception of Pakistan and Vietnam, on average, a greater share of urban households than rural ones used transport. SharesofExpenditureonEnergy,Food,and TransportbyUserHouseholds Where uptake is not universal, the value of average expenditure on energy for a given group would not accurately indicate the average expenditure of those households actually consuming the item. The lower the rate of uptake, the greater the difference between the average expen- diture for all households and the average expenditure for user house- holds: the average expenditure share of users is equal to the average expenditure share of all households divided by the uptake rate. Because energy pricing policies may take into account the importance of expendi- ture for those households using the energy source, it is helpful to address the extent of usage by households that actually use that energy source as well as food and transport affected by energy prices. Tables 3.10 and 3.11 show the average expenditure shares for energy sources by rural and urban user households, respectively; table B.16 60 Expenditure of Low-Income Households on Energy Table3.10 SharesofRuralHouseholdExpenditureonVarious EnergySources,Food,andTransport,byQuintile:UserHouseholds (%) Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 1.6 ND 3.9 3.5 4.8 6.3 1.7 2.0 56 14 2.6 2 1.4 ND 3.7 2.7 3.6 5.9 1.4 1.9 51 18 2.8 Bangladesh 3 1.3 ND 6.2 2.8 3.3 5.5 1.3 2.1 49 17 3.0 4 1.0 ND 8.2 2.6 3.8 4.7 1.1 2.3 45 16 3.3 5 0.8 ND 6.9 2.2 3.2 3.5 1.2 2.5 39 13 3.3 All 1.3 ND 6.5 2.6 3.4 5.3 1.4 2.1 49 16 3.0 1 1.8 5.7 ND 5.6 NA 7.1 1.8 1.9 50 33 4.5 2 1.5 4.8 ND 5.1 NA 6.2 1.5 1.7 51 29 3.4 Cambodia 3 1.3 4.3 ND 4.9 NA 5.6 1.4 1.6 54 26 3.5 4 1.1 3.8 ND 4.5 NA 4.9 1.3 1.8 56 21 3.0 5 0.7 2.4 ND 3.8 NA 3.2 1.1 2.3 50 15 1.3 All 1.3 2.9 ND 4.2 NA 5.4 1.4 1.8 52 25 2.9 1 2.2 7.0 5.9 3.9 NA 9.0 2.3 3.4 55 19 2.9 2 2.0 6.4 6.6 3.7 NA 8.0 2.2 3.8 52 20 2.7 3 1.8 6.1 6.4 3.7 NA 7.2 2.5 4.4 49 21 3.1 India 4 1.6 5.2 6.8 3.5 NA 6.1 3.3 5.5 45 19 3.6 5 1.3 3.9 7.0 3.3 NA 4.1 5.4 7.8 41 15 4.7 All 1.9 4.7 6.8 3.6 NA 7.3 2.9 4.7 49 19 3.3 1 2.4 4.4 4.6 3.6 2.0 4.3 2.7 4.9 55 18 2.9 2 2.6 3.6 4.8 3.6 1.1 3.8 3.4 6.1 56 14 3.5 Indonesia 3 2.5 4.0 4.7 3.9 2.7 3.2 3.8 6.9 56 14 3.9 4 2.6 4.0 4.6 3.9 3.7 2.7 4.4 7.7 56 12 4.0 5 2.2 2.8 4.2 3.5 1.2 2.1 4.7 7.6 53 12 4.2 All 2.5 3.4 4.6 3.7 2.2 3.5 3.6 6.4 55 15 3.6 1 2.8 -- 11 1.3 NA 4.7 2.8 2.8 40 33 5.0 2 2.4 5.2 7.6 3.7 NA 4.1 2.4 2.5 37 30 4.8 Kenya 3 2.2 5.2 5.8 4.1 NA 3.7 2.3 2.3 35 27 4.6 4 2.0 3.5 9.9 3.6 NA 3.3 2.2 2.3 33 24 4.6 5 1.8 4.5 10 2.2 NA 2.7 2.6 2.7 30 17 5.0 All 2.2 4.4 9.7 2.7 NA 3.5 2.4 2.5 35 26 4.8 (continued) Extractive Industries for Development Series 61 Table3.10 SharesofRuralHouseholdExpenditureonVarious EnergySources,Food,andTransport,byQuintile:UserHouseholds (%)(continued) Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 1.2 2.1 1.5 4.8 3.7 4.9 1.4 4.1 48 19 3.0 2 1.1 2.5 2.5 4.6 3.1 4.5 1.6 4.3 44 20 3.1 Pakistan 3 1.0 2.3 3.7 4.4 2.9 4.3 1.9 4.5 42 20 3.3 4 1.0 3.1 4.8 4.3 3.3 3.7 2.9 5.3 39 21 3.4 5 0.8 2.9 6.1 4.2 2.5 3.6 4.3 6.4 35 19 3.8 All 1.0 2.7 4.4 4.4 3.0 4.2 2.4 4.9 42 20 3.3 1 1.1 3.9 7.0 3.3 -- 2.3 7.3 8.3 37 18 4.2 2 1.6 3.4 7.4 3.1 -- 1.8 8.0 9.6 37 14 4.1 Thailand 3 1.0 2.7 7.9 3.0 -- 1.6 8.5 10 36 11 4.6 4 0.5 2.2 8.5 2.9 -- 1.3 9.0 11 34 9.4 5.1 5 0.3 1.4 9.5 2.4 2.4 0.6 9.7 11 28 6.6 4.9 All 1.1 2.5 8.0 3.0 2.4 1.8 8.5 9.9 35 13 4.6 1 2.0 ND -- 6.9 NA 6.8 2.0 2.1 26 36 6.6 2 1.8 ND 3.7 5.5 NA 5.5 1.8 1.9 25 36 5.8 Uganda 3 1.6 ND 9.4 4.4 NA 4.7 1.7 1.8 26 34 5.3 4 1.4 ND 8.5 3.7 NA 4.0 1.7 1.8 28 31 5.6 5 1.3 ND 9.0 3.5 NA 3.0 1.9 2.4 31 21 5.2 All 1.6 ND 8.3 4.0 NA 4.8 1.8 2.0 27 31 5.5 1 0.9 8.3 4.4 2.7 NA 5.7 3.0 4.2 37 28 1.5 2 0.7 8.7 4.3 2.8 NA 4.8 4.2 5.7 38 21 1.3 Vietnam 3 0.6 8.1 4.9 2.9 NA 4.0 6.1 7.7 40 16 1.3 4 0.6 7.5 5.2 2.8 NA 3.5 8.3 10 39 12 1.1 5 0.5 6.1 5.6 2.9 NA 2.7 9.4 12 37 7 0.9 All 0.7 7.3 4.9 2.8 NA 4.4 6.1 7.5 38 18 1.2 Source: Authors' calculations. Note: NA = fuel not available; -- = no household reported expenditure; ND = no question was asked concerning the fuel; P = purchased; NP = nonpurchased. shows the average expenditures for user households averaged across both rural and urban areas. The shares for petroleum products and for modern energy are calculated by averaging over households that use at least one petroleum product or modern source of energy, respectively, and hence are not equal to the sum of the expenditures by user households over the separate fuels. 62 Expenditure of Low-Income Households on Energy Table3.11 SharesofUrbanHouseholdExpenditureonVarious EnergySources,Food,andTransport,byQuintile:UserHouseholds (%) Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 1.4 ND -- 3.2 4.1 6.0 1.4 2.6 62 8.3 2.5 2 1.2 ND 5.6 3.2 3.5 5.1 1.2 3.1 60 9.1 2.7 Bangladesh 3 1.1 ND 7.1 3.4 3.8 4.4 1.2 3.9 57 8.0 2.9 4 1.1 ND 4.2 3.7 4.5 3.4 1.1 4.9 55 7.3 3.3 5 0.7 ND 5.2 3.1 3.2 1.6 1.5 5.5 42 4.1 3.6 All 1.0 ND 5.2 3.3 3 3.2 1.3 4.6 51 6.7 3.3 1 2.0 -- ND 4.9 NA 7.4 2.0 2.2 61 24 0.9 2 1.6 4.5 ND 6.7 NA 6.2 1.7 2.6 59 23 0.7 Cambodia 3 1.3 3.7 ND 4.3 NA 5.4 1.5 2.7 60 21 1.0 4 1.2 2.9 ND 4.5 NA 4.4 1.5 3.5 60 15 1.6 5 0.7 2.2 ND 4.3 NA 1.9 1.9 5.3 47 10 0.6 All 1.3 2.3 ND 4.4 NA 3.7 1.8 4.2 53 17 0.8 1 2.5 7.5 7.4 4.5 NA 7.0 2.8 4.9 53 8.7 4.0 2 2.2 7.1 5.3 3.7 NA 6.1 2.9 5.3 53 9.9 2.4 3 2.3 6.9 4.9 3.8 NA 5.3 3.8 6.6 50 11 2.6 India 4 2.5 5.8 5.2 3.9 NA 4.3 5.2 8.4 47 10 3.1 5 2.3 3.8 5.6 4.0 NA 2.5 6.4 9.9 37 5.9 3.7 All 2.3 4.3 5.6 3.9 NA 4.3 5.7 8.9 41 7.8 3.4 1 3.8 1.5 3.5 4.3 3.9 4.0 3.9 7.3 60 7.8 2.7 2 3.6 -- 4.2 4.2 4.5 3.1 4.2 8.1 58 7.9 3.5 Indonesia 3 3.2 3.6 4.2 4.1 6.6 2.9 4.4 8.2 57 9.4 4.2 4 2.8 3.2 3.8 4.2 4.9 2.1 4.6 8.7 55 7.9 4.9 5 2.0 2.1 3.8 4.1 2.2 1.7 4.6 8.5 48 6.8 4.9 All 2.7 2.3 3.8 4.1 2.9 2.9 4.5 8.4 53 7.9 4.6 1 3.8 -- -- 2.2 NA 5.0 3.8 3.8 54 16 5.0 2 3.9 -- -- 4.6 NA 6.1 3.9 4.5 49 13 5.2 Kenya 3 4.3 11 5.6 3.3 NA 4.5 4.4 4.7 48 9.3 6.2 4 3.6 4.2 4.6 3.4 NA 3.3 3.7 4.3 45 8.3 6.5 5 2.8 2.6 8.8 2.1 NA 1.9 4.0 4.8 35 6.8 7.1 All 3.2 2.7 8.8 2.4 NA 2.9 4.0 4.6 40 7.8 6.9 (continued) Extractive Industries for Development Series 63 Table3.11 SharesofUrbanHouseholdExpenditureonVarious EnergySources,Food,andTransport,byQuintile:UserHouseholds (%)(continued) Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 0.7 3.4 2.6 5.3 3.5 4.9 1.8 6.5 51 12 2.9 2 1.3 3.4 2.7 5.2 3.0 4.5 2.5 7.0 49 12 3.2 Pakistan 3 1.1 3.6 3.2 5.1 2.8 4.0 2.9 7.3 46 11 3.1 4 1.5 3.9 4.3 5.2 2.6 3.4 4.0 8.0 44 11 3.8 5 0.9 4.0 6.6 4.7 2.0 2.7 6.4 9.8 36 10 3.7 All 1.1 3.8 5.8 5.0 2.4 3.9 5.1 8.5 42 11 3.5 1 0.6 4.0 7.2 3.9 -- 2.3 7.7 9.1 42 14 4.1 2 1.1 2.9 6.8 3.8 -- 1.8 7.4 9.7 41 11 4.6 Thailand 3 0.8 2.4 7.1 3.8 -- 1.6 7.4 9.7 39 10 5.0 4 0.7 1.7 7.3 3.6 1.9 1.1 7.3 9.2 38 8.6 5.4 5 0.4 1.1 8.6 3.3 2.8 0.6 8.5 9.8 31 7.0 5.4 All 0.7 1.6 7.9 3.5 2.7 1.6 7.9 9.6 35 8.8 5.3 1 2.6 ND -- -- NA 7.8 2.6 2.6 41 23 6.7 2 1.8 ND -- 4.4 NA 6.8 1.8 2.0 38 22 5.5 Uganda 3 2.0 ND 3.0 6.7 NA 5.8 2.1 2.4 41 17 4.5 4 1.6 ND 3.9 3.8 NA 4.9 1.6 2.3 43 14 5.3 5 1.2 ND 9.2 3.2 NA 3.1 2.0 3.3 36 12 5.4 All 1.5 ND 8.8 3.4 NA 4.2 2.0 2.9 38 14 5.3 1 1.4 11 4.7 3.7 NA 4.4 5.0 6.9 48 18 1.2 2 1.9 8.4 4.6 3.7 NA 4.3 7.1 8.8 50 11 1.6 Vietnam 3 1.8 8.6 5.2 3.7 NA 3.6 9.3 12 47 8.3 1.0 4 1.6 7.6 5.7 3.9 NA 3.1 11 14 43 5.2 1.1 5 0.7 5.5 5.6 4.3 NA 2.2 10 14 37 2.9 1.0 All 1.4 6.4 5.6 4.0 NA 3.2 10.0 13 41 6.4 1.1 Source: Authors' calculations. Note: NA = fuel not available; -- = no household reported expenditure; ND = no question was asked concerning the fuel; P = purchased; NP = nonpurchased. For rural and urban quintiles, the share of expenditure on electricity by users tended to fall at the upper quintile levels. This suggests that the increasing share by quintile group for all households was in large part due to increasing uptake. The declining expenditure share for electricity suggests that the income elasticity of demand for user households may be less than unity; this hypothesis could be explored from surveys where quantities purchased are available. 64 Expenditure of Low-Income Households on Energy For LPG, the expenditure share by user households also tended to decline at higher quintile levels, suggesting that the quantity purchased by these households did not increase markedly with quintile expendi- ture, thus resulting in a declining share. This finding is in strong contrast to the share for all households, which tended to increase with quintile level, suggesting that the increased uptake rate was the dominant factor underpinning the increase in share for all households. Uptake rates for kerosene were generally high, so the pattern of expenditure share by user households was similar to that for all house- holds. For gasoline and diesel, where uptake rates were much lower, the share of expenditure by user households was markedly higher than for all households. For those households that have decided to use these fuels, their importance in the budget was significant even in low-quintile groups--indicating that, for a subgroup of households, increases in the price of these fuels would have had a large effect on household welfare. The share of expenditure for those households that purchased at least one petroleum product was higher than for the group of all households, indicating that there were some households that did not purchase any petroleum products at all. The difference was largest for the lowest quin- tiles. The expenditure share of transport is high in some countries. The bottom two quintiles in Kenya, Thailand, and Uganda and in rural Cam- bodia stand out in this regard. In Uganda, the expenditure share in every quintile except one exceeded 5 percent--a much larger share than that spent by users on modern energy. MainEnergySourceforCooking A household's choice of primary cooking fuel is of interest because, as with lighting, this is a major activity for which a choice of energy sources exists and fuel substitution is possible. Among the energy choices for cooking, the factors leading to the decision to use biomass as the main source of energy are particularly complex--cooking with biomass may be time consuming, not only in terms of getting the fire started but also because, if not purchased, the fuel must be collected. Also, traditional use of biomass creates considerable indoor air pollution, which is injuri- ous to health; nonetheless, many households continue to use it given its low (or nonexistent) cost and their limited cash income. Cooking practices and cultural preferences also influence the decision to rely on biomass. Extractive Industries for Development Series 65 Six of the nine surveys examined (all except those in Bangladesh, Indonesia, and Vietnam) asked about the main cooking fuel used. The definition of "main" could indicate the fuel used for the longest period per unit time, or for generating the most heating power, or costing the most if fuels are purchased; the surveys do not use a consistent or precise definition. Generally, it is understood that "main" refers to the fuel used for the longest amount of time. One complication is that many house- holds use multiple fuels, and which should be classified as the main one was likely not clear to some. The following results should be interpreted with these limitations in mind. Figure 3.9 shows that biomass was by far the most commonly used main cooking fuel in five of the six countries. The one exception was Thailand, where LPG was widely used. When the data are analyzed by area (figure 3.10), a clear divide between rural and urban households becomes evident in India, Kenya, and Pakistan. The most common main cooking fuels in urban areas are LPG in India (60 percent), kerosene in Kenya (45 percent), and natural gas in Pakistan (67 percent). Table 3.12 presents the results by rural and urban quintiles, while table B.17 gives the national statistics by quintile. This paper groups answers according to five energy sources--electricity, kerosene, bio- mass, LPG or natural gas, and other (mainly coal and coke). The results Figure3.9 MainCookingFuelAcrossAllHouseholds % of all households 100 93 96 82 80 73 70 59 60 40 37 20 0 Cambodia India Kenya Pakistan Thailand Uganda Thailand LPG Biomass Source: Authors' calculations 66 Expenditure of Low-Income Households on Energy Figure3.10 MainCookingFuelinRuralandUrbanAreas % of households 97 96 98 100 94 86 89 78 80 72 60 49 46 41 40 31 25 20 12 0 Cambodia India Kenya Pakistan Thailand Uganda Thailand LPG Biomass Rural Urban Source: Authors' calculations indicate that in the countries for which information was available, elec- tricity was almost never used as the main cooking source, even by the highest urban quintile groups. Thailand was the one exception to this: 15 percent of households in the highest urban quintile used electricity as their main cooking fuel. When these figures are compared to the propor- tion of households with electricity uptake, it is clear that uptake was not related to the non-use of electricity as the main energy source for cook- ing: apart from Uganda, the uptake rates at the top quintile levels were substantial. Electricity is widely used in Asia for cooking rice by those households connected to electricity, and is used increasingly everywhere for reheating food and limited cooking using microwave ovens by those who own them. However, for other cooking activities, even high-income households seem to prefer a gaseous fuel to electricity. The use of biomass as the main cooking fuel was surprisingly high in some urban areas: the third quintile in India had a 60 percent rate, that in Pakistan a 45 percent rate, and that in Thailand a 20 percent rate. Ker- osene was the main cooking fuel for some upper-quintile urban house- holds in India (16 percent in the fourth quintile), Kenya (50 percent in the top quintile), and Uganda (5 percent in the top quintile). Elsewhere, it was scarcely ever the main cooking source. In Thailand, LPG was the Extractive Industries for Development Series 67 Table3.12 MainEnergySourceforCooking:PercentageofRural/ UrbanHouseholdsUsingThatSource Quintile Electricity Kerosene Biomass LPG/naturalgas Other Rural 1 0.0 0.0 99 0.1 0.7 2 0.0 0.0 99 0.1 0.7 3 0.0 0.0 98 0.5 1.0 4 0.0 0.1 98 1.3 0.7 5 0.3 0.1 88 10 1.2 Cambodia All 0.1 0.1 97 2.3 0.9 Urban 1 0.6 0.0 97 0.6 1.7 2 0.4 0.0 97 1.0 1.3 3 0.0 0.0 97 2.0 1.1 4 0.2 0.2 91 5.8 2.8 5 0.9 0.2 52 45 1.9 All 0.6 0.1 72 25 1.9 Rural 1 0.0 0.5 94 0.4 5.6 2 0.0 0.7 93 1.7 5.0 3 0.0 0.8 90 4.9 4.0 4 0.1 1.8 82 13 3.4 5 0.1 3.6 57 38 1.9 All 0.0 1.3 86 8.6 4.2 India Urban 1 0.2 6.7 80 2.8 11 2 0.1 7.2 75 9.2 8.2 3 0.1 11 59 21 9.2 4 0.2 16 36 43 5.3 5 0.3 9.4 8.1 80 2.2 All 0.2 11 25 60 4.2 Rural 1 0.0 0.4 100 0.1 0.0 2 0.0 0.6 99 0.0 0.1 3 0.2 1.1 98 0.0 0.5 4 0.2 1.8 97 0.2 0.7 5 0.9 9.9 85 3.4 0.6 Kenya All 0.2 2.6 96 0.7 0.4 Urban 1 0.0 13 86 0.2 0.5 2 0.0 19 81 0.2 0.1 3 0.8 36 62 0.5 0.9 4 0.5 43 54 0.9 1.9 5 2.6 50 27 19 1.0 All 1.8 45 41 12 1.1 68 Expenditure of Low-Income Households on Energy Quintile Electricity Kerosene Biomass LPG/naturalgas Other Rural 1 0.0 0.6 96 2.7 1.2 2 0.0 0.4 96 3.0 0.9 3 0.0 0.7 96 2.7 0.6 4 0.0 0.7 92 6.0 1.3 5 0.2 1.0 88 10 0.7 Pakistan All 0.1 0.7 94 4.8 0.9 Urban 1 0.0 0.5 65 33 0.7 2 0.0 1.6 51 47 0.7 3 0.0 0.8 45 54 1.1 4 0.0 1.6 33 65 0.3 5 0.2 0.9 13 85 1.1 All 0.1 1.1 31 67 0.9 Rural 1 0.6 0.3 80 19 0.4 2 0.6 0.2 61 38 0.3 3 1.1 0.3 43 55 0.1 4 2.2 0.1 24 74 0.1 5 1.6 0.3 47 51 0.2 Thailand All 1.5 0.3 46 49 2.9 Urban 1 1.6 0.7 63 35 0.1 2 2.7 0.2 41 56 0.1 3 4.3 0.2 20 75 0.1 4 9.2 0.2 7.5 83 0.0 5 14 0.2 2.1 83 0.1 All 10 0.2 12 78 0.1 Rural 1 0.0 0.1 100 0.0 0.3 2 0.2 0.2 99 0.0 0.2 3 0.0 0.5 99 0.0 0.7 4 0.0 0.3 98 0.0 1.4 5 0.2 2.8 91 0.1 5.7 Uganda All 0.1 0.8 98 0.0 1.6 Urban 1 0.0 0.0 100 0.0 0.0 2 0.0 0.0 99 0.0 0.9 3 0.0 0.0 100 0.0 0.3 4 0.0 0.8 98 0.0 1.5 5 1.4 5.4 83 1.2 9.2 All 0.8 3.5 89 0.7 6.0 Source: Authors' calculations. Note: Biomass includes firewood, charcoal, dung, and agricultural waste. The category of LPG and natural gas is combined in the questionnaire for Pakistan; in Cambodia, India, Ke- nya, and Uganda, households do not use natural gas. For Cambodia, those households that reported using a combination of LPG and electricity as their main energy source for cooking are categorized under LPG. "Other" includes coal and coke. Extractive Industries for Development Series 69 most common main cooking fuel at higher quintile groups in both rural and urban areas. In Cambodia, LPG was important in the top urban quintile. The use of LPG as the main cooking fuel was low in all other rural areas except at the highest quintile level in India, but its importance increased steadily with rising income in urban India, reaching 80 percent for the top quintile. In Pakistan, where natural gas is widely available in urban areas, the uptake rates for gas were almost identical to the rate of choice of natural gas for cooking. Where households had uptake of gas, it was their main cooking fuel. MainEnergySourceforLighting Lighting is another activity that can be powered by different energy sources, of which electricity and--in its absence--kerosene are usually the dominant choices. No question on main lighting fuel was asked in the surveys in Bangladesh, Indonesia, and Thailand (this last presum- ably because of Thailand's near-universal access to electricity). In Kenya, detailed data on energy sources used other than electricity and kerosene revealed that firewood was important for lower quintiles; in Cambodia, batteries were an important third main energy source. Figure 3.11 shows the percentage of households using electricity as their main lighting source, and, for the three countries for which electric- ity use was limited, the share of households using kerosene for lighting. The percentages of households using electricity match the uptake rates. Figure 3.12 shows the same statistics for rural and urban households separately. Tables 3.13 and 3.14 show the results by quintile for rural and urban households, respectively, while tables B.18 and B.19 give the percentages averaged across the country. Generally, the share of households using electricity as a main lighting source corresponded to the rate of uptake of this energy source; absent its uptake, kerosene was the most widely used source of lighting, consumed disproportionately by the poor. In India, Pakistan, and Vietnam, the proportion of households indicating that something other than electricity or kerosene was their main lighting source was very small in all quintiles. Only in some quintiles in Cambo- dia, Kenya, and Uganda did more than 10 percent of households indicate an energy source other than electricity or kerosene as their main lighting source. Batteries were the main lighting source for at least one-quarter of households for the top four quintiles in rural Cambodia. Batteries were also important in urban Cambodia, except for the bottom and top 70 Expenditure of Low-Income Households on Energy Figure3.11 MainLightingSourceforAllHouseholds % of all households 100 96 83 85 80 76 65 60 55 40 19 16 20 10 0 a ya an da a ya da di di a n st n am di n n bo In Ke i ga et n bo Ke ga am P ak U Vi am U C C Electricity Kerosene Source: Authors' calculations. 3.12MainLightingSourceinRuralandUrbanAreas % of households 97 99 100 92 95 91 86 77 80 65 60 60 55 55 51 47 41 40 27 20 11 4 4 0 a a an a a a a m a di di nd nd ny ny di na st bo bo In Ke Ke ga ga ki et am am Pa U U Vi C C Electricity Kerosene Rural Urban Source: Authors' calculations. Extractive Industries for Development Series 71 Table3.13 MainEnergySourceforLightinginIndia,Pakistan, Uganda,andVietnam:PercentageofRural/UrbanHouseholdsUsing ThatSource Elec- Kero- Elec- Kero- Quintile tricity sene Other Quintile tricity sene Other Rural 1 33 67 0.5 Urban 1 58 42 0.9 2 46 54 0.7 2 71 28 1.2 3 56 43 0.4 3 81 18 0.8 India 4 69 31 0.5 4 91 8.7 0.5 5 86 14 0.4 5 98 1.7 0.2 All 55 45 0.5 All 92 7.2 0.4 1 69 29 1.8 1 92 6.5 1.2 2 72 26 2.1 2 95 3.7 1.5 Pakistan 3 77 22 1.6 3 97 2.1 1.1 4 82 17 0.9 4 97 1.4 1.3 5 87 12 0.8 5 98 0.4 2.1 All 77 22 1.5 All 97 1.7 1.6 1 0.2 89 11 1 0.6 99 0.0 2 0.6 95 4.4 2 1.4 97 1.3 Uganda 3 1.6 96 2.3 3 9.2 87 3.5 4 2.6 96 1.7 4 26 69 4.3 5 15 80 4.6 5 57 38 4.8 All 4.0 91 4.7 All 41 55 4.2 1 90 8.0 2.4 1 96 3.2 1.0 2 95 2.9 1.9 2 97 2.2 1.3 Vietnam 3 96 2.8 1.4 3 100 0.0 0.0 4 98 1.0 1.5 4 100 0.0 0.1 5 99 0.6 0.8 5 100 0.0 0.3 All 95 3.3 1.7 All 99 0.3 0.3 Source: Authors' calculations. Note: In Vietnam, the categories for lighting are grid electricity; LPG, vegetable oil, or kero- sene (categorized here as "kerosene"); and batteries, generators, and other (categorized here as "other"). quintile groups. Firewood was the main lighting source for the low- est quintile in rural and urban Kenya. Also in Kenya, as comparison of table A.2 and table 3.14 shows, the use of electricity as the main lighting source was most closely linked to having access to grid electricity. 72 Expenditure of Low-Income Households on Energy Table3.14 MainEnergySourceforLightinginCambodiaandKenya: PercentageofRural/UrbanHouseholdsUsingThatSource Quintile Electricity Kerosene Batteries Other Rural 1 1.6 80 16 2.5 2 3.5 71 24 1.3 3 6.4 64 29 1.1 4 13 53 34 0.5 5 34 34 32 0.6 Cambodia All 11 60 27 1.2 Urban 1 6.2 85 6.6 2.6 2 16 69 13 2.3 3 34 50 14 2.0 4 52 33 15 0.8 5 91 5.7 2.8 0.2 All 65 27 7.7 0.9 Rural 1 0.3 78 17 4.6 2 0.7 90 6.3 3.2 3 1.8 93 3.7 1.8 4 4.0 91 2.0 3.4 5 13 79 1.5 6.8 All Kenya 3.8 86 5.9 3.9 Urban 1 0.2 87 8.9 3.9 2 15 79 0.8 5.0 3 22 76 0.9 0.8 4 38 60 0.3 2.0 5 65 32 0.2 2.2 All 51 47 0.5 2.2 Source: Authors' calculations. StylizedEnergyFacts The findings of a group of earlier energy survey studies compared to a series of stylized facts were presented in table 2.14. These studies were based mainly on surveys carried out in the late 1990s or early 2000s, while the current study is based on later surveys from a period centering around 2005. Of the nine countries that formed the basis of the current analysis, only India and Vietnam were covered in these earlier studies. It is useful to assess whether the earlier findings held true in this later period and for a largely different group of countries. The same patterns Extractive Industries for Development Series 73 can be checked (see the section on common findings in chapter 2 for the specific questions asked), apart from expenditure on diesel, which was not identified in these surveys (table 3.15). The surveys included here show fairly consistent patterns. The share of expenditure on all energy for the lowest quintile was greater than 5 percent in every country, except in rural Kenya. When biomass is excluded, the share of expenditure on modern energy exceeded 5 percent for the bottom rural quintile only in Pakistan, indicating that the exclu- sion of biomass--for which values are imputed for collected fuels--can change the apparent importance of energy to the household budget. The previous studies showed a mixed picture, which may have depended in part on the treatment of biomass. When all sources of energy are considered, urban households spent a greater proportion of their total expenditures except in India, Indonesia, and Thailand. When only modern sources of energy are considered, the share was higher in urban than in rural areas everywhere except Thai- land. Previous studies had also largely indicated a greater importance of energy for urban households. The share of kerosene declined with income group in all cases, except for rural Indonesia (where kerosene was heavily subsidized at the time Table3.15 PatternsofEnergyUseBasedonSelectedSurveys U-shapedresponse Electricityrises Gasolinerises Kerosenefalls Urban>rural Share>5% LPGrises Countryandsurveyyear Bangladesh, 2005 Y Y Y Y ND Y N Cambodia, 2003­04 Y Y Y Y Y ND ND India, 2004­05 Y N Y Y Y Y N Indonesia, 2005 Y N N N Y Y N Kenya, 2005­06 N Y N Y Y Y N Pakistan, 2004­05 Y Y Y N Y Y N Thailand, 2006 Y N Y N N N N Uganda, 2005­06 Y Y Y Y ND Y Y Vietnam, 2006 Y Y Y Y Y Y N Source: Authors' calculations. Note: Y = yes; N = no; ND = no question was asked concerning the fuel. 74 Expenditure of Low-Income Households on Energy of the survey and, equally important, not rationed; kerosene was also heavily subsidized in India but was tightly rationed) and urban Kenya, conforming to the universal pattern found in earlier surveys. The share of LPG increased in six of the seven countries where data were avail- able. This result was in contrast to the previous surveys where the share increased in only three of the seven countries for which there was infor- mation. The share of electricity increased with quintile level in six countries, while in Indonesia and Pakistan it neither increased nor decreased; it showed a downward trend in Thailand. This mixed pattern was also identified in the earlier surveys, where the expenditure share increased in four of the eight cases for which information was presented. The clearest pattern of household energy expenditure was that of gasoline and diesel. In the current study, seven out of eight countries for which there were data showed an increasing share, Thailand being the one notable exception. In the earlier surveys, the share had similarly increased in every country for which information was available. Although the surveys examined here showed increasing shares of LPG and of gasoline and diesel and decreasing share of kerosene with income, only in Uganda did these opposing trends produce a U-shaped expen- diture share for petroleum products, primarily in rural areas (tables 3.5 and 3.6). In some countries, the share of total expenditure on petroleum products increased, while in Bangladesh it decreased monotonically with income. Earlier surveys had shown a U-shaped response in the majority of cases, but it appears that the relative strengths of the demands for the different petroleum products varied sufficiently among countries to rule out a general pattern. However, it may well be that, even in Bangladesh, as incomes increase further, the share of kerosene will decline to such a low level that the increasing share of gasoline will produce an increasing share for petroleum products. ImplicationsforUniversalPriceSubsidiesfor PetroleumProducts Patterns of household energy use and consumption guide policies for energy subsidies. Komives and others (2005) propose three dimensions of subsidy performance: · Benefit incidence (how well the subsidy targets benefits to poor, as opposed to other, households) Extractive Industries for Development Series 75 · Beneficiary incidence (what proportion of poor households as a whole receive the subsidy) · Materiality (how significant the amount of the subsidy received by poor households is) It would be possible to have a scheme in which the subsidies were well targeted (most of the benefits go to poor households) but with low beneficiary incidence because only a few poor households actu- ally received the subsidy. Alternatively, a subsidy might reach only poor households and most poor households but be small in value relative to household income. One useful measure is the benefit-targeting indicator (), defined as the ratio of the share of total benefits received by poor households to the proportion of households that are poor. If the indicator takes a value of unity, the scheme is neutral, and the poor receive benefits in propor- tion to their numbers. A subsidy is progressive if is greater than 1, and regressive if it is less than 1, with nonpoor households receiving a larger share of the total subsidy pool than their proportion in the population. Beneficiary incidence is measured by the exclusion rate (E)--the percent- age of poor households that do not receive the subsidy--while material- ity is defined as the average value of the subsidy received by poor house- holds benefiting from the subsidy as a percentage of their household income. The benefit-targeting indicator can be shown to be equal to the prod- uct of five ratios: = (Ap / AH) × (Up / UH) × (Tp / TH) × (Rp / RH) × (Qp / QH) (1) (2) (3) (4) (5) where A = percentage of households that have potential access to the energy source U = percentage of households with access that are connected to the energy source T = share of households that are connected that are eligible for the subsidy R = average rate of subsidization for eligible households Q = average quantity consumed by subsidy recipients p = group of poor households H = group of all households 76 Expenditure of Low-Income Households on Energy It is possible to compute for hypothetical price subsidies for petro- leum products under certain restricted conditions using the data ana- lyzed in this paper. The simplifying assumptions are that the surveyed households faced similar unit prices for the petroleum product in ques- tion, the hypothetical subsidy is universal, and the subsidy is constant per unit of the petroleum product and not a function of the amount purchased. The first assumption--essentially pan-territorial pricing--is a weak one for countries with liberalized pricing, particularly Uganda. Because many poor households live in areas with poor transport infra- structure, they tend to pay more for fuels. If that is the case, this assump- tion will lead to an overestimation of the quantity of fuel consumed by the poor and make the subsidy look more progressive than it would be in practice. The product of (1) and (2) is the proportion of poor house- holds that consume the petroleum product divided by the proportion of all households that consume the product. The ratio of the propor- tions of all poor and all users that would receive a subsidy, (3), is equal to unity because of the universality of petroleum product subsidies. The ratio of the average rate of subsidization for eligible poor and all eligible households, (4), is also unity because the subsidy is a constant amount per unit quantity, independent of the quantity purchased. The final term, (5), is the ratio of the average quantity consumed by eligible (user) poor households to that consumed by all eligible (user) households. For the purpose of this illustration, the ratio of expenditures is taken as a proxy for the ratio of quantities, assuming the price should be more or less the same for all users and therefore a common factor in the two sets of expenditures. Multiplying the percentage of households using the energy source by the average quantity consumed by user households is equal to the average consumption of all households in the group (users and non-users). Hence, the benefit-targeting indicator would be equal to the average consumption of the poor relative to the average consumption of all households. Simulated and E based on the foregoing are shown in figures 3.13 and 3.14, respectively. The poor were defined as those households with per capita income in the bottom 40 percent of the population. The simu- lation illustrates the projected performance of the two indicators were the government to introduce a flat-rate universal subsidy. The results show a clear pattern for . Focusing on kerosene con- sumed only by households, a subsidy on kerosene would be mildly pro- gressive in some countries and only modestly regressive in the others. The proportion of the poor that would not have benefited from such a Extractive Industries for Development Series 77 Figure3.13 SimulatedFlat-RateUniversalPriceSubsidy: Benefit-targeting indicator 1.6 1.4 Progressive 1.2 1.0 0.8 0.6 0.4 No data No data No data 0.2 0.0 a h a a an nd m a di ya es si di nd na In st la ne en bo ad ga ki ai et do K am gl Pa Th Vi U an In C B Kerosene LPG Gasoline + diesel Source: Authors' calculations. Figure3.14 SimulatedFlat-RateUniversalPriceSubsidy:E Exclusion rate 100 90 80 70 60 50 40 30 20 No data No data No data 10 0 a sh a a an nd m a di ya si di nd de na In st ila ne en bo ga ki a et a do K am gl Pa Th Vi U an In C B Kerosene LPG Gasoline + diesel Source: Authors' calculations. 78 Expenditure of Low-Income Households on Energy subsidy (because of not consuming kerosene) would generally be low, although there are notable exceptions. In Pakistan, Thailand, and Viet- nam, more than half the poor households did not use kerosene and the percentage of exclusion would have been correspondingly high. These findings should be interpreted with caution, because the data analyzed exclude kerosene consumed by other users. If diesel prices are higher, subsidized kerosene is inevitably diverted to the automotive sec- tor and added to diesel fuel because kerosene is a nearly perfect substi- tute for diesel (Kojima and Bacon 2001). When this diversion is taken into account, a kerosene subsidy can become highly regressive. A study in India found that subsidized kerosene consumed by households was evenly shared between the rich and the poor, but up to as much as half of subsidized kerosene might have been diverted to nonhousehold users (ESMAP 2003a). For LPG and gasoline/diesel, would be low, indicating high regres- sivity. The main exception to this pattern was Thailand where the much higher per capita consumption and income (twice that of the next high- est country, Indonesia, when measured at PPP) led to a wider-scale use of these fuels and a smaller difference in consumption between low-income and other households. The range for could be between 0 and 1.5, with a value above 1 signaling progressivity. Excluding Thailand, was 0.28 or smaller in every country for LPG and gasoline/diesel. Even in Thai- land, was 0.51 for LPG and 0.38 for gasoline/diesel. The exclusion fig- ures indicate that the great majority of the poor would be excluded from receiving any direct benefit from subsidies on LPG or gasoline and diesel. These simulated results based on the expenditure patterns of a variety of countries with different fuel-use patterns and price levels suggest that subsidies on transportation fuels and LPG are likely to be strongly regres- sive in countries with low to moderate income levels. Extractive Industries for Development Series 79 Chapter 4 Conclusions This paper analyzed nine household expenditure surveys conducted in the few years centered around 2005 in Asia and Africa. The importance of energy in household expenditure was confirmed in all the countries studied. When the imputed value of nonpurchased food was included, the share of energy expenditure in both rural and urban areas ranged from 6 to 14 percent; if the value of nonpurchased food was large, as in Cambodia, Kenya, Uganda, and Vietnam, and was excluded, the shares of rural energy expenditure were markedly higher, nearly doubling the expenditure share in rural Cambodia and Uganda. The share of expenditure devoted to modern energy sources was also generally large. In urban areas, with the exception of Uganda, the share ranged between 4 and 13 percent. In rural areas, the share was somewhat lower, ranging from about 2 percent in Bangladesh, Cambodia, Kenya, and Uganda up to 10 percent in Thailand. The use of modern sources in the total consumption of energy increased at higher income levels in both rural and urban areas. However, even for the top quintile, reliance on tra- ditional energy sources (biomass of all forms) was about 50 percent of all energy expenditure in Bangladesh, Cambodia, and Uganda. At the lowest quintile level in rural areas in Bangladesh, Cambodia, India, and Uganda, three-quarters of energy expenditure was on traditional fuels; with the exception of India, this proportion declined to less than 50 percent at the top quintile. Expenditure shares of petroleum products did not follow the pat- tern exhibited with modern energy, on which urban households spend more than rural ones. Rural households on average spent more than urban households in Bangladesh, Pakistan, Thailand, and Uganda. At the opposite end of the spectrum, urban households spent nearly twice as much as rural on petroleum products in Kenya and Vietnam. The share for petroleum products was greater than 5 percent in urban Thailand and Vietnam. For the bottom 40 percent, the spending on petroleum products was concentrated on kerosene, with the exception of Thailand, Vietnam, and urban Pakistan: for this latter group of households, the 80 Expenditure of Low-Income Households on Energy expenditure share ranged from 0.6 percent in urban Pakistan to 5.7 per- cent in urban Thailand. In those countries where the expenditure shares were already high in the middle of the last decade, subsequent oil price increases in 2007 and 2008 likely hit the poor hard. The analysis of energy expenditures by quintile group gave useful information on the relative importance of energy to the poorest house- holds in each country. In Bangladesh, Cambodia, India, and Uganda, the share of expenditure on energy in rural and urban areas was greatest for the bottom quintile; this was also true of urban areas in Indonesia and Thailand. Only Kenya, Pakistan, and Vietnam showed no such pattern, with the top quintile having the highest energy expenditure share in both rural and urban areas in Pakistan and Vietnam. The share of expenditure on modern energy generally rose with income in all the Asian countries. However, there was no clear evidence that the share for modern energy increased steadily with quintile level in Kenya and Uganda. There was no consistent pattern for the share of petroleum products with respect to income. A subset of the surveys allowed some comparisons to be made between rural and urban households at comparable income levels. Since rural and urban households may have quite different average expendi- tures, it was necessary to identify pairwise quintile comparisons where the total household expenditure levels were similar. This set of compari- sons showed that, although total energy expenditure shares were similar, their composition was quite different. As expected, the share for modern energy was higher, and the share for biomass lower, in urban areas at the same income level. The expenditure on electricity was higher in urban areas, but the share for gasoline and diesel was higher in rural areas. The picture for kerosene, regarded as predominantly a rural fuel, showed no consistent pattern. The limited evidence on LPG suggested that it is more important as an energy source to urban than rural households. The pattern of uptake revealed a large difference between the Asian and African countries surveyed. In Asia, with the exception of Cambodia, uptake of electricity for urban households was high above the bottom quintile, and almost universal at the top quintile. In Kenya and Uganda, urban uptake was less than 40 percent in the fourth quintile and less than 70 percent for the top quintile--even though total expenditure (excluding nonpurchased food) for the top urban quintile in Kenya ($914 in 2005 dollars at PPP) and Uganda ($715) was well above that for the top quintile in Bangladesh ($567), India ($471), or Indonesia ($452). It is apparent that uptake of electricity in urban Kenya and Uganda was Extractive Industries for Development Series 81 not limited by income alone but by the generally poor power sector infrastructure, a common problem in Sub-Saharan Africa. Rural uptake of electricity, while rising with income, was much lower than urban uptake everywhere except Thailand and Vietnam. However, in Kenya, the highest rural quintile had an uptake rate of only 21 percent; in Uganda, the highest rural quintile had an uptake rate of 15 percent; and in Cam- bodia, the fourth rural quintile had an uptake rate of 13 percent. Predict- ably, where a household had uptake of electricity, this was its primary source of energy for lighting. The uptake of kerosene, an important fuel for lighting where house- holds do not have electricity, was very high in all rural areas except in Pakistan, Thailand, and Vietnam--but still averaging 40 percent of rural households in both Pakistan and Vietnam--and in most countries, it did not decline much with income. In urban areas, kerosene was important at lower income levels, except again in Pakistan, Thailand, and Vietnam. LPG, by contrast, was not used much by low-income urban households, except in Vietnam, but its use increased at higher incomes. It was gener- ally not used much in rural areas, but was used by half the households in the top quintile in India, and by 70 percent of households in the top quintile in Vietnam. In Bangladesh and Pakistan, where natural gas is found in urban areas, the uptake rate increased rapidly with quintile level, reaching 50 percent of the top quintile in Bangladesh and 80 per- cent of this group in Pakistan. The share of urban households using biomass was high in most coun- tries, and exceeded 80 percent even at the top quintile level in Bangla- desh, Cambodia, and Uganda; in rural areas, the uptake rate was even higher. The top rural quintile had an uptake rate of more than 90 percent in five countries. Even in Thailand, which was the highest-income coun- try of the nine and had the lowest uptake rate, close to half of all rural households were using biomass. An interesting finding is the extent to which the rich in the survey countries were using biomass as their primary cooking fuel. These results should be interpreted with caution because households use a portfolio or fuel-stacking approach to cooking as income rises, and some house- holds using two or more sources of energy might not have found it easy to name their primary cooking fuel. Only in high-income countries do households use only gas, only electricity, or some combination of both for cooking. What is surprising is how many rich households said bio- mass was their primary cooking fuel. Households do not abandon bio- mass use altogether for a variety of reasons, including cost, the fact that 82 Expenditure of Low-Income Households on Energy modern fuel supplies are not always reliable or are time consuming to acquire where they live, and because of cooking practices and cultural preferences. Figure 4.1 shows the percentage of households in the top two quintiles in the six countries surveyed that reported biomass as their main cooking fuel. The persistent use of biomass even in urban areas and even as monthly household expenditure reached upwards of $800, valued in 2005 dol- lars at PPP, shows that steps to move households away from biomass will need to address a variety of concerns and problems. This is particularly true of LPG, which, apart from electricity, is the most realistic modern fuel option in rural areas for cooking and heating. There are economies of scale in LPG delivery, which also requires good road infrastructure. Figure 4.1 UseofBiomassasMainCookingFuelbyHouseholdsin theTopTwoQuintiles % of households 100 98 98 98 92 91 88 91 87 88 82 83 80 60 57 55 54 52 47 40 36 33 27 24 20 13 8.1 7.5 2.1 0 Cambodia India Kenya Pakistan Thailand Uganda Rural quintile 4 Rural quintile 5 Urban quintile 4 Urban quintile 5 Monthlyhouseholdexpendituresin2005$atPPP Quintile Cambodia India Kenya Pakistan Thailand Uganda Rural 4 291 215 391 445 755 311 Urban 4 337 252 395 518 853 329 Rural 5 757 335 675 623 1,406 558 Urban 5 993 471 993 904 1,588 784 Source: Authors' calculations. Note: There may have been data entry problems for biomass in rural Thailand during the survey. Extractive Industries for Development Series 83 If LPG is delivered only once every so many days, a backup cylinder (at $20­30 each) becomes essential. Delivery may also not be regular and reliable in a low-volume market, particularly if it is remote. The cost of cylinder management rises with declining cylinder size, but large cylin- ders mean large refill payments, a problem for households with irregular cash income flow. These challenges all too often exist against the back- drop of more readily available and much cheaper biomass, which is also suited for cooking traditional meals--for example, in an earlier Mexican study, many considered fuelwood essential for tortilla making (Masera, Saatkamp, and Kammen 2000). For policy to shift household fuel use from traditional biomass to cleaner cooking fuels, it would make sense to examine first how the urban rich could be persuaded to make this shift, because they are most likely to be able to afford it and have ready access to the LPG service infrastructure. If there are distortions in the market--a lack of competi- tion; an inadequate regulatory framework for the industry; poor enforce- ment of regulations; or any combination of these factors resulting in high prices, low quality of service, or both--the national energy ministry should take the lead in addressing them. The principal problems may, however, lie outside the energy sector: port congestion and slow customs clearance incurring high demurrage charges, bad road infrastructure, or the cost of doing business discouraging investment in bottling plants. Identifying and addressing these issues would require the involvement of other government ministries and agencies. Universal price subsidies for petroleum products are common, and many governments resorted to such subsidies as oil prices soared to his- toric heights in 2008 (Kojima 2009b). A recent estimate suggests that global pretax petroleum product subsidies increased from $57 billion in 2003 to $519 billion by mid-2008. After falling to $136 billion in mid- 2009, the total is projected to rise to $240 billion in 2010. When tax reductions are also included, the estimate for 2010 amounts to 1 percent or more of global gross domestic product (Coady and others 2010). Simulation of universal flat-rate price subsidies for petroleum products suggested that, when direct effects on households are considered, such subsidies would be regressive for LPG, gasoline, and diesel in all coun- tries where data were available, and for kerosene in half the countries. The exclusion rate is very high for LPG, gasoline, and diesel. It is low for kerosene in six countries where that energy source is widely used. These findings might suggest that a kerosene price subsidy could be pro-poor under certain circumstances. However, because kerosene is a nearly 84 Expenditure of Low-Income Households on Energy perfect substitute for diesel, when the price of subsidized kerosene is lower than that of diesel, the former is almost universally diverted to the automotive sector, benefiting businesses and higher-income households and potentially making the kerosene subsidy regressive even in countries where an analysis of household energy use might suggest it would be progressive. Further insights into subsidy policy may be gleaned from examining cash expenditures on food. The expenditure share of purchased food constituted one-third or more of total household expenditures for every urban quintile and as much as 60 percent for the bottom four urban quintiles in Cambodia and the bottom two in Bangladesh. In rural areas, cash expenditures on food in every quintile comprised 50 percent or more of total household expenditure in Cambodia and Indonesia and one-third or more in Bangladesh, India, Pakistan, and Vietnam. The indirect effects of higher energy prices--especially of gasoline and diesel which are used to transport food--on the welfare of these households would be high. In India, Indonesia, Kenya, Pakistan, Thailand, and Viet- nam, the ratio of expenditures on purchased food to those on petroleum products declined monotonically with income in both rural and urban areas. The expenditure shares of purchased food were at least an order of magnitude larger than those for petroleum products among the poor everywhere except Thailand, and were in fact two orders of magnitude larger among the urban poor in Pakistan. The indirect effects of higher oil prices through higher food prices could be larger than the direct effects of higher oil prices for many households. If that is the case, the policy response to help the poor cope with higher gasoline and diesel prices might more productively focus on assistance, ideally through tar- geted cash transfers, for food purchase--and, more generally, the basket of goods that the poor consume--than subsidizing fuel prices. This analysis of household surveys leaves several unanswered ques- tions and suggests avenues for future investigation. Some issues will require more detailed surveys, but others could be based on the surveys used here. Most questions relate to the changing patterns of energy con- sumption at progressively higher income levels. Although this study has confirmed the presence of fuel stacking, with households initially adding extra fuels as income rises without dropping other fuels, it is less clear how the quantities of the different fuels change. In particular, further information could be sought on the following: Extractive Industries for Development Series 85 · The extent to which the demand for electricity is income elastic, and evidence on how fuel-use levels change once electricity is added to the household portfolio · The extent to which the demand for LPG (and natural gas where avail- able) is income elastic, and evidence on how fuel-use levels change once LPG is added to the household portfolio · How the use of purchased versus nonpurchased biomass changes as income rises · Whether charcoal and other forms of biomass exhibit different expen- diture and use patterns at increasing income levels · How total household biomass usage may change as modern forms of energy are introduced into the household budget · Why the share of expenditure on gasoline and diesel is higher for rural households than for urban at similar income levels 86 Expenditure of Low-Income Households on Energy Appendix A SurveyDataSetsand EstimationofTotal HouseholdExpenditure The data used for this study cover countries in Asia and Sub-Saharan Africa. Nationally collected household-level socioeconomic survey data formed the basis of the analysis. These large surveys provide detailed household-level information on the consumption patterns of energy in every income class. A bottom-up approach, summing all expenditure components, was used to compute the total household expenditure in each country using a uniform procedure. Monthly total household expenditure was defined as the sum of the following: · Food expenditures: Expenditures for all food items using the 7-day recall period (if available) were prorated to a 30-day level, to provide a "monthly" value. · Nonfood expenditures: Expenditures on nonfood items were pro- rated to a 30-day level where required. Where both monthly and annual expenditures were provided, the following practice was fol- lowed: --Monthly: For expenditures on items such as fuel and light, enter- tainment, non-institutional medical, personal, toiletries, consumer services, rent, and commuting, the monthly recall values were used. --Annual: Expenditures on such items as clothing, bedding, foot- wear, education, medical (institutional), durable goods, life insur- ance premiums, vehicle insurance premiums, and membership fees were prorated from their annual levels to 30-day levels. The surveys provided information on certain large expenditure items (such as furniture; household appliances including refrigerators, air conditioners, washing machines, televisions, DVD players; expensive jewelry; automobiles; personal computers; ceremonies; and taxes and Extractive Industries for Development Series 87 cesses). Expenditures on these items--which ranged from 0 percent to greater than 90 percent of total expenditure in some cases, and which averaged about 1­5 percent of total household expenditure--were removed. This was done for two reasons: first, to avoid misrepresentation of households in their respective income quintiles since large expendi- tures are not made regularly, and second, to enable cross-country com- parison fuel shares given that the various categories of large expenditure items were treated differently, in that they were included in some surveys but not in others. The data set was inspected for the presence of any outliers in energy consumption values. This inspection involved examining households whose consumption levels and shares in total household expenditure for energy-related variables (kerosene, LPG, gasoline, diesel, electricity, natu- ral gas, coal, firewood, charcoal, other biomass) were very different from what could be considered a reasonable household consumption level based on the levels of consumption of other households in the survey. The analysis was based on alternative definitions of total household expenditure: · For all countries, a measure of total household expenditure excluding durable items as explained above · For those countries (Cambodia, Kenya, Uganda, and Vietnam) where the value of nonpurchased food formed a substantial portion of total expenditure, a second measure excluding this item from the total The data sets were divided into five population quintiles based on monthly per capita expenditure levels derived from the above two defini- tions of total household expenditure. Each quintile contained the same number of individuals and not households. As well as national quintiles, separate urban and rural quintile groups were defined by drawing house- holds from the national quintile depending on whether they resided in an urban or rural area. The numbers of people in these quintile groups were not the same and depended on the relative numbers of urban and rural households in the nationally defined quintile. Average expenditures of energy-related variables were computed at the household level. The various categories of energy sources identified in the different surveys were aggregated where necessary to the follow- ing items to allow a comparison among surveys: kerosene, LPG, gasoline and diesel, electricity, natural gas, and biomass. The expenditure on total petroleum products was defined as the sum of a household's expendi- tures on kerosene, LPG, gasoline, and diesel. Expenditure on modern 88 Expenditure of Low-Income Households on Energy energy was defined as the sum of expenditures on petroleum products, natural gas (where available), and electricity. Expenditure on biomass was defined as the sum of expenditures on firewood, charcoal, and other forms of biomass. Expenditure on total energy products was defined as aggregate expenditures on all fuels, including biomass, and electricity. Where the same item appears in more than one question, it may be possible to check internal consistency. For example, the percentage of households using a specific source of energy as their primary source for an activity, say lighting, should be the same as or smaller than the per- centage that use that energy source for any number of activities, of which lighting is one. Defining "use" to mean reporting positive expenditures (cash or imputed) led to internal inconsistencies, violating this simple rule, for the uptake of electricity in Kenya and Uganda, of biomass in Kenya, and of LPG in Thailand. For all cases, alternative means of calcu- lating uptake were used for internal consistency. Where expenditures are lumpy and a given energy source appears to be paid for less frequently than once a month--such as LPG in Thailand and possibly electricity in Kenya and Uganda--average expenditures are not necessarily affected as long as the sample size is large enough: zero expenditures of some households will be offset by others that paid for the energy source to cover more than one month of consumption; on aver- age, the expenditures should be equal to what households would spend for one month's consumption. However, the share of households using the energy source would be underestimated if positive expenditures were taken as an indication of uptake. Bangladesh(HIES2005) The data for this study are from the Household Income and Expenditure Survey (HIES) conducted by the government's Bureau of Statistics. The fieldwork for this survey was conducted between January and December 2005. The sample size was 10,054 households, 64 percent of which were based in rural areas. Based on the weights provided in the survey, this corresponded to 139 million people, of whom 75 percent were in rural areas, and 29 million households. The actual population of Bangladesh in 2005 was 153 million, and the proportion of the population living in rural areas in the same year was 74 percent (World Bank 2009). The large expenditure items, excluded from the definition of total household expenditures, averaged slightly more than 2 percent of aggre- gate total household expenditure. The value of nonpurchased food Extractive Industries for Development Series 89 accounted for 15 percent of aggregate total expenditure for rural house- holds and 4 percent for urban households. Overall, 10,080 households were interviewed. The cleaned data set contained observations from 10,054 households after excluding those with duplicate observations, or for which there was no information on major expenditure categories. Nine households that had reportedly spent more than 35 percent of their total monthly household expenditure on either gasoline or diesel were also excluded. Data on household expenditure on kerosene, firewood, electricity, gas, gasoline, diesel, lubricant oil, "coal and charcoal," cow dung, jute sticks, other agricultural wastes (paddy, hag, pressed sugarcane, dried corn plants), and other unclassified fuels which included matches and candles were available as were the imputed values of all nonpurchased fuels. The category "gas" represents both natural gas and biogas since the survey questionnaire combined both expenditures. Expenditure on total petroleum products was defined as the sum of a household's expenditures on kerosene, diesel, gasoline, and lubricant oil. Expenditure on biomass was defined as the sum of expenditures on firewood, dungcake, jute sticks, other agricultural wastes (paddy, hag, pressed sugarcane, dried corn plants) and coal and charcoal. Expenditure on modern energy was defined as the aggregate of expenditures on total petroleum products, natural gas, and electricity. Total expenditure on energy sources was defined as the aggregate of expenditures on modern energy, biomass, and other unclassified fuels which included matches and candles. Cambodia(CambodiaSocio-EconomicSurvey, 2003­04) The data for this study are from the Cambodia Socio-Economic Sur- vey conducted by the National Institute of Statistics of the Cambodian Ministry of Planning. The fieldwork for the survey was conducted between October 2003 and January 2005. The sample size was 14,572 households, 80 percent of which were based in rural areas. Based on the weights provided in the survey, this corresponded to approximately 12.6 million people, of whom 85 percent were in rural areas, and 2.5 million households. The actual population of Cambodia in 2003 and 2004 was 13.5 million and 13.7 million, respectively. The proportion of the population living in rural areas in these years was 81 percent (World Bank 2009). 90 Expenditure of Low-Income Households on Energy Large expenditure items, which averaged slightly more than 5 percent of total household expenditures, were removed. The value of nonpur- chased food comprised 20 percent of the budget of rural households and 7 percent of that of urban households. Overall, 15,000 households were interviewed. The cleaned data set contained observations from 14,572 households after excluding house- holds that had missing or differently coded information on either food expenditures or most of the other expenditure categories. Thirty-three households were removed because their expenditure on electricity, kero- sene, or firewood comprised more than 40 percent of their total expen- diture. However, 16 households that gave no information on their main lighting or cooking fuel were retained for the rest of the analysis since fuel consumption information was available. Data on household expenditures on kerosene, firewood, charcoal, electricity, LPG, batteries, and other unspecified fuels were available. Imputed values of any nonpurchased fuels were not reported. Expendi- tures on diesel and gasoline were also not reported. Expenditure on total petroleum products was defined as the sum of a household's expenditures on kerosene and LPG. Expenditure on modern energy was defined as the sum of expenditures on petroleum products and electricity. Expenditure on biomass was defined as the sum of cash expenditures on firewood and charcoal. Total expenditure on energy sources was defined as the aggregate of expenditure on modern energy, biomass, batteries, and other unclassified fuels. India(NationalSampleSurvey,2004­05) The data for this study are from the 61st round of the National Sample Survey of India conducted by the government's National Sample Survey Organization. The fieldwork was conducted between July 2004 and June 2005. The sample size was 121,630 households, 64 percent of which were based in rural areas. Based on the weights provided in the survey, this corresponded to 959 million people, of whom 75 percent were in rural areas, and 203 million households. The actual population of India in 2004 and 2005 was 1.08 billion and 1.09 billion, respectively. The proportion of the population in rural areas in these years was 71 percent (World Bank 2009). Large expenditure items, which averaged slightly more than 2 percent of total household expenditures, were removed. The value of nonpur- Extractive Industries for Development Series 91 chased food comprised 11 percent of the budget of rural households and 1 percent of that of urban households. Overall, 124,644 households were interviewed. The cleaned data set contained observations from 121,630 households: 3,014 households that reported having paid prices that were far from those prevailing for the rest of the sample for LPG, kerosene, or electricity were dropped. Data on household expenditures on LPG, subsidized and rationed kerosene sold through the Public Distribution System (PDS), market (non-PDS) kerosene, diesel, gasoline, lubricant oil, coke, coal, "firewood and chips," dungcake, electricity, charcoal, candles, matches, gobar gas (biogas), and other unspecified fuels were available. The imputed val- ues of firewood and chips, dungcake, gobar gas, and other unspecified nonpurchased fuels were also provided. Expenditure on total petroleum products was defined as the sum of a household's expenditures on LPG, kerosene (PDS and non-PDS), diesel, gasoline, and lubricant oil. Expen- diture on biomass was defined as the sum of expenditures on firewood and chips, charcoal, dungcake, biogas, and other unspecified fuels. Expenditure on modern energy was defined as the sum of expenditures on petroleum products and electricity. Total expenditure on energy sources was defined as the aggregate of expenditures on modern energy, biomass, coke, coal, candles, and matches. Indonesia(SUSENAS,JanuaryPanelModule 2005) The data for this study are from the National Socio-Economic Survey (SUSENAS) conducted by BPS-Statistics Indonesia. The SUSENAS com- prises a series of large-scale socioeconomic surveys initiated in the 1960s and fielded annually. The main survey consists of two parts: the core questionnaire, which is administered to about 200,000 households, and a module administered to about 65,000 households. There are three module sections--consumption, health, and social--each of which is repeated every three years. The data for this study were taken from the consumption module, which provides detailed information on house- hold-level expenditure patterns. A panel version of the consumption module is administered in January to March every year to about 10,000 households. The panel consumption module for 2005 was used because the full module data set for the survey conducted later that year was not avail- able. The fieldwork for the panel consumption module was conducted 92 Expenditure of Low-Income Households on Energy between January and March 2005. The sample size was 9,925 house- holds, 57 percent of which were based in rural areas. Using the weights provided in the survey, this corresponded to 206 million people, of whom 55 percent were in rural areas, and 52 million households. The actual population of Indonesia in 2005 was 221 million, and the propor- tion of the population living in rural areas in the same time period was 52 percent (World Bank 2009). Large expenditure items, which averaged slightly more than 3 percent of total household expenditures, were removed. The value of nonpur- chased food accounted for 11 percent of the budget of rural households and 3 percent of that of urban households. Overall, 10,575 households were interviewed. The cleaned data set contained observations from 9,925 households. Seventy-four households were dropped because their expenditure on a particular fuel appeared to be an outlier, while another 576 households that seem to have paid prices for LPG, kerosene, gasoline, or electricity that were far from those prevailing, or that lacked data on important expenditure items, were dropped. Data on household expenditures were available for LPG, kerosene, city gas (natural gas), automotive diesel, automotive gasoline, coal/briquette, "firewood and other fuels," and electricity. "Firewood and other fuels" included biomass fuels not included in other categories. No data were included for the value of nonpurchased fuels. The questionnaire asked separately about fuels used for electricity generation, but the authors of this study were not able to obtain the data; hence, fuels used for power generation are not included in the results presented here. Expenditure on total petroleum products was defined as the sum of a household's expenditures on kerosene, LPG, automotive diesel, and auto- motive gasoline. Expenditure on modern energy was defined as the sum of expenditures on petroleum products, city gas, and electricity. Expen- diture on biomass was defined as expenditures on firewood and other fuels. Total expenditure on energy sources was defined as the aggregate of expenditures on modern energy, biomass, and coal/briquette. Kenya(KenyaIntegratedHouseholdBudget Survey2005­06) The data for this study are from the Kenya Integrated Household Budget Survey conducted by the Kenya National Bureau of Statistics within the Ministry of Planning and National Development. Extractive Industries for Development Series 93 The fieldwork for this survey was conducted between May 2005 and April 2006. The sample size was 12,996 households, 65 percent of which were based in rural areas. Based on the weights provided in the survey, this corresponded to 35 million people, of whom 80 percent were in rural areas, and 6.9 million households. The actual population of Kenya in 2005 and 2006 was 36 million and 37 million, respectively. The proportion of the population living in rural areas in these years was 79 percent (World Bank 2009). Large expenditure items, which averaged slightly more than 1 percent of total household expenditures, were removed. The value of nonpur- chased food comprised 25 percent of the budget of rural households and 7 percent of that of urban households. Overall, 13,114 households were interviewed. The cleaned data set contained observations from 12,996 households after excluding house- holds that had duplicate or inconsistent observations, lacked informa- tion on food expenditures or some other expenditure category, or whose expenditure on an individual energy source comprised more than 40 per- cent of their total household expenditure (21 households). Data on household expenditures on kerosene/paraffin, electricity, LPG (referred to as gas in the survey questionnaire), firewood, charcoal, other unspecified cooking fuels, gasoline, and diesel were available. The imputed values of all nonpurchased fuels were also available. Expenditure on total petroleum products was defined as the sum of a household's expenditures on kerosene, LPG, diesel, and gasoline. Expen- diture on modern energy was defined as the sum of expenditures on petroleum products and electricity. Expenditure on biomass was defined as the sum of expenditures on firewood and charcoal; there was no sepa- rate category for other forms of biomass (animal and agricultural wastes) in section J of the survey on fuel and power expenditures. Total expen- diture on energy sources was defined as the aggregate of expenditures on modern energy, biomass, and other unspecified cooking fuels. Comparison of reported expenditures and quantities consumed for firewood in section J showed that many households assigned a value of zero to nonpurchased biomass. To define consumption of biomass in order to calculate uptake, section H of the survey--which asked whether households used purchased firewood, collected firewood, animal wastes, straws and stalks, and charcoal--was used. Because animal wastes may be used for housing, only animal wastes used for boiling, heating, and cooking were considered for this purpose. The differences are shown in table A.1. 94 Expenditure of Low-Income Households on Energy TableA.1 PercentageofHouseholdsUsingBiomassinKenya Positiveexpenditures Consumptionof Useoffirewood, onfirewoodand firewoodand charcoal,&animal& Quintile charcoal charcoal agriculturalwastes Rural Urban Nat'l Rural Urban Nat'l Rural Urban Nat'l 1 19 52 20 58 66 58 99 86 99 2 31 71 34 59 76 60 99 92 98 3 35 72 41 58 77 61 98 83 96 4 49 76 56 66 79 69 98 86 95 5 54 53 54 67 54 60 93 61 76 All 38 61 44 62 63 62 97 70 91 Source: Authors' calculations. Similarly, responses in sections J and H were compared to estimate the uptake of electricity. Table A.2 shows the results based on positive expenditures on electricity as reported in section J; the percentage of households citing grid electricity as one of the sources of electricity in section H; and the percentages of households citing grid electricity, own, a neighbor's, or the community's electricity generation, or solar power among their sources of electricity in section H. It is clear that expendi- tures during the preceding 30 days on electricity cannot be equated with the uptake of electricity, and the last set of numbers were used instead in tables 3.8, 3.9, and B.11. TableA.2 PercentageofHouseholdsUsingElectricityinKenya Grid,owngeneration, Positiveexpenditures Gridelectricityasa orsolarpanelsas Quintile onelectricity sourceofelectricity sourcesofelectricity Rural Urban Nat'l Rural Urban Nat'l Rural Urban Nat'l 1 0.1 0.2 0.1 0.3 0.2 0.3 0.6 0.5 0.6 2 0.4 12 1.3 0.6 16 1.7 1.2 16 2.2 3 0.6 9.2 2.0 1.6 22 4.8 3.3 23 6.3 4 1.9 18 5.7 4.2 38 12 8.1 39 15 5 6.2 37 23 13 65 40 21 66 45 All 1.8 28 8.3 3.7 51 16 6.6 52 18 Source: Authors' calculations. Extractive Industries for Development Series 95 Pakistan(HIES2004­05) The data for this study are from the 2004­05 HIES, conducted by Paki- stan's Federal Bureau of Statistics. The fieldwork for this survey was con- ducted between July 2004 and June 2005. The sample size was 14,700 households, 61 percent of which were based in rural areas. Based on the weights provided in the survey, this corresponded to 130 million people, of whom 68 percent were in rural areas, and 19 million households. The actual population of Pakistan in 2004 and 2005 was 152 million and 156 million, respectively. The proportion of the population living in rural areas in these years was 65 percent (World Bank 2009). Large expenditure items, which averaged slightly less than 1 percent of total household expenditures, were removed. The value of nonpurchased food comprised 14 percent of the budget of rural households and 2 per- cent of that of urban households. Overall, 14,744 households were interviewed. The cleaned data set contained observations from 14,700 households after excluding 44 households for which there was no information on food expenditures, fuel expenditures, other nonfood expenditures, or household size. Data on household-level expenditures on LPG, kerosene, natural gas, coal, firewood, dungcake, electricity, charcoal, other forms of biomass (bagasse, cotton sticks, sawdust, shrubs, weeds, tobacco sticks, and so on), and "diesel and gasoline" were available. The imputed values of all nonpurchased fuels were also available. Expenditures on diesel and gaso- line along with minor repairs such as punctures were reported as a single item. Expenditure on total petroleum products was defined as the sum of a household's expenditures on LPG, kerosene, and the combined expenses for "diesel and gasoline." Expenditure on biomass was defined as the sum of expenditures on firewood, charcoal, dungcake, and other forms of bio- mass. Expenditure on modern energy was defined as the sum of expen- ditures on petroleum products, natural gas, and electricity. Total expen- diture on energy sources was defined as the aggregate of expenditures on modern energy, biomass, coal, candles, and matches. Thailand(HouseholdSocio-EconomicSurvey 2006) The data for this study are from the Household Socio-Economic Sur- vey conducted by Thailand's Economic and Social Statistics Bureau within the National Statistical Office. The fieldwork for the survey was 96 Expenditure of Low-Income Households on Energy conducted in calendar year 2006. The sample size was 44,888 house- holds, 38 percent of which were based in rural areas. Based on the weights provided in the survey, this corresponded to 60 million people, of whom 70 percent were in rural areas, and 18 million households. The actual population of Thailand in 2006 was 63 million. The proportion of the population living in rural areas in the same time period was 67 per- cent (World Bank 2009). Large expenditure items, which averaged slightly less than 3 percent of total household expenditures, were removed. The value of nonpurchased food comprised 11 percent of the budget of rural households and 4 per- cent of that of urban households. Overall, 44,918 households were interviewed. The cleaned data set contained observations from 44,888 households because the households that had duplicate observations, or for which there was no information on food expenditures, were dropped. In addition, 27 households whose expenditure on unleaded gasoline (octanes 91 or 95), high-speed diesel, automotive LPG, or electricity comprised more than 40 percent of their total household expenditure were excluded. Data on household expenditures on electricity, cooking gas (LPG), "gas (LPG) for other purposes," "charcoal and wood," kerosene, unleaded gasoline (octane 91), unleaded gasoline (octane 95), gasohol (mixture of gasoline and ethanol), automotive compressed natural gas, automotive LPG, and high-speed diesel were available. The imputed values of non- purchased fuels were also available. Expenditure on gasoline and diesel was defined as the sum of expen- ditures on unleaded gasoline (octanes 91 and 95), gasohol, and high- speed diesel. Expenditure on total petroleum products was defined as the sum of a household's expenditures on cooking gas (LPG), LPG gas for other purposes, kerosene, automotive LPG, and gasoline and diesel. Expenditure on biomass was defined as expenditures on "charcoal and wood." Expenditure on modern energy was defined as the sum of expen- ditures on petroleum products, automotive compressed natural gas, and electricity. Total expenditure on energy sources was defined as the aggre- gate of expenditures on modern energy and biomass. The percentage of households reporting positive expenditures for LPG was significantly smaller than that reporting LPG as their main cooking fuel; that in turn was smaller than the percentage that owned an LPG cooking stove. One possible explanation is that the size of the LPG cyl- inder commonly used by households is 15 kilograms, which lasts more than a month. The questionnaire asked about monthly expenditures on Extractive Industries for Development Series 97 LPG, but if households were asked how much they had spent on LPG during the preceding 30 days, many might have reported zero expendi- tures. Because the differences are very large, the uptake tables report the percentages of households owning LPG cooking stoves rather than those reporting positive expenditures. Uganda(UgandaNationalHouseholdSurvey 2005­06) The data for this study are from the Uganda National Household Survey conducted by the Uganda government's Bureau of Statistics. The field- work for the survey was conducted between May 2005 and April 2006. The sample size was 7,414 households, 77 percent of which were based in rural areas. Based on the weights provided in the survey, this corre- sponded to 27 million people, of whom 85 percent were in rural areas, and 5.2 million households. The actual population of Uganda in 2005 and 2006 was 29 million and 30 million, respectively. The proportion of the population living in rural areas in these years was 87 percent (World Bank 2009). Large expenditure items, which averaged slightly more than 1 percent of total household expenditures, were removed. The value of nonpur- chased food comprised 31 percent of the budget of rural households and 7 percent of that of urban households. Overall, 7,421 households were interviewed. The cleaned data set contained observations from 7,414 households because those households that had duplicate observations, or for which there was no information on either food expenditures or some other expenditure categories, were dropped. Also, seven households whose expenditure on gasoline and diesel comprised more than 40 percent of their total household expendi- ture were excluded. Data on household expenditures on paraffin (kerosene), firewood, charcoal, electricity, generator/lawn mower fuels, and "diesel and gaso- line" were available. The imputed values of nonpurchased fuels were also available. The questionnaire asked about expenditures on fuels used to operate lawn mowers and electricity generators, but the type of fuel was not specified; this study assumed that the fuels were either gasoline or diesel. Expenditure on total petroleum products was defined as the sum of a household's expenditures on kerosene, generator/lawn mower fuels, and combined expenses for diesel and gasoline. Expenditure on biomass was 98 Expenditure of Low-Income Households on Energy defined as the sum of expenditures on firewood and charcoal. Expendi- ture on modern energy was defined as the sum of expenditures on petro- leum products and electricity. Total expenditure on energy sources was defined as the aggregate of expenditures on modern energy and biomass. For the top rural quintile and all urban quintiles except the second, the percentages of households reporting electricity as their primary lighting source were higher than those reporting positive expenditures on electricity. Unlike in Kenya, the household survey in Uganda did not contain sufficient supplementary questions to probe this discrep- ancy further. The only related question in the survey asked whether the household owned one or more solar panels. Only 0.2 percent of house- holds owned them. To compute the uptake rate for electricity in Uganda, the higher of the two--the percentage of households reporting positive expenditures or that citing electricity as the primary lighting source--was selected for each quintile. The rural and urban averages, as well as all national quintiles, were computed based on the uptake rates for the rural and urban quintiles. Vietnam(LivingStandardMeasurementSurvey 2006) The data for this study are from the Vietnam Household Living Standards Survey conducted by the General Statistics Office of Vietnam, with tech- nical assistance from the World Bank. The fieldwork for the survey was conducted between January and December 2006. The sample size was 9,127 households, 75 percent of which were based in rural areas. This corresponded to 82 million people, of whom 73 percent were in rural areas, and 19 million households. The actual population of Vietnam in 2006 was 84 million, with 73 percent living in rural areas (World Bank 2009). Large expenditure items, which averaged 14 percent of total house- hold expenditures, were removed. No values for imputed rent were pro- vided in the data set. The value of nonpurchased food comprised 16 per- cent of the budget of rural households and 3 percent of that of urban households. Overall, 9,189 households were interviewed. The cleaned data set contained observations from 9,127 households, because households with duplicate observations or for which there was no information on food expenditures or inconsistent information in some other expenditure cat- egory, as well as four households whose expenditure shares for gasoline Extractive Industries for Development Series 99 or "coal and firewood" comprised more than 40 percent of their total household expenditure, were removed. Data on household-level expenditures on LPG, paraffin (kerosene), "gasoline and lubricants," agricultural by-products (biomass), electricity, and "coal and firewood" were available. Expenditure on diesel was not recorded. The imputed values of all nonpurchased fuels were also avail- able. Expenditure on total petroleum products was defined as the sum of a household's expenditures on LPG, kerosene, and gasoline and lubricants. Expenditure on biomass was defined as the sum of expenditures on "coal and firewood," and agricultural by-products (biomass). Expenditure on modern energy was defined as the sum of expenditures on petroleum products and electricity. Total expenditure on energy sources was defined as the aggregate of expenditures on modern energy and biomass. 100 Expenditure of Low-Income Households on Energy Appendix B AdditionalResults This appendix provides results beyond those presented in the main report, notably material that highlights differences when nonpurchased food is included/excluded from total expenditure in Cambodia, Kenya, Uganda, and Vietnam. · Tables B.1­B.4 show total household expenditure with and without nonpurchased food for rural and urban quintiles in Cambodia, Kenya, Uganda, and Vietnam. · Table B-5 shows national statistics for shares of expenditure on energy, food, and transport. · Tables B.6­B.9 show expenditure shares when total household expen- ditures exclude nonpurchased food in Cambodia, Kenya, Uganda, and Vietnam, respectively. · Table B.10 shows national statistics for expenditure shares by quintile. · Table B.11 shows national statistics for shares of households consum- ing energy, food, and transport. · Tables B.12­B.15 show shares of households in Cambodia, Kenya, Uganda, and Vietnam, respectively, consuming energy, food, and trans- port when expenditures on nonpurchased food are excluded from total household expenditure. · Table B.16 shows expenditure shares when only those households reporting positive expenditures on a given item are considered. · Table B.17 shows the main energy source for cooking by quintile; tables B.18 and B.19 show the main lighting source. Extractive Industries for Development Series 101 TableB.1 TotalHouseholdExpenditurebyQuintileIncludingand ExcludingNonpurchasedFood:Cambodia(2005$atPPP) Including Excluding Including Excluding Rural nonpur- nonpur- Urban nonpur- nonpur- quintile chasedfood chasedfood quintile chasedfood chasedfood 1 140 85 1 140 99 2 187 139 2 199 143 3 228 185 3 256 206 4 291 257 4 337 297 5 757 749 5 993 964 All 310 266 All 650 631 Source: Authors' calculations. TableB.2 TotalHouseholdExpenditurebyQuintileIncludingand ExcludingNonpurchasedFood:Kenya(2005$atPPP) Including Excluding Including Excluding Rural nonpur- nonpur- Urban nonpur- nonpur- quintile chasedfood chasedfood quintile chasedfood chasedfood 1 139 82 1 126 83 2 229 158 2 202 148 3 295 223 3 276 233 4 391 334 4 395 336 5 675 646 5 993 914 All 343 273 All 729 689 Source: Authors' calculations. TableB.3 TotalHouseholdExpenditurebyQuintileIncludingand ExcludingNonpurchasedFood:Uganda(2005$atPPP) Including Excluding Including Excluding Rural nonpur- nonpur- Urban nonpur- nonpur- quintile chasedfood chasedfood quintile chasedfood chasedfood 1 118 63 1 127 72 2 181 109 2 201 109 3 240 161 3 247 178 4 311 239 4 329 271 5 558 495 5 784 715 All 280 205 All 586 557 Source: Authors' calculations. 102 Expenditure of Low-Income Households on Energy TableB.4 TotalHouseholdExpenditurebyQuintileIncludingand ExcludingNonpurchasedFood:Vietnam(2005$atPPP) Including Excluding Including Excluding Rural nonpur- nonpur- Urban nonpur- nonpur- quintile chasedfood chasedfood quintile chasedfood chasedfood 1 179 126 1 199 148 2 239 192 2 260 214 3 288 258 3 325 293 4 377 353 4 422 400 5 568 559 5 749 737 All 304 264 All 556 547 Source: Authors' calculations. TableB.5 SharesofHouseholdExpenditureonVariousEnergy Sources,Food,andTransport:AllHouseholds(%) Bangladesh Cambodia Indonesia Pakistan Vietnam Thailand Uganda Kenya India Expenditureitem Kerosene 1.0 1.0 1.5 2.3 2.1 0.3 0.0 1.5 0.3 LPG ND 0.2 1.1 0.2 0.1 0.2 0.6 ND 2.6 Gasoline and diesel 0.1 ND 0.8 1.0 0.2 1.0 6.1 0.2 3.1 Petroleumproducts 1.1 1.2 3.4 3.8 2.5 1.6 6.7 1.7 5.9 Electricity 1.1 0.8 2.4 3.4 0.2 3.8 3.1 0.4 3.0 Natural gas 0.3 NA NA 0.0 NA 0.6 0.0 NA NA Modernenergy 2.5 2.0 5.8 7.2 2.7 6.0 9.8 2.0 9.0 Biomass 4.7 4.8 5.4 1.6 1.4 3.1 0.6 4.4 3.1 Totalenergy 7.3 6.8 12 8.8 4.1 9.0 10 6.5 12 Purchased food 49 52 47 54 36 42 35 29 39 Nonpurchased food 12 18 7.9 7.8 21 10 8.7 24 12 Totalfood 61 70 55 62 57 52 44 53 51 Transport 2.5 0.2 2.4 2.4 3.2 3.0 1.7 2.0 0.6 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel. Nonpur- chased items, including cashfree biomass, are included. Extractive Industries for Development Series 103 TableB.6 SharesofHouseholdExpenditureonVariousEnergy Sources,Food,andTransportinCambodia:AllHouseholds(%) Petro- Gaso- Nat- leum Energy Quin- Kero- line& Elec- ural Bio- prod- Food, Trans- tile sene LPG diesel tricity gas mass ucts Modern Total NP port 1 2.9 0.0 ND 0.0 NA 12 2.9 3.0 15 60 0.3 2 1.9 0.0 ND 0.1 NA 7.9 1.9 2.0 10 67 0.3 3 1.4 0.0 ND 0.3 NA 6.5 1.4 1.7 8.2 68 0.2 Rural 4 1.1 0.1 ND 0.6 NA 5.3 1.2 1.8 7.1 67 0.3 5 0.5 0.4 ND 1.5 NA 3.1 0.9 2.4 5.6 56 0.1 All 1.6 0.1 ND 0.5 NA 7.0 1.7 2.1 9.2 64 0.3 1 2.7 0.0 ND 0.0 NA 9.2 2.7 2.8 12 67 0.1 2 1.8 0.0 ND 0.6 NA 8.4 1.8 2.4 11 67 0.0 1.6 2.9 9.2 Urban 3 1.4 0.1 ND 1.3 NA 6.2 70 0.4 4 0.7 0.3 ND 2.3 NA 4.7 1.0 3.2 7.9 67 0.1 5 0.1 1.3 ND 3.9 NA 1.7 1.4 5.3 7.0 49 0.1 All 0.6 0.8 ND 2.8 NA 3.7 1.4 4.2 8.0 58 0.1 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; NP = nonpurchased. Quintiles are based on total per capita expenditure excluding nonpurchased food. TableB.7 SharesofHouseholdExpenditureonVariousEnergy Sources,Food,andTransportinKenya:AllHouseholds(%) Petro- Gaso- Nat- leum Energy Quin- Kero- line& Elec- ural Bio- prod- Food, Trans- tile sene LPG diesel tricity gas mass ucts Modern Total NP port 1 3.7 0.0 0.0 0.0 NA 1.2 3.8 3.8 5.0 52 2.1 2 3.2 0.0 0.0 0.0 NA 1.8 3.2 3.2 5.0 51 3.0 3 2.7 0.0 0.0 0.0 NA 1.7 2.8 2.8 4.5 48 3.2 Rural 4 2.2 0.0 0.3 0.1 NA 2.2 2.5 2.6 4.7 43 4.0 5 1.8 0.3 0.6 0.2 NA 1.7 2.7 2.8 4.5 35 4.7 All 2.7 0.1 0.2 0.1 NA 1.7 3.0 3.0 4.8 46 3.3 1 4.4 0.0 0.0 0.0 NA 2.6 4.4 4.4 7.0 58 1.5 2 3.9 0.0 0.0 0.5 NA 4.8 3.9 4.4 9.3 54 2.3 4.1 4.5 8.3 Urban 3 4.1 0.0 0.0 0.4 NA 3.7 52 2.3 4 3.3 0.1 0.0 0.4 NA 3.0 3.4 3.8 6.8 49 4.9 5 2.5 0.6 0.8 0.9 NA 1.1 3.9 4.8 5.9 38 5.9 All 2.9 0.4 0.5 0.7 NA 1.9 3.8 4.5 6.4 42 5.2 Source: Authors' calculations. Note: NA = fuel not available. Quintiles are based on total per capita expenditure excluding nonpurchased food. 104 Expenditure of Low-Income Households on Energy TableB.8 SharesofHouseholdExpenditureonVariousEnergy Sources,Food,andTransportinUganda:AllHouseholds(%) Petro- Gaso- Nat- leum Energy Quin- Kero- line& Elec- ural Bio- prod- Food, Trans- tile sene LPG diesel tricity gas mass ucts Modern Total NP port 1 3.6 ND 0.0 0.2 NA 12 3.6 3.8 16 29 1.5 2 2.8 ND 0.0 0.2 NA 9.0 2.8 3.0 12 36 2.2 3 2.4 ND 0.2 0.1 NA 7.0 2.5 2.7 9.6 38 2.5 Rural 4 1.9 ND 0.4 0.2 NA 5.3 2.2 2.4 7.7 39 2.8 5 1.4 ND 0.6 0.6 NA 3.1 2.1 2.6 5.7 38 3.1 All 2.4 ND 0.2 0.2 NA 7.4 2.7 2.9 10 36 2.5 1 3.6 ND 0.0 0.1 NA 11 3.6 3.7 15 36 1.6 2 2.8 ND 0.0 0.0 NA 10 2.8 2.8 13 42 1.6 2.2 2.6 9.6 Urban 3 2.2 ND 0.0 0.4 NA 7.0 40 2.2 4 1.7 ND 0.0 0.7 NA 5.6 1.7 2.4 8.0 44 2.0 5 0.9 ND 0.6 1.5 NA 2.8 1.5 3.1 5.9 38 2.9 All 1.3 ND 0.4 1.2 NA 4.1 1.7 2.9 7.0 39 2.7 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; NP = nonpurchased. Quintiles are based on total per capita expenditure excluding nonpurchased food. TableB.9 SharesofHouseholdExpenditureonVariousEnergy Sources,Food,andTransportinVietnam:AllHouseholds(%) Petro- Gaso- Nat- leum Energy Quin- Kero- line& Elec- ural Bio- prod- Food, Trans- tile sene LPG diesel tricity gas mass ucts Modern Total NP port 1 0.6 0.1 1.7 3.3 NA 8.7 2.4 5.7 14 46 0.8 2 0.4 0.8 2.5 3.3 NA 5.7 3.7 7.0 13 47 0.8 3 0.3 2.0 3.2 3.2 NA 4.3 5.5 8.7 13 46 0.8 Rural 4 0.3 4.1 4.1 3.1 NA 2.6 8.5 12 14 44 0.7 5 0.2 4.8 4.6 3.0 NA 1.4 9.5 12 14 39 0.5 All 0.4 2.0 3.0 3.2 NA 5.0 5.4 8.6 14 45 0.7 1 0.7 0.8 1.7 4.1 NA 5.1 3.2 7.3 13 50 0.3 2 0.7 1.9 2.1 3.7 NA 4.4 4.6 8.3 13 53 0.4 7.7 12 14 Urban 3 0.5 3.8 3.3 3.9 NA 2.4 50 0.6 4 0.4 5.3 4.3 3.9 NA 1.4 10 14 15 44 0.5 5 0.1 5.0 5.0 4.3 NA 0.5 10 14 15 38 0.5 All 0.3 4.6 4.4 4.1 NA 1.3 9.3 13 15 42 0.5 Source: Authors' calculations. Note: NA = fuel not available; NP = nonpurchased. Quintiles are based on total per capita expenditure excluding nonpurchased food. Extractive Industries for Development Series 105 TableB.10 SharesofHouseholdExpenditureonVariousEnergy Sources,Food,andTransport,byQuintile:AllHouseholds(%) Petro- Gaso- Nat- leum Energy Food Quin- Kero- line& Elec- ural Bio- prod- Trans- tile sene LPG diesel tricity gas mass ucts Modern Total P NP Total port 1 1.5 ND 0.0 0.4 0.0 6.2 1.5 1.9 8.1 57 12 70 1.8 2 1.2 ND 0.0 0.6 0.1 5.7 1.2 1.9 7.7 53 15 68 2.1 Bangladesh 3 1.1 ND 0.0 1.1 0.1 5.3 1.1 2.3 7.6 50 14 64 2.5 4 0.8 ND 0.1 1.5 0.3 4.3 0.9 2.8 7.1 48 12 60 2.8 5 0.5 ND 0.3 1.9 0.9 2.5 0.8 3.6 6.1 40 7.1 47 3.1 All 1.0 ND 0.1 1.1 0.3 4.7 1.1 2.5 7.3 49 12 61 2.5 1 1.6 0.0 ND 0.1 NA 6.6 1.6 1.7 8.4 50 27 77 0.2 2 1.3 0.0 ND 0.2 NA 5.8 1.3 1.6 7.4 51 24 76 0.2 Cambodia 3 1.1 0.1 ND 0.4 NA 5.2 1.2 1.6 6.9 54 20 74 0.2 4 0.9 0.1 ND 0.8 NA 4.5 1.0 1.8 6.4 56 15 71 0.2 5 0.3 0.7 ND 2.2 NA 2.4 1.0 3.2 5.6 49 6.9 56 0.1 All 1.0 0.2 ND 0.8 ND 4.8 1.2 2.0 6.8 52 18 70 0.2 1 2.1 0.0 0.0 1.3 NA 8.7 2.2 3.4 13 54 8.5 63 1.5 2 1.9 0.2 0.1 1.8 NA 7.7 2.2 3.9 12 52 9.8 62 1.8 3 1.7 0.6 0.3 2.2 NA 6.5 2.5 4.7 12 49 11 59 2.3 India 4 1.5 1.4 0.7 2.7 NA 4.7 3.6 6.3 11 46 8.9 55 2.7 5 0.8 2.5 2.2 3.5 NA 1.4 5.6 9.1 11 38 3.4 41 3.2 All 1.5 1.1 0.8 2.4 NA 5.4 3.4 5.8 12 47 7.9 55 2.4 1 2.3 0.0 0.2 2.6 0.0 3.5 2.6 5.2 8.6 55 14 69 1.2 2 2.6 0.0 0.6 3.2 0.0 2.4 3.3 6.5 8.9 57 9.5 66 1.8 Indonesia 3 2.6 0.1 0.9 3.6 0.0 1.6 3.8 7.4 9.0 56 8.5 64 2.4 4 2.5 0.2 1.3 3.8 0.0 0.9 4.3 8.2 9.1 55 5.6 61 2.8 5 1.6 0.5 1.8 3.7 0.0 0.3 4.4 8.1 8.4 49 3.6 53 3.4 All 2.3 0.2 1.0 3.4 0.0 1.6 3.8 7.2 8.8 54 7.8 62 2.4 1 2.2 0.0 0.0 0.0 NA 1.0 2.2 2.2 3.1 39 30 70 1.5 2 2.2 0.0 0.0 0.1 NA 1.5 2.2 2.3 3.8 38 28 66 2.2 Kenya 3 2.3 0.0 0.0 0.1 NA 1.6 2.3 2.4 4.0 37 24 60 2.6 4 2.1 0.0 0.1 0.2 NA 1.8 2.2 2.4 4.3 36 19 55 3.3 5 1.9 0.4 0.7 0.5 NA 1.2 3.0 3.5 4.7 32 11 43 4.8 All 2.1 0.1 0.2 0.2 NA 1.4 2.5 2.7 4.1 36 21 57 3.2 106 Expenditure of Low-Income Households on Energy Petro- Gaso- Nat- leum Energy Food Quin- Kero- line& Elec- ural Bio- prod- Trans- tile sene LPG diesel tricity gas mass ucts Modern Total P NP Total port 1 0.5 0.1 0.1 3.4 0.3 4.4 0.7 4.4 8.8 48 10 59 2.6 2 0.4 0.1 0.3 3.6 0.3 3.9 0.8 4.8 8.7 45 12 57 2.8 Pakistan 3 0.4 0.2 0.4 3.7 0.4 3.5 1.0 5.1 8.7 43 12 55 3.0 4 0.3 0.3 0.9 4.0 0.7 2.7 1.5 6.2 8.9 41 11 52 3.2 5 0.1 0.4 2.5 4.2 0.9 1.7 3.1 8.2 9.9 36 7.6 43 3.3 All 0.3 0.2 1.0 3.8 0.6 3.1 1.6 6.0 9.0 42 10 52 3.0 1 0.0 0.3 4.6 3.2 0.0 1.6 5.0 8.2 9.8 38 17 54 0.8 2 0.0 0.7 5.7 3.2 0.0 1.0 6.4 9.6 11 38 12 50 1.0 Thailand 3 0.0 0.8 6.2 3.1 0.0 0.6 7.0 10 11 37 9.0 46 1.5 4 0.0 0.7 6.4 3.1 0.0 0.3 7.1 10 11 35 6.3 42 2.1 5 0.0 0.4 7.0 2.9 0.0 0.0 7.4 10 10 30 3.2 33 2.4 All 0.0 0.6 6.1 3.1 0.0 0.6 6.7 9.8 10 35 8.7 44 1.7 1 1.8 ND 0.0 0.1 NA 6.6 1.8 1.9 8.5 26 35 61 0.9 2 1.7 ND 0.0 0.1 NA 5.4 1.7 1.8 7.2 25 34 59 1.4 Uganda 3 1.6 ND 0.1 0.1 NA 4.6 1.7 1.8 6.4 27 31 58 1.7 4 1.4 ND 0.2 0.2 NA 4.0 1.6 1.8 5.9 30 25 55 2.2 5 1.1 ND 0.6 0.9 NA 2.6 1.6 2.5 5.1 33 12 44 2.6 All 1.5 ND 0.2 0.4 NA 4.4 1.7 2.0 6.5 29 24 53 2.0 1 0.4 0.3 1.2 2.4 NA 5.3 1.9 4.3 9.7 38 25 63 0.6 2 0.3 0.9 2.0 2.8 NA 4.4 3.3 6.0 11 39 18 57 0.6 Vietnam 3 0.3 2.2 2.9 2.9 NA 3.3 5.4 8.3 12 41 13 54 0.6 4 0.3 4.0 3.9 3.2 NA 2.2 8.2 11 14 40 7.3 47 0.6 5 0.1 4.7 4.7 3.7 NA 0.9 9.6 13 14 37 2.5 39 0.5 All 0.3 2.6 3.1 3.0 NA 3.1 5.9 9.0 12 39 12 51 0.6 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; P = pur- chased; NP = nonpurchased. Extractive Industries for Development Series 107 TableB.11 PercentageofAllHouseholdsConsumingVarious EnergySources,Food,andTransport Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 92 ND 0.6 12 0.4 99 92 94 100 92 69 2 88 ND 0.5 22 1.9 99 88 94 100 91 77 Bangladesh 3 86 ND 0.4 35 3.9 99 86 95 100 87 83 4 80 ND 1.0 52 7.1 99 81 93 100 81 87 5 63 ND 5.2 68 27 98 66 93 100 71 91 All 81 ND 1.6 39 8.4 99 82 94 100 84 82 1 90 0.2 ND 1.5 NA 93 90 91 99 83 4.6 2 88 0.5 ND 3.8 NA 93 88 90 100 84 5.3 Cambodia 3 85 1.4 ND 8.6 NA 94 86 91 100 79 6.6 4 77 4.2 ND 18 NA 94 80 90 100 74 8.0 5 45 29 ND 53 NA 86 73 92 100 47 13 All 76 8.0 ND 18 NA 92 83 91 100 72 7.8 1 95 0.5 0.5 32 NA 97 95 98 99 45 53 2 95 2.7 1.9 48 NA 98 96 99 100 50 68 3 91 8.8 4.4 60 NA 94 94 100 100 52 75 India 4 82 25 11 74 NA 81 94 100 100 47 79 5 46 66 36 92 NA 38 93 99 100 26 79 All 79 24 13 64 NA 78 94 99 100 43 72 1 89 0.2 4.9 69 0.4 82 89 98 100 82 43 2 90 0.3 13 84 0.2 66 91 99 100 74 52 Indonesia 3 92 1.8 20 91 0.2 51 94 100 99 68 59 4 93 5.2 31 94 0.2 34 96 100 100 54 62 5 80 24 46 94 1.2 15 95 98 100 39 70 All 89 7.1 25 87 0.5 47 93 99 100 62 58 1 76 0.0 0.1 0.6 NA 99 76 76 98 93 30 2 87 0.1 0.2 2.2 NA 98 87 87 100 96 45 Kenyaa 3 90 0.2 0.5 6.3 NA 96 90 90 99 95 54 4 88 0.9 1.5 15 NA 95 89 90 100 95 66 5 83 14 7.2 45 NA 76 90 92 100 94 77 All 85 4.3 2.5 18 NA 91 87 88 99 95 58 1 40 2.7 7.6 70 7.4 90 47 98 100 58 87 2 38 5.1 11 76 11 87 48 98 100 62 90 Pakistan 3 36 7.6 12 81 16 83 47 99 100 61 91 4 29 9.6 20 86 24 73 49 99 100 57 91 5 17 12 39 93 47 48 56 99 100 43 89 All 30 8.0 20 83 24 74 50 99 100 55 90 108 Expenditure of Low-Income Households on Energy Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 1.1 45 66 97 0.0 68 68 99 99 93 19 2 0.7 69 78 99 0.0 57 81 100 99 90 25 Thailandb 3 0.8 81 80 99 0.0 40 84 100 99 83 31 4 0.5 85 78 99 0.0 21 84 100 99 69 39 5 0.1 80 78 100 0.0 6.9 82 100 100 46 46 All 0.6 74 77 99 0.0 35 81 100 100 74 34 1 88 ND 0.0 1.1 NA 96 88 88 99 96 13 2 95 ND 1.1 2.0 NA 97 95 95 100 95 24 Ugandac 3 96 ND 1.2 3.5 NA 97 96 96. 99 92 32 4 94 ND 2.6 7.4 NA 97 94 96 99 85 39 5 82 ND 6.3 33 NA 86 83 93 99 65 50 All 91 ND 2.6 11 NA 94 91 94 99 82 36 1 44 3.4 27 88 NA 95 60 98 99 92 39 2 44 11 46 95 NA 93 73 100 100 89 47 Vietnam 3 41 27 59 96 NA 84 81 99 100 84 51 4 32 54 73 99 NA 65 90 100 100 70 51 5 19. 82 84 99 NA 36 96 100 100 47 51 All 35 38 60 96 NA 73 81 99 100 75 48 Source: Authors' calculations. Note: NA = fuel not available; ND = no question was asked concerning the fuel; P = pur- chased; NP = nonpurchased. a. For biomass, the percentages shown are a combination of those households reporting positive expenditures (upper three urban quintiles) and those reporting biomass as the primary cooking fuel (the remaining quintiles). b For LPG, the percentages are those households that own an LPG cook stove. c. Data for electricity are computed from tables 3.8 and 3.9. Extractive Industries for Development Series 109 TableB.12 PercentageofHouseholdsConsumingVariousEnergy Sources,Food,andTransportinCambodia Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 90 0.3 ND 0.6 NA 94 90 90 99 95 4.1 2 90 0.3 ND 1.7 NA 93 90 91 100 87 6.0 3 87 0.7 ND 4.6 NA 94 87 89 100 80 5.5 Rural 4 81 2.8 ND 14 NA 94 82 89 100 72 7.9 5 58 17 ND 38 NA 90 72 90 100 52 11 All 82 3.6 ND 11 NA 93 85 90 100 78 6.8 1 96 0.0 ND 0.9 NA 95 96 97 99 84 6.6 2 85 0.3 ND 10 NA 94 85 92 99 79 4.3 78 94 Urban 3 77 2.5 ND 23 NA 97 100 71 12 4 52 9.9 ND 51 NA 95 62 94 100 45 9.7 5 14 56 ND 90 NA 80 69 98 100 20 17 All 39 33 ND 64 NA 87 71 96 100 39 13 Source: Authors' calculations. Note: NA = fuel not available; ND = survey did not ask for information about the fuel; P = purchased; NP = nonpurchased. Quintiles are based on total per capita expenditure exclud- ing nonpurchased food. TableB.13 PercentageofHouseholdsConsumingVariousEnergy Sources,Food,andTransportinKenya Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 74 0.1 0.1 0.5 NA 99 74 74 98 97 26 2 88 0.0 0.2 1.4 NA 98 89 89 99 97 47 3 90 0.1 0.4 3.3 NA 98 90 91 100 98 56 Rural 4 89 0.8 2.3 8.7 NA 98 90 90 100 97 68 5 86 6.3 5.1 23 NA 92 89 90 100 96 78 All 86 1.2 1.4 6.6 NA 97 86 87 99 97 54 1 73 0.0 0.0 0.9 NA 95 73 73 98 87 25 2 91 0.0 0.0 8.0 NA 91 91 91 100 78 33 87 88 Urban 3 87 0.1 0.0 23 NA 83 99 82 46 4 86 0.7 0.4 33 NA 86 86 89 100 86 65 5 81 21 9.0 65 NA 62 90 93 100 90 79 All 83 14 5.8 52 NA 70 89 91 100 88 71 Source: Authors' calculations. Note: NA = fuel not available; ND = survey did not ask for information about the fuel; P = purchased; NP = nonpurchased. Quintiles are based on total per capita expenditure exclud- ing nonpurchased food. 110 Expenditure of Low-Income Households on Energy TableB.14 PercentageofHouseholdsConsumingEnergy,Food, andTransportinUganda Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 88 ND 0.3 1.8 NA 95 88 88 98 100 13 2 95 ND 0.2 1.5 NA 96 95 95 98 98 23 3 97 ND 1.8 2.2 NA 98 97 97 100 98 33 Rural 4 96 ND 2.9 4.1 NA 97 97 97 99 92 40 5 90 ND 6.6 15 NA 89 91 95 100 75 49 All 94 ND 2.3 4.6 NA 95 94 94 99 92 33 1 98 ND 0.0 5.7 NA 92 98 98 95 94 12 2 92 ND 0.0 0.0 NA 96 92 92 98 81 18 94 97 Urban 3 94 ND 0.3 5.8 NA 94 95 74 30 4 86 ND 1.2 14 NA 96 86 92 100 58 39 5 70 ND 5.9 46 NA 84 73 91 100 42 51 All 76 ND 4.3 34 NA 87 78 92 99 48 47 Source: Authors' calculations. Note: NA = fuel not available; ND = survey did not ask for information about the fuel; P = purchased; NP = nonpurchased. Quintiles are based on total per capita expenditure exclud- ing nonpurchased food. TableB.15 PercentageofHouseholdsConsumingVariousEnergy Sources,Food,andTransportinVietnam Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 46 1.3 28 88 NA 96 61 98 99 98 38 2 44 8.3 45 95 NA 96 72 99 100 94 49 3 42 23 59 96 NA 90 80 99 100 89 54 Rural 4 40 51 72 98 NA 73 91 100 100 80 56 5 28 74 78 99 NA 53 95 100 100 68 56 All 41 26 53 95 NA 85 77 99 100 88 50 1 40 7.4 32 85 NA 88 62 94 97 86 23 2 35 18 39 94 NA 77 68 99 100 71 37 85 100 Urban 3 33 45 62 99 NA 63 100 52 39 4 22 65 76 100 NA 44 91 100 100 42 43 5 12 90 88 100 NA 22 97 100 100 30 47 All 19 71 77 99 NA 38 91 100 100 40 44 Source: Authors' calculations. Note: NA = fuel not available; P = purchased; NP = nonpurchased. Quintiles are based on total per capita expenditure excluding nonpurchased food. Extractive Industries for Development Series 111 TableB.16 SharesofExpenditureonVariousEnergySources,Food, andTransport:UserHouseholds(%) Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 1.6 ND 3.9 3.4 4.3 6.2 1.6 2.0 57 14 2.6 2 1.4 ND 3.8 2.9 3.5 5.8 1.4 2.1 53 17 2.8 Bangladesh 3 1.2 ND 6.3 3.0 3.7 5.3 1.3 2.4 50 16 3.0 4 1.0 ND 7.8 3.0 4.4 4.3 1.1 3.0 48 14 3.3 5 0.7 ND 6.1 2.7 3.2 2.6 1.3 3.8 40 10 3.4 All 1.2 ND 6.0 2.9 3.5 4.8 1.4 2.7 49 14 3.0 1 1.8 5.7 ND 5.5 NA 7.1 1.8 1.9 50 33 4.3 2 1.5 4.8 ND 5.5 NA 6.2 1.5 1.7 52 29 3.2 Cambodia 3 1.3 4.1 ND 4.7 NA 5.5 1.4 1.7 54 25 3.1 4 1.1 3.5 ND 4.5 NA 4.8 1.3 2.0 56 21 2.8 5 0.7 2.3 ND 4.1 NA 2.8 1.4 3.4 49 14 1.0 All 1.3 2.5 ND 4.3 NA 5.2 1.5 2.2 52 25 2.4 1 2.2 7.1 5.9 4.0 NA 8.9 2.3 3.5 55 19 2.9 2 2.0 6.6 6.5 3.7 NA 7.9 2.3 4.0 52 19 2.7 3 1.9 6.4 6.3 3.7 NA 6.9 2.7 4.7 49 20 3.1 India 4 1.9 5.5 6.3 3.6 NA 5.8 3.9 6.4 46 19 3.5 5 1.8 3.8 6.0 3.8 NA 3.6 6.1 9.2 38 13 4.1 All 1.9 4.4 6.1 3.7 NA 6.9 3.6 5.9 47 18 3.3 1 2.6 2.4 4.5 3.7 2.9 4.2 2.9 5.3 56 17 2.8 2 2.8 3.6 4.6 3.8 3.1 3.7 3.6 6.6 57 13 3.5 Indonesia 3 2.8 3.8 4.5 3.9 4.0 3.1 4.0 7.4 56 12 4.0 4 2.7 3.5 4.1 4.1 4.4 2.6 4.5 8.2 55 10 4.6 5 2.0 2.2 3.8 3.9 2.2 2.0 4.6 8.3 49 9 4.8 All 2.6 2.5 4.1 3.9 2.7 3.4 4.0 7.3 54 13 4.1 1 2.9 -- 11 1.3 NA 4.8 2.9 2.9 40 32 5.0 2 2.5 5.2 7.6 4.4 NA 4.4 2.5 2.6 38 29 4.8 Kenya 3 2.6 7.4 5.8 3.5 NA 3.9 2.6 2.7 37 25 4.8 4 2.3 3.8 9.5 3.4 NA 3.3 2.5 2.7 36 20 5.1 5 2.3 3.0 9.3 2.1 NA 2.2 3.3 3.8 33 12 6.2 All 2.5 3.1 9.2 2.4 NA 3.3 2.8 3.0 36 22 5.4 1 1.1 2.3 1.7 4.9 3.6 4.9 1.4 4.5 48 18 3.0 2 1.1 2.7 2.6 4.7 3.1 4.5 1.7 4.9 45 19 3.2 Pakistan 3 1.0 2.6 3.6 4.6 2.8 4.2 2.1 5.2 43 19 3.3 4 1.0 3.3 4.6 4.6 2.7 3.7 3.1 6.2 41 19 3.5 5 0.8 3.3 6.4 4.5 2.0 3.5 5.4 8.3 36 18 3.7 All 1.0 3.0 5.1 4.6 2.4 4.2 3.1 6.1 42 19 3.4 112 Expenditure of Low-Income Households on Energy Petro- Gaso- Nat- leum Food Quin- Kero- line& Elec- ural Bio- prod- Modern Trans- tile sene LPG diesel tricity gas mass ucts energy P NP port 1 1.1 3.9 7.1 3.3 -- 2.3 7.3 8.3 38 18 4.2 2 1.6 3.3 7.3 3.2 -- 1.8 7.9 9.6 38 13 4.1 Thailand 3 1.0 2.6 7.7 3.2 -- 1.6 8.3 10 37 11 4.7 4 0.6 2.0 8.1 3.1 1.9 1.3 8.4 10 36 9.1 5.3 5 0.3 1.2 8.9 2.9 2.5 0.6 9.0 10 30 6.8 5.3 All 1.0 2.2 8.0 3.1 2.5 1.8 8.3 9.8 35 12 4.9 1 2.1 ND -- 6.9 NA 6.9 2.1 2.1 26 36 6.6 2 1.8 ND 3.7 5.3 NA 5.6 1.8 1.9 26 36 5.8 Uganda 3 1.6 ND 8.6 4.7 NA 4.8 1.7 1.9 28 33 5.2 4 1.5 ND 8.3 3.8 NA 4.2 1.7 1.9 30 29 5.5 5 1.3 ND 9.1 3.3 NA 3.0 2.0 2.7 33 18 5.3 All 1.6 ND 8.5 3.6 NA 4.7 1.8 2.2 29 29 5.4 1 0.9 8.9 4.4 2.8 NA 5.6 3.1 4.4 38 27 1.5 2 0.8 8.6 4.4 2.9 NA 4.7 4.5 6.1 39 21 1.3 Vietnam 3 0.8 8.2 4.9 3.0 NA 4.0 6.7 8.4 41 15 1.2 4 0.8 7.6 5.4 3.2 NA 3.4 9.1 11 40 10 1.1 5 0.6 5.7 5.6 3.7 NA 2.5 9.9 13 37 5.3 0.9 All 0.8 6.8 5.1 3.2 NA 4.3 7.3 9.0 39 17 1.2 Source: Authors' calculations. Note: NA = fuel not available; -- = no household reported expenditure; ND = no question was asked concerning the fuel; P = purchased; NP = nonpurchased. Extractive Industries for Development Series 113 TableB.17 MainEnergySourceforCooking:PercentageofAll HouseholdsUsingThatSource Quintile Electricity Kerosene Biomass LPG/naturalgas Other 1 0.1 0.0 99 0.1 0.7 2 0.0 0.0 99 0.2 0.8 Cambodia 3 0.0 0.0 98 0.7 1.0 4 0.1 0.1 97 1.9 1.0 5 0.5 0.2 76 22 1.4 All 0.2 0.1 93 5.6 1.0 1 0.0 0.7 93 0.5 5.8 2 0.0 1.2 91 2.3 5.3 3 0.0 2.3 86 7.2 4.7 India 4 0.1 5.7 69 21 3.9 5 0.2 7.3 25 65 2.1 All 0.1 3.8 70 22 4.2 1 0.0 0.9 99 0.1 0.0 2 0.0 2.0 98 0.0 0.1 Kenya 3 0.3 6.4 93 0.1 0.6 4 0.2 11 87 0.4 0.9 5 1.8 31 55 12 0.8 All 0.6 13 82 3.5 0.6 1 0.0 0.6 91 7.5 1.1 2 0.0 0.6 87 12 0.9 Pakistan 3 0.0 0.7 83 16 0.7 4 0.0 1.0 73 26 1.0 5 0.2 0.9 47 51 0.9 All 0.1 0.8 73 25 0.9 1 0.7 0.3 78 20 0.4 2 0.8 0.2 58 41 0.3 Thailand 3 1.8 0.3 38 60 0.1 4 4.8 0.2 18 77 0.1 5 11 0.5 5.3 84 0.0 All 4.1 0.3 37 59 0.2 1 0.0 0.1 100 0.0 0.3 2 0.2 0.2 99 0.0 0.3 Uganda 3 0.0 0.5 99 0.0 0.7 4 0.0 0.4 98 0.0 1.4 5 0.7 3.9 88 0.6 7.1 All 0.2 1.2 96 0.2 2.4 Source: Authors' calculations. Note: Biomass includes firewood, charcoal, dung, and agricultural waste. The category of LPG and natural gas is combined in the questionnaire for Pakistan; in Cambodia, India, Ke- nya, and Uganda, households do not use natural gas. For Cambodia, those households that reported using a combination of LPG and electricity as their main energy source for cooking are categorized under LPG. "Other" includes coal and coke. 114 Expenditure of Low-Income Households on Energy TableB.18 MainEnergySourceforLightinginIndia,Pakistan, Uganda,andVietnam:PercentageofAllHouseholdsUsingThat Source Quintile Electricity Kerosene Other 1 34 66 0.5 2 48 52 0.7 3 60 40 0.5 India 4 75 24 0.5 5 94 5.9 0.3 All 65 34 0.5 1 73 26 1.7 2 76 22 2.0 Pakistan 3 82 17 1.5 4 87 12 1.0 5 93 5.9 1.5 All 83 15 1.5 1 0.2 89 10 2 0.6 95 4.3 Uganda 3 2.3 95 2.4 4 6.4 91 2.1 5 33 63 4.7 All 10 85 4.6 1 90 7.7 2.3 2 95 2.8 1.8 Vietnam 3 97 2.3 1.1 4 98 0.7 1.0 5 99 0.2 0.5 All 96 2.5 1.3 Source: Authors' calculations. Note: In Vietnam, the categories for lighting are grid electricity; LPG, vegetable oil, or kero- sene (categorized here as "kerosene"); and batteries, generators, and other (categorized here as "other"). Extractive Industries for Development Series 115 TableB.19 MainEnergySourceforLightinginCambodiaand Kenya:PercentageofAllHouseholdsUsingThatSource Quintile Electricity Kerosene Batteries Other 1 1.9 80 16 2.6 2 4.2 71 24 1.4 Cambodia 3 9.1 62 28 1.2 4 18 50 31 0.5 5 54 24 22 0.5 All 19 55 24 1.1 1 0.3 79 16 4.6 2 1.7 89 5.9 3.4 Kenya 3 5.0 90 3.3 1.7 4 12 83 1.6 3.1 5 40 54 0.8 4.4 All 16 76 4.5 3.5 Source: Authors' calculations. 116 Expenditure of Low-Income Households on Energy References Andriamihaja, N., and G. Vecchi. 2007. "An Evaluation of the Welfare Impact of Higher Energy Prices in Madagascar." Africa Region Working Paper Series 106. Washington, DC: World Bank. www.worldbank.org/afr/wps/wp106.pdf. Bacon, Robert and Masami Kojima. 2006. "Coping with Higher Oil Prices." ESMAP Report 323/06. Washington, DC: ESMAP/World Bank Group. www.esmap.org/filez/pubs/372007122243_ CopingWithHigherOilPrices_323-06.pdf. Bacon, Robert, Soma Bhattacharya, and Masami Kojima. 2009. "Changing Patterns of Household Expenditures on Energy: A Case Study of Indonesia and Pakistan." Extractive Industries and Development Series #6. Washington, DC: World Bank. http://siteresources.worldbank.org/ INTOGMC/Resources/eifd6_changing_household_patterns.pdf. Bascand, Geoff, Jeff Cope, and Diane Ramsay. 2006. "Selected Issues in the Measurement of New Zealand's Saving(s)." Paper prepared for Reserve Bank workshop on "Housing, Savings, and the Household Balance Sheet," November 14, 2006. www.stats.govt.nz/research-reports/selected-issues- measurement-nz-savings.htm. Coady, David, and David Newhouse. 2006. "Ghana: Evaluating the Fiscal and Social Costs of Increases in Domestic Fuel Prices." In Aline Coudouel, Anis Dani, and Stefano Paternostro, eds., Poverty and Social Impact Analysis of Reforms, pp. 387­413. Washington, DC: World Bank. http:// siteresources.worldbank.org/INTPSIA/Resources/490023-1120845825946/ PSIACASESTUDIES_BOOK.pdf. Coady, David, Moataz El-Said, Robert Gillingham, Kangni Kpodar, Paolo Medas, and David Newhouse. 2006. "The Magnitude and Distribution of Fuel Subsidies: Evidence from Bolivia, Ghana, Jordan, Mali, and Sri Lanka." IMF Working Paper WP/06/247. Washington, DC: International Monetary Fund. www.imf.org/external/pubs/ft/wp/2006/wp06247.pdf. Coady, David, Robert Gillingham, Rolando Ossowski, John Piotrowski, Shamsuddin Tareq, and Justin Tyson. 2010. "Petroleum Product Subsidies: Costly, Inequitable, and Rising." IMF Staff Position Note 10/05. www.imf.org/external/pubs/ft/spn/2010/spn1005.pdf. El Said, Moataz, and Daniel Leigh. 2006. "Fuel Price Subsidies in Gabon: Fiscal Cost and Distributional Impact." IMF Working Paper WP/06/243. Washington, DC: International Monetary Fund. www.imf.org/external/pubs/ ft/wp/2006/wp06243.pdf. Extractive Industries for Development Series 117 ESMAP (Energy Sector Management Assistance Program). 2003a. India: Access of the Poor to Clean Household Fuels. Report 263/03. Washington, DC: World Bank. www.esmap. org/filez/pubs/26303India.pdf. ------. 2003b. "Household Fuel Use and Fuel Switching in Guatemala." Technical Report. Washington, DC: World Bank. www.esmap.org/filez/ pubs/03603FueluseGuatemala.pdf. --. 2003c. "Household Energy Use in Developing Countries: A Multicountry Study." Technical Report. Washington, DC: World Bank. www.esmap.org/filez/pubs/ HouseholdEnergyUseinDevelopingCountries.pdf. --. 2005. Household Energy Supply and Use in Yemen. Volume 1: Main Report, Volume 2, Annexes. Report 315/05. Washington, DC: World Bank. www.esmap.org/filez/ pubs/31505Vol1ForWebsite.pdf. Kojima, Masami. 2009a. "Changes in End-User Petroleum Product Prices." Extractive Industries for Development Series #2. Washington, DC: World Bank. http:// siteresources.worldbank.org/INTOGMC/Resources/ei_for_development_2.pdf. --.2009b. "Government Response to Oil Price Volatility: Experience of 49 Developing Countries." Extractive Industries for Development Series #10. Washington, DC: World Bank. Kojima, Masami and Robert Bacon. 2001. "Abuses in Fuel Markets: How to Protect Consumers and Public Health." Viewpoint 237. http://rru.worldbank.org/documents/ publicpolicyjournal/237Kojim-831.pdf. Komives, Kristin, Vivien Foster, Jonathan Halpern, and Quentin Wodon. 2005. Water, Electricity and the Poor. Who Benefits from Utility Subsidies? Washington, DC: World Bank. Kpodar, Kangni. 2006. "Distributional Effects of Oil Price Changes on Household Expenditures: Evidence from Mali." IMF Working Paper WP/06/91. Washington, DC: International Monetary Fund. www.imf.org/external/pubs/ft/wp/2006/wp0691.pdf. Masera, Omar R., Barbara D. Saatkamp, and Daniel M. Kammen. 2000. "From Linear Fuel Switching to Multiple Cooking Strategies: A Critique and Alternative to the Energy Ladder Model." World Development 28(12): 2083­2103. World Bank. 2003. "Iran: Medium Term Framework for Transition. Converting Oil Wealth to Development." Report 25848-IRN. Washington, DC. http://go.worldbank.org/A8LQ9H2ZB0. --. 2005. Angola: Assessing the Impacts of Phasing out Fuel and Utility Price Subsidies, Volume 1: Policy Briefing. Poverty and Social Impact Analysis. Washington, DC. http:// siteresources.worldbank.org/INTPSIA/Resources/490023-1120841262639/Angola_ PSIA_vol1_English.pdf. --. 2009. World Development Indicators. Online subscription database. Extractive Industries for Development Series #19 Mineral Resource Tenders: Application, Guiding Principles, and Methodology #18 Rockets and Feathers: Asymmetric Petroleum Product Pricing in Developing Countries #17 Environmental Governance in Petroleum Producing Countries: Findings From a Comprehensive Survey #16 Expenditure of Low-Income Households on Energy: Evidence from Africa and Asia #15 Petroleum Markets in Sub-Saharan Africa: Analysis and Assessment of 12 Countries #14 Toward Strengthened EITI Reporting: Summary Report and Recommendations #13 The Aluminum Industry in West and Central Africa: Lessons Learned and Prospects for the Future #12 Engagement with Civil Society: An EITI Implementation Case Study #11 Changes in CO2 emissions from Energy Use: A Multicountry Decomposition Analysis #10 Government Response to Oil Price Volatility: Experience of 49 Developing Countries #9 Guidance Note for Task Team Leaders: Mainstreaming Gender into Extractive Industries Projects #8 Mining for Equity: Gender Dimensions of the Extractive Industries #7 Financial Surety: Guidelines for the Implementation of Financial Surety for Mine Closure #6 Changing Patterns of Household Expenditures on Energy: A Case Study of Indonesia and Pakistan #5 Emerging Players in Global Mining #4 Mining Cadastres: Promoting Transparent Access to Mineral Resources #3 Extractive Industries Value Chain: A Comprehensive Integrated Approach to Developing Extractive Industries #2 Changes in End-User Petroleum Product Prices: A Comparison of 48 Countries #1 Vulnerability to Oil Price Increases: A Decomposition Analysis of 161 Countries The World Bank Oil, Gas, and Mining Policy Division The World Bank Group's role in the oil, gas, and mining sectors focuses on ensur- ing that its current interventions facilitate the extractive industries' contribution to poverty alleviation and economic growth through the promotion of good gover- nance and sustainable development. The Oil, Gas, and Mining Policy Division serves as the Bank's global sector manage- ment unit on extractive industries and related issues for all the regions of the world. It is part of the Oil, Gas, Mining, and Chemicals Department, a joint World Bank/ International Finance Corporation department. Through loans/credits/grants, technical assistance, policy dialogue, and analytical work, the Division leads a work program with multiple activities in more than 70 countries, of which almost half are in Sub-Saharan Africa. More specifically, the Division: · Advises governments on legal, fiscal, and regulatory issues and on institutional arrangements as they relate to natural resources, as well as on good governance practices. · Assists governments in setting up environmental and social safeguards in projects in order to promote the sustainable development of extractive industries. · Helps governments formulate policies that promote private sector growth and foreign direct and domestic private sector investments. · Advises governments on how to increase the access of the poor to clean commercial energy and to assess options for protecting the poor from high fuel prices. The Oil, Gas, and Mining Policy Division serves as a global technical advisor that supports sustainable development by building capacity and providing extractive industry sector-related advisory services to resource-rich developing country gov- ernments. The Division also carries out an advocacy role through its management of the following global programs: · The Extractive Industries Transparency Initiative (EITI) Multi-Donor Trust Fund, which supports countries in implementing EITI programs. · The Global Gas Flaring Reduction (GGFR) Public-Private Partnership, which brings governments and oil companies together to reduce gas flaring. · The Communities and Small-Scale Mining (CASM) Partnership, which promotes an integrated approach to addressing issues faced by artisanal and small-scale miners. · The Gender and Extractive Industries Program, which addresses gender issues in extractive industries. · The Petroleum Governance Initiative (PGI), which promotes petroleum governance frameworks, including linkages to environmental and community issues. The World Bank 1818 H Street, N.W. Washington, D.C. 20433 USA www.worldbank.org/ogmc (or /oil or /gas or /mining) www.ifc.org/ogmc (or /oil or /gas or /mining)