34541 _______________ Economic Analysis of Solar Home Systems: A Case Study for the Philippines ________________ Peter Meier Copyright Page ______________ Economic Analysis of Solar Home Systems: A Case Study for the Philippines February 10, 2003 ________________ Peter Meier CONTENTS Teri ­ check why forward, abstract format is different FOREWORD iv ABSTRACT v ACKNOWLEDGMENTS......................................................................................................vi CURRENCY EQUIVALENTS.............................................................................................vii ABBREVIATIONS AND ACRONYMS ...............................................................................vii Chapter 1. BACKGROUND...................................................................................................3 Chapter 2. ASSESSING BENEFITS AS AVOIDED COSTS...................................................9 ASSUMPTIONS.................................................................................................................. 9 RESULTS ........................................................................................................................ 11 VAT EXEMPTION ........................................................................................................... 17 CAPITALGRANT SUBSIDY............................................................................................... 18 CHAPTER 3. ASSESSING BENEFITS AS CONSUMER SURPLUS CHANGES..................21 THE DEMAND CURVE..................................................................................................... 21 ISSUES............................................................................................................................ 22 OTHER STUDIES.............................................................................................................. 24 CHAPTER 4. RESULTS OF THE CONSUMER SURPLUS CALCULATIONS....................27 ASSUMPTIONS................................................................................................................ 27 RESULTS ........................................................................................................................ 30 IMPACT OF KEROSENE PRICE .......................................................................................... 32 SENSITIVITY ANALYSIS.................................................................................................. 33 RESULTS FOR LARGER SYSTEMS.................................................................................... 36 CHAPTER 5. EXTERNALITIES.........................................................................................41 LOCAL AIR POLLUTION.................................................................................................. 41 HEALTH BENEFITS ......................................................................................................... 42 GLOBAL EXTERNALITIES ............................................................................................... 42 CHAPTER 6. SUBSIDY STRUCTURE...............................................................................45 CHAPTER 7. RISK ANALYSIS..........................................................................................51 CHAPTER 8. AGGREGATE ECONOMIC BENEFITS ........................................................55 Chapter 9. CONCLUSIONS.................................................................................................57 REFERENCES 59 ii TABLES IN TEXT: Table 1: 1998 Average Annual Expenditure (Pesos) for Lighting and Electricity......................10 Table 2: PV System Capital Costs (US$)...............................................................................11 Table 3: Comparison of System Costs (installed, US$)...........................................................11 Table 4: Economic Analysis for 20Wp Systems (Poorest HH): Avoided Costs Only ...............12 Table 5: Reconciliation of Economic and Financial Flows, NPVs...........................................14 Table 6: Economic Analysis: 40Wp Systems .........................................................................15 Table 7: Economic Analysis: 75Wp Systems .........................................................................16 Table 8: Impact of VAT Exemption on Consumer FRR: System Only.....................................17 Table 9: Extending VAT to Batteries....................................................................................17 Table 10: Impact of Subsidy Structure on FIRR.....................................................................18 Table 11: Calculation Lumen-hour Per Day (QKERO)...............................................................27 Table 12: Utilization of PV-system Output............................................................................28 Table 13: Consumer Surplus Estimate, 20Wp System for Poorest Households .........................29 Table 14: Adjustment to Higher World Oil prices: Poorest Households ..................................32 Table 15: Switching Values, 20Wp System ...........................................................................34 Table 16: Pessimistic Scenario..............................................................................................36 Table 17: Poor Households: 40Wp System............................................................................37 Table 19: Pessimistic Scenario Comparisons: ERR ...............................................................40 Table 20: Comparison of Consumer Surplus Estimates, 75Wp Systems...................................40 Table 21: Cost of Avoided Carbon (kerosene replacement only) .............................................43 Table 22: Increase in ERR....................................................................................................44 Table 23: Proposed Subsidy v. Economic Benefit ..................................................................45 Table 24: Aggregate Impact on Government Revenue............................................................46 Table 25: Input Assumptions for Risk Analysis .....................................................................51 Table 26: Aggregate Economic Flows: APL-I ........................................................................55 Table 27: Summary of Results..............................................................................................57 iii FIGURES IN TEXT: Figure 1: Annual Consumer Cash Flows, 20Wp System.........................................................19 Figure 2: The Demand for Lumen-hours................................................................................21 Figure 3: Beneficiaries: 20Wp Systems .................................................................................30 Figure 4: Sensitivity of ERR to Allocation Fraction ...............................................................31 Figure 5: Adjusting to Higher Kerosene Prices:.....................................................................33 Figure 6: Sensitivity to PV System Life.................................................................................35 Table 18: Non-poor Households: 75Wp Systems ...................................................................38 Figure 7: Beneficiaries, 40Wp Systems .................................................................................39 Figure 8: Beneficiaries, 75Wp systems..................................................................................39 Figure 9: Capital Subsidy, Government Revenue and FRR.....................................................46 Figure 10: Demand and Supply for Subsidies (40Wp systems).................................................47 Figure 11: Impact of Demand Elasticity ................................................................................48 Figure 12: Dealer Margins and FRR (75Wp Systems)............................................................48 Figure 13: Input Probability Distributions..............................................................................52 Figure 14: Distribution of Rate of Return..............................................................................53 iv FOREWORD The recent decreases in the capital costs of photovoltaic systems are making off-grid solar home systems a real alternative for electrification in rural areas where low load densities make the costs of conventional grid extension uneconomic. Such systems are now being proposed in many countries, and the successful experience in Sri Lanka and elsewhere shows a high willingness to pay for such systems, given the transformation of lifestyles made possible by electric lighting, radio and TV viewing. The economic benefits thus extend beyond the mere replacement costs of the kerosene-based lighting that solar homes replace. Building on a detailed survey of rural household energy use in the Philippines, this paper addresses some of the issues involved in the economic analysis of off-grid systems for the proposed Philippines Rural Power Project, including the development of plausible assumptions for demand curves and consumer surplus calculations, the reconciliation of economic and financial flows (in the presence of both subsidies and taxes), the problems of income effects and budget constraints of poor rural households, the design and impact of subsidies, and the analysis of risk and uncertainty. It is hoped that the analytical approaches developed in this paper will also be useful for analysis of other technologies in rural applications, or solar home systems use in other countries. Mohammad Farhandi Acting Sector Manager East Asia and Pacific Region The World Bank v ABSTRACT This paper deals with the economic analysis of solar photovoltaic home systems for off- grid application in the Philippines. Solar homes systems financing is one of several components of the proposed Philippines Rural Project. Each of the three candidate systems is evaluated separately: 20Wp systems for the poorest households, 40Wp systems for poor households, and 75Wp systems for non-poor households, with baseline data on household energy use before electrification based a 1998 Survey. The economic analysis using only replacement costs (of kerosene, dry cells, and rechargeable auto-batteries) suggests that only the smallest 20Wp system is economic. However, when increases in consumer surplus are correctly assessed, all three size classes show economic rates of return substantially above the 15% hurdle rate. Demand curves are derived for both lighting and TV-viewing hours. PV-based systems provide a much greater level of service for given levels of household budget constraints, and the survey data confirms high willingness to pay for a system that provides TV viewing and clean bright lighting. The paper uses a concave shape of the demand curve (which is more conservative, and also more reasonable from the standpoint of theory, than the usual assumption of linearity). The assumption of a constant budget constraint for energy expenditures in the poorest households further ensures conservative results. A detailed reconciliation of economic and financial flows is provided, which provides the basis for an assessment of the rationale for, and impact of, alternative levels of subsidy. The conventional switching value sensitivity analysis is augmented by a probabilistic risk assessment using Monte Carlo simulation. Risk factors assessed include shorter than expected system lives, capital costs, market penetration, kerosene prices, and uncertainty in the demand curve parameters. Additional benefits from carbon reduction further enhance the economic rate of return. A range of further environmental benefits ­ such as the reduction of burn injuries and deaths from kerosene lamps, and indoor air pollution benefits are noted, though not quantified. The paper concludes that solar home systems provide high economic returns for areas with low load density for which grid extension is uneconomic, with very small risk of hurdle rates not being achieved. vi ACKNOWLEDGMENTS The writer thanks Anil Cabraal, who provided much of the data on which this work is based, and whose ideas and access to prior work made this study possible. Also gratefully acknowledged is the support of Selina Shum, the Task Manager for the Philippines Rural Power Project. The paper has also benefited from the critical reviews of Susan Bogach and Charles Feinstein. Support from Teresita G. Velilla for preparation of the document for publication is gratefully acknowledged. Funding for this study was provided by the Asia Alternative Energy Program (ASTAE), managed by the Energy and Mining Sector Development Unit in the East Asia and Pacific Region of the World Bank. World Bank and donor partners established ASTAE in 1992 to support the transition to environmentally sustainable energy use in developing countries in Asia. ASTAE's strategic objective during the past decade has been to mainstream alternative energy in World Bank energy sector activities, by promoting the preparation and implementation of renewable energy and energy efficiency components in World Bank energy sector projects in Asia. This strategic objective has been achieved. By year 2000, about 15 percent of the Bank's power sector lending in Asia was for alternative energy components/projects. vii CURRENCY EQUIVALENTS Currency Unit = Philippine Peso (P) 1 PHP = US$ 0.0196 US$1 = PHP 51 ABBREVIATIONS AND ACRONYMS APL Adaptable Program Loan ASTAE Asia Alternative Energy Program DOE Philippines Department of Energy EASEG Energy and Mining Sector Development Unit, East Asia and Pacific Region ERR Economic Rate of Return FRR Financial Rate of Return GEF Global Environment Facility GHG Greenhouse Gas ICR Implementation Completion Report LPG Liquid Petroleum Gas NEA National Electrification Administration NPV Net Present Value P Peso PCF Prototype Carbon Fund PV Photovoltaic SHS Solar Home System VAT Value Added Tax (10%) WTP Willingness to Pay CHAPTER 1. BACKGROUND This report presents an economic analysis of solar photovoltaic-based home systems (SHS) for application in the Philippines.1 These are expected to be one technology option supported under the proposed Philippines Rural Power Project. The purpose of using SHS is to provide a means for rural consumers, who are unlikely to gain access to grid electricity services, to obtain affordable electricity services through off-grid means.2 According to the National Electrification Administration (NEA), in 2000 there were over 2.5 million households without access to grid electricity. Even when 100 percent of barangay (villages) are electrified, NEA estimates that about 1 million households are likely to be too dispersed and/or too far to be economically connected to the distribution grid network. Studies in the Philippines and elsewhere find that solar PV can be the least cost solution to providing basic electricity services for lighting, communications and other household/community needs etc. in areas with small dispersed populations and remote from the grid.3 It is expected that in early years of the implementation of the stand-alone renewable energy systems component, solar PV options are likely to dominate because of the following: v The Philippines has private firms and NGOs capable of delivering products/services and an extensive rural financial network for credit delivery; 1 A SHS kit usually consists of the PV module, controller, battery, several fluorescent lamps along with cables and support structure. Electricity available from a PV system is proportional to the size of the PV module(s) and sunlight availability and brightness. Each household/facility has its own unit. Electricity is stored in a rechargeable battery for use when needed. 2 In addition to individual SHS, other candidate renewable energy technologies to provide off-grid services include: · Battery charging stations of 300 Wp serving 10 households. Each would cost around $3,500- $4,500. Households would bring the batteries to the charging station to be recharged once every week or ten days. · Small wind or wind-PV systems for household service. These are appropriate in areas with good wind resources. Typically these systems have lower unit costs compared to SHS, but applicability is more site-specific. · Pico-hydro units. These are small, typically 200-500 W units that operate with a head of 1 ­ 1.5 meters. Low cost units are available from China and Vietnam. Their quality is low and annual repair costs are generally comparable to the initial costs of about $25-75 per unit. Units are installed in streams with power lines, usually strung on trees or bamboo poles, taking power to households. Safety and voltage control are concerns because such units often lack a regulator. 3 See: Cabraal et al, Best Practices in Household Electrification Programs in Developing Countries, World Bank Technical Series No. 324, 1996. 4 Chapter 1. Introduction ______________________________________________________________________________________ v There are positive experiences and good implementation models emerging from other countries.4 About 3,500 SHS and 260 battery charging stations have been installed in the Philippines during the last 20+ years5; v There is far less experience with other technology options in the Philippines; and v The Philippines enjoys good to excellent solar resources throughout the archipelago, even during the rainy season. The Philippines Rural Power Project is proposed for implementation as an Adaptable Program Loan (APL), whose first phase (APL-I) is expected to cover a five year period. For rural households that are too remote or are too dispersed to be cost-effectively connect to the grid, the Project includes a subcomponent to make available for direct purchase various capacities of solar PV systems through private vendors and NGOs. Recognizing the generally low incomes of these rural consumers and the still high capital costs of PV systems, the project will provide, through GEF and government funds, subsidies to lower the cost to consumers, and financing to spread out the payments. The suppliers would offer small PV system options (e.g. 20-75 Wp) sufficient to provide 4 These experiences have shown that several important conditions are needed for such programs to succeed. These include: competitive marketplace; consumer credit availability; significant private investment; GEF and other grant support especially in the early years to help companies build the sales and service infrastructure, increase affordability particularly for poorer households, help mitigate the higher perceived risk of consumers due to the newness of the technology, and to improve the profitability for companies; one-time subsidy is more effective to improve affordability than interest rate subsidy; initial pace of implementation is likely to be slow; variety of products should be offered to meet varying consumer needs/affordability; consumer satisfaction requires correct expectations; product quality/reliability and effective/responsive after sales service is essential; and marketing and sales are key activities. For details, see Cabraal et al (1996), Martinot et al (2001). 5 About 3,957 systems of various PV applications are located in the country. These installations are largely attributed to the initial efforts of the Philippine-German Solar Energy Program (PGSEP) in the 1980s, whose main objective was to demonstrate the technical viability of using PV for electrification. The project likewise demonstrated and tested various PV applications ranging from telecommunication, battery charging stations, PV-powered video cinemas, refrigerators, incubators, streetlights and others. Currently, an estimated 3,455 solar home systems have been installed in various locations in the country. With the real costs of PV project development above the affordability level of most of the rural population (the SHS considered in this report cost between US$350-700), international cooperation is necessary in the realization of such projects. Systems have been installed through private companies, local cooperatives (multi-purpose, agricultural, credit, etc.) as well as Rural Electric Cooperatives. Because of the limited market, Philippines PV systems tend to be more expensive than products available in other Asian countries. Experiences in the Philippines have not been all positive as many donor or government-funded programs have not paid adequate attention to sustainability. Chapter 2. Assessing Benefits as Avoided Costs 5 ______________________________________________________________________________ basic services to households. Competing vendors would be enticed to do business through incentives that include assistance in market development and capacity building, product promotions and other risk-reducing activities funded by the proposed GEF grants in order to reduce the critical barriers of PV market development. GEF grants would be leveraged with government subsidies to help reduce the initial cost of the PV systems and make them affordable for the rural poor. Further, to remove the barrier of credit access, this subcomponent would provide a line of credit to financial intermediaries (such as rural banks and micro-finance institutions) to enable them to provide consumer loans for the PV systems and financing of incremental working for dealers. The economic analysis presented in this paper covers the PV systems sub-component during its first phase of implementation during which about 10,000 PV systems of various sizes would be financed. CHAPTER 2. ASSESSING BENEFITS AS AVOIDED COSTS Assumptions The simplest approach is to assess economic benefits as the avoided costs of the services in non-electrified households that would be replaced by the PV system. These services include lighting (provided largely by kerosene lamps and some by candles and torch cells), TV and radio (provided largely by dry cells and rechargeable batteries). Such an analysis provides a lower bound for the economic benefits, because it does not account for the fact that the PV system provides a greater level of service: for example, a 20Wp system is capable of providing 10 times as many lumens of better quality light as the kerosene lamp(s) it replaces. The PV system assumed varies according to income group: 20Wp for the poorest, 40Wp for the poor, and 75Wp for the non-poor. 6 The expenditure information for estimating the avoided costs is available from a detailed survey of non-electrified households (Table 1).7 There is considerable variation in expenditure among households in each income category. In the poorest group, 98% of households use kerosene for lighting, but only 5.1 % use car batteries. Nevertheless, for that group of households that do use such batteries, annual energy expenditure will increase by P1,893 per year, greater than the expenditure for kerosene (P675/year). These households therefore devote a substantial proportion of their total annual income for energy, and are the households that would be the most likely candidates for the small (20Wp) PV systems. We note that in APL-I the target is only 2,000 systems for this size category, as against 90,301 households in this income group that incur the high cost of 6 The threshold for non-poor is income of P55,470/year; but it may be assumed that in APL-I, it is the better off households in each class that would be the first candidates for PV systems. Similarly, for the poorest households, we may assume that the initial candidates are those close to the poverty threshold (of P 22,382/HH/year), and similarly in the poor group, we may expect candidate households to be near the upper income limit of the range, i.e. close to P55,470/year. This means that the present expenditures for kerosene and battery charging are also likely to be higher than the averages used in this report, in turn implying that actual economic benefits are also likely to be higher. 7 Barnes et al, Rural Electrification and Development in the Philippines: Measuring the Social and Economic Benefits, ESMAP Report 255/02, May 2002 10 Chapter 2. Assessing Benefits as Avoided Costs ______________________________________________________________________________________ batteries, plus another 81,113 households that subscribe to local generator services.8 This implies a modest market penetration assumption of 2.2% of those households using batteries. The corresponding market penetration assumption in the poor and non-poor groups is 9.5% and 1.5%, respectively. The risk of having overestimated market penetration assumptions is therefore small. Table 1: 1998 Average Annual Expenditure (Pesos) for Lighting and Electricity (as averages for households that use each device) Poorest (1) Poor (2) Non-poor(3) Kerosene 675 823 966 n=845,198 n=506,782 n=366,761 Dry Cells 566 626 730 n=560,614 n=332,528 n=273,231 Candles 202 197 297 n=33,463 n=17,321 n=15,927 Power generators 896 1334 1796 n=81,113 n=86,784 n=82,148 Car batteries (Charging) 1143 1290 1349 n=90,301 n=73,345 n=68131 New Battery every 2 yrs 750 800 900 n=90,301 n=90,301 n=90,301 Average across all households in income category 1562 2330 4937 N=864,451 N=517,072 N=382,139 Target PV systems in APL-I 2,000 7000 1000 (1) households with income below the first income decile (2) above first income decile, but below poverty threshold (3) above the poverty threshold of P55,470/year n=number of households using each device; N=number of households in each income category Table 2 shows the breakdown of the assumed PV system capital costs. The assumed dealer gross margin is 100%, which is based on the Sri Lanka experience where high margins and good access to consumer credit have resulted in attractive business environment, and a successful PV electrification program. In contrast, gross margins are very low in China due to lower costs of labor and local materials, and lower expectations of profits by the rural PV companies. Gross margins are also low in Indonesia where there is strong pressure to keep prices low due to lack of consumer financing. 8 We assume that a household would use one or the other, but not both, since the data suggest that the annual cost of subscription to a local generator (P896 in the poorest group) is significantly below that of the car battery alternative (P1,893). Chapter 2. Assessing Benefits as Avoided Costs 11 ______________________________________________________________________________ Table 2: PV System Capital Costs (US$) Hardware Source 20 Wp 40 Wp 75 Wp PV module Imported 85.00 140.00 240.00 steel mounting frame & pole Local 5.00 7.00 7.00 storage battery Local 22.00 25.00 25.00 charge controller (BCU) Imported 22.00 25.00 25.00 12Vdc/10W lamps Local 6.00 9.00 9.00 wirings & installation accessories Local 6.00 8.00 8.00 Total Hardware Costs 146.00 214.00 314.00 Gross Margins 146.00 214.00 314.00 Total Retail Cost 292.00 428.00 628.00 VAT and Duties 29.93 43.62 63.63 Local Transport 4.38 6.42 9.42 Installation 4.38 6.42 9.42 Installed Cost 330.69 484.47 710.47 Assumptions: Unit costs are assumed to be OEM prices to PV systems integrator. OEM prices for local items such as batteries are World Bank estimates. Module costs are based on international prices plus shipping to Manila in container loads The PV system prices taken here for the Philippines are higher than prices in other countries that benefit from large volumes (e.g. in Sri Lanka, where 10,740 systems were sold in 2001). Given that module prices continue to decline steadily (20% decline for every doubling of cumulative output in real terms), the risk of higher capital costs over the APL project period is low. Table 3: Comparison of System Costs (installed, US$) Philippines Indonesia Sri Lanka India Kenya 20 Wp System 331 302 37 Wp System 270 40Wp System 484 303 419 307 45 Wp System 509 50Wp System 300-408 480 360 822 60Wp 588 75-80 Wp 710 686 Prices include taxes and duties Results Table 4 shows the results of the economic analysis using replacement costs as the measure of benefits. Rows [1]-[6] break down the (financial) first cost of the PV system by funding source;9 we then subtract taxes and transfers to derive the economic cost 9 10% down payment; GEF and government grants as defined below; consumer finance at 24% over 5 years. 12 Chapter 2. Assessing Benefits as Avoided Costs ______________________________________________________________________________________ (Row 12). For these poorest households we assume a single hurricane lamp at $5 (51 pesos per US$) with 3-year life.10 Table 4: Economic Analysis for 20Wp Systems (Poorest HH): Avoided Costs Only [unit] NPV 1 2 3 4 5 6 7 8 9 10 1 costs of PV system 2 down payment [peso] 719 827 3 GEF [peso] 2217 2550 4 Govt.Grant [peso] 6957 8000 5 loan principal [peso] 4485 5158 6 finanancial cost [peso] 14378 16535 7 finance [peso] 5476 1879 1879 1879 1879 1879 8 loan repayments [peso] -5476 -1879 -1879 -1879 -1879 -1879 9 lessVAT&duties [peso] -1301 -1497 10 less income tax on margin [peso] -813 -934 11 less transfers [peso] -863 -993 12 economic capital cost [peso] 11401 13111 0 0 0 0 0 13 14 O&M costs 15 bulbs [peso] 508 100 100 100 100 100 100 100 100 100 16 controller [peso] 388 600 17 battery [peso] 3473 1500 1500 1500 1500 18 financial cost to consumer [peso] 4370 0 100 1600 100 1600 700 1600 100 1600 100 19 less VAT [peso] -437 0 -10 -160 -10 -160 -70 -160 -10 -160 -10 20 economic O&M costs [peso] 3933 0 90 1440 90 1440 630 1440 90 1440 90 21 Total economic costs [peso] 15333 13111 90 1440 90 1440 630 1440 90 1440 90 22 23 benefits at avoided costs 24 kerosene consumption [litres] 260 51 51 51 51 51 51 51 51 51 25 kerosene [peso] 3432 675 675 675 675 675 675 675 675 675 26 battery&charging expenditure [peso] 9625 1893 1893 1893 1893 1893 1893 1893 1893 1893 source: 27 dry cell expenditures [peso] 2878 566 566 566 566 566 566 566 566 566 28 hurricane lamp [peso] 577 255 255 255 255 29 petromax lamp [peso] 0 0 0 0 30 wick, gauzes [peso] 467 92 92 92 92 92 92 92 92 92 31 total, financial [peso] 17239 255 3277 3277 3532 3277 3277 3532 3277 3277 3532 32 kerosene duties [peso] -208 -41 -41 -41 -41 -41 -41 -41 -41 -41 33 VAT [peso] -1355 -26 -255 -255 -281 -255 -255 -281 -255 -255 -281 34 avoided costs, economic [peso] 15676 230 2981 2981 3210 2981 2981 3210 2981 2981 3210 35 36 Net economic flows [peso] 343 -12881 2891 1541 3120 1541 2351 1770 2891 1541 3120 37 ERR [ ] 15.6% 38 net financial impact on consumers 39 PV system [peso] 10565 827 1979 3479 1979 3479 2579 1600 100 1600 100 40 Replacement [peso] 17239 255 3277 3277 3532 3277 3277 3532 3277 3277 3532 41 net flow [peso] 6674 -572 1298 -202 1553 -202 698 1932 3177 1677 3432 42 FRR [ ] 159.8% Given their magnitude, the treatment of dealer margins is important, for a share of this margin represents transfer payments. Some part (assumed to be 60%) of this margin reflects the real (current) cost of doing business, and counts as an economic cost. Another part of the margin reflects the return on equity capital invested up front. That part which is equal to the assumed opportunity cost of capital (15%) counts as an economic cost, but any excess (and dealers typically expect 20%+ post-tax returns) emerges in the reconciliation of financial and economic costs as a transfer payment (and 10 O&M costs for kerosene lanterns are taken as $0.15/month for wick, glass replacement, etc. This is based on the 1987 survey data of $0.12/month for Rwanda and Burundi (R. Van der Plas and A. B. de Graff, Comparison of Lamps for Domestic Lighting in Developing Countries, World Bank, Industry and Energy Department Working Paper, Energy Series Paper #6, June 1988. Chapter 2. Assessing Benefits as Avoided Costs 13 ______________________________________________________________________________ is so recorded in row 11 of Table 4). Moreover, the net margin (i.e. net of actual expenses) may be assumed to be subject to corporate income tax (at 32%), a financial flow recorded in row 10 of Table 4. The resulting economic rate of return (ERR) is 15.6%, which is marginally above the hurdle rate used in the Philippines. But the FRR to the consumer is a very high 159.8%, reflecting the large subsidy component. This suggests that if properly marketed, PV systems with the given GEF and government grant components should find high market acceptability. We examine below alternative subsidy structures. Table 5 shows the reconciliation of economic and financial flows expressed as NPVs at 15% (the discount rate used by the Government). The (negative) net economic benefits of +P343 reflect that the ERR (15.6%) is slightly above the discount rate. The financial institutions (FIs) appear as beneficiaries because the consumer loan rate (24%) exceeds the discount rate.11 11 The present value of the loan principal is P4,485, whereas the present value of the stream of loan repayments (principal and interest) is P5,476. These become equal when the consumer finance rate is exactly equal to the discount rate, but here the difference is booked as a net transfer to the financial institutions (P 991). We assume this is taxed at the corporate tax rate, with a net surplus of P 674 a shown in the last row of Table 5. 14 Chapter 2. Assessing Benefits as Avoided Costs ______________________________________________________________________________________ Table 5: Reconciliation of Economic and Financial Flows, NPVs (20Wp systems, at 15% discount rate) Reconciliation of financial and economic flows consumer dealers govt FIs GEF total benefits(avoided costs) 17239 -1562 15676 purchase price -14378 14378 0 loan 4485 -4485 0 GEF grant 2217 -2217 0 GovtGrant 6957 -6957 0 VAT[on equipment] -1301 1301 0 Income tax on dealers profit -813 813 0 Income tax of FIs profit 317 -317 0 economic cost -11401 -11401 consumer finance -5476 5476 0 O&M costs -4370 437 -3933 total 6674 863 -5651 674 -2217 343 beneficiaries and funders 10 6.7 5 0.9 0.7 1000pesos 0.3 0 -2.2 NPV@15%, -5 -5.7 -10 consumer dealers govt FIs GEF total Table 6 shows the corresponding results for 40Wp systems. The ERR declines to 11.7%, and the FRR to the consumer declines to 17.9% -- which shows (as one might expect) that the level of subsidy (P 5,000 for the 40Wp systems as opposed to P 8,000 for the 20Wp systems) strongly influences the consumer's FRR. Chapter 2. Assessing Benefits as Avoided Costs 15 ______________________________________________________________________________ Table 6: Economic Analysis: 40Wp Systems (poor households) poor households: 40Wp system 15% =discount rate [unit] NPV 1 2 3 4 5 6 7 8 9 10 costs of PV system down payment [peso] 2107 2424 GEF [peso] 4435 5100 Govt.Grant [peso] 4348 5000 loan principal [peso] 10184 11712 finanancial cost [peso] 21074 24236 finance [peso] 12435 4266 4266 4266 4266 4266 loan repayments [peso] -12435 -4266 -4266 -4266 -4266 -4266 lessVAT&duties [peso] -1907 -2194 less income tax on margin [peso] -1191 -1370 less transfers [peso] -1265 -1455 economic capital cost [peso] 16711 19217 0 0 0 0 0 O&M costs bulbs [peso] 763 150 150 150 150 150 150 150 150 150 controller [peso] 485 750 battery [peso] 3704 1600 1600 1600 1600 financial cost to consumer [peso] 4953 0 150 1750 150 1750 900 1750 150 1750 150 less VAT [peso] -495 0 -15 -175 -15 -175 -90 -175 -15 -175 -15 economic O&M costs [peso] 4457 0 135 1575 135 1575 810 1575 135 1575 135 Total economic costs [peso] 21168 19217 135 1575 135 1575 810 1575 135 1575 135 benefits at avoided costs kerosene consumption [litres] 326 64 64 64 64 64 64 64 64 64 kerosene [peso] 4317 849 849 849 849 849 849 849 849 849 battery&charging expenditure [peso] 10627 2090 2090 2090 2090 2090 2090 2090 2090 2090 dry cell expenditures [peso] 3183 626 626 626 626 626 626 626 626 626 hurricane lamp [peso] 1154 510 510 510 510 petromax lamp [peso] 0 0 0 0 wick, gauzes [peso] 934 184 184 184 184 184 184 184 184 184 total, financial [peso] 20541 510 3813 3813 4323 3813 3813 4323 3813 3813 4323 kerosene duties [peso] -261 -51 -51 -51 -51 -51 -51 -51 -51 -51 VAT [peso] -1590 -51 -290 -290 -341 -290 -290 -341 -290 -290 -341 avoided costs, economic [peso] 18690 459 3471 3471 3930 3471 3471 3930 3471 3471 3930 Net economic flows [peso] -2478 -18758 3336 1896 3795 1896 2661 2355 3336 1896 3795 ERR [ ] 11.7% net financial impact on consumers PV system [peso] 19495 2424 4416 6016 4416 6016 5166 1750 150 1750 150 Replacement [peso] 20541 510 3813 3813 4323 3813 3813 4323 3813 3813 4323 net flow [peso] 1046 -1914 -603 -2203 -93 -2203 -1353 2573 3663 2063 4173 FRR [ ] 17.9% KWh 312 61 61 61 61 61 61 61 61 61 P/kWh(levelised)(economic) 68 Reconciliation of financial and economic flows consumer dealers govt FIs GEF total benefits(avoided costs) 20541 -1851 18690 purchase price -21074 21074 0 loan 10184 -10184 0 GEF grant 4435 -4435 0 GovtGrant 4348 -4348 0 VAT[on equipment] -1907 1907 0 Income tax on dealers profit -1191 1191 0 Income tax of FIs profit 720 -720 0 economic cost -16711 -16711 consumer finance -12435 12435 0 O&M costs -4953 495 -4457 total 1046 1265 -1885 1531 -4435 -2478 beneficiaries and funders 2 1.0 1.3 0 1000pesos -2 -1.9 -2.5 NPV@15%,-4 -4.4 -6 consumer dealers govt FIs GEF total 16 Chapter 2. Assessing Benefits as Avoided Costs ______________________________________________________________________________________ Table 7: Economic Analysis: 75Wp Systems non-poor households:75Wp system 15% =discount rate [unit] NPV 1 2 3 5 6 7 8 9 10 costs of PV system (non-poor households)4 down payment [peso] 3092 3556 GEF [peso] 4990 5738 Govt.Grant [peso] 0 0 loan principal [peso] 22840 26266 finanancial cost [peso] 30922 35561 finance [peso] 27888 9567 9567 9567 9567 9567 loan repayments [peso] -27888 -9567 -9567 -9567 -9567 -9567 lessVAT&duties [peso] -2799 -3219 less income tax on margin [peso] -1747 -2010 less transfers [peso] -1857 -2135 economic capital cost [peso] 24519 28197 0 0 0 0 0 O&M costs bulbs [peso] 763 150 150 150 150 150 150 150 150 150 controller [peso] 485 750 battery [peso] 3936 1700 1700 1700 1700 financial cost to consumer [peso] 5184 0 150 1850 150 1850 900 1850 150 1850 150 less VAT [peso] -518 0 -15 -185 -15 -185 -90 -185 -15 -185 -15 economic O&M costs [peso] 4666 0 135 1665 135 1665 810 1665 135 1665 135 Total economic costs [peso] 29185 28197 135 1665 1 3 5 1 6 6 5 8 1 0 1 6 6 5 1 3 5 1 6 6 5 1 3 5 benefits at avoided costs kerosene consumption [litres] 379 75 75 75 75 75 75 75 75 75 kerosene [peso] 5013 986 986 986 986 986 986 986 986 986 battery&charging expenditure [peso] 11435 2249 2249 2249 2249 2249 2249 2249 2249 2249 dry cell expenditures [peso] 3712 730 730 730 730 730 730 730 730 730 hurricane lamp [peso] 577 255 255 255 255 petromax lamp [peso] 1931 1065 1065 1065 wick, gauzes [peso] 1687 332 332 332 332 332 332 332 332 332 total, financial [peso] 24735 1320 4371 4371 4626 5436 4371 4626 4371 5436 4626 kerosene duties [peso] -303 -60 -60 -60 -60 -60 -60 -60 -60 -60 VAT [peso] -1934 -132 -331 -331 -357 -438 -331 -357 -331 -438 -357 avoided costs, economic [peso] 22498 1188 3981 3981 4210 4939 3981 4210 3981 4939 4210 Net economic flows [peso] -6687 -27009 3846 2316 4075 3274 3171 2545 3846 3274 4075 ERR [ ] 8.7% net financial impact on consumers PV system [peso] 36165 3556 9717 11417 9717 11417 10467 1850 150 1850 150 Replacement [peso] 24735 1320 4371 4371 4626 5436 4371 4626 4371 5436 4626 net flow [peso] -11429 -2236 -5346 -7046 -5091 -5981 -6096 2776 4221 3586 4476 FRR [ ] 1.6% KWh 585 115 115 115 115 115 115 115 115 115 P/kWh(levelised)(economic) 5 0 Reconciliation of financial and economic flows consumer dealers govt FIs GEF total benefits(avoided costs) 24735 -2238 22498 purchase price -30922 30922 0 loan 22840 -22840 0 GEF grant 4990 -4990 0 GovtGrant 0 0 0 VAT[on equipment] -2799 2799 0 Income tax on dealers profit -1747 1747 0 Income tax of FIs profit 1615 -1615 0 economic cost -24519 -24519 consumer finance -27888 27888 0 O&M costs -5184 518 -4666 total -11429 1857 4442 3433 -4990 -6687 beneficiaries and funders 10 5 4.4 3.4 1.9 0 1000pesos -5 -5.0 -6.7 NPV@15%, -10 -11.4 -15 consumer dealers govt FIs GEF total Chapter 2. Assessing Benefits as Avoided Costs 17 ______________________________________________________________________________ For the 75Wp system (Table 7) the ERR declines further to 8.7%; more importantly the FRR for the consumer is a very low 1.6%. However, the PV system will bring much higher levels of service, reflected in the high willingness-to-pay of non-poor households for PV systems in other countries: this consumer benefit ­ the increase in consumer surplus -- is not captured here. However, the results of Table 7 do raise questions about the proposed subsidy structure: the government grant is proposed to be extended only to the poorest and poor groups. But because of the capital intensity of the larger PV systems, the net result for the larger 75Wp systems is an increase in Government revenue, as the increase in PV-system related VAT and income tax on large dealer margins exceeds the loss of revenue from VAT and kerosene duty of the un-electrified household. Indeed, it should be noted that even if consumer surplus benefits were included (as in the next section), the (positive) fiscal balance for government remains unchanged for the 75Wp systems. VAT exemption Table 8 shows the impact of the proposal to exempt PV systems from VAT. Clearly, elimination of VAT on the PV modules is tantamount to an increase of the capital grant ­ e.g. in the case of 20Wp systems, from P8,000 to P9,477 (or from P5,000 to P7,194 for 40Wp systems). Since this directly affects the first-year cash flows, the FRR shows significant increases. However, in the case of 75Wp systems, the VAT exemption would still not be sufficient to bring the consumer FRR (based on replacement costs only) to a value comfortably above the hurdle rate. Table 8: Impact of VAT Exemption on Consumer FRR: System Only VAT 10% VAT VAT exempt Change Amount(1) (Pesos) 20Wp 1,497 160% 308% +148% 40Wp 2,194 17.9% 26.5% +9.2% 75Wp 3,219 1.6% 3.6% +2.7% If the VAT exemption were extended to replacement batteries (Table 9), the FIRR would increase further, but not significantly. Table 9: Extending VAT to Batteries FRR, FRR, system system only plus batteries 20Wp 308% 314% 40Wp 26.5% 28% 75Wp 3.6% 4.1% 18 Chapter 2. Assessing Benefits as Avoided Costs ______________________________________________________________________________________ This analysis suggests that: v Extending the exemption to batteries has an insignificant effect on consumer FIRR (which would be true even if the FIRR considered consumer surplus benefits as well); and v The VAT exemption on PV modules is monetarily equivalent to increasing the capital grant: a P1,000 decrease in VAT is equivalent to a P1,000 increase in capital grant (though there may be slight differences in transaction costs not reflected in our calculations).12 Capital grant subsidy The present proposal for subsidies is P8,000 for 20Wp systems, P5,000 for 40Wp systems, and zero for 75Wp systems. These subsidy levels reflect the perception that it is the poorest households, and therefore 20Wp systems, that are most worthy of subsidy support. However, as shown in Table 10, the subsidy on the 20Wp system is over twice that for 40Wp systems as a fraction of the upfront financial cost, which in turn has a disproportionate effect on the FIRR. Table 10: Impact of Subsidy Structure on FIRR Proposed Financial first Percent of FRR subsidy cost first cost 20Wp 8000 16,535 48% 160% 40Wp 5000 24,235 21% 17.9% 75Wp 0 35,560 0% 1.6% This raises the obvious question: is the subsidy proposed for the poorest households too large? For poor households (40Wp) the 17.9% FRR appears reasonable, and for the non- poor households, experience in other countries suggests high willingness to pay (based on consumer surplus) to offset the narrowly defined FIRR based on replacement cost. 12 However, Mostert (Rural Electrification Subsidy Principles: A Reference Manual, Report to the World Bank, February 2002), argues that a direct capital subsidy has a stronger incentive effect on purchases than tax rebates having the same subsidy value, because it is more transparent for investors and voters (though this lack of transparency may be also be seen as an advantage). On the other hand, an advantage of tax rebates is the avoidance of a separate bureaucracy for the administration of subsidy payments. A detailed study of subsidy policy for rural electrification in the Philippines is currently underway (Patalinghug et al., 2002). Chapter 2. Assessing Benefits as Avoided Costs 19 ______________________________________________________________________________ Clearly, it is reasonable that even the poor households should experience some negative cash flow in the first year, as an incentive to take care of the system. However, as shown in Figure 1, even at the P8,000 subsidy level, there remain two further years of negative cash flow (for battery replacements in years (1-5) when the loan payments are still being made). Any reduction of the subsidy would therefore worsen the magnitude of these negative years. Figure 1: Annual Consumer Cash Flows, 20Wp System 4 3 2 P 1000 1 0 -1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 14 15 However, before we can properly assess the subsidy structure as a whole (in Section 6), we need a better characterization of the economic benefits, to which we turn in the next Section. CHAPTER 3. ASSESSING BENEFITS AS CONSUMER SURPLUS CHANGES The Demand Curve Experience in other countries shows high willingness to pay for the improved level of lighting provided by PV. Yet as noted, the analysis of the previous section does not take into account the benefits of the higher levels of services provided (e.g. in the case of lighting, more lumens delivered, or in the case of radio/cassette, more hours of listening). These benefits are represented by the increase in consumer surplus. Figure 2: The Demand for Lumen-hours price A PKERO B C PPV D E QKERO QPV lumen-hours Estimating consumer surplus requires knowledge of the demand curve. This can most easily be illustrated by the demand curve for lumen-hours shown in Figure 2. Initially, lumens are provided by kerosene lamps, with QKERO lumen-hours consumed at the price PKERO. The total benefit is the area under the demand curve to the point (i.e. the area 22 Chapter 3: Consumer Surplus Benefits _____________________________________________________________________________________ A+B+D); the cost of delivering QKERO lumen-hours is the price of kerosene times quantity consumed (the area B+D), leaving the area A as the net benefit of consuming QKERO lumens ­ the so-called consumer surplus. Suppose lumens can be delivered from a PV system at the (lower) cost PPV. At this lower cost, the quantity of lumens consumed increases to QPV. And at this level of consumption the corresponding consumer surplus is A+B+C, with an increase (over kerosene-based lighting) of B+C. This represents the net benefit of the PV system. Issues The main difficulty of this approach is that only one point of the demand curve is known with reasonable certainty, namely the point , which is readily measurable from survey data of unelectrified households. Data from other countries may give guidance on point , since it is possible to observe the changes in lumen consumption where PV systems are already in place (e.g. from the successful PV-program in Sri Lanka). But extrapolation to the Philippines necessarily involves some uncertainty. Even with an estimate of point in hand, calculation of the area C requires knowledge of the shape of the demand curve in the range of QKERO to QPV. It is often assumed that the demand curve is linear,13 but if the demand curve is concave (with respect to the origin), then all other things equal the area C will be much smaller.14 The curve as drawn reflects the assumption made in this analysis, namely of constant elasticity (as derived below), which results in more realistic estimates of consumer surplus benefits. A second issue concerns the assumption that the demand curve is independent of any changes in the price and consumption of goods or services that may be complements to or substitutes for electricity (World Bank, 2002). The demand curve could be expected to shift to the right (i.e. greater demand for given price) if complements for electricity consumption--such as the availability of the services of electrical appliances--become less expensive. The demand curve would shift to the left (i.e. lower demand at a given price) if fuels that may substitute for electricity become cheaper ­ though that is hardly likely in the case of lighting, which is the main focus of this analysis. 13 For example, the consumer surplus estimates in the World Bank's analysis of the Philippines survey data assumes such linearity. 14 The World Bank report, op.cit., notes that if the demand curve is convex with respect to the origin, then calculation of the area C assuming linearity would under-estimate the true value. But it would be rare for a demand curve to have a convex shape, so the assumption of linearity is much is much more likely to result in an over-estimate of the consumer surplus than an under-estimate. Note also that the shape of the demand curve above point , or to the right of point , does not affect the calculation of changes in consumer surplus (which occurs between these two points). Chapter 3: Consumer Surplus Benefits 23 _____________________________________________________________________________ A third issue relates to income effects: as income increases, the aggregate demand curve shifts to the right. Therefore, rather than using a single aggregate demand curve, we estimate separate demand curves for each of the categories of the three-way income classification used in the survey ­ poorest, poor, and non-poor. This is the conservative assumption, for if over the project lifetime incomes increase, then economic benefits would also increase.15 Thus assumptions on these matters are unavoidable, given the lack of empirical data from cross-country studies. However, calculation of the relevant switching values can provide an indication of the robustness of returns to the assumptions made (see example below, Table 15). This analysis simplifies a number of important aspects. First, the quantity of lumens is not the sole determinant of consumer utility, for a lumen delivered from a smelly and possibly dangerous kerosene lamp would be valued less than a lumen from a PV system. Thus, as in the case of electricity, the improved quality of PV-generated lumens is an additional value not captured in the above analysis. Kerosene lighting also contributes to indoor air pollution damage costs, whose avoidance is a benefit to the PV system. The incremental contribution to indoor pollution levels from kerosene lighting may be difficult to identify given the much higher contribution from wood-based cooking stoves, but it is certainly not zero. Kerosene lighting also causes a significant number of burn injuries, house fires and related deaths: these impacts are discussed further below. There exist other indirect benefits that can plausibly be claimed as a benefit of PV systems, for which there is significant logical rationale and anecdotal evidence, but which still lack the necessary research studies to permit monetization. These include: · Income effects: for example traders in India who used solar lanterns at their roadside stalls found that the quality of lighting and absence of kerosene fumes attracted more customers during the main early evening business hours, with 50% increases in their daily income (Rs 50-100/day)16; and · Educational benefits: the Philippines survey data shows that members of electrified households attain about two years more formal education than their 15 However, the benefits are subject to the upper bound of the watt-hours that the given system initially purchased would produce. One of the few advantages of kerosene is that output can be increased by small increments (purchase of a an additional lamp) as lighting demand rises. 16 Meritec Ltd., India Renewable Resources Development Project, Review of Project for Implementation Completion Report, April 2002., p.41. 24 Chapter 3: Consumer Surplus Benefits _____________________________________________________________________________________ non-electrified counterparts, resulting in earnings increases of $37- $45/household.17 PV solar-homes systems may result in a somewhat smaller impact on reading and studying habits than full electrification, but again, the effect is unlikely to be zero. It should be noted that the consumer surplus calculations are done in terms of financial prices (since these are the prices to which consumers respond and make their decisions). Financial and economic flows must therefore be reconciled in order to assess economic benefits, in the manner shown in Box 1. Other studies The general framework as depicted in Figure 2 has been applied in similar projects elsewhere. For example, in the Implementation Completion Report (ICR) for the Indian Renewable Resources Development Project,18 the economic analysis uses an estimate of benefits of 20Rs/kWh (about 40 US Cents/kWh at the current exchange rate) for the consumer surplus, resulting in a 30% ERR without global environmental benefits (and 108% with global benefits). With only the cost of replacement kerosene as the measure of benefits, the ERR was lower at 15% (60% with global benefits). On the other hand, the Sri Lanka Renewable Energy for Rural Development Project19 omits an economic analysis altogether for the case of the solar homes component: In this case [solar homes] there is little difference between a financial and economic computation. It can safely be assumed that the economic rate of return would be higher than 12% since the economic benefits are likely to be much higher than the mere replacement cost of kerosene. For example the indirect benefits of replacing kerosene with solar lighting ­ between quality of lighting, higher safety and freedom from air pollution are not captured in the financial benefit evaluation. Given these factors and uncertainty about the valuation of actual benefits, a separate EIRR calculation for solar has not been presented here. 17 World Bank, 2002, op.cit., para 6.25-6.27. 18 Meritec Ltd., India Renewable Resources Development Project, Review of Project for Implementation Completion Report, April 2002. 19 World Bank, Project Appraisal Document, May 24, 2002 (Report 23886-CE) Chapter 3: Consumer Surplus Benefits 25 _____________________________________________________________________________ Box 1: Reconciling Economic and Financial Costs The consumer surplus calculations are done using financial costs (since these are the costs actually seen by the consumer). However, the financial and economic costs require careful enumeration. The figure below illustrates the reconciliation in terms of the demand curve for lumen-hours, in a situation in which the cost of PV systems is both subject to subsidy of (equal to PPV[econ] -PPV[subs] equivalent to the area QPV), as well as taxed (e.g. VAT on sales), which raises the price from PPV[subs] to the retail price PPV (equal to the area D*+E*). As drawn, the assumption is that the subsidy exceeds related taxes: but as noted in the text, for 75Wp systems, taxes exceed the subsidies. For kerosene, the total (financial) cost to the consumer is B*+B+D*+D. Of this, the area B* represents taxes (10% VAT, customs duty etc). For PV, the total cost to the consumer is D+D*+E+E*, of which D*+E* represents taxes and duties (VAT, duties on imported components, etc). price A PKERO B* PKERO [economic] B C PPV[econ] PPV D* E* PPV[SUBS D E QKERO QPV lumen-hours In tabular format, the reconciliation is as follows: kerosene PV difference consumer surplus A A+B*+B+C B+B*+C taxes and duties (VAT) B* D*+E* D*+E*-B* PV system subsidy -QPV -QPV economic cost B+D*+D D+E+dQPV +E+dQPV -B-D* total benefit to consumer (area A+B*+B+D*+D A+B+B*+C+ C+E*+E under demand curve) D*+D+E*+E CHAPTER 4. RESULTS OF THE CONSUMER SURPLUS CALCULATIONS Assumptions Calculation of the demand curve requires numerical estimates of four quantities (in the case of a linear demand curve): PKERO, PPV, QKERO and QPV. The price of kerosene-based lumen-hours, PKERO, follows from survey consumption data and the reconciliation of kerosene consumption and lighting output shown in Table 11. Table 11: Calculation Lumen-hour Per Day (QKERO ) 20Wsystem 40WpSystem 75WpSystem annual kerosene consumption [litres/year] 51 64.2 74.5 annual kerosene cost peso/year 675 849 986 daily consumption litres/day 0.14 0.18 0.20 hurricane lamp litres/hour 0.04 0.04 0.04 operating 3.49 4.40 0.50 hours/day lumens/hour 40 40 40 lumen-hours/day 140 176 20 litres/day 0.14 0.18 0.020 petromax lamp litres/hour 0.08 operating 2.3 hours/day lumens/hour 400 lumen-hours/day 921 litres/day 0.184 total lumen-hours/day lumen-hours/day 140 176 941 Note: Petromax lamps consume about twice that of a hurricane lamp, but produces 10 times the lumen output (i.e. 400 lumens/hour v.40). Non-poor households are assumed to have one Petromax and one hurricane lamp, whose utilisation is taken as 2.3 and 0.5 hours per day, respectively (so as to reconcile with the known data on average daily consumption). To estimate the quantity of lumen-hours available from the PV system for lighting (and other uses) requires assumptions about consumer behavior that may be derived from the survey data shown in Table 12. The resulting watt-hour calculations must reconcile with the available watt-hours based on the peak output of each system. This reconciliation then provides a basis not only for estimating the quantity of lumen-hours available (QPV), but also the proportion of the total PV system cost that may be allocated to lighting. For example, for the poorest households, 40% of the output (and cost) is allocable to lighting. 28 Chapter 4: Results of Consumer Surplus Calculations _____________________________________________________________________________________ Table 12: Utilization of PV-system Output PV system size [Wp] 20 40 75 100 Solar Radiation [peak sunlight hours] 4.2 4.2 4.2 4.2 Power available [W-hours] 84 168 315 420 Lights Number [ ] 2 2 4 6 Watts [watts] 8 8 8 8 Hours [hours] 2 4.5 2 3 Sub total Wh [W-hours] 32 72 64 144 TV Number [ ] 1 1 1 1 Watts [watts] 15 15 45 45 Hours [hours] 3.2 4 5 4 Sub total Wh [W-hours] 48 60 225 180 Radio Number [ ] 1 1 1 1 Watts [watts] 5 5 5 20 Hours [hours] 0 6 4 4.5 Sub total Wh [W-hours] 0 30 20 90 Fans Number [ ] 0 0 0 0 Watts [watts] 0 0 10 20 Hours [hours] 0 0 10 5 Sub total Wh [W-hours] 0 0 0 0 Total Consumption [W-hours] 80 162 309 414 fraction allocated to lighting [ ] 0.400 0.444 0.207 0.348 Source: Mission estimates To represent the shape of the demand curve we use the standard functional specification with constant elasticity Q =Q0 P0 P Eq.[1] from which the price-elasticity of lumen-hours can be calculated for each income group by simple transformation as ­1.2, -1.1 and ­0.81 for poorest, poor and non-poor categories, respectively.20 This results in the concave demand curve illustrated in Table 13. The high value of price elasticity is consistent with the observed behavior of significant increases in lighting demand when lighting price can be lowered. log QKERO QPV 20The elasticity follows as = log PPV PKERO Chapter 4: Results of Consumer Surplus Calculations 29 _____________________________________________________________________________ Table 13: Consumer Surplus Estimate, 20Wp System for Poorest Households source: Note: results calculated over 15-year assumed lifetime for PV system (but only years 1-10 shown for clarity) 30 Chapter 4: Results of Consumer Surplus Calculations _____________________________________________________________________________________ Results The resulting consumer surplus in area C is significantly lower than that which follows from the common assumption of linearity.21 For example, for the poorest households using the assumed elasticity of ­1.2, the area C calculates to P1,053 (as an annual quantity), as opposed to P2,723 assuming linearity.22 Using exactly the same procedure we also calculate the consumer surplus benefits associated with TV viewing ­ which, as noted, accounts for significant costs of battery charging (or subscription to local diesel generators) in non-electrified households.23 The ERR taking into consideration the consumer surplus benefits is 45%, which, as expected, is significantly higher than when only replacement costs are considered. Figure 3 shows the beneficiaries (as NPVs over the 15 year lifetime, for a single system). Sources of funds and subsidies are shown as negative quantities in this chart. Figure 3: Beneficiaries: 20Wp Systems 21 The area (C+E) under the demand curve Eq.[1] follows as the definite integral QPV C + E = Po Q Q1 +1 / 1/ dQ = Po QPV Qo 1/ Qo1 / 1 + 1 / QKERO QKERO 22 If the area C is zero, then the consumer surplus reduces to the change in financial costs (i.e. area B+D=881 for kerosene, and D+E=831 for PV, a gain of 50). 23 Barnes et al, 2002. Table 6.4 states that viewing hours per month in non-electrified households using battery is "1.85 hours per month." This implies 4 minutes per day ­ which hardly justifies the high cost of battery. We therefore make the assumption that with battery, an average of 1 hour per day of TV viewing is possible, as opposed to 3.2 to 5 hours per day for PV systems (as per Table 10). Chapter 4: Results of Consumer Surplus Calculations 31 _____________________________________________________________________________ We note the predominance of TV benefits (P33,670) over those of lighting (P9,300), a result that follows from the allocation of PV-system output (of Table 13). Figure 4 shows the sensitivity of ERR to this assumption ­ which changes little.24 Figure 4: Sensitivity of ERR to Allocation Fraction 50 40 30 ERR 20 10 0 0.4 0.5 0.6 fraction allocated to Lighting Of interest is the robustness of the economic returns to other key uncertainties: · The capital cost of the PV system · The TV viewing hours provided by the PV system · The lumen-hours provided by the PV system (QPV) · The kerosene price (as influenced by changes in the world oil price) · The actuallife of the system 24 While our assumptions appear plausible, and consistent with observed behavior elsewhere, there is a lack of reliable survey information. Improving the database has high priority, and resources should be set aside in the Philippines project to collect the necessary information by follow-up surveys conducted one, two, three and five years after installation. Ideally, households seeking PV-system grants should be made to complete a survey on their pre-electrification habits as well. Such an on-going data collection process would also facilitate the ICR. 32 Chapter 4: Results of Consumer Surplus Calculations _____________________________________________________________________________________ Impact of kerosene price Higher kerosene prices imply higher avoided costs, and hence larger economic benefits. Indeed, since the survey year of 1998, when oil prices were atypically low (Singapore spot prices for kerosene in mid 1998 were around 16 $/bbl), oil prices have almost doubled, and then fallen some (the present price is around 26$/bbl). However, for the poorest and poor households with little disposable income, increases in kerosene price would most likely imply a downward adjustment of consumption rather than an increase in annual kerosene expenditure. Table 14 compares the daily lumen-hours at 16 and 30$/bbl, which for poorest households under an expenditure constraint implies a reduction in daily lumen-hours from 140 to 111. Table 14: Adjustment to Higher World Oil prices: Poorest Households Singapore kerosene spot price [$US/bbl] 16 30 [Pesos/bbl] 640 1200 [Pesos/litres] 4.03 7.55 freight&margins@1998 values [Pesos/litres] 8.40 8.40 Taxes [Pesos/litres] 0.80 0.80 retail price [Pesos/litres] 13.23 16.75 expenditure constraint [pesos/year] 675 675 Consumption [Pesos/litres] 51 40 ratio of consumption [ ] 0.79 lumen-hours/day [lumen-hours/day] 140 111 With the replacement cost of kerosene assumed constant, the ERRs based on avoided costs obviously do not change. But one might expect that the consumer surplus estimates of benefit would increase, because the initial kerosene lumen-hours decrease. Figure 5 depicts this adjustment graphically: the areas U+V=675 for the 16$/bbl price must equal U+Y for the 30$/bbl price (since the kerosene expenditure is assumed constant). Chapter 4: Results of Consumer Surplus Calculations 33 _____________________________________________________________________________ Figure 5: Adjusting to Higher Kerosene Prices: Lumen-hour demand curve for poorest households price X PKERO (30$/bb) Y Z PKERO (16$/bb) U V W PPV QKERO=111 QKERO=140 QPV ($30/bbl) ($16/bbl) lumen-hours But even an increase to 30$/bbl affects the consumer surplus-based ERR estimates very little ­ an additional 1% on ERR. This is because the area Y+Z is quite small as a percentage of the total gain in consumer surplus (Y+Z+U+V+W). We conclude that the economic benefits are not sensitive to world oil price changes, under the conservative assumption of a household expenditure constraint. Sensitivity Analysis Figure 6 shows the sensitivity of FIRR and ERR to PV system life. For the ERR based on consumer surplus, returns are robust down to 7 years for the large systems, and 4 years for the smaller systems: this is because the consumer surplus benefits of increased lighting and TV viewing are so large (and the first few years of which determine the magnitude of the ERR). For both the replacement-cost based ERR, and the FRR, the curves are relatively flat in the 10-15 year range. The sensitivity of the economic returns to the other three assumptions is measured by calculating the relevant switching values (Table 15), which are the values of the variable in question that make the ERR equal to the hurdle rate ­ assumed here as the opportunity cost of capital at 15%. 34 Chapter 4: Results of Consumer Surplus Calculations _____________________________________________________________________________________ Table 15: Switching Values, 20Wp System base value switching ERR at (as in Table 14 ) value switching value Capital cost (equipment) 7300 38690 15% Q(TV) as viewing- 3.2 .28 15% hours/day Q(PV), as lumen-hours/day 1120 140 33% The switching values confirm the robustness of the benefits of 20Wp systems. We have already noted that the risks of PV equipment costs increasing are very small ­ but even in the remote case that these would increase over time, a 530% increase would be needed to bring the ERR to the hurdle rate. Indeed, this subsumes the possible impact of Peso devaluation, given that 73% (see Table 2) of the equipment cost is imported: even a 100% change in the exchange rate would do little to impair the (consumer surplus-based) economic returns of the PV system. The hurdle rate is reached if the TV viewing hours were reduced to .28 hours per day, rather than the 3.2 hours assumed in the base case. Such a reduction would imply serious PV system problems ­ but for output to be permanently 10% below the reference value would imply significant defects in manufacture or maintenance -- which experience elsewhere (e.g. Sri Lanka) suggests is quite unlikely. Similarly for lighting output ­ even if it were reduced to that of kerosene, the ERR is still 33%. Chapter 4: Results of Consumer Surplus Calculations 35 _____________________________________________________________________________ Figure 6: Sensitivity to PV System Life 20Wp system 2 FRR 1.5 1 return of 0.5 ERR, CONSUMER SURPLUS HURDLE RATE=15% rate 0 ERR, REPLACEMENT COSTS -0.5 -1 3 4 5 6 7 8 9 10 11 12 13 14 15 PV system life, years 40Wp system 0.6 ERR, CONSUMER SURPLUS 0.4 0.2 HURDLE RATE=15% FRR return of ERR, REPLACEMENT COSTS rate 0 -0.2 -0.4 3 4 5 6 7 8 9 10 11 12 13 14 15 PV system life, years 75Wp system 0.4 ERR, CONSUMER SURPLUS 0.2 HURDLE RATE=15% ERR, REPLACEMENT COSTS return FRR of 0 rate -0.2 -0.4 3 4 5 6 7 8 9 10 11 12 13 14 15 PV system life, years 36 Chapter 4: Results of Consumer Surplus Calculations _____________________________________________________________________________________ This switching values analysis varies one assumption at a time. But suppose that all of the assumptions in the reference case were optimistic? Table 16 shows such a pessimistic scenario, with a 25% capital cost increase for PV system equipment and a 30% reduction in lumen-output and a 30% reduction of TV viewing hours. These assumptions taken together reduce the ERR to 26.8%, still comfortably above the hurdle rate. Table 16: Pessimistic Scenario Pessimistic Reference Capital cost 9125 7300 QTV 2.24 3.2 QPV 784 1120 ERR 26.8% 45% This analysis suggests that despite the various uncertainties, the economic benefits of the solar PV system are robust for the case of the 20Wp systems that are targeted at the poorest households. Results for Larger Systems Tables 17 and 18 replicate the above analysis for the larger 40Wp and 75Wp systems, with baseline ERRs of 46.6% and 45%, respectively. Chapter 4: Results of Consumer Surplus Calculations 37 _____________________________________________________________________________ Table 17: Poor Households: 40Wp System source: 38 Chapter 4: Results of Consumer Surplus Calculations _____________________________________________________________________________________ Table 18: Non-poor Households: 75Wp Systems Chapter 4: Results of Consumer Surplus Calculations 39 _____________________________________________________________________________ Inspection of the distributions of the economic benefits shown in Figures 7 and 8 reveal several important points. First, for both systems, the consumer surplus benefits for TV/radio exceed those for lighting ­ which is a reflection of the high willingness to pay for TV ­ and particularly for color TV, which is made possible by the 75Wp system. Second, in the case of the 75Wp systems, the lighting benefits are only half those of the 40Wp systems. This is because non-poor households are assumed to use petromax lamps, which provide significantly more lumens per peso of kerosene expenditure than do hurricane lamps (used by the poor). Consequently the improvement in lighting service for poor households (40Wp systems) is very much greater than that for the non-poor.25 Figure 7: Beneficiaries, 40Wp Systems 60 TV/radio:35.2 49.3 40 1000pesos 20 lighting:17.6 NPV@15%, 0 -1.9 1.3 1.5 -4.4 -20 consumer dealers govt FIs GEF total Figure 8: Beneficiaries, 75Wp systems 40 35.5 30 TV/radio:21.8 20 1000pesos 10 lighting:8.9 NPV@15%, 4.4 0 3.4 1.9 -5.0 -10 consumer dealers govt FIs GEF total 25 Note the very different shape and position of the lighting demand curves in Tables 17 and 18. 40 Chapter 4: Results of Consumer Surplus Calculations _____________________________________________________________________________________ The switching values analysis for the larger systems differs little to that displayed in Table 15 for the 20 Wp systems. Table 19 compares the results for the three systems under the three sets of assumptions: benefits at replacement costs, and consumer surplus benefits of lighting under reference and pessimistic assumptions. Table 19: Pessimistic Scenario Comparisons: ERR Benefits @ avoided costs only Consumer surplus benefit Reference Pessimistic scenario 20 Wp systems 15.6% 45% 27 40 Wp systems 11.7% 47% 25 75 Wp systems 8.7% 32% 16 Table 20 compares the estimates of consumer surplus with expenditures before and with the PV systems, and with household income, for the case of 75Wp systems. The total consumer surplus estimate is $US119, or $9.90 per household/month. This is substantially lower than the estimate of $56.4/household/month in the ESMAP report for the Philippines (World Bank, 2002) ­ but the latter is for full electrification, rather than for the lower levels of service provided by a 75Wp PV system. Though our estimates of net benefits are high, they are supported by observed willingness-to-pay in other countries. The increases as a fraction of household income are in proportion to the dramatic improvement in the quality of life experienced by households previously using kerosene and batteries. Table 20: Comparison of Consumer Surplus Estimates, 75Wp Systems Before With PV system peso peso $US lighting expenditure 1811 1473 CS 1757 34 TV expenditure 2979 5639 CS 4291 84 total 4790 13160 119 HHincome 60000 60000 as% 8% 22% We conclude that the economic benefits of PV systems are robust even without consideration of externalities, and substantially above risk-adjusted hurdle rates. The risks of returns falling to unsatisfactory levels are quite small and certainly unrelated to the physical and economic aspects of the project design. CHAPTER 5. EXTERNALITIES PV-system generated electricity avoids the air-emissions associated with kerosene lighting. The avoided damage costs represent a benefit, and need to be considered in the economic analysis. Other negative impacts of kerosene whose avoidance is not included in the above analysis include occasional burn injuries, bad odors, and the inconvenience of having to buy and store the fuel. Their monetisation, even if possible, would be so small that they would make no material difference to the economic analysis. However, in the case of avoided air emissions, the avoided damages do make a significant difference to the economic analysis, notably in the case of carbon emissions.26 Local Air Pollution It has been argued that kerosene lamps represent a small fraction of total indoor air emissions, particularly in poor households that use fuel-wood for cooking. Even where LPG is introduced to replace fuel-wood (or where improved cook-stoves with chimneys are introduced), kerosene-lamp based emissions would represent a small part of the total. And if the fraction of emissions is small, health damages may be smaller still given that wood-fuel emissions generally contain higher fractions of carcinogenic compounds. Indeed, the World Bank survey noted weak evidence of a health effect for grid-based rural electrification.27 The argument that these damages represent a small fraction of the total does not necessarily also constitute an argument that they are not worth considering (particularly given some of the problems in self-reporting of symptoms or illnesses in surveys). Rather, the problem is that uncertainties in the shape of the damage function make estimates speculative. For example, if the reduction of emissions occurs to the right of any significant threshold effects, then even a small marginal reduction may be significant. 26 As noted earlier, for example in the case of Indian PV solar homes, the ERR was claimed to have increased from 30% to 108% when global environmental benefits were considered. 27 The survey noted that (based on self-reported illnesses and symptoms): A relationship between rural electrification and avoided school days missed during a three-month period [could be established]. However, there did not seem to be any similar health benefits for adults in terms of avoided lost work days. These results may suggest the need for more detailed and reliable health questions in future surveys. (World Bank, op.cit.,p.73.) 42 Chapter 5: Externalities _____________________________________________________________________________________ On the other hand, if, say, replacement of wood-fuel based cooking by LPG lowers emissions to the left of any threshold effect, then further reductions of kerosene-based emissions would be unimportant. Given the high level of uncertainty, and the absence of a detailed heath damage study, local air-emission effects are not considered in this economic analysis. However, it is unlikely that these benefits are actually zero. Health Benefits A significant social and human cost of kerosene lighting is that of burn injuries. While these are hard to quantify, their widespread incidence make them more than just anecdotal: in Sri Lanka, it is estimated that 50% of burn injuries are attributable to accidents involving kerosene lanterns and candles, of which a substantial fraction occurs to children.28 Similar problems are reported in many other countries.29 In the Philippines, there have been numerous reports of deaths in fires caused by kerosene lamps:30 indeed such instances may well be most frequent among the poorest families where home-made kerosene lamps are reported. This is an important area for further research, for burn injuries and child deaths have high social costs whose avoidance should be counted as one of the benefits of rural electrification in general (and solar PV systems in particular). Indeed, such direct costs may be easier to ascertain, and most likely of greater magnitude, than those related to local air-pollution. Global Externalities While the treatment of GEF grants in the financial analysis (where they represent a source of funds) is straight forward, their treatment in the economic analysis needs care. The Bank's Procedures (OP10.04) for the calculation of the ERR to be reported in the Project Appraisal Document (PAD) requires that the global environmental benefit be 28 At the Burns Unit of the Lady Ridgeway Children's Hospital, established in December 2000, by August 2001, 176 children had been treated for burn wounds. The majority of the victims were from the unelectrified rural villages of Chilaw, Puttalam and Karapitiya dependant upon kerosene for lighting (Sri Lanka Sunday Observer, 23 September 2001). 29 See, e.g., Measuring the Impact of Energy Interventions on the Poor--An Illustration from Guatemala, V. Foster and J. Tre at http://www.ppiaf.org/conference/section1-paper4.pdf. 30 See e.g. Twin girls killed in Bulacan fire - December 23, 2001: ... Sunday December 23, 2001, Philippines, ... 6, from the fire, but they both sustained severe burns in ... fire officer, said Buena left a lighted kerosene lamp ... www.inq7.net/met/2001/dec/23/met_6-1.htm - 22k Chapter 5: Externalities 43 ______________________________________________________________________________ taken as the amount of the GEF grant (in the years the grants are actually received). As we note below (Section 8), this results in the project ERR increasing from 44.6% to 53.2%. As a matter of theory, however, there are problems with this approach, notwithstanding that GEF grants (or carbon purchases by the Prototype Carbon Fund) do reflect the actual global willingness-to-pay for carbon reduction. In our case, the proposed GEF subsidy for APL-I is $2.50 per Wp for systems of 50Wp or less, and $1.50 per Wp for systems larger than 50Wp. Clearly the actual carbon benefit is not a function of the installed capacity of the PV systems built, but of GHG emissions avoided (and is therefore certainly not inversely proportional to Wp installed!).31 Table 21 shows the implied cost of avoided carbon for the three different systems, which are significantly different. Table 21: Cost of Avoided Carbon (kerosene replacement only) GEFGrant 15year $/tonC carbon saving [Peso] [$] [Kg] [$/ton] 20Wp 2550 50 1443 35 40Wp 5100 100 1816 55 75Wp 5738 113 9710 12 These data take as a basis of the avoided carbon the quantity of kerosene that is displaced by the PV system, using an emission factor of 0.688 Kg C per litre of kerosene.32 It might be argued that the PV system also displaces the fossil fuel that is required to manufacture dry cells and to generate the electricity for auto-battery charging, but many additional (and possibly controversial) assumptions would be so required. In addition, since the larger systems provide a higher level of electricity service than the kerosene that is replaced (which would be provided by some mix of grid-based fossil and hydro generation, or by diesel generation in minigrids), the avoided cost of carbon may in fact be lower than the estimates implied by Table 21. 31 Indeed, whatever GEF's accounting procedures, it should be obvious that there can only be one value of carbon damage costs, because all carbon emissions, from whatever their location on earth, make an identical contribution to actual global damage costs. 32 This emission factor is based on the values for kerosene given in IPCC, Reference Manual and Workbook for National GHG Inventories, 1996. Calorific value is 44.75 TJ/103 tonnes; carbon emission 19.6 ton C/TJ ; and density of kerosene 0.7855 (kg/litre). 44 Chapter 5: Externalities _____________________________________________________________________________________ Table 22 shows the impact on ERR of the GEF subsidy. For 20 and 40Wp systems, the impact on ERR using 15$/ton of carbon (the estimate that is widely used as a basis for global carbon reduction benefits) will give values that lie between the two ERR estimates; and for the larger 75Wp systems, slightly above the GEF value (which has an implied value of only 12$/ton). Table 22: Increase in ERR Baseline ERR GEF method (consumer (OP 10.04) surplus) 20Wp 45.0% 50.4% 40Wp 46.6% 56.4% 75Wp 32.2% 38.5% CHAPTER 6. SUBSIDY STRUCTURE Having established a better estimate of the economic benefits using the consumer surplus approach, we now return to the question of the subsidy structure.33 We note the relationship between size of government capital grant subsidy, and the net economic benefit achieved by this subsidy. Clearly the proposed subsidy design meets one of the first tests of efficiency: subsidies leverage significantly larger economic benefits. Table 23: Proposed Subsidy v. Economic Benefit (using consumer surplus method) Proposed NPV, net Subsidy, as FRR NPV of net subsidy economic % of to consumer government benefits benefit tax revenue 20Wp 8,000 34,562 23 160% -5,634 40Wp 5,000 44,235 11 17.4% -1,866 75Wp 0 32,548 0 0.9% 4,662 However, we also note that the consumer FRR, based on avoided costs, show very large variations across the three types of systems; and that, in the case of the 75Wp systems, the net effect on Government revenue is an increase, largely the result of tax receipts from the assumed dealer margins, and the VAT of PV system modules. Given the fungibility between capital subsidy and VAT noted earlier,34 what really matters is the net flow to/from government. However, as noted in Table 24, the net revenue from the 75Wp systems offsets 18% of the government's cost to the other categories. Whether this is intended or not requires discussion with the Government. 33 See also W. Mostert (Rural Electrification Subsides in the Philippines, Volume 2: Rural Electrification Subsidy Principles, A Reference Manual. Report to the Department of Energy, National Electrification Administration and the World Bank, 2001) Section 5.6: Policy Recommendations for the Design of SHS subsidy systems. Mostert argues that subsidies should be designed to maximize the ratio of market expansion to free riders, and calls for a three-phase approach: a first phase in which a market priming subsidy is launched; a second phase when no subsidies are give; and a third phase where a market deepening subsidy is introduced. The proposed subsidies to be given during APL-I fall into the first category. 34 Though again see Mostert (2002) on the incentive value of these two approaches. 46 Chapter 6: Subsidy Structure _____________________________________________________________________________________ Table 24: Aggregate Impact on Government Revenue 20Wp 40Wp 75Wp Total Targets [number 2000 7000 1000 10000 systems] NPVGovtImpact [P/system] -5651 -1885 4442 aggregate impact [P million] -11.3 -13.2 4.4 -20 Figure 9 shows the impact of subsidy level on FRR and aggregate government revenue. A revenue neutral subsidy would be around P4,750 ­ but the FRR climbs only to around 5%; and even at a subsidy of P 8,000, the FRR is still around 10%. We conclude that even with higher subsidy, the FRR based on avoided cost is nowhere close to the hurdle rate. It follows that if non-poor households do indeed show the high willingness-to-pay, then offering them a subsidy (or a VAT reduction) will make little difference: they become "free riders".35 Figure 9: Capital Subsidy, Government Revenue and FRR NPV 6 0.1 as 4 0.08 1000P <------GOVERNMENT REVENUE FRR -------> 2 0.06 FRR Revenue, 0 0.04 consumer Government on -2 0.02 impact -4 0 0 1000 2000 3000 4000 5000 6000 7000 8000 Capital subsidy, pesos These calculations do not take into account the price elasticity of demand for solar home systems. Figure 10 shows the demand and supply curves for subsidies: the x-axis shows 35 The reverse "Ramsey Pricing" principle calls for subsidies to be focused on those products or services with the highest elasticity of demand in order to achieve maximum market expansion per invested subsidy amount (Mostert, 2001). Since the estimated price elasticity (of lumen-hour demand) is highest for the poorest families (20Wp systems, -1.2), and smallest for the non-poor (75Wp systems, - 0.81), the emphasis on subsidies for the smallest systems meets this criterion. Chapter 6: Subsidy Structure 47 the number of rural households obtaining access to electricity from SHS per year;36 the supply curve shows the relationship between the number of customers and subsidy rate for a given annual subsidy budget. As drawn we assume an annual subsidy budget for 40Wp systems of P3.5million, at which level we expect first year demand be 700 40Wp systems: hence the demand curve is drawn to intersect the supply curve at this point. One must concede, however, that while the supply curve is known (given knowledge of the costs of the SHS systems and the subsidy), the estimate of demand (at this price) is subject to considerable uncertainty. However, since the FRR for 40Wp systems is 17.4%, purely on a replacement cost basis is seems unlikely that demand would be significantly less. Figure 10: Demand and Supply for Subsidies (40Wp systems) 35 30 25 1000P SUBSIDY BUDGET=3.5m 20 system, of cost 15 INELASTIC DEMAND 10 5 0 200 400 600 800 1000 1200 number of households per year However an additional dimension of uncertainty is the elasticity of demand. For example, if the first-year subsidy budget were cut by half, to P1.75million (see Figure 11), then with the relatively inelastic demand curve of Figure 10, the impact on sales would be relatively small, falling from 700 systems to about 650 systems. But if the demand curve is more elastic, then the same cut in subsidy budget would result in only 500 systems sold. 36 The demand curve for subsidies shown in Mostert (2001), p.25 has an upward slope because of his definition of the y-axis as the quantum of subsidy rather than the price. We prefer the conventional representation so as to avoid confusion with other demand curves shown in this report. 48 Chapter 6: Subsidy Structure _____________________________________________________________________________________ Figure 11: Impact of Demand Elasticity 35 30 25 SUBSIDY BUDGET=1.75m 1000P SUBSIDY BUDGET=3.5m 20 system, of ELASTIC DEMAND cost 15 INELASTIC DEMAND 10 5 0 200 400 600 800 1000 1200 number of households per year A related issue is dealer margin, for which we use a figure of 100% (based on the Sri Lanka experience). These are healthy margins, that may well be competed down over time ­ though in the initial years of the project such margins may well be necessary until consumer acceptance is demonstrated. Figure 12 shows that the FRR does indeed respond to cuts in margin: but in the early years, even if margins are cut to 60-70%, the FRR is still less than 10%. Figure 12: Dealer Margins and FRR (75Wp Systems) 0.15 0.1 FRR consumer0.05 0 1 0.9 0.8 0.7 0.6 0.5 0.4 dealer margin Chapter 6: Subsidy Structure 49 In any event, if indeed the willingness to pay for the large 75Wp systems is insufficient to overcome the poor FRR based on replacement costs only, then non-poor households can simply buy the smaller 40Wp systems, whose FRR is satisfactory. The VAT exemption would also be worth about P3,200, which would still mean a net increase in Government revenue for this category of system. CHAPTER 7. RISK ANALYSIS Although the sensitivity analyses and switching values suggest robust economic returns, this needs to be verified in a more general risk analysis, in which all of the major input assumptions are specified as probability distributions, and the economic returns are estimated as a probability distribution (that reflects the likely circumstances in which all input variables vary simultaneously, rather than just one at a time). Table 25 (and Figure 13) shows the probability distributions assumed for the input variables. Since 40Wp systems account for the bulk of the installations, the analysis is for this system size class. Table 25: Input Assumptions for Risk Analysis Assumption Probability distribution Basis Equipment costs Triangular capital costs may be lower than estimated, (imported PV given worldwide cost decreases from scale modules) economies and learning curve effects Project Life Triangular module life may be lower than the 15years assumed in the basecase Allocation fraction Uniform (from 0.3 to Uncertainty in consumer behaviour (share of PV system 0.5) output devoted to Lighting) Lumen-hours/day Normal (mean= 2520, Uncertainty in consumer behaviour (as it = 200) affects the calculation of consumer surplus) TV-hours per day Normal (mean=4, =1) Uncertainty in consumer behaviour (as it affects the calculation of consumer surplus) 52 Chapter 7: Risk Analysis _____________________________________________________________________________________ Figure 13: Input Probability Distributions source: Chapter 7: Risk Analysis 53 _____________________________________________________________________________ The resulting probability distribution (of 1000 trials) for the ERR of the 40Wp systems is shown in Figure 14. The probability of the ERR falling below 15%, the hurdle rate, is less than 1%. The average ERR of the 1000 trials is 47.7% (as opposed to 46.6% using "most likely" (median) estimates of assumptions). Figure 14: Distribution of Rate of Return 0.4 @ref=0.466 @avg=0.477 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 1.2 CHAPTER 8. AGGREGATE ECONOMIC BENEFITS We next aggregate the costs and benefits over the entire project. For each of the three size/income-classes we estimate the number of systems installed each year of the period of APL-I; and the benefits of each cohort extended over their respective lifetimes (See Table 26). The baseline ERR for the SHS component of the project is 44.6%.37 When the GEF subsidies are treated as economic benefits, the ERR increases to 53.2%. Table 26: Aggregate Economic Flows: APL-I source: 37 This is the weighted average of the individual ERRs presented previously: as may be seen from Table 26, the ERRs for each size category are the same as in the earlier individual tables. Since 70% of the systems are assumed to be in the 40Wp size (ERR 46.6%), the overall ERR is fairly close to this value. CHAPTER 9. CONCLUSIONS Solar PV systems for individual homes are found to be economic. Under the most conservative assumptions the ERR for APL-I as a whole is 44.6%. The consumer- surplus based ERRs are 45% for the smallest 20Wp systems, 47% for 40Wp systems, and 32% for 75Wp systems. Table 27 summarizes the results (including the FRR from the consumer's perspective, the ERR using only avoided costs, and the ERR when also considering global carbon benefits). Table 27: Summary of Results FRR ERR ERR ERR Consumer Benefits at Benefits based Consumer perspective avoided on increase in surplus benefits economic consumer plus GEF costs only surplus subsidy as benefit 20Wp systems 160% 15.4% 45% 50.4% 40WP systems 17.4% 11.4% 46.6% 56.4% 75Wp systems 1.6% 8.7% 32.2% 38.5% Entire project 44.6% 53.2% We note that for the smallest 20Wp systems, the ERR when using avoided costs is marginally above the hurdle rate of 15%; and with the proposed subsidy schemes, the FRR (from the consumer's perspective) are high (and consistent with the high personal discount rates observed in poor families). If the marketing and institutional aspects of the project are competently handled, the probability of realizing the economic benefits is therefore also high. The low consumer-based financial rate of return for the non-poor (looking only at replacement costs) is offset by their high willingness to pay for the very much greater level of service of the larger PV-systems, as observed in almost all other countries with similar projects (as e.g. in India and Sri Lanka). Moreover, if the willingness-to-pay for such systems does not offset the low FRR, the non-poor households can always buy the smaller 40Wp systems, whose FRR is satisfactory. The result of net economic benefit is robust with respect to input assumptions in the plausible range. A switching values analysis shows that increases in initial cost, problems in system performance, and assumptions about the shape of the demand curve all pose relatively small risks to achieving the project benefits. With 10,000 systems as the goal for APL-I, the risks of having overestimated market size is small (requiring sales to 2.2%, 9.5% and 1.5% of poorest, poor and non-poor households, respectively, that use 58 Chapter 9: Conclusions _____________________________________________________________________________________ batteries for radio/TV, or 1.1%, 4.75% and 0.75%, respectively, of households that use either batteries or connected to small diesel generators). The input assumption with the most significant potential effect on ERR is the life of the PV module lives, which we assume at 15 years in the reference case. However, because of the large magnitude of (consumer surplus) benefits, significant erosion of ERR occurs only if model lives are below 7-8 years, which based on experience in other countries seems unlikely. The benefits are consistent with those estimated in other countries for similar projects (e.g. 30% ERR as estimated by the ICR for a similar project component in India). They reflect high willingness to pay for the improved levels of lighting service, and the significantly higher levels of TV viewing. While the benefits may be smaller than those for full grid extension (or mini-grid) based-electrification, for areas where such electrification strategies are not cost-effective, PV-powered solar homes represent an economically efficient alternative. Finally, the results of Table 27 do not include a range of further benefits that include the avoidance of significant number of burn injuries and fires; the benefit to families of higher levels of educational achievement; and the benefit of attaining higher levels of family income. REFERENCES Barnes, D.F. A. Domdom, V. and H. Peskin, Rural Electrification and Development in the Philippines: Measuring the Social and Economic Benefits, ESMAP Report 255/02, May 2002. Cabraal, A., M. Cosgrove-Davies and L. Schaeffer, "Best Practices in Household Electrification Programs in Developing Countries, World Bank Technical Series No. 324, 1996. Madecor Environmental Management Systems Inc., 2001. Market Assessment for Rural Electrification, Report to Philippines Department of Energy, Final Report. Martinot, E., A. Cabraal, S. Mathur, World Bank/GEF Solar Home Systems Projects: Experiences and Lessons Learned 1993-2000. Renewable and Sustainable Energy Reviews 5(1):39-57, 2001. Meritec, 2002. India: Renewable Resources Development Project, Review of Project for Implementation Completion Report, Report for the World Bank. Mostert, W. 2001a. Rural Electrification Subsides in the Philippines, Volume 1: Policy Recommendations for optimized use of re-subsidies and the national lifeline rate, Report to the Department of Energy, National Electrification Administration and the World Bank. Mostert, W. 2001b. Rural Electrification Subsides in the Philippines, Volume 2: Rural Electrification Subsidy Principles, A Reference Manual. Report to the Department of Energy, National Electrification Administration and the World Bank. Mostert, W., 2002. Rural Electrification Subsidy Principles: A Reference Manual. AFFREI, World Bank, Washington, DC. Patalinghug, E., N. Mendoza and R. Alonzo, 2002. Rationalisation of Subsidy Policy for Rural Electrification, Report to the World Bank and the Department of Energy, Government of Philippines. Smith, W., 2001. Designing output-based aid schemes: A Checklist. Output-Based Aid, p.91-117. Van der Plas, R. and A. B. de Graff, 1988. Comparison of Lamps for Domestic Lighting in Developing Countries, World Bank, Industry and Energy Department Working Paper, Energy Series Paper #6, Washington, DC. 60 Chapter 9: Conclusions _____________________________________________________________________________________ World Bank, 2002. Sri Lanka Renewable Energy Development, Project Appraisal Document, Report 23886-CE, Washington, DC.