DIRECTIONS IN DEVELOPMENT -15490, 'Jn. 1011( *I~q rj* q Taxing Ba s by Taxing Goods Pollution Control with P4sumptive Charges f GUNNAR S. ESKELAND SHANTAYANAN DEVARAJAN XE- - gs *>' tl. Atb ',, k _ t < _ _ S.~ ~ ~ ~ ,sh-!.4 ,, .-9. DIRECTIONS IN DEVELOPMENT Taxing Bads by Taxing Goods Pollution Control with Presumptive Charges Gunnar S. Eskeland Shantayanan Devarajan The World Bank Washington, D.C. O 1996 The International Bank for Reconstruction and Developmentt/THE WORILD BANK 1818 H Street, N.W. Washington, D.C. 20433 All rights reserved Manufactured in the United States of America First printing January 1996 The findings, interpretationis, and conclusions expressed in this study are entirely those of the authors and shoultd not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. Gunnar S. Eskeland is senior economist in the World Bank's Policy Research Department, Public Economics Division. Shantayanan Devara an is chief of that division. Libranl ofCoogrcss Catalogiozg-iti-Puiblicati(ln Data Eskeland, Gunnar S. Taxing bads by taxing goods: pollution control with presumptive charges / Gonniar S. Eskeland, Shantayanan Devarajan. p. cm. - (Directions in developmiienit) Includes bibliographical references. ISBN 0-82 1 3-3457-3 1. Environmental impact charges-Mathematical models. 2. Pollution-Economic aspects-Mathematical models. 3. Environmental policy-Economic aspects-Mathematical models. I. Devarajan, Shantavanan, 1954- . 11. Title. Ill. Series: Directions in development (Washington, D.C.) HJ5316.E578 1996 336.2-dc2O 95-49782 CIP Contents Preface v I The Economics of Pollution Control: A Basic Analytical Framework The Textbook Case: An Emission Fee 3 When Monitoring is Costly: Demand Reduction and Technical Controls 3 Presumptive Charges: Problems and Promises 10 Notes 14 2 Pollution from Mobile Sources 15 Using Indirect Instruments: Cleaner Cars, Fewer Trips 15 Market-Based versus Regulatory Demand Management 20 Market-Based and Other Inducements for Cleaner Cars and Fuels 27 Effects on Income Distribution 29 Monitoring Technologies and the Need for Detailed Program Evaluation 30 Notes 33 3 Control of Pollution from Industrial Sources 36 General Economic Policies, Input and Output Taxes, and Regulation 37 Energy Conservation and Interfuel Substitution 39 Exploration of General Equilibrium Effects 46 Notes 53 4 Conclusion 55 Bibliography 57 iii Preface Solving environmental problems, in both developing and industrial countries, appears to be more challenging than merely applying a fee on polluters-a recommendation first put forth in Pigou's Economlics of Wel- fare (1920). One reason why the textbook remedy-levying an emission fee on the polluter-has not been applied is that it requires the monitor- ing of individual emissions, which may be unrealistic in many situa- tions. In fact, most existing pollution control policies employ indirect instruments-those that tax or regulate activities associated with emis- sions-rather than direct instruments based on individual emissions. An important feature of the textbook solution is that the emission fee gives the polluter an incentive to choose the optimal mix of cleaner technologies and reductions in the scale of output or in the use of inputs. When indirect instruments are used, this optimal combination does not come about by itself. For instance, emission standards lead to cleaner cars and have modest monitoring requirements (annual inspections will do), but they do not effectively provide incentives for reducing the number of miles driven. The purpose of this book is to show that indirect instruments designed to reduce the scale of output can be important complementary measures in a cost-effective pollu- tion control program. Examples of such instruments are taxes on out- put or on polluting inputs, called presimiptive because their target is the pollution presumed to be associated with the activity. A combination of the two types of instruments-those that reduce output and those that reduce emissions per unit of output-can mimic fairly well the effect of an optimal emission fee without the latter's monitoring requirements. Specifically, when regulations requiring the adoption of cleaner technologies are in place, the addition of a tax on an input asso- ciated with emissions can yield substantial benefits for a pollution con- trol program. These complementary instruments require very little or no additional resources for monitoring and enforcement. Furthermore, V vi TAXING BADS BY TAXING GOODS schemes that make use of presumptive charges can be (and are likely to be) refined as a country's monitoring capacity develops, so they are not in conflict with "first-best" instruments. Unlike the situation with emission fees, the optimal use of indirect instruments requires knowledge concerning the "cleaner" technolo- gies (to determine the ease with which emissions per unit of output can be reduced) and the sensitivity of demand to prices (as an indica- tion of the ease with which output or input use can be reduced). We rely on case studies of Mexico City, Chile, and Indonesia to illustrate the choice between the two types of instruLments and the interaction between them. A recurring theme throughout the book is that taxation of fuel use can be a powerful indirect instrument for controlling air pollution because of the association between fuel use and emissions. In the case of automobiles, we show that failing to employ gasoline taxes in Mex- ico City would significantly reduce welfare, even when regulatory standards are in place. In the case of point-source pollution, we calcu- late that there is significant potential for altering the fuel mix of indus- tries in Indonesia and Chile by taxing the "dirtier" fuels. Furthermore, for Indonesia we use a general equilibrium model to show that the consequences of such a tax are similar to what is indicated in partial equilibrium models (although somewhat dampened). In sum, we advocate taxing a "bad" (pollution) by taxing goods (fuels) as part of a program to address air pollution when monitoring of emissions is prohibitively expensive. Chapter 1 lays out our basic analytical framework. Chapter 2 treats the case of mobile-source pollution through an examination of gasoline taxes and regulatory policies in Mexico City. Chapter 3 addresses point- source pollution and the potential for altering the fuel mix in industries in Indonesia and Chile, based on firm-level data. A general equilibrium model of Indonesia portrays the economywide consequences of changes in fuel taxes. Chapter 4 contains our concluding remarks. The book distills the main results of a World Bank research project, "Pollution and the Choice of Policy Instruments." We are grateful to Emmanuel Jimenez, who developed the project jointly with Gunnar Eskeland and who provided advice, guidance, and comments. A short- er version of this book was presented at the 50th World Congress of the International Institute for Public Finance, Cambridge, Mass., August 21-25, 1994, and appears in Bovenberg and Cnossen (1995). We thank Charles Ballard, Maureen Cropper, Gordon Hughes, and David Wheeler for helpful comments. Peggy Pender and Hedy Sladovich provided production assistance. 1. The Economics of Pollution Control: A Basic Analytical Framework To illustrate some important principles in the economics of pollution control, we begin with the textbook example of a polluting factory or an automobile-let us call it a plant-that produces a good (say, bread or passenger-kilometers). Because the pollution emitted will harm people other than the plant manager, the marginal social cost of pro- duction exceeds its marginal private cost. The textbook solution to this problem is to levy a fee on pollutant emissions equal to the difference between the two costs (see box 1.1). The imposition of the emission fee gives the plant manager an incentive to reduce emissions in two ways: by adopting cleaner technologies and by reducing output. Moreover, he will choose a combination of cleaner technologies and reduced out- put that is optimal not only for himself but also for society as a whole. Note that the policymaker need not know what this optimal combina- tion is-just as the purchaser of bread need not know the optimal com- bination of inputs required to produce it. To make an emission fee work, however, the government has to mon- itor the plant's emissions. For many pollution problems, especially those caused by a large number of polluters, this is virtually impossi- ble. In this case, the government has two alternatives: tax the good pro- duced (or used) by the polluter, or ask the polluter to adopt a cleaner technology. Below we show that each method contributes toward achieving the benefits of an emission fee, and that a good combination of the two can mimic the effects of an emission fee reasonably well. In this situation, knowledge about control technologies and elasticities of demand for the polluting good is valuable to the policymaker. (Such information is redundant if emissions can be monitored; polluters will use their knowledge-and will volunteer the best response.) The issue treated here is distinct from another that is popular in the literature: the use of a market in pollution quotas or tradable permits. The use of emission quotas or permits, whether tradable or not, I 2 TAXING BADS 13Y TAXING GOODS Box 1.1 Behind the Textbook: Why Intervention Is Necessary Under certain conditions, an undisturbed market will induce firms to use inputs and supply outputs efficiently. But the conditions under which the market, without an intervening authority, will induce a firm to contain pol- lution efficiently are rare-mainly because the benefits of reduced residuals are public, shared by many. The Coase theorem states that a firm will control pollution optimally if it negotiates with the beneficiaries of pollution control, irrespective of whether the firm initially holds the right to pollute or, alternatively, the vic- tims hold the right to a clean environment. Let us first examine the situation in which polluters hold the right to pollute and victims must "bribe" pol- luters in order to reduce pollution. If the "victims" are numerous (as in Mexico City, with 19 million people, or Greater Jakarta, with 15 million), a voluntary club to purchase emission reductions is likely to suffer from the free-rider problem: nonmembers as well as members enjoy the benefits of the emission reductions. Then the purchases made by a voluntary club will be backed by few, and pollution reduction will be suboptimal. The situation is not much better if polluters initially hold no right to pollute but have to buy emission permits from the victims of pollution. An individual holding such rights will sell his with a light heart (even at a low price), since he will suffer only a minuscule share of the consequences in terms of pollution increases. Others will also suffer, without sharing in the revenues from this transaction. In both cases it is the "publicness" of the environmental good that makes voluntary negotiations so impotent, and in both cases applica- tion of authority can solve the problem. If we view the municipality as a citizen-minded club in which member- ship is compulsory, the pollution control agency can be seen as its repre- sentative, with power to negotiate with polluters. In that situation the Coase theorem ensures efficient pollution control and optimal pollution levels, with fees or permits that reflect a "social contract." requires the pollution control agency to monitor individual firms' emissions. Hence, if monitoring of emissions is difficult, the problem cannot be solved by allowing trade in quotas. Nevertheless, only under certain conditions will problems with monitoring make the use of market forces in a control program a poor proposition. Note further that we are concerned here with cost-efficient pollution control; that is, we do not attempt to quantify the benefits from pollution reduction. Rather, we take the targeted level of pollution reductions as given and look for the least costly means of achieving it. (Benefit estimates can be useful as priority weights in cost-effectiveness analysis when multiple THE ECONOMICS OF POLLUTION CONTROL 3 pollutants are addressed. Box 1.2, later in the chapter, presents a sum- mary of such estimates.) The Textbook Case: An Emission Fee Figure 1.1 illustrates the case of a polluting firm that produces a good whose demand curve is given by p = d(q). PMC is the private marginal cost of production, and SMC is the social marginal cost. The gap between PMC and SMC is the result of emissions associated with production, which affect the general public, not just each user according to his or her own emissions. The classic examples are a factory that emits toxic wastes or an automobile that sends fumes into the atmosphere when burning gasoline. We will use the familiar example of the automobile. The problem with the situation in figure 1.1a, showing the case in which there is no intervention to reduce pollution, is that q*, the point at which the automobile owner is in equilibrium, is not a social opti- mum. Consumer surplus, which we will take as the measure of social welfare, is the area under the demand curve that lies above the SMC curve (shaded area A), less area B. By reducing production (moving q* to the left), we can reduce area B and so increase consumer surplus. Therefore q* is not a social optimum. As noted earlier, the standard solution to this problem is to levy a fee on emissions. When facing such a fee, the motorist has two choices: he can drive less and reduce total emissions, or he can choose technical solutions, such as a catalytic converter, that will reduce emissions per liter of gasoline. (In both cases the polluter will pay fees for the emis- sions that remain.) For any given emission fee, we can be sure the motorist will choose the socially optimal mix of these two ways of reducing emissions. To the extent that he reduces his emissions per liter of gasoline consumed, the external costs (the excess of SMC over PMC) are lowered. However, the catalytic converter would also raise his PMC. If the emission fee is set to reflect the difference between PMC and SMC, the resulting equilibrium will be that in figure 1.1b. (We assume here that tax revenues are transferred to consumers as lump-sum income.) Shaded area A has grown, while B has disap- peared, so that social welfare, A - B, has increased.' When Monitoring is Costly: Demand Reduction and Technical Controls Consider now the case in which it is difficult to monitor individual emissions and therefore difficult to levy an emission fee. Is it possible 4 TAXING BADS BY TAXING GOODS Figure 1.1 A Polluting Firm: Four Scenarios (a) No intervention (b) A pollution tax Consumer surplus (A - B) (or output tax and abatement) Consumer surplus = A Price Price (j)) (P) SMC -------SMC Demand-PMC Quantity Quantity (c) Pollution tax on output (d) Mandated abatement Consumer surplus = A Consumer surplus = (A - B) Price Price (P) (P) A \SMC A-M-------------SMC '\ PMC x IMC' - . - - <- - - -PMC Quantity Quantity to achieve some or all the benefits of an emission fee by using other means to curtail demand or induce technical controls? The first possibility is illustrated in figure 1.1c, where we have imposed a tax on output (or on a variable input such as gasoline) equal to the difference between social and private marginal costs. The tax induces consumers to reduce their consumption by manipulating the price that they face.2 Consumer surplus, in this case, will be area A. Soci- ety has made net gains equal to B in figure l.la by eliminating the con- sumption for which marginal social costs exceeded marginal social ben- efits. Note that the social savings from such a strategy (area B in figure 1.1a) depend crucially on the slope of the demand curve: the steeper the THE ECONOMICS OF POLLUTION CONTROL 5 curve, the smaller the gains from demand-management instruments such as input and output taxes. However, the size of the optimal tax is independent of the slope of the demand curve whenever the difference between SMC and PMC is constant. In figure 1.1c we assume that a tax on a commodity such as gasoline would suppress demand in a "neutral" fashion-not changing the amount of pollution per liter of gasoline consumed. In figure 1.1b we assume that an emission control technology, such as a catalytic con- verter, is available. The application of the technology would increase the private marginal costs from PMC to PMC' but would also reduce pollution per unit so much as to position the new social marginal cost curve, SMC, below the original one. Figure I.Id displays what happens if the adoption of such a tech- nology is induced directly-for instance, by regulation. First, as social marginal costs are shifted downward, area A expands and area B shrinks. Second, since consumers now face a private marginal cost that includes the costs of catalytic converters, consumption is also reduced (in comparison with figure 1.1a), and so area B is further reduced. Con- sumer surplus (A - B) will necessarily be greater than in the uncon- trolled case. Finally, consider what happens when mandated abatement is com- bined with a gasoline tax that is optimal under a specific abatement technology. We introduce a tax to suppress the consumption that is excessive, taking into account the reduced emission coefficient. The new equilibrium is identical to that achieved by an emission tax in fig- ure 1.1b. In comparison with figure 1.1c, we have introduced man- dated abatement to make consumption cleaner. Taxes on the polluting good are therefore lower, and we can benefit from additional con- sumption. Note that area A is bigger and the gasoline tax per liter of gasoline is smaller. However, the tax rate per unit of pollution is the same in figures 1.1a and c; the difference in tax rates per liter merely reflects the difference in emissions per liter. When a polluter can "go in two different directions"-abatement or reduced activity levels-to reduce pollution, the problem is best described in three dimensions. Figure 1.2 portrays the cost structure in terms of private marginal costs and social marginal costs. The two curves along the output line are traditional marginal cost curves with no abatement applied: the PMC curve, with no inclusion of the dam- ages caused by pollution, and the SMC curve, including the external costs of the pollution. Along the abatement axis, we can see that pri- vate marginal costs increase monotonously, while the social marginal costs first decrease, then increase. The reason is that abatement is effec- 6 TAXING BADS BY TAXING GOODS Figure 1.2 Determination of Abatement A i Cost, price i _ _ __ _ _ __ _ __ _ _ __ _ __ _ / >expansion Output tive in reducing emissions, and initially at such a rate that the reduced damages are more than sufficient to justify the private costs of abate- ment. When the SMC surface eventually turns upward, it means that zero pollution is "too little"-because reductions eventually cost too much in terms of abatement costs. The "optimal expansion path," depicting the projection vertically under the "valley bottom" of social marginal costs, describes the socially optimal way of producing out- put, for a range of output levels. In figure 1.3 the cost structure is brought together with demand. The traditional demand curve, now a "demand surface," is sliced vertical- ly along the lines of zero abatement and along the optimal expansion THE ECONOMICS OF POLLUTION CONTROL 7 Figure 1.3 Determination of Output and Abatement Cost, price 4 Demand surface Abatement I Optimal - _. __expansion path B A: No intervention A B: A presumptive charge in output only Output C: An emission tax, or a skillful combination D: An emission standard only path. A is a useful point of departure: it shows output and price with zero abatement-the equilibrium if there are no environmental stan- dards or charges. Point B is the equilibrium point if a presumptive charge is levied on output: it stimulates no abatement, but it does cause a reduction in demand. Point D, by contrast, is an equilibrium with an abatement requirement (i.e. an emission standard); abatement is stimulated, but output will be "too high," since consumers will be confronted with a price that only includes the increased costs of abate- ment, not the external costs associated with the remaining emissions. 8 TAXING BADS BY TAXING GOODS Point C displays optimal abatement and output, which can be reached with the use of direct instruments such as charges levied directly on emissions, or approximated with skillful combinations of abatement requirements and presumptive charges levied on output. As we have emphasized, emission reductions can be obtained by a combination of instruments that make the polluting activity cleaner or that reduce its scale. Which combination of instruments will be used depends, as figures 1.1-1.3 illustrate, on the elasticity of demand for the polluting good and the ease with which the activity can be made clean- er. The reduction in emissions will be achieved mostly by technical con- trols if these cost little per unit of emission reductions and if demand elasticities are low (that is, if points A and D are far from each other, as in figure 1.3).3 By contrast, emission reductions will be attained mostly by demand reductions if demand elasticities are high and control effec- tiveness is low (see figure 1.4). A typical example of the former might be reductions of the waste load in domestic sewage. It would be hard to argue that the generation of human waste loads is sensitive to prices; thus all pollution reductions have to be achieved through technical con- trols. An example of demand reduction might be the control of carbon dioxide (CO2) to prevent climate change. Since no control technologies for reducing CO2 emissions from fossil fuel burning are available, all emission reductions will have to be achieved through changes in demand for fuels (including shifts toward fuels with less CO, per calo- rie, such as natural gas). Both demand reductions and technical controls will be important in obtaining reductions in the two remaining cases. In one case, polluters have two important ways of responding to emission taxes, and emission reductions will be cheap. (A good example in this category is air pollution from urban transport, at least in early stages.) In the other case, polluters can do little to reduce emissions, and emis- sion reductions will be costly. Initially, ozone-depleting substances such as chlorofluorocarbons (CFCs) appeared to be in this category, but tech- nical alternatives have been developed for most important uses. In gen- eral, the costs of obtaining emission reductions increase as one moves toward the southeast of the figure. Figure 1.4 characterizes responses that deliver emission reductions under emission taxes. Without such direct instruments, the policy- maker will generally need separate tools to induce technical controls and demand reductions. Then, in order to know which tools can deliv- er the bulk of emission reductions, the policymaker needs to know the elasticity of demand and the efficiency of technical controls. This is a very important difference between the world of direct instruments based on perfect monitoring and the world of indirect instruments. THE ECONOMICS OF POLLUTION CONTROL 9 Figure 1.4 How Most Emission Reductions Are Delivered When... High Low - Demannd reduction Both Simply put, if the pollution control agency knows little about individ- ual emissions, then it should know much about other things, such as what characterizes polluters and what options are available to them. Figure 1.5 summarizes wlhich indirect instrument will be most impor- tant in providing emission reductions, on the basis of the same under- lying conditions as in figure 1.4. The lower left cell, where demand elasticities are high and control effectiveness is low, represents the case in which it is most important to tax goods and services used in polluting activities. We call it a strat- egy of taxing bads by taxing goods. To follow the examples above, levying taxes on fuels will be a good strategy for reducing CO, emis- sions. Similarly, to reduce discharges from domestic sewage, technical solutions that treat or redirect waste loads are necessary, and indirect instruments mandating either these technologies or connection to col- lective systems can stimulate such responses. For cars, as we shall show later, it is important both to make trips cleaner and to discourage trips-instruments that do both are necessary unless taxes can be levied on individual emissions. A simple and balanced, although blunt, solution is a combination of emission standards with periodic testing, to ensure that cars are reasonably clean, and gasoline taxes or other demand management instruments, to discourage the use of cars. A more sophisticated solution is to tax the owner according to pre- Figure 1.5 How Emission Reductions Are Stimulated by Indirect Instruments When... I High Low inBoth Stimulation of technical controls Tax on polluting goods Both 10 TAXING BADS BY TAXING GOODS sumed accumulated emissions. The latter proposal-which would give much better incentives to owners and users-has yet to be applied in practice. It should be noted that combinations of instruments are useful in all the cells; the emphasis here is on which instruments are most impor- tant. Another way of reading the grid is to ask when it is costly to apply only a narrow range of instruments. For instance, a program consisting only of instruments stimulating emission control technolo- gies will be a costly one, and unnecessarily so, if demand elasticities are high. (Such narrow programs are the rule, rather than the excep- tion, in practice.) Similarly, a program consisting only of input taxes will be costly if emission control technologies would be effective in reducing emissions per unit of input. Presumptive Charges: Problems and Promises A Pigovian tax on emissions that is equal to external damages would generally be the first-best policy prescription. Box 1.2 summarizes con- servative estimates for the health benefits of air pollution reductions in Santiago, Chile.4 We have argued that taxes can also be levied on pol- luting goods and services, in presumption of emissions associated with their use. However, because the link between external damages and the use of polluting goods and services may be weak, marginal emissions may vary among polluters paying the same presumptive charges. In brief, the incentives provided by presumptive taxes will inherit any weakness in the association between the tax base (such as fuel consumption) and damages. On the control side, because each option typically involves several pollutants, benefits and control costs can best be compared by strategy rather than by pollutant. The analysis shows that health benefits alone can justify all four strategies. The two with the highest benefit-cost ratios are the fixed-source strategy (mostly because of its particle reductions) and the car strategy (mostly for its ozone reductions). We use an example to illustrate the problems-and the strategies- associated with presumptive instruments. Let emissions from polluter i, e', be determined by the quantity consumed of a polluting good, x', and also by a parameter describing the type of technology, ai. The cost to the polluter of a price increase of $1 for the polluting good is equal to the number of units used, x'. The change in tax proceeds received by the government (assuming that other goods are untaxed) is x' + txt, where a subscript stands for partial derivatives, so that x, denotes the marginal change in demand in response to the tax increase.5 Thus, if we assume THE ECONOMICS OF POLLUTION CONTROL 11 Box 1.2 A Limited Treatment of Benefits To assess the damages associated with pollutants, one needs to know whom they are affecting and how they increase costs to producers or reduce con- sumer welfare. The main motivating factor in efforts to control air pollution is often the associated health effects. The most important health considera- tion seems to be the systematic effects of dust, or the smallest, "respirable" particles, on illness and premature mortality (Ostro 1994). World Bank (1994) estimated the benefits of a specific control program in Santiago, Chile. The table shows point estimates (the ranges of uncertainty are not easily quantifiable) for the main air pollutants in that city. The con- trol scenario consists of various strategies (for point sources, for buses, for trucks, and for cars) that produce specified reductions in emissions. A dis- persion model is used to estimate the air pollution reductions that these strategies produce in different parts of the city and the resulting reductiolns in population exposure. Ideally, one would value such effects by willing- ness to pay, but a simplistic valuation by lost productiv e days and treatment costs was chosen. Both the dose-response estimates (which include only demonstrated acute effects) and the valuation reflect conservative assump- tions, leading to lower bounds for the health benefits. The results indicate that emissions of respirable particulates (PM-10) should be given the high- est priority, at $18,000 per ton. This would change somewhat if one could adequately assess the role of sulfur oxides (SOx) and nitrogen oxides (NOx) in particulate formation and if one could include nonhealth benefits. For instance, SOx would be credited with crop loss and with damage to the ecosystem and materials (costs that are not included here). Benefits for Santiago, Chile, Control Scenario Reduction of Amwbient exposure under Healtli bensefits Ben.efits per toll pollutant control scenario (nillions of dollars) emitted (dollars) PM-10 8.43iLg/m3 (70/o) 70 18,192 Ozone 0.0365 ppm (33%.) 33 495 (voc) 1,315 (NOx) NOx 0.013 ppm (33%) 0.7 59 SOx 0.16[Lg/m3 (0.3%) 0.1 82 Note: voc, volatile organic compounds; hLg/n3, micrograms per cubic meter; ppm, parts per million. The control scenario yields an improvement, not a complete removal. Souirce: World Bank 1994. 12 TAXING BADS BY TAXING GOODS that income is worth the same whether it accrues to the government or to the polluter, the social cost of raising the tax rate is tx,'. The marginal change in emissions from polluter i in response to a change in the tax rate will be e. xi + e', a,. Assuming, perhaps conservatively that technol- ogy does not change in response to a price increase for the polluting input (na, = 0), the marginal social cost of emission reductions produced by a price change for the polluting good is (1.1) t/e' or, simply, the tax rate divided by the marginal emission factor, aver- aged over polluters (the latter may depend on technology as well as on consumption). There are two problems with such a presumptive charge on a pol- luting good. One is related to the failure of t effectivelv to stimulate other initiatives that can reduce emissions, here represented by the variable a'.6 As noted above, one may address this problem by using separate instruments (incentives or regulation) to induce the applica- tion of cleaner technologies, depending on how attractive these are and on their monitoring and enforcement requirements. At given abatement levels, however, marginal cost expressions like equation 1.1 apply as long as the emission factor reflects the abatement level. Thus, when taking into account the estimated marginal emission factors, appropriately averaged if there are many polluters, one can reduce emissions by raising the price of the polluting input at a marginal cost indicated by equation 1.1. The other problem with the presumptive tax levied on a polluting input is that it is "unfair," and in a way that is also inefficient: it dis- courages the use of the polluting good with the same strength (as mea- sured by the tax rate) for polluters who may have different emission factors. Thus, while a polluter with a high emission factor could reduce emissions cheaply by reducing her use of the polluting good, she will have only average inducements for doing so as long as she faces the same tax on the polluting good as others. This may be a prob- lem even if separate strategies can be pursued to induce cleaner tech- nology, since emission factors may vary even wheni separate instru- ments push abatement optimally. Thus, if one recommends presumptive taxation of polluting inputs and goods in lieu of high monitoring and enforcement costs, these weaknesses have to be kept in mind. An important point is that an incentive scheme based mainly on presumptive charges can be com- plemented and refined as technical and institutional capacity allows. THE ECONOMICS OF POLLUTION CONTROL 13 Because of these weaknesses, presumptive taxes are not themselves part of an ideailized incentive scheme (a hypothetical scenario with cost- less monitoring). They may provide a setting conducive to institution- al and policy development in the direction of the ideal system-a set- ting sensitive to the practical challenge of developing monitoring capacity. The argument rests on the observation that presumptive charges are both "unfair" and "transfer-intensive." "Unfairness" will induce polluters that are or can be cleaner than average to self-report, citing their lower emission factors to justify refunds of parts of the taxes paid or simply lower tax rates. Incentives to do so are high, since the taxpaying polluter regards the transfer she makes to the govern- ment as a cost and any savings as 100 percent genuine savings. The pollution control agency can announce, when it is institutionally and technically capable, that special treatment, such as partial refunds, is available for those who prove cleaner than average. It would then be up to the polluter to take the initiative. She can support the claim that she is cleaner than average by self-reporting, by credible monitoring, and by submitting to special audit regimes. Such a policy refinement would provide incentives both for pollution abatement technologies and for monitoring and auditing technologies. As polluters with lower emission coefficients opt out of the "basket case" treatment, the tax rate for the remaining polluters will increase (since their weighted- average emission factors increase). Starting from a basis of crude pre- sumptive instruments, the policy stance described here could lead in the direction of a very sophisticated emission fee system. This "policy trajectory" argument carries some risks, however. The incentives provided by blunt presumptive charges (and those pro- vided by environmental regulation, which will also be based on aver- ages) may remain rigid and blunt, rather than continuously refined. The pollution control agency (and the treasury), infatuated by pro- ceeds from taxes on polluting goods, may resist calls for special treat- ment, since they come from polluters who would like to pay less. Also, rigidity and tiniformity may be seen as virtues in some bureaucracies. Polluters, if they expect a negative or undifferentiated response, will not have an additional incentive to become cleaner than the average. The latter risk is higher if the taxes are perceived as ordinary indirect taxes, rather than as a specific charge that polluters are invited to avoid through specific avenues. It should be remembered, however, that pre- sumptive charges serve a purpose in pollution control even if they remain blunt instruments, as long as they do not block development of the more refined incentive schemes. 14 TAXING BADS BY TAXING GOODS Notes 1. The assumption essential to this argument is that the value of public rev- enue is equal to that of private incomes. Assuming (costless) lump-sum trans- fers is one way of ensuring this equivalence, and an assumption of marginal utilities of income independent of the price of the polluting good allows the use of consumer surplus as a measure of welfare costs. For analysis in a con- text with (optimal) distortionary taxes, see Sandmo (1975). 2. If an input tax is envisaged, the argument in this section assumes fixed coefficients of that input to output; that is, the gasoline tax can reduce the use of the car but cannot make car use less fuel-intensive. This assumption and its relaxation are discussed in detail in the next two chapters. 3. Technical controls and cleaner consumption are here used svnonymous- ly; what is essential is that emissions are reduced per unit of input or output. Technical controls show high effectiveness if they can shift emissions down- ward so that the SMC curve is shifted downward sharply. Because the SMC curve denotes the sum of PMC and the external costs due to emissions, this shift is possible only if emissions can be reduced sharply without a major increase in PMC. 4. On the importance of health concerns for air pollution control policies, see Freeman (1982) and Pearce and Markandva (1986). On health effects, see Ostro (1994) for a review and, for estimates from developing countries, Alberini, Harrington, and McConnell (1994), Alberini and others (1995), Eskeland and others (forthcoming), and Ostro and others (1995). 5. The argument also holds for another setting for nondistortionary taxa- tion: when all goods are otherwise optimally taxed at uniform ad valorem rates (no distortionarv taxation). In that case, the tax rate here referred to as the tax rate, t. on the poliuting good should be interpreted as a special additional unit rate. Note that we assume that the production cost of the polluting good is given; thus the change in the user cost is the same as the change in the tax rate. 6. Replacement of older, dirtier machinerv by newer, cleaner machinerv will often be accelerated by increased fuel prices, but this does not mean that fuel prices are effective in stimulating cleaner technologies. 2. Pollution from Mobile Sources A pollution control agency able to levy taxes on each individual's emis- sions need not know anything about the technical and behavioral changes that could produce pollution reductions. The agency would be in a position analogous to that of a buyer of fruit who is confident of his ability to assess the produce upon arrival. The buyer pays whoever offers good produce; he has no interest in the background of the suppliers or the history of their shipments. But what if fruit for the city mysteriously landed on the mayor's doorstep at night? He would be hesitant about paying those claiming to be responsible the following morning, just as he would be hesitant about paying a rainmaker without further evidence. Similarly, if a social planner possessed data on how much pollution each individual caused throughout the year, a year-end tax bill based on emissions would easily provide appropriate incentives for pollu- tion reduction; only those who had polluted would contribute to the pollution control program. For many important pollution categories, however, including motor vehicles, emissions are not monitored con- tinuously-and will not be, in the near future. In this situation, the planner needs to investigate which sectors are polluting, what options exist withini each sector, and how he can best stimulate them. Using Indirect Instruments: Cleaner Cars, Fewer Trips Let us initially accept the proposition that the planner has a monitor- ing problem: he does not know how much pollution each individual emits throughout the year and thus cannot, with full precision and without cost, levy a charge directly on individual emissions. With indi- rect instruments, he will often have to apply separate strategies to pro- mote the various ways by which polluters can curb emissions. We may view the pollution control agency as an agent who purchases emission reductions on behalf of the public (see box 1.1 in the last chapter). To 15 16 TAXING BADS BY TAXING GOODS such an agent, cleaner cars and fewer trips are alternative suppliers of emission reductions, and she should buy from each up to the point at which emission reductions, at the margin, cost the same from each.' We now consider a traditional control program that focuses on mak- ing cars cleaner and ask whether and how such a program can be com- bined with a uniform gasoline tax. The role of the gasoline tax is to pro- vide polluters with an incentive to reduce their demand for the polluting good. The principal weakness of such a reform is that it dis- courages vehicle use uniformly, ignoring differences in emission rates. The practical reason for suggesting such a modest reform is the gener- al suspicion that administrative and technical systems for emission test- ing are still vulnerable (see box 2.3, below), raising doubts as to whether they should be used as major tax-collecting devices. The beau- ty of the fuel tax is its administrative simplicitv: all countries interfere in fuel markets (some with a tax, others with a subsidy), so their abili- ty to manipulate this price is proven. The collection of a pollution-moti- vated charge in fuel markets requires little or no new monitoring. Since each liter of gasoline consumed produces a certain amount of pollution, the gasoline demand curve shows the price at which con- sumers can reduce pollution by reducing their gasoline consumption. When consumer surplus is used as a measure of the welfare costs of reducing consumption, the part of the demand curve that is above the gasoline supply curve (marginal production costs) shows the welfare costs of pollution reductions provided through demand reductions. Formally, if welfare depends on gasoline consumption and other con- sumption, and emissions depend on gasoline consumption, the marginal welfare costs of adjusting the gasoline tax rate, per unit of associated emission reduction, is:2 a3n' ie _ tdx __t (2.1) 1-----_tdr at at e .dx e, where w' is welfare (which, for simplicity, does not depend on emis- sions), t is the gasoline tax rate, e is emissions, and e. is the marginal emission coefficient for gasoline. The formula is most easily explained in terms of figure 2.1. A tax change, dt, leads to an additional welfare cost equal to the shaded trapezoid, approximated by the rectangle t * dx. Emission reductions will equal c_dx, and dx cancels out in the expression for marginal costs, which is the ratio between the effect on welfare and the effect on emis- sions. The term dx is determined by the slope of the demand curve, or by the demand elasticity. Thus, when this term cancels out, it means that the marginal welfare cost of providing emission reductions POLLUTION FROM MOBILE SOURCES 17 Figure 2.1 The Cost of an Increase in the Gasoline Tax Cost (pesos) p +t+d- dti p+ tt Demand x (gasoline consumption, in liters) dx Note: Values are as follows: p, producer price of gasoline; t, gasoline tax rate; dt, change in gasoline tax rate; dx, corresponding change in gasoline consumption1. through tax rate changes at a given tax rate is independent of the demand elasticity. Of course, the emission reductions obtained by a given tax change are not independent of the demand elasticity, but the marginal cost at any given tax rate is. As an example, if the elasticity is large, the emission reductions are large, but so are the costs from sac- rificed consumption. Thus, the demand elasticitv* would not affect the costs at which the emission reductions are delivered. Now that we have an expression for the welfare costs of reducing emissions with a gasoline tax change, a cost-effective program would seek emission reductions from cleaner cars and fewer trips such that the marginal costs are the same for both "deliveries": (2.2) tlex = c where c, is the marginal cost of emission reductions via technical con- trols ("cleaner cars"). This formula quantifies the principle that cleaner cars and fewer trips deliver emission reductions in such a way as to equalize the marginal costs. 18 TAXING BADS BY TAXING GOODS Table 2.1 Abatement Measures by Cost-Effectiveness, and Matching Gasoline Tax Rates, Mexico City Clulmllalfive Mlarginal emiissioni Curmlnative Matchliing cost rednictionis cost, abate- gasoline (US$ (thousa171tns 11117n1t otnly taX (celnts Measiure per toW) of tonis) (US$, mttillionis) per liter) LPG retrofit, gasoline trucks -379 90 0 -4.4 CNG retrofit, minibuses -248 148 0 -2.8 CNG retrofit, gasoline trucks -225 231 0 -2.4 Gasoline vapor recovery -80 275 0 -0.8 Re-engining of buses 140 299 3 1.4 1992 standards for minibuses 181 391 20 1.7 Mandatory j&Mi for high-use vehicles 209 545 52 1.8 1993 standards, gasoline trucks 264 632 75 2.1 Tier I standards, taxis 322 641 78 2.5 Re-engining of R-100 buses 482 651 83 3.7 Taxi replacement, 1993 standards 510 714 115 3.7 Centralized i&M, cars 651 771 152 4.4 1993 standards, cars 669 883 227 4.0 Diesel reformulation 699 893 234 4.2 Lowering of vapor pressure to 7.5 836 904 243 4.9 Provision of regular unleaded gasoline 923 954 289 5.1 Decentralized i&M, cars 1,034 1,018 356 5.3 Replacement of gasoline trucks 1,114 1,096 442 5.0 5% additives in regular unleaded gasoline 1,201 1,116 467 5.3 Lower vapor pressure in premium unleaded gasoline 1,313 1,128 482 5.6 Road paving (1,000 km) 1,335 1,136 498 5.7 1991 standards, cars 1,367 1,180 508 5.4 Reduction of sulfur to 0.1% in diesel fuel 1,371 1,187 569 5.3 Tier 1 standards, cars 1,629 1,201 578 6.2 U.S. specifications for diesel 2,097 1,207 601 7.9 11%.o additives in regular unleaded gasoline 2,447 1,219 613 9.0 5% additives in premium unleaded gasoline 13,487 1,222 643 49.0 11%.;, additives in premium unleaded gasoline 14,728 1,226 686 53.2 POLLUTION FROM MOBILE SOURCES 19 Note to table 2.1 Note: i Rn, liquefied petroleum gas; CNG, compressed natural gas; I&M, inspection and maintenance. The first and second columns give, respectively, the height and length of the steps on the technical control cost curve shown in figure 2.2. The match- ing gasoline tax (equation 2.2) shown in the last column increases monotonously per gram of pollutants but not per liter of gasoline, since emissions per liter (cx) are brought down progressively by controls. Source: Eskeland 1994b. What if a pollution control program did not follow this principle? For instance, suppose it pursued technical controls but with no or only mod- est taxation of gasoline (as in the United States). Consumers would be loaded with technical control requirements that would be quite costly at the margin, compared with the costs at which they could have sacrificed some nonessential trips. They would be better off if asked to abstain from some of their least essential trips, in return being freed from some of the costliest controls (with total emissions held at the same level). A technical control cost curve was constructed for a study on air pol- lution in Mexico City (Eskeland 1992,1994b; World Bank 1992). Table 2.1 shows the cost-effectiveness ranking of the proposed interventions for making cars and fuels cleaner, together with the rates for a gasoline tax that would optimally match these measures (equation 2.2). It may be noticed that the gasoline tax rate, in cents per liter, does not increase proportionately with the technical control costs. The reason is that the emission factor per liter of gasoline (e'l) declines as one climbs further along the control cost curve because abatement technologies are applied, making cars progressively cleaner. Thus, while the gasoline tax follows the control measures in terms of dollars per ton emitted (as stated in equation 2.2), it increases less than proportionately when measured in cents per liter. Whether the optimal discouragement of trips matters much quanti- tatively depends on the slope of the demand curve, in comparison with the slope of the technical control cost curve. Eskeland and Feyzioglu (1994a) used panel data from thirty-one Mexican states over seven years to estimate the relationships of demand to income and prices of optimal car stocks, gasoline consumption per car, and total gasoline consumption. The results are in line with a rich body of empirical liter- ature in terms of long-term elasticities (which are the most important from the perspective of efficiency)-in the range of around 1 for income elasticities, and -0.5 to -1 for price elasticities. The results stand out, however, in displaying quite rapid adjustment in the short 20 TAXING BADS BY TAXING GOODS Table 2.2 Elasticities for Total Gasoline Consumption Itenm Short-run elasticity Long-runii elasticitw Price of gasoline -0.79 -0.8 Price of new car -0.03 -0.03 Income 0.98 1.02 SoIrce: Eskeland and Fevzioglu 1994a. term, so short-term elasticities are larger, in absolute value, than what is often found (table 2.2). Figure 2.2 shows two supply curves for emission reductions in Mex- ico City. The lower curve results from cost minimization when an adjustable gasoline tax is included in the toolbox of the control agency; the higher would result if the control agency were prevented from using a gasoline tax. The gasoline tax rates reflected in the lower curve are given by equation 2.2 and are listed in table 2.1. Without the gaso- line tax, an otherwise well-designed control program would become 23 percent costlier, at an additional cost to the citizens of Mexico City of $110 million a year. Thus, since demand for the polluting good is quite sensitive to pricing, demand management is important for the solution to this problem. The proposed program removes about 70 percent of weighted emissions from the vehicle fleet. The gasoline tax that is opti- mal given such a target is about 25 percent; it removes 20 percent of emissions, under the assumption of a price elasticity of -0.8.3 Market-Based versus Regulatory Demand Management Once it is realized that the demand for polluting goods and services may be manipulated to reduce pollution, what is the best way of doing so? Initiatives to economize on polluting trips might include gasoline taxes, regulatory restrictions on driving, parking fees, urban zoning and tolls (toll rings), and subsidies to public transport. As a demand management instrument, a gasoline tax has the shortcoming of allowing only a limit- ed differentiation by geographic area and no variation by time of day. Apart from that, it has many advantages. It may be used alone if sharp geographic and temporal differentiation is not important, and it may be supplemented or replaced by toll rings, peak-load charges, parking fees, and public transport policies when such variation is necessary. When consumers sacrifice trips in response to demand management instruments, they incur welfare costs.4 When a trip is sacrificed be- cause of a marginal increase in the gasoline price, the value of the sac- POLLUTION FROM MOBILE SOURCES 21 Figure 2.2 Supply Curves for Reduction of Emissions from Transport in Mexico City, with and without a Gasoline Tax Marginal cost of emission reductions (dollars per ton) 2,600 - Fuel I 2,100 improvements Emission \ stan ards - 1,600- \\Passenger cars. 1,100 _ Tax'is ,i \(elceet 1 Srengthened Gasoline l' I 600 Minibu es trucks n spection of _ _ passenger cars ITarget 100 Inspection of high-use vehiCleSa freduction 0 ~~~~~~Retrofitting (natural gas and LPG)E -400 Cumulative emission reductions 1.2 (millions of weighted tons) - - - - Technical controls only C ontrols, matched with gasoline tax m Welfare cost when tax is excluded Note': Calculations are based on -0.8 elasticity of demand for gasoline. a. Including taxis, light-duty trucks, and minibuses. Source: Eskeland 1994b. rificed unit to the consumer is the retail price of gasoline. Thus, a gaso- line price increase will screen out, systematically, the trips that are worth the least to consumers, sparing inframarginal units of gasoline (and trips) that are worth more. This property of the gasoline tax allows it to reduce trips at the lowest possible welfare cost.5 The prob- lem of how to reduce demand at the lowest possible cost is, of course, analogous to the traditional problem of how to take advantage of market-based mechanisms to help contain the costs of emission reduc- tions brought about with direct instruments (box 2.1). 22 TAXING BADS BY TAXING GOODS In Mexico, gasoline is now taxed, but a driving ban is also used in an attempt to curtail driving. The capital's "day without a car" pro- gram, based on the vehicle's license plate number, bans each car from driving on a specific workday. Thus the maximal potential effect on demand that the regulation could have would be to remove 20 percent of workday driving, and the effect would be less if people adjusted in ways other than merely canceling trips. To quantify the costs and benefits of the regulation, one needs to quantify the demand reduction (if any) that it causes and the welfare Box 2.1 How a Market Can Help Even if emissions are monitorable, the policymaker will often be uncer- tain-and in error-about what it will cost individual polluters to reduce their pollution discharges. Information about the control costs for individ- ual polluters would be useful to the planner, since he can minimize the aggregate control costs only if individual pollution reductions are so dis- tributed that the resulting marginal control costs are equalized for all pol- luters. To stimulate such a distribution of reduction efforts, the planner needs information about individual control costs. One way for information to be revealed is through market-based instruments, such as tradable emis- sion permits or emission taxes. This simple point and its qualifications are illustrated by the figure. There are two polluters, and the planner has determined a target for total emission reductions, equal to the length of the horizontal axis. Emission reductions from polluter 1 are measured from left to right, and emission reductions from polluter 2 from right to left. The planner has an estimate of each polluter's marginal control cost curve, represented by the solid upward- and downward-sloping curves, respectively, but he assesses these with an error. We shall assume that the true control costs (the broken curves) are known only to the respective polluters. If the planner were to set individual quotas for pollution so as to mini- mize his expectation about total control costs, he would choose quotas, q,, i2, that would equalize the expected marginal costs, as shown by the intersection of the solid lines. Real costs, however (the area under the broken lines), would be minimized by the reductions e1, q2, where the broken lines intersect. The shaded area could have been saved had the dis- tribution of emission reductions between polluters been determined by the true costs rather than by the planner's estimate. As polluters know their own cost curves, they can redistribute quotas among themselves, assuming that trade in quotas is allowed. Thus, starting from point q,, q2, polluter 1 would be happy to purchase emission permits POLLUTION FROM MOBILE SOURCES 23 costs of obtaining these reductions. When demand falls because of a marginal price increase, we know that the sacrificed consumption is worth the commodity's market price to the consumer. Thus (referring to figure 2.1), when one consumer reduces his demand for gasoline by I liter, we know that there is not another consumer who uses gasoline but who has a lower marginal willingness to pay. Demand reductions resulting from regulation will rarely have this selection quality. The "day without a car" program may curtail trips in households with a very high willingness to pay, and it may block a equal to the difference between il, and q2, for a unit price of P, in order to be able to pollute more. Polluter 2 would be glad to sell, so a market in permits can assist the planner in finding the least costly way of controlling pollution. It follows that the scope for reaping savings through the use of market- based instruments requires that there be several different polluters and some room for reallocation of emission reductions among them. Special cases in which the market cannot assist in lowering control costs are those in which there is only one polluter, those where the errors with which the planner assesses individual control costs are perfectly positively correlated, and those in which either no or 100 percent emission reductions are sought. Use of Market-Based Instruments for Reducing Emissions Costs for polluter 1 Costs for polluter 2 (C1) (C2) A A C1 CX, C2 C p 24 TAXING BADS BY TAXING GOODS household's Tuesday driving, even if the household could more easily have sacrificed other trips. Both effects result because the regulation does not allow "trading" of the rationed commodity: trips. As shown in figure 2.3, the regulation may curtail inframarginal as well as marginal trips. The figure illustrates a comparison of a regulatory demand reduction with a gasoline tax change calibrated to reduce demand by the same amount. The welfare costs of the regulation will be at least as high, and possibly much higher, depending on the shape of the demand schedule and the selection of trips that are squeezed out by the regulatory measure. A key assumption in this argument is that if the regulation achieves any emission reduction at all, it will do so through its impact on aggregate gasoline consumption. Then, as we argued above, using market forces to allocate any reduction (in gaso- line consumption, this time, rather than in emissions) will help contain the total costs of the reductions. In order to estimate the demand reduction caused by the regulation, a demand function was estimated based on data from Mexico City for the periods before the regulation (Eskeland and Feyzioglu 1994b). Since the demand function is estimated in periods without regulation, it can be used to simulate counterfactual demand scenarios for subse- quent periods as if no regulation had been implemented. The result, which was indeed surprising, is shown in figure 2.4. One would expect actual consumption to be lower than in the counterfactual simulation, but this occurs only in the first two quarters. After that, actual con- sumption exceeds and stays above the levels forecast by the simulation model.6 Moreover, the indicated confidence intervals (obtained by shifting forecast values up and down by one standard deviation) show the extremely low probability of observing the actual string of con- sumption levels if demand had not been shifted upward by the regu- lation. (The probability of observing one value outside the confidence interval is about 3 percent; observing many is much less likely.) Thus we can safely conclude that, after an initial period of adjustment, gaso- line consumption was not curtailed by the regulation but rather was boosted, or at best held constant. (This result is supported by the appropriate Chow test.) The result is not intuitive, and to try to explain it, one may think through the ways households can respond to the regulation. One household complies as expected-by managing without its car on Tuesday, sharing rides with friends, using the metro, and postponing or canceling trips. Of course, if the trip is only deferred to another day, total demand reduction would be zero, but otherwise one would expect demand reductions to be positive. Another household, finding POLLUTION FROM MOBILE SOURCES 25 Figure 2.3 Costs of an Increase in the Gasoline Tax versus Regulation Cost of tax increase Cost (pesos) p + t + dt- p mandnan iq . q (gasoline consumption, in liters) dq Welfare cost of driving ban Cost (pesos) p + Heand q i n , t Li + + % = q (gasoline consumption, in liters) 26 TAXING BADS BY TAXING GOODS that way of complying too costly, decides to buy an additional car. If the household has more drivers' licenses than cars, its total driving is likely to increase; the household provides negative demand reductions in response to the regulation. Thus, for the market as a whole, one might see total demand reductions close to zero or even negative, if there are many households in the latter group.7 Having found that the demand reductions, and thus the emission reductions, offered by the regulation are zero or negative, we do not need to attempt the difficult task of calculating the welfare costs. These costs are definitely positive, and if they do not produce benefits, the regulation is not an attractive proposal (rather, removing it is). In Santiago, Chile, a similar regulation is used as an emergency mea- sure in high-pollution periods, and its effect is thus less likely to be jeopardized by households buying additional vehicles. Because the Santiago regulation applies only to old, uncontrolled cars, it helps make cleaner cars more attractive. However, this particular form of regulation has more weaknesses than most rationing schemes. For Figure 2.4 Gasoline Consumption under Mexico City's "Day without a Car" Program Log of consumption (liters) 4.75 Simulated (+ ISD) 4.70 I 4.65 o / -' sSimulated s R s (/without 4.60 . , regulation), Simulated 4.55 g/ (-1SD) 4.50 4.45 1987 1988 1989 1990 1991 1992 POLLUTION FROM MOBILE SOURCES 27 example, it allows households to purchase "implicit driving permits" by buying cars, thus tying up the nation's capital unproductively. In addition, the limits involved relate to certain trips, rather than to a household's total driving, thereby introducing more limitations than necessary, even for within-household reallocations.8 If pollution (or congestion) problems show sharp variations spatially or by time of day, demand management instruments such as provision and subsidization of mass transport, parking policies, and so on may be used to complement the effects of gasoline prices. In transport projects, by contrast, time savings are the main benefits produced. Krupnick (1992) found that the inclusion of emission benefits in the analysis of projects designed to reduce congestion and produce time saving is not difficult in principle and should be done. He also found, however, that congestion- and time-saving benefits would probably dwarf pollution benefits in such projects. Thus, the inclusion of emission benefits is not likely to tilt the balance in favor of large investments in transport infra- structure if the projects would otherwise be difficult to justify. Among the stylized facts cited by Krupnick is that developing coun- try cities with air pollution problems invariably are growing rapidly. Thus he recommends against using modeling frameworks that do not incorporate long-term phenomena such as residential developments. This warning is, of course, relevant if one considers the empirical find- ings of short-term studies, such as the many models of travel mode choice. These studies often conclude that choice of travel mode responds very little to changes in parameters such as travel costs and travel times (elasticities usually are between zero and 0.2 in absolute value), so that better or cheaper mass transport will do little to reduce the use of private cars. The results from Swait and Eskeland's (1995) mode choice model for Sao Paulo are in line with others from industri- al as well as developing countries. They can be read as a warning against the belief that mass transport can perform wonders in curtail- ing private transport. It is important to remember, however, that these models represent household behavior in a very restricted setting. In the long run, when hlouseholds choose where to live, where to work, how many cars to own, and so on, behavior may indeed be more sensitive to gasoline prices, fares, and travel times than these models indicate. Market-Based and Other Inducements for Cleaner Cars and Fuels We can distinguish between the "physical" changes that can deliver emission reductions (such as a tune-up) and the policy instruments 28 TAXING BADS BY TAXING GOODS used to induce those changes. For cars in use, standards and manda- tory inspection and maintenance programs are ordinarily the most important policy instruments. Higher gasoline prices will bring about some additional attention to maintenance and tuning and may even increase the market share of smaller cars. However, gasoline price increases can only trigger those options that are driven by fuel effi- ciency considerations. Emission standards and gasoline taxes are indirect instruments: the first is a regulation addressing a characteristic of the equipment (the car's emission factor), the second, a market-based instrument address- ing a proxy for the car's utilization. Emission standards have to be designed to stimulate those physical solutions that can cost-effectively improve emissions for the average car.9 Other instruments might trig- ger options attractive only for selected cars. Furthermore, the binary nature of compliance versus noncompliance gives rise to some weak- nesses; cars that are in noncompliance are pushed to become cleaner, but owners of cars in compliance have no incentive to reduce emis- sions, even though some of them might be able to do so cheaply. Sim- ilarly, standards may inadvertently require costly changes for vehicles that are close to but not in compliance, even though those changes may yield very low benefits in comparison with costs. Could there exist a large number of vehicles in use for which eco- nomic ways of reducing emissions might be devised, when standards set for the average vehicle do not stimulate these reductions? Clearly, there could.10 To illustrate, we refer to a feature frequently found in the economics of pollution controls: fixed costs. A tune-up, a retrofit with a catalytic converter, and conversion to natural gas all have one thing in common: a fixed cost is paid up front, but emission reductions are realized as a flow through subsequent periods, proportional to the uti- lization rate and the lifetime of the equipment. Such vehicle modifica- tions are more cost-effective the higher the vehicle's utilization. Thus a standard set to induce measures that are cost-effective for a vehicle traveling the average 8,000 kilometers per year (for Mexico City) will fail to induce many of the control options that are costlier, but more cost-effective, for cars used more intensively Of course, standards for vehicles in use can be complemented with a wide range of instruments, which could target some of these options directly. However, technical constraints limit how finely these instruments may be designed. Two possible vehicle modifica- tions-conversion to natural gas and retrofitting with a catalytic con- verter-may serve as illustrations. Conversion gives the owner access to a fuel-natural gas-that can be priced independently. POLLUTION FROM MOBILE SOURCES 29 Thus, adjustments in the relative price of natural gas may be used to induce high-use vehicles to self-select for conversion. A similar scheme to induce retrofitting with a catalytic converter is not feasi- ble; a lower price for unleaded gasoline cannot be used to induce such conversions, since unleaded gasoline can be used with or with- out a catalytic converter. Moreover, the prices of leaded and unleaded gasolines may need to be kept close together to avoid encouraging misfueling, as use of leaded gasoline in a car with a catalytic con- verter poisons the catalytic converter and irreversibly damages its ability to reduce emissions.11 One could look for other ways of select- ing high-use cars to be retrofitted with catalytic converters-for instance, by introducing separate requirements for taxis and delivery trucks. Such classifications of vehicles are surely rough proxies for utilization; taxis in Mexico City travel about nine times the annual average mileage for personal vehicles and may thus be useful as bases for indirect instruments, if that is administratively feasible. Since taxis have to have licenses issued by the city, making special requirements for them is usually simple.'2 The important lesson is not that taxis and delivery trucks should be cleaner than most cars (although this will often be desirable). More fundamentally, the absence of comprehensive information based on continuous emission monitoring leads the agency to a detailed exami- nation of where and how emissions can be squeezed. The agency then considers the economics of the reductions themselves, under alterna- tive inducement mechanisms. A good agency tries to allow maximum flexibility for polluters, leaving room for solutions it had not imagined or had thought uneconomic. It may still be the case, however, that sep- arate indirect instruments must be used to stimulate separate groups of polluters or to pursue separate avenues. In that case, much respon- sibility for analysis rests with the agency. Effects on Income Distribution Finally, there is the issue of income distribution effects. Usually, other policies can be found that redistribute income more effectively, and pollution control strategies that minimize total costs can be chosen. However, the analytical issues are not complex, as the cost of control strategies generally will be distributed among households according to consumption patterns for pollution-intensive goods (box 2.2). Thus, strategies for income-elastic goods-such as private vehicles-will, to a greater extent, be paid for by wealthier households, while strategies for necessities will have a less progressive incidence. This will be true 30 TAXING BADS BY TAXING GOODS Box 2.2 Distribution of Effects by Income Group Does concern for income distribution dictate the choice between pollution control instruments? Before analyzing the income distribution effects of con- trol strategies, one should remember that other instruments may exist that are better tailored for redistributive goals than is the modification of pollution control programs. If so, pollution control programs can focus on efficiency. A textbook on welfare economics would recommend that a polluter face the social marginal costs of her choices. Polluters who pay both their own control costs and fees for the remaining emissions will have a self-interest in finding the most effective means of pollution control. This idea-which is based on efficiency considerations, not equity or other income distribu- tion concerns-has flavored the so-called "polluter-pays principle," adopted as a policy guideline by the Organisation for Economic Co- operation and Development (OECD 1975) and by many countries. (The OECD formulation specifies that polluters pay for abatement, but it is silent about paying for damages.) In comparison with an uncontrolled environment, both taxes and regu- lation introduce private costs for polluters, although with taxes, part of the costs are merely transfers to the government. Under both systems, costs are distributed among households according to consumption patterns: for a polluting good, the share of the control costs paid by the rich will be higher, the greater is their role in consumption of that good. As the table shows, higher-income households in Jakarta and Mexico City spend a higher share of their income on their automobiles (counting purchase, maintenance, repairs, and fuels) than do the poor. Thus, policies being paid for by vehicle-owning households will, in general, have a progressive impact across households (Eskeland and Kong 1994). Studies of the income distribution effects of control programs in the United States in the 1970s yielded some important insights (see Harrison 1977). One was the proposal of a "two-car strategy" designed to allow "dirt- ier" cars in areas where air pollution was not a problem. In the United States whether one chooses regulatory or tax-based strategies, but the dis- tributed costs are higher for tax-based strategies, since these include transfers to the government. (Then, distributional assessments should include how the revenues are used.) Monitoring Technologies and the Need for Detailed Program Evaluation Although continuous monitoring of emissions at the source (say, by a running meter, as for billing water or electricity consumption) is not POLLUTION FROM MOBILE SOURCES 31 this proposal was defended as being both more efficient and more equi- table because the many poor rural households that depend on their private vehicles (and will not benefit from improvements in air quality) would be exempt from the controls. The proposal was not adopted originally, but geographic differentiation has later become more important in U.S. pro- grams, as states and local jurisdictions have obtained more freedom. Some strategies will influence the firms' costs as well, with cost increases filtering through the industries' transactions. These strategies will increase costs of goods and services more evenly, so that differenti- ation by the consumption patterns of income groups is generally less important. Eskeland and Kong (1994) found that in Indonesia a "trans- port strategy" has a more progressive distributional impact than a "man- ufacturing strategy" and that an "energy strategy" is the least progres- sive of the three. Share of Household Expenditure for Transport, by Income Quintile, Indonesia, and Mexico (percent) First Third Fifth Transport mode and site quintile quintile quintile Private transport Java, excluding Jakarta 0.2 0.8 4.8 Jakarta 0 0.03 3.1 Rural Mexico 0.3 1.3 2.8 Mexico City 1.2 2.1 3.7 Public transport Jakarta 2.4 6.8 6.2 Mexico City 11.2 12.1 9.3 Sources: Eskeland and Kong 1994; Kanninen and Hanemann forthcoming. yet possible, several "windows" give approximate information about emission rates. New car models are tested through a "model driving cycle" designed to mimic typical driving. New-car standards allow certification if the estimated average emissions per mile do not exceed a predefined limit. Periodic tests, used in mandatory inspection and maintenance programs, measure the percentage of various gases in the exhaust and thus provide a measure of emissions per liter of gasoline for vehicles in use. More recently, technology for remote roadside mea- surement has been developed and tested. As yet, however, it can only provide the percentages of hydrocarbons and carbon monoxide in the exhaust and has yet to be used for enforcement purposes. The tech- 32 TAXING BADS BY TAXING GOODS nology has provided useful corrections to the picture of the vehicle fleet given by the periodic nonrandom tests used in mandatory pro- grams. These approaches to monitoring have two important limita- tions: they give emission rates, rather than cumulative flows, and the measurement may not be representative of the vehicle's in-use emis- sions (box 2.3). The recent demonstration of remote-sensing techniques has empha- sized the problem of representativeness. Lawson and others (1990) found that the length of time since the last periodic test had little influ- ence on whether a car's emission rate complied with the standard.13 Thus, it appears, many "repairs" to make polluting cars pass the man- datory tests have no lasting effect on emission rates. The measurement of emission rates rather than cumulative emissions is quite important with cars, since one car may travel 8,000 kilometers per year and anoth- er 70,000 kilometers. (These are the mileages for the average private car Box 2.3 Monitoring Emission Rates or Emissions: An Analogy A system built on monitoring of emission rates has incentive problems sim- ilar to those of a production system with productivity measures rather than monitoring of production. Consider a tea plantation. If we were not moni- toring how much each tea plucker collects through the day, we could spot- check her speed during a one-minute observation. The problem of repre- sentativeness is clear: if the pluckers could sense when they were being observed, the monitoring would be of little value. (It would test who can best speed up for a limited time rather than who brings in the most leaves through a day.) Another problem of representativeness arises if the speed measurements are subject to random fluctuations and other disturbances- for example, if cloud cover and bushes vary greatly or if the inspector can be easily corrupted and distracted. The problem of monitoring the picking rate rather than the cumulative harvested amount would not be so impor- tant if we (and the pluckers) felt the measure was representative and if all the workers worked the same hours. But what if some worked a half-day or less, and we did not know who? In effect, we would end up rewarding part-time workers with full-time pay, and our tea would be costly. (Who would then work full-time?) If we could add a proxy for hours worked to the incentive scheme, using hourly pay by productivity class, incentives would improve dramatically. For emission reduction programs, incentives based on emission rates could in the same way be improved if they were complemented by some measure of "hours worked," such as gasoline consumption. POLLUTION FROM MOBILE SOURCES 33 and the average taxi, respectively, in Mexico City.) Measuring only their emission rates leads to equal incentives to clean up both, although the social returns from reducing emission rates differ by a factor of nine. Periodic tests and the institutional and technical systems that sup- port them are improving, reducing the scope for both poor measure- ment and weaknesses due to corruption.14 The representativeness of the tests is being gradually improved through better technology and more resources (loaded mode tests, longer-cycle tests, and closer vehi- cle inspection). Other, but less likely, developments that could improve representativeness are surprise roadside tests and a technology for reading cumulative emissions since the last test.15 Given present monitoring capacity, an obvious way to improve the applied systems would be to read the emission rate from the car's annual test, multiply it by the cumulative mileage (on a tamperproof odometer), and apply an emission fee based on the result. The fee could be paid on testing, or it could be paid uniformly at the gas sta- tion as a presumptive tax, to be refunded in part to owners of vehicles found to be "cleaner" than was presumed in the prepaid tax. The efficiency gains from such a reform would come through several channels. First, all owners would have continuous incentives to drive less, and owners of more polluting cars would, appropriately, have more of those incentives than others. Second, all owners would have incentives to make their cars cleaner, but owners who use their cars rarely would feel less of this pressure than owners with higher mileage. As a consequence, society would waste few resources cleaning or scrap- ping cars that are rarely used. Third, the car market would facilitate exchange of vehicles, to make sure that households and firms which use their vehicles intensively would end up with the cleaner ones. Notes 1. This perspective is examined in detail in Eskeland (1994b), from which much of this material is selected. We use the term gasoline tax, but the same principles apply to other polluting inputs, such as diesel fuel. 2. Detailed calculations are shown in Eskeland (1994b). An important point is that income is here valued equally whether it accrues to the public sector or to private households. This treatment therefore abstracts from income distri- bution effects and revenue effects. 3. In Mexico City alone such a gasoline tax would collect between $300 mil- lion and $350 million in tax revenues annually. A gasoline tax motivated by pollution control objectives should, ideally, be higher in urban areas, but the geographic differences may have to be limited for administrative reasons. Gasoline may also be taxed for general revenue and other purposes, of course. 4. We use trips and gasoline as metaphors for polluting goods and inputs. 34 TAXING BADS BY TAXING GOODS Consequently, if gasoline is saved by using more fuel-efficient cars, the argu- ment applies as if trips were saved. Other fuels should be taxed according to emissions, as well. 5. When calculating the welfare costs of manipulating consumption, we need to emphasize that we employ the "compensation criterion": we assume no value (or penalty) associated with the transfer of money from private to public hands, or from poor to rich. Thus the analysis ignores any concern for income distribution, as would be correct if transfers could be arranged cost- lessly with the use of other policy instruments. 6. Simulations are performed with actual price and income developments inserted in the estimated demand model. In all calculations, seasonality and stationarity are taken into account. 7. There is anecdotal evidence that many households bought used cars; that is, Mexico City imported old cars from the rest of the country. If this did hap- pen, use may have shifted toward cars that are dirtier per liter than the aver- age car, which would make the consequences for emissions worse than indi- cated by aggregate consumption. 8. Rationing nontradables in this case, trips-thus has higher welfare costs than rationing tradables with low resale costs, such as, say, alcohol and tobacco. 9. Regulations, or their enforcement, can allow for some differentiation, which may be desirable for efficiency or equity reasons. For instance, regula- tions might be applied only in polluted areas (see Harrison 1977, an early statement of this point) or to certain model years, weight classes, and so on. Cars that cannot pass the emission test in Mexico City could be used in other areas of the country-a very desirable feature in a poor country where ambi- ent air pollution is the least of the problems in rural villages. 10. Anderson (1990) provides an interesting practical review of strategies applied to reduce emissions from in-use vehicles. McConnell and Harrington (1992, 1993) evaluated strategies for detecting high-emitting vehicles. They found that the continued presence of such vehicles, despite standards and scheduled testing, is due partly to monitoring problems but also to caps allow- ing a vehicle to operate when a specified tune-up cost has been incurred. The efficiency of programs to make cars and fuels cleaner depends on how well they screen vehicles with respect to relevant parameters. As an example, Alberini, Harrington, and McConnell (1993) analyzed a buyback program designed to get cars with higher emissions off the streets. They found that the program selected for some desirable characteristics (poor operating condition) but also for some undesirable ones (low expected remaining lifetime mileage). Krupnick, Walls, and Collins (1993) compared various instruments, including a corporate average fuel economy (CAFE) standard. They found that a tougher CAFE standard was not cost-effective, even if the emission control objective in question was for greenhouse gases. 11. Measures that discourage misfueling (such as nozzle-size regulations, enforcement, and octane differences) can easily fail if the price difference between leaded and unleaded gasoline is large. 12. For example, the "tier 1 standard" yields emission reductions at a cost of $322 per ton for taxis in Mexico City and $1,629 per ton for other cars (see table 2.1). 13. Harrington and McConnell (1993) find that the use of remote (and thus unscheduled) sensing techniques is cost-effective in comparison with man- POLLUTION FROM MOBILE SOURCES 35 dated periodic testing under plausible assumptions. It would also provide additional emission reductions cost-effectively in areas with established inspection and maintenance programs. 14. For instance, computerized testing equipment may print automatically a certification sticker based on the test result. Such advances may be of little help, however, if random or other disturbances allow the gas sensor to be held away from the tailpipe. 15. Of course, the development of such a technology could be stimulated by offering car owners a rebate on the gasoline tax based on the reading of their cumulative emission meter (it would have some interesting incentive effects). Other illustrations of monitoring techniques and problems may be cited. In Mexico City surprise roadside tests were, perhaps laudably, resisted as an enforcement measure. It was felt that the public should be spared police inter- ruption unless on specific suspicion. In Jakarta policemen have been autho- rized to fine smoking vehicles on the basis of visual assessment. In Los Angeles the public may call authorities to report smoking buses. 3. Control of Pollution from Industrial Sources Perhaps the most important feature distinguishing industrial polluters from vehicles is the much greater heterogeneity of the former. Although motor vehicles certainly differ in important respects (age, size, fuel, combustion and emission control systems, maintenance, and utilization), they may easily be placed in a small number of categories. In each category-new personal gasoline vehicles and used personal gasoline vehicles, for example-there will be a high number of pol- luters that are reasonably homogenous (almost all use an internal com- bustion engine to produce transport services), and we can know each group reasonably well by sampling a small subset. Industrial sources are necessarily more heterogeneous. Cars primar- ily pollute the air, while firms pollute all media, and with a wider range of pollutants. Firms' technologies, equipment, and use of inputs vary much more, mainly because of their more varied outputs. Thus, a priori, it would appear more dangerous for the planner to assume he knows the emission control opportunities open to firms than it would be in the case of vehicles. Another important difference is that firms' quantitative impact varies mnuch more, so that a few sources may dominate as contributors to a pol- lution problem. In Java two pulp factories represent 37 percent of indus- trial discharges (measured as biological oxygen demand) in one river, and two other factories account for 92 percent in another stream. According to a registry of the ninety largest polluters in the Valley of Mexico, two firms emit 95 percent of total dust emissions, and 10 per- cent of firms emit 99.5 percent. By contrast, even the most polluting of vehicles will-individually-play only a minor role. Even if one selected the most-polluting vehicles in Mexico City, more than 100,000 vehicles would be needed to account for 50 percent of total emissions. These two features distinguishing industrial polluters, or fixed-point sources as they are often called in the literature, have implications for 36 CONTROL OF POLLUTION FROM INDUSTRIAL SOURCES 37 pollution control policy. The greater heterogeneity of sources means that the planner will have a weak basis for assessing individual emis- sions and control options. Thus, if the planner tried, through sampling and a rough categorization of polluters, to make prescriptions for each-in terms of allowable emissions or of abatement technologies to be applied-it would be on the basis of grossly inadequate information. The resulting emission reduction would, consequently, come at a high cost. By contrast, the greater dominance of a few polluters makes it eas- ier to devise strategies for cost-effective emission reduction. The plan- ner might, for instance, commission engineering studies of a firm's or a sector's emissions and control opportunities and so bridge part of the knowledge gap between the sector and himself. In industrial as well as developing countries, pollution control agen- cies issue standards for industrial pollution control whereby a firm's "allowable" emissions or practices are determined on the basis of rough categories (such as "textiles," as reported by the firm). Stan- dards and regulation in developing countries are, at best, determined on the basis of a study that has sampled a few firms in the category. More often, the standards are based not on local studies but on studies from other countries, without much questioning of their adequacy or relevance for local conditions. Also, plant-specific studies may be done on large polluters, attempting to take into account factory-specific and recipient-specific conditions. Such studies then form the basis for plant-specific regulation, allowable time for attaining compliance, and so on. Thus there is ample evidence that the agency recognizes the costs associated with the information gap. We argue in this chapter that policy instruments are available for dealing with such problems but that their full potential is not exploited. General Economic Policies, Input and Output Taxes, and Regulation If the planner, or society, has an interest in total emissions but lacks information about individual emissions, he will inevitably find him- self seeking proxies not only for individual emissions but also for abatement opportunities and costs.1 Thus he is immediately taken far away from a situation with full information, in which he would include only information on individual emissions, or damages, in the incentives scheme with which he confronts polluters. In the case of vehicles (chapter 2), the planner conducted a study to learn about the determinants of emissions before deciding to work with manufactur- 38 TAXING BADS BY TAXING GOODS ers to encourage production of cleaner vehicles, to intervene in fuel markets to tax fuels and make them cleaner, and to use periodic test- ing to encourage proper maintenance and use. Similarly, for firms the planner needs to have an idea about how emissions are generated in order to know where and how to squeeze emissions. Kopp (1992) describes the planner as needing an expanded- although traditional-model of the firm. He includes residuals among the variables for which demand and supply functions are specified. As an example, if the expanded model shows that emissions are highly sensitive to the characteristics of a firm's capital equipment, equipment could be a useful point of intervention-and monitoring-in the incen- tive scheme. Similarly, should the expanded elasticity matrix indicate that emissions-say, of sulfur-are sensitive to the relative prices of sul- fur-rich fuels, taxation of these could provide emission reductions in a way that greatly economizes on monitoring and enforcement. Economic studies of pollution often use quite superficial models of emission generation. For instance, marginal emission factors may be assumed to be fixed by sector according to either output or fuel use, abstracting from much of the variation among plants in characteristics and behavior. They will therefore be inadequate for analyzing policies aimed at modifying the determinants of emission factors at the firm level. By contrast, such models have credibility in analysis of policies that would mainly change incentives at broader levels. In the case of output-based emission factors by sector, the models are adequate for simulating the effects of sector-specific taxes, trade policy, and indus- trial policy. In the case of emission factors linked to fuel use, the mod- els may also simulate well the effects of energy-pricing policy and con- servation initiatives. A major research effort in the World Bank (Hettige and others 1995) has estimated sector-specific emission coefficients on the basis of U.S. manufacturing data. Using such emission coefficients, Ten Kate (1993) studied how industrial pollution had developed in Mexico in the period 1950 to 1989. Growth in manufacturing over the period was remarkable-a tenfold increase in output. But Ten Kate observed that pollution had increased also because of an uninterrupted change in output composition, increasing the share of the more-polluting sectors. Surprisingly, Ten Kate found no breaks in this trend that could be asso- ciated with the distinct changes in economic policy over the years, from an emphasis on import substitution, through the period domi- nated by rising petroleum revenues and public sector expansion, to stabilization and liberalization in the 1980s. He found evidence that Mexico, through low domestic energy pricing, had "escaped" the envi- CONTROL OF POLLUTION FROM INDUSTRIAL SOURCES 39 ronmental benefits associated with rising world energy prices after 1973. Mexican industry increased its average energy intensity by 6 per- cent from 1970 through 1990, a period in which OECD countries reduced their average energy intensity by 38 percent. Apart from this observation, Ten Kate found that rapid expansion in a few highly pol- luting sectors, such as agro- and petrochemicals, drove much of the increase in average emission coefficients. Since those are sectors with predominant public sector influence, and domestic energy pricing has relied on public subsidies (equivalent to 4-7 percent of gross domestic product, or GDP ), Ten Kate left open the possibility that domestic pol- icy had a major influence on the increased emissions. Energy Conservation and Interfuel Substitution While the differences between sectors in terms of their emission inten- sities are important, it is highly desirable to provide more sophisticated incentives in order to induce responses along a greater range of possi- bilities than simply varying sectoral output levels.2 One way of doing that, short of monitoring emissions from each firm, is to exploit the link to energy use, which is particularly strong for air pollution. For industrial as well as for other polluting activities, it is cost- effective to pursue emission reductions through technical controls, as well as through demand reductions. The marginal costs of emission reductions through these two means will be equalized if emission taxes (or tradable permits) are used. Otherwise, presumptive taxes on polluting goods should be used, complementing mandated technical controls to equalize marginal costs. Cost-effective programs (and, among these, optimal programs) are characterized by tax levels on pol- luting goods (t) which, when normalized by marginal emission coef- ficients (e'), equal marginal costs of emission reductions through tech- nical controls: tl/e. = c,. The formula emphasizes that a polluting good should be taxed in proportion to its emission coefficient as long as emissions themselves are not taxed. For firms, we shall not here develop technical control cost curves, as we did for automobiles. We shall, however, investigate the scope for "purchasing" emission reductions via taxation of a particularly impor- tant category of polluting goods-fuels-using a rather simple model of how air pollution emissions from manufacturing respond to price changes:3 (3.1) tdpl ,, ax,, all];)a'( - 40 TAXING BADS BY TAXING GOODS where et, is emissions of pollutant mi, summation is over a set of inputs categorized as polluting (fuels), 'e,Z,/dxi(W) is the marginal emission coefficient for fuel i when a, characterizes the technology, and axi/apj is the price responsiveness of demand of fuel i with respect to a price p,. We thus turn to the topic of emission coefficients and demand elastic- ities for fuels in manufacturing. On the link between fuel use and emissions, it is worth noting that the technically based literature (summarized, for instance in u.s. EPA 1986) usually projects emissions by constant emission factors, specified in grams of pollutant per unit of fuel. These will generally vary by measurement, in part according to identified characteristics (for burn- ers, according to configuration and firing conditions; for coal, accord- ing to ash content, and so on) but will be averaged when applied to a population of polluting sources.4 The use of fuel consumption as the basis for emission models has traditionally been motivated mostly by technical reasoning. Experi- ence about emission rates from different types of equipment and pro- cesses is then based on sample testing or theory, and fuel throughput may be viewed as a proxy for the utilization of the equipment. An alternative check, to our knowledge previously untried, is to regress reported emission intensities against information on energy use and other possible determinants. For PM-10 (small dust particles), arguably the air pollutant of greatest concern in many developing countries (see Ostro 1994), the energy cost share is highly significant, with a positive sign, in explaining variations in emission intensities.5 This holds whether or not other variables, such as dummies for the Standard Industrial Classification (sic) industry code, cost shares for material inputs, and the like, are included among the explanatory vari- ables. Moreover, many sic industry dummies that are not significant in explaining variations in PM-10 emission intensities when the energy cost share is not included in the model are significant when energy is included. Those industries can be said to represent "clusters" in terms of PM-10 emission intensities, but only if the variation due to differ- ences in energy intensity is taken into account. A similar test has yet to be performed using data on different types of fuel as the explanatory variables, since such data have not been merged with emission data. Given that different fuels pollute differently, the scope for interfuel substitution and energy conservation will be an important determinant of the effectiveness with which fuel taxes can be used to reduce emis- sions. To estimate the substitutability of different fuels, we have resorted to both econometric and engineering studies. Econometric studies do not provide information on emissions and often require abstractions CONTROL OF POLLUTION FROM INDUSTRIAL SOURCES 41 where specificity would be desirable for the purposes of this study. (See, for example, Kopp 1992 on the eclectic use of estimates based part- ly on econometric studies, partly on technical studies and expertise.) The experience of the OECD countries since the 1973 oil price shock indicates that the possibilities for energy conservation and interfuel substitution are quite high. Between 1971 and 1988 manufacturing out- put in OECD countries rose 62 percent, while energy use (measured in tons of oil equivalent, or TOE) remained unchanged, implying a 38 per- cent reduction in average energy intensity (Bacon 1992). By contrast, in Mexico, where energy prices were suppressed domestically, the ener- gy intensity of manufacturing increased during the same period (Ten Kate 1993). Using separability assumptions and flexible functional forms, Fuss (1977) and Pindyck (1979) estimated own-price elasticities for individual fuels in manufacturing to be in the range -0.7 to -2.2. Own-price elasticities for aggregate energy were estimated at -0.5 by Fuss and -0.7 by Pindyck. The whole set of own- and cross-price elas- ticities is relevant for the responsiveness of emissions of each pollutant to a specific fuel price, but the price elasticity for aggregate energy has a more immediate interpretation: a price elasticity for aggregate ener- gy of 0.5 implies that an equiproportionate price increase of 20 percent for all fuels (and electricity) would reduce demand for all fuels by 10 percent and, consequently, emissions of all pollutants by 10 percent.6 The evidence from developing countries has been more limited. An important study on Indonesian manufacturing (Pitt 1985) used plant- level data over three years (1976-78). Pitt estimated the own-price elas- ticity for aggregate energy to be between -0.1 and -0.8, overlapping the Fuss-Pindyck range. He was able to estimate models that distin- guished between subsectors of manufacturing, and, in general, he found higher elasticities for the more energy-intensive sectors. Moss and Tybout (1994) studied how much of the change in the energy intensity of manufacturing in Chile has been due to shifts in output levels among subsectors and how much to changes in input use at the plant level. They concluded, perhaps surprisingly, that much of the change has come from shifts in input use at the plant level. This is an important finding, since many econometric studies based on aggre- gation over firms and sectors are unable to distinguish flexibility with- in firms and sectors from shifts between energy-intensive sectors and other sectors. A subsequent study (Guo and Tybout 1994) estimated interfuel substitutability for selected subsectors on the basis of firm- level observations and found that there is considerable substitutability but that it varies by subsectors and across plants. (The estimates are also sensitive to the choice of model for estimation.) 42 TAXING BADS BY TAXING GOODS Eskeland, Jimenez, and Liu (1994) estimated, for manufacturing, own- and cross-price elasticities for different fuels in Chile and Indonesia, with a view toward assessing the potential of fuel taxes for reducing emissions. A model with separability between an energy aggregate and other inputs yields various results, depending on the assumed policy experiment. (The methodology follows that of Fuss 1977.) One possibility is to assume that aggregate energy is held con- stant; another is to assume that output is constant but that substi- tutability between energy, materials, and labor occurs. When aggre- gate energy is held constant (the "energy submodel"), the relative price of one energy source (a fuel, or electricity) is increased, while the price and use of aggregate energy is constant. When only output is constant, the price of the energy aggregate changes when an individu- al fuel price changes. For Chile the authors report on estimates based on both assumptions, while for Indonesia only the energy submodel could be estimated. Table 3.1 presents estimates of Chilean price elasticities of demand for the first case, in which aggregate energy is held constant. The six energy sources in the original study have been aggregated into three to provide a simpler display of policy tools. Note that all the own-price elasticities are negative and are fairly large in absolute value. At such a general level, the results are in line with earlier research (for instance, Pindyck 1979). In the full elasticity matrix, twenty-four of the thirty-six elasticity estimates are significantly different from zero at the 5 percent level and an additional two at the 10 percent level. The fossil fuels are sorted according to the intensity of their TSP (total suspended particles, or "dust") emissions, with coal, fuel oil, and "grouped fuels" (the lat- ter includes wood and coke) comprising heavy fuels, while diesel and natural gas comprise light fuels. The elasticities in table 3.1 are calcu- Table 3.1 Energy Group Elasticities, with Energy Constant, Chile Price of Price of Price of DeOiaind electricityl heavy ftuels light fite/s For electricity -0.98 0.77 0.22 For heavy fuels 0.61 -0.89 0.28 For light fuels 0.20 0.53 -0.74 Note: Heavy fuels are coal, fuel oil, and grouped fuels; light fuels are natural gas and diesel. Elasticities are combinations of elasticity estimates in a six-fulel matrix. Most of the elasticities are significant at a 5 percent level. Souurce: Eskeland, Jimenez, and Liu 1994. CONTROL OF POLLUTION FROM INDUSTRIAL SOURCES 43 Table 3.2 Emission Elasticities, with Energy Constant, Chile Price of Price of Price of Emissiotns electricitv heavy fuels liglht Jiels Electricitv, polluting SOx 0.07 -0.19 0.11 TSP -0.14 -0.09 0.24 Electricity, non polluiting SOx 0.50 -0.58 0.07 TSP 0.48 -0.73 0.25 Souirce: Eskeland, Jimenez, and Liu 1994. lated under the assumption that prices for various fuels in a group increase proportionally.7 As an example, if prices for heavy fuels increase by 10 percent, demand for electricity and light fuels will increase by 7.7 and 5.3 percent, respectively, while demand for heavy fuels will fall by 8.9 percent. What are the implications of these elasticities for the responsiveness of emissions to price changes? The answer depends to a great extent on what happens in the electricity sector, and there are good reasons for treating that sector separately. First, the electricity sector is reasonably easy to sujbject to separate policy instruments, including those based on emission monitoring. Second, since electric power plants may consist in part of tall-stack units located far away from the most polluted and pop- ulated areas, emissions from the sector should perhaps not be treated in the same way as emissions generated directly in the manufacturing sec- tor (as each power plant could be treated separately). Table 3.2, for these reasons, provides emission elasticities under alternative assumptions. The assumption "electricity, polluting" assumes that electricity at the margin is produced 50 percent by coal-fired, uncontrolled plants and 50 percent by nonpolluting technologies. The alternative assumption is that the electric power sector is nonpolluting at the margin. Again using a 10 percent increase in prices for heavy fuels as an example, SOx emissions would be reduced by 5.8 percent and TSP emissions by 7.3 percent under the assumption that the electricity sec- tor is nonpolluting at the margin. Price increases for light fuels would, as expected, increase emissions under all assumptions, since they would lead to substitution toward more-polluting energy sources. Increases in electricity tariffs would increase emissions significantly if electricity is assumed to be nonpolluting. If electricity is assumed to be polluting at the margin, the tariff increases would lead to a small 44 TAXING BADS BY TAXING GOODS increase in SOx emissions and to a small reduction in TSP emissions. To sum up, price increases for heavy fuels are quite powerful in reducing emissions if and only if the electricity sector is nonpolluting at the mar- gin. The reason, as can be seen in table 3.1, is that much of the induced demand response is toward electricity. Table 3.3 shows the calculated group demand elasticities for Indone- sia. In comparison with those for Chile, there are differences, but neg- ative own-price elasticities and positive or small cross-price elasticities again dominate. The own-price elasticity for heavy fuels is larger in absolute value, making tax instruments more powerful. Emission elasticities are given in table 3.4. As in the case of Chile, price increases for heavy fuels reduce emissions for both SOx and TSP, but the magnitudes of the effects are greater for Indonesia. Referring back to tables 3.1 and 3.3, one sees that Indonesia has a higher own- price elasticity for heavy fuels (in absolute value) and a smaller cross- price elasticity with electricity. Thus, emissions respond more to an increase in prices of heavy fuels. Under the assumption that the elec- tricity sector is nonpolluting at the margin, emission elasticities with respect to the price of heavy fuels are -1.02 for SOx and -1.52 for rsr. Even under the assumption that electricity is polluting, emission elas- ticities with respect to the prices of heavy fuels for Indonesian manu- facturing are quite large, in the range of -0.8 for SOx and TSP. These estimates are based on the assumption that the price of the aggregate "bundle" of energy does not change when a fuel price rises.8 However, unless compensatory price changes are made for other fuels, firms will also alter their use of aggregate energy by substituting labor and materials (still holding output constant). For Chile the necessary deflators for outputs and other inputs could be constructed, and a short-run model with three aggregate inputs-energy, labor, and mate- rial-could be estimated. Table 3.3 Energy Group Price Elasticities, with Energy Constant, Indonesia Price of Price of Price of Demntid electricityI lcavy ftietls lighti fiels For electricity -1.02 0.26 0.76 For heavy fuels 0.24 -1.41 1.17 For light fuels -0.01 0.37 --0.35 Note: See notes to table 3.1. Source: Eskeland, Jimnenez, and Liu 1994. CONTROL OF POLLUTION FROM INDUSTRIAL SOURCES 45 Table 3.4 Emission Elasticities, with Energy Constant, Indonesia Price of Price of Price, of Emfissions lehctricity iheazy filel; light fuels ElectricitY/, pol0lting SOx 0.02 0.81 0.79 rsp -0.21 -0.86 1.07 Electricit., nionpolluting SOx 0.22 -1.02 0.80 TSP 0.26 -1.52 1.25 Source: Eskeland, Timnenez. and LiLl 1994. Table 3.5 displays the own- and cross-price elasticities in this aggre- gate model. The own-price elasticity of aggregate energy is consider- able, at 0.63, between Pindyck's 0.7 and Fuss's 0.5. This indicates that demand for all fuels-and emissions of all air pollutants-would fall by 6.3 percent if all energy prices increased by 10 percent. We may also notice that the cross-price elasticity between energy and labor is posi- tive and substantial, indicating that subsidies to energy would be a poor strategy if increased employment were a policy objective. With aggregate energy adjusting, price elasticities for individual fuels will be modified-own-price elasticities are higher in absolute value, while cross-price elasticities are somewhat dampened and may Table 3.5 Manufacturing: Elasticities for Aggregate Inputs, Chile Aggregate Pricet o'f energy mlaterial DePatnid price Wage inpluts For energy -0.63 1.57 -(1.95 (0.36) (0.49) (0.31) For labor 0.52 -1.89 1.37 (0.16) (0.34) (0.30) For materials -0.07 0.32 -0.25 (0.02) (0.07) (0.08) Note: Figures in parentheses are standard errors. All coefficients are significant at the 5 percent level. Souirce: Eskeland, Jimenez, and LiLl 1994. 46 TAXING BADS BY TAXING GOODS Table 3.6 Emission Elasticities, with Energy Adjusting, Chile Priec Price of Price of All of- leaz!vy light enerrg Et?fissionis electricity fuiels fuels prices EleXctricitly, polilittitig SOx -0.15 -0.52 0.04 --0.63 Tsp -0.37 -0.42 0.16 -0.63 ElectricitYl, nlolpolluting SOx 0.28 -(.91 -0.01 -0.63 TSP 0.25 -1.06 0.18 -0.63 Source: Eskeland, Jimenez, and Liu 1994. switch sign if they are positive in the energy submodel. The intuition is that it is easier to change demand for a taxed fuel when one can vary the use of other (nonenergy) factors as well. Energy substitution also reduces the tendency for demand to spill over to other fuels. The implications for policy are seen in the emission elasticities in table 3.6. The greater own-price elasticities and the lower cross-price elasticities increase the elasticities of emissions with respect to the price of heavy fuels. Under these assumptions, a tax increase for heavy fuels can be a potent instrument for pollution control, whether or not electricity is polluting at the margin. Exploration of General Equilibrium Effects Up to now, we have been discussing the effects of fuel taxes on emis- sions in a partial equilibrium context. For transport, we emphasized the role fuel taxes could play in reducing output in the polluting activity, while other instruments could be used to make each "polluting plant" less polluting per unit of output. In the applications to industrial emis- sions, by contrast, the emphasis has been on the extent to which the same amount of aggregate output from manufacturing can be delivered with less use of energy, or with the use of less-polluting types of ener- gy. In this section we extend these explanations with a computable gen- eral equilibrium (CGE) model of the Indonesian economy. Increased input taxes, if uncompensated, will alter output levels, since they change production costs. In the case of a fuel tax, fuel- intensive producers and sectors will initially experience the greatest increase in costs and will thus be under more pressure either to change their input mix or to raise output prices. The latter will occur to a great CONTROL OF l'OLLUTION FROM INDUSTRIAL SOURCES 47 extent if fuel-intensive producers have little flexibility in input choice, and one would expect to see reduced equilibrium output levels in fuel- intensive sectors. The new equilibrium, with a different output combi- nation and different emission levels, will reflect the opportunities for change open to all agents in the economy. The CGE model of the Indonesian economy is designed to highlight the features that are important for energy use (Devarajan and Lewis 1994). In particular, the model distinguishes among three types of pri- mary energy (oil, gas, and coal) and three energy products (electricity, gas, and refined oil products) and incorporates the existing system of taxes and subsidies in the economy. The rest of the model is a fairly standard CC;E model, with twenty-six productive sectors, four house- holds, five labor groups, and fixed, sector-specific capital. The model enables us to go deeper than the rather abstract "aggre- gate manufacturing output" discussed above. Table 3.7 shows the greater degree of specificity in terms of outputs. Eight sectors consti- tute manufacturing in a traditional sense, while six "sectors" convert petroleum resources to refined products and liquefied natural gas (LNG). Four sectors are engaged in extraction of natural resources, and two sectors represent agricultural activities. The first column shows emissions of TSP from each sector, as modeled on the basis of fuel use. In manufacturing, three sectors stand out: chemicals and fertilizer, nonmetallic minerals, and basic metals. Importantly, two "miscella- neous" sectors (electricity and transport) are the two largest in terms of TSP emissions, while construction ranks equally with a manufactur- ing sector such as food processing. The second column shows the direct emission coefficients, or total emissions from the sector divided by the sector's gross output. The CGE model allows us to evaluate the importance of intermediate deliveries between sectors. The third column presents the sum of the direct and indirect emission coefficients. That is, it includes not just the direct effect on emissions of increasing output in a sector by one unit but also the effects of the interindustry demand stimulated by this increase in output. These are important when looking at the economy with a blunt planning perspective. The column informs us, for instance, that an expansion of a sector such as metal products and machinery would increase output from polluting sectors and thus increase pollu- tion, even though the sector itself does not have a very high emission coefficient. This may be relevant if major industrial expansion plans are considered and adequate policy instruments are riot in place to influ- ence the levels of pollution directly. If polluting suppliers to the indus- try in question face appropriate tax levels, however, expansion plans 48 TAXING BADS BY TAXING GOODS Table 3.7 TSP Emissions by Sector in the Computable General Equilibrium Model, Indonesia Tonal eni issionis (tolls Direct Totnl CGE Sector per yJear) coefficient coefficient coefficient Manuiifactuiring Food processing 44.7 0.003 0.015 - 0.006 Textiles and leather 9.8 0.003 0.039 0.003 Wood and furniture 10.2 0.004 0.024 - 0.012 Paper and related industries 3.5 0.003 0.042 -0.002 Chemicals and fertilizer 157.5 0.041 0.095 0.016 Nonmetallic minerals 220.0 0.136 0.200 0.076 Basic metals 164.5 0.090 0.166 0.028 Metal products and machinery 14.8 0.002 0.056 0.002 Refininlg Gasoline 0.1 0.000 0.013 -0.095 High-speed diesel 0.1 0.000 0.013 -0.017 Industria] diesel 0.0 0.000 0.013 -0.211 Kerosene 0.1 0.000 0.013 -0.171 Fuel oil 0.1 0.000 0.013 -0.570 LNG 2.8 0.001 0.004 -0.003 Agricuiltutre anid tnatural resources Food agriculture 0.9 0.000 0.007 -0.010 Traded agricultlre 4.1 0.000 0.010 -0.010 Oil extraction 6.4 0.001 0.004 - 0.000 Natural gas 1.1 0.001 0.004 -0.017 Coal mining 39.1 0.515 0.603 3.484 Other mining 5.5 0.004 0.003 - 0.020 MiscellanIeoUs Electricity and gas 1,668.9 0.999 1.225 0.586 Water supplv 0.2 0.002 0.058 0.006 Construiction 42.2 0.002 0.043 -0.005 Transport 692.2 0.073 0.088 0.020 Trade and storage 5.0 0.000 0.015 -0.007 Services 6.0 0.000 0.025 0.( Total 3,037.3 Soiirce: Lewis 1994. CONTROL OF POLLUTION FROM INDUSTRIAL SOURCES 49 for a downstream industry need only pass the test of paying for the effects of its own emissions (that is, according to direct coefficients). The calculation of the total coefficient in the third column assumes implicitly that resources for expanding output are available in perfect- ly elastic supply in this economy. This is why all the total coefficients are larger than the direct coefficients. In reality, however, when a sector expands output, it draws resources from other sectors by bidding up factor prices. As a result, some other sectors will contract, and the over- all effect on pollution may be positive or negative. The fifth column dis- plays this overall effect when these resource limitations are brought into the picture.9 Note that some of the coefficients are negative because the higher factor costs resulting from the output expansion in the sec- tor concerned lead to a contraction in more-polluting sectors. In partic- ular, some of the petroleum product sectors-such as kerosene and industrial diesel-actually have a negative coefficient because an increase in their output crowds out other, more polluting products. In general, of course, taxing the polluting activities themselves according to their pollution and damage levels is a good idea. In the case of fossil fuels, since pollution is related to the use of these fuels (to the outputs from the refining sectors, strictly speaking), taxation of fuels will be a tax on polluters and will have much of the desired effect in terms of inducing using sectors (and users of using sectors' outputs) to reduce polluting inputs. We turn, therefore, to the impact of changes in tax rates on energy use and emissions. What would be the effect of eliminating the com- plex system of taxes and subsidies in Indonesia's energy sector? The answer, not reported here in detail, is that emissions of most pollutants would increase. The main reason is that many of the less-polluting fuels are subsidized at the outset (kerosene and tvo types of diesel, in particular), and some of the heavier fuels are taxed. Thus, elimination of taxes and subsidies makes the more-polluting fuels more attractive, contributing to emissions. Table 3.8 General Equilibrium Arc Elasticities, Indonesia Price of Price of Pricc of Demand eltctricitiy heeavy fiels lightfuels For electricity -0.41 -0.04 0.04 For heavy fuels -0.08 -0.43 0.45 For light fuels 0.00 0.02 -0.14 Source: Eskeland, Jimenez, and Liu 1994. 50 TAXING BADS BY TAXING GOODS To evaluate the impact of changes in tax rates, we display in table 3.8 general equilibrium arc elasticities for grouped energy sources. As before, own-price elasticities are negative, and cross-price elasticities are positive or small. Compared with the partial equilibrium elastici- ties for the manufacturing sector (see table 3.3), the general equilibri- um (tax) elasticities are all smaller in absolute value. To obtain a better understanding of this, we turn to the differences between the models. The general equilibrium model is different from the partial equilibri- um model in three important ways. First, the model attempts to describe the Indonesian economy as a whole, rather than only the input side of the manufacturing sector. This implies coverage of additional sectors, but it also means that manufacturing industries may adjust their output levels in addition to their energy use. Second, the supply of fuels is not perfectly elastic, implying that users will face a price increase smaller than the increase in the fuel tax. Refining is modeled as one sector trans- forming crude oil into a range of petroleum products, with a joint out- put, constant elasticity of transformation technology. Third, the model incorporates budget constraints for the various "households." Impor- tantly, when energy tax rates are changed, the model adjusts income tax rates so that government revenues are presered. How might these differences be expected to influence the respon- siveness reflected in the elasticities? The added flexibility in aggregate energy use and in output composition should make the CGE model more flexible than the partial equilibrium model, but the final result would also depend on the flexibility within the additional, nonmanu- facturing sectors.10 Fuel supply that is imperfectly elastic, by contrast, would tend to make quantities less responsive to fuel tax changes in the CGE model than in the partial equilibrium model. In particular, imperfect elasticity of transformation between petroleum products limits substitution between fuels, since suppliers will prevent user prices for fuels from moving too far apart. A similar effect hurts the effective flexibility that users have in their choice between heavy fuels and electricity: CGE (cross-tax) elasticities are close to zero, as opposed to 0.25 in partial equilibrium. The reason is that in the CGE model elec- tricity producers also face higher costs when the prices of heavy fuels increase. They will raise electricity rates and thus dampen the result- ing change in relative user prices between heavy fuels and electricity. Finally, the compensatory changes in income tax rates is an assump- tion chosen to give specific meaning to the allowed flexibility in out- put composition. In comparison with an alternative-for example, let- ting the government keep the proceeds from a fuel tax increase-the result may not be very different: it would depend on whether the CONTROL OF POLLUTION FROM INDUSTRIAL SOURCES 51 marginal consumption basket for households is more or less energy- intensive than that of the government. The result is that the CGE model displays notably less flexibility than the partial equilibrium model for the manufacturing sector. In accor- dance with the above discussion, the added sectors are as flexible in their energy use as manufacturing, so the imperfectly elastic supply of fuels, the joint production in refining, and the cost pass-through in electricity production are responsible for this lower flexibility. Table 3.9 displays the resulting elasticities for emissions of TSP and SOx. In this model, the electricity sector chooses its inputs endoge- nously, and the line "electricity, nonpolluting" simply assumes that tight emission controls apply when electricity is based on fossil fuels. As table 3.9 shows, the general equilibrium model nevertheless dis- plays a significant response of emissions to changes in fuel and energy prices. For example, assuming that electricity is polluting, a 10 percent price change for heavy fuels will reduce emissions of TSI' by 4 percent and of SOx by 2 percent. For larger price increases, the emission reduc- tions can be expected to increase in the same proportion (or more), although the empirical basis for such estimates is weaker. The elastici- ties are, however, noticeably smaller than for the partial equilibrium model. An important contributing factor is that the sectors supplying fuels are modeled as imperfectly elastic in their supply, thereby limiting the changes in fuel prices caused by changes in fuel tax rates. This reflects assumptions about technology and about the trade regime for petroleum products. If Indonesia were to trade crude oil and products without barriers, users would face almost perfectly elastic fuel supplies. Another lesson in instrument choice is seen by observing that the cross- price elasticities between heavy fuels and electricity would be positive Table 3.9 General Equilibrium Arc Emission Elasticities, Indonesia Price of Price of Price of Emissions elcctricitiy heavyfilels light fuels Electricityt, pollitting SOx -0.04 -0.20 0.14 TsP -0.08 -0.41 0.41 Elcctricity, nonoslfit)luing SOx 0.06 -0.19 0.07 TSP 0.06 -0.55 (.50 Sourcc: Eskeland, Jimenez, and Liu 1994. 52 TAXING BADS BY TAXING GOODS if taxes on heavy fuels were not applied to the electricity sector's pur- chases. Such an exemption, or refunds, would be justified if the elec- tricity sector were applying tight emission controls. Thus, if the elec- tricity sector is "nonpolluting" and can purchase heavy fuels without paying the emission-presumptive taxes, the SOx and TSP emission elas- ticities with respect to the price of heavy fuels would be larger in abso- lute value than the indicated 0.19 and 0.55. In sum, the evidence points to significant potential for using fuel taxes to combat air pollution. Partial equilibrium analysis and general equilibrium simulation show significant responsiveness of pollutant emissions with respect to changes in tax rates for fuels and electricity. The general equilibrium model serves to remind us of the importance of sectors other than transport and manufacturing and also of the importance of the technical and policy conditions governing supply of fuels and electricity. Clearly, a restrictive trade regime for fuels can limit an economy's responsiveness to fuel and energy tax changes. Furthermore, if the electricity sector can be emission controlled and (then) exempted from presumptive fuel taxes, the scope for substitu- tion in using sectors will be enhanced, and emissions will respond more strongly to fuel taxes. Apart from these insights, the general equi- librium simulations show us much the same types of effects as those observed in the partial equilibrium models for manufacturing (and for transport, although the latter were from another country). It is worth noting that much of the relevant flexibility lies in substi- tution between heavy fuels (coal and fuel oil) and electricity. This flex- ibility renders taxation of heavy fuels a more powerful instrument if pollution from the electricity industry is low or is taken care of by other instruments. Thus, if the costs of monitoring and controlling emissions from electricity plants are such that the power sector will not be too polluting at the margin, it would lend additional strength to simple fuel tax instruments in the pursuit of emission reductions from other users of energy. Our analytical framework indicates that each fuel should be taxed according to its direct emission coefficient, with damage estimates pro- viding weights by pollutant. Whether emphasis should be placed on sulfurous emissions, dust emissions, or other pollutants depends on local conditions. Many studies now point toward emphasis on dust (in particular, the smaller particles, PM-10) in urban air pollution. How- ever, taxation either of sulfur or of dust will often single out the same fuels (called heavy fuels here), at least until the taxation scheme and the market operators become sophisticated enough to distinguish more finely and encourage fuels with lower ash and sulfur content. CONTROL OF POLLUTION FROM INDUST'RIAL SOLIRCES 53 When tax rates are levied according to each fuel's characteristics, the responsiveness of emissions to price changes will be greater than dis- played in this section, where uniform price changes for a broader range of fuels were considered. Although the general lesson from this section is similar to the title of the paper-tax "bads" by taxing bad goods more-there are many possible avenues toward emission reductions that cannot be stimulated with fuel taxes. Thus, we highlight the value of complementing this instrument with others and of refining the instruments as institutional and technical capacity develops. An important question will always be: at what point can the pollution control agency offer partial refunds of fuel taxes to users who prove that they pollute less than the average user of the fuel? Only at that point will fuel taxes, in themselves, add encouragement for control technologies to the encouragement they provide for fuel economy, demand reduction, and fuel switching. At earlier stages, and perhaps for a long time, other instruments that stim- ulate cleaner technology (such as emission standards, which are pop- ular in the real world) will be valuable supplements in a program of emission-presumptive fuel taxes. Notes 1. As noted in chapter 1, the proxies addressed bv indirect instruments are chosen not only to be related to emissions but also to emphasize the choices the polluter may make that actually change emissions. 2. Indeed, general equilibrium simulations have shown surprisingly little responsiveness in economies' sectoral shares, although a common conjecture is that this impression will change as models' ability to capture dynamic effects improves. See, for instance, Bell and Srinivasan (1984), Dahl, Devarajan, and van Wijnbergen (1986), and Go (1994) on the literature on tax and tariff reform and Hazilla and Kopp (1990) and Jorgenson and Wilcoxen (1990) on the impact of pollution control policies. Boyd, Krutilla, and Viscusi (1995) perform sensitivity analysis. 3. The simplifying assumption is that the marginal emission coefficient itself is not changed bv the price change: that is, *)cti/p= 0. See the discussion of equation 1.1. 4. A fuel may have different emission factors for different users, and the marginal emission coefficient for a population of users exposed to a fuel price change would then be a weighted sum of their individual emission factors, with their respective shares in total fuel demand change as weights. We apply an averaged emission factor and an aggregate demand elasticity 5. We are grateful to David Wheeler and associates, who, with the collabo- ration of the U.S. Bureau of the Census and EPA, have for the first time merged data on firms' costs, toxic releases, air emissions, and water pollutant dis- charges. We also thank Manjula Singh, who performed the regressions, for the 54 TAXING BADS BY TAXING GOODS opportunity to present these results. By emission intensities, we understand a measure of emissions divided by output. We have not performed the test for other air pollutants, but the energy cost share is an important determinant also for some other important pollutant measures. Among the main air pollutants, one would expect, on the basis of technical reasoning, the link betwveen PM-IO and energy use to be of "intermediate" strength-weaker than for SOx and CO2, for which strict proportionality bv fuel should hold (atoms are simplv passed through), but stronger than for volatile organic compounds, NOx, and CO, for which the transformations are very sensitive to firing conditions. 6. These models assume that different fuels form a separable homothetic aggregate called energy, so that fuel shares in the energy bill do not change unless relative prices between fuels change. 7. Aggregation is as follows: when several prices increase in the same pro- portion (say, prices for heavy fuels), a fuel's demand elasticity with respect to this aggregate price change is the sum of its elasticities with respect to the indi- vidual prices. The elasticity for an aggregate fuel (say, demand for heavv fuel) is the quantity-weighted sum of the elasticities for individual fuels in the aggregate. 8. The experiment entails making compensatory price changes for other fuels, so that the price of aggregate energy is held constant. 9. These coefficients are calculated as follows. Government demand for a sector's output is increased exogenously (financed through lump-sum taxa- tion). The model solves for the new set of output and factor prices that equate demand and supply in the new equilibrium (the one with the exogenously increased demand for output of one sector). The emissions associated with the new production structure are then estimated and compared with emissions in the old production structure. The difference is labeled the CGE coefficient in table 3.7. 10. The manufacturing sectors of the CGE model are calibrated with a view to the estimated parameters of the economiietric model, but no close mapping can be achieved because of the differences in structure. Most importantly, since the manufacturing sector in the estimated partial equilibrium model discussed above is aggregate manufacturing, it should display greater input demand flexibility than each manufacturing sector displays in the general equilibrium model, if the two are to approximate the same technology. (Some changes in output mix will be recorded as such in the CGE but will be masked as changes in input mix, holding "output constant" or "energy constant" in the partial equilibrium model.) 4. Conclusion The literature on pollution control has debated the merits of market- based instruments versus command-and-control approaches and has concluded that the former is generally preferred. This study has looked at the problem from a different angle-one of choosing a mix of instruments when continuous monitoring of emissions is impossi- ble. When viewed from this perspective, some of the command-and- control policies appear more sensible because they focus on processes, equipment, and machinery, giving the government valuable informa- tion about where it can intervene. For example, if an agencV cannot monitor emissions continuously but can inspect machinery periodical- ly and tax fuel use, emission standards and presumptive fuel taxes would be used. However, a focus on equipment, processes, and input use is only a partial solution to the problem, and this book has addressed the impli- cations of recognizing this fact. First, when continuous emission mon- itoring is feasible (and not costly), the government should monitor emissions only and not evaluate equipment, processes, and input use. Second, in the more realistic case in which continuous monitoring is costly, the government should focus on the factors causing emissions: the equipment and processes, on the one hand, and the scale of the activity, on the other. Pollution policy consists of incentives to make all activities cleaner (for instance by addressing equipment and process- es) and incentives to reduce the scale of the more-polluting activities. When only on-e set of policies is in place-for example, only standards requiring catalytic converters in cars-the introduction of demand- management instruments (such as a gasoline tax) can significantly improve welfare. These principles were illustrated using empirical information gleaned from case studies of Mexico City, Chile, and Indonesia. In ana- lyzing the options for controlling automotive air pollution in Mexico 55 56 TAXING BADS BY TAXING GOODS City, we showed that "cleaner cars" and "fewer trips" are two ways of achieving emission reductions and that therefore the contribution of each should be equated at the margin. Specifically, we observed that a gasoline tax, which will induce fewer trips, is part of a cost-effective control strategy. We calculated the size of such a tax, as well as the costs of not adopting it. Noting that industrial pollution sources are much more heteroge- neous, we examined the potential for fuel demand management in the industrial sector by estimating fuel demand elasticities based on firm- level data from Chile and Indonesia. The results indicated that firms have the flexibility to respond to fuel-price increases. They can switch from dirty to clean fuels, and they can reduce overall energy intensity. Finally, simulations with a general equilibrium model of Indonesia revealed that these fuel taxes will lead to similar responses when all of the adjustments in the economy take place. In fact, the economy shifts toward cleaner fuels and becomes less energy-intensive, both of which benefit the environment. The mobile-source example and the fixed-source example both illus- trate how effectiveness depends on the quality of monitoring. In Mex- ico City, while fuel taxes provide 20 percent of emission reductions, 80 percent is achieved by technical controls, mostly depending on testing and periodic inspections. For fixed sources, fuel and energy taxes work much better if the electricity sector can be kept tightly controlled and when highly controlled sectors can be rewarded with partial refunds of their presumptive tax payments. To be sure, these conclusions about indirect instruments for pollu- tion control (when emission monitoring is a problem) should be viewed as improvements and reform steps, rather than final solutions. As monitoring technologies and capacity improve, the instruments should shift toward policies based on actual monitoring of emissions. In the meantime, however, instruments aimed at increasing the marginal cost of polluting activities-such as taxes on marketed out- puts and inputs, and standards for equipment and processes-can reduce pollution cost-effectively, with modest monitoring require- ments. More refined instruments based on emission monitoring may be phased in as new technologies and information become available. Bibliography An asterisk (*) denotes a study performed as part of the World Bank research project. Alberini, Anna, Winston Harrington, and Virginia McConnell. 1993. "Determinants of Participation in Accelerated Vehicle Retirement Programs." Discussion Paper QE93-18. Resources for the Future, Washington, D.C. . 1994. "Estimating an Emission Supply Function from Accel- erated Vehicle Retirement Programs." Discussion Paper 94-09. Resources for the Future, Washington, D.C. Alberini, Anna, Alan Krupnick, Maureen Cropper, and Winston Har- rington. 1994. "Air Quality and the Value of Health in Taiwan." Prepared for the Eastern Economic Association, Boston, Mass. Resources for the Future, Washington, D.C. *Alberini, Anna, Gunnar S. Eskeland, Alan Krupnick, and Gordon McGranahan. 1995. "Determinants of Diarrheal Disease in Jakarta." Policy Research Working Paper Series 1568. World Bank, Policy Research Department, Washington, D.C. Anderson, Robert. 1990. "Reducing Emissions from Older Vehicles." Research Study 053. American Petroleum Institute, Washington, D.C. *Bacon, Robert. 1992. "Measuring the Possibilities of Interfuel Substi- tution." Policy Research Working Paper Series 1031. World Bank, Policy Research Department, Washington, D.C. Balcer, Yves. 1980. "Taxation of Externalities, Direct versus Indirect." Journal of Puiblic Economics 13:121-29. Beaton, S. P., G. A. Bishop, and D. H. Stedman. 1991. "Automobile Emissions in Mexico City." Report to Los Alamos Laboratories. University of Denver, Colo. Bell, Clive, and T. N. Srinivasan. 1984. "On the Uses and Abuses of Economywide Models in Development Policy Analysis." In Moshe 57 58 TAXING BADS BY TAXING GOODS Syrquin, Lance Taylor, and Larry E. Westphal, eds., Economic Struc- ture and Perforrmance: Essays in Honor of Hollis B. Chenery. Orlando, Fla.: Academic Press. Bovenberg, Lans, and Sijbren Cnossen, eds. 1995. Public Economics and the Environment in an Imperfect World. Boston, Mass.: Kluwer. Boyd, Roy, Kerry Krutilla, and Kip Viscusi. 1995. "Energy Taxation as a Policy Instrument to Reduce CO2 Emissions: A Net Benefit Analy- sis." Journal of Environmental Economics and Managemiient 29(1): 1-24. Dahl, Henrik, Shantayanan Devarajan, and Sweder van Wijnbergen. 1986. "Revenue-Neutral Tariff Reform: Theory and an Application to Cameroon." CPD Paper 1986-25. World Bank, Policy Research Department, Washington, D.C. *Eskeland, Gunnar S. 1992. "Attacking Air Pollution In Mexico City." Finance and Development 29(4):28-30. Eskeland, Gunnar S. 1994a. "The Costs and Benefits of an Air Pollu- tion Control Strategy for Santiago, Chile." In World Bank, "Chile: Managing Environmental Problems: Economic Analysis of Selected Issues." Report No. 13061-CH. World Bank, Latin America and the Caribbean Region, Washington, D.C. Also available from the World Bank Public Information Center. - . 1994b. "A Presumptive Pigovian Tax: Complementing Regu- lation to Mimic an Emission Fee." World Bank Economic Review 8(3):373-94. *Eskeland, Gunnar S., and Tarhan Feyzioglu. 1994a. "Is Demand for Polluting Goods Manageable? An Econometric Study of Car Own- ership and Use in Mexico." Policy Research Working Paper Series 1309. World Bank, Policy Research Department, Washington, D.C. - . 1994b "Rationing Can Backfire: The 'Day without a Car' Pro- gram in Mexico City." World Bank, Policy Research Department, Washington, D.C. *Eskeland, Gunnar S., and Emmanuel Jimenez. 1991. "Curbing Pollu- tion in Developing Countries." Finanice and Development 28(1):15-18. *. 1992. "Policy Instruments for Pollution Control in Developing Countries." World Bank Researchi Observer 7(2):145-69. *Eskeland, Gunnar S., and Chingying Kong. 1994. "Protecting the Environment and the Poor: A Welfare Function Approach to Air Pollution Control on Java." Presented at the annual meeting of the American Economic Association, January. World Bank, Policy Research Department, Washington, D.C. *Eskeland, Gunnar S., Emmanuel Jimenez, and Lili Liu. 1994. "Energy Pricing and Air Pollution: Econometric Evidence from Manufactur- BIBLIOGRAPHY 59 ing in Chile and Indonesia." Policy Research Working Paper Series 1323. World Bank, Policy Research Department, Washington, D.C. *Eskeland, Gunnar S., Tarhan Feyzioglu, Jose Miguel Sanchez, Carlos Francisco Aranda, and Bart Ostro. Forthcoming. "Air Pollution and Health Effects: A Study of Respiratory Illness among Children in Santiago, Chile." Policy Research Working Paper Series. World Bank, Policy Research Department, Washington, D.C. Freeman, A. Myrick III. 1982. Air and Water Pollution Control: A Bene- fit-Cost Assessment. New York: Wiley. Fuss, M. A. 1977. "The Demand for Energy in Canadian Manufactur- ing." Journal of Econometrics 5: 89-116. Go, Delfin. 1994. "External Shocks, Adjustment Policies and Invest- ment in a Developing Economy: Illustrations from a Forward- Looking CGE Model of the Philippines." Jouirnal of Development Eco- nomics 44 (August):229-61. Greenwald, Bruce C., and Joseph E. Stiglitz. 1986. "Externalities in Economies with Imperfect Information and Incomplete Markets." Quarterly Journal of Economics 101:229-64. *Guo, Charles C., and James R. Tybout. 1994. "How Relative Prices Affect Fuel Use Patterns in Manufacturing: Plant-Level Evidence from Chile." Policy Research Working Paper Series 1297. World Bank, Policy Research Department, Washington, D.C. Harrington, Winston, and Virginia McConnell. 1993. "Modeling In- Use Vehicle Emissions and the Effects of Inspection and Mainte- nance Programs." Discussion Paper QE93-02. Resources for the Future, Washington, D.C. Harrison, David Jr. 1977. "Controlling Automotive Emissions: How to Save More Than $1 Billion Per Year and Help the Poor Too." Pub- lic Policy 25(4):527-53. Hazilla, Michael, and Raymond Kopp. 1990. "Social Costs of Envi- ronmental Quality Regulation: A General Equilibrium Analysis." Journral of Political Economy 98(4):853-73. Hettige, Hemamala, Paul Martin, Manjula Singh, and David Wheeler. 1995. "The Industrial Pollution Projection System." Policy Research Working Paper 1431. World Bank, Policy Research Department, Washington, D.C. Jorgenson, Dale W., and Peter J. Wilcoxen. 1990. "Environmental Reg- ulation and U.S. Economic Growth." Rand Joiurnal of Economnics 21(2):314-40. *Kanninen, Barbara, and Michael Hanemann. Forthcoming. "Welfare Effects of Environmental Policies." Policy Research Working Paper Series. World Bank, Policy Research Department, Washington, D.C. 60 TAXING BADS BY TAXING GOODS *Kopp, Raymond J. 1992. "Economic Incentives and Point Source Emis- sions: Choice of Modeling Platform." Policy Research Working Paper 920. World Bank, Policy Research Department, Washington, D.C. *Krupnick, Alan J. 1992. "Measuring the Effects of Urban Transporta- tion Policies on the Environment: A Survey of Models." Policy Research Working Paper Series 1030. World Bank, Policy Research Department, Washington, D.C. Krupnick, Alan, and Margaret A. Walls. 1992. "The Cost Effectiveness of Methanol for Reducing Motor Vehicle Emissions and Urban Ozone." Journal of Policy Analysis and Management 11(3):373-96. Krupnick, Alan J., Margaret A. Walls, and Carol T. Collins. 1993. "Global Warming and Urban Smog: Cost-Effectiveness of CAFE Standards and Alternative Fuels." Energy Journal 14(4): 75-98. Lawson, Douglas R., Peter J. Groblicki, Donald H. Stedman, Gary A. Bishop, and Paul L. Guenther. 1990. "Emissions from In-Use Motor Vehicles in Los Angeles: A Pilot Study of Remote Sensing and the Inspection and Maintenance Program." Journal of the Air and Waste Management Association 40:1096-1105. *J. D. Lewis. 1994. "Energy Pricing, Economic Distortions and Air Pol- lution in Indonesia." Draft. Harvard Institute of International Devel- opment, Cambridge, Mass., and World Bank, Washington, D.C. McConnell, Virginia, and Winston Harrington. 1992. "Cost-Effective- ness of Enhanced Motor Vehicle Inspection and Maintenance Pro- grams." Discussion Paper QE92-18-REV. Resources for the Future, Washington, D.C. -. 1993. "Cost Effectiveness of Remote Sensing of Vehicle Emis- sions." In Richard F. Kosobud, William A. Testa, and Donald A. Hanson, eds., Cost Effective Control of Urban Smog. Chicago, Ill.: Fed- eral Reserve Board of Chicago. *Moss, Diana, and James R. Tybout. 1994. "The Scope for Fuel Substi- tution in Manufacturing Industries: A Case Study of Chile and Colombia." World Bank Economic Reviezo 8(1):49-74. OECD (Organisation for Economic Co-operation and Development). 1975. The Polluter Pays Principle. Paris. *Ostro, Bart. 1994. "Estimating the Health Effects of Air Pollution: A Method with an Application to Jakarta." Policy Research Working Paper Series 1301. World Bank, Policy Research Department, Wash- ington, D.C. *Ostro, Bart, Jose Miguel SAnchez, Carlos Aranda, and Gunnar S. Eske- land. 1995. "Air Pollution and Mortality: Results from a Study of Santiago, Chile." Policy Research Working Paper Series 1453. World Bank, Policy Research Department, Washington, D.C. BIBLIOGRAPHY 61 Pearce, David W., and Anil Markandya. 1986. "Environmental Policy Benefits: Monetary Valuation." Prepared for OECD Conference on Environment and Economics, June 1984. Organisation for Econom- ic Co-operation and Development, Paris. Pigou, A. C. 1920. The Economics of Welfare. London: Macmillan. Pindyck, Robert S. 1979. The Structure of World Energy Demand. Cam- bridge, Mass.: MIT Press. Pitt, Mark. 1985. "Estimating Industrial Energy Demand with Firm Level Data: The Case of Indonesia." Energy Jouirnal 6(2):25-39. Roberts, Marc J., and Michael Spence. 1976. "Effluent Charges and Licenses under Uncertainty." Journal of Public Economics 5:193-208. Sandmo, Agnar. 1975. "Optimal Taxation in the IPresence of External- ities." Scandinavian Journal of Economics 77: 86-98. . 1976. "Direct versus Indirect Pigovian Taxation." European Economic Review 7: 337-49. * Swait, Joffre, and Gunnar S. Eskeland. 1995. "Travel Model Substitu- tion in Sao Paulo: Estimates and Implications for Air Pollution Control." Policy Research Working Paper 1437. World Bank, Policy Research Department, Washington, D.C. 'Ten Kate, Adriaan. 1993. "Industrial Development and the Environ- ment in Mexico." Policy Research Working Paper Series 1125. World Bank, Policy Research Department, Washington, D.C. U.S. EPA (Environmental Protection Agency). 1986. Compilation of Air Pollutant Emission Factors. Volume 1: Stationary Point and Area Sources. Supplement A. AP-42. Research Triangle Park, N.C. Viscusi, W. Kip, Wesley A. Magat, Alan Carlin, and Mark Dreyfus. 1994. "Environmentally Responsible Energy Pricing." Energy Jour- nal 15(2):23-42. Weitzman, Martin L. 1974. "Prices vs. Quantities." Review of Economic Studies 41: 477-91. Wijkander, Hans. 1985. "Correcting Externalities through Taxes on/ Subsidies to Related Goods." Journal of Public Economics 28:111-25. World Bank. 1992. World Development Report 1992: Development and the Environment. New York: Oxford University Press. . 1994. "Chile: Managing Environmental Problems: Economic Analysis of Selected Issues." Report 13061-CH. World Bank, Latin America and the Caribbean Country Department 1, Washing- ton, D.C. Directions in Development Begun in 1994, this series contains short essays, written for a general audience, often to summarize published or forthcoming books or to highlight current development issues. Africa's Management in the 1990s and Beyond: Reconciling Indigenous and Transplanted Institutions Building Human Capitalfor Better Lives Financing Health Care in Sub-Saharan Africa through User Fees and Insur- ance (also available in French) Imrzprovinig Early Childlhood Developmnernt: An Integrated Program for the Philippines (with a separate supplement) Investing in People: The World Bank in Action (also available in French and Spanish) Meeting the Infrastructure Challenge in Latin America and the Caribbean (also available in Spanish) MIGA: The First Five Years and Future Challenges Nurturing Development: Aid and Cooperation in Today's Changing World Nutrition in Zimbabzve: An Update Private and Public Initiatives: Working Togetherfor Health and Education Private Sector Participation in Water Supply and Sanitation in Latin America Reversing the Spiral: The Population, Agriculture, and Environmtient Nexus in Sub-Saharan Africa (with a separate supplement) A Strategyfor Managing Water in the Middle East and North Africa (also available in Arabic and French) Taxing Bads by Taxing Goods: Pollution Control with Presumptive Charges Toward Sustainable Management of Water Resources Unshackling the Private Sector: A Latin American Story The Uruguay Round: Widening and Deepening the World Trading System Water Supply, Sanitation, and Environmental Sustainiability: The Financing Challenge S - '- V i i- SCV9i i iilIXVl iiii i FT- <-1^,]r C - 6_ 1llEll~~~~~t l7. T:li11 ll{i .:is s i