Europe and Central Asia Region Environmentally and Socially Sustainable Development Infrastructure and Energy ____________________________________________ Coping with the Cold Heating Strategies for Eastern Europe and Central Asia's Urban Poor _______ Julian A. Lampietti Anke S. Meyer The World Bank, 2002 www.worldbank.org/eca/environment The findings, interpretations, and conclusions expressed here are those of the author(s) and do not necessarily reflect the view of the Board of Executive Directors of the World Bank or the governments they represent. Cover photo by: Svend Erik Mikkelsen. Sellers of fuelwood stoves on a street market in Vanadzor, Armenia. Contents Foreword v Abstract vi Acknowledgements vii Executive Summary 1 1 What is Unique about ECA 5 The cold winters 5 The crumbling legacy of central planning 5 Falling household incomes 6 2 Household Energy Use 9 Energy prices 10 Conclusion 12 3 Household Demand for Heat 13 Household heating patterns and fuel choices 13 How much heat do households consume? 14 How much do households spend on heating? 14 Demand for heat in selected countries 15 How will households respond to price and income signals? 15 Conclusion 16 4 Rethinking Heat Supply 17 The prevailing practice: Least-cost heating options for full heat service 17 The new reality: Lower heat demand 19 The importance of building-based efficiency measures 21 Conclusion 21 5 Providing Clean Heat in Fiscally-Sustainable Ways 23 What we know so far 23 Policy instruments to encourage clean choices 24 Investments in heating technology 25 Institutional challenges 27 Conclusion 28 iii Annexes 29 1 Purchasing Power Parity and Exchange Rate Conversions 30 2 Data Assumptions 32 3 Household Energy Consumption Summary Statistics 35 4 Social Costs of Heating Options 42 5 How to Estimate the Demand for Heat 44 6 Validity of Heat Demand Model 48 7 Fixed Effects of Different Heating Techniques 49 8 Key Technical Characteristics of District Heating Systems and Housing Stock in Eastern Europe and Central Asia 50 References 57 Figures 1-1 Observed mean temperature January 1961 to January 1990, degrees Celsius 6 1-2 Relative changes in energy prices and incomes in Eastern Europe and Central Asia, 1991­2000 6 2-1 Per capita energy consumption in Eastern Europe and Central Asia 11 2-2 Share of energy spending in household budgets in Eastern Europe and Central Asia 11 2-3 Expenditure elasticity of energy demand 11 3-1 Urban household heating fuel choices by welfare (income) quintile 13 3-2 Predicted per capita heat and nonheat energy consumption in selected countries 14 3-3 Predicted heat expenditure as a percentage of household expenditures 15 3-4 Demand for heat in selected countries 15 4-1 Costs of different heat supply options in Moldova 18 4-2 Annual costs of different heating options for full heat service in Yerevan, Armenia 18 4-3 Cost of district heating per capita at various effective indoor temperatures compared with national GDP and official wages per capita in Moldova 19 4-4 Fuel costs as share of total heat costs for different heat supply options and demand levels, Yerevan, Armenia 20 4-5 Yerevan: Average cost of heating for high and low demand 20 Tables 2-1 Urban network energy use in Eastern Europe and Central Asia 9 2-2 Urban non-network energy use in Eastern Europe and Central Asia 10 2-3 Energy prices in Eastern Europe and Central Asia 10 3-1 Self-reported temperatures and heating expenditures in Armenia, 2000 15 4-1 Payback times for energy efficiency investments in buildings 21 4-2 Typical cost of heat metering and individual controls in apartments 21 iv Foreword Europe and Central Asia (ECA) is unique among Also, consistent with the expectation that the developing regions in that the cold winters poor already have cut heat consumption close to necessitate additional expenditures on heat. This the minimum needed to avoid health problems study is the first comprehensive examination of and chosen dirtier fuels to further save money, heat demand in ECA. For the urban poor in our the survey data show that the demand of the region, covering heating expenditures has poor is less income and less price elastic than become a major challenge. Problems are com- that of the nonpoor. This implies greater propor- pounded by the rapid deterioration or collapse tionate welfare losses to the poor and a more of central heating services in some countries. active search for substitutes if heating prices The challenge ahead is to design policies and increase. It suggests the possibility of designing investments that enable all people (poor and price-based heating subsidies that benefit the nonpoor) to access clean, affordable heating. poor more than the nonpoor. However, in tar- The authors use household survey data from geting subsidies, subsidy design must be based selected countries in the region to analyze on an understanding of income-linked access household energy consumption and heating rates to clean energy networks. If the poor lack patterns. The analysis provides several empirical network access, the bulk of network-based subsi- findings with significant policy relevance relat- dies will be captured by the nonpoor and there- ing to household expenditures on heat, the fore subsidies for non-network solutions may income and price elasticity of heat demand, and result in better poverty targeting. fuel choices. In the countries studied, nonpoor By providing new insights into how much people obtain heat at a cost of between $30 and energy people demand for heating and how $50 per year while poor people spend between much they pay for it, this study raises awareness $25 and $40 a year. The nonpoor also enjoy a in the Bank, as well as in the region, on the links higher quality heat supply at only slightly between heating, poverty alleviation, and envi- greater cost than the poor. This suggests that ronmental sustainability. By increasing our heating policy or investment interventions that understanding of heat demand, particularly at result in higher costs than existing systems will low income levels, it improves our ability to face substantial implementation resistance assess the advantages and disadvantages of vari- among the poor. Indeed, although the absolute ous supply options and to identify policies that cost differences are small, proportionally the are most likely to lead to outcomes that are poor spend almost twice as much of their house- acceptable from a social, fiscal, and environmen- hold budgets on heating as do the nonpoor. tal point of view. Pradeep Mitra Chief Economist Europe and Central Asia Region v Abstract Heating is a critical issue for people's livelihoods vides new insights into how much energy peo- in Eastern Europe and Central Asia. The region's ple demand for heating and how much they pay cold climate, the legacy of central planning, and for it and makes recommendations on how to the drop in household incomes over the past 10 design policies and investments that enable all years influence profoundly the design of heating people (poor and nonpoor) to access clean, strategies for the urban poor. This paper pro- affordable heating. vi Acknowledgements This report was written by Julian A. Lampietti were provided by Alexandre Marc (Sector (Task Team Leader) and Anke S. Meyer. The Manager ECSSD) and Henk Busz (Sector report was conceived by Laszlo Lovei (formerly Manager ECSIE). Sector Leader ECSEG) and completed under the The report benefited from reports prepared guidance and direction of Lee Travers (Sector by Environmental Resource Management and Leader ECSIE) and David Craig (formerly Sector COWI A/S on the "Urban Heating Strategy Leader ECSEG). Development for Republic of Armenia," COWI The authors are indebted to a number of A/S on the "Development of Heat Strategy for individuals that contributed to the research the Kyrygz Republic," SWEDPOWER/FVB on effort. These include Gediz Kaya, Irina "Strategic Heating Options for Moldova," and Klytchnikova, Brian Kropp, Haiyong Liu, and Kantor Management Consultants on "Surveys of Xun Wu. Editorial assistance was provided by a Metered District Heated Consumers in Estonia team including Paul Holtz, Bruce Ross Larson, and the Slovak Republic." and Stephanie Rostron. Peer review comments vii Executive Summary Heating is a critical issue for people's livelihoods per year while poor people spend between $25 in Eastern Europe and Central Asia. The region's and $40 a year. The nonpoor also enjoy a higher cold climate, the legacy of central planning, and quality heat supply at only slightly greater cost the drop in household incomes over the past 10 than the poor. This suggests heating policy or years influence profoundly the design of heating investment interventions that result in higher strategies for the urban poor. costs than existing systems will face substantial This paper uses survey data from selected implementation resistance among the poor. countries in the region1 to study how people heat Indeed, although the absolute cost differences their homes. It provides new insights into how are small, proportionally the poor spend almost much energy people demand for heating and twice as much of their household budgets on how much they pay for it. The reader must keep heating as do the nonpoor. in mind that heating is a local issue and solutions Second, consistent with the expectation that depend on the local circumstances. Therefore the the poor already have cut heat consumption guidance offered in this report must be adapted close to the minimum needed to avoid health based on analysis of local conditions. problems and chosen dirtier fuels to further save Analysis of the survey data revealed that money, the demand models show that the poor almost all households use electricity, with small are less income and less price elastic than the differences between the poor and the nonpoor. nonpoor. This implies greater proportionate wel- Poor people are much less likely to use district fare losses to the poor and a more active search heat and gas and much more likely to use wood for substitutes if heating prices increase. This and coal. Unfortunately, the data cannot tell us suggests the possibility of designing price-based whether the poor are more likely to use dirtier heating subsidies that benefit the poor more fuels because they lack access to the clean net- than the nonpoor. However, in targeting subsi- work fuels or because the prices of clean net- dies, subsidy design must be based on an under- work fuels are higher. However, fuel choices by standing of income-linked access rates to clean the poor correlate highly with fuel prices, and energy networks. If the poor lack network those prices have been consistently lower for the access, the bulk of network-based subsidies will dirtier fuels over the past decade. be captured by the nonpoor and therefore subsi- dies for non-network solutions may result in Household demand for heat better poverty targeting. Third, the analysis shows that such demand The analysis of household demand for heat in becomes much more elastic at consumption lev- Armenia, Kyrgyz, and Moldova provides several els above 500 Kgoe and a price of $0.20 per kilo- empirical findings with significant policy rele- gram of energy equivalent (equal to $0.017 per vance. These findings relate to expenditures on kWh). Because the long run marginal cost of heat, the income and price elasticity of heat clean energy sources is everywhere above that demand, and fuel choices. cost and unlikely to fall, network heat suppliers First, in the countries studied, nonpoor peo- recovering full costs will be operating in an ple obtain heat at a cost of between $30 and $50 inelastic portion of the consumer demand 1 2 Coping with the Cold curve. This inflection point will vary by country, Building-level boilers or individual heat tech- but is useful to estimate because it provides poli- nologies with low fixed cost and high variable cy guidance on the price above which consumer (fuel) cost may be less environmentally friendly, welfare begins to drop quickly and complemen- but may be more attractive in areas with low tary interventions to address this drop may be demand and low population density. Both dis- needed. trict and building-based heating systems require In the countries studied, poor people cope heat-metering devices to give consumers, espe- with unreliable district heating and rising ener- cially the poor, control over heating expendi- gy prices by substituting less expensive dirty tures. The individual heat technologies are energy, including wood, coal, and kerosene. But much easier for individual consumers to control there are private and social costs associated with and also reduce the need for institutional reform poor people's heating choices. Private costs needed to provide a demand-driven service. include the opportunity cost of the time spent Careful planning is required to make sure collecting heating material (especially wood) heating systems are affordable but also fully and illnesses and labor productivity losses asso- integrated into the national energy sector strate- ciated with insufficient heating. Social costs gies. For example, investments in district heat include air pollution from the burning of dirty may be justified because it is a byproduct from fuels and the environmental costs associated cogeneration plants critical to the national with deforestation and the loss of biodiversity. power supply system. Heating is the single most These costs must be taken into account when important use of energy in the residential and evaluating the economic implications of alterna- building sector; therefore the broader impact of tive heating policies and investments. heating fuel choices on energy networks requires careful consideration. Solving heating problems in poor countries--and poor towns Policies and instruments for poor people Experience restructuring district heating systems The challenge is to design policies and invest- in Eastern Europe, particularly in Poland and ments that enable all people (poor and non- the Baltics, has shown that they can be poor) to access clean, affordable heating. In an modernized--approaching efficiency, cost, and urban environment this is particularly difficult service levels experienced in the market because whole communities are affected by economies of the colder areas in Western and these choices. Therefore it is critical that the Northern Europe. In high-density urban areas choices allow poor people to opt in to the district heating is typically the most comfort- degree they wish to get the heat they want able, energy efficient, cost effective, and envi- because they might not use and will not pay for ronmentally friendly heating mode, particularly the wrong investments. when supplied from combined heat and power Policy instruments such as regulations, taxes, plants. It is available year-round and can be con- and subsidies coupled with institutional reform trolled individually by each consumer, and pay- and investments in technology offer a way for- ment for heat is usually based on consumption. ward. Those instruments can be used to encour- Investment strategies in poor countries must age the poor to make clean choices. Investing in carefully consider the advantages and disadvan- new technology or reengineering existing tech- tages of different types of heating systems. nology enables governments to do this in fiscal- District heating may be environmentally friend- ly sustainable ways. If the goal is to provide ly and very cost-effective in areas with a high access to clean and affordable heating, invest- heat load and high population density. ments and policy instruments must be explicitly However, high fixed costs may make them too funded to cover the difference between house- expensive in poor countries where households hold expenditures and the cost of supply. consume less heat than these systems are usually If the focus is on promoting clean non-net- designed to supply and they have lower heat work fuels, targeted vouchers for equipment and expenditures than required for cost recovery. possibly fuel may be promising instruments. If Executive Summary 3 the focus is on promoting access and use of net- ments must be coupled with innovative finan- work energy, lifeline tariffs can be effective, as cial instruments that enable consumers, particu- long as the size of the blocks is set to minimize larly the poor, to distribute capital costs over a capture by the nonpoor and the government longer period. explicitly compensates utilities for any social In addition to policy and investment instru- transfers they are asked to provide. ments, there is considerable room to increase For new investments, there is a role for pub- the institutional efficiency of heating service lic sector intervention in either increasing access delivery. This can be achieved through a combi- to low-cost clean non-network energy or extend- nation of training and commercialization of ing clean energy networks into poor areas. heating service providers, promoting effective Network investments must be coupled with collective action at the apartment building or investments in metering and control options community level, and encouraging participation and with consumption-based billing, allowing of private sector service providers. users to choose the amount of heat and levels of comfort and spending. Particularly promising, Note especially in areas where large increases in clean fuel prices are expected, are investments in effi- 1. Survey data comes from the following countries: ciency and insulation that can produce substan- Armenia, Croatia, Kyrgyz Republic, Latvia, Lithuania, Moldova, and Tajikistan. tial reductions in consumption. These invest- CHAPTER 1 What is Unique about ECA Heating is a critical issue for people's livelihoods clothes, and food are necessary to survive during in Eastern Europe and Central Asia. This paper the cold winters. provides new insights into the links between heating, poverty alleviation and environmental The crumbling legacy of central planning sustainability by taking a closer look at house- hold demand for heating. Under central planning, the region's govern- Urban areas in the region have three unique ments provided almost universal access to infra- features that distort patterns of development structure services. For example, close to 100 per- and limit household choices when it comes to cent of households have electricity connections. living conditions. The first is the region's cold In urban areas, space heating and in many cases climate, which necessitates high spending on domestic hot water supply were also part of the heat, winter clothing, and food. The second is cradle-to-grave centrally planned system. In the the legacy of central planning, which provided 1950s large, centralized district heating became almost universal access to infrastructure the system of choice in most developed coun- services--many of which are rapidly deteriorat- tries, including Eastern Europe and Central Asia, ing. The third is the drop in household incomes because it had the potential of efficiently using over the past 10 years. the waste heat recovered from power generation These factors influence profoundly the through combined heat and power (CHP) plants. design of heating strategies for the urban poor Users of district heating systems in centrally in Eastern Europe and Central Asia. The purpose planned economies had no influence over when of this paper is to facilitate the design of policies and how much heat was provided. They could and projects that provide poor households with be reasonably assured, however, that heat would access to clean, affordable heat. The study covers be provided for free as soon as outside tempera- only the urban poor because they have fewer tures dropped below 8o Celsius for at least five affordable heating options than do the rural days. Heating systems would then be opera- poor, resulting in more severe price shocks dur- tional until temperatures were above 8o Celsius ing the transition from central planning. for at least five days. Rooms would be heated to at least 20o Celsius most of the time and, lacking The cold winters individual controls, consumers would respond to overheating by opening windows--even in Average temperatures in the region are well the winter. below those in most other regions (Figure 1-1). Even before the 1990s, district heating sys- During the coldest days of the winter tempera- tems suffered from a lack of maintenance and tures often drop below minus 20o Celsius in financing. As a result temperatures within a dis- many places, and as a result heating is required trict heating system--and within buildings-- for five to seven months in most places. People could be quite different from one area to the at the same income level as in other regions are next. Moreover, breaks in hot water pipes worse off in Eastern Europe and Central Asia became more frequent, requiring that the affect- because additional expenditures on heat, warm ed part of the system be shut down for repairs. 5 6 Coping with the Cold Figure 1-1 Observed mean temperature January 1961 to January 1990, degrees Celsius Source: Intergovernmental Panel on Climate Change. Financial problems created by the collapse of the increasing income polarization, and in many centrally planned economies were aggravated by countries urban poverty has reached alarming the increase in primary energy prices in these levels. countries starting in the early 1990s. The costs of providing heat began to soar, and one govern- Figure 1-2 Relative changes in energy prices ment after another decided to raise residential and incomes in Eastern Europe and Central heat tariffs closer to supply costs. Higher heat tar- Asia, 1991­2000 iffs coincided with the lower household incomes caused by the contraction in economic activity. Index (1991=100) While not having control over the amount 250 of heat consumed may have been acceptable when heat was essentially free of charge, it Clean fuels 200 became untenable as prices rose. Coupled with late or non-payment of salaries and pensions as well as a loss of entitlements, many households 150 Dirty fuels responded by not paying their heating bills, falling behind in their payments or switching to less expensive heating fuels. 100 Falling household incomes 50 Average GDP per capita Between 1991 and 1996 real incomes dropped by 14 percent a year in Eastern Europe and 0 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Central Asia. Between 1996 and 2000 real incomes grew slightly, by just under one percent Source: Author's calculations from International Energy Agency data and World a year. Such changes have been accompanied by Bank data. What is Unique about ECA 7 While real incomes have stabilized, real ener- natural gas, and kerosene) and dirty fuels (coal, gy prices have been rising. Governments have wood, and diesel). The price of clean fuels rose been eliminating energy subsidies, pushing utili- much faster (110 percent between 1991 and ties to raise prices in an attempt to improve cost 2000) than that of dirty fuels (45 percent). Thus recovery. Many of the price increases have been energy, particularly from clean fuels, has become substantial--for example, between 1991 and a relatively more expensive component of 2000 the price of electricity jumped by an aver- consumption. age of 177 percent throughout Eastern Europe and Central Asia.1 Note The changes in energy prices and incomes between 1991 and 2000 are shown in Figure 1-2. 1. These data cover Armenia, Azerbaijan, Estonia, The figure separates price changes in clean (liq- Georgia, Kazakhstan, Kyrgyz Republic, Latvia, Lithuania, Moldova, Tajikistan, and Uzbekistan. uefied petroleum gas, electricity, district heat, CHAPTER 2 Household Energy Use Heat is just one of many forms in which energy Almost all households use electricity, with small is consumed by households. This chapter starts differences between the poor and the nonpoor. by examining overall energy consumption But poor people are much less likely to use dis- because this provides key insights into the issues trict heat and gas. Are the nonpoor more likely that need to be addressed to create a better policy to use network energy because they have better environment and highlights important issues access to the network, or is it because they make related to heating and access to clean energy different choices? Although this question can- infrastructure. The seven countries considered in not be answered with the data from household this chapter--Armenia, Croatia, Kyrgyz Republic, surveys, it points to the need for country- Latvia, Lithuania, Moldova, and Tajikistan--were specific analysis to identify supply constraints-- selected based on the availability of recent (end such as network location and capital equipment 1997 or later) household survey data with suffi- (such as gas heaters)--that limit poor people's cient questions about energy expenditure pat- access to clean network energy. terns. Purchasing power parity and exchange rate If poor people are not using network energy, conversions, data assumptions, and summary what are they using? Primarily dirty non-net- statistics for the data presented in this chapter work energy. Wood and coal use are consistent- are found in annexes 1, 2, and 3. ly higher among the poor--except in Tajikistan, Between 1990 and 1997 the region's per capi- where coal is heavily subsidized for everyone ta commercial energy consumption fell by one- (Table 2-2). Except in Latvia, the nonpoor are third (World Bank 2001). While much of this more likely to use liquefied petroleum gas drop can be attributed to the collapse of indus- (LPG), the cleanest non-network energy. The try, the decline in subsidized infrastructure poor may favor dirty non-network energy services--coupled with higher prices and because it is less expensive or because they do increasing poverty--may explain what appears not have the resources to spend on appliances to be a fundamental shift in energy consump- that enable them to use network energy, such as tion and spending among urban households. gas stoves. Burning dirty fuels has social costs-- Separating network and non-network energy mainly air pollution and deforestation--that use provides insight into this shift (Table 2-1). require careful, country-specific analysis to Table 2-1 Urban network energy use in Eastern Europe and Central Asia (percent) District heating Central gas Electricity Country Poor Nonpoor Poor Nonpoor Poor Nonpoor Armenia, 1999 11 14 4 16 97 99 Croatia, 1997 15 39 19 30 99 100 Kyrgyz Republic, 1999 17 55 13 33 100 99 Latvia, 1997 70 83 57 68 99 100 Lithuania, 1998 31 46 47 56 85 94 Moldova, 1999 17 57 37 70 65 89 Tajikistan, 1999 1 1 3 6 100 100 Source: Author's calculations from household survey data. 9 10 Coping with the Cold Table 2-2 Urban non-network energy use in Eastern Europe and Central Asia (percent) Liquefied propane gas Kerosene Coal Wood Country Poor Nonpoor Poor Nonpoor Poor Nonpoor Poor Nonpoor Armenia 17 27 14 11 n/a n/a 47 50 Croatia 44 45 3 7 1 1 51 26 Kyrgyz Republic 24 39 31 17 60 31 46 22 Latvia 37 28 n/a n/a <1 <1 1 2 Lithuania n/a n/a n/a n/a <1 <1 1 2 Moldova 6 7 n/a n/a 9 5 12 9 Tajikistan n/a n/a <1 1 11 18 47 32 n/a: Not available from household survey. Source: Author's calculations from household survey data. assess the size of these costs and evaluate the is in place, and almost all fuels are available in economic implications of raising the price of all countries. Resolving this issue on a country- clean energy (annex 4). specific basis is important for the design of pro- poor heating strategies because it will influence Energy prices policy and investment decisions. More informa- tion on this is provided in the final chapter. Countries in the region have taken different Different pricing policies and differences in approaches to reforming energy prices. In the income, climate, and the availability of substitute countries in our sample the average price of a fuels lead to very different consumption patterns kilogram of oil equivalent (kgoe) is $0.25 (Table across countries (Figure 2-1). In relatively wealthy 2-3).1 But in some countries (such as Croatia) Croatia energy consumption is 325 kgoe per capi- some energy prices are much higher, while in ta per year, while in poor Tajikistan it is only 75 others (Kyrgyz Republic, Tajikistan) they are kgoe per capita per year. On average the nonpoor much lower. Non-network LPG tends to be consume one-third more energy per capita than expensive, while coal and wood tend to be do the poor (160 kgoe compared with 118 kgoe). cheaper. Network electricity tends to be expen- These consumption figures likely underestimate sive, while the prices of network central heat actual energy consumption, however, because the and gas generally fall between those of electrici- self-reported data include a number of very low ty and wood. values and do not include district heat. Two competing hypotheses explain the Households spend a large portion of their region's energy use patterns. The first is that budgets on energy--from 3 percent in Tajikistan poor people choose non-network energy because to about 12 percent in Armenia and Moldova it is less expensive; the second is that they do (Figure 2-2).2 (These expenditures include dis- not have access to network energy. But two fac- trict heat, electricity, coal, LPG, kerosene, wood, tors suggest that if there is an access issue, it is a and central gas.) In all countries except Latvia local one: high network energy use before the the poor spend a larger share of their household transition indicates that network infrastructure budgets on energy than do the nonpoor.3 This Table 2-3 Energy prices in Eastern Europe and Central Asia, (2001 U.S. dollars per kilogram of oil equivalent, most recent year available) Country LPG Kerosene Coal Electricity Wood Central gas District heat Armenia 0.59 0.40 0.07 0.56 0.16 0.12 0.18 Croatia 0.88 0.17 0.21 0.94 0.11 0.25 n/a Kyrgyz Republic 0.22 0.50 0.01 0.05 0.08 0.07 0.04 Latvia 0.35 0.21 0.12 0.60 0.17 0.76 0.35 Lithuania 0.26 0.21 0.13 0.60 0.11 0.20 0.28 Moldova 0.41 0.19 0.10 0.45 0.10 0.11 0.15 Tajikistan 0.33 0.11 <0.01 0.03 <0.01 0.06 0.13 n/a: Not available. Source: Author's calculations from household survey data and International Energy Agency data. Household Energy Use 11 Figure 2-1 Per capita energy consumption Following the methodology in Subramanian in Eastern Europe and Central Asia and Deaton's (1996) study on the demand for food calories, a preliminary assessment of the kgoe per capita per year tradeoff between income and energy expendi- 400 ture can be made by running a simple log-log Nonpoor 350 regression to estimate the expenditure elasticity of energy demand. In doing so, however, it must Poor 300 be remembered that differences in rates of 250 change in household spending across countries may be confounded by local differences in poli- 200 cy and physical infrastructure. The results of the regression for poor and 150 nonpoor consumers are shown in Figure 2-3. 100 Although there is variation across countries, the results show that relative to income, poor peo- 50 ple's energy expenditures are consistently more elastic than those of the nonpoor. A 10 percent 0 Armenia Kyrgyz Croatia Moldova Tajikistan Lithuania Latvia increase (decrease) in income results in an 8 per- Republic cent increase (decrease) in energy expenditure Source: Author's calculations based on household surveys. for poor people and a 5 percent increase (decrease) for nonpoor people. In an environ- Figure 2-2 Share of energy spending in ment of falling incomes the poor appear to be household budgets in Eastern Europe cutting back energy expenditures (as a percent- and Central Asia age of income) faster than the nonpoor, proba- bly by consuming less expensive, dirtier fuels. Percentage of total spending 15 Thus heating policies designed to encourage the Poor poor to use clean fuels must either increase their Nonpoor incomes or reduce their expenditures (through 12 subsidies or investments in efficiency). 9 Figure 2-3 Expenditure elasticity of energy demand 6 Expenditure elasticity 1.0 Nonpoor 3 Poor 0.8 0 Armenia Kyrgyz Croatia Moldova Tajikistan Lithuania Latvia Republic 0.6 Source: Author's calculations from household survey data. 0.4 finding is the opposite of Freund and Wallich's (1996) finding in Poland. Clearly relative impacts will vary by country, but policy 0.2 designed to raise energy prices must disaggre- gate impact by income class and, especially 0.0 where energy consumption levels are quite Armenia Kyrgyz Croatia Moldova Tajikistan Lithuania Latvia Republic high, anticipate substitution of cheaper, dirtier fuels by poor households. Source: Author's calculations from household survey data. 12 Coping with the Cold Conclusion When cheaper dirty fuels are available, the social costs associated with consuming them This chapter has shown the considerable hetero- may warrant public intervention. In countries geneity in household energy use, consumption, with high-energy expenditure elasticities, policy- and spending across Eastern Europe and Central makers who want to encourage the poor to use Asia--highlighting the importance of country- clean fuels must either increase incomes or specific analysis. That poor people are less likely reduce the relative cost of clean fuels (through to use network energy than nonpoor people subsidies or investments in efficiency). raises questions about the role of the public sec- tor in providing subsidies or investing in new or Notes rehabilitating old infrastructure. Unless clean energy networks are made accessible to the poor, 1. Comparing energy consumption patterns requires the bulk of any energy subsidies will go to the converting different fuels to equivalent energy val- ues. Conversions to kilograms of oil equivalent nonpoor. (kgoe) are based on mean values of fuel energy con- As noted, the elasticity of energy expenditure tent relative to oil, so the exact heat content of a relative to total expenditure is 0.8 for the poor given fuel will vary depending on its quality and effi- and 0.5 for the nonpoor. The poor spend a larger ciency in combustion. This paper uses the following portion of their incomes on energy yet consume equivalence values: 1 kilowatt-hour of electrici- less energy in absolute terms. Energy pricing ty=0.085 kgoe; 1 cubic meter of central gas=0.833 kgoe; 1 kilogram of LPG=1.059 kgoe; 1liter of policy discussions tend to focus on the cleaner kerosene=0.824 kgoe; 1 kilogram of wood=0.376 fuels because of their greater importance in kgoe; 1 kilogram of coal=0.541 kgoe. The source is quasi-fiscal activity. A failure to recognize the the International Energy Agency. significant impact of energy prices on poor peo- 2. These are reported expenditures, not subject to ple's welfare and their option to substitute adjustment for arrears and nonpayments. 3. The result in Latvia may be explained by the large cheaper, dirty fuels may simply replace one bad number of households on the central heating with another. network. CHAPTER 3 Household Demand for Heat Studying how people heat themselves when left Household heating patterns and fuel to their own devices provides insights into how choices much energy they demand for heating and how much they are willing to pay for it. Anecdotal Figure 3-1 shows the heating fuel choices of evidence suggests that households--particularly households not on district heating networks.1 poor households--have a wide variety of heat- When free to choose, the poor are more likely to ing strategies. For example, in Russia it is report- use dirty fuels such as wood (Armenia) and coal ed that "wearing winter clothing indoors or (Moldova), while the nonpoor rely on clean sleeping under a multitude of blankets only fuels such as electricity and central gas. keeps one warm for so long. One popular reme- These patterns have important implications dy is stuffing rags into an empty can, dousing for heating interventions. First, as incomes fall, them in vegetable oil, and setting them on fire" people buy dirtier heating fuels. Second, while (Filipov 2001). cash transfers2 may offset the welfare effects of Figure 3-1 Urban household heating fuel choices by welfare (income) quintile Clean fuels (electricity, central gas, kerosene) Percent of households Combination clean and dirty 120 Dirty fuels (wood, coal) 100 80 60 40 20 0 Bottom 2 3 4 Top Bottom 2 3 4 Top Bottom 2 3 4 Top Armenia Moldova Kyrgyz Republic Quintile and country Note: Excludes district heating. Source: Author's calculations from 1999 household survey data. 13 14 Coping with the Cold higher heating prices, they will not stop house- based on a minimum consumption level per holds from using dirtier fuels if the prices of household. those fuels are not raised as well. Thus thought Annual nonheat energy consumption ranges should be given to designing heating policies from 50 kgoe per capita in Armenia to about 125 that take into account the social costs of burn- kgoe in the Kyrgyz Republic. Annual predicted ing dirty fuels. These include the health costs heat consumption ranges from 40 kgoe per capi- associated with not having enough heat and the ta in Armenia to 175 kgoe in Moldova to 180 resulting productivity losses, the health costs kgoe in the Kyrgyz Republic. Thus heat con- associated with burning dirty fuels, the environ- sumption accounts for 40­60 percent of total mental costs associated with deforestation, and energy consumption. Differences across coun- the opportunity costs of time spent collecting tries are driven by differences in climate and heating material--especially wood (annex 4). energy pricing policies. The average temperature during the heating season is highest in Armenia How much heat do households consume? (2.60 Celsius), followed by Moldova (0.60) and the Kyrgyz Republic (­2.90). Energy prices are Household heat consumption was estimated by highest in Armenia, followed closely by developing a model to predict household heat Moldova, and are substantially lower in the and nonheat energy consumption, then sub- Kyrgyz Republic. tracting nonheat from the total (for details on the model and its validity see annexes 5 and How much do households spend on 6).3 Figure 3-2 presents heat consumption heating? results on a per capita basis.4 The figure reveals variations in household heat consumption--in To calculate heating expenditures, we multiply Armenia and the Kyrgyz Republic the poor predicted heat consumption by the price of a consume less heat per capita than do the non- household's primary heating fuel. These calcula- poor.5 That the results are confounded by tions indicate that heating accounts for 5­10 household size complicates the design of pro- percent of household spending and for 20­40 poor heating tariffs such as lifelines, which are percent of energy spending. On average the poor spend almost twice as much of their house- Figure 3-2 Predicted per capita heat and hold budgets on heating as do the nonpoor nonheat energy consumption in selected (Figure 3-3). In absolute terms nonpoor house- countries holds spend $30­50 a year on heating and poor Predicted nonheat households spend $25­40. Predicted heat These results are important for three reasons. Kyrgyz Republic First, that the poor spend a larger share of their nonpoor budgets on heating suggests that it is possible to design a heating subsidy that benefits them Kyrgyz Republic poor more than the nonpoor. Second, that heat is a large share of energy spending suggests higher Moldova nonpoor heating prices will considerably reduce house- hold welfare unless inexpensive substitutes are Moldova poor available. Third, poor people are unlikely to pay for heating systems that cost more than $25­40 Armenia nonpoor a year because they can find less expensive ways to heat themselves. They might, however, be willing to pay slightly more for heating systems Armenia poor that are substantially more convenient. 0 50 100 150 200 250 300 350 400 One of the factors complicating this analysis kgoe per year is that we do not have data on actual heat con- sumption. But in a recent survey Armenian Note: Excludes households on district heat. Source: Author's calculations. apartment dwellers were asked to estimate their Household Demand for Heat 15 Figure 3-3 Predicted heat expenditure as a Table 3-1 Self-reported temperatures and heat- percentage of household expenditures ing expenditures in Armenia, 2000 Reported Percent of total expenditure on heat mean 12 Reported expenditure mean (US$ per US$ Type of temperature heating per 10 household (0C) season) degree Poor with district heating 15.62 17 1.09 8 Poor Nonpoor with distict heating 16.51 21 1.27 Poor without district 6 heating 14.53 9 0.62 Nonpoor Nonpoor without district heating 15.61 17 1.09 4 Source: Author's calculations from Armenia 2001 household survey data. 2 households alter their heating strategies quickly in response to price changes in the range of 0 Armenia Kyrgyz Republic Moldova $0.01­0.20 per kgoe--and that for households without substitution opportunities, welfare loss- Note: Excludes households on district heat. es will be greater when the price rises above $0.2 Source: Author's calculations. per kgoe. In these cases it will be particularly previous year's spending on heating and their important to design policies that cushion the average indoor temperature during the heating blow of energy price increases on the poor. season. Self-reported spending ranged from $10­20 a year, which is of the same order of How will households respond to price and magnitude as the model results (Table 3-1). In income signals? addition, poor households with full control of their heating arrangements keep their apart- The model can also be used to estimate the ments at lower temperatures and spend less income and price elasticity of heat demand for than do households on the district heating the three countries. The income elasticity of network--suggesting that the poor lose the most when they cannot regulate their heating use. Figure 3-4 Demand for heat in selected countries Demand for heat in selected countries Exchange rate (U.S. dollars per kgoe) 0.8 Armenia We expect a heat demand function to be kinked, sloping steeply around the minimum amount Kyrgyz Republic 0.7 needed for survival and then rapidly leveling off Moldova 0.6 as the quantity of heat goes from necessity to luxury. Identifying the location of this kink is 0.5 important because at prices above it demand is 0.4 inelastic and welfare losses are large--while at prices below it demand is more elastic and wel- 0.3 fare losses are smaller. 0.2 A scatter plot of predicted household heat consumption against price per kgoe for 0.1 Armenia, the Kyrgyz Republic, and Moldova 0.0 suggests a function of precisely this shape 0 500 1,000 1,500 2,000 Predicted kgoe per year consumed for heating (Figure 3-4). There is a steep downward slope below 250 kgoe and above $0.2 per kgoe fol- Note: Excludes district heating. lowed by a rapid flattening out. It appears that Source: Author's calculations. 16 Coping with the Cold demand is between 0.1 and 0.2, meaning that a particularly important to design policies that ten percent increase (decrease) in income will cushion the blow of energy price increases. produce a one percent increase (decrease) in Third, on average the poor spend almost energy consumption for heating by the poor twice as much of their household budgets on and about a two percent increase (decrease) by heating as do the nonpoor. That the poor spend the nonpoor. That the three data sets produce a larger share of their budgets on heating sug- similar results and are consistent with econom- gests that it is possible to design a heating sub- ic theory increases our confidence in the sidy that benefits them more than the nonpoor. model. The difficulty with designing such a subsidy is As expected, there is much greater variation that the nonpoor tend to have higher access in price response by income group and country. rates to clean energy networks. Therefore they Price elasticity is ­0.4 in Armenia and ­0.2 in the are more likely to capture the bulk of the sub- Kyrgyz Republic and Moldova, meaning that a sidy if it is passed through the network without ten percent increase in price will produce about first increasing the access rate of the poor. a four percent decrease in consumption in Finally, the data from Armenia, the Kyrgyz Armenia and about two percent in Kyrgyz and Republic, and Moldova suggest that demand Moldova. In Armenia and Moldova the poor are becomes much more elastic at prices below less price elastic than the nonpoor. That the $0.20 per kgoe (equal to $0.017 per kWh). poor are less income and less price elastic than Although such inflection points will vary by the nonpoor suggests that they will have greater country, they provide policy guidance on the welfare losses from price increases unless they price above which consumer welfare begins to can find less expensive substitutes. drop quickly and complementary interventions to address this drop may be needed. Conclusion Notes The information in this chapter is important for designing pro-poor heating policies and invest- 1. Most of the analysis in this chapter is limited to a ments. First, unless there is a significant sample of urban households from Armenia (1999), the Kyrgyz Republic (1999), and Moldova (1999). improvement in heat quality, poor people are 2. Direct cash transfers are discussed on page 28. unlikely to pay for heating systems that cost 3. The consumption and expenditure results in this more than $25­40 a year because they can find chapter are not identical to those in the previous less expensive ways to heat themselves. Thus, chapter because the analysis in this chapter focuses cost recovery strategies must take into account only on a subsample of urban households for which consumer perceptions of system quality, which heating information is available. 4. While heat is a public good at the household is a function of cost and convenience. level, larger (poor) households tend to consume more Second, the poor are less income and less energy than smaller (non-poor) households. There price elastic than the nonpoor, suggesting that are on average two more people in poor than in non- they will have greater welfare losses from price poor households. Also there is not much differentia- increases unless they can find substitutes. For tion in living area because commercial real estate markets are not well developed in the sample poor people these substitutes tend to be dirtier countries. fuels, and there are social costs associated with 5. In Moldova the difference is not statistically sig- the use of these fuels. In these cases it will be nificant at the five percent level. CHAPTER 4 Rethinking Heat Supply International financial institutions (IFIs) such as rooms of a dwelling, and reduced service, mean- the World Bank and the European Bank for ing a lower temperature in one or several rooms. Reconstruction and Development (EBRD) were These results are compared with typical expendi- the main funding sources for rehabilitation ture levels reported in chapter 3, and conclu- investments for district heating systems in many sions are then drawn regarding the implementa- cities in Eastern Europe and Central Asia in the tion of financially and environmentally sustain- 1990s. The experience in restructuring Soviet- able and affordable heating strategies that take type district heating systems in Eastern Europe, into account the fixed and variable costs and particularly in Poland (see Box 2 in Annex 8 and investment requirements of various heat supply World Bank 2000c) and the Baltics,1 has shown options. that, through a combination of investments, The heat supply options compared in this institutional improvements and sector reform, chapter range from highly centralized district those district heating systems can be heating networks fed by cogeneration plants or modernized--approaching efficiency, cost, and heat-only boilers to building boilers that supply service levels as in Western and Northern only one or a few buildings with heat to decen- Europe. There, district heating is considered the tralized (individual) heating where each most comfortable, efficient, environmentally dwelling has its own heat source. Each of these friendly heating mode; it is available year-round heating options can be based on a wide range of and can be controlled individually by each con- fuels and comes with very different levels of effi- sumer, and payment for heat is usually based on ciency and environmental performance. metered consumption. These solutions and experiences are not fully The prevailing practice: Least-cost heating applicable, however, when devising solutions to options for full heat service the heating problems of households in extreme- ly poor countries or in many small, poor towns Before the transition, consumers in Eastern in other countries of the region. The informa- Europe and Central Asia connected to central tion from the previous chapters shows that heating expected that every room in their living many poor urban households consume less heat quarters would be heated to about 20o Celsius and have lower heat expenditures than usually for 24 hours during the official heating season. associated with a district heating system. Even In the following we call this "full heat service." though district heating systems can be the most Under such conditions, and with the typical cost-effective heating mode given a high heat high population densities in many suburban load, their high fixed costs make them potential- areas with high-rise residential buildings, the ly very expensive for consumers demanding less heating system that typically provides heat at heat. the lowest cost is a district heating system sup- In the remainder of this chapter the typical plied from cogeneration plants. Comparative costs of various heat supply options are com- studies ("heat plans") have been carried out in pared for two levels of heat demand: full service, many cities in Eastern and Western Europe con- meaning provision of about 18o Celsius2 in all firming this result for greenfield development as 17 18 Coping with the Cold Figure 4-1 Costs of different heat supply ness to more decentralized options. Heating options in Moldova costs of about US$0.16 per kgoe would result in an annual household heating bill of $160, Heating costs (U.S. dollars per kgoe) assuming heat consumption of 1000 kgoe (or 10 0.5 Based on heat Gcal) per flat. demand Minimum and maximum according to The costs of modernized district heating sys- linear densities in six cities effective tems in various countries and cities have been indoor 0.4 temperature well researched during the preparation of feasi- of 20 degrees Celsius bility studies. The resulting costs per unit of heat (A) delivered at the building entrance usually fall 0.3 within a fairly similar range of $0.20­0.35 per District heating with gas fuel cost kgoe, leading to annual household heating bills 0.2 of $200­900, depending on dwelling size, specif- Local boiler with gas fuel cost ic heat consumption, and heat tariff level. The costs of modern heating options other 0.1 Based on 1999/ District heating than district heating are less well known in 2000 heat with zero fuel cost demand Eastern Europe and Central Asia. According to (B) studies recently carried out in Armenia, those 0 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 heating options result in annual costs per house- Linear heat load density, kgoe per meter hold of $135­324 (Figure 4-2). Options based on natural gas have high investment costs and low Source: Based on SwedPower/FVB 2001. fuel costs, while the opposite holds for heating based on electricity, kerosene, LPG, and wood well as modernization of existing district heat- stoves. For all heating options represented in ing systems. Figure 4-2, investments have been included to Figure 4-1 gives a graphic illustration of this ensure that the equipment would be functional point with an example from Moldova. The fig- over a lifetime of 20 years.4 As a result, the costs ure shows the costs per unit of heat depending on the "linear heat load density." This is the Figure 4-2 Annual costs of different heating total connected load divided by the total length options for full heat service in Yerevan, Armenia of the network, with the length of the network defined as the total length of the route (in con- U.S. dollars per household trast to the length of single pipes, which would 350 cover both supply pipes and return pipes). Full Capital heat service results in a high heat load with the 300 Operations and maintenance typically high population density in urban Fuel Moldova (high linear heat load density = A). 250 Under these circumstances it is more cost-effec- 200 tive to build a district heating network and a rather capital-intensive cogeneration plant that 150 delivers heat almost for free, rather than invest in an extensive natural gas distribution system 100 with separate gas-boilers for one building or a small group of buildings.3 The exact location of 50 the cost curves depends on local costs and cir- cumstances, particularly the price at which 0 cogenerated power can be sold to the grid and CHP Large Small Block Ind. Ind. Solid fuel LPG Kerosene HOB HOB B. NG electric nat. stove stove stove the cost of fuels for heat generation. However, stove gas stove when the heat demand is much less, such as currently in Moldova (low linear heat load den- Note: The calculations are based on a comfort level of 17o Celsius and 110 heating days. sity = B), district heating loses its competitive- Source: Based on COWI 2002a. Rethinking Heat Supply 19 per apartment are lowest for wood stoves, build- Figure 4-3 Cost of district heating per capita at ing-based natural gas boilers, and apartment- various effective indoor temperatures compared with national GDP and official wages based natural gas heaters. But the current natur- per capita in Moldova al gas tariff for small consumers is only about 17 percent higher than that for large consumers, U.S. dollars per year and so does not reflect the higher distribution 300 costs. Moldova, GDP 2000Q1 per capita 250 The new reality: Lower heat demand Moldova disposable income per capita, 2000Q1 200 The analysis so far has focused on heating Chisinau, official wages per capita, 2000Q1 options based on full heat service. But since the early 1990s many consumers in Eastern Europe 150 and Central Asia have received less than full Cost of district heating heat service as systems started to deteriorate due 100 Moldova official wages per capita, 2000Q1 to lack of maintenance. More important, declin- Rest of country, ing incomes have led many consumers to drasti- 50 official wages per capita, 200Q1 cally reduce their heat consumption, with lower supply temperatures, shorter heating seasons, Fuel component (natural gas) 0 and less area heated. However, only for those 0 4 8 12 16 20 households not on the network or that have dis- Effective indoor temperature, degrees centigrade connected from the network would this result in lower heat bills (see Chapter 3). Those still con- Note: The cost of district heating is estimated using the current price of $0.18 per kgoe. nected to district heating experienced rising Source: SwedPower/FVB 2001. heat tariffs and thus higher expenditures despite declining service levels. The reason is when share of total costs and which are modular are supply-driven, inflexible district heating systems much easier to adapt to lower heat demand, as lose customers, heat not consumed does not demonstrated in Chapter 3. With electrical heat- materialize as fuel savings at the heat generation ing, for example, fuel accounts for about 85 per- plant. Instead it is consumed somewhere else in cent of total costs (Figure 4-4). Therefore while the system, for example, in the form of heat electrical heating has a high unit cost, it may be losses or higher temperatures and open win- less expensive for the household to heat with dows. Therefore, utilities are typically not able because it is so much more flexible. In many to reduce costs in the short to medium term in countries of the region, however, the already proportion to the decline in demand, let alone overburdened electrical distribution network trying to reduce staff or fixed costs and the would have to be strengthened to be able to cope remaining customers have to bear even higher with additional heat loads. This would cause costs. In Bulgaria this vicious circle could be additional investments, reflected in higher elec- observed in 1996­99. Since then, customers tricity tariffs. Typically, district heating systems have slowly started to reconnect due to efforts to can only be adapted to a lower heat load in the meter heat consumption and bill customers medium to long term, when replacement invest- accordingly (see Box 3, Annex 8). ment and modernization of the system configura- Figure 4-3 shows the cost of district heating tion take place. This is an option worth pursuing in Moldova at various temperature levels5 and for those district heating systems that can be compares it with various measures of income in shown to be viable even at lower heat demand. the country. Outside the capital city of Centralized options are cheaper than electric Chisinau, the official per capita wage would heating or wood stoves when providing full heat barely cover the fuel costs of district heating service. Now that incomes have fallen, con- supplied at 14o Celsius. sumers, particularly the poor demand lower More flexible options--such as individual heat indoor temperatures and heat less living space. technologies--for which fuel accounts for a larger Under these circumstances individual options 20 Coping with the Cold Figure 4-4 Fuel costs as share of total heat When the costs of different heating options costs for different heat supply options and are compared with what households are current- demand levels, Yerevan, Armenia ly paying for heat, between $25 and $50 in Armenia (see Chapter 3), it appears that only the Percent 100 most basic heating could be considered. In fact, Survival (low demand) most people would be willing to pay more for a Consolidation (high demand) convenient heat supply during the entire heat- 80 ing season--according to the Armenia survey, between $50 and $100.6 This would not entirely bridge the gap between the supply cost of most 60 heating options and expected consumer pay- ments. But the remaining gap may be narrow 40 enough to make the financial support by the government to the poorest consumers feasible. Figure 4-5 makes another interesting point. 20 In the short-term, it may be possible to provide affordable heating with centralized heating sys- tems by emulating how consumers use individ- 0 Combined Large Small Block Ind. Ind. Solid fuel Liquid Kerosene heat heat heat natural electric natural stove propane stove ual heating systems. The Armenia Urban and only only gas stove gas gas Heating Strategy proposes that during the sur- power boiler boiler stove stove vival phase only one or two of the vertical risers Source: COWI 2002a. supplying each apartment are kept connected, are less expensive than centralized options delivering a temperature of about 17o Celsius in because they tend to be modular. Figure 4-5 those one or two rooms. Adopted for an entire illustrates this for Armenia. The result is corrob- centrally heated area, this should cut down con- orated for Moldova in Figure 4-5 comparing sit- siderably on fuel costs that have a cost share of uation A with situation B. 70­80 percent (see Figure 4-4). It is hoped that this would enable the heating company to cover its full cost and bill and, more importantly, col- Figure 4-5 Yerevan: Average cost of heating lect accordingly. This is however only an interim for high and low demand strategy, suggested in Armenia to buy time for putting in place the basis for a more market-dri- U.S. dollars per flat per year 350 ven heat supply. Survival (low demand) Armenia is not typical for the region because Consolidation (high demand) 300 of its relatively low heating requirements, its extended natural gas distribution system and 250 that its only cogeneration plant is industry- based and not critical for the power system. In 200 Moldova and Kyrgyzstan, however, the com- bined heat and power plants in Chisinau and 150 Bishkek are relatively modern and they are needed for the power system. There are thus 100 additional factors favoring maintenance of the district heating system. Careful planning at 50 many levels is required however to make heat from those systems as affordable as possible. 0 Comined Large Small Block Ind. Ind. Solid fuel Liquid Kerosene heat heat heat natural electric natural stove propane stove Parts of the centrally supplied DH system that and only only gas stove gas gas are not economic to supply must be shut down; power boiler boiler stove stove minimum investment plans to make heat sup- Note: Percent of population purchasing heating services at different prices: ply and consumption more efficient must be 80% at US$50 per year, 60% at US$70 per year, 40% at US$100 per year. Source: COWI 2002a, ERM 2002. devised; financing sources must be identified; Rethinking Heat Supply 21 Table 4-1 Payback times for energy efficiency the type of heating system, such as repairing or investments in buildings replacing broken windows and doors, and insu- Payback lating roofs and walls. Other measures are target- Investment time (years) ed at reducing the losses caused by a building's Reducing drafts, weather stripping windows 1­2 internal heating system, such as insulating Installing heat meters at the building level 1­2 Installing meters for domestic hot water at pipes, balancing risers, controlling the tempera- the apartment level 1­2 ture of the heating system, metering at the Insulating pipes for domestic hot water 3­5 building and apartment levels, and allowing Installing controls for heating and domestic hot water for individual buildings 4­5 individual control of heat consumption. Installing radiator control valves 5­6 Table 4-1 lists the most common investments Insulating heating pipes 10 for reducing heat requirements and lowering Insulating roofs 25 Insulating outside walls 50 heating bills, as well as the payback times for Source: Based on SwedPower/FVB 2001. those investments, based on full heat service. However, the expected savings cannot always be taken for granted. In many cases where build- and management and institutional measures to ings are under-heated, households tend to make the remaining truly least-cost district heat- increase comfort first and save energy later, as ing systems viable both for producers and con- experiences from Lithuania's Energy Efficiency sumers must be identified. The latter requires and Housing Project show (see World Bank rebalancing tariffs between electricity and heat 2002). and commercializing the utilities.7 For pilot projects to be carried out in Armenia, the costs of improving buildings' The importance of building-based efficiency internal facilities to improve the centralized heat measures supply and to enable households to regulate and control heat and pay for it based on consump- The importance and feasibility of improving the tion were estimated to be $200­400 per apart- efficiency of delivering heat to consumers was ment (for equipment and installation, including highlighted in the previous sections of this new piping and radiators), depending on the chapter. For poor consumers it may however be size of the building. Typical investment and even more important to have their buildings recurrent costs that have been observed for improved in order to cut down on the heating metering and control installations in more than necessary to warm them up. As is well known, two million dwellings in Poland are reported in buildings in Eastern Europe suffer from bad con- Table 4-2. struction and neglected maintenance, particular- ly in common areas (for details see Annex 8). Conclusion These shortcomings lead to high energy con- sumption for heating that could be reduced con- When putting together information on the costs siderably by energy-saving investments in those of various heating options and the effective low buildings. Some measures are independent of level of heat demand in many of the poorer Table 4-2 Typical cost of heat metering and individual controls in apartments Cost per unit Units required Cost per apartment (US$) per apartment (US$) Total investment cost per apartment 145 Heat meter, building level 285 N/A 8a Heat cost allocator (HCA) 4.20 4 17 Thermostatic radiator valve 30 4 120 Annual cost of billing and N/A 15 serviceper apartment 1.50/HCA + 8.50/apartment N/A: Not applicable. a. Assumes 35 apartments per building. Source: Supplier information for Poland. 22 Coping with the Cold reduced by an estimated five to ten percent, on aver- countries in Eastern Europe and Central Asia, it age. The renovation of the transmission and distribu- becomes obvious that the traditional provision tion networks and installation of variable speed of heat at a full service level of 18o Celsius is too pumps has led to significant energy savings, again expensive in several countries of the region to estimated in the order of up to ten percent heat and enable district heat utilities to achieve cost pumping losses. Very dramatic reductions in water losses have also been achieved through the switch recovery. The high fixed costs of centralized from direct to indirect domestic hot water connec- heating systems make them relatively slow to tions, amounting to a decrease of over 85 percent in react to a heterogeneous heat demand. Tallinn, of almost 90 percent in Tartu and over 90 Decentralized heating options are less risky in percent in Parnu. The heat consumption in buildings this respect since they are modular. equipped with renovated substations has been esti- Set against this must be the social costs asso- mated to have been reduced by about 24 percent, on average" (World Bank 2000a: 7). ciated with using low-efficiency heating appli- 2. The effective indoor temperature would be 20oC, ances together with environmentally problemat- considering 2oC additional from appliances and body ic fuels such as wood or coal and power system temperature. reliance on cogeneration plants. In those cities 3. If gas is used as heating fuel in a more decentral- where incomes are growing, investments in high ized way, substantial investment could be needed in the network to enable it to carry a larger load than it efficiency and environmentally benign central- was originally designed for, as well as at the block ized heating may be justified. However, central- and building levels, for example in metering. ized heating justified on these grounds must be 4. The analysis is based on a cash-flow methodology equipped with metering and heat control where all future cash-flows are discounted by a dis- devices needed to give the poor control over count factor of ten percent a year. their expenditures. Energy efficiency invest- 5. A reduction of indoor temperature by 1o Celsius reduces heat consumption by about six percent. ments help achieve that goal but often involve 6. Households were asked how much they would be high initial costs poor consumers cannot afford. willing to pay for an improved heating system with Here governments should consider providing the following characteristics. It would provide support, such as with financial schemes that enough heat to heat each occupied room in an apart- enable consumers to distribute initial costs over ment, to a minimum of 16°C on a reliable 24­hour a day basis, for as many weeks per year as desired by a number of years, or even with outright grants the household; it would be installed at no cost to the that enable poor consumers to overcome those household; households could control the amount of initial costs. heat consumed using controls inside the apartment; bills for the improved service would be based on Notes meter readings of the actual amount of heat con- sumed and payments would be spread out over 12 months. The survey results were as follows: 80 per- 1. For example, in the Estonia District Heating cent of households agreed with payments of $50, 60 Project, considerable energy efficiency improvements percent with $70 and 40 percent with $100 (see ERM were achieved: "The Project has made efficiency 2001, section 7.5). gains in the areas of heat production, transmission, 7. For details see Swedpower/FVB (2001) and COWI distribution and consumption. In the production A/S (2002b). process, the specific fuel consumption has been CHAPTER 5 Providing Clean Heat in Fiscally- Sustainable Ways The challenge ahead is to design policies, insti- political pressure, some governments continue tutions and investments that enable all people to pump money into antiquated district heating (poor and nonpoor) to access clean, affordable systems that are providing a failing service. On heating. In an urban environment this is partic- the other side, consumers are refusing to pay ularly difficult because whole communities are their bills as the quality of district heating affected by these choices. Therefore it is critical erodes. that the choices allow poor people to opt in to Poor people cope with failing district heating the degree they wish to get the heat they want supplies and rising energy prices by substituting because they might not use and will not pay for less expensive dirty energy, including wood, the wrong investments. coal, and kerosene. But there are private and social costs associated with poor people's heat- What we know so far ing choices. Private costs include the opportuni- ty cost of the time spent collecting heating The transition has brought difficult choices to material (especially wood) and illnesses and governments trying to rationalize budgets and labor productivity losses associated with insuffi- to households trying to maintain living stan- cient heating. Social costs include air pollution dards. Energy utilities, which evolved in a cen- from the burning of dirty fuels and the environ- tral planning framework characterized by mas- mental costs associated with deforestation and sive price distortions and direct state control the loss of biodiversity. These costs must be over resource allocation decisions, are caught in taken into account when evaluating the eco- the middle. World Bank estimates of annual nomic implications of alternative heating poli- quasi-fiscal deficits in the power sector range cies and investments. from $34 million in Moldova1 to $188 million Policy instruments such as regulations, taxes, in Georgia (3.6 percent of GDP) to $1 billion in and subsidies coupled with investments in tech- Serbia. nology and institutional changes offer a way for- Unable to cover the costs of operations and ward. Policy instruments can be used to encour- maintenance, many centralized heating systems age the poor to make clean choices. Investing in have started deteriorating significantly. In an new technology or reengineering existing tech- effort to rescue the systems, governments often nology enables governments to do this in fiscally raise prices. But the absence of meters on these sustainable ways. If the goal is to provide access systems and the difficulty of disconnecting non- to clean and affordable heating, investments and paying customers make it difficult to enforce policy instruments must be explicitly funded to payment. People, particularly poor people, do cover the difference between household expendi- not like paying for heating systems that do not tures and the cost of supply. Finally, greater allow them to control their expenditures. There emphasis on commercialization of utilities, more is an additional challenge of asking people to involvement of the private sector and of house- pay for a service that used to be free and that holds themselves through community-driven keeps deteriorating. The net result is a low-level development activities are necessary to improve equilibrium where on one side, often due to the daily life of poor households in cold climates. 23 24 Coping with the Cold Policy instruments to encourage clean poor than poor households benefit from such choices subsidies. In addition, the nonpoor are likely to consume more clean energy than the poor, so Regulations the bulk of the subsidy ends up going to the nonpoor. Regulations involve designing and enforcing rules that limit household heating choices. They Lifeline tariffs influence household behavior by indirectly rais- ing the costs of using dirty energy. Examples Restricting fuel subsidies to an initial block of include setting limits on pollution emissions, consumption is less costly than providing regulating access to forest resources by enforcing across-the-board subsidies but preserves some of restrictions on the cutting of fuel wood, and set- their politically attractive universal protection ting standards that prohibit the use of certain (Lovei 2000). However, these function only technologies (such as wood stoves in high-rise when consumption can be controlled and is buildings). This means that the new choices are metered at the household level--preconditions only affordable if they involve decreasing con- not met by most existing systems and rather sumption or if clean fuels become less expen- expensive to retrofit, given the vertical piping of sive. In general, because the poor rely more on district heating and gas in buildings in the for- dirty fuels this approach places an additional merly centrally planned economies. In terms of burden on them. Regulations can also be diffi- targeting, a lifeline tariff depends on the share cult to enforce and are often expensive. Such of the poor connected to the utility. Because regulations might only be a way forward if more nonpoor than poor people use clean net- investments are made in more efficient tech- work energy, there will likely be leakage prob- nologies, such as improved stoves. lems. One solution is to design a block structure that includes a fixed fee for a very low mini- Taxes mum needs level of consumption followed by a significantly higher price for following blocks. An alternative to regulation is to tax dirty fuels to encourage households to make clean choices. Vouchers Taxes on dirty fuels are simple to administer (unless the fuel being taxed is traded illegally) Vouchers (or grants) are lump sum subsidies pro- but politically difficult to implement. They may vided to consumers based on personal or house- not make sense on equity grounds because they hold characteristics and tied to certain behavior. result in higher prices for poor households. It For example, during the winter poor households may be possible to return money raised by tax- may be given vouchers for kerosene that can be ing dirty fuels to consumers in the form of direct redeemed at their leisure. The greatest danger payments or by cutting another tax. Such with vouchers is the risk of voucher devaluation efforts, however, make such taxes much more through trading on secondary markets. Another difficult and costly to administer. problem with vouchers is how to target the poor. Solutions to both problems include mak- Subsidies ing vouchers nontradable and rationing their delivery. While more difficult to administer than Across-the-board fuel subsidies can help ensure taxes and subsidies, vouchers may well be fiscal- that poor people have access to clean, affordable ly less expensive if they can be effectively heat such as from gas or electricity. Though targeted. politically attractive, these types of subsidies are costly, and among poor people coverage is limit- Direct cash transfers ed to the share of connected households. When it comes to clean network energy other than Another approach is to provide poor households electricity, evidence presented earlier on connec- with direct cash transfers or untied lump sums. tion rates (Table 2.1) suggests that more non- Such transfers give households complete free- Providing Clean Heat in Fiscally Sustainable Ways 25 dom in deciding how to use the money. affordability reasons, district heating even if Transfers are usually based on eligibility criteria. modern, flexible, and well managed will most In practice, the coverage of the poor achieved by probably not be the least-cost heating system. these programs seldom rises above 60 percent Furthermore, the investments required may not (Lovei 2000). The problem is that this instru- be affordable for those cities most in need of ment does not ensure conditional behavior, them. such as burning clean energy. In Armenia it was In smaller towns, building- or apartment found that despite an additional transfer to help based individual heating options would normal- meet electricity payments in the face of a price ly be expected to be least cost and therefore increase, consumption of electricity dropped preferable over district heating. Individual con- and consumption of wood increased--particu- sumers decide how many points of service are larly among the poor (Lampietti and others needed, procure the needed equipment, and 2001). In addition, the transfers are usually too arrange for fuel supply. In densely built urban small to allow poor households to cover all their environments, however, individual heating is basic needs. relatively expensive--usually more expensive than any form of central heating at full heat ser- Investments in heating technology vice levels. It can also have negative environ- mental impacts, including air pollution and pos- It is almost impossible to find clean heat supply sibly deforestation. Any broader intervention in options that deliver full service at $25­40 a year. the heating sector needs to recognize these Even reduced service from clean fuels comes tradeoffs. with a price tag that many poor households can Finally, governments, IFIs and other decision ill afford. Thus it is critical to recognize that if makers need to keep in mind that heating is a the goal is to provide access to clean heating, local issue and solutions depend very much on investments must be explicitly funded to cover the local circumstances. Therefore the solutions the difference between household expenditures and recommendations in this report can offer and the cost of supply. only broad guidelines and need to be adapted There are two investment strategies. One is on the basis of local analyses. Decision makers to continue or increase reliance on large-scale, also need to realize that heating has important centralized heating technologies. The other is to linkages with the energy sector in general and encourage smaller, less centralized technologies the power sector in particular through com- such as gas-fired boilers that could be supplied bined heat and power facilities. Since heating is under a variety of institutional settings. One the most important energy use of the residential danger of the current system is that it encour- and building sector, the fuel sources and impacts ages the buildup of vested interests with a stake of heating on energy networks need to be better in maintaining the status quo. integrated in national energy sector strategies. Central-planning-style district heating usual- ly offers one fixed level and quality of heating, Centralized options impeding flexibility. In an environment where the government can no longer afford broad sub- In many of the poorer countries of Eastern sidization of heat consumption (directly or indi- Europe and Central Asia and in many smaller rectly), heat supply options need to be towns of the region district heating systems are revamped and if necessary newly designed in a in dire need of extensive renovation if they are way that allows consumers to choose from a to be operational for the next 15-20 years. This range of heating levels with corresponding pay- makes the investment decision similar to a deci- ment levels. sion about a greenfield development. Chapter 4 If properly managed and provided basic shows the conditions that make modern district investments are made, centrally provided dis- heating the least-cost heating option. But while trict heating can be just as flexible as individual the costs of heat from a modern system will be heating. However, if heat demand is expected to lower than those from an old system, they will be low for the foreseeable future because of still not be affordable for many families. Full 26 Coping with the Cold heating service from an existing system tends to the number of radiators (see Box 3, Annex 8, for cost between $200 and $900 a year. The cost of the experience in Bulgaria). heat from a building boiler tends to be similar to that from district heating (see Chapter 4), but it Efficient stoves would be easier to increase capacity should demand increase over time. This decreases the Individual heating options can be clean and effi- financing needs and risks created by unused cient as well as flexible. In some countries capacity. Thus, even if a district heating system (Georgia, Mongolia) improved stoves for wood would be the most cost-effective under assump- and coal have been developed and commercially tions of full heat service, new investment should distributed. These stoves use much less fuel, be targeted first at block heating systems that burn much cleaner, and do not cost much more could later be joined and supplied from a more than a regular, inefficient stove (for Mongolia central generation facility, preferably on the see ESMAP 2001). Electric heaters are generally basis of cogeneration. not a feasible heating option for poor people Smaller-scale solutions such as building boil- because high electricity tariffs make anything ers could also be implemented more easily, but the most basic heating very expensive. In based on the decisions of just one building, addition electric heating may impose large which tends to be more homogeneous than an demand on the power networks that will require entire community or municipality. A coopera- major investments. If cheap nighttime electrici- tive or condominium could contract with, say, a ty can be provided, and the needed time-based gas company or an entrepreneur for the delivery meters installed, partial electric heating might of heat services. become an affordable option. Thus investing in efficient technology can produce substantial Meters and control options reductions in consumption, with this strategy being particularly important in places where All centrally provided heat supply options can prices are still very low but are bound to be fitted with meters and control options that increase. make the systems more flexible and allow users to choose the amount of heat and levels of com- Better insulation of buildings fort and spending. Many consumers in Eastern Europe and Central Asia have invested in meters Most buildings in Eastern Europe and Central and control devices, considerably reducing their Asia use two to three times as much heat as heat expenditures, increasing their comfort, or buildings in comparable climates in Western both (see Box 2, Annex 8, for the experience in Europe. Improving the tightness of the building Poland and JP 2002 more generally for the inter- shell lowers the requirements for heating and so national experience). Whether and how much the cost of achieving a minimum or desired consumers can actually save depends on the comfort level. Such measures are rather expen- level of over- or underheating and the relation- sive, however, and unless very bad conditions ship between the system's fixed and variable are remedied (such as broken windows), they are costs. Only variable costs can be reduced in pro- not as cost-effective as many measures on the portion to reduced consumption. In systems supply side. Payback times of 5­10 years are typ- where underheating is common, many con- ical; exterior insulation has an even longer pay- sumers tend to increase their comfort rather back time. than save energy (see World Bank 2002). In gen- However, if poor consumers could receive eral, however, individual metering and control financial support for improved insulation, this can save 15­20 percent of heat energy. might enable them to participate in communal In some countries where individual meters heating services (such as building boilers) that are not yet in place, a crude approximation of a they could otherwise ill afford. A revolving fund flexible district heating system has been used. could be established to help the population Consumers are allowed to disconnect some of finance small heat-saving investments. This their radiators, and payment is then based on fund, which would be preferentially geared Providing Clean Heat in Fiscally Sustainable Ways 27 toward poor households, would finance up to heat supply options. Institutional challenges 100 percent of such investments (depending on need to be considered carefully as well. District the financial condition of the debtor) through heating requires an organization with advanced loans of up to, say, $200. The loan would be technical, financial, and organizational capacity. administered by local bodies (such as the munici- But financial and organizational capacity is usu- pality or household associations) and would ally in very short supply in municipal utilities. If automatically be repaid over several years from district heating is to survive without continuous the savings achieved on the heating bills, government subsidies, at a minimum, the heat assessed with some appropriate algorithm. Part supply companies need to be commercialized. of the savings (say, 70 percent) could be used to This means a streamlined organizational struc- repay the loan, with the rest benefiting the ture, efficient operation (rationalization) based household until the loan is repaid. Financing on contractual obligations, full cost recovery schemes such as this, appropriately communicat- through enforcement of payment and no under- ed, would offer substantial incentives to the poor taking of "social" obligations. for improving their condition (see Kantor 2001). District heating is often protected from com- For very poor households outright grants could petition by outdated norms and inappropriate be considered, especially if this would enable regulations being applied to less centralized heat them to participate in community-based activi- supply options. Old norms are often still in place ties to improve heating services. that require enormous reserve capacity (say, 100 percent), making it extremely costly to provide Capital and recurrent costs heat from building boilers. There is a tendency to apply the regulatory supervision, especially tariff The main barrier to poor households accessing setting, in place for district heating also to non- clean (modern) infrastructure services may well district heating. However, these smaller-scale be the high initial costs--that is, the connection heating systems are much more open to compe- to network energy sources and purchase of neces- tition and could be provided entirely by the pri- sary appliances. More centralized heating systems vate sector on the basis of commercial contracts. tend to have higher capital costs relative to vari- An appropriate enabling framework would con- able and fuel costs (see Figure 4-4). For network sist of arrangements to protect consumers from energies (that is, in a utility context) much of the monopolistic pricing and enforce safety and capital cost is initially expensed by the utility and environmental standards, and provide for some charged to consumers over a fairly long period support mechanisms to ensure access to heating through monthly fees. In principle, this approach services for poor households. In addition, if con- makes the service more affordable for house- sumers associate, for example in condominiums, holds. But many poor people find themselves these may act either as suppliers or contract on increasingly unable to afford high monthly pay- behalf of their members for heating services. This ments for network energy. Innovative financial would provide more bargaining power vis-à-vis instruments need to be explored to distribute any heat supplier. costs over a longer period. Microfinance instru- ments have been successfully used in some coun- Community development tries, including the Kyrgyz Republic and Romania, to finance small building boilers. Collective action at the building or community level offers one promising approach to bringing Institutional challenges down the costs of heat supply. Groups of house- holds acting together can produce significant Commercialization of district heating and private economies of scale in consumption, reduce trans- provision of heating services action costs in collections, and provide guaran- tees to service providers. In fact, such collective The costs of providing heat and the flexibility interface is indispensable for any central heating that a heat supply option offers are not the only option, since individual connections and discon- considerations when making decisions about nections are technically difficult and expensive. 28 Coping with the Cold While promising, collective action is not as holds to make clean heating choices and commonly observed as one might expect. People improve cost recovery. If the focus is on promot- are often observed collaborating on a reactive ing clean non-network fuels, then targeted rather than proactive basis. For example, they vouchers for equipment and possibly fuel may organize the repair of an elevator when it breaks be promising instruments. If the focus is on pro- or a roof when it leaks, but rarely do they gather moting access and use of network energy then money from residents to arrange service con- lifeline tariffs, as long as the size of the blocks is tracts, such as heating, in advance. When asked set to minimize leakage to the non-poor and the why they do not engage in more collective government explicitly compensates utilities for behavior residents typically indicate that it is not any social transfers they are asked to provide, their responsibility to get involved in the run- may offer a promising alternative. ning of the heating system and that free riding In terms of investments, earlier chapters indi- will encourage the state to supply heat. cate that there is a role for public sector inter- Community development activities that will vention in either increasing access to low-cost encourage effective collective action include clean non-network energy or extending clean building trust between community members, energy networks into poor areas. Network increasing organizational and management investments must be coupled with investments skills, and promoting transparent procurement in metering and control options and with con- and implementation. These activities can be sumption-based billing, allowing users to choose supported through investments in training in the amount of heat and levels of comfort and collective decision-making, conflict resolution spending. Particularly promising, especially in and financial management as well as develop- areas where large increases in clean fuel prices ment of standard by-laws for cooperative organi- are expected, are investments in efficiency zations and codes of conduct for managers. improvements and insulation that can produce In addition, financial issues are key concerns substantial reductions in consumption. These in organization of collective action. People do investments must be coupled with innovative not want to handle the difficult task of manag- financial instruments that enable consumers, ing non-payment, nor do they want to be particularly the poor, to distribute capital costs responsible for depriving their neighbors of over a longer period. heating. Collective actions that focus on labor In addition to policy and investment instru- contributions rather than financial participation ments, there is considerable room to increase may be more successful, especially in poor areas. the institutional efficiency of heating service Where financial participation is required (for delivery. This can be achieved through a combi- example, via the collection of user fees or repair nation of enterprise commercialization and charges), it is important to recognize potential training of heating service providers, promoting implementation problems in buildings where effective collective action at the community there are many poor households and a great level, and encouraging participation of private variation in household incomes. sector service providers. Conclusion Note There is a broad range of policy and investment 1. Probably in 1998 (about 2.3 percent of GDP); since instruments available to encourage poor house- then reduced considerably to US$7­8 million through privatization. ANNEXES 29 Annex 1. Purchasing power parity and exchange rate conversions To provide consistency in comparison of poverty rates according to the same standards across the countries, US$2.15 per capita-day and US$4.30 per capita-day, household consumption in local currencies were brought to 1999 values by using PPP adjusted rates. Country Survey Period 1996 PPP CPI 1996 CPI PPP Exchange Exchange Average Survey Rate Adjusted Rate* Reference to the Survey Period Period** Armenia Nov 99 ­ Jan 00 128.54 118.7 146.3 158.47 Kyrgyz Sept ­ Dec 99 3.47 130.4 303.3 8.07 Croatia Oct 97 4.16 104.3 117 4.66 Moldova Feb ­ May 99 1.16 120.9 191 1.83 Tajikistan May 99 56.94 100 411.5 234.29 Lithuania Sep 98 1.55 124.6 142 1.77 Latvia Sep 97 0.26 117.6 133.2 0.29 * The rates that were used in World Bank 2000b. ** 1996 PPP exchange rates were adjusted for domestic inflation by multiplying the 1996 PPP exchange rate by the ratio of CPI (Consumer Price Index) in the country in the survey period (average for the period) to the average CPI in 1996. The exchange rate was used to convert the expenditures reported in the surveys from local currency to PPP USD. In calculation of expenditure on energy or other monetory values such as unit prices of different fuel types, we used inflated exchange rates rather than PPP adjusted values. There are two reasons for that. First, we wanted to see the differences in energy policies by looking at the differences between unadjusted values. Second, the comparison among countries, such as the rate of energy expenditure in household budget, did not require such adjustment. Nevertheless, after converting variables to US dollars by using the rates within the survey period, they were inflated to 1999 levels for consistency. Country Survey Period Local Currency Survey Inflation US$ Rate, Year US$ Inflated to Rate 1999 Armenia Nov 99 ­ Jan 00 Armenian Dram 525.48 · 525.48 Kyrgyz Sept ­ Dec 99 Kyrgyz Soms 43.84 · 43.84 Croatia Oct 97 Croatian Kunas 6.101 0.97 5.877 Moldova Feb ­ May 99 Moldovan Lei 9.99 · 9.99128 Tajikistan May 99 Tajik Rubles 1128.00 · 1128.00 Lithuania Sep 98 Lituanian Litas 4.00 4.00 Latvia Sep 97 Latvian Lats 0.58 0.97 0.558 · Survey period is the same as baseline year. Lithuania uses "fixed currency rate" policy. 30 Poverty in Eastern Europe and Central Asia (1996 U.S. dollars adjusted for purchasing power parity) National Urban mean Urban head Urban mean per head count per capita count at capita spending Country at $2.15 a spending $2.15 a day among the poor (survey year) day (dollars a day) (percent) (dollars a day) (percent) Armenia (1999) 38 2.6 41 1.6 Croatia (1997) <1 16.2 <1 - Kyrgyz Rep. (1999) 47 2.9 44 1.4 Latvia (1997) 7 5.9 6 1.6 Lithuania (1998) 2 8.8 1 1.8 Moldova (1999) 35 5.6 13 1.6 Tajikistan (1999) 69 2.2 63 1.4 1 The seven countries included in the analysis were selected based on the availability of recent (1997 or later) household survey data with sufficient survey questions about energy expenditure patterns. Studying household energy consumption patterns requires data on prices, quantities, and expenditures. Because these data are not available from all the household surveys, some assumptions are required. These assumptions and summary statistics for the data presented in this paper are provided in Annexes 2 and 3. Source: Author's calculations based on household survey data. 31 Annex 2. Data Assumptions Variables and their definition Fuel Abbr. Unit Use* Price Quantity Expenditure Liquid Gas LPG kg useLPG pLPG_kg qLPG_kg expLPG Kerosene KER l useKER pKER_l qKER_l expKER Coal COL kg useCOL pCOL_kg qCOL_kg expCOL Wood WOD kg useWOD pWOD_kg qWOD_kg expWOD Electricity ELE kwh useELE pELE_kwh qELE_kwh expELE Central Heating CH gc useCH pCH_gc qCH_gc expCH Central Gas CG m3 useCG pCG_m3 qCG_m3 expCG * Use is defined as positive consumption of fuel Legend: : Reported in the survey c: calculated means dividing expenditure by either quantity or price. Thus calculated prices are actually average costs. : from outside source ·: missing / not available M: Monthly S: Seasonal A: Annual Armenia Fuel Use Price Quantity Expenditure Source Period Source Period LPG A c KER A c COL A c WOD A c ELE * M M CH * c S CG * c M Kyrgyz Fuel Use Price Quantity Expenditure Source Period Source Period LPG S c KER c S COL c S S WOD c S ELE c M CH c S CG c M S 32 Croatia Fuel Use Price Quantity Expenditure Source Period Source Period LPG A A KER · A A COL A A WOD A A ELE M M CH · · M CG A A Moldova Fuel Use Price Quantity Expenditure Source Period Source Period LPG · · KER · · · · COL S c WOD S c ELE M c CH c M CG M M Tajikistan Fuel Use Price Quantity Expenditure Source Period Source Period LPG · · · · KER c M COL · · WOD · · ELE c M CH c S CG c S Lithuania Fuel Use Price Quantity Expenditure Source Period Source Period LPG · · · KER · · · · COL C A WOD C A ELE C A CH C A CG C A 33 Latvia Fuel Use Price Quantity Expenditure Source Period Source Period LPG c M KER · · · · COL c S WOD c M ELE c M CH c S CG c S 34 Annex 3. Household Energy Consumption Summary Statistics Armenia Variable Obs Mean Std. Dev. Min Max useLPG 1350 0.24963 0.432959 0 1 useKER 1350 0.117778 0.322464 0 1 useCOL 1350 0 0 0 0 useWOD 1350 0.491852 0.500119 0 1 useELE 1350 0.985926 0.11784 0 1 useCH 1350 0.137037 0.344014 0 1 useCG 1350 0.140741 0.347883 0 1 Variable Obs Mean Std. Dev. Min Max pLPG_oe 331 0.589047 0.059046 0.44925 0.7188 pKER_oe 133 0.400775 0.107283 0.230949 0.692848 pCOL_oe 0 pWOD_oe 573 0.164662 0.047353 0.050612 0.278368 pELE_oe 1350 0.559712 0 0.559712 0.559712 pCH_oe 900 0.180348 0.053126 0.147419 0.265964 pCG_oe 1350 0.116512 0 0.116512 0.116512 Variable Obs Mean Std. Dev. Min Max qLPG_oe 1340 18.16422 38.66249 0 211.8 qKER_oe 1336 3.540856 13.90509 0 123.6 qCOL_oe 1350 0 0 0 0 qWOD_oe 1341 151.6617 195.8701 0 1128 qELE_oe 1347 136.2212 103.6615 0 611.065 qCH_oe 0 qCG_oe 1317 111.1877 363.6023 0 2744 Variable Obs Mean Std. Dev. Min Max expLPG 1337 10.74096 23.26625 0 137.0176 expKER 1313 1.104202 4.847675 0 57.09066 expCOL 1350 0 0 0 0 expWOD 1255 22.4295 30.83066 0 159.8539 expELE 1348 65.7214 54.71257 0 342.0682 expCH 1296 6.066617 19.36755 0 102.7632 ExpCG 1317 12.21801 39.99641 0 274.0352 35 Kyrgyz Variable Obs Mean Std. Dev. Min Max UseLPG 2344 0.367321 0.482178 0 1 UseKER 2342 0.190009 0.392391 0 1 UseCOL 2345 0.352239 0.47777 0 1 UseWOD 2346 0.252344 0.43445 0 1 useELE 2352 0.992347 0.087165 0 1 useCH 2348 0.498722 0.500105 0 1 useCG 2351 0.301999 0.459223 0 1 Variable Obs Mean Std. Dev. Min Max pLPG_oe 859 0.219054 0.049807 0.080773 1.220566 pKER_oe 2364 0.470876 0 0.470876 0.470876 pCOL_oe 705 0.006539 0.002238 7.03E-06 0.021082 pWOD_oe 2364 0.078325 0 0.078325 0.078325 pELE_oe 2288 0.051957 0.003561 0.048841 0.063651 pCH_oe 2364 0.040267 0 0.040267 0.040267 pCG_oe 224 0.070621 0.058126 0.005532 0.79069 Variable Obs Mean Std. Dev. Min Max qLPG_oe 2063 13.94205 25.47925 0 148.26 qKER_oe 2326 0.558355 1.522229 0 11.6261 qCOL_oe 2228 392.1522 653.3833 0 3246 qWOD_oe 2343 67.46366 152.391 0 1164.898 qELE_oe 2311 202.7521 144.6981 0 1321.714 qCH_oe 0 qCG_oe 1862 8.281684 26.06614 0 183.26 Variable Obs Mean Std. Dev. Min Max expLPG 2068 8.977586 16.18402 0 93.0657 expKER 2348 0.26292 0.717712 0 5.474452 expCOL 2362 10.65464 17.48504 0 91.24088 expWOD 2361 1.74794 3.966445 0 30.41363 expELE 2311 10.96501 8.970309 0 82.11679 expCH 2347 5.603208 8.929994 0 63.86861 expCG 2332 1.838003 3.652303 0 24.63504 36 Croatia Variable Obs Mean Std. Dev. Min Max useLPG 1726 0.445539 0.497169 0 1 useKER 1726 0.067787 0.251453 0 1 useCOL 1726 0.011008 0.104371 0 1 useWOD 1726 0.303013 0.459694 0 1 useELE 1726 0.998841 0.034031 0 1 useCH 1726 0.352839 0.477992 0 1 useCG 1726 0.284473 0.451294 0 1 Variable Obs Mean Std. Dev. Min Max pLPG_oe 1726 0.878676 0 0.878676 0.878676 pKER_oe 0 pCOL_oe 1726 0.2064 0 0.2064 0.2064 pWOD_oe 1726 0.113133 0 0.113133 0.113133 pELE_oe 1726 0.938337 0 0.938337 0.938337 pCH_oe 0 pCG_oe 1726 0.248947 0 0.248947 0.248947 Variable Obs Mean Std. Dev. Min Max qLPG_oe 1721 41.47822 57.79245 0 317.7 qKER_oe 1709 85.66804 387.821 0 2472 qCOL_oe 1713 0.189492 3.197121 0 54.1 qWOD_oe 1720 274.6549 498.1277 0 2368.8 qELE_oe 922 298.5378 224.3949 0 1591.2 qCH_oe 0 qCG_oe 1455 192.3617 641.2096 0 4198.32 Variable Obs Mean Std. Dev. Min Max expLPG 1719 31.34008 43.2622 0 340.8491 expKER 1713 40.60634 179.6126 0 1274.475 expCOL 1713 0.420034 7.128711 0 136.3396 expWOD 1719 6.685913 16.20486 0 101.948 expELE 1720 303.4076 198.7088 0 1255.688 expCH 1707 15.99929 84.84733 0 615.5378 expCG 1715 86.78488 224.2414 0 1546.539 37 Moldova Variable Obs Mean Std. Dev. Min Max useLPG 1051 0.0647 0.246113 0 1 useKER 0 useCOL 1051 0.055186 0.228451 0 1 useWOD 1051 0.091342 0.288231 0 1 useELE 1051 0.856327 0.350925 0 1 useCH 1051 0.506185 0.5002 0 1 useCG 1027 0.652386 0.476445 0 1 Variable Obs Mean Std. Dev. Min Max pLPG_oe 1051 0.4093 0 0.4093 0.4093 pKER_oe 0 pCOL_oe 1051 0.093242 0 0.093242 0.093242 pWOD_oe 1051 0.097159 0 0.097159 0.097159 pELE_oe 1051 0.447449 0 0.447449 0.447449 pCH_oe 1051 0.150131 0 0.150131 0.150131 pCG_oe 1051 0.111262 0 0.111262 0.111262 Variable Obs Mean Std. Dev. Min Max qLPG_oe 0 qKER_oe 0 qCOL_oe 1037 7.51557 39.39946 0 324.6 qWOD_oe 1049 35.59276 128.02 0 1278.4 qELE_oe 1049 69.98328 67.69999 0 428.4 qCH_oe 0 qCG_oe 1044 309.6949 851.0945 0 4998 Variable Obs Mean Std. Dev. Min Max expLPG 0 expKER 0 expCOL 1037 0 0 0 0 expWOD 1050 3.162615 11.84604 0 124.2083 expELE 1045 52.55473 41.62928 0 240.2095 expCH 497 87.65471 44.8206 8.006982 240.2095 expCG 1047 25.07784 46.62881 0 399.9488 38 Tajikistan Variable Obs Mean Std. Dev. Min Max useLPG 0 useKER 544 0.011029 0.104536 0 1 useCOL 544 0.165441 0.37192 0 1 useWOD 544 0.347427 0.476591 0 1 useELE 544 0.998162 0.042875 0 1 useCH 544 0.012868 0.112807 0 1 useCG 544 0.056985 0.232028 0 1 Variable Obs Mean Std. Dev. Min Max pLPG_oe 0 pKER_oe 544 0.110119 0.018012 0.096829 0.134485 pCOL_oe 544 3.41E-05 0 3.41E-05 3.41E-05 pWOD_oe 544 0.0002 0 0.0002 0.0002 pELE_oe 544 0.026074 0 0.026074 0.026074 pCH_oe 535 0.133536 0.061392 0.027944 0.400684 pCG_oe 544 0.060024 0 0.060024 0.060024 Variable Obs Mean Std. Dev. Min Max qLPG_oe 0 qKER_oe 517 14.25632 49.50227 0 439.4666 qCOL_oe 0 qWOD_oe 0 qELE_oe 534 258.7201 371.0144 0 2325.6 qCH_oe 0 qCG_oe 520 32.35328 71.87463 0 472.6241 Variable Obs Mean Std. Dev. Min Max expLPG 0 expKER 517 1.432569 4.909651 0 42.55319 expCOL 0 expWOD 0 expELE 534 6.745936 9.673927 0 60.6383 expCH 519 0.035564 0.439801 0 7.092199 expCG 520 1.941974 4.314203 0 28.36879 39 Lithuania Variable Obs Mean Std. Dev. Min Max UseLPG 0 UseKER 0 UseCOL 5179 0.003862 0.062029 0 1 UseWOD 5179 0.022398 0.147989 0 1 UseELE 5179 0.928172 0.258229 0 1 UseCH 5179 0.430006 0.495124 0 1 UseCG 5179 0.546438 0.497887 0 1 Variable Obs Mean Std. Dev. Min Max pLPG_oe 5179 0.263279 0 0.263279 0.263279 pKER_oe 0 pCOL_oe 5179 0.130776 0 0.130776 0.130776 pWOD_oe 5179 0.106782 0 0.106782 0.106782 pELE_oe 5179 0.588235 0 0.588235 0.588235 pCH_oe 5179 0.000327 0 0.000327 0.000327 pCG_oe 5179 0.208884 0 0.208884 0.208884 Variable Obs Mean Std. Dev. Min Max qLPG_oe 0 qKER_oe 0 qCOL_oe 20 11249.16 8122.379 607.6788 35786.29 qWOD_oe 116 6400.556 5219.525 140.4732 25285.18 qELE_oe 4785 113.8147 74.71105 1.122 586.5 qCH_oe 0 qCG_oe 2800 179.4595 266.6966 8.760862 2626.822 Variable Obs Mean Std. Dev. Min Max ExpLPG 0 ExpKER 0 ExpCOL 20 1471.124 1062.215 79.47 4680 ExpWOD 116 683.4636 557.3509 15 2700 ExpELE 4785 66.94982 43.94768 0.66 345 ExpCH 2221 347.5222 183.1078 4.8 1276.53 ExpCG 2800 37.48614 55.70853 1.83 548.7 40 Latvia Variable Obs Mean Std. Dev. Min Max UseLPG 5269 0.292086 0.454765 0 1 UseKER 0 UseCOL 5269 0.001708 0.041298 0 1 useWOD 5269 0.018789 0.135793 0 1 useELE 5269 0.998292 0.041298 0 1 useCH 5269 0.805276 0.396026 0 1 useCG 5269 0.664832 0.472094 0 1 Variable Obs Mean Std. Dev. Min Max pLPG_oe 5269 0.345334 0 0.345334 0.345334 pKER_oe 0 pCOL_oe 5269 0.121209 0 0.121209 0.121209 pWOD_oe 5269 0.166873 0 0.166873 0.166873 pELE_oe 5269 0.600867 0 0.600867 0.600867 pCH_oe 5269 0.353782 0 0.353782 0.353782 pCG_oe 2432 0.764921 0.123135 0.761839 6.606911 Variable Obs Mean Std. Dev. Min Max qLPG_oe 4218 37.59387 114.6067 0 622.9412 qKER_oe 0 qCOL_oe 3915 690.3813 553.7428 1.774803 4449.432 qWOD_oe 100 2392.791 1639.365 0 7734.857 qELE_oe 3924 138.947 111.7738 0 897.557 qCH_oe 0 qCG_oe 4169 40.87255 53.1887 0 531.4257 Variable Obs Mean Std. Dev. Min Max expLPG 5253 10.3426 35.61531 0 206.5179 expKER 0 expCOL 5261 0.020445 1.482935 0 107.5614 expWOD 5217 1.677232 18.96123 0 322.6841 expELE 5249 62.1124 67.57205 0 498.6546 expCH 5255 55.28749 87.49977 0 548.563 expCG 5235 8.120874 12.07784 0 121.9029 41 Annex 4. Social Costs of Heating Options Health and productivity costs of not having enough heat In many recent surveys in Eastern Europe, households have complained about insufficient heat from dilapidated district heating systems and the resulting increases in illnesses. For example, in Sevastopol, Ukraine, it was reported that in 56 percent of households somebody had gotten sick because indoor temperatures were too low. A better heat supply would significantly reduce the number of sick days and so increase productivity. Moldova's heat supply is also severely constrained, with many households subjected to indoor temperatures of just 5-10º Celsius during the heating season. Nearly three-quarters of survey respondents connected to district heating said that they were too cold last winter. Among the 40 percent of urban households surveyed that are not connected to district heating, the average household heats only two rooms for less than five hours. About a third of urban households reported that at least one family member got sick during the winter, and many believed that low indoor temperatures were the reason. These families lose income and have to bear the costs of treating the illnesses. National productivity and GDP also suffer. Health costs of burning dirty fuels Air pollution, particularly indoor air pollution, is an increasingly important environmental and public health issue. A number of studies have established the possible effects of wood stoves and other dirty fuels on respiratory illness, particularly in young children and the elderly (Honicky and others, 1991; Xu and others, 1989). Airborne particulate matter from the burning of wood and coal is associated with chronic obstructive pulmonary disease, acute respiratory diseases in children, low birth-weight, higher infant and perinatal mortality, pulmonary tuberculosis, naso-pharyngeal and laryngeal cancer, and even lung cancer. Because the poor are more likely to burn wood and coal in their homes, they are also more likely to be exposed to higher levels of particulate matter. Indoor air pollution has both direct and indirect costs. Direct costs include time spent visiting doctors, sick leave for patients and individuals caring for them, and spending on medicine and health care. Indirect costs include pain and suffering. Treating an episode of respiratory illness in Armenia, including the doctor's visit, medicine, food, and lost labor costs, runs an average of $9. An individual in a wood-burning household is 2.5 times more likely to report an episode of respiratory disease than is an individual in a household that does not use wood--suggesting that poor people bear the brunt of the social cost from the burning of dirty fuels. Indoor air pollution levels may not result solely from heating, but may also be due to cooking with dirty fuels. Policies to improve indoor air quality must take that into account as well. It was also calculated that in Armenia indoor urban exposure to smoke causes the annual loss of 3,467 life years per 100,000 children under five and 120 life years per 100,000 women (Environmental Resources Management, 2001). The resulting economic cost to 42 these women and children is estimated to be $3.2 million a year. Details of these calculation can be found in Environmental Resources Management 2001. Environmental costs of deforestation Forests are a valuable resource for their timber, watershed protection functions, and biodiversity values, and for sequestering carbon. Yet they are one of the most mismanaged resources in developing countries because they are often seriously undervalued. Many of their environmental benefits do not enter markets, and poor governance has often encouraged illegal activities. Moreover, many policies and investments aimed at addressing problems in one sector, such as energy, may affect forests in ways that are not well understood or are disregarded. In almost all countries wood is among the least expensive fuels, so when the price of another fuel increases, wood may be substituted. For example, in January 2000 the government of Armenia eliminated the increasing block tariff for electricity in favor of a single price, leading to a 47 percent increase in the price of electricity for residential consumers. When households were asked how they responded to this change, 80 percent reported that they substituted away from electricity--and more than 60 percent said that wood was the primary substitute. The increased reliance on wood is particularly common among the urban poor, who now use wood for heating and cooking. Although deforestation is likely to continue even with good economic management and governance, in Eastern and Central Europe and the former Soviet Union over the past 10 years it may partly be the result of sector policy spillovers coupled with lack of governance. Energy policies have often led to rapid increases in electricity prices in countries with more or less open access to forest resources. Opportunity cost of collecting wood The poor often must spend time collecting wood and other materials to keep themselves warm. This is time that they could spend earning money or that their children could spend in school. The decision of whether to collect wood or to buy it depends on the opportunity cost of one's time and the price of wood. Thus collecting wood is a substitute for buying wood--when the price of wood rises, the poor will spend more time collecting it. In Armenia only 20 percent of poor urban apartment dwellers buy wood, while more than 27 percent use it. The difference is that many households collect and cut their own wood rather than purchasing cut wood. Poor households spend nearly twice as much time collecting wood than do nonpoor households. Over the course of a year a poor household might spend more than 10 person-days collecting and cutting the 200 kgoe of wood needed for heating. Assuming a daily wage for unskilled labor of $2-3, this is equivalent to $20-30 a year in heating costs, which is consistent with the spending data presented above. 43 Annex 5. How to Estimate the Demand for Heat Separating the demand for heat from nonheat energy is difficult in survey data because households consume a mix offuels for a variety of purposes. For example, one household may use wood for heating and cooking in the winter and LPG for cooking in the summer. Another may use electricity for heating and gas for cooking in the winter and electricity for air conditioning and gas for cooking in the summer. One approach to identifying heat consumption is to use norms to net out basic needs, then study the residual. But that approach obscures the variations in consumption and spending patterns that are of interest here. To solve that problem, we developed a new approach to estimating heat demand. The approach relies on two subsamples: households that are connected to the central heating network and report that central heating is their only source of heat, and households that have no central heat. For the first group, all non-central heat energy consumption is for nonheating purposes such as lighting and cooking. Comparing the total energy consumption (not including central heat) of these two groups of households makes it possible to isolate the energy used for heating. A scatter-plot illustrating this relationship for the three countries is presented below. We exploit this natural experiment in our data to develop and estimate a nested heat demand model. The model specification and interpretation of the coefficients is presented in the box below. 44 Energy Consumption Scatterplots A. Armenia 1400 1200 including CH Connection 1000 (not Heat) 800 No CH connection year 600 per Central 400 Heat Consumption 200 KGOE 0 0 500 1000 1500 2000 2500 3000 3500 4000 Household Income per year B. Kyrgyz 4000 3500 CH Connection including 3000 2500 No CH connection (not heat) 2000 year 1500 central per 1000 Heat Consumption 500 KGOE 0 0 500 1000 1500 2000 2500 3000 3500 Household Income per year C. Moldova 4000 including 3000 CH Connection No CH connection (not Heat) 2000 year per Central1000 Heat Consumption KGOE 0 0 1000 2000 3000 4000 5000 Household Income per year Source: Author's calculations. 45 Household heat demand specification We start with the following reduced form equation: qOEi = 0 +1incomei +2 p _ nhi +3hhsizei +4 CHi (1) i where energy consumption (qOE) for household (i) is a function of income, price index of non heat energy (p_nh), and central heat (CH). For households with central heating (CHi=0) the equation becomes qOEi = 0 +1incomei +2 p _ nhi +3hhsizei (2) and for households without central heating (CHi=1) the equation becomes qOEi = 0 +1incomei +2 p _ nhi +3hhsizei +4 (3) i Since households with central heating do not consume other fuels for heating purpose, the difference between the two equations, or 4 , can be interpreted as a measurement of heating consumption for households without central i heating. Therefore, qOE _ heatingi =4 (4) i Suppose, restricting i to be households without central heating, the demand function for heating can be specified as qOE _ heatingi = 0 + 1incomei + 2 p _ hi + ' (5) where p_h is the price of heat and X'B is a vector including number of rooms, housing type, and temperature. substituting qOE _ heating i = 4 into (5) yields i 4 = 0 + 1incomei + 2 p _ hi + ' (6) i Since we don't know 4 equation (6) can not be estimated directly. However, the coefficients for equation (6) can be i estimated by linking (6) with (1) through 4 . Substituting (6) into (1) leads the following new equation i qOEi = 0 +1incomei +2 p _ nhi +3hhsizei + 0CHi + 1CHiincomei + 2CHi p _ hi + ' ...........(7) which can be estimated directly, and the coefficients for demand function for heating are coefficients for CHi,CHiincomei,CHi p _ hi and ', respectively. The variables included in the model are explained in the box. The nested heat demand model fits the data well in all three countries. F-statistics are highly significant and the R- square is in the range of 0.4 to 0.5. All of the variables, except temperature, have the expected sign and are statistically significant at the five percent level, increasing confidence in the model. Energy consumption increases with both income and household size and it decreases with energy price. Households with central heat consume less energy than those without. Heat consumption increases with income and decreases with price. Households with more rooms consume more heat. Households living in apartments consume less heat than in houses. Standard diagnostic tests were performed to verify the validity of the Ordinary Least Squares procedure. These tests, outlined in detail in Annex 6 of this chapter, indicate the empirical results are reliable. The main disadvantage of this approach is that we are measuring the demand for energy for heating rather than for heat itself. But we cannot measure the demand for heat directly because we do not have data on indoor temperatures or the efficiency of heating appliances. This lack of data prevents us from directly exploring how much variation there is in actual heat consumption between the poor and nonpoor. 46 Heat demand estimation results Armenia Kyrgyz Moldova Description Coef. t-stat Coef. t-stat Coef. t-stat Per capita exp. 17.59 2.91 54.03 2.01 17.52 0.36 Quintiles Non-heat energy price index -125.73 -2.42 -3485.13 -6.01 -1291.20 -13.04 Household size 17.69 4.20 69.14 4.23 18.87 0.48 Central heat(0) 196.88 5.27 377.37 2.65 576.25 3.03 otherwise (1) CH x per capita 0.06 3.87 0.31 3.51 0.14 1.45 expenditure CH x price of primary heat fuel -445.03 -9.62 -6123.74 -6.87 -1443.27 -3.49 CH x number of rooms 34.64 5.27 166.18 8.45 74.75 2.19 CH x apartment (1), other (0) -67.33 -4.89 -512.62 -6.63 -40.52 -0.12 CH x monthly temp. 3.74 1.34 13.84 0.69 -108.37 -2.83 (Nov-Feb) Constant 146.07 2.98 317.35 1.89 730.01 2.74 Rsquared 0.50 0.47 0.41 Fstatistic F(9, 734) = 81.69 F(9, 904) = 90.24 F(9, 399) = 30.72 N 744 914 409 Source: Author's calculations. We can, however, use a fixed effects model on 2001 Armenia data to capture average differences in the total energy consumed by households using different heating technologies--electric heaters, gas stoves, kerosene stoves, and wood stoves (see annex 4). Surprisingly, the analysis reveals that households using electric heaters consume 87 kgoe less energy a year than the average, while those using wood consume about 100 kgoe more. These findings suggests that, contrary to popular belief, electricity might be the most environmentally friendly form of heating for users, and it may be more efficient from the perspective of private consumers. This unusual result has several possible explanations. People using electric heaters may maintain a lower indoor temperature because electricity is more expensive. Or they may consume less energy because electric heaters are more efficient. Finally, it may be easier to manage energy consumption with electric heaters because, unlike wood stoves, they can be turned on and off and moved from room to room. 47 Annex 6. Validity of Heat Demand Model A nested OLS model is used to identify the heat demand for households. To verify the validity of the OLS assumptions of this model, a series of statistical tests have been conducted. Basically, for an OLS model, we assume the following conditions to be held: · Linearity of the regression model · The matrix of regressors is full-rank · Homoscedasticity The linearity assumption is not as narrow as it might first appear. In the classic OLS regression, linearity suggests that the parameters and the disturbance enter the equation is a linear fashion. Therefore, the relationship among the regressors does not have to be linear. A typical verification of this assumption is to use other alternative function forms to test the robustness of the linear specification. A log-linear regression model has been specified and estimated. The point estimates are very similar to those of our OLS specification. Therefore, we have no particular reason to distrust the validity of linearity. Several multi-collinearity tests have been conducted and they unanimously reject the hypothesis that the regressors in the model are collinear. Hence, the matrix of our regressors is shown to be full-rank. Homoscedasticity is not particularly restrictive in the context of our analysis. Even if the homoscedasticity is violated, i.e. the error terms in the regression are heteroscedastic, the point estimates obtained are still consistent. In other words, given a large sample, our estimates based on heteroscedastic OLS regression are still valid and unbiased. In this study, Cook-Weisberg test for heteroscedasticity using fitted values of dependent variables have revealed heteroscedasticity problems. This might cause inefficiency but the estimates are still unbiased given that other OLS assumptions are not violated. In addition, alternative specifications have to be used to test the robustness of the OLS estimates, and the point estimates cross different specification are very consistent to the ones we obtained from OLS regression. Particularly, this shows that our estimates on price elasticity and income elasticity are valid. 48 Annex 7. Fixed Effects of Different Heating Techniques Model: yij=a+xijb+vi+eij , where i is the heating device index and j is the household index. a+vi stands for the different intercepts of demand function across different heating techniques. Obviously, this term measures the average of the households who use a specific heating device, i. . xtreg qOE_oe welf p_nh hhsize CH CHxe CHxp CHxs if h_dev~=2, fe; Fixed-effects (within) regression Number of obs = 822 Group variable (i) : h_dev Number of groups = 4 R-sq: within = 0.1532 Obs per group: min = 43 between = 0.5014 avg = 205.5 overall = 0.1454 max = 349 F(7,811) = 20.96 corr(u_i, Xb) = 0.0881 Prob > F = 0.0000 ------------------------------------------------------------------------------ qOE_oe | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---------+-------------------------------------------------------------------- welf | -73.54967 12.59018 -5.842 0.000 -98.26285 -48.83648 p_nh | -.1238701 .0404127 -3.065 0.002 -.2031959 -.0445442 hhsize | 18.80239 3.083842 6.097 0.000 12.74914 24.85564 CH | -89.51655 28.22341 -3.172 0.002 -144.9161 -34.117 CHxe | .0000478 .0000172 2.771 0.006 .0000139 .0000816 CHxp | .1129281 .084609 1.335 0.182 -.0531504 .2790065 CHxs | 1.848796 .5355006 3.452 0.001 .7976657 2.899927 _cons | 294.9749 18.84356 15.654 0.000 257.987 331.9628 ------------------------------------------------------------------------------ sigma_u | 94.457774 sigma_e | 137.78127 rho | .31972649 (fraction of variance due to u_i) ------------------------------------------------------------------------------ F test that all u_i=0: F(3,811) = 45.80 Prob > F = 0.0000 49 Annex 8. Key Technical Characteristics of District Heating Systems and Housing Stock in Eastern Europe and Central Asia 1 Box 1: What is district heating? Usually, the term describes a system supplying heat produced centrally in one or several locations to a non- restricted number of customers. It is distributed on a commercial basis by means of a distribution network using hot water or steam as a medium. Often, the heat is also used for provision of domestic hot water and industrial purposes, such as process heat. Although most people understand district heating to be large centralized urban heating systems, many national statistics also include very small heating systems. Furthermore, systems that supply only steam for industrial purposes are usually also called district heating systems. In fact, the term district heating system is usually linked with the activities of the respective district heating company. Such a company usually operates larger centralized district heating networks, smaller isolated "block" heating systems (supplying only a small number of buildings), and even individual boilers in single buildings. Source: ESMAP 2000. Heat production. In the Russian and Eastern European schemes, district heating is supplied from cogeneration plants and/or heat-only-boiler (HOB) plants. Where cogeneration is used, the peak and reserve capacity requirements are covered by HOBs. Typically, large district heating systems have from one to three cogeneration plants and several hundred HOBs. A CHP plant is a technically more efficient way to produce heat and power than separated power generation and heat production. In Eastern Europe, the typical CHP plant efficiencies are around 70-75 percent as compared with 80-90 percent in Western Europe. The efficiency of the older HOBs in Eastern Europe are only 60-80 percent. However, with the introduction of modern automation and control systems, replacement of burners and cleaning of boiler surfaces, the efficiency can be typically increased to 85 percent. The efficiencies of new boilers are even higher, over 90 percent. District heating transmission and distribution networks. Typical factors leading to poor efficiency and various other operational problems in the DH networks in Eastern Europe include high levels of leakages, due to external and internal corrosion of pipes as well as insufficient pipe insulation, and the use of constant flow technology. Network leakages are common due to both internal and external corrosion of pipes. It is not uncommon for water to infiltrate pipeline channels from the outside and high ground water to cause external corrosion when pipe insulation material becomes wet and ventilation in pipeline channels is poor. In systems with high water losses, make-up water has to be added. In the worst cases, the networks may have to be refilled a hundred times a year or more, as compared to one or two times per year in well- maintained Western DH systems. Where water treatment is not adequate, poor quality of make-up water corrodes the pipes from the inside. Heat losses are also typically high due to inadequate insulation. The thickness of insulation is less than in Western countries. In DH systems where pipelines run above ground, e.g., in Armenia, theft of insulation material has become a problem. 1This section is based on ESMAP 2000, Gochenour 2001, and Martinot 1997. 50 The dominant mode of network operation in Eastern Europe has been constant flow regime. Basically, constant flow means that heat supply and heat demand are being adjusted by manually varying the flow temperature, typically in the range of 70-130ºC, based on the ambient outdoor temperature. The adjustment of heat supply (and thus consumption level) in the typical constant flow DH system is carried out centrally at the heat production plants. Heat distribution to individual buildings depends entirely on the hydraulic balance of the network, leading to inaccurate heat distribution to buildings, e.g., too high indoor temperatures, particularly during spring and fall. In a constant flow system, each hydraulic section of the DH network system can be supplied typically by heat from only one location, which does not generally allow for the heat load to be dispatched from the least-cost production source. Since the early 1990s, renovation and modernization of DH systems has started in Eastern Europe. The achievements of the World Bank financed investments project in Estonia (see endnote 1 in chapter 4) or in Poland (see Box 2 below) are quite representative. Consumer installations. Hot water, usually both for space heating and domestic hot water, is transported through pipeline networks to substations from where heat is distributed to consumers. The substations may be located within the individual buildings or, larger ones, serve a group of buildings through secondary networks, which typically involve four 4 pipes ­ two for space heating and two for domestic hot water. Those secondary networks usually experience high losses, and the technical lifetimes of those networks is short. In the typical Eastern European schemes, the larger substations traditionally dominate, while in Western Europe, most substations are installed in individual buildings. Both direct and indirect consumer connections are used in Eastern Europe. Indirect connection means that heat or domestic hot water is transferred by heat exchangers from the primary to the secondary network; the systems are thus hydraulically separated. Direct connection means that the water circulating in the DH network is introduced directly to the consumer installations, i.e., building pipes and radiators. In systems with direct domestic hot water connections, the hot tap water is supplied directly from the DH pipes and needs to be made up at the point of supply. There are about 300 cities in the FSU which utilize direct domestic hot water systems. The advantages of indirect heat and hot water transfer include, for example, more efficient network regulation, better protection against corrosion and reduced need for make-up water. Metering. Virtually no heat or hot water metering existed before 1990 in residential, commercial and public sector buildings in most countries in Eastern Europe. There was very little point in installing meters because consumers could not regulate the heat supply. The lack of regulation and metering resulted in too low or too high room temperatures and further losses of heat from the opening of windows to cool the sometimes overheated rooms. Since 1990, many countries have made substantial progress in installing regulation and metering equipment. Many countries (Poland, Hungary, Bulgaria) have introduced mandatory metering at the building level, and in many cities in the Czech Republic, Hungary and Poland, for example, the metering rate is now close to 100 percent. In many countries of the FSU (excluding the Baltics), however, very few (under 1 percent) residential buildings have meters. 51 Box 2: Experiences with heating metering and billing reform in Poland With partial support from a World Bank loan over 1991­99, the four Polish cities of Warsaw, Kracow, Gdansk and Gdynia undertook renovations of their heat supply systems, disseminated building-level heat meters for existing buildings, and reformed the heat tariff from a square- meter based tariff to a two-part tariff charged at the building level. Results in Four Cities 1991/92 1999 Change (percent) Household heat bill subsidy (%) 67 <5 (1994) na Heat bill charged to households (1999 US$/m2) 13.7 6.2 -55 Heated floor area (million sq m) 63.8 68.6 7 Heat energy sold (gcal/sq m) 0.27 0.22 -18 Energy savings na na 22 The Government of Poland implemented energy sector reforms under which payment for heat gradually became the responsibility of households, and they began to use heat more efficiently. Households (or companies operating as their agents) invested in radiator valves (TRVs), heat allocation meters, better windows and some insulation. The internal piping systems of buildings generally were not changed--single-pipe vertical systems are still in place, but radiator bypass pipes have been added where not already in place. A key result was that the costs of heating a given apartment area fell by 55 percent, due to efficiency improvements by consumers, and to technical, operational and management improvements in the heat supply companies. This reduction in costs helped to make the removal of the subsidy less burdensome to households. Nationwide, household heat subsidies, provided by municipal governments, have been reduced from 78 percent in 1991 to zero by the end of 1997. Installation of building-level heat meters has been mandatory for all buildings since 1999. Use of heat allocation meters has become a popular way to allocate heat bills within buildings--a total of 5.5 million were installed as of 1997 in about 30 percent of the dwellings nationwide. (Apartment-level heat meters are generally considered too expensive.) More than 10 companies have been formed and compete in the market for billing services--including allocation meter installation, meter reading, billing and maintenance. Energy savings, reflected in customer heat bills, stemming from the reform (including savings from private investments spurred by the reform) typically range from 20 to 40 percent. Water quality improvements, however, were required before the metering could be effective. It also should be noted that apartment heat levels were generally adequate in Poland before the reform--in other cases (e.g. Lithuania), energy efficiency gains may be harvested more in terms of improved comfort level instead of energy savings. Source: World Bank 2000c Operation and maintenance. Maintenance in Eastern Europe has typically concentrated on repairing damage that has occurred and not on preventing it, although there are exceptions. Repair works of production plants and networks are usually carried out in summer, typically during a two to four week period. During this period, the water circulation in DH networks is totally shut off, with the result that consumers do not obtain even domestic hot water. In most Central and Eastern European and FSU systems, DH networks are tested by pressure once a year to reveal weak pipelines and leaks. For example, in Kyivenergo's systems (Kyiv, Ukraine), this has proven to be 52 efficient because typically about ten breakages are repaired during the heating season while some four hundred are repaired outside the heating season. Buildings and customer installations. Cities in CEE/FSU were designed with a central heat and hot water supply in mind. Multifamily buildings, which vary in height from three to more than 20 stories, provide the majority of total dwellings in urban areas: 73 percent in Estonia, 50 percent in Lithuania, and approximately 80 percent in the Russian Federation, for example, but only 35 percent in Armenia. Throughout the 1950s and 1960s, most multi-family buildings constructed were no more than five stories high. Later, in the 1970s and 1980s, more nine-story and 16-story buildings were constructed, often on the very outskirts of cities. This pattern of development has typically led to larger population densities further away from city centers than in the centers themselves. 2 There are several types of multifamily residential building construction common to most CEE/FSU countries: · Brick. These were mostly constructed from 1950 to 1975, with 4 to 12 stories, and have radiators for the heating system. · Large block. These were mostly constructed from 1955 to 1970, with 4 to 12 stories, and have radiators for the heating system · Prefabricated concrete panel. These have been constructed from 1960 to the present using both one-layer and three-layer panels. Building sizes range from 5 to 22 stories. Heating systems in older buildings of 5 and 9 stories used radiators while most 12- story buildings that emerged in the 1970s used heated wall panels; convectors were used in modern 17-story and 22-story buildings. · Wood. These are single-family houses and multifamily buildings of two to four stories. Low thermal requirements in construction standards, a historical lack of attention to quality in construction materials and practices, and a poor record of operation and maintenance have led to high thermal losses in residential buildings. The most recent Soviet norms (1984­87) permitted heat transmission values more than twice those of Germany and Great Britain, and about five times those of Sweden for the same period. Actual heat losses in residential buildings are estimated to be 25­40 percent higher than design values. Deficiencies found most frequently are leaky windows and doors, uneven heat supply within buildings, and missing or insufficient basement and roof insulation. Joints between panels consist of rubber molding and cement mortar have deteriorated and permit air and rain to leak through. Although building designs and methods are very similar throughout CEE/FSU, seemingly identical buildings display enormous differences in actual construction and consequently in thermal properties; for instance, up to 40 percent in the Russian city of Ryazan. Radiator systems are either one-pipe or two-pipe vertical systems (one-pipe systems are most common). 2However, in the centers of the cities itself, the building density (and therefore the need for heat) can be extremely high (such as in Budapest), but as the buildings are quite old, they are equipped with individual heating systems. A later retrofit would have entailed high construction costs. 53 Most space heat for multi-family buildings built in the last four decades is supplied from DH systems. Buildings connected to DH or with other central heating facilities don't have chimneys.Typically, heating pipes supplying radiators within buildings are vertically arranged one- or two-pipe systems. In the more common one-pipe systems, the hot water flowing through one radiator continues through several more before returning to the source. Most radiators do not have control valves, or if they do, the valves have usually become broken or non-functional. The piping layout and the lack of control make consumption-based billing or individual cut-off of non-paying consumers very difficult. In contrast, in Western Europe (at least in newer buildings), pipes are arranged horizontally, so that each radiator and DHW source in an apartment is supplied in one single loop; two-pipe systems are standard. Radiators are typically equipped with control valves, mostly of the electronic kind, and hot water consumption is always metered. In many countries, individual heat consumption and not only building heat consumption is metered, and consumers are charged partially on a square-meter basis and partially on a consumption-basis. Technical measures for reducing heat losses in buildings include additional insulation on roofs, exterior walls, and basement ceilings; hot water and heat pipe insulation; window replacement, renovation, or simple weather stripping; improved caulking and sealing of building panel joints; new building entrance doors; and improvements to building ventilation systems. In particular, studies in CEE/FSU have highlighted the high thermal losses associated with building ventilation, leaky windows, and the low thermal insulation properties of exterior walls. Heat balancing valves for balancing the heat flows within the building also can reduce heat losses by eliminating overheated sections of a building. Although many of the above retrofit strategies are straightforward (i.e., windows, insulation, and heating equipment renovation), heat metering and controls in buildings are of special importance, pose special problems, and deserve greater attention. Measures for metering and regulation of heat demand include both (1) building-level heat meters, valves, and automatic control systems for controlling the heat entering the building and (2) apartment-level heat meters and radiator valves for controlling the heat in individual apartments. Building-level meters for measuring total building consumption are an essential part of any retrofit strategy. The question of metering at the apartment level, however, is more complex. Based on experience in the Nordic countries, there is little doubt that households in collectively metered buildings (i.e., with building-level meters but not apartment-level meters) consume more heat per square meter than households in buildings with apartment-level meters. Occupants tend to be more responsive when they can see (and have to pay for) their individual consumption. Controls are equally important. If the occupants of each unit are to be responsible for their own consumption, they must have control over the amount of heat they actually use. 3 3For a comparative analysis of heat metering and billing options in Western and Eastern Europe, Korea and China, see the report by JP-Building Engineering Ltd (2002), prepared for the World Bank and the Chinese Ministry of Construction. 54 DH coverage of urban households in the 1990s. Though DH is essentially an urban form of heating and needs relatively large heat loads and heat densities to be competitive (see Chapter 4), it can also be found in smaller towns and even villages. The highest shares of households connected to district heating can be found in the colder, more developed and more urbanized countries of the FSU, i.e., Russia, the Baltics, Ukraine and Belarus. Between 50 and 70 percent of all households in these countries are connected to DH systems (see figure below). If only urban households are taken into account, the share of households connected to DH goes up, and in some countries like the Baltics in bigger cities the share approaches 90 percent. Percentage of Households Served by District Heating in the late 1990s Latvia Russian Federation Ukraine Lithuania Estonia Belarus Slovak Republic Poland Czech Republic Romania DH share urban households Bulgaria DH share all households Croatia Moldova Kyrgyz Republic Hungary Slovenia Yugoslavia Armenia Georgia 0 10 20 30 40 50 60 70 80 Percent Source: Based on Euroheat&Power data (see www.euroheat.org), except author's calculation for Armenia, Kyrgyz Republic and Georgia. In many countries the share of households provided with DH has actually gone down during the 1990s. It is well-known that this happened in Georgia and Armenia due to economic or natural disasters. And practically everywhere, the newly rich and many other families who can afford it, switch to individual heating systems, mostly on the basis of natural gas, because they are more reliable and can be controlled by the consumer. In Estonia, for example, this trend was particularly strong. Since the mid-nineties, many families, particularly in Bulgaria (see Box 3 below), but also in Romania and Moldova, elected to disconnect all or some of their radiators as a consequence of increasing tariffs for district heat, declining service levels and ability to pay. Instead, these households use mostly individual electric heaters or fuel stoves. In some instances, e.g., in Romania, households got together, managed to secure small loans and purchased boilers that would supply central heating to their building. 55 Box 3: Disconnections in Bulgaria Declining affordability as operating subsidies are phased out faster than consumer's incomes are rising and coverage by the social safety net is inadequate resulted in over 30 percent of residential consumers opting to disconnect partially or completely from the DH systems. For Sofia and Pernik, which cover together 66 percent of all 578,000 Bulgarian households consuming DH, the two figures below show the disconnections and (re-)connections from 1996-1998. The disconnection rate in Sofia has steadily increased from 10 percent in 1996 to 26 percent of all DH consumers in the first half of 1999. During the same period, the disconnection rate in Pernik increased from 34 percent in 1996 to 49 percent of DH household consumers. In Sofia, average wages and the share of households eligible for targeted income support for energy consumption are lower than in Pernik which helps to explain that in Sofia, about equal shares of space were disconnected partially and totally, whereas partial disconnections dominated in Pernik. Pernik - Annually Sofia - Annually Disconnected/Connected HH Heating Disconnected/Connected HH Heating Space (cub.m.) Space (cub.m.) 500,000 7,000,000 450,000 400,000 6,000,000 350,000 5,000,000 300,000 4,000,000 250,000 200,000 3,000,000 150,000 2,000,000 100,000 50,000 1,000,000 0 0 1996 1997 1998 1996 1997 1998 Disconn.HH(cub.m) Connected HH (cub.m) Disconn.HH (cub.m.) Connected HH (cub.m.) Disconnections generate "free-rider" problems among consumers, since the disconnected premises still technically absorb heat and thus divert it from the neighboring apartments remaining connected to the system. In most cases the heat supply thus remains unchanged. The disconnections of heating space result in increasing the bill of the remaining connected customers who are metered. In cases where the heat supply to households is not metered, individual consumers pay in established proportions to their heated space. Disconnected heating space in this situation only increases the losses to the DH supplier since the bill for the heat, technically absorbed by the disconnected space, is not distributed among the remaining active consumers. Therefore, the government established the following disconnection rules: Consumers who disconnect their entire heating space do not pay anything for DH, whereas consumers who disconnect a portion of their heating space pay for heating the new reduced household space plus 25 percent of the regular heating expenditures for the disconnected section. Source: Bulgaria 1999 56 References Bulgaria 1999. Disconnections from District Heating District Heated Consumers in Estonia and the Slovak Systems. Background Paper prepared for ECSIE. Republic. Final Consultant Report to the World COWI A/S, 2002a. Urban Heating Strategy Development Bank. Athens. for Republic of Armenia. Draft Final Consultant Lampietti, J. A., A. A. Kolb, S. Gulyani, and V. Report to the Government of Armenia. Avenesyan. 2001. Utility Pricing and the Poor. COWI A/S, 2002b. Development of Heat Strategies for Washington, D.C.: World Bank Technical Paper the Kyrgyz Republic. Draft Final Consultant No. 497. Report. Lovei, L. et al. 2000. Maintaining Utility Services for the Environmental Resources Management. 2002.Urban Poor--Policies and Practices in Central and Eastern Heating Strategy for Armenia: Demand Analysis. Europe and the Former Soviet Union. Washington Draft Report. London. DC, World Bank. ESMAP (Energy Sector Management Assistance Martinot, E. 1997. Investments to improve the energy Program). Forthcoming. Improved Space Heating efficiency of existing residential buildings in coun- Stoves for Ulan Baatar (Mongolia). Implementation tries of the former Soviet Union. Studies of Completion Report. Energy Sector Management Economies in Transition No. 24. The World Assistance Program (ESMAP) and World Bank, Bank. Washington, D.C. Washington, D.C. Subramanian, S. and A. Deaton. 1996. "The Demand ------. 2000. Increasing the Efficiency of Heating for Food and Calories." Journal of Political Systems in Central and Eastern Europe and the Economy, 104:1. Former Soviet Union. Report 234/00, Energy Sector SwedPower/FVB, 2001. Strategic Heating Options for Management Assistance Program (ESMAP) and Moldova. Final Consultant Report. World Bank, Washington, D.C. World Bank.2002. Lithuania Energy Efficiency and Filipov, D. 2001. "In Russia, Arctic Blasts Prove Housing Pilot Project--Implementation Completion Deadly." Boston Globe. 10 January. Report. Report No. 22868. Washington, D.C. Freund, C. and C. Wallich. 1996. The Welfare Effect of ------. 2001. World Development Report 2000/2001: Raising Household Energy Prices in Poland. The Attacking Poverty. New York: Oxford University Energy Journal, 17:1. Press. Gochenour, C. 2001. District Energy Trends, Issues, and ------. 2000a. Estonia: Implementation Completion Opportunities. World Bank Technical Paper No. Report for a District Heating Rehabilitation Project. 493. Report No. 20631. Honicky, R. E. and J. S. Osborne 1991 "Respiratory ------. 2000b. Making Transition Work for Everyone: effects of wood heat: clinical observations and Poverty and Inequality in Europe and Central Asia. epidemiologic assessment." Environmental health Washington, D.C. perspectives 95: 105­10. ------. 2000c. Poland: Implementation Completion JP-Building Engineering Ltd (2002 forthcoming). Report for a Heat Supply Restructuring Project. Heat Metering and Billing: Technical Options, Report No. 20394. Policies and Regulations. Draft Final Consultant Xu, Z. et al. 1989 "Smoking, Air Pollution, and the Report to the World Bank. High Rates of Lung Canger in Shenyang, China." Kantor Management Consultants. 2001. Heating Journal of National Cancer Institute, 81, 1800­06. Supply Options for the Poor: Surveys of Metered 57