47107 W O R L D B A N K W O R K I N G P A P E R N O . 1 6 0 Residential Electricity Subsidies in Mexico Exploring Options for Reform and for Enhancing the Impact on the Poor Kristin Komives Todd M. Johnson Jonathan D. Halpern José Luis Aburto John R. Scott THE WORLD BANK W O R L D B A N K W O R K I N G P A P E R N O . 1 6 0 Residential Electricity Subsidies in Mexico Exploring Options for Reform and for Enhancing the Impact on the Poor Kristin Komives Todd M. Johnson Jonathan D. Halpern José Luis Aburto John R. Scott THE WORLD BANK Washington, D.C. Copyright © 2009 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First Printing: January 2009 printed on recycled paper 1 2 3 4 5 12 11 10 09 World Bank Working Papers are published to communicate the results of the Bank's work to the development community with the least possible delay. The manuscript of this paper therefore has not been prepared in accordance with the procedures appropriate to formally-edited texts. Some sources cited in this paper may be informal documents that are not readily available. The findings, interpretations, and conclusions expressed herein are those of the author(s) and do not necessarily reflect the views of the International Bank for Reconstruction and Devel- opment/The World Bank and its affiliated organizations, or those of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank of the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/The World Bank encourages dissemination of its work and will normally grant permission promptly to reproduce portions of the work. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, Tel: 978-750-8400, Fax: 978-750-4470, www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW,Washington, DC 20433, USA, Fax: 202-522-2422, email: pubrights@worldbank.org. ISBN-13: 978-0-8213-7884-7 eISBN: 978-0-8213-7886-1 ISSN: 1726-5878 DOI: 10.1596/978-0-8213-7884-7 Library of Congress Cataloging-in-Publication Data Residential electricity subsidies in Mexico: exploring options for reform and for enhancing the impact on the poor/Kristin Komives . . . [et al.]. p. cm. -- (World Bank working paper; no. 160) Includes bibliographical references and index. ISBN 978-0-8213-7884-7 (alk. paper)--ISBN 978-0-8213-7886-1 (electronic: alk. paper) 1. Electric utilities--Subsidies--Mexico. 2. Electric utilities--Rates--Mexico. 3. Subsidies-- Mexico. I. Komives, Kristin. HD9685.M62R47 2009 333.793'23--dc22 2008051806 Contents Acronyms and Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2. Residential Electricity Subsidies and Tariffs in Mexico . . . . . . . . . . . . . . . . . . . . . . 5 3. Distributional Incidence of Residential Electricity Subsidies . . . . . . . . . . . . . . . . 15 4. Assessment of Alternative Subsidy Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 5. Synthesis of Findings and Ways Forward . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 APPENDIXES A: Evolution of Residential Electricity Subsidies and Tariffs in Mexico . . . . . . . . . . . . 51 B: Residential Electricity Tariffs in 2006 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 C: Tariff Reclassifications, 2002­06. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 D: Analysis of Subsidies Using CFE Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 E: Analysis of Subsidies Using ENIGH Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 F: Tariff Reform Scenarios. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 iii Acronyms and Abbreviations CAS Chile's Means-tested Program (Caracterización Social) CFE Mexico National Electricity Commission (Comisión Federal de Electricidad) DAC High Consumption Residential Tariff (Tarifa Doméstica de Alto Consumo) ENIGH Mexico National Household Income and Expenditure Survey (Encuesta Nacional de Ingresos y Gastos de los Hogares) GDP Gross domestic product HH Household IBT Increasing block tariff IPP Independent power producer kWh Kilowatt-hour LFC Central Light and Power, Mexico City (Luz y Fuerza del Centro) MP Mexican peso MWh Megawatt-hour OECD Organisation for Economic Co-operation and Development Oportunidades Mexico's conditional cash transfer program for the poor Oportunidades Oportunidades program focused on energy (established 2007) Energéticas Pidiregas Deferred financing mechanism to support private sector energy sector investment in Mexico (Programas de Inversión de Impacto Diferido en el Gasto) RES Residential Electricity Schedule SEDESOL Ministry of Social Development (Secretaría de Desarrollo Social) SENER Ministry of Energy (Secretaría de Energía) SHCP Ministry of Finance (Secretaría de Hacienda y Crédito Público) VDT Volume differentiated tariff Currency Equivalents Currency Unit = Mexican Peso Exchange Rate (Exchange Rates Effective April 3, 2008) 1 US Dollar = 10.557 Mexican Pesos Executive Summary This report addresses a pressing issue in Mexico's electricity sector--the large and growing subsidies to residential consumers and their regressive incidence across different segments of the population. It responds to requests from the Ministry of Energy (SENER) to provide a preliminary assessment of alternatives to the current subsidy system, building on prior collaboration between the Government of Mexico and the World Bank on the distribu- tional impact of public spending, the performance of conditional cash transfer programs and other poverty-targeted programs, and related work on pricing and subsidies for infra- structure services. This study was designed as the first phase of a multiphase program of collaborative analytical work. This first phase provides estimates of the distributional and fiscal performance of alternative subsidy targeting mechanisms to help inform discussion and deliberations on feasible goals and practical approaches over the medium term. A sec- ond phase would address transition paths, specific compensatory mechanisms, and deci- sion processes for pursuing the options that the Mexican authorities deem most promising. Electricity subsidies in Mexico are among the largest in the world (US$9 billion in 2006) and have absorbed a significant proportion of public resources. Subsidies in 2006 were equivalent to about 1 percent of gross domestic product and were more than one-third of total electricity sector revenues. Over two-thirds of electricity subsidies go to residential consumers, and the volume of subsidies to residential customers increased by 46 percent between 2002 and 2006 in real terms. Tariff subsidies of this magnitude impact heavily on the performance of the electric- ity sector and on Mexican society more generally. First, the cost of the subsidies needs to be covered in some way--if not by con- sumers, then by government; if not directly by the government, then by reduced spending on system expansion to meet growing demand,on service quality upgrades, and on operations and asset maintenance. Fiscal transfers to the utilities divert resources from priority social and economic programs and reduce fiscal space for financing priority investments in the electricity sector. The 2006 federal budget for investment in electricity (excluding the Programas de Inversión de Impacto Diferido en el Gasto [Pidirigas]) was 40 billion pesos, only slightly larger in nominal terms than it was in 2001 and less than half the total volume of subsidies. Second,subsidiesdistortpricesignals,elevatingdemandabovewhatitwouldbeif electricity was priced at marginal or average cost. Massive underpricing also mutes incentives for customers to take energy saving measures, such as replacing old equipment and appliances. Higher customer demand requires additional power supply, especially to meet periods of peak demand. Because elevated residential electricity demand tends to lower load factors for the power system, it leads to reduced efficiency in the use of generation and transmission facilities and thus to a higher marginal cost of supplying all customer classes. This has negative implica- tions for economic competitiveness. Subsidiesalsoengenderenvironmentalcosts.Elevateddemandleadstoincremen- tal emissions from power plants, including local pollutants responsible for poor vii viii Executive Summary air quality (such as particulates and ozone precursors) and global pollutants linked to climate change. Further, standard merit order dispatch rules exacerbate the ten- dency for marginal generating plants (those with low capital costs, but high fuel costs, such as small fuel-oil and diesel-fired facilities) to generate higher levels of pollutants. In rural areas, electricity subsidies for irrigation pumping exacerbate overexploitation of groundwater resources. Moreover, the bulk of subsidies go to the non-poor. Numerous reports in the last sexenio have pointed out the regressive distributional incidence of these subsidies among residen- tial customers. Cognizant of this problem, the Mexican authorities expressed a strong inter- est in evaluating the impact of a shift from quantity-based targeting through the tariff structure to means testing, building upon the poverty profiling and verification systems established under Oportunidades, Mexico's well-functioning, means-tested conditional cash transfer program. The pilot effort for such a system was recently implemented with Oportunidades Energéticas, which, with its 3 billion peso annual budget, is small relative to existing tariff subsidies. This program, begun in 2007, did not replace or modify existing tariff-based subsidies, but rather operates in parallel with them. Among the alternatives to the current subsidy system, the study examines the use of means testing to allocate subsidies and modifications in electricity tariff structures based on international experience. It also assesses the impact of achieving lower unit costs of sup- ply on subsidies. Each alternative is assessed in terms of its potential to reduce the overall magnitude of electricity subsidies and improve distributional outcomes. Residential Electricity Subsidies and Tariffs in Mexico Electricity subsidies as reported in Mexico are financial subsidies that result from below- cost pricing, with cost measured as accounting costs. Subsidies are measured as the differ- ence between the price of electricity paid by consumers and the average cost of supply. This approach to measuring subsidies does not capture the economic cost of electricity provi- sion, which could differ significantly from the accounting costs. Accounting costs are reported by the two operating utilities--the Mexico National Electricity Commission (Comisión Federal de Electricidad, CFE) and Central Light and Power (Luz y Fuerza del Cen- tro, LFC)--according to government accounting standards. CFE's subsidies have largely been financed by a bookkeeping transfer. The federal government essentially reimburses the CFE for providing subsidies to its customers by discounting the taxes and dividends (aprovechamiento1) that CFE would otherwise have to pay the government. Since 2002, the volume of subsidies has exceeded the notional amount of aprovechamiento and has there- fore begun to erode CFE's capital base. For LFC, the electricity provider for the Valle de Mexico and surrounding areas, its financial situation has been so dire that the federal gov- ernment provides the company with a direct cash subsidy to cover its mounting operating deficits and customer subsidies. The current system of residential subsidies is not consistent with the basic precepts prescribed in Mexico's legal framework for the electricity sector. The legal framework gov- 1. Equal to 9 percent of net fixed assets. Executive Summary ix erning electricity tariffs in Mexico establishes that tariffs should cover costs and promote efficient consumption. The tariffs of state-owned enterprises are to be set "in accordance with economic efficiency and financial health criteria" (Article 26, Reglamento de la Ley de Empresas Estatales). Electricity tariffs in particular are to be "fixed, adjusted, and restruc- tured in a way that covers financial and public service expansion requirements, and rational energy consumption"(Article 31, Ley de Servicio Público de Energía Eléctrica). Current res- idential electricity tariffs are far from meeting these objectives. Because only a tiny pro- portion of residential customers pay the marginal cost of electricity service, current tariffs achieve neither the cost recovery nor the efficient consumption provisions of the current legislation. With respect to subsidies, the legal framework establishes that subsidies "shall be tem- porary" and subject to the criteria of "objectivity, equity, transparency, publicity, and selec- tivity." The law further states that, "in programs of direct benefit to individuals or social groups, amounts [of the subsidy] and percentages shall be determined on the basis of redis- tributive criteria that must privilege the lower income population and aim at equity amongst regions and states"(Article 75, Ley Federal de Presupuesto y Responsabilidad Hacendaria). In actuality, the regressive nature of residential electricity subsidies in Mexico is widely known, and the situation has deteriorated further since the 2002 tariff revision. Mexico's current system of tariff-based subsidies is among the most complex in the world, and is largely a product of accretion rather than deliberate policy. Electricity subsi- dies were first introduced in Mexico in 1973 in response to persistent inflation, when the single electricity tariff was changed into a three-part, increasing block tariff, with sub- sidized rates for the first two blocks. The first"summer subsidy"(Tariff 1A) was introduced in 1974, providing additional subsidized rates to customers living in hotter areas (1A was defined as regions with more than four months of average temperatures above 25 degrees Celsius [°C]). Successive climate-based tariffs with increasingly subsidized rates over larger volumes were introduced in 1988 (Tariff 1B > 28 °C; 1C > 30 °C), 1990 (1D > 31 °C), 1995 (1E > 32 °C), and 2002 (1F > 33 °C). Today, Mexico has an extremely complex tariff sys- tem with over 112 different billing possibilities for residential consumers. Among the most important changes undertaken as part of the 2002 tariff revision was the introduction of a new tariff schedule for households consuming large volumes of elec- tricity known as the High Consumption Residential Tariff (Tarifa Doméstica de Alto Con- sumo, DAC), which was to raise tariffs slightly above long-run marginal costs for roughly the top 5 percent of customers in each of the seven tariff categories. The 2002 tariff reform had no long-lasting impact on subsidy volumes--although residential subsidies dropped slightly in real terms in 2002, by 2003 they had risen above 2001 levels. The primary reasons for the rebound in subsidies were: (a) the introduction of yet a more highly subsidized summer tar- iff category (Tariff 1F), (b) the reclassification of large numbers of consumers (3.4 million between 2002 and 2006) to more highly subsidized tariff schedules, and (c) increasing elec- tricity supply costs. The unit cost of providing electricity service is a key driver of both the magnitude and distributional incidence of electricity subsidies.Given the way that subsidies are measured-- tariff minus accounting cost--the magnitude of subsidies is directly related to the cost of supplying electricity. Reducing costs can offset, at least partly, the need for raising tariffs in an effort to reduce subsidies. Any cost reduction would result in a greater proportion of subsidies going to low-income households. Since 2002, CFE's average cost per Megawatt- hour (MWh) delivered has increased by about 20 percent, while LFC's costs were over twice x Executive Summary as high as CFE's in 2006. While fuel costs (mainly natural gas and fuel oil) rose by nearly 70 percent during this period, other factors contributing to relatively high electricity sup- ply costs in Mexico include high personnel costs (wages, benefits, pensions) and large losses (both technical and commercial). For example, between 2002 and 2006, retirement pay- ments and wages and benefits increased by 35 percent and 24 percent, respectively. While transmission losses have declined in percentage terms for both CFE and LFC in recent years, distribution losses have been rising and are especially high for LFC--distribution losses for CFE increased from 11 percent to 11.6 percent between 2000 and 2005, while those of LFC have exceeded 30 percent since 2005. Distributional Incidence of Electricity Subsidies Estimating electricity consumption and subsidies by income level in Mexico is challenging due to the lack of comparable data. Data from CFE provide information on the distribu- tion of costs and electricity consumption within each tariff class and can be used to assess the distribution of subsidies by electricity consumption level. Independent of CFE statis- tics, household survey data from the Mexico National Household Income and Expenditure Survey (Encuesta Nacional de Ingresos y Gastos de los Hogares, ENIGH) on electricity expen- ditures by household can be used to estimate the distribution of subsidies by income level. The level of electricity consumption is often used as a proxy for income; however, in practice, this correlation is not perfect. Not all poor households consume small amounts of electricity due to a range of factors, such as old and inefficient appliances or because they live in very hot climates. For different reasons, rich households do not all consume a lot of electricity (for example, seldom-used second homes or single professionals who spend long hours away from their home). CFE data show an 80-fold difference in annual consump- tion between average consumption in the 1st consumption decile and the 10th consump- tion decile. By comparison, ENIGH data find that there is only a threefold difference in average monthly electricity consumption between the poorest income decile (139 kWh per month) and the richest income decile (380 kWh per month) during the summer months. There are a number of other differences between these two data sources that lead to some differences in the empirical estimates generated with each source. Nonetheless, the over- arching conclusions and general observations about subsidy incidence that emerge from the two sets of data are strikingly similar. The distributional incidence of residential electricity subsidies is highly regressive, is most pronounced for the climate-based tariff categories (the summer subsidies),and appears to have worsened over time. In 2005, the bottom three consumption deciles accounted for roughly 16 percent of electricity subsidies while the top three deciles accounted for 39 per- cent.The proportion of subsidies received by the top consumption decile has increased since the 2002 tariff review.In 2003,the top consumption decile accounted for 4.4 percent of total subsidy benefits, while by 2005 this percentage had increased to 10 percent. Likewise, DAC consumers (those paying the highest rates) accounted for 5 percent of total customers in 2002, but by 2005 this figure had dropped to just 2 percent. Estimates of subsidy distribution by income decile show that residential subsidies dis- proportionately benefit non-poor households,with the subsidies provided through climate- based tariffs (that is, 1A­1F) having the most regressive distributional incidence.While the bottom three income deciles account for about 21 percent of total subsidies, the top three Executive Summary xi income deciles account for 38 percent. Among Tariff 1F customers (the most highly subsi- dized customer group), more than one-quarter of total subsidies go to the top income decile alone. In contrast, ENIGH data suggest that the pilot program Oportunidades Energéticas has a very progressive distribution of resources across income classes, with nearly 75 percent of the payments going to the bottom three income deciles. While the majority of residential electricity subsidies benefit middle- and upper-income households, such subsidies also represent a significant proportion of household income for low-income consumers. Any reform of the current electricity tariff system should therefore consider and mitigate the impact that such changes would have on the poor. Assessment of Alternative Subsidy Mechanisms What options does Mexico have for reducing the burden of the large and growing electric- ity subsidies on public finances and redressing their highly regressive incidence? Four options were assessed with regard to their impact on distributional incidence and aggre- gate volume--two modifications of the tariff structure, the introduction of means testing, and a reduction in supply costs (alone or in combination with the tariff changes). The assessment of each option is based on simulations, using the same databases and estima- tion methods as were employed in evaluating the distributional performance of the cur- rent system. The options were chosen to reflect, on the one hand, current thinking in Mexico about possible directions for changes in the tariff system and, on the other hand, to evaluate options that have been shown to be promising in other countries.2 This sort of simulation-based assessment provides a first-order approximation of possible end states. It does not define specific time horizons or precise transition paths. The purpose is not to present all feasible options but rather to promote policy discussion by demonstrating four distinct alternatives to the current subsidy situation. Minor Modification of Existing Tariff Structures Among the electricity tariff reform proposals currently under discussion in Mexico is the modification of the DAC tariff to increase its coverage and slightly lower the rate to reflect long-run marginal costs. The simulation assumed that the top 10 percent of customers in each tariff class would pay the new DAC tariff. This option would reduce the overall mag- nitude of subsidies provided to residential customers by a mere 4 percent with little change in the distributional incidence across consumption deciles.Extending the DAC to more cus- tomers has very little impact because it does not modify the subsidies provided to 90 per- cent of electricity consumers.Moreover,the potential revenue gain associated with applying a DAC tariff to the top decile is less than otherwise expected because consumption in the top decile would decline in response to the increase in prices for "new" DAC customers. A National Volume Differentiated Tariff (VDT) This option involves moving from the current system of multiple increasing block tariffs (IBTs) to a much simpler national VDT system. A VDT preserves the goal of an IBT, which 2. The simulations were specified based on discussions with government officials in January 2008. xii Executive Summary is to provide greater subsidies to low-volume consumers. The difference is that each cus- tomer class pays one price per kWh consumed and that price is set by the customer's monthly consumption level (as is now done only with the DAC tariff). The price for all kWhs consumed is thus higher for large-volume consumers than for small-volume con- sumers. Unlike the IBT, however, the VDT's largest discounts are limited to only the lowest- volume consumers, because large-volume consumers receive no highly subsidized blocks of electricity. The VDT simulation results indicate that the implementation of a VDT sys- tem would reduce the total volume of residential electricity subsidies by about half and would slightly improve the distributional incidence of subsidies. These results are consis- tent with studies from other countries that show that moving from an IBT to a VDT pro- duces a neutral or slightly progressive subsidy distribution. When the VDT is combined with a 20 percent cost reduction (such that those paying the top tariff are no longer sub- sidized), the volume of subsidies could be reduced by more than 80 percent and would dra- matically improve the targeting of remaining subsidies to the poor (Table ES1). Table ES1. Performance Indicators for VDT Scenarios Consumption Deciles Income Deciles (CFE data) (ENIGH data) VDT with 20% VDT with 20% VDT Cost Reduction VDT Cost Reduction Total volume of subsidies as a 45% 16% 48% 11% percent of current Percent of subsidies going to 32% 54% 30% 43% first 4 deciles Ratio of subsidies for deciles 1­3 0.56 4.94 0.53 1.23 to subsidies for deciles 8­10 Means Testing Building on the existence and success of Mexico's means-tested program,this option assumes that all Oportunidades participants would receive a discount on electricity consumption (either the first 100 kWh, 200 kWh, or all consumption), while all other consumers would pay a uniform volumetric tariff set at 80 percent of CFE average cost (the same level as the top tariff in the VDT scenario). As with the VDT scenario, the impact of these tariff modifi- cations alone was simulated, as was the impact of the means-tested discount in combination with cost reductions. In combination with cost reductions, the tariff with means-tested dis- counts of various sizes yielded by far the most progressive subsidy distribution of all the alter- natives. The simulation of a discount on 100 kWh resulted in 66 percent of total subsidy benefits going to the lowest two income deciles, and would be even larger (82 percent) if all households in the lowest income quintile were covered by Oportunidades. Reducing Supply Costs If the costs of service provision could be reduced, for example through improvements in productive efficiency, this alone would significantly reduce the magnitude of subsidies without raising tariffs or changing the tariff structure. The simulation with ENIGH data Executive Summary xiii indicates that a reduction in average costs of 10 to 20 percent would permit aggregate residential subsidies to decline by 21 and 42 percent, respectively. Similarly, simulations using CFE data found that a 20 percent reduction in costs would reduce subsidies by 35 percent. Moreover, the simulations show that such cost reductions would slightly improve the distributional incidence of residential subsidies, by consumption decile and by income decile. Synthesis of Findings and Ways Forward The alternatives assessed in this study are stylized illustrations of the many options that exist. They could be employed in combination to enhance their impact: for example, a VDT could be combined with a means-tested discount,by offering the most discounted tariff in theVDT only to those eligible for Oportunidades. Likewise, the coverage of a means-tested discount could be improved by offering the discount to all households in the country's most marginal communities.3 The simulations do not provide an exhaustive review of possible tariff/sub- sidy options, but rather illustrate how different measures (changes in the DAC, shifting from IBTs to VDTs, means-targeted discounts, and reduction in supply costs) would perform in terms of reducing subsidy volumes and improving distributional incidence. Six clear findings emerge from the evaluation of alternatives to the current tariff/subsidy scheme for residential consumers of electricity. There is Relatively Little to be Gained from Minor Tinkering With the Current Tariff Structure Mexican electricity tariff structures are among the most complex in the world by design and by natural accretion. The complexity lends itself to reclassification and misclassifica- tion of specific consumer groups seeking to obtain still greater discounts on ever larger vol- umes on consumption.A minor tariff "reform"that centers around reducing the DAC tariff slightly, and expanding its coverage to the 10 percent of highest-volume consumers in each tariff category, has almost no impact on the distribution of subsidy benefits and has min- imal impact on the total volume of subsidies. Moreover, history suggests that tariff adjust- ments of this type are likely to be followed by intense pressure from those most affected by the higher tariffs to introduce new, more highly discounted tariff categories, negating the impact of the change. Combined With Cost Reductions, the Means-tested Discount Produces the Most Progressive Distribution of Subsidy Benefits, but a National VDT is Also a Significant Improvement Over the Current Tariff System The VDT and the means-tested discount scenarios both produce progressive subsidy dis- tributions when combined with 20 percent cost reductions. The VDT does a good job of targeting subsidies to low-volume consumers,and achieves a slightly-more-progressive-than- neutral distribution of subsidies across income groups. The means-tested discount on 3. Geographic targeting is also possible, based on the concentration of "marginal" groups. xiv Executive Summary 100 kWh is highly progressive--almost all the benefits flow to the lowest income quintile of the population. Extending the discount to 200 kWh or to all consumption of targeted households makes the tariff structure even more progressive and reduces errors of exclusion. Improving the targeting performance of Oportunidades itself would also improve distrib- utional incidence. A means-tested discount has a disadvantage over the VDT: poor households that for some reason are excluded from the mean-tested discount (for example, because they do not qualify for the Oportunidades program) would receive no discount on their electric- ity bill. With a VDT, by contrast, households have the possibility of obtaining a discount by lowering their consumption. This provides a safety net for all households. On the other hand, the means-tested discount would have the most dramatic impact on the sub- sidy budget, because it effectively excludes non-poor households from subsidies and because the subsidy to the targeted poor can be limited to the first 100 kWh or 200 kWh of electricity consumed. Changes in the Tariff Structure That Do Not Eliminate Subsidies for Most Customers Will Not Substantially Improve Distributional Incidence When theVDT and the means-tested discount scenarios are not combined with cost reduc- tions, they leave in place a subsidy of 20 percent per kWh for untargeted customers. The more these untargeted customers consume, the larger their total subsidy.While the subsidy per kWh is higher for low-volume consumers (VDT) or poor households (means test), the largest total subsidy still goes to the largest-volume consumers. As a result, while there is some improvement over the current tariff system, the overall distribution of subsidies remains regressive even with rather significant changes in the tariff structure. To realize major improvements in the progressivity of subsidy distribution, it is critical to eliminate or significantly reduce subsidies for the majority of consumers. Reducing Subsidies Would Entail Significant Increases in Customer Bills, Especially for Those in the Middle-income and Consumption Deciles For the subsidy options assessed, the largest increases in tariffs are found among those in the middle-income or consumption deciles. For the poorest, smaller absolute increases are still significant compared to household income. Any transition strategy toward an improved tariff and subsidy model needs to address how to mitigate the socioeconomic impact of rising utility charges for various strata of the population. Reducing supply costs, or reducing consumer bills through end-use effi- ciency improvements, would help to limit the increases in bills that would be needed to significantly reduce the magnitude of subsidies to residential customers. Other options include using some of the funds saved on consumption-based subsidies to compensate middle-class customers through tax rebates in the medium term, to increase transfers provided to the poor through programs like Oportunidades or geographically targeted programs (Ministry of Social Development, Secretaría de Desarrollo Social, SEDESOL micro-regiones), or through increased transfers to the most affected states (such as those with warm summer climates). Executive Summary xv Reducing Costs Can Offset Part of the Need for Raising Tariffs and Any Cost Reduction Would Result in a Greater Proportion of Subsidies Going to Low-income Households Improving supply efficiency and reducing costs would have a slightly progressive effect on subsidy distribution since total costs (and therefore also the absolute impact of cost reduc- tions) are greatest among the largest-volume consumers.Low-income and low-consumption households would see their subsidy shares rise slightly, while the share going to the rich and large-volume consumers would decline. Cost reductions would also have a greater impact on reducing the total volume of subsidies compared to minor changes in the existing tariff structure. The simulation with ENIGH data indicates that a reduction in average costs of 10 or 20 percent would permit aggregate residential subsidies to decline by 21 to 42 percent, respectively. Similar results are obtained using a consumption scalar with CFE data. Better Information About Electricity Consumption Among the Poor is Needed This study used two different sources of information to assess the performance of alter- native tariff and subsidy instruments--CFE billing data and ENIGH household survey data. Neither source is ideal for examining the incidence of subsidies or the impact of tar- iff reforms on the poor. Having improved information about electricity consumption among the poor would be very useful for calibrating the design of any alternative subsidy program. Mexico's Electricity Subsidies in a Broader Social and Economic Context The current electricity subsidy system in Mexico delivers the bulk of subsidy benefits to the middle and upper-middle classes and to households consuming relatively large amounts of electricity. One cannot argue that generalized electricity subsidies like these are justified by positive externalities associated with electricity consumption. To the con- trary, electricity subsidies inflate demand, raise systemwide costs, contribute to the pro- duction of greenhouse gas emissions, and stimulate environmentally unfriendly behaviors, such as overpumping of groundwater. Nor is the "merit good" argument for subsidizing electricity--the idea that a good should be subsidized because of its importance to well- being--very strong compared to other competing sectors (such as water and sanitation, for example). Given the pronounced regressivity of the current system of residential subsidies in Mexico, one cannot justify the existing subsidies as a pro-poor policy intervention. Since the late 1990s, Mexico's social policy has moved increasingly toward the development of targeted, means-tested social programs, but the electricity sector has yet to adopt this prin- ciple. It is clear from the simulations presented here that it would be possible to improve the targeting of electricity subsidies to the poor through significant changes in tariff levels and structure, and even more so if combined with reductions in the unit cost of service. Using means testing based on the Oportunidades program in any new tariff or subsidy structure would seem to be a logical extension of developments that have been made in other sectors in Mexico. xvi Executive Summary It will be important to consider, however, whether the goal of subsidizing electricity for the poor is to keep the price of electricity low for the poor (as prices for other customers rise) or whether electricity is to be used as a more general social policy vehicle for augment- ing the purchasing power of the poor. If the latter is the policy goal, then it is important to recognize that electricity service is a suboptimal choice of subsidy delivery instrument. First, some poor households (less than 4 percent of the overall Mexican population) do not have electricity connections and are thus excluded from any subsidy delivered through electric- ity tariffs. Second, electricity service does not have all the characteristics of a good targeting mechanism--it is an undifferentiated good that is used by non-poor households as well as the poor. Third, because there is not a perfect correlation between electricity consumption and income, it is difficult to use electricity bills to identify the poor and deliver subsidies to them. Straight quantity-targeted subsidies delivered through the tariff structure will not be well targeted. Finally, Mexico has many other social programs that are better targeted to the poor than electricity subsidies. If the goal is to increase transfers to the poor, it would be better to use these mechanisms for reaching the poor. The idea that other social transfer programs could do a better job of reaching the poor than electricity subsidies is supported by inter- national experience. One study of the targeting of social policy instruments (Coady, Grosh, and Hoddinott 2003) shows that social funds, cash transfer programs, and work programs are much more likely than electricity subsidies to do a good job of targeting the poor. CHAPTER 1 Introduction exican electricity subsidies are among the largest in the world in absolute terms M (Table 1.1). They are equivalent to 1 percent of gross domestic product (GDP), 0.8 percent of total federal public expenditure, and 40 percent of sector revenue (Table 1.2). Large tariff subsidies have important implications for the electricity sector and for Mexican society more generally. First, the cost of the subsidies needs to be covered in some way--if not by consumers, then by government; if not by the government, then by reduced expenditure on system expansion, service quality upgrading, and on operations and asset maintenance. Fiscal transfers to the utilities divert resources from priority social and economic programs and reduce fiscal space for financing essential investments in the electricity sector. The 2006 federal budget for investment in electricity (excluding the Programas de Inversión de Impacto Diferido en el Gasto, Pidiregas) was 39,404 million pesos, only slightly larger in nominal terms than it was in 2001. Investment in areas where the Pidiregas scheme is not applicable, such as distribution, has been constrained. Second, subsidies distort price signals, elevating demand above what would otherwise prevail if electricity was priced at marginal or average cost. Massive underpricing also reduces incentives for customers to take energy saving measures, such as replacing old equipment and appliances. Higher customer demand requires additional power supply, especially to meet periods of peak demand. Because elevated residential electricity demand tends to lower load factors for the power system, it leads to reduced efficiency in the use of generation and transmission facilities and thus a higher marginal cost of supplying all cus- tomer classes. This has potentially negative implications for economic competitiveness. Finally, subsidies do not only carry monetary costs. They also impose environmental costs. Subsidies result in additional emissions from power plants, as demand for electricity 1 2 World Bank Working Paper Table 1.1. Electricity Subsidies as a Percent of GDP and Gross National Expenditure, 2005 Electricity Subsidy 2005 Subsidy as a Subsidy as a Percent of Country name (current US$) Percent of GDP Gross National Expenditure China 6,000,000,000 0.3% 0.3% Indonesia 1,500,000,000 0.5 0.5 Mexico 8,876,000,000 1.1 0.8 India 10,000,000,000 1.2 1.2 South Africa 4,000,000,000 1.7 1.6 Saudi Arabia 5,500,000,000 1.8 2.7 Russian Federation 14,000,000,000 1.8 2.1 Iran, Islamic Rep. 3,500,000,000 1.8 2.0 Egypt, Arab Rep. 2,500,000,000 2.8 2.7 Ukraine 3,500,000,000 4.2 4.3 Source: World Bank GDF and World Development Indicators (WDI) Central Database (April 2007); IEA (2006). Table 1.2. Total Mexican Electricity Subsidies in Relation to other Scalars (2002­06) Electricity subsidies as a percent of: 2002 2003 2004 2005 2006 Total electricity revenues 44.4% 44.9 43.6 42.9 38.1 Federal investment budget for electricity 173.1% 203.6 237.2 235.1 243.8 GDP 1.0% 1.1 1.1 1.1 1.0 Total federal public expenditure 0.7% 0.8 0.8 0.8 0.7 Source: INEGI, Banco de Información Estadística; SHCP, DGYPT Subsecretaria de Ingresos; SENER "Estadísticas de Electricidad 2000­2006." grows in response to the subsidized prices. Further, standard merit order dispatch rules (plants with low capital costs, but high fuel costs, such as small fuel-oil and diesel-fired facilities) exacerbate the tendency for marginal generating plants to generate higher levels of pollutants. In rural areas, electricity subsidies for irrigation pumping cause overexploitation of groundwater reservoirs. In short, there are many reasons to be concerned about the effects of large and grow- ing electricity subsidies. There is broad agreement in Mexico that something must be done about this pattern, but also recognition that making a lasting change will be very challeng- ing. Subsidies to residential and agricultural customers are deeply entrenched, and cost recovery levels are very low (Table 1.3). Although subsidies per unit are greater in agricul- ture, the total volume of subsidies going to residential customers is much larger. Two-thirds of all electricity subsidies in Mexico are directed at residential electricity customers, and this share has increased over time. The subsidies provided to residential customers have increased by 46 percent since the last tariff reform in 2002. Time and again, efforts to reduce subsidies have been followed by the creation of new and more highly sub- sidized tariff categories to offset the burden on those residential customers adversely affected by the last tariff change. Residential Electricity Subsidies in Mexico 3 Table 1.3. Distribution of Electricity Subsidies, by Customer Class, 2006 Sales Subsidies Sector (Gigawatt hours) (2006 million pesos) Price/Cost Ratios Residential 44.5 63,369 0.41 Commercial 13.2 5,476 0.92 Public services 6.6 2,887 0.77 Agriculture 8.0 9,211 0.30 General, medium tension 65.3 12,478 0.83 General, high tension 37.9 4,213 0.97 Total 175.4 97,633 0.68 Source: Aburto 2007. Includes LFC and CFE. Sales taken from CFE web page. Subsidies and ratios compiled from SHCP data. Subsidies to LFC customs computed as LFC costs minus LFC revenues. Numerous studies undertaken during the last sexenio documented the regressive inci- dence of these subsidies to residential customers: the bulk of the subsidies go to the non- poor. Cognizant of this problem, Mexican authorities have expressed a strong interest in evaluating the potential gains and impacts of a shift from quantity-based targeting through the tariff structure to means testing, building upon the poverty profiling and verification systems established under Oportunidades, Mexico's well-functioning, means-tested con- ditional cash transfer program. The pilot effort for such a system was recently implemented with Oportunidades Energéticas, which, with its 2,999-million-pesos annual budget, is small relative to existing tariff subsidies. This small program, begun in 2006, did not replace or modify the existing tariff-based subsidies but rather operates in parallel with it. This report responds to interest on the part of the Ministry of Social Development (Secretaría de Desarrollo Social, SENER), the Ministry of Finance (Secretaría de Hacienda y Crédito Público, SHCP), the Ministry of Social Development (Secretaría de Desarrollo Social, SEDESOL), and other federal agencies in assessing the performance of alternative subsidy mechanisms for residential customers and other options to reduce the magnitude of subsidies and to improve their distributional impact. In addition to examining the impact of expanding the use of means testing in the allocation of subsidies, this study also assesses the impact of modifying electricity tariff structures to improve distributional out- comes and reduce the overall magnitude of electricity subsidies. Chapter 2 of the report describes the evolution of residential electricity subsidies and tariffs in Mexico and discusses the main factors that have contributed to the exceptional growth of subsidies in recent years. Chapter 3 provides estimates of the distributional inci- dence of the subsidies among residential customers. Chapter 4 identifies several alternatives to the present subsidy mechanism, and their effects on the total volume and distributional incidence of subsidies are estimated. Chapter 5 discusses the implications of the analysis and directions for further work to better inform public policy in this area. CHAPTER 2 Residential Electricity Subsidies and Tariffs in Mexico Defining and Measuring Subsidies The electricity subsidies reported in Mexico are financial subsidies that result from below- cost pricing, and are measured as the difference between the price of electricity paid by con- sumers and the national average cost of supply.4 Costs are measured as accounting costs as reported by the two operating utilities--the Mexico National Electricity Commission (Comisión Federal de Electricidad, CFE) and Central Light and Power (Luz y Fuerza del Cen- tro, LFC)--according to government accounting standards. The subsidies to CFE have been financed by a bookkeeping transfer. The federal government essentially reimburses CFE for providing subsidies to its customers by discounting the taxes and dividends (aprovechamiento5) that CFE would otherwise have to pay the government. Since 2002, the volume of subsidies has exceeded the notional amount of aprovechamiento and has thus begun to erode CFE's capital base (World Bank 2005). In the case of LFC, the country's second electricity provider covering the Valle de Mexico and surrounding areas, the gov- ernment provides the company with a direct cash subsidy to cover its mounting operating deficits and customer subsidies. It is important to recognize that this accounting approach to measuring subsidies ignores the fact that the economic cost of electricity provision may be significantly higher 4. There are regional differences in costs across Mexico, related to fuel mix and technology in gener- ation and grid density and transmission voltages in transmission. These cost differences are not taken into account in subsidy calculation. Cost differences are lower now than in the past due to the well-developed grid. The regions with highest average cost are likely southern Baja California and southern Mexico, but this is difficult to verify because CFE no longer publishes regional cost data. 5. Equal to 9 percent of net fixed assets. 5 6 World Bank Working Paper than the accounting cost. Moreover, although households benefit from financial subsidies in the form of lower bills, they may not capture the full value of the subsidy. When account- ing costs are unnecessarily high due to supply inefficiencies, utilities, their employees, and suppliers capture part of the subsidy benefit. This is currently not taken into account in estimating the value of electricity subsidies received by residential electricity customers in Mexico--there is an implicit assumption that electricity customers are the sole beneficia- ries of the subsidies. Legal Framework for Tariffs and Subsidies The legal framework governing electricity tariffs in Mexico establishes that tariffs should cover costs and promote efficient consumption. The tariffs of state-owned enterprises are to be set "in accordance with economic efficiency and financial health criteria" (Article 26, Reglamento de la Ley de Empresas Estatales). Electricity tariffs in particular will be "fixed, adjusted, and restructured in a way that tends to cover financial and public service expan- sion requirements, and rational energy consumption" (Article 31, Ley de Servicio Público de Energía Eléctrica). Current residential electricity tariffs are far from meeting these objec- tives. Because only a small proportion of residential customers pay the marginal cost of electricity service, current tariffs achieve neither the cost recovery nor the efficient con- sumption provisions of the current legislation. With respect to subsidies, federal law establishes that subsidies "shall be temporary" and shall be subject to criteria of "objectivity, equity, transparency, publicity, and selectiv- ity." Moreover, "in programs of direct benefit to individuals or social groups, amounts and percentages [of the subsidy] shall be determined on the basis of redistributive criteria that must privilege lower-income population and aim at equity amongst regions and states" (Article 75, Ley Federal de Presupuesto y Responsabilidad Hacendaria). In contrast, numer- ous reports produced during the last sexenio documented that residential electricity sub- sidies are highly regressive, and the analysis presented below confirms that the tariff revision of 2002 did not redress this. A Brief History of Tariffs and Subsidies Explicit subsidies for residential customers were introduced in 1973 when the tariff was changed into a three-block, increasing block tariff (IBT), with subsidized prices for the first two blocks.6 IBTs provide "quantity-targeted" subsidies, meaning that the discount a cus- tomer receives depends on the quantity consumed. IBTs are the most common tariff struc- ture among electricity utilities in developing countries. One study found that utilities in nearly 70 percent of countries applied IBTs, usually with three to four consumption blocks, and a first block of between 50 kWh and 100 kWh per month (Komives and others 2005). In addition to quantity-targeted electricity subsidies, Mexico also offers "summer subsidies"--extra discounts to residential customers living in warm areas, where the demand for air-conditioning is high. The first "summer subsidy" (Tariff 1A) was introduced in 1974 for customers living in areas where temperatures exceeded 25 degrees Celsius (°C) for four months of the year. Successive climate-based tariffs with increasingly subsidized rates over 6. See Appendix A for a detailed history of residential tariffs over the last four decades. Residential Electricity Subsidies in Mexico 7 larger volumes of consumption were introduced in 1988 (Tariff 1B > 28 °C; 1C > 30 °C), 1990 (1D > 31 °C), 1995 (1E > 32 °C), and 2002 (1F > 33 °C). In 2002, an attempt was made to reduce residential electricity subsidies. Three billing schedules were introduced in each tariff zone--low, moderate, and high. The highest- volume consumers (roughly the top 5 percent of customers in each tariff zone before the tariff change) were charged a High Consumption Residential Tariff (Tarifa Doméstica de Alto Consumo, DAC), which included a fixed fee plus a uniform volumetric consumption charge, set slightly above estimated long-run marginal cost. For customers in the moder- ate billing schedule, prices in the top IBT blocks were increased. Even following the 2002 tariff review, the entire volume of actual consumption was still subsidized for 75 percent of customers. Another 20 percent of customers were to receive subsidies for the majority of their consumption. Only the 5 percent of residential customers paying DAC tariffs were expected to cover the costs of the service they received. The 2002 tariff reform measures did not have a lasting impact on residential subsidy volumes. As Figure 2.1 shows, subsidies to residential customers dropped slightly in real terms in 2002 but by 2003 they had risen above 2001 levels. Part of the reason for this was the introduction in 2002 of a yet larger summer subsidy--Tariff 1F. Since 2003, residential subsidies have continued to rise. The following Chapter discusses the factors that have con- tributed to the persistent increase in residential subsidies. The accretion of modifications to the tariff structure introduced since the 1970s (see Appendix A) has created one of the most complex tariff systems in the world, with over 112 different billing possibilities for residential customers (see Appendix B). This is certainly more complex than the tariffs applied in other Latin American countries. For example, Peru, Figure 2.1. Subsidies to Residential Customers (2006 constant billion pesos) 70 60 pesos 50 of 40 billiones 30 2006 20 Constant 10 0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Source: Aburto 2007, based on data from SHCP. 8 World Bank Working Paper where electricity provision Table 2.1. Average Residential Electricity Prices (2006) outside of Lima is supplied by state-owned companies, Country US$ / kWh has only 1 residential tariff Republic of Korea 0.089 with three different billing Mexico, all Residential Tariffs (excluding DAC) 0.090** options according to con- United States 0.096 sumption level. Uruguay has Colombia 0.098 one national utility provid- Chile 0.109 ing electricity and only two Mexico, all Residential Tariffs 0.101** residential tariffs. In other Uruguay 0.117 Latin American countries that Turkey 0.119 have more than one electric- France 0.137 ity provider, each of the elec- Brazil 0.143 tricity utilities has between Spain 0.154* 1 and 3 residential tariffs United Kingdom 0.158 (CIER 2005). Japan 0.196* Compared to other Germany 0.198* Organisation for Economic Co-operation and Develop- *Last available data. **2006 average. ment (OECD) and middle- Source: IEA, Key World Energy Statistics 2006; World Bank income countries in Latin Electricity Benchmarking Database for Latin America and the America, average residential Caribbean; Mexico, 1 and DAC tariffs, computed from data electricity tariffs in Mexico on CFE's web page. Data are for first quarter and include are very low (Table 2.1).7 taxes (except in the United States). Among Latin American countries, Mexico's tariffs sit in the bottom third of the distribution. Among OECD countries, even the highest average tariff paid by consumers in Mexico is among the lowest. Growth of Residential Electricity Subsidies The factors responsible for the growth of residential electricity subsidies in recent years are described below. Summer Subsidies and Tariff Reclassifications One of the main factors driving the growth of residential subsidies has been the subsidized tariffs for warm areas--the summer subsidies. Each new tariff category has provided greater discounts over larger volumes of consumption for customers in a particular cli- matic zone. These new lower tariffs to warm areas are not justified based on lower aver- age costs in warmer regions. The goal rather is simply to extend greater subsidies to households in areas where air-conditioning and fans are common household fixtures. The difference between the subsidies provided through Tariff 1 and the most recent summer subsidy tariff (Tariff 1F) is particularly noteworthy (Table 2.2). Each month, a 7. A 15 percent VAT is added to the electricity bills in Mexico, but this is not included in the average tariff figures listed in comparison tables or in the analysis of tariff and subsidies in Mexico. Residential Electricity Subsidies in Mexico 9 customer in the Tariff 1F zone can consume 10 times the amount of electricity as a customer in the Tariff 1 zone (an additional 2,250 kWhs per month) before paying the unsubsidized DAC tariff. According to CFE estimates, this is equivalent to the energy needed to run three large air coolers (aire lavado), or two new 1.5-ton window air-conditioning units 24 hours a day. Households that run their air-conditioners only at night or that have smaller units could run two to three times this many air-conditioners with their extra share of subsidized electricity. The subsidy per kWh provided to customers in Tariff 1F is also much larger than that pro- vided to customers of Tariff 1, even at the same level of consumption. For the first 75 kWhs consumed, those in Tariff 1 pay one-third more per kWh than those in Tariff 1F. For the next 75 kWh, customers in Tariff 1 pay two-thirds more than those in Tariff 1F. Table 2.2. Comparison of Charges: Tariff 1 and Tariff 1F, 2006 (Mexican pesos) Billing Schedule Tariff 1 Summer Tariff 1F Summer Low Range: Up to 140 kWh per month Up to 1200 kWh per month Minimum charge 15.88 Minimum charge 13.88 0­75: 0.635 / kWh 0­300: 0.448 / kWh >75: 0.749 / kWh >175: 0.586 / kWh Moderate Range: 140­250 kWh per month 1200­2500 kWh per month 0­75: 0.635 / kWh 0­300: 0.448 / kWh 75­125: 1.043 / kWh 300­900: 0.743 / kWh >125: 2.213 / kWh >900: 1.39 / kWh High Range: More than 250 kWh: More than 2500 kWh: All consumption: DAC All consumption: DAC The introduction of Tariff 1F resulted in a loss of revenue from all customers located in this tariff category. But from 2002 to 2006, there were also many other reclassifications of cus- tomers to more subsidized tariffs. Requests for reclassification are ostensibly based on new climatic information but do not appear to have been subject to independent verification. Appendix C reports the number of customers reclassified each year since the 2002 tar- iff revision, plus the static cost of that reclassification (in the year of reclassification only). Tables 2.2 and 2.3 summarize this information. The number of reclassifications was espe- cially high (nearly 3 million customers reclassified) during the period immediately following the 2002 Tariff Review. The bulk of the reclassifications were from tariff 1E to the new 1F and from 1B to 1C. Most reclassified communities moved to the next, more highly subsi- dized tariff category (for example, from 1 to 1A or from 1D to 1E), but some communities managed to jump two or three tariff levels (for example, from 1B to 1D). The estimation of the cost of these reclassifications (as reported in Table 2.2 and Appendix C) are single-year estimates (that is, in the year of reclassification). Of course, reclassifications represent a permanent loss of revenue to the electricity utilities. Although the pace of reclassifications has slowed in recent years, the cost of these reclassifications will grow exponentially over time as a greater percentage of total consumption is billed out at more subsidized rates. 10 World Bank Working Paper Table 2.3. Number of Customers Reclassified and the Cost of Reclassification Cost in additional Subsidies, in Year Customers Reclassified Year of Reclassification Only (million pesos) 2002 1,309,347 524.9 2003 1,648,626 374.4 2004 209981 135.2 2005 243913 66.7 2006 71255 34.5 Source: Aburto, based on data from CFE and Hacienda. Table 2.4. Number of Residential Customers Reclassified between Tariff Categories, 2002­06 2002 2003 2004 2005 2006 Total 1 to 1A 20,577 16,219 77,452 23,775 138,023 1A to 1B 33,991 42,558 2485 25,958 104,992 1B to 1C 639,560 876,783 5990 103,188 40,605 1,666,126 1C to 1D 11,076 275,134 25,084 16,340 421 328,055 1D to 1E 34,882 55,846 161,740 3,246 5,814 261,528 1E to 1F 564,208 14,275 578,483 Note: The table includes most common reclassifications only. The full list of reclassifications is reported in Appendix C. Source: Aburto, based on data from CFE and Hacienda. Table 2.4 illustrates this compounding effect for the years 2002­06. While total elec- tricity sales to residential customers have gone up by 14 percent, the increase has not been evenly distributed across tariff zones (Figure 2.2). Sales in Tariff 1 remained stable dur- ing this period, and sales in 1A and 1B have gone down slightly. Sales in the more sub- sidized 1C, 1D, 1E, and 1F all increased. By 2006, 9 percent of all residential electricity sales were to customers in Tariff 1F, which comprises 3 percent of residential customers. In comparison, the unsubsidized DAC tariff accounted for only 8.4 percent of all elec- tricity sales, and for only 2 percent of customers. After 2002, many of the 5 percent of cus- tomers to whom the DAC tariff was originally applied managed to move into another tariff category, either as a result of being reclassified or by reducing their consumption below the DAC threshold. Given this shift in energy sales to more subsidized tariff categories, it is not surprising that the average price charged to CFE residential customers has risen more slowly than the average price charged to LFC customers (where no reclassification is possible because all customers fall under Tariff 1).8 If one attributes this increasing gap in average tariff to the reclassifications, then the cumulative cost of the reclassifications between 2002 and 2006 8. In 2000 and 2001, LFC and CFE average tariffs were nearly identical. Residential Electricity Subsidies in Mexico 11 Figure 2.2. Electricity Sales by Tariff Category, in 2002 and 2006 0.45 0.40 0.35 0.30 sales 0.25 2002 2006 energy 0.20 of %0.15 0.10 0.05 0.00 DAC 1 1A 1B 1C 1D 1E 1F Tariff category Source: Based on Aburto 2007, with CFE statistics. appears to be around 15 billion pesos (Aburto 2007). To put this figure in perspective, this is about 23 percent of the total value of residential electricity subsidies in 2006 and 38 per- cent of the 2006 federal investment budget for electricity. Accounting Costs and Inefficiency At the same time that CFE has lost potential revenue through tariff reclassifications, costs at CFE and LFC have been rising. Since 2002, CFE's average cost per Megawatt-hour (MWh) provided to all customer groups has increased by 20 percent in real terms (Table 2.5). The main factors driving the increase in costs were power purchases and fuel costs, employee retirement payments, and wages and benefits. While the proportion of mainte- nance and depreciation to unit costs, and taxes and financing costs to unit costs, fell between 2002 and 2006 in real terms, power purchases and fuel costs increased by 70 per- cent over this five-year period. This was due at least in part to high gas prices in recent years and to the policy (since 1994) of expanding generating capacity through natural-gas-fired combined-cycle plants. Increased purchases of power from independent private providers may also be driving this increase, in which case one would have expected to see this offset by cost reductions in CFE operations. Instead, the contributions to retirement payments and to wages and benefits at CFE increased by 35 percent and 24 percent, respectively, between 2002 and 2006. LFC costs per MWh are roughly double those of CFE. Some of this difference is attrib- utable to important structural and functional differences between CFE and LFC--the 12 World Bank Working Paper Table 2.5. CFE Accounting Costs per MWh, 2002­06 (2006 pesos) Percent Change, Cost Concept 2002 2003 2004 2005 2006 2002­06 Wages and benefits 110 118 135 133 136 24% Retirement payments 86 99 96 104 116 35 Power purchases and fuels 350 514 535 612 595 70 Maintenance and depreciation 249 265 220 206 208 -16 Taxes and financial cost 327 350 314 294 304 -7 Total cost 1,121 1,345 1,300 1,350 1,359 24 Source: Prepared by Petrelec from CFE sources. latter is essentially a distribution utility while the former is a vertically integrated electricity utility handling generation, transmission, and distribution. Moreover, LFC's customer mix is heavily weighted toward low-tension customers, because there is little industry in its ser- vice area. Nonetheless, even given these differences, LFC has much higher personnel costs. The contribution of wages and benefits and of retirement payments to unit costs in LFC is nearly four times greater than in CFE. Overall, LFC's unit costs have seen a 45 percent increase in real terms over the last five years (Table 2.6). Table 2.6. LFC Accounting Costs per MWh, from 2002 to 2006 (2006 pesos) Percent Change, Cost Concept 2002 2003 2004 2005 2006 2002­06 Wages and benefits 452 487 522 540 579 28% Retirement payments 634 461 554 494 548 -14 Power purchases and fuels 808 1,008 1,407 1,337 1,682 108 Maintenance and depreciation 142 149 151 152 147 3 Taxes and financial cost 1.21 1.59 1.20 1.09 1.04 -14 Total cost 2,038 2,107 2,635 2,524 2,956 45 Source: Aburto 2007 from CFE sources. Examining trends in unit costs is one way of assessing efficiency. Another indicator of efficiency is statistics on transmission and distribution losses. Transmission losses as a per- centage of received energy have been declining in percentage terms in both CFE and LFC as a result of an ambitious CFE investment program, financed through the Pidiregas scheme. Reported losses are now under 2 percent for CFE, which is an internationally com- petitive standard. They are under 3 percent for LFC, the level achieved by CFE only in 2002. However, distribution losses in both companies have been increasing and are extremely high in the case of LFC. CFE's distribution losses rose from 11 percent in 2000 to 11.6 per- cent in 2005. As a point of comparison, a good international standard for distribution losses, for a utility with CFE's load and geographic characteristics, would be around 8 per- Residential Electricity Subsidies in Mexico 13 cent. LFC's distribution losses have also been increasing and are very high, exceeding 30 per- cent since 2005. Increasing real unit costs and high and increasing distribution losses are both bad for electricity subsidy budgets. Unless these are matched by real increases in average tariffs, the increasing electricity supply costs lead to increasingly higher subsidies. The creation of summer subsidy programs and the reclassification of customers to more subsidized tariffs work against efforts to bring prices in line with costs, achieve financial sustainability, and manage the federation's fiscal exposure in the sector. The fact that virtually all residential customers of both utilities are subsidized also means that the 3.56 million new customers that were connected between 2002 and 2006 were an added financial burden to the utili- ties and also increased the overall subsidy budget. CHAPTER 3 Distributional Incidence of Residential Electricity Subsidies he failure to halt the growth in subsidies, despite the intentions of the 2002 tariff T review, is troubling and reason to consider serious changes in pricing and subsidy policies for electricity. Yet, the public policy concern is not just the magnitude of subsidies but also their distributional incidence: Who benefits from these subsidies and how are they distributed across different income groups? Areelectricitysubsidiesdeliveredthroughatariffsystemdominatedbyincreasing block tariffs and "summer subsidies" a good way to subsidize the poor and to keep their electricity expenditures low? To address these questions, it is necessary to evaluate the distributional incidence of resi- dential electricity subsidies. Usually, "distributional incidence" refers to the distribution of subsidies by income level--how pro-poor is the subsidy expenditure? However, evaluating the distributional incidence of electricity subsidies is complicated by the fact that utility databases--a good source of information about electricity consumption--do not contain information about the income level of residential customers: it is not possible to identify which customers are poor and which are not. There is generally a strong correlation between income and electricity use because wealthier households have larger homes with more electricity-consuming appliances. But this does not mean that all of the lowest-volume electricity consumers are poor, or that all poor households consume little electricity. The vacation homes of rich households, for example, may have low consumption levels during much of the year. The same is true for the apartments of single professionals who work long hours away from the home. Likewise, 15 16 World Bank Working Paper consumption by poor households can be higher than expected in some cases, because they have old, inefficient appliances, because they have electricity-using home-based businesses, or because multiple poor households share a single electricity connection. In Mexico, with its climatic variations by region, it is also common that poor households in hot areas con- sume more electricity than non-poor households in more temperate zones. For all these reasons, a strict relationship between income and electricity consumption cannot be assumed. With this in mind, two kinds of data are used to assess the distributional incidence of residential electricity subsidies. The resulting estimates are not directly comparable, but rather can be legitimately used to answer specific questions: CFEdataonthedistributionofcostsandelectricityconsumptionwithineachtar- iff class are used to examine the distribution of subsidies by consumption level. What percentage of subsidies goes to lower-volume consumers? Information on electricity expenditure from the Mexico National Household Income and Expenditure Survey (Encuesta Nacional de Ingresos y Gastos de los Hogares, ENIGH) household survey is used to estimate the distribution of subsidies by income quintile. What percentage of the subsidies goes to the poor? And how does the distributional incidence of electricity subsidies compare to that of other government expenditures? Appendixes 4 and 5 describe the methodologies used with each data source. Do Subsidies Favor Low-volume Consumers? Figure 3.1 presents the distribution of electricity consumption by consumption decile in 2005. The consumption per connection ranges from an average of only 5 kWh per connec- tion per month in the first consumption decile to 80 times more (404 kWh per connection per month) in the highest consumption decile. Four hundred kWhs is typical of the con- sumption level in temperate regions of developed countries. Five kWh per month is an extremely low level of consumption, even for very poor households. According to CFE esti- mates, this is equivalent to using one 60-watt lightbulb for about 2.5 hours per day. It is likely that this first decile includes many vacation homes with irregular consumption levels, con- nections with malfunctioning meters, and inactive accounts. Average consumption in the second decile--38 kWh per household per month--is closer to what one might expect in a poor rural home in Mexico.9 Consumption levels vary dramatically by seasonal tariff zones (Figure 3.1). As one would expect, consumption is much higher among customers paying lower tariffs in the more highly subsidized geographic areas. Average consumption in the 5th decile of Tariff 1 is just 97 kWh, whereas average consumption in the same decile for consumers in Tariff 1F is 277 kWh. The difference between consumption levels in decile 10 is even 9. According to ENIGH 2005, the average household in the poorest decile of rural households in Mex- ico has two lightbulbs and a television set. Assuming that these households use these appliances at 75 per- cent of the average usage level in Mexico, expected consumption would be 38 kWh per month. Residential Electricity Subsidies in Mexico 17 Figure 3.1. Monthly Electricity Consumption, by Tariff Category and Consumption Decile Consumption by tariff Category 5000 4500 4000 3500 1240 1F 3000 1E 895 1D Month/ 2500 1C 1B kWh 677 1A 2000 667 1 1500 511 463 420 588 348 365 1000 333 277 284 249 266 419 235 202 216 286 184 500 176 161 229 151 190 125 126 164 209 334 107 137 171 145 75 94 112 101 121 166 56 86 80 116 139 52 56 100 65 87 171 270 017 13 8943 31 27 43 141 30 50 66 91 105 113 1 2 3 4 5 6 7 8 9 10 Decile Source: Aburto 2007 using 2005 CFE data. larger: the largest volume consumers in Tariff 1 use 270 kWh per month on average, com- pared to 1240 in Tariff 1F. To calculate the subsidies accruing to each electricity connection, electricity bills were compared to the cost of service received. A two-part cost function (fixed charge per connec- tion plus variable cost per kWh) is derived from CFE data (see Appendix D). CFE costs are used as the standard cost estimate for all customers to help approximate the value of the subsidy that households receive, rather than the portion of the subsidy that accrues to a utility with higher costs (LFC in this case). Table 3.1 shows the distribution of residential electricity subsidies by consumption decile. The distribution of subsidies is heavily skewed toward above-average users of electricity. The subsidy received per kWh is highest in the lowest consumption deciles, but the total value of the subsidy received per connection per month is smallest in these low- consumption deciles. The first five consumption deciles receive only 25 percent of the total volume of subsidies. The average monthly subsidy per connection increases through decile 8, then declines somewhat in deciles 9 and 10. These top two deciles receive 14.2 and 10 percent of the sub- sidies, respectively. The slightly lower proportion of subsidies in the highest deciles is a result of reforms instituted in the 2002 tariff review (the introduction of the DAC tariff and the intermediate consumption schedule). One study (Aburto 2003), conducted just after the tariff reform and using a similar methodology, found that the top decile of consumers received only 4.4 percent 18 World Bank Working Paper Table 3.1. Distribution of Residential Electricity Subsidies by Consumption Decile Average Total Average Average Consumption Subsidy Subsidy Subsidy Subsidy Cumulative kWh/ Mpesos/ Pesos/ Pesos/ Distribution Subsidy Decile HH-month yr kWh HH-year % % 1 5 1,402 9.03 550 2.6 2.6 2 38 2,980 2.56 1,169 5.6 8.2 3 64 4,003 2.05 1,571 7.5 15.8 4 85 4,801 1.85 1,884 9.0 24.8 5 111 5,737 1.70 2,251 10.8 35.6 6 131 6,463 1.62 2,536 12.2 47.7 7 148 6,981 1.55 2,739 13.1 60.9 8 181 7,961 1.44 3,124 15.0 75.8 9 224 7,529 1.10 2,954 14.2 90.0 10 404 5,314 0.43 2,085 10.0 100 Source: Aburto 2007 using 2005 CFE data. of residential electricity subsidies. Their share of the total subsidy appears to have doubled since then, suggesting that large-volume consumers have gained back some of the subsidy share they lost in 2002. In the medium term, therefore, the 2002 tariff revision did little to either shift subsidies to low-volume consumers or to restrain the growth of the overall vol- ume of subsidies. The very large seasonal subsidies are a major factor contributing to the highly regressive distribution of the subsidies. From Figure 3.2 it is clear that the more recent the tariff, the more regressive the distribution. The distribution of subsidies across consumption deciles becomes increasingly regressive going from Tariff 1 to Tariff 1F. Among households subject to Tariff 1F, the most recently created tariff category, the highest consumption decile receives an astounding 20 percent of total residential subsidies. Importantly, the subsidies provided under all tariff categories (even Tariff 1, which does not have a special summer subsidy) are skewed toward higher-volume consumers. The com- mon element in each tariff is the use of an increasing block tariff (IBT) structure for two of the three consumption schedules. Only those in the top consumption schedule pay a DAC tariff, which is a flat volumetric tariff plus a fixed charge. The other two groups are billed with an IBT of either two or three blocks, with a minimum charge. It is widely believed that IBTs are an effective way of targeting subsidies to the poor and to low-volume consumers. However, a global review of the distributional conse- quences of delivering subsidies through IBTs found that they are almost universally regressive and usually provide larger subsidies to large-volume and high-income con- sumers (Komives and others 2005). The experiences in Mexico are consistent with this finding. IBTs do deliver the largest subsidy per kWh to the smallest-volume consumers (pro- vided there is no fixed charge). However, the total subsidy per month is higher for larger- volume consumers because they consume so many more subsidized kWhs--a smaller Residential Electricity Subsidies in Mexico 19 Figure 3.2. Distribution of Electricity Subsidies, by Tariff Zone and Consumption Decile, 2005 Distribution of Electricity Subsidies by Tariff Zone & Consumption Decile (2005) 25.00 20.00 decile 15.00 by arehs 10.00 Subsidy 5.00 0.00 10 9 1F 8 1E 7 Tariff 1D 6 5 Decile Category 1C 4 1B 3 1A 2 1 1 1 1A 1B 1C 1D 1E 1F Source: Aburto 2007 using 2005 data from CFE. subsidy per unit over more units of consumption translates into a larger total subsidy per year. IBTs become even more regressive when a fixed charge is added to the tariff because the lowest-volume consumers then pay the highest price per kWh. One of the reasons that IBTs produce such a regressive distribution of subsidy ben- efits is that almost all of the consumption of large-volume users remains subsidized. This is the case in Mexico, where very few residential customers pay an unsubsidized unit price for their electricity consumption. In theory, one could improve the distrib- utional incidence of an IBT by raising the last block of an IBT to a price higher than average cost, such that large-volume consumers "pay back" the subsidy they receive on their first units of consumption. In practice, however, this rarely happens because there is much pressure to retain large blocks with highly subsidized consumption. The recent "summer subsidy" tariffs in Mexico are examples of IBTs with extremely large subsi- dized consumption blocks. Even in other tariff zones, however, the IBT blocks are large enough that most consumption is subsidized. In Mexico during the summer months only 2 percent of consumers in Tariff 1, 1 percent of those in Tariff 1C, and even fewer customers in the other tariff categories end up paying more than what it costs to serve them (Aburto 2007). 20 World Bank Working Paper In sum, Mexico's existing residential tariff structures disproportionately benefit larger- volume consumers, both because of the summer subsidies but also because of the use of IBT structures for most customers. The modifications instituted through the 2002 tariff review (the DAC and the higher price in the third block of the IBT for the intermediate consumption schedule) were insufficient to make any significant difference in the distrib- utional incidence of the subsidies. Subsidies for the Poor? The above analysis makes it clear that residential electricity subsidies in Mexico dispro- portionately benefit large-volume consumers and those living in warm regions. This Chap- ter assesses whether the current subsidy system delivers the majority of benefits to the non-poor. Assessing the distributional incidence of subsidies across income rather than consumption deciles also permits comparison of the benefit incidence of electricity subsi- dies to that of other public spending programs in Mexico. For this analysis, the ENIGH 2006 household survey was the primary source of information. The methodology used is described in Appendix E. ENIGH does not collect information on electricity consumption but rather on house- hold expenditures for electricity. To estimate the volume of subsidies received by each income decile, it was first necessary to infer the electricity consumption of each house- hold. Two approaches were used: (a) using reported household electricity expenditure and the applicable tariff schedule, and (b) using information about appliance ownership.10 Both estimates suggest that the electricity consumption curve is much flatter across income deciles than across consumption deciles (Figure 3.3).11 When electricity con- sumption is estimated using reported expenditure, there is nearly a threefold difference between the poorest decile (139 kWh per month) and the richest (380 kWh per month). Estimates based on appliance ownership suggest that the rich consume four times more electricity than the poor. Both approaches generate similar estimates of electricity con- sumption among the poorest groups. As one would expect, at all income levels, con- sumption is higher among households subject to the more subsidized tariffs--those living in warmer regions (Table 3.2). 10. ENIGH collects information on electricity-using appliances, from lights to air-conditioning. Applying CFE's estimates of average use of these appliances, we were able to estimate electricity con- sumption in each household. 11. There are several reasons why average consumption derived from household expenditure surveys is higher than average consumption levels derived from CFE histograms. First, estimates using ENIGH data only include households, not vacation homes, inactive contracts, and so forth. Second, the reported results are only for households with non-zero expenditure on electricity. Thus, those with nonfunction- ing meters or illegal connections are not included. Third, the ENIGH sample is not designed to be repre- sentative of the different tariff zones. Fourth, households may mis-report their electricity expenditure (for example, by reporting bimonthly rather than monthly consumption or systematically understating or over- stating expenditure). It is worth noting, however, that average reported expenditure among the poor in the ENIGH survey is very similar to average reported expenditure in a recent Oportunidades Energéticas survey. Moreover, the consumption estimates based on appliance ownership are consistent with reported expenditure levels, at least among the poor. Residential Electricity Subsidies in Mexico 21 Figure 3.3. Estimated Electricity Consumption (kWh per month) by Income Decile Using Two Different Methodologies 600 500 400 kWh 300 estimated200 100 0 1 2 3 4 5 6 7 8 9 10 household deciles kWh from spending kWh from appliances Source: Scott 2007 using ENIGH data. Table 3.2. Estimated Electricity Consumption (kWh per Month), by Income Decile and Tariff Zone Income Tariff Zone Decile Average 1 1A 1B 1C 1D 1E 1F 1 146 115 124 165 222 297 538 345 2 189 129 176 196 309 415 500 804 3 210 146 190 187 360 419 508 640 4 212 143 212 200 329 426 516 545 5 244 159 229 204 372 416 551 791 6 248 167 209 214 380 506 553 806 7 273 167 269 245 411 516 627 1023 8 277 168 245 237 429 606 704 807 9 336 189 262 246 489 581 765 1209 10 391 229 284 349 613 641 901 1405 Average 230 163 221 223 423 506 629 951 Source: Scott 2007 using ENIGH data to estimate consumption based on reported expenditure. Consumption among households reporting non-zero electricity expenditure. 22 World Bank Working Paper Table 3.3 presents the distribution of electricity subsidies by income decile. It com- pares the distributional incidence of subsidies provided through the tariff structure to the subsidies provided through the Oportunidades Energéticas program. The subsidies deliv- ered through the tariff structure are regressive, with the poorest 40 percent of households capturing only about 30 percent of the subsidies. This is significantly less than they would receive if subsidies were randomly distributed to all utility customers. In contrast, the rich- est 40 percent of households receive 50 percent of the subsidies. As expected, the subsidized summer tariff schedules are more regressive than Tariff 1. The poorest 10 percent of households in Tariff 1 receive 9.5 percent of the subsidies deliv- ered to households in that tariff group, whereas the poorest 10 percent of households in Tariff 1F receive less than 1 percent of the subsidies given to this customer group. The last row of Table 3.3 presents a measure of the relative regressivity of the tariffs delivered to each tariff class. The measure divides the percentage of subsidies delivered to households in the first four deciles of each tariff by 40, which is the percentage of subsidies that the poorest four deciles would receive with a universal or random distribution of subsidies. When the result is greater than 1, the subsidy is progressive and the poor receive more than they would with a random distribution of subsidies. Only the means-tested Oportunidades Energéticas subsidy accomplishes this goal: the poorest 40 percent of households receive over 85 percent of these subsidies. When the result is less than 1, the subsidy is regressive, and the poor do worse than they would with random distribution. Tariffs 1­1F are all regressive, but Tariffs 1C, 1E, and 1F perform particularly poorly. In these tariff groups, the poorest 40 percent of households receive about half what they would if subsidies were distributed randomly or equally to all households. Tariffs 1­1F each consist of a series of IBTs plus a flat volumetric DAC tariff for the highest-volume consumers. In general, IBTs perform poorly in distributional terms. Table 3.4 examines the targeting performance of IBTs in other countries, using the measure of regressivity presented in Table 3.3. The best performance among the examples listed in the table is in Hungary, where there is universal access to electricity and little difference in con- sumption between poor and non-poor. The poorest performance is in countries with very low coverage, like Rwanda and Cape Verde, where many households are excluded from subsidies simply because they do not have an electricity connection. Mexico performs somewhere between Cape Verde and India (0.61), even though it has much higher cover- age levels than either of these countries. This is an indication of just how regressive the tariff structure in Mexico is. When evaluating the significance of subsidies for poor households, one must look beyond just the distributional incidence of subsidies. Another important measure is the error of exclusion, the percentage of poor households that do not receive any subsidy at all (Table 3.4). With IBTs, virtually all connected households generally receive a subsidy, so errors of exclusion are usually driven by those households that are still unconnected. In Mexico, with a 96 percent coverage rate, the error of exclusion is relatively low. Another consideration in assessing the impact of subsidies is their magnitude relative to household income. Mexico's electricity subsidies represent a larger percentage of income among the poor than among the non-poor (Figure 3.4). For the poorest 20 percent of households, the subsidies are equivalent to about 5 percent of income, for the richest only 1 percent. This means that the subsidies do contribute to reducing inequality in purchasing power in Mexico. It may seem surprising that even the most regressive tariffs (in absolute Table 3.3. Distribution of Electricity Subsidies by Income Decile in each Tariff Category (ENIGH 2006) Oportunidades Cumulative Subsidies in Tariff Structures Deciles Energéticas Total Total 1 1A 1B 1C 1D 1E 1F 1 35.0 5.7% 5.7% 9.5% 8.1% 13.7% 2.3% 2.9% 4.6% 0.5% 2 21.1 7.6 13.3 10.1 10.6 12.9 4.8 6.8 6.5 4.4 3 17.6 8.0 21.3 10.0 10.2 9.8 6.5 7.2 8.0 5.8 4 10.1 8.9 30.2 10.4 12.3 10.7 6.8 13.5 6.8 6.9 5 6.6 10.3 40.5 10.8 9.5 9.6 9.8 9.3 12.3 10.7 6 3.8 10.5 51.0 10.9 10.7 9.6 12.3 6.2 10.4 8.7 7 2.9 12.0 62.9 11.6 9.5 9.2 12.2 13.9 12.9 13.4 Residential 8 1.7 11.5 74.5 10.2 10.3 9.8 15.2 10.9 12.3 8.3 9 0.7 13.3 87.8 8.9 11.9 10.2 17.0 15.5 15.0 15.8 10 0.6 12.2% 100.0% 7.8% 6.9% 4.6% 13.1% 13.8% 11.1% 25.6% Electricity Subsidy to deciles 1­4/40 2.10 0.61 0.76 0.83 0.84 0.45 0.69 0.53 0.43 Subsidies Source: Scott 2007. in Mexico 23 24 World Bank Working Paper Table 3.4. Targeting Performance of Subsidies Delivered through IBTs Country with IBT for Index of Regressivity: Electricity (1.0 = neutral distribution) Error of Exclusion (%) Rwanda 0.35 87.2 São Tomé and Principe 0.41 76.8 Cape Verde 0.48 75.6 India: Average of state IBTs 0.70 21.0 Peru 0.82 59.9 Hungary 0.98 1.7 Index of Regressivity: Share of subsidies to poorest 40 percent of HHs/40. Error of Exclusion: Percent of poorest 40 percent of households not receiving subsidy. Source: Komives and others 2005. terms) are less regressive than the distribution of income in Mexico, and are especially sig- nificant for the poor in the warm localities where the most regressive tariffs are applied. This situation arises because of the high electricity consumption of households in hot local- ities, including the poor, and the large subsidies they receive per kWh. Estimated monthly electricity consumption among the poor in tariff zones 1D, 1E, and 1F is two to three times greater than consumption among the poor in tariff zone 1 or 1A (Table 3.2). As a result, Figure 3.4. Electricity Subsidies as a Percentage of Household Income, by Income Quintiles 24% 22% 20% 18% 16% income 14% 12% 10% 8% subsidy/current 6% 4% 2% 0% 1 2 3 4 5 household quintiles Total RES 1 1A 1B 1C 1D 1E 1F RES = Residential electricity subsidies. Source: Scott 2007 using ENIGH (2006) and CFE tariff data. Residential Electricity Subsidies in Mexico 25 electricity subsidies represent 20 percent of income for the poorest 40 percent of house- holds in tariff categories 1E and 1F. In contrast, Tariff 1 subsidies represent 3.4 percent of income for the poorest 40 percent of households in this tariff area. Putting this observation into perspective, the large volumes of consumption among the poor in tariff zones 1D­1F strongly reflects demand response to very low electricity prices. Low prices elevate consumption at all income levels and raise systemwide costs. If prices were to increase, households would find ways to reduce their electricity consump- tion. This price elasticity of demand is taken into account in the simulations of alternative subsidy mechanisms in Chapter 4 of this report. Thus far the discussion has centered on the question of how well electricity subsidies reach the poor. More germane in terms of social policy is whether electricity subsidies are the most effective way of improving the living standards of the poor. If the goal is to aug- ment the purchasing power of poor households, how effective are electricity subsidies likely to be relative to subsidies for other goods or programs? To target the poor, it is best to sub- sidize goods that all poor households use (universal reach) and that non-poor households either do not use or spend less on than the poor (inferior goods). Electricity is a nondif- ferentiated good that is used by both the poor and the non-poor. Unconnected households that use no electricity are likely to be poor. On this basis, electricity consumption does not appear to be a good choice of a mechanism for delivering subsidies to the poor. It is not surprising to find that residential electricity subsidies delivered through the tariff structure are poorer distributive instruments than other social programs in Mexico. Figure 3.5 compares the redistributive effectiveness of electricity subsidies to that of some Figure 3.5. Concentration Coefficients (Quasi-Gini) for Electricity Subsidies and other Redistributive Instruments in Mexico P. Empleo Temporal OEC IMSS-Oportunidades School Breakfasts Total targeted Health uninsured SSA Primary Education Lower Sec Education Tariff 1 PROCAMPO Total Social Spending LICONSA milk subsidy Total untargeted Total Electricity Total RES Housing credit uninsured Upper Sec Education Health insured IMSS Pensions IMSS Tariff 1F Health insured ISSSTE Tertiary Education Housing credit insured Pensions ISSSTE -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 Progressive (Pro-Poor) Neutral Regressive (Pro-Rich) RES = Residential electricity subsidies. Source: World Bank (2004); Scott (2004); Scott (2007). 26 World Bank Working Paper of the principal redistributive programs currently implemented in Mexico. The figure reports the concentration coefficient (or quasi-Gini) for each program. The Oportunidades Energéticas subsidy is allocated through the most effective (administrative) targeted instru- ment available in Mexico (Oportunidades); this makes it highly progressive. Tariff 1 is slightly more regressive than social spending in general in Mexico. Tariff 1F, on the other hand, is among the most regressive instruments presented in Figure 3.5. Overall, electric- ity subsidies (subsidies delivered through tariffs plus Oportunidades Energéticas) are mod- erately regressive. This suggests that, if the main goal of electricity subsidies is to serve as an instrument of social policy, there are other better ways to raise the disposable incomes of the poor. As just one example, Figure 3.5 shows that temporary employment programs in Mexico are highly progressive. Coady, Grosh, and Hoddinott (2003) also found, in a review of targeted programs around the world, that all of the public works programs in their sample were progressive. These programs are able to successfully target the poor through self-exclusion: the non-poor are not interested in the low-wage jobs these programs provide. The same review showed that 82 percent of cash transfer programs and 83 percent of social funds produced a pro- gressive distribution of benefits.12 Cash transfer programs performed well because they, like Mexico's Oportunidades program, were usually targeted with means testing or a proxy means test. The social fund programs achieved good targeting by directing resources toward goods and services that disproportionately benefit the poor. 12. See summary of Coady, Grosh, and Hoddinott (2003) results presented in Komives and others (2007). CHAPTER 4 Assessment of Alternative Subsidy Mechanisms A s indicated, Mexican residential utility subsidies are large and continue to grow. They constitute a significant burden to public finances and are regressively dis- tributed, benefiting larger-volume consumers and upper-income households dis- proportionately. Current tariff levels and structures elevate demand and raise the cost of electricity provision for all consumers. What options does Mexico have for reducing the size and regressivity of the electricity subsidies? Four options are assessed. They were selected to reflect, on the one hand, current thinking in Mexico about the possible direction of tariff changes and, on the other hand, policies and approaches that have proven promising in other countries. Possible scenarios were discussed with Mexican government officials in January 2008 and the comments received were used to make the final choice of the scenarios. The intention is not to present all possible or feasible options. The goal is rather to pro- mote meaningful policy discussion by demonstrating what could be achieved with four dis- tinct alternatives to the current subsidy mechanism in terms of their impact on both the magnitude and distribution of subsidies. The four scenarios are as follows: 1. Modification of the High Consumption Residential Tariff (DAC): The first scenario is a conservative option, similar to efforts that have been made in the past. The DAC tariff is expanded to cover the top 10 percent of consumers in each tariff category rather than the 2 percent currently paying the DAC rates, while the tariff level is low- ered to 90 percent of its current level.13 Tariffs for other consumers would remain the same. 13. It is generally believed that the current DAC tariffs are slightly higher than long-run marginal cost. 27 28 World Bank Working Paper 2. Volume-differentiated tariff (VDT): A second option is to make a major change in the tariff structure, eliminating the multitude of increasing block tariffs and replacing them with a unified national tariff system of just three flat volumetric tariffs with dif- ferent levels of discount. 3. Means testing: The third option examined is also a major change to the tariff structure. Residential customers in Mexico pay the same flat volumetric tariff, with the exception of poor households that are targeted (through qualification for the Oportunidades program, or an alternative means-testing mechanism) to receive a discount. 4. Cost reduction: The last scenario considered is a reduction of supply costs.14 Cost reduction (for example, through technical, operational, and management/gover- nance improvements) has the potential to reduce the magnitude of the subsidies without altering the tariff structure. Two hypothetical scenarios are analyzed: a reduction of average unit cost by 10 percent and by 20 percent.15 Of the four options assessed, the first (the modified DAC) is the most similar to what has been done in the past and is what some policymakers are considering for the future. The VDT and means-tested discount represent major departures from the current tariff and subsidy system, for consumers and for the utilities, while implementing a cost reduc- tion scenario would likely require significant managerial, labor, and corporate changes and a multiyear investment program. None of these changes could be made overnight, but rather need to be phased in over time, and implementing any one would require a care- fully planned transition. The modeling of transition paths is beyond the scope of this report. The purpose of these simulations is to present a first assessment of what might be possible to achieve with different types of reforms in terms of the overall level of subsidies and their distributional incidence. Once a decision is made to move in a particular direc- tion, planning and timing of the transition path would become critical. An important feature of these simulations is the incorporation of price elasticity of demand in the analysis. Benefit incidence analysis of policy change often ignores the fact that price changes affect consumer behavior. As a result, these analyses tend to overesti- mate impacts on revenues, subsidies, and distributional incidence. With electricity tariff reform, one would expect electricity consumption to fall if prices rise and to increase if prices drop. To capture the effect of changes in electricity consumption on the results of tariff reform, one needs to incorporate demand elasticity into the calculations. An estimate of short-run price elasticity of demand that was derived from CFE's demand model (recently published by SENER): -0.14.16 Over time, of course, price is not the only factor affecting demand. Incomes may rise or the prices for electrical appliances could fall. Where possible, the estimates of consumption levels after a tariff reform also account for possible changes in these other components of the demand function over the period of the simula- tion (see Appendix D). 14. Cost reductions are based on CFE costs because, as explained in Chapter 3 and Appendixes 4 and 5, CFE costs are used as the basis for calculating electricity subsidies. 15. Because subsidies are measured in this study as the difference between price paid and CFE average cost (see Appendixes 4 and 5), the cost reduction amounts to a 10 or 20 percent reduction of CFE costs. 16. This estimate is slightly lower than the -0.16 estimate produced by SHCP in 2003. Residential Electricity Subsidies in Mexico 29 Simulations of scenarios are conducted, whenever possible, with both CFE and ENIGH data. The two analyses are not directly comparable for several reasons. The CFE data are from 2005, and the ENIGH data are from 2006.17 The CFE analysis takes into considera- tion annual consumption (and thus the tariffs in both the summer and winter months), whereas the ENIGH data only consider summer consumption levels. The CFE analysis assesses subsidy distribution across consumption deciles, while the ENIGH analysis does so across income deciles. Moreover, as seen earlier, the two data sets have different esti- mates of the number of CFE customers and of average consumption levels. Despite these differences, the same broad policy messages about the likely effect of different scenarios for subsidy reform emerge from both analyses. The two data sets complement each other and increase confidence in the overarching conclusions of the study. A Modified DAC Tariff Among the options currently under consideration for reducing electricity subsidies is mod- ifying the DAC tariff, reducing the Kilowatt-hour (kWh) charge to better reflect long-run marginal cost, and applying it to a larger group of large-volume consumers. It is widely rec- ognized that the DAC tariff is above long-run marginal cost. From an economic perspec- tive this is inefficient; efficient levels of consumption are obtained when customers pay the marginal cost of their consumption for each kWh consumed. From this perspective, cur- rent DAC tariffs are too high, although how much too high is not known. This simulation (the parameters of which were discussed with SHCP) assumes that the existing DAC tariffs would be replaced with two new DAC tariffs (Table 4.1).18 The new DAC tariffs would preserve the fixed charge of the existing system, but the volumetric charge would be reduced by 10 percent.19 The minimum consumption threshold for the DAC tar- iffs is reduced such that the top 10 percent of all customers in each tariff category pay this new DAC rate. The combined result of these two changes would be lower prices for those currently paying DAC and higher prices for those just under the current DAC thresholds. Table 4.2 presents the simulated results of this change to the DAC on Table 4.1. DAC Tariffs Used in the Simulation the magnitude of subsidies received New DAC . . . Winter Summer by each household and on the distri- bution of subsidies by consumption For Tariffs 1­1E decile. Compared to the current tar- Fixed charge 51.024 51.024 iff system, the modified DAC would Volumetric charge 1.809 1.809 reduce the total value of the subsidies For Tariff 1F provided to residential customers by Fixed charge 51.395 50.653 approximately 2,500,000 pesos per Volumetric charge 1.748 1.856 year, or only about 4.5 percent of the 17. The most recent data available at the time the study was started were used for the analysis. 18. The first new tariff is designed to reflect costs in the national interconnected system. The second reflects costs in Baja California, which is not connected to the rest of the national system. Tariff 1F users are concentrated in Baja California and Sonora. 19. The new DAC tariffs are modeled on the current DAC for the Baja California and North East regions. 30 World Bank Working Paper Table 4.2. Modified DAC Scenario: Subsidy Distribution by Consumption Decile Avg. Subsidy Avg. Subsidy Subsidy Cumulative Decile (pesos/kWh) (pesos/HH per year) Distribution (%) Distribution (%) 1 9.03 550 2.8 2.8 2 2.56 1,169 5.9 8.6 3 2.05 1,571 7.9 16.5 4 1.85 1,884 9.4 26.0 5 1.70 2,251 11.3 37.2 6 1.62 2,536 12.7 50.0 7 1.55 2,739 13.7 63.7 8 1.44 3,124 15.7 79.4 9 1.10 2,954 14.8 94.2 10 0.26 1,158 5.8 100 Total 1.23 1,994 100 Note: Average consumption is estimated after the price change. current volume of subsidies. The impact of this tariff modification on the volume of sub- sidies is quite small because it does nothing to modify the subsidies provided to 90 percent of electricity consumers. Moreover, the potential revenue gain associated with applying a DAC tariff in the top decile is less than expected because consumption in the top decile would drop as a result of the price effect.20 Nor does the modified DAC significantly alter the distribution of the subsidies across consumption deciles: the only real difference is for the largest 10 percent of electricity consumers (in decile 10), who in this scenario capture only 6 percent of the subsidies compared to 10 percent with the current tariff. The ENIGH analysis confirms that extending a modified DAC tariff to the top 10 per- cent of customers, and making no other changes to the tariff structure would have only a modest impact on the volume and distribution of subsidies. If the wealthiest 10 percent of households pay the DAC, subsidies decline by 16 percent, and subsidy distribution would remain regressive with the wealthiest 30 percent of households capturing more of the sub- sidies than the poorest 30 percent.21 Two Options for a Unified National Tariff Structure The modified DAC tariff option discussed above largely preserves the current structure of residential tariffs in Mexico--seven distinct tariff structures for different climatic zones, three billing schedules for each tariff, seasonally differentiated tariffs, and increasing block 20. In this simulation, the assumption is that average consumption in the top decile would drop from 404 kWh per month to 368 kWh per month in response to the price increase. 21. Applying the DAC tariffs to the wealthiest customers (as opposed to the largest-volume consumers) would be challenging in practice. It would require using some form of means testing to identify wealthy customers. Residential Electricity Subsidies in Mexico 31 tariffs for most customers. One possible direction of change for the future is to move toward a single national electricity tariff. The next two scenarios assess the performance of two different national tariff structures that offer some degree of discount for all customers while eliminating specific regressive features of the existing tariff structure, namely the extreme seasonal subsidies and the increasing block tariffs. The first scenario introduces a volume-differentiated tariff (VDT) that targets low-volume consumers for additional dis- counts. The second scenario uses means testing to offer added discounts to the poorest of customers. Volume Differentiated Tariff Like an increasing block tariff (IBT), a volume differentiated tariff (VDT), is a quantity- targeted tariff--the discounts given to customers depend on how much electricity a cus- tomer uses. The difference between an IBT and a VDT is that each customer class pays just one price per kWh consumed (as is now the case with the DAC tariffs in Mexico). In this way, larger-volume consumers are excluded from receiving the deep discounts normally provided in the first blocks of an IBT. Figure 4.1 shows one example of a VDT composed of three tariff schedules, each with a different level of discount. The basic tariff schedule (1.10 pesos per kWh) is for customers consuming less than 100 kWh per month. The intermediate tariff schedule applies to cus- tomers consuming between 101 kWh and 200 kWh per month. Since they pay 1.40 pesos per kWh for all units of consumption, a customer consuming 200 kWh per month would pay a total of 280 pesos per month. The standard tariff schedule (1.75 pesos per kWh) applies to all remaining customers, those consuming more than 200 kWh per month. Figure 4.1. Example of a VDT with Three Tariff/Schedules 2.00 1.80 1.60 1.40 1.20 kWh per 1.00 0.80 Pesos 0.60 0.40 0.20 0.00 0 0 40 80 120 160 200 240 28 320 360 400 440 480 kWh consumed Basic Intermediate Standard 32 World Bank Working Paper Simulations conducted in other countries suggest that switching from an IBT to a VDT can improve the poverty targeting of electricity subsidies.22 In Cape Verde, for example, moving from an IBT with a first block of 40 kWh to a VDT tariff with one price for those consuming less than 40 kWh and another for those consuming more was shown to more than double the share of subsidy benefits going to the poorest 40 percent. The same simu- lation in Rwanda also indicated that moving to a VDT was a progressive change, but the improvement was not as large as in Cape Verde, and subsidy distribution remained very regressive even with the VDT. Part of the difference between the two cases is that connec- tion rates among the poor are lower in Rwanda than in Cape Verde. In Mexico, with near universal coverage, the application of a VDT has strong potential to increase the progres- sivity of subsidies. The VDT in the simulation is composed of three different flat volumetric tariffs (as in Figure 4.1). The standard VDT tariff (that is, the highest) is set equal to 80 percent of CFE's current average accounting cost, which means that in this simulation all customers receive at least a 20 percent subsidy on all units of consumption.23 The intermediate and basic tar- iffs provide larger discounts. These prices are 10 percent and 30 percent less than the stan- dard tariff, respectively (see Appendix F for tariffs used in the simulation).24 More than 90 percent of those in Tariff 1 would fall into one of these more deeply discounted tariffs (basic or intermediate). Tariff 1F has the lowest percentage of customers that would fall into the basic or intermediate tariff schedule: 55 percent in the winter and 20 percent in the summer. All others would receive the standard tariff schedule's 20 percent subsidy on all units of consumption. Even though this scenario preserves some degree of subsidies for all customers (and even increases subsidies for the 2 percent of customers who are now paying the DAC), moving to this national VDT would go much farther toward reducing the volume of sub- sidies than the modified DAC tariff scenario presented above. In the analysis with CFE data, the total volume of subsidies would decrease by more than 50 percent and the price- to-cost ratio would increase to 0.71 (compared to 0.44 with the DAC scenario). The aver- age subsidy received by the smallest-volume consumers would not change significantly. The annual subsidy received by all others would drop, both because of the price increase and because households would reduce consumption in response to the tariff increase. Larger-volume consumers would still receive larger annual subsidies than smaller-volume consumers, but the difference is much less dramatic than with the current tariff structure (Figure 4.2). Overall, the VDT represents a slightly progressive change in the distribution of subsidies: 32 percent of subsidies would go to the first four consumption deciles (com- pared to 25 percent today and 26 percent with the modified DAC scenario). The VDT simulation with ENIGH data also found that total subsidy volume declines dramatically relative to the current mechanism. The reduction in subsidies would be felt 22. These examples were reported in Komives and others (2005). 23. This simulation sets the standard tariff at 80 percent of cost (and thus preserves a general subsidy for all customers) in order to later be able to differentiate between the effect of tariff change and the effect of removing subsidies for the majority of customers. This discussion appears later in the chapter, when the analysis of moving to a VDT in which those paying the standard tariff are no longer subsidized is presented. 24. Since the cost function in analysis with the CFE data includes a measure of fixed plus variable cost, the tariff structures used in the simulation also have a fixed and a variable charge, each set at 80 percent of average cost. Residential Electricity Subsidies in Mexico 33 Figure 4.2. Average Subsidy per Household per Year in each Consumption Decile, with Three Different Tariff Structures 3500 3000 year 2500 per hh 2000 per 1500 subsidy 1000 Avg 500 0 1 2 3 4 5 6 7 8 9 10 Consumption decile Current VDT Modified DAC Source: Based on Aburto 2007 with CFE data. in all income groups. The average subsidy per household per month in the lowest income decile would go down by more than 55 percent, in decile 5 by 62 percent, and in decile 10 by 52 percent. The resulting distribution of subsidy benefits across income groups would not be dramatically different from the distribution today. Means-tested Discounts With a VDT, like an IBT, the quantity of electricity consumed determines the amount of subsidy each consumer receives. An alternative to consumption-based subsidy allocation mechanisms is to use a means test to determine eligibility for subsidies. Means testing can take many forms, but always involves using some information about households to assess household poverty status. It may be done using income information, or proxies for house- hold income such as information about the quality of housing or socioeconomic charac- teristics of the household. It is a much more labor-intensive approach to determining who is eligible for subsidies than using quantity targeting, because means testing requires an administrative decision about eligibility for subsidies. With quantity-targeted subsidies, households indirectly decide whether or not to be subsidized when they decide how much electricity to use. One way to reduce the administrative cost of allocating subsidies through means-testing is to piggyback on an existing means-testing program. The Oportunidades Energéticas pro- gram, for example, uses the means-testing system established under Oportunidades to 34 World Bank Working Paper determine eligibility for subsidies. In Chile, the national water subsidy,25 which is admin- istered by the water utility through water bills, similarly uses a national means-testing instrument (the ficha CAS--Caracterización Social) to determine eligibility for the water subsidy. Using an existing means-tested program to target water or energy subsidies has the advantage that it spreads the administrative cost of means testing over a number of sectors/programs and can therefore be quite cost-effective. In contrast, establishing sector- specific means-testing systems are usually prohibitively expensive because administrative costs absorb a large proportion of the subsidy "budget." Means-tested water and electricity subsidies are much more progressively distributed than quantity-targeted programs like IBTs or VDTs. Komives and others 2005 found that the share of subsidy benefits accruing to the poorest 40 percent was twice as large on aver- age in means-tested programs compared to IBTs or VDTs (Table 4.3). However, a dis- advantage of means testing is the prevalence of errors of exclusion among the poor. In an effort to exclude the non-poor from receiving subsidies, means-tested programs exclude poor households as well. Table 4.3. Pro-Poor Targeting Performance of Means-tested Electricity Subsidies Measure of Progressive Distribution: Percent Subsidies to Poorest Error of Exclusion 40 Percent of HHs / 40 (% of poor HHs not Case (1.0 = neutral distribution) receiving subsidy) Argentina: Average of provincial-level 1.50 94 means-tested subsidies Tiblisi, Georgia: Limited free 1.20 75 electricity allowance for means-tested beneficiaries Bogotá, Colombia: Combined use of 1.35 17 means testing and geographic targeting (simulated) Cape Verde: Means-tested subsidy on 1.46 93 first 40 kWh per month (simulated) Source: Komives and others 2005. In Mexico, such errors of exclusion are one risk of relying on the Oportunidades veri- fication systems for determining eligibility for electricity subsidies. Although Oportu- nidades is the best-targeted social program in the country, some poor households are excluded. As one example, poor households living in communities that lack health centers and schools do not qualify for Oportunidades. This is because to be eligible one must not only qualify as poor through the proxy means test, but also must send children to school and make regular visits to health centers. Two targeting approaches are used in the simu- lations of means-tested subsidies presented here. In the first, only Oportunidades recipi- ents qualify for the means-tested subsidy. In the second, the subsidy is provided to all 25. The program offers a discount for 15 cubic meters of water per eligible household. Residential Electricity Subsidies in Mexico 35 households in Oportunidades and all other households in the first income quintile, thus simulating the potential reach of Oportunidades with improved targeting. Simulations of means-tested electricity subsidies in Mexico used ENIGH data because it is not possible to identify poor households based on the CFE customer database. The tar- iff structure used in these simulations is a national tariff--this time a flat volumetric tariff for all households in the country, except those that qualify as poor through a means test. The standard tariff is equal to 80 percent of CFE average cost (the same level as the stan- dard tariff in the VDT scenario), which means that all households receive at least a 20 per- cent discount on each kWh of consumption. The households classified as "poor" in the means test receive an additional discount of 40 percent on all units of consumption.26 Using this national tariff with means-tested discounts would result in a significant reduction of the total cost of subsidies. When the extra discounts are provided to Oportu- nidades recipients only, the total subsidy bill is 42 percent of the current subsidy volume. If Oportunidades recipients plus all other households in the first quintile of the income dis- tribution receive the extra discount, the cost of the subsidies increases slightly but is still only 48 percent of the current total cost. To further reduce costs for this subsidy option, one could also limit the volume of consumption for which the discount is provided. For example, if all targeted households received a subsidy on only the first 100 kWh of elec- tricity they consume each month (and thereafter paid the standard tariff),27 the subsidy bill would drop to below 40 percent of the current level. Figure 4.3 compares the average subsidy per household per month with the current tariff and under the means-tested discount and VDT simulations. The subsidies provided to the top deciles are lower with the means-tested discount than the VDT, and both are dramatically lower than with the current tariff. In the lower-income deciles, the means- tested discount results in larger subsidies for the poor than does the VDT. When all house- holds in the first quintile are targeted (the improved targeting scenario), the average subsidy provided to the poorest is very similar to what they receive today. In general, a national tariff with a means-tested discount for the poorest households is the most pro- gressive option examined here. The better the reach and accuracy of the means-testing pro- gram, the more significant the improvement would be. Interestingly, limiting the volume of consumption on which targeted poor households receive an extra discount actually makes the means-tested subsidy less progressive. Figure 4.4 demonstrates this with simulations of discounts for Oportunidades recipients only. When targeted poor households receive an extra discount only on 100 kWh of consumption, the poorest deciles receive a lower percentage of the total subsidy than they would if the dis- count were provided on the full volume of poor households' consumption. This occurs because many poor households in Mexico (particularly those in the hotter regions) are rel- atively large consumers of energy. When the discount is limited to only 100 kWh, much of their consumption would not be subject to this extra discount. This analysis shows that 26. Because the simulations with ENIGH data use a variable measure of cost for the cost function (see Appendix D), the standard tariff in this scenario is 1.75 pesos per kWh (20 percent less than the current CFE average cost of 2.19 per kWh). The discounted tariff is 1.10 pesos per kWh. 27. This would amount to offering targeted poor households an IBT in which they receive an extra discount on the first 100 kWh per month, and thereafter would pay as much as all other customers for their electricity. 36 World Bank Working Paper Figure 4.3. Average Subsidy per Household per Year in each Income Decile, with Different Tariff Structures 350 300 (pesos) 250 month 200 per hh 150 per 100 subsidy 50 Avg 0 1 2 3 4 5 6 7 8 9 10 Income decile Current Means-tested discount on all consumption, Oportunidades only Means-tested discount on all consumption with improved targeting VDT Source: Based on Scott 2007--ENIGH data. there is a tradeoff in designing a means-tested discount of this sort between reducing subsidy volumes and directing subsidies to the poor. Limiting the subsidy to 100 kWh or 200 kWh would help bring down the cost of the subsidies, but would also reduce the progressivity of the tariff structure. Cost Reductions The fourth option for reducing subsidies and improving their progressivity is cost reduc- tions. Rather than reduce subsidies by raising tariffs, an alternative is to reduce them by lowering supply costs. From the analysis in previous Chapters of this report, it is clear that there is scope for lowering costs at LFC, but probably also at CFE. No attempt has been made to determine what magnitude of cost reductions would be feasible at each utility, or exactly how the cost reductions could be obtained. To examine how cost reductions affect the magnitude and distribution of subsidies, two hypothetical scenarios are analyzed: a reduction of CFE average cost by 10 percent and 20 percent. The tariffs do not change in these scenarios: Tariffs 1­1F and the DAC tariffs remain at their current levels. Residential Electricity Subsidies in Mexico 37 Figure 4.4. Percent of Subsidies Accruing to each Income Decile, under Three Different Means-tested Discount Scenarios 20% 18% decile in 16% hhs 14% by 12% 10% received 8% 6% 4% subsidies of 2% % 0% 1 2 3 4 5 6 7 8 9 10 Income decile Extra discount on 100 kWh Extra discount on 200 kWh Extra discount on all consumption Note: All scenarios assume that subsidies go to Oportunidades recipients only. Source: Based on Scott 2007--ENIGH data. Cost reductions of 20 percent would have a more significant effect on the magnitude of the subsidies and on subsidy distribution than the modified DAC tariff (but not as great as the VDT or means-tested discount). Cost reductions result in a smaller subsidy per kWh for all customers at all levels of consumption, compared to the DAC scenario, which affected only 10 percent of customers. The result, using CFE data, is an estimated 35 percent reduc- tion in the total annual volume of subsidies. Larger-volume consumers would see their total subsidies decline more than smaller-volume consumers. As a result, there is a slightly progressive change in the distribution of subsidies by consumption decile. Simulations using ENIGH 2006 data confirm that a 20 percent cost reduction would produce a 42 percent reduction in the total magnitude of the subsidies. This level of cost reduction would also have a progressive effect on the distribution of subsidies by income decile. It would increase the share of subsidies going to the poorest 30 percent of house- holds by 3 percentage points (from 21 to 24 percent) and reduce the share of subsidies going to the richest 30 percent of households by 4 percent (from 37 percent to 33 per- cent). Nonetheless, the bulk of the subsidies would still go to the non-poor after such a change. A 10 percent reduction in average cost would reduce the magnitude of electricity subsidies by 21 percent. Subsidy reductions would be larger for richer customers, who 38 World Bank Working Paper also tend to be larger-volume consumers, than for poorer customers. As a result, even a 10 per- cent reduction in costs would produce a slight improvement in the distribution of subsidies. Combining Tariff Modifications with Cost Reduction Thus far, the simulations examined the effect on subsidies of modifying tariffs (but pre- serving some degree of subsidy for all or most customers) and of reducing supply costs. Combining changes in tariff level and structures with supply cost reductions would pro- duce more dramatic reductions in subsidy volumes and improvements in distributional incidence, without having to raise residential tariffs further. To demonstrate this, the VDT and means-tested discount scenarios described above are combined with a 20 percent reduction in CFE average cost. Since the standard tariff in these scenarios was set at 80 per- cent of CFE current average cost, a 20 percent cost reduction means that residential cus- tomers who pay the standard tariff would no longer receive any subsidies. Subsidies would be limited to VDT customers in the basic or intermediate tariff group (that is, those con- suming less than 200 kWh per month) and, in the means-tested discount scenario, to those who are shown to be poor through the means test. The national VDT, coupled with a 20 percent reduction in CFE average cost, would reduce the volume of subsidies to less than a fifth of their current level, compared to half their current level for the VDT alone or two-thirds their current level with cost reduction alone (Figure 4.5). Also, a much larger percentage of subsidies is delivered to the house- holds in the first three consumption deciles when tariff change and cost reduction are com- bined. The analysis with ENIGH data confirms that simultaneously moving to a VDT and reducing supply costs would make the distribution of subsidies progressive across income deciles. The complete elimination of subsidies for higher-volume consumers (in this case Figure 4.5. Outcome Indicators for VDT Scenario, Alone and in Combination with 10 and 20 Percent Reduction in CFE Average Costs Ratio of total subsidies for deciles 1-3 to total subsidies for deciles 8-10 Proportion of subsidies going to first 4 deciles Total volume of subsidies as a proportion of current 0.0 0.1 0.2 0.3 0.4 0.5 VDT with no cost reduction VDT with 10% cost reduction VDT with 20% cost reduction Source: Aburto 2007 with CFE data. Residential Electricity Subsidies in Mexico 39 through a combination of price increases and cost reductions) is critical to achieving a more progressive distribution of subsidies. Similar conclusions emerge from the analysis of the means-tested discount in combi- nation with cost reductions: subsidies delivered through a tariff with means-tested dis- counts cost less and are more progressive when subsidies for untargeted consumers are eliminated through cost reductions. Figure 4.6 shows how the magnitude and distribution of subsidies across income deciles would change when the means-tested discount on 100 kWh per household per month is combined with 20 percent reductions in average supply cost. The improvement in distributional incidence is dramatic. These results point to the criticality of (gradually) eliminating subsidies for the majority of customers if the policy goal is to significantly improve the distribution of electricity sub- sidies. As long as large-volume and richer consumers continue to receive subsidies on every kWh consumed, the total subsidies these groups receive will be larger than those received by the poorer and smaller-volume consumers. Retaining subsidies for the larger-volume and richer consumers also substantially increases the cost of the subsidy program. There are good reasons, therefore, to gradually eliminate general subsidies in the residential electricity tariffs and replace them with limited subsidies targeted at a subset of consumers. This could be done through price increases alone, but, since tariffs are highly subsidized, the required price increases would be very large. In the simulations presented here, general subsidies are elim- inated through a combination of price increases and reductions in the cost of supply, which means that the burden of reducing subsidies is shared by consumers and the utilities. Figure 4.6. Cumulative Distribution of Subsidies across Income Deciles: Means-tested Discount on 100 kWh (with improved targeting) Alone, and with 20 Percent Cost Reduction 102 100 80 subsidy of % 60 40 Cumulative 20 0 1 2 3 4 5 6 7 8 9 10 Income decile Tariff change alone With 20% cost reduction Source: Scott 2007. 40 World Bank Working Paper Comparing the Options Figures 4.7, 4.8, and 4.9 compare the different scenarios presented with respect to their impact on the magnitude of subsidies and on two indicators of distribution. The modified DAC scenario has the smallest impact on the volume of subsidies because only the bills of 10 percent of households are touched in the tariff reform. VDT and means-tested discount scenarios both reduce subsidy volumes by more than half. Among the tariff change scenar- ios without cost reduction, the greatest reduction comes from the means-tested discount on 100 kWh. This scenario limits both the value of the subsidy to which each recipient is enti- tled and the number of households targeted for this discount. When the discount on 100 kWh is combined with a cost reduction of 20 percent (such that untargeted households receive no subsidy and those targeted only receive a discount on their first 100 kWh of consump- tion each month), residential electricity subsidies are nearly eliminated. The VDT in com- bination with cost reduction is the second-best performer in terms of subsidy volume. In this scenario, only households consuming less than 200 kWh receive the subsidy. Figure 4.7. Magnitude of Subsidies as a Percentage of Current Subsidy Volume under Different Scenarios Discount on 100 kWh and 20% cost reduction VDT and 20% cost reduction Discount on 100 kWh Discount on all kWh VDT 20% cost reduction Modified DAC 0 10 20 30 40 50 60 70 80 90 100 % of baseline CFE ENIGH Note: The means-tested discount scenarios presented here all simulate improved targeting (Oportunidades recipients plus all households in the first quintile). With respect to the distribution of subsidies across consumption deciles, all of the options perform better than the existing subsidy scheme. The best-performing option is clearly the VDT combined with cost reduction (it was not possible to simulate distribution of a means-tested subsidy across consumption deciles). This is the only scenario simulated with CFE data in which the bottom four deciles receive a larger share of the subsidies than they would in a random or neutral distribution of subsidy benefits (54 percent compared Residential Electricity Subsidies in Mexico 41 to 40 percent of the subsidy benefits). In all other cases, the first four consumption deciles receive less than 40 percent of the subsidy benefits. Cost reductions and the VDT without cost reduction (which preserves a degree of subsidy for all customers on all units of con- sumption) do, nonetheless, represent an improvement over the current tariff structure. The modified DAC scenario hardly changes the existing situation. With respect to distribution of subsidies across income deciles, one scenario actually per- forms worse than the existing tariff structure--the means-tested discount restricted to only the first 100 kWh. This occurs because most of the poor consume much more than 100 kWh per month. For similar reasons, the distribution of subsidies with the VDT alone is not much different from the distribution achieved under the present tariff/subsidy system. Cost reduc- tions of 20 percent have a larger progressive impact than the VDT, but the only scenarios that manage to distribute at least 40 percent of the subsidies to the poorest 40 percent of consumers are those that offer a means-tested discount on all consumption and/or that eliminate subsi- dies to most consumers by combining a tariff change with cost reductions. The means-tested discounts combined with cost reductions produce the most progressive distribution. A risk with the means-tested discounts concerns errors of exclusion. Poor households that are, for some reason, excluded from the targeting scheme will receive no subsidies. Improving targeting and reducing errors of exclusion becomes especially important when prices are increased and general subsidies are reduced. Figure 4.8. Percent of Subsidies Going to First Four Income (ENIGH) or Consumption (CFE) Deciles under Different Scenarios 100% 90% 80% deciles 70% 4 first 60% to 50% going 40% subsidies 30% of % 20% 10% 0% Means-test Current tariff Modified DAC VDT 20% cost Means-test VDT and 20% Means-test discount (100 structure reduction discount (all cost reduction discount (100 kWh) kWh) kWh) and 20% cost reduction ENIGH CFE Note: The means-tested discount scenarios presented here all simulate improved targeting (Oportunidades recipients plus all households in the first quintile). 42 World Bank Working Paper Figure 4.9. Ratio of Subsidies to First Three Deciles Compared to Last Three Deciles in Different Scenarios Income deciles (ENIGH) Consumption deciles (CFE) VDT and 20% cost reduction 4.94 20% cost reduction 0.56 VDT 0.53 Modified DAC 0.44 Current tariff structure 0.4 Discount (100 kWh) and 20% cost reduction 153 VDT and 20% cost reduction 1.23 Discount (all kWh) 1.01 20% cost reduction 0.73 Current tariff structure 0.57 VDT 0.56 Discount (100 kWh) 0.48 0 1 2 3 4 5 6 7 Note: The means-tested discount scenarios presented here all simulate improved targeting (Oportunidades recipients plus all other households in the first quintile). Toward Transition These simulations of alternative subsidy options cannot be implemented overnight. With the exception of the conservative tariff reform scenario (the modified DAC), it is likely that a tran- sition period of perhaps 5 to 10 years would be needed to phase in tariff changes. Although it is beyond the scope of this report to consider possible transition paths, it is useful to have an understanding of which customers would be most affected by a transition to the subsidy sce- narios presented above and how much their bills would increase. There are important differ- ences in this regard across scenarios. Table 4.4. VDT Scenario: Average Increase in For the VDT scenarios, Monthly Bills, by Consumption Decile Table 4.4 shows that those least affected by the price increases Percent of Current Pesos Consumption Decile Monthly Bill per Month would be the smallest-volume con- 1 50% 7 sumers (who would qualify for the largest discount in the VDT) and 2 132% 29 the largest-volume consumers (who 3 126% 44 are already paying high prices). 4 123% 58 Households in consumption deciles 5 124% 72 2­8 would see their monthly bills 6 136% 103 more than double. The very-small- 7 136% 121 volume consumers would see their 8 115% 136 bills rise by 50 percent. The top 9 55% 115 decile of consumers would, on 10 -11% -66 average, have smaller bills after Source: Aburto 2007 using CFE data. the change than before. Residential Electricity Subsidies in Mexico 43 For the VDT and means-tested options, the largest absolute increases in bills are felt in the middle- to upper-middle-income households (Table 4.5). Increases for the richest are less dramatic (and in some cases could even go down) because many of these house- holds are already paying cost-reflective tariffs. The very poorest also face slightly lower price increases because they consume less electricity than other groups and/or because they qualify for the means-tested discount. Not surprisingly, as a proportion of household income, the price increases would be greatest among the poor and lowest among the rich. For households in the lowest deciles, the increases in bills would equal around 3 percent of household income. The burden is slightly higher for the means-testing scenarios when the subsidy is limited to only 100 kWh per month. Moreover, when targeting is done based on Oportunidades alone, some of the poorest households do not qualify for the discount. The burden on the poor could be reduced by extending the discount to 200 kWh or all consumption and by improving the targeting of Oportunidades. More precise and accurate information about electricity con- sumption among the poor would be useful in the final design of a means-tested discount. Both tables indicate that the tariff increases modeled in the simulations are large rel- ative to current bills. Price increases of this magnitude would undoubtedly need to be introduced gradually and consideration would need to be given to compensatory mecha- nisms for the poor and middle-income groups. Many options for compensating and pro- tecting the poor and middle class exist, and countries that have already undergone major sector reform and tariff increases have chosen different strategies. Supply cost reductions and reform of tariff-based subsidies (to improve targeting) are basic elements in nearly all major reforms. The source of the cost reductions and the particular direction of the tariff modifications differ by country, depending on the starting position and particular prob- Table 4.5. Change in Average Electricity Bills as a Percent of Household Income, by Income Decile for Three Subsidy Scenarios Means-tested Discount on Means-testing Discount on 100 kWh Subsidies 100 kWh (Oportunidades VDT (Oportunidades) plus poorest quintile) Income % of HH Pesos per % of HH Pesos per % of HH Pesos per Decile Income Month Income Month Income Month 1 2.9 86 3.8 113 3.5 104 2 2.4 113 3.0 146 2.8 135 3 1.9 118 2.5 148 2.5 148 4 1.8 128 2.3 159 2.3 159 5 1.7 144 2.1 174 2.1 174 6 1.5 137 1.8 166 1.8 166 7 1.3 144 1.5 172 1.5 172 8 1.1 147 1.3 174 1.3 174 9 0.9 161 1.1 185 1.1 185 10 0.2 90 0.3 111 0.3 111 Source: Scott 2007. 44 World Bank Working Paper lems and opportunities in each context. Ghana, for example, started with a VDT tariff and thus the decision was made to cut subsidies by moving from five tariff groups to only three and to focus subsidies only on the tariff applicable to the lowest-volume consumers. Arme- nia, which started with IBTs in the tariff structure, chose to eliminate the blocks and cre- ate a flat volumetric tariff. A number of countries have also gone beyond these basic reform elements to address concerns about the poor and middle class. One option is to help households invest in improving energy efficiency so that tariff increases coincide with reduction in electricity consumption. Another option is to combine tariff and subsidy modifications with trans- fer mechanisms outside the electricity sector. In Armenia, for example, compensation for electricity price increases was provided through a separate targeted cash transfer program (the Poverty Family Benefit Program). In Brazil, electricity tariffs retained some quantity- targeted and means-tested subsidies even after reform, but the poor were also assisted by the Bolsa Familia cash transfer program. Similarly, in a number of Eastern European coun- tries such as Lithuania and Ukraine, governments established cash transfer programs to help households whose expenditures on basic utilities exceeded a set percentage of house- hold income after price reforms in multiple sectors. The simulations presented in this study are not an exhaustive list of options that the Government of Mexico has for reducing subsidies and better targeting them to low-income households. The options assessed represent potential elements of what could be a future reform strategy. There is no reason a VDT could not be combined with a means-tested dis- count, for example, by offering the most discounted tariff in the VDT only to those in Oportunidades. Likewise, the coverage of a means-tested discount could be improved by offering the discount to all households in the country's most marginal communities.28 Moreover, a complete tariff reform strategy could include additional measures such as those used in the cases described briefly above. Nonetheless, having a sense of the fiscal and distributional impacts of possible elements of a reform package is a critical first step to selecting and implementing meaningful reform. The reform options simulations in this study give a first-order assessment of several promising alternatives for residential elec- tricity subsidies reform in Mexico. 28. The intention was to also simulate the impact of using geographic targeting, but this was not fea- sible because not enough of the most marginal communities are included in the ENIGH sample. CHAPTER 5 Synthesis of Findings and Ways Forward Key Findings Six clear findings emerge from the evaluation of alternatives to the current tariff/subsidy scheme for residential consumers of electricity. There is Relatively Little to be Gained From Minor Tinkering With the Current Tariff Structures Mexico's electricity tariff structures are among the most complex in the world, by design and by natural accretion. The complexity lends itself to reclassification and misclassifi- cation of specific consumer groups seeking to obtain still greater discounts on ever larger volumes on consumption. A minor tariff "reform" that revolves around reducing the High Consumption Residential Tariff (DAC) tariff slightly and expanding its coverage to the 10 percent of highest-volume consumers in each tariff zone has limited impact on the distribution of subsidy benefits. Moreover, it has minimal impact on the total vol- ume of subsidies. After the small number of affected customers adjusts consumption in response to the new (higher) tariffs, the analysis presented here suggests that the total subsidy budget will decline by only 4.5 percent. Moreover, history suggests that tariff adjustments of this type are likely to be followed by intense pressure and lobbying from those most affected to introduce new, more highly discounted tariffs, which would ulti- mately reduce the impact of the reform. The most that one can say for this type of con- servative tariff reform is that it would reduce consumption among the largest-volume consumers, which would have some limited impact on dampening electricity demand. But to have a significant impact on subsidy volumes or distributional incidence requires more far-reaching changes. 45 46 World Bank Working Paper Combined With Cost Reductions, the Means-tested Discount Simulation Produces the Most Progressive Distribution of Subsidy Benefits, but a National Volume Differentiated Tariff (VDT) is Also a Significant Improvement Over the Current Tariff System The VDT and the means-tested discount options produce progressive subsidy distributions when combined with 20 percent cost reductions. The VDT does a good job of targeting sub- sidies to low-volume consumers, and achieves a slightly-more-progressive-than-neutral dis- tribution of subsidies across income groups. The means-tested discount on 100 kWh is highly progressive--almost all the benefits accrue to the lowest-income quintile of the pop- ulation. Extending the discount to 200 kWh or to all consumption of targeted households results in an even more progressive distribution of subsidies while significantly reducing errors of exclusion. Improving the targeting performance of Oportunidades would further improve the distributional incidence of an electricity subsidy targeted to Oportunidades beneficiaries. A means-tested discount has one disadvantage over the VDT for the poor--poor households that are for some reason excluded from the mean-tested discount (for example, because they do not qualify for the Oportunidades program) would receive no discount on their electricity bill. They would therefore pay as much as everyone else. With a VDT, by contrast, households have the possibility of obtaining a discount by simply lowering their consumption level, thus providing some natural safety net for all households. Changes in the Tariff Structure That Do Not Eliminate Subsidies for Most Customers Will Not Substantially Improve Distributional Incidence In the simulations presented here, when the VDT and the means-tested discount are not combined with cost reductions, they leave in place a subsidy of at least 20 percent per kWH for all customers. The more customers consume, the larger their total subsidy. The subsidy per kWh may be higher for low-volume consumers (VDT) or poor households (means test), but the largest total subsidy still goes to the largest-volume and richest consumers. As a result, while there is some improvement over the current tariff system, the overall dis- tribution of subsidies remains regressive even with far-reaching changes in tariff structures. The lesson here is that to make major improvements to subsidy distribution it is critical to aim to eventually eliminate subsidies for the majority of consumers. Reducing Subsidies Will Entail Significant Increases in Customer Bills, Especially for Those in the Middle-income and Consumption Deciles Even the Smaller Increases Among the Poorest Still Represent a Significant Reduction in Their Disposable Income Among the subsidy options assessed, the largest increases in tariffs are found among those in the middle-income and consumption deciles. For the poorest, smaller absolute increases are still significant in comparison to household income. Any transition strategy toward an improved tariff and subsidy model needs to address how to mitigate the socioeconomic impact of rising electricity charges for various strata of Residential Electricity Subsidies in Mexico47 the population. Reducing consumer bills through end-use efficiency improvement would help to limit the increases in charges that would be needed to significantly reduce the mag- nitude of subsidies to residential customers. Other options include using some of the funds saved on consumption-based subsidies to compensate middle-class customers through tax rebates in the medium term, to increase transfers provided to the poor through programs like Oportunidades or geographically targeted programs (SEDESOL micro-regiones), or through increased transfers to the most affected states (such as those with warm summer climates), which could then decide themselves how to compensate those who would face increased charges. Reducing Costs Can Partially Offset the Need for Raising Tariffs, and Any Cost Reduction Would Itself Result in a Greater Proportion of Subsidies Accruing to Low-income Households Improving supply efficiency and reducing costs would have a slightly progressive effect on subsidy distribution since total costs (and the absolute impact of cost reductions) are greatest among the largest-volume consumers. Low-income and low-consumption households would see their subsidy shares rise slightly, while the share going to the rich and large-volume con- sumers would decline. Cost reductions would also have a greater impact on reducing the total volume of subsidies compared to minor changes in the existing tariff structure. The simulation with ENIGH data indicates that a reduction in average costs of 10 or 20 percent would permit aggregate residential subsidies to decline by 21 or 42 percent, respectively. Simulations by consumption decile indicate that a 20 percent reduction in costs would reduce subsidies by 35 percent. Moreover, unit cost reductions can also moderate the magnitude of tariff increases that are needed to eliminate general subsidies. If no progress is made in reducing supply costs, eliminating general subsidies would require greater increases in prices for customers not targeted for subsidies. Politically, it could be more difficult for consumers to support major tariff increases if they were not accompanied by meaningful efforts to reduce utility supply costs. Better Information About Electricity Consumption Among the Poor is Needed This study used two different sources of information to assess the performance of alterna- tive tariff and subsidy instruments--CFE billing data and ENIGH household survey data. Neither source is ideal for examining the incidence of subsidies or the impact of tariff reforms on the poor. CFE databases have no information on household income levels. The ENIGH survey asks households to recall their electricity expenditure but does not record electricity consumption levels from household electricity bills. Both problems could be remedied. CFE could presumably identify Oportunidades recipients in its customer data- base. The ENIGH survey (and other household surveys) could improve information on electricity consumption by asking to see customer bills and recording one or several months of consumption. Having this improved information about electricity consump- tion among the poor would be very useful for calibrating the design of any alternative sub- sidy program. 48 World Bank Working Paper Electricity Subsidies in a Broader Social and Economic Context The current electricity subsidy system in Mexico delivers the bulk of subsidy benefits to the middle and upper-middle classes and to households consuming relatively large amounts of electricity. One cannot argue that general electricity subsidies like these are justified by positive externalities associated with electricity consumption. To the contrary, electricity subsidies inflate demand, raise systemwide costs, contribute to the emission of local pol- lutants and greenhouse gases, and stimulate environmentally damaging behaviors, such as overpumping of groundwater. Nor is the "merit good" argument for subsidizing electricity--that a good should be subsidized because of its importance to individual well-being--very strong in comparison to other competing sectors (such as water and sanitation, health, and education, for example). Given the regressivity of residential subsidies in Mexico at present, one cannot justify the existing system as a pro-poor policy intervention. Since the late 1990s, Mexico's social policy has moved more and more toward reliance on targeted, means-tested programs, but the electricity sector has yet to adopt this approach. It is clear from the simulations pre- sented here that it would be possible to improve the targeting of electricity subsidies to the poor through tariff structure changes, possibly in combination with cost reduction. Using means testing based on the Oportunidades program in any new tariff or subsidy structure would seem to be a logical extension of developments in other sectors in Mexico. It is important to consider whether the goal of subsidizing electricity for the poor is to keep the price of electricity low for the poor (as prices for other customers rise) or whether electricity is to be used as a more general social policy vehicle for delivering subsidies to the poor. If the latter is the policy goal, then it is important to recognize that electricity service is a suboptimal choice of subsidy delivery instrument. First, some poor households (less than 4 percent of the overall Mexican population) do not have electricity connections and are thus excluded from any subsidy delivered through electricity utilities. Second, electric- ity service does not have all the characteristics of a good targeting mechanism--it is an undifferentiated good that is used by non-poor households as well as the poor. Third, because there is not a perfect correlation between electricity consumption and income, it is difficult to use electricity bills alone to identify the poor and deliver subsidies. Straight quantity-targeted subsidies delivered through the tariff structure will not be well targeted. Finally, Mexico has many other social programs that are better targeted to the poor than electricity subsidies. If the goal is to increase transfers to the poor, it would be better to use these mechanisms for reaching the poor. The idea that other social transfer programs could do a better job of reaching the poor than electricity subsidies is supported by inter- national experience. The Coady, Grosh, and Hoddinott (2003) study of the targeting of social policy instruments, cited above, shows that social funds, cash transfer programs, and work programs are much more likely than electricity subsidies to do a good job of target- ing the poor. APPENDIXES APPENDIX A Evolution of Residential Electricity Subsidies and Tariffs in Mexico 51 52 World Bank Working Paper Period Subsidies Tariff Structures 1964­70 Tariff set to cover costs. Increased effi- Tariff 1: Fixed charge + single consump- ciency in the sector cancels out low tion charge. levels of inflation. 1971­73 Subsidies begin, as pace of inflation increases. 1973 Subsidies formally introduced into the Tariff 1: tariff structure. Fixed charge eliminated and replaced by small minimum charge. Three block IBTs introduced with subsidies in first two blocks (0­50, 50­100, 100+). 1974 Summer subsidies introduced. Tariff 1A: Extended second subsidized block of Tariff 1 for areas where the tem- perature is > 25 °C for four months of the year. 1988 Additional summer subsidies introduced. Tariffs 1B­1C Extensions of Tariff 1A for zones with temps > 28 °C and > 30 °C, respectively. 1990 Tariffs adjusted, new summer subsidy Tariffs 1, 1A, 1B, 1C changed: introduced. Blocks of IBT extended. Prices of last block increased. Tariff 1D introduced for zones with temps > 31 °C. 1995 New summer subsidy introduced. Tariff 1E introduced for zones with temps > 32 °C. Feb 2002 Tariff review. Three billing schedules introduced for each tariff area: low (75 percent of customers), moderate (25 percent of customers), and excessive con- sumption (5 percent of customers). Those with moderate consumption pay more than cost in the third block of the IBT. Those with excessive consumption pay a flat DAC tariff, which is above long-run marginal cost. April 2002 New summer subsidy introduced. Tariff 1F introduced for zones with temps > 33 °C. APPENDIX B Residential Electricity Tariffs in 2006 53 54 World Bank Working Paper Domestic Electricity Tariffs (Average May­October 2006) Season: Tariff Summer/ Range Tariff Type Non-summer Tariff Blocks kWh per kWh Básico 1­75 0.610 <140 Intermedio 76­140 0.724 1 Básico 1­75 0.610 >140 Intermedio 76­125 1.006 Excedente 126­ 2.126 Básico 1­100 0.530 <150 Intermedio 101­150 0.631 S Básico 1­100 0.530 >150 Intermedio 101­150 0.796 Excedente 151­ 2.126 1A Básico 1­75 0.610 <150 Intermedio 76­150 0.724 NS Básico 1­75 0.610 >150 Intermedio 76­125 1.006 Excedente 126­ 2.126 Básico 1­125 0.530 <225 Intermedio 126­225 0.631 S Básico 1­125 0.530 >225 Intermedio 126­200 0.796 Excedente 201­ 2.126 B Básico 1­75 0.610 <175 Intermedio 76­175 0.724 NS Básico 1­75 0.610 >175 Intermedio 76­150 1.006 Excedente 151­ 2.126 Básico 1­150 0.530 <300 Intermedio 151­300 0.631 S Básico 1­150 0.530 >300 Intermedio 151­450 0.796 Excedente 451­ 2.126 C Básico 1­75 0.610 <175 Intermedio 76­175 0.724 NS Básico 1­75 0.610 >175 Intermedio 76­150 1.006 Excedente 151­ 2.126 Residential Electricity Subsidies in Mexico 55 Domestic Electricity Tariffs (Average May­October 2006) (Continued) Season: Tariff Summer/ Range Tariff Type Non-summer Tariff Blocks kWh per kWh Básico 1­175 0.530 <400 Intermedio 176­400 0.631 S Básico 1­175 0.530 >400 Intermedio 176­600 0.796 Excedente 601­ 2.126 D Básico 1­75 1­75 0.610 <200 Intermedio 76­175 0.724 NS Básico 1­75 1­75 0.610 >200 Intermedio 76­175 76­175 1.006 Excedente 176­ 2.126 Básico 1­300 0.436 <750 Intermedio 301­750 0.561 S Básico 1­300 0.436 >750 Intermedio 301­900 0.718 Excedente 901­ 2.126 E Básico 1­75 0.610 <250 Intermedio 76­200 76­200 0.724 Excedente 201­250 2.126 NS Básico 1­75 1­75 0.610 >250 Intermedio 76­200 76­200 1.006 Excedente 201­ 2.126 Básico 1­300 1­300 0.436 <1200 Excedente 301­1200 0.561 Básico 1­300 1­300 0.436 S Intermedio Bajo 301­1,200 301­1200 0.718 >1200 Intermedio Alto 1201­2500 1.338 Excedente 2500­ 2.126 F Básico 1­75 1­75 0.610 <250 Intermedio 76­200 76­200 0.724 Excedente 201­250 2.126 NS Básico 1­75 1­75 0.610 >250 Intermedio 76­200 76­200 1.006 Excedente 201­ 2.126 Source: CFE. APPENDIX C Tariff Reclassifications, 2002­06 Reclassifications in 2002 Tariff Number of Cost From To Customers (million pesos) 1 1A 20,577 0.5 1 1B 1,504 0.2 1 1C 448 0.1 1 1D 1A 1B 33,991 4.7 1A 1C 1A 1D 1B 1C 639,560 332.0 1B 1D 3,161 0.3 1C 1D 11,076 1.3 1D 1E 34,822 12.9 1E 1F 564,208 182.9 Sum 1,309,347 534.9 57 58 World Bank Working Paper Reclassifications in 2003 Tariff Number of Cost From To Customers (million pesos) 1 1A 16,219 0.2 1 1B 1 1C 1 1D 1A 1B 421,558 70.4 1A 1C 3,086 0.2 1A 1D 1B 1C 876,783 220.0 1B 1D 1C 1D 275,134 63.4 1D 1E 55,846 20.2 1E 1F Sum 1,648,626 374.4 Reclassifications in 2004 Tariff Number of Cost From To Customers (million pesos) 1 1A 1 1B 1 1C 1 1D 1A 1B 2,485 0.4 1A 1C 1A 1D 1B 1C 5,990 2.0 1B 1D 407 0.1 1C 1D 25,084 5.4 1D 1E 161,740 100.0 1E 1F 14,275 27.3 Sum 209,981 135.2 Residential Electricity Subsidies in Mexico 59 Reclassifications in 2005 Tariff Number of Cost From To Customers (million pesos) 1 1A 77,452 8.9 1 1B 9,390 1.7 1 1C 956 0.1 1 1D 4,985 0.9 1A 1B 25,958 7.5 1A 1C 1A 1D 2,398 0.7 1B 1C 103,188 42.2 1B 1D 1C 1D 16,340 3.6 1D 1E 3,246 1.1 1E 1F Sum 243,913 66.7 Reclassifications in 2006 Tariff Number of Cost From To Customers (million pesos) 1 1A 23,775 5.0 1 1B 1 1C 1 1D 1A 1B 1A 1C 1A 1D 1B 1C 40,605 25.0 1B 1D 640 0.4 1C 1D 421 0.1 1D 1E 5,814 4.0 1E 1F Sum 71,255 34.5 60 World Bank Working Paper Reclassifications in 2002­06 Tariff Number of Cost From To Customers (million pesos) 1 1A 138,023 14.6 1 1B 10,894 1.9 1 1C 1,404 0.2 1 1D 4,985 0.9 1A 1B 483,992 83.0 1A 1C 3,086 0.2 1A 1D 2,398 0.7 1B 1C 1,666,126 621.2 1B 1D 4,208 0.8 1C 1D 328,055 73.8 1D 1E 261,468 138.2 1E 1F 578,483 210.2 Sum 3,483,122 1,145.7 Source: Aburto 2007, using data provided by CFE. APPENDIX D Analysis of Subsidies Using CFE Data Methodology for Measuring Subsidies and Subsidy Distribution The methodology used to measure subsidies and to estimate subsidy distribution across consumption deciles is, with minor exceptions,29 the same as was used in the 2003 analy- sis of Mexican residential electricity subsidies commissioned by the World Bank. Data The following data from 2005 were used for this study: The32residentialtariffschedules(fromtheMexicoNationalElectricityCommission [CFE] web page); Aggregate annual data on the number of customers, sales, revenues (minus taxes and fees), and subsidies for each of the seven basic tariffs (1, 1A, 1B, 1C, 1D, 1E, 1F) and for the six High Consumption Residential Tariffs (DAC) grouped together (provided by the Finance Ministry); Histogramsofelectricityconsumersforeachbasictariff,withintervaldistributions of customers and sales (provided by the Finance Ministry); and Monthlysales andrevenues by tariff (from the CFE web page). 29. In 2003, Tariff 1F and the DAC tariffs had just been established and there was insufficient informa- tion on them to represent them adequately. Therefore, Tariffs 1E and 1F were worked together as a single tariff. For the DAC tariffs, estimators were used to infer their impact upon each basic tariff. In the current study Tariff 1F and the DAC tariffs are fully analyzed. A second difference between the previous study and this study is that the 2003 study included households without electricity in the first consumption decile. The current study examines only the distribution of subsidies among customers with electricity connections. 61 62 World Bank Working Paper The monthly sales and revenue information was used to turn the annual histograms for all of the basic tariffs except Tariff 1 into seasonal histograms (summer and winter season).30 The basis for the subsidy analysis was, thus, 13 histograms corresponding to the seven basic tariffs. No separate histogram for Central Light and Power (LFC) customers was available. However, since consumption and price statistics for LFC customers are very similar to Tar- iff 1 customers from CFE, the assumption was made that CFE's histogram for Tariff 1 could be used for LFC customers as well. These sources of information are not entirely consistent. There are several reasons for this, including nontechnical losses, billing cycles, and errors. Nevertheless, differences are very small and have no significant effect on the analysis. Building Subsidy Histograms The histograms provided by the Finance Ministry had information on only the number of customers and sales for each histogram interval. To estimate consumption, revenues, and subsidies by interval in each of the 13 histograms, the following steps were performed: 1. Interval consumption midpoints were used to compute total annual sales from the bottom up. These sales estimates differed somewhat from true sales, due to the skew- ness of the customers and sales distributions in the histogram. An adjustment coef- ficient was therefore applied to the midpoints in order to reproduce actual sales (which are known). These new interval "midpoints" are good estimates of actual average consumption per interval. 2. Average monthly bills per season were determined for each interval based on this estimate of average consumption and the known tariff schedules. 3. Average customer bills multiplied by the number of customers per interval yields (sea- sonal) revenues for that particular interval. The sum of all intervals equals total rev- enues for that histogram. This total was compared with the official statistic for revenues and, again, a skewness adjustment factor was applied to make them coincide. 4. The last step was to estimate subsidies by interval in each histogram. Subsidies were defined as the difference between cost and revenue.31 It is generally accepted that the cost functions for residential electricity follow the structure of the DAC tariffs (that is, consist of a fixed cost plus a volumetric cost component), but that the DAC tar- iffs are above cost. To estimate the cost function for each tariff group, first, the DAC fixed charge was adopted as the estimate of fixed cost. The energy cost (the variable portion of the cost function) was then independently estimated for each of the 13 histograms. The DAC volumetric charge was used as the initial estimate of variable cost. This charge was then lowered until the histogram reproduced the known total value for subsidies for each particular tariff and season.32 30. Tariff 1 does not vary by season and the consumption pattern is fairly consistent across the year. 31. CFE accounting costs were used as costs for both CFE and LFC customers, because LFC account- ing costs are elevated by inefficiency. 32. The energy costs that make histogram-generated subsidies match total reported subsidies for the seven basic tariffs were found to be 70 percent to 80 percent of the energy charges for the DAC tariffs. This is consistent with estimates from the Treasury Department regarding the relationship between actual costs and the levels of the DAC tariffs' charges. Residential Electricity Subsidies in Mexico63 Adding DAC Customers These histograms did not include DAC customers; they only included customers in the basic and intermediate consumption schedules of each tariff. Information on revenue and subsidies was not available for DAC customers by tariff, but rather for all DAC customers as a group. It was thus necessary to add DAC customers to the histograms for each tariff. To do this, assumptions had to be made about the distribution of DAC customers across basic tariff classes: the numbers of households paying DAC tariffs in each basic tar- iff group was estimated in proportion to the overall number of households and the aver- age consumption per customer, for each basic tariff. The estimates range from 2.3 percent of customers paying DAC in the Tariff 1 zone to 0.81 percent of customers for Tariff 1F. Next, since there are six regional DAC tariffs, and these regions do not correspond completely with the basic tariff zones, it was necessary to produce an estimate of the aver- age DAC tariff applied to customers in each tariff group. To do this, the largest 50 cities in the country were examined, and their basic tariffs specified. They were then grouped by basic tariff and the applicable DAC tariffs weighted by population for those large cities involved. This generated the average DAC tariff to be applied for each basic tariff group. With this information, it was possible to compute aggregate variables (sales, revenues) for DAC customers in each basic tariff and to create "basic tariff + DAC" histograms for each tariff and each season. Throughout this process, statistical tests were performed to assess the accuracy of the information in the resulting histograms. Tests consisted of computing average consump- tion, and annual revenues and subsidies, and comparing these partial results with actual figures available from CFE's annual statistics. The statistical errors for global statistics were 0.2 percent for annual revenues, 0.3 percent for annual subsidies, and less than 0.1 percent both for annual sales and annual costs. Determining Deciles and Subsidies Accruing to Each Decile The histograms provided by the Finance Ministry had variable numbers of intervals. To permit the analysis of subsidy by consumption decile, the original histograms had to be modified into histograms with only 10 intervals, one per decile. For Tariffs 1A to 1F, the summer and winter "basic tariff + DAC" histograms were added (a weighted sum) to establish annual histograms. These histograms together with the Tariff 1 histogram constitute the seven annual histograms over which deciles of cus- tomers were identified. The consumption intervals so defined were then applied back to seasonal histograms for each basic tariff. In most cases it was necessary to interpolate the Kilowatt-hour (kWh) breakpoints for each decile, because the intervals in the original histograms did not necessarily correspond to the decile breakpoints. Linear interpolation was used to establish the boundaries. For each decile, interval average values were computed for households and sales. This was followed by the computation of revenues, costs, and subsidies for each interval. In each case the end result was an abbreviated histogram with only 10 intervals corresponding to customer deciles. APPENDIX E Analysis of Subsidies Using ENIGH Data S ixty-five percent of households surveyed answered a survey question about electric- ity expenditure; most of the other 35 percent have electricity connections but either failed to report the data or are not paying for the electricity they receive. For the 65 percent of households that reported electricity expenditures, electricity consumption was estimated by applying the tariff applicable to each household (based on information received from the Mexico National Electricity Commission [CFE]). It was necessary to eliminate 7 percent of the Mexico National Household Income and Expenditure Survey [ENIGH] households from this analysis because CFE listed two tariffs for their localities and it was unclear which applied to the surveyed households. In the analysis of the remain- ing households, the value-added tax was subtracted from reported expenditure before calculating electricity use. A benefit incidence analysis of the Residential Electricity Schedule (RES) entails a number of methodological and data difficulties, which include the following. How Much of the Subsidy Benefits Paying Consumers Compared to Nonpaying Consumers and Producers? It is not clear what proportion of the residential electricity subsidy should be imputed as benefits to paying electricity consumers. The subsidy compensates consumers for the com- paratively high costs of electricity production and distribution in Mexico (Tables 2.5 and 2.6). There is evidence that these excess costs derive in part from production inefficiencies, electricity leakages, and high labor costs associated with low productivity and generous worker benefits (Hernández 2007; Guerrero and others 2007; Carreón-Rodríguez, Jimenez, and Rosellon 2007; World Bank 2005). Part of the subsidy thus benefits consumers who are 65 66 World Bank Working Paper illegally connected to the network, not paying for all or any of their electricity consumption, and electricity sector workers and pensioners. Unfortunately, the sector is opaque in its cost structure, and the total RES is reported in official documents without any disaggregation or explanation of estimation methods and sources, so the estimate of the incidence of resi- dential electricity subsidy is based on full costs. Tariff Structure The RES tariff structure is complex, varying by consumption level, locality (temperature), and season (summer/non-summer), with a wide range of cost per kWh (0.4 to 2.7 Mexican pesos [MP] on average in 2006, Appendix B). There are currently seven residential elec- tricity subsidy tariff types, each structured into three to five increasing block tariff (IBTs), plus a volume differentiated tariff (VDT) for high-volume electricity consumers: High Consumption Residential Tariff (DAC). 1. Tariff 1 is the principal tariff type (largest coverage), and has no seasonal variation. The other six tariffs, 1A to 1F, allocate subsidies for localities with high summer tem- peratures, increasing by alphabetic order both in temperature and subsidy rates. Each of the latter has different summer and non-summer schedules. Non-summer tariffs per Kilowatt-hour (kWh) in Tariffs 1A to 1F coincide with Tariff 1, but the ceilings of the intermediate and upper blocks increase for higher-temperature tar- iffs (from 140 kWh per month in Tariff 1, up to 250 kWh in Tariff 1F). In the case of summer tariffs, the tariffs per kWh are lower for the basic and intermediate blocks, andtheceilingsincreasemuchmoresharply,especiallyforthefourhigher-temperature tariffs (1C to 1F), with a subsidized upper-intermediate block reaching 2500 kWh per month in 1F. 2. For each of the seven tariff types, the DAC tariff comes into force when a moving average of consumption over the previous year reaches beyond a specific threshold, which increases from 250 for Tariff 1 to 2500 for Tariff 1F (Appendix B). This tariff has a small fixed charge (about 60 MP in 2006), plus a constant tariff per kWh (2.1 to 2.7 MP per kWh), which varies regionally (Baja California, Baja California Sur, North West, North-North East, South & Peninsula, and Center),33 and for the two Baja California peninsula states, also seasonally (with higher rather than lower summer tariffs). These tariffs exceed full marginal production costs by at least 20 percent on Ministry of Finance (SHCP) estimates, and are reported in house- hold electricity bills as fully unsubsidized. Household-level Data on Electricity Spending and Estimation of Consumption The only available source of household data to analyze the incidence of the RES by income groups is the ENIGH income and expenditure survey. We use the most recent available ENIGH survey, for 2006. This reports high-quality household income data, and household 33. The classification of some localities into regions does not appear to be determined by their geographic locations. Residential Electricity Subsidies in Mexico67 expenditures on electricity, but does not report kWh consumption. We have used the tariff structure to deduce the latter from the former, using the algorithm presented in Appendix C. In this imputation we make the following assumptions and information: 1. It is only possible to use the households in ENIGH that report electricity spending, representing 65 percent of the total ENIGH sample. Most of the other 35 percent have electricity connections, but it is not known how many of these fail to report because they are connected illegally (and thus fully, but illegally subsidized), and how many where paying consumers who failed to report this data. 2. It is assumed that households correctly report monthly electricity spending to ENIGH interviewers, despite the fact that most households are billed bimonthly. 3. Since electricity bills include a 15 percent value-added tax, this tax is subtracted from electricity spending reported by households so that pretax expenditures are used in the noted imputation algorithms. 4. It is assumed that electricity spending reported in ENIGH corresponds to summer tariffs. ENIGH interviews during August­November. Households are asked about consumption in the previous month (July­October), and since they mostly pay their bills bimonthly and with a monthly lag, we assume that households are reporting consumption for May/June­August/September, which corresponds to the summer months used by CFE in the definition of tariffs. Average tariffs as reported by CFE for May­October 2006 are used. 5. Tariff types are defined at the locality level, so to assign households to their corre- sponding tariff type we obtained from INEGI the locality codes corresponding to all ENIGH households, and from CFE (through Hacienda) the tariff types corre- sponding to each of these localities. For some unexplained reason, the CFE list reported multiple tariffs for some localities, corresponding to 7 percent of ENIGH households. These were eliminated from the analysis. 6. For each tariff type, we assume households are paying DAC tariffs when reported spending is above the spending threshold implied by the DAC kWh threshold priced according to the corresponding DAC tariff (average May­October 2006). 7. ENIGH is also used to estimate the incidence of the Oportunidades Energéticas pay- ments, because this source reports Oportunidades transfers. Distribution by Income-ordered Household Deciles and Quintiles We present the results of the distributive analysis by household deciles and quintiles ordered by total current income per capita. Even by quintiles, the averages reported for Tariffs 1E and 1F for the first quintile should be interpreted with care because they are based on small numbers of cases. Cost per kWh To estimate per-household subsidies, an average cost per kWh of 2.19 pesos is assumed, derived from official (SHCP and CFE) reports of the total residential electricity subsidy by CFE, electricity sales for domestic use, and average electricity price to domestic users (Table E.1). For consistency, this cost is applied to all households, despite the fact that 68 World Bank Working Paper reported average cost per Table E.1. Cost per KWh (2006) kWh for LFC is significantly CFE LFC Total higher (3.82 pesos). This cost Average Cost per kWh (MP) 2.19 3.82 2.42 per kWh is applied for all Average Price per kWh (MP) 0.96 1.11 0.98 households independently of Price/Cost ratio 44% 29% 41% tariff type, geographic loca- Total Subsidy (million MP) 46,683 17,288 63,971 tion, size of locality, season, Total Revenue (million MP) 36,637 7,082 43,719 or level of consumption. Of Total Cost (million MP) 83,320 24,370 107,690 these variables, geography and Total Sales (MWh million) 38,072 6,381 44,452 size of localities must imply significant cost differences, Source: CFE, SHCP. but the cost data are lacking to take these into account. Given the strong correlation of poverty and population dispersion in Mexico, ignoring the implicit subsidy to small and remote localities leaves out an important source of progressiv- ity (pro-poor) in the electricity subsidy. On the other hand, ignoring geographic cost differ- ences in the incidence analysis might be justified if the equity objective of electricity subsidies is defined in terms of equal access to electricity, rather than equal financing per kWh. Table E.2 compares aggregate electricity subsidy, consumption, and spending data by tariff type obtained from the ENIGH survey with the corresponding data from administrative records for all households served by the national electricity network, reported by CFE. The distribution of the subsidy, consumption, and spending by tariff types of ENIGH is broadly consistent with administrative data. There is a very significant difference, however, in average spending (and thus consumption and subsidy) levels per household. (Note also that the average DAC price is slightly below cost according to administrative data, but above cost in ENIGH). This may reflect misreporting by households in ENIGH of bimonthly bills as monthly spending, but it may also reflect non-revenue consumption, reported in ENIGH but not in the administrative records. Table E.2. Residential Electricity Subsidies, Consumption, and Spending by Tariff Type: Administrative and Household Survey Data Average Subsidy Electricity Consumption Electricity Spending Users Price Tariff/ Million MP/HH kWh/HH Million MP/HH Million MP per Subsidy MP Distribution (monthly) MWh Distribution (monthly) MP Distribution (monthly) Households Distribution KWh Administrative Data (CFE 2006) Total RES 63,971 100.0% 202 44,452,408 100.0% 141 43,719 100.0% 138 26.348 100.0% 0.984 1 24,893 38.9% 143 16,146,697 36.3% 93 14,224 32.5% 82 14.493 55.0% 0.881 1A 2,896 4.5% 160 1,811,376 4.1% 100 1,493 3.4% 82 1.508 5.7% 0.824 1B 7,404 11.6% 195 4,709,015 10.6% 124 4,004 9.2% 106 3.158 12.0% 0.850 1C 13,332 20.8% 277 8,709,386 19.6% 181 7,767 17.8% 161 4.018 15.3% 0.892 1D 3,227 5.0% 331 2,038,791 4.6% 209 1,713 3.9% 176 0.813 3.1% 0.840 1E 5,659 8.8% 470 3,430,456 7.7% 285 2,651 6.1% 220 1.004 3.8% 0.773 1F 6,423 10.0% 680 3,862,778 8.7% 409 2,935 6.7% 311 0.787 3.0% 0.760 DAC 137 0.2% 20 3,743,909 8.4% 550 8,933 20.4% 1312 0.568 2.2% 2.386 OEC (2007) 2,919 4.4%* 50 4.864 Survey Data (ENIGH 2006) Total RES 62,007 100.0% 319 48,393,315 100.0% 249 55,230 100.0% 285 16.173 100.0% 1.141 1 20,046 32.3% 176 17,005,975 35.1% 149 21,153 38.3% 186 9.494 58.7% 1.244 1A 1,972 3.2% 273 1,468,338 3.0% 203 1,585 2.9% 219 0.603 3.7% 1.080 1B 6,779 10.9% 279 4,981,331 10.3% 205 5,289 9.6% 217 2.027 12.5% 1.062 1C 17,287 27.9% 644 11,082,745 22.9% 413 9,562 17.3% 356 2.238 13.8% 0.863 1D 4,153 6.7% 786 2,593,447 5.4% 491 2,130 3.9% 403 0.440 2.7% 0.821 1E 4,021 6.5% 1098 2,268,973 4.7% 620 1,475 2.7% 403 0.305 1.9% 0.650 1F 8,669 14.0% 1618 5,094,962 10.5% 951 3,674 6.7% 686 0.446 2.8% 0.721 DAC -919 -1.4% -124 3,897,543 8.1% 524 10,361 18.8% 1393 0.620 3.8% 2.658 OEC (2007) 2,348 3.6%* 50 3.914 *Of total domestic electricity subsidies. Source: CFE, ENIGH (2006). APPENDIX F Tariff Reform Scenarios Part 1. Methodology Used for the Simulations (CFE Data) Predicting Changes in Electricity Consumption An important part of the simulation methodology is predicting the change in electricity consumption that will result when prices are increased. To calculate price elasticity of demand, the Mexico National Electricity Commission (CFE) demand model for residen- tial electricity was used. This model structure (without coefficients) was published by the Ministry of Energy (SENER) in December 2007.34 The demand model has two compo- nents. The first estimates the percentage of homes in Mexico that have electricity--the number of residential users. The second model estimates the energy demand per user: VUt = F (VUt , SATt, CP/Vt, Pt, PREDt, T, VIVUR) -1 Where: VUt -1 = sales per residential users in the period t-1 SATt = coefficient of saturation of users (residential users / homes) CP/Vt = private consumption per household (indicator of disposable income) Pt = average real price for residential electricity PREDt = relative price of electrical appliances T = time VIVUR = variable that measures increases in nontechnical losses in Central Light and Power (LFC) (0 in 2001 with gradual increases up to 1 in 2007). 34. Secretaría de Energía, "Prospectiva del Sector Eléctrico 2007­2016," pp. 167­9. 71 72 World Bank Working Paper This is a double-log model, which ensures constant elasticity. To estimate price elasticity of demand, a database of 30 years was created for each of the variables in the demand model (1977 to 2006). A regression was then run to obtain the coefficients of the model, includ- ing both short-run and long-run elasticity: -0.14 and -0.31, respectively. These estimates are close to those published by Ministry of Finance (SHCP) in 2003 (-0.16 and -0.33, respectively). To project future demand, price elasticity is not the only important variable. Changes in the number of users, increases in income, and reductions in the relative price of electri- cal appliances will also affect demand. For the projection of demand in the Table F.1. Electricity Demand Model simulations, therefore, changes in Explanatory Variables Average these variables were also taken into Annual Rate of Change (last 10 years) account. It was assumed the past rates of growth or reduction over the last 10 sales / household 0.29% years would continue into the future. saturation (customers / households) 0.16% A 10-year period was taken as the base private consumption / household 0.86% for the future projections, assuming real price of electricity 2.73% that the simulations would be imple- electrodomestics' price index -4.92% mented over time (Table F.1). General Approach to Simulations The simulations involved two basic steps. The first was to simulate the application of the new cost functions and tariffs without taking into account likely changes in con- sumption. In the second step, the demand model described above was used to calculate how electricity demand would change for each interval of consumption.35 Based on this information, changes in the distribution of users and of subsidies were calculated. For the scenario involving only cost reductions, prices for users would not change and thus the second step was not necessary. Simulation Details Modified High Consumption Residential Tariff (DAC). The tariffs for the modified DAC simulation were established in collaboration with Dr. Pedro Luna of the Ministry of Finance. It was decided that an average of the summer and winter DAC for the North East region of the country would be used as the basis for the new DAC. The only exception is for users in Tariff 1F, where an average of the summer and winter DAC tariffs for the North East and Baja California were used as the basis for the new DAC for this user group. The DAC tariffs in the simulation were set at 90 percent of these base levels and these tariffs were applied to the top 10 percent of customers in each tariff group. The threshold of the 10th decile was determined from the histograms for each tariff group. These thresholds dif- fered in the summer and in the winter (Table F.2). 35. The CFE data are presented in histograms with summary information for different intervals of consumption. Price changes and consumption changes were thus calculated for each interval of con- sumption rather than for individual customers. Tariff Reform Scenarios 73 Volume Differentiated Tariff Table F.2. Threshold of 10th Decile in kWh per (VDT). The VDT tariff structure was Connection per Month, for Each Tariff Group proposed by the authors. It includes three different tariffs for consumers Tariff Summer Winter with different levels of consumption. 1 190 190 The first tariff subsidizes consumption 1A 200 200 of those consuming less than 100 Kilo- 1B 250 225 watt-hours (kWh) per month. The 1C 450 300 second subsidizes consumption for 1D 600 300 those consuming between 100 kWh 1E 850 400 and 200 kWh. Table F.3 displays 1F 1200 400 the exact tariffs and level of discount applied for each tariff. Table F.3. VDT Simulation Tariff, for Analysis with CFE Data Tariff For Consumers Using Fixed Charge Variable Charge per kWh Basic consumption 0­100 kWh 16.328 1.013 pesos tariff (40% of std tariff) (70% of standard tariff) Intermediate tariff 100­200 kWh 36.737 1.302 pesos (90% of std tariff) (90% of standard tariff) Standard tariff Above 200 kWh 40.819 1.447 pesos Cost Reduction. For the reduced cost scenario, and for the modified DAC plus cost reduction scenario, the fixed and variable components of the tariff and season-specific cost functions were each reduced by 20 percent. For the VDT with cost reduction scenario, however, it was necessary to create a new systemwide cost function because the existing tariff groups disappear in this simulation. The cost function for Tariff 1 was used as the base. That function is: C = 51.024 + 1.809*Q. After applying the 20 percent cost reduction, the cost function became: C = 40.819 + 1.447*Q. Part 2. Methodology Used for the Simulations (ENIGH Data) Predicting Changes in Electricity Consumption To predict the changes in electricity consumption as a result of the simulated tariff changes, the short-run price elasticity estimates are derived from the CFE demand model (see Scott 2007, Annex 40). Details of the Simulations Modified DAC Tariff. For the simulation of the modified DAC tariff, the tariff and threshold for the 10th decile that were used in the simulation with CFE data (see Appen- dix D) where applied to the ENIGH data. Those customers consuming more than the thresholds indicated in Appendix D were spread among different income deciles. 74 World Bank Working Paper Cost Reduction. In the analysis with ENIGH data, the cost function applied only has a variable component (no fixed cost). A 20 percent reduction on CFE's official estimate of average cost (2.19 pesos) is assumed. The resulting average cost is 1.75 pesos per kWh. VDT Tariff with Cost Reduction. The following VDT tariff was applied for this scenario. To the effect of a VDT without cost reductions, the same tariff was applied, but with the standard tariff equal to 2.19 pesos and then the same 80 percent and 63 percent discounts applied to the basic and intermediate tariffs (Table F.4). Table F.4. Assumptions for VDT Tariff Tariff For consumers using Fixed charge Variable charge per kWh Basic consumption tariff 0­100 kWh None 1.10 pesos (63% of standard tariff) Intermediate tariff 100­200 kWh None 1.45 pesos (80% of standard tariff) Standard tariff Above 200 kWh None 1.75 pesos Means-tested Discount with Cost Reduction. For the means-tested discount scenario, most households receive no discount--they pay either 1.75 pesos (in cost reduction sce- nario) or 2.19 pesos per kWh (without cost reduction). Those who qualify for the dis- count receive a discount only on the first 100 kWh of consumption (Table F.5). Two scenarios for discount recipients were tested. In the first, only those households that qual- ify for Oportunidades receive a discount. In the second, all households in the first income decile of ENIGH are also included in the discount recipient group to simulate an improve- ment in the targeting of this means-tested program. Table F.5. Assumptions for Means-tested Discount For consumers Fixed charge Variable charge per kWh Oportunidades 0­100 kWh None 1.10 pesos (or Oportunidades + 100 + None 1.75 pesos first income quintile) Standard tariff All others None 1.75 pesos Bibliography Aburto, José Luis. 2003."Mexico Power Sector Subsidies."Consulting Report for the World Bank. ------. 2007. "Mexico: Electricity Subsidies--Options for Enhancing the Impact on the Poor." Consulting Report for the World Bank. Carreón-Rodriguez, V. G., Armando Jimenez, and Juan Rosellon. 2007."The Mexican Elec- tricity Sector: Economic, Legal and Political Issues."In David G.Victor, ed., The Political Economy of Power Sector Reform: The Experiences of Five Major Developing Countries. Cambridge University Press. Comisón de Integración Energetica Regional (CIER). 2005."Tarifas Eléctricas en los Países de la CIER 2005." Montevideo, Uruguay. http://www.cier.org.uy. Coady, D., M. Grosh, and J. Hoddinott. 2003. "The Targeting of Transfers in Developing Countries: Review of Experience and Lessons."Social Protection Discussion Paper, The World Bank, Washington, D.C. Guerrero, I., L. F. López-Calva, and M. Walton. 2007."The Inequality Trap and its Links to Low Growth in Mexico." Paper presented at World Bank-Harvard Rockefeller Center Conference on Equity and Growth in Mexico. Mexico City, November 27­28, 2006. Hernández,C.2007."La Reforma Cautiva: Inversión,Trabajo y Empresa en el Sector Eléctrico Mexicano." Centro de Investigación para el Desarrollo, AC. International Energy Agency (IEA). 2006."World Energy Outlook 2006."Paris: OECD/IEA. Komives, Kristin, Vivien Foster, Jonathan Halpern, and Quentin Wodon. 2005. "Water, Electricity, and the Poor: Who Benefits from Utility Subsidies?" The World Bank, Washington, D.C. Ministry of Energy (SENER). 2007."Estadísticas de Electricidad 2000­2006."Mexico City, Mexico. 75 76 World Bank Working Paper Samaniego-Breach, Richard. 2005. "Quality, Efficiency, and Expenditure Review of the Electricity Sector of Mexico." Revised Final Report. Consulting Report for the Mexico Infrastructure Public Expenditure Review. Scott, John Roberto. 2004."Eficiencia redistributiva de los programas contra la pobreza en México." Documento de Trabajo 307, División de Economía, CIDE, Mexico City. ------. 2007. "Electricity Subsidies--Options for Enhancing the Impact on the Poor." Background report prepared for the World Bank. World Bank. 2004. "Mexico Public Expenditure Review." Report No. 27894--MX. Wash- ington, D.C. ------.2005."Mexico: Infrastructure Public Expenditure Review (IPER)."Washington,D.C. ------. 2006."Mexico's Competitiveness: Reaching Its Potential." Washington, D.C. ------.2005."Mexico: Income Generation and Social Protection for the Poor."Washington, D.C. ------. 2007. GDF & WDI Central Database. Washington, D.C. http://go.worldbank.org/ T9V3X1EL00. Eco-Audit Environmental Benefits Statement The World Bank is committed to preserving Endangered Forests and natural resources. We print World Bank Working Papers and Country Studies on 100 percent postconsumer recy- cled paper, processed chlorine free. The World Bank has formally agreed to follow the rec- ommended standards for paper usage set by Green Press Initiative--a nonprofit program supporting publishers in using fiber that is not sourced from Endangered Forests. For more information, visit www.greenpressinitiative.org. In 2008, the printing of these books on recycled paper saved the following: Trees* Solid Waste Water Net Greenhouse Gases Total Energy 355 16,663 129,550 31,256 247 mil. * Pounds Gallons Pounds CO2 Equivalent BTUs 40' in height and 6-8" in diameter Residential Elecricity Subsidies in Mexico is part of the World Bank Working Paper series. These papers are published to communicate the results of the Bank's ongoing research and to stimulate public discussion. This report addresses a pressing issue in Mexico's electricity sector--the large and growing subsidies to residential consumers and the highly regressive distribution of these subsidies across income classes. Today, Mexico's complex tariff system includes over 100 different billing possibilities for residential customers, with many increasing block tariffs spread over eight climate-based tariff zones. Only the top 2 percent of residential customers cover the cost of the electricity they receive. Roughly 16 percent of total residential electricity subsidies go to the smallest consumers (the bottom three consumption deciles), while the top three consumption deciles receive 39 percent. This report explains the growth of subsidies, the current distributional incidence, and then uses utility and household survey data to estimate the distributional and fiscal performance of alternative subsidy mechanisms. The goal is to help inform discussion in Mexico about how subsidies might be reduced and redirected toward the poor. The findings also offer lessons for other countries that are planning tariff reforms in their electricity sectors. World Bank Working Papers are available individually or on standing order. Also available online through the World Bank e-Library (www.worldbank.org/elibrary). ISBN 978-0-8213-7884-7 THE WORLD BANK 1818 H Street, NW Washington, DC 20433 USA Telephone: 202 473-1000 Internet: www.worldbank.org SKU 17884 E-mail: feedback@worldbank.org