The International Yearbook of Environmental And Resource Economics 2006 2007: A Survey of Current Issues (New Horizons in Environmental Economics)

The International Yearbook of Environmental and Resource Economics 2006/2007 NEW HORIZONS IN ENVIRONMENTAL ECONOMICS S...

Author: Tom Tietenberg

12 downloads 1237 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

The International Yearbook of Environmental and Resource Economics 2006/2007

NEW HORIZONS IN ENVIRONMENTAL ECONOMICS Series Editors: Wallace E. Oates, Professor of Economics, University of Maryland, USA and Henk Folmer, Professor of General Economics, Wageningen University and Professor of Research Methodology, Groningen University, The Netherlands This important series is designed to make a signiﬁcant contribution to the development of the principles and practices of environmental economics. It includes both theoretical and empirical work. International in scope, it addresses issues of current and future concern in both East and West and in developed and developing countries. The main purpose of the series is to create a forum for the publication of high-quality work and to show how economic analysis can make a contribution to understanding and resolving the environmental problems confronting the world in the twenty-ﬁrst century. Recent titles in the series include: Econometrics Informing Natural Resources Management Selected Empirical Analyses Phoebe Koundouri The Theory of Environmental Agreements and Taxes CO2 Policy Performance in Comparative Perspective Martin Enevoldsen Modelling the Costs of Environmental Policy A Dynamic Applied General Equilibrium Assessment Rob B. Dellink Environment, Information and Consumer Behaviour Edited by Signe Krarup and Cliﬀord S. Russell The International Yearbook of Environmental and Resource Economics 2005/2006 A Survey of Current Issues Edited by Henk Folmer and Tom Tietenberg The Greening of Markets Product Competition, Pollution and Policy Making in a Duopoly Michael Kuhn Managing Wetlands for Private and Social Good Theory, Policy and Cases from Australia Stuart M. Whitten and Jeﬀ Bennett Amenities and Rural Development Theory, Methods and Public Policy Edited by Gary Paul Green, Steven C. Deller and David W. Marcouiller The Evolution of Markets for Water Theory and Practice in Australia Edited by Jeff Bennett Integrated Assessment and Management of Public Resources Edited by Joseph C. Cooper, Federico Perali and Marcella Veronesi Climate Change and the Economics of the World’s Fisheries Examples of Small Pelagic Stocks Edited by Rognvaldur Hannesson, Manuel Barange and Samuel F. Herrick Jr The Theory and Practice of Environmental and Resource Economics Edited by Thomas Aronsson, Roger Axelsson and Runar Brännlund The International Yearbook of Environmental and Resource Economics 2006/2007 A Survey of Current Issues Edited by Tom Tietenberg and Henk Folmer

The International Yearbook of Environmental and Resource Economics 2006/2007 A Survey of Current Issues Edited by

Tom Tietenberg Mitchell Family Professor of Economics, Colby College, USA

Henk Folmer Professor of Research Methodology, Groningen University and Professor of General Economics, Wageningen University, The Netherlands NEW HORIZONS IN ENVIRONMENTAL ECONOMICS

Edward Elgar Cheltenham, UK • Northampton, MA, USA

© Tom Tietenberg, Henk Folmer 2006 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited Glensanda House Montpellier Parade Cheltenham Glos GL50 1UA UK Edward Elgar Publishing, Inc. 136 West Street Suite 202 Northampton Massachusetts 01060 USA

A catalogue record for this book is available from the British Library

ISSN: 1 460 7352 ISBN–13: 978 1 84542 723 8 (cased) ISBN–10: 1 84542 723 8 (cased) Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall

Contents List of ﬁgures List of tables List of contributors Preface Tom Tietenberg and Henk Folmer Editorial board 1

The incidence of pollution control policies Ian W.H. Parry, Hilary Sigman, Margaret Walls and Roberton C. Williams III

2

Geographical information systems (GIS) and spatial analysis in resource and environmental economics Ian Bateman, Wanhong Yang and Peter Boxall

vi vii viii ix x 1

43

3

Disclosure strategies for pollution control Susmita Dasgupta, Hua Wang and David Wheeler

93

4

Environmental policy under imperfect competition Till Requate

120

5

Transport and the environment Piet Rietveld

208

6

The Faustmann face of optimal forest harvesting Richard J. Brazee

255

289

Index

v

List of figures 1.1 Burden of a product tax 3 1.2 Abatement under an emissions tax 4 2.1 Distribution of household income levels in the counties of the Canadian province of Ontario 45 2.2 Representing real-world features as vector or raster data 48 2.3 GDP generated in ﬂood-prone areas of the Netherlands 50 2.4 GIS viewshed calculation of viewable area from one point on a speciﬁed property 52 2.5 Modelled emissions of CO and NO2 in Birmingham, UK; cumulative frequency of persons in ﬁve ethnic groups with respect to exposure to CO pollution 55 2.6 Conventional approach to aggregating WTP data; GIS-based approach to aggregation allowing for spatial variation in population and income distribution and distance decay and income inﬂuences upon marginal WTP 59 2.7 Isochrone map for a single recreational site; isochrone map for all wildlife parks in the UK 63 2.8 Examples of types of neighbourhood analysis for raster data 66 2.9 Oﬃcial counts of recreational visits to UK woodlands and visits predicted by GIS-based models 69 2.10 GIS-generated map of the marginal value of predicted woodland recreation demand for potential forest sites in Wales 72 2.11 A local indicators of spatial association map for expert ratings of the intensity of random camping in the East Slopes Region of Alberta, Canada 78 2.12 Two illustrations of ﬁrst-order contiguity binary weights matrices constructed from a 3 3 regular lattice 82 2.13 Spatial distribution of ﬁshing package prices at remote tourism sites in Ontario 84 2.14 Spatial autocorrelograms of the natural logarithm of weekly prices at ﬂy-in accessible tourism sites in Ontario 85 5.1 The chain from transport activity to valuation of disturbance 210 5.2 Competition between transport modes in urban areas 242

vi

List of tables 2.1 Selected websites containing information about geographic data sources 2.2 The present non-user’s beneﬁts of preserving the present condition of Broadland aggregated across the UK using various procedures 2.3 Regression models relating oﬃcial counts of visitors to woodland sites to predictions of the number of visitors obtained from GIS-based analyses 3.1 Indonesia’s PROPER results, 1995–97 3.2 The Philippines’ EcoWatch results, 1997–98 3.3 China’s GreenWatch results, 1999–2000 3.4 Performance rating programs: changes in compliance status 5.1 Life-cycle emissions for gasoline-fuelled cars with respect to fuel production, vehicle production and in-service use 5.2 Transport and economic trends in Europe, 1970–2000 5.3 International comparison of transport ﬁgures, including CO2 emissions 5.4 Environmental eﬀects of transport 5.5 Examples of noise management policies in transport 5.6 Vehicle pollution emissions 5.7 Contribution of mobile versus non-mobile sources to pollution, USA 5.8a Average emissions of CO2 and NOx per traveller-km per transport mode; urban transport, 2000, the Netherlands 5.8b Average emissions of CO2 and NOx per traveller-km per transport mode; medium-distance transport, 2000, the Netherlands 5.9 Average versus marginal environmental costs of public transport during the peak and oﬀ-peak period 5.10 Life-cycle greenhouse gas emissions for alternative fuels 5.11 Comparison of valuation methods used in the ﬁeld of transport and the environment 5.12 Examples of policy instruments for containing the environmental intrusion of transport 5.13 The shares of transport modes in passenger transport in various countries vii

46

60

70 101 101 102 103 209 211 213 215 217 218 220 222

222 225 227 230 235 243

Contributors Ian Bateman, Centre for Social and Economic Research on the Global Environment (CSERGE), University of East Anglia, UK and Adjunct Professor of Agricultural and Resource Economics, University of Western Australia, Perth Peter Boxall, University of Alberta, Canada Richard J. Brazee, University of Illinois at Urbana-Champaign, USA Susmita Dasgupta, Senior Economist, Development Research Group, World Bank, USA Ian W.H. Parry, Resources for the Future, USA Till Requate, University of Kiel, Germany Piet Rietveld, Vrije University Amsterdam, The Netherlands Hilary Sigman, Rutgers University and NBER, USA Margaret Walls, Resources for the Future, USA Hua Wang, Senior Economist, Development Research Group, World Bank, USA David Wheeler, Lead Economist, Development Research Group, World Bank, USA Roberton C. Williams III, University of Texas at Austin and NBER, USA Wanhong Yang, University of Guelph, Canada

viii

Preface As a discipline, Environmental and Resource Economics has undergone a rapid evolution over the past three decades. Originally the literature focused on valuing environmental resources and on the design of policy instruments to correct externalities and to provide for the optimal exploitation of resources. The relatively narrow focus of the ﬁeld and the limited number of contributors made the task of keeping up with the literature relatively simple. More recently, Environmental and Resource Economics has broadened its focus by making connections with many other subdisciplines in economics as well as the natural and physical sciences. It has also attracted a much larger group of contributors. Thus the literature is exploding in terms of the number of topics addressed, the number of methodological approaches being applied and the sheer number of articles being written. Coupled with the high degree of specialization that characterizes modern academic life, this proliferation of topics and methodologies makes it impossible for anyone, even those who specialize in Environmental and Resource Economics, to keep up with the developments in the ﬁeld. The International Yearbook of Environmental and Resource Economics: A Survey of Current Issues was designed to ﬁll this niche. The Yearbook publishes state-of-the-art papers by top specialists in their ﬁelds who have made substantial contributions to the area which they are surveying. Authors are invited by the editors, in consultation with members of the editorial board. Each chapter is critically reviewed by the editors and by experts in the ﬁeld. The editors would like to thank Wallace Oates for getting this project started. We also owe a special debt of gratitude to Sarah West, Michael Finus, Mark Cohen, Anthony Heyes, Juan Pablo-Montero, Thorsten Bayndir-Upmann, Bernard Sinclair-Desgagné, Dani Shefer and KarlGustaf Löfgren for helping us to shape this collection of essays. Tom Tietenberg Henk Folmer

ix

Editorial board EDITORS Henk Folmer, Wageningen University and Groningen University, The Netherlands Tom Tietenberg, Colby College, USA

EDITORIAL BOARD Kenneth Arrow, Stanford University, USA Scott Barrett, Johns Hopkins University, USA Nancy Bockstael, University of Maryland, USA Peter Bohm, Stockholm University, Sweden Lans Bovenberg, Tilburg University, The Netherlands Trond Björndal, University of Portsmouth, UK Carlo Carraro, University of Venice, Italy Partha Dasgupta, University of Cambridge, UK Ariel Dinar, The World Bank, USA Shelby Gerking, University of Central Florida, USA Lawrence Goulder, Stanford University, USA Eiji Hosoda, Keio University, Japan Per-Olov Johansson, Stockholm School of Economics, Sweden Bengt Kriström, Swedish University of Agricultural Sciences, Sweden Karl-Gustav Löfgren, University of Umeå, Sweden Juan-Pablo Montero, Catholic University of Chile, Chile Adolf Mkenda, University of Dar es Salaam, Tanzania Wallace Oates, University of Maryland, USA Charles Perrings, York University, UK Alan Randall, The Ohio State University, USA Michael Rauscher, Rostock University, Germany Kathleen Segerson, University of Connecticut, USA Bernard Sinclair-Desgagné, HEC Montreal, Canada V. Kerry Smith, North Carolina State University, USA Robert Solow, MIT, USA Alistair Ulph, University of Southampton, UK Michael Young, CSIRO land and water, Australia x

1. The incidence of pollution control policies* Ian W.H. Parry, Hilary Sigman, Margaret Walls and Roberton C. Williams III 1.

INTRODUCTION

Economic evaluations of pollution control policies have traditionally focused on pure eﬃciency eﬀects – either a comparison of their economic costs and environmental beneﬁts, or a comparison of their costs relative to those of alternative control policies (e.g., Cropper and Oates 1992, Morgenstern 1997, Hahn 2005). However, the distribution of policy costs and beneﬁts across households and ﬁrms is receiving increasing attention among researchers and policymakers.1 One reason is concern about whether a policy is ‘fair’ or not. Another is political feasibility – a policy justiﬁable on eﬃciency grounds may be impractical if it imposes a disproportionate burden on a politically inﬂuential group. Often the two are critically related; for example, political opposition to higher fuel taxes, carbon taxes, or other emissions taxes in the USA is frequently based on the claim that such taxes fall most heavily on low-income groups. The purpose of this chapter is to summarize what is actually known, and not known, about the incidence of beneﬁts and costs from pollution taxes, and alternative emissions control measures, across household income groups. Distributional issues have many diverse dimensions and we omit a number of topics, some of which have been comprehensively reviewed elsewhere. We do not discuss racial incidence. Evidence on this is discussed in Hamilton (2005); the main ﬁnding is that, in addition to income, other factors such as lack of participation in local decision-making explain the disproportionate burden of environmental risks borne by minorities. We only brieﬂy touch on incidence across consumers versus producers and across capital versus labor; Fullerton and Metcalf (2002) have recently reviewed the extensive public ﬁnance literature on such issues. And we do not discuss regional incidence within a country, nor inter-generational and inter-country incidence.2 Finally, most of our discussion applies to local

1

2

Yearbook of environmental and resource economics

and global air pollution policies in the USA and Europe, as that is where the bulk of empirical eﬀorts have been focused.3 The chapter is organized as follows. Section 2 provides a conceptual framework for understanding and measuring the burden on diﬀerent income groups from the costs of alternative emissions control instruments. Section 3 summarizes various empirical studies on how the costs of emissions taxes, emissions permits, and command-and-control policies are distributed across households. Section 4 discusses the distributional pattern of beneﬁts from emissions control policies. Section 5 brieﬂy discusses three ways in which distributional considerations might be integrated into traditional cost–beneﬁt analyses of environmental policies. Section 6 summarizes the main ﬁndings from the review and lists important topics for future research.

2.

CONCEPTUAL ISSUES

This section provides a highly simpliﬁed theoretical framework to help interpret results from more sophisticated empirical models discussed in Section 3. Our main focus is on pollution taxes in a competitive, partial equilibrium setting; we also touch on measurement issues, other forms of regulation, and complications due to non-competitive pricing and general equilibrium eﬀects. 2.1

Pollution Taxes

2.1.1 Product taxes To start with, consider the taxation of a single polluting commodity X (e.g., gasoline, electricity) which is produced by competitive ﬁrms under constant returns and consumed by all i 1 . . . N individuals in the economy. Individuals diﬀer according to income level Ii. Prior to the introduction of the tax, the product price px p0x, where p0x is the producer price, consumption for household i is X0i , and consumer surplus is triangle abc in Figure 1.1. With a speciﬁc tax tx the consumer price is p1x p0x tx; the burden of the tax for household i, prior to recycling of revenues, is the consumer surplus loss, trapezoid decb in Figure 1.1. This consists of the ﬁrst-order tax payment rectangle degb equal to txX1i , and a second-order loss from the reduction in consumption, triangle ecg equal to (Xi/px)(tx)2/2 (assuming linear demand over the relevant range). Alternatively, the burden is the tax payment with no change in consumption rectangle dfcb equal to txX0i , less triangle efc equal to (Xi/px)(tx)2/2, the saving in spending, net of forgone consumption beneﬁts, from the reduction in consumption.

The incidence of pollution control policies

Price

3

a

Demand

p1

d

e

f

X

p0 X

b

g

Xi1

c

Xi0

Consumption Figure 1.1

Burden of a product tax

We deﬁne the (initial) own-price elasticity of demand for household i by xi =(Xi /px) p0x / X0i . Using this and the above expressions, the burden to household i (Bi) can be written: Bi tx X1i (tx p0x ) 2 p0x X0i xi 2 txX0i (tx p0x ) 2p0x X0i xi 2

(1.1)

For small tax changes the second-order eﬀect from the behavioral response to the tax is small relative to the ﬁrst-order eﬀect of the price increase (ecg is small relative to degb, or efc is small relative to dfcb, in Figure 1.1); thus it is reasonable to ignore the second-order eﬀect. In this case the burden, expressed relative to income, is approximately given by: Bi Ii (tx p0x )s0xi

(1.2)

where s0xi p0xX0i Ii is the (initial) budget share. In this case, whether policy costs are progressive/proportional/regressive (i.e. whether the burden relative to income rises, is constant or falls with income) depends on whether the budget share is lower/the same/greater for low-income households than for high-income households.

4

Yearbook of environmental and resource economics

2.1.2 Emissions taxes Eﬃcient environmental taxes are levied on emissions, rather than a consumption good, and thus ultimately may aﬀect ﬁnal product prices across a range of industries. Suppose a tax of tE is levied on pollution emissions and that emissions are released during the production of j 1 . . . M consumption goods. Firms producing good j will reduce emissions per unit of output until the incremental abatement cost (e.g., from end-of-pipe treatment or from substituting cleaner inputs in production) equals the emissions tax. In Figure 1.2, the tax-induced abatement is therefore e0j ej , where ej denotes emissions per unit of output. Assuming ﬁrm costs are fully passed forward to consumers (see below), the price of good j is: p1j tEej cj (e0j ej ) p0j

(1.3)

where tE ej is the emissions tax payment per unit of output, rectangle acdb in Figure 1.2, cj () is the resource cost from abatement per unit of output, area 0ab, and p0j is the initial producer price. Using analogous expressions to (1.1) for the burden of price increases, aggregating over all goods and dividing by income, the burden of the emissions tax can be expressed: Bi Ii

M

M

j1

j1

pˆj sji0 12pˆj2 0ij sji0

pj (p1j p0j ) p0j (tE ej cj (·))p0j ,

ˆ

sji0 p0j Xji0 Ii

(1.4a)

(1.4b)

Marginal cost Marginal abatement cost

a

tE

b 0

ej0⫺ ej Emissions abatement

Figure 1.2

Abatement under an emissions tax

c d

The incidence of pollution control policies

5

ˆ

where pj is the proportionate increase in price of product j and s0ji is household i’s (initial) budget share for good j. As before, the burden amounts to ﬁrst-order surplus losses from the increase in prices at initial consumption, less second-order gains from the reduction in consumption. And ignoring the second-order terms is reasonable when the proportionate change in 2 product prices is small ( pj is small relative to pj).

ˆ

ˆ

2.1.3 Revenue recycling Many empirical studies do not consider recycling of environmental tax revenues in other tax reductions, even though overall incidence impacts are very sensitive to the form of recycling. Consider, for example, the case of one polluting good. With p small, the burden with revenue recycling for household i would be

ˆ

Bi Ii psi0 i

ˆ

(1.5)

where i is the rebate from revenue recycling as a proportion of household i’s income. Clearly, the regressivity of the tax could be reduced if the rebate were larger as a fraction of income for low-income households than for high-income households (for example, if revenues ﬁnanced an increase in income tax thresholds). 2.1.4 Indexing of transfer payments Many transfer programs (e.g., social security) are indexed for price changes, suggesting that low-income households may receive some compensation, even prior to recycling of revenues. This is a signiﬁcant consideration, taken into account in some studies. However, not all lowincome households receive transfer payments; moreover, low-income families often have relatively high budget shares for energy and polluting goods, implying that they will be under-compensated from price indices that weight these goods according to budget shares of the average household. 2.2

Measurement Issues

We now comment brieﬂy on the measurement of variables that enter into the net burden formula (1.4). 2.2.1 Household expenditure on ﬁnal products This is available from household data sources, such as the Consumer Expenditure Survey (CES) in the USA.

6

Yearbook of environmental and resource economics

2.2.2 Household income Measurement of household income is tricky. In principle lifetime income is a better measure of an individual’s well-being than current income. For example, Harvard MBA students may have low current income, but are not poor in a lifetime context given their high, expected future earnings. In addition, a reasonably well-oﬀ person may appear poor in a particular year due to transitory factors, such as temporary unemployment, illness, extended maternity leave, etc. Lifetime measures of income are designed to remove the confounding eﬀects of similar people being at diﬀerent stages of their lifecycle, and short-term variation in earnings; incidence estimates based on lifetime income tend to be signiﬁcantly less regressive than those based on annual income. However, lifetime income is far more diﬃcult to measure than current income, as it requires tracking households over extremely long time periods. Earlier studies by Poterba (1989) of federal excise taxes and Metcalf (1994) of state and local sales taxes, proxied lifetime income by annual consumption, based on the Friedman–Modigliani permanent income and lifecycle models; with perfect lifecycle consumption smoothing, current consumption is proportional to lifetime income. However, evidence suggests that the lifetime consumption trajectory is not ﬂat but has an inverted-U shape, due in part to liquidity constraints (e.g., Bull et al. 1994, Zeldes 1989, Souleles 1999 and 2002), and that consumption is responsive to changes in the timing of income (Shapiro and Slemrod 1994); these ﬁndings are inconsistent with the lifetime income hypothesis. Some more recent studies therefore use econometric methods relating income to education, age, and other demographic variables to construct more sophisticated measures of lifetime income.4 Even this approach is not entirely satisfactory, as it does not consider all potentially important factors determining income and it implicitly assumes that the status of a person observed at a single point in time stays the same forever;5 some authors even argue against using the lifetime income concept in policy analysis at all (Barthold 1993). Given the controversy surrounding income measurement, studies often report results for a range of alternative income deﬁnitions. 2.2.2 Change in product prices The tax component of changes in ﬁnal goods prices can be obtained using input–output tables that trace all intermediate goods going into ﬁnal goods production (e.g., from the Bureau of Economic Analysis), and data on emissions factors for polluting inputs (e.g., from EPA 1996). Direct estimates of the abatement cost component of price increases might be unavailable, and are often ignored, which is reasonable so long as the proportionate emissions reduction is modest.6

The incidence of pollution control policies

7

2.2.3 Demand elasticities Estimates of product demand elasticities for diﬀerent household income groups have only recently begun to emerge from analyses of micro data (e.g., West 2004). Previous empirical studies that included behavioral responses usually assumed proportionate demand changes were the same across all households. One subtle issue is that the second-order component of the change in household welfare diﬀers according to whether it is measured by consumer surplus, equivalent variation, or compensating variation; in the ﬁrst case the welfare change is the relevant area under the uncompensated demand curve, while in the second and third cases it is measured with respect to compensated demand curves. For individual product price increases there is usually very little diﬀerence between the three welfare measures, so long as the share of spending on this good is a small fraction of income;7 but this may not be the case when a wide range of product prices is simultaneously increased, and the relevant budget share is more substantial. 2.3

Other Control Instruments

For simplicity, in the discussion below we assume an emissions policy aﬀects the price of just one commodity X. 2.3.1 Tradable emissions permits Tradable permits have essentially the same distributional eﬀects as an emissions tax that would induce the equivalent emissions reduction, if the permits were fully auctioned. That is, if the equilibrium permit price is , then analogous expressions to equations (1.3) and (1.4) apply again, with tE replaced by 8. Eﬀects are quite diﬀerent, however, if permits are given out for free (Dinan and Rogers 2002, Parry 2004).9 The reason is that, rather than going to the government, permit rents are reﬂected in higher ﬁrm equity values because ﬁrms receive an asset with market value for free. The permit cap acts rather like a binding production quota, or a cartel where members agree to limit their production; in all cases output is reduced below free market levels and product prices and ﬁrm proﬁts are increased. Ultimately permit rents accrue to households in the form of dividends or capital gains; in terms of equation (1.4a), there is now an income term, i/Ii, subtracted from the right-hand side, where i is proﬁt income for household i arising from permit rents. To the extent that wealthy households receive a greater share of their income from capital than poor households, that is i/Ii increases with Ii, the creation of scarcity rents is regressive.

8

Yearbook of environmental and resource economics

In fact, it is possible that high-income households are actually made better oﬀ under grandfathered permits (excluding environmental beneﬁts), while low-income households are substantially worse oﬀ! To see this, suppose, for simplicity, that half the population has high income (H), the other half has low income (L), and high-income households own all the capital. Suppose also that we can ignore second-order eﬀects. Then the burden for high- and low-income households is given by: BH (p1x p0x )XH0 H,

BL (p1x px0 )XL0

(1.6)

where p1x p0x e c(e0x ex ) is the product price increase and H is capital or proﬁt income per high-income household. The latter equals the permit rents e(X0L X0H ) , equal to the permit price times emissions e(X0L X0H ) . Low-income households receive no compensation and are unambiguously worse oﬀ (BL 0). However high-income households could be better oﬀ overall (BH 0); in our highly simpliﬁed example, this occurs when permit rents are large enough relative to abatement costs so that

e(XH0 XL0 ) ( e c)XH0 ) or ec XH0 XL0. Several subtle considerations weaken, though not necessarily overturn, the prospects for this perverse equity eﬀect. First, the share of permit rents, vis-à-vis the share of abatement costs, in the product price increase typically diminishes at higher levels of abatement, at least with linear (though perhaps not convex) marginal abatement costs; that is, e (rectangle acdb in Figure 1.2) decreases in size relative to c() (triangle 0ab). Thus the prospects for rent income to overcompensate high-income households for the price increase diminishes with the level of abatement. Second, a signiﬁcant portion of the permit rents (around 35 percent in the USA) will still go, albeit indirectly, to the government rather than owners of capital, via corporate taxes paid on additional proﬁts, and personal income and capital gains taxes on additional household income. Third, capital income is not exclusively earned by high-income households; even low-income households may have some capital income in the form of retirement accounts.10 Fourth, as discussed below, the price eﬀects of tradable permits, and hence the rent transfer from consumers to shareholders, is sensitive to assumptions about returns to scale, market structure, and possible diﬀerences between marginal and non-marginal production technologies. 2.3.2 Command-and-control regulation Suppose ﬁrms must satisfy a performance standard that imposes a limitation of e on emissions per unit of output. With homogeneous ﬁrms each ﬁrm’s abatement cost per unit is c(e0 e) and the product price is p1x c(e0 e) p0x. That is, there is no tax revenue/permit rent component

The incidence of pollution control policies

9

to the price increase.11 Thus the policy avoids the adverse distributional consequence of freely allocated permits that arises from the transfer of rent income to high-income households; whether the policy is progressive or regressive essentially depends on budget shares across income groups. In fact, for a given emissions reduction, low-income households could be worse oﬀ under grandfathered permits than under the performance standard; although the pure abatement costs are lower under permits, the diﬀerence could be more than oﬀset by the price eﬀect of permit rents (Goulder et al. 1999). The same qualitative result could still apply, though is less likely, if ﬁrms are heterogeneous (and abatement costs are not minimized across ﬁrms), or under an end-of-pipe technology mandate (where ﬁrms cannot exploit the least-cost combination of measures to reduce emissions per unit of output, including substitution of clean for dirty inputs). 2.4

Complicating Factors

So far we have assumed that all policy costs are fully passed forward to consumers, the standard approach taken in input–output analyses discussed below. However, there are various complications that may aﬀect the impact of environmental policies on product prices, and policies may also aﬀect prices in (economy-wide) factor markets. 2.4.1 Upward-sloping supply curves Even in the long run supply curves may still be upward sloping (rather than perfectly elastic as assumed above), due to rising marginal costs of using an input, such as a scarce natural resource, or industry-speciﬁc capital. In this case part of the burden of an emissions tax or other regulation will come at the expense of reducing producer prices and producer surplus, or pure ﬁrm proﬁts, rather than higher product prices. This ultimately passes some of the burden back to shareholders in the form of lower equity values or dividend payouts. The eﬀect is progressive since wealthier households earn a larger share of their income from capital. 2.4.2 Non-competitive pricing The assumption of competitive pricing may be unrealistic; this is particularly the case for electricity generation, which is a major contributor to global and local air pollution. Currently, well over half of generated electricity in the USA is subject to regulated prices, though this fraction will diminish in future with continued restructuring (Brennan et al. 2002). In states where generation prices are regulated, the opportunity cost of using grandfathered permits to cover emissions is not passed on in higher prices (Burtraw et al. 2001, p. 7). But even in deregulated markets, abatement

10

Yearbook of environmental and resource economics

costs may in part come at the expense of proﬁts (and ultimately shareholders) rather than being fully passed forward, if ﬁrms have market power (e.g., Borenstein et al. 2002). Dominant ﬁrms may exercise market power in regional electricity markets when fringe competitors become capacity constrained at peak periods, and congestion on the grid prevents the import of power from other regions. 2.4.3 Multiple production technologies In electricity generation diﬀerent technologies are often used to supply baseload and peak demand. At peak periods infra-marginal production is often from coal plants, which have a high emissions intensity, while marginal production is often from natural gas plants, which have lower emissions intensity (in the case of carbon and NOx) or zero emissions (in the case of SO2). Consequently, abatement costs for coal plants at peak periods will not be fully passed on in higher prices; they will, at least in part, come at the expense of rents earned on infra-marginal production (in Parry’s 2005 analysis of the SO2 trading program this dampens the eﬀect on product prices by about 45 percent). Again, some of the burden is passed back to owners of capital. 2.4.4 Changes in economy-wide factor prices A recent general equilibrium analysis by Fullerton and Heutel (2004) explores the extent to which pollution taxes aﬀect the economy-wide rate of return from capital relative to that from labor. They study a closed economy with labor and capital in ﬁxed aggregate supply but mobile across industries, and two competitive, constant returns industries producing a clean good and a dirty good. They show that if the elasticity of substitution between polluting inputs and labor is the same as that for polluting inputs and capital, then a pollution tax will usually lower the relative return of the factor that is used more intensively in the dirty sector. Polluting industries tend to be relatively capital intensive (Antweiler et al. 2001, p. 879), implying that emissions taxes may enact a transfer from (wealthy) households with a relatively high capital income share to (poor) households with a relatively high labor income share. In this regard, most empirical studies of pollution control incidence may overstate policy regressivity, as they do not account for changes in economy-wide factor prices.12

3.

EMPIRICAL STUDIES

This section begins by discussing the sizeable empirical literature on environmental taxes, and the much more limited one on emissions permits; for

The incidence of pollution control policies

11

these policies the primary incidence eﬀects are transparent, assuming tax payments or permit rents are fully passed on in higher prices. We then summarize older literature on command-and-control policies and overall federal environmental programs in the USA. Here, price eﬀects must be inferred from estimates of abatement costs, which may be inaccurate since analysts and researchers often have imperfect information on ﬁrms’ costs. Care is needed in comparing studies as they may measure incidence and household income diﬀerently, some studies account for behavioral responses to price rises induced by the policies while others do not, and some rely on a partial equilibrium framework while others take a general equilibrium approach. 3.1

Environmental Taxes

A substantial literature exists on a variety of energy taxes, including gasoline and carbon taxes, and this forms the bulk of the work reviewed here. A general ﬁnding is that, prior to revenue recycling and on the basis of annual income, most environmental taxes look regressive because lowerincome households tend to spend a disproportionately larger fraction of their income on energy, which is a necessity good. Using lifetime income, taking account of increases in prices of other goods for which energy is an input, and recycling revenues can mitigate this regressivity, at least in part. Our review captures some of the most important published work in the area but is not meant to be exhaustive; rather, we provide a ﬂavor for the literature, and note consensus on results if and where it exists. 3.1.1 Gasoline taxes A gasoline tax is, for the most part, a ﬁnal product tax and thus has eﬀects as we described in section 2.1.1 above. Poterba’s (1989) study of gasoline taxes (and other federal excise taxes) was among the ﬁrst to emphasize the quantitative signiﬁcance of diﬀerent measures of income for the degree of regressivity (see also Poterba 1991). He used CES data and computed the budget share on gasoline for each household. He found that the budget share of the bottom income quintile was 5.3 times that for the top income quintile when annual income is used, but 1.5 times that for the top income quintile when lifetime income, as proxied by annual consumption expenditure, is used. Using data from the CES, West and Williams (2004) examine the incidence of an increase in the (federal and state) gasoline tax from its current level of about $0.40 per gallon to $1 per gallon, again with consumption as a proxy for lifetime income. Unlike Poterba, they account for behavioral responses in their incidence calculations, by econometrically estimating

12

Yearbook of environmental and resource economics

gasoline demand elasticities by income quintile. They ﬁnd that the gasoline tax is generally regressive prior to revenue recycling. Regressivity is reduced if revenues are returned through an equal percentage reduction in the marginal tax rate on labor income for each income group. This occurs because labor income is a greater fraction of total income for low-income households than for high-income households. The regressivity of the gasoline tax increase is eliminated altogether if revenues are returned in equal lump-sum transfers for all households; in fact, in this scenario, the bottom two quintiles are actually better oﬀ overall. The authors also ﬁnd that ignoring demand responses (i.e. computing rectangle dfcb rather than trapezoid decb in Figure 1.1), or assuming the same gasoline demand elasticity across diﬀerent income groups, makes the tax increase appear more regressive; this is because lower income groups have more elastic demands for gasoline, implying a disproportionate reduction in their burden from demand responses. Finally, they ﬁnd little discrepancy in incidence eﬀects between welfare measures based on consumer surplus and equivalent variation. Although revenues from a future fuel tax increase might ﬁnance other tax cuts, revenues from the current fuel tax are earmarked for highways. Interestingly, Wiese et al. (1995), using a computational general equilibrium model, ﬁnd that existing gasoline taxes are actually progressive, as highway spending bids up the demand for manual labor and relative wages of the bottom income quartile. Progressivity declines if some of the tax receipts are instead diverted to the general government and deﬁcit reduction. Bento et al. (2005) use the 2001 National Highway Transportation Survey (NHTS), a large US household survey dataset, to estimate a random coeﬃcients model of vehicle choice and miles traveled, and combine that with a model of vehicle scrappage as well as a model of Bertrand competition (across manufacturers) in the new car market. They simulate a 10, 30 and 50 cent per gallon increase in the gasoline tax under scenarios when revenues are rebated to households in proportion to their gasoline tax payments and in proportion to income. The authors ﬁnd that with tax-based recycling, the impact of the tax across income groups is close to proportional. With income-based recycling, on the other hand, low-income households pay more as a percentage of income than high-income households.13 3.1.2 Other energy taxes and carbon taxes The studies just discussed assume that gasoline is a ﬁnal good, directly consumed by households; this is reasonable, because household consumption accounts for the bulk of gasoline use. However, it is not reasonable for many other goods that might be taxed on environmental grounds. For example, direct household consumption accounts for only about two-ﬁfths of electricity sales; the remainder is split about equally between industrial

The incidence of pollution control policies

13

and commercial users, and is eﬀectively an intermediate good in the production of consumer products. In this case it is potentially important to account for increases in prices of other ﬁnal goods that are indirectly aﬀected by the tax. In their study of taxes on electricity, coal, natural gas, gasoline and other reﬁned petroleum products, Casler and Raﬁqui (1993) compute price eﬀects on 89 ﬁnal goods consumed by households, using input–output tables. They assume that taxes are fully passed forward to consumers, and that ﬁrm’s input–output ratios and household product demands are ﬁxed. Price calculations are combined with CES data on the 89 commodities by income quintile, and income is measured on an annual basis. They ﬁnd that the greater the share of output from the taxed good that is an intermediate, as opposed to ﬁnal, good, the less regressive the tax. Overall, the tax burden to income ratio for the lowest quintile is only modestly larger than that for the top quintile across the various taxes. Bull et al. (1994) use the same data sources and a similar approach to analyse a tax based on energy content, i.e., a Btu (British thermal unit) tax, and a tax based on carbon content. However, they consider a broader range of household income measures than Casler and Raﬁqui, including annual income, annual consumption and lifetime income.14 Like Casler and Raﬁqui, they ﬁnd that on the basis of annual income the direct components of Btu and carbon taxes look quite regressive, while the indirect components are less regressive. On the basis of lifetime income, the direct component remains regressive but the indirect component becomes mildly progressive; overall, the taxes look much less regressive on a lifetime income basis than on an annual income basis. In a third study employing similar data and methods, Metcalf (1999) analyses a revenue-neutral package of environmental taxes, including a carbon tax, an increase in motor fuel taxes, taxes on various stationary source emissions, and a virgin materials tax; revenues from this package amount to 10 percent of federal income tax revenue. Prices of energy goods – electricity, natural gas, fuel oil and gasoline – increase substantially under these measures (by 14 to 27 percent), while prices of all other consumer goods increase by less than 5 percent. Although the taxes disproportionately hit low-income groups, Metcalf shows that the overall package can be made distributionally neutral (under a range of diﬀerent income measures) through careful targeting of income and payroll tax reductions. Cornwell and Creedy (1997) use data from the1984 Australian Household Expenditure Survey to estimate parameters of a linear expenditure system for diﬀerent income groups and then use these parameters to calculate compensating and equivalent variations resulting from a carbon tax. The authors assume that the carbon tax is fully passed forward to consumers and

14

Yearbook of environmental and resource economics

that the prices of goods increase in proportion to their carbon content. They ﬁnd that the tax is regressive, on the basis of annual income: both compensating and equivalent variation as a fraction of income fall as income rises. They also show how a ‘minimum income guarantee’ could be increased and oﬀset the regressive eﬀects of the tax. Brännlund and Nördstrom (2004) use data from Sweden to analyse a carbon tax with revenues recycled in a reduction in the general value-added tax (VAT), or in a reduction in the VAT on public transport.15 The authors use the Swedish Household Expenditure Survey, combined with aggregate data from the National Accounts, to estimate the demand for non-durable goods. They obtain price elasticities that vary by income quintile, then simulate the eﬀects of the tax policies. They ﬁnd that the tax is regressive under the ﬁrst recycling scenario, but less regressive in the second. Larger diﬀerences show up regionally in the second scenario – city dwellers beneﬁt from the policy while rural households are hurt. Unlike other studies that consider proposed carbon taxes, Wier et al. (2005) examine the existing CO2 tax in Denmark, based on actual taxes paid directly and indirectly by households. They use input–output tables from Denmark for the year 1996, assuming taxes are fully passed through to consumers in higher product prices, and a consumer expenditure survey of over 3400 households. On the basis of annual income, Wier et al. ﬁnd that (excluding use of revenues) the CO2 tax is regressive: households in the lowest income decile paid approximately 0.8 percent of their disposable income in taxes, while households in the highest decile paid less than 0.3 percent. Again, the direct component of the tax accounts for most of the regressivity. Using total expenditures as a proxy for lifetime income, Wier et al. ﬁnd that the regressivity is greatly reduced, though not entirely eliminated. 3.1.3 Motor vehicle taxes Walls and Hanson (1999) combine motor vehicle emissions per mile data from a remote sensing experiment in California with vehicle mileage data by household obtained from the NHTS to study the replacement of existing vehicle registration fees by taxes on total yearly emissions, emissions per mile, and vehicle miles traveled.16 The authors construct a measure of lifetime income for each household in their dataset using a relationship between education and demographic data and lifetime income estimated earlier by Rogers (1993). They ﬁnd that, regardless of whether annual or lifetime income is used, the tax on emissions per mile is more regressive than the emissions tax, which in turn is more regressive than the tax on miles driven; this is because ownership of older, dirtier vehicles with higher emission rates is

The incidence of pollution control policies

15

disproportionately concentrated among lower-income households. They also ﬁnd that the mileage tax looks more regressive than existing registration fees, which are based on vehicle value, on an annual income basis, though not on a lifetime income basis; this is because the mileage/income relation is weaker than the vehicle value/income relation on an annual basis, but this diﬀerence disappears under the lifetime income measure. Finally, adverse distributional eﬀects of any of the three new taxes are diminished if revenues raised are used to reduce existing registration fees, which are themselves regressive. West (2004) integrates behavioral responses into an incidence analysis of motor vehicle taxes and subsidies based on an econometric model of household vehicle choice and mileage estimated with CES data. The policies she considers are a tax on vehicle size, a mileage tax, and a subsidy to vehicle ‘newness.’ She ﬁnds that households in the lower income deciles have more elastic demands for miles traveled than those in the higher income deciles. Consequently, looking at estimated tax payments as a share of lifetime income (proxied by annual consumption) without considering behavioral responses overstates the true regressivity of a mileage tax (see above). West ﬁnds, however, that the mileage tax is still regressive, even after accounting for behavioral responses. Interestingly, the tax payments as a share of income, or consumer surplus change as a share of income, become larger from the lowest decile to the middle deciles, but then fall after that, and drop sharply for the top decile. Some of this impact is due to lowincome households who do not own a vehicle: the regressivity of the tax is greater when only households who own vehicles are considered. Finally, West also ﬁnds that both the subsidy for new vehicles and a tax on vehicle size are signiﬁcantly regressive, more so than the mileage tax.17 3.2

Emissions Permits

Dinan and Rogers (2002) provided the ﬁrst major contribution on the incidence of emissions permits; they considered a program to reduce US carbon emissions by 15 percent below 1998 levels (at a permit price of $100 per ton of carbon). They extend the modeling framework of Casler and Raﬁqui (1993) to incorporate behavioral responses (assumed uniform across households), indexing of transfer payments (e.g., social security), and they allocate to households additional burdens from the eﬀect of higher product prices on reducing real factor returns and compounding eﬃciency costs of pre-existing factor tax distortions (e.g., Goulder et al. 1999). Dinan and Rogers’s results show that distributional eﬀects hinge crucially on whether permits are grandfathered or auctioned, and whether revenues from permit auctions, or from indirect taxation of permit rents, are

16

Yearbook of environmental and resource economics

used to cut payroll taxes, corporate taxes, or provide lump-sum transfers. For example, they estimate that households in the lowest income quintile would be worse oﬀ by around $500 per year under grandfathered permits while, due to large indirect increases in the value of their stockholdings, households in the top income quintile would be better oﬀ by around $1000. If instead the permits were auctioned with revenues returned in equal lump-sum rebates for all households, they estimate that low-income households would on net be better oﬀ by around $300 while high-income households would be worse oﬀ by around $1700.18 Parry (2004) estimates a simple, calibrated, analytical model with household income proxied by consumption to examine the incidence of emissions permits, among other control instruments, to control power plant emissions of SO2, carbon and NOx. He ﬁnds that using grandfathered emissions permits to reduce carbon emissions by 10 percent, and NOx emissions by 30 percent, can be highly regressive; the top income quintile is made better oﬀ while the bottom income quintile is made much worse oﬀ. The SO2 cap imposed by the 1990 Clean Air Act Amendments, which has reduced emissions by roughly 45 percent, is also regressive but much less so than the carbon and NOx policies. This result underscores the point that permit rent relative to abatement costs, and hence the relative transfer to wealthy households, is smaller at higher levels of emission reductions. Rose and Oladosu (2002) use a computable general equilibrium model with 41 production sectors, four factors of production, and ten income brackets to analyse a carbon permit trading system that reduces US emissions by 7 percent below 1990 levels (the original US target under the Kyoto Treaty) for a permit price of $128 per ton. With auctioned permits, and prior to revenue recycling, the lowest income bracket experiences a burden to income ratio 73 percent larger than that for the top income bracket, reﬂecting the former’s larger budget shares for energy goods. When revenues are returned in income tax cuts, the policy is approximately proportional overall (burden to income ratios are similar across diﬀerent households). 3.3

Command-and-Control Regulations

We now turn to some older literature on the incidence of US federal pollution regulations, which have historically been command-and-control programs. These older studies do not consider behavioral responses, and measure income on an annual rather than lifetime basis, so they likely overstate the degree of regressivity. Unlike literature discussed above, some of these older studies considered both policy beneﬁts and costs; we discuss beneﬁt estimates in the next section.

The incidence of pollution control policies

17

Dorfman and Snow (1975) looked at the costs of pollution controls incurred as a result of all federal environmental programs, by government, industry and households. Government control costs are assumed to cause proportional increases in income taxes paid by households; industry costs are assumed to be fully passed forward in ﬁnal product prices and are allocated to households based on their total consumption expenditures; and household pollution control costs are higher prices for vehicles resulting from emissions regulations, allocated on the basis of vehicle ownership. The study found the aggregate costs of federal regulations to be regressive with burden to income ratios of roughly 2 percent and 1 percent for the bottom and top income quintiles respectively. A major factor was that emissions standards drive up the price of lower-cost vehicles, purchased by poor households, by about the same amount in absolute terms as expensive ones, purchased by wealthy households; added costs relative to income are therefore greater for low-income households.19 However, in a similar study conﬁned to the 1970 Clean Air Act, Gianessi et al. (1979) ﬁnd a more complex incidence pattern when smaller household groups are disaggregated. They estimate that the bottom income decile incurs a lower burden to income ratio than other groups, as they had the lowest vehicle ownership rate.20 A drawback of these early studies is that they match total abatement costs to total consumption spending by households. Robison (1985) improves on this by matching estimates of price increases for 78 consumption goods with spending on these goods by household class; price increases are obtained using input–output tables and abatement costs (for all media) from the Pollution Abatement Costs and Expenditures (PACE) data. Results show that the regulations imposed on industry are regressive: the poorest 5 percent of the population paid about 1 percent of their income for such costs, the next 5 percent of the population paid about 0.6 percent of their income, with the burden continually declining to the richest 5 percent, which paid only about 0.2 percent of their income. Lake et al. (1979) examine costs of the 1972 Clean Water Act. They conduct an extensive analysis of the implied burden on households from diﬀerent mechanisms for ﬁnancing municipality water treatment expenses (taxes, sewer fees, bond issues, etc.). Overall costs of municipal wastewater treatment remain regressive, despite the federal grant program: the lowest income decile’s burden to income ratio is three times that for the highest income decile. Lake et al. also analyse control costs for ten ﬁnal industries with high levels of wastewater discharge, and assign these costs to households based on their consumption patterns. Control costs are highly regressive: burden to income ratios are 2.4 percent and 0.13 percent respectively for the bottom and top income decile.

18

Yearbook of environmental and resource economics

Collins (1977) also analyses the federal grant program embodied in the Clean Water Act, focusing on a region in the Midwest rather than the whole nation. He estimates the amount and distribution of both the subsidies, and the implied tax increases, by income class. He ﬁnds that the highest income category beneﬁts the most from the grant program and that middle-income groups incur the greatest cost; the lowest-income groups get a slight positive beneﬁt. High-income groups beneﬁt primarily because of the implied subsidies to polluting ﬁrms. Ostro (1981) replicates Collins’s work for the Boston metropolitan area, but obtains opposite results. In particular, he ﬁnds that all of the lower- and middle-income groups beneﬁt from the grant program and the top four income groups lose. These ﬁndings are a result of assuming that the municipal subsidy accrues to groups in proportion to their water usage and population and a larger number of people are in the lower-to-middle-income groups than the higher-income groups. The author speculates that similar results might be found for other urban areas.21 For a number of reasons it would be useful to update and reﬁne the type of analysis in these earlier, comprehensive studies of federal environmental regulations. In the last 25 years or more the income distribution has become more unequal, household expenditure patterns have changed, ﬁrms have often found new, radically more eﬃcient ways to reduce pollution, and new laws and regulations have imposed new costs. In addition, limitations on the earlier work can be overcome; for example, using recent data, we can more accurately estimate lifetime income for various demographic groups, based on their current income, educational attainment and age. 3.4

Comparison across Instruments

Studies have generally focused on one policy instrument in isolation, rather than considering a broad range of alternative instruments. One exception is Parry (2004), who ﬁnds that the burden imposed on low-income households from control of power plant emissions can be lower under performance standards and technology mandates than under grandfathered permits (assuming homogeneous ﬁrms). For a given emissions reduction, the command-and-control policies cause higher abatement costs than emission permits as they fail to optimally exploit all the diﬀerent margins for emission reductions; however, the overall product price increase is larger under emission permits, as it also reﬂects the large rents created under the policy. More research is needed on how robust this result might be to incorporating heterogeneity in abatement costs among ﬁrms, diﬀerent levels of pollution reduction, non-competitive pricing in the electricity sector and other factors.

The incidence of pollution control policies

3.5

19

Summary

Most empirical studies ﬁnd that the costs of environmental policies are borne disproportionately by lower income groups. This appears to hold across a range of policy instruments, especially grandfathered emissions permits, though the bulk of empirical work has focused on taxes, particularly energy taxes. The ﬁnding is less pronounced for taxes on intermediate products than for taxes on ﬁnal goods, and when some measure of lifetime income is used rather than annual income, though measuring lifetime income remains problematic, especially in cross-section studies. Perhaps the most important ﬁnding, and one that deserves more attention in future research, is the potential for revenue recycling (from taxes and auctioned permits) to mitigate the burden on low-income households. More attention should also be paid to the extent to which the burden of environmental policies is passed backwards in reduced returns to owners of capital, rather than forward to consumers in higher product prices.

4. WHO BENEFITS FROM ENVIRONMENTAL POLICIES? An abundant ‘environmental justice’ literature examines the distribution of existing environmental risks, but we should be cautious about using ﬁndings from this literature to make inferences about the distribution of welfare gains from policy, for at least four reasons. First, due to a lack of data, the bulk of the environmental justice literature uses measures of environmental risk that do not adequately account for the degree of exposure and factors aﬀecting individuals’ susceptibility to pollution-induced illness. Second, when policies create non-uniform environmental improvements, the existing risk distribution will inaccurately predict distributional beneﬁts from the policy change. Third, to translate physical beneﬁts into welfare gains we need to measure how diﬀerent households value environmental quality. Finally, we also need to account for the possible eﬀect of changing environmental quality on market prices or wages, as these price changes also aﬀect household welfare. All of these issues are taken up in this section. 4.1

Evidence on the Distribution of Prevailing Environmental Risks

Studies have evaluated the existing distribution of environmental risks for many diﬀerent pollutants at many diﬀerent geographical scales. They merge data on environmental conditions with census data on the characteristics of local populations.22

20

Yearbook of environmental and resource economics

4.1.1 Findings Empirical literature on the distribution of existing environmental risks is so large that it is diﬃcult to do justice to the range of methods and results. The discussion here is necessarily cursory because of the broader focus of this chapter and the availability of several comprehensive surveys (Bowen 2002, Pearce 2003, Hamilton 2005, and Ringquist 2005). Early studies, which focused mainly on air pollution in the USA, generally found a negative association between environmental risk and income (e.g., Freeman 1972, Zupan 1973, Asch and Seneca 1978, Harrison and Rubinfeld 1978). In the 1980s, studies also began to focus on waste management facilities and, in the 1990s, on toxic releases, using newly available data from the Toxic Release Inventory (TRI). Most of these studies also ﬁnd that lower income groups suﬀer more exposure to hazardous waste management facilities (e.g., Hamilton 1993 and 1995, Yandle and Burton 1996) and to toxic releases (e.g., Brooks and Sethi 1997, Arora and Cason 1999, Ash and Fetter 2004). However, the negative relation between income and environmental risk is not a universal ﬁnding. For example, Anderton et al. (1994) report no link between poverty rates and the location of waste management facilities in US cities, and Harrison and Antweiler (2003) do not ﬁnd an association between income and pollution releases in Canada. Ringquist (2005) conducts a meta-analysis to identify the sources of diﬀerences in the results; he ﬁnds that studies report a more negative association with income when they focus on the location of facilities, rather than the level of pollution, and a greater association with poverty when the exposure area is narrowly deﬁned. In addition, as research has progressed, it has evolved from descriptive studies into causal analyses of the distribution of pollution.23 When studies report no eﬀect of income, they may not mean no unconditional correlation between environmental risk and income, but lack of an additional eﬀect beyond the eﬀects of race, population density, education levels, etc. For incidence analysis, however, we may care about the unconditional relationship. Literature on other countries is sparse. Pearce (2003) concludes that ‘while the evidence is very limited, the data for the United Kingdom suggest that the existing distribution of risks is biased towards the poor’ (p. 23). Hamilton (2005) provides a detailed survey of studies from the USA and other OECD countries with a focus on risks from hazardous waste, toxic chemical releases, and contaminated sites. As in the USA, the studies he surveys mostly ﬁnd greater pollution in poor neighborhoods, but exceptions arise. 4.1.2 Measures of environmental quality An important concern with this literature is whether environmental quality is appropriately measured. With a slight modiﬁcation from Pearce (2003),

The incidence of pollution control policies

21

consider a hierarchy of measures where each level builds on the previous one: (i) existence of a polluting facility, (ii) emissions of pollutants, (iii) ambient concentrations, (iv) exposure (which depends on ambient concentrations as ﬁltered through an individual’s behavior), and (v) personal risk (which depends on exposure as well as personal characteristics such as age and prior health status). Unfortunately, exposure and personal risk are not directly observed, and therefore most studies use measures (i), (ii), or (iii). Existence of a polluting facility is obviously not an ideal measure of emissions, and local emissions may diﬀer greatly from ambient pollution because of the timing of releases, possible import/export of emissions from/to neighboring regions, and factors such as topography and wind that aﬀect dispersion rates. In the past, concentration data were limited to the small number of air pollutants for which a sizable monitoring network is available; more recently, ambient measures have been simulated by running emissions data through dispersion models (e.g., Shadbegian et al. 2005, Ash and Fetter 2004). However, a casual examination of the empirical literature suggests that the choice of emissions versus concentration does not aﬀect the association between environmental risks and income, because neither measure depends on characteristics of exposed individuals. In contrast, the diﬀerence between (iii) and (iv) may well skew the results because rich and poor households likely diﬀer in their intentional and unintentional averting behavior. For example, low-income individuals may face greater risks from proximity to groundwater contamination because they are less likely to use ﬁlters or bottled water and may have more exposure to air pollution if they are more likely to work in outdoor occupations, such as construction. The distinction between (iv) and (v) may also be important because health status, and hence susceptibility to pollution-induced illness, bears a close relationship to income.24 Thus the widespread practice of using measures (i) to (iii) instead of (iv) or (v) might lead to serious understatement of the income/risk association. A few studies do work with measures of exposure or personal risk. For example, Brajer and Hall (1992) ﬁnd a correlation between income and exposure, where the latter is measured by combining pollution data with a behavioral model of time spent in various indoor and outdoor activities. Hamilton and Viscusi (1999) ﬁnd elevated cancer risks for minorities from contaminated sites to be cleaned up under the Superfund program, using EPA risk measures that depend in part on behavior (they do not report risks by income class). However, the quality of the risk measures in these types of studies depends critically on the accuracy of the behavioral modeling; more econometric work on such behaviors is needed.

22

Yearbook of environmental and resource economics

4.2 Evidence on the Distribution of Improvements from Environmental Policies Even if lower-income households are disproportionately exposed to existing environmental risks, appropriately measured, they may not beneﬁt disproportionately from environmental policies if the resulting environmental improvements are non-uniform. A limited number of empirical studies have addressed the distribution of improvements from pollution control policies. One diﬃculty in this literature is that the researcher needs to assume a counterfactual scenario representing what would have happened in the absence of policy change. This problem has been handled in a wide variety of ways. One approach is to focus on changes in environmental quality over time that might be largely associated with policies. For example, Kahn (2001) examines changes in air quality at monitoring stations in the Los Angeles basin and ﬁnds that improvements have been greater in low-income census tracts. Such comparisons eﬀectively assume that, without policy intervention, pollution would be unchanged. The implication of this assumption for the distribution of policy eﬀects is unclear. For example, if pollution would have grown worse without policy intervention, it might have exacerbated inequality in existing risks; however, worsening pollution might also have spread the risks to relatively clean (and wealthy) neighborhoods. A second approach is to simulate policy eﬀects through some more sophisticated counterfactual. For example, Shadbegian et al. (2005) examine the distribution of beneﬁts (and costs) of the US SO2 allowance-trading program. They compare the actual pattern of emissions in 1995 to a ‘without-policy’ baseline that applies a utility’s emissions rate before the policy (1993) to its output with the policy (1995). They ﬁnd that the poor received per capita beneﬁts that were 5 percent lower than the average beneﬁts over the whole population: that is, the distribution of ambient concentration reductions was slightly regressive. A third approach is to examine the intensity of regulatory eﬀort. Hamilton and Viscusi (1999) ﬁnd some evidence of more aggressive policy responses for low-income communities in the Superfund program: target risk levels for cleanup are set lower (i.e., a more extensive cleanup is chosen) when the median income within 1 mile of the site is lower than average. This association may reﬂect behavioral assumptions – regulators could think that low-income residents practice less averting behavior and thus require more protection to achieve the same level of safety – but could also reﬂect special attention to low-income communities because of environmental equity concerns.

The incidence of pollution control policies

23

Gray and Shadebegian (2004) also study the intensity of regulatory eﬀort, as measured by inspection and enforcement eﬀort for air and water pollutants. They ﬁnd that facilities with larger poor populations within 50 miles receive more frequent inspections and less frequent enforcement actions. Both these variables potentially measure plants’ compliance, as well as the government’s eﬀort, and so are open to several interpretations. Nonetheless, a possibility is that the government devotes more attention to inspections near the poor and thus needs to devote less attention to enforcement, again supporting the view that government eﬀort is progressively distributed. 4.3

Relationship between Environmental Improvements and Welfare

Both the large environmental justice literature and the more limited literature on the distribution of policy improvements is focused almost entirely on environmental risks. However, changes in the pattern of environmental risks may give a misleading picture about changes the distribution of household welfare relative to income. This disparity between risks and welfare may arise because individuals may value risks diﬀerently and because they may be partly compensated for changes in risks via market price or wage adjustments. 4.3.1 Diﬀerential valuation of the environment Most plausible forms of households’ utility functions have willingness to pay (WTP) for environmental improvements that rise with income;25 however, the implication of this positive relationship for the distribution of environmental beneﬁts depends critically on the income elasticity of WTP for environmental improvements (Ebert 2003). If this elasticity is greater than unity, the value of a uniform environmental improvement rises relative to income. Although a positive income elasticity seems extremely likely a priori, whether the elasticity is greater than one is an empirical issue.26 Incidentally or intentionally, many studies provide information on this elasticity. Contingent valuation (CV) studies, which survey households on their WTP for speciﬁc environmental goods, often ask about income (among other household characteristics). In a survey of such studies in Europe, Kriström and Riera (1996) conclude that the income elasticity of WTP is positive but less than one; Hokby and Soderqvist (2003) ﬁnd similar results from a review of CV studies for a range of environmental services in Sweden. Pearce (2003) argues that a range of 0.3 to 0.7 ‘seems about right’ based on his review of international CV studies. A few revealed preference studies also measure income elasticities of WTP but reach highly conﬂicting results. For example, Sieg et al. (2004)

24

Yearbook of environmental and resource economics

estimate a general equilibrium model of property price changes in response to air quality improvements and ﬁnd an income elasticity of WTP of more than 4. In contrast, Boyle et al. (1999) ﬁnd no association between the value of lake water clarity in Maine and income among the high-income households, suggesting a WTP elasticity of zero! Another source of information on the income elasticity of WTP is the hedonic wage analyses and surveys used to value health risk reductions. Viscusi and Aldy (2003) summarize income elasticities from four previous meta-analyses of value of statistical life studies (Liu et al. 1997, Mrozek and Taylor 2002, Bowland and Beghin 2001, and Miller 2000) and then conduct their own meta-analysis on studies from ten countries. Their preferred estimate for the income elasticity is .0.46–0.49.27 4.3.2 Market responses to environmental changes Household welfare is also aﬀected by changes in market prices that might accompany improvements in environmental quality. In particular, environmental improvements might be capitalized into housing prices or built into wages through wage premiums for workers with jobs in polluted areas (Roback 1982, Blomquist et al. 1988).28 Empirical evidence on capitalization in housing prices is widespread from the extensive literature on the hedonic valuation of environmental amenities (see for example a meta-analysis of studies of air pollution by Smith and Huang 1995). Chay and Greenstone (2005) ﬁnd a substantial housing price eﬀect from recent reductions in total suspended particulates, but also ‘a precisely estimated zero’ eﬀect on wages, suggesting that the housing market is the principal source of compensation for environmental improvements. Property ownership is skewed toward well-oﬀ households; in the USA in 2003, 73 percent of households above the poverty line owned homes, compared with 43 percent of households below it (US Census Bureau 2004, Table 964). These diﬀerences in ownership rates could have dramatic eﬀects on the distribution of beneﬁts. Households that own their home receive the beneﬁt of any unanticipated environmental improvement that is fully capitalized, whereas households that rent may end up paying higher rental rates. In fact renters may even be made worse oﬀ on net if increased rental payments outweigh their valuation of the local environmental improvement. Evidence of relative out-migration of poorer families from areas of environmental improvements supports this possibility (e.g., Sieg et al. 2004, Banzhaf and Walsh 2004), though migration only indicates harm for renters; when families who own homes leave, they take with them capital gains from the environmental improvement.

The incidence of pollution control policies

4.4

25

Integrated Cost and Beneﬁt Studies

Two older studies attempt to integrate cost and beneﬁt distributions to study overall incidence. Gianessi et al. (1979) assume the beneﬁts of the 1970 Clean Air Act are proportional to emissions reductions.29 Overall, they ﬁnd beneﬁts fall short of policy costs (discussed in Section 3 above) for all income groups, though the pattern of net losses is only mildly regressive. Dorfman (1977) integrated environmental beneﬁts into the Dorfman and Snow (1975) cost analysis of all federal environmental regulations (see above), using a survey asking people whether they were willing to pay certain dollar amounts for ‘cleanup of the natural environment.’ The survey results indicated that wealthier people were willing to pay more relative to income; in fact, beneﬁts exceeded costs for the top income quintile, but were less than costs for the bottom income quintile. However, more recent aggregate (rather than distributional) studies ﬁnd a much more favorable cost–beneﬁt comparison for the Clean Air Act, due to stronger evidence on the link between pollution and health eﬀects; this underscores the need for updating the earlier analyses. 4.5

Summary

Despite the extensive literature, the existing distribution of environmental risks is still controversial; many studies ﬁnd that the poor face higher risks, but this result appears to be sensitive to the nature of the risk studied, with more recent studies chipping away at this as stylized fact. However, if exposure and health sensitivity considerations were adequately considered, it would likely tilt existing risks toward the poor. But even if the poor disproportionately bear environmental risks, it does not necessarily follow that environmental policies have progressive beneﬁts. Although there are some cases where actual policies have tended to skew beneﬁts toward the poor (e.g., Superfund), this ﬁnding does not apply in other cases (e.g., the SO2 trading program). Translating environmental improvements into welfare further muddies the picture. On the one hand, most available evidence suggests an income elasticity of WTP below unity, implying that the same emissions reduction for rich and poor households would represent a larger share of the poor households’ income. On the other hand, capitalization of environmental improvements into housing values may disproportionately reduce beneﬁts to low-income households. Much work remains to be done on all of these issues.

26

Yearbook of environmental and resource economics

5. INCORPORATING DISTRIBUTIONAL CONSIDERATIONS INTO ENVIRONMENTAL COST–BENEFIT ANALYSIS The standard practice among environmental economists – as in other ﬁelds of economics – is to keep eﬃciency and distributional issues separate. Very few cost–beneﬁt analyses include more than a cursory mention of distribution, while incidence studies often ignore eﬃciency. There are often important methodological reasons for this separation, because models or approaches that work well for eﬃciency analysis can be completely inadequate for equity analysis, and vice versa.30 And if the question is simply whether or not to implement a new regulation or tax, this decoupled approach can work well; policymakers can evaluate evidence on the aggregate net beneﬁt and on distributional eﬀects from diﬀerent studies, and decide for themselves what weight to put on each. But this approach is inadequate in other cases; for example, in order to judge the optimal stringency of a particular policy intervention one must either ignore distribution entirely, or utilize an approach that integrates eﬃciency and equity.31 Below, we discuss three such approaches: using a social welfare function or set of distributional weights; imposing constraints on the losses that can be imposed on particular groups; and using ‘distribution-neutral’ analysis. 5.1

Social Welfare Function/Distributional Weights

This approach allows the value of a dollar of beneﬁts or costs to vary based on who receives that beneﬁts or bears that cost. One way to do this is to evaluate policy based on how it aﬀects social welfare, which is a function of the utility of each individual (i.e., social welfare is given by the function W(U1, U2, U3), where Ui is utility for individual i). Under this approach, concavity with respect to income in either the individual utility functions (i.e., decreasing marginal utility per dollar of income), or in the social welfare function (i.e., decreasing marginal social welfare per util) causes the marginal social welfare per dollar for any given individual to be decreasing in that individual’s income. A simpliﬁed version of this approach assigns a ‘distributional weight’ for each individual, and then calculates the weighted sum of costs and beneﬁts.32 Distributional weights (or values for marginal social welfare per dollar) that decrease with income imply that society cares about equity as well as eﬃciency. It is common to set these weights using an isoelastic function, which implies that the weight on individual i is equal to (1 ) 1 i (Y1 Y ) , where Y is mean income and represents society’s i

The incidence of pollution control policies

27

aversion to inequality. For 0 the weights are constant, equivalent to an unweighted cost–beneﬁt measure; 1 gives weights inversely proportional to income; and implies zero weight on all but the very poorest individual.33 This approach is only slightly more complex than standard (unweighted) cost–beneﬁt analysis; the social welfare function looks much like an individual utility function, and thus the intuition and techniques developed for consumer utility maximization translate easily into social welfare maximization. And the distributional weights have an obvious and intuitive interpretation. Recent environmental applications of this approach include Fankhauser et al. (1997) on the aggregation of projected climate change damages across rich and poor nations; Mayeres (2001) on various transportation policies; Parry (2004) on the social costs of emissions permits; and literature on optimal environmental taxes with heterogeneous agents (e.g., Cremer et al. 1998 and 2003, Eskeland 2000, Mayeres and Proost 1997 and 2001, and Pirttilä and Tuomala 1997). The major diﬃculty, of course, comes in determining what set of distributional weights to use. The choice of weights can have a dramatic eﬀect on the magnitude of the change in social welfare, and can even alter the sign of that change. One method, analogous to using revealed preference to infer individual utility functions, is to infer distributional weights based on the tradeoﬀs made in other government decisions.34 However, these estimates may be an unreliable indicator of society’s true preferences, because policy decisions are not strictly the result of benevolent optimization, but are determined at least in part by the interplay of powerful interest groups. Alternatively, a set of weights could be chosen that appear reasonable, but that choice is inherently arbitrary, and may lead to the acceptance of policies that would perform very badly on an unweighted cost–beneﬁt analysis (e.g., Harberger 1976). One possible response to this problem is to consider a range of diﬀerent social welfare functions (e.g., by varying the inequality aversion parameter), and look for robust qualitative results (e.g., instrument A always yields greater social welfare than instrument B). But such robust results may not exist, and there is little consensus over the appropriate range of social welfare functions to consider.35 5.2

Distributional Constraints

This approach uses an unweighted cost–beneﬁt measure, but rules out policies that would make particular groups worse oﬀ. For example, when evaluating an automobile emissions testing program, one might impose the constraint that low-income households cannot be made worse oﬀ.

28

Yearbook of environmental and resource economics

This would rule out a policy that simply imposes a stringent emissions limitation, because many older cars owned by low-income households would fail that test. Some real-world testing programs satisfy this distributional constraint by exempting low-income households. But a more eﬃcient way to satisfy the constraint would be for the government to oﬀer to buy back those older cars, rather than simply banning them, or to compensate low-income households in some other way (such as with a tax credit). We are not aware of any environmental study that explicitly incorporates distributional constraints based on household income. But some studies do impose other distributional constraints. For example, recent papers by Bovenberg and Goulder (2001) and Bovenberg et al. (2004) examine pollution control costs under the constraint that proﬁts of particular industries do not fall. The rationale for that constraint is that politically powerful industries could block any policy that would make them worse oﬀ. In these studies, politically powerful industries are still subject to emissions controls, but they receive compensation through rents obtained from the allocation of free emission permits. Distributional constraints are straightforward to implement, as long as the set of constraints is not overly complex. But the choice of constraints will typically be somewhat arbitrary. Bovenberg and Goulder’s (2001) analysis, for example, imposes the constraint that utilities and fossil fuel producers cannot suﬀer a net loss from policies to reduce carbon emissions, but imposes no such constraint on other industries or consumer groups. Another problem is that satisfying the distributional constraints could potentially be very costly in terms of economic eﬃciency. In theory, distributional constraints could lead to the rejection of policies that generate tiny losses for a favored group, even if they produce huge beneﬁts for other (non-favored) groups. A more practical example is the policy mentioned above that exempts low-income households from auto emissions testing. This policy is ineﬃcient because it fails to exploit potential reductions in emissions from those households’ cars that, in the absence of distributional concerns, would probably be the most cost-eﬀective reductions available. And in the Bovenberg and Goulder, and Bovenberg et al. papers eﬃciency losses result because compensation schemes erode the amount of government revenue collected from auctioned permits (or emission taxes) that can be used to ﬁnance cuts in other distortionary taxes. Moreover, in these studies the fraction of permits (or of emissions tax revenue) that must be used for compensation increases with the level of abatement; indeed, beyond a certain reduction in emissions, full compensation becomes infeasible.

The incidence of pollution control policies

5.3

29

Distribution-neutral Analysis

The distribution-neutral approach evaluates a particular policy by creating a hypothetical ‘neutralized’ version of that policy – one with no distributional eﬀects – and seeing if that policy passes a cost–beneﬁt test. As proposed by Kaplow (1996, 2004), this entails two steps. The ﬁrst is to calculate what changes to the income tax schedule would exactly compensate each income group for the eﬀects of the policy, thereby neutralizing the distributional eﬀects of the policy. This would require a compensating income tax cut for any income group that would suﬀer a loss from the policy, and a compensating tax increase for any group that would receive a gain. Thus the combination of the policy in question and the compensating changes in the tax schedule would leave each income group no better or worse oﬀ than in the status quo. The second step is to estimate whether that combination – the policy in question plus the compensating income tax changes – would increase or reduce net government revenue. If it would increase net revenue, then it is possible to make all income groups better oﬀ (by distributing the excess revenue as a lump-sum transfer). But if it would reduce net revenue, then the opposite result holds: the government could not possibly compensate everyone who is made worse oﬀ by the policy, even if it were to tax away all of the gains from those who beneﬁt from the policy.36 Williams (2004) provides the ﬁrst empirical application of this approach, using it to calculate the optimal gasoline tax. The gasoline tax is regressive, so neutralizing its distributional eﬀects requires making the income tax schedule more progressive, which has an eﬃciency cost.37 Consequently, the distribution-neutral approach yields a lower optimal gasoline tax rate (which Williams estimates at $0.91 per gallon in 1997 dollars) that is lower than the eﬃciency-maximizing tax rate (estimated at $1.03 per gallon). Thus, this approach provides a way to bring distributional considerations into cost–beneﬁt analysis, while still avoiding the problems inherent in the previous two approaches. It is similar in concept to the Hicks–Kaldor criterion in that it looks for a potential Pareto improvement, but it accounts for the eﬃciency costs of redistribution, whereas the Hicks–Kaldor criterion implicitly assumes that redistribution is costless. However, just because the income tax schedule could be adjusted to prevent anyone being made worse oﬀ doesn’t necessarily mean that it will be (of course, the same criticism applies to Hicks–Kaldor). A more serious problem is that the compensation might not even be possible; if there is heterogeneity within an income group, such that some people in this group gain from a policy while others lose, then the income tax cannot neutralize distributional eﬀects within that group.38

30

Yearbook of environmental and resource economics

Another potential problem is that distribution-neutral analysis can be complicated to implement when the welfare eﬀects of a policy are a nonlinear function of income. The cost would be nonlinear if the income elasticity of demand for the polluting good varied across the income distribution (as is the case for gasoline, which has an income elasticity less than one over most of the income distribution, but greater than one near the bottom of the income distribution, where owning a car is a luxury). The environmental beneﬁts are nonlinear if the income elasticity of WTP varies over the income distribution, or if exposure to a pollutant varies nonlinearly with income. In any of these cases, the hypothetical compensating change in the income tax schedule will also be nonlinear, and evaluating the eﬃciency eﬀects of that compensating change will be complicated. Surprisingly, Kaplow (2004) shows that distribution-neutral analysis yields the same result as one would get by ignoring both distributional eﬀects and interactions with pre-existing tax distortions. Thus he argues that imposing distribution neutrality might actually simplify cost–beneﬁt analysis. However, Williams (2005) shows that this result holds only under a restricted speciﬁcation for utility, which is frequently rejected in empirical studies. 5.4

Conclusions

Given the drawbacks of each of these approaches, they should be used with caution. In most cases, it is better to provide separate measures of the eﬃciency and equity eﬀects of a policy than to attempt to integrate them into a single measure. However, there are some problems, such as calculating socially optimal policies, for which an integrated measure may be appropriate. In such cases, the best approach for a given problem will depend on the characteristics of the problem. If distributional eﬀects matter primarily because of political concerns (i.e., imposing too heavy a burden on a particular group will cause that group to block the policy), then it is straightforward to represent those concerns as distributional constraints. On the other hand, if distributional eﬀects matter primarily for equity reasons, then using distributional weights or distribution-neutral analysis will likely be the best approaches. Because it does not rely on an arbitrary assumption about society’s aversion to inequality, distribution-neutral analysis can provide the most rigorous and objective results, but only if it is applied without imposing unrealistic restrictions on preferences – and it can be very complicated to implement without those restrictions. Often, a simpler approach will be needed, and in such cases the best option is probably to use distributional weights based on an isoelastic function, and to consider a wide range of possible values for the inequality aversion parameter.

The incidence of pollution control policies

31

6. LESSONS LEARNED AND TOPICS FOR FUTURE RESEARCH Although in general (with some exceptions) low-income households appear to bear a disproportionate share of existing environmental risks, policies that reduce environmental risks are not necessarily progressive. For example, the geographic pattern of emissions reductions may be uneven, and induced changes in property values may indirectly beneﬁt the wealthy at the expense of the poor. Moreover (again with some exceptions), the costs of environmental policies tend to fall most heavily on poorer groups through increases in product prices, as energy goods are necessities. And the type of emissions control instrument can be critically important; freely allocated tradable emission permits may actually hurt the poor the most, as they transfer income to shareholders via scarcity rents created at the expense of higher prices. On the other hand, emissions taxes (or auctioned emission permits) oﬀer the opportunity to oﬀset regressive eﬀects, if revenues are recycled to ﬁnance progressive changes to the tax system. Revenue-raising environmental policies have, however, proven extremely diﬃcult to implement in the USA; recent legislation has instead favored grandfathered emissions permits where policy rents accrue to regulated ﬁrms.39 Although the chance of new environmental or energy taxes being introduced in the next few years appears very remote, it is conceivable that this situation may change further down the road, given continuing pressures to ‘do something’ about US greenhouse gas emissions and looming deﬁcit problems from retiring baby boomers. Although literature on incidence of pollution control policies has been expanding rapidly in recent years, the above discussion reveals that existing analysis is very thin, or even non-existent, in a wide range of critical areas. We ﬁnish by summarizing a variety of topics in particular need of attention. Empirical studies on the extent to which the costs of environmental policies are passed forward into higher prices of consumer products would be extremely valuable; at present, empirical analyses typically assume 100 percent pass-through based on the assumption of competitive, constant returns production.40 Also on the cost side, there is a need for more unifying analyses that compare the incidence of a broad range of alternative control instruments for a particular pollutant (taxes, permits, technology mandates, etc.) on a consistent basis. And how relative incidence eﬀects of policies depend on factors such as the extent of pollution reduction, abatement cost heterogeneity among ﬁrms, changes in relative factor prices, non-competitive pricing, and diﬀerences in emissions intensity between marginal and infra-marginal production all need to be explored. This

32

Yearbook of environmental and resource economics

research eﬀort would aid policymakers in gauging when the choice of instrument is important for distributional concerns and when it is not. A related issue is the extent to which adverse distributional impacts might be oﬀset via targeted tax reductions from the recycling of revenue raised from emissions taxes and auctioned permits. It would also be useful to examine the eﬃciency gains forgone by targeting tax reductions to speciﬁc income groups, in place of more broadly based tax reductions. Again, research on these issues would inform policymakers about the strength of the economic case for using revenue-raising policies in preference to grandfathered permit trading programs. On the beneﬁt side, research should address the distribution of physical environmental beneﬁts from policy changes (as opposed to the existing distribution of environmental risks), ideally with account taken of how health eﬀects depend on exposure and personal characteristics. And to quantify the pattern of welfare gains from environmental improvements, these estimates need to be merged with estimates of the willingness to pay for environmental quality across households, as well as estimates of potentially oﬀsetting eﬀects from price adjustments in housing markets. Moreover, very few studies integrate both the beneﬁt side and cost side of particular environmental regulations, in order to obtain the distribution of overall welfare eﬀects across household groups. The last comprehensive study of the incidence of federal environmental regulations uses data from the 1970s. Finally, further development of alternative approaches for incorporating distributional considerations into cost–beneﬁt analyses is needed. Many problems require an integrated analysis of eﬃciency and equity, but the choice of distributional weights or distributional constraints is arbitrary. Distribution-neutral analysis avoids that problem, but is diﬃcult to implement except under very restrictive assumptions. Progress in this area – a less arbitrary means of determining distributional weights or constraints, a simpler and more general method for distribution-neutral analysis, or an entirely new approach altogether – would be very valuable.

NOTES * 1. 2.

The authors are grateful to Spencer Banzhaf, Henk Folmer, Adam Rose, Tom Tietenberg, and two reviewers for excellent comments and suggestions. For example, in 1994 the Clinton Administration directed the EPA to study the pattern of environmental hazards across diﬀerent income and racial groups and to explore options for reducing disparities. The latter two issues are particularly contentious in climate policy. Estimated future damages from atmospheric accumulation of greenhouse gases are highly sensitive to assumptions about long-range discount rates, and whether diﬀerent weights are attached

The incidence of pollution control policies

3. 4.

5. 6. 7. 8.

9.

10.

11.

12.

13. 14.

33

to the welfare of poor nations that are most vulnerable to climate change (see e.g., Portney and Weyant 1999, Azar and Sterner 1996, Carraro 2000, Stevens and Rose 2002). We do not cover incidence of solid waste policies and noise pollution; for some discussion of these issues see Kinnaman and Fullerton (2000), and Feitelson et al. (1996). See for example Fullerton and Rogers (1993), Casperson and Metcalf (1994), Walls and Hansen (1999), Slesnick (1994), and the discussion in Metcalf (1999). An alternative approach is cohort analysis (e.g., Gale et al. 1996), which considers only households at similar stages of their lifecycle; however, this approach does not address the problem of annual income variation. For example, if we observe that a person has 14 years of education, we have no way of knowing whether they will go to college for two more years and earn a Bachelor’s degree or not. If abatement costs are unknown they could be approximated by tE (e0j ej ) 2, if it is reasonable to assume marginal costs are linear over the relevant range (this is easily seen from Figure 1.2). If not, abatement costs can be bounded by {0, tE(e0 e)}. See Willig (1976). From the Slutsky equation, the diﬀerence between the uncompensated and compensated own-price demand elasticity equals the income elasticity times the budget share. For each unit of emissions ﬁrms must either buy a permit from other ﬁrms, or forgo selling one of their permits; either way the cost to the ﬁrm is , and ﬁrms will abate until the incremental cost equals . This leaves aside some complications, such as transactions costs to permit trading and uncertainty over control costs. Nearly all permits have been given out for free in existing programs, including those to reduce the lead content of gasoline, ozone-depleting chemicals, and utility emissions of SO2 in the USA, and the CO2 trading program introduced in the European Union in January 2005. For the USA, total stock and bond ownership across households can be measured from the CES; Dinan and Rogers (2002) and Parry (2004) compute that the top income quintile owns 86 percent and 53 percent of total stocks, respectively. The diﬀerence is that Parry includes retirement capital, which is more evenly dispersed among income groups than non-retirement ﬁnancial capital, while Dinan and Rogers exclude it. Ideally information should be used on stock ownership in polluting ﬁrms, rather than total stock ownership across all ﬁrms, but the former is particularly diﬃcult to obtain, given households own most of their stocks indirectly through large institutional investors. This is because there is no binding quota imposed on economy-wide emissions; new ﬁrms are free to enter the market without having to buy emissions permits from incumbent ﬁrms. Similarly, no rents would be created if the government took an existing command-and-control system and allowed ﬁrms to trade credits, unless the government also imposed a cap on economy-wide emissions below the initial level. However, results from general equilibrium incidence models are notoriously complex and ambiguous. For example, to the extent that the net return on capital is determined on world capital markets, this will cushion the potential fall in the relative return to capital. On the other hand, allowing for imperfect mobility of capital across industries may increase the burden that can be borne by capital. And more generally results also depend on the relative substitutability of labor and capital for polluting inputs, about which very little is known. Theoretical results on the general equilibrium incidence of pollution mandates have similar ambiguities (Fullerton and Heutel 2005). Although Bento et al. allow for scrappage and changes in vehicle holdings, their simulations only consider impacts in the ﬁrst year after the policy and not longer-run impacts. Future work by the authors will conduct a more long-run analysis. The lifetime measure is obtained by calculating a ‘typical’ consumption path for various subgroups deﬁned by age and education, and constructing lifetime consumption for each individual in the sample by scaling their current consumption relative to the average for their age/education subgroup. A similar exercise is carried out to obtain measures of lifetime direct and indirect energy taxes paid; lifetime incidence is lifetime taxes paid as a percentage of lifetime consumption.

34 15. 16.

17.

18. 19.

20. 21.

22.

23.

24.

Yearbook of environmental and resource economics For revenue neutrality in the second scenario, they end up with an ad valorem transport subsidy of 23 percent. In principle, the emissions tax is the most eﬃcient one, as it encourages abatement measures to reduce emissions per unit of fuel, improvements in fuel economy and reduced driving. The other two taxes do not optimally exploit all of these behavioral responses. Emission taxes could be implemented based on data collected during periodic emissions rate inspections, and mileage data collected during the same inspection, or on a continuous basis via Global Positioning Systems. Ideally, an emissions tax would also vary by location, time of day and temperature, as all these factors inﬂuence marginal pollution damages. At the present time, it is probably still impractical to implement a perfect emissions tax. In a related study, West (2005) analyses the same taxes as Walls and Hanson (1999), but integrates diﬀerential behavioral responses across households. Results are broadly consistent with earlier ones, though the degree of regressivity is mitigated somewhat because lower-income households are more price responsive than upper-income households. On the other hand they ﬁnd auctioned carbon permits to be regressive if revenues are used to cut payroll taxes, and highly regressive if they are used to cut corporate taxes. These ﬁndings are similar to those in an earlier study by Harrison (1975), though that study looked only at vehicle emissions regulations, and also accounted for tax eﬀects and stockholder burdens from reduced proﬁts for vehicle manufacturers. Harrison estimated burden to income ratios of 1.5 percent for the lowest income quintile and 0.8 percent for the top income quintile. Consistent with Dorfman and Snow, other groups in the lower half of the income distribution suﬀered a burden to income ratio about double that of deciles in the upper half of the distribution. There is a very small literature on incidence eﬀects of energy eﬃciency standards for home appliances (e.g., refrigerators, washers, dryers), and demand-side management programs. Sutherland (1991) argued that there is a positive net burden from appliance standards (higher product prices exceed discounted savings from improved energy eﬃciency) and that the burden is borne disproportionately by poorer households. However, according to estimates by Stoft (1993) the net burden of a given standard is quantitatively very small. Sutherland (1994) ﬁnds that wealthier households are more likely to participate in voluntary utility demand-side management programs (e.g., home energy audits, rebates for the purchase of energy-eﬃcient appliances) but that such programs do not signiﬁcantly reduce electricity consumption. He speculates that the higherincome households who participate are probably being subsidized to undertake conservation measures that they would take anyway, suggesting such programs have regressive eﬀects. Initially, studies used convenient but approximate measures of aﬀected populations, such as counties or postal codes. As techniques for working with geographic information became more sophisticated, researchers were able to measure the population residing within certain distances from pollution sources or monitoring stations. However, tradeoﬀs arise in the choice of spatial scales (Bowen 2002): with smaller geographical units localized inequities are less likely to be overlooked, but measures of individuals’ overall exposure may be less reliable. For example, an individual may work near a polluting facility and have considerable exposure, but live in a distant area and therefore be treated as having minimal exposure. For example, Been and Gupta (1997) and Wolverton (2002) look at the siting of new facilities (hazardous waste facilities and TRI facilities, respectively) to distinguish inequities in the siting process from the decision of households to move near facilities. They both ﬁnd that poor neighborhoods are less likely to host new facilities. For example, asthma, which greatly increases risk from air pollution, is much more prevalent in lower-income individuals; a recent survey in the USA found adult asthma frequency of 9.8 percent among persons with family incomes of less than $15 000 relative to only 5.9 percent among persons with family incomes of greater than $75 000 (CDC 2001).

The incidence of pollution control policies 25.

26. 27. 28.

29.

30.

31.

32.

33. 34.

35. 36.

37.

35

The use of WTP in estimating welfare impacts of policies for diﬀerent income groups is routine among economists, although highly contentious outside of the economics profession (e.g., Heinzerling and Ackerman 2002). It raises conceptual concerns and political objections, especially when the environmental quality in question aﬀects risks of death as it seems to imply that some people are ‘worth’ more than others. A positive income elasticity of WTP is broadly consistent with the U-shaped relationship between environmental quality and per capita income estimated by some studies in the ‘Environmental Kuznets Curve’ literature (Israel and Levinson 2004). Valuations of illness are now beginning to supplement the large literature on death. For example, Dickie and Hubbell (2004) estimate an income elasticity of WTP to avoid acute respiratory disease of 0.5 based on survey data. One caveat is that housing prices (or wages) in areas unaﬀected by environmental quality improvements may also change; if dirty regions become cleaner, while clean regions are unaﬀected, the advantage of living in the latter areas are diminished and outward migration from them may lower their property values. Ideally, the distributional implications of these other price changes should also be taken into account. Beneﬁt estimates come from an earlier study where national estimates were allocated to local areas based on an index that accounted for local pollution, the population at risk and land areas. Local pollution was estimated based on information about industrial activity, vehicle ownership and home heating. Aggregate time-series datasets, for example, can be suﬃcient for eﬃciency analysis, but do not include the information needed to measure distributional eﬀects. Similarly, distributional analyses often ignore quantity responses to price changes, which are key to estimating eﬃciency eﬀects. Studies that ignore distribution often justify this as an implementation of the Kaldor–Hicks criterion; if a policy passes an unweighted cost–beneﬁt test, then it represents a potential Pareto improvement, because those who beneﬁt from the policy could compensate those who lose. But whether the policy improves overall social welfare will also depend on its distributional eﬀects. It is often argued that environmental policy should not consider distribution, but should focus exclusively on eﬃciency, because tax and transfer policy is a much more direct way to achieve distributional goals. But redistribution through taxation of labor and capital, and means-tested beneﬁts, has limits because it involves eﬃciency costs. To the extent that equity goals are not fully addressed through the tax and beneﬁt system, the distributional eﬀects of environmental policies remain a concern. Note that this weighted cost–beneﬁt measure provides a ﬁrst-order approximation to the change in social welfare if the distributional weight on any given individual equals the marginal social welfare per dollar of income (i.e., if the weight for individual i equals (W/Ui)(Ui/Yi), where Yi is that individual’s income). These weights approximate an isoelastic social welfare function if utility is roughly proportional to income. In this case, 0 implies a utilitarian social welfare function, while implies a Rawlsian social welfare function (Atkinson and Stiglitz 1980, p. 340). For example, Gruber and Saez (2002) calculate a set of distributional weights that are consistent with the eﬃciency cost of raising income tax revenue from diﬀerent income groups. It is not always possible to solve this ‘inverse optimum problem’; for example, Ahmad and Stern (1984) could not ﬁnd a positive set of weights that might justify the commodity tax system in India, which implies that commodity tax reform could make all income groups better oﬀ. For example, Parry (2004) uses values of 0, 0.5 and 1 for the inequality aversion parameter , Cremer et al. (1998) uses 0.1 and 1.9, and Fankhauser et al. (1997) use 0, 1 and . This approach implicitly takes into account the eﬃciency costs of redistribution, because the compensation comes not from individual-speciﬁc lump-sum taxes and transfers, but rather from changes in the income tax schedule, and thus may change the excess burden of the income tax. The compensating change in the income tax schedule depends on the distribution of beneﬁts as well as the distribution of costs. However, Williams (2004) assumes that the

36

38.

39. 40.

Yearbook of environmental and resource economics external beneﬁts of lower gasoline consumption are proportional to income, so compensating for those beneﬁts does not aﬀect the progressivity of the income tax. This problem arises in Williams (2004), because at any given income level, some households use more gasoline than others, and bear more of the burden of increases in the gasoline tax. Thus, while the tax schedule can be adjusted so that for any level of income, the average household at that level is exactly compensated for the increased gasoline tax, this will undercompensate households that use more gasoline than the average for their income level, and will overcompensate those that use less than average. The McCain–Lieberman bill to reduce carbon emissions, and recent EPA rulings to reduce utility emissions of SO2, NOX and mercury, would primarily grandfather permits to existing emission sources. There is an empirical literature on the pass-through of taxes into consumer prices in contexts other than environmental policy (see Fullerton and Metcalf 2004, section 2.6).

REFERENCES Ahmad, Ehtisham and Nicholas Stern (1984), ‘The theory of reform and Indian indirect taxes,’ Journal of Public Economics, 25, 259–98. Anderton, Douglas L., Andy B. Anderson, John Michael Oakes and Michael R. Fraser (1994), ‘Environmental equity: the demographics of dumping,’ Demography, 31, 229–48. Antweiler, Werner, Brian R. Copeland and M. Scott Taylor (2001), ‘Is free trade good for the environment?’, American Economic Review, 91, 877–908. Arora, Seema and Timothy N. Cason (1999), ‘Do community characteristics inﬂuence environmental outcomes? Evidence from the Toxics Release Inventory,’ Southern Economic Journal, 65, 691–716. Asch, P. and J. Seneca (1978), ‘Some evidence on the distribution of air quality,’ Land Economics, 54, 218–57. Ash, Michael and T. Robert Fetter (2004), ‘Who lives on the wrong side of the environmental tracks? Evidence from the EPA’s Risk-Screening Environmental Indicators Model,’ Social Science Quarterly, 85, 441–62. Atkinson, Anthony B. and Joseph E. Stiglitz (1980), Lectures on Public Economics, New York: McGraw-Hill. Azar, C. and Thomas Sterner (1996), ‘Discounting and distributional considerations in the context of global warming,’ Ecological Economics, 19, 169–84. Banzhaf, H. Spencer and Randall P. Walsh (2004), ‘Testing for environmental gentriﬁcation: migratory responses to changes in environmental quality,’ working paper, Resources for the Future, Washington, DC. Barthold, Thomas (1993), ‘How should we measure distribution?’, National Tax Journal, 46, 291–9. Been, Vicki and Francis Gupta (1997), ‘Coming to the nuisance or going to the barrios: a longitudinal analysis of environmental justice claims,’ Ecological Law Quarterly, 24, 1–56. Bento, Antonio, Lawrence Goulder, Emeric Henry, Mark Jacobsen and Roger von Haefen (forthcoming), ‘Distributional and eﬃciency impacts of gasoline taxes: an econometrically-based multi-market study,’ American Economic Review Papers and Proceedings. Blomquist, Glenn C., Mark C. Berger and John P. Hoehn (1988), ‘New estimates of quality of life in urban areas,’ American Economic Review, 78, 89–107.

The incidence of pollution control policies

37

Borenstein, Severin, James B. Bushnell and Frank A. Wolak (2002), ‘Measuring market ineﬃciencies in California’s restructured wholesale electricity market,’ American Economic Review, 92, 1376–405. Bovenberg, A. Lans and Lawrence H. Goulder (2001), ‘Neutralizing the adverse industry impacts of CO2 abatement policies: what does it cost?’ in C. Carraro and G. Metcalf (eds), Behavioral and Distributional Eﬀects of Environmental Policy, Chicago: University of Chicago Press. Bovenberg, A. Lans, Lawrence H. Goulder and Derek J. Gurney (forthcoming), ‘Eﬃciency costs of meeting industry-distributional constraints under environmental permits and taxes,’ RAND Journal of Economics. Bowen, William (2002), ‘An analytical review of environmental justice research: what do we really know?’, Environmental Management, 29, 3–15. Bowland, Bradley J. and John C. Beghin (2001), ‘Robust estimates of value of a statistical life for developing economies,’ Journal of Policy Modeling, 23, 385–96. Boyle, Kevin J., P. Joan Poor and Laura O. Taylor (1999), ‘Estimating the demand for protecting freshwater lakes from eutrophication,’ American Journal of Agricultural Economics, 81, 1118–22. Brajer, Victor and Jane V. Hall (1992), ‘Recent evidence on the distribution of air pollution eﬀects,’ Contemporary Economic Policy, 10, 63–71. Brännlund, Runar and Jonas Nordström (2004), ‘Carbon tax simulations using a household demand model,’ European Economic Review, 48, 211–33. Brennan, Timothy J., Karen L. Palmer and Salvador A. Martinez (2002), Alternating Currents: Electricity Markets and Public Policy, Resources for the Future, Washington, DC. Brooks, Nancy and Rejiv Sethi (1997), ‘The distribution of pollution: community characteristics and exposure to air toxics,’ Journal of Environmental Economics and Management, 32, 233–50. Bull, Nicholas, Kevin Hassett and Gilbert Metcalf (1994), ‘Who pays broad-based energy taxes? computing lifetime and regional incidence,’ The Energy Journal, 15, 145–64. Burtraw, Dallas, Karen L. Palmer, Ranjit Bharvirkar and Anthony Paul (2001), ‘The eﬀect of allowance allocation on the cost of carbon emissions trading,’ discussion paper 01-30, Resources for the Future, Washington, DC. Carraro, Carlo (ed.) (2000), Eﬃciency and Equity of Climate Change Policy, Boston: Kluwer Academic Publishers. Casler, Stephen D. and Aisha Raﬁqui (1993), ‘Evaluating fuel tax equity: direct and indirect distributional eﬀects,’ National Tax Journal, 46, 197–205. Casperson, Erik and Gilbert Metcalf (1994), ‘Is a value added tax regressive? Annual versus lifetime incidence measures,’ National Tax Journal, XLVII, 731–47. Center for Disease Control (CDC) (2001), ‘Self-reported asthma prevalence among adults – United States, 2000,’ Center for Disease Control Morbidity and Mortality Weekly Report, 50, 682–6. Chay, Kenneth and Michael Greenstone (2005), ‘Does air quality matter? Evidence from the housing market,’ Journal of Political Economy, 113, 376–424. Collins, Robert (1977), ‘The distributive eﬀects of public law 92–500,’ Journal of Environmental Economics and Management, 4, 344–54, Cornwell, Antonia and John Creedy (1997), ‘Measuring the welfare eﬀects of tax changes using the LES: an application to a carbon tax,’ Empirical Economics, 22, 589–613.

38

Yearbook of environmental and resource economics

Cremer, Helmuth, Firouz Gahvari and Norbert Ladoux (1998), ‘Externalities and optimal taxation,’ Journal of Public Economics, 70, 343–64. Cremer, Helmuth, Firouz Gahvari and Norbert Ladoux (2003), ‘Environmental taxes with heterogeneous consumers: an application to energy consumption in France,’ Journal of Public Economics, 87, 2791–815. Cropper, Maureen L and Wallace E. Oates (1992), ‘Environmental economics: a survey,’ Journal of Economic Literature, XXX, 675–740. Dickie, Mark and Bryan Hubbell (2004), ‘Family resource allocation and the distribution of health beneﬁts of air pollution control,’ working paper, US Environmental Protection Agency, Washington, DC. Dinan, Terry M. and Diane Lim Rogers (2002), ‘Distributional eﬀects of carbon allowance trading: how government decisions determine winners and losers,’ National Tax Journal, LV, 199–222. Dorfman, Robert (1977), ‘Incidence of the beneﬁts and costs of environmental programs,’ American Economic Review, 67, 333–40. Dorfman, Nancy S. and Arthur Snow (1975), ‘Who will pay for pollution control? The distribution by income of the burden of the national environmental protection program, 1972–1980,’ National Tax Journal, XXVII, 101–15. Ebert, Udo (2003), ‘Environmental goods and the distribution of income,’ Environmental and Resource Economics, 25, 435–59. Environmental Protection Agency (EPA) (1996), National Air Quality and Emissions Trends Report 1995, Washington, DC: EPA. Eskeland, Gunnar (2000), ‘Environmental protection and optimal taxation,’ discussion paper, World Bank, Washington, DC. Fankhauser, Samuel, Richard Tol and David Pearce (1997), ‘The aggregation of climate change damages: a welfare theoretic approach,’ Environmental and Resource Economics, 10, 249–66. Feitelson, Eran J., Robert E. Hurd and Richard R. Mudge (1996), ‘The impact of airport noise on willingness to pay for residences,’ Transportation Research: Part D: Transport Environment, 1, 1–14. Freeman, A. Myrick (1972), ‘The distribution of environmental quality,’ in Alan Kneese and Blair Bower (eds), Environmental Quality Analysis, Johns Hopkins University Press, Baltimore, 243–80. Fullerton, Don and Diane Lim Rogers (1993), Who Bears the Lifetime Tax Burden? Washington, DC: Brookings Institution. Fullerton, Don and Garth Heutel (2004), ‘The general equilibrium incidence of environmental taxes,’ working paper, Department of Economics, University of Texas. Fullerton, Don and Garth Heutel (2005), ‘The general equilibrium incidence of environmental mandates,’ working paper, Department of Economics, University of Texas. Fullerton, Don and Gilbert E. Metcalf (2002), ‘Tax incidence,’ in A.J. Auerbach and M. Feldstein (eds), Handbook of Public Economics, 4, Elsevier, New York. Gale, William, Scott Houser and J. Karl Scholz (1996), ‘Distributional eﬀects of fundamental tax reform,’ in H. Aaron and W. Gale (eds), Economic Eﬀects of Fundamental Tax Reform, Washington, DC: Brookings Institution. Gianessi, Leonard P., Henry M. Peskin and Edward N. Wolﬀ (1979), ‘The distributional eﬀects of uniform air pollution policy in the United States,’ Quarterly Journal of Economics, 93, 281–301. Goulder, Lawrence H., Ian W.H. Parry, Dallas Burtraw and Roberton C. Williams

The incidence of pollution control policies

39

(1999), ‘The cost-eﬀectiveness of alternative instruments for environmental protection in a second-best setting,’ Journal of Public Economics, 72, 329–60. Gray, Wayne B. and Ronald J. Shadbegian (2004), ‘Optimal pollution abatement – whose beneﬁts matter and how much?’, Journal of Environmental Economics and Management, 47, 510–34. Gruber, Jon and Emmanuel Saez (2002), ‘The elasticity of taxable income: evidence and implications,’ Journal of Public Economics, 84, 1–32. Hahn, Robert W. (2005), In Defense of the Economic Analysis of Regulation, Washington, DC: AEI–Brookings Joint Center for Regulatory Studies. Hamilton, James T. (1993), ‘Politics and social costs: estimating the impact of collective action on hazardous waste facilities,’ RAND Journal of Economics, 24, 101–25. Hamilton, James T. (1995), ‘Testing for environmental racism: prejudice, proﬁts, political power?’, Journal of Policy Analysis and Management, 14, 107–32. Hamilton, James T. (2005), ‘Environmental equity and the siting of hazardous waste facilities in OECD countries: evidence and policies,’ in T. Tietenberg and H. Folmer (eds), International Yearbook of Environmental and Resource Economics 2005/2006, Cheltenham, UK and Northampton, MA: Edward Elgar. Hamilton, James T. and W. Kip Viscusi (1999), Calculating Risks? The Spatial and Political Dimensions of Hazardous Waste Policy, Cambridge, MA: MIT Press. Harberger, Arnold C. (1976), ‘On the use of distributional weights in social cost–beneﬁt analysis,’ Journal of Political Economy, 86, S87–S120. Harrison, David (1975), Who Pays for Clean Air? The Costs and Beneﬁt Distribution of Automobile Emissions Standards, Cambridge, MA: Ballinger. Harrison, David, Jr and Daniel L. Rubinfeld (1978), ‘The distribution of beneﬁts from improvements in urban air quality,’ Journal of Environmental Economics and Management, 5, 313–32. Harrison, Kathryn and Werner Antweiler (2003), ‘Incentives for pollution abatement: regulation, regulatory threats and nongovernmental pressures,’ Journal of Policy Analysis and Management, 22, 361–82. Heinzerling, Lisa and Frank Ackerman (2002), Pricing the Priceless: Cost Beneﬁt Analysis of Environmental Protection, Washington, DC: Georgetown Environmental Law and Policy Institute, Georgetown University. Hokby Stina and Tore Soderqvist (2003), ‘Elasticities of demand and willingness to pay for environmental services in Sweden,’ Environmental and Resource Economics, 26, 361–83. Israel, Debra and Arik Levinson (2004), ‘Willingness to pay for environmental quality: testable empirical implications of the growth and environment literature,’ Contributions to Economic Analysis and Policy, 3(1), article 2. Kahn, Matthew E. (2001), ‘Beneﬁciaries of Clean Air Act legislation,’ Regulation, 24, 34–9. Kaplow, Louis (1996), ‘The optimal supply of public goods and the distortionary cost of taxation,’ National Tax Journal, 49, 513–33. Kaplow, Louis (2004), ‘On the (ir)relevance of distribution and labor supply distortion to government policy,’ Journal of Economic Perspectives, 18,159–75. Kinnaman, Thomas C. and Donald Fullerton (2000), ‘The economics of residential solid waste management,’ in Tom Tietenberg and Henk Folmer (eds), International Yearbook of Environmental and Resource Economics 2000/2001, Cheltenham, UK and Northampton, MA: Edward Elgar.

40

Yearbook of environmental and resource economics

Kriström, Bengt and Pere Riera (1996), ‘Is the income elasticity of environmental improvements less than one?’, Environmental and Resource Economics, 7, 45–55. Lake, Elizabeth, William M. Hanneman and Sharon Oster (1979), Who Pays for Clean Water? The Distribution of Water Pollution Control Costs, Boulder, CO: Westview Press. Liu, Jin Tan, James K. Hammitt and Jin Long Liu (1997), ‘Estimating hedonic wage function and value of life in a developing country,’ Economics Letters, 57, 353–58. Mayeres, Inge (2001), ‘Equity and transport policy reform,’ discussion paper, Center for Economic Studies, K.U. Leuven, Belgium. Mayeres, Inge and Stef Proost (1997), ‘Optimal tax and public investment rules for congestion type of externalities,’ Scandinavian Journal of Economics, 99, 261–79. Mayeres, Inge and Stef Proost (2001), ‘Marginal tax reform, externalities, and income distribution,’ Journal of Public Economics, 79, 343–63. Metcalf, Gilbert E. (1994), ‘The lifetime incidence of state and local taxes: measuring changes during the 1980s,’ in J. Slemrod (ed.), Tax Progressivity and Income Inequality, Cambridge, MA: Cambridge University Press. Metcalf, Gilbert E. (1999), ‘A distributional analysis of green tax reforms,’ National Tax Journal, 52, 665–81. Miller, Ted R. (2000), ‘Variations between countries in values of statistical life,’ Journal of Transport Economics and Policy, 34, 169–88. Morgenstern, Richard D. (1997), Economic Analyses at EPA: Assessing Regulatory Impact, Washington, DC: Resources for the Future. Mrozek, Janusz R. and Laura O. Taylor (2002), ‘What determines the value of life? A meta-analysis,’ Journal of Policy Analysis and Management, 21, 253–70. Ostro, Bart D. (1981), ‘The distributive eﬀects of Public Law 92-500,’ Journal of Environmental Economics and Management, 8,196–8. Parry, Ian W.H. (2004), ‘Are emissions permits regressive?’, Journal of Environmental Economics and Management, 47, 364–87. Parry, Ian W.H. (forthcoming), ‘Fiscal interactions and the costs of pollution control from electricity,’ RAND Journal of Economics. Pearce, David (2003), ‘Conceptual framework for analyzing the distributive impacts of environmental policies,’ prepared for the OECD Environment Directorate Workshop on the Distribution of Beneﬁts and Costs of Environmental Policies, Paris. Pirttilä, Jukka and Matti Tuomala (1997), ‘Income tax, commodity tax and environmental policy,’ International Tax and Public Finance, 4, 379–93. Portney, Paul R. and John P. Weyant (eds) (1999), Discounting and Intergenerational Equity, Washington, DC: Resources for the Future. Poterba, James M. (1989), ‘Lifetime incidence and the distributional burden of excise taxes,’ American Economic Review, 79, 325–30. Poterba, James M. (1991), ‘Is the gasoline tax regressive?’ in David Bradford (ed.), Tax Policy and the Economy 5, Cambridge, MA: National Bureau of Economic Research. Ringquist, Evan J. (2005), ‘Assessing evidence of environmental inequities: a metaanalysis,’ Journal of Policy Analysis and Management, 24, 223–47. Roback, Jennifer (1982), ‘Wages, rents and the quality of life,’ Journal of Political Economy, 90, 1257–78. Robison, H. David (1985), ‘Who pays for industrial pollution abatement?’, Review of Economics and Statistics, 67, 702–6.

The incidence of pollution control policies

41

Rogers, Diane Lim (1993), ‘Measuring the distributional eﬀects of corrective taxation,’ paper presented at a National Tax Association session of the Allied Social Science Association Meetings, Boston, 3–5 January. Rose, Adam and Gbadebo Oladosu (2002), ‘Greenhouse gas reduction policy in the United States: identifying winners and losers in an expanded permit trading system,’ The Energy Journal, 23, 1–18. Shadbegian, Ronald J., Wayne Gray and Cynthia Morgan (2005), ‘Beneﬁts and costs from sulphur dioxide trading: a distributional analysis,’ mimeo, University of Massachusetts, Dartmouth. Shapiro, Matthew and Joel Slemrod (1994), ‘Consumers’ response to the timing of income: evidence from a change in tax withholding,’ American Economic Review, 85, 274–83. Sieg, Holger, V. Kerry Smith, H. Spencer Banzhaf and Randy Walsh (2004), ‘Estimating the general equilibrium beneﬁts of large changes in spatially delineated public goods,’ International Economic Review, 45, 1047–77. Slesnick, Daniel T. (1994), ‘Consumption, needs, and inequality,’ International Economic Review, 35, 677–703. Smith, V. Kerry and Ju-Chin Huang (1995), ‘Can markets value air quality? Metaanalysis of hedonic property value models,’ Journal of Political Economy, 103, 209–27. Souleles, Nicholas S. (1999), ‘The response of household consumption to income tax refunds,’ American Economic Review, 89, 947–58. Souleles, Nicholas S. (2002), ‘Consumer response to the Reagan tax cuts,’ Journal of Public Economics, 85, 99–120. Stevens, Brandt and Adam Rose (2002), ‘A dynamic analysis of the marketable permits approach to global warming policy: a comparison of spatial and temporal ﬂexibility,’ Journal of Environmental Economics and Management, 44, 45–69. Stoft, Steven (1993), ‘Appliance standards and the welfare of poor families,’ The Energy Journal, 14, 123–8. Sutherland, Ronald J. (1991), ‘Market barriers to energy-eﬃciency investments,’ The Energy Journal, 12, 15–34. Sutherland, Ronald J. (1994), ‘Income distribution eﬀects of electric utility DSM programs,’ The Energy Journal, 15, 103–18. US Census Bureau (2004), Statistical Abstract of the United States: 2004–2005, Washington, DC. Viscusi, W. Kip and Joseph E. Aldy (2003), ‘The value of a statistical life: a critical review of market estimates throughout the world,’ Journal of Risk and Uncertainty, 27, 5–76. Walls, Margaret and Jean Hanson (1999), ‘Distributional aspects of an environmental tax shift: the case of motor vehicles emissions taxes,’ National Tax Journal, 52, 53–65. West, Sarah E. (2004), ‘Distributional eﬀects of alternative vehicle pollution control policies,’ Journal of Public Economics, 88, 735–57. West, Sarah E. (2005), ‘Equity implications of vehicle emissions taxes,’ Journal of Transport Economics and Policy, 39, 1–24. West, Sarah E. and Roberton C. Williams III (2004), ‘Estimates from a consumer demand system: implications for the incidence of environmental taxes,’ Journal of Environmental Economics and Management, 47, 535–58. Wier, Mette, Katja Birr-Pedersen, Henrik Klinge Jacobsen and Jacob Klok (2005),

42

Yearbook of environmental and resource economics

‘Are CO2 taxes regressive? Evidence from the Danish experience,’ Ecological Economics, 52, 239–51. Wiese, Arthur M., Adam Rose and Gerald Schluter (1995), ‘Motor-fuel taxes and household welfare: an applied general equilibrium analysis,’ Land Economics, 71, 229–43. Williams, Roberton C., III (2004), ‘An estimate of the second-best optimal gasoline tax, considering both eﬃciency and equity,’ working paper, University of Texas at Austin. Williams, Roberton C., III (2005), ‘Optimal commodity taxes, public good provision and pollution taxes: reconciling results from representative-agent and multiple-agent models,’ working paper, University of Texas at Austin. Willig, Robert D. (1976), ‘Consumer’s surplus without apology,’ American Economic Review, 66, 589–97. Wolverton, Anne (2002), ‘Does race matter? An examination of a polluting plant’s location decision,’ mimeo, US Environmental Protection Agency, Washington, DC. Yandle, T. and D. Burton (1996), ‘Re-examining environmental justice: a statistical analysis of historical hazardous waste landll siting patterns in metropolitan Texas,’ Social Science Quarterly, 77, 477–92. Zeldes, Stephen (1989), ‘Consumption and liquidity constraints: an empirical investigation,’ Journal of Political Economy, 97, 305–46. Zupan, J. (1973), The Distribution of Air Quality in the New York Region, Baltimore, MD: Johns Hopkins University Press.

2. Geographical information systems (GIS) and spatial analysis in resource and environmental economics Ian Bateman, Wanhong Yang and Peter Boxall 1.

INTRODUCTION

Most, if not all, of the problems addressed by resource and environmental economists have a spatial dimension by nature. Assets such as natural resources and disamenities such as pollution emissions can be located in space. That location deﬁnes a myriad of proximities with other resources, the complexity of which poses a substantial empirical challenge to economic analyses of resource and pollution management decisions. Meeting this challenge constitutes one of the major tasks which economists must address if they are to provide meaningful inputs to the decision process. Recent years have witnessed what we suggest are the ﬁrst steps in a revolution in the way in which economists incorporate spatial complexity within their analyses. This revolution is fuelled by the importation from other disciplines (notably applied geography) of both innovative techniques of spatial analysis and new tools for facilitating such analyses, in particular geographical information systems (GIS). While still in its early days, the use of GIS within environmental and resource economics is growing rapidly. Examples of applications already range from simple data preparation and mapping tasks to more complex data integration and manipulation exercises. These have aﬀorded demonstrable and signiﬁcant improvements in analytical rigour (for overviews of examples see Bateman et al. 2002a; 2003). However, as economists seek to address more complex problems, so the advances in data availability and interrogation aﬀorded by GIS increasingly need to be supplemented by more sophisticated spatial analysis methods. This chapter has dual but highly interrelated objectives. First, in the following section, we set out to provide an overview of the functionality of a GIS, illustrating this with reference to existing and where appropriate likely 43

44

Yearbook of environmental and resource economics

future applications to environmental and resource economists. Second, in Section 3, we review advances in selected areas of spatial analysis considered to be of particular interest to this audience. Due to the limitations of space both reviews are somewhat introductory, as either could readily ﬁll an entire volume. However, throughout we seek to indicate appropriate further readings. Section 4 provides a few pointers to future directions and concludes.

2.

AN OVERVIEW OF GIS FEATURES

Chang (2002) deﬁnes a GIS as a computer system for capturing, storing, querying, analysing, and displaying geographically referenced data’. While the design and functionalities of major GIS software packages such as ArcView, Arc/Info, or MapInfo vary somewhat, they are all grounded in a common set of knowledge regarding spatial data representation, analysis and modelling. This section introduces some features of GIS that could be useful in resource and environmental economics. Discussion of features is supported by case study overviews where appropriate. The section has been grouped under two broad headings: the ﬁrst concerns the acquisition, processing and visualization of spatial data; the second discusses the spatial analysis of physical and socioeconomic information. The theme of spatial analysis is expanded upon in Section 3. 2.1

The Acquisition, Processing and Visualization of Spatial Data

Traditionally resource and environmental economists use data organized in spreadsheet formats or charts and if the data are spatial, then an identiﬁer such as location coordinates or name of a county or other spatial unit is included as a column in the table. Using GIS to examine these data, the spatial location information, such as county name in a national dataset, can be linked to the same spatial identiﬁer in another spatial layer such as county boundaries to create a new GIS data layer.1 This new layer can then be used to display data visually, explore spatial patterns, generate new data variables, and possibly support economic modelling under a spatial framework. Figure 2.1 illustrates results from such a simple process where the counties of Southern Ontario are associated with average household income statistics from the most recent Canadian census. The boundary layers of the counties (one GIS layer) have been merged with the income data using the county name as the spatial identiﬁer. This process shows the utility of GIS in examining income data on a spatial dimension.

45

Geographical information systems and spatial analysis 82ºW

80ºW

78ºW

80ºW

46ºN

46ºN

Georgian Bay Lake Huron

44ºN

44ºN

Lake Ontario

Lake Erie 42ºN

100 km 82ºW

Average household income 2001($) 20 000 – 27 500 (1st quartile) 27 500 – 30 000 (2nd quartile) 30 000 – 32 500 (3rd quartile) 32 500 – 47 500 (4th quartile)

Figure 2.1 Distribution of household income levels in the counties of the Canadian province of Ontario Data acquisition Researchers typically have three options for acquiring spatial data for GISbased analysis. The ﬁrst is to obtain data for free from various public or private sources. The most common source of such data is the Internet. We provide some examples of websites in Table 2.1, where data are freely available. For example, the US National Spatial Data Infrastructure Clearinghouse Network Website provides links to GIS data sources across the world. Most of the US data at this site are free. A second option is to purchase data from private vendors. These vendors provide both data that are not available from free sources and GIS products customized to a purchaser’s requirements. For example, the Geolytics website in Table 2.1 provides a portal for obtaining customized US census data. A third, increasingly popular option is to explicitly generate GIS data. For example, researchers interested in toxic sources in a community can use a global positioning system (GPS) device to receive satellite signals from which the spatial location of any ﬁeld data record can be ﬁxed. Using this approach, survey data can readily be converted into GIS data layers (Williams et al.

46

http://geogratis.cgdi.gc.ca/

http://data.geocomm.com/

http://www.geoplace.com/

http://www.geographynetwork.com/

http://www.censuscd.com/

http://www.teleatlas.com/

http://www.activesol.co.uk/

GeoGratis

The GIS Data Depot

GEOWorld Data Directory

Geography Network

Geolytics

TeleAtlas

Active

Adapted from Longley et al. (2001).

http://www.fgdc.gov/

National Spatial Data Infrastructure Clearing house Network

Source:

http://www.data-store.co.uk/

The Data Store

Table 2.1 Selected websites containing information about geographic data sources

Provide user-friendly geographic and demographic information and systems, private data vendor

Provide location-based GIS data services across America, private data vendor

Customize US census data for easy access and use, private data vendor

Global online data and map services, private data vendor

List of GIS data companies

Free GIS and geospatial data depot

Distribute geospatial data of Canada; most of the data are free

Worldwide list of data sources. Most of the US data are free

UK, Europe and worldwide data catalogue, a mix of free and private data

Geographical information systems and spatial analysis

47

1996; Williams 1997). For example, Lovett et al. (1998) generate new data to which they apply GIS modelling techniques to link incinerator emission plumes to soil PCB (polychlorinated biphenyl) and dioxin toxicity levels. Environmental economists are also increasingly generating spatial data from survey information. For example, Bateman et al. (1996; 2000) use survey respondents’ stated postal addresses to spatially identify recreational visitor outset locations. As discussed subsequently, this information can then be used to generate travel distances, costs and times for valuation studies or to add spatial decay variables to estimated valuation functions. Data acquisition can also be done through the conversion of information from paper maps into digital data layers through a process called digitizing. The resulting digital information is then converted into a GIS data layer. This method of acquiring GIS data has been used in agricultural economic studies. For example, in many studies of soil erosion it is common to combine soil characteristics data with other spatial data such as topographic information. In some study areas only paper maps of these data are available. Thus, digitizing the relevant map information and relating the various layers using the GIS is an eﬃcient way to develop models of soil loss within a reasonable time frame (Millward and Mersey 1999). However, copyright issues can be an impediment to widespread digitizing. Resource and environmental economists use a wide range of environmental (physical or ecological) and socioeconomic data. If these data are in GIS format, then the data will either be in vector or raster formats (Figure 2.2). Vector data use points, lines, or polygons to represent spatial features and topologies. Raster data use standardized grid cells called pixels to represent spatial features and topologies (Longley et al. 2001). Vector data are mostly created from digitized maps, while raster data are frequently acquired from satellite images. In most cases the data for resource and environmental economic research may be from vector data sources such as census data and soil data. However, in recent years raster data have become popular due to the increasing availability of satellite imagery and the improvements in imagery processing technology. For example, satellite data such as land use images have been used in economic research concerning changes in forest cover (Mertens and Lambin 1997) and the resultant economics costs and beneﬁts arising from changes in carbon sequestration (Brainard et al. 2003). In GIS data acquisition it is essential to obtain associated metadata (literally data about data). The metadata have basic information about GIS datasets such as the title, geographical coverage, accuracy, scale, map projection information and provenance (Chang 2002). This information is required for evaluating the suitability of datasets for speciﬁc research tasks and in subsequent GIS processing of that data, a task to which we now turn.

48

5 5 5 5 5 5 5 5 3 3 5 5 5 5 5 5 5 5 5 5 5 1 5 2 2 3 5 5 5 5 5 5 5 5 5 5 1 5 2 2 3 5 5 5 5 5 5 5 5 5 5 1 5 2 2 5 3 5 5 5

5 5 5 5 5 1 5 5 2 2 5 3 5 5 5 5 5 5 5 5 1 5 5 2 2 5 3 5 5 5

Lake

5 5 5 5 5 5 1 5 2 2 5 3 5 5 5

Cropland

5 5 5 5 5 5 5 5 5 2 3 5 5 5 5

Representing real-world features as vector or raster data

Adapted from Lo and Yeung (2002).

Figure 2.2

Source:

Real world

5 5 5 5 5 5 5 3 5 5 5 5 5 5 5 5 5 5 5 5 5 5 1 2 2 5 3 5 5 5

5 5 5 5 1 1 1 1 5 2 5 3 5 5 5 5 5 5 1 5 5 5 5 1 2 3 5 5 5 5

1 5 5 1 5 5 5 5 5 5 3 5 5 5 5 1 1 1 5 5 5 5 5 5 5 3 5 5 5 5 5 5 5 5 5 5 5 5 5 5 3 5 5 5 5

5 5 5 5 5 5 5 5 5 5 3 5 5 5 5 4 4 5 5 5 5 5 5 5 5 3 5 5 5 5

Highway

River

4 4 4 5 5 5 5 5 5 5 3 5 5 5 5 4 4 4 4 5 5 5 5 5 3 5 5 5 5 5

4 4 4 4 4 5 5 5 5 3 5 5 5 5 5 4 4 4 4 4 5 5 5 3 5 5 5 5 5 5

4 4 4 4 4 4 5 5 3 5 5 5 5 5 5 4 4 4 4 4 4 5 3 5 5 5 5 5 5 5

Forest

4 4 4 4 4 5 5 5 5 3 5 5 5 5 5

Cropland

Cropland

Vector data

5 5 5 5 1 5 1 5 2 2 5 3 5 5 5

Raster data 5 5 5 5 1 5 5 1 2 2 5 3 5 5 5 4 4 4 4 4 4 5 3 5 5 5 5 5 5 5 4 4 4 4 4 4 3 5 5 5 5 5 5 5 5 4 4 4 4 4 4 3 5 5 5 5 5 5 5 5

4 4 4 4 4 5 3 5 5 5 5 5 5 5 5

1. River 2. Lake 3. Highway 4. Forest 5. Cropland

Geographical information systems and spatial analysis

49

Data processing Data processing refers to the assembly of acquired GIS data into a uniﬁed database using a common software format, data model and map projection. Typically the data from various external sources are encoded in many diﬀerent formats that are software-speciﬁc such as Arc/Info interchange ﬁles, ArcView shape ﬁles, MapInfo interchange formats, and ERDAS IMAGINE ﬁle formats (Longley et al. 2001). Most GIS packages are able to directly read data ﬁles from other software or interchange ﬁle formats. For further analysis it is important to convert data into the format speciﬁc to the software that the researchers are using. As discussed earlier, the acquired data could also be either in vector or raster formats. For some analyses where there are layers of data in diﬀerent formats, the layers must be converted into the same format or data model (DeMers 2003). For example, a land slope grid in a raster format could be converted into a vector data layer before it is overlaid with a vector soil layer.2 Alternatively, the vector soil layer could be converted into a grid layer and overlaid with the slope raster layer. The desired conversion routine will depend on the choice of ﬁnal data model (either vector or raster) to be used by the researcher. In GIS data processing researchers also need to handle map projections of the acquired data. Map projection refers to the process of systematically transforming positions on the earth’s spherical surface to a ﬂat map while maintaining spatial relationships (Robinson et al. 1995). Data in diﬀerent projections cannot be combined for analysis and must be converted to a common projection. Most GIS packages have the capability to perform such conversions. However, standard or customized map projection parameters need to be identiﬁed or provided to complete the re-projection. In most cases these parameters are listed in the metadata or are obtained by contacting the data providers. Simple data visualization: mapping The simplest and most common reason for using GIS to view spatial information is to create maps. Such outputs can be considerably eﬀective in conveying information both across interdisciplinary frontiers and into the policy arena. For example, well before its general publication (Brouwer and Kind 2005), the map shown in Figure 2.3 illustrating downstream assets at ﬂood risk in the Netherlands was adopted by the Dutch water ministry as a prima facie case for investing more than 30 million euros in upstream ﬂood diversion schemes as a highly eﬃcient alternative to costly downstream managed realignment of rivers (Ministerie van Verkeer en Waterstaat 2000). However, many resource management questions are complex and require an understanding of multidimensional

50

Yearbook of environmental and resource economics Area below sea level protected from flooding by dike enclosures

12.84

Area above sea level without dike enclosures

10.67 0.80

69.70 5.72 6.35 15.71 78.56

7.67

27.16

21.06 0.03

Source:

Brouwer and Kind (2005).

Figure 2.3 GDP generated in ﬂood-prone areas of the Netherlands (billions of euros, 2000) spatial patterns, in turn requiring the acquisition of multiple spatial data layers, an issue to which we now turn. Most GIS packages have easy-to-use icons or menus to guide the process of constructing maps. Colour schemes or symbols in GIS can usually easily be changed to reveal spatial patterns of points, lines or polygon features. Spatial data can also be easily classiﬁed or re-classiﬁed to examine spatial relationships between diﬀerent layers. However, some basic knowledge of cartography is useful. For example, following Chang (2002), a thematic map should have the following components: a title, legend, orientation, a scale, and a clear deﬁnition of map source. Furthermore, the colour scheme and type of symbols used in maps need to be carefully designed to produce a balanced result. A simple rule of thumb is that no more than six colours are readily distinguishable within a given map although this may be somewhat

Geographical information systems and spatial analysis

51

extended through judicious use of shading schemes or, in the case of blackand-white maps, symbols. Poorly designed maps have the potential to convey false and even deliberately misleading impressions, a fact entertainingly and thoughtfully demonstrated by Monmonier (1991) in his deﬁnitive book How to Lie with Maps. Viewsheds One of the most interesting features of GIS are their ability to calculate the area that can be observed from a given viewpoint, an area known as the ‘viewshed’. Such viewsheds have been used in a variety of contexts, for example to quantify the viewable areas from spatially explicit objects such as a hiking trail, roads and residences. The three-dimension analysis features in some GIS packages can be used to derive people’s cognition of scenic landscape components such as relief, depth of view, horizon characteristics and shape (Bateman 1994; Baldwin et al. 1996). Basic viewshed information is calculated by inputting data concerning natural topographic relief, ecological barriers and man-made obstructions such as barriers and buildings (e.g. Din et al. 2001; Paterson and Boyle 2002). Furthermore, this information can be supplemented with land use data and aerial photography information to build up data concerning the type and quality of views and the level and nature of visual amenities and eyesores (Lake et al. 2000a; 2000b; Bateman et al. 2002a; 2004). For example, Lake et al. report signiﬁcant negative impacts upon property prices from visual disamenity caused by views of factories and roads and positive (if weaker) impacts of views of parkland. Figure 2.4 gives a visual representation of a viewshed as constructed for one point on a given property. Such variables have been primarily used by economists for hedonic pricing studies of non-market, typically environmental goods (Bateman et al. 2004). Weighting functions can be applied to give greater emphasis to objects which are nearby than to those that, although occupying the same viewing angle, are located at a greater distance to the viewing point (Lake et al. 1998; 2000a,b). With modern computing power such calculations can be conducted for multiple faces and ﬂoors of a property and repeated across any number of properties to swiftly generate highly detailed information concerning views. For example, in undertaking the analysis underpinning the hedonic pricing study described by Bateman et al. (2004), viewsheds were calculated for each face of a property at all ﬂoor levels and repeated across a dataset of more than 10 000 properties. Virtual reality GIS (VRGIS) One of the most attractive features of GIS is their ability to interface and integrate with other software advances. One of the most exciting areas of present and potential future development is in the linkage between GIS and

52

Yearbook of environmental and resource economics

Viewpoint at property

Viewshed Source:

Bateman et al. (2002a).

Figure 2.4 GIS viewshed calculation of viewable area from one point on a speciﬁed property virtual reality (VR) software. VR allows the researcher to construct images of hypothetical future landscapes or other environments. This software has already been used by environmental economists within contingent valuation (CV) studies. For example,3 EFTEC and CSERGE (1998) show that VR images can help accurately convey policy change impacts, in this case the eﬀects of abstracting water from rivers. A split-sample design is used with one sample, presented purely with textual and numeric information regarding abstraction impacts, while a second sample was additionally presented with VR-generated images. Results showed that the non-VR sample gave scope-insensitive willingness-to-pay (WTP) estimates with non-zero intercept values (i.e. extrapolation of responses suggested that individuals had some anomalous ‘warm-glow’ value for any scheme). However, the use of VR images removed this problem, with respondents reporting zero WTP values for relatively small levels of abstraction and signiﬁcant and progressively increasing WTP for larger abstractions. Recent years have seen researchers harnessing the power of VR software and combining this with the spatial data manipulation strengths of GIS to create hybrid VRGIS tools. For example, VRGIS has been used to visualize stream ﬂows before and after dam construction and assess the impacts of ﬂood events (Gamboa and Santos 1996), generate visualizations of urban scenes (Martin and Higgs 1997), or provide vivid images of landscape and

Geographical information systems and spatial analysis

53

associated landscape changes (Mitas et al. 1997; Appleton et al. 2002; Appleton and Lovett 2003). In a recently completed (and currently unpublished) study, researchers at the University of East Anglia4 use VRGIS technology to generate a series of ‘ﬂythroughs’ (three-dimensional pathways across VR-generated landscapes) to illustrate the status quo and various alternative landscapes arising from a set of coastal defence options. These were used to convey choice alternatives within a choice modelling experiment. Split-sample experimental techniques were used to contrast the choices of respondents presented with map, numeric and VR ﬂythrough representations with choices made by respondents who are not shown VR images. Results show that, compared to other treatments, respondents exposed to VR ﬂythroughs reported higher degrees of certainty regarding the consequences of policy change, were signiﬁcantly more likely (p 0.01) to select alternatives to the present-day situation (i.e. they were less subject to the types of status quo anomalies discussed by Tversky and Kahneman 1991) and exhibited higher implicit WTP for such options. While interesting questions arise regarding the extent to which such advanced images inﬂuence preferences and values (in essence there are parallels to the inﬂuence upon values exerted by marketing and advertising), such approaches do seem to heighten comprehension of policy impacts and would therefore generally be accepted as oﬀering considerable promise for future developments in decision support. 2.2 GIS-Based Spatial Analysis of Physical and Socioeconomic Information GIS is becoming more powerful in performing spatial analysis of data frequently collected by geographers, biologists and ecologists. In particular, quantifying spatial patterns and relationships among environmental features that can aﬀect socioeconomic variables may be of considerable interest to economists. In this section we outline a number of GIS features that have this potential. Spatial patterns or distributions Many resource and environmental studies concern patterns and distributions across multiple dimensions including geographic, economic, social, cultural and ethnic space. Economists increasingly recognise this diversity by extending analyses beyond the traditional conﬁnes of narrowly deﬁned eﬃciency and into wider realms of equity and cultural objectives. A recent example focusing upon an urban setting is provided by Brainard et al. (2002), who examine air quality patterns within a major UK city and its distribution across socioeconomic and ethnic dimensions. Data on

54

Yearbook of environmental and resource economics

concentrations of two pollutants, carbon monoxide (CO) and nitrogen dioxide (NO2), were combined with information concerning land use and pollution sources within a proprietary model assuming a Gaussian dispersion of pollutants. The resultant output of air pollution measures was interpolated to 20 20 m cells using standard routines within the GIS which also generated maps such as those shown in the upper panel of Figure 2.5. Both maps indicate higher levels of pollutants around major roads and the central business district, although associations with outlying industrial areas can clearly be seen on the NO2 map. The GIS was then also used to extract pollution measures and various census derived socioeconomic and ethnicity indicators for each 2020 m cell. Statistical analyses indicated highly signiﬁcant associations between pollution levels and measure of economic deprivation and ethnicity. The lower panel of Figure 2.5 illustrates an association between levels of CO and cumulative frequency of ﬁve ethnic groupings. As can be seen, the overall distribution of the non-immigrant, native ethnic group (labelled as ‘ 90 per cent white’) enjoys substantially lower levels of CO pollution than does the Afro-Caribbean group. Interestingly, multiple regression modelling showed that, even after controlling for various economic deprivation indices, signiﬁcant associations with some but not all ethnic groupings could still be identiﬁed. This suggests that a sole focus upon eﬃcient redistributions of incomes and other socioeconomic indicators may not lead to evenly distributed improvements in access to environmental quality across ethnic groups. This implies that ‘trickle-down’ philosophies, assuming that gains in eﬃciency lead to improvements in equity, need careful qualiﬁcation and that explicit consideration of the equity implications of cost–beneﬁt analyses may well be justiﬁed. GIS routines have also been frequently applied to the analysis of spatial patterns and distributions within rural settings, and numerous studies have illustrated the importance of these factors. For example, the spatial arrangement of habitat, land cover, or eﬄuent discharge has been shown to have an important, and often dramatic, eﬀect on species diversity, natural assimilative capacity and nutrient cycles (Turner et al. 1993). Similarly, the total quantity or surface area of wetlands in a landscape may not be as important as their spatial pattern in providing environmental and ecological beneﬁts (Bockstael 1996). Furthermore, it may not be the total area of forested land in a region that inﬂuences species abundance and diversity, but the size, shape and degree of conﬂicting land uses found along the edges of the forested areas (Wadsworth and Treweek 1999). Several GIS routines can be applied to quantify some of the spatial patterns of landscape features. These involve the development of measurement metrics that can be used as inﬂuential variables in an economic study.

55

Geographical information systems and spatial analysis

CO emissions (mg/m3) 0.4

Birmingham London

0 2 4 6 km

NO2 emissions (␮g/m3) 5.2

1.6

10.9

2.9

16.6

4.1

22.3

5.4

28.0

6.6

33.7

7.8

39.3

Cumulative percentage of persons

100 90 80 70 60 50 >90% white

40

50–90% white

30

Asian

20

Mixed

10

Afro-Caribbean

0 0.5

1.5

2.5 3.5 CO emissions (mg/m3)

4.5

5.5

Source: Adapted from Brainard et al. (2002).

Figure 2.5 Upper panel: modelled emissions of CO and NO2 in Birmingham, UK. Lower panel: cumulative frequency of persons in ﬁve ethnic groups with respect to exposure to CO pollution The development of these metrics depends on the type of spatial features under examination – speciﬁcally whether they are points, lines or polygons. For spatial objects classiﬁed as points in a data layer, quadrant analysis in GIS can classify the point pattern to assess the degree to which they are clustered. One of the point pattern indices commonly used is the variance–mean ratio (VMR), which is calculated by dividing the variance

56

Yearbook of environmental and resource economics

of the point frequencies by the mean number of points in each sub-area examined. VMR values equal to, greater than or less than 1.0 indicate random, clustered or uniform patterns respectively. DeMers (2003) has described statistical tests of point pattern hypotheses using the VMR values. One potential application of such assessments of clusters is to quantify point pollution source patterns for environmental economic analysis. ‘Nearest neighbour’ analysis, on the other hand, can be used to determine the proximity of points and so identify those that are closest together and those that are signiﬁcantly closer than others (DeMers 2003). Such analyses are helpful to agricultural and resource economists studying the spread of disease between farms where distance to neighbouring farms is one of the important factors in determining the risk exposure (Woolhouse 2003). A useful measure for linear spatial features such as roads is to calculate the density of the linear feature in an area (DeMers 2003). For example, the length of roads per square kilometre (km/km2) was used by Haener et al. (2004) as an index to assess levels of access for big game hunting to forested habitats in various spatial units in Alberta. GIS can also be used to measure the shape of polygon features. Based on the fact that a circle has perfect convexity, a convexity index for a polygon can be computed as the ratio of its perimeter to its area, multiplied by a constant that is determined by the size of the circle that would inscribe the irregular polygon. This index is designed to provide positive values that range from 1 to 100 where the latter value indicates a perfectly circular polygon (DeMers 2003). For example, in a study examining the inﬂuence of wetland amenities on sale prices of residential properties in Portland, Oregon, GIS was used to consider the shape of the nearest wetland as well as its proximity to the property as variables inﬂuencing its price (Mahan et al. 2000). The size and shape of wetlands have also been used as variables in an expert survey model to prioritize salt marsh restoration actions (Johnston et al. 2002). In both cases the shape of a wetland was found to be signiﬁcant in terms of its impacts on property prices or conservation decisions. Analysis of polygon features can also involve the calculation of fragmentation indices. These indices were originally developed and used by ecologists interested in habitat quality for wildlife species. For example, FRAGSTATS is a program developed to be used with GIS to calculate a variety of indices associated with habitat fragmentation on landscapes (McGarigal and Marks 1994). Environmental economists have also used fragmentation measures in hedonic pricing models (Bateman et al. 2002a). For example, Geoghegan et al. (1997) examined measures of percentage of open space, diversity, and fragmentation of land uses around land parcels in the Patuxent Watershed in Maryland, USA. They calculated diversity and fragmentation indices by obtaining 25 land-cover class images from the

Geographical information systems and spatial analysis

57

Maryland Oﬃce of Planning. The diversity index was designed to signify the extent to which the landscape was dominated by a few or many land uses, while the fragmentation index was set to represent the variance in landscape parcel types within an area. Spatial pattern analysis involving combinations of point, line and polygon features could oﬀer even more spatial measures with potential uses for researchers. A point and polygon overlay analysis can be used to calculate how many points occur in various polygons within a study area. A line and polygon analysis can be used to estimate the speciﬁc type of polygon intersected by line features. Polygon overlays can be used to determine which polygon is touched by other polygons, or how much of a polygon in one layer overlaps with polygons in other map layers (DeMers 2003). Spatial and temporal trends GIS can be used to study spatial land features and their changes over time (Pan et al. 1999; Mertens and Lambin 1997; Irwin 2000). For example, forest cover in a region may change because of ﬁres, logging activities, and re-planting. GIS can overlay land-use layers for diﬀerent years and calculate the spatial changes. Here raster forest layers can be classiﬁed as binary, where 1 and 0 indicate forest and non-forest respectively. Contrasting forest layers across years provides a spatial and temporal trend highlighting areas of aﬀorestation, deforestation and no change (Heywood et al. 2002). A more sophisticated spatial trend analysis is to impose a mathematical functional form to derive a spatial surface. The estimated surface will represent the spatial trend of the land features. For example, site-speciﬁc crop yields may have very detailed yield points in a ﬁeld yet the yield pattern is inﬂuenced by global factors such as temperature and precipitation and local factors such as soil fertility and moisture conditions. In order to separate the inﬂuence of these factors it may be advantageous to convert the yield pattern into a global and a local trend. In this case a quadratic or a cubic functional form could be used in GIS to ﬁt the yield pattern. The functional form will represent the spatial trend of the yield pattern at global scale and the residuals will be associated with local eﬀects (Dessaint and Caussanei 1994). More complex, multivariate functions can also be applied to spatially distributed predictors via GIS. Bateman and Lovett (1998) estimate models relating tree growth and consequent timber yield to a variety of spatially distributed variables including elevation, rainfall, soil type, etc. These variables include some which were custom-generated by GIS for this analysis. For example, a digital elevation model (DEM) was interrogated to provide estimates of land slope. Then, by examining the variation in slopes, measures of the topographical shelter (i.e. the natural shelter provided by local

58

Yearbook of environmental and resource economics

relief) conditions were obtained. A considerable variety of such variables were found to be signiﬁcant predictors of tree growth and timber yield, and the resultant model was then used to estimate spatial trends in yield across a large study area (the entire country of Wales). Subsequent research extended this approach to provide estimates of carbon sequestration across the study area (Bateman and Lovett 2000). A similar methodology was also applied to estimate GIS-based, multi-sector models of agricultural output and consequent market and shadow values for the same case study area (Bateman et al. 1999a). This and beneﬁt transfer estimates for recreation values (discussed later) were subsequently synthesized within a GIS-based cost–beneﬁt analysis of optimal locations for land use change as discussed by Bateman et al. (2003). A relatively overlooked but arguably important spatial trend is the impact of distance upon WTP values for spatially deﬁned natural resources. In the UK concern regarding this issue was focused by an assessment for the Environment Agency (EA) of water company schemes to abstract water from rivers. These schemes were leading to low ﬂows and consequent loss of non-market values. Assessments such as that conducted for the River Kennet near London calculated total WTP values by assuming that all households within a given distance (d ) of the river had the same value for avoiding abstraction losses (that value being set at the mean WTP estimated from a CV study), while households beyond that distance had zero WTP. The upper panel in Figure 2.6 illustrates the eﬀects of this assumption; spatial trends in population are ignored and aggregate WTP is given by the area of the rectangle shown. The use of GIS allows us to import census data to relax the assumption of constant population distribution and to recognize the diﬀering socioeconomic characteristics of that population (shown in the left-hand graph of the lower panel of Figure 2.6). We can then apply a model of marginal WTP, recognizing its positive correlation with income and the decaying eﬀect of distance upon such values (right-hand graph). To illustrate the impact of using GIS to incorporating distance decay and income eﬀects within WTP aggregation, Bateman et al. (2000; 2002) apply this method to data from a CV survey of a wetland preservation scheme and contrast this result with total value estimates obtained via the simpler EA approach.5 A further sensitivity analysis was undertaken to examine the robustness of both aggregation approaches to the truncation of the upper and lower 2.5 per cent of WTP responses so as to allow for possible outlier or strategic bidding eﬀects. Table 2.2 presents results from this comparison. As can be seen, the EA approach of assuming constant WTP up to some set distance has in this case produced total WTP estimates which are more than six times larger than under the GIS based approach (in fact the former values can be manipulated up or down just by varying the

59

Geographical information systems and spatial analysis

WTP mean WTP

d Population: millions of households

Distance from asset

WTP total popn low income high income

Distance from asset

high income low income

Distance from asset

Source: Adapted from Bateman et al. (2000, 2002b).

Figure 2.6 Upper panel: conventional approach to aggregating WTP data. Lower panel: GIS-based approach to aggregation allowing for spatial variation in population and income distribution (lefthand graph) and distance decay and income inﬂuences upon marginal WTP (right-hand graph) assumptions regarding the distance across which WTP is constant). Furthermore, these values are highly sensitive to truncation. By contrast, the GIS-based approach uses data-driven socioeconomic and spatial trends within the WTP bid function. This yields values which are sensitive to the distribution of population and income and allow marginal WTP to decay with increasing distance. The resultant values are not only much lower but also are robust to truncations. Distance functions GIS functions provide convenient ways for calculating distances between diﬀerent spatial objects and interpolating data between a variety of objects

60

Yearbook of environmental and resource economics

Table 2.2 The present non-user’s beneﬁts of preserving the present condition of Broadland aggregated across the UK using various procedures (£ million/annum) Aggregation approach Aggregation assuming all households have constant WTP to a given distance and zero thereafter (EA approach) GIS-based aggregation using WTP bid functions (sensitive to population and income distribution and distance decay in WTP)

Untruncated

Truncated

159.7

98.4

25.4

24.0

Source: Adapted from Bateman et al. (2000; 2002b).

including points, lines and polygon features. For example, the distance from a toxic source to vulnerable features such as schools and hospitals could be used to measure potential pollution impacts upon such clusters of human activity (Ihlanfeldt and Taylor 2001). For example, Lake et al. (1998; 2000a,b) used distance functions as inputs to noise estimation models applied within hedonic pricing analyses. Similarly, Powe et al. (1997) approximated forest amenity impacts upon property prices by constructing an index variable that measures the ratio of forest acreage to the squared distance away from each property. In order to examine the relationship between prices of waterside properties and indicators of nearby water quality, Leggett and Bockstael (2000) used GIS to assign measures of faecal coliform counts taken at sampling stations to property locations using an inverse distance-weighted average of coliform counts based on data from the nearest monitoring stations. The distance function can also be used to measure travel distances between visitors’ homes and recreational sites in travel costs models (Bateman et al. 1996). These distance measures can be Euclidean straight lines or measured along road and trail networks in digitized GIS layers (see subsequent discussions of connectivity). One important use of the distance function in GIS is the ability to create buﬀers of diﬀerent sizes around points, lines or polygons. This ability to buﬀer spatial objects provides a way to generate new polygons that represent inﬂuencing regions of or upon the study objects. Some examples of buﬀering in economic studies include Boxall et al. (1996), who classiﬁed canoe routes (linear features) in Manitoba by the amount of the forested area along both shores that were subjected to historic forest ﬁres. Using GIS, they generated a buﬀer of 500 m along the canoe routes and assessed how much of this buﬀer area was burnt. The resulting measures were used

Geographical information systems and spatial analysis

61

in a model of recreational site choice to assess the inﬂuence of historic ﬁres on recreation behaviour. Yang and Weersink (2004) imposed various buﬀers between 5 and 20 meters on a river system to delineate the riparian area and estimate the economic costs and water quality beneﬁts of these buﬀers for prioritizing the targeting of riparian buﬀers to achieve environmental goals. Boxall et al. (2005) buﬀered rural residences to assess the inﬂuence of oil and gas facilities on property values. In this case the residence (a point object) served as the centre and the buﬀer around each residence (a circle) became a new polygon layer. Connectivity Connectivity is a measure of accessibility between spatial objects or entities. This measure can be used to quantify the quality of habitats by examining, say, the number of connections (links) between patches of forest. GIS have the capacity to enumerate these connections and to calculate relevant measures. For example, a gamma index for measuring connectivity is calculated by dividing the number of links by the maximum number of links available between diﬀerent spatial objects (DeMers 2003). In applications, the connection measure can be made more sophisticated. For example, habitat connectivity can be measured as either species dispersal success or search time, based on immigration into all habitat patches on the landscape (Tischendorf and Fahrig 2000). Pedestrian connectivity is introduced as a measure of both the directness of route and the route distance for the pedestrian for each home-destination trip (Randall and Baetz 2001). Simple connectivity analyses typically adopt the assumption that all connected nodes are of equal size or importance. But in reality nodes may have diﬀerent degrees of attraction and therefore inﬂuence connectivity. For example, a lake is likely to attract more waterfowl than a pond. This leads to the construction of the gravity model where the concept of gravitational attraction is incorporated to measure interactions between two nodes in a GIS coverage. The model may have diﬀerent functional forms. The simplest form of the model is Lij K(PiPj)/d 2, where Lij is the interaction between nodes i and j, Pi and Pj are the magnitudes of node i and j respectively, d is the distance between the two nodes, and the constant K relates the equation to the types of object being studied such as population size, etc. As shown in the equation, the larger the magnitude of the nodes, the greater the level of interactions and hence connectivity between nodes (DeMers 2003). Another useful GIS application in connectivity is routing, an operation which involves ﬁnding the path with the minimum impedance (the ‘diﬃculty’ of travel) between nodes on a network (Chang 2002). GIS-based routing algorithms have been used to provide a spatial framework for environmental analyses such as that provided by the Agricultural Nonpoint

62

Yearbook of environmental and resource economics

Source Pollution (AGNPS) model which is used to characterize the transport of pollutants in watersheds (Stocks and Wise 2000). An obvious application of such impedance-based routing features is within analyses of recreational values as clearly gravel roads provide higher levels of impedance than paved roads. This ability of GIS functions to assess routing and impedances can be used to estimate recreation travel times and so provide inputs to travel cost analyses. Bateman et al. (1996, 1999b) use such GIS-based routines to generate ‘isochrones’ (literally lines of constant travel time) around recreational sites. The left-hand map in Figure 2.7 shows an isochrone map for a single site (here one-minute-width isochrones have been amalgamated for the purposes of reproduction). Jones et al. (2002) extend this methodology by generating isochrones not only for a large set of woodland recreation sites but also for a highly extensive set of potential substitute and complementarity amenities. The lower panel map in Figure 2.7 shows isochrones around one such set of alternative recreational sites (wildlife parks). Subsequent analyses discovered that complementary (visitincreasing) relationships existed between a given woodland site and lakes, heathland areas, coastal beaches, National Trust properties and urban attractions.6 Conversely, substitute (visit-reducing) relationships exist between visits at a given woodland site and other such sites and canalside walks (both of which provide very similar recreational experiences). A further connectivity application is in ‘allocation’ – the study of the spatial distribution of resources through a network (Chang 2002). A common measure of allocation is the area covered by a public facility within a service distance or response time. For example, the allocation facility can be used to deﬁne the market extent of a water supply plant where the maximum service area through the network is deﬁned (Church 2002). Similarly, the response time of a ﬁre truck, ambulance or other emergency response vehicle to reach a chemical spill or pollution incident could deﬁne a service area for a ﬁre station or hospital (Chang 2002). Spatial overlay Arguably one of the most important yet relatively simple features of GIS is the integration of various layers of spatial data through the process of overlays. Spatial overlay for raster data layers involves diﬀerent arithmetic operations, and researchers need to be cautious in preparing the input grids and interpreting output grids. For example, imagine that four data layers are pertinent to the selection of a nuclear waste site: geology, population density, road and rail accessibility buﬀers. Suppose also, for simplicity, that these can be classiﬁed using a binary system where 1 indicates suitable sites and 0 denotes unsuitable sites. An overlay of the four layers applying a multiplication operation will produce a binary grid where 1 indicates locations which

Geographical information systems and spatial analysis

Travel time (minutes) < = 9.9 10 – 29.9 30 – 59.9

Main roads

60 – 89.9

Study site

> = 90 0

10

20

30

40

50

km

Location of wildlife parks Travel time (minutes) 0–10 11–20 21–30 31–45 46–60 61–90 91–120 121–180 Over 180 0

50

100

200

300 km

Sources: Upper panel: adapted from Bateman et al. (1996, 2002); lower panel: adapted from Bateman et al. (2002a) and Jones et al. (2002).

Figure 2.7 Upper panel: isochrone map for a single recreational site. Lower panel: isochrone map for all wildlife parks (potential substitutes) in the UK

63

64

Yearbook of environmental and resource economics

meet all four criteria and 0 represents areas with at least one unsuitable factor. Alternatively, an overlay of the four layers using an addition operation will produce a grid with values ranging from 0 to 4, where 0 indicates locations with all factors unsuitable and 4 represents places with all factors suitable. These two overlay operations represent diﬀerent decision criteria, with the addition operation being the more ﬂexible (Heywood et al. 2002). There are three common spatial overlay operations for vector data: point-on-polygon; line-on-polygon; and polygon-on-polygon (DeMers 2003). For example, a point data layer on toxic sources could be overlaid with a census population layer to link the residential area information. A line feature such as a planned route layer can be overlaid with a vegetation layer to uncover the inﬂuence of the route on vegetation types. Similarly, the overlay of a soil polygon layer with a land use polygon layer reveals the soil types associated with each land use type (see e.g. Bateman and Lovett 1998 or Bateman et al. 1999a). There are other types of vector overlays between spatial objects. For example, two point data layers could be overlaid and the distance between pairs of points calculated. Such operations are used by Lovett et al. (1998) for relating air pollution emission sources to pollution monitoring stations. Line and line coverage overlays can be used to determine places of intersection between the two layers. For example, a canoe route layer can be overlaid with a stream and lake layer in GIS to determine the number and length of portages canoeists must undertake to follow each route (Boxall et al. 1996). Despite their versatility of application, there are complications involved in polygon-on-polygon integration in vector data. An initial problem concerns cases where overlay data are at the same spatial scale but are held using diﬀerent spatial units (or boundaries). Such problems often arise when integrating environmental and socioeconomic data layers. For example, a farm ﬁeld boundary may be not consistent with a soil boundary. When the two layers are overlaid, the farm ﬁeld may have some soil types completely within the ﬁeld boundary while other soil types are only partially within it. A second complication is when researchers overlay data at diﬀerent spatial scales. For example, STATSGO in the USA is a small-scale (less detailed) soil data layer for each state within which data are highly generalized with large polygons. When this soil data layer is overlaid with a large-scale (more detailed) land slope data layer with small polygons, each soil polygon will contain more than one slope polygon. In both cases ‘aggregation’ of data layers is needed to extract information from the layer with small spatial units to the layer with large spatial units. Researchers need to be aware of the implications of aggregation procedures and select appropriate statistical measures. In the farm soil example

Geographical information systems and spatial analysis

65

mentioned above, an area-weighted average or maximum of soil parameters such as organic matter could be estimated for each farm ﬁeld. In the STATSGO case, an area-weighted average or maximum slope could be computed for each soil polygon. The two aggregation schemes have diﬀerent impacts on research outputs and should be selected by keeping in mind the objectives of the relevant modelling exercise (see Yang et al. 2003; 2005). In some cases mean values are more informative (e.g. for subsequently calculating total soil organic matter levels) whereas for other situations maximum values identify the issues of interest (e.g. identifying farms that are particularly disadvantaged by land slope problems). Because of these diﬀerent impacts of aggregation schemes, studies employing these procedures need to be fully documented to ensure reproducibility of results. Neighbourhood operations or analysis Neighbourhood operations or analyses typically refer to the assignment of raster data in neighbouring cells to some designated focal cell. Neighbourhood cells can be in a ﬁrst- (adjacent) or higher-order (two or more cells away from the focal cell) conﬁguration (Chang 2002). The assignment of data can be achieved using diﬀerent statistical measures such as minimum, maximum, average, total, medium, mode, range and diversity (Burrough and McDonnell 1998). As shown in Figure 2.8, the 3 3 grid in each lattice contains a central base or focal cell, here shaded in grey. The values of neighbouring cells (including the focal cell) are used to assign a value to the focal cell based on diﬀerent statistical measures. For example, the GIS ‘total’operation assigns the sum of the values of all cells in the neighbourhood into the focal cell. Conversely, the ‘diversity’ operation compares the value of the focal cell with those of neighbouring cells to assign the total number of diﬀerent values to that of the focal cell. Alternatively, the ‘mode’ operation assigns the most frequently appearing value in the neighbourhood into the focal cell. These functions are useful in many ways. For example, in the analysis of wildlife habitats it is sometimes convenient to calculate the total number of species in a neighbourhood and use this number as a measure of importance in prioritizing protection locations. The method could also be used to derive diversity in vegetation types in a neighbourhood (Chang 2002). Case study combining GIS functions: GIS as the last/best hope for beneﬁt function transfer As we have demonstrated in the preceding discussions, GIS oﬀer a highly ﬂexible and practical array of functions for applied research in environmental and resource economics. However, the greatest empirical gains are achieved when these various functions are combined to more accurately model the complexity of real-world environments. In this subsection we

66

Yearbook of environmental and resource economics

2

3

4

2

3

4

2

1

5

5

1

6

5

1 31 5

3

4

6

3

4

6

3

4

6

2

3

4

R

A

G

3

4

1

5

3

4

6

TA

DIVERSITY

E

1

4

L

N

2

3

MIN

X

A

2

3

4

1

3

5

3

4

6

STD DEV

TO

MEAN

1 3.4 5 3

4

6

M

O

D

E

M

2

3

4

2

4

2

3

4

1

6

5

1 1.5 5

1

3

5

3

4

6

3

3

4

6

3 4

6

Source: Adapted from Burrough and McDonnell (1998).

Figure 2.8

Examples of types of neighbourhood analysis for raster data

provide a brief case study example illustrating the synthesis of a selection of these functions. The case study concerns the transfer of models describing the recreational beneﬁt value of one type of open-access resource – woodlands in Great Britain. This draws upon surveys conducted both by the researchers and the UK Forestry Commission amounting to more than 13 000 interview records obtained from woodland sites across the country. In discussing this we draw upon previous work (described in Bateman et al. 1996; 1999b and 1999c; 2003; Bateman and Jones 2003; Brainard et al. 1997, 1999, 2001; Jones et al. 2002; and Lovett et al. 1997). Beneﬁt transfer typically involves the inference of values for some resource site which policymakers are interested in (the ‘policy site’) based upon prior research estimating values for similar sites elsewhere (the ‘survey’ sites). One of the more sophisticated approaches to beneﬁt transfer is to estimate a value function based on data from a set of survey sites and then use this function to predict values at the unsurveyed policy sites. In eﬀect the assumptions here are that all sites have a common set of predictor

Geographical information systems and spatial analysis

67

variables and, while the level of these variables may change across sites, the coeﬃcient values estimated for the survey sites apply to the policy sites. One of the criticisms of prior studies (which have typically failed to support the hypothesis that such functions can be transferred) is that they rely upon a very limited set of predictor variables and that even these are poor approximations of the complex environments which characterize a recreational site. GIS directly address this criticism by allowing the researcher to generate an extensive set of high-quality predictor variables for both survey sites (to feed into the estimation of beneﬁt functions) and policy sites (to provide the level of predictor variables at those sites, to which coeﬃcients estimated at survey sites may be applied allowing the derivation of recreation values for policy sites). These joint operations of estimating beneﬁt function models for survey sites and transferring these to policy sites involve a number of the GIS functions described previously. In our work on developing a GIS-based beneﬁt function transfer methodology we have attempted to incorporate in an accurate yet readily reproducible manner the complexity of environmental and socioeconomic factors which determine recreational visit decisions, in this case to open-access woodlands. An initial task was to link records concerning visitor outset locations to the spatial coordinates of destination sites. This required not only GIS data acquisition functions but also certain data processing operations such as the conversion of visitor outset postcode records to their spatial coordinate equivalents. This was achieved by using the GIS to link spreadsheet records of visitor survey responses and addresses, through a postcode database and on to a spatial grid. For modelling purposes we are interested not only in visits to the set of woodland sites for which we have survey records, but also to other potential woodland recreation sites. This requirement was addressed through further data acquisition routines importing images from satellite sources until a full coverage of Great Britain was assembled. A second task was to apply GIS distance and connectivity functions to generate accessibility measures from each outset location to each destination. Here connectivity impedance routines were used to incorporate data on the full road network with data on the quality and congestion of roads and resultant road speeds. Robust statistical modelling of visitation requires that we consider not only the decision of a particular individual to visit a given site, but also the decision of another individual not to make such a visit. While we did not have direct survey information from non-visitors, such relationships can be inferred by contrasting the distribution and characteristics of the population (considered below) with the availability of both survey sites and (from secondary data sources) non-surveyed woodland sites. To capture this information, GIS connectivity operations were iterated across a

68

Yearbook of environmental and resource economics

high-resolution (500 m cell size) regular grid covering the entirety of Great Britain. For each cell, accessibility was measured to every woodland across the country. Data-driven inverse weighting routines were tested to replicate the functional form of the prominence given to more accessible sites. Spatial distribution routines were also used to incorporate further weights allowing for diﬀering attraction values for woodlands according to their size. Data concerning the facilities and attractions oﬀered at woodlands were also incorporated into the analysis. Of course it is not only the distribution of other woodland sites which determine visits to any given survey site. A further important determinant of visitation rates is the substitute and complementarity relationships which may exist with regard to other attractions. Therefore alongside the measures of accessibility to other woodlands mentioned previously, GIS data acquisition, distance function and connectivity routines were used to assess the impact upon woodland visitation of a highly diverse set of attractions. This included open-access countryside attractions such as coastal beaches, heathlands, national parks, etc., open-access man-made attractions such as castles and historic houses, and developed attractions requiring entrance fees such as National Trust properties, theme parks, zoos and wildlife parks and urban attractions. As adjustment needed to be made for the obviously uneven distribution of population across such a large study area, further data acquisition and spatial overlay functions were used to import data from the UK census. This also permitted the incorporation of data on the spatially varying characteristics of that population including its demographic, socioeconomic and ethnic nature. GIS spatial overlay functions were used to compile these diverse datasets into a single uniﬁed database. Beneﬁt value functions were estimated using count data models applied using multi-level modelling techniques which controlled for the impact upon error structures of repeated observations being obtained from a given forest site. Results showed, perhaps not surprisingly, that location is vital to the determination of visitor numbers and corresponding values. Reducing travel times by locating recreation sites near to areas of high population was by some margin the single most important factor inﬂuencing visits. By contrast, site facilities, other than the basic provision of walking tracks given at all sites, only exerted a weak inﬂuence upon visit numbers. However, the proximity of other attractions proved highly signiﬁcant in determining visits. While the presence of other woodlands acted as substitutes, reducing visits, numerous complementary relationships were identiﬁed including boosts to visitor numbers from nearby open-access sites including inland water attractions, coastal beaches and heathland areas. Developed attractions requiring entrance fees also

Geographical information systems and spatial analysis

69

boosted woodland visits, including National Trust sites and urban attractions. A number of socioeconomic, demographic and ethnicity variables also proved signiﬁcant; for example visits were higher in areas with higherincome and retired populations. Beneﬁt function transfer testing is typically achieved by omitting certain sites and using functions based upon the remaining subset to estimate values for those omitted sites, these values then being compared with those estimated directly from data collected at those sites. While this is a reasonable procedure, arguably this type of internal validation lacks the objective weight of comparison with some external criterion measure. Furthermore, some policymakers remain sceptical regarding non-market values. Consequently, in a separate analysis, we compared our estimated visit numbers with oﬃcial visitor counts. Figure 2.9 graphs our predicted visitor numbers against oﬃcial estimates. As can be seen, there is a good correspondence between these ﬁgures. Table 2.3 reports simple OLS (ordinary least squares) models of this relation, ﬁrst with an intercept term and then, as this is clearly insigniﬁcant, by dropping this constant. As can be seen, the slope coeﬃcient is insigniﬁcantly diﬀerent from unity (with a 180 000 160 000

Official estimates

140 000 120 000 100 000 80 000 60 000 40 000 20 000 0 0

20 000

40 000

60 000

80 000 100 000 120 000 140 000 160 000 180 000 Predicted visits

Figure 2.9 Oﬃcial counts of recreational visits to UK woodlands and visits predicted by GIS-based models; linear trendline for model including intercept added

70

78.0%

8105.9 1.144 6740.7 0.119

s.e. 1.203 9.650

t 0.240 0.000

Sig (p) Not included 1.021

Coeﬃcient

91.4%†

n/a 0.060

s.e.

n/a 17.014

t

Model without intercept

n/a 0.000

Sig (p)

Note: †Estimates of R2 for models without constants are not comparable with those that include an intercept. Instead this measure expresses the proportion of the variability in the dependent variable about the origin explained by the model.

R2 (adj.)

Constant Predicted visits

Coeﬃcient

Model with intercept

Table 2.3 Regression models relating oﬃcial counts of visitors to woodland sites (the dependent variable) to predictions of the number of visitors obtained from GIS-based analyses

Geographical information systems and spatial analysis

71

small degree of variance). In eﬀect we cannot reject the hypothesis that our GIS-based transfers are providing good estimates of actual recreation demand. The above results suggest that, when performed using GIS to capture the complexity of the real-world environment, beneﬁt function transfers may yield acceptable approximations of demand and values at policy sites. Furthermore, although the initial GIS manipulations required to produce digital map layers of pertinent substitutes and complementarities may be analytically demanding, once these are produced they do not need to be reconstructed from scratch for future analysis. Rather, they can be reused and only occasionally updated (say every few years) to allow for changes in road networks, population distribution and new attractions. In essence, therefore, we have a policy-usable tool which appears capable of delivering the objective of beneﬁt transfers – an acceptable degree of accuracy in predicting visits and values at policy sites. Once derived and suitably tested, GIS beneﬁt transfer functions can also be used to assist in the fundamental task of economic analysis: identifying the optimal allocation of limited resources. An example of this undertaking is given in Bateman et al. (2003) through the construction of GIS value maps for recreation demand, timber yield, carbon sequestration, agricultural values and cost–beneﬁt analysis of land use change. Figure 2.10 illustrates a map of potential recreation demand values generated by transferring GIS-generated travel cost functions estimating the beneﬁts of locating recreational woodlands in diﬀerent locations across the entirety of Wales. The pattern shown conﬁrms to prior expectations, with values being highest for sites located in areas of high populations (e.g. around Cardiﬀ in the south of the country) and with good road infrastructure access (e.g. the area in the north-east of the country which can readily be accessed by populations from the large conurbations of Liverpool and Manchester). Conversely, recreation values are lowest in the upland middle and coastal western areas, where local population density is low and accessibility is poorer. Such maps are ideally suited for allocating resources so as to optimize economic values. Unfortunately, as Bateman et al. show, actual planting of forests has been guided not by economic values including nonmarket recreation beneﬁt, but rather by a desire to minimize market land purchase costs. This has led to concentrations of woodland in the lowestvalue central areas of the country, a situation which constitutes a classic market failure. The above gives an overview of functions through which GIS can help incorporate spatial complexity within applied analyses. However, of equal importance are the methodological techniques for spatial analysis to which such tools can be applied. In the following section we provide a brief

72

Yearbook of environmental and resource economics

in £000s < £60 £60 to £99

Roads

0

£100 to £199

Motorway

£200 to £299

Dual carriageway

≥ £300

Single carriageway

10

20

30

40

50 km

Source: Bateman et al. (2003).

Figure 2.10 GIS-generated map of the marginal value of predicted woodland recreation demand for potential forest sites in Wales (£ per site per annum) 72

Geographical information systems and spatial analysis

73

overview of some pertinent aspects of spatial analysis which may be of use to environmental and resource economists.

3. SPATIAL STATISTICAL ANALYSES AND MODELLING In many resource and environmental economic studies there is a need to integrate biophysical and economic data. Typically this information is formed by some natural or economic process that results in diﬀerent spatial dispersion, or the data are collected at quite diﬀerent scales. The potential divergence in dispersion and scales of these data sources usually results in the presence of various spatial eﬀects. Anselin and Getis (1992) suggest that these eﬀects have two interrelated implications which complicate any understanding of spatial data. They list the ﬁrst as Tobler’s First Law of Geography (Tobler 1979), which states that ‘everything is related to everything else, but near things are more related than distant things’. The second is that the size and conﬁguration of the spatial units of analysis generate relationships within or among the variables being considered that are results of the nature of the spatial units employed as well as the variables under examination. Spatial eﬀects can be classiﬁed into spatial dependence and spatial heterogeneity. The former refers to the propensity for observations close in space to be more highly correlated than those observations further apart. Examples of spatially dependent phenomena include knowledge spillovers resulting from the spatial diﬀusion of ideas or technologies among economic agents located in space (e.g. farmers); and contagion eﬀects such as the spread of diseases or toxins from a point of origin. Dependent relationships between spatially referenced data can arise from the nature of the variable(s) under study, as well as the size, shape and conﬁguration of the spatial units employed in the analysis. For example, small spatial units experience a greater probability that nearby units will be spatially dependent on them than larger spatial units. If units are spatially long and narrow, the chances of spatial dependence with nearby units will also be greater than if units are more compact. Spatial heterogeneity results when there is lack of spatial uniformity of the eﬀects of spatial dependence (Anselin and Getis 1992). For example, spatially homogeneity would describe a set of data that exhibits a spatial dependence structure that is consistent across space, or in other words the correlation among neighbouring units is the same everywhere. When these data are spatially heterogeneous, the degree and direction of dependence diﬀers among the observations across space. For example, crop yields in an

74

Yearbook of environmental and resource economics

agricultural watershed may be spatially heterogeneous because of the presence of diﬀerent soil types and other landscape characteristics (Yang et al. 2003, 2005). Crop yields within locations in a given ﬁeld, however, may be more homogeneous due to the similarity of the soil types and conditions. Heterogeneity can also result when spatial relationships between the variables under study are inconsistent across space. In essence, spatial heterogeneity is a special case of spatial dependence and can be a complex realization of the nature of the variables being examined. Techniques for the analysis of spatial data range from the simple description of patterns and information (exploratory spatial data analysis or ESDA) to spatial regression (model-driven inference and conﬁrmatory techniques). Typically these analyses take place using software available outside of GIS. However, the processes are clearly facilitated through the use of GIS because the results are displayed or mapped using the GIS and can be better understood as a result.7 We consider these approaches below but note that both can typically be employed in a research endeavour. Exploratory Spatial Data Analysis (ESDA) Exploratory data analysis refers to the manipulation and reduction of data to gain insights and formulate or explore hypotheses concerning spatial relationships. This approach for non-spatial information was pioneered by Tukey (1977). When data are spatially referenced, however, researchers must go beyond traditional approaches. Use of GIS for this is a key component of ESDA. An important advantage in using GIS in this process is the ability to visualize the spatial data. The visual display of data from a number of diﬀerent perspectives can facilitate the discovery of errors in the data prior to formal analysis, as well as determine various ways in which relationships could be modelled in subsequent stages of the analysis. Consider a dataset consisting of an attribute for each case and its location identiﬁer. One obvious class of techniques for initial exploration of this information includes the mapping of this single attribute information using GIS and the additional mapping of layers of diﬀerent spatial information, perhaps other attributes associated with each location. For example, one could envision exploring the eﬀects of point source pollution on human health parameters in some region through the development of maps in which one attribute, incidence rates of some human health outcome, is mapped along with a second attribute, the pollution point sources. Further exploration involves the use of related non-spatial techniques: the so-called standard exploratory data analysis techniques such as measures of the centre of a distribution of the attribute values (the median), measures of the spread of values about the median (quartiles or quintiles

Geographical information systems and spatial analysis

75

etc.), measures of central tendency (mean and mode) and outliers. Perhaps the most popularly used exploratory data analysis technique is the box plot, which is a graphical summary illustration of all of these measures of the distribution of attribute values. In ESDA these measures can be linked with the GIS such that areas with attribute values above or below the median are mapped, or spatial units that lie in the upper or lower quartiles and/or that contain outliers are mapped (see Figure 2.1 for an example). These techniques have the ability to assist researchers in gaining insights into potential relationships among variables and can result in the construction of hypotheses to be tested in conﬁrmatory studies (see below). However, the simple ESDA techniques described above suﬀer from the perspective that the attribute values are merely considered as observations upon some phenomenon, and beyond a simple mapping of the values, are rarely considered as spatial in nature. Furthermore, these techniques only apply to the univariate case. When the attribute data are formally considered spatial in nature, then other techniques can be employed to describe spatial properties. One method is smoothing, in which a map of many small spatial units is treated to reveal patterns (such as a trend) that may be unclear from a simple mapping of values. Typically this involves spatial averaging in which the attribute value for one spatial unit is averaged with that of its neighbours (see Figure 2.8 and Haining et al. 1998). For example, median-based ‘averaging’ was used by Pickle and Su (2002) to spatially smooth the proportion of women who ever smoked cigarettes in each county based on a ‘smoothing window’ of up to 30 neighbouring counties in the USA. In this study, the smoothing procedure was implemented to stabilize health patterns from sparsely populated areas and permitted the identiﬁcation of clusters of counties with similar health values. Another set of methods involves the determination of trends or gradients in the map distribution of attributes (e.g. increases in pollution from north to south). Haining (1993) describes various approaches to do this, such as kernel estimation, spatially lagged box plots, and the use of transects through data coupled with various scatterplot techniques. Clearly the spatial trending and neighbourhood GIS facilities described previously (Figure 2.8) have relevance here. The next group of techniques responds to the criticism that exploratory analyses generally do not include measures of ﬁt and tests of signiﬁcance in spatial association. To do this, one must formally consider the spatial eﬀects (dependence and heterogeneity) described above, as these eﬀects can lead researchers to make spurious conclusions regarding spatial trends and patterns detected with the exploratory techniques described above. For example, statistical tests of spatial patterns in the presence of spatial dependence can appear to be signiﬁcant (null hypothesis is rejected).

76

Yearbook of environmental and resource economics

Formal consideration of spatial eﬀects involves testing the assumption of stationarity or structural stability of an attribute over the entire space or region under study. The most common spatial eﬀect dealt with by analysts in this arena is spatial autocorrelation, which refers to the propensity for attribute values in neighbouring areas to be similar (positive autocorrelation) or dissimilar (negative autocorrelation). Another spatial eﬀect rarely examined by economists is spatial cross-correlation, which refers to the relationship between more than one attribute across space.8 A simple initial ESDA technique for examining the existence of spatial autocorrelation is a scatterplot, where each attribute value is plotted on the vertical axis against the average of the attribute values in adjacent spatial areas on the horizontal axis. Upward right-sloping scatter points reveal positive spatial autocorrelation while upward left-sloping scatter indicates negative autocorrelation. The literature contains a number of formal treatments of spatial autocorrelation at the ‘global’ or whole-map perspective. Examination of autocorrelation involves the computation of statistics and their levels of signiﬁcance that represent tests for the propensity of the attribute values to cluster over the study region. Anselin et al. (2004) refer to global examination of autocorrelation as tests for ‘clustering’. The most common global statistics which economists may be familiar with are the joint count statistic for nominal data, and the Moran’s I and the Geary’s c statistics for assessing spatial autocorrelation using interval or ratio data. The details of how these statistics are calculated can be found in Cliﬀ and Ord (1981). Depending solely on global measures of autocorrelation to assess the propensity of some attribute to cluster over space can be limiting in that these statistics assume that measures from the whole dataset represent situations in all parts of the area under study. Thus, recent developments in the spatial association literature have focused on local rather than global measures of spatial autocorrelation. This approach to examining spatial information is concerned with diﬀerences or similarities in one (univariate) or two (bivariate) attribute values among neighbouring spatial units. Hence the local measures focus more on identifying speciﬁc clusters of similar attribute values rather than the propensity of these values to cluster over the study region. The techniques, called local indicators of spatial association (or LISA) by Anselin (1995), involve dissection of the global statistics, such as Moran’s I, into local constituencies. On the basis of local measures of spatial autocorrelation, areas of the study region in which the attribute values of neighbouring spatial units cluster into ‘high–high’ (hot spots) or ‘low–low’ (cold spots) valued areas can be identiﬁed by assessing the frequency and direction of speciﬁc clusters of attributes in the study region. Furthermore, as developed by Anselin et al. (2004) in the GeoDa software

Geographical information systems and spatial analysis

77

package, clusters of ‘high–low’ and ‘low–high’ attributes associated with spatial units can also reveal the presence and location of outliers. The development of tools which assess local spatial relationships should be of considerable utility to resource and environmental economists since assessing the level, direction and signiﬁcance of association among spatial units in a study area may reveal interesting patterns of environmental and human interactions. Figure 2.11 provides a univariate LISA cluster map derived using the GeoDa software package. The study involved collecting forestry experts’ ratings of the number of visits by recreationists in non-designated camping areas in the Eastern Slopes region of the province of Alberta, Canada. Much of this region outside of Banﬀ and Jasper National Parks is undeveloped public lands which are freely accessible by campers. There are signiﬁcant concerns regarding forest ﬁres, both for the ﬁre potential caused by campers and for evacuation purposes in the event of ﬁres. The spatial unit in this example was 25 km2 cells which constitute a regular (equal-sized square cells) lattice. Areas of ‘high–high’ values identify clusters of cells with correlated high ratings and thus represent hotspots of recreation activity. The value of the global Moran’s I was 0.46, which indicated the presence, over the entire region, of positive spatial autocorrelation among the ratings. The ‘high–high’ clusters, however, identify speciﬁc areas of the mountain region where ﬁre management and detection resources should be allocated. Conversely, the ‘low–low’ areas represent clusters of correlated low recreation ratings and would represent areas where ﬁre management eﬀorts may not be as important concerning recreation use. Spatial Regression The exploration of spatial data in conjunction with consideration of expected relationships based upon theory naturally leads to the development and formal testing of models. In economics, spatial econometrics has become an important component of the economist’s toolbox, and in resource and environmental applications the construction of hedonic price regressions is the most frequently used of these econometric tools. The development of spatial econometrics, led largely by Luc Anselin, involves regression techniques that explicitly recognize a number of problems inherent in the analysis of spatial data. Perhaps the most important problem is that of spatial autocorrelation described above, which, if present, results in the violation of the assumption of the classical regression model that the data consist of independent observations. This violation can lead to ineﬃcient or biased estimates and misleading inference. To address this problem the literature points to two spatial regression approaches: (1) a

78

Yearbook of environmental and resource economics LISA clusters High–high Low–low Low–high High–low

0

60

120 km

Canada

Alberta

Study area

Figure 2.11 A local indicators of spatial association (LISA) map for expert ratings of the intensity of random camping in 25 km2 cells in the East Slopes Region of Alberta, Canada

Geographical information systems and spatial analysis

79

spatial lag model in which the dependent variable exhibits positive spatial autocorrelation; and (2) a spatial error model where the errors in the regression are spatially autocorrelated. Each of these is brieﬂy addressed below as there are a number of detailed treatises in the literature (e.g. Anselin 1988). To illustrate these regression approaches, consider n observations of some spatially referenced dependent variable, denoted by the vector y and n observations of m independent variables represented by the n by m matrix X. The traditional OLS model can be expressed as follows: yX

(2.1)

where is a vector of coeﬃcients, and is a vector of random errors with elements distributed N(0, 2I). The assumption that the variance of is 2I does not allow the researcher to account for the existence of spatial patterns in the dependence of residuals. For instance, the error terms resulting from applying this model may be linked to adjacent spatial units. One way to account for this spatial dependence could be to include the mean of the dependent variable in adjacent spatial units for each of the n observations as an explanatory variable in the regression model. This involves a linear transformation of the vector y into another vector of adjacent means which can be written as Wy, where W is an n by n symmetric matrix called a spatial weights matrix. The elements which form W are speciﬁed by the researcher such that the realm of relevant neighbours for each element of y deﬁnes the form of spatial dependence. The construction of W has an obvious link to the neighbourhood operation in GIS described earlier. The speciﬁc development of W is discussed in detail below. Incorporating the additional Wy term on the right-hand side of the regression model gives: y X Wy

(2.2)

where represents the coeﬃcient for the spatially adjacent mean variable. For obvious reasons this model is often called a spatial autoregressive model. Important estimation issues arise with this model since including the Wy term on the right-hand side imparts a degree of simultaneity, and this must be explicitly accounted for in either a maximum likelihood framework or through the use of instrumental variables (Anselin 1988). If the spatially lagged dependent variable is present in the data-generating process and ignored in the model, omitted variable speciﬁcation error results, with the implication that the OLS estimates will be biased and inconsistent. Including Wy thus allows for correct interpretation of the signiﬁcance of exogenous variables contained in X after the spatial eﬀects have been

80

Yearbook of environmental and resource economics

accounted for. If is known, further adjustments can include a ﬁltering of the spatial process which results from subtracting Wy from both sides of equation (2.2). This results in the following: yWy X

(2.3)

A diﬀerent form of the spatial autoregressive model results when the error term is considered autoregressive rather than the dependent variable. This model is identiﬁed in the following equation: yX

(2.4)

where W, represents a coeﬃcient for the error lag term W, and is an uncorrelated homoskedastic disturbance term. The inspiration for this model arises from time series models where both autoregressive and moving average models may be used to assess the behaviour of some dependent variable observed at regular time intervals (Fotheringham et al. 2000). In this model spatial error dependence is typically treated as a nuisance in the sense that it reﬂects spatial autocorrelation in measurement errors or in ignored exogenous variables that otherwise are not crucial to the model and/or spill over across the spatial units being examined (Anselin and Bera 1998). Ignoring spatial error dependence, when present in data, results in ineﬃcient regression estimates. In spatial regression a key component is the construction of the spatial weights matrix, W. Griﬃth (1996) suggests that the construction of spatial weights matrices ‘mostly is done in an ad hoc manner, and seems to be governed primarily by convenience and/or convention’. However, Bell and Bockstael (2000) show that speciﬁcation of this matrix in a hedonic pricing model had a greater impact on parameter estimates than choice of estimation technique. Thus construction of W is likely a key factor in spatial regression analysis and requires the use of GIS functions. The weights matrix expresses for each observation i (row) those locations j (columns) that belong to its neighbourhood set as non-zero elements. Thus the diagonal elements of W (wii ) are 0 (since an observation cannot be a neighbour to itself), and the non-diagonal terms (wij) can be non-zero. The speciﬁc value of wij, however, can depend on the researcher’s choice of the data model underpinning the statistical analysis. The choice of the raster or vector data model refers to the form or view of spatial reality chosen by the researcher in order to allow analysis from the GIS perspective. The most common data model in social science research is where spatial data are discrete objects and are represented in GIS as vector data. These objects typically correspond to economic agents such as businesses or

Geographical information systems and spatial analysis

81

jurisdictions and have discrete locations in space such as coordinates, addresses, census tracts or counties. Furthermore, in some studies these features can be associated with a lattice overlaid over the study area using GIS and this in essence converts the vector data to a raster format. In such models the simplest form of the weights matrix is the ‘binary contiguity’ matrix where units are speciﬁed as neighbours of unit i if they are contiguous or adjacent to i as shown in Figure 2.12. In this case wij 1 for each spatial unit which touches or adjoins i. Further, when comparison of spatial parameters is of interest, the rows of the weights matrix are typically standardized where wijs wij /j wij such that j wijs 1. Thus, resulting elements in the standardized row hold values between 0 and 1. In many environmental and resource economics studies spatial units (such as ﬁelds, habitats, ecosystems, counties or some other administrative units that share a common border with each other) would represent ﬁrst-order neighbours. This would be delineated through the GIS using the layer that contains the location data for the borders of the spatial units. Some studies using object data artiﬁcially create spatial units by using GIS to lay a lattice of polygons over the study area and then utilizing the cells in this lattice as the spatial units to deﬁne an observation. In a regular (equal area) grid lattice there are three possible conﬁgurations which can be used to deﬁne neighbours. Two of these are illustrated in Figure 2.12, where the shaded cells represent the relevant neighbours9 for the cell labelled ‘5’. The choice of conﬁguration has implications for the structure of the binary weights matrix, as further shown in Figure 2.12. This diﬃculty of choosing relevant neighbours has led some researchers to suggest the use of hexagons rather than squares as the spatial unit since hexagons share all six borders with other hexagons and the distance between the nearest centroids (the middle points of the spatial unit) of the hexagonal cells is the same in a hexagonal lattice (Deutsch 1970). This would not be the case in a square lattice utilizing the queen’s conﬁguration to deﬁne neighbours as shown in Figure 2.12. Another method used to construct the weights matrix involves the use of distances between centroids of areal units or the speciﬁc locations of the economic agents in the study area. This approach has been called distancebased contiguity (Anselin and Bera 1998) and an obvious use of GIS in this exercise is the calculation of the distances between agents using the distance function described above. One approach involves the construction of distance bands around each agent whereby wij 1 if dij c, where d is the distance between agent i and j and c is the distance cut-oﬀ (e.g. see Walker et al. 2000). Choosing the value of c, however, is obviously a critical element in this exercise and should be governed by some underlying theory. In a majority of empirical studies the distance band appears to have been

82

Yearbook of environmental and resource economics Weights matrix 0 1 0 1 0 0 0 0 0

1 0 1 0 1 0 0 0 0

0 1 0 0 0 1 0 0 0

1 0 0 0 1 0 1 0 0

0 1 0 1 0 1 0 1 0

0 0 1 0 1 0 0 0 1

0 0 0 1 0 0 0 1 0

0 0 0 0 1 0 1 0 1

0 0 0 0 0 1 0 1 0

0 0 0 0 1 1 1 0 1

0 0 0 0 1 1 0 1 0

Rook configuration for cell ‘5’

Spatial units 1

2

3

4

5

6

7

8

9

1

2

3

4

5

6

7

8

9

Bishop configuration for cell ‘5’ 1

2

3

4

5

6

7

8

9

Queen configuration for cell ‘5’ 1

2

3

4

5

6

7

8

9

Weights matrix 0 1 0 1 1 0 0 0 0

1 0 1 1 1 1 0 0 0

0 1 0 0 1 1 0 0 0

1 1 0 0 1 0 1 1 0

1 1 1 1 0 1 1 1 1

0 1 1 0 1 0 0 1 1

0 0 0 1 1 0 0 1 0

Figure 2.12 Two illustrations of ﬁrst-order contiguity binary weights matrices constructed from a 3 3 regular lattice chosen on an arbitrary basis (Haining 1993), or based upon the ﬁt of the underlying model (e.g. Kim et al. 2003). A further approach in calculating the non-zero elements in the weights matrix involves the use of parameters in conjunction with the distances between agents. The idea here is to calculate weights that reﬂect the degree

Geographical information systems and spatial analysis

83

of spatial interaction among agents, or to impose some decaying eﬀect of neighbours as a function of their distance from each other. The approach called generalized Cliff–Ord weights after its authors (Cliﬀ and Ord 1981), called generalized Cliﬀ–Ord weights, involves the calculation wij [dij ] [bij ], where dij is the distance between spatial units i and j, bij is the share of common boundary between agents i and j in the perimeter of i, and and are parameters. However, a simpler and more popular approach, developed by Can and Megbolugbe (1997) and widely used in the resource and environmental economics literature, involves using just the dij term. A common method here is some inverse distance weights function such as wij 1/dij, where the eﬀect of agent j on i is a declining function of the distance between them. This approach has been widely used in spatial hedonic studies including, for example, Gawande and Jenkins-Smith (2000) in an examination of the eﬀects of nuclear waste transport on property values and by Boxall et al. (2005) in an analysis of the impacts of sour gas wells on rural residential property values. Additional formulations for declining distance weights involve decay functions such as the negative exponential function (wij edij). The choice of the weighting parameter(s) in these more complex speciﬁcations of the weights matrix is obviously an important process of estimating the spatial model. Anselin (2002) notes that estimating these coeﬃcients jointly with the regression parameters is a complex undertaking, and typically joint estimation leads to a number of serious identiﬁcation problems. Usually, however, these parameters are chosen a priori, or are selected based on a series of preliminary tests of model ﬁt using various speciﬁcations. To illustrate the principles and issues described above we describe the study by Hunt et al. (2005) which investigates the eﬀects of logging operations on remote tourism prices in Ontario. The authors specify the following hedonic price model: pp(S, D,F,E) where p is the market price of a ﬁshing trip as charged by a tourism operation, S is a vector of the amenities of the site, D is the distance of the site from a ﬂoat plane airbase, F is a vector of ﬁshing quality measures and E is a vector of other environmental attributes. This type of model is well known in the resource and environmental literature although most studies utilize property sales rather than prices charged by accommodation or business operations. For example, a number of empirical applications have assessed the impacts on property prices of powerlines (des Rosiers 2002), nuclear waste transport (Gawande and Jenkins-Smith 2000), water quality (Leggett and Bockstael 2000), and road, rail and air traﬃc noise (Day et al. 2003, 2004).

84

Yearbook of environmental and resource economics

Hunt et al. (2005) ﬁrst used GIS to locate tourism enterprises in Ontario and then associated their prices with their locations. GIS was also used to derive independent variables such as the location and amounts of logging in the vicinity of the tourism operations and water quantity variables. Figure 2.13 shows the location of the sites stratiﬁed by their price. The authors noted the existence of a distinct spatial pattern in the prices: as one moves from south-east to north-west the prices become higher. Diagnostic tests indicated that while both a spatial lag and spatial error were present in the data, a spatial lag model was chosen since the 2 value of the LM test for the spatial lag greatly exceeded that value from the LM test for the spatial error model (see Anselin and Rey 1991). The lag was deﬁned in which the price at each enterprise was a decaying function of the distance to all of the other enterprises. Thus the weights matrix was developed using a negative power function (dij) and the authors chose the parameter

N

Price of a seven-day fishing package in Canadian dollars 390.00–703.02 703.03–991.63 991.64–1176.63 1176.64–1879.60 area of undertaking 100 50

0

100

200

300

400

500 km

Source: Hunt et al. (2005).

Figure 2.13 Spatial distribution of ﬁshing package prices at remote tourism sites in Ontario

85

Geographical information systems and spatial analysis

from a grid search to minimize the statistical test for spatial error dependence in the spatial lag model. The authors found that the spatial lag model resulted in parameter estimates that were quite diﬀerent from those estimated with OLS. To examine this further, the Moran’s I statistics were calculated for all sites within various distance bands of each other (i.e. 0–100 km, 101–200 km etc.). Figure 2.14 shows the results in the form of an autocorrelogram. The ln_price relationship identiﬁes the spatial pattern of prices among the ‘near’ establishments have prices that are positively correlated. As the distances among establishments become greater, the relationship among the enterprise prices becomes more strongly negatively correlated. This autocorrelation pattern is still apparent (but reduced) among the predicted values from an OLS regression. This results from the fact that some of the independent variables also contained a non-random spatial pattern of values. The inclusion of these independent variables helped to remove some of the spatial

1 ln_price OLS residual MLE residual

0.8 0.6 0.4

Moran’s I

0.2 0 ⫺0.2 ⫺0.4 ⫺0.6 ⫺0.8

10

0– 99 0– 19 9 20 0– 29 30 9 0– 39 40 9 0– 49 9 50 0– 59 9 60 0– 69 70 9 0– 79 80 9 0– 89 90 9 0– 10 999 00 –1 11 099 00 -1 12 199 00 –1 13 299 00 –1 14 399 00 –1 49 9 15 00 ⫹

⫺1

Distance interval in km

Source: Data and information from econometric models in Hunt et al. (2005).

Figure 2.14 Spatial autocorrelograms of the natural logarithm of weekly prices at ﬂy-in accessible tourism sites in Ontario

86

Yearbook of environmental and resource economics

autocorrelation among the residuals. However, the spatial dependence appears to have been successfully purged from the data using the spatial lag model as the predicted prices from this model displayed virtually no spatial autocorrelation (Figure 2.14).

4.

CONCLUDING REMARKS

As demonstrated throughout our review of the literature and empirical studies, the functionality provided by GIS can considerably enhance the incorporation of spatial issues within applied environmental and resource economics. However, it must be emphasized that GIS is not a universal panacea for improving data analysis. Indeed, the quality of results obtained depends upon a range of factors common to any quantitative analysis, such as the accuracy of the input information and the appropriateness of the data structures used to store it. Furthermore, as highlighted in the previous section, the choice of analytical methods is of crucial importance and if not accorded suﬃcient weight can negate all of the data and functionality advantages aﬀorded by using GIS. Notwithstanding the above caveats, there is, we believe, considerable scope for the continuing development of GIS applications in the ﬁeld of environmental economics. The techniques directly address many of the limitations in data handling and modelling that have restricted previous investigations. As illustrated throughout this chapter, the functionality provided by GIS packages allows the researcher to incorporate spatial complexity directly within applications. Provided that this is combined with suitable methods for spatial analysis, then it is our belief that continuing advances in computing power and GIS functionality will stimulate development in further areas of environmental economic research in the future. This chapter has sought to highlight to environmental and resource economists the great potential that GIS techniques oﬀer for incorporating the spatial dimension into applied studies. The diversity of studies discussed illustrates the great ﬂexibility and applicability of such techniques to a range of issues. Provided that developments in spatial analysis is heeded, then such application oﬀers the potential to signiﬁcantly enhance the ability of economists to successfully incorporate the complexity of the environment within their empirical analyses. Indeed, the promise of GIS is to turn the spatial dimension from one to be either ignored or inadequately represented into a key element of empirical economic investigations of the real world.

Geographical information systems and spatial analysis

87

NOTES 1. A layer is a thematic set of spatial data stored in GIS such as roads and political boundaries (Chang 2002). 2. Spatial overlay is deﬁned as combining the geometry and attributes of two or more spatial layers to create an output layer in GIS (Chang 2002). 3. See also Simpson et al. (1997), who envision future landscapes in environmentally sensitive areas of Scotland. 4. This research is a collaboration between CSERGE and the Tyndall Centre, UEA. Further results are given in Bateman et al. (2006). 5. The EA have since revised its approach. For more recent GIS analysis of WTP distance decay see Bateman et al. (2005). 6. Note that this analysis adjusts for population distribution so the relationship with urban areas reﬂects the facilities of such areas rather than their higher populations. 7. There is controversy in the literature about how much GIS should facilitate these analyses. The literature suggests that GIS was developed more for the management of spatial data than for analysis of those data (Fotheringham et al. 2000). 8. Spatial cross-correlation could be signiﬁcant in regional economic studies where, for example, consumption in one spatial unit may be related to income levels in neighbouring units. 9. Note that the conﬁgurations illustrated refer to the movement of chess pieces on the gridlatticed chess board. Thus the ‘queen conﬁguration’ in Figure 2.12 follows the possible one cell moves of the queen chess piece from the cell labelled ‘5’.

REFERENCES Anselin, L. (1988), Spatial Econometrics: Methods and Models, Dordrecht: Kluwer Academic. Anselin, L. (1995), ‘Local indicators of spatial association – LISA’, Geographical Analysis, 27, 93–115. Anselin, L. (2002), ‘Under the hood: issues in the speciﬁcation and interpretation of spatial regression models’, Agricultural Economics, 27, 247–67. Anselin, L. and A. Bera (1998), ‘Spatial dependence in linear regression models with an introduction to spatial econometrics’, in A. Ullah and D. Giles (eds), Handbook of Applied Economic Statistics, New York: Marcel Dekker, pp. 237–89. Anselin, L. and A. Getis (1992), ‘Spatial statistical analysis and geographic information systems’, Annals of Regional Science, 26, 19–33. Anselin, L. and S. Rey (1991), ‘Properties of tests for spatial dependence in linear regression models’, Geographical Analysis, 23, 112–31. Anselin, L., I. Syabri and Y. Kho (2004), ‘GeoDa: An introduction to spatial data analysis’, accessed 29 Nov. 2004, http://sal.agecon.uiuc.edu/pdf/geodaGA.pdf. Appleton, K., A. Lovett, G. Sünnenberg and T. Dockerty (2002), ‘Visualising rural landscapes from GIS databases: a comparison of approaches, options and problems’, Computers, Environment and Urban Systems, 26(2–3), 141–62. Appleton, K. and A. Lovett (2003), ‘GIS-based visualisation of rural landscapes: deﬁning “suﬃcient” realism for environmental decision-making’, Landscape and Urban Planning, 65(3), 117–31. Baldwin, J.A.D., P.F. Fisher, J.D. Wood and M. Langford (1996), ‘Modelling environmental cognition of the view with GIS’, in Proceedings of the Third International Conference/Workshop on Integrating Geographic Information

88

Yearbook of environmental and resource economics

Systems and Environmental Modeling, CD-ROM, National Center for Geographic Information and Analysis, Santa Barbara, CA. Bateman, I.J. (1994), ‘The contingent valuation and hedonic pricing methods: problems and possibilities’, Landscape Research, 19(1), 30–32. Bateman, I.J. and A.P. Jones (2003), ‘Contrasting conventional with multi-level modelling approaches to meta-analysis: an illustration using UK woodland recreation values’, Land Economics, 79(2), 235–58. Bateman, I.J. and A.A. Lovett (1998), ‘Using geographical information systems (GIS) and large area databases to predict yield class: a study of Sitka spruce in Wales’, Forestry, 71(2), 147–68. Bateman, I.J. and A.A. Lovett (2000), ‘Modelling and valuing carbon sequestration in softwood and hardwood trees, timber products and forest soils’, Journal of Environmental Management, 60(4), 301–23. Bateman, I.J., G.D. Garrod, J.S. Brainard and A.A. Lovett (1996), ‘Measurement, valuation and estimation issues in the travel cost method: a geographical information systems approach’, Journal of Agricultural Economics, 47(2), 191–205. Bateman, I.J., C. Ennew, A.A. Lovett and A.J. Rayner (1999a), ‘Modelling and mapping agricultural output values using farm speciﬁc details and environmental databases’, Journal of Agricultural Economics, 50(3), 488–511. Bateman, I.J., J.S. Brainard, G.D. Garrod and A.A. Lovett (1999b), ‘The impact of journey origin speciﬁcation and other measurement assumptions upon individual travel cost estimates of consumer surplus: a geographical information systems analysis’, Regional Environmental Change, 1(1), 24–30. Bateman, I.J., A.A. Lovett and J.S. Brainard (1999c), ‘Developing a methodology for beneﬁt transfers using geographical information systems: modelling demand for woodland recreation’, Regional Studies, 33(3), 191–205. Bateman, I.J., I.H. Langford, N. Nishikawa and I. Lake (2000), ‘The Axford debate revisited: a case study illustrating diﬀerent approaches to the aggregation of beneﬁts data’, Journal of Environmental Planning and Management, 43(2), 291–302. Bateman, I.J., A.P. Jones, A.A. Lovett, I.R. Lake and B.H. Day (2002), ‘Applying geographical information systems (GIS) to environmental and resource economics’, Environmental and Resource Economics, 22, 219–69. Bateman, I.J., A.P. Jones, A.A. Lovett, I.R. Lake and B.H. Day (2002a), ‘Applying geographical information systems (GIS) to environmental and resource economics’, Environmental and Resource Economics, 22, 219–69. Bateman, I.J., R.T. Carson, B. Day, W.M. Hanemann, N. Hanley, T. Hett, M. JonesLee, G. Loomes, S. Mourato, E. Özdemiroglu, D.W. Pearce, R. Sugden and J. Swanson (2002b), Economic Valuation with Stated Preference Techniques: A Manual, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Bateman, I.J., A.A. Lovett and J.S. Brainard (2003), Applied Environmental Economics: a GIS Approach to Cost–Beneﬁt Analysis, Cambridge: Cambridge University Press. Bateman I.J., B.H. Day and I. Lake (2004), ‘The valuation of transport-related noise in Birmingham’, technical report to the Department for Transport, published online at http://www.dft.gov.uk/stellent/groups/dft_econappr/documents/ divisionhomepage/032865.hcsp. Bateman, I.J., S. Georgiou and I. Lake (2005), ‘The aggregation of environmental beneﬁt values: a spatially sensitive valuation function approach’, Centre for Social and Economic Research on the Global Environment working paper EDM 2005–04, University of East Anglia, p.65. Bateman, I.J., A.P. Jones, S. Jude and B.H. Day (2006), ‘Reducings gains/loss asym-

Geographical information systems and spatial analysis

89

metry: a virtual reality choice experiment (VRCE) of land use change’, Centre for Social and Economic Research on the Global Environment working paper, University of East Anglia, p.40. Bell, K.P. and N.E. Bockstael (2000), ‘Applying the generalized-moments estimation approach to spatial problems involving microlevel data’, Review of Economics and Statistics, 82, 72–82. Bockstael, N.E. (1996), ‘Modeling economics and ecology: the importance of a spatial perspective’, American Journal of Agricultural Economics, 78, 1168–80. Boxall, P.C., D.O. Watson and J. Englin (1996), ‘Backcountry recreationists’ valuation of forest and park management features in wilderness parks of the western Canadian Shield, Canadian Journal of Forest Research, 26, 982–90. Boxall, P.C., W.H. Chan and M.L. McMillan (2005), ‘The impact of oil and natural gas facilities on rural residential property values: a spatial hedonic analysis’, Resource and Energy Economics, 27(3), 248–69. Brainard, J.S., A.A. Lovett and I.J. Bateman (1997), ‘Using isochrone surfaces in travel cost models’, Journal of Transport Geography, 5(2), 117–26. Brainard, J.S., A.A. Lovett and I.J. Bateman (1999), ‘Integrating geographical information systems into travel cost analysis and beneﬁt transfer’, International Journal of Geographical Information Systems, 13(3), 227–46. Brainard, J.S., I.J. Bateman and A.A. Lovett (2001), ‘Modelling demand for recreation in English woodlands’, Forestry, 74(5), 423–38. Brainard, J.S., A.P. Jones, I.J. Bateman, A.A. Lovett and P.J. Fallon (2002), ‘Modelling environmental equity: access to air quality in Birmingham, England’, Environment and Planning A, 34, 695–716. Brainard, J.S., A.A. Lovett and I.J. Bateman (2003), Social and Environmental Beneﬁts of Forestry, Phase 2: Carbon Sequestration Beneﬁts of Woodland, Edinburgh: The Forestry Commission. Brouwer, R. and J. Kind (2005), ‘Cost–beneﬁt analysis and ﬂood control policy in the Netherlands’, in R. Brouwer and D.W. Pearce (eds), Cost–Beneﬁt Analysis and Water Resources Management, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Burrough, P.A. and R.A. McDonnell (1998), Principles of Geographical Information Systems, Oxford: Oxford University Press. Can, A. and I. Megbolugbe (1997), ‘Spatial dependence and house price index construction’, Journal of Real Estate Finance and Economics, 14, 203–22. Chang, K. (2002), Introduction to Geographic Information Systems, New York: McGraw-Hill. Church, R.L. (2002), ‘Geographical information systems and location science source’, Computers and Operations Research, 29, 541–62. Cliﬀ, A. and J.K. Ord (1981), Spatial Processes: Models and Applications, London: Pion. Day, B.H., I.J. Bateman and I. Lake (2003), ‘Nonlinearity in hedonic price equations: an estimation strategy using model-based clustering’, CSERGE working paper EDM 04-02, Centre for Social and Economic Research on the Global Environment, University of East Anglia. Day, B.H., I.J. Bateman and I. Lake (2004), ‘Omitted locational covariates in hedonic analysis: a semiparametric approach using spatial statistics’, CSERGE working paper EDM 04-04, Centre for Social and Economic Research on the Global Environment, University of East Anglia. DeMers, M.N. (2003), Fundamentals of Geographic Information Systems, New York: John Wiley and Sons.

90

Yearbook of environmental and resource economics

Dessaint, F. and J.P. Caussanei (1994), ‘Trend surface analysis: a simple tool for modeling spatial patterns of weeds’, Crop Protection, 13, 433–7. Deutsch, E.S. (1970), ‘On parallel operations on hexagonal arrays’, IEEE Trans Comp, C-19(10), 982–3. Din, A., M. Hoesli and A. Bender (2001), ‘Environmental variables and real estate prices’, Urban Studies, 38, 1989–2000. EFTEC and CSERGE (S. Mourato, E. Özdemirogˇlu, I.J. Bateman and A.A. Lovett) (1998), ‘Valuing preferences for changes in water abstraction from the River Ouse’, report to Yorkshire Water, Bedford, by the Economics for the Environment Consultancy (EFTEC), London. Fotheringham, A.S., C. Brunsdon and M. Charlton (2000), Quantitative Geography: Perspectives on Spatial Data Analysis, London: Sage. Gamboa, M. and M.A. Santos (1996), ‘A GIS for dam safety management’, in Proceedings of the Workshop on Dams and Safety Management at Downstream Valleys, Lisbon, Portugal, 13-15 November, pp. 173–8. Gawande, K. and H. Jenkins-Smith (2000), ‘Nuclear waste transport and residential property values: estimating the eﬀects of perceived risks’, Journal of Environmental Economics and Management, 42, 207–33. Geoghegan, J., L. Wainger and N. Bockstael (1997), ‘Spatial landscape indices in a hedonic framework: an ecological economics analysis using GIS’, Ecological Economics, 23, 251–64. Griﬃth, D.A. (1996), ‘Some guidelines for specifying the geographic weights matrix contained in spatial statistical models’, in S.L. Arlinghaus, D.A. Griﬃth, W.C. Arlinghaus, W.D. Drake and J.D. Nystuen, Practical Handbook of Statistics, New York: CRC Press, pp. 65–82. Haener, M.K., P.C. Boxall, W.L. Adamowicz and D.H. Kuhnke (2004), ‘Aggregation bias in recreation site choice models: resolving the resolution problem’, Land Economics, 80, 561–74. Haining R.P. (1993), Spatial Data Analysis in the Social and Environmental Sciences, Cambridge: Cambridge University Press. Haining R.P., S.M. Wise and J. Ma (1998), ‘Exploratory spatial data analysis in a GIS environment’, The Statistician, 47, 457–69. Heywood, I., S. Cornelius and S. Carver (2002), An Introduction to Geographical Information Systems, 2nd edn, New York: Addison Wesley Longman. Hunt, L.M., P.C. Boxall, J. Englin and W. Haider (2005), ‘Remote tourism and forest management: a spatial hedonic analysis’, Ecological Economics, 53, 101–13. Ihlanfeldt, K.R. and L.O. Taylor (2001), Assessing the Impacts of Environmental Contamination on Commercial and Industrial Properties, Tallahassee, FL: Department of Economics, Florida State University. Irwin, E.G. (2000), ‘Using spatial data and methods to study rural-urban change’, paper presented at Rural Policy: Issues, Data Needs, and Data Access Conference, Washington, DC. Johnston, R.J., G. Magnusson, M.J. Mazzotta and J.J. Opaluch (2002), ‘The economics of wetland ecosystem restoration and mitigation: combining economic and ecological indicators to prioritize salt marsh restoration actions’, American Journal of Agricultural Economics, 84, 1362–70. Jones, A.P., I.J. Bateman and J. Wright (2002), ‘Estimating arrival numbers and values for informal recreational use of British woodlands’, report to the Forestry Commission, published at http://www.forestry.gov.uk. Kim, C.W., T.T. Phipps and L. Anselin (2003), ‘Measuring the beneﬁts of air quality

Geographical information systems and spatial analysis

91

improvement: a spatial hedonic approach’, Journal of Environmental Economics and Management, 45, 24–39. Lake, I.R., I.J. Bateman, B.H. Day and A.A. Lovett (2000a), ‘Improving land compensation procedure via GIS and hedonic pricing’, Environment and Planning C, 18, 681–96. Lake, I.R., I.J. Bateman, B.H. Day and A.A. Lovett (2000b), ‘Using GIS and largescale digital data to implement hedonic pricing studies’, International Journal of Geographical Information Systems, 14, 521–41. Lake, I., A.A. Lovett, I.J. Bateman and I.H. Langford (1998), ‘Modelling environmental inﬂuences on property prices in an urban environment’, Computers, Environment and Urban Systems, 22(2), 121–36. Leggett, C.G. and N.E. Bockstael (2000), ‘Evidence of the eﬀects of water quality on residential land prices’, Journal of Environmental Economics and Management, 39, 121–44. Lo, C.P. and A.K.W. Yeung (2002), Concepts and Techniques of Geographic Information Systems, New Jersey: Prentice Hall. Longley, P.A., M.F. Goodchild, D.J. Maguire and D.W. Rhind (2001), Geographic Information Systems and Science, Chichester: John Wiley and Sons. Lovett, A.A., J.S. Brainard and I.J. Bateman (1997), ‘Improving beneﬁt transfer demand functions: a GIS approach’, Journal of Environmental Management, 51(4), 373–89. Lovett, A.A., C.D. Foxall, D.J. Ball and C.S. Creaser (1998), ‘The Panteg Monitoring Project: comparing PCB and dioxin concentrations in the vicinity of industrial facilities’, Journal of Hazardous Materials, 61, 175–85. Mahan, B.L., S. Polaksy and R. Adams (2000), ‘Valuing urban wetlands: a property price approach’, Land Economics, 76, 100–13. Martin, D. and G. Higgs (1997), ‘The visualization of socio-economic GIS data using virtual reality tools’, Transactions in GIS, 1, 255–66. McGarigal, K. and B.J. Marks (1994), ‘Fragstats: spatial pattern analysis program for quantifying landscape structure’, technical report PNW-GTR351, US Dept. of Agriculture Forest Service Paciﬁc Northwest Research Station, Portland, OR. Mertens, B. and E.F. Lambin (1997), ‘Spatial modelling of deforestation in southern Cameroon’, Applied Geography, 17, 143–62. Millward, A.A. and J.E. Mersey (1999), ‘Adapting the RUSLE to model soil erosion potential in a mountainous tropical watershed’, Catena, 38, 109–29. Ministerie van Verkeer en Waterstaat (2000), ‘Anders omgaan met water: waterbeleid in de 21e eeuw. Ministerie van Verkeer en Waterstaat, Rijkswaterstaat (RWS), Hoofdkantoor van de Waterstaat (HKW) Uitg: Den Haag: Ministerie van Verkeer en Waterstaat, RWS. Mitas, L., W. Brown and H. Mitasova (1997), ‘Role of dynamic cartography in simulations of landscape processes based on multi-variate ﬁelds’, Computers and Geosciences, 23, 437–46. Monmonier, M. (1991), How to Lie with Maps, Chicago: University of Chicago Press. Pan, D., G. Domon, S. de Blois and A. Bouchard (1999), ‘Temporal (1958–1993) and spatial patterns of land use changes in Haut-Saint-Laurent (Quebec, Canada) and their relation to landscape physical attributes’, Landscape Ecology, 14, 35–52. Paterson, R.W. and K.J. Boyle (2002), ‘Out of sight, out of mind? Using GIS to incorporate visibility in hedonic property value models’, Land Economics, 78(3), 417–25. Pickle, L.W. and Y. Su (2002), ‘Within-state geographic patterns of health insurance

92

Yearbook of environmental and resource economics

coverage and health risk factors in the United States’, American Journal of Preventative Medicine, 22, 75–83. Powe, N.A., G.D. Garrod, C.F. Brunsdon and K.G. Willis (1997), ‘Using a geographic information system to estimate a hedonic price model of the beneﬁts of woodland access’, Forestry, 70(2), 139–49. Randall, T.A. and B.W. Baetz (2001), ‘Evaluating pedestrian connectivity for suburban sustainability’, Journal of Urban Planning and Development, 127, 1–15. Robinson, A., J. Morrison, P.C. Muehrcke, J. Kimerling and S.C. Guptill (1995), Elements of Cartography, 6th edn, New York: John Wiley and Sons. des Rosiers, F. (2002), ‘Power lines, visual encumbrance and house values: a microspatial approach to impact measurement’, Journal of Real Estate Research, 23, 275–301. Simpson, I., D. Parsisson, N. Hanley and C. Bullock (1997), ‘Envisioning future landscapes in the environmentally sensitive areas of Scotland’, Transactions of the Institute of British Geographers, 22, 307–20. Stocks, C.E. and S. Wise (2000), ‘The role of GIS in environmental modelling’, Geographical and Environmental Modelling, 4, 219–35. Talen, E. and L. Anselin (1998), ‘Assessing spatial equity: an evaluation of measures of accessibility to playgrounds’, Environment and Planning A, 30(4), 595–614. Tischendorf, L. and L. Fahrig. (2000), ‘How should we measure landscape connectivity?’, Landscape Ecology, 15 (7), 633–41. Tobler, W. (1979), ‘Cellular geography’, in S. Gale and G. Olsson (eds), Philosophy in Geography, Dordrecht: Reidel, pp. 379–86. Tukey, J.W. (1977), Exploratory Data Analysis, Reading, MA: Addison-Wesley. Turner, M.G., W.H. Romme, R.H. Gardnerl, R.V. O’Neill and T.K. Kratz (1993), ‘A revised concept of landscape equilibrium: disturbance and stability on scaled landscapes’, Landscape Ecology, 8(3), 213–27. Tversky, A. and D. Kahneman (1991), ‘Loss aversion in riskless choice: a referencedependent model’, The Quarterly Journal of Economics, 106(4), 1039–61. Wadsworth, R. and J. Treweek (1999), Geographic Information Systems for Ecology: An Introduction, Harlow: Longman. Walker, R., E. Moran and K. Anselin (2000), ‘Deforestation and cattle ranching in the Brazilian Amazon: external capital and household processes’, World Development, 28, 683–99. Williams, V.S. (1997), ‘Using the GSMCAD program with GPS for data collection in the ﬁeld and as a quick and eﬃcient way of creating Arc/Info geologic map coverages’, US Geological Survey open-ﬁle report 97-269, Denver, CO. Williams, V.S., G.I. Selner and R.B. Taylor (1996), GSMCAD, a new computer program that combines the functions of the GSMAP and GSMEDIT programs and is compatible with Microsoft Windows and Arc/Info: US Geological Survey Open-File Report 96-007, Denver, CO. Woolhouse, M.E.J. (2003), ‘Foot-and-mouth disease in the UK: what should we do next time?’, Journal of Applied Microbiology, 94(s 1), 126–30. Yang, W., M. Khanna, R. Farnsworth and H. Onal (2005), ‘Is geographical targeting cost-eﬀective? The case of the Conservation Reserve Enhancement Program in Illinois’, Review of Agricultural Economics, 27, 70–88. Yang, W. and A. Weersink (2004), ‘Cost-eﬀective targeting of riparian buﬀers’, Canadian Journal of Agricultural Economics, 52, 17–34. Yang, W., M. Khanna, R. Farnsworth and H. Onal (2003), ‘Integrating economic, environmental and GIS modeling to determine cost eﬀective land retirement in multiple watersheds’, Ecological Economics, 46, 249–67.

3. Disclosure strategies for pollution control Susmita Dasgupta,* Hua Wang and David Wheeler* 1.

INTRODUCTION

This chapter provides a chronicle of experience, as well as an analytical perspective. For the past decade, we have planned and implemented pollution disclosure strategies with many environmental agencies in Asia, Latin America and Africa. In the chapter, we draw on our own experience and the professional literature to address several basic questions about pollution disclosure: what is the rationale for it? How has it worked in practice? What factors have determined its success or failure? What future developments seem likely, and what complementary research agenda looks promising? We begin by deﬁning the limits of our exposition. Pollution is a very general term for byproducts of human activity that damage health and ecosystems by contaminating the air, water and soil. We focus principally on polluting emissions from factories, both because our own experience lies in this domain, and because the professional literature has provided an extensive treatment of the topic. On occasion, however, we draw on illustrative cases from a broader set of activities that generate negative social externalities. Our topic is actually broader than ‘disclosure,’ which carries a connotation of exposure or revelation. The term is commonly used only because widely disseminated public information about speciﬁc polluters is relatively new, and the ﬁrst release of such information has a revelatory aspect. To be eﬀective, however, public disclosure has to sustain itself in ongoing public information programs. Although we devote some attention to voluntary disclosure, we focus principally on government-mandated disclosure programs. We begin the chapter with a brief introduction to these programs, which have been termed the ‘third wave’ of environmental regulation.1 In section 2, we introduce the topic by summarizing the recent history of pollution disclosure programs. Section 3 discusses the rationale for pollution disclosure. Section 4 describes current disclosure programs and their 93

94

Yearbook of environmental and resource economics

results, while section 5 considers a set of important implementation issues. Section 6 considers possible future developments in this area, while section 7 provides a summary, conclusions and suggestions for future research.

2. PUBLIC DISCLOSURE: THE ‘THIRD WAVE’ OF REGULATION In discussions of national pollution regulation since the 1960s, it has become conventional to describe three ‘waves’ of development: ﬁrst, quantitative restrictions on emissions, backed by civil and criminal penalties; second, market-based instruments such as pollution charges and trading of emissions permits; third, and most recently, public disclosure of polluters’ emissions. This typology is analytically convenient, and it does convey useful information about the relative growth rates of the three approaches in the 1970s, 1980s and 1990s. In no sense, however, has each ‘wave’ swamped its predecessor. Most regulation is still quantity-based; market instruments continue to ﬁnd limited application, and public revelations about pollution have played some role in regulation for centuries.2 Nevertheless, it is reasonable to assert that public pollution disclosure has rapidly progressed since the 1980s. What was formerly the province of occasional press scandals has become a broad international movement toward government-supported or -mandated systems that monitor, evaluate and disclose polluting emissions from factories and other sources. What explains this recent development, when the technical means for implementing disclosure have been available for decades? In fact, governmental pollution disclosure began as a ‘niche’ activity in the USA, occupying a domain that could not be claimed by conventional regulators for institutional and technical reasons. By the late 1980s, two forces had merged to promote disclosure. The ﬁrst was the growth of ‘rightto-know’ laws, based on the proposition that people exposed to pollutants had the right to know the risks posed by their exposure. The second impetus was pragmatic: political leaders and regulators concluded that conventional regulatory instruments could not be applied to the hundreds of toxic pollutants whose harmful eﬀects had been documented. Faced with an obvious threat to public welfare, the federal government enacted the Toxics Release Inventory (TRI) to mandate public reporting of toxic emissions by thousands of polluting facilities. Two international trends have emerged in the wake of TRI. First, a number of countries have enacted similar programs that focus on public reporting of emissions volumes. They include Canada’s National Pollutant Release Inventory (initiated in 1992); Australia’s National Pollutant

Disclosure strategies for pollution control

95

Inventory (2000); Europe’s Pollutant Emission Register (2000); and Japan’s Pollutant Release and Transfer Register (2001). The same model has been adopted by some rapidly industrializing and former socialist economies. Examples include South Korea’s Pollutant Release and Transfer Register (initiated in 1999); Mexico’s Registro de Emisiones y Transferencia de Contaminantes (2001); and the Czech Republic’s Pollutant Release and Transfer Register (under development). Although their pollutant coverage diﬀers signiﬁcantly, these programs have a common characteristic: they report emissions, leaving the interpretation of the health and environmental implications to other government agencies or non-governmental organizations.3 Existing public emissions inventories focus mainly on toxic pollutants, which are so numerous that many remain uncovered by conventional regulations.4 The second international trend, catalyzed by collaboration between the World Bank and developing-country environment agencies, has featured a diﬀerent approach that addresses regulatory problems in developing countries. Corruption and weak enforcement have made it diﬃcult for regulatory institutions to control pollution in many of these countries. While weakness in conventional regulation clearly opens the door for community pressure via public disclosure, the appeal of emissions inventories in developing countries is limited by a second problem: public interpretation of information on emissions is hampered by generally low levels of education and a relative scarcity of technically informed NGOs that can play an interpretative role. Confronted by these problems, a number of developingcountry environment agencies have opted for environmental performance rating programs, in which they interpret the signiﬁcance of emissions as well as disclosing them. Since performance benchmarks are needed for ratings, developing-country programs have focused on conventional pollutants for which clear regulatory standards exist. Environmental performance ratings have been particularly well received in Asia. Indonesia began its PROPER program in 1995, followed by the Philippines’ EcoWatch in 1997, China’s GreenWatch (2000), and similar programs in Vietnam (2002) and India (2002, 2004). To summarize, public pollution disclosure began with emissions inventories in OECD and rapidly industrializing countries, because conventional instruments could not be used to regulate hundreds of toxic pollutants. Disclosure adapted to institutional and educational constraints in the developing countries of Asia, emerging as public performance rating programs in several countries. In 2005, after 15 years of steady growth, pollution disclosure systems are well established in countries that account for the majority of the world’s population and a very large share of world industrial production. As these systems approach ﬁrst-stage maturity, a general

96

Yearbook of environmental and resource economics

accounting is in order: what is the conceptual rationale for such systems? How well have they performed, through what channels have they aﬀected polluters, and what are the prospects for continued eﬀectiveness?

3.

THE RATIONALE FOR PUBLIC DISCLOSURE

Why should public disclosure reduce pollution? Drawing on an extensive literature, we address this question with a model of ‘equilibrium pollution’ (Baumol and Oates, 1988; Shavell, 1992; Pargal and Wheeler, 1996; Wang and Wheeler, 2003). In this model, the implicit ‘price’ of pollution is determined at the intersection of a ﬁrm’s demand and supply schedules for environmental services. The supply schedule reﬂects the terms on which stakeholders are prepared to allow the ﬁrm to pollute (or, equivalently, use local environmental services). The ﬁrm’s demand schedule, or response of its emissions to the price of pollution, reﬂects its marginal cost of abatement. The environmental supply schedule describes the relationship between the ﬁrm’s emissions and the expected penalty or compensation exacted by aﬀected agents such as regulators, representatives of neighboring communities, ‘green’ consumer organizations, liability-conscious lenders and concerned shareholders. As a factory pollutes more, aﬀected agents impose higher costs. Extensive research has investigated the roles of regulators and ﬁnancial agents in this process (Heyes, 2000; Wang and Wheeler, 2005; Harrington, 1988; Heyes and Rickman, 1999; Yaeger, 1991; Hawkins, 1983; Gray and Deily, 1996; Laplante and Rilstone, 1996; Harrison, 1995; Feinstein, 1989; Mixon, 1994; Dasgupta et al., 2004; Dasgupta et al., 2001; Hamilton, 1995; Lanoie and Laplante, 1994; Lanoie et al., 1998; Muoghalu et al., 1990; Gupta and Goldar, 2003; Hamilton, 1999). Research on the role of neighboring communities in developed and developing countries has also become quite extensive. Signiﬁcant factors that aﬀect the ‘pricing of pollution’ by local communities include income, education, level of civic activity, legal or political recourse, media coverage, NGO presence, the eﬃciency of existing formal regulation, local employment alternatives, and the total pollution load faced by the community in relation to the absorptive capacity of the local environment (Pargal and Wheeler, 1996, Blackman and Bannister, 1998; Arora and Cason, 1998; Wheeler et al., 1997; Kathuria, 2004; Hamilton, 1999; Wang and Wheeler, 2003). In the equilibrium pollution model, public disclosure aﬀects emissions because it changes the environmental demand and/or supply schedules. Although most attention has focused on the supply schedule, recent research has suggested a demand-side eﬀect as well. Conventional economic

Disclosure strategies for pollution control

97

theory assumes that the ﬁrm is perfectly informed about its own production processes, emissions and abatement investment options. However, the empirical literature on ﬁrms’ responses to public disclosure requirements suggests that the conventional model needs modiﬁcation. Without incentives to control emissions, ﬁrms have no economic reason to invest in learning about the relationships linking emissions to production and investment decisions. This changes when public disclosure alters the incentive environment by aﬀecting the environmental supply schedule faced by the ﬁrm. In previously unregulated environments, the induced learning and consequent shift in the ﬁrm’s environmental demand schedule may be particularly signiﬁcant. Such disclosure-related ‘learning shocks’ confronted toxic polluters in the USA when the Toxics Release Inventory (TRI) was initiated in 1987, and industrial polluters in Indonesia and the Philippines when public performance ratings were introduced in 1995 and 1997. Numerous case studies have documented the eﬀect of TRI, which required many ﬁrms to measure their toxic emissions for the ﬁrst time. Once these emissions were ‘priced’ by stakeholders’ response to disclosure, ﬁrms began learning how to reduce the toxic intensity of their production processes (Konar and Cohen, 1997; Gerde and Logsdon, 2001; US EPA, 2003). A similar learning process has been documented by Afsah et al. (2000) in a survey of plant managers’ responses to PROPER, Indonesia’s environment performance rating program. Anecdotal evidence from PROPER and a related program in India suggests that disclosure can also promote useful learning across ﬁrms.5 A good rating for one ﬁrm in a ﬁeld of competitors establishes the feasibility of cleaner production, encouraging the other ﬁrms to invest more resources in improving environmental performance. In the equilibrium pollution model, disclosure-induced learning produces a ﬂattening and/or leftward movement of the ﬁrm’s environmental demand schedule and a reduction in equilibrium emissions. Pollution disclosure also aﬀects the environmental supply schedule, by changing the information available to regulators, local communities, consumer organizations and market agents. The eﬀect of disclosure on the supply schedule, and the equilibrium price of pollution, depends on the direction and magnitude of the diﬀerence between ex-ante and ex-post beliefs about a ﬁrm’s emissions. Although the conventional assumption is that disclosure always entails a negative shock, this is not necessarily the case. If disclosure reveals environmental performance that is better than expected, the ﬁrm may be rewarded in product and capital markets, or by better relationships with regulators and local communities. In the converse case, of course, disclosure raises the price of current emissions by increasing penalties from regulators, communities and market agents.6

98

Yearbook of environmental and resource economics

The recent professional literature includes many papers on related topics. A theoretical analysis by Stephan (2002) provides a detailed exposition of the supply-side factors that we have identiﬁed. Disclosure decisions by proﬁt-maximizing ﬁrms are analysed by Barth et al. (1997), Cormier and Magnan (1999), Kirchhoﬀ (2000), Nash et al. (2000), Berthelot et al. (2003), Cormier and Magnan (2003), Sinclair-Desgagné and Gozlan (2003), Eisner (2004) and Gozali et al. (2002). In a representative analysis of the determinants of voluntary disclosure, Shweiki (1996) notes the positive incentive to disclose by ﬁrms that believe their environmental performance exceeds expectations.7 Shweiki suggests that regulators encourage voluntary disclosure by enacting measures that ease concerns about criminal or civil liability.

4. POLLUTION DISCLOSURE PROGRAMS AND RESULTS Although theoretical work on disclosure has burgeoned in the literature, empirical work on its impact has been hindered by the proprietary nature of plant-level data on non-environmental production factors.8 In this section, we provide brief descriptions of existing pollution disclosure programs and review what is known about their impact on emissions. 4.1

Emissions Inventories

In the OECD and rapidly industrializing economies, mandated emissions disclosure programs have all been quantitative inventories. Most focus on toxic pollutants that are not covered by conventional regulation, but some include formally regulated pollutants as well. The US Toxics Release Inventory (TRI) is probably the best known, and certainly the most-studied of these programs. Since its inception in 1987, TRI has expanded to include approximately 650 toxic chemicals that are emitted to air, water and land by US polluters. The system includes reports from about 20 000 industrial facilities. Between 1988 and 1999, the US EPA reported an overall decrease of 43 percent in national releases of toxics reported under TRI.9 Since trend data prior to TRI’s inception are not available, we cannot infer that TRI is responsible for the entire reduction in emissions.10 Nor, since toxic chemicals diﬀer by orders of magnitude in their relative hazard intensity, can we infer that the 43 percent reduction produced a corresponding reduction in exposure risk.11 The evidence on TRI’s impact is mixed in the empirical literature. Hamilton (1999) ﬁnds that TRI disclosures reduce emissions more rapidly in plants for which toxic hazard or aﬀected community characteristics are consistent with signiﬁcant inward movement of the environmental supply

Disclosure strategies for pollution control

99

schedule in the equilibrium emissions model. For example, TRI disclosure is associated with relatively large emissions reductions in communities with relatively high levels of income, education and documented political activity. However, Bui and Mayer (2003) ﬁnd little evidence that TRI-reported hazards have reduced house prices in aﬀected communities. Since property values should discount the eﬀects of known hazards, Bui and Mayer suggest that the impact of emissions inventories may be reduced by the public’s inability to process complex information on hazardous emissions. Canada’s National Pollutant Release Inventory (NPRI) was established in 1992, and currently covers 323 toxic chemicals. In 2002, 4530 facilities reported toxic emissions under this program.12 In an analysis of trends since 1997, the 2000 NPRI found that total releases of 17 toxic substances were up slightly (4.5 percent). Of the 17 substances, emissions decreased for ﬁve and increased for seven, while ﬁve showed little trend. Australia’s National Pollutant Inventory (NPI) is based on the US TRI, and was ﬁrst published in 2000. It currently reports emissions of 90 toxic substances from 3400 facilities. A trend analysis in 2004 reported that of the 90 substances, 50 had reduced emissions since the inception of NPI.13 Several other emissions inventories have been established in recent years: the European Pollutant Emission Register currently covers emissions of 50 toxic and non-toxic pollutants from about 10 000 industrial facilities in 16 European countries.14 Japan’s Pollutant Release and Transfer Register covers emissions of over 350 toxic chemicals from 34 517 polluters.15 South Korea’s Pollutant Release and Transfer Register covers 146 toxic chemicals emitted by about 1200 facilities.16 Mexico and the Czech Republic are both developing emissions inventories, and Mexico produced its ﬁrst state-level summary of results (from Aguascalientes) in 2003.17 Outside the USA, empirical research on the impact of public pollution inventories is sparse. Controversy about the eﬀectiveness of the US TRI persists, although most published research supports the view that its impact on toxic emissions has been signiﬁcant. Elsewhere, programs are either too new or too sparsely documented for any ﬁrm conclusions to be drawn. 4.2

Environmental Performance Ratings

Several national and state-level government agencies in Asian developing countries have implemented an alternative disclosure model that operates through performance ratings rather than publication of emissions data. Current programs include Indonesia’s PROPER, the Philippines’ EcoWatch, China’s GreenWatch, Vietnam’s Environmental Information Disclosure System for Hanoi, and two industrial environmental performance rating systems in Uttar Pradesh State, India. All of the Asian programs

100

Yearbook of environmental and resource economics

have adopted the basic performance rating model pioneered by Indonesia’s PROPER system in 1995. These models all include the following elements: 1.

2.

3.

4.

5.

They focus on conventionally regulated pollutants, rather than toxic chemicals, because they are intended to complement traditional regulatory instruments that have not brought most factories into compliance with local emissions standards. They adopt locally recognized benchmarks for grading performance. In most cases, local emissions standards and environmental reporting requirements provide the criteria for acceptable performance, and ISO 14000-type standards provide the criteria for excellent performance.18 For covered pollutants, they compare audited factory emissions reports with benchmark standards and grade the results in a few overall categories: above-standard (signiﬁcantly better than legal requirements, often in two classes, with the highest class conforming to ISO 14000 requirements); acceptable (compliance with all legal requirements); and below-standard (generally in two classes: non-compliance with some requirements, but not ﬂagrant; ﬂagrant non-compliance). They assign locally understood color codes or other graphic symbols to performance grades, so that the results can be easily disseminated by the media and understood by the public. The practice has been to avoid controversy by restricting higher ratings to ﬁrms that are in compliance with regulations for all covered pollutants. While this approach avoids accusations of ‘tolerating pollution’ that could undermine public conﬁdence in the ratings, it has the unavoidable side-eﬀect of restricting possible trade-oﬀs based on relative abatement costs (i.e., factories cannot compensate for non-compliance in the case of a high-abatement-cost pollutant by exceeding compliance norms for a low-abatement-cost pollutant). The performance rating agencies maintain close communication with audited facilities throughout the process. Typical modes of contact include pre-audit meetings for explanation of the system; post-audit communication of preliminary ratings to factories for comments; provision of a grace period before public disclosure, during which companies can attempt to improve performance; and detailed explanation of the ﬁnal ratings to audited facilities, with explicit suggestions for improving their ratings in the next disclosure round.

Indonesia’s environment agency initiated this model by instituting the PROPER system in 1995, assigning color codes to denote performance in ﬁve categories: world-class (gold), above-standard (green), compliant (blue), non-compliant (red), and ﬂagrantly non-compliant (black). Table 3.1

101

Disclosure strategies for pollution control

presents PROPER’s initial ratings for 146 large-scale polluters in 1995, along with ratings two years after disclosure. Since its inception, PROPER’s coverage has expanded to several hundred polluters.19 Performance rating systems are intended to provide incentives for improvement, since they reward above-standard performance as well as punishing below-standard performance.20 However, the Indonesian results suggest that the stick dominated the carrot during the ﬁrst phase of implementation. The main change in Table 3.1 is a signiﬁcant movement away from red to blue, signaling a strong interest in achieving basic compliance with regulations. However, the few ﬂagrantly non-compliant polluters remained black (and their number increased by one, after a facility was downgraded from red to black because its dishonest reporting was exposed by the neighboring community). Similarly, two ﬁrms improved to green status but none proved willing or able to upgrade to gold. These results are in line with what we know about marginal abatement costs: once PROPER applied public pressure, many non-compliant ﬁrms found that they could upgrade one step, to basic compliance, without incurring huge costs. Faced with substantially higher incremental costs and no deﬁnite reward for achieving red status, the black-rated plants retained their outcast status. Similarly, few blue plants proved willing to incur the cost of upgrading to green or gold status, at least during the ﬁrst two years of program implementation. The Philippines’ environment agency followed the PROPER experiment closely, and adopted a nearly identical model for its EcoWatch program in 1997. Since its pilot phase in 1997, EcoWatch has expanded to a mandated national program for thousands of factories. Table 3.2 provides evidence on the impact of EcoWatch during its ﬁrst implementation phase. These results look very similar to the initial results for PROPER: no movement out of black, no movement into green or gold, but very signiﬁcant Table 3.1

1995 1997

Table 3.2

1997 1998

Indonesia’s PROPER results, 1995–97 Black

Red

Blue

Green

Gold

Total

2 3

90 54

47 80

7 9

0 0

146 146

The Philippines’ EcoWatch results, 1997–98 Black

Red

Blue

Green

Gold

Total

9 10

39 9

4 26

0 0

0 0

52 45

102

Table 3.3

Yearbook of environmental and resource economics

China’s GreenWatch results, 1999–2000 Black (%)

Zhenjiang, Jiangsu Province (%) 1999 2000 Hohhot, Inner Mongolia (%) 1999 2000

Red (%)

Yellow (%)

Blue (%)

Green (%)

Total facilities 91

11 3

14 12

44 23

28 61

3 1 56

11 5

66 33

19 47

4 15

0 0

movement from red to blue. Again, the initial impact of the program was a large shift from non-compliant to compliant status, probably because many factory managers found that they could achieve blue status at modest cost.21 Shortly after the Philippines adopted EcoWatch, China began experimenting with a similar model in GreenWatch. The national environment agency co-sponsored pilot programs with local authorities in Zhenjiang, a city in China’s relatively developed coastal province of Jiangsu, and Hohhot, a city in Inner Mongolia, a relatively poor interior province. Since its inception in 1999, GreenWatch has expanded to cover approximately 5000 factories in three provinces: Jiangsu, Anhui and Inner Mongolia. The national environment agency is currently considering a plan to mandate full national coverage during the next two years. Table 3.3 provides information on results for GreenWatch in Zhenjiang and Hohhot. China’s color-coding diﬀers somewhat above black and red, as follows: yellow (compliant); blue (above-standard); green (world-class). Table 3.3 results are presented in percentage form, to facilitate comparison between the two cities. The initial GreenWatch results appear stronger at the low end than those for Indonesia and the Philippines. While the same broad movement from non-compliant to compliant status is apparent, there is also signiﬁcant movement out of black. In addition, many factories moved from compliant (yellow) to above-standard (blue) status, although world-class status clearly remained diﬃcult to achieve.22 Table 3.4 condenses the evidence in Tables 3.1–3.3 into two compliance categories, adding comparable data from the Vietnam performance rating program. In all four countries, performance rating systems were intended to strengthen, but not replace, weak conventional regulatory systems. The tabulated results suggest that initial concerns about compliance were well founded.

103

Disclosure strategies for pollution control

Table 3.4

Performance rating programs: changes in compliance status

Country/ program

Factories by compliance status

% compliance status

Compliance Total factories change rated

Non- Compliant % non- % compliant compliant compliant Indonesia PROPER 1995 1997 Philippines EcoWatch 1997 1998 Vietnam EIDS Hanoi 2001 2002 China Green Watch Zhenjiang 1999 2000 China Green Watch Hohhot 1999 2000

92 57

54 89

63 39

37 61

24%

146 146

48 19

4 26

92 42

8 58

50%

52 45

45 38

5 12

90 76

10 24

14%

50 50

23 14

68 77

25 15

75 85

10%

91 91

43 21

13 35

77 38

23 62

39%

56 56

Before implementation of performance rating, compliance rates were 37 percent in Indonesia, 8 percent in the Philippines, 10 percent in Vietnam, 75 percent in Zhenjiang, China and 23 percent in Hohhot, China. After implementation of performance ratings, the compliance rate increased by 24 percent in Indonesia, 50 percent in the Philippines, 14 percent in Vietnam, 10 percent in Zhenjiang, China (from a high base), and 39 percentin Hohhot, China. In light of the evident regulatory problems in all four countries, these improvements suggest that performance ratings had a very signiﬁcant eﬀect on polluters. The results are also compatible with results from market event studies in four developing countries – Argentina, Chile, Mexico and the Philippines – that suggest strong stock market sensitivity to environmental news in the media (Dasgupta et al., 2004; Dasgupta et al., 2001; Gupta and Goldar, 2003). In fact, the market responses in these countries are much larger than those reported for US and Canadian ﬁrms: Stock price gains average 20 percent in response to good news and losses range from 4–15 percent in the wake of bad news.

104

Yearbook of environmental and resource economics

After nearly a decade of implementation, environmental performance ratings appear to have had a signiﬁcant, consistently positive impact on regulatory compliance in several large Asian countries. Important interpretative questions remain, however. First, how robust are these results? As in the case of emissions inventories, impact analysis is hampered by the absence of information on prior emissions, compliance trends and other production factors at the plant level. In Indonesia and the Philippines, anecdotal evidence suggests that prior compliance rates were stagnant at low levels, and that subsequent emissions reductions were in fact responses to performance ratings. The only systematic survey evidence has been gathered by Afsah et al. (2000), who interviewed a large sample of factory managers about their responses to Indonesia’s PROPER. Their results strongly suggest that PROPER provided the impetus for rapid changes by many polluters. The situation in China is less clear-cut, since conventional regulation was more strongly established there, and other programs for emissions reduction were implemented concurrently with GreenWatch. Initial conditions in Vietnam and India seem closer to those in Indonesia and the Philippines, but the evidence about ﬁrms’ responses remains anecdotal at this point. If we attribute the observed improvements to public performance ratings, through what channels did these ratings aﬀect polluters? In our previous discussion of ‘equilibrium pollution,’we identiﬁed four agents whose responses aﬀect the environmental supply schedule faced by the polluter: regulators, neighboring communities, consumers and market agents. Financial-market event studies have suggested high sensitivity to adverse environmental disclosures in both developed and developing countries, and at least one study (Konar and Cohen, 1997) has suggested that market sensitivity has prompted ﬁrms to reduce emissions after public disclosure. We have no equivalent evidence for performance ratings, but this link certainly seems plausible. Numerous cross-section studies by the present authors and others have established an empirical link between community characteristics and the environmental performance of neighboring factories, suggesting that community feedback aﬀects polluters even when formal regulation is absent or ineﬀective. The community action link to public disclosure is strongly supported by Afsah et al. (2000), whose survey of factory managers identiﬁes local communities as the strongest external source of pressure to reduce emissions after disclosure. Regulators themselves may also play some role, since bad performance ratings focus public and regulatory attention on ﬂagrantly non-compliant ﬁrms. However, the results for Indonesia and the Philippines in Tables 3.1 and 3.2 suggest that this factor may have modest importance. In both countries, the worst-rated polluters remained in that status throughout the initial implementation period.

Disclosure strategies for pollution control

105

In our discussion of equilibrium pollution, we have also considered the potential impact of disclosure on the polluter’s environmental demand schedule. If prior regulation has been weak or absent, polluters have had no previous incentive to invest in learning about the emissions characteristics of alternative technologies. By administering a ‘learning shock,’ disclosure may induce such learning by polluters, with a consequent change in the environmental demand function and reduced equilibrium emissions. Relevant evidence is scarce for developing countries, but the PROPERrelated survey of Afsah et al. (2000) again provides strong support. In the survey, factory managers identiﬁed induced learning as the primary determinant of reduced emissions, along with community pressure.

5. IMPLEMENTATION ISSUES FOR PUBLIC DISCLOSURE As we have learned in our own work with environmental agencies, public disclosure challenges conventional regulatory thinking because it operates through perceptions and diﬀuse channels of inﬂuence. Research has shown that regulators’ interactions with ﬁrms are far from determinate, because enforcement of regulations is frequently tempered by negotiations over deadlines and applicable emissions standards (Gray and Deily, 1996; Hawkins, 1983; Harrington, 1988; Harrison, 1995; Heyes, 2000; Heyes and Rickman, 1999; Wang and Wheeler, 2005; Yeager, 1991). However, these interactions are more clearly deﬁned and understood than the myriad potential interactions that aﬀect the ﬁrm’s environmental supply schedule under public disclosure. In this section, we consider a number of issues that make regulation via public disclosure as much an art as a science. We begin with common problems for emissions inventories and performance ratings, and then consider problems that are unique to the latter. 5.1 Common Problems for Emissions Inventories and Performance Ratings The ﬁrst common problem is selection of pollutants subject to the reporting requirements. Although most emissions inventories have been conﬁned to toxics in industrialized countries, there is no technical reason for this limitation. As the European Pollutant Emission Register system has demonstrated, public emissions inventories can also be used as complements to conventional regulation for many pollutants. Conversely, ﬁrstgeneration performance rating systems have been largely conﬁned to conventional pollutants because existing regulations provide compliance

106

Yearbook of environmental and resource economics

benchmarks, even if they have been weakly enforced. However, there is no reason in principle why benchmarks could not be developed for previously unregulated toxic pollutants. Ideally, selection of pollutants should be grounded in a rigorous evaluation of potential beneﬁts (from risk studies) and abatement costs, as well as the relative costs of monitoring and evaluation of monitoring evidence. In reality, however, the public nature of disclosure introduces factors related to community and media perceptions. Disclosure’s eﬀect depends heavily on whether its information is valued by the agents who aﬀect the ﬁrm’s environmental supply schedule. In this domain, disclosure for visibly hazardous, well-understood pollutants is likely to have more impact than disclosure for substances that are known to scientists but not to the public. Emissions inventories have an additional burden in this context that is not shared by performance ratings, since the latter summarize complex information in simple, color-coded ratings. By contrast, even the simplest emissions inventories cover hundreds of toxics that are little known to the public. In the absence of credible governmental interpretation, the media and the public may have a double problem: inability to sort through myriad risk factors, and unwillingness to bank on interpretations by environmental NGOs that have a clear interest in maximizing public pressure on all polluters. The previously cited ﬁndings of Bui and Mayer (2003) suggest that proliferation of covered pollutants may weaken the impact of emissions inventories. For countries that have not yet begun such inventories, this highlights the importance of strategic choice in coverage selection. Related to the coverage problem is the potential problem of legal liability. Initially, the designers of emissions inventory programs were worried that forced disclosure of many chemical emissions would inadvertently reveal trade secrets and prompt a wave of lawsuits that would undermine the programs. Somewhat surprisingly, such lawsuits have been very rare. However, our own private communications with staﬀ members of the US EPA have suggested other concerns that have strongly limited the scope for public disclosure, at least in the USA. Under existing law, aside from proprietary technology disclosure problems, TRI is not subject to legal challenge by private ﬁrms as long as it simply audits and publishes the ﬁrms’ own emissions reports. However, TRI could immediately be challenged on scientiﬁc grounds if it included estimates of potential health or economic damage as part of its information provision to the public. Highly reliable evidence on short- and long-term hazards remains scarce for the hundreds of toxic chemicals covered by TRI. The system’s developers have chosen to avoid explicit risk judgments for reporting facilities, since these would undoubtedly provoke systemcrippling legal challenges by ﬁrms with high hazard ratings. For the same

Disclosure strategies for pollution control

107

reason, EPA has never implemented a performance rating system, although staﬀ members have privately expressed strong interest to the authors of this chapter. 5.2

Additional Problems for Environmental Performance Ratings

Performance rating systems are based on benchmarks from which the ratings agency judges the relative performance of polluters. Obvious candidates for benchmarks include national emissions standards, formal reporting requirements, and ISO 14000 criteria. First-generation performance rating systems have been conﬁned to conventionally regulated pollutants, mostly because this allows them to use national emissions standards as benchmarks. Inevitably, the choice of benchmarks reﬂects an implicit assessment of beneﬁts and costs. Performance rating systems aim to reduce pollution by providing incentives to improve environmental performance over time. To induce meaningful change, however, they must strike a balance: reliance on overly strict national regulatory standards that ignore abatement costs is a recipe for failure, if political leaders conclude that compliance will be tantamount to bankruptcy for many ﬁrms. By implication, eﬀective benchmarks should encourage improvement by identifying a substantial percentage of facilities as sub par, while demonstrating that comparable facilities have already achieved acceptable performance. Indonesia provides a good example of eﬀective benchmarking, based on emissions standards that are in line with local capabilities and abatement costs. Adopting these standards for benchmarking, PROPER identiﬁed 54 percent of reporting factories as non-compliant in 1995, leaving scope for improvement while providing a strong demonstration eﬀect from substantial numbers of compliant ﬁrms. Local benchmarks are highly relevant for local agents who are consumers of public disclosure information. However, the case may be diﬀerent for international agents (ﬁnancial institutions, ‘green’ consumer organizations, etc.). Local regulatory benchmarks may be acceptable to such agents if they are primarily interested in local regulatory compliance. Financial institutions, for example, may focus on limitation of reputational or ﬁnancial liability from association with ﬁrms whose emissions do not comply with local legal codes. However, other international agents (green consumer groups, etc.) may have universal standards in mind when they judge environmental performance. For these consumers of disclosure information, compliance with local regulations may be insuﬃcient if the regulations are judged to be too lax. In practice, this does not seem to pose a major problem for performance ratings, because most national standards reﬂect developing-country norms established by the United Nations Environment Program.

108

Yearbook of environmental and resource economics

A more serious problem may arise for performance ratings in countries with great regional diﬀerences in development levels. China provides an illustration, since some of its coastal provinces are considerably more advanced than some interior provinces. In planning for national implementation of performance ratings, China’s GreenWatch program is currently struggling with the problem of national benchmarks in this context. If the benchmarks are set to pressure coastal factories for change, they may be unrealistically high for many factories in the interior. On the other hand, realistic benchmarks for interior facilities may provide insuﬃcient pressure for coastal facilities. Another problem for performance ratings concerns the orchestration of public relations. Eﬀective choices of dissemination media and modes of presentation depend on local capabilities and customs.23 Since the media are always attracted to novelty, the timing of disclosure is also important. In Indonesia, the environment agency has sustained media interest in PROPER by disclosing for one industry sector at a time, leaving fairly extended periods between disclosures. The length of periods between disclosure rounds is also important in this context. Disclosure every year may be too disruptive for industry, but disclosure every ﬁve years may allow industry too much of a ‘cooling-oﬀ’ period between bouts of media pressure. Public credibility is another important issue for performance rating systems. In principle, ratings disclosure can be accompanied by full disclosure of the supporting information. In the Philippines, existing right-toknow laws are compatible with this strategy. In other countries, however, the legal system may impose conﬁdentiality requirements that prevent release of the underlying data. In such cases, how can the rating agency preserve credibility? Communities near polluters provide the ﬁrst check, since they have many ways of detecting errors in the ratings.24 However, full credibility for an ‘inside’ ratings process warrants some form of auditing by outsiders who are trusted by the community. In Indonesia, this role is played by a committee of auditors who are respected academics, NGO representatives, and staﬀ members from government agencies that are not connected to the environment agency. Performance rating systems also impose requirements on environment agencies that must be faced realistically. All current rating systems are in developing countries, where the implementing agencies suﬀer from a variety of institutional weaknesses. Although some proponents of public disclosure have argued that disclosure avoids costs associated with conventional regulation, we believe that this proposition is overstated. A serious public disclosure program requires strict attention to monitoring, data entry, data analysis, and accurate reporting to ﬁrms and the public. In fact, the information requirements of a public disclosure program are very

Disclosure strategies for pollution control

109

similar to the requirements of eﬀective conventional regulatory programs that focus on emissions.25 What is distinctive about disclosure is the level of information integrity that it forces the environment agency to maintain. If public ratings prove faulty because the information has been mishandled, the program will rapidly lose credibility and eﬀectiveness. Such discipline is an extremely powerful driver for institutional reform, but it does not come easily to agencies that have traditionally been sealed oﬀ from public scrutiny. Realistically, implementing agencies have to learn the disclosure business in progressive steps, starting with limited pilot programs, and extending coverage as institutional capability improves.

6.

POSSIBLE FUTURE DEVELOPMENTS

The international trend toward public disclosure of industrial pollution is part of a general movement toward public rating of the social performance of private and public agents. Remarkably, the evidence we have cited suggests that the trend is nearly as pronounced in developing countries as in industrial societies. In this chapter, we have discussed the rationale for pollution disclosure, its recent history in developed and developing countries, and some design issues for disclosure programs. After over a decade of conceptual, empirical and implementation-related work on pollution disclosure, we believe that its genesis and rapid growth can be explained in relatively simple terms. The key element, in our view, is the same interplay between evolving technologies and tastes that has sparked rapid product diﬀerentiation in the consumer sector: ‘Standard brands’ have diversiﬁed enormously, as consumers and producers have responded to the reduced incremental cost of variety made possible by new information and communication technologies. Now we are witnessing the same development in the public sector. The trend is somewhat less evident, however, because legal norms still require uniform design, interpretation and application of regulatory instruments. In the conventional regulatory domain, one size continues to ﬁt all, at least within broadly deﬁned categories (for example, variations in the strictness of pollution regulation are legally permitted across air- and watersheds in some cases, but non-uniform application within those areas would be considered prejudicial). However, citizens who inhabit areas covered by uniform regulations may have great diﬀerences in desired levels of environmental quality and willingness to pay for it. Public information about polluters has become popular so quickly because it ﬁlls the same gap as product diﬀerentiation in the private sector. Regulatory agencies in a democracy tend to implement and enforce regulations that reﬂect the

110

Yearbook of environmental and resource economics

preferences of the median voter. If regulation is the only instrument for changing polluter behavior, then overall environmental quality will also reﬂect median preferences and willingness to pay. Public information provides an additional instrument for individuals or groups who would prefer less pollution than the level allowed by conventional regulatory practice. They wield this instrument through many channels, including stockholder meetings, local boycotts, demonstrations or news stories, pressure on lenders to deny ﬁnancing, and moral suasion directed at employees through local actions and professional associations. To summarize, the evidence suggests that public disclosure has created a potent form of ‘product diﬀerentiation’ in the public sector. Using public information, multiple agents, with multiple objectives, have joined regulatory agencies in inﬂuencing the behavior of polluters in countries as varied as China and Canada. Important results have included rapid movement toward compliance with regulatory standards in countries where regulatory agencies are weak, and common improvements beyond compliance standards in countries where the agencies are strong. How will this trend manifest itself in the future? If present trends continue, most polluters of signiﬁcant size in all countries may soon be ‘barcoded’, and information about their economic activities and comparative environmental performance may be readily available on the web. If this occurs, the globalization of regulatory product diﬀerentiation will be felt with full force. The contours of this future can already be discerned in sectors where international public scrutiny is farther advanced. An example is provided by Nike which, under pressure from NGOs about its labor practices in poor countries, has taken extraordinary measures to support thirdparty auditing and public rating of its factories on a website.26 Other producers have been forced to follow suit, at least to some degree, in order to preserve their brand reputations and market shares. As this trend continues, we are witnessing the advent of ‘virtual regulation’ via the Internet. As it progresses, we can expect an ampliﬁcation of eﬀorts by Western NGOs and labor unions to promote improved environmental performance in Third World factories. In the current negotiating climate, environmental provisions of free trade agreements are largely limited to general provisions for monitoring and encouraging compliance with environmental regulations in the signatory countries. With the advent of international factory ‘bar coding’, however, it will become possible for governments to get much more speciﬁc about non-compliance. In one possible future regime, weak regulation in developing countries could be supplemented by strict regulation at international borders. Importing countries could levy compensatory charges on speciﬁc non-compliant factories, rebating the charges to governments in the exporting countries.27

Disclosure strategies for pollution control

111

Even if national governments refuse to intervene, environmental and consumer groups will become increasingly capable of identifying products from speciﬁc non-compliant factories as targets for avoidance in the market.28 In this realm of virtual regulation, ﬁnancial institutions, professional societies and other institutions will become natural focal points for pressure on the managers and employees of non-compliant polluters. On the other hand, virtual regulation may prove to be a chimera if it places too much reliance on the eﬀorts of private agents. One beneﬁt of formal, governmental regulation is its exploitation of scale economies and operational comparative advantage. Government regulators are hired full-time to focus on compliance, while citizen eﬀorts in the virtual domain remain largely uncompensated (at least directly). Virtual regulation may therefore lack the consistency of formal regulation, as well as posing the risk that beneﬁts will ﬂow disproportionately to well-organized pressure groups. In addition, the proliferation of information about polluters may have the same eﬀect as the proliferation of cable channel oﬀerings has had on television viewers: a form of information overload and fatigue, in which too many social and environmental performance ratings vie for the limited attention of the public. As advertisers have discovered, increasing stimulation may be needed to get the same reaction. Ultimately, familiarity may breed passivity in the virtual regulatory domain. To circumvent this problem, global promoters of virtual regulation may follow the practice of the commercial media, which maintain consumer interest by tailoring their products ever more narrowly as information bandwidth grows. Although global public disclosure will remain public, its most powerful impact may come via dissemination in specialized channels, for interested regulators, communities and market agents.

7.

SUMMARY AND CONCLUSIONS

In this chapter, we have presented a perspective on pollution disclosure which reﬂects our experience as both researchers and practitioners. When we began to develop these ideas and seek country partners in 1992, we had no idea whether such programs would be acceptable, feasible and eﬀective in developing countries. From the outset, we have been consistently surprised by the rate at which these programs have actually developed. After hundreds of presentations in dozens of countries, we have yet to encounter a hostile audience. Numerous countries in Latin America and Africa have experimented with disclosure, although the primary domain for development to date has been in Asia. In ten years, public environmental performance ratings have moved from concept to eﬀective implementation in Indonesia, the Philippines, China, Vietnam and India – countries that, together,

112

Yearbook of environmental and resource economics

account for the majority of the population and industrial capacity in lowincome countries. We are optimistic about the future of such programs, but we have identiﬁed a number of implementation problems that remain diﬃcult to handle. In closing, we would like to discuss two additional issues that may be of interest to readers. First, for those who have an interest in promoting sustainable development, a pragmatic question: how can international assistance promote the development of institutions that can implement such programs eﬀectively? The evidence on how regulatory capability can be developed is sparse, but the World Bank’s indicators of institutional and policy development provide grounds for some optimism. Productive public policy is clearly correlated with economic development, but the evidence indicates considerable variation in the relationship. Some excellent economic performers have very low per capita incomes, because their adoption of growth-promoting policies is relatively recent (examples include Mongolia and Uganda). While general policy indicators predict environmental policy performance very well, some countries with low overall policy ratings have proven capable of focused eﬀorts to protect critical environmental assets. The most pronounced outliers are often countries where speciﬁc natural resources are important determinants of tourist revenue, such as the Maldives, the Seychelles, Belize, Ecuador and Bhutan. Apparently, even poorly administered societies can strengthen regulation when environmental damage is clear, costly and concentrated in a few sites. But these exceptions aside, it seems unlikely that broader environmental regulation will outpace more general institutional reform. A full response to the challenge of promoting disclosure will therefore require serious attention to long-run development of public-sector administrative and decision-making capacity. We believe that the international community can play a valuable role by ﬁnancing appropriate training, policy reforms, information-gathering and public environmental education. We would also like to identify topics for further research in this area. As we have noted in the chapter, initial results from both emissions inventories and performance ratings appear promising but far from conclusive. In the relatively data-rich environments of the OECD countries, there is clearly scope for more empirical work on the impact of emissions inventories on polluters and public health. The prospects may be particularly bright in the European Union, since the European Pollutant Emission Register covers conventionally regulated pollutants whose emissions have been recorded by national environment agencies for decades. The EU experience oﬀers good prospects for a rigorous panel study of disclosure as a large-scale policy experiment, with a clearly deﬁned initiation date, plentiful ex ante and ex post data on factories’ environmental performance, and the potential

Disclosure strategies for pollution control

113

existence of a control group of factories that have not been covered by the disclosure system. Only China oﬀers some prospect of a comparable study for developing countries, since its conventional regulatory system gathered emissions data from GreenWatch factories before disclosure, and many conventionally regulated plants have not been covered by GreenWatch during its initial implementation. Accessing such data would be quite diﬃcult in China, however. At a more modest level, the advent of public information in disclosure systems creates the potential for a great deal of interesting empirical work in all the countries where performance ratings have been implemented. As the ratings systems continue to generate relative performance information, questions to be explored include: what is the diﬀerence between the shortand long-term impacts of disclosure? Does disclosure produce sustained improvements, or do ﬁrms tend to regress after the initial ‘learning shock’? How do diﬀerences in ﬁrm and neighboring-community characteristics aﬀect the response to disclosure? A related topic for theoretical and empirical research is the scope for complementary implementation of disclosure and other regulatory instruments. In the emerging regulatory environment, the conventional regulator represents the median voter while public disclosure implicitly enfranchises other agents whose interests may diverge from median preferences. We know very little about the theoretical properties of this dual system: under what conditions are its multiple incentives compatible or divergent? How have polluters and regulators actually responded to the simultaneous implementation of conventional and disclosure-based systems? Here again, the emissions information provided by the European Pollutant Emission Register oﬀers a potentially powerful resource for comparative research. Ultimately, we believe that the most interesting and challenging new work in this area may join with research on other forms of ‘virtual regulation.’ Regulatory monitoring and enforcement have traditionally been considered public-sector monopolies, and most of the public policy literature remains focused on optimal policies for government agents. However, the advent of virtual regulation implies a diminishing (albeit necessary) role for public agents in the total regulatory environment. In the new regime, how can public and private agents most productively leverage their own and their counterparts’activities? What can virtual regulators learn from advertisers, who have worked on this terrain for many years? Under what conditions should governments subsidize virtual regulators in the private sector, and when should government focus on maintaining or expanding its own regulatory capacity? In this new domain, the set of potentially interesting research questions is obviously quite extensive. Because pollution disclosure is already well

114

Yearbook of environmental and resource economics

advanced, and its results relatively well documented, we believe that research in this area can provide important insights for researchers in related ﬁelds as the domain of virtual regulation continues to expand.

NOTES *

1. 2.

3.

4. 5. 6.

7. 8.

The authors wish to acknowledge an enormous debt to many colleagues in the World Bank and their partner countries who have contributed to the work reviewed in this chapter. The ﬁndings, interpretations and conclusions are entirely those of the authors, and do not necessarily represent the view of the World Bank, its executive directors, or the countries they represent. Thanks also to the editors and reviewers for their useful comments on a previous draft. The ﬁrst and second waves have been quantitative and market-based regulation, respectively. For further discussion, see Tietenberg and Wheeler (2001). Most famously, the stench of the Thames in 1858 drove the British Parliament to enact sewerage charges and construction of the city’s ﬁrst treatment system (Glick, 1980). Histories of environmental protection in the USA frequently cite the impact of news about the Donora Fog, a severe air pollution event in Donora, Pennsylvania that killed 20 people and hospitalized over 7000 in 1948 (Davis, 2002). In Brazil, a similar role was played by the Cubatao Valley pollution crisis of January 1985. During one 48-hour period, 15 inches of rain poured onto hillsides that had been denuded by air pollution from the valley’s unregulated industrial complex. Hundreds of mudslides cascaded into the valley, and one broke a large ammonia pipeline, releasing gas that injured many residents and forced a mass evacuation. Faced with widespread media attention and public criticism, São Paulo State declared an emergency and mandated forceful action by CETESB, the state’s pollution control agency (CETESB, 1986, 1990). For an eﬀective example of NGO interpretation of TRI results, see Environmental Defense’s Scorecard website at http://www.scorecard.org. Although the US EPA does not rate polluters explicitly, it does provide information on toxic hazard weights that can be used for interpretation (US EPA, 2002). For an assessment of the EPA’s methodology, see Toﬀel and Marshall (2004). Over time, formal regulations are being extended to more toxics. In the USA, for example, the Clean Air Act now requires the US EPA to regulate toxic air emissions from 82 categories of major industrial sources, as well as solid waste combustion facilities. The Indian case, co-developed by one of the authors with Indian colleagues, involves a cooperative disclosure agreement in Uttar Pradesh State between local distilleries and the State Pollution Control Board. Increased penalties will reduce emissions, and the sectoral and locational pattern of pollution reduction will reﬂect the relative ability of interested agents to apply pressure. The results will not guarantee ‘optimal’ pollution reduction from a social welfare perspective, since the process will not be administered by a central agent that could, at least in principle, weigh overall marginal beneﬁts and costs. In practice, public disclosure seems to have arisen as a complement to conventional regulation when the latter has been judged insuﬃcient to attain consensus environmental norms. Among developing countries, public disclosure programs have been successfully implemented by both authoritarian and democratic governments. In section 5, we discuss the role of expectations in the design of appropriate benchmarks for performance rating systems. Such data are collected by national census bureau, but are only made available to researchers under special provisions. For example, the authors’ own research on industry– environment links at the US Census Bureau required special clearances and nondisclosure agreements. Such restrictions shield empirical research on this topic from the standard scholarly requirement of third-party replicability.

Disclosure strategies for pollution control 9. 10.

11.

12. 13. 14. 15. 16. 17. 18.

19.

20. 21.

22. 23. 24.

25.

115

The evidence suggests that the decline in toxic releases has not been accompanied by a matching decline in generation of toxics. The decline seems largely due to recycling, rather than conversion to production using safer chemicals. Rigorous empirical analysis of policy interventions requires ex-ante and ex-post data on performance, for both experimental and control groups. Programs such as TRI cannot provide such information, since they mandate compliance by all ﬁrms in a setting where ex-ante performance data are unavailable. The authors are presently collaborating on an experimental public performance rating program in China that incorporates all elements needed for a rigorous empirical analysis. The results are not yet available. In this context, we should note that the test will not necessarily be relevant for judging emissions inventory programs such as TRI, which report emissions without interpreting them. Since 1988, some high-volume pollutants such as sodium hydroxide have been removed from reporting requirements because they have been judged insuﬃciently hazardous. Reductions in sodium hydroxide accounted for a signiﬁcant percentage in early toxic reductions credited to TRI. Source: http://www.ec.gc.ca/pdb/npri/2002Highlights Source: Media Release by Dr David Kemp, Minister for the Environment and Heritage, 30 January 2004. Source: http://eper.cec.eu.int/eper Source: http://www.env.go.jp/chemi/prtr Source: http://www.me.go.kr Mexican source: http://www.cec.org/takingstock/highlights The International Standards Organization (ISO) has incorporated environmental management principles into its most recent (14000 series) standards for judging ﬁrms’ performance. The ISO 14000 standards evaluates ﬁrms’ adoption of environmental management techniques that are believed to reduce polluting emissions. For detailed descriptions and analysis of the PROPER program, see Afsah et al., (2000) and World Bank (2000). More recent comparisons are unavailable because PROPER’s operations were disrupted by Indonesia’s ﬁnancial crisis in the late 1990s. The program recovered as the government returned to a relatively sound ﬁnancial footing, but consistent and comparable data are not yet available for the post-crisis period. Top-rated ﬁrms have no disclosure-related incentive for further improvement, but assignments of top ratings have proven so rare that this potential problem has not emerged in practice. For further discussion of EcoWatch, see World Bank (2000). Like PROPER, EcoWatch was aﬀected by ﬁnancial and political changes during the late 1990s. EcoWatch began during the presidency of Fidel Ramos. It was not supported by the administration of Ramos’s successor, Joseph Estrada, but it has been restored and strongly supported by the current administration of Gloria Macapagal-Arroyo. For further discussion of China’s GreenWatch program, see Wang et al. (2004). In poor rural communities, for example, radio broadcasts or community meetings may be superior to dissemination via newspapers and television. Community feedback is undoubtedly a strong force for discipline in this context. As we noted previously, one Indonesian factory was downgraded from blue (compliant) to black (ﬂagrantly non-compliant) after a neighboring community revealed its false reporting. The factory had been dumping pollutants into the neighboring river at night, to avoid detection by government inspectors, but local residents were perfectly aware of the situation. Disclosure programs operate principally through informal regulation by communities and markets, so monitoring and enforcement costs are shared by these actors. By contrast, governments absorb all monitoring and enforcement costs in conventional regulation. In addition, conﬁdential transactions between inspectors and plant managers under conventional regulation are susceptible to corruption in many developing countries. These elements provide important advantages for public disclosure but, as we note above, its information requirements (and costs) are not signiﬁcantly diﬀerent from those

116

26. 27.

28.

Yearbook of environmental and resource economics imposed by conventional regulatory programs (market- or quantity-based) that target emissions directly. For details, see Nike’s monitoring and compliance web page at http://www.nike.com/ nikebiz/nikebiz.jhtml?page=25&cat=compliance Recent interpretations of the WTO conventions suggest that such measures could be acceptable under existing agreements. Past controversies have focused on the distinction between products and processes in environment-related trade measures. In one wellknown case, the WTO seemed to allow measures directed at products only, ruling against US restrictions on tuna imports from countries whose ﬁshing ﬂeets did not employ dolphin-safe methods. However, more recent decisions have been more receptive to process distinctions. For example, a ruling against US shrimp import restrictions based on use of turtle-safe methods was rejected, not on process grounds, but because the USA had given remedial technical assistance to some Caribbean nations without providing it to Asian competitors. Under the current conditions, a per-unit emissions charge levied equally on domestic and foreign producers without reference to nationality might well be accepted by the WTO. Technology seems to be outpacing the conventional imagination in this sphere. Large retailers such as Walmart are close to individually bar-coding every unit of every product. This, along with embedded microchips capable of communicating directly with point-of-sale terminals, will probably eliminate checkout lines within a few years. Shopping carts will then become programmable, allowing for automatic warnings about products from factories that fail consumers’ social or environmental criteria. One further example, just to reinforce the point: an NGO in India is currently injecting microchip transmitters into individual trees in protected areas, so that they can be tracked by satellite if they are illegally logged.

REFERENCES Afsah, S., A. Blackman and D. Ratunanda (2000), ‘How do public disclosure pollution programs work? Evidence from Indonesia,’ discussion paper 00-44, Resources for the Future. Arora, S. and T. Cason (1998), ‘Do community characteristics inﬂuence environmental outcomes? Evidence from the Toxics Release Inventory,’ Journal of Applied Economics, 1(2), 413–53. Barth, M.E., McNichols, M. and G.P. Wilson (1997), ‘Factors inﬂuencing ﬁrms’ disclosures about environmental liabilities,’ Review of Accounting Studies, 2, 35–64. Baumol, M. and W.E. Oates (1988), The Theory of Environmental Policy, New York: Cambridge University Press. Berthelot, S., D. Cormier and M. Magnan (2003), ‘Environmental disclosure research: review and synthesis,’ Journal of Accounting Literature, 22, 1–44. Blackman, A. and G.J. Bannister (1998), ‘Community pressure and clean technology in the informal sector: an econometric analysis of the adoption of propane by traditional Mexican brick makers,’ Journal of Environmental Economics and Management, 35, 1–21. Bui, L. and C. Mayer (2003), ‘Regulation and capitalization of environmental amenities: evidence from the Toxic Release Inventory in Massachusetts,’ The Review of Economics and Statistics, 85(3), 693–708. CETESB (1986), ‘Restoring the Serra do Mar’ 1990, ‘Cubatao: A change of air.’ Cormier, D. and M. Magnan (1999), ‘Corporate environmental disclosure strategies: determinants, costs and beneﬁts,’ Journal of Accounting and Finance, 14 (4), 429–51.

Disclosure strategies for pollution control

117

Cormier, D. and M. Magnan (2003), ‘Does disclosure matter?,’ CA Magazine, 136(4), 43–5. Dasgupta, S., B. Laplante and N. Mamingi (2001), ‘Pollution and capital markets in developing countries,’ Journal of Environmental Economics and Management, 42, 310–35. Dasgupta, S., J. Hong, B. Laplante and N. Mamingi (2004), ‘Disclosure of environmental violations and the stock market in the Republic of Korea,’ World Bank Development Research Group working paper no. 3344. Davis, D. (2002), When Smoke Ran Like Water: Tales of Environmental Deception and the Battle Against Pollution, New York: Basic Books. Eisner, M.A. (2004), ‘Corporate environmentalism, regulatory reform and industry self-regulation: toward genuine regulatory reinvention in the US,’ Governance: An International Journal of Policy, Administration and Institutions, 17(2), 145–67. Feinstein, J.S. (1989), ‘The safety regulation of US nuclear power plants: violations, inspections and abnormal occurrences,’ Journal of Political Economy, 97(1), 115–54. Gozali, N., J. How and P. Verhoeven (2002), ‘The economic consequences of voluntary environmental information disclosure’, http://www.iemess.org/iemess 2002/ proceedings/. Gerde, V. and J. Logsdon (2001), ‘Measuring environmental performance: use of the toxics release inventory (TRI) and other US environmental databases,’ Business Strategy and the Environment, 10(5), 269–85. Glick, T.F. (1980), ‘The great stink of 1858,’ in L.J. Bilsky (ed.), Historical Ecology: Essays on Environment and Social Change, Port Washington, NY: Kennikat Press, p. 122. Gray, W. and Deily, M.E. (1996), ‘Compliance and enforcement: air pollution regulation in the U.S. steel industry,’ Journal of Environmental Economics and Management, 31, 96–111. Gupta, S. and B. Goldar (2003), ‘Do stock markets penalise environmentunfriendly behavior? Evidence from India,’ mimeo. Hamilton, J.T. (1999), ‘Exercising property rights to pollute: do cancer risks and politics aﬀect plant emission reductions?,’ Journal of Risk and Uncertainty, 18(2), 105–24. Hamilton, T. (1995), ‘Pollution as news: media and stock market reaction to the Toxics Release Inventory data,’ Journal of Environmental Economics and Management, 28, 98–113. Harrington, W. (1988), ‘Enforcement leverage when penalties are restricted,’ Journal of Public Economics, 37(1), 29–53. Harrison, K. (1995), ‘Is cooperation the answer? Canadian environmental enforcement in comparative context,’ Journal of Policy Analysis and Management, 14(2), 221–44. Hawkins, K. (1983), ‘Bargain and bluﬀ: compliance strategy and deterrence in the enforcement of environmental regulations,’ Law and Policy Quarterly, 5(1), 35–73. Heyes, A. (2000), ‘Implementing environmental regulation: enforcement and compliance,’ Journal of Regulatory Economics, 17(2), 107–29. Heyes, A.G. and N. Rickman (1999), ‘Regulatory dealing – revisiting the Harrington paradox,’ Journal of Public Economics, 72(3), 361–78. Kathuria, V. (2004), ‘Informal regulation of pollution in a developing country: empirical evidence from India,’ SANDEE working paper no. 6-04.

118

Yearbook of environmental and resource economics

Kirchoﬀ, S. (2000), ‘Green business and blue angels,’ Environmental and Resource Economics, 15(4), 403–20. Konar, S. and M.A. Cohen (1997), ‘Information as regulation: the eﬀect of community right to know laws on toxic emissions,’ Journal of Environmental Economics and Management, 32, 109–24. Lanoie, P. and B. Laplante (1994), ‘The market response to environmental incidents in Canada: a theoretical and empirical analysis,’ Southern Economic Journal, 60, 657–72. Lanoie, P., B. Laplante and M. Roy (1998), ‘Can capital markets create incentives for pollution control?,’ Ecological Economics, 26, 31–41. Laplante, B. and P. Rilstone (1996), ‘Inspections and emissions of the pulp and paper industry in Quebec,’ Journal of Environmental Economics and Management, 31(1), 19–36. Mixon, F.G. (1994), ‘Public choice and the EPA: empirical evidence on carbon emissions violations,’ Public Choice, 83(1), 127–37. Muoghalu, M.I., H. Robison and J.L. Glascock (1990), ‘Hazardous waste lawsuits, stockholder returns, and deterrence,’ Southern Economic Journal, 357–70. Nash, J., J. Ehrenfeld et al. (2000), ‘ISO 14001 and EPA Region I’s Startrack programs: assessing their potential as tools in environmental protection,’ in A.R. Edwards (ed.), Transforming Environmental Protection for the 21st Century, Washington, DC: National Academy of Public Administration, 2.1–2.90. Pargal, S. and D. Wheeler (1996), ‘Informal regulation of industrial pollution in developing countries: evidence from Indonesia,’ Journal of Political Economy, 104(6), 1314–27. Shweiki, O. (1996), ‘Environmental audit privilege and voluntary disclosure rule: the importance of federal enactment,’American Criminal Law Review, 33(4), 1219–49. Shavell, S. (1992), ‘A Note on Marginal Deterrence,’ International Review of Law and Economics, 12(1), 133–49. Sinclair-Desgagne, B. and E. Gozlan (2003), ‘A theory of environmental risk disclosure,’ Journal of Environmental Economics and Management, 45(2), 377–93. Stephan, M. (2002), ‘Environmental information disclosure programs: They work, but why?,’ Social Science Quarterly, 83(1), 190–205. Tietenberg, T. and D. Wheeler (2001), ‘Empowering the community: information strategies for pollution control,’ in Henk Folmer (ed.), Frontiers of Environmental Economics, Cheltenham, UK and Northampton, MA: Edward Elgar. Toﬀel, M. and J. Marshall (2004), ‘Improving environmental performance assessment: a comparative analysis of weighting methods used to evaluate chemical release inventories,’ Journal of Industrial Ecology, 8(1–2), 143–72. US Environmental Protection Agency (US EPA) (2002), User’s Manual for RSEI Version 2.1 [1988–2000 TRI Data], Washington, DC: US EPA. US EPA (2003), ‘How are the Toxics Release Inventory data used ? – government, business, academic and citizen uses’, EPA report no. 260-R-002-004, May. Wang, H. and D. Wheeler (2003), ‘Equilibrium pollution and economic development in China,’ Environment and Development Economics, 8(3), 451–66. Wang, H. and D. Wheeler (2005), ‘Financial incentives and endogenous enforcement in China’s pollution levy system,’ Journal of Environmental Economics and Management, 49(1), 174–96. Wang, H., J. Bi, D. Wheeler, J. Wang, D. Cao, G. Lu and Y. Wang (2004), ‘Environmental performance rating and disclosure: China’s GreenWatch program,’ Journal of Environmental Management, 71(2), 123–33.

Disclosure strategies for pollution control

119

Wheeler, D., H. Hettige, S. Pargal and M. Singh (1997), ‘Formal and informal regulation of industrial pollution: evidence from Indonesia and the US,’ The World Bank Economic Review, 11(3), 433–50. Yaeger, P. (1991), The Limits of the Law: The Public Regulation of Private Pollution, Cambridge: Cambridge University Press. World Bank (2000), Greening Industry: New Roles for Communities, Markets and Governments, Oxford: Oxford University Press.

4. Environmental policy under imperfect competition* Till Requate 1.

INTRODUCTION

It is well known that due to their static eﬃciency economists prefer marketbased instruments of environmental policy such as emission taxes and tradable permits, to command and control. According to the Pigouvian rule, the optimal price of pollution should be equal to marginal social damage. Thus, since competitive ﬁrms equalize their marginal abatement costs to the price of pollution, notably an emission tax rate or a price for tradable permits, a socially optimal allocation can be decentralized. This is because (a) marginal abatement costs are levelled out among all the polluters, and (b) marginal abatement costs are equalized to marginal damage. At an early stage, Buchanan and Stubblebine (1962) and Buchanan (1969) challenged the Pigouvian paradigm by pointing out that a monopolist distorts the market allocation by holding down output. Therefore, a Pigouvian tax established to regulate emissions by a polluting monopolist would exacerbate the distortion. Starting from this observation, Buchanan (1969) launches a general attack against emission taxes in imperfectly competitive markets. He writes ‘This note is presented as a contribution to the continuing dismantling of the Pigouvian tradition in applied economics’ and ‘the whole approach of the Pigouvian tradition is responsible for many confusions in applied economics that are slowly to be clariﬁed . . . making the marginal private cost as faced by the decision-taking unit equal to marginal social cost does not provide the Aladdin’s Lamp for the applied welfare theorist, and the sooner he recognizes this the better.’ Finally, on the relationship of Pigouvian taxes and market structure he writes: ‘It is necessary to distinguish, however, between the relevance of market structure for the emergence of externality and the relevance of market structure for the application of Pigouvian norms . . . it is necessary to limit the Pigouvian correctives on the tax side to situations of competition.’ This statement certainly overshoots the mark with regard to the problem of regulating a polluting monopolist, because when environmental damage is large, a zero tax (or the 120

Environmental policy under imperfect competition

121

absence of any other kind of regulation) may be much worse than setting the Pigouvian level of a tax (equal to marginal damage), even if the market structure is monopolistic. D.R. Lee (1975) supports Buchanan’s view in principle but does not reject emission taxes outright as a means of mitigating the externality. He was the ﬁrst to point out that, compared to the tax rate to be imposed on a competitive ﬁrm, a tax charged on a ﬁrm exercising market power has to be reduced by a term including P(Q)qi, where P and qi, denote the inverse demand function and ﬁrm i’s output, respectively. As is well known, P(Q)qi represents the diﬀerence in marginal revenues between a competitive and an oligopolistic (or monopolistic) ﬁrm. While Lee concentrated on a standard-and-charges approach where the emission target is assumed to be given, Barnett (1980) was the ﬁrst to rigorously solve the problem of determining the second-best optimal emission level and the corresponding second-best optimal emission tax to be imposed on a monopolist when pollution is evaluated with reference to a social damage function. Barnett’s article was a milestone in the theory of pollution regulation under imperfect competition and opened up a new avenue for research. There is hardly any paper on the theory of pollution control on imperfectly competitive ﬁrms that does not refer to Barnett’s contribution. In this chapter I survey the theory of pollution control on ﬁrms exercising market power at some point in the market process. For this purpose I set up a general model where the ﬁrms’ technologies, including abatement opportunities, are represented by their cost functions. I distinguish two cases. In the ﬁrst case, pollution is proportional to output and ﬁrms have no opportunity to reduce pollution other than by reducing output. In the second case, ﬁrms can in principle decide independently on output and emissions. In this case I write a typical ﬁrm’s cost function as C(q, e), which is interpreted as the cost incurred by the ﬁrm for producing q units of output with no more than e units of emissions. This representation of the ﬁrms’ technologies is used throughout the whole chapter, except when I focus on pure market power in a market of tradable permits. Here the ﬁrms’ technologies are simply represented by their abatement cost functions. For the greater part of this chapter, I will discuss both the comparative static eﬀects of diﬀerent pollution control instruments and the rules for determining the ﬁrst- or second-best optimal levels of certain policy instruments, in particular emission taxes. I start with the case of pure monopoly; besides taxes I also study other instruments such as absolute and relative standards. Next I summarize pollution control policies in the standard types of oligopoly model: Cournot competition, Bertrand competition with homogeneous goods, price competition with diﬀerentiated commodities, Cournot competition with free entry, and, ﬁnally, monopolistic

122

Yearbook of environmental and resource economics

competition. Then I turn to permit trading when only two or a small number of ﬁrms engage in imperfect competition on the output market. In a further section I consider market power on some input market and investigate the consequences for optimal or second-best optimal policy setting. An assumption frequently made in all these models is that the regulator can only use environmental policy to combat two or at best three market imperfections: the ﬁrms create an externality, they keep up prices (or hold down output) by virtue of exercising market power, and ﬁnally, in models with free entry, a non-optimal number of ﬁrms enter the market. The main conclusion in all this literature is that with both types of market power, in the output or in an input market, second-best optimal policies lead to allocations where the ﬁrms’ marginal abatement costs fall short of marginal damage. This implies that, under a tax policy, the second-best optimal tax rate is smaller than marginal damage. There are two exceptions to this rule. The ﬁrst is the case of Cournot competition with free entry, in which the second-best optimal tax may exceed marginal damage to mitigate excessive market entry. The second is where a monopolist has market power over an abating input or an advanced abatement technology. In a scenario like this, a regulator can raise the demand for abating inputs or advanced abatement technology by raising the tax rate. Further, I study monopoly power on the market for tradable permits. For this purpose I elaborate a generalized version of Hahn’s inﬂuential model (1984). Hahn presented a simple set-up for a permit market with one large price-setting ﬁrm and several small price-taking ﬁrms. This model has been extended in several directions: market power on both the permit and the output market, non-compliance by either the small ﬁrms or the large ﬁrm, etc. The Hahn model has also been subjected to a number of experimental investigations which I brieﬂy summarize. All the models mentioned so far rely on the assumption that market power is exercised in a closed domestic market. Since the literature on environmental policy in open economies has been treated in detail elsewhere, in particular by Ulph (1997a) in this series, I will not attempt to present a complete survey on this large sector of the literature. However, I would like to highlight the link between the theories of environmental policy under imperfect competition and environmental policy as trade policy. Several environmental and trade theorists have pointed out that in the absence of trade policy environmental policy instruments can be (ab)used as trade policy since they can have an impact on the terms of trade when the country applying those instruments is large (see Markusen, 1975). In particular, a country hosting large ﬁrms with market power in international markets has to take account of several oﬀsetting eﬀects when calculating the unilaterally optimal emission tax rate. If domestic consumers are

Environmental policy under imperfect competition

123

also served by domestic ﬁrms engaging in imperfect competition on international markets, the regulator has to take account of domestic consumer surplus and weigh this either against the terms-of-trade or against the rentshifting eﬀect. Due to these oﬀsetting aspects caution is required in talking about ‘eco-dumping’ whenever a government sets a domestic tax rate below marginal damage. The point is that in the pure model of a polluting monopolist, where pollution is proportional to output, the regulator can even implement the ﬁrst-best allocation by choosing the appropriate tax rate. In this case, the regulator must set the tax rate below marginal damage since he has to take into consideration the monopolistic behavior of the ﬁrm. Hence, in an international trade model with imperfect competition, the regulator must take account of the dead weight loss generated at home. Accordingly, as Duval and Hamilton (2002) have pointed out, not every issue that makes the tax rate lower than marginal damage is to be interpreted as eco-dumping. To highlight these issues I extend the basic model used in this survey to the case of international trade, I suggest alternative decompositions of the unilateral optimal emission tax, and I discuss rival interpretations of such decompositions with respect to terms-of-trade versus the rent-shifting eﬀect. In discussing important results from the literature in detail, I shall usually adapt the models of other authors to my notation and assumptions, in order to present an integrated treatment of all the diﬀerent cases. I will also add some new material, in cases where I have found a gap in the literature that needs to be closed. Two such cases are optimal standards in Cournot oligopoly with free entry and market power on factor markets. Needless to say, there are many other gaps in the literature that need to be closed. The remainder of this chapter is organized as follows. In the next section I set up some general assumptions on the ﬁrms’ cost functions, market demand, and the social damage function that will be used throughout this chapter. In section 3 I treat the pure regulation of a monopolist. Besides the tax instrument I also study absolute standards, relative standards and tax/subsidy schemes. Further, I summarize several extensions of that basic model taken from the literature. Since there are so many contributions on the regulation of polluting oligopolies, I have decided to split up the treatment of oligopoly into several sections. In section 4 I investigate emission taxes in the classical Cournot model, i.e. quantity-setting oligopoly with a ﬁxed number of ﬁrms. I extend that model to tax/subsidy schemes, and at the end of the section I summarize several extensions taken from the literature, such as pre-investment, ecological tax reform, and other issues. In section 5 I study emission taxes in price-setting oligopoly, where I treat both cases, i.e. genuine Bertrand competition with homogeneous commodities and the case of diﬀerentiated commodities. Section 6 deals with

124

Yearbook of environmental and resource economics

environmental policy in models with market entry. Besides the classical Cournot model with free entry, I also present a version of the Dixit–Stiglitz model and brieﬂy summarize results from emission taxes in Salop’s model of the circular city. In section 7 I present a model for an output duopoly with permit trading and brieﬂy discuss extensions to the case of more than two ﬁrms. In section 8 I treat models with market power on an input market, considering the cases of pure monopsony and Cournot oligopsony. In subsection 8.5 I brieﬂy summarize some results on models of market power for a clean input. Though brief, this section is important because it yields results where the second-best optimal emission tax may exceed marginal damage. In section 9 I study models involving market power in the permit market. The section presents a model that generalizes the seminal model suggested by Hahn (1984) to cover the case of several ﬁrms with market power. I survey the numerous extensions of the Hahn model, including experimental evidence. In section 10 I deal with the regulation of market power in models of international trade. Section 11 provides a summing-up, indicates some gaps in the literature and suggests directions for further research.

2.

SOME BASIC ASSUMPTIONS

For most of this chapter I intend to discuss partial equilibrium models with one consumption good and one pollutant. In sections 5.2 and 6.3 I will also consider models with several private commodities, modifying assumptions on preferences and demand accordingly. There are n1 ﬁrms producing a homogeneous commodity. In the production process the ﬁrms generate a pollutant emitted into the environment. Most of the time I will assume that the pollutant is generated in this industry only (i.e., by the n ﬁrms under consideration). This assumption does not hold in all industries, of course. CO2, for example, is generated by many diﬀerent industries. In the chemical industry, by contrast, several hazardous pollutants are generated by the production of one commodity only. In our framework, notably when considering imperfect competition, assuming the pollutant to be industry-speciﬁc makes the analysis more interesting and is of greater empirical relevance. As a welfare measure I will adopt the standard cost/surplus concept employed throughout the literature on industrial organization. Thus welfare is deﬁned as gross consumer beneﬁt minus production costs minus the monetarized damage from pollution, to be deﬁned more precisely below. For i1, . . ., n I will use qi and ei to denote the quantity produced and supplied on the commodity market and the emissions

Environmental policy under imperfect competition

125

dumped into the environment by ﬁrm i, respectively. Aggregate (or industry) output and total emissions are denoted by Q : ni1 qi and E : ni1ei, respectively. 2.1

Preferences

Preferences on the part of the consumers (or society) are represented by an inverse demand function P and a social damage function D, where the latter captures both the disutility that consumers suﬀer and the economic damage that other industries suﬀer from the pollution generated by the n ﬁrms under consideration. I make the following assumption about demand: Assumption 1 The inverse demand function P : → depends on aggregate industry output Q only. Moreover, ˜ 0| with P (Q ˜ ) 0}. (i) it is twice continuously diﬀerentiable for all Q {Q (ii) P is strictly decreasing, and P is suﬃciently bounded; in particular, for all Q 0 we have: P"(Q) Q 1 P(Q)

(4.1)

(4.1) says that P is not too convex. This is suﬃcient to guarantee the secondorder conditions for several maximization problems, in particular for social optimum, proﬁt maximization of monopoly and, for the oligopolistic ﬁrm in Cournot–Nash equilibrium. It is also suﬃcient to guarantee the stability and uniqueness of Nash equilibrium when we consider the Cournot game in section 4. For the damage function I adopt the following assumption from the literature: Assumption 2 The damage from pollution depends on total pollution E only. The social damage function D(E) is twice continuously diﬀerentiable, increasing and (weakly) convex, i.e. D(E)0. Where certain results require strict convexity, I will make explicit mention of this fact. 2.2

Technologies

The ﬁrms’ technologies are represented by their reduced cost functions. This assumes that all factor markets are perfectly competitive and – both here and in the models of imperfect competition in the output market – are not inﬂuenced by any strategic behavior of the ﬁrms in other markets.

126

Yearbook of environmental and resource economics

We will make alternative assumptions about those technologies. In the ﬁrst assumption, pollution is proportional to output and ﬁrms do not have any further abatement technologies: Assumption 3 Each ﬁrm’s cost function Ci: → is (i) twice continuously diﬀerentiable for qi 0. (ii) Moreover, Ci (qi ) 0 and C"i (qi ) 0. (If C"i (qi ) 0 is required, this will be explicitly mentioned.) (iii) Pollution is proportional to output, i.e. ei iqi, with i 0, for all i 1, . . ., n. Alternatively, I assume that ﬁrms have technologies where pollution can be substituted for by using more of other abating inputs, which in turn incurs higher costs. I assume that those abatement opportunities are already incorporated in the reduced cost function written as Ci(qi, ei), i.e., the cost depends on both ﬁrm i’s output qi and its emissions ei, thus satisfying the following assumption: 2 Assumption 4 For each ﬁrm i {1, . . ., n} its cost function Ci: → is twice continuously diﬀerentiable and satisﬁes the following properties (we omit the superscript i): (i) Cq 0, Cqq 0, Cee 0, Cqe 0 (ii) For all q there exists an emission level e(q) such that Ce(q, e(q))0, and Ce(q, e) 0 if e e(q), and Ce(q, e)0 if e e(q). (iii) The Hessian of C is positive deﬁnite. In particular, C satisﬁes

Cqq Cee [Cqe]2 0

(4.2)

This assumption implies that the variable cost function is strictly convex. In particular, we have increasing marginal costs for ﬁxed emission levels, abatement costs are convex for each ﬁxed output, and output and emissions are complements (Cqe 0), which implies that the marginal abatement cost increases with more output (Ceq 0). Moreover, for each output level there is a cost-minimizing emission level e(q) that would be chosen by the ﬁrms in the absence of regulation (i.e. Ce(q, e(q))0). In such a case, we can deﬁne a further reduced cost function by C˜ (q): C(q,e(q)). In principle the cost function could contain ﬁxed costs. Where these are important, for example in models with free entry, we will also explicitly mention the fact. Note: Some authors, such as Barnett (1980), Conrad (1993), and Duval and Hamilton (2002), specify certain inputs that can be used to reduce pollution. These authors write the cost function as C(q, w( )), where w( ) is

Environmental policy under imperfect competition

127

the price of a polluting input including a tax rate (or a corresponding permit price). Using Shepard’s Lemma it is possible to recover each ﬁrm’s level of emissions, which are assumed to be proportional to the polluting input x, by the relationship eCw(q, w( )), where is an emission coeﬃcient. 2.3

Welfare and First-Best Allocations

To evaluate the utility and the harm of a given allocation (q1, . . ., qn, e1, . . ., en) for society, we deﬁne a social welfare function W, which is additively separable into consumers’ gross beneﬁt, production cost, and the social damage caused by the pollution: W(q1, . . ., qn, e1, . . ., en ):

Q

P(z)dz

n

Ci(qi ei) D(E)

(4.3)

i1

0

where Q ni1 qi and E ni1ei. If there are no abatement technologies we simply obtain W(q1, . . ., qn ):

Q

P(z)dz

0

n

Ci(qi) D(E)

(4.4)

i1

n

where total emissions are now determined by E iqi. i1 Under Assumption 3, and if the ﬁrms’ cost functions are strictly convex, the socially optimal allocation is characterized by the following set of ﬁrst order conditions:1 P(Q)Ci (qi)iD (E)

i1, . . ., n

Thus the marginal willingness to pay equals marginal production costs plus marginal social damage times the emission coeﬃcient of the respective ﬁrm. Under Assumption 4, the social optimum is characterized by the following alternative set of ﬁrst-order conditions: Ciq (qi, ei ) P(Q) i 1, . . ., n

(4.5)

Cie (qi, ei ) D(E) i 1, . . ., n

(4.6)

This means that the ﬁrms’ marginal cost of producing a further unit of the marketable commodity equals the consumers’ marginal willingness to pay, and the marginal abatement costs Cie (qi , ei ) are equal to the marginal

128

Yearbook of environmental and resource economics

social damage. In particular, marginal costs and marginal abatement costs are equal across all the ﬁrms. We use (q*i , e*i ) i1, . . . n to denote the socially optimal allocation and Q* for the corresponding aggregate output. The corresponding total emissions are denoted by E*. 2.4

Decentralization under Perfect Competition

It is easy to see that under perfect competition the social optimum can be implemented by charging a tax rate that satisﬁes the Pigouvian rule, i.e. equals the marginal damage resulting from the socially optimal allocation. This can be seen as follows. The ﬁrms take the price for the output commodity as given and maximize i(qi, ei)pqi Ci(qi, ei) ei which leads to the following ﬁrst-order conditions: Cie (qi, ei ) p P(Q) i 1, . . ., n

(4.7)

Ciq (qi, ei ) i 1, . . ., n

(4.8)

D(E*)

(4.9)

If the regulator sets

we see that (4.7) and (4.8) imply (4.5) and (4.6). Thus the socially optimal allocation results if the market equilibrium is unique, which is guaranteed by our assumptions. Alternatively, the regulator could issue an amount of tradable permits L equal to the socially optimal total emission level E*.2 In the latter case, it does not matter whether the permits are auctioned oﬀ or issued for free (grandfathered). I assume throughout this chapter that the society is indiﬀerent about the redistribution of tax revenues. This implies that collecting tax revenues in this industrial sector is not a government objective. This in its own is tantamount to assuming that no other distortionary taxes (and hence no marginal costs of public funds) exist, and that there is no additional technology that the government can buy in order to reduce the aggregate emissions E once these have been dumped into the environment by the ﬁrms. Thus collected taxes will be redistributed to the consumers in a lump-sum way. Only in sections 3.5.8 and 4.6.6, where I brieﬂy discuss ecological tax reforms in the presence of imperfect competition, do I deviate from this assumption.

129

Environmental policy under imperfect competition

Further, I assume that the emissions generated by each ﬁrm can be perfectly monitored without cost by the regulatory authorities. Accordingly, the ﬁrms will pay a tax bill that is exactly equivalent to the amount of pollutants they emit. In the case of permits, ﬁrms cannot emit more than the number of permits allows them to. Otherwise I assume that a high penalty has to be paid (boiling-in-oil policy). Thus there is no room for moral hazard. One departure from this assumption is the summary of models with non-compliance in permit markets in section 9.

3. IMPERFECT COMPETITION IN THE OUTPUT MARKET: MONOPOLY In this section and the following I study models of imperfect competition in the output market, starting with monopoly. I assume that a single ﬁrm operates in the output market and generates a pollutant that is industryspeciﬁc (or a local pollutant). Thus regulation refers to this single ﬁrm only and does not aﬀect other industries. A real situation represented by such a model is a chemical or pharmaceutical ﬁrm specializing on the production of a certain output and generating either a speciﬁc or a local pollutant. 3.1

Taxation of Monopoly when there is no Abatement Technology

If the monopolist’s technology satisﬁes Assumption 3 and it is subject to an emission tax , its proﬁt depends on output only and can be written as (q)P(q)q C(q) q The ﬁrst-order condition for proﬁt maximization is then given by (4.10)

P(q)qP(q)C(q) 0

The solution is denoted by qM( ). Setting n1 in (4.4), and writing welfare as a function of the tax rate, we obtain: W( ) :

q M ( )

P(z)dz C(qM ( )) D(qM ( ))

(4.11)

0

Diﬀerentiating (4.11), with respect to , we obtain [P(qM ( )) C(qM ( )) D(qM ( ))]

dqM ( ) 0 d

(4.12)

130

Yearbook of environmental and resource economics

Using (4.10) by substituting for P(q)C(q) P(q)q and ‘solving’ (4.12) for the tax rate,3 we get

D(qM ( ))

P(qM ( ))qM ( )

(4.13)

P(qM ( )) ·

(4.14)

which can also be rewritten as

D(qM ( ))

where P(qM( ))qM( )/P(qM( )) is demand elasticity. By substituting (4.14) into (4.10), we immediately see that in this case the optimal emission tax rate leads to the ﬁrst-best outcome. The simple reason for this is that the regulator has direct control over output by virtue of qe/, and since qM( ) is decreasing in .4 Thus the regulator can induce the ﬁrm to produce the socially optimal outcome, i.e. qM( )q*. We summarize this as Proposition 1 If emissions are proportional to output (or a monotonic function of output) and there is no abatement technology, there exists an emission tax rate that implements the social optimum. The optimal tax rate is lower than marginal damage. Note that the tax rate can also be negative because the monopolist creates two market imperfections by holding down output and generating a negative externality through pollution. If the tax rate is negative, the ﬁrst imperfection dominates the second, and the regulator mitigates underprovision of the output market by subsidizing pollution, which seems to be more a theoretical option rather than a real one.5 3.2

Taxation of a Monopolist with Abatement Technology

I now consider the case where the monopolist’s cost function satisﬁes Assumption 4. Again, the monopolist is subject to an emission tax, its proﬁt is given by (q, e) P(q)qC(q, e) e Proﬁt maximization then leads to the following ﬁrst-order conditions: P(q)qP(q) Cq(q, e) 0 Ce(q, e)

(4.15) (4.16)

Environmental policy under imperfect competition

131

This means that marginal revenue equals the marginal cost for producing one more unit of output, and marginal abatement cost is equal to the tax rate. I use q( ) and e( ) to denote the solution to (4.15) and (4.16). The environmental authority then seeks to maximize welfare as a function of the tax rate: W( ) :

q( )

P(z)dz D(e( )) C(q( ),e( ))

(4.17)

0

leading to the ﬁrst-order condition W( )[P(q( ))Cq(q( ), e( ))]q( ) [Ce(q( ), e( ))D (e( ))] e ( ) 0

(4.18) (4.19)

where again for convenience I have written q( ): dq/d and so on. Employing (4.15) and (4.16) and ‘solving’ for the tax rate gives us the following optimality condition for the second-best optimal emission tax:

D(e( )) P(q( ))q( ) D(e( ))

q( ) e( )

P(q( )) q e

(4.20) (4.21)

where q/e is the reaction of output to the relaxation of an emission standard and is the price elasticity. To determine the signs of q and e, we diﬀerentiate (4.15) and (4.16) with respect to , yielding q

Cqe

0 Cee (P 2P) (CqqCee [Cqe]2 )

and e C1 ee

[Cqe]2

0 2 [Cee] (P 2P) Cee (CqqCee [Cqe]2 )

where we have made use of Assumptions 1 and 4, in particular Cqe 0, to sign the expressions. In this case, the second-best optimal tax rate does not lead to the ﬁrstbest allocation, as can be seen from substituting (4.20) into the monopolist’s ﬁrst-order conditions of proﬁt maximization. The reason for this result is that the monopolist can now independently decide on output and emissions, while the regulator has only one instrument available. This result was ﬁrst established by Barnett (1980). Again, the second-best tax rate falls short of marginal damage, as can be seen from (4.20), and it can be even

132

Yearbook of environmental and resource economics

negative. Note that the second-best optimal emission level can exceed the ﬁrst-best level. This is due to the two market imperfections and the lack of a suﬃcient number of instruments. A phenomenon like this should not be confused with (environmental or) eco-dumping.6 It is merely a consequence of an insuﬃcient number of instruments and hence of a second-best optimal setting. As a special instance, we may brieﬂy consider the case where C(q, e) is additively separable into the cost of production and abatement, i.e. C(q, e)CP(q)CA(e). Since Cqe 0, we obtain

D (e( ))

(4.22)

and the second-best optimal tax rate equals marginal damage, as is also optimal under perfect competition. We can summarize this result as follows: Proposition 2 If C(q, e) is additively separable, the optimal emission tax rate is independent of the commodity market structure and equal to marginal damage, thus satisfying the Pigouvian rule. 3.3

Emission Standards

Absolute emission standard As an alternative to charging a price for emissions, the regulator can set an (absolute) emission standard. If the unregulated monopolist’s emission level exceeds the optimal emission level, an emission standard is equivalent to an emission tax. Conversely, if the unregulated emission level falls short of the optimal level of pollution, which is the case when the distortion from the monopolist’s market power exceeds the distortion resulting from pollution, the shadow price of pollution is negative, and thus the ﬁrst-best allocation (in the case where no abatement technology exists) cannot be implemented by a standard. The reason is simply that the regulator cannot induce the monopolist to increase output by setting an emission standard that does not bite. Emission permits Grandfathering a quota of emission permits is clearly equivalent to setting an absolute emission standard. The same applies to auctioning oﬀ such a quota of permits. The monopolist would bid zero (or epsilon) and obtain all the permits. Relative emission standard In reality, relative emission standards are more common than absolute standards. Under this kind of standard, a ﬁrm is restricted with respect to its emissions per unit of output: eq for some number 0. If emissions

133

Environmental policy under imperfect competition

are proportional to output, i.e. eq, either the standard is not binding (i.e. ) or the ﬁrm cannot meet the standard and has to terminate production in the short run. If the ﬁrm’s technology satisﬁes Assumption 4 and the standard is binding, the ﬁrm’s cost function can be written as C(q, q). The monopolist’s ﬁrst-order condition is then represented by P(q)P(q)qCq(q, q)Ce(q, q)0

(4.23)

The comparative static eﬀect of relaxing the standard is now given by Ce q[Cqe Cee] dq d P(q)q 2P(q) [Cqq 2Cqe 2Cee] We see that the sign depends on the shape of the cost function. The denominator is negative, while the sign of the numerator is ambiguous. With a similar welfare-maximizing procedure to the one above, we ﬁnd that the second-best optimal standard is now characterized by the following relationship:

dq dq Ce (q, q) D(q) 1 q d P(q) d

(4.24)

Equation (4.24) implicitly deﬁnes the second best optimal standard, denoted by *. We see that the two terms (/q)(dq/d)D(E) and P(q)dq/d oﬀset each other, no matter what the sign of dq/d is. In general the sign of dq/d is ambiguous. In the normal case, however, we would expect dq/d 0 to hold. In this case we see, as usual, that the term accounting for monopoly power is negative, whereas the multiplier 1(/q)(dq/d) exceeds one. Thus the imperfect instrument represented by a relative standard causes the regulator to set the emission target at a stricter level compared to the second-best optimal emission level, resulting from a second-best optimal emission cap or a second-best optimal emission tax. The reason is that the polluting ﬁrm can comply to the standard not only by reducing emissions but also by increasing output. 3.4

Regulating Emissions and Output Simultaneously

It can easily be seen that if the regulator has two regulation instruments, namely an emission tax and a subsidy on output , the ﬁrst-best allocation can be achieved by setting the tax rate equal to the Pigouvian level and the subsidy equal to P(q*)/(q*), where (q*) is again demand elasticity at the optimal output quantity q*. This policy mix seems to be more of a theoretical possibility than a real option, since subsidies are frequently not feasible or

134

Yearbook of environmental and resource economics

are simply not allowed due to free trade treaties or other agreements. In section 4.4, where we discuss tax-subsidy schemes for oligopolistic competition, we will see, however, that a tax-subsidy scheme can be mimicked by a tax/tax-refunding scheme as exempliﬁed by the NOX tax system in Sweden. 3.5

Extensions

3.5.1 Some remarks on previous work As mentioned in the introduction, Buchanan (1969) refuses to apply emission taxes to the case of a monopolistic polluter. Siebert (1976) criticizes Buchanan’s conclusions by setting up a formal model where emissions are initially proportional to output but can be reduced separately by an abatement technology. Siebert indicates that under such a technology both the competitive and the monopolistic ﬁrm will choose the same level of abatement activity if they are subject to the same tax rates. However, this does not imply, as Siebert claims, that the second-best optimal tax rate is equal to the Pigouvian level. Prior to Barnett, Asch and Seneca (1976) and Smith (1976) had also arrived at what is basically the same conclusion, i.e. that taxing emissions by a monopolist requires setting the tax rate below marginal social damage. Parallel to Barnett, Misiolek (1980) generalizes the ideas of both Asch and Seneca and Smith. Oates and Strassmann (1984) extend Barnett’s conclusions to other forms of market structure. 3.5.2 Simultaneous regulation of monopoly and competitive firms Innes et al. (1991) study a situation where one ﬁrm, which is a monopolist in a given commodity market, and other ﬁrms, engaging in perfect competition on a diﬀerent output market, emit the same pollutant. This is certainly a realistic scenario for major pollutants such as SO2 and NOX or greenhouse gases such as CO2. The authors study the structure of the second-best optimal tax rate when both industries can only be regulated by a uniform tax. If we use eC( ) and qC( ) to denote emissions and output and CC(, ) and PM() to denote the aggregate cost function and the inverse demand function of the competitive sector and eM( ), qM( ),CM(,), and PM() to denote emissions, output, cost, and the inverse demand function of the monopolist, respectively, we can write welfare as:7 qM ( ) W( ) PM (z)dz CM (qM ( ),eM ( ))

0

qC ( )

PC (z)dz CC (qC ( ),eC ( ))

0

D(eM ( ) eC ( ))

Environmental policy under imperfect competition

135

Maximizing welfare with respect to the uniform emission tax the secondbest optimal tax rate is characterized by the following formula:

D(eM ( )) P(qM ( )) qM ( )

dqM ( )d

deM ( )d deC ( ) d

(4.25)

The intuition for this result is as follows: We would expect the secondbest optimal tax rate to be determined by the beneﬁts from production in both the monopoly and the competitive sector and to be a weighted sum from the pure second-best optimal monopoly tax and the Pigouvian tax resulting from regulating the competitive sectors. This is indeed the case through the multiplier [dqM( )/d ]/[deM( )/d deC( )/d ]. If the reaction of the competitive sector on a tax increase is small, i.e. deC( )/d is small, the multiplier will be close to [dqM( )/d ]/[deM( )/d ]. In that case the regulator can neglect the competitive sector, and (4.25) will be approximately represented by (4.20). If the reaction of the competitive sector on a tax increase is large compared to the reaction of the monopolist, the multiplier will be small and (4.25) will be close to the Pigouvian rule. Furthermore, the authors show that a discriminating tax system would be better than a uniform tax. They also show that a system of tradable permits outperforms the uniform tax if the initial allocation of permits is chosen appropriately. We will return to this issue in section 9, where we study models of market power in markets for tradable permits. 3.5.3 Several local monopolies Requate (1993b) considers the case of regulating several local monopolists. Under pure monopoly, a market for permits does not make much sense since there is only one ﬁrm on the demand side for permits. It often happens, however, that there are several ﬁrms, each of which exercises monopoly power in a local commodity market and generates the same kind of pollution. The sum of emissions caused by all the ﬁrms generates a negative externality for an extensive region, a country, or even the whole world. Typical and highly relevant examples of this kind of market structure are the utility industries, especially in Europe, where the ﬁrms have had (and in some regions still have) local monopoly power, and each ﬁrm emits pollutants such as SO2 and NOX, or CO2 as a greenhouse gas. If several monopolists are subject to regulation, issuing permits does make sense. Requate (1993b) shows that a suitable Pigouvian tax or a suitable number of tradable permits are equivalent tools for maintaining an aggregate pollution level below the unregulated laissez-faire level. This corresponds to a charges-and-standards approach. As with perfect competition, both policies lead to the same allocation. This will not be surprising if the permit

136

Yearbook of environmental and resource economics

market is competitive. However, the result still holds if the number of ﬁrms is small so that price-taking behavior on the permit market cannot be expected. In this case Requate assumes that the ﬁrms negotiate about both the allocation of permits and transaction prices. Since there is no strategic interaction on any commodity market, the ﬁrms have an incentive to achieve a cost-eﬃcient allocation of permits among themselves, leading to equal marginal abatement costs across ﬁrms. This does not hold in general for ﬁrms interacting in the output market, as we shall see in section 7. Requate also studies the second-best optimal uniform tax and the second-best optimal discriminating taxes. The latter case is equivalent to several single monopolists. In the ﬁrst case, the second-best optimal tax rate is given by i1 Pi (qi )qi (qi ) E

n

D(E)

(4.26)

where Pi(qi) represents the inverse demand function in market i. The optimal number of permits leads to a market price for permits that is equal to the second-best optimal tax rate determined by (4.26). Moreover, if the regulator can pay discriminatory subsidies, he can implement the ﬁrst-best outcome by charging the Pigouvian tax. If discriminatory subsidies are not legally feasible, we are back in the world of second-best. 3.5.4 Rent-seeking Misiolek (1988) points out that prior to production a monopolist tends to expend considerable resources to establish its monopoly position. This kind of behavior is also referred to as rent-seeking. A considerable share of these resources are usually a waste in social terms. Misiolek argues that the regulator should account for such rent-seeking behavior. In order to lower the incentives for rent-seeking, the second-best tax rate should be higher than the optimal tax rate in cases where the monopolist just arrives out of the blue. Misiolek shows that if we take rent-seeking into account, the second-best optimal emission tax rate can be higher, lower or equal to marginal damage. 3.5.5 When emissions inﬂuence consumer demand Ebert and von dem Hagen (1998) extend Barnett’s (1980) basic model by assuming that both consumer demand and production costs are aﬀected in a negative way by pollution. This can be simply modelled by writing inverse demand and cost functions as P(Q, E) and C(Q, E), respectively. Whereas consumers have less beneﬁt from the output good in a more heavily polluted environment, reﬂected by PE 0, the eﬀect on cost and marginal cost may go in opposite directions. In my view, only CE 0, CXE 0 is relevant, reﬂecting the fact that the ﬁrm faces positive marginal abatement

Environmental policy under imperfect competition

137

costs. In the opposite case, the ﬁrm would have an incentive to internalize the ‘externality’ of pollution on itself since the ﬁrm is the only polluter (this is the Coase theorem applied to a single agent). Moreover, the authors allow for two policy instruments, an output tax and an emission tax. Not surprisingly, if both instruments are available, the ﬁrst-best allocation can be implemented where both tax rates satisfy the Pigouvian rule (being equal to total marginal social damage). If either the emission tax or the output tax is the only instrument, the second-best optimal tax rates may exceed or fall short of marginal damage if CXE 0 holds, and, surprisingly, will otherwise exceed marginal damage. 3.5.6 Taxation of a durable goods monopolist Runkel (2004) extends Barnett’s model to the case of a durable goods monopolist, as developed by Bulow (1986). However, Barnett’s rule still holds: the second-best optimal tax rate falls short of marginal damage. 3.5.7 Overall regulation Laﬀont (1994) considers the overall regulation of a monopolist. Alongside an emission tax and a subsidy on output he also suggests a lump-sum tax for the extraction of the monopoly rent. This is especially important in the presence of existing distortionary taxes. Laﬀont also studies regulation under asymmetric information. Following the methodology of adverse selection applied to regulating a monopolist with unknown costs, as suggested by Baron and Myerson (1982), Laﬀont describes the second-best optimal policy scheme that induces the monopolist to produce second-best optimal output combined with the second-best optimal pollution level. 3.5.8 Environmental tax reform Bayindir-Upmann (2000) studies the eﬀects of an ecological tax reform in the case of monopoly. In his model the monopolist produces an output by means of two inputs: capital and emissions (or a polluting input). The government can levy both capital taxes and emission taxes. Bayindir-Upmann ﬁnds that if the initial tax rate on the non-polluting input (in his model: capital) is low, an environmental tax reform yields a triple dividend: it leads to a decrease of emissions, raises demand for non-polluting inputs, and raises the ﬁrms’ proﬁts. If the initial capital tax is extremely high, a double dividend does not exist. Here, it is the environmental dividend that fails to materialize. In other words, a tax shift from the non-polluting to the polluting input increases pollution while keeping revenues constant. High initial tax rates hamper economic activities and, thereby, induce the side eﬀect of low demand for environmentally harmful production factors. If such a tax is lowered, the total distortion on the economy decreases and

138

Yearbook of environmental and resource economics

demand for all goods and production factors increases. This occurs even when the tax on the environmentally harmful factor increases. In this case, an ecological tax reform stimulates production and thus increases proﬁts. This emerges, however, at the expense of environmental quality. By contrast, Fullerton and Metcalf (2002), who consider a similar model with labor and emissions, come to the opposite conclusion. The existence of monopoly power has two oﬀsetting eﬀects on welfare. On the one hand, a revenue-neutral tax reform that lowers the labor tax and increases the tax on the polluting input (or emissions) lowers the monopolist’s proﬁt which, however, leads to an increase of labor supply. This partially oﬀsets pre-existing labor supply distortion. On the other hand, the ecological tax reform raises output prices, and this, together with the pre-existing monopoly distortion, further exacerbates the labor supply distortion. The authors show that for empirically relevant parameters the second eﬀect dominates the ﬁrst. Thus, monopoly power does not raise the probability of achieving a double dividend from a revenue-neutral ecological tax reform. The more optimistic result in Bayindir-Upmann (2000) may be driven by both the inelastic supply of capital and the initially low tax rate on capital. Thus, in contrast to a labor market, there is no distortion resulting from the supply of capital. We will return to ecological tax reforms and the double dividend hypothesis in section 4.6.6.

4. ENVIRONMENTAL POLICY FOR COURNOT OLIGOPOLY After this discussion of the polar cases of market structure, namely perfect competition and monopoly, I now turn my attention to the wide array of oligopoly models. Of those the one discussed in the literature in most detail is the Cournot model, where ﬁrms use quantities as strategic variables. Accordingly, I shall be giving this model a relatively large amount of space in this survey. The Cournot model also links the perfect competition and monopoly models in a very natural way. The most relevant example of this kind of market structure with pollution regulation is the energy sector, where a homogeneous product, electricity, is produced and ﬁrms compete (at least locally) à la Cournot. Other examples are salt eﬄuents from salt mining, where a small number of ﬁrms pollute a river. This was the case with the river Rhine at the border between France and Germany, or with the rivers Elbe and Werra in former Eastern Germany. There are many other examples involving near-homogeneous products. In this section I start the analysis by studying the tax instrument and I also brieﬂy discuss standards. Since in oligopoly models the analysis of permits

Environmental policy under imperfect competition

139

is more complicated than in other models, I treat that instrument in a separate section (7). The ﬁrst researcher to investigate emission taxes for Cournot oligopoly was Levin (1985), who mainly studies the comparative statics eﬀects of increasing an emission tax in an asymmetric Cournot oligopoly. However, Levin does not take damage from pollution into account and so he does not discuss the structure of the (second-best) optimal tax rate. To the best of my knowledge, Ebert (1992) was the ﬁrst to present a secondbest analysis for Cournot oligopoly, restricting his attention, however, to the case of symmetric ﬁrms. Later, Simpson (1995) also discusses the case of asymmetric duopoly. I begin with a more general model and derive the main results produced by Levin, Ebert and Simpson as special cases. Unless otherwise stated, I assume in this section that an environmental authority can set a uniform emission tax rate per unit of eﬄuent only. Here, the authority neither has the power to regulate output distortions, say, by subsidizing output (cf. Cropper and Oates, 1992), nor can it charge individual taxes. The informational structure is such that the government knows what technologies are out there but does not necessarily need to know exactly which ﬁrm has which technology. In section 4.4, I brieﬂy consider the case where the regulator has two instruments at his disposal, a uniform emission tax and a subsidy on output. 4.1

The General Framework

I consider a quantity-setting Cournot game with n ﬁrms, where the number of ﬁrms is assumed to be exogenously given. The ﬁrms produce a homogeneous commodity. Its market price is determined by aggregate output according to pP(Q). The ﬁrms’ technologies satisfy Assumption 4. Thus, if a ﬁrm is subject to an emission tax, its proﬁt is given by i(qi, ei, qi)P(Q)qi Ci(qi, ei) ei

(4.27)

where qi (q1, . . ., qi1, qi1, . . ., qn) is the strategy proﬁle of the remaining n1 competitors. Firms are assumed to play Nash equilibrium, which due to our model assumptions is unique and stable. We also assume that the Nash equilibrium yields positive amounts of output and emissions, denoted by (q*1 ( ), . . ., q*n ( ),e*1 ( ), . . ., e*n ( ) ). Thus the equilibrium strategies are interior solutions to the ﬁrms’ non-cooperative proﬁt maximization problems and thus satisfy the following ﬁrst-order conditions for all i 1, . . ., n: P(Q* ( ))q*i ( ) P(Q* ( )) Ciq (q*i ( ),e*i ( )) 0

(4.28)

Cie (q*i ( ),e*i ( ))

(4.29)

140

Yearbook of environmental and resource economics

In the following we omit the asterisks and for simplicity write for short qi( ), ei( ) and so on. 4.2

Comparative Statics

In order to study the impact of a tax raise on output and emissions, we can diﬀerentiate the 2n equations (4.28) and (4.29) with respect to to obtain [Pqi P]Q [P Ciqq]qi Ciqeei 0

(4.30)

Ciqe qi Cieeei 0

(4.31)

where for convenience I have again written qi dqi d and so on. Solving the 2n equations for qi and ei yields Ciee[P qi P]Q Ciqe qi C P A ee

ei

(4.32)

i

Ciqe[Pqi P]Q Ciqq P CieeP Ai

(4.33)

where Ai CiqqCiee [Ciqe]2 is the Hessian of the cost function. In contrast to the monopoly case we cannot unambiguously sign qi , since the sign of the ﬁrst term of the numerator of (4.32) is ambiguous. However, by summing over (4.32) and rearranging we obtain Q

n Ci [Pq P] Ciqe ee i · 1 Ci P Ai Ciee P Ai i1 ee i1 n

1

0

(4.34)

Studying these terms, we see that by virtue of Assumptions 1 and 4 Q is negative. This gives rise to the following result: Proposition 3 Under Assumptions 1 and 4, aggregate output is decreasing in . This result is also obtained by Levin (1985) and Ebert (1992) proceeding from more restrictive assumptions. By contrast, signing E is not possible in general, as summing over (4.33) yields E

n Ci [Pq P] Ciqq P qe i Q· i i P A C P A C i i ee i1 ee i1 n

(4.35)

141

Environmental policy under imperfect competition

Here the ﬁrst term is negative, but under Assumption 1 the second term is positive, such that the sign of the whole expression cannot be determined unambiguously. Below, I consider some special cases where we can sign the change of aggregate emissions as a function of the emission tax. It will transpire, however, that in general the sign of dE/d is indeed ambiguous. What about the ﬁrms’ proﬁts? One can show that where ﬁrms are suﬃciently symmetric proﬁts go down if the tax rate increases. Simpson (1995) shows that if ﬁrms are suﬃciently diﬀerent, one ﬁrm may beneﬁt from taxation. Carraro and Soubeyran (1996) indicate detailed conditions under which both uniform and ﬁrm-speciﬁc taxes lead to an increase or decrease in proﬁts and market shares. This eﬀect can be illustrated by a simple duopoly example with constant marginal costs c1 c2, pollution proportional to output, i.e. ei iqi , and linear demand: P(Q)1Q. In this case, proﬁts are given by i (12(cii )cj j )2/9. Thus di/d 0 if and only if j 2i. Thus ﬁrm i beneﬁts from a higher tax rate if and only if the other ﬁrm pollutes double as much per unit of output as ﬁrm i. 4.3

Second-Best Taxation

Let us now consider the regulator’s problem. He maximizes W( )

Q( )

0

P(z)dz

n

Ci(qi( ),ei( ) ) D(E( ))

(4.36)

i1

Diﬀerentiating (4.36) with respect to and employing (4.28) and (4.29) (assuming an equilibrium with all ﬁrms producing positive quantities), we obtain W( )

n

n

n

1

1

Cqi (qi, ei )ei D(E) ei [P(Q) Ciq(qi, ei)]qi i i i1

n

n

i1

i1

P(Q)qi qi [ D(E)]ei

(4.37)

where we have again written qi for dqi /d and so on. Setting W( )0 and ‘solving’ for (note again that the right-hand side also depends on ), we obtain for i 1, . . ., n: Cie (qi, ei ) * D(E)

P(Q) ni1 qiqi E

(4.38)

where E ni1 ei. As we have seen above, E′ can be positive if ﬁrms are extremely asymmetric. But this implies that the second-best optimal tax rate may also exceed marginal damage. The main ﬁndings can be summarized as follows:

142

Yearbook of environmental and resource economics

Proposition 4 Let Assumptions 1, 2, and 4 hold. Then the following is true: (i) If ﬁrms are suﬃciently similar, in the sense that the diﬀerences of their costs |Ci (q, e)Cj(q, e)| (including derivatives) are suﬃciently small (leading to similar equilibrium output and emission levels), then (a) aggregate emissions are decreasing in the emission tax rate; (b) the second-best optimal tax rate falls short of marginal pollution damage; (c) both ﬁrms’ proﬁts fall if the tax rate increases. (ii) If ﬁrms are suﬃciently asymmetric, in the sense that their cost functions are suﬃciently diﬀerent, then (a) it is possible that for particular tax-rate intervals aggregate emissions are increasing; (b) the second-best tax rate may exceed marginal damage; (c) in a duopoly, one ﬁrm at most can beneﬁt from a tax raise. In an oligopoly model with n 2, at least one ﬁrm will incur decreasing proﬁts, but several ﬁrms may enjoy increasing proﬁts when the tax goes up. Thus we see that if ﬁrms are symmetric or suﬃciently similar, we obtain the same results in qualitative terms as in the case of monopoly. If ﬁrms are asymmetric, some perverse eﬀects may arise. The intuition is that if the marginal cost diﬀerential between the ﬁrms is diﬀerent from the diﬀerence in emission coeﬃcients, taxation changes the cost structure between the ﬁrms. This can not only lead to a situation where one ﬁrm gains whereas the other ﬁrm suﬀers from a tax increase, but can also cause aggregate pollution to rise and the tax rate to exceed marginal damage. For the last two eﬀects to arise, however, asymmetry of ﬁrms does not suﬃce. We also need an inverse demand function which has an extreme curvature, i.e. which is either suﬃciently convex or suﬃciently concave (see Levin, 1985 and section 4.3.1B below). 4.3.1

Some special cases

A Symmetric firms If ﬁrms are symmetric, i.e. Ci(·,·) C(·,·) for all i, uniqueness of equilibrium requires a symmetric equilibrium, i.e., qi q Q/n. In this case using (4.31), (4.35) becomes Cqe n[Cqe]2 E Cn C Q Cn

0 Cee[PQ (n 1)P] A ee ee ee and hence eE/n 0.

143

Environmental policy under imperfect competition

The second-best optimal tax rate from (4.38) simpliﬁes to

* D(E) P(Q)q

dq de

(4.39)

P(Q) dq D(E) n de

or:

(4.40)

where dq/de is the reaction of output to the relaxation of an emission cap. In all these equations, the second term is clearly smaller than zero. Therefore, as in the monopoly case, the tax rate falls short of marginal social damage. Since the oligopolistic industry output Q( *) is less than the competitive output, and since e(q, )/q 0, aggregate emissions E( *) are smaller than they are in the case of perfect competition. Hence, the (second-best) optimal emission tax rate is lower for oligopoly compared to the Pigouvian tax rate under perfect competition. Moreover, (4.40) suggests that the second-best optimal tax rate increases if the number of ﬁrms increases.8 This is quite intuitive. The higher the number of ﬁrms, the closer the market outcome is to the competitive outcome and the closer the tax rate is to the Pigouvian level. This is also emphasized by Katsoulacos and Xepapadeas (1996). B No abatement technology (the models by Levin, (1985) and Simpson (1995) This case mainly serves to illustrate that raising an emission tax can indeed cause more pollution. Therefore I assume, as in the Levin model (1985), that ei iqi for each ﬁrm. Let H: ni1 i. The ﬁrms’ cost functions Ci(qi) depend on output only. Moreover, for simplicity I assume Ci C for all i. If ﬁrms face a uniform tax on emissions, their ﬁrst-order proﬁt maximization conditions in Nash equilibrium are given by P(Q)qi P(Q)Ci (qi) i 0

i 1, . . ., n

(4.41)

Carrying out the comparative statics exercise yields E P 1 C

2 PQ (n 1)P C H [PH P E]

n

i

i1

(4.42)

where E ni1i qi. This shows that the sign of E is negative (positive) as P[EH Q

2i ] ( )

i2 [(n 1)P C ] H2P H·E

144

Yearbook of environmental and resource economics

Thus aggregate emissions increase with a rise in tax if the inverse demand function is either suﬃciently concave or suﬃciently convex. Since aggregate output deﬁnitely drops as the tax rate goes up, we can conclude that in Levin’s example a small emissions subsidy will increase welfare since output goes up while emissions go down. However, we cannot conclude from Levin’s result that it is optimal to subsidize pollution whenever E(0) 0. The reason that emissions go up while industry output goes down must be that some ﬁrms with high pollution raise their output whereas those with low pollution cut down on production. However, if the tax rate is set suﬃciently high, so that each ﬁrm’s output goes down, then aggregate emissions also have to go down, compared to the laissez-faire level. It should not be diﬃcult to ﬁnd examples with suﬃciently steep damage functions where a suﬃciently high tax improves welfare in comparison with the laissez-faire level. C Symmetric ﬁrms with pollution proportional to output (Ebert’s Model) In the ﬁrst part of his paper, Ebert (1992) assumes that emissions are completely determined by output, i.e. eq.9 In this case, the emission tax leads to the socially optimal outcome:

* D(E* )

P(Q*)Q* n

P(Q*) D(E*) n

(4.43) (4.44)

Note that the term P(Q)Q/n is equal to the optimal subsidy on output, if we could subsidize output directly.10 Hence we can rewrite (4.43) as * D(E*) /, i.e. the tax equals marginal damage minus the optimal subsidy divided by the emission coeﬃcient, which seems reasonable. The main insights gained in this subsection can therefore be summarized as follows: Proposition 5 If oligopoly is symmetric and ﬁrms do not have an abatement technology, i.e., pollution is determined completely by output, then there is always an emission tax rate that implements the social optimum. (b) In a case like this, the tax can also be charged on output. The optimal output tax tout (or subsidy if it is negative) would then be given by (a)

Q* tout D(Q*) P(Q*) n The last statement is easily veriﬁed.

(4.45)

145

Environmental policy under imperfect competition

4.4

Emission Taxes cum Subsidy System

4.4.1 The basic system We have seen that if emissions are completely determined by output, and ﬁrms are symmetric, the regulator can simultaneously stir emissions and output to the optimal levels. If the ﬁrms can decide separately on output and emissions, the regulator will need at least two instruments to regulate output and emissions. However, if ﬁrms are asymmetric, even the two instruments lead to second-best optimal allocation only. In this section we will consider the case where the regulator has two instruments, an emission tax and a subsidy on output . In this case the ﬁrms’ ﬁrst-order proﬁt maximization conditions, which constitute the Nash equilibrium, are given by P(Q)qi P(Q) Ciq (qi, ei ) 0 Cie (qi, ei )

In the case of a uniform tax/subsidy system, we obtain the following formula for the second-best optimal tax/subsidy system:

D(E) P(Q)

P(Q)

n

n

q qi Q i1 i

q q i i1 i n

E Q dE Q

q qi E i1 i

q q i i1 i n

E Q E Q

Q

(4.46)

E

(4.47)

Furthermore, one can show that qi /

0, ei /

0, qi /! 0, and ei/! 0 if the ﬁrms are not too diﬀerent. In this case too, Q/

0, E/

0, Q/! 0, and E/! 0 hold. If ﬁrms are symmetric, the formulas (4.46) and (4.47) boil down to

D(E*) Q* P(Q) n

(4.48) (4.49)

In this case the optimal tax/subsidy system ( , ) in fact implements the social optimum. This is not surprising as we have two targets q and e, identical for all ﬁrms, and two instruments.

146

4.4.2

Yearbook of environmental and resource economics

Extensions of tax/subsidy schemes

Tax/tax-refunding schemes As set out in section 3.4, subsidies on output are not a realistic policy option for coping with market power. Gersbach and Requate (2004), however, show that in a Cournot model a tax/taxrefunding scheme may mimic such a tax subsidy scheme. They study emission taxes with a system of refunding tax revenues back to ﬁrms according to market shares, as is the case for the NOX emission tax system in Sweden. If ﬁrms compete à la Cournot, the ﬁrms’ objective function is then determined by qi

i (qi, ei, qi, e i ) P(Q)qi Ci (qi, ei ) ei " Q

n

ei i1

where " with 0 "1 is the share of tax revenues to be refunded to ﬁrms. Gersbach and Requate show that if market failure through pollution exceeds the market externality through market power, an optimal tax/taxrefunding scheme mimics an optimal tax/tax-subsidy system. They also show that in symmetric oligopoly, except where the marginal distortion from oligopolistic behavior exceeds the marginal social damage, it is possible to ﬁnd a tax rate and a refunding share " that implement the social optimum. Since the optimal share " is usually smaller than 1, complete refunding, as exercised in Sweden for NOX emissions, is, however, generally not optimal. Leontief technologies Requate (1993a) considers an asymmetric duopoly with linear technologies, i.e. ﬁrms have constant marginal costs c1 c2, and emissions are proportional to output, that is, ei iqi . The interesting case is where the ﬁrm with the lower private cost has the higher emission coeﬃcient, i.e. 1 2. Since emissions are determined by output, Requate ﬁnds that in this case also, there exists a tax/subsidy system that implements the social optimum. Optimal non-linear tax/subsidy schemes We have seen in section 4.4 that a regulator can decentralize the ﬁrst-best allocation by a uniform tax/subsidy system only if ﬁrms are symmetric. Kim and Chang (1993) propose an ingenious non-linear tax/subsidy scheme that works for asymmetric ﬁrms engaging in imperfect competition. Though their scheme is non-discriminatory, it nevertheless leads to the ﬁrst-best allocation. This system, which is a modiﬁed version of the one proposed by Loeb and Magat (1993) to regulate utilities, can even be employed if the regulator is imperfectly informed about the ﬁrms’ technologies. The system works as follows:11 the regulator proposes a

Environmental policy under imperfect competition

147

non-linear tax/subsidy function T(qi, ei, q~ i, e~ i ), where q~ i ji qj and e~ i ji ej represents the aggregate output and emissions of the other ﬁrms with the exception of ﬁrm i. The proﬁt of a typical ﬁrm is then given by i (qi,ei,qi,ei ) P(Q)qi Ci (qi,ei ) T(qi,ei,q~ i,e~ i ) which leads to the following ﬁrst-order conditions in Cournot–Nash equilibrium: P(Q) P(Q)qi Cie (qi, ei ) Tq (qi, ei, q~ i, e~ i )

(4.50)

Cie (qi, ei ) Te (qi, ei, q~ i, e~ i )

(4.51)

i

i

From this we see that if we choose the function T() as follows: T(qi, ei, q~ i, e~ i ) D(ei e~ i ) D(e~ i ) qiP(Q)

qi

0

P(x q~ i )dx

then the Nash equilibrium conditions induce the ﬁrst-best outcome.12 What makes this mechanism so attractive is that it is not only nondiscriminatory but also reduces the regulator’s informational burden. The regulator only needs to know the damage and the demand functions; he requires no knowledge of private ﬁrm data. Of course, each advantage is oﬀset by some disadvantage. The ﬁrms do not know exactly how large their ﬁnal costs will be since their tax bill (or subsidy) depends not only on their own decision but also on the choices about output and emissions made by the other ﬁrms. Note, however, that this is also the case under a system of tradable permits. 4.5

Standards

As I shall be dealing with regulation by permits in a separate section (7), I here brieﬂy discuss the two kinds of standards that have already been introduced in section 3. Actually, for symmetric oligopoly, the results are not very diﬀerent from the monopoly case. For simplicity, I will only treat the symmetric case here, restricting my attention to instances where the ﬁrms’ technologies satisfy Assumption 4. As under an absolute emission standard no ﬁrm is allowed to emit more than e units of emissions, the Cournot–Nash equilibrium of a symmetric industry is determined by just one equation: P(nq) P(nq)q Cq (q,e) 0

(4.52)

148

Yearbook of environmental and resource economics

This yields symmetric individual output q(e) . It is easy to show that if the standard is binding, output goes up when the standard is relaxed, i.e. q(e) 0. The regulator again maximizes welfare with respect to e, yielding Ce (q, e) D(ne) P(Q)qq(e) Accordingly, the marginal abatement costs are once again lower than marginal damage. If the right-hand side of the last equation is positive, taxes and standards are equivalent, which is not surprising in the light of the monopoly model. A relative standard restricts emissions per units of output: e q. If the standard is binding, the ﬁrms’ proﬁts can be written as: P(Q)qC(q, q), as in the monopoly case, and the Nash equilibrium condition in symmetric oligopoly is given by P(Q)qP(Q)Cq(q, q)Ce(q, q)0

(4.53)

Again, the comparative statics eﬀect dq/d is ambiguous, although we intuitively would expect dq/d 0, i.e. a stricter standard would lead to lower output per ﬁrm. The second-best optimal standard satisﬁes the following condition:

dq dq Ce (q, q) D(E) 1 q d P(Q) d We see that there is no general way of determining whether or not the marginal abatement costs exceed or fall short of marginal damage. In the ‘normal’ case, i.e. dq/d 0, the strategic term P(Q)(dq)/(d) is negative but the multiplier 1 (dq)/(d) is greater than 1. The marginal abatement cost may now exceed marginal damage. The intuition is the same as in the case of pure monopoly: the relative standard can be met by reducing emissions or extending output. The latter strategy is not conductive to the protection of the environment. In order to work against such a strategy, the regulator using an ineﬃcient instrument has to devise a stricter standard than in the case of an absolute emission standard. If we have the ‘perverse’ eﬀect of dq/d 0, the strategic term is positive but the multiplier 1(dq)/(d) is smaller than 1. Thus the two eﬀects (dq)/(d) and P(Q)(dq)/(d) always work in opposite directions.

Environmental policy under imperfect competition

4.6

149

Further Extensions of Emission Taxation in Cournot Oligopolies

4.6.1 Emission taxes and endogenous market structure In the model with constant marginal costs c1 c2 and emission coeﬃcients 1 2, referred to above, Requate (1993a) also discusses endogenous market structure. That is, depending on the size of the tax rate, the outcome can be either monopoly or duopoly. For this purpose, Requate writes the damage function as D(E, s), with Ds 0 and DEs 0. The parameter s determines the slope of the damage function and can be interpreted as a damage parameter where higher s leads to both higher damage and higher marginal damage.13 As is intuitively obvious, I show that it is socially optimal for ﬁrm 1 only to produce if the damage function is suﬃciently ﬂat, for ﬁrm 2 only to produce if the damage function is suﬃciently steep, and for both ﬁrms to produce in the case of moderately steep damage functions. I use [ _s , s ] to denote the interval of damage parameters at which it is optimal for both ﬁrms to produce. I show that the ﬁrst-best allocation can only be induced by an emission tax if the damage parameter is suﬃciently low, i.e. for s _ _s (where _s is some parameter smaller than s ), in which case only ﬁrm 1 will _ _produce, or if the damage parameter is suﬃciently high, i.e. for s s (where s is some parameter greater than s), in which case only the cleaner ﬁrm will produce. As mentioned above, the ﬁrst-best allocation can be restored for each s by a suitable tax/subsidy system. 4.6.2 Inter-ﬁrm externalities Yin (2003) presents an important extension of the oligopoly model by considering externalities between the producers. This model is based on the assumption that the ﬁrms’ costs are raised by total pollution. For both Cournot and price competition with diﬀerentiated commodities, Yin ﬁnds that the second-best optimal tax rate exceeds marginal damage (which here is determined by the ﬁrms’ productivity and proﬁt loss caused by the externality) if the externality is considerable and the number of competing ﬁrms is large. More speciﬁcally, this is the case if and only if the tax eﬀects on total pollution and on total output go in opposite directions. Interestingly, when output increases by raising the tax rate, there is no trade-oﬀ between environmental quality and consumer surplus. 4.6.3 Pre-investment Carlsson (2000) extends the duopoly model proposed by Simpson (1995) to allow for pre-investment in abatement capital with a view to reducing future emissions from production. The ﬁrms’ proﬁt functions are given by i (q1, q2 ) P(q1 q2 )qi Ci (qi, xi ) rxi tei (qi, xi )

150

Yearbook of environmental and resource economics

where qi denotes output, xi the amount of abatement capital to be invested in the ﬁrst period, r the price of capital, and e(qi, xi) the emission function. Firms play one-, two- or three-stage games, respectively. In the ﬁrst game, ﬁrms play open-loop strategies by committing simultaneously and irreversibly to the levels of both investment and production. The second game is the usual two-stage game where ﬁrms invest in the ﬁrst stage and compete à la Cournot in the second stage. In the third game, one ﬁrm is a Stackelberg leader, with the ﬁrst-mover advantage of choosing the level of capital. The authors ﬁnd that if abatement capital enhances the marginal costs of production, then in the open-loop game the second-best optimal tax rate falls short of marginal damage for ‘normal’ parameters. The same holds true for the closed loop game, except in the case where ﬁrms are extremely diﬀerent. By contrast, little can be said about the Stackelberg case. 4.6.4 Dynamic model with accumulating pollutants Benchekroun and Van Long (1997) extend the duopoly models proposed by Simpson (1995) and Katsoulacos and Xepapadeas (1996) to the case of an accumulating stock pollutant leading to a dynamic model with an inﬁnite time horizon. Two ﬁrms compete à la Cournot in the ﬁnal output market, and their emissions are subject to taxation. Firms are symmetric and emissions are proportional to output. The ﬁrms may employ either open-loop or Markov-perfect (feedback) strategies. Benchekroun and Van Long ﬁnd that in both cases a time-independent tax rule exists, i.e. a tax rule depending on the current stock of pollution only, which leads to a socially optimal outcome. This is not surprising in the light of Proposition 5 from the static model. The tax rate may be negative, i.e. it turns into a subsidy in the initial time interval when the pollution stock is still low. Surprisingly, however, this subsidy induces the ﬁrms to produce less than they would in the case of laissez-faire. The reason is that they know that if they produce more, then the subsidy will be reduced in future and/or will (sooner) be converted into a tax. 4.6.5 Durable goods Runkel (2002, 2004) investigates the taxation of oligopolists producing a durable good that creates waste after it has expired. He ﬁnds that if the durability of the good is exogenous, the second-best tax rate in the second period falls short of marginal damage (as usual), whereas in the ﬁrst period the emission tax rate may be higher or lower than marginal damage. If durability is endogenous, over-internalization may also occur. 4.6.6 Ecological tax reforms under imperfect competition Imperfect competition in the output market has also been considered in the recent literature on ecological tax reforms. After the enthusiasm about the

Environmental policy under imperfect competition

151

existence of a double dividend was dampened by Bovenberg and De Mooij (1994) and several other papers by Bovenberg and co-authors14, some researchers have tried to rescue the idea of a double dividend by introducing imperfect competition into the models. In a model with monopolistic competition on the output market, Marsiliani and Renström (1997) show that besides the environmental dividend a second dividend on the labor market caused by boosting employment may or may not arise as a result of a revenue-neutral ecological tax reform. Holmlund and Kolm (2000) come to a similar conclusion. Bayindir-Upmann (2004) investigates a general equilibrium model with Cournot competition on the output market for a dirty consumption good. Assuming sticky nominal wages and consumers displaying Cobb–Douglas preferences with respect to clean and dirty goods, Bayindir-Upmann shows that a double dividend resulting in both a reduction in consumption of the dirty good and more employment can be achieved if both the initial labor tax rate and the share of income spent on the dirty good are suﬃciently low. However, for the more interesting and more relevant case involving a high initial labor tax rate and a high consumption level for the dirty good, a double dividend does not exist. This is not because raising the tax rate on the dirty good further increases unemployment, but rather because consumption of the dirty good rises so that the environmental dividend is lost. This conversely implies that lowering the tax rate on the dirty good will improve environmental quality but at the same time raise unemployment. The range of parameters for which a double dividend exists shrinks with the degree of competition. Hence the results are less optimistic than in the case of pure monopoly, which is discussed in Bayindir-Upmann (2000). This result is also consistent with the ﬁndings of Fullerton and Metcalf (2002). It is worth mentioning that many authors of ecological tax-reform literature only consider the comparative statics eﬀects of a revenue-neutral ecological tax reform and do not investigate the second-best optimal tax structure.15 4.6.7 Financial structure of ﬁrms and emission taxes Damania (2000) presents an oligopoly model in which he explicitly takes account of the ﬁrms’ ﬁnancial structure. He considers a regime in which the regulator ﬁrst makes a commitment to his tax rate, and then the ﬁrms play a three-stage game. In the ﬁrst stage they choose the ﬁnancial structure (level of leverage), in the second they choose their level of abatement (or their abatement technology), and ﬁnally they decide on the level of output. Damania shows that higher tax rates can lead to higher output and emission levels. In that model the reason is that an increase of emission taxes

152

Yearbook of environmental and resource economics

increases the probability that ﬁrms go bankrupt, so ﬁrms will focus only on the solvent states of the world. This in turn encourages ﬁrms to increase output and emission levels. One result of this may be that increasing emission taxes leads to the adverse eﬀect of raising emissions. 4.6.8 Dynamic approach with capital accumulation in clean and dirty technology Stimming (1999) studies the dynamic implications of emission taxes by setting up a Cournot duopoly model where ﬁrms can accumulate capital for dirty and clean technologies, respectively. She ﬁnds that, for both a tax and a permit regime, a stricter policy induces ﬁrms to reduce investment in the dirty technology, whereas the eﬀect on the clean technology is ambiguous. As expected, output and emissions go down. A perverse eﬀect – stricter environmental policy increasing aggregate emissions – can result if the ﬁrms are regulated diﬀerently, one by a tax, the other by a standard. 4.6.9 Dominant ﬁrm with a competitive fringe Besides perfect competition and symmetric Cournot oligopoly, Conrad and Wang (1993) also consider the case where the market is governed by both a dominant ﬁrm and a competitive fringe. The model is thus similar to Innes et al. (1991), to be discussed in section 9. The authors only consider the comparative statics eﬀects, they do not investigate the second-best optimal tax rule. The results are as expected. If both the dominant and the competitive ﬁrms are subject to the same tax rate, a tax increase induces a decrease of both output and emissions from both types of ﬁrm. If the fringe ﬁrms are foreign ﬁrms not subject to taxation, the eﬀect of an increase in domestic tax may be oﬀset by an increase of foreign emissions (the leakage eﬀect). The authors also investigate abatement subsidies in models of symmetric oligopoly and monopoly, alongside their dominant ﬁrm model. They ﬁnd that, contrary to the case of perfect competition with a ﬁxed number of ﬁrms, increasing the subsidy leads to an increase in output. Hence emissions may rise or fall. 4.6.10 Diﬀerentiated emission taxes Van Long and Soubeyran (2005) study asymmetric Cournot oligopoly with diﬀerentiated taxes. They also take account of the marginal costs of public funds such that the regulator also has a motive for collecting taxes. They ﬁnd that high-cost ﬁrms should be taxed at a higher rate. Moreover, the optimal tax structure leads to an increase in market concentration as measured by the Herﬁndahl index.

Environmental policy under imperfect competition

5.

153

EMISSION TAXES IN PRICE-SETTING DUOPOLY

In this section I treat oligopoly models where ﬁrms use prices as strategic variables. I start by outlining the case of genuine Bertrand competition, i.e. price competition where commodities are homogeneous and ﬁrms incur constant marginal costs without being capacity-constrained. Then in section 5.2, I discuss the case of price competition with diﬀerentiated commodities. I have elected to restrict analysis to the duopoly case for several reasons. In the genuine Bertrand case, more than two ﬁrms do not lead to any further insights. In the case of diﬀerentiated commodities, there are no general symmetric models of price-setting oligopoly with diﬀerentiated commodities. Moreover, asymmetric models with more than two ﬁrms would require a large amount of notational clutter without yielding any essential additional insights. 5.1

Bertrand Duopoly with Homogeneous Commodities

Requate (1993c) studies price competition among ﬁrms that supply a homogeneous good and have constant marginal costs without being capacity-constrained. This kind of competition is usually referred to as real Bertrand competition. For this purpose I use the same model as in section 4.6.1, i.e. two ﬁrms producing with constant (asymmetric) marginal costs c1 c2 and emitting a pollutant proportional to output, i.e. ei iqi. I consider both cases, where the ﬁrm with lower production cost, i.e. ﬁrm 1, is also the cleaner ﬁrm, i.e. 1 2, and the case where it is the worse polluter, i.e. 1 2. The ﬁrst case is not very interesting since ﬁrm 2 will be out of business for any level of tax rates. Therefore we focus on the second case, i.e. 1 2. Tough competition (of this kind) implies that the tax can always induce only one ﬁrm to produce whenever this is socially optimal. However, the regulator is not able to enforce optimal allocation between the two ﬁrms when it is socially optimal for both ﬁrms to produce. I shall brieﬂy outline the argument. The ﬁrms have the same technologies as in section 4.6.1 but now set prices denoted by p1 and p2 rather than quantities. To determine the ﬁrm’s demand, I follow the standard Bertrand model: if ﬁrms charge prices p1 and p2, ﬁrm i’s demand is given by

G(pi) Gi (pi, pj ) : G(pi)/2 0

if pi pj if pi pj if pi pj

(4.54)

154

Yearbook of environmental and resource economics

where G(p) is the market demand function.16 This deﬁnition is based on the assumption that consumers are perfectly informed about the prices, that they always buy from the cheaper ﬁrm if prices diﬀer and split up equally if prices are equal, and that the ﬁrms are not capacity-constrained. The ﬁrms’ demand functions will be diﬀerent if we consider regulation by permits, which naturally imposes capacity constraints on the ﬁrms. This will lead us to Bertrand–Edgeworth rather than Bertrand competition and will require a rationing rule. I will return to that case in section 7.3. In a price-setting game like this, it is well known that for symmetric ﬁrms there is a unique Bertrand–Nash equilibrium where ﬁrms charge a price equal to marginal cost. Where, say, c1 c2 and c2 p1m, in which pm 1 is ﬁrm 1’s monopoly price, I follow the industrial organization literature and take p1 p2 c2 as the unique Bertrand–Nash equilibrium price. Hence I will call p min{pm i , cj} the Bertrand equilibrium price if ci cj. Suppose now that a uniform linear tax is imposed on emissions such that the ﬁrm’s marginal cost amounts to ci i . If pim( ) : arg maxp{[p (ci i )]G(p)} denotes the monopoly price under an emission tax, the market price under Bertrand competition is given by ˇ

p( ) min{cj j, pm i ( )}

(4.55)

From this it can readily be seen that the regulator can always induce only one ﬁrm to produce if it is optimal to do so. In this case, the regulator can even induce the ﬁrst-best outcome by setting the tax equal to

P(q*i ( ))q*i ( )

max ci cj iD(i q*i ),D(i q*i ) j

i

when it is optimal for only ﬁrm i to produce. Note that the ﬁrst term in brackets refers to the case where the ﬁrm to be regulated engages in limit pricing, while the second term refers to monopoly pricing. As mentioned above, the ﬁrst-best allocation cannot be induced if it is optimal for both ﬁrms to produce, because in this case the social optimum requires the following relationship to hold: c1 1D(1q*1 2q*2 ) c2 2D(1q*1 2q*2 ) Setting the tax rate equal to marginal damage would induce identical marginal costs. The demand would then split up equally among the ﬁrms, and this does not necessarily correspond to the socially optimal allocation. Requate (1993c) characterizes the second-best optimal tax for such a case

155

Environmental policy under imperfect competition

in more detail. The tax rate turns out to be discontinuous as a function of the damage function’s slope. Note also that, contrary to the Cournot case, the ﬁrst-best allocation cannot be induced by a tax-subsidy system in the Bertrand case. 5.2

Price-Setting Duopoly with Diﬀerentiated Commodities

Lange and Requate (1999) study price-setting duopoly with diﬀerentiated commodities. They ﬁnd that the results obtained previously, i.e. that the second-best optimal tax rate should be set below marginal social damage, is not corrupted under imperfect price competition unless ﬁrms are extremely diﬀerent. This holds true irrespective of whether the commodities are substitutes or complements. Only if the ﬁrms are extremely diﬀerent with respect to both their technology and their market demand functions is it possible for the second-best optimal tax rate to exceed marginal social damage. 5.2.1 Outline of the model There are two ﬁrms oﬀering diﬀerentiated commodities. The utility of a representative consumer is given by a separable quasi-linear utility function uU(q1, q2)q0 D(e1, e2)

(4.56)

where commodity 0 is a numeraire with the price normalized to 1. The subutility function U is quasi-concave and monotonically increasing. Further assumptions are imposed on the Marshallian demand functions for the two commodities 1 and 2. We use e1 and e2 to denote the ﬁrms’ emissions, as deﬁned below. D(e1, e2) denotes the disutility (or damage) caused by the pollution from the two ﬁrms. The utility-maximizing consumer clearly sets Ui U(q1, q2)/qi pi where pi is the price of commodity i. Let Gi(p1, p2) denote the solution of this problem, i.e. the demand for commodity i. The two commodities are substitutes (complements) if Gij Gi pj 0( 0) holds. Further, the authors assume that Gij Gi pj 0 , and that ﬁrm i’s revenue pi G i is concave in pi. The ﬁrms produce with constant marginal costs ci #0, and generate pollution ei iqi proportional to output. If the ﬁrms have to pay an individual tax i (which may be uniform as a special case), the ﬁrms’ proﬁt is given by ˇ

i(p1, p2)(pi ci i i) Gi (p1, p2)

ˇ

ˇ

(4.57)

156

Yearbook of environmental and resource economics

5.2.2 Diﬀerentiated emission taxes The Nash equilibrium of the simultaneous price-setting game is then determined by the following set of equations: Gi (p1, p2 ) (pi ci i i )Gii (p1, p2 ) 0 for i 1,2

(4.58)

The comparative static eﬀects of diﬀerentiated taxes are quite intuitive. Proposition 1 (i) pi / i 0 for i1, 2. (ii) pi / j 0 if commodities are substitutes and pi / j 0 if commodity i is a complement to commodity j. The second-best optimal set of diﬀerentiated taxes can be characterized as follows: pi Gi D

i D ei iGii ei ii

(4.59)

where i Gii pi G i denotes the demand elasticity for commodity i. Note that formula (4.59) provides the same structure for the second-best optimal tax rates as in the pure monopoly case (see section 3). Of course, the two rules for 1 and 2 given by (4.59) are not independent of each other. But if both marginal damage and demand elasticity are known or can be determined empirically, rule (4.59) is easy to handle if taxes can be diﬀerentiated, especially if the two ﬁrms supplying diﬀerent commodities emit diﬀerent pollutants. 5.2.3 Uniform emission tax If the regulator can set a uniform tax only, the comparative static eﬀects are less clear-cut. Lange and Requate obtain the following result: Proposition 2 (i) If the commodities are substitutes, both prices will increase if the tax rate goes up. (ii) If the commodities are complements and the ﬁrms have asymmetric cost and demand structures, one of the prices may go down. The second-best optimal tax rate is now more complicated: 1

2

D dG D dG G dG G dG 1 e d 2 e2 d G1 d G2 d

1 1

1 dG2

1dG d 2 d

1

1

2 2

2

(4.60)

Environmental policy under imperfect competition

157

~

If the pollutant is uniform, i.e. if D (e1, e2 ) D (e1 e2 ) implying ~ De1 De2 D , we obtain ~

D

2 2 G1 dG1 GG 2 dG 1 d

G1 d

2 dG1 2 1 d 2dG d

(4.61)

If ﬁrms are suﬃciently similar, i.e. G1 G2 and G11 G22 as well as 1 2, the formula for the optimal tax rate approximates i ~

D G i iGi

(4.62)

Note that this is independent of whether the two commodities are complements or substitutes. If ﬁrms are extremely diﬀerent, however, the second term on the righthand side of (4.61) taking account of the strategic interaction between the ﬁrms may be positive, resulting in a second-best optimal tax rate that exceeds marginal damage. Lange and Requate (1999) present a numerical example where this is in fact the case. The reason for the high tax rate in that example is ﬁrm 1’s extreme advantage with respect to private cost. In order to oﬀset this advantage and to cut down emissions, the regulator has to choose a tax rate that is higher than marginal damage.

6. MONOPOLISTIC COMPETITION AND FREE ENTRY In this section I summarize models of imperfect competition where the number of ﬁrms is determined endogenously. The industrial organization literature oﬀers three prototype models of this kind: Cournot competition with free entry, the Dixit–Stiglitz model of product diﬀerentiation, and Salop’s model of the circular city. Katsoulacos and Xepapadeas (1995), Requate (1997), and S.-H. Lee (1999) all investigate the Cournot model with free entry by polluting ﬁrms. Lange and Requate (1999) discuss the two remaining models of monopolistic competition. In all these papers, the authors proceed on the assumption that an emission tax is the only policy instrument available to the regulator. Therefore it is important to emphasize that this instrument now has to deal with three market imperfections: ﬁrms pollute, their prices are higher than is socially optimal, and the number of ﬁrms is not optimal (in general). In particular, the Cournot model leads to excessive market entry by ﬁrms in the absence of regulation.

158

Yearbook of environmental and resource economics

Therefore it may be the case that the second-best optimal tax rate exceeds marginal social damage, which contrasts with the results obtained so far. 6.1

Emission Taxes in Cournot Oligopoly with Free Entry

The material of this subsection is based on Requate (1997). Katsoulacos and Xepapadeas (1995) study a special case of this model, assuming linear demand and a cost function additively separable into output and emissions. S.-H. Lee (1999) replicates the results of Requate (1997) focusing on the case where pollution is proportional to output. Here, we will study both cases, i.e. where ﬁrms have abatement technologies and where emissions are completely determined by output. For the latter case we obtain the neat result that the second-best optimal tax rate equals marginal damage if demand is linear. 6.1.1 Basic assumptions and the ﬁrms’ behavior Throughout this section I assume that ﬁrms are symmetric.17 Since the number of ﬁrms is determined endogenously by a zero-proﬁt condition, we need to explicitly take account of ﬁxed costs. Hence we could in principle split up the costs into variable costs () and ﬁxed costs F, that is: C(q) v(q) F if emissions are proportional to output, that is, e q, and C(q, e) v(q, e)F if the cost function satisﬁes Assumption 4. Further, I only consider the case where the regulator moves ﬁrst by making a commitment to his tax rate. In the second stage, ﬁrms decide whether or not to enter the market, and in the third stage they engage in Cournot competition. For the case where emissions are proportional to output, the Nash equilibrium condition in the last stage (assuming the existence of an interior Cournot–Nash solution with output q* 0) is given by18 P(Q*)q* P(Q*) C(q*) 0

(4.63)

In the second stage of the regulation game, a number n* of ﬁrms enters the market until all ﬁrms earn zero proﬁts: P(n*q*)q* C(q*) F q* 0

(4.64)

If abatement is possible, i.e. if the ﬁrms’ cost functions satisfy Assumption 4, a Nash equilibrium in the last stage (again assuming an interior equilibrium q* 0, e* 0) is characterized by P(Q*)q* P(Q*) Cq (q*, e*) 0

(4.65)

Ce (q*, e*)

(4.66)

Environmental policy under imperfect competition

159

Free entry in the second stage yields P(n*q*)q* C(q*,e*) F e* 0

(4.67)

6.1.2 The government’s problem Employing the usual procedure of welfare maximization leads to the following second-best optimal tax formula:

D(E) P(Q)

Q·(dqd ) dEd

(4.68)

The sign of dE/d is ambiguous in general. However, one can derive the following result: Proposition 6 Under symmetric oligopoly with free entry where the ﬁrms’ technologies satisfy Assumption 4, the second-best optimal emission tax rate exceeds marginal social damage if (i) Cqe 0 and P is (weakly) convex, or if (ii) Cqe qCeee0 and P is linear. In the case of emissions proportional to output, one can even show that the following holds: Proposition 7 In a symmetric oligopoly with free entry and emissions proportional to output, the second-best optimal emission tax rate is given by PQPq

D(E) [C 2P]

(4.69)

The second-best optimal tax rate (a) exceeds marginal damage if demand is strictly concave, (b) falls short of marginal damage if demand is strictly convex, (c) is equal to marginal damage if demand is linear. We see that if no abatement technology exists, the second-best optimal emission tax rate equals marginal damage for linear demand despite imperfect competition. Thus we obtain the same result as under perfect competition, which again is interesting in the light of Buchanan’s (1969) early attack on the Pigouvian tax rule. This result also contrasts with Katsoulacos and Xepapadeas (1995), who ﬁnd that for linear demand the second-best optimal

160

Yearbook of environmental and resource economics

tax rate exceeds marginal damage if the marginal abatement costs are independent of the level of output. Note that in the case where no abatement technology exists we consider a situation where the regulator has only one instrument to regulate two market imperfections: the wrong quantity of output (and hence pollution), plus excessive entry by ﬁrms. Little can be said in general about whether output is too high or too low. On the one hand, individual ﬁrms hold down output due to imperfect competition; on the other, individual ﬁrms do not account for the social damage caused by pollution, which in this case is strictly proportional to output. If the number of ﬁrms were ﬁxed or regulated by another device, such as a license scheme (as is the case in many taxi markets), the regulator could implement the ﬁrst-best outcome precisely by taxing either emissions or output. In the case of free entry, however, the potential rents earned by virtue of imperfect competition attract more ﬁrms. Since there is excessive entry by ﬁrms, they produce and thus pollute to a higher degree than is optimal, and they dissipate ﬁxed costs. To mitigate this excess entry eﬀect, the regulator has to set the second-best optimal tax rate higher than in the case where the number of ﬁrms is exogenous. Whether this tax rate is higher or lower than marginal damage depends, as we have seen, on the curvature of the inverse demand function. Note that in the case where demand is linear and emissions are proportional to output (Proposition 7), we can neither conclude that the emission tax implements the ﬁrst-best outcome nor that the second-best optimal emission tax rate regulating an oligopoly with an endogenous number of ﬁrms is the same as for regulating a competitive market. Since under imperfect competition the ﬁrms price higher than marginal cost, they produce less and accordingly also pollute less than under perfect competition. Hence, in oligopoly with an endogenous number of ﬁrms and linear demand, the second-best optimal tax rate has to be set lower than in the case of perfect competition. The results summarized so far have been derived under the assumption that all the existing ﬁrms are identical. Requate (1997) also discusses the case of several possible technologies. Even though generically only one technology will survive, there may exist other potential, possibly cleaner technologies that may be used if a suitable environmental policy is implemented. Requate (1997) shows that for at most one particular tax rate two diﬀerent ﬁrms can be active at the same time. However, although from a social point of view it may be optimal for diﬀerent types of ﬁrms to share the market, a regulator will not in general be able to enforce the desired technology mix by setting an appropriate Pigouvian tax. The reason is that

Environmental policy under imperfect competition

161

under free entry a multiplicity of equilibria exists. The various equilibria, however, lead to diﬀerent levels of pollution, some with excessive, others with too little pollution.19 6.2

Standards

In this section we brieﬂy investigate standards in oligopoly with free entry. We start with absolute emission standards and then look at relative standards. I only consider the case where the ﬁrms’ technologies satisfy Assumption 4. 6.2.1 Absolute emission standard In addition to the ﬁrst-order Nash equilibrium condition (4.52), the zeroproﬁt condition must again hold. The comparative statics exercise now yields: dq PCqeq Ce[P Pq] de Pq[2P P q Cqq] dn (n 1)PCqeq Ce[(n 1)P nP q Cqq ] 0 Pq[2P P q Cqq] de The last eﬀect implies that a stricter emission standard leads to market exit. By contrast, the eﬀect on output is ambiguous. The second-best optimal standard satisﬁes the following condition:

dq Ce (q,e) D(E) 1 ne dn P(Q)q de de This gives us the following result: Proposition 3 If a Cournot oligopoly with free entry is regulated by an absolute emission standard, then (i) the second-best optimal standard can induce marginal abatement costs to be greater or smaller than marginal damage; dq (ii) if 0, then marginal abatement costs will exceed marginal damage. de 6.2.2 Relative standards Next I consider the eﬀect of a relative standard restricting units of emissions per output: eq. If the standard is binding, the ﬁrms’ proﬁts can be

162

Yearbook of environmental and resource economics

written as P(Q)qC(q, q). The equilibrium is now given by the Nash condition (4.53) and the zero-proﬁt condition. The comparative statics exercise yields: dq Ce[P Pq] P[Ce q(Cqe Cee )] d P[Pq 2P (Cqq 2Cqe Cee )] dn Ce[nPq (n 1)P (Cqq 2Cqe Cee )] d Pq[Pq 2P (Cqq 2Cqe Cee )]

(n 1)P[Ce q(Cqe Cee )] Pq[Pq 2P (Cqq 2Cqe Cee )]

Now both dq/d and dn/d are completely ambiguous. However, the change of total output is given by dQ d[nq] Ce[P (Cqq 2Cqe Cee )] P[Ce q(Cqe Cee )] d d Pq[Pq 2P (Cqq 2Cqe Cee )] This term cannot be signed in general but is positive for quadratic cost functions.20 The second-best optimal relative standard now satisﬁes the following condition:

dQ P(Q) dq Ce (q, q) D(E) 1 Q d d If we compare the term for dq/d to dQ/d, we see that dQ/d is larger than dq/d if P0 or if |P| is suﬃciently small. Hence, if dq/d is positive (which is the case for example for quadratic cost functions) the term (/Q)[(dQ)/(d)] D(E) dominates the term P(Q)(dq)/(d) which accounts for the distortion resulting from imperfect competition. The intuition here is that the oligopoly rent attracts more ﬁrms than is optimal, and due to the relative standard more ﬁrms induce more pollution. Therefore the standard has to be set more strictly than in the case where the number of ﬁrms is exogenous, meaning that the marginal abatement costs exceed marginal damage. 6.3

The Dixit–Stiglitz Model

In this subsection, I present the implications for environmental policy in the prototype model of imperfect price competition with free entry, i.e. the

Environmental policy under imperfect competition

163

Dixit–Stiglitz–Spence model. I extend that model by assuming that the diﬀerentiated commodity is produced by emitting a pollutant that is proportional to output. The pollutant is subject to taxation. The social damage is measured in units of the numeraire commodity. We establish that the second-best optimal tax rate is always lower than marginal damage. Contrasting with the Cournot model this suggests that the Dixit–Stiglitz model does not lead to excess entry in a way requiring the regulator to set a tax rate above marginal damage. Moreover, we ﬁnd that the more competition we have, the closer we should set the second-best optimal tax rate to marginal social damage. These results hold true under rather general conditions. On the one hand, the result is quite intuitive and is in line with what we already know about the regulation of monopolistic ﬁrms. On the other, the result is not completely obvious in the light from the ﬁndings from the Cournot model with free entry. 6.3.1 Outline of the model In the Dixit–Stiglitz model, the representative consumer draws utility from n1 commodities, a compound commodity I that is supplied in n diﬀerent varieties, and a numeraire commodity 0. The consumer also suﬀers from n the aggregate level of pollution E i1 ei, where ei is the amount of pollution generated by ﬁrm i. The damage from pollution, measured in units of the numeraire commodity, is denoted by D(E). Thus we can write the utility as uU

i

qi

1

, q0 D(E)

where q0 M ni1 piqi E is the consumption of the numeraire commodity, i.e. gross income M minus expenditures for the commodities i 1, . . ., n, plus tax revenues that are redistributed to the consumer in a lump-sum way. The price for commodity i is denoted by pi, while the price for commodity 0 is normalized to 1. Moreover, it is usually assumed that the numeraire commodity 0 and the compound commodity I are normal goods. Then utility maximization leads to the following relationship: piU0

qi i

11

q1 UI i

for i1, . . ., n

where U0 and UI are the partial derivatives of utility with respect to the numeraire and the compound commodity, respectively. If n is large, a

164

Yearbook of environmental and resource economics

change of price pi and thus a change in demand for commodity i has little eﬀect on nj1 qpj and hence little eﬀect on U0 and UI.21 Accordingly, the demand for commodity i can be approximated by 1 1

qi (pi ) k·pi

(4.70)

where k is a constant. The ﬁrms produce at constant marginal cost c 0 and incur ﬁxed costs F 0. To keep the model simple, pollution is assumed to be proportional to output. Therefore without loss of generality we can identify pollution with output. If the government charges a tax on pollution, a typical ﬁrm’s proﬁt – if it decides to enter the market – is given by 1 p1

i (pi c )qi (pi ) F (pi c )k · pi

F

Proﬁt maximization leads to the monopoly price pi

(c )

Zero proﬁt through free entry implies F s(c )

qi

(4.71)

where s 1/1. Thus a symmetric equilibrium consisting of a price p, a ﬁrm’s output q, and the number of ﬁrms n as endogenous variables is represented by the following equations: (c )

(4.72)

F s(c )

(4.73)

p q

UI (A, B) (c ) ns U0 (A, B)

(4.74)

165

Environmental policy under imperfect competition

where, using (4.71) and employing symmetry in the consumption of the diﬀerentiated goods, the expressions A and B are deﬁned as:

A I npq n q D(nq) I n F c

F D n F s(c ) s(c )

Bn1/ q We are now ready to investigate the comparative statics eﬀects of raising the tax rate and look for the structure of the second-best optimal tax rate. 6.3.2 The eﬀect of increasing the tax rate One interesting aspect is the question of how an introduction or increase of an emission (or output) tax aﬀects the endogenous variables of the model. Unfortunately, these eﬀects are quite ambiguous. It may even be the case that both the number of ﬁrms and total emissions will rise as a result of increasing the emission tax.22 If we assume a fully quasi-linear utility function by specifying U(q0 D(E), qI)q0 D(E)V(qI), where qI n1/ q, we can say rather more about the relationship between the size of the tax rate and the number of ﬁrms. Proposition 4 Denoting the elasticity (V(qi )qi )V(qi ), we obtain

of

marginal

utility

by

dn/d

0 if and only if 0 1 Note, however, that if we assume that demand for each particular product decreases as the number of ﬁrms goes up, i.e. product diversity increases (the price being held ﬁxed), it follows that23 1 0

(4.75)

If we further assume that 1,24 we obtain n 0. Both assumptions together also guarantee the existence of a ﬁnite number of products ( ﬁrms) in the ﬁrst-best allocation, given an arbitrary social damage function. Hence dn/d

0 is more likely to hold, although dn/d 0 cannot be excluded.

166

Yearbook of environmental and resource economics

6.3.3 The second-best optimal tax rate We are now able to make a general statement about the second-best optimal tax rate: Proposition 8 If in the Dixit-Stiglitz model pollution is proportional to output and the emission tax is the government’s only regulatory device, the second-best optimal emission tax rate will be smaller than marginal social damage. Interestingly, it is not possible to obtain a rule similar to the case of monopoly or oligopoly stating that the ‘tax rate is equal to marginal social damage plus a (negative) term taking account of imperfectly competitive behavior’. Rather, one can show that (c )/(cD) is smaller than 1, implying

D. In contrast to the Cournot model, this suggests that the Dixit-Stiglitz model does not lead to excess entry in a way requiring the regulator to set a tax above marginal social damage. Note that a complete analysis of excess entry at this stage is not possible. In general there are several conﬂicting forces that lead to a deviation from the social optimal level in the number of ﬁrms. Spence (1976) analyzes the problem of excess entry in a standard model of monopolistic competition, i.e. without social damage caused by production. Even in this context, the question of excess entry cannot be answered generally but only for some special cases. In addition, in our model the impact of an increase in the tax rate on the number of ﬁrms is ambiguous. Hence we are not able to compare the number of ﬁrms to the socially optimal level. For the case of quasi-linear utility functions, Lange and Requate (1999) show the following neat convergence result: Proposition 9 Assume that 1 0 for close to 1. Then, the higher the degree of competition, i.e. the better substitutes the goods are (the closer is to 1), the closer the second-best optimal tax rate is to marginal social damage. In the limit, i.e. 1, when the commodities become perfect substitutes, the second-best optimal tax rate coincides with marginal social damage. The results obtained in this section conﬁrm generally accepted wisdom on second-best taxation for imperfect competition, i.e. the second-best optimal tax rate falls short of marginal social damage but converges to it as competition gets tougher. On the one hand, this result is satisfactory since it does not contradict our knowledge about second-best taxation of a monopolist; after all, pure monopolies rarely exist since every monopolist competes with other ﬁrms in some way. On the other hand, the result is not trivial, since from Cournot competition with free entry we know that the

Environmental policy under imperfect competition

167

second-best optimal emission tax rate may exceed marginal damage with a view to mitigating excess entry. 6.4

Salop’s Model of the Circular City

However, if the goods under consideration are physically identical but only diﬀer with respect to their location, the conclusion in Proposition 8 does not necessarily hold in general. To show this, Lange and Requate (1999) discuss an extended version of Salop’s model of the circular city. For this purpose the authors slightly modify Salop’s original model by relaxing the usual assumption of unit demand, but rather assuming downward sloping demand in order to obtain variable aggregate output and thus variable pollution (otherwise environmental policy such as taxation would not be very interesting). 6.4.1 Outline of the model Each consumer has elastic demand for the consumption good supplied by n ﬁrms located on a circle with perimeter 1. Consumers are also located uniformly on the circle with density 1. The utility of a consumer with distance x to the closest ﬁrm is given by ux U(q)pqtx where q is the quantity of the commodity consumed, p is its price, and t is the consumer’s marginal transportation costs. Let q(p) denote the consumer’s Marshallian demand and V(p) : U(q(p)) pq(p) the consumer’s gross indirect utility function, i.e. the utility disregarding the transportation costs. To obtain a concave revenue function the authors assume that 2(q) 2 qq 0 holds. As in the last section, the ﬁrms are identical, they produce with constant marginal costs c0 and pollute proportional to output. For simplicity we again identify pollution with output. Given that the ﬁrms are located at equal distances around the circle and all potential competitors oﬀer the good at price p, the demand for ﬁrm i’s good is given by

Gi (pi, p) q(pi )

V(pi ) V(p) 1 n t

where the second term is the share of consumers buying at ﬁrm i. If the ﬁrms are subject to an emission tax which in this case can also be charged on output, a ﬁrm’s proﬁt is determined by

168

Yearbook of environmental and resource economics

i (pi , p) [pi c ]Gi (pi , p) ˇ

ˇ

[pi c ]q(pi )

V(pi ) V(p) 1 n t

Proﬁt maximization leads to the following ﬁrst-order condition which is also the Nash equilibrium condition in the second stage of the game, once the ﬁrms have entered the market.

i G i (p , p) [p c ]q(p ) V(pi ) V(p) 1 n i i i t pi [pi c ]q(pi )

V(pi ) t 0

(4.76)

Using Roy’s identity, i.e. V(pi)q(pi), and the symmetry of equilibrium we can write: ppi and qq(p). Further we write q:q(p). Thus, the ﬁrst order Nash equilibrium condition (4.76) becomes q (p c ) [ q q2 nt ] 0

(4.77)

q (p c ) n F

(4.78)

Zero proﬁts yield

6.4.2 The eﬀect of increasing the tax rate Now we investigate the impact of raising the emission tax rate. Diﬀerentiating the system (4.77) and (4.78) with respect to the tax rate , and solving for both pdp/d and ndn/d yields (for details of the algebra see Lange and Requate, 1999): p n

2q

(2q2nt

q 2p2nt (p c )(q 3qq nt )

(p c )(2(q) 2 qq) F [2q 2q2nt (p c )(q 3qq nt )]

(4.79) (4.80)

The denominators are clearly negative by the ﬁrms’ second-order conditions of proﬁt maximization. Thus p 0. Since we have further assumed that the ﬁrms’ revenue is concave, i.e. 2(q)2 qq 0 we obtain n 0 Thus the price rises and the number of ﬁrms fall if the emission tax is raised, as we would have expected.

Environmental policy under imperfect competition

169

6.4.3 The second-best optimal tax rate We now turn to the regulator’s problem. If we again assume that the tax is the only instrument available to the regulator, she or he maximizes welfare deﬁned as follows

W( ): U(q(p)) 2nt

1 2n

sds cp(p) nF D(q(p))

(4.81)

0

where p and n, determined by (4.77) and (4.78), are functions of the tax rate. Note that the second term on the right-hand side of (4.81) takes account of the consumers’ transportation costs. It is easy to calculate that this term is equal to t/(4n). Diﬀerentiating (4.81) with respect to the tax rate and solving for yields

D(E)

q n t F 2n q q t pq 4n2

(4.82)

Since up to now we have not made use of the zero-proﬁt condition in this subsection, we immediately obtain the following result: Proposition 10 If the number of ﬁrms is ﬁxed or if a regulator has direct control over it, for example by a license scheme, the second-best optimal tax rate is given by

D

q 2n q q t

(4.83)

and thus falls short of marginal damage. Let us now get back to the case where the regulator has no direct control over n. Since the second term of (4.82) is clearly negative and both n 0 and p 0 hold, the second-best tax rate falls short of marginal damage if F t/4n2. However, Lange and Requate (1999) show by example that in equilibrium the last inequality may also be reversed. Thus the second-best optimal tax rate may exceed of fall short of marginal damage. Thus, paralleling the Cournot model, there is excessive entry in the extended version of the Salop model. Since the second-best optimal emission tax has to correct for three market imperfections, i.e. pollution, too little output per ﬁrm, and too many ﬁrms entering the market, the latter may be suﬃciently strong such that the emission tax has to be set higher and above marginal damage.

170

7.

Yearbook of environmental and resource economics

PERMIT TRADING IN OLIGOPOLY

In standard oligopoly theory, the ﬁrms’ technologies are represented by their cost functions. In doing so, one implicitly assumes that the ﬁrms behave as price-takers in all factor markets. If an industry is regulated by issuing tradable permits, those permits can be considered an additional input. If the ﬁrms behave competitively on the market for permits, the price for permits has the same impact as an emission tax on the ﬁrms. In the fully competitive model, the regulator could issue a suitable number of permits instead of charging an emission tax. Firms, however, will only behave as price-takers if the number of ﬁrms is large. Since in oligopoly the number of ﬁrms is small by deﬁnition, competitive behavior on the permit market can only be justiﬁed if there are other ﬁrms outside the oligopolistic industry that operate on the same permit market. For pollutants arising in many diﬀerent production processes, such as CO, CO2, NOx and others, this will certainly be a realistic assumption. However, in some industries, for example the chemical industry, pollutants are emitted that are speciﬁc to that industry, i.e. the few competitors on the output market are the only emitters of a certain pollutant. In this section, I discuss this kind of industry structure. In the next subsection I start with a duopoly model, then going on to discuss why it is diﬃcult to extend the approach to more than two ﬁrms. 7.1

Cournot Duopoly

I begin my discussion with Cournot competition, presenting an adapted version of von der Fehr’s (1993) model.25 Requate (1993a) independently develops a similar model, assuming, however, that the ﬁrms’ technologies are linear.26 We proceed on the assumption that the government issues (grandfathers) a number of permits denoted by L. The process in the economy can be divided into two stages. Initially, the ﬁrms hold an endowment (l1, l2) of permits with l1 l2 L. In the ﬁrst stage, they are allowed to trade, i.e. one ﬁrm sells some or all permits to the other ﬁrm. Firms thus end up with a new allocation of permits e(e1, e2), where e1 e2 L. In the second stage, ﬁrms engage in Cournot competition and choose quantities q1(e), q2(e) given the allocation of permits e(e1, e2). Writing Q(e)q1(e)q2(e), a Cournot–Nash equilibrium (q1(e), q2(e)) then satisﬁes the following conditions for i1, 2: P(Q(e))qi (e) P(Q(e)) Ciq (qi (e),ei ) 0 if Cie (qi (e),ei ) 0

[ 0 if qi(e) 0]

Environmental policy under imperfect competition

and

P(Q(e) )qi (e) P(Q(e)) Ciq (qi (e), eˆi ) 0 if ∃ eˆi ei such that

Cie (qi (e), eˆi ) 0

171

[ 0 if qi (e) 0] i 1, 2

Note that we have to allow for corner solutions here because one ﬁrm might buy all the permits from the other ﬁrm. If ﬁrm i produces, i.e., qi(e) 0, and the permit constraint is binding, the ﬁrst-order (Nash equilibrium) conditions reduce to P(Q(e))·qi (e) P(Q(e)) Ciq (qi (e),ei ) 0

for i1, 2

(4.84)

To ﬁgure out how the ﬁrms will trade the permits in the ﬁrst stage, we have to study the gains from trade. For this purpose we use Ni (e1, e2) to denote the proﬁt of ﬁrm i if the ﬁnal allocation of permits in the ﬁrst stage is (e1, e2) and both ﬁrms choose Nash quantities in the second stage. Observe that, starting from any allocation (l1, l2), a gain from trade will be fully conditional on the existence of an allocation (e1, e2) such that N N N N 1 (l1, l2 ) 2 (l1, l2 ) 1 (e1, e2 ) 2 (e1, e2 )

In this case there exists a real number T that is interpreted as a transfer payment from ﬁrm 1 to ﬁrm 2 (which may of course be negative), such that N N 1 (e1, e2 ) T 1 (l1, l2 ) N N 2 (e1, e2 ) T 2 (l1, l2 )

This is the same procedure as in the model for several local monopolists (see subsection 3.5.3). We do not need to bother about how the ﬁrms ﬁgure out T. For example, they could agree on the Nash bargaining solution. The maximum gain from trading permits is then determined by N max [ N 1 (e1, e2 ) 2 (e1, e2 )] e1,e2

s.t. e1 e2 L, e1 0, e2 0

(4.85)

On the assumption that ﬁrms behave as proﬁt-maximizers, it is natural to make the following assumption: Assumption 5 Firms trade permits in the ﬁrst stage such that the ﬁnal allocation (e*1, e*2) solves (4.85). Note that this assumption also allows for the case of one ﬁrm buying all the other ﬁrm’s permits and thus resulting in a monopoly on the market.

172

Yearbook of environmental and resource economics

Observe further that the solution of (4.85) does not depend on the initial allocation (l1, l2), in contrast, of course, to the ﬁnal proﬁts (net transfer payments). But we need not be concerned about this as the distribution of proﬁts among the ﬁrms does not aﬀect welfare. Thus, by virtue of the institutional permit-market framework, cooperation becomes feasible to a limited degree. Since by anti-trust laws it is usually forbidden for ﬁrms to sign binding contracts to maximize joint industry proﬁts, ﬁrms can do no better than choosing Nash quantities. However, if ﬁrms buy or sell pollution permits, they implicitly commit either to a direct capacity constraint – in case that pollution is strictly proportional to output and no additional abatement technologies exist – or to extremely high production costs which amounts to an indirect capacity constraint. Hence, by trading permits, the ﬁrms can achieve joint maximization of Cournot–Nash proﬁts. I shall not work out the maximization problem of (4.85) here. It is important to note, however, that for no solution of (4.85) can it be the case that both ﬁrms hold permits and at the same time the permit constraint is not binding on one of the ﬁrms. Otherwise a ﬁrm i would have idle pollution capacity and would engage at the same time in Cournot competition with the other ﬁrm. In this case, ﬁrm i could increase its proﬁts by buying all the permits from ﬁrm j, thus establishing a monopoly position. On the other hand, monopoly does not necessarily maximize total industry proﬁts, as, if both ﬁrms incur sharply increasing marginal production costs, it might be more proﬁtable for both of them to share both the permits and production with one another, despite Cournot competition in the second stage. Solving the regulator’s problem is quite a complicated matter due to the sequential nature of the ﬁrms’ game. One can show, however, that if the ﬁrms trade the permits so that they both hold permits in equilibrium, the regulator’s optimal permit supply satisﬁes the following conditions:

D(L) P(Q(e(L))) Q(L)Q(L) q1 (e(L) )

q2 (e(L))

q2 (e(L)) ei

q1 (e(L)) ei

Cie (q2 (e(L)),e2 (L)) where is the Lagrange multiplier with respect to the constraint e1 e2 L and qi/ej is the reaction of ﬁrm i with respect to output if ﬁrm j obtains

Environmental policy under imperfect competition

173

more permits. Note that if the term in curled brackets is positive for one ﬁrm, it must be negative for the other ﬁrm, since qi/ej qi/ei. This implies that in general we will have C1e (q1 (L),e1 (L)) C2e (q2 (L),e2 (L)) i.e. the ﬁrms do not level out their marginal abatement costs. In the ﬁrst stage of the game, for strategic reasons the ﬁrms commit to an ineﬃcient distribution of permits and thus to an ineﬃcient cost structure in order to commit to lower output, thus extracting a higher rent in the second stage of the game when they engage in Cournot competition. The Lagrange multiplier can be interpreted as the opportunity cost of shifting a permit from one ﬁrm to the other. Accordingly, it is the oligopolistic industry’s shadow price for pollution. Von der Fehr (1993) emphasizes that allowing for trade may in fact be welfare-decreasing since it leads to monopolization in the permit market. Sartzetakis and McFetridge (1999) oﬀer a graphical analysis illustrating how permit trading in duopoly aﬀects and shifts the ﬁrms’ reaction curves in the output market. 7.2

Welfare Comparison between Permits and Taxes

The fact that, in general, marginal abatement costs are not equalized across ﬁrms if they trade permits strategically does not however necessarily imply that regulating duopoly by permits is, in general, more ineﬃcient than regulation by emission taxes. To see this, consider the simple linear model developed in Requate (1993a), where ﬁrm 1 has lower marginal costs c1 c2 but emits more pollutants per unit of output, i.e. 1 2. Assume that the social damage function is relatively steep so that it is socially optimal for the ‘cleaner’ ﬁrm 2 to serve almost the whole market, whereas ﬁrm 1 should produce only very little. Under a permit regime it may be optimal to issue a small number of permits so that ﬁrm 2 does indeed buy almost all the permits, while under taxes the worse polluting ﬁrm 1 produces too much. For the case of linear technologies assumed in section 4.6.1, Requate (1993a) fully characterizes the regions where taxes and permits lead to higher welfare, depending on a damage parameter that determines the slope of the damage function. The bottom line of that analysis is that, in general, neither policy is superior to the other, i.e. for some parameters the second-best optimal permit policy yields higher welfare than the secondbest optimal tax policy, whereas for other damage parameters the opposite is true.

174

7.3

Yearbook of environmental and resource economics

Price Competition

If ﬁrms engage in price competition we end up with Bertrand–Edgeworth rather than Bertrand competition, since ﬁrms face either increasing marginal costs or they are capacity-constrained if they have linear technologies and emissions are proportional to output. It is well known that under Bertrand–Edgeworth competition, pure strategy equilibria do not exist in general. For the case of linear technologies, however, Requate (1993a) shows that the ﬁrms trade the permits in such a way that the Cournot outcome is a pure strategy equilibrium (see Kreps and Scheinkman, 1983). Again, the welfare comparison with the tax regime is ambiguous. 7.4

Permit Trading and Subsidies on Output

Requate (1993a,c) shows that, in the case of linear technologies, subsidies on output and permits to regulate emissions always lead to the ﬁrst-best allocation, irrespective of whether ﬁrms engage in quantity (Cournot) or in price (Bertrand) competition. The intuition is the same as in the case of taxes. The regulator has two instruments for dealing with two distorted decisions. The result does not hold, of course, if ﬁrms are asymmetric and their technologies satisfy Assumption 4. 7.5

More than Two Firms

In the last few subsections we have considered a model where the ﬁrms trade the permits in the ﬁrst stage and engage in Cournot competition in the second. In the ﬁrst stage, the permits are traded in such a way that the joint Nash proﬁts to be earned in the second stage are maximized. The question is whether this procedure can be generalized in a natural way to apply to more than two ﬁrms. Let there be n ﬁrms, and let (l1, . . ., ln) be an initial allocation of permits n n with i1 li L. Let e(e1, . . ., en) be a feasible allocation, i.e., i1 ei L. Finally, let N i (e) denote the proﬁt for ﬁrm i if each ﬁrm holds ei (many) permits and all the ﬁrms have chosen Cournot–Nash quantities. Let us further assume that the ﬁrms achieve the cooperative outcome in the ﬁrst stage by solving n

max e

Ni(e) i1

s.t. ni1 ei L

and let e* (e*1, . . ., e*n ) be the corresponding maximizer.

Environmental policy under imperfect competition

175

Such a procedure, however, presumes that the ﬁrms can commit to their share e*i of permits and do not engage in further trade with other ﬁrms once that allocation of emissions has been set. The government could set an aggregate pollution quota for certain pollutants, and the ﬁrms may cooperatively agree on the degree to which each ﬁrm is allowed to pollute within a certain period of time. The agreement allows for transfer payments between the ﬁrms. Once the agreement has been signed, no further trade is allowed. Thus the ﬁrms commit themselves to an allocation of permits that remains unaltered for a certain period of time. However, for most actually existing permit trading regimes the institutional framework is diﬀerent. An agreement about the allocation of permits cannot be enforced, and it may be proﬁtable for any two given ﬁrms to engage in a further trade of permits and improve upon an allocation e* that n maximizes i1 N i (e) . In other words, the core of permit allocations may be empty. Weigel (1992) provides numerical examples of a Cournot market with three ﬁrms facing linear demand and quadratic cost functions, where the core is indeed empty. The point is that, by trading permits, two ﬁrms impose a negative monetary (as opposed to a real) externality on the third ﬁrm. The negative monetary externality materializes because, by trading the permits, two ﬁrms can make the cost structure more eﬃcient for each other, thus gaining a greater market share and inducing the market price to fall, which hurts the third ﬁrm. Proposing a solution concept for a ﬁnal allocation of permits if there are more than two ﬁrms is an unresolved problem requiring further research. 7.6

Extensions

Sartzetakis (1997b) also investigates the interaction between permit markets and oligopolistic output markets. In contrast to the model presented above, the author assumes that the permit market is competitive, while on the output market the ﬁrms engage in imperfect competition. Trading permits has oﬀsetting eﬀects by equalizing marginal abatement costs, but it can make ineﬃcient ﬁrms more proﬁtable. Sartzetakis shows, however, that the net welfare eﬀect of permit trading is positive compared to the non-trade situation. In an extension of our basic model with completely inelastic permit supply, von der Fehr (1993) also considers the case where the regulator uses an increasing supply function of permits. The duopolists can now act strategically on both the input and the output market. In a two-stage game the ﬁrms ﬁrst buy the permits, then go on to engage in quantity competition in the second stage. The main result is that if the ﬁrms’ quantities are strategic substitutes, the ﬁrms over-invest in emission permits in comparison to the

176

Yearbook of environmental and resource economics

ﬁrst-best outcome. The reason is that the ﬁrms commit themselves to a lowcost structure by shifting their reaction curves outwards. This is the usual eﬀect when ﬁrms can invest in the ﬁrst period to achieve lower production costs in the second. For price competition the result is ambiguous. But even in that case, over-investment in emission permits may occur.

8.

MARKET POWER IN INPUT MARKETS

In this section I study situations where polluting ﬁrms have market power in some input market. For this purpose I study models where pollution is proportional to one input. I begin with monopsony and then discuss brieﬂy the case of a quantity-demanding (Cournot) oligopsony. The ﬁrms may exercise market power either in the market for the dirty input or in a market for a clean input. 8.1

Monopsony Power over a Polluting Input

I consider a ﬁrm that is now a price-taker in an output market where p is the price of the output good. The output is produced by (at least) two inputs qf(x1, x2), one of which, say input 1, causes pollution proportional to its quantity ex1. The ﬁrm has market power in the factor market of input 1. To model this, I use w1(x1) to denote the supply function in the upstream sector producing input 1. That sector is assumed to be competitive, i.e. the suppliers of input 1 act as price-takers. Let Ci(xi) denote the cost function of the representative ﬁrm in sector i1, 2, where we assume Ci 0 and C"i 0. The market for input 2 is assumed to be competitive, with the factor price denoted by w2. Assuming that input 1 is subject to an environmental tax , this ﬁrm’s proﬁt can be written as (x1, x2)pf(x1, x2)[w1(x1) ]x1 w2x2 The ﬁrst-order conditions for the monopsonist’s proﬁt maximum are given by f p x w1 (x1 ) w1 (x1 )x1

(4.86)

f p x w2

(4.87)

1

2

The upstream input suppliers’ ﬁrst-order conditions correspond quite simply to Ci (xi ) wi. From this we can even derive the supply function

Environmental policy under imperfect competition

177

for good 1 as w1(x1) Ci (x1 ). Obviously, w1(x1) is upward-sloping due to increasing marginal costs of the input suppliers. To ensure that the second-order condition of the monopsonist is satisﬁed, we assume 2w1 (xi ) w1 (x1 ) 0. This is a similar condition to (1) for the monopolist’s inverse demand function. To characterize the social welfare function, we assume a small open economy with respect to the output market. This allows us to neglect consumer surplus. This procedure does not restrict the validity of the results in any way. Thus we have Wpf(x1, x2)C1(x1)C2(x2)D(x1) Taking into account the monopsonist’s choice of inputs, determined by (4.86) and (4.87) and denoted by x1( ) and x2( ), the regulator maximizes W with respect to the tax rate. Solving for the optimal tax rate we obtain

D(x1 )

w1 (x1 )x1

This gives us the following result: Proposition 5 If a polluting ﬁrm has monopsony power on a market for a polluting input, the ﬁrst-best allocation can be achieved by levying an input tax. The optimal tax is lower than marginal damage. Note that for the size of the tax rate the result does not change if the regulator taxes the output of the input supplier. Thus, although the input supplier is a price-taker, the distortion of the monopsonist requires setting the tax rate below marginal damage. This result is important, as it suggests that it is not suﬃcient to look at the market structure of the polluting ﬁrm alone. 8.2 Second-Best Analysis: When Monopsony Power is Exercised over a Clean Input Things are slightly diﬀerent if the ﬁrm exercises monopsony power over a non-polluting input but uses another dirty input supplied on a competitive market. For this purpose we can simply rewrite the above model by assuming that pollution is equal to the second input: ex2. Dealing with the welfare-maximizing exercise as above, one can determine the second-best tax rate as

D(x2 )

w1 (x1 )x1dx1 d

dx2 d

178

Yearbook of environmental and resource economics

Diﬀerentiating (4.86) and (4.87) with respect to the tax rate, we obtain dx2/d

0, i.e. the output of the polluting input goes down, and thus pollution goes down as the emission tax rises. The eﬀect on the clean input is, however, ambiguous. We obtain dx1/d

0 if and only if f12 2f(x1, x2) / (x1x2) 0, which is the case for most production functions used in applied work (in particular in CGE models), such as CES or nested Leontief/CES functions. This gives rise to the following result for the second-best tax rate: Proposition 6 Consider a monopsonist exercising monopsony power in a market for a clean input and using a polluting input that it buys in a competitive factor market. If the regulator can only target the emissions of the polluting input, the second-best optimal tax rate falls short of (exceeds) marginal damage if and only if f12 0 (f12 0). This result can be interpreted as follows: If f12 0, the level eﬀect due to a rise in the emission tax dominates the substitution eﬀect, such that the ﬁrm reduces dirty and clean inputs as the tax rate rises. Hence the welfare-maximizing regulator does not want to set the tax too high because the monopsonist produces too little anyway. If f12 0 holds, the substitution eﬀect dominates the level eﬀect. In that case, the welfaremaximizing regulator will want to set a tax rate that exceeds marginal damage in order to encourage the monopsonist to substitute the input clean for the dirty one. 8.3

Oligopsonies

The results do not change dramatically when we move from monopsony to oligopsony. Hence, we will only outline the results here. The model can be generalized à la Cournot by assuming that the factor price on an input market depends on the total input demand of several ﬁrms: w1 w1 (x11 ... xn1 ), where xj1 is the input demand of ﬁrm j. We can summarize the results as follows: Proposition 7 Assume n ﬁrms have market power in a factor market. Assume the factor is homogeneous so that there is only one price. (i) If ﬁrms are symmetric and exercise market power in a market for a polluting input and if no further abatement technologies exist, the ﬁrstbest allocation can be achieved by implementing either an input or an emission tax. The optimal tax rate is lower than marginal damage. Raising the tax rate lowers pollution.

Environmental policy under imperfect competition

179

(ii) If either the situation is the same as in (i) except that ﬁrms are symmetric, or if ﬁrms exercise market power in a market for a clean input, the ﬁrst-best allocation can neither be achieved by a uniform input nor by a uniform emission tax. If ﬁrms are not too diﬀerent, the secondbest optimal tax rate is lower than marginal damage, and raising the tax rates lowers pollution. (iii) If ﬁrms are suﬃciently diﬀerent with respect to their cost structure, perverse eﬀects may arise, i.e. pollution may be increased be raising the tax rate, and the second-best tax rate may exceed marginal damage. 8.4

Mixed Structures

We have seen that in situations with market power in an input market, the second-best optimal tax rate usually falls short of marginal damage. Hence it is clear that if a ﬁrm exercises market power in both the output and the input market, the eﬀects will ‘add up’.27 The more market power there is, the lower the second-best tax rate will be. This also holds if we have either a vertical monopoly or a vertical monopsony chain. Since in this case the distortions work into the same direction and the total distortion increases, the second-best optimal emission tax rate designed to regulate a vertical monopoly or monopsony chain is lower than in the case where there is market power in one market only. As set out above, the Pigouvian tax rule also fails to hold for a competitive ﬁrm if this ﬁrm sells to a monopsonistic downstream ﬁrm or if it buys some input from a monopolistic upstream ﬁrm. Thus the regulator needs to be conversant with the complete vertical industry structure when determining the second-best optimal level of his tax rate. According to my knowledge, the literature has been silent on market structures as outlined in this section. 8.5

Market Power in Markets for Clean Inputs and Clean Technology

David and Sinclair-Desgagné (2005) and Requate (2005) study market power in an upstream market for abating inputs or for cleaner technology, respectively. David and Sinclair-Desgagné consider a competitive polluting industry where emissions e(x, w) are a function of output x and abatement input w. The latter is supplied by an upstream industry, the ﬁrms of which engage in Cournot competition. If the polluting downstream ﬁrms are taxed and the tax is the only instrument at the regulator’s disposal, the second-best optimal tax rate will exceed marginal damage. David and Sinclair-Desgagné (2005) further show that a voluntary approach to pollution abatement may be doomed to failure unless some limitations are imposed on the ecoindustry’s market behavior.

180

Yearbook of environmental and resource economics

Requate (2005) discusses a model with a monopolistic R&D ﬁrm developing new technology for a polluting downstream sector. He also ﬁnds that the second-best optimal tax rate exceeds marginal social damage if the regulator can make an ex ante commitment to the level of his tax rate. In both models, the intuition for the high tax rate is that the upstream sector sets its prices too high, which leads to an ineﬃciently low purchase of the abating input in the model by David and Sinclair-Desgagné and to an ineﬃciently low purchase of the advanced abatement technology in Requate’s model. By raising the tax rate the regulator enhances demand for the clean input or the new technology, respectively.

9.

MARKET POWER IN THE PERMIT MARKET

So far we have mainly studied imperfect competition in the output market. Permit markets have either been assumed to be competitive, or noncompetitive permit markets have been modelled by bargaining between two or several ﬁrms (see section 7). If the permit market is small, i.e. if there are only a small number of traders, there is no competitive demand side. Hence it is not possible to model a non-competitive permit market with a small number of ﬁrms, as in a Cournot model and to apply the standard models on market power in factor markets. However, there are markets with many small ﬁrms and a few big ﬁrms, e.g. in the US market for SO2 permits (see Howe, 1994). In his inﬂuential paper, Hahn (1984) sets up a stylized model with many small ﬁrms and a single ﬁrm exercising market power in the market for tradable permits. In this section I present a generalized version of the Hahn model allowing for the presence of several large ﬁrms. 9.1

A Model of Permit Trading with Large and Small Firms

To model oligopoly power on the permit market, I divide the set of ﬁrms participating in the permit market into a set of large ﬁrms i 1, . . ., n and a set of small ﬁrms in 1, . . ., m. This divide is exogenous, which is certainly a weakness of this approach. But to date the literature has not provided a viable alternative. Moreover, I intend to neglect the output market in this section. Hence we can represent the ﬁrms, whether small or large, by their abatement cost functions Ci(ei). Each ﬁrm owns an initial endowment of permits, denoted by êi . Accordingly, we are studying a system of grandfathered permits. A typical ﬁrm’s total costs can then be written as Ci(ei) [ei êi] where is the market price for permits.

Environmental policy under imperfect competition

181

The small ﬁrms act as price-takers and thus set their marginal abatement costs equal to the market price of permits: Ci (ei ) Accordingly, emissions ei of the small ﬁrms can be written as ei() and can be interpreted as the factor demand for permits. Summing up these demands for all the small ﬁrms we obtain the demand in the competitive sector: Ec ()

nm

ei()

in1

If we invert this curve, we obtain the competitive sector’s inverse demand function for permits: ()(Ec)1() The number of permits employed by the competitive sector is the amount left by the large ﬁrms. Hence, n

ei

Ec L

i1

and the market price for permits will be

L

ei n

i1

Now we can write the total costs of the large ﬁrms as

Ci (ei ) L

ei· [ei eˆ i] n

i1

The permits market with large and small ﬁrms is now modelled in a Cournot-like way. The ﬁrst-order condition for the cost minimization of a typical large ﬁrm can now be written as

Ci (ei ) L

ei· [ei eˆ i] n

i1

182

Yearbook of environmental and resource economics

Thus, if in equilibrium ei êi, then Ci (ei ) , and if ei êi then Ci (ei ) . This observation leads us immediately to the following result: Proposition 8 Consider a permit market with grandfathering. (i) If in equilibrium a large ﬁrm is a net buyer (seller), the large ﬁrm sets its marginal abatement cost higher (lower) than the permit price. (ii) Only if the large ﬁrms obtain an initial endowment corresponding to the eﬃcient ﬁnal allocation will permit trading lead to an eﬃcient outcome. The intuition for (i) is the following: a large ﬁrm that wants to buy additional permits buys fewer than optimal in order to keep the permit price low. Thus the net buyer behaves as an oligopsonist (or as a monopsonist if n 1). A large ﬁrm that wants to sell spare permits sells less than the eﬃcient amount in order to keep the permit price high. Thus the net seller behaves as an oligopolist (or a monopolist for n 1). Trade will only lead to an eﬃcient outcome if the large ﬁrms’ initial endowments correspond to their eﬃcient emission levels. In this case, the large ﬁrms do not participate in trade and cannot distort the market price for permits. This does not generally imply, however, that large ﬁrms should not participate in trade. Since trade always goes in the right direction, it will improve eﬃciency, i.e. the ﬁnal allocation is less ineﬃcient than the initial allocation. Where legally and informationally feasible, the regulator should, however, allocate approximately as many permits to the large ﬁrms as they will ultimately need, even after trade. Maeda (2003) obtains the same results for the case of n 2 and illustrates the model by simulating the international trade in carbon dioxide allowances, investigating which country is likely to have market power. 9.2

Extensions of the Hahn Model

There are several extensions to this kind of model. They take into account output markets (Malueg, 1990; Innes et al., 1991; and Misiolek and Elder, 1989) and intertemporal permit trading (Hagem and Westskog, 1998 and Sartzetakis, 1997a, 1997b), or they study the possibility of non-compliance (van Egteren and Weber, 1996; Malik, 2002; and Chavez and Stranlund, 2003). 9.2.1 Including the Output Market Innes et al. (1991) extend the Hahn model by assuming that a large ﬁrm has market power in both the output and the permit market. They argue that,

Environmental policy under imperfect competition

183

in contrast to Hahn’s result, the big ﬁrm should participate in trade, and they show that a regime of tradable emission permits is welfare-superior compared to a regime where both the monopolist and the competitive ﬁrms are regulated by a uniform tax. Moreover, they show that there exists an initial allocation of permits that leads to the same ﬁnal allocation of permits as a discriminating tax system. Under a discriminating tax, the monopolist needs to be regulated by a lower tax rate than the competitive ﬁrms in order to mitigate his market power. Allowing the output monopolist to exercise monopoly power in the permit market as well means that he virtually faces a lower price for permits than the competitive ﬁrms, which mimics the discriminating tax system. This leaves us with the question of what is more diﬃcult: ﬁnding exactly the right initial allocation of permits or ﬁnding exactly the right discriminating tax rates. Both may be equally diﬃcult. However, it may be easier to adjust the tax rates than to reallocate the permits after ﬁrms have engaged in trade. Misiolek and Elder (1989) extend the Hahn model in a diﬀerent way by assuming that the (only) large ﬁrm acts as a price-taker in the output market. Nevertheless, the large ﬁrm takes into consideration how buying permits from the (small) rival ﬁrms raises those rivals’ costs, aﬀects the output price, and increases the large ﬁrm’s market share. Sartzetakis (1997a) takes a similar approach to Misiolek and Elder, assuming, however, that (two) ﬁrms engage in imperfect (Cournot) competition in the output market. However, there is still one ﬁrm assumed to have market power in the permit market, whereas the second ﬁrm is a price-taker in the permit market. In the output market, by contrast, the ﬁrms are symmetric with respect to their behavior, as both of them play the simultaneous quantity-setting game à la Cournot–Nash. Sartzetakis then discusses a two-stage game where the large ﬁrm sets the price for permits in the ﬁrst stage. In the second stage, ﬁrms simultaneously decide on both abatement – including how many permits to buy or sell – and on the quantities of output. Sartzetakis shows that market power in the permit market can reduce competition in the output market, a feature also observed in cases where ﬁrms have other strategic options for raising their rivals’ costs (see Salop and Sheﬀman, 1983, 1987). The regulator can mitigate this anticompetitive tendency by issuing fewer permits to the powerful ﬁrm and more to the weak ﬁrm. The question remains why the two ﬁrms display diﬀerent behavior in the permit market but symmetric behavior in the output market. In a companion paper Sartzetakis (1997b) shows that allowing Cournot oligopolists to engage in trade is welfare-improving compared to commandand-control, where each ﬁrm faces the same absolute emission standard. Sartzetakis assumes interior solutions such that marginal abatement costs

184

Yearbook of environmental and resource economics

equalize. Malueg (1990), by contrast, allows for corner solutions and ﬁnds that permit trading in oligopoly may be welfare-reducing. 9.2.2 Intertemporal permit trading Hagem and Westskog (1998) draw on Hahn’s (1984) model by including the intertemporal aspect. In their model there is only one large ﬁrm and it is always a seller of permits. The authors study a two-period model and compare two diﬀerent permit schemes. In the ﬁrst scheme, all ﬁrms can bank and borrow permits in an unlimited way. In the second, one permit allows its holder to emit a constant stream of emissions in each period, a system the authors refer to as a durable quota system. Under the ﬁrst system, all ﬁrms eﬃciently allocate emissions over time. The monopolist, however, as in the model described above, sells too few permits, so that his marginal abatement costs are smaller than those of the competitive ﬁrms. Depending on the initial allocation of permits, one or the other system may lead to lower total abatement costs in industry. 9.2.3 Market power and non-compliance Van Egteren and Weber (1996) extend the Hahn model by considering the possibility of non-compliance, i.e. ﬁrms emitting pollution in excess of the number of permits they hold. Firms are audited with a certain probability and are ﬁned if any cheating is discovered. For the competitive ﬁrms, the incentive to cheat is higher, the higher the market price is for permits. Hence the authors ﬁnd that if the (only) large ﬁrm is compliant, a redistribution of the initial allocation of permits from the competitive ﬁrms to the big ﬁrm increases the total of violations. If the large ﬁrm is non-compliant, then clearly the ﬁrm is less likely to cheat if it receives a higher initial allocation of permits. Malik (2002) extends the model proposed by van Egteren and Weber (1996) by investigating the eﬃciency consequences of non-compliance combined with the market power of one ﬁrm. He shows that in the presence of market power non-compliance by the small price-taking ﬁrms is potentially desirable. Conversely, in the presence of non-compliance, some market power by the (only) big ﬁrm is desirable. The reason is that the monopolistic ﬁrm retires some permits and thus reduces some of the excess pollution emitted in the case of non-compliance. In my view, the recommendation to maintain monopoly power in order to mitigate non-compliance by competitive ﬁrms seems rather strange, as the regulator could in principle mimic the monopolist’s behavior by issuing a smaller number of permits. Chavez and Stranlund (2003) complement the work of van Egteren and Weber by endogenizing the enforcement of compliance. Whereas Hahn (1984) ﬁnds that the (only) large ﬁrm should obtain an allocation of

Environmental policy under imperfect competition

185

permits such that it does not want to trade, Chavez and Stranlund suggest that the large ﬁrm should be a buyer (seller) of permits when monitoring costs are increasing (decreasing) in the ﬁrm’s initial endowment of permits. 9.3

Market Power through Innovation

Fischer et al. (2003) consider a model where a single ﬁrm, called the innovator, is able to invent a new technology that leads to lower (marginal) abatement costs. Since the innovator needs fewer permits after innovation, the equilibrium price for permits decreases. This gives the innovator a degree of market power. Fischer et al. (2003) do not study the consequences of the allocation of permits. They ﬁnd, however, that free permits provide fewer incentives for innovation than auctioned permits. The reason is the endowment eﬀect. The value of the innovator’s permit endowment depreciates through the invention of new technology. Montero (2002a) also studies investment incentives in diﬀerent kinds of policy instruments, notably tradable permits and two kinds of standards, emission and performance standards. Besides allowing for perfect and imperfect competition in the output market, he – like Innes et al. (1991) – models imperfect competition in both the output and the permit market. In the latter case, ﬁrms negotiate on the permit price, as proposed in section 7. The ﬁrms employ the Nash bargaining solution, taking the level of R&D as given and anticipating the expected output decisions in the last stage of the game. Since ﬁrms have market power in both markets, trading permits has a strategic eﬀect with respect to the output market. This causes the ﬁrms to invest more under an emission standard and a regime of auctioned permits than in a regime of free permits. Montero (2002b) studies both Cournot and price competition with diﬀerentiated products, allowing for R&D spill-overs and again assuming imperfect competition in both the output and the permit market. Besides emission standards and permits he also analyzes emission taxes. Montero ﬁnds that no strategic eﬀect results from levying an emission tax, as the marginal costs of both ﬁrms are constant. Taxes can provide more, less, or the same incentive to invest in innovation than/as both emission standards and auctioned permits, whereas free permits provide fewer incentives to innovate than taxes. With Bertrand competition in the output market, taxes provide a higher incentive than an emission standard, which in turn provides a higher incentive than free permits. Auctioned permits again can oﬀer more, less or the same incentive than/as taxes. Montero concludes that in the Cournot case, either emission standards, taxes, or auctioned permits can provide the highest incentive, whereas in the Bertrand case this holds either for taxes or for auctioned permits.

186

9.4

Yearbook of environmental and resource economics

Results from Experimental Studies

Given the lack of empirical data for analyzing the eﬃciency of permit markets, laboratory experiments on permit trading have become an attractive substitute for empirical ﬁeld investigations on markets for emission allowances. In particular the Hahn model on monopoly power in permit markets has attracted considerable attention from experimental economists. To my knowledge, Brown-Kruse and Elliott (1990) and Brown-Kruse et al. (1995) were the ﬁrst to test for market power in emission-trading experiments. In their experimental treatments a single seller or a single buyer is confronted with ten buyers and ten sellers, respectively. Godby (2000) extends this approach including the product market. By setting the product market price either exogenously or endogenously, he is able to mimic the models developed by Hahn (1984), Misiolek and Elder (1989), and Innes et al. (1991). The experimental design of Godby’s (2000) ﬁrst series of treatments follows that of Brown-Kruse and Elliott (1990) and Brown-Kruse et al. (1995). Whereas the large ﬁrm has a production capacity of ten units, each of the ten fringe ﬁrms can produce only one unit. Pollution is proportional to output. Thus under a laissez-faire policy, the market would produce and pollute 20 units. The number of emission permits allocated to industry, however, is ten. Godby carries out four diﬀerent treatments. In the ﬁrst two treatments the product market price is given exogenously, and all the permits are either allocated to the large ﬁrm (treatment 1) or equally distributed to the fringe ﬁrms (treatment 2). Thus the big ﬁrm is a net seller in the ﬁrst case and a net buyer in the second. Since the product market price is exogenous, Godby calls the big ﬁrm’s manipulation of the permit market ‘simple manipulation’. In the other two treatments the output price is determined endogenously. Following Misiolek and Elder (1989), Godby calls this kind of market power ‘exclusionary manipulation’, because if the large ﬁrm is a net seller, it has an incentive to sell fewer permits than is socially optimal for two reasons: it not only wants to keep the price for permits high and thus earn high revenues from selling permits; it also wants to increase the product market price by increasing the production costs of the fringe ﬁrms and thus reducing supply in the product market. If the big ﬁrm is a net buyer of permits, there are two oﬀsetting eﬀects. On the one hand, the large ﬁrm has an incentive to buy fewer permits than is socially eﬃcient in order to hold down the market price for permits. On the other hand, it wants to raise the competitors’ costs by buying more permits than would be eﬃcient. For the case of exclusionary manipulation, Godby again conducts two experiments. In one case, all the permits are allocated to the large ﬁrm (treatment 3) while in the second the fringe ﬁrms initially receive all the permits (treatment 4).

Environmental policy under imperfect competition

187

Godby extends this design by reducing the number of fringe ﬁrms from ten to ﬁve and enhancing their capacity from one to two units. Thus he also gives limited market power to the small ﬁrms. The reason for this modiﬁcation is the empirical observation that the market for NOX permits in Ontario was governed by an electricity utility that demanded about 50 percent of the permits, whereas ﬁve other ﬁrms, such as producers of iron and steel, cement, etc., demanded about 10 percent each. Godby (2000) replicates several of the results produced by Brown-Kruse and Elliott (1990) and Brown-Kruse et al. (1995). In particular, he ﬁnds that the prediction of the Hahn model is well conﬁrmed by the experimental outcomes in treatments 1 and 2, where the price on the output market is taken as exogenously given. When the large ﬁrm is a seller, the eﬃciency gain through trade over the initial allocation is approximately 60 – 70 percent of the gains that are theoretically achievable through perfect competition. If the large ﬁrm is a buyer, the eﬃciency loss through market power turns out to be much smaller (approximately 80 – 95 percent of the maximal possible gains). In both treatments the eﬃciency loss is greater when there are only ﬁve fringe ﬁrms with high capacity. Under exclusionary manipulation, by contrast, permit trading results in negative eﬃciency. Godby establishes an eﬃciency of approximately 40 percent in treatment 3, and of 120 to 140 percent in treatment 4 (in the latter case, the eﬃciency loss is even larger for the case of ten small fringe ﬁrms). In other words, permit trading turns out to be inferior to command-and-control. Godby (2002) replicates the analysis of Godby (2000) with one large ﬁrm holding a capacity of ten units and ﬁve fringe ﬁrms with a capacity of two units each. Compared to his earlier paper he adds two more treatments where the permits are allocated proportionally to the ﬁrms’ capacities, i.e. the dominant ﬁrm gets ﬁve, whereas the fringe ﬁrms receive one permit each. Using statistical techniques, Godby (2002) arrives at basically the same results as in Godby (2000): in all treatments the hypothesis that the dominant ﬁrm exercises market power cannot be rejected when looking at the end points of the permit double-auction price series. Product market prices also indicate convergence toward the market power benchmarks. Observed quantities deviate from competitive levels in the theoretically predicted direction. Moreover, when exclusionary manipulation is possible, permit trade leads to eﬃciency losses relative to the command-and-control benchmark. Finally, Godby ﬁnds evidence that players engage in speculative behavior. Muller et al. (2002) carry out similar experiments. They generate market power on the seller or buyer side by aggregating ﬁve sellers and ﬁve buyers, respectively. They also observe market power in double auctions. Their

188

Yearbook of environmental and resource economics

main conclusion is that, contrary to the proposals by Smith (1981), the double-auction design is not as robust with regard to market power. In contrast to Godby (2000, 2002), who avoids any framing, Carlén (2003) conducts a framed experiment, mimicking the international carbon trade where the big trader – a buyer – is interpreted as the USA. Moreover, in contrast to most other laboratory studies, the participants have no chance of gaining experience. The authors argue that this setting comes closer to international permit trading in the ﬁeld. In contrast to the other experiments referred to here, Carlén does not ﬁnd evidence for distortions through potential market power. Bohm and Carlén (1999) additionally introduce an information structure to mimic more realistic ﬁeld conditions for the carbon dioxide emission-permit trading program. However, they do not ﬁnd that changes in the information structure signiﬁcantly aﬀect market eﬃciency. Finally, Cason et al. (2003) also conduct a framed ﬁeld experiment in order to mimic permit trading of nitrogen allowances among sewage treatment plants. The novelty of their study is the introduction of asymmetric information by letting one or two large emitters know the abatement costs of the small emitters, whereas the small emitters do not know the costs of the big ones. The authors also test for the impact of diﬀerent initial allocations. They ﬁnd that in a monopoly situation with one large seller the market price for permits is larger than in the duopoly treatment, although the diﬀerence is not signiﬁcant. However, in sharp contrast to the ﬁndings of Brown-Kruse et al. (1995) and Godby (2000, 2002), they ﬁnd that the prices, proﬁts and transaction volumes are much closer to the competitive equilibrium prediction than to the monopoly or duopoly prediction. In conclusion, it is worth pointing out that the subject pools participating in these experiments were mainly students and not real decision-makers.

10. ENVIRONMENTAL POLICY, IMPERFECT COMPETITION AND INTERNATIONAL TRADE A survey on environmental policy under imperfect competition would certainly not be complete without addressing the issue of imperfect competition on international markets. Environmental policy in open economies has become a topic of major interest in environmental economics since Markusen’s (1975) seminal paper. Assuming that all markets are competitive, Markusen makes the point that, in the absence of tariﬀs, for example due to a free trade agreement, emission taxes can be used to inﬂuence the terms of trade and can thus serve as a substitute for trade policy. As a consequence an exporting country would like to over-internalize environmental

Environmental policy under imperfect competition

189

damage in order to improve its terms of trade. But this implies that with the competitive trade model it is not possible to explain what both environmentalists and economists refer to as ‘environmental dumping’ (cf. Rauscher, 1994). Brander and Spencer (1985) set oﬀ a new direction of research on trade theory by showing that, under imperfect quantity competition, the optimal policy consists in making exports cheaper (through subsidies) rather than improving the terms of trade by making them more expensive (through export taxes or import tariﬀs). Thus trade policy does not aim at improving the terms of trade by making export goods more expensive, but rather tries to increase market shares at the expense of worsening the terms of trade. This is the celebrated rent-shifting eﬀect. Conrad (1993), Barrett (1994) and Kennedy (1994) were the ﬁrst to discover that, in the presence of imperfect (Cournot) competition, emission taxes can be used to indirectly subsidize exports by under-internalizing even the domestic environmental damage. Whereas Conrad looks at a model where two governments support their domestic industries to increase their market share in a third country’s market, Kennedy (1994) studies a similar model of a closed two-country economy. Barrett (1994), studying both quantity and price competition, ﬁnds that under imperfect price competition, the optimal unilateral policy should over-internalize marginal damage.28,29 In a series of follow-up papers, several authors (among others Ulph, 1994a, 1994b, 1996a, 1996b; Nannerup, 1998; Markusen et al., 1993, 1995; Simpson and Bradford, 1996; Hamilton and Requate, 2004; and others) extend the Conrad–Kennedy–Barrett type of model by including imperfect information, pre-investment in cost reduction, R&D, choice of location, and other decisions to be made before ﬁrms engage in imperfect quantity or price competition.30 Since environmental policy in the presence of international trade has been surveyed extensively elsewhere (see Althammer and Buchholz, 1995, 1999; Ulph, 1997a, 1997b; Rauscher, 1997; and Duval and Hamilton, 2002), it is not my concern here to fully summarize the results of this literature. That literature, however, contains rival and sometimes even contradictory interpretations of the decomposition of the unilaterally optimal tax rate. Accordingly, I wish to highlight the structure of the unilateral second-best optimal tax rate that has to target both imperfect competition at home, which harms domestic consumers, and the market power of domestic ﬁrms on the international market, which favors domestic welfare. For this purpose, I shall concentrate on Cournot competition and regulation by taxes.31 It has become fashionable to talk about ecological (or environmental) dumping whenever a country sets its emission tax below marginal damage. Partially following Rauscher (1994) and Duval and Hamilton (2002), I argue that this view is not entirely appropriate.32

190

Yearbook of environmental and resource economics

10.1

Extension of the Basic Model to International Trade

As is usual, I assume that there are only two countries (governments), one domestic and one foreign. Furthermore, I assume that there is a free trade agreement such that the governments are not allowed to impose tariﬀs or to subsidize their ﬁrms directly. I extend my basic model by denoting the world inverse demand function by Pw(Qw), where Qw is world total output. Furthermore, I write the domestic inverse demand function as Pd(Yd), where Yd is the domestic level of consumption. Moreover, there are nd domestic and nf foreign ﬁrms. Ci(qi, ei) represents the cost function of a typical ﬁrm from country i. Within one country, all ﬁrms have identical technologies (inducing identical cost functions). Moreover, I denote the domestic damage by Dd(Ed), where Ed sddnded sdfnfef is eﬀective (or ambient) pollution arriving in the domestic country caused by emissions from both domestic and foreign ﬁrms. Domestic emissions are multiplied by the emission coeﬃcient sdd that indicates how much domestic pollution aﬀects the home country, whereas sdf indicates how much of the foreign ﬁrms’ emissions arrive in the domestic country. Analogously, I denote foreign damage by Df(Ef), where eﬀective foreign pollution is given by Ef sffnfef sfdnded. 10.1.1 Cooperative environmental policy In this subsection I brieﬂy study the benchmark case where governments agree on their environmental policies in a cooperative way. In order to avoid problems of imperfect competition in the permit market, I assume that the governments use taxes as policy instruments. To achieve a fully cooperative solution it is sometimes necessary for one country to compensate another country for potential welfare losses caused by cooperative environmental policy.33 Hence I assume that transfer payments between governments are possible to achieve the cooperative outcome. This assumption allows us to ignore participation constraints, thus simplifying matters considerably.34,35 In this case, the objective of the two governments is represented by

max td , tf

Qw

Pw (Q)dQ ndCd (qd , ed ) nf C f (qf , ef ) Dd (Ed ) Df (Ef )

0

Omitting the algebra, the cooperative tax rates are given by: td Dd (Ed )sdd Df (Ef )sfd Pw (Qw )qd

qd ed

qf tf Df (Ef )sff Dd (Ed )sdf Pw (Qw )qf e f

(4.88)

(4.89)

Environmental policy under imperfect competition

191

This implies that, when regulating its own domestic industry, each country takes into account the marginal damage inﬂicted by its own industry on both the domestic and the foreign country. Each country also takes into account the oligopolistic distortion caused by its own industry on the world market. These formulas are essentially equivalent to formula (4.39) in section 4 and can also be found in Duval and Hamilton (2002). 10.1.2 Non-cooperative environmental policy Let us now turn to the more interesting case of non-cooperative policy setting. In this case we have to add the market value of exports and imports to domestic welfare. The institutional set-up in this scenario is that domestic and foreign ﬁrms compete imperfectly à la Cournot in an international market that may also consist of third-country markets. The governments are not allowed to subsidize their ﬁrms directly. Hence they will conceivably attempt to use environmental policy as trade policy. The objective function of the domestic government is now given by W

Yd

Pd (Y)dY ndCd (qd, ed ) Pw (Qw )[nd qd Yd ] Dd (Ed ) (4.90)

0

taking into account that domestic ﬁrms compete imperfectly in the international market. The ﬁrms’ ﬁrst-order conditions are familiar from section 4 and need not be repeated here. Diﬀerentiating (4.90) with respect to the tax rate, setting the derivative equal to zero, and solving for the tax rate yields the following formula for the unilaterally optimal (non-cooperative) domestic tax rate: d d

nc d sdD(Ed ) sf D(Ed ) qd

qf

nd nf

Pw (Qw )[ndqd Yd] d nde

nf

ef

nd e d qd

Pw (Qw )qd e d

(4.91)

This decomposition, suggested by Duval and Hamilton (2002), consists of four parts: the ﬁrst term represents the domestic marginal damage caused by the domestic ﬁrms. The second term represents the leakage eﬀect, i.e. the domestic marginal damage caused by the foreign ﬁrms multiplied by the total indirect reaction of foreign ﬁrms’ emissions to a domestic tax rise and divided by the total indirect reaction of domestic ﬁrms with respect to emissions. In the normal case, i.e. if domestic and foreign ﬁrms are not too

192

Yearbook of environmental and resource economics

diﬀerent, we obtain both ed / 0 and qd / 0. The latter eﬀect causes the world market price to rise (the terms-of-trade eﬀect), and therefore induces the foreign ﬁrms to increase output and thus also to increase emissions, i.e. ef / 0. Hence the second term of (4.91) is negative. Thus, even without the strategic aspects represented by the third and the fourth term, the domestic government should under-internalize the world’s social damage for two diﬀerent reasons: ﬁrst, it does not take into account the damage to the foreign countries, i.e. it neglects the term Df (Ef )sdf in (4.88). Second, in the absence of international coordination, the government should even under-internalize the domestic marginal damage, since by raising the domestic emission tax, production and thus pollution shifts from home to abroad and then comes back across the border through wind or water. Thus the second term represents protection against transboundary pollution. Note that this leakage eﬀect is always present, even if markets are competitive. Even though, the reaction of the ﬁrms, and thus the weight factor [nf (ef)/( )]/[nd (ed)/( )], depends on the market structure. Note that taking the leakage eﬀect into account is sometimes also called the ‘not-in-my-backyard incentive’ (see Markusen et al., 1995).

Strategic aspects: Taking into account imperfect competition and ‘terms of trade versus rent-shifting’ Let me now turn to the third and fourth terms. These can be interpreted in diﬀerent ways: Duval and Hamilton (2002) interpret the third term as the terms-of-trade eﬀect and the fourth term as the imperfect competition eﬀect. The latter is clearly negative for the same formal reasons as in section 4. However, the interpretation is diﬀerent, as I will argue below. Let us ﬁrst study the terms-of-trade eﬀect. The numerator of

qf q e nd d nf nd d

is negative since qd / is negative. The latter causes world output to fall and thus the market price for the polluting good to rise. Although the foreign ﬁrms react to this by increasing output,36 one can show that the total eﬀect is negative under normal conditions (i.e. if the inverse demand function is not too convex and the ﬁrms’ cost functions are not too diﬀerent).37 The denominator is clearly negative, as argued above. Since Pw (Qw ) 0, the whole third term is positive if and only if the domestic country is an exporter of the polluting good. Finally let us study the fourth term. According to Duval and Hamilton (2002) this term corresponds to the usual imperfect competition eﬀect that

Environmental policy under imperfect competition

193

is always present, even in the absence of trade. Duval and Hamilton further claim that the regulator has to take account of too little production and hence this eﬀect has nothing to do with eco-dumping. The question is what is meant by ‘too little production’. In a closed economy, the regulator takes into account too little production in order to mitigate the dead-weight loss resulting from market power. But even if the traded good is a pure export good, i.e. if Yd 0, as for example is the case in Conrad’s model and also in one version of Brander and Spencer’s original model, the fourth term will not simply vanish. Accordingly, in my view, the interpretation of the fourth term must be diﬀerent: the regulator wants to help the export industry to improve the terms of trade. But due to market power, the ﬁrms can partially achieve this on their own. This help-yourself eﬀect is represented by the fourth term, which then has to be subtracted from the terms of trade eﬀect. (By inspection it is easy to verify that the fourth term is negative.) Let us now present an alternative decomposition of the second-best optimal tax rate. For this purpose we rewrite (4.91) as follows: n

ef

f d d

nc d sdD(Ed ) sf D(Ed ) n ed d

Pw (Qw )qd

qf q (nd 1) d nf

Pw (Qw )Yd

nd e d

q qd n f f

nd e d

nd

(4.92)

A similar decomposition has been suggested by Althammer and Buchholz (1999). Recall that qd denotes total domestic output, while Yd denotes domestic consumption. Now we can interpret the third term as the strategic market eﬀect, which can be positive or negative since qd / is negative and qf / is positive, but the numerator of the multiplier

qf q e (nd 1) d nf nd d

is less than the total change of output (which would be negative). Thus the numerator can be positive or negative. This means that the strategic market eﬀect can be either a terms-of-trade eﬀect or (!) a rent-shifting eﬀect.

194

Yearbook of environmental and resource economics

If the number of domestic ﬁrms is relatively large, i.e. nd 1 nd, then qf q Q 0 (nd 1) d nf

If in addition qd Yd 0, i.e. the domestic country is an exporter of the polluting good, the third and the fourth term together correspond to the terms-of-trade eﬀect, and it will be optimal for the domestic government to make the output good more expensive to improve the terms of trade. If, by contrast, the number of domestic ﬁrms nd is small, especially if nd 1, the third term of (4.92) boils down to

qf e Pw (Qw )qd nf d

which is exactly the familiar rent-shifting eﬀect. This term is clearly negative. If in addition, the good is a pure export good, i.e. Yd 0, then the third and the fourth term boil down to the implicit Brander–Spencer–Conrad subsidy represented by

qf e Pw (Qw )qd nf d

If at all, it is this term which may be interpreted as the environmental dumping eﬀect. In this case, the government wants to make the exported good cheaper in order to gain a higher market share in the international market and thus shifting rents from the foreign to the domestic ﬁrms.38 Let us ﬁnally interpret the fourth term in (4.92). This term takes into account the welfare loss of domestic consumers due to imperfect competition. Its sign is clearly negative, since total output decreases if the domestic tax rate goes up. However, this eﬀect is bound to be smaller in an open economy than in the closed economy. The reason is another leakage eﬀect: if the domestic government lowers the emission tax to increase output for the sake of increasing the domestic consumer surplus, the foreign ﬁrms will react by reducing their output, which harms the domestic consumers. Therefore, the regulator makes less of an eﬀort to increase domestic consumption in an open economy than in a closed economy. The bottom line from this analysis is that it depends crucially on the number of domestic ﬁrms whether the (domestic) government sets out to use the emission tax to improve the terms of trade or elects the opposite by

Environmental policy under imperfect competition

195

increasing international market shares in order to shift rents from the foreign to the domestic ﬁrms. So allowing for more than just one ﬁrm in each country, as done by both Althammer and Buchholz (1995, 1999) and Duval and Hamilton (2002), gives useful insights which are not possible in the one-ﬁrm-per-country models proposed by Conrad (1993), Barrett (1994), and follow-up papers such as Ulph (1994a, 1994b, 1996a, 1996b) and A.Ulph and D. Ulph (1996). Note ﬁnally that, in this section, we have been investigating only quantity competition. It is not possible to extend this analysis in a simple way to the model of price competition à la Barrett for the case of more than one ﬁrm in each country. The reason is that imperfect price competition requires diﬀerentiated products. However, apart from the Dixit–Stiglitz model, which does not really model the strategic interaction between ﬁrms, and Salop’s model of the circular city, there exist no symmetric models of imperfect price competition with more than two competing ﬁrms. 10.1.3 Extensions to models with endogenous numbers of ﬁrms There are a small number of recent papers that endogenize the number of ﬁrms in each country. Gürtzgen and Rauscher (2000) consider a model with the Dixit–Stiglitz type of monopolistic competition. They ﬁnd that a stricter environmental domestic standard need not lead to an increase in domestic pollution. In our parlance, this means that the second term in (4.91) or in (4.92) need not necessarily be negative. Moreover, the standard improves the terms of trade in the domestic country. Finally, there is a negative term accounting for imperfect competition (similar to the closedeconomy model suggested by Lange and Requate, 1999). Thus in total, the marginal abatement costs may exceed or fall short of marginal damage. Bayindir-Upmann (2003) studies a Cournot model where the domestic number of ﬁrms is ﬁxed while the foreign number of ﬁrms is determined endogenously by a zero-proﬁt condition. He shows that an increase in the domestic emission tax leads to an increase in the number of foreign ﬁrms but leaves total output and total pollution unaﬀected (domestic pollution goes down if the pollutant is not fully transboundary). Thus by raising the domestic tax rate, domestic production is crowded out at a one-to-one rate by foreign producers. An increase of the foreign emission tax, by contrast, leads to a decrease in the number of foreign ﬁrms and a reduction in total pollution. Hence the second-best optimal tax structure for the foreign government corresponds to (4.91) whereas for the domestic government the terms-of-trade eﬀect vanishes. This is the case because the domestic government is not able to inﬂuence the foreign ﬁrms’ costs. Consequently, the domestic government is not able to inﬂuence the terms of trade. Raising the tax rate would only result in losing market shares to the foreign country.

196

Yearbook of environmental and resource economics

11. CONCLUSIONS AND PROPOSALS FOR FURTHER RESEARCH In this chapter I have summarized the main issues and results relating to environmental policy when ﬁrms subject to environmental regulation exercise market power on at least one market. The bulk of the literature refers to situations where the ﬁrms have market power on the output market, the most prominent of these models being monopoly, diﬀerent kinds of oligopoly, and monopolistic competition. The majority of models summarized in this chapter work on the assumption that a regulator has only one instrument for the regulation of various market imperfections, excessive pollution, insuﬃcient output, and – sometimes – excessive market entry. One typical result is that the second-best optimal price for pollution is below the marginal social damage caused by the pollution. Roughly speaking, the reason is that, in comparison to a market structure with perfect competition, ﬁrms exercising market power hold down output and thus pollute less. How relevant is this insight? We have seen that the second-best optimal tax rate depends crucially on the size of the demand elasticity of the relevant market in which market power is exercised. Is it feasible to set particular tax rates for each market in which polluting ﬁrms exercise market power? In many cases, ﬁrms with market power in a particular market emit the same pollutant as other ﬁrms engaging in perfect competition in some other market. Is it feasible to set discriminatory taxes in this case? If it is, a whole array of diﬀerent tax rates may be subject to lobbying eﬀort by the industry thus regulated. Note that Innes et al. (1991) is one of the few papers that studies simultaneous regulation of both a large monopolistic ﬁrm and many competitive small polluting ﬁrms. Moreover, we have seen two interesting theoretical ﬁndings. First, in Cournot models with free entry, the Pigouvian tax equal to marginal damage seems to be a good rule of thumb. Second, as Misiolek (1988) has pointed out, even in the monopoly case, rent-seeking eﬀorts lead to (second-best optimal) tax rates higher than the monopoly tax rule, as was originally suggested by Barnett (1980). So even under imperfect competition, the Pigouvian rule is not such a bad option. Accordingly, from a normative point of view, it seems to be a good strategy to stick to the Pigouvian rule and to encourage more competition through tough anti-trust laws or – if monopoly power cannot be excluded – by direct control of market power. From my survey it has become evident that not every pollution-control instrument has been investigated for each kind of market structure. For example, in models of monopolistic competition it is largely the impact of taxes and second-best tax rules that has been investigated. It is certainly not

Environmental policy under imperfect competition

197

necessary to ﬁll up all the gaps by investigating each instrument for each kind of market structure. However, the observations I have outlined above indicate several directions for further research. The ﬁrst one is empirical. It would be important to measure the market power (performance) of polluting ﬁrms in the output market in order to assess how serious the distortion from applying the Pigouvian rule to imperfectly competitive markets will be, and also for determining the second-best optimal tax levels in practice, if this is necessary. Determining the size of the second-best optimal tax rates would also require the empirical estimation of demand elasticities for the output commodities. A second avenue of research would be to investigate the relationship between anti-trust and environmental policy. As a matter of fact, in most countries competition law prevents a monopolist from simply charging the monopoly price (e.g. according to Art. 82 of the European Treaty abuse of market power is not allowed). Some industries with potential monopoly power, notably utilities, are subject to even stricter regulations. Given these institutional settings, the simple second-best optimal tax formulas, as developed by Barnett (1980) and many others, including myself, do not apply. Third, it would be interesting to investigate whether there are incentives for ﬁrms to (ab)use voluntary environmental agreements and commitments in order to bypass anti-trust laws. (Note that caps on emissions imply caps on output and thus raise prices.) One paper aiming in this direction is Conrad, 2001, who addresses a similar question in a strategic international trade framework. Moreover, there are very few papers studying the simultaneous regulation of both market power and pollution. In this case, the most simple rule is to tax emissions and subsidize output. Subsidies, however, are prohibited by several international agreements (EU Treaty, WTO rules etc.). Therefore more sophisticated mechanisms have to be developed for regulating a number of market imperfections. Besides Laﬀont (1994), Kim and Chang (1993) present one of the few papers proposing mechanisms for regulating both imperfections on the output market and externalities caused by pollution. Certainly more needs to be done on this issue. Besides market power on the output market we have also summarized models with market power on factor markets. One important factor market is the market for emission permits, which has attracted much attention from researchers since Hahn’s inﬂuential paper (1984). One unsatisfactory feature in this and several follow-up models is the exogenous and rather arbitrary division of ﬁrms into powerful and competitive ﬁrms. Here an issue for further research would be the question of how to endogenize the degree of market power depending on ﬁrm size. This question is certainly not speciﬁc to permit markets, though it is of special interest there since big

198

Yearbook of environmental and resource economics

and small ﬁrms interact in the same market. Possible tools for solving this problem could include techniques from the theory of multi-unit auctions. Moreover, it has become clear from the analysis of factor markets that knowledge about market power in a particular market to be regulated is not enough. Strictly speaking, the regulator has to know the whole vertical structure, including the degree of market power at each step of the production chain, in order to determine the optimal tax rate, or the level of some other policy instrument. On this point, too, we still know relatively little about the relationship between optimal tax rules and market power at diﬀerent stages of the production process. A further point which has only been treated in the literature with reference to tradable permits is compliance. As summarized above, several authors, such as van Egteren and Weber (1996), Malik (2002) and Chavez and Stranlund (2003), have investigated the consequences of optimal permit policies when some ﬁrms exercise market power in the market for tradable permits and some ﬁrms do not fully comply to the regulations (i.e. if they emit in excess of the number of allowances they hold). This problem, however, is not speciﬁc to the permit instrument. Thus there is need for more general consideration of the relationship between the optimal level of pricing instruments and the optimal level of monitoring. There is also need to investigate the relationship between the pricing instrument and penalty fees in case of non-compliance. In all the models studied here, pollution and the damage resulting from it is assumed to be deterministic. However, in many cases random elements may determine either emissions or damage. For example, accidental hazardous emissions may result from insuﬃcient care on the ﬁrms’ part. For those cases we have a diﬀerent tool-box of environmental policy rules, i.e. liability rules such as strict liability or negligence. To my knowledge, the market structure of output markets has never been an issue in the literature discussing these rules. But even if emissions are determined completely by both output and the ﬁrms’ abatement eﬀort, the resulting damage may be inﬂuenced by natural shocks to resilience and nature’s capacity for reducing pollution. Sometimes even the weather conditions are crucial in determining the size of the damage. In connection with stochastic damages, too, diﬀerent forms of market structure have yet to be taken into account. In particular, we do not know how optimal emission taxes look like if ﬁrms have market power and damages are stochastic. Besides the need for further theoretical and much more empirical work, experimental work may be useful to pre-test diﬀerent forms of regulation in the laboratory. Surprisingly, there is a considerable amount of experimental research on the Hahn model, i.e. a permit market with one big and many small ﬁrms. There are, however, almost no experimental contributions on

Environmental policy under imperfect competition

199

alternative forms of regulation. In particular it might be interesting to test mechanisms designed to regulate several distortions, such as the elegant mechanism suggested by Kim and Chang (1993). The interesting issue here is that the price of pollution and the implicit subsidy on output does not only depend on the setting of the regulator, but also on the (strategic) behavior of the other ﬁrms. Since those mechanisms are relatively complex, it is far from clear whether the ﬁrms will fully understand the regulation scheme and whether they will act in a way that accords with what the theory predicts. If it turns out that ﬁrms behave diﬀerently from theoretical predictions, this has further consequences for the proper design of regulatory multi-issue schemes. Although this chapter has become rather long and I have tried to cover the static models of complete information as fully as possible, I have only been able to touch on the literature on environmental policy under imperfect competition including (a) asymmetric information (see e.g. Kim and Chang, 1993; Laﬀont, 1994), (b) dynamic modelling of accumulating pollutants (e.g. Benchekroun and Van Long, 1997), (c) R&D decisions prior to production (see e.g. Montero, 2002a, 2002b; Fischer et al., 2003; and Requate, 2005), and several other issues. In particular, the last-named issue is a promising ﬁeld of research, since environmental regulation has a crucial impact on the direction taken by technological progress, and R&D typically creates market structures where only a small number of ﬁrms conduct R&D projects and thus engage in imperfect competition. Finally, maybe a caveat is in place. Despite the need for further research in the various directions outlined above, distinguishing too many structural forms of competition and oﬀering too many second-best rules for diﬀerent instruments could do more harm than good to the authority of economics as a discipline. Politicians need simple and clear-cut rules. Complicated rules depending on too many parameters, such as marginal damage, demand elasticities in the output market, market conduct, and many more, may either not be applied at all, or they run the risk of being manipulated by lobbyists. This gives rise to a ﬁnal avenue of research, i.e. the dimension of political economy, and the question of which instruments and which institutional frameworks deciding on the level of those instruments are most eﬀective in resisting the inﬂuence of political interest groups.

NOTES *

The author is grateful to Gunter Bahr, Albrecht Bläsi, Henk Folmer, Andrew Jenkins, Thomas Lontzek, Dagmar Nelissen, Thure Traber, Andreas von Döllen and two anonymous referees for carefully reading the manuscript, pointing out errors and making

200

1. 2. 3. 4. 5. 6. 7. 8.

9. 10. 11. 12. 13. 14. 15. 16. 17.

18. 19. 20. 21. 22. 23.

Yearbook of environmental and resource economics helpful suggestions, to Thorsten Bayindir-Upmann for helpful discussions, and to Tom Tietenberg for valuable indications on further literature. If ﬁrms have constant marginal costs, it is socially optimal that only those ﬁrms with the lowest marginal social cost (including marginal social damage) produce. In this case, the ﬁrst-order condition (4.8) is substituted by Cie (qi, ei ) , where is the competitive price for tradable permits. Note that we do not properly solve for since the right-hand side of (4.13) also depends on . This can be seen easily by diﬀerentiating (4.10) with respect to . Solving for dqM( )/d

yields: dqM ( ) d [Pq 2P C] 0 by Assumptions 1 and 3. Of course he can also tax or subsidize output with the same eﬀect. Although there is no unique deﬁnition in the literature, eco-dumping usually refers to underinternalizing the social damage of pollution. See Rauscher (1994) for a careful discussion of the concept. In the model of Innes et al. (1991), the monopolist and the competitive ﬁrms operate on the same market. Thus the competitive ﬁrms act as a competitive fringe to the monopolist. Inspection of (4.40) shows that the direct eﬀect of increasing n on the second-best tax rate is positive. To show that the direct eﬀect dominates the indirect eﬀects, we need to diﬀerentiate (4.40) totally with respect to . In order to establish the result d */dn 0, we have to assume that the third derivatives of the ﬁrms’ cost functions Ci have the ‘right’ sign or are suﬃciently bounded. Assumptions like these are always necessary when doing comparative statics of second-best instruments with respect to exogenous parameters. In fact, Ebert’s model is slightly more general by assuming that e is an increasing function of q. See section 4.4 below, where we study a tax/subsidy system in more detail for the case where the ﬁrm can separately decide on output and emissions. Kim and Chang model the ﬁrms’ cost functions in a slightly diﬀerent way. I have adapted their model and notation to the model used throughout this survey. Note that the term D(e~ i ) is not necessary for eﬃciency. It does, however, guarantee that each ﬁrm only pays for the additional damage it generates over and above the other ﬁrms’ emissions. It prevents the ﬁrms’ tax burden from becoming too high. The steepness of the damage function (the slope of the marginal rates of substitution of the preferences) is also of signiﬁcance in related models, for example in Weitzman’s (1974) paper on regulation under imperfect information. See also Baumol and Oates (1988). I have not listed all these papers here because this is not the central concern of this chapter. For an excellent survey see Bovenberg (1999). This does not hold for the most inﬂuential paper on the double dividend by Bovenberg and de Mooji (1994) or for other papers by Bovenberg and co-authors. I use the letter G for demand instead of D as usually found in textbooks, since I reserve D for the damage function. In models of free entry this is a standard assumption. It can be justiﬁed by arguing, ﬁrst, that only ﬁrms using technologies inducing the lowest average costs will prevail, and, second, that if several technologies are randomly drawn from a continuous distribution of parameters, it is extremely unlikely that two diﬀerent technologies will induce the same average cost. Our assumptions guarantee that an equilibrium exists and that it is unique and stable. This phenomenon does not depend on the assumption of imperfect competition but also materializes under perfect competition. See Requate (1995), Requate and Unold (2001, 2003). If C(q, e)[q e]2/2, then the last term boils down to Cqe Cee , which is smaller than zero if the relative standard bites, i.e. . All the other terms of the numerator are positive under the assumptions we make throughout this chapter. See Tirole (1988), pp. 298–9. Lange and Requate (1999) provide a numerical example for such a case. Dixit and Stiglitz (1977), p. 298 refer to this assumption in their original model as the ‘natural’ case.

Environmental policy under imperfect competition 24.

25. 26. 27. 28. 29.

30. 31. 32. 33. 34. 35.

36. 37. 38.

201

This assumption corresponds to price elasticity of demand for the compound commodity that is greater than 1. Assume for a moment that n ﬁrms collude and jointly maximize their proﬁts. Then 1 is a necessary condition for guaranteeing the existence of a joint proﬁt maximum. In fact, von der Fehr’s model is slightly more general since he allows for quantity competition with diﬀerentiated products. This means that ﬁrms have constant marginal costs and emissions are proportional to output. The eﬀects, of course, do not necessarily add up in a linear way. In contrast to Conrad and Kennedy, Barrett uses standards. Conrad (1996a) extends Conrad (1993) by assuming that the good supplied on the world market by oligopolistic ﬁrms is also consumed at home. Conrad (1996b) extends Barrett’s model by also studying taxation in a price-setting duopoly model with diﬀerentiated commodities. Simpson and Bradford (1996) use a model of imperfect competition and R&D prior to market competition to challenge the Porter hypothesis proposed in Porter (1990, 1991). Ebert (1999) studies the strategic use of relative standards in open economies where the regulated domestic ﬁrms exercise market power in an international market. Rauscher (1994) provides a detailed discussion on the meaning of ecological dumping in the framework of a competitive model. One country might be much more seriously aﬀected by pollution than another. This problem has often been ignored in the literature. In reality, direct transfer payments are not so common, but examples do exist. For instance, Germany and the Netherlands made direct payments to France for abating eﬄuents from salt mining. In other cases, indirect payments are made by negotiating multiple issues simultaneously. Our assumptions guarantee downward-sloping reaction functions. Under perfect competition we need no further conditions to show this eﬀect. The ‘normal’ conditions refer to Cournot competition, where perverse eﬀects can arise under extreme asymmetries of the ﬁrms or extreme curvatures of the inverse demand function. Interestingly, Duval and Hamilton (2002) do not mention the rent-shifting eﬀect, while van Long and Soubeyran (1999) – using almost the same model and a decomposition similar to (4.92) – only mention the rent-shifting eﬀect but do not refer to a terms-oftrade eﬀect.

REFERENCES Althammer, W. and W. Buchholz (1995), ‘Strategic trade incentives in environmental policy,’ Finanzarchiv 52, 293–305. Althammer, W. and W. Buchholz (1999), ‘Distorting environmental taxes: the role of market structure,’ Jahrbücher für Nationalökonomie und Statistik 219, 257–70. Asch, P. and J.J. Seneca (1976), ‘Monopoly and external cost: an application of second best theory to the automobile industry,’ Journal of Environmental Economics and Management 3, 69–79. Barnett, A.H. (1980), ‘The Pigouvian tax rule under monopoly,’ American Economic Review 70, 1037–41. Baron, D. and R. Myerson (1982), ‘Regulating a monopolist with unknown costs,’ Econometrica 50, 911–30. Barrett, S. (1994), ‘Strategic environmental policy and international trade,’ Journal of Public Economics 54, 325–38. Baumol, W.J. (1972), ‘On taxation and the control of externalities,’ American Economic Review 64, 307–21.

202

Yearbook of environmental and resource economics

Baumol, W.J. and W.E. Oates (1988), The Theory of Environmental Policy, Cambridge, MA: Cambridge University Press. Bayindir-Upmann, T. (2000), ‘Do monopolies justiﬁably fear environmental tax reforms,’ Finanzarchiv 57, 459–84. Bayindir-Upmann, T. (2003), ‘Strategic environmental policy under free entry of ﬁrms,’ Review of International Economics 11, 379–96. Bayindir-Upmann, T. (2004), ‘On the double dividend under imperfect competition,’ Environmental and Resource Economics 28, 169–94. Benchekroun, H. and N. van Long (1997), ‘Eﬃciency inducing taxation for polluting oligopolists,’ Journal of Public Economics 70, 325–42. Bohm, P. and B. Carlén (1999), ‘Emission quota trade among the few: Laboratory evidence of joint implementation among committed countries,’ Resource and Energy Economics 21, 43–66. Bovenberg, L. (1999), ‘Green tax reforms and the double dividend: an updated reader’s guide,’ International Tax and Public Finance 6, 41–60. Bovenberg, L. and R. de Mooij (1994), ‘Environmental levies and distortionary taxation,’ American Economic Review 84, 1085–9. Brander, J. and B. Spencer (1985), ‘Export subsidies and international market share rivalry,’ Journal of International Economics 16, 227–42. Brown-Kruse, J. and S.R. Elliott (1990), ‘Strategic manipulation of pollution permit markets: an experimental approach,’ University of Colorado, Boulder, Department of Economics working paper. Brown-Kruse, J., S.R. Elliott and R. Godby (1995), ‘Strategic manipulation of pollution permit markets: an experimental approach,’ Department of Economics, McMaster University working paper 95–10, Hamilton, Canada. Buchanan, J.M. (1969), ‘External diseconomies, corrective taxes, and market structure,’ American Economic Review 59, 174–7. Buchanan, J.M. and W.C. Stubblebine (1962), ‘Externality,’ Economica 29, 371–84. Bulow, J. (1986), ‘An economic theory of planned obsolescence,’ Quarterly Journal of Economics 101, 729–49. Carlén, B. (2003), ‘Market power in international carbon emissions trading: a laboratory test,’ The Energy Journal 24, 1–26. Carlsson, F. (2000), ‘Environmental taxation and strategic commitment in duopoly models,’ Environmental and Resource Economics 15, 243–56. Carraro, C. and A. Soubeyran (1996), ‘Environmental taxation, market share, and proﬁts in oligopoly,’ in C. Carraro, Y. Katsoulacos and A. Xepapadeas (eds), Environmental Policy and Market Structure, Dordrecht: Kluwer Academic Publishers. Cason, T.N., L. Gangadharan and C. Duke (2003), ‘Market power in tradeable emission markets: a laboratory testbed for emission trading in Port Phillip Bay, Victoria,’ Ecological Economics 46, 469–91. Chavez, C. and J. Stranlund (2003), ‘Enforcing transferable permit systems in the presence of market power,’ Environmental and Resource Economics 25, 65–78. Conrad, K. (1993), ‘Taxes and subsidies for pollution-intense industries as trade policy’, Journal of Environmental Economics and Management 25, 121–35. Conrad, K. (1996a), ‘Optimal environmental policy for oligopolistic industries under intra-industry trade,’ in C. Carraro, Y. Katsoulacos and A. Xepapadeas (eds), Environmental Policy and Market Structure, Dordrecht: Kluwer Academic Publishers.

Environmental policy under imperfect competition

203

Conrad, K. (1996b), ‘Choosing emission taxes under international price competition,’ in C. Carraro, Y. Katsoulacos and A. Xepapadeas (eds), Environmental Policy and Market Structure, Dordrecht: Kluwer Academic Publishers. Conrad, K. (2001), ‘Voluntary environmental agreements vs. emission taxes in strategic trade models’, Environmental and Resource Economics, 19 (4), 361–81. Conrad, K. and J. Wang (1993), ‘The eﬀect of emission taxes and abatement subsidies on market structure,’ International Journal of Industrial Organization 11, 499–518. Cropper, M. and W. Oates (1992), ‘Environmental economics: a survey,’ Journal of Economic Literature 30, 675–740. Damania, R. (2000), ‘Financial structure and the eﬀectiveness of pollution control in an oligopolistic industry,’ Resource and Energy Economics 22, 21–36. David, M. and B. Sinclair-Desgagné (2005), ‘Environmental regulation and the ecoindustry,’ Journal of Regulatory Economics, 28, 141–55. Dixit, A. and J. Stiglitz (1977), ‘Monopolistic competition and optimal product diversity,’ American Economic Review 67, 297–308. Duval, Y. and S. Hamilton (2002), ‘Strategic environmental policy and international trade in asymmetric oligopoly markets,’ International Tax and Public Finance 9, 259–71. Ebert, U. (1992), ‘Pigouvian taxes and market structure: the case of oligopoly and diﬀerent abatement technologies,’ Finanzarchiv 49, 154–66. Ebert, U. (1999), ‘Relative standards as strategic instruments in open economies,’ in E. Petrakis, E. Sartzetakis and A. Xepapadeas (eds), Environmental Regulation and Market Power – Competition, Time Consistency and International Trade, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Ebert, U. and O. von dem Hagen (1998), ‘Pigouvian taxes under imperfect competition if consumption depends on emissions,’ Environmental and Resource Economics 12, 507–13. Fischer, C., I. Parry and W. Pizer (2003), ‘Instrument choice for environmental protection when technological innovation is endogenous,’ Journal of Environmental Economics and Management 45, 523–45. Fullerton, D. and G.E. Metcalf (2002), ‘Cap and trade policies in the presence of monopoly and distortionary taxation,’ NBEW working paper 8901. Gersbach, H. and T. Requate (2004), ‘Emission taxes and the design of refunding schemes,’ Journal of Public Economics 88, 713–25. Godby, R. (2000), ‘Market power and emission trading: theory and laboratory results,’ Paciﬁc Economic Review 5, 349–63. Godby, R. (2002), ‘Market power in laboratory emission permit markets,’ Environmental and Resource Economics 23, 279–318. Gürtzgen, N. and M. Rauscher (2000), ‘Environmental policy, intra-industry trade and transfrontier pollution,’ Environmental and Resource Economics 17, 59–71. Hagem, C. and H. Westskog (1998), ‘The design of a dynamic tradeable quota system under market imperfections,’ Journal of Environmental Economics and Management 36, 89–107. Hahn, R. (1984), ‘Market power and transferable property rights,’ Quarterly Journal of Economics 99, 753–65. Hamilton, S. and T. Requate (2004), ‘Vertical contracts and strategic environmental trade policy,’ Journal of Environmental Economics and Management 47, 260–69. Holmlund, B. and A.-S. Kolm (2000), ‘Environmental tax reform in a small open economy with structural unemployment,’ International Tax and Public Finance 7, 315–90.

204

Yearbook of environmental and resource economics

Howe, C. (1994), ‘Taxes versus tradeable discharge permits: a review in the light of the US and European experience,’ Environmental and Resource Economics 4, 151–69. Innes, R., C. Kling and J. Rubin (1991), ‘Emission permits under monopoly,’ Natural Resource Modelling 8, 321–43. Katsoulacos, Y. and A.P. Xepapadeas (1995), ‘Pigouvian taxes under oligopoly,’ Scandinavian Journal of Economics 97, 411–20. Katsoulacos, Y. and A.P. Xepapadeas (1996), ‘Emission taxes and market structure,’ in C. Carraro, Y. Katsoulacos and A. Xepapadeas (eds), Environmental Policy and Market Structure, Dordrecht: Kluwer Academic Publishers, pp. 3–22. Kennedy, P.W. (1994), ‘Equilibrium pollution taxes in open economies with imperfect competition,’Journal of Environmental Economics and Management 27, 49–63. Kim, J.C. and K.B. Chang (1993), ‘An optimal tax/subsidy for output and pollution control under asymmetric information in oligopoly markets,’ Journal of Regulatory Economics 5, 183–97. Kreps, D. and J. Scheinkman (1983), ‘Quantity precommitment and Bertrand competition yield Cournot outcomes,’ Bell Journal of Economics 14, 326–37. Laﬀont, J.J. (1994), ‘Regulation of pollution with asymmetric information,’ in C. Dosi and T. Tomasi (eds), Nonpoint Source Pollution Regulation: Issues and Analysis, Dordrecht: Kluwer Academic Publishers. Lange, A. and T. Requate (1999), ‘Emission taxes for price setting ﬁrms: Diﬀerentiated commodities and monopolistic competition,’ in E. Petrakis, E. Sartzetakis and A. Xepapadeas (eds), Environmental Regulation and Market Power – Competition, Time Consistency and International Trade, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Lee, D.R. (1975), ‘Eﬃciency of pollution taxation and market structure,’ Journal of Environmental Economics and Management 2, 69–72. Lee, S.-H. (1999), ‘Optimal taxation for polluting oligopolists with endogenous market structure,’ Journal of Regulatory Economics 15, 293–308. Levin, D. (1985), ‘Taxation within Cournot oligopoly,’ Journal of Public Economics 27, 281–90. Loeb, M. and W.A. Magat (1993), ‘A decentralized method for utility regulation,’ Journal of Law and Economics 22, 399–404. Maeda, A. (2003), ‘The emergence of market power in emissions rights markets: the role of initial permit distribution,’ Journal of Regulatory Economics 24, 293–314. Malik, A. (2002), ‘Further results on permit markets with market power and cheating,’ Journal of Environmental Economics and Management 44, 371–90. Malueg, D.A. (1990), ‘Welfare consequences of emission trading programs,’ Journal of Environmental Economics and Management 18, 66–77. Markusen, J.R. (1975), ‘International externalities and optimal tax structures,’ Journal of International Economics 5, 531–51. Markusen, J.R., E.R. Morey and N. Olewiler (1993), ‘Environmental policy when market structure and plant locations are endogenous,’ Journal of Environmental Economics and Management 24, 68–86. Markusen, J.R., E.R. Morey and N. Olewiler (1995), ‘Competition in regional environmental policies when plant locations are endogenous,’ Journal of Public Economics 56, 55–77. Marsiliani, L. and T. Renström (1997), ‘Imperfect competition, labour market distortions and the double dividend hypothesis,’ University of Birmingham Department of Economics discussion paper no. 97-26.

Environmental policy under imperfect competition

205

Misiolek, W. (1980), ‘Eﬄuent taxation in monopoly markets,’ Journal of Environmental Economics and Management 7, 103–7. Misiolek, W. (1988), ‘Pollution control through price incentives: the role of rent seeking costs in monopoly markets,’ Journal of Environmental Economics and Management 15, 1–8. Misiolek, W. and H. Elder (1989), ‘Exclusionary manipulation of markets for pollution rights,’ Journal of Environmental Economics and Management 16, 156–66. Montero, J.P. (2002a), ‘Permits, standards, and technological innovation,’ Journal of Environmental Economics and Management 44, 23–44. Montero, J.P. (2002b), ‘Market structure and environmental innovation,’ Journal of Applied Economics 5, 293–325. Montgomery, W.D. (1972), ‘Markets in licenses and eﬃcient pollution control programs,’ Journal of Economic Theory 5, 395–418. Muller, R.A., S. Mestelman, J. Spraggon and R. Godby (2002), ‘Can double auctions control monopoly and monopsony power in emissions trading markets?’ Journal of Environmental Economics and Management 44, 70–92. Nannerup, N. (1998), ‘Strategic environmental policy under incomplete information,’ Environmental and Resource Economics 11, 61–78. Oates, W.E. and D.L. Strassmann (1984), ‘Eﬄuent fees and market structure,’ Journal of Public Economics 24, 29–46. Pigou, A.C. (1938), The Economics of Welfare, London: MacMillan and Co. Porter, M.E. (1990), The Competitive Advantage of Nations, New York: Free Press. Porter, M.E. (1991), ‘America’s green strategy,’ Scientiﬁc American 264, 168. Rauscher, M. (1994), ‘On ecological dumping,’ Oxford Economic Papers 46, 822–40. Rauscher, M. (1997), International Trade, Factor Movements and the Environment, Oxford: Oxford University Press. Requate, T. (1993a), ‘Pollution control in a Cournot duopoly via taxes or permits,’ Journal of Economics 58, 255–91. Requate, T. (1993b), ‘Equivalence of eﬄuent taxes and permits for environmental regulation of several local monopolies,’ Economics Letters 42, 91–5. Requate, T. (1993c), ‘Pollution control under imperfect competition: asymmetric Bertrand duopoly under linear technologies,’ Journal of Institutional and Theoretical Economics 149, 415–42. Requate, T. (1995), ‘Incentives to adopt new technologies under diﬀerent pollutioncontrol policies,’ International Tax and Public Finance 2, 295–317. Requate, T. (1997), ‘Green taxes in oligopoly if the number of ﬁrms is endogenous,’ Finanzarchiv 54, 261–80. Requate, T. (2005), ‘Timing and commitment of environmental policy, adoption of new technology, and repercussions on R&D,’ Environmental and Resource Economics, 31, 175–99. Requate, T. and W. Unold (2003), ‘On the incentives of environmental policy to adopt advanced abatement technology – will the true ranking please stand up?’ European Economic Review 47, 125–46. Requate, T. and W. Unold (2001), ‘On the incentives of policy instruments to adopt advanced abatement technology if ﬁrms are asymmetric,’ Journal of Institutional and Theoretical Economics 157, 536–54. Runkel, M. (2002), ‘A note on emissions taxation in durable good oligopoly,’ Journal of Industrial Economics 50, 235–6. Runkel, M. (2004), ‘Optimal emissions taxation under imperfect competition in a durable good industry,’ Bulletin of Economic Research 56, 115–31.

206

Yearbook of environmental and resource economics

Salop, S.C. and D.T. Scheﬀman (1983), ‘Raising rivals’ costs,’ American Economic Review 73, 267–71. Salop, S.C. and D.T. Scheﬀman (1987), ‘Cost-raising strategies,’ Journal of Industrial Economics 26, 19–34. Sartzetakis, E. (1997a), ‘Raising rivals’ costs strategies via emission permits markets,’ Review of Industrial Organization 12, 751–65. Sartzetakis, E. (1997b), ‘Tradeable emission permits regulations in the presence of imperfectly competitive product markets: welfare implications,’ Environmental and Resource Economics 9, 65–81. Sartzetakis, E.S. and D.G. McFetridge (1999), ‘Emission permits trading and market structure,’ in E. Petrakis, E. Sartzetakis and A. Xepapadeas (eds), Environmental Regulation and Market Power – Competition, Time Consistency and International Trade, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Siebert, H. (1976), ‘Emissionssteuern im Monopol. Eine Anmerkung,’ Zeitschrift für die gesamte Staatswissenschaft 132, 679–82. Simpson, D. (1995), ‘Optimal pollution taxation in a Cournot duopoly,’ Environmental and Resource Economics 6, 359–69. Simpson, D. and D. Bradford, III (1996), ‘Taxing variable cost: environmental regulation as industrial policy,’ Journal of Environmental Economics and Management 30, 282–300. Smith, V.K. (1976), ‘A note on eﬄuent charges and market structure,’ Journal of Environmental Economics and Management 2, 309–11. Smith, V. (1981), ‘An experimental study of decentralized institutions of monopoly restraint,’ in G. Horwich and J. Quirk (eds), Essays in Contemporary Fields of Economics, West Lafayette, IN: Purdue University Press. Spence, A.M. (1976), ‘Product selection, ﬁxed costs, and monopolistic competition,’ Review of Economic Studies 43, 217–36. Stimming, M. (1999), ‘Capital accumulation games under environmental regulation and duopolistic competition,’ Journal of Economics 69, 267–87. Tirole, J. (1988), The Theory of Industrial Organization, Cambridge, MA: MIT Press. Ulph. A. (1994a), ‘Environmental policy, plant location and government protection,’ in C. Carraro (ed.), Trade, Innovation, Environment, Dordrecht: Kluwer Academic Publishers, pp. 123–63. Ulph, D. (1994b), ‘Strategic innovation and strategic environmental policy,’ in C. Carraro (ed.), Trade, Innovation, Environment, Dordrecht: Kluwer Academic Publishers, pp. 205–28. Ulph, A. (1996a), ‘Environmental policy and international trade when governments and producers act strategically,’ Journal of Environmental Economics and Management 30, 265–81. Ulph, A. (1996b), ‘Environmental policy instruments and imperfectly competitive international trade,’ Environmental and Resource Economics 7, 333–55. Ulph, A. (1996c), ‘ Strategic environmental policy and international trade – the role of market conduct,’ in C. Carraro, Y. Katsoulacos and A. Xepapadeas (eds), Environmental Policy and Market Structure, Dordrecht: Kluwer Academic Publishers, pp. 94–127. Ulph, A. (1997a), ‘International trade and the environment: a survey of recent economic analysis,’ in H. Folmer and T. Tietenberg (eds), Yearbook of Environmental and Resource Economics, Cheltenham, UK and Lyme, MA, USA: Edward Elgar, pp. 66–96.

Environmental policy under imperfect competition

207

Ulph, A. (1997b), ‘Environment policy and international trade,’ in C. Carraro and D. Siniscalco (eds), New Directions in the Economic Theory of the Environment, Cambridge: Cambridge University Press, pp. 147–92. Ulph, A. and D. Ulph (1996), ‘Trade, strategic innovation and strategic environmental policy – a general analysis,’ in C. Carraro, Y. Katsoulacos and A. Xepapadeas (eds), Environmental Policy and Market Structure, Dordrecht: Kluwer Academic Publishers, pp. 181–208. Van Egteren, H. and M. Weber (1996), ‘Marketable permits, market power, and cheating,’ Journal of Environmental Economics and Management 30, 161–73. Van Long, N. and A. Soubeyran (1999), ‘Pollution, Pigouvian taxes, and asymmetric international oligopoly,’ in E. Petrakis, E. Sartzetakis and A. Xepapadeas (eds), Environmental Regulation and Market Power, Competition, Time Consistency and International Trade, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 175–94. Van Long, N. and A. Soubeyran (2005), ‘Selective penalization of polluters: an infconvolution approach,’ Economic Theory 25, 421–54. von der Fehr, N.-H. (1993), ‘Tradeable emission rights and strategic interaction,’ Environmental and Resource Economics 3, 129–51. Weigel, O. (1992), ‘Schadstoﬀvermeidung im Oligopol – Eine numerische Untersuchung,’ Institute of Mathematical Economics, MA thesis, University of Bielefeld. Weitzman, M.L. (1974), ‘Prices vs. quantities,’ Review of Economic Studies 41, 477–91. Yin, X. (2003), ‘Corrective taxes under oligopoly with inter-ﬁrm externalities,’ Environmental and Resource Economics 26, 269–77.

5.

Transport and the environment Piet Rietveld

1.

INTRODUCTION

Transport is a major contributor to some of our environmental problems, at both the local and the global level. In many countries substantial eﬀorts have been made to reduce these problems, and often with substantial success. Nevertheless environmental problems related to transport appear diﬃcult to curb. Long-run projections indicate substantial scope for further growth in transport demand, of both passengers and freight (Schafer, 1998; OECD, 2002). And opportunities to reduce emissions from transport activity usually lead to various rebound eﬀects, which means that the ﬁnal improvement in environmental performance of transport is disappointing or even absent. Hence it is no surprise that the theme of sustainable transport has attracted substantial attention among researchers during the last decade, and one may expect that it will continue to do so. One of the indications of the research interest in the theme is that it generated its own scientiﬁc journal, Transport and Environment (Transportation Research D), which ﬁrst appeared in 1996. The notion of sustainable transport – useful as it may be to mobilize attention to environmental problems related to transport – has some limitations that deserve special mention. A main problem is that it tends to focus on transport per se, without incorporating the environmental eﬀects of production and consumption activities it is meant to support. Thus themes such as spatially varying externalities in production activities are not incorporated. This may lead to sub-optimization since the spatial organization of production and consumption has far-reaching eﬀects on total demand for transport. When this spatial pattern is just assumed to be given, the development of a more environmentally friendly transport system may easily lead to under- or over-investment from a systems perspective. Interactions between choice of location, land use and transport therefore deserve attention. An example is the improvement of the energy performance of horticulture exports without considering relocation possibilities to regions where owing to natural circumstances production is less energy intensive. Another example is the optimization of waste transport 208

209

Transport and the environment

Table 5.1 Life-cycle emissions for gasoline-fuelled cars with respect to fuel production, vehicle production and in-service use (g/km) Life-cycle stage

CO2

CO

NOX

HCs from paint

SO2

Particulates

Fuel production Vehicle production Vehicle use

47.0 54.5 186.3

0.061 0.021 3.371

0.174 0.160 0.224

0.388 0.105 0.299

0.185 0.493 0.020

0.011 0.016 0.005

Total

287.8

3.453

0.558

0.792

0.699

0.032

ﬂows versus the system-wide optimization of waste treatment and transport taking into account environmental economies of scale in treatment (see Bartelings, 2003). A related perspective concerns the usual analysis of environmental eﬀects of transport in terms of emissions per vehicle-km, which is incomplete since it does not incorporate the life-cycle eﬀects, also called ‘well-towheels’ eﬀects. An example of what the latter would imply is given by Khare and Sharma (2003), in Table 5.1, although it should be noted that this table is not yet complete since emissions related to the disposal of used vehicles are missing. Standard analysis focuses on vehicle-use eﬀects, whereas life-cycle analysis would incorporate the environmental eﬀects of fuel production and vehicle production. As the table shows, the vehicle-use eﬀects are indeed the larger component in the case of CO2 emissions, having a share of about 2/3. For CO, the vehicle-use component is even close to 100 per cent, but for other aspects such as NOX and particulates the vehicle-use eﬀects have much lower shares, close to 10 per cent or even lower. Such ﬁgures are important inputs for policy analysis, since they suggest that for some types of emissions, the transport activities themselves may not be the most appropriate point of departure for policy intervention. This life-cycle perspective may also have implications for the attractiveness of biofuels. Estimates of emissions in forestry and agriculture to produce biofuels suggest that the environmental performance in this sector is much lower than in the transport sector itself due to a lack of emission regulation for agricultural machinery (Johansson, 1995). So from a life-cycle perspective, improving the environmental performance in forestry and agriculture can be an eﬀective way to reduce emissions due to transport activities. The analysis of transport impacts on the environment entails the use of a number of steps (Friedrich and Bickel, 2002), starting with transport activities and ending with the valuation of disturbances (see Figure 5.1).

210

Yearbook of environmental and resource economics

Transport activities

Response of receptors

Physical impact

Figure 5.1.

Transport and chemical conversion of pollutants

Emissions

Concentration /deposition

Valuation of disturbance

The chain from transport activity to valuation of disturbance

Depending on the type of emission, the complexity of the individual steps may vary, but it is clear that the problem of estimating the environmental costs of transport not only involves valuation issues as such, but also the earlier steps that mainly take place in the physical domain. The structure of this contribution is as follows. In section 2 I discuss trends in transport and their environmental eﬀects from the perspective of decoupling. A more detailed view on environmental eﬀects of transport is given in section 3, followed by a comparison between transport modes in section 4. The potential contribution of technological change to the environmental performance of the transport sector is discussed in section 5. Section 6 is devoted to valuation issues of environmental eﬀects in transport. A survey of transport policy opportunities is given in section 7. Section 8 concludes.

2. TRENDS IN TRANSPORT AND THEIR ENVIRONMENTAL EFFECTS Decoupling the growth in energy use and economic activities is one of the aims of the Kyoto protocol. Long-term trends in the EU indeed indicate that energy consumption is growing much more slowly than GDP (see Table 5.2). However, transport is a striking exception in this development. Energy use in transport is growing faster than the economy. As a consequence the share

211

Transport and the environment

Table 5.2

Transport and economic trends in Europe (EU15), 1970–2000

Population (millions)1 GDP (billion US$ constant 1995) Energy consumption by sector (Mtoe):2 Industry Transport Other3 Total Passenger transport (billion passenger-km) Freight transport (billion tonne-km)

1970

2000

% change, 1970–2000

342 3965

378 8270

11 109

309 145 318 772 2142 1407

325 317 410 1052 4839 3078

5 119 29 36 126 119

Notes: 1 Only six countries were part of the European Community in 1970, but for comparison purposes the data for 1970 and 2000 relate to the 15 countries (EU15) that were members of the European Union in 2000. 2 The energy consumption ﬁgures in the ﬁrst column are for 1971 (not 1970). 3 This category includes energy use in the agricultural, commercial, public service and residential sectors. Sources: European Commission (2003); OECD (2002).

of energy consumed in transport has grown from 19 per cent to 30 per cent in the last three decades. To understand the mechanisms behind the increase of energy use I use the following decomposition (see also Schipper and Fulton, 2003): Energy use (Energy use/transport volume)(transport volume/ GDP) (GDP/Pop) Pop Table 5.2 reveals, for both passenger and freight transport, that the transport intensity of production (transport volume/GDP) has slightly increased. An obvious explanation is that the real generalized costs of transport are decreasing in the long run (see for example Rietveld and Vickerman, 2004) and that institutional barriers to trade were declining during recent decades (Maddison, 2001). The latter meant that as transaction costs in international trade were reduced, international trade volumes increased faster than domestic trade, implying an increase in transport distances. The energy intensity in transport (energy use/transport volume) is about constant. This is a striking result since during recent decades substantial eﬀorts have been made to improve the eﬃciency of energy use in transport. This means that countervailing forces must have

212

Yearbook of environmental and resource economics

been at work. For example, the increase in energy eﬃciency of the combustion engine in cars has been accompanied by the gradual increase in the weight of cars and the increasing use of energy-consuming devices such as air-conditioning (Brink and Van Wee, 2001). A recent development is the large-scale adoption of SUVs (standard utility vehicles) on the roads (Plaut, 2004). In addition there has been a gradual shift from slow to fast transport modes – in particular aviation, which implies higher energy use per kilometre travelled (see Brewer, 1991; Peeters, 1997). Further, GDP per capita increased substantially in the EU, although population growth itself has been modest (11 per cent during 30 years). Thus the major force behind the increase of energy use in transport is the relatively high responsiveness of transport to economic growth and the occurrence of rebound eﬀects within the transport sector already mentioned in section 1 that oﬀset the potential eﬃciency gains that were made possible by technological progress. An important implication of these aggregate ﬁgures is that within the transport sector there is no evidence of the environmental Kuznets curve (for a review see Stern et al., 1996). This curve displays an inverted U shape: at lower levels of economic development emissions increase with production whereas when production displays further growth, emissions start to decrease. This means that as economic activity increases, the energy performance gradually improves to such an extent that the total environmental burden will decrease. When the environmental Kuznets curve holds there is not much reason for concern about economic growth because countervailing mechanisms oﬀ-set adverse environmental eﬀects. But – as the ﬁgures in Table 5.2 show – in the transport sector these mechanisms are not strong enough. We conclude that Table 5.2 gives reasons for concern regarding the longterm environmental consequences of transport. One should be aware, however, that for a substantial number of pollutants much progress has been made. For example, the phasing out of leaded gasoline in Europe and the USA has tremendously reduced lead emissions in transport. Similarly the use of catalytic converters has led to a strong reduction of NOX emissions of road transport, as far as they are using gasoline. Thus in other ﬁelds, outside the energy domain, large improvements have been made. The main remaining sustainability problems that are diﬃcult to solve in the long run seem to be the local ones related to noise, accidents, etc., and the global ones related to energy use. A further observation is that behind the aggregate EU ﬁgures there are substantial diﬀerences between individual countries (Tapio, 2005). Even larger are the diﬀerences between countries in various parts of the world. An example is given in Table 5.3.

213

Transport and the environment

Table 5.3 International comparison of transport ﬁgures, including CO2 emissions EU15 Population Passenger cars (millions) Motorization (cars/1000 persons) Mode of travel (billion p km) Car Air (domestic) Rail Transportation of goods (billion tonne-km) Truck Inland waterways Transport CO2 emissions

375 169.0 451

USA Japan China Russia 270 126 1250 131.8 51.2 4.2 488 406 3

145 19.7 135

3676 260 281

6216 767 23

723 76 389

NR 80 370

NR 56 81

1254 121 872

1499 521 1771

301 0 278

548 NR 219

140 66 137

Note: NR: not reported. Source: Rothengatter (2003).

Table 5.3 shows that in per capita terms the volume of CO2 emissions in the USA is more than twice as high than in the EU15 (1771/270 versus 872/375). The per capita levels in Europe and Japan are comparable. Of particular importance is the position of China, and with it other developing countries with high economic growth such as India. The global development of greenhouse emissions will be strongly determined by these countries if they continue to grow as at present. It is no coincidence that in their review of transport problems at the world level WCBSD (2001) makes a clear distinction between developed and developing countries. The developed countries are characterized by high income levels, high degrees of urbanization, low demographic growth, and high mobility levels that tend to display moderate growth. Developing countries experience much lower incomes, rapidly increasing degrees of urbanization, high demographic growth and much lower but often rapidly increasing mobility levels. Local environmental conditions and traﬃc safety in the developing countries are often a reason for concern. So the challenges to the transport sector are clearly considerably bigger in the developing countries than in the developed world. A related driving force of transport demand is lifestyle. This involves socioeconomic and demographic variables, possibly combined with certain data representing attitudes or preferences (Kitamura, 1988). In the developed countries the continuous increase in life expectancy and early retirement in combination with low birth rates has a bearing on travel demand

214

Yearbook of environmental and resource economics

patterns. On the other hand, reduction in infant mortality and better health care aﬀect the age distribution of the population and also has some bearing on travel demand. A ﬁnal remark is that much of the policy debate on sustainable transport is based on the implicit assumption that all economic sectors should contribute more or less proportionally to the achievement of decoupling goals. From an economic perspective this is diﬃcult to defend, however, since eﬃciency generally requires the removal of ﬁxed targets for speciﬁc sectors. The point is that such ﬁxed targets do not take into account diﬀerences in the marginal costs to achieve the reduction in emissions, and there are indications that the marginal costs in some transport sectors are high because of the high levels of taxation, making them less promising as a target for environmental policy than other sectors. For some important transport sectors such as road transport in Europe it is clear that taxes on energy use are already high, higher than on energy use in other sectors. The share of production and distribution costs in the price paid by the consumer of gasoline is only about 35 per cent in several European countries (Rietveld and van Woudenberg, 2005). The application of tradable permits across all sectors and countries would reveal where the best opportunities exist for reduction of energy use (Kling, 1994; for an application at the city level, see Goddard, 1997). A striking feature of the transport sector is the large imbalance in the treatment of various transport modes. While road transport is already subject to considerable taxes in most countries, aviation, rail and inland navigation (barges) are confronted with much lower taxes on energy, or do not even pay any energy-related tax. This obviously also leads to distortions, implying that the notion of a level playing ﬁeld within the transport sector is a chimera in many countries. Given the diﬀerences in the initial situation of treatment within the transport sector, a diﬀerentiated treatment of transport actors would be called for, where in particular the aviation sector would deserve attention given its relatively high growth rate. I conclude that while aiming for sustainable transport one should be aware of risks of overshooting since it is cheaper to harvest the sustainability gains in other sectors. This obviously depends on the current level of taxes, and tax structures tend to vary among countries; for example, fuel levels are considerably higher in the EU than in the USA. The risk of overshooting is most prominent in the case of greenhouse gas emissions operating at the global level for those subsectors where taxes are already high, or standards are already very strict (Vermeend and Vaart, 1998; Dijkstra and Janse, 2001). For the environmental themes occurring at the local level (such as noise and CO), this overshooting risk is much less probable since here the transport sector’s contribution as a main source of the problem is evident.

Transport and the environment

3.

215

ENVIRONMENTAL EFFECTS OF TRANSPORT

Environmental eﬀects of transport not only relate to the movement of vehicles, but also to standing vehicles, the existence of infrastructure and the broader life-cycle costs, leading to a wide range of eﬀects (see Table 5.4). Some of the elements in Table 5.4 relate to marginal costs of transport, whereas others have a ﬁxed-cost character. The eﬀects of moving vehicles imply positive marginal external costs per vehicle-km. This also holds for the upstream processes via the impacts of energy production. Most of the other eﬀects have the character of ﬁxed costs implying zero marginal external costs, at least in the short run. There are intermediate cases, however. For example, part of infrastructure maintenance is use dependent, and also the decision to scrap a car depends on the intensity of its former use. The quality of a car degrades by two processes: one is time dependent, the other is use dependent. The relative weight of the two depends on external features such as temperature and air humidity. The marginal contribution of driving a kilometre to the vehicle scrapping costs equals the probability that an extra kilometre driven leads to the scrapping of the vehicle times the environmental costs involved in scrapping. The mechanism behind this is that use of cars leads to accident risks and also to an increase of maintenance Table 5.4

Environmental eﬀects of transport

Types of eﬀects

Examples

Eﬀects of moving vehicles

Noise Emissions of pollutants Climate change Disturbance to nature People killed by transport Animals killed by transport

Eﬀects of standing vehicles

Loss of quality of public space in cities and rural areas

Eﬀects of existence of infrastructure

Severance/barrier eﬀects Landscape deterioration Ground sealing Habitat reductions

Eﬀects of upstream and downstream processes (life-cycle approach)

Moving vehicles: impacts of energy production Vehicle production, scrapping: air pollution Infrastructure maintenance: surface renewal

216

Yearbook of environmental and resource economics

and repair activities. Since part of the vehicles are ﬁnally scrapped because of ageing rather than use-dependent costs, part of the total construction costs remain ﬁxed. The distinction between marginal and average costs is important in view of the usual policy recommendation to apply marginal cost pricing, although this recommendation did not remain unchallenged (see for a debate Rothengatter, 2003a and Nash, 2003). The main points in the current debate concern equity issues – marginal cost pricing may imply that consumers of transport services do not pay their way – and costs coverage problems when marginal costs are lower than average costs. Also there may be uncertainty in the measurement of marginal costs, and in the cases of distorted markets, simple marginal cost pricing no longer guarantees eﬃcient outcomes. This would call for second-best pricing strategies (cf. Verhoef, 1996). After this discussion of ﬁxed and marginal environmental costs I now discuss some of the environmental eﬀects in more detail. Noise Noise nuisance may have a large impact on human well-being. It leads to temporary disruptions of activities such as listening and conversation. In addition it may also lead to more permanent problems such as stress and physical illness. Some 50 per cent of Dutch residents experience noise problems (Brons et al., 2003). A frequently used indicator of noise nuisance is Leq, the equivalent noise level deﬁned as the average sound energy measured in decibels for a certain observer. Its use is explained for example in Schipper (1999), Gillen (2003) and Dhingra et al. (2003). It depends among other things on the number of vehicles, their speed, distance between observer and infrastructure, meteorological conditions, shielding factors and absorption by land cover. Of special importance is that some summary measures apply weighting per part of the day, with weights that may be very high during the night, a point applied in particular in the case of airports (Gillen, 2003). Quinet (2003, p. 368) emphasizes that the exact levels of noise nuisance depend strongly on speciﬁc local details so that application of standard models for noise nuisance may lead to misleading results: ‘it is the technical and not the economic stage which is the weak link in the evaluation process’. This closely ties in with the valuation chain between transport activity and valuation of disturbance presented in section 1 (Figure 5.l) where both physical and valuation aspects are incorporated. Compared with other types of environmental eﬀects, noise is very local. Hence space-speciﬁc measures are often used to limit the damage. Some examples are given for road/rail and airports in Table 5.5.

217

Transport and the environment

Table 5.5

Examples of noise management policies in transport

Noise strategy

Roads/rail

Airports

Infrastructure (dis)investments

Location of line infrastructure ﬁne-tuned with location of existing residential areas

Relocation of airport near city to location further away

Infrastructure design

Underground alternatives Noise shields

Noise shields

Infrastructure use

Limitation of traﬃc volumes in urban areas

Airport curfews Limitations on use of speciﬁc runways

Vehicle design Vehicle-use restrictions

Technological standards Speed limits

Technological standards Speciﬁc routing and ﬂight level instructions for landing and departing aircraft

Physical planning

Restrictions on residential construction in zones near line infrastructure

Restrictions on new residential construction near airports

Air Pollution Air pollution in transport concerns a broad group of pollutants such as CO, NOX, O3, particulate matter (PM), lead (Pb) and SO2. A fairly complete list of emissions and their eﬀects is given by Litman (2002); see Table 5.6. For some of these pollutants, transport is the major source. For example EPA (2001) estimates that in the USA 79 per cent of all CO emissions areas are due to transport (see Table 5.7). In urban areas this may be as high as 95 per cent. For a range of pollutants ambient air quality standards, measured in terms of maximum allowable concentrations per time unit have been introduced (Holmen and Niemeier, 2003). This makes sense since for some pollutants distance decay is strong in the chain from emissions to concentration/deposition (see Figure 5.1) so that even within urban areas substantial variations in concentrations can be observed: high close to main roads, and lower elsewhere (Van Wee, 2005). When these thresholds are passed at certain places, special measures have to be taken (see section 7). In densely populated regions these ambient standards tend to become a major factor hampering further transport growth and stimulating the introduction of transport-calming measures such as the reduction of speed on expressways passing through metropolitan areas (Holmen and Niemeier, 2003).

218

Dust particles created by vehicle movement

Various compounds. Some are toxic; all contribute to ozone formation

Unburned fuel. Forms ozone

A variety of organic compounds that form aerosols

VOCs that are toxic and carcinogenic

Major urban air pollution problem resulting from NOX and VOCs combined in sunlight

Road dust

Nitrogen oxides (NOX)

Hydrocarbons (HCs)

Volatile organic compounds (VOCs)

Toxics (e.g. benzene)

Ozone (O3)

NOX and VOC

Fuel production and internal combustion engines

Fuel production and internal combustion engines

Fuel production and internal combustion engines

Internal combustion engines

Vehicle use

Human health, plants, aesthetics

Human health, risks

Human health, ozone precursor

Human health, ozone precursor

Human health, ozone precursor

Regional

Very local

Local and regional

Regional

Regional

Human health, aesthetics Local

Human health, aesthetics Local and regional

Diesel engines and other sources

Very local

Scale

Particles consisting of fuel and carbon

Harmful eﬀects

Fine participates (PM10, PM2,5)

Source Human health, climate change

Description

Emission

Carbon monoxide A toxic gas which undermines the Internal combustion (CO) ability of blood to carry oxygen engines

Vehicle pollution emissions

Table 5.6

219

A gas with signiﬁcant greenhouse gas properties

Durable chemicals formerly widely used, now with use restrictions owing to environmental risks

Methane (CH4)

Chloroﬂuorocarbons (CFCs)

Litman (2002).

By-product of combustion

Carbon dioxide (CO2)

Source:

Lung irritant; causes acid rain

Sulphur oxides (SOX)

Climate change

Climate change

Human health risks, acid rain

Vehicles (especially those Ozone depletion with older air conditioner units)

Fuel production and internal combustion engines

Fuel production and internal combustion engines

Diesel engines

Global

Global

Global

Regional

220

Yearbook of environmental and resource economics

Table 5.7 Contribution of mobile versus non-mobile sources to pollution, USA (in %)

PM2,5 NOX CO

On road, mobile

Non-road mobile

Non-mobile

10 34 53

19 22 26

71 44 21

Source: Environmental Protection Agency (2001).

CO2 As opposed to the other materials mentioned thus far, carbon dioxide does not have direct adverse health eﬀects. However, it is considered to be one of the main sources of global warming and hence has received much policy attention. An important example is the Kyoto protocol in which targets have been formulated for CO2 reductions for each country. It appears that where policies aiming at the reduction of speciﬁc pollutants such as lead or NOX have been clearly eﬀective during recent decades, it has not been so easy to reduce CO2 emissions in transport. What has already been outlined in Table 5.2 for Europe also holds for the world as a whole: the share of transport in world carbon emissions is about 25 per cent, and this share tends to increase. Prospects to improve carbon dioxide emissions by the use of alternative fuels will be discussed in section 4. Traﬃc Safety Traﬃc safety as such is not included in this review of transport and the environment. However, it deserves to be mentioned since safety issues and environmental eﬀects of transport are linked. One of the links is that environmental policies usually aﬀect safety levels – intended or unintended. In most policy domains the policies will be mutually beneﬁcial. For example, speed limits to reduce emissions will also improve homogeneity of speeds and reduce seriousness of accidents, thus having positive safety eﬀects (see Rietveld et al., 1998 for a social cost–beneﬁt analysis of speed limits where safety and environmental eﬀects are treated in an integrated way). There are exceptions, however: policies to reduce emissions by making vehicles lighter may make them more vulnerable at collisions. There is also another link between traﬃc accidents and the environment: traﬃc accidents will have eﬀects on emissions when they lead to disturbances of the normal traﬃc

Transport and the environment

221

ﬂow. Thus incident management, primarily motivated by the desire to help traﬃc victims and to reduce the nuisance of other road users, may also have positive environmental eﬀects. Hazardous Materials Another aspect of transport and the environment that deserves brief mention concerns the transport of hazardous materials such as explosives and radioactives (Dennis, 1996). Transport of hazardous materials is an unavoidable consequence of presently used production technologies. Similar to noise, the eﬀects tend to be local. The development of policies to address hazardous materials is strongly helped by quantitative risk assessment. However, this tool has problems with addressing relevant themes such as the issue of diﬀerences among the public in the perception of diﬀerent types of risks, and perception errors (Waters, 2003). Severance Transport links are built to facilitate certain transport ﬂows, but at the same time they may function as barriers for perpendicular transport ﬂows. In an urban setting large roads with heavy traﬃc will discourage crossing nonmotorized transport ﬂows (Handy, 2003). The same holds true for expressways when the number of tunnels and bridges is insuﬃcient. Expressways, railway lines and canals may also function as barriers that reduce the habitat of species and thus may aﬀect biodiversity due to ecosystem fragmentation. Thus severance has a meaning in both rural and urban contexts. This subject of severance is among the less-researched themes in the ﬁeld of transport and the environment.

4. COMPARISONS OF ENVIRONMENTAL PERFORMANCE BETWEEN TRANSPORT MODES One of the policy options with respect to transport and the environment is to achieve a change in modal choice towards environmentally benign transport modes. For this purpose it is important to be able to compare modal alternatives in order to ﬁnd out which shifts are beneﬁcial for the environment. An example of such a comparison is contained in Tables 5.8a and 5.8b. These tables show that public transport performs better than the car in terms of average CO2 emissions. The diﬀerence usually ranges between factors of 2 and 5. In the case of NOX a similar pattern emerges, although here the bus is dominated by the car.

222

Yearbook of environmental and resource economics

Table 5.8a Average emissions of CO2 and NOX per traveller-km per transport mode; urban transport, 2000, the Netherlands (gram per traveller-km)

CO2 NOX

Car

Bus

Tram

Metro

200 0.65

110 0.85

65 0.15

60 0.15

Source: Essen et al. (2003).

Table 5.8b Average emissions of CO2 and NOX per traveller-km per transport mode; medium-distance transport, 2000, the Netherlands (gram per traveller-km)

CO2 NOX

Car

Bus

Local train (electric)

Intercity

190 0.55

85 0.65

65 0.2

35 0.1

Source: Essen et al. (2003).

It is important to underline that a number of issues are involved in such intermodal comparisons of transport modes that may lead to biased perceptions on which transport modes are ‘environmentally benign’ and hence would deserve policy support (see for example Stanley and Watkiss, 2003). A ﬁrst, and obvious, observation is that since technological standards are not ﬁxed, emission coeﬃcients will change in the course of time. This may imply that where in the past a certain transport mode is supposed to have superior environmental performance, it may well lose this advantage when the pace of technological development in other modes is faster. An example is inland navigation (barges), which has a number of advantages, but technological progress is slower than for road transport. A second, equally obvious point is that there is a multitude of emission types, so that it is not easy after all to determine which of two modes is performing better. This would call for an overall index of environmental performance, a subject closely related to the environmental cost theme addressed in section 6. Third, in freight transport, discussion mostly takes place in terms of emission per tonne. However, this appears to depend strongly on the type of good concerned, such as dry bulk, wet bulk, containerized goods, etc.

Transport and the environment

223

(RIVM, 2000). Thus the discussion has to be reﬁned by considering various freight categories separately. An aggregate comparison of emissions per tonne-km transported by road and by rail leads to misleading results since the mixture of these freight types across transport modes diﬀers strongly. Fourth, the ﬁgures obtained depend of course on load factors and empty vehicle movements.These are factors that must not be assumed to be given when comparing transport modes, because a shift from one mode to the other may well lead to opportunities to improve the average load factors for a certain mode. A more general way of addressing this point is that measures of environmental performance of transport modes are based on average costs, whereas marginal costs would make more sense. This issue will be spelled out in more detail below. Fifth, a similar reasoning applies to average vehicle size. There are substantial economies of scale in transport from an environmental perspective (Rovers, 1999). Adding a load unit to a push tug has an eﬀect on energy use that is clearly less than proportional. Shifts in volumes transported per transport mode may stimulate the size of vehicles and hence are beneﬁcial for the mode that gains in its modal share, whereas the other mode will lose. Sixth, it is of course very unfair only to compare the environmental performance of transport modes without taking into account other aspects. An important point is obviously that the penetration rates and network structures of transport modes are not equal. This means that some transport modes imply larger detours than other transport modes. Most trips in road transport imply rather low detour factors – deﬁned as the ratio between the actual number of kilometres travelled and the distance as the crow ﬂies – no higher than 1.20. But hub-and-spoke network structures in aviation may lead to larger detours. And short sea shipping between the Baltic and the Mediterranean may be three times longer than overland transport. Along similar lines, the lack of penetration of rail or inland waterways will usually lead to parts of transport movements that take place via the road. This will add to the environmental burden of the total trip. A fundamental limitation of these comparisons is further that they assume origin and destination of the trip as given. Other demand responses may be ignored. This may lead to an overly simple view. For example, although it makes sense to compare the environmental performance of a bicycle trip to a supermarket and a car trip to the same supermarket, it may well be that once a consumer decides to take the car, he chooses another shop, located at a greater distance, but oﬀering a more attractive choice. Similarly, it makes sense to compare a rail trip from Brussels to Vienna as

224

Yearbook of environmental and resource economics

the destination of a holiday trip with a ﬂight for the same destination. But one should not overlook the point that once the plane is considered as an alternative, other destinations may also be considered, located much further away, such as Athens or Lisbon. A clear illustration of this phenomenon is provided in the tourism market in Europe, where the emergence of low-cost carriers had a negative eﬀect on long-distance rail travel demand to places that are not served by these carriers. A further observation is that the demand side obviously considers in the ﬁrst place the factors directly aﬀecting proﬁt or welfare of the relevant actors. Hence price and speed of the transport modes are obvious determinants of preferences for particular transport modes. Information on environmental parameters of transport modes has to be combined with the other price and quality features before meaningful policy advice can be given. For freight transport the overall tendency is that the cross-price elasticity is rather low (Graham and Glaister, 2004). After this general discussion of intermodal comparisons of environmental performance, I discuss two aspects in more detail that relate to behavioural aspects. The ﬁrst considers the behaviour of suppliers of transport services, the other the behaviour of the demand side. Starting with the supply side, one of the reasons why environmental performance in transport is much lower than technically feasible is that vehicles tend to have load factors that are far below 100 per cent. For example, one out of three containers moved in the world is empty, a main reason being that international trade ﬂows are unbalanced. Occupation rates are also low in passenger transport. Private vehicles have on average 1.5 persons on board, implying a load factor of almost 40 per cent, and public transport companies in most countries have diﬃculty in achieving load factors above 30–40 per cent. Aviation performs relatively well, with average load factors around 75 per cent owing to the extensive use of yield management techniques. The low occupation rates in public transport may point at X-ineﬃciencies, but are also the result of speciﬁc features of the industry such as temporal variations in demand leading to peaks during the morning and afternoon. These peaks, in combination with economies of vehicle size, lead to the purchase of large vehicles that do not match levels of demand during the rest of the day. The large discrepancy of occupation rates during the peak (high, though usually only in one direction of the transport services) and the rest of the day (low in both directions) in combination with the public service obligation to provide a minimum level of service during the rest of the day makes the system ineﬃcient. This situation leads to environmental costs that depend strongly on the peak versus the oﬀ-peak part of the day (see Table 5.9).

225

Transport and the environment

Table 5.9 Average versus marginal environmental costs of public transport during the peak and the oﬀ-peak period

Peak demand Oﬀ-peak demand

Average costs

Marginal costs

Low High

High Low

During the peak period public transport has a low average cost owing to the high load factor, but marginal costs are high since expansion of demand calls for a proportional increase in capacity. In the oﬀ-peak period the situation is the opposite. The average costs are high, but the marginal costs are very low since additional travellers can well be absorbed with the given capacity. The square root principle formulated by Mohring (1976) for a welfare-maximizing transport ﬁrm during the oﬀ-peak period implies a supply response in terms of frequency of operations with an elasticity of 0.5 when demand increases. The background of this formula is the tradeoﬀ between the costs of operations that are proportional to frequency and scheduling costs that are inversely proportional to frequency. Empirical research for the Netherlands suggests an elasticity that is somewhat lower (0.35, as estimated in Rietveld, 2002). Since environmental costs are proportional to frequency of services, this result for the costs of operations as a function of travel demand also holds for the environmental costs. We conclude that the use of average load factors of public transport modes may lead to a strongly misleading view on the environmental costs and beneﬁts of modal shift. Public transport is an attractive mode to shift traﬃc to during the oﬀ-peak period, not during the peak. The second behavioural aspect concerns the demand side. It is often argued that governments should stimulate a shift towards environmentally benign transport modes. This can be done by measures such as investing in faster public transport or by subsidizing public transport. Making public transport faster means that modal shift will occur, but also other changes will take place. As outlined in Van Wee and Rietveld (2003), the major impact of an overall increase in railway speeds will be an increase in railway use, but the shift from the car to the train is limited. The increase in travel by rail consists of only 20 per cent of people who shifted from the car. These are people who already travel by train but now make use of it more intensively, and people who make trips they did not make before. For the environment the speed improvement is probably disadvantageous since the additional travellers imply higher transport volumes and hence higher emissions. In addition, the speed increase will also have an unfavourable eﬀect on the environmental performance per km (Peeters, 1997). A similar

226

Yearbook of environmental and resource economics

result is found for making public transport cheaper. Here also the modal shift from car to train is limited in size and the major behavioural eﬀect consists of new travellers. The only situation where public transport subsidies or even free public transport for certain target groups may have positive eﬀects on the environment is when the marginal costs of public transport are very low, and even after the subsidy the load factor is probably still low.

5. TECHNOLOGICAL DEVELOPMENTS FOR TRANSPORT Similarly to the approach adopted in section 2, a simple equation of total transport emissions of a certain mode reads as follows: Emissions(Emissions/Vehicle-km) (Vehicle-km/Vehicle) (Vehicles/Household) (Households/Population) (Population) This formula means that the long-run development of emissions is driven by demographic phenomena such as population growth and household formation, vehicle ownership (the average number of vehicles per household), intensity of use and technological features. Behind these phenomena are factors such as pricing policies, income change, physical planning, etc. At the world level the population is still increasing, and in countries with a stagnant population household size is still decreasing, putting an upward pressure on car ownership. The number of cars per household is increasing all over the world. In a country such as the USA that has already reached high car ownership per household it is still increasing, though at a more modest rate than most other countries (Bunch, 2000). In a country such as the Netherlands the number of cars equals the number of households, where about 20 per cent of all households do not own cars, about 60 per cent own one car and the remaining 20 per cent own two or more cars. The annual number of kilometres driven per vehicle has been remarkably stable in many countries during recent decades, even though patterns of vehicle ownership have changed markedly. This seems to be a subject that has received little attention in the research literature. It may be just a coincidental outcome of several countervailing forces, but it remains surprising that it is relatively stable in so many diﬀerent countries (IRU, 2000). In the present section I will focus on the ﬁrst element of the chain implied by the above equation: the development of emission technology. What are the prospects for a strong decline of emissions per vehicle-km owing to technological change? The potential for improvements in fuel eﬃciency is

227

Transport and the environment

Table 5.10

Life-cycle greenhouse gas emissions for alternative fuels

Fuel

Life-cycle greenhouse gas emission (g/km CO2 equivalent)

Gasoline Reformulated gasoline Diesel Liqueﬁed petroleum gas Compressed natural gas Methanol from coal Methanol from natural gas Methanol from wood Ethanol from sugar cane Ethanol from corn Ethanol from wood Liquid hydrogen

222–282 222–283 173–266 180–203 164–253 424–426 250–252 65–81 70–123 90–263 65–81 29–88

Source: Intergovernmental Panel on Climate Change (1996).

given in Table 5.10, which shows that substantial gains can be obtained in the sphere of CO2 emissions if alternative fuels are adopted. A closer look at vehicle technology for road transport reveals that during recent decades technology has changed considerably, but not in a revolutionary way. The internal combustion engine has been substantially improved and this has led to a potential increase in energy eﬃciency of cars, which was, however, oﬀ-set by the introduction of larger and heavier vehicles (see for example Brink and Van Wee, 2001). This is an example of a rebound eﬀect mentioned earlier. The penetration of diesel in various parts of the world varies, Europe having high shares in passenger transport, owing to a favourable ﬁscal treatment. The quality of diesel engines is increasing, which makes them gradually more popular. This is favourable for carbon dioxide emissions since diesel has a higher fuel eﬃciency, but for other pollutants the traditional gasoline engine, also known as the Otto motor, is doing better. Alternative fuels would improve the carbon dioxide performance of the internal combustion engine. This indeed has been implemented at a largescale basis in Brazil, where ethanol has become an important fuel for road transport. However, from a cost perspective this is not (yet) attractive, implying the need to subsidize ethanol in road transport. A less spectacular route has been followed during the 1990s by introducing mixes of fossil and biofuels in countries such as Brazil, South Africa and the USA, sometimes called ‘gasohol’. This may well be an eﬃcient way of dealing with

228

Yearbook of environmental and resource economics

waste produced in agriculture and forestry, but biomass from dedicated energy crops is expected to remain too expensive as long as current oil prices do not increase dramatically (Johansson, 1999). Electric vehicles were developed as early as the nineteenth century, but their scale of operation has been limited thus far. Their obvious advantages are that they are comfortable, have zero local emission, and produce little noise. A major disadvantage is that electricity has to be stored and that batteries are heavy and expensive, which implies that battery electrical vehicles will probably be conﬁned to niche markets for vehicles to be used only for short-distance trips. A more recent development concerns the hybrid electrical vehicle (Sperling, 2003). This combines an ordinary, but downsized combustion engine with a small pack of batteries. The combination is used in situations where the engine’s capacity is insuﬃcient, as in the case of acceleration and hill climbing. Electricity is generated by the engine. The combination of the two components leads to a substantial improvement of fuel economy of 25–50 per cent. Hybrid cars are more expensive to produce and some researchers conclude that it is not cost-eﬀective in improving the fuel economy or lowering emissions (Lave and MacLean, 2002) since there are cheaper alternatives by further improving the fuel economy and emissions of traditional combustion engines. Nevertheless, several countries have used ﬁscal tools to make the hybrid car an attractive alternative to the consumer and this seems to pave the way towards gradual market adoption of the hybrid car. This development might well mean that the electric car will not make it in the future, but that the less revolutionary hybrid car will be a next step. It is interesting to note the correspondence with the mixed fuel option mentioned above. Both are examples of gradual changes where the given technology of the combustion engine is taken as the point of departure. And in both cases the mixed option seems to be the most favourable one, not only in the short run, but possibly also in the medium and long run. This result is in line with the standard theories on innovation and technological development, where returns on learning by doing and network externalities lead to irreversibility and lock-in situations where it is very hard to achieve a breakthrough with new technologies (David, 1985; Arthur, 1989). An example of such a new technology would be the introduction of zero emission vehicles based on the fuel cell. This technology is based on hydrogen as a fuel input. It is potentially superior to the combustion engine since it would reduce energy use and improve environmental performance (Geerlings, 1998), but the above-mentioned defence mechanisms of conventional technology make it hard to achieve a breakthrough. In addition, there are two major bottlenecks hindering its introduction relating to the costs of producing it and the problem of storing H2. Given its light weight,

Transport and the environment

229

hydrogen is very voluminous, which makes it diﬃcult to store, since compression leads to safety risks related to explosion.

6.

VALUATION AND COST–BENEFIT ANALYSIS

The valuation of environmental eﬀects of transport has attracted much attention during recent decades. The value of a change in environmental quality is deﬁned as the amount of money an individual would have to pay (or the amount of compensation needed) to keep him equally well oﬀ. Given the wide variety of environmental eﬀects involved it is no surprise that there is not a unique method being used. Table 5.11 gives a concise review of methods available in the ﬁeld of transport and the environment and their strengths and weaknesses. Broader reviews can be found among others in Freeman (1993) and Perman et al. (2003). The implicit prices method refers to the use of actual travel data to estimate the value of time in transport, a non-market good that has a large impact on travel behaviour (see for example Small, 1992 and Wardman, 2002). This approach is closely related to the household production approach where time use is explicitly modelled as an input in household production and consumption processes (Jara Diaz, 2003). This value of time is needed for the application of the travel cost method, a point of discussion being to what extent usually adopted values apply in the case of recreational travel (Perman et al., 2003). Hedonic pricing approaches can be used in the case of local eﬀects such as noise and the natural quality of the immediate vicinity of housing. Examples can be found in Schipper (1999) for noise and Kruk (2005) for wetlands. For broader environmental eﬀects of transport, and for transport safety, stated preference methods have to be used (see for example Hoevenagel, 1994, and Blaeij et al., 2003). A more limited domain is covered by the cost of illness method that directly measures the value of resources needed in the medical sector as a consequence of illness and accidents plus the value of foregone production as a consequence of illness (Evans, 2000). An outlier in the table is the use of shadow prices based on government-imposed standards for emissions and other environmental amenities as is the case with the valuation of CO2 emissions based on the Kyoto protocol. The main diﬀerence is that the other approaches are based on the notion of consumer sovereignty and thus express consumer preferences, whereas public sector standards depend on political processes. The reason that such standards are nevertheless used in valuation is that consumer-preference-based values are diﬃcult to obtain for some of the global environmental eﬀects of transport. We discuss the valuation of some environmental eﬀects of transport in more detail. The pollutants emitted by transport lead to various health

230

Yearbook of environmental and resource economics

Table 5.11 Comparison of valuation methods used in the ﬁeld of transport and the environment Method

Information needed

Reliability

Domain of application

Implicit prices; household production approach

Observed choices

Adequate

Travel time

Travel cost

Observed travel choices

Depends on assumptions on value of travel time for recreational purposes

Use value of nature and landscape

Hedonic prices

Market outcomes

Depends on assumption of well-functioning housing market

Noise, external safety, quality of the immediate vicinity of housing

Stated preference, contingent valuation

Hypothetical choices

Various biases may occur

Travel time, noise, use value of nature and landscape, emissions, safety, non-use values

Cost of illness approach

Data on medical costs; value of non-worktime

Adequate

Safety, health problems due to emissions; but not the value of pain and suﬀering

Shadow price, based on standards set by public sector

Prevention costs

Adequate

Non-use values of nature and landscape, emissions

problems, which may eventually lead to death. McCubbin and Delucchi (1999, 2003) give a review of valuation approaches around the health eﬀects of motor vehicle air pollution. The typical approach is to estimate the value of risk reduction by means of stated preference approaches. For example, people may be willing to pay 40 euros for a reduction of a mortal risk from 5 in 100 000 to 4 in 100 000 per year. If 100 000 persons had the same valuation, their joint willingness to pay would be 4 million euros for an expected

Transport and the environment

231

risk reduction of one victim. The resulting value of a statistical life as 4 million euros is then used in the valuation of transport-induced health eﬀects. Of special relevance is that the expected duration of a life does matter. This is solved by Moore and Viscusi (1988) by taking a certain average remaining life expectancy people would have if they did not become a victim. For example, if people had a remaining life expectancy of 20 years, the value of a statistical life-year would become 200 000 euros, assuming a zero discount rate. A positive discount rate would obviously lead to a higher value of a statistical life-year due to the discounting of future values. There are several problems connected to this approach, some of which have received ample attention in the literature; others have remained somewhat neglected. First, the approach is based on small probabilities, and respondents tend to misinterpret these. As indicated by Kahnemann and Tversky (1979), people tend to overestimate small risks. This means that much eﬀort has to be devoted to informing respondents in an understandable way on the actual risk levels they are exposed to. A second problem is that the above approach is based on expected value formulations. However, in risky situations such as this, respondents may well be risk averse, implying that they not only consider expected utility values of income-accident bundles, but also standard deviations. A simple example would be the wellknown Markowitz formula, where the ﬁnal valuation is the sum of expected utility E(U) and the standard deviation SD(U); see Markowitz (1987). The consequence would be that the estimate for the risk parameter in the utility function does not just reﬂect the ‘disutility of being dead’, but that it also captures risk aversion. Another problem is that it is not so obvious how to transform the value of statistical life into a value of a life-year. The point is that in estimations of the value of statistical life the valuation is certainly not proportional to the expected length of the remaining life (Blaeij, 2003). This suggests that the values of statistical life seem to consist of two components: a ﬁxed part related to being the victim of an accident or a certain disease, and a second part related to the expected length of the remaining life. This would imply that the ultimate health valuation costs should preferably be treated in a disaggregate manner: one via the expected number of people aﬀected, and the other via the expected number of life-years lost due to the health problems. A further point that is now well recognized in the literature is that the value of statistical life depends on the context in which it is estimated. For example Jones-Lee and Loomes (2003) compare the value of statistical life in road transport versus public transport and indicate that the latter is probably larger than the former, although they also warn against overestimating this gap. Important aspects that may lead to diﬀerent values concern the

232

Yearbook of environmental and resource economics

voluntariness of the activity. People may have a higher willingness to accept risks when they perceive the decision whether or not to take the risk as a free choice (for example car driving) compared with the possibility that they may get cancer due to air pollution. Also the mode of risk – accident or disease – and context appear to matter. Thus it is not possible simply to apply estimates of statistical life obtained in a certain context and apply these to another context. Cost–Beneﬁt Analysis and Compensation Cost-beneﬁt analysis (CBA) is an important tool in many countries to guide the selection of transport policy alternatives (Quinet and Vickerman, 2004). Valuation of environmental eﬀects is an important element in these. CBA is not without its problems, such as on what grounds to choose a particular discount rate and how to deal with risk. Further, applications of CBA usually take place within a certain limited spatial domain where a public body has its area of jurisdiction (for example country, state, region). Depending on the transport plan being evaluated, there may well be important spatial spill-overs that are not taken on board. These spill-overs may relate to environmental issues, but they may also concern positive eﬀects such as customers of a transport link being located outside the area of jurisdiction. An obvious problem with CBA is that there are environmental eﬀects for which estimates of their valuation are weak or absent. In particular, for the ‘soft’ aspects of environmental eﬀects such as habitat loss, landscape values, etc. no well-founded estimates are available. There are basically four ways to proceed in such a case; one may: 1 2 3 4

do additional research; carry out a sensitivity analysis based on the available data; carry out a meta-analysis based on studies in other countries and propose a value transfer on the basis of this; or accept the non-availability of the parameters needed and leave the ﬁnal decision to the policy-makers.

The actual choice depends strongly on the political culture in countries, but the use of meta-analytical approaches tends to be of increasing relevance (see for example Miller, 2000). There is another problem with CBA that also limits its usefulness in real-world decision-making. It concerns the problem of equity. The Hicks–Kaldor criterion on which CBA is based says that the most attractive alternative is the one where the sum of net beneﬁts across all persons

Transport and the environment

233

aﬀected is largest. When this sum is positive it means that the gains of the winners are large enough to compensate the losers. This seems to be an attractive property, but the problem is that this compensation has a hypothetical character, which of course brings little comfort to the losers. Actual compensation of losers would increase the social and political feasibility of otherwise infeasible transport policies. For example, if according to the Hicks–Kaldor principle the project could hypothetically compensate the losers, some persons may, in actual fact, be much worse oﬀ if actual transfers do not take place. One can imagine that depending on the political constellation this may not be acceptable. Depending on the degree of inequality aversion, one places more or less emphasis on distributional implications of policy prescriptions. Public and political acceptability of road pricing, for example, is generally seen as the greatest barrier to implementation (Jones, 1998). Actual compensation seems therefore an attractive alternative, as long as it is not at the cost of eﬃciency. Now suppose that rather than requiring hypothetical compensation, we augment the Hicks–Kaldor principle such that actual compensation is required to pass the Hicks–Kaldor test. This implies that a compensation package is now part of the project evaluation procedure. Suppose, further, that one is able to determine the compensating variations for all aﬀected agents correctly, then the question arises as to how compensation should take place. With a tax here, a compensation there, it seems that government can readily compensate, meanwhile trying to get a preferred income distribution. From the public ﬁnance literature, however, one knows that transfers, in general, are at the cost of eﬃciency (for a discussion see Rosen, 1992). Also, the excess burden associated with these transfers then would ask for new compensation schemes. Still, in such cases, it is in principle possible to evaluate projects and compensation schemes on the basis of welfare economic principles. In the sequel we brieﬂy discuss two types of compensation: ﬁnancial compensation and physical compensation. Financial Compensation Government could use tax revenues to compensate the losers. For example, residents may receive compensation for intrusion due to road construction activities. And property taxes may be reduced for dwellings that are aﬀected by traﬃc noise. However, in general, ﬁscal compensation creates a deadweight loss for society due to the distortionary character of most taxes, reducing the eﬃciency of a proposed scheme (cf. Rosen, 1992). Another potential problem is that compensation may discourage victims to take the necessary measures to reduce the noise costs.

234

Yearbook of environmental and resource economics

Physical Compensation Physical compensation is another operational method to compensate losers. Rather than compensation in monetary terms, one could consider compensation in physical form, mainly aimed at alleviating the (noise) eﬀects of the introduction of a new airport infrastructure, new roads and/or new railroads. Possible examples are installing double-glazed windows, and other sound-dampening materials and the construction of tunnels to reduce noise in vulnerable areas. Government could use the tax revenues of these harmful activities to pay for the amendments made to dwellings. For these cases, physical compensation is potentially eﬃcient and practically feasible as well. In the context of political decision-making processes, however, there is a clear risk that physical compensation becomes excessive (see also Flyvbjerg et al., 2003).

7.

TRANSPORT-RELATED POLICIES

There is a wide range of policies to address environmental problems related to transport. Table 5.12, partly based on Button and Rietveld (2000), gives an outline according to two major dimensions: ● ●

type of measure: market based versus command-and-control; point of impact: vehicle, fuel, infrastructure and traﬃc.

Market-based Incentives A look at Table 5.12 reveals that there are few market-based incentives that are directly linked to environmental intrusion of transport activities. A rare example of a direct incentive is the use of noise penalties for planes landing and taking oﬀ at certain airports. Calthrop and Proost (2003) show that the direct emission tax is the welfare-maximizing instrument giving the incentive to arrive at the right combination of emission technology and travel demand. There are probably three reasons why direct instruments are seldom used. First, there is a measurement problem with intrusion. Vehicles seldom have monitoring tools on board to measure it. Second, measurement would be costly and might be subject to fraud. And third, there are indirect tools available that may also do a rather good job in approximating the costs of intrusion. Examples of such indirect tools are given in column 2 of Table 5.12. An often-used point of impact is the vehicle. When vehicle ownership taxes vary according to the type of vehicle, this helps the adoption of environmentally

235

●

●

●

●

●

●

●

●

●

Traﬃc

Payment for noise intrusion: airports Kilometre charge for road use, diﬀerentiated according to

Emission fees

●

●

●

Congestion charges Parking charges Subsidies for less polluting modes

Compensation for intrusion at time of infrastructure construction

Diﬀerential fuel taxation High fuel taxes

Diﬀerential vehicle taxation Tax allowance for new vehicles Tradable car ownership permits

Indirect

Infrastructure

Fuel

Vehicle

Direct

Market

●

●

●

●

●

●

Physical restraint of traﬃc Designated routes

Noise shields

Fuel composition Phasing out of highpolluting fuels

Emission standards

Direct

●

●

●

●

●

●

●

●

●

●

Speed limits Restraint on vehicle use

Bus lanes Airport location Motorway expansion Railway construction

Fuel economy standards

Compulsory inspection and maintenance Mandatory use of lowpolluting vehicles Compulsory scrapping of old vehicles

Indirect

Command

Table 5.12 Examples of policy instruments for containing the environmental intrusion of transport

236

Other

Spatial structure

environmental standard of vehicle

Direct

Table 5.12 (continued) Indirect

Market Direct

●

●

●

●

●

●

Restrictions on settlement densities Location of new settlements Institutional change Education Stimulation of telecommuting Policies in non-transport ﬁelds

Indirect

Command

Transport and the environment

237

friendly vehicle types. For example, Alberini et al. (1995, 1996) suggest the possibility that vehicle ownership taxes that increase with age, based on their bad environmental performance, will stimulate scrapping of old vehicles. The vehicle ownership tax was used to speed up the introduction of the catalytic converter after 1990 and is now being used to speed up the adoption of hybrid vehicles. In countries where ﬁxed vehicle taxes are substantial, and even more so when there are substantial taxes on the purchase of new cars, this may easily lead to rapid adoption of environmentally friendly car types because the discounts are suﬃcient to oﬀ-set the price diﬀerence between ordinary cars and the new car types. This has been documented for example by Bergh et al. (2004) for the introduction of cars with a catalytic converter in the Netherlands. Calthrop and Proost (2003) show that for plausible values of parameters such a subsidy is inferior compared with the ﬁrst-best emission tax, but that the welfare loss is low. The second market-based indirect instrument relates to fuel taxes. In line with the formula given at the top of section 5, the elasticity of fuel demand with respect to the fuel price D is the sum of the elasticities of fuel eﬃciency per km FE, kilometres travelled per car KM number of and cars CAR with respect to the fuel price: D FE KM CAR As indicated by Johansson and Schipper (1997), the total elasticity D is about 0.7, and among the three components the largest part is FE with a value of about 0.4. The values for KM and CAR appear to be about 0.2 and 0.1, respectively. This means that the main eﬀect of fuel taxes is an improvement of the fuel eﬃciency of cars given its incentive to the drivers to buy more fuel-eﬃcient cars and hence its stimulus for car manufacturers to produce such cars. More recent work of Brons et al. (2005) on this issue tends to lead to somewhat lower values: 0.45 (0.20 0.125 0.125) (see also Espey, 1998, for a review). It should be noted that governments in countries with high taxes have substantial scope for diﬀerentiation, for example between gasoline, diesel and LPG, and also between variants of gasoline, such as variants with and without lead. The higher tax on leaded fuel has signiﬁcantly speeded up the reduction of lead emissions by gasoline cars in transport. For those pollutants that are closely linked to the fuel composition (for example CO2) the use of indirect market-based incentives can be a very eﬀective instrument to contain environmental intrusion. However, there are a number of remaining themes where such a direct link between fuel use and emissions does not hold. These tend to be the local emissions of noise and disturbance of the attractive nature of residential

238

Yearbook of environmental and resource economics

and natural areas by moving and parked vehicles. Here, ﬁnancial instruments are probably less eﬀective. For example, the current practice of congestion charges and parking charges is not very eﬀective in incorporating the external environmental costs. There may even be unintended negative eﬀects, as can be seen in the case of introduction of parking fees in order to reduce parking congestion. This may well lead to the situation that shortterm parking is stimulated at the detriment of long-term parking, leading to an increase in vehicle movements and implying that the environmental quality is lower in the end (for an analysis of the duration of parking see Calthrop, 2001). Similarly, congestion charges imposed on certain roads may induce certain road users to change route choice, with adverse eﬀects on emissions and traﬃc safety (Emmerink et al., 1995). A kilometre charge is another instrument to be considered. Calthrop and Proost (2003) show that an undiﬀerentiated kilometre charge does a bad job in arriving at optimal levels of travel demand and pollution. It does not stimulate the adoption of an environmentally friendly vehicle technology and it will unnecessarily reduce travel demand. However, if the kilometre charge were diﬀerentiated, one could get much closer to the ﬁrst-best optimum. A recent development in Europe is the introduction in countries such as Switzerland and Germany of a kilometre charge for expressway use by freight transport. These charges do not show diﬀerentiation according to environmental aspects, but the charging technology used can quite easily be adjusted to incorporate these environmental eﬀects and also expressway congestion. It may well be that this initiative will pave the way for more reﬁned kilometre charges, not only for trucks but also for passenger traﬃc. This would bring the situation much closer to the ﬁrst-best pricing approach. In the meantime, research on the potential eﬀects of a kilometre charge is a promising theme. One relevant dimension is the eﬀect on route choice. The point is that the present examples of kilometre charges focus on expressways, and hence stimulate truck drivers to choose other roads. This not only increases noise nuisance in urban areas, but also leads to higher traﬃc safety risks. The experiences with truck drivers making detours to avoid the expressway charges suggests that the values of time for freight transport that are usually employed in cost–beneﬁt analyses are too high. From a welfare economic perspective it is not a good idea to introduce a kilometre charge only for expressways. A full network introduction would certainly be better to avoid negative eﬀects on local environmental conditions. But here again there is a need for a thorough analysis of route choice behaviour from a welfare economic perspective on what the optimal prices would be in a network context, and how drivers would react to that (Verhoef, 1996).

Transport and the environment

239

Another theme is how to achieve the appropriate balance between ﬁxed and variable taxes on cars. Some countries in Europe and elsewhere have high vehicle ownership taxes. This leaves much scope for ‘variabilization’ of taxes, implying that the variable costs become both higher and more diﬀerentiated in order to achieve a large eﬀect on car use and its environmental consequences (Ubbels et al., 2002). The common thought is that increasing variable costs will reduce car use, leaving car ownership unaﬀected. However, it is not impossible that such a reorientation of car-related taxes – ﬁxed costs will decrease – will stimulate car ownership, and this will of course partly oﬀset the original negative eﬀects of car use, and in addition increase the negative eﬀects of standing vehicles presented in Table 5.4. This calls for an integrated analysis of car ownership and car use in the context of ﬁxed and variable taxes, a subject that is relatively unexplored (see for example, De Jong, 1990 and Bunch, 2000). Subsidies are another example of an indirect market-based instrument. As already indicated in section 4, a public transport subsidy to counter the negative environmental eﬀects of car use is a rather blunt tool to improve the environmental performance of the transport system, because of its unintended side eﬀects. The problem is that such a subsidy not only stimulates the modal shift between car and public transport, but also between non-motorized transport modes and public transport. In addition it leads to a substantial number of new travellers and travellers making more and longer public transport trips. Such a subsidy obviously has a second-best character. Motivations related to equity and social exclusion are probably more convincing to defend such subsidies. Standard Setting Technological change is in the long run a major factor inﬂuencing the environmental burden of transport. Emission standards have been used extensively as a policy tool in the past to address environmental problems (Stopher, 2003). For most types of pollutants, vehicles that are now entering the market are no doubt superior to vehicles that were introduced some 30 years ago. Although from an economic perspective ﬁrst-best pricing would be a superior instrument compared with emission standards, the diﬃculty of measuring the actual emissions of vehicle use make the imposition of standards an attractive second-best tool. A well-known example of standard setting is the CAFE (corporate average fuel economy) standard system in the USA. As outlined by Greene (1998), this system has worked well in stimulating car manufacturers to introduce more fuel-eﬃcient cars on the market at relatively low costs. This does not necessarily mean that the standards have been set at a welfare-maximizing level, but that the

240

Yearbook of environmental and resource economics

CAFE system was a cost-eﬀective way to improve the average fuel economy in the USA. Of some importance is the rebound eﬀect of improved fuel eﬃciency on the number of kilometres travelled. If fuel eﬃciency increases lead to reductions in the costs per km travelled, car users will start to drive longer distances. It is an empirical question to what extent these eﬀects are substantial. Greene (1998) indicates that the pertaining demand elasticities are relatively small – see the values mentioned above – so that the rebound eﬀect is not a serious problem. For an opposing view see Litman (2005). Physical Planning, Transport and the Environment According to the well-known study of Newman and Kenworthy (1989), land use and spatial structure have a substantial impact on travel patterns and energy use. Later research conﬁrms that there is indeed an impact, but that its size is limited. For example Handy (1992) ﬁnds that high-density environments lead to fewer trips and less car use, mixed land use has a slightly negative eﬀect on car use and decentralization of work leads to shorter commuting trips, but the share of the car increases. The last can be understood since decentralization of jobs following decentralization of workers brings jobs closer to workers, but it reduces the possibilities for public transport since such networks are usually best in bringing people to the central city. It appears that personal features are a stronger determinant of travel patterns than neighbourhood features. In addition, part of the neighbourhood impacts that are found on travel patterns occur via the interrelationship between personal features and neighourhood features (Meurs and Hajer, 2001). Obviously people are not distributed randomly in the city. After correcting for individual features there is still an impact of type of dwelling and type of neighbourhood on travel pattern, although the size of the eﬀect tends to be smaller. The conclusion is that physical planning has eﬀects on travel patterns, but that the eﬀect should not be exaggerated. It should be noted that transport costs obviously have a conducive role. The higher they are, the higher the eﬀectiveness of spatial strategies will be. Since in the Netherlands transport costs are already rather high, one may expect that in most other countries the conclusion on the eﬀectiveness of physical planning will not be more positive. A similar result was obtained in a simulation study of future land use patterns in the Netherlands (Hilbers et al., 2000). It was found that spatial policy aiming at residential construction within or immediately linked to existing cities leads to the smallest number of kilometres travelled of the residents concerned (about 43 km per person per day). On the other hand a strongly dispersed pattern of construction would lead to about 50 km per

Transport and the environment

241

person per day. The largest diﬀerence takes place in the share of public transport, being 10 per cent in the ﬁrst case and 6 per cent in the second. The conclusion is that eﬀects of diﬀerent spatial strategies on traﬃc volumes must not be exaggerated. In terms of total number of car kilometers, compact city strategies – implying the increase of urban densities by residential construction in and near existing urban areas – lead to traﬃc volumes that are about 20 per cent smaller compared with diﬀused spatial development. These ﬁgures hold for the new residents. Given the long lifetime of housing, the existing housing stock is large compared with the expansion during a 15-year period. Therefore the eﬀect on total mobility of the whole population is much smaller. Thus spatial policies will only bear fruit when they are pursued in a consistent way during a suﬃciently long period. In addition, their eﬀects will depend on the strategies carried out in other policy ﬁelds, in particular pricing of transport. Physical planning has a substantial impact on the formation of choice sets of people. For example, in the case of compact city strategies, it allows many residents to make short trips for various types of travel motives. However, adding nearby choice alternatives to a choice set is apparently not suﬃcient to arrive at substantially diﬀerent travel patterns as long as the costs of transport are not aﬀected. Not only in residential areas, but also near workplaces, physical planning can be used in order to reduce environmental nuisance. One of the ways to do this is to impose limitations on parking near workplaces in cities where there is public transport of suﬃcient quality. An example is the Dutch ABC policy during the 1990s in which national government implemented a rather stringent location policy with respect to ﬁrms at new locations. The aim of this policy was to stimulate modal shift away from the car to public transport by getting ‘the right ﬁrm in the right place’. The policy was based on a diﬀerentiation of locations: ‘A locations’ are those that are easily accessible by public transport and not by car (for example, locations close to intercity railway stations in city centres); here only a very small number of parking places per worker were allowed. ‘B locations’ are easily accessible via both public transport and car; here intermediate parking standards applied. ‘C locations’ consist of the remaining category, and here no parking constraints applied. The rationale behind these parking standards is that ﬁrms that attract many workers or customers per m2 are only allowed at A locations. At the other extreme, ﬁrms with small numbers of workers per m2, and that rely on freight transport, are only allowed at C locations. B locations are for intermediate cases. The ABC location policy was a rather stringent one, and it generated a number of tensions between the central government that formulated the standards and the local governments that had to apply them.

242

Yearbook of environmental and resource economics

This proved so problematic that the ﬁxed parking standards were ﬁnally abolished in 2001. Another source of tension arising from the ABC policy concerned the competitive position of new locations with restricted parking versus existing ones where parking standards were less strict. This led to a distortion of competition that discouraged the development of new sites. Nevertheless the ABC parking policy is often considered as an innovative approach that can also be adopted in other countries (Hall, 1998; WCBSD, 2001). The novel element is the integration of transport policy and physical planning in order to achieve a modal shift in an environmentally friendly direction. Alternative ways of aﬀecting employer-provided parking seem to be an interesting ﬁeld of policy to deal with urban transport externalities (Calthrop, 2001; Shoup, 1997). Policies with Respect to Non-motorized Transport Modes At the local level non-motorized transport modes play a large role. As illustrated by Figure 5.2, the low start-up costs of biking and walking make it attractive competitors against the car for short-distance trips. Start-up costs relate to time, money and eﬀorts to get a vehicle moving, and in the case of the car these involve elements such as parking fees and time needed to walk to the car, to ﬁnd a parking place, etc. The generalized costs in Figure 5.2 are deﬁned as the sum of the monetary costs, the money-metric time used for a trip and possibly other eﬀorts needed to make the trip measured in a money-metric way. Policies aiming at parking, relative speeds and detour factors will aﬀect the shares of these modes.

Generalized costs Car

Bicycle

Pedestrian Distance

Figure 5.2

Competition between transport modes in urban areas

243

Transport and the environment

Table 5.13 The shares of transport modes in passenger transport (in number of trips) in various countries (%) Country

Bicycle

Walking

Public transport

Car

Other

30 20 12 11 10 10 9 8 5 5 1 1

18 21 22 24 29 39 31 12 30 28 10 9

5 14 16 9 20 11 13 14 12 16 14 3

45 42 49 53 38 36 39 62 47 42 74 84

2 3 1 0 1 4 8 4 6 9 1 3

Netherlands Denmark Germany Finland Switzerland Sweden Austria England/Wales France Italy Canada USA Source: Pucher (1998).

As Table 5.13 shows, there are large variations between countries in terms of the shares of motorized versus non-motorized transport modes in passenger trips. In Europe there are several countries where this share reaches 40–50 per cent, whereas in the USA and Canada it is between 10 and 20 per cent. The reasons behind this relate partly to the supply of public transport: there is a substantial substitution between biking and local buses and trams. The factors that explain use of non-motorized transport modes have been analyzed among others by Pucher (1998), Rodriguez and Joo (2004), and Rietveld and Daniel (2004). Physical (rough terrain) and climate circumstances (wind, temperature) explain part of the variations in nonmotorized transport between cities. Also economic and cultural variables are relevant: income and ethnic background of city residents matters in their mode choice. For example, people with a foreign ethnic background appear to cycle less in the Netherlands. From a policy perspective it appears that local policies aiming at providing safe, direct and travel alternatives without delays for walking and biking, for example, by means of cycle lanes, help. In addition, policies aimed at making the car less attractive by means of parking restrictions and parking charges appear eﬀective. Institutional Change A last aspect that deserves attention is the theme of institutional change and regulatory reform in transport. In various parts of the world, institutional

244

Yearbook of environmental and resource economics

changes and regulatory reform have been high on the political agenda during the past two or three decades. The changes have led to a very strong increase in eﬃciency and consumer welfare in aviation (see Schipper, 1999), whereas regulatory reform in the railway sector had mixed eﬀects, ranging from positive in the USA to mainly negative in the EU (see Nash, 2003). In the ﬁeld of eliminating trade restrictions at borders we also observe substantial welfare increases. A point of overall concern is the sustainability aspect of these institutional changes and regulatory reforms. Where it is possible to give several convincing examples of positive eﬀects on eﬃciency, the environmental eﬀects seem to be much less favourable, and often negative. For example, it is evident that the very dynamic development of the aviation sector after deregulation had negative eﬀects on the environment (Schipper, 1999). Institutional changes and regulatory reform obviously had a strong focus on eﬃciency goals during the past several decades. An important theme is therefore how institutional change can be achieved that explicitly promotes sustainability. An example is the change needed to make the Kyoto protocol eﬀective. At the world level the governance structures to promote sustainability are still weak. As a special point of attention we mention the problem of how to deal with international transport in this context. Also the implementation of strategies of internalizing the full costs of transport in the aviation sector requires attention. Another example of an institution that has negative eﬀects is the Mannheim convention in Western Europe dating back to the nineteenth century. This convention, which is still in force, says that water transport on the river Rhine should not be tolled, and one of the policy conclusions has been that excise taxes on barges are not allowed. This obviously hampers the use of market-based incentives to improve the environmental performance of inland water transport. Still another example of institutional change concerns the way the various actors deﬁne their responsibilities in transport. The common approach is that households and ﬁrms make their transport decisions without reference to broader environmental goals and other externalities. This leads to a situation where in network analytical terms a user equilibrium is achieved that may be rather far away from the system optimum (Sheﬃ, 1985). It is the responsibility of governments at various levels to implement measures to bring user equilibrium closer to the system optimum. One of the problems governments meet is that the social acceptability of such policies is often limited, leaving the transport system performance at poor levels. It appears in the meantime that certain actors – voluntarily, or under some pressure of governments – adopt more active roles. For example, employers are stimulated to develop mobility management plans covering parking policies, carpooling and special public transport services in order to keep the workplaces

Transport and the environment

245

accessible for the workers. And the retail sector in congested cities ﬁnds that it has to take an active role in mobility matters in the ﬁeld of parking, reduced public transport fares and delivery of goods to the shoppers’ homes if it wants to stay in business. Telecommuting Telecommuting (or teleworking) has long been promoted as an instrument to reduce transport demand. The overall contribution of teleworking to a reduction of travel demand has been low, however. There are several reasons for this. First of all, teleworking is only applicable to a limited range of jobs. Second, employers and employees often perceive certain disadvantages (incentives to work, value of social contacts) when work takes place at home. Further, if telecommuting takes place various rebound eﬀects may occur. For example, when people really start to work at home some days each a week they may decide to relocate to a place that is further away from the job location, implying that their total commuting distance over the whole week does not decrease in proportion to the number of days worked at home; it might even increase. There is also a branch of literature on the so-called constant travel time budget (Golob et al., 1981; Mokhtarian and Salomon, 1999; Mokhtarian and Chen, 2004), which says that the average time per day spent on travelling by certain socioeconomic groups happens to be constant over a long period. The rationale behind this is that as there appear opportunities to reduce travel distances in one ﬁeld, there will be compensatory behaviour in other ﬁelds, implying that in the end the total time spent travelling remains largely unaﬀected. It does not rule out the possibility that modal changes take place, however, implying that the environmental burden changes, given the diﬀerent environmental performance of the various modes. Although the constant travel time budget hypothesis is nothing more than a crude empirical regularity, which is not guaranteed to hold true, it is nevertheless useful since it reminds us of the many ways in which people may adjust their behaviour as a response to certain technological opportunities. Time saved in commuting may well be used for other transport activities, and when it implies that the car is standing in front of the house when it not used for commuting, it may well be used for other purposes or by other household members. This does not mean to say that teleworking is useless or may be ignored. Detailed analysis of time use data reveals that most teleworkers work at home only part of the day, which means that they do not save on commuting distances (De Graaﬀ and Rietveld, 2004). Instead they may leave home later in the morning after teleworking for some time, thus reducing congestion in the morning peak. The reverse happens in the

246

Yearbook of environmental and resource economics

afternoon. Employees may leave at the normal time in the afternoon and continue working at home during the evening. In that case teleworking may well aggravate the seriousness of the afternoon traﬃc peak. In both cases the main eﬀects of teleworking are on congestion, the environmental eﬀects being less relevant. Non-transport Policy Measures Since transport is a derived demand, there is considerable scope for aﬀecting transport demand by measures in other policy ﬁelds within the reach of the public sector. For example, policies on the location and scale of public facilities have an immediate eﬀect on the distances their visitors and workers have to travel. This holds true for a wide range of facilities such as hospitals, schools, libraries, etc. Decisions on the optimal size of these facilities obviously have implications on their spatial density and hence on travel distances. Tendencies towards increases in the scale of such facilities thus have adverse eﬀects on transport demand.

8.

CONCLUSIONS

The theme of transport and the environment calls for an integrated approach where not only travel behaviour is considered but also the interrelationship between transport and economic activities, in particular those relating to land use. Another direction of integration concerns the need to adopt a life-cycle approach in order to address the overall impacts of transport on the environment. Trends in transport and its environmental eﬀects indicate that – although a good number of environmental problems have been solved in the meantime – some are very diﬃcult to solve. In particular energy use and CO2 emissions tend to grow at a pace comparable to that of GDP. Technological improvements in transport have been large, but this has not led to evident improvements of the energy intensity of transport due to countervailing developments. As a consequence the share of transport in world energy demand is increasing. The economic growth in Asia will lead to large additions in transport-related energy demand at the world level. The environmental eﬀects of transport are broader than those of moving vehicles plus the upstream and downstream processes implied by the lifecycle approach. In particular the eﬀects of parked vehicles on the quality of urban neighbourhood and the severance and barrier eﬀects of infrastructure seem to be under-researched aspects. These themes are also diﬃcult to address by means of technological measures.

Transport and the environment

247

A common approach to the environmental eﬀects of transport is to compare modes from the viewpoint of their environmental performance per km. This certainly leads to important insights but it has the risk of simple policy recommendations on modal shift towards environmentally friendly transport modes. As has been explained extensively in section 4, this may lead to disappointing or even counterproductive eﬀects, especially in public transport if the distinction between marginal and average costs is not considered. Also the distinction between peak and oﬀ-peak is essential in this respect. Technological developments can be both revolutionary and gradual. The present trends towards mixed fossil and biofuels and of the hybrid car seem to indicate that in the short and medium term the gradual paradigm will dominate. Cost–beneﬁt analysis is a common tool to evaluate policies aiming at enhancing the environmental performance of transport. Social acceptability due to the unequal distribution of costs and beneﬁts of measures often hampers the introduction of measures that have a positive net eﬀect. This calls for the explicit inclusion of compensatory measures in cost–beneﬁt analysis. Among the three compensatory schemes considered – ﬁscal, lumpsum and physical – there is no clear best candidate. The least that should be done is to incorporate the additional costs or ineﬃciencies due to compensation in cost–beneﬁt analysis in order to avoid the situation that ﬁrst decisions are taken based on a favourable cost–beneﬁt ratio, after which additional measures are taken that make the project relatively expensive. A direct emission tax is the ﬁrst-best instrument to address environmental eﬀects of transport. However, due to measurement and implementation problems, indirect emission taxes are usually preferred. The current policy trend seems to be moving towards a kilometre charge which would improve the possibility of direct emission taxes. Also the use of standard setting in countries such as the USA had large impacts on the environmental performance of transport. Physical planning and parking policies are important policy directions to address the environmental intrusion of transport at the local level and stimulate opportunities for non-motorized transport modes. Experiences with these instruments indicate that the eﬀects, although positive, are smaller than is often thought. New Research Themes At various places in this chapter we have mentioned topics that are interesting ﬁelds of research (and policy) for the future. We conclude this section by pointing out in some more detail a limited number of research themes that seem to be relevant for the coming period.

248

Yearbook of environmental and resource economics

An interesting theme for future research seems to be the relationship between climate change, transport and the environment. This is obviously a two-sided relationship. The theme of mitigation of climate change by reducing greenhouse gas emissions has already been addressed extensively in the research community. The other side of the coin is that changes in climate may have substantial impacts on our transport systems. For example, if climate change leads to weather conditions with stronger variations in wind speeds, this may well impact on availability of runway capacity of airports. Along similar lines, changes in climate may make road transport more vulnerable to delays and unreliability. Also the attractiveness of non-motorized transport modes will change when climate changes. This theme can be addressed by intensifying the research eﬀorts on the impact of weather on travel behaviour. Also in freight transport there may be substantial eﬀects. For example, inland navigation in Northwest Europe may be strongly aﬀected when the Alp region loses its glaciers, implying that the river Rhine will become a pure rain-fed river, making it much more vulnerable to water-level ﬂuctuations. Along similar lines, a theme that may be expected to play a large role in the future is the relationship between transport and health. Campaigns in several countries to stimulate non-motorized transport modes in order to improve local environmental conditions have often yielded negligible eﬀects, probably because of free-riding with respect to health eﬀects owing to a reduction of emissions. However, some health eﬀects take place in the private domain, for example when the use of non-motorized transport modes contributes to one’s individual health. We discussed some examples of the positive eﬀects of bicycle lanes on bicycle use and hence on health in the context of cost–beneﬁt analysis. This may well become an important issue in cost–beneﬁt analysis of infrastructure networks (see, for example, Saelensminde, 2004). An immediate countervailing eﬀect is the higher traﬃc safety risk per km experienced by pedestrians and cyclists. Thus there is need for an integrated analysis of health-related aspects. It makes sense to focus on transport-oriented measures in a study on transport and the environment. Nevertheless, it should be realized that unintended side eﬀects of other policy ﬁelds on transport may well be substantial. For example, the structure of income taxes will aﬀect the use of company cars, company-provided parking, allowances for commuting costs, all potentially having considerable impacts on commuting and its environmental eﬀects. An example in another policy ﬁeld concerns policies with respect to the scale and location of public facilities such as hospitals and educational institutions. Explicit consideration of the travel implications of such policies is often lacking. This is another example of the need to consider sustainable transport from a broader perspective than transport only.

Transport and the environment

249

ACKNOWLEDGEMENT I would like to thank Henk Folmer and Tom Tietenberg and an anonymous referee for helpful and constructive comments.

REFERENCES Alberini, A., W. Harrington and V. McConnell (1995), ‘Determinants of participation in accelerated vehicle-retirement programs’, The RAND Journal of Economics, 26 (1), 93–112. Alberini, A., W. Harrington and V. McConnell (1996), ‘Estimating an emissions supply function from accelerated vehicle retirement programs’, The Review of Economics and Statistics, 78 (2), 251–65. Arthur, W. Brian (1989), ‘Competing technologies, increasing returns and lock-in by historical events’, Economic Journal, Royal Economic Society, 99 (394), 116–31. Bartelings, H. (2003), ‘Municipal solid waste management problems: an applied general equilibrium analysis’, PhD thesis, Wageningen University. Bergh, J.C.J.M. van den, Leeuwen, E. van, F.H. Oosterhuis, P. Rietveld and E.T. Verhoef (2004), Social Learning by Doing in Innovations for Sustainable Transport: An Analysis of Successes and Failures, Amsterdam: Vrije Universiteit. Blaeij, A.T. de, R.J.G.M. Florax, P. Rietveld and E.T. Verhoef (2003), ‘The value of statistical life: a meta-analysis’, Accident Analysis and Prevention, 35 (6), 973–86. Brewer, G.D. (1991), Hydrogen Aircraft Technology, Boca Raton, FL: CRC Press. Brink, van den R.M.M. and B. van Wee (2001), ‘Why has car-ﬂeet speciﬁc fuel consumption not shown any decrease since 1990? Quantitative analysis of Dutch passenger car-ﬂeet speciﬁc fuel consumption’, Transportation Research D, 6 (2), 75–93. Brons, M., P. Nijkamp, E. Pels and P. Rietveld (2003), ‘Railroad noise: economic valuation and policy’, Transportation Research D, 8, 169–84. Brons, M., E. Pels, P. Rietveld and P. Nijkamp (2005), ‘Price elasticities of gasoline demand, fuel eﬃciency, car use and car ownership’, Vrije Universiteit, working paper, Amsterdam. Bunch, D.S. (2000), ‘Automobile demand and type choice’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport I: Transport Modeling, Amsterdam: Pergamon, pp. 463–79. Button, K. and P. Rietveld (2000), ‘A meta analysis of the impact of infrastructure policy and regional development’, in P. Nijkamp, J. Poot and H. Kohno (eds), Regional Cohesion and Competition in the Age of Globalisation, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 309–25. Calthrop, E. (2001), ‘Essays in urban transport economics’, Katholieke Universiteit Leuven, thesis no. 151. Calthrop, E. and S. Proost (2003), ‘Environment pricing in transport’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Amsterdam: Elsevier, pp. 529–46. David, P.A. (1985), ‘Clio and the economics of QWERTY’, American Economic Review, 75 (2), 332–7. Dennis, S.M. (1996), ‘Estimating risk costs per unit of exposure for hazardous materials transported by rail’, Logistics and Transportation Review, 32, 351–75.

250

Yearbook of environmental and resource economics

Dhingra, S.L., K.V.K. Rao and V.M. Tom (2003), ‘Environmental assessment for sustainable transport’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Amsterdam: Elsevier, pp. 309–29. Dijkstra, W. and P. Janse (2001), ‘Eﬀects of green taxes in transport’, research paper prepared by the CE Institute, Delft. Emmerink R.H.M., P. Nijkamp and P. Rietveld (1995), ‘Is congestion-pricing a ﬁrst-best strategy in transport policy? A critical review of arguments’, Environment and Planning B, 22, 581–602. Environmental Protection Agency (EPA) (2001), National Air Quality and Emissions Trends Report, 2000, EPA 454/R-01-004, Washington, DC: EPA. Espey, M. (1998), ‘Gasoline demand revisited: an international meta analysis of elasticities’, Energy Economics, 20, pp. 273–95. Essen van, H., O. Bello, J. Dings and R. van den Brink (2003), To Shift or Not to Shift, That’s the Question. The Environmental Performance of the Principal Modes of Freight and Passenger Transport in the Policy-Making Context, Delft/ Bilthoven: CE/Rijksinstituut voor Volksgezondheid en Milieu. European Commission (2003), Energy: Issues, Options and Technologies, report 20624, Brussels: European Commission, DG Research www.europa.eu.int/ comm/public_opinion/archives/special.htm. Evans, A. (2000), ‘The economic appraisal of road traﬃc safety measures in Great Britain’, ECMT Roundtable, 117. Flyvbjerg, B., N. Bruzelius and W. Rothengatter (2003), Megaprojects and Risk. An Anatomy of Ambition, Cambridge: Cambridge University Press. Freeman, A. Myrick, III (1993), The Measurement of Environmental and Resource Values: Theory and Methods, Washington, DC: Resources of the Future. Friedrich, R. and P. Bickel (2002), Environmental External Costs of Transport, Berlin: Springer-Verlag. Geerlings, H. (1998), Meeting the Challenge of Sustainable Mobility, Berlin: Springer-Verlag. Gillen, D. (2003), ‘The economics of noise’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 81–95. Goddard, H.C. (1997), ‘Using tradeable permits to achieve sustainability in the world’s large cities. Policy design issues and eﬃciency conditions for controlling vehicle emissions, congestion and urban decentralization with an application to Mexico City’, Environmental and Resource Economics, 10, 63–99. Golob, T.F., M.J. Beckman and Y. Zahavi (1981), ‘A utility-theory travel demand model incorporating travel budgets’, Transportation Research-B, 15B, 375–89. Graaﬀ, T. de and P. Rietveld (2004), ‘ICT and substitution between out-of-home and at-home work: the importance of timing’, Environment and Planning A, 36, 879–96. Graham, D.J. and S. Glaister (2004), ‘Road traﬃc demand elasticity estimates; a review’, Transport Reviews, 24, 261–74. Greene, D.L. (1998), ‘Why CAFE worked’, Energy Policy, 26 (8), 595–613. Hall, P. (1998), Cities in Civilization: Culture, Innovation and Urban Order, London: Weidenfeld and Nicolson, London. Handy, S. (1992), How Land Use Patterns Aﬀect Travel Patterns: A Bibliography, Council of Planning Librarians, University of Berkeley, CA, USA. Handy, S. (2003), ‘Amenity and severence’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 117–40. Hilbers, H. (2000), Verstedelijkingstoets [Valuation or Urbanization Alternatives], The Hague: Department of Economic Aﬀairs.

Transport and the environment

251

Holmen, B.A. and D.A. Niemeier (2003), ‘Air Quality’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 61–76. Hoevenagel, R. (1994), Contingent Veluation Method: Scope and Validity, Amsterdam: Vrije University. Intergovernmental Panel on Climate Change (IPCC) (1996), Climate Change 1995 – Impacts, Adaptation and Mitigation of Climate Change, Cambridge: Cambridge University Press. International Road Federation (2000), World Road Statistics 2000, Geneva: IRF. Jara Diaz, S.R. (2003), ‘On the goods–activities technical relations in the time allocation theory’, Transportation, 30, 245–60. Johansson, O. and L. Schipper (1997), ‘Measuring the long run fuel demand of cars’, Journal of Transport Economics and Policy, 31, 277–92. Johansson, B. (1995), ‘Strategies for reducing emissions of air pollutants from the Swedish transportation sector’, Transportation Research A, 29, 371–85. Johansson, B. (1999), ‘The economy of alternative fuels when including the cost of air pollution’, Transportation Research D, 4D (2), 91–108. Jones, P. (1998), ‘Urban road pricing, public acceptability and barriers to implementation’, in K.J. Button and E.T. Verhoef (eds), Road Pricing, Traﬃc Congestion and the Environment, Cheltenham, UK and Lyme, USA: Edward Elgar. Jones-Lee, M. and G. Loomes (2003), ‘Valuation of safety’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 451–62. Jong, G.C. de (1996), ‘A disaggregate model system of vehicle holding duration, type choice and use, Transportation Research B, 30 (4), 263–76. Kahnemann, D. and A. Tversky (1979), ‘Prospect Theory: an analysis of decision under risk’, Econometrica, 47, 263–91. Khare, M. and P. Sharma (2003), ‘Fuel options’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 159–82. Kitamura, R. (1988), Life Style and Travel Demand, Washington: Transportation Research Board. Kling, C.L. (1994), ‘Emission trading vs. rigid regulations in the control of vehicle emissions’, Land Economics, 70 (2), 174–88. Krik, R. (2005), ‘Valuation of wetlands by means of hedonic pricing methods’, Vrije Universitiet Ph D dissertation, Amsterdam. Lave, L.B. and H.L. MacLean (2002), ‘An environmental-economic evaluation of hybrid electric vehicles: Toyota’s Prius vs. its conventional internal combustion engine Corolla’, Transportation Research D, 7D (2), 155–62. Litman, T.A. (2002), ‘Transportation demand management and “win–win” transportation solutions’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 805–15. Litman, T. (2005), ‘Eﬃcient vehicles versus eﬃcient transportation. Comparing transportation energy conservation strategies’, Transport Policy, 12, 121–9. Maddison, A. (2001), The World Economy: A Millennial Perspective, Paris: OECD. Markowitz, H.M. (1987), Mean Variance Analysis in Portfolio Choice, Oxford: Basil Blackwell. McCubbin, D.R. and M.A. Delucchi (1999), ‘The health cost of motorcycle related air pollution’, Journal of Transport Economics and Policy, 33, 253–86.

252

Yearbook of environmental and resource economics

McCubbin, D.R. and M.A. Delucchi (2003), ‘The health eﬀects of motor vehiclerelated air pollution’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 411–28. Meurs, H. and R. Hajer (2001), ‘Spatial structure and mobility’, Transportation Research D, 6D (6), 429–46. Miller, T.R. (2000), ‘Variations between countries in valuation of statistical life’, Journal of Transport Economics and Policy, 34, 169–88. Mobility (2001), World Mobility at the End of the Twentieth Century and its Sustainability, World Business Council for Sustainable Development, Massachusetts Institute of Technology and Charles River Associated Incorporated. Mohring, H. (1976), Transportation Economics. Cambridge, MA: Ballinger. Moore, M.W. and J.K. Viscusi (1998), ‘Doubling the estimated value of life: results using new occupational fatality data’, Journal of Policy Analysis and Management, 7, 476–90. Mokhtarian, P. and C. Chen (2004), ‘TTB or not TTB that is the question: a review and analysis of the empirical literature on travel time (and money) budgets’, Transportation Research Part A, 35, 643–75. Mokhtarian, P. and I. Salomon (1999), ‘Travel for the fun of it’, Access, 15, 20–31. Nash, C. (2003), ‘Marginal cost and other pricing principles for user charging in transport: a comment’, Transport Policy, 10 (4), 345–8. Newman, Peter W.G. and Jeﬀrey R. Kenworthy (1989), Cities and Automobile Dependence: A Sourcebook, Aldershot, UK and Brookﬁeld, VT: Gower Technical. OECD (2002), Energy Balances of OECD Countries, 1999–2000, Paris: OECD. Peeters, P. (1997), Slow Motion, Delft: Delft University Press. Perman, R., Y. Ma, J. McGilvray and M. Common (2003), Natural Resource and Environmental Economics, Harlow, UK: Pearson Education. Plaut, P.O. (2004), ‘The uses and users of SUV’s and light trucks in commuting’, Transportation and Research D, 9D (3), 175–83. Pucher, J. (1998), ‘Urban transport in Germany: providing feasible alternatives to the car’, Transport Reviews, 18 (4), 285–310. Quinet. E. (2003), ‘Evaluation of environmental impacts’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 357–75. Quinet, E. and R. Vickerman (2004), Principles of Transport Economics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Rietveld, P. (2002), ‘Why railway passengers are more polluting in the peak than in the oﬀ-peak; environmental eﬀects of capacity management by railway companies under conditions of ﬂuctuating demand’, Transportation Research D, 7, 347–56. Rietveld, P. and V. Daniel (2004), ‘Determinants of bicycle use in the Netherlands: do municipal policies matter?’, Transportation Research A, 38 (7), 531–50. Rietveld, P. and R. Vickerman (2004), ‘Transport in regional science: the “death of distance” is premature’, Papers in Regional Science, 83, pp. 229–48. Rietveld, P. and S. van Woudenberg (2005), ‘Why fuel prices diﬀer’, Energy Economics, 27, 79–92. Rietveld, P., A. van Binsbergen, T. Schoemaker and P. Peeters (1998), ‘Optimal speed limits for various types of roads: a social cost beneﬁt analysis for the Netherlands’, in R. Roson and K.A. Small (eds), Environment and Transport in Economic Modelling, Dordrecht: Kluwer, pp. 206–25.

Transport and the environment

253

RIVM (2000), Milieu winst op het spoor [Environmental Advantages of Rail Travel], Bilthoven. Rodriguez, D.A. and J. Joo (2004), ‘The relationship between non-motorized mode choice and the local physical environment’, Transportation and Research D, 9D (2), 151–73. Rosen, H. (1992), Public Finance, Toronto: Irwin Publishers. Rothengatter, W. (2003a), ‘How good is ﬁrst best? Marginal cost and other pricing principles for user charging in transport’, Transport Policy, 10 (2), 121–30. Rothengatter, W. (2003b), ‘Environmental concepts – physical and economic’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 9–35. Rovers, J.H.M (1999), ‘The impact of order-driven production on freight transport’, thesis, Vrije Universiteit, Twigt, Waddinxveen. Saelensminde, K. (2004), ‘Cost–beneﬁt analyses of walking and cycling track networks, taking into account insecurity, health eﬀects and external costs of motorized traﬃc’, Transportation Research Part A: Policy and Practice, 38 (8), 593–606. Schafer, A. (1998), ‘The global demand for motorized mobility’, Transportation Research Part A: Policy and Practice, 32 (6), 455–77. Schipper, L.J. and L. Fulton (2003), ‘Carbon dioxide emissions from transportation: trends, driving factors and forces for change’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 203–24. Schipper, Y. (1999), Market Structure and Environmental Costs: A Welfare Analysis of European Air Transport Reform, Tinbergen Institute thesis, Thela thesis, no. 217, Amsterdam. Sheﬃ, Y. (1985), Urban Transportation Networks. Equilibrium Analysis with Mathematical Programming Methods, New Jersey: Prentice-Hall. Shoup, D. (1997), ‘Evaluating the eﬀects of cashing out employer paid parking, eight case studies’, Transport Policy, 4, 201–16. Small, K.A. (1992), Urban Transportation Economics, New York: Harwood Academic Publishers. Sperling, D. (2003), ‘Cleaner vehicles’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 185–203. Stanley, J. and P. Watkiss (2003), ‘Transport energy and emissions: buses’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 227–47. Stern, D.I., M.S. Common and E.B. Barbier (1996), ‘Economic growth and environmental degradation: the environmental Kuznets curve and sustainable development’, World Development, 24 (7), 1151–60. Stopher, P.R. (2003), ‘History of environmental legislation’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 497–514. Tapio, P. (2005), ‘Towards a theory of decoupling: degrees of decoupling in the EU and the case of road traﬃc in Finland between 1970 and 2001’, Transport Policy, 12, 137–51. Ubbels, B., P. Rietveld and P. Peeters (2002), ‘Environmental eﬀects of a kilometre charge in road transport: an investigation for the Netherlands’, Transportation Research Part D, 7, 255–64. Verhoef, E.T. (1996), The Economics of Regulating Road Transport, Cheltenham, UK and Lyme, USA: Edward Elgar.

254

Yearbook of environmental and resource economics

Vermeend, W. and J. van der Vaart (1998), Greening Taxes, Deventer: Kluwer. Wardman, M. (2001), ‘A review of British evidence on time and quality service valuations’, Transportation Research, E37 (107), 128. Waters, W.G. II (2003), ‘Transportation of hazardous goods and materials’, in D.A. Hensher and K.J. Button (eds), Handbook of Transport and the Environment, Oxford: Elsevier, pp. 707–25. Wee, B. van (2005), ‘Cost–beneﬁt analysis and health impacts of road traﬃc emission: an overview of literature and a categorisation of policy measures’, presented at the 8th NECTAR conference, Las Palmas. Wee, B. van and P. Rietveld (2003), ‘Openbaar vervoer en milieu: mythen en feiten’, [‘Public transport and the environment: myths and facts’], Milieutijdschrift ArenA, 9 (5), 74–8. World Business Council for Sustainable Development (WBCSD) (2000), World Mobility at the End of the Twentieth Century and its Sustainability, Geneva: WBCSD.

6. The Faustmann face of optimal forest harvesting Richard J. Brazee* 1.

INTRODUCTION

Forests are important natural resources in countries and regions of the world. Forests cover approximately 30 percent of the earth’s land area (FAO 2000). Without human interventions forests would be shaped solely by natural elements including the sun, climate, soil nutrients, topography, weather and ﬁres. Using primarily vegetative manipulations including harvesting, planting, thinning and burning, humans manage forests to meet personal and social objectives. Given the key role of vegetative manipulation, the fundamental question of forestry economics is: when should a tree or a stand of trees be harvested? Answers range from immediate harvest to never harvest. Important speciﬁc harvesting questions include: 1 2 3

When should a stand of trees1 be harvested? Should deforestation rates in the Amazon be slowed? Is proﬁt maximization consistent with sustainable harvesting?

A secondary long-standing question is: how can small private landowners be encouraged to more actively manage their forests? The proportion of forest land held by small private owners ranges across countries from virtually none to virtually all. Correspondingly the importance of encouraging small private landowners to actively manage their forests ranges from irrelevant to crucial (Kuuluvainen et al. 1996). The goal of this chapter is to provide a modeling introduction to optimal forest harvesting for economists. A single model, the Faustmann model, serves as foundation for a simple introduction to optimal harvesting (Faustmann 1849 [1995]). The Faustmann model is suﬃcient to span much of the current range of optimal harvesting analysis. Speciﬁcally, the Faustmann model and selected extensions are used to illustrate analyses of several forestry policy and management questions of interest to forest economists. 255

256

Yearbook of environmental and resource economics

As with all introductions, this one is incomplete and omits important research in optimal forest harvesting.2 To be consistent with historical development, with the dominant stream of analysis in forestry economics, and to ease exposition, this chapter focuses on determining optimal rotation ages for timber production. Non-timber goods and services are included in speciﬁc sections. Research that is primarily empirical is not included. Policy and management issues presented highlight questions important in forestry, but less important outside of forestry. Given the focus on optimal harvesting, this chapter under-represents two-period harvest models. The arguments of many papers, particularly recent papers, were not included due to the complexity of the analysis presented. To have included these papers would have both greatly lengthened an already long chapter and made exposition much more diﬃcult. The rest of the chapter is organized into ﬁve sections. In section 2 optimal harvest models are characterized and a simple Faustmann model is reviewed. In section 3 silvicultural eﬀort, labor supply, selected public policies and nontimber goods and services are incorporated into the simple Faustmann model presented in section 2. In section 4 dynamic programming is used to address parameter ﬂuctuations and catastrophic forest risks. In section 5 optimal control theory is used to analyze dynamic forest models, forest investments within a portfolio and credit constraints. Uneven-aged management and other selected omitted issues are brieﬂy discussed in section 6.

2. DEVELOPMENT OF OPTIMAL HARVESTING MODELS Characterization of Optimal Harvesting Models Although planting and harvesting are inherent activities in both forestry and agricultural production, some other inherent characteristics of forest production diﬀer signiﬁcantly from other inherent characteristics of agricultural production. Given the time period required for trees to grow from regeneration to maturity, forest stands are a durable asset rather than an annual ﬂow. In contrast to agricultural decision-making, forestry decisionmaking has a time horizon of years and decades rather than weeks, months or a few years. The decision to harvest a stand of trees more closely resembles the decision to sell a durable asset, such as a house or a ﬁnancial asset, than to harvest most agricultural commodities. After harvest forest landowners face a signiﬁcant decision: should they re-plant the site of the recently harvested stand, do nothing and allow trees to regenerate naturally, or actively use the land for non-forestry purposes? If trees are planted or

The Faustmann face of optimal forest harvesting

257

allowed to regenerate naturally, the land will either be occupied for a long period with trees, or prior to a minimum harvest age the landowner will incur signiﬁcant costs to remove the regenerated trees. In contrast to many ﬁnancial assets, many durable assets and agricultural commodities, stands of trees often serve multiple purposes. Stands may be harvested for timber to be used for lumber, veneer, pulp or ﬁrewood. Before harvest, most stands provide a range of non-market goods and services including a variety of aesthetic, ecological and recreational goods and services. As will be discussed for many sets of relative valuation of diﬀerent forest goods and services, it is optimal to never harvest some stands (Bowes and Krutilla 1985). If we consider only one rotation of trees, that is, seedlings are planted and at some point in the future the resulting stumpage3 is harvested and the land is not re-planted, then the optimal forestry-harvesting problem resembles the wine-aging problem. In both problems an owner must make a one-time decision regarding when to sell/harvest/drink an asset that is increasing in value. If a wine-maker maximizes the net present value (NPV) of his wine, then he will sell it when the increase in the discounted beneﬁts from delaying selling equals the discounted costs from delaying selling. Similarly, if a forest landowner maximizes the NPV of her stumpage, then she will sell the stumpage when the increase in the discounted beneﬁts from delaying selling equals the discounted costs from delaying selling. A key diﬀerence between a stand of trees and wine is that aging trees occupy land while aging wine is stored in barrels and bottles. If a wine-maker wants to increase wine production, he purchases or grows more grapes, which may require the purchase or rental of more land, and buys or constructs more barrels and more bottles. In general, if a forest owner wants to increase her production, she purchases or rents more land, and purchases and plants more seedlings. Historically land is much more expensive than barrels or bottles, and growing forests require much more land than wine production requires. In contrast to analyses of wine aging in which additional grapes or additional land may be purchased, many analyses of forest harvesting assume a binding land constraint, that is, analyses assume that forest land is ﬁxed and focus on determining optimal harvesting on the managed land. With a binding land constraint, the forest landowner must consider the opportunity cost of land. This arises from future rotations of trees, or from non-forestry uses. European settlement of North America and recent deforestation in the Amazon are two examples in which high opportunity costs of non-forestry uses contributed to the destruction of forests. However, simple as this answer to the optimal harvesting problem appears, the path to it was convoluted. First, the answer has been independently derived several times in the eighteenth to twentieth centuries (Löfgren

258

Yearbook of environmental and resource economics

1983, Scorgie and Kennedy 1996). Second, several famous economists have presented incorrect answers to the optimal harvesting problem. The most common error was ignoring the opportunity cost of land (Samuelson 1976). Third, despite the fact that he was neither the ﬁrst independent discoverer nor the most famous discoverer – one of the later independent discoverers was Bertil Ohlin (Ohlin 1921 [1995]) – Martin Faustmann, a German forester, is credited with correctly formulating the optimal harvesting question in the context of a land valuation paper (Faustmann 1849 [1995]). Pressler provided the solution to the question Faustmann posed (Löfgren 1983, Pressler 1860 [1995]). Fourth, Faustmann’s paper, published in 1849, remained in relative obscurity for over one hundred years. Only in the late 1950s was the Faustmann model broadly recognized as the correct model for the optimal harvesting problem (Bentley and Teeguarden 1965, Gaﬀney 1957). Until Samuelson’s 1976 paper which demonstrated the consistency of the Faustmann model with market equilibria, many non-forestry economists doubted the validity of the Faustmann model. The Faustmann model is now the dominant paradigm in optimal harvesting literature. By 2000 over 300 papers had extended the model. Eighty-ﬁve percent of these papers were published between 1980 and 2000 (Newman 2002). Basic Faustmann Model Forest decision-making is inherently intertemporal since a series of decisions must be made over a long time horizon. The Faustmann model predates the development of modern dynamic optimization methods such as optimal control theory and dynamic programming. A key to solving the optimal rotation problem before the development of modern methods was to frame the optimal harvesting problem so that it could be solved using static optimization techniques. In the Faustmann model the time horizon is inﬁnite and all parameters, including stumpage price, regeneration costs, the discount rate and the timber volume function, are known and constant over time. Forest landowners maximize the NPV of land. These assumptions imply that the same harvesting age is optimal in every rotation, and that the optimal harvesting problem can be formulated as single rotation multiplied by a geometric series. For mathematical simplicity several additional assumptions are made in simple Faustmann models. A key assumption is that even-aged management is employed for the stand. This implies that all trees are regenerated and harvested simultaneously. Other simplifying assumptions include that the stand is initially without trees, that is, the land is bare,4 there is a single output, the land area is ﬁxed and homogeneous, and the stand is not thinned before harvest.

The Faustmann face of optimal forest harvesting

259

The initial Faustmann optimal harvesting formulation is (Faustmann 1849 [1995], Johansson and Löfgren 1985, Pressler 1860 [1995], Samuelson 1976):5 Maximize NPV(T) L(0,T) e rTPQ(T) C e rT (e rTPQ(T) C) w.r.t. T e 2rT (erTPQ(T) C) . . . (e rTPQ(T) C)(1 e rT e 2rT . . .) erTPQ(T) C (6.1) 1 erT where C is stand regeneration cost, L(0,T) is the maximum NPV of bare land with stumpage harvested at age T, commonly referred to as the soil expectation value, P is stumpage price, r is the discount rate, T is harvest age, and Q(T) is stumpage harvested at age T; that is, Q(T) is the volume function for the stand. In the initial Faustmann model, the opportunity cost of land is the discounted value of future rotations. If harvest in a rotation is delayed, then the harvests in all future rotations will also be delayed, which reduces the present value of the future rotations. The opportunity costs of land may also be from non-forestry uses. Let W be the NPV of land from the most proﬁtable non-forestry use, and V be the maximum NPV of the land from either non-forestry uses or the soil expectation value, then the objective function in (6.1) may be generalized to: NPV(T)erT PQ(T)CerT V

Maximize w.r.t. T

(6.2)

where Vmax{W, L(0,T)}. If the forest land market is in equilibrium, then the NPV of land in the non-forestry use equals the soil expectation value, i.e. WL(0,T) (Samuelson 1976). To ﬁnd the optimal harvest age (6.2) is diﬀerentiated with respect to T and the derivative is set equal to zero.

erT rPQ(t) P

dQ(T) rV 0 dT

(6.3)

A suﬃcient but not strictly necessary condition for optimality is a concave stumpage volume function. A useful manipulated ﬁrst-order condition is: dQ(T) P rPQ(T) rV dT

(6.4)

260

Yearbook of environmental and resource economics

In general it is optimal to harvest when the marginal beneﬁts of delaying harvest equal the marginal costs of delaying harvest. More speciﬁcally, it is optimal to harvest when the capital gain due to stumpage growth from delaying harvest equals the interest forgone from the sum of stumpage and cleared land from delaying harvest. If the marginal beneﬁts of delaying harvest on the left-hand side of (6.4) are greater than the marginal costs of delaying harvest on the right-hand side of (6.4), then stumpage is not harvested. The stand is allowed to grow until the marginal beneﬁts of delay equal the marginal costs of delay. If the marginal beneﬁts of delaying harvest are less than the marginal costs of delaying harvest, then stumpage should be harvested immediately. If after the ﬁrst rotation non-forest uses are more valuable than forestry uses, then W replaces V in (6.4).6 If after the ﬁrst rotation forestry is the best use, then L(0, T) replaces V and (6.4) may be rewritten as: dQ(T) erTPQ(T) C r(PQ(T) C) rPQ(T) r dT 1 e rT 1 erT

P

(6.5)

The numerator of the right-hand side is now the interest forgone from delaying the harvest of the current rotation, while the denominator of the right-hand side increases the interest forgone to reﬂect all future rotations. The results in (6.3)–(6.5) describe the ‘Faustmann rule’. The harvest rule described by (6.5) highlights a long-running debate regarding sustainability and ﬁnancial models of forest management (Shepard 1925). Equation (6.5) requires the growth rate of the value of unharvested stumpage increase faster than the rate of discount. In fact many species growth rates are low relative to the discount rate. Whether Faustmann models support sustainable harvests is place and time speciﬁc. Comparative statics results for parameter shifts in simple Faustmann models are well known (Chang 1983, Clark 1990). An increase in regeneration costs lengthens the rotation as the landowner delays incurring the next rotation’s regeneration costs. An increase in stumpage price and under most circumstances an increase in the discount rate reduce the optimal harvest age as the opportunity cost of delaying harvest increases (Binkley 1987). Factors that favor sustainable harvests include low discount rates, high rates of stumpage growth, increases in stumpage prices, and low non-forestry land values. High non-forestry land values based on agricultural subsidies and the need to clear a portion of the land to claim title contribute to deforestation in the Amazon (Southgate 1990). A long period of increasing stumpage prices and low non-forest land values contribute towards the large increase in forest biomass in the USA since 1920.

The Faustmann face of optimal forest harvesting

261

The Faustmann Model as Foundation The primary strength, as well as the primary weakness, of the Faustmann model is that it characterizes the essence of the optimal rotation problem and solution in a simple model. The simplicity and clarity of the model’s assumptions and conclusions allow the Faustmann model and the Faustmann rule to be broadly understood. The Faustmann rule is similar to the Hotelling rule, which is used to characterize allocations of depletable resources over time (Hotelling 1931), in the sense that both serve as one of a small set of foundational concepts within the ﬁeld of natural resource economics. The simplicity of the Faustmann model is also its most important weakness. By its very simplicity, the model must omit many factors that aﬀect optimal rotation age. To remedy the omission of factors, the Faustmann model has been extended to include several important factors including silvicultural eﬀort, labor supply, taxes, subsidies, non-timber goods and services, risk and regulatory constraints. Fortunately, the Faustmann model’s simplicity and clarity allow it to serve as the foundation for the analysis of many forest policy and management issues. The bulk of the remaining text brieﬂy introduces several important forest policy and management issues, and then describes how extensions of the Faustmann model address these issues. Each issue is presented in its simplest meaningful context with easier-to-model issues presented ﬁrst.

3. SILVICULTURAL EFFORT, AMENITIES AND PUBLIC POLICIES Silvicultural Eﬀort and Labor Supply In the Faustmann model presented in section 2, ﬁxed regeneration costs are the only non-land input included in the model. Fixing the regeneration costs and the level of silvicultural eﬀort are strong assumptions. The magnitude of silvicultural eﬀort in the form of labor and other inputs to assist regeneration and to manage a stand during a rotation signiﬁcantly impacts stumpage volume at age T. Possible silvicultural management activities include site preparation, pre-commercial thinning and protection from insects, disease and ﬁre. Some of the best-known extensions to the Faustmann model include silvicultural eﬀort (Samuelson 1976, Chang 1983). Possibly the simplest form of explicitly including silvicultural eﬀort is to make the stumpage volume function dependent on the magnitude of

262

Yearbook of environmental and resource economics

silvicultural eﬀort. Let E be silvicultural eﬀort at the start of each rotation, and C(E) be regeneration costs.7 The objective function in condition (6.2) becomes: Maximize w.r.t. E & T

NPV(E,T)ert PQ(E, T)C(E)erT V

(6.6)

where the volume function Q(E,T) is a function of both silvicultural eﬀort, E, and rotation age, T. The NPV of bare land depends on two choice variables, silvicultural eﬀort and rotation age. Silvicultural eﬀort impacts the NPV of bare land in two ways. First, for any given age, greater silvicultural eﬀort results in greater stumpage volume. Second, changing the size of regeneration costs directly impacts the magnitude of NPV through both the initial regeneration costs and with inﬁnite future rotations the costs of every subsequent rotation. To illustrate this impact, note that with inﬁnite rotations (6.6) becomes: Maximize w.r.t. E & T

NPV(E,T)

erTPQ(E, T) C(E) 1 e rT

(6.7)

The optimal level of silvicultural eﬀort and the optimal harvest age are found by diﬀerentiating NPV(E,T) with respect to E and T, and then setting both partial derivatives equal to zero:8 erTPQE (E, T) CE (E) 0 (6.8) 1 erT erT [ rPQ(E, T) rC(E) PQT (E, T)(1 erT )] 0 NPVT (E,T) (1 erT ) 2 NPVE (E,T)

where subscripts E and T represent partial derivatives with respect to E and T, respectively. The ﬁrst condition of (6.8) requires that the discounted marginal revenue product of silvicultural eﬀort equal the marginal cost of silvicultural eﬀort. The second condition of (6.8) is analogous to (6.3) and requires that the interest forgone from delaying harvest equal the increase in total revenue from an increase in stumpage volume from delaying harvest. Equation (6.8) diﬀers from (6.3) in that the result in (6.8) is dependent on the size of the silvicultural eﬀort, while in (6.3) silvicultural eﬀort is ﬁxed. Periodic or continuous management costs would add another cost term to (6.6), (6.7) and, depending on the functional form of the management costs, possibly (6.8) (Chang 1983).9 Its impact is illustrated by inclusion of a property tax in the next section. Harvest costs, road building and

The Faustmann face of optimal forest harvesting

263

maintenance costs at the time of harvest are subtracted from harvest revenue (Heaps and Neher 1979). Taxes and Subsidies Due to the long production period for timber, taxes have a tremendous potential impact on the proﬁtability of timber production, the value of forest land and the optimal rotation age. A landowner who regenerates a stand needs to wait for a commercial thinning or harvest before any revenue is generated from timber production. Common lags between regeneration investments and harvest revenues range from ten years in the tropics to 150 years in high latitudes (Kuuluvainen and Salo 1991). Forest economists have long recognized the impact of taxes on optimal rotation age (Chang 1982).10 A property tax is an annual payment based on the assessed land value. Assessments may be based on either forestry or non-forestry land uses. Introducing a property tax into the Faustmann model adds a term to the NPV functions in (6.6) and (6.7). With the inclusion of a property tax, (6.7) becomes: Maximize w.r.t. E & T

NPV(E, T)

e rTPQ(E, T) C(E) –r 1 e rT

(6.9)

where is the instantaneous tax payment, and /r is the discounted value of an inﬁnite stream of instantaneous tax payments. Note that if the tax payments are ﬁxed, then they do not enter the ﬁrst-order conditions, and the inclusion of a property tax does not change either the optimal rotation age or the optimal level of silvicultural eﬀort. Similar to all forestry taxes, a property tax reduces the NPV of the land. Tax assessments may be based on either forestry or non-forestry land uses. When assessments are based on non-forestry uses, a previously positive NPV may become negative after the property tax is included. When tax assessments are based on the NPV of forestry uses, a previously positive NPV from forest uses decreases, but remains positive (Chang 1982, Johansson and Löfgren 1985). Given the potential impact of property taxes on forest land values and the transference of land from forestry to non-forestry uses, forest economists have analyzed taxes that are collected at harvest rather than throughout a rotation. Forest economists have analyzed yield taxes on harvest revenues, severance taxes on quantities harvested, and productivity taxes on net harvest revenues as substitutes for a property tax (Amacher et al. 1991, Chang 1982, Klemperer 1983). These taxes all reduce net harvest revenues. With the inclusion of a yield tax (6.7) becomes:

264

Yearbook of environmental and resource economics

Maximize w.r.t. E & T

NPV(E,T)

e rTPQ(E,T) C(E) 1 e rT

(6.10)

where is the yield tax rate. With multiple rotations yield, severance and most forms of productivity taxes impact both optimal rotation age and the optimal level of silvicultural eﬀort. With a yield tax the necessary conditions are: erTPQE (E, T) CE (E) 0 (6.11) 1 erT rT rT [rPQ(E, T) rC(E) PQT (E, T)(1 e )] e NPVT (E, T) 0 (1 e rT ) 2

NPVE (E, T)

In general the impacts of yield tax on the optimal level of silvicultural eﬀort and optimal rotation age are ambiguous. A negative cross-partial derivative of the volume function, QET (E,T) 0, implies a negative cross-partial derivative of the NPV function, NPVET (E,T) 0. With NPVET(E,T) 0, the inclusion of a yield tax reduces the optimal level of silvicultural eﬀort and increases the optimal rotation age. With QET (E,T) 0, and NPVET (E, T) 0, the impacts of a yield tax on the optimal level of silvicultural eﬀort and optimal rotation age are ambiguous. Governments often provide incentives to small private landowners to promote timber management on their land. Some of these incentives are knowledge based. For example, government foresters may assist landowners with management plans and information about regeneration, thinning and harvesting. Financial incentives are also commonly used, with regeneration subsidies being the most common. Regeneration subsidies reduce regeneration costs (Chang 1983), and (6.7) becomes: Maximize w.r.t. E & T

NPV(E, T)

erTPQ(E, T) C(E)(1 ) (6.12) 1 e rT

where is the subsidy rate. In general the impacts of regeneration subsidy on the optimal level of silvicultural eﬀort and optimal rotation age are ambiguous. Similar to a yield tax when QET(E,T) 0 which implies NPVET(E, T) 0, the impacts of regeneration study are unambiguous. The inclusion of a regeneration subsidy increases the optimal level of silvicultural eﬀort and reduces the optimal rotation age. With QET(E,T) 0 which implies NPVET(E,T) 0, the impacts of regeneration subsidies on silvicultural eﬀort and rotation age are ambiguous.

The Faustmann face of optimal forest harvesting

265

Non-timber Goods and Services In addition to timber, forests produce many other goods and services. Non-timber goods and services include recreation, wildlife habitat and aesthetics. Many non-timber goods and services are also non-market goods and services. Standard non-market valuation techniques may be employed to value forest non-market goods and services. There are several approaches to incorporate non-timber goods and services into the analysis. Hartman (1976) used a simple approach by extending the Faustmann model to include non-market goods and services. Let A(E,t) be the sum of the value of all non-timber goods and services from age t stumpage. The Faustmann NPV function of (6.7) becomes: T

erTPQ(E, Maximize NPV(E, T) w.r.t. E & T

T) C(E) ertA(E, t)dt 1 erT

0

(6.13)

The numerator of the Faustmann NPV function now includes the discounted sum of the value of all non-market goods and services during one rotation. The other terms in the numerator remain the discounted timber revenues and regeneration costs during one rotation. The denominator remains the geometric series that captures the impact of inﬁnite rotations. Other than the assumption that A(E,t)0, no general statements can be made about the shape of the non-timber goods and services function. Diﬀerent shapes are possible with diﬀerent goods and services (Calish et al. 1978). For example, an existence value for forest land could be constant over a rotation if such a value depended solely on whether the land was forested or not. Some wildlife species such as white-tailed deer and ruﬀed grouse prefer younger stands, and the non-timber value of forested land declines as the rotation ages. Some wildlife species such as spotted owl prefer very old stands, and the non-timber value of forested land increases as the rotation ages. Other goods and services have diﬀerent shapes over a rotation. If values from diﬀerent goods and services are combined, as in A(E,t), then it is even more diﬃcult to describe the shape than it is for one good or service. The inclusion of a non-timber value function in the NPV function introduces the possibility of a non-convex feasible set (Strang 1983, Swallow et al. 1990). Non-convexity of the feasible set makes maximizing the NPV function more diﬃcult since the optimal conditions only describe a local rather than a global optimum. The local optimal levels of silvicultural eﬀort and harvest age are found by partially diﬀerentiating NPV(E,T)

266

Yearbook of environmental and resource economics

with respect to E and T, and then setting both partial derivatives equal to zero: T

NPVE (E, T)

e rTPQE (E, T) CE (E) e rtAE (E, t)dt 1 erT

0

0

(6.14)

NPVT (E, T)

T

erT r( PQ(E, T) C(E) ) (PQT (E, T) A(E, T) ) (1 e rT ) r e rtA(E, t)dt 0

(1 e rT ) 2

0

The ﬁrst condition of (6.14) generalizes the ﬁrst condition of (6.8) and requires that the discounted marginal revenue product of silvicultural eﬀort, the discounted sum of discounted marginal timber revenues and the marginal value non-timber services and services equals the marginal cost of silvicultural eﬀort. The second condition of (6.14) generalizes the second condition of (6.8) and requires that the sum of interest forgone from net harvest revenues and from opportunity costs of land from net delaying harvest equal the increase in revenue from timber from stumpage growth and the marginal value of the non-timber services quantity multiplied by the impacts on future rotations from delaying harvest. The solution that maximizes the NPV of bare land may be identiﬁed from all combinations that satisfy (6.14). In (6.13) and (6.14) the forest owner captures not only timber harvested but the value of the non-timber goods and services too. However, many of these are externalities. Visual beauty, carbon sequestration and hiking are externalities jointly produced with other forest goods and services. As with all externalities, there is a diﬀerence between private and social net beneﬁts, with the landowner only capturing a fraction of the value of the externalities. If the landowner captures no such value, then the goods and services which generate the externalities will not enter her objective, and her objective function will be (6.6) or (6.7). Although having no inﬂuence on harvest decisions, the landowner’s neighbors will enjoy the full beneﬁts of the externalities. If the landowner captures some, but not all of the value of the externalities, then her objective in (6.13) needs to be modiﬁed to reﬂect the proportion of the externalities that the landowner captures.11 Land Quality and Multiple Stand Management In the Faustmann model homogeneous land quality is assumed. The assumption of homogeneous land quality is particularly misleading with the inclusion of non-timber goods and services. A key strategy in managing

The Faustmann face of optimal forest harvesting

267

for more than one good or service is to exploit diﬀerences in land quality. Many public forest managers are legally required to employ ‘multiple use management’ to manage forests for more than one use. A long-standing multiple use question is whether all stands should be managed for all forest goods and services, or whether some stands should specialize in a subset of forest goods and services as long as all parcels taken together provide the full range of uses. It should not be surprising to economists that the ﬂexibility of being able to implement diﬀerent management regimes on heterogeneous stands has potential gains from specialization (Vincent and Binkley 1993). Equations (6.13) and (6.14) represent an extreme case in which every stand is managed identically. To manage some hectares for a subset of goods and services, the manager must ﬁrst consider diﬀerent combinations of management strategies and then ﬁnd the combinations of management strategies that maximize the NPV. Next the manager compares the diﬀerent combinations of strategies regimes and chooses the combination with the highest NPV for the entire management area. This approach allows for specialization and a range of management strategies. Usually some stands will never be harvested, while others will be harvested intensively. The possibility that specialization will be optimal increases when there is a difference in stand ages at the start of the planning period, as some stands may already be specialized. The US Forest Service’s planning process in response to the passage of NEPA in 1976 is an example of combining management strategies. The US Forest Service constructed several scenarios for each national forest, and then solicited public input before selecting a preferred scenario. Regulatory Constraints In the Faustmann model presented in section 2, no regulatory constraints are imposed on the forest landowner. In some regions or countries (Nordic Countries, Northern Europe, Canada and some US states), forest landowners are regulated through government-imposed forest practice acts, whose primary goal is to ensure a minimum level of forest quality on the regulated land. Two common forest practice act regulations are regeneration success and minimum harvest age. The former usually require a minimum number of live trees per hectare one or two years after harvest; they seek to ensure adequate forest cover, that is, a minimum density of trees on forested land. With minimum harvest age regulations, the forest landowner is not legally able to harvest until stumpage reaches a speciﬁed age. Harvest age constraints seek to ensure that trees are a minimum size before harvest (Hyytiäinen and Tahvonen 2001).

268

Yearbook of environmental and resource economics

Regeneration success and minimum age constraints may be included in the Faustmann model. A Lagrangian function, $(), with (6.7) and constraints is: Maximize w.r.t. E & T

e rTPQ(E, T) C(E) 1 e rT (C(E) K) %(T Tmin )

$(E, T, , %)

(6.15)

where K is the minimum expenditure required to insure that the regeneration success regulation is met, Tmin is the minimum permissible harvest age, and and % are multipliers on the non-negativity constraints, C(E) K, and T Tmin. Diﬀerentiating $(E,T,,%) with respect to E and T and setting both partial derivatives equal to zero provides the Kuhn–Tucker conditions: $E (E, T, , %) $T (E, T, , %)

erTPQE (E, T) CE (E) CE (E) 0 1 e rT

e rT [r(PQ(E, T) C(E)) PQT (E, T)(1erT )] % 0 (1erT ) 2 (C(E) K) 0

%(T Tmin ) 0

(6.16)

If the constraints are non-binding, then the ﬁrst two conditions of (6.16) are identical to (6.8). With a binding regeneration expenditure constraint, the ﬁrst condition of (6.16) requires that the discounted marginal revenue product of silvicultural eﬀort equal the marginal cost of silvicultural eﬀort plus the marginal value of the regeneration constraint. With a binding regeneration expenditure constraint, the second condition of (6.16) requires that the interest forgone from delaying harvest equal the increase in net revenue from an increase in stumpage volume from delaying harvest plus the marginal value of the harvest age constraint. If the regeneration regulation is binding, then 0, and the constraint requires that the level of silvicultural eﬀort is higher than the level of silvicultural eﬀort that maximizes NPV. Depending on the shape of the volume function, rotation length may be greater than or less than the rotation age that maximizes NPV. Rotation age increases or decreases depending on whether the impact of delaying harvest to postpone required additional regeneration expenses outweighs the impact from an increase in stumpage volume due to the higher level of silvicultural eﬀort. If the minimum harvest age regulation is binding, then % 0, and the harvest age will be older than the harvest age that maximizes NPV. Depending on the shape of the volume function, silvicultural eﬀort may be

The Faustmann face of optimal forest harvesting

269

greater than or less than the silvicultural eﬀort that maximizes NPV. Silvicultural eﬀort increases or decreases depending on whether the impact of an increase in stumpage volume at harvest outweighs the impact from increased discounting of harvest revenues.

4.

RISK AND UNCERTAINTY

Stumpage Price Risk and Uncertainty The Faustmann model assumes that all parameters are known and constant. In fact all parameters change over time (McConnell et al. 1983, Hardie et al. 1984, Newman et al. 1985). The impacts on optimal harvesting decisions of the variations of some parameters are more important than the variations of other parameters. Variations in regeneration costs are usually relatively small and known at the time of regeneration. In the absence of a catastrophic event, timber volume functions are known with a high degree of accuracy (Avery and Burkhart 2001). For many species prediction errors are usually under 10 per cent and often under 1 percent. Variations in stumpage prices and discount rates may dramatically impact optimal harvesting decisions. For some species variations in stumpage prices are as high as 30 per cent of the mean stumpage price (Ulrich 1981). Given the long production period between regeneration and harvest, a 1 per cent change in the discount rate may greatly change the optimal rotation age and possibly the sign of the land expectation value (Campbell and Dawson 1989). Variations in parameters may be separated into long-run trends and short-term stochastic ﬂuctuations around trends. From 1988 through 1999 risk and uncertainty, primarily stochastic stumpage price ﬂuctuations and catastrophic risks including ﬁre, high winds and disease, were analyzed in over half of the forestry economics papers published in the primary peerrefereed journals in which forestry economics is published. Before 1988 only a handful of papers in the primary forestry economics journals had analyzed risk and uncertainty (Newman 2002). Several approaches are used to address stumpage price risk and uncertainty. Many papers use a ‘stopping rule’ approach12 to determine when harvest should occur (Brazee and Mendelsohn 1988, Brock et al. 1979, Gong 1999, Haight 1991, Norstrom 1975, Lohmander 1987). An incomplete list of approaches (Brazee and Newman 1999) includes analysis of risky irreversible investments in multiple stand management (Dixit and Pindyck 1994, Yin and Newman 1999), life-cycle models with two ages and with or without overlapping generations (Amacher et al. 1999) and comparison of performance with other assets (Washburn and Binkley 1990, Heikkinen 1999).

270

Yearbook of environmental and resource economics

Given the inﬁnite or arbitrarily large dimensionality and the inherent dynamics of stochastic problems, static methods do not easily generate solutions. Dynamic optimization methods are preferred (Neher 1990). Dynamic controls are either open-loop or closed-loop controls. With an open-loop control, expected stumpage prices are used and the expected NPV would be maximized once at the beginning of the planning horizon. Information available after the beginning of the planning horizon is ignored. With a closed-loop control, known stumpage prices are used when possible. The expected NPV is re-maximized every period as the most recent stumpage price becomes known. With maximization each period, the path of control variables from closed-loop solutions always perform at least as well as the path of control variables from open-loop solutions. For better performance and to incorporate more information, asset sale and search models use closed-loop controls. The introduction of stochastic prices, and the use of a closed-loop dynamic program control requires reformulating the forest landowner’s problem as a dynamic rather than static problem. The landowner now maximizes the expected NPV of her land. Landowners become aware of stumpage prices as the prices are realized, i.e., the stumpage price at time t is ﬁrst known at time t. After the stumpage price t is known, the landowner decides whether to harvest or not harvest in period t. In a closed-loop formulation, the landowner’s objective may be stated as functional equations for every time t. The general form of the functional equation for the landowner’s objective is:

Maximize E(NPV(a,t)) f(P(t) ) [P(t)Q(a) E(NPV(0,t)]dp(t) R(a,t) R(a,t) BF(R(a,t) ) [E(NPV(a 1,t 1) ] (6.17) where a is stumpage age, B (1r)1, E() is the expectation operator, NPV(a, t) is the net present value of land occupied by age a stumpage at time t, NPV(0,t) is the value of bare land at time t,13 P(t) is stumpage price at time t, R(a,t) is the reservation price of age a stumpage at time t, f(P(t)) is the probability density function for stumpage prices at time t and F(P(t)) is the cumulative distribution function for stumpage prices at time t. The reservation price is the minimum price at which age a stumpage will be harvested at time t. The landowner maximizes the expected value of age a stumpage at time t. The integral term of the right-hand side of the landowner’s objective function is the expected value of harvest revenues and bare land when harvest occurs at time t. The integrand is the probability density for each stumpage price at time t multiplied by the expected beneﬁts from harvest at

The Faustmann face of optimal forest harvesting

271

the speciﬁed stumpage price. The reservation price for age a stumpage at time t is the lower limit of the integrand. If the stumpage price at time t is at or above the reservation price for age a stumpage at time t, harvest will occur. If the stumpage price at time t is below the reservation price for age a stumpage at time t, harvest will not occur. The second term of the right-hand side of the landowner’s objective function is the discounted expected value of not harvesting at time t. The discounted expected value of not harvesting at time t is the value of cumulative distribution function at the reservation price multiplied by the discounted expected value of a1 stumpage at time t 1. The expected NPV of age a stumpage at time t is maximized by diﬀerentiating (6.17) with respect to R(a, t) and setting the derivative equal to 0:14 E(NPV(a,t)) 0 R(a,t) f(R(a, t)) [ R(a,t)Q(a) E(NPV(0,t) ) BE(NPV(a 1,t 1))]

(6.18)

Dividing by f(R(a, t)) and rearranging provides: R(a, t) Q(a)E(NPV(0,t))BE(NPV(a 1, t1)

(6.19)

Stumpage will be harvested when the left-hand side of (6.19), the expected beneﬁts of harvesting, is greater than or equal to the right-hand side of (6.19), the discounted expected beneﬁts of not harvesting. The reservation price is the price at which the expected beneﬁts of harvesting equal the discounted expected beneﬁts of not harvesting. The beneﬁts of harvesting are stumpage price at time t multiplied by the volume of age a stumpage, plus the discounted expected value of bare land. The expected beneﬁt of delaying harvest is the discounted expected value of age a1 stumpage at time t1. Two general results similar to non-forestry asset sale and search models are: 1.

2.

In the absence of information-gathering costs but with a known distribution of stumpage prices, the landowner’s NPV under stochastic prices will be at least as large as the NPV under a non-stochastic mean price of the stumpage price distribution. Increasing the spread of the stumpage price distribution may increase or reduce expected age at harvest.

The simple stochastic model presented in this section is consistent with the Faustmann model. Previous papers have relaxed many of the assumptions of the model presented. Previous research lines include:

272

1.

2.

3.

Yearbook of environmental and resource economics

Early analyses of stochastic stumpage prices addressed optimal behavior on a single stand. Firms were assumed to be price takers facing a distribution of stumpage prices. Little attention was paid to how the ﬂuctuating stumpage prices were generated. Several later studies have aggregated all ﬁrms’ output into a timber supply function. Market equilibrium at every time t is determined by equating timber supply with an exogenously ﬂuctuating demand function. That is, stumpage demands rather than stumpage prices are stochastic (Gong 1999). The distribution of stumpage prices. Several authors have studied whether stumpage prices exhibit a random distribution or exhibit a distribution that is the sum of a random distribution plus a drift term that tends to draw stumpage prices back to a mean stumpage price. With a distribution that drifts to the mean stumpage price, the market is ineﬃcient and a single landowner may gain from behaving optimally. Without a drift term, the market is eﬃcient, and there are no gains from a reservation price approach (Abiltrup et al. 1997, Gong 1999, Gong and Löfgren 2005, Washburn and Binkley 1990).15 Shifts or ﬂuctuations in other parameters besides stumpage price. Heikkenen (1999) analyzes stochastic discount rates. If the stumpage price is held constant, then (6.17)–(6.19) collapse to the Faustmann model. With a constant stumpage price, NPV and reservation price for each age are constant over time. With the time argument dropped, (6.17) and (6.19) become:

Maximum NPV(a)arg max [R(a) Q(a)NPV(0), BNPV(a1)] or (6.17a) R(a) Q(a)NPV(0)B(NPV(a1)) (6.19a) Stumpage will be harvested the ﬁrst time (6.19a) is satisﬁed, that is, stumpage will be harvested when the marginal beneﬁts of harvesting are ﬁrst equal to the discounted marginal beneﬁts of not harvesting. Since stumpage price is held constant, (6.17a) and (6.19a) imply that the reservation price for age a stumpage price equals stumpage price. With a concave volume function, reservation prices decline monotonically with age. Substituting P for R(a), expanding NPV(a 1) to highlight the optimality of harvesting at age a1, and multiplying through by B1 1r gives: [PQ(a)NPV(0)] (1r)pQ(a1)NPV(0)

(6.19b)

Expanding and collecting terms implies: P[Q(a1) Q(a)]r[pQ(a)NPV(0)] which is identical to (6.3) with T replaced by a.

(6.19c)

The Faustmann face of optimal forest harvesting

273

Catastrophic Risk and Uncertainty Trees are subject to physical risks throughout their lifespan. Timber stands are subject to catastrophic risks, such as ﬁre, wind, animals and diseases, between regeneration and harvesting (Reed 1984, Routledge 1980, Thorsen and Helles 1998, Yin and Newman 1996). For most species in temporal or boreal forests, the period between regeneration and harvesting is measured in decades. The potential of catastrophic risks impacts both the optimal rotation period and the optimal level of silvicultural eﬀort. It is diﬃcult to make general statements regarding the impacts of catastrophic risks. The lack of general comments is due in part to the number of diﬀerent ways that catastrophic risks may impact incentives. The existence of catastrophic risks may reduce harvest age, to avoid future expected risks, or may increase harvest age, since the stand has survived risks to reach its current age. Similarly, risks may increase silvicultural eﬀort to reduce the rotation age, and hence the exposure to risks, or risks may reduce silvicultural eﬀort due to expected losses during the rotation. Several factors contribute to the range of impacts of catastrophic risk including: 1.

2.

3.

4.

5.

The probability of catastrophic risk may change with the age of the stand. Older stands are usually more vulnerable to ﬁre, wind and disease. Younger stands are usually more vulnerable to grazing. Salvage value of a stand varies after catastrophe. A crown ﬁre is destructive. Expected salvage values after a crown ﬁre are often low. Diseases and insects may take months or years to destroy a stand. Expected salvage values may be high. High winds may minimally damage a stand or may totally destroy it. Salvage values may be high or low. Landowner silvicultural practices that may reduce risk. Controlled burning and clearing a stand’s understory may decrease the probability of a stand-destroying ﬁre. Fencing may reduce grazing by excluding wildlife. Some patterns of regeneration and thinning reduce the risk of wind damage. Other landowners’ silvicultural practices that may reduce joint risk. If a catastrophic risk is a public bad, then the behavior of other landowners may impact the probability of catastrophic risk. Actions by several landowners often reduce the risk from ﬁre, insects and disease. Financial insurance may be available. Given long production periods ﬁnancial insurance has potentially signiﬁcant impacts.

274

Yearbook of environmental and resource economics

5. DYNAMIC FOREST MODELS, INVESTMENTS AND CREDIT CONSTRAINTS Timber Supply and Multiple Stand Management The Faustmann model is dimensionless. With an exogenous price the Faustmann model can be used to determine the optimal harvest age of an even-aged stand of any size ranging from a single tree to millions of hectares. This implies that in a multiple stand model, a simple Faustmann model can determine optimal harvest age, but cannot determine the impacts of harvest on timber supply. A timber supply function is needed to move from stand level analyses to market level analyses. Many presentations of the Faustmann model assume a Faustmann ‘normal forest’. Previous papers have demonstrated conditions under which a forest might converge to such a ‘forest’ (Mitra and Wan 1995, 1996, Salo and Tahvonen 2002a). A ‘normal forest’ has an equal land area in each age class. A Faustmann normal forest is one in which there are T* age classes, where T* is the optimal Faustmann rotation length. That is, a Faustmann normal forest has as many age classes as the optimal harvest age, and each age class occupies an equal land area. Rather than assuming the age class distribution, it is straightforward to aggregate over stands to determine aggregate timber supply (Berck 1979, Johannsson and Löfgren 1985). Aggregate timber supply at time t is the sum of timber harvested on every stand at time t. Let j represent the jth of the N stands that are harvested at time t. Aggregate timber supply at time N t is j1 Qj (t) .16 Similarly timber supply for a single landowner with multiple stands can be calculated by restricting the aggregation to the landowner. A landowner with multiple stands may expand the NPV function of (6.1) to include multiple stands.17 Let i represent the ith of M stands, that is, i 1, 2 . . . M. If the stands are managed independently and are of equal size, then the landowner’s NPV function becomes: Maximize w.r.t. T

NPV(Z,T)

M

L(ai, T) i0

er(Ta ) PQ(T) M

i

i0

er(Ta ) M

i

i0

erTPQ(T) C 1 erT

PQ(T) C 1 erT

(6.20)

where Z is a 1 by M vector of stand ages at time 0, L(ai,T) is the land expectation value of the ith stand with age ai stumpage at time 0, and

275

The Faustmann face of optimal forest harvesting

NPV(Z, T) is the NPV of all of the landowner’s M stands. Similar to the single stand model described in (6.1)–(6.5), with the assumption of constant parameters, harvest age, T, is constant for every rotation regardless of the initial age of the stumpage. The Faustmann harvest age is determined by taking the derivative of (6.20) with respect to T and setting it equal to zero. To minimize algebra, the expanded objective function in the ﬁrst line of (6.20) is used: M

dNPV(Z,T) T

d

L(ai,T) i0

dT

0

M

er(Ta ) i

i0

e rTPQ(T) C dPQ(T) dT 1 erT

rPQ(T) r

d

e rTPQ(T) C 1 e rT dT

(6.21)

The ﬁrst three terms on the right-hand side of (6.21), that is, those terms within both the curly and square brackets, are analogous to (6.5). That is, the marginal beneﬁts from delaying harvest equal the marginal costs of delaying harvest. The sum of these ﬁrst three terms equals zero at the Faustmann harvest age. By the deﬁnition of Faustmann’s rule, the last term within the curly brackets on the right-hand side of (6.21) equals 0 at the Faustmann harvest age. Since 000, these statements imply that condition (6.21) is satisﬁed at the Faustmann harvest age. This result holds for every stand i regardless of the initial age of stumpage. Several of the assumptions made in this section may be relaxed: 1.

2.

3.

The assumption of homogeneous land for all stands may be relaxed. With heterogeneous land, both timber volume functions and Faustmann harvest ages will diﬀer across stands (Bowes and Krutilla 1985). The assumption of one species in all stands may be relaxed. Similar to heterogeneous land with multiple species, both timber volume functions and Faustmann harvest ages will diﬀer across species (Jacobsen 2004). Multiple stands may generate economies of scale in the production of some forest goods and services. Wildlife habitat, water quality, forest preserves, management costs and harvesting costs are examples of

276

4.

Yearbook of environmental and resource economics

forest goods and services that often exhibit economies of scale. With economies of scale, it is no longer optimal to manage the stands independently. Often economies of scale cannot be realized on one landowner’s property. Several possible strategies for capturing economies of scale that can only be realized on multiple properties exist. These strategies are similar to managing for economies of scale for non-forest goods and services. First, landowners could choose to manage their land collectively. The joint objective function and resulting solution are similar to (6.20) and (6.21), but with many landowners rather than with one landowner.18 Second, a government could provide ﬁnancial incentives through a combination of taxes, subsidies or regulation to capture the economies of scale. The landowner’s objective function might resemble (6.10) or (6.12). If silvicultural eﬀort is included as a second choice variable, eﬃcient public policy design usually requires the use of two policy instruments (Baumol and Oates 1988). Third, a government or NGO could purchase the land and manage it jointly. Landowners, who maximize NPV, should sell if an oﬀer for the land at the end of rotation based on non-forestry uses exceeds the value from forestry uses. That is, if W L(0,T), the landowner should sell.

Dynamic Forest Models With the exception of the analyses of risk in section 4, the Faustmann model and the previously presented modiﬁcations address the optimal harvesting problem using a static framework. Given that timber harvesting is usually inherently dynamic, strong assumptions are needed to develop useful static frameworks. Stumpage prices, interest rates and costs experience time trends. The existence of time trends makes the construction of additional useful meaningful static optimization frameworks diﬃcult or impossible. Although dynamic models of stands or forests are quite appealing, theoretically deriving optimal dynamic controls is diﬃcult. One signiﬁcant problem is the inclusion of regeneration. When trees are regenerated, the new stand will be available ready for harvest at some future time. That is, current harvesting decisions inﬂuence available stumpage in the future. This decision-making connection between current and future harvests makes current harvest decisions a lagged control variable of future harvests. Lagged control variable problems are theoretically diﬃcult to solve (Kamien and Schwartz 1991). The diﬃculties in determining lagged optimal controls are increased with the existence of inﬁnite rotations, which imply an inﬁnite number of lagged control variables. 19

The Faustmann face of optimal forest harvesting

277

One approach that eases the technical diﬃculties of deriving dynamic optimal controls is to separate the harvesting and planting decisions. Several possible assumptions may justify this. The easiest approach is to ignore replanting altogether by either assuming costless natural regeneration, or by assuming that the landowner is only concerned with the current rotation and not with any future rotation. Given that the average length of landownership is less than 30 years (Kuuluvainen and Salo 1991) and most rotations in temperate or boreal forests are much longer than 30 years, forest stands are often sold or become an intergenerational bequest. If market price is either low or unresponsive to the biological state of the forest at sale, and if bequest motives are weak, then given the small value of future rotations to the landowner, she may not include future rotations when maximizing NPV. Another possible approach formally splits the harvesting and planting decisions into two separate processes, modeled with two control variables (Berck 1979, Brazee and Mendelsohn 1990). This split is plausible if replanting is viewed as an investment and if the landowner can buy or sell forest land over time. This approach contrasts with the initial inﬁnite rotation Faustmann model in which regeneration immediately follows harvest as described by (6.7), but is consistent with the more recent single rotation Faustmann model with land rents possibly from non-forestry uses described by (6.6). To illustrate this approach consider a dynamic model in which stumpage is divided into older growth and newer growth. Older growth is stumpage of any age that is alive at time 0; that is, the stumpage is alive before current planning starts. Newer growth is stumpage of any age that is regenerated after time 0; that is, the stumpage is regenerated during the planning period.20 The objective function for older growth is: N

Maximize NPV(H(t), X0, Z0 ) e rt [P(t)Y(am (t), t) VH]H(t)]dt 0 w.r.t. H(t) (6.22) where am(t) is the age of the oldest trees at time t, H(t) is the harvest rate of hectares of older growth harvested at time t, N is the end of the planning period, VH is the per hectare value in its forestry or non-forestry use after harvest, X0 is the initial number of hectares of older growth, Y(am(t), t) is the volume function for the oldest trees21 at time t and Z0 is the initial age class distribution on the initial hectare of older growth. The objective is to maximize the discounted net scarcity rent from harvesting older growth. The state equation for older growth is: dX(t) H(t) dt

(6.23)

278

Yearbook of environmental and resource economics

where X(t) is the stock of hectares of older growth at time t. In (6.23) the change in the stock of hectares of older growth equals the number of hectares of older growth harvested. Harvest rate, H(t), and the stock level, X(t), are assumed to be nonnegative. Applying optimal control theory (Kamien and Schwartz 1991), the manipulated necessary condition for optimal harvest of older growth for a continuous period, that is, H(t) 0 for a time interval, is:

dP(t) Y(a(t), t) da(t) Y(a(t), t) Y(a(t), t) P(t) t dt a(t) dt

r[P(t)Y(a(t),t) VH]

(6.24)

Equation (6.24) requires that continuous harvesting occur when the growth rate in forest value equals the discount rate. Although derived from a dynamic model, the interpretation of (6.24) is analogous to the interpretation of (6.4).22 The left-hand side of (6.24) is the marginal beneﬁts of delaying harvest, while the right-hand side is the marginal costs of delaying harvest. The marginal beneﬁts of delaying harvest are potential capital gains from a changing stumpage price plus tree growth and from changing the age of the oldest stumpage. The marginal costs of delaying harvest are the interest forgone on total revenue and cleared land. If the marginal beneﬁts of delaying harvest on the left-hand side of (6.24) are greater than the marginal costs of delaying harvest on the right-hand side of (6.24), then stumpage is not harvested. The stand is allowed to grow until the marginal beneﬁts of delay equal the marginal costs of delay. If the marginal beneﬁts of delaying harvest are less than the marginal costs of delaying harvest, then stumpage is harvested immediately. Dividing both sides of condition (6.24) by P(t)Y(a(t), t) VH gives:

dP(t) Y(a(t),t) da(t) Y(a(t),t) Y(a(t),t) P(t) t dt a(t) dt P(t)Y(a(t),t) VH

r

(6.25)

Condition (6.25) requires that harvesting occur when the growth rate in net forest beneﬁts equals the discount rate. The shape of the harvesting path determined by (6.25) describes a simple Hotelling extraction path (Hotelling 1931) with two additional components. First, the per hectare value of land rent, VH, is included. This contrasts with simple Hotelling models which have no storage costs of maintaining a stock in situ (Berck 1981, Vousden 1973). Second, individual

The Faustmann face of optimal forest harvesting

279

stands of older growth may increase in volume before harvest. This also contrasts with simple Hotelling models in which exhaustible resources do not increase in volume from stock growth. Condition (6.25) describes the shape of the optimal harvest path during an era in which older growth is continuously harvested. As with all singlestate-variable/single-control-variable optimal control problems, the solution is a two-point boundary problem which requires two additional pieces of information to determine the absolute level of the harvesting path. Z0, the initial age class distribution of older stumpage, is assumed to be known, which provides one of the needed pieces of information. The second piece of information may be provided by the terminal stumpage price. The terminal stumpage is either the intersection of the demand curve with the price axis or the production costs of a substitute for older growth. If the substitute is limitless, it is a backstop technology (Dasgupta and Heal 1979, Nordhaus 1973, Solow 1974). Here newer growth serves as a backstop technology. If the supply of land is inﬁnitely elastic, then the potential supply of newer growth is also inﬁnitely elastic. Assuming a competitive land market, the production costs of planted stumpage are determined by equating the soil expectation value and the value of land from the best non-forest alternative, that is, by equating L(0, T) and W, and then by solving for stumpage price. Expanding L(0, T) provides: erTP(T)Q(T) C W 1 e rT

(6.26)

Solving for stumpage price implies: P(T)

CerT W(erT 1) Q(T)

(6.27)

The costs of producing age T stumpage are the sum of the compounded regeneration costs and land rents forgone during a rotation of newer growth divided by harvest volume. Since the Faustmann harvest age maximizes the NPV of bare land, if newer growth is harvested at the Faustmann harvest age, then the production costs of newer growth are minimized. If newer growth is harvested before the Faustmann harvest age, then production costs of newer growth will be initially higher than if newer growth were harvested at the Faustmann harvest age. If regeneration is adequate, then the production costs of newer growth will fall until a minimum is reached at the Faustmann harvest age (Dasgupta and Heal 1979). If newer growth is not harvested until after the Faustmann harvest age, then regeneration should be delayed until the newer growth can be harvested at the

280

Yearbook of environmental and resource economics

Faustmann harvest age. With the second piece of information provided by (6.27), determining the level of the optimal harvesting path is possible. Forests in an Investment Portfolio For many landowners forests are often only one of several types of assets. Forest decisions may be incorporated into a more general decision-making process that includes other investments and assets. A simple way to simultaneously manage forest and ﬁnancial assets is by adding one new control variable, savings, and one new state variable, ﬁnancial assets, to the dynamic model in (6.22)–(6.23) (Kuuluvainen 1990, Tahvonen 1998, Tahvonen et al. 2001). To simplify the model, regeneration is ignored and older growth and newer growth are combined into one variable, stumpage. A concave utility function is added to reﬂect the need for a steady stream of consumption. The objective function of (6.22) becomes: NPV(H(t),Z, F0 ) Maximize w.r.t. H(t), S(t) N ertU [(M(t) S(t) (P(t)Y(am (t),t) VH)H(t)]dt 0

(6.28) where F(t) is the stock of the ﬁnancial asset at time t, F0 is the initial stock of the ﬁnancial asset, M(t) is the rate of exogenous income at time t, S(t) is the net savings rate at time t, and U() is an instantaneous utility function. The landowner maximizes discounted utility from consumption over her lifetime given initial wealth and initial forest stocks by controlling harvesting and net savings. With H(t) and X(t) redeﬁned for all stumpage rather than just older growth, the state equation (6.23) is retained. The stock of the ﬁnancial asset at time t is determined by the initial stock, the path over time of the net saving rate, and the discount rate. The state equation for the ﬁnancial asset is: dF(t) r(F(t) S(t)) (6.29) dt The change in the stock of the ﬁnancial asset at time t equals the sum of the interest accrued or owed on the ﬁnancial asset at time t and the net saving level at time t. Similarly to the previous model, H(t) and X(t) are assumed to be nonnegative. F(t) and S(t) are unconstrained and may be positive or negative. After applying optimal control theory and several steps of algebra, a useful condition emerges when H(t) 0:

281

The Faustmann face of optimal forest harvesting

Y(a(t),t) da(t) Y(a(t), t) dP(t) Q(a(t),t) P(t) t dt a(t) dt rP(t)Y(a(t),t) rVH

(6.30)

Condition (6.30) is analogous to (6.24) with H(t) and X(t) redeﬁned to include all stumpage and not just older growth. That is, the left-hand side of (6.30) is the marginal beneﬁts of delaying harvest, while the right-hand side is the marginal costs of delaying harvest. The marginal beneﬁts of delaying harvest are a sum of the ﬁnancial capital gain from a changing stumpage price and a physical capital gain from tree growth and from changing the age of the oldest stumpage. The marginal costs of delaying harvest are the interest forgone on total revenue and cleared land. Credit Constraints The Faustmann model assumes that the landowner has unlimited access to credit. In fact limits on a landowner’s ability to borrow are universal. Few are granted unlimited credit. A credit-constrained landowner could harvest in the absence of being able to borrow. Substituting harvesting for borrowing is called the ‘Volvo principle’ or the ‘Volvo theorem’ and provides alternative motivation for harvest than maximization of NPV. The Volvo theorem suggests that landowners harvest timber to meet non-forestry expenses such as the purchase of a new Volvo rather than choosing to harvest at the age that maximizes NPV (Johansson and Löfgren 1985). Credit constraints can be introduced into the dynamic model described by (6.26), (6.28) and (6.29). A simple debt ceiling would require that F(t) Fmin, where Fmin is a minimum asset level that describes a debt ceiling. Several authors have studied debt ceilings and other credit constraints (Tahvonen 1998, Tahvonen and Salo 1999, Tahvonen et al. 2001).23 These studies show that when landowners are credit constrained consumption and production are no longer separate, individual owner characteristics inﬂuence harvest decisions, and the impacts of shifting parameters become more complex.

6. UNEVEN-AGED MANAGEMENT AND SELECTED OMITTED ISSUES Many stands are even-aged. Some stands are uneven-aged; that is, the stand has trees of many ages. Species usually determines whether a stand is evenaged or uneven-aged. Species that are shade intolerant and need bare land to successfully regenerate are more likely to be even-aged, while species that

282

Yearbook of environmental and resource economics

are shade tolerant and able to regenerate through vegetation are less likely to be even-aged. Even-aged management is stressed because most species grown for commercial timber are even-aged. Recently, interest in uneven-aged management has increased. Within parts of the European Union, continuous cover management is legislatively mandated (Price 2003). The primary intent of continuous cover management is to manage uneven-aged forests. The impetus for continuous cover forests appears to be forest health. The focus on forest health of continuous cover management contrasts with the historical primary focus on timber harvesting under uneven-aged management. Another uneven-aged management regime strategy that focuses on short-term timber production is diameter limit cutting (DLC). Under DLC strategy all trees over a speciﬁed diameter are harvested simultaneously, while to the extent possible no tree under the speciﬁed diameter is harvested. The objective of DLC, similar to the Faustmann objective, is to maximize the NPV from timber sales. DLC management may actually reduce forest health by harvesting the large, healthy trees and not the small, relatively unhealthy trees. For further details on the maximization of NPV under DLC management, see Nautiyal (1983) and Chang (1982). By design this chapter has presented issues that are important in forestry. However, several important issues in forestry are also important in other ﬁelds of economics. In general these issues can be approached using concepts from basic economic theory or methods developed in other ﬁelds of economics. Some of the signiﬁcant omitted issues are: 1.

2.

3.

Sustainable forestry Recently some consumers and forest product ﬁrms have successfully demanded timber products from forests that are managed sustainably. A response to these demands has been the creation of several national and international organizations that certify timber and non-timber goods that have been produced by sustainable management practices. As with other green goods, there are interesting economic questions regarding incentives, monitoring and competing standards (Overdevest 2004). Trade in timber products The export of timber products such as log, lumber and pulp is an important component of several national economies including Canada, Finland, Indonesia, Norway and Sweden. Many other countries are large net importers of timber products. Since trade in timber products is a post-harvest activity, existing trade models may be used. Spatial analysis Historically, location has had a big role in forest management, and this importance should continue. Forests are often on lands that have small opportunity costs. The size and weight of logs

The Faustmann face of optimal forest harvesting

4.

5.

283

makes transporting them expensive and provides incentives for mills to locate in or near forested areas. Landowner motivation Although many studies have examined if and why small private landowners manage and/or harvest timber on their forested properties, explanations to date have not resulted in universally successful programs. Perhaps producer and/or consumer theory in conjunction with other methods may generate a better understanding of the motivations of small, private forest landowners (Kuuluvainen et al. 1996). Financial economics and risk and uncertainty As described in section 4, the majority of peer-refereed papers in forest economics journals during the period 1988–2000 have addressed some form of risk or uncertainty. To date advances in the application of dynamic optimization techniques to ﬁnancial assets have not been fully incorporated into forest economics studies. The further incorporation of these techniques into forest economics studies should enhance the quality of economic analyses of forest management.

NOTES *

1. 2. 3. 4. 5. 6.

7. 8. 9.

I am indebted to Henk Folmer, Karl-Gustav Löfgren and Tom Tietenberg for comments that signiﬁcantly improved the chapter. Any remaining errors are mine. I thank Henk Folmer and Tom Tietenberg for their truly extraordinary patience in waiting to receive an extremely late manuscript. A stand is a land area managed as one unit for trees. A stand is the most common forest management unit. I ask all authors whose research was omitted or under represented to accept my profuse apologies. Stumpage is literally ‘timber on the stump.’ Often timber harvesters contract to purchase a stand’s stumpage. Bare land is short for bare mineral soil, land without any vegetative cover. Most evenaged species require bare mineral soil to regenerate. If a landowner, who maximizes NPV, voluntary regenerates a stand, then there exists age s, such that the NPV from the stand is greater than the opportunity cost of the land. Note in this, the Faustmann model with constant parameters and the initially bare land, the landowner makes a one-time decision whether the land is to be forested or nonforested in perpetuity. In more complex Faustmann models including those with either shifting or stochastic parameters, it may be optimal to shift land between forest and other uses. If E0, then C(E) 0, and the stand is regenerated naturally. If labor supply is inﬁnitely elastic, d 2 C(E)/dE2 0, regeneration costs are linear and equal the marginal cost of silvicultural eﬀort multiplied by level of silvicultural eﬀort. For convenience the Jacobian matrix is assumed to be negative semi-deﬁnite. Pre-commercial thinnings are speciﬁc to species and climate, and will not be explicitly modeled. An easy way to implicitly include pre-commercial thinnings is to include additional silvicultural eﬀort for thinning, discount the net costs of the pre-commercial thinnings to time zero and include these costs in C(E). Commercial thinnings can be included

284

10.

11. 12. 13. 14.

15. 16.

17. 18. 19. 20.

21. 22. 23.

Yearbook of environmental and resource economics as net revenues received during the rotation. Naslund (1969), Brazee and Bulte (2000), and Lu (2004) provide more developed models of thinning. The discussion of taxes and subsidies in this section is incomplete. A complete discussion is beyond the simple models of this chapter and would require several additional pages. Some important omissions are: ● The design of optimal self-ﬁnancing forestry taxes (Amacher 1999, Amacher and Brazee 1997). ● Tax avoidance strategies with stochastic stumpage prices (Thorsen 1999). ● Two-period rather than rotational analyses of a broad set of taxes (Koskela 1989a, 1989b). Interest in forest management on the rural–urban interface is an example of signiﬁcant externalities from the maintenance of forested and other wildlands (Alavapati et al. 2005). A stopping rule is a set of conditions under which it is optimal to terminate a process. In forest economics stopping rules are used to determine when to harvest, that is, when to stop waiting to harvest, and when to invest. E(NPV(0,t)) CBE(NPV(1, t 1)). E(NPV(1, t 1) follows from (6.17) with a 1, and t1 replacing t on the right-hand side of the condition. Note that the ﬁrst two terms of (6.18) follow from the application of the fundamental theorem of calculus, that is, by diﬀerentiating the integral in (6.17) with respect to its lower endpoint. The third term follows from diﬀerentiating the cumulative distribution function F(R(a, t)) with respect to its argument, which produces the probability density function f(R(a, t)). Recent work suggestions that potential gains from optimal behavior are negligible with demand ﬂuctuations. A noteworthy comparative static result is that increasing stumpage price usually reduces stumpage volume at harvest, that is, the short-run per curve is backward-bending (Binkley 1987, Clark 1990). If land is able to enter and exit forestry through a competitive land market, then the long-run aggregate supply curve may be upward-sloping throughout its entire range as more land is devoted to forestry when the stumpage price increases. The approach in this section is simplistic. More sophisticated approaches are found in Berck (1979, 1981), and Johannson and Löfgren (1985). The extension to multiple landowners can be accomplished by identifying each stand with its owner. Heaps (1981, 1984) and Salo and Tahvonen (2002a, 2002b) have solved more sophisticated models some with inﬁnite rotations. Older growth and newer growth are used rather than old and new growth because of the perceived restrictiveness of the latter terms. Old growth often refers to a mature stand, whose volume is not increasing due to slow growth and natural mortality. New growth usually refers to a very young stand. Here older and newer growth deﬁne whether the stumpage was regenerated before or during the planning period. With a concave volume function, it is easy to show it is always optimal to harvest the oldest trees ﬁrst. If parameters are held constant over time, (6.22)–(6.24) derive the Faustmann rule. Other authors have studied credit constraints in two-period life-cycle models (Kuuluvainen 1990, Kuuluvainen and Salo 1991).

REFERENCES Abiltrup, J., J. Riis and B.J. Thorsen (1997), ‘The reservation price approach and informationally eﬃcient markets’, Journal of Forest Economics 3, 229–45.

The Faustmann face of optimal forest harvesting

285

Alavalapati, J.R.R., D.R. Carter and D.H. Newman (eds) (2005), ‘Economic and policy perspectives of the wildland–urban interface’, Forest Policy and Economics (special issue), 7 (5). Amacher, G.S. (1999), ‘Government preferences and public harvesting’, American Journal of Agricultural Economics 81, 15–28. Amacher, G.S. and R.J. Brazee (1997), ‘Designing forest taxes with varying government preferences and budget targets’, Journal of Environmental Economics and Management 32, 323–40. Amacher, G.S., R.J. Brazee and T.A. Thomson (1991), ‘The eﬀect of forest productivity taxes on timber stand investment and rotation length’, Forest Science 37, 1099–118. Amacher, G.S., R.J. Brazee, E. Koskela and M. Ollikainen (1999), ‘Taxation, bequests, and short and long run timber supplies: an overlapping generations problem’, Environmental and Resource Economics 13, 269–88. Avery, T.E. and H. Burkhart (2001), Forest Measurements, San Francisco: McGrawHill. Baumol, W.J. and W.E. Oates (1988), The Theory of Environmental Policy, Cambridge: Cambridge University Press. Bentley, W.R. and D.E. Teeguarden (1965), ‘Financial maturity: a theoretical review’, Forest Science 11, 76–87. Berck, P. (1979), ‘The economics of timber: a renewable resource in the long run’, Bell Journal 10, 447–62. Berck, P. (1981), ‘Optimal management of renewable resources with growing demand and stock externalities’, Journal of Environmental Economics and Management 8, 105–17. Binkley, C.S. (1987), ‘When is the optimal economic rotation longer than the maximum sustained yield’, Journal of Environmental Economics and Management 14, 152–8. Bowes, M.D. and J.V. Krutilla (1985), ‘Multiple use management of public forestlands’, in A.V. Kneese and J.L. Sweeney (eds), Handbook of Natural Resource and Energy Economics, Amsterdam: North-Holland. Brazee, R.J. and E. Bulte (2000), ‘Optimal harvesting and thinning with stochastic prices’, Forest Science 46, 23–31. Brazee, R.J. and R. Mendelsohn (1988), ‘Timber harvesting with ﬂuctuating prices’, Forest Science 34, 359–72. Brazee, R.J. and R. Mendelsohn (1990), ‘A dynamic model of timber markets’, Forest Science 36, 255–64. Brazee, R.J. and D.H. Newman (1999), ‘Observations on recent forest economics research on risk and uncertainty’, Journal of Forest Economics 5, 193–200. Brock, W., M. Rothschild and J. Stiglitz (1979), Stochastic Capital Theory, Madison, WI: Department of Economics, University of Wisconsin. Calish, S., R.D. Fight and D.E. Teeguarden (1978), ‘How do non-timber values aﬀect Douglas-ﬁr rotations?’, Journal of Forestry 76, 217–22. Campbell, G.E. and J.O. Dawson (1989), ‘Growth, yield, and value projections for black walnut interplantings with black alder and autumn olive’, Northern Journal of Applied Forestry 6, 129–32. Chang, S.J. (1981), ‘Determination of the optimal growing stock and cutting cycle for an uneven-aged stand’, Forest Science 27, 739–44. Chang, S.J. (1982), ‘An economic analysis of forest taxation’s impact on optimal rotation age’, Land Economics 58, 310–23.

286

Yearbook of environmental and resource economics

Chang, S.J. (1983), ‘Rotation age, management intensity and the economic factors of timber production: do changes in stumpage price, interest rate, regeneration cost and forest taxation matter?’, Forest Science 29, 267–78. Clark, C.W. (1990), Mathematical Bioeconomics: The Optimal Management of Renewable Resources, 2nd edn, New York: Wiley & Sons. Dasgupta, P.S. and G.M. Heal (1979), Economic Theory and Exhaustible Resources, Cambridge: Cambridge Economic Handbooks. Dixit, A.K. and R.S. Pindyck (1994), Investment Under Uncertainty, Princeton: Princeton University Press. FAO (2000), ‘Global forest resources assessment 2000: main report’, FAO Forestry paper. Faustmann, M. (1849 [1995]), ‘Calculation of the value which forest land and immature stands possess for forestry’, Journal of Forest Economics 1, 7–44. Gaﬀney, M.M. (1957), Concepts of Financial Maturity of Timber and Other Assets, Raleigh, NC: North Carolina State University Agricultural Economics Information Series No. 62. Gong, P. (1999), ‘Optimal harvest policy with ﬁrst-order autoregressive price process’, Journal of Forest Economics 5, 413–39. Gong, P. and K.-G. Löfgren (2005), ‘Market and welfare implications of adaptive harvest strategy’, presented at the conference ‘Faustmann and the Optimal Stopping Time in Forestry and Beyond’, Baton Rouge, LA, 21 April. Haight, R.G. (1991), ‘Feedback thinning policies for uneven-aged stand management with stochastic prices’, Forest Science 36, 1015–31. Hardie, I.W., J.N. Daberkow and K.E. McConnell (1984), ‘A timber harvesting model with variable rotation lengths’, Forest Science 30, 511–23. Hartman, R. (1976), ‘The harvesting decision when the standing forest has value’, Economic Inquiry 14, 52–8. Heaps, T. (1981), ‘The qualitative theory of optimal rotations’, Canadian Journal of Economics 14, 686–99. Heaps, T. (1984), ‘The forestry maximum principle’, Journal of Economic Dynamics and Control 7, 131–51. Heaps, T. and P.A. Neher (1979), ‘The economics of forestry when the rate of harvest is constrained’, Journal of Environmental Economics and Management 6, 297–319. Heikkenen, V.-P. (1999), ‘Cutting rules for ﬁnal fellings: a mean variance portfolio analysis’, Journal of Forest Economics 5, 269–84. Hotelling, H. (1931), ‘The economics of exhaustible resources’, Journal of Political Economy 39, 137–75. Hyytiäinen, K. and O. Tahvonen (2002), ‘The eﬀects of legal restrictions and recommendations in timber production: the case of Finland’, Forest Science 47, 443–54. Jacobsen, J.B. (2004), Economic Aspects of Uneven-aged Mixed Species Forestry, dissertation, Copenhagen: KVL. Johansson, P.O. and K.-G. Löfgren (1985), The Economics of Forestry and Natural Resources, New York: Basil Blackwell. Kamien, M.I. and N.L. Schwartz (1991), Dynamic Optimization: The Calculus of Variations and Optimal Control in Economics and Management, 2nd edn, Amsterdam: North Holland. Klemperer, W.D. (1983), ‘Ambiguities and pitfalls in forest productivity taxation’, Journal of Forestry 81, 16–9.

The Faustmann face of optimal forest harvesting

287

Koskela, E. (1989a), ‘Forest taxation and timber supply under price uncertainty: perfect capital markets’, Forest Science 35, 137–59. Koskela, E. (1989b), ‘Forest taxation and timber supply under price uncertainty: credit rationing in capital markets’, Forest Science 35, 160–72. Kuuluvainen, J. (1990), ‘Virtual price approach to short-term timber supply under credit rationing’, Journal of Environmental Economics and Management 19, 109–26. Kuuluvainen, J. and J. Salo (1991), ‘Timber supply and life cycle harvest of nonindustrial private forest owners: an empirical analysis of the Finnish case’, Forest Science 37, 1011–29. Kuuluvainen, J., H. Karppinen and V. Ovaskainen (1996), ‘Landowner objectives and nonindustrial private timber supply’, Forest Science 42, 300–309. Lohmander, P. (1987), The Economics of Forest Management Under Risk, dissertation, Umea, Sweden: Swedish University of Agricultural Sciences. Löfgren, K.-G. (1983), ‘The Faustmann–Ohlin theorem: a historical note’, History of Political Economy 15, 261–4. Lu, F. (2004), Optimization of Forest Management Decision Making Under Conditions of Risk, dissertation, Umea, Sweden: Swedish University of Agricultural Sciences. McConnell, K.E., J.N. Daberkow and I.W. Hardie (1983), ‘Planning timber production with evolving prices and costs’, Land Economics 59, 292–9. Mitra, T. and H.Y. Wan, Jr. (1985), ‘Some theoretical results on the economics of forestry’, Review of Economic Studies 52, 263–82. Mitra, T. and H.Y. Wan, Jr. (1986), ‘On the Faustmann solution to the forest management problem’, Journal of Economic Theory 40, 229–49. Naslund, B. (1969), ‘Optimal rotation and thinning’, Forest Science 15, 446–51. Nautiyal, J.C. (1983), ‘Towards a method of uneven-age forest management based on the theory of ﬁnancial maturity’, Forest Science 29, 47–59. Neher, P.A. (1990), Natural Resource Economics: Conservation and Exploitation, Cambridge: Cambridge University Press. Newman, D.H. (2002), ‘Forestry’s golden rule and the development of the optimal forest rotation literature’, Journal of Forest Economics 8, 5–28. Newman, D.H., C.E. Gilbert and W.F. Hyde (1985), ‘The optimal forest rotation with evolving prices’, Land Economics 61, 347–53. Nordhaus, W.D. (1973), ‘The allocation of energy resources’, Brookings Papers 3, 529–79. Norstrom, C.J. (1975), ‘A stochastic model for the growth period decision in forestry’, Swedish Journal of Economics 77, 329–37. Ohlin, B. (1921 [1995]), ‘Concerning the question of the rotation period in forestry’, Journal of Forest Economics 1, 89–114. Overdevest, C.A. (2004), ‘Codes of conduct and standard setting in the forest sector: constructing markets for democracy?’, Industrial Relations/Relations Industrielles, 59, 172–98. Pressler, M.R. (1860 [1995]), ‘For the comprehension of net revenue silviculture and the management objectives derived thereof’, Journal of Forest Economics 1, 45–88. Price, C. (2003), ‘The economics of transformation from even-aged to uneven-aged forestry’, in F. Helles, N. Strang and L. Wichmann (eds), Recent Accomplishments in Applied Forest Economics Research, the Netherlands: Kluwer Publications. Reed, W.J. (1984), ‘The eﬀects of risk of ﬁre on the optimal rotation of a forest’, Journal of Environmental Economics and Management 11, 180–90.

288

Yearbook of environmental and resource economics

Routledge, R.D. (1980), ‘The eﬀect of potential catastrophic mortality and other unpredictable events on optimal forest rotation policy’, Forest Science 26, 389–99. Salo, S. and O. Tahvonen (2002a), ‘On equilibrium cycles and normal forests in optimal harvesting of tree vintages’, Journal of Environmental Economics and Management 44, 1–22. Salo, S. and O. Tahvonen (2002b), ‘On the optimality of a normal forest with multiple land classes’, Forest Science 48, 530–42. Samuelson, P. (1976), ‘Economics of forestry in an evolving society’, Economic Inquiry 14, 466–92. Scorgie, M. and J. Kennedy (1996), ‘Who discovered the Faustmann condition?’, History of Political Economy 28, 77–80. Shepard, W. (1925), ‘The bogey of compound interest’, Journal of Forestry 23, 251–9. Solow, R.M. (1974), ‘Intergenerational equity and exhaustible resources’, Review of Economic Studies 41, 29–46. Southgate, D. (1990), ‘The causes of land degradation along “spontaneously” expanding agricultural frontiers in the Third World’, Land Economics 66, 93–101. Strang, W.J. (1983), ‘On the optimal forest harvesting decision’, Economic Inquiry 21, 576–83. Swallow, S.K., P.J. Parks and D.N. Wear (1990), ‘Policy-relevant nonconvexities in the production of multiple forest beneﬁts’, Journal of Environmental Economics and Management 19, 264–80. Tahvonen, O. (1998), ‘Bequests, credit rationing and in situ values in the Faustmann–Pressler–Ohlin forestry model’, Scandinavian Journal of Economics 100, 781–800. Tahvonen, O. and S. Salo (1999), ‘Optimal forest rotation with in situ preferences’, Journal of Environmental Economics and Management 37, 106–28. Tahvonen, O., S. Salo and J. Kuuluvainen (2001), ‘Optimal forest rotation and land values under a borrowing constraint’, Journal of Economic Dynamics and Control 25, 1595–627. Thorsen, B.J. (1999), ‘Progressive income taxes and option values: the case of a farmer who owns a forest’, Journal of Forest Economics 5, 217–34. Thorsen, B.J. and F. Helles (1998), ‘Optimal stand management with endogenous risk of sudden destruction’, Forest Ecology and Management 108, 287–99. Ulrich, A.H. (1981), U.S. Timber Production, Trade, Consumption and Price Statistics 1950–80, USDA miscellaneous publication no. 1408. Vincent, J.R. and C.S. Binkley (1993), ‘Eﬃcient multiple-use forestry may require land-use specialization’, Land Economics 69, 370–76. Vousden, N. (1973), ‘Basic theoretical issues of resource depletion’, Journal of Economic Theory 6, 126–43. Washburn, C.L. and C.S. Binkley (1990), ‘Informational eﬃciency of markets for stumpage’, American Journal of Agricultural Economics 72, 394–405. Yin, R. and D.H. Newman (1996), ‘The eﬀect of catastrophic risk on forest investment decisions’, Journal of Environmental Economics and Management 31, 186–97. Yin, R. and D.H. Newman (1999), ‘A timber producer’s entry, exit and other related decisions under uncertainty’, Journal of Forest Economics 5, 305–20.

Index abatement costs 107 abatement eﬀects emission taxes 4–5 heterogeneity in 18 abatement technologies 129–32, 125–7, 129–30, 143–4, 160 ABC policy, Netherlands 241–2 abstraction impacts 52 Active 46 aggregation procedures 58–60, 64–5 agricultural economic studies 47 Agricultural Nonpoint Source Pollution (AGNPS) model 62 air pollution, transport 217–20 air quality patterns 53–4 air travel 212, 214, 223, 224–5, 244 airport noise 216–17, 234 Alberta forested habitats 56 non-designated camping areas 77, 78 allocation studies 62 Amazon, deforestation 260 ambient pollution 21 Anhui province, China 102 annual income 13, 14–15, 16, 19 Anselin, Luc 77 anti-trust 197 Arc/Info 44, 49 ArcView 44, 49 Argentina, pollution control 103 Asia, pollution control 95, 99–100, 104, 111–12 auctioned emission permits 15–16, 28, 31, 32, 132, 185, 187–8 Australia, pollution control 95–6, 99 Australian Household Expenditure Surveys 13–14 benchmark standards, pollution 100, 106, 107 benefit function transfer, GIS 65–73

Bertrand competition 12, 121–2, 123–4, 174, 185 Bertrand Duopoly 153–5 Bertrand–Edgeworth competition 154, 174 Bertrand–Nash equilibrium 154 binary contiguity/weights matrices 81–3 binding land constraint 257 biofuels 209, 227–8 biophysical data, integration of 73 Birmingham, modelled emissions 55 border trade restrictions 244 Boston, water treatment 18 boundary problems, spatial overlays 64 box plot 75 Brazil, alternative fuels 227–8 buﬀering 60–61 calibrated analytical models 16 California, vehicle emissions experiment 14–15 Canada income/pollution study 20 pollution control 94, 99 capital intensive industries 10 capitalization, housing prices 24 carbon dioxide 54 permit trading 188 tax 14, 134 vehicle emissions 209, 213, 217, 220–22, 227–8, 246 carbon monoxide 124, 218 carbon taxes, empirical studies 11, 12–14 Cardiﬀ, recreation demand values 71, 72 cartography 50–51 case studies combining GIS functions 65–73 hedonic price models 83–6 289

290

Yearbook of environmental and resource economics

catalytic converters 212, 237 census data 19, 44, 58, 68 chemical industry 124, 129 Chile, pollution control 103 China emissions 213 pollution control 95, 99, 102, 103, 104, 108, 113 chlorofluorocarbons (CFCs) 219 Clean Air Act (1970) 25 amendments (1990), US 16, 17 clean inputs/technology dynamic approach 152 market power 179–80 monopsony power 177–8 Clean Water Act (1972) 17–18 clusters assessment of 56–7, 60 testing for 76–7 coal plants 10 coal taxes 13 Coase theorem 137 Cobb–Douglas preferences 151 colour schemes, maps 50–51 combustion engine 212, 227, 228 command-and-control regulations 8–9 empirical studies 16–18 commuting time 245–6 compensated demand curves 7 compensation, traﬃc pollution 233–4, 247 competitive firms, regulation of 134–5 compliance enforcement 184–5 pollution 101, 107 tradable permits 198 computational general equilibrium model 12, 16 conceptual issues, pollution control policies command-and-control regulation 8–9 complicating factors 9–10 measurement issues 5–7 pollution taxes 2–5 tradable emissions permits 7–8 congestion charges 238 reduction 245–6 connectivity functions, GIS 61–2, 67–8

constant travel time budget 245 Consumer Expenditure Survey (CES), US 5, 11–12 consumer preference-based values 229–30 consumer preferences 125 consumption activities, environmental eﬀects of 208–10 consumption eﬀects of taxation 2–3, 31 contagion eﬀects 73 contingent valuation (CV) studies 23, 52, 58–9 conventional pollutants 105–6, 107 conventional regulatory domain 109–10, 113 cooperative environmental policy 190–91 copyright issues 47 corporate average fuel economy (CAFE) 239–40 cost of illness method 229–30 cost–benefit analysis, environmental eﬀects of transport 232–3, 247 see also environmental cost–benefit analysis county-based datasets 44–5 Cournot competition 121–4, 151, 166–7, 172–4, 179, 183, 189–95, 196 Cournot Duopoly 152, 170–73 Cournot oligopoly 183–4, 195 comparative statics 140–41 emission taxes cum subsidy system 145–7 emission taxes with free entry 158–61 general framework 139–40 overview 138–9 second-best taxation 141–4 special cases 142–4 standards 147–52 Cournot–Nash equilibrium 125, 147 credit constraints, landowners 281 crop yields 73–4 CSERGE 52–3 customized data 45 Czech Republic, pollution control 95, 99

Index data acquisition 45–8, 67–8 Data Depot, The 46 data processing 49, 67 Data Store, The 46 decay functions 83 decentralization under perfect competition 128–9 demand elasticities, products 7 demand, influence of emissions 136–7 demand side factors, transport 224–6 Denmark, taxation 14 development, regional diﬀerences 108 diameter limit cutting (DLC), timber 282 diesel engines, fuel eﬃciency of 227 diﬀerential valuation of environment 23–4 diﬀerentiated commodities 155–7 price competition 185 digital elevation model (DEM) 57–8 digitalizing 47 dioxin toxicity levels 47 direct incentives 234–7 direct taxes 247 dirty technology 152 disclosure-related learning 97, 105 distance functions, GIS 59–61, 67–8 distance-based contiguity approach 81–2 distributional eﬀects, pollution control policies beneficiaries 19–25 conceptual issues 2–10 empirical studies 10–19 cost–benefit analysis 26–30 overview 1–2 research directions 31–2 diversity indices 56–7 Dixit–Stiglitz model 124, 157, 162–3, 195 eﬀect of tax rate increase 165 outline 163–5 second-best optimal tax rate 166–7 dominant firms with competitive fringe 152 durable assets, forests as 256–8 durable goods, taxation of 137, 150 dynamic forest models 276–80 dynamic model with accumulating pollutants 150

291

Ebert’s model 144 eco-dumping 123, 132 econometric models 15 economic data, integration of 73 economic development, eﬀect on emissions 212–13 economic eﬃciency costs, distributional constraints 28 economies of scale, forest goods and services 275–6 economy-wide factor price changes 10 EcoWatch, Philippines 95, 99, 101–2, 103, 104 EFTEC 52 electric vehicles 228 electricity generation 9–10 electricity taxes 13 emissions aﬀecting consumer demand 136–7 comparisons of 212–13, 221–6 eﬀects of public disclosure 96–7 fixed targets for reduction 214 incentives for polluters 97, 98, 101, 107, 113, 234–9 long-run development of 226 proportional to output 130, 160 regulation of 133–4 emissions inventories 94–6, 98–9, 104 common problems 105–7 emissions permit market, market power extensions of Hahn model 182–5 framework 172 market power through innovation 185 relations with oligopolistic output markets 175–6 results from experimental studies 186–8 trading model with large/small firms 180–82 emissions permits empirical studies 15–16 see also auctioned permits; grandfathered permits; tradable permits emissions standards 185 absolute standards 132, 147–8, 161 oligopolies with free entry 161–2 relative standards 132–3, 148, 161–2 transport 239–40

292

Yearbook of environmental and resource economics

emissions taxes 4–5, 28, 31, 32 basic system 145 Bertrand Duopoly 153–5 Cournot oligopoly with free entry 158–61 diﬀerentiated 152, 156 dominant firm with competitive fringe 152 durable goods 150 dynamic approach 150, 152 endogenous market structure 149 extensions 146–7 and financial structure of firms 151–2 inter-firm externalities 149 pre-investment 149–50 price-setting duopoly 155–7 reforms under imperfect competition 150–51 uniform 157–8 emissions technology 226–7 empirical studies, pollution control policies command-and-control regulations 16–18 comparison across instruments 18 emissions permits 15–16 environmental taxes 11–15 overview 10–11 summary 19 employer-provided parking 242–3 endogenous market structure 149 endogenous numbers of firms 195 energy eﬃciency, transport 211–12 energy sector 138 energy taxes, empirical studies 11–14 energy use taxes 214 transport 210–12 environmental agencies, requirements on 108–9 Environment Agency, UK 58–9 environmental cost–benefit analysis conclusions 30 distribution-neutral analysis 29–30 distributional constraints 27–8 equity implications of 54 social welfare function/distributional weights 26–7 environmental data 47

environmental demand schedule 105 environmental eﬀects of transport 210–21, 246–7 environmental improvements distribution of 22–3 welfare relationship 23–4 Environmental Information Disclosure System, Hanoi 99 environmental justice literature 19, 23 environmental performance ratings 99–105, 112 common problems 105–9 environmental policies, beneficiaries of environmental improvements distribution 22–3 environmental risks distribution 19–21 improvements/welfare relationship 23–4 integrated cost-benefit studies 25 summary 25 environmental policy for Cournot oligopoly 138–52 environmental quality measures of 20–21 over time 22 environmental risks distribution findings 20 measures of environmental quality 20–21 environmental supply schedules 96, 98–9,106 environmental taxes, empirical studies 10–19 equilibrium pollution model 96–8, 99, 104–5 equity problem 232–3 ERDAS IMAGINE file formats 49 ethanol 227 ethnic distribution of pollutants 53–4 Euclidean straight lines 60 Europe contingent valuation (CV) studies 23 kilometre charges 238 passenger transport 227 road transport 214 European Union emissions 213, 214, 220 energy consumption 210–11

Index forest management 282 pollution control 95, 99, 105, 112–13 even-aged management 258–60 exclusionary manipulation, permits market 186–7 exploratory spatial data analysis (ESDA) 74–7 exports, emission taxes as subsidies 189 externalities 266 factor markets 197–8 farms, spread of disease between 56 Faustmann model basic model 258–60 as foundation, optimal forest harvesting 261 overview 255–6 see also optimal forest harvesting Faustmann, Martin 258 final products household expenditure on 5 taxes on 19 financial compensation, traﬃc pollution 233 financial structure, firms/emissions taxes 151–2 firm size and market power 186–7, 195 first-best allocations 127–8 first-best pricing 239 first-generation performance rating systems 105–6, 107 fixed costs of transport 215–16 fixed targets, emissions 214 fixed taxes 239 ‘flythroughs’ 53 forest amenities 60 forest layers 57 Forestry Commission, UK 66 forests in investment portfolios 280–81 formally regulated pollutants 98, 100 fossil fuels 227–8 fragmentation indices 56–7 FRAGSTATS 56 free data 45 free entry 122, 158–61,196 free trade agreements, environmental provisions 110–11 freight transport 222–3, 238

293

Friedman–Modigliani income and lifecycle models 6 gamma indexes 61 gasohol 227–8 gasoline taxes calculation of optimal 29 empirical studies 11–12 Gaussian dispersion of pollutants 54 Geary’s c statistics 76 Generalized Cliﬀ–Ord weights 83 GeoDa software 76–7 geographic data sources 45–8 Geographic Data Technology 46 geographical information systems (GIS) overview of features 44–73 spatial statistical analyses and modelling 73–86 Geography Network 46 Geolytics website 45, 46 GEOWorld Data Directory 46 global positioning system (GPS) devices 45 globalization of regulatory product diﬀerentiation 110 government control costs 17 government mandated disclosure programs 93–4 government regulation, benefits of 111 grandfathered permits 8, 9, 15–16, 18, 19, 31, 132, 170, 182 gravity models 61 GreenWatch, China 95, 99, 102, 103, 104, 108, 113 habitat connectivity 61 Hahn permit trading model 122, 124, 186, 198–9 intertemporal permit trading 184 market power/non-compliance 184–5 output market 182–4 harvest age regulations 267–9 hazardous materials, transportation of 221 hedonic pricing models 56, 80 pricing 60, 77 studies 51

294

Yearbook of environmental and resource economics

hedonistic wage analyses/surveys 24 Herfindahl index 152 Hicks–Kaldor criterion 29, 232–3 high-density environments, transport in 240 Hohhot, Inner Mongolia 102, 103 homogeneous land quality 266–7 Hotelling rule 261, 278–9 household expenditure, final products 5 household income measurement 6 household pollution control costs 17 household production approach 229–30 household welfare changes 7 housing, capitalization in 24, 25 How to Lie with Maps 51 hub-and-spoke network structures, aviation 223 hybrid electric vehicles 228, 237 hydrocarbons emissions 218 hydrogen 228–9 imagery processing technology 47 impedance-based routing 61–2 imperfect competition basic assumptions 124–9 Cournot oligopoly 138–52 emission taxes in price-setting duopoly 153–7 environmental policy 188–95 market power in input markets 176–80 market power in permit market 180–88 monopolistic competition/free entry 157–69 monopoly 129–38 overview 120–24 permit trading in oligopoly 170–76 research directions 196–9 implicit prices method 229–30 incidence studies 26, 31–2 income elasticity, WTP 23–4, 30 income tax, compensation via 29 income-based recycling 12 India, pollution control 95, 96, 99 indirect incentives 234–9 Indonesia, pollution control 95, 99–102, 103, 104, 107, 108

industrialized countries, emissions inventories 105 industrializing countries, emissions inventories 95 industry control costs 17 industry-specific pollutants 124, 129 information requirements, public disclosure programs 108–9 infrastructure eﬀects of transport 215–16 inland navigation transport 214, 222, 244 Inner Mongolia 102 innovation, market power through 185 input markets, market power in markets for clean inputs/technology 179–80 mixed structures 179 monopsony power, polluting inputs 176–7 oligopsonies 178–9 second-best analysis 177–8 input–output tables 13, 14, 17 inside rating processes, credibility of 108 institutional change, transport 243–5 inter-firm externalities 149 intermediate products, taxes on 19 international standards, emissions 100, 107 international trade, imperfect competition in 188–90 cooperative environmental policy 190–91 endogenous numbers of firms 195 non-cooperative environmental policy 191–5 Internet data 45–6 virtual regulation 110–11 intertemporal permit trading 184 inverse demand functions 125 inverse weighting routines 60, 68 investment portfolios, forests in 280–81 ISO 14000 100, 107 iso-elastic functions 26–7 isochrone maps 62, 63

Index Japan emissions 213 pollution control 95, 99 Jiangsu province, China 102 joint count statistic for nominal data 76 journeys, origins/destinations of 223–4 Kennet (river), England 58 Kernel estimation 75 kilometre/mileage charges 238 knowledge-based incentives 264 knowledge spillovers 73 Kuznets curve 212 Kyoto Protocol 16, 210, 220, 229, 244 labor intensive industries 10 labor supply, silvicultural eﬀort 261–3 Lagrange multiplier 172–3, 268 land quality 266–7 land-use layers 57 non-forestry 256–8, 259–60, 263 patterns, eﬀects on transport 240–41 landowner motivation 283 landscape features, spatial patterns of 54–5 large firms, permit trading model 180–82 leaded gasoline, phasing out of 212 legal liability problem, disclosure 106–7 legal system, confidentiality requirements 108 Leontief technologies 146, 178 Leq 216 Levin (1985) model 143–4 life-cycle eﬀects of transport 209, 215–16 life-cycle emissions 227 lifestyle and transport demand 213–14 lifetime measures of income/consumption 6, 13–15, 16, 19 line-on-polygon 64 line/line coverage overlays 64 linear expenditure system parameters 13–14 linear spatial features, measurement of density 56

295

load factors, transport 223, 224 local indicators of spatial association (LISA) 76–7 local monopolies, regulation of 135–6 Los Angeles, air quality study 22 Maine, study in 24 management costs, forests 262–3 manipulation, permit market 186–7 Manitoba, canoe routes 60–61 Mannheim convention 244 map projection 49 MapInfo 44, 49 mapping 49–51, 74, 75 maps, paper 47 marginal abatement costs 101 marginal transport costs 215–16, 225–6 market imperfections 122 market power 121–2 input markets 176–80 and non-compliance 184–5 permit market 180–88 regulation 197 through innovation 185 market responses to environmental changes 25 market sensitivity, environmental disclosure 104 market structure 120–21 market-based incentives, traﬃc pollution 234–9 Markov-perfect feedback strategies 150 Markowitz formula 231 Marshallian demand functions 155, 167 Maryland, Patuxent Watershed 56–7 measurement issues, net burden formula 5–7 median preferences/willingness to pay 109–10 meta-analysis 20, 24 metadata 47, 49 methane emissions 219 Mexico, pollution control 95, 103 mileage/kilometre charges 238 mixed land use, transport 240 mobility management plans 244–5 monopolistic competition 157–8, 196–7

296

Yearbook of environmental and resource economics

Dixit–Stiglitz model 162–7 emission standards 161–2 emission taxes in Cournot oligopoly with free entry 158–61 Salop’s circular city model 167–9 monopoly 129–38 monopsony power clean inputs 177–8 polluting inputs 176–7 Moran’s/statistics 76, 77, 85 motor vehicle emissions testing programs 27–8 motor vehicle taxes, empirical studies 14–15 motorized transport modes 243 moving vehicles eﬀects of transport 215–16 multiple production technologies 10 multiple stand management, forests 266–7, 274–6 multivariate functions, GIS 57–8 Nash equilibrium 139, 143, 158, 161–2, 168, 171–2, 174, 185 national emissions standards 107 National Highway Transportation Survey (NHTS), US 12 National Pollutant Inventory, Australia 94–5, 99 National Pollutant Release Inventory, Canada 94, 99 National Trust, UK 62, 68, 69 natural gas plants 10 natural gas taxes 13 nearest-neighbour analysis 56 negative exponential function 83 neighbourhood operations/analysis 65, 66 net present value (NPV) land 259–60, 262–72, 274–5, 279–81, 282 wine 257 Netherlands ABC policy 241–2 car ownership 226, 237 flood risk 49–50 land use patterns 240–41 public transport 225 transport emissions 222 network structures, transport 223

Nike 110 nitrogen 16, 54, 134, 209, 212, 218, 221–2 permit trading 188 nitrogen dioxide 54 nitrogen oxide 16, 218 noise estimation models 60 noise nuisance, transport 216–17, 233–4, 237–8 non-competitive pricing 9–10 non-compliance 101–2, 107, 111, 184–5 non-convex feasibility sets 265–6 non-cooperative environmental policy 191–5 non-designated camping areas 77, 78 non-governmental organizations (NGOs), role in disclosure strategies 95, 106, 108, 110 non-market goods/services 257, 265–6 non-market valuation techniques 265–6 non-motorized transport modes 239, 242–3, 247 non-spatial techniques 74–5 non-timber goods/services 265–7 non-transport policy measures 246 Ohlin, Bertil 258 oligopolies, permit trading Cournot Duopoly 170–73 extensions 175–6 multiple firms 174–5 price competition 174 and subsidies on output 174 welfare comparison permits/taxes 173 oligopsonies 178–9 Ontario eﬀects of logging on tourism 83–6 household income statistics 44–5 open access resources, recreational benefits of 65–73 open-loop strategies 150 opportunity cost of land 257–8, 259–60 optimal allocation of limited resources 71 optimal forest harvesting, use of Faustmann model

Index basic model 259–60 catastrophic risk/uncertainty 273 characterization of optimal harvesting models 256–8 credit constraints 281 dynamic forest models 276–80 forests in an investment portfolio 180–81 as foundation 261 land quality/multiple stand management 266–7 non-timber goods/services 265–6 overview 255–6 regulatory constraints 267–9 silvicultural eﬀort/labour supply 261–3 stumpage price risk/uncertainty 269–72 taxes/subsidies 263–4 timber supply/multiple stand management 274–6 uneven-aged management/selected issues 281–3 optimal non-linear tax/subsidy schemes 146–7 ordinary least squares (OLS) models 69, 79–80, 85–6 Otto motor 227 out-migration, low-income families 24 output market, imperfect competition in emissions influencing consumer demand 136–7 emissions standards 132–3 environmental tax reform 137–8 local monopolies 135–6 overall regulation 137 pollution proportional to 144 rent-seeking 136 simultaneous regulation 133–5 taxation of durable goods monopolist 137 taxation of monopoly 129–32 output, regulation of 133–4 ozone 218 Pareto improvement 29 parking charges 238 companies 241–2

297

partial equilibrium models 124 Patuxent Watershed, Maryland 56–7 pedestrian connectivity 61 penetration rates, transport 223 perfect competition, decentralization under 128–9 performance agencies/grades, pollution 100–102 performance standards 185 personal risk 21 perverse equity eﬀect 8 Philippines, pollution control 95, 97, 99, 101–2, 103, 104, 108 physical compensation, traﬃc pollution 234 physical information, GIS-based spatial analysis of 53–73 physical planning, transport/environment 240–42 Pigouvian taxes 120–21, 128, 133, 134, 135–6, 137, 159–61, 179, 196, 197 point and polygon overlay analysis 57 point-on-polygon/polygon-on-polygon 64 policy change impacts 52–3 policy measures, transport-related 234–46 Pollutant Emission Register, Europe 95, 99, 105, 112–13 Pollutant Release and Transfer Register Czech Republic 95, 99 Japan 95, 99 South Korea 95, 99 pollutants ambient air quality standards 217 pollutants, selection of 105–6 polluters disclosure-related learning 97, 105 eﬀects of community feedback 104 polluting emissions 93 polluting inputs, monopsony power over 176–7 Pollution Abatement Costs and Expenditures (PACE) data 17 pollution control policies beneficiaries of 19–25 conceptual issues 2–10 distributional considerations 26–30

298

Yearbook of environmental and resource economics

empirical studies 10–19 lessons/future research 31–2 overview 1–2 pollution disclosure programs emissions inventories 98–9 environmental performance ratings 99–105 pollution disclosure strategies future developments 109–11 implementation issues 105–9 overview 93–4 programs/results 98–105 public disclosure 94–6 rationale for public disclosure 96–8 summary/conclusions 111–14 pollution taxes 2–5 pollution, pricing of 96 polychlorinated biphenyl (PCB) toxicity levels 47 polygon features, analysis of 56–7 Portland, influence of wetland amenities 56 power plant emissions 16, 18 pre-investment 149–50 price competition 174, 185 price-setting duopolies diﬀerentiated emission taxes 156 outline of model 155 uniform emission tax 157–8 product demand elasticities 7 product diﬀerentiation 109–10 product price change measurement 6 product taxes 2–4 production intensity, transport 211–12 production, environmental eﬀects of 208–9 progressive policies/taxes 12, 31 PROPER program, Indonesia 95, 97, 99–102, 103, 104, 105, 107, 108 property ownership, US 24 public credibility, performance rating systems 108 public disclosure strategies 94–6 rationale for 96–8 public facilities, location/scale 246 public relations, orchestration of 108 public sector standards 229–30 public transport 221–2, 224, 225–6, 239, 241

quadrant analysis, GIS 55–6 rail noise 217 rail transport 214, 222, 223, 225–6, 244 random coeﬃcients model 12 Raster data 47–8, 49, 65 real Bertrand competition 153–5 rebound eﬀects, transport 208, 212, 240, 245 recreation demand values 71 recreational benefits, open access resources 65–73 recreational site choice models 61 recreational sites, isochrone maps 62–3 recycling, revenue taxes 5 reduced cost functions 125–6 regeneration of forests 261–4, 267–9, 276–80 Registro de Emisiones y Transferencia de Contaminantes, Mexico 95, 99 regression models 70 regressive taxes 12, 14, 16–17 regulation emissions/output simultaneously 133–4 local monopolies 135–6 monopoly/competitive firms simultaneously 134–5 pre-testing 198–9 third wave of 95–6 regulators, role of 104, 105 regulatory capability, development of 112 regulatory constraints, forest landowners 267–9 regulatory eﬀort, intensity of 22–3 regulatory reform, transport 243–5 remote sensing experiments 14–15 rent-seeking 136 rent-shifting 192–5 research directions, pollution control 112–14 pollution control policies 31–2 proposals for 196–9 revenue recycling 5, 31, 32 revenue-neutral environmental taxes 13, 138 Rhine (river) 244 right-to-know laws 94, 108

Index risk reduction, value of 230–32 risk-averting behavior 21, 231 risk, forest management 273, 283 road noise 217 route choice 238–9 routing 61–2 rural analysis of spatial patterns/distribution 54 Russia, emissions 213 safety issues, traﬃc 220–21 Salop’s circular city model 195 eﬀect of tax rate increase 168 outline 167–8 second-best optimal tax rate 169 satellite imagery 47, 67 scale problems, spatial overlays 64 scatterplot techniques 75, 76 scenic landscape components 51 second-best optimal emission tax 121 second-best optimal price for pollution 196 second-best taxation 141–4, 166–7, 169 sensitivity analysis 58–60 shadow prices 229–30 Shepard’s Lemma 127 silvicultural eﬀort 261–3 simple data visualization 49–51 simple manipulation, permit market 186 Simpson (1995) model 143–4 site-specific soil yields 57–8 small firms, permit trading model 180–82 smoothing procedures 75 social optimum 128–9 social welfare functions 26–7, 127–8 socioeconomic data 47 socioeconomic distribution of pollutants 53–4 socioeconomic information, GIS-based spatial analysis of 53–73 ‘soft’ environmental eﬀects 232 soil data 64–5 South Korea, pollution control 95, 99 spatial autocorrelation 76–8 spatial autoregressive models 79–80 spatial averaging 75 spatial cross-correlation 76

299

spatial dependence 73–4, 75–6 spatial distribution routines 68 spatial errors models 79–80 spatial hedonic studies 83 spatial heterogeneity 73–4, 75–6 spatial lag models 77, 79–80, 84–6 spatial overlay functions, GIS 62–5, 68–9 spatial patterns/distributions 53–7 spatial regression 77–86 spatial spill-overs 232 spatial statistical analyses/modelling exploratory spatial data analysis (ESDA) 74–7 overview 73–4 spatial regression 77–86 spatial strategies, eﬀects on traﬃc volumes 241–2 spatial trend analysis 57–9 spatial weight matrices 79–80 speed limits 220 split-sample experimental techniques 52–3 spreadsheet records, linking of 67 Stackleberg leader 150 standard cost/surplus concept 124–5 standard exploratory data analysis techniques 74–5 standard setting, transport pollution 239–40, 247 standard utility vehicles (SUVs) 212 standard-and-charges approach 121, 135–6 standing vehicles eﬀects of transport 215–16 stated preferences methods 229–32 static optimisation techniques 258–60, 270 statistical life valuation 231–2 STATSGO 64 stopping rule approach, forest harvesting 269 stumpage price risk, forest harvesting 269–72 subsidies basic system 145 extensions 146–7 output 174 public transport 225–6, 239 timber management 263–4

300

Yearbook of environmental and resource economics

sulphur dioxide 16, 22, 134 sulphur oxides 219 Superfund program 21, 22, 25 supply-side factors, disclosure 98 sustainable harvests 260, 282 sustainable transport 208–10 policy debate 214 risks of overshooting 214 Sweden, nitrogen oxide taxes 134, 146 Swedish Household Expenditure Surveys 14 symbols, maps 50–51 symmetric firms 142–3, 144, 159–60 tax-based recycling 12 taxation distribution of revenues 128–9 eﬀect of increases 165, 168 energy use 214 land use 263–4 of monopoly 129–32, 137 reform of 137–8 schemes 146–7 see also carbon taxes; emissions permits, energy taxes; environmental taxes; gasoline taxes, income taxes; motor vehicle taxes; tradable emissions permits technological development in transport 226–9 technologies 125–7 mix of 160–61 see also abatement technologies; clean inputs/technology; dirty technology telecommuting/teleworking 245–6 temporal trend analysis 57–9 thematic maps 50–51 third wave of regulation 93–6 timber products, trade in 282 timber supply 274–6 Tobler’s First Law of Geography 73 toxic emissions 218 pollutants 95, 98–9, 106 Toxics Release Inventory (TRI), US 20, 94, 97, 98–9, 106 tradable emissions permits 7–8, 31, 214

compliance 198 experimental studies 186–8 market power in markets for 135–6 in oligopoly 170–76 regime 183 trade policy environmental policy as 122–3 terms of trade 189, 192–5 trade secrets, disclosure of 106 traﬃc flows/safety 220–21 transects through data 75 transfer payments between governments 190–91 indexing of 5 Transport and Environment 208 transport demand 213–14, 246 transport flows 221 transport modes 221–6 competition between 242 imbalance in treatment of 214 price/speed of 224 transport responsibilities 244–5 transport trends 246 transport, environmental eﬀects conclusions 246–8 environmental performance between modes 221–6 overview 208–10 passenger transport 243 policies 234–46 technological developments 226–9 trends/environmental eﬀects 210–21 valuation/cost–benefit analysis 229–34 transport-calming measures 217 transport-related policies institutional change 243–5 market-based incentives 234–9 non-motorized transport modes 242–3 non-transport 246 physical planning 240–42 standard setting 239–40 telecommuting 245–6 transportation of hazardous materials 221 travel costs analysis/models 60, 62 travel patterns, determinants of 240–42 travel times 224–5

Index trickle-down philosophies 54 two point data layers 64 UK air quality patterns 53–4 broadland preservation 60 distance impact on WTP 58–9 distribution of risks 20 open-access resources 66–73 recreational sites 62 uncertainty, forest management 269–72, 273, 283 uncompensated demand curves 7 uneven-aged forests, management of 281–3 United Nations Environment Program 107 University of East Anglia 53 unweighted cost–benefit analyses 27–8 upstream flood diversion schemes, Netherlands 49–50 upward-sloping supply curves 9 US air pollution 20 car ownership 226 carbon emissions 15–16, 213, 214, 217, 220 corporate average fuel economy 239–40 federal pollution regulations 16–18 forest biomass 260 legislation 31 opposition to fuel taxes 1 pollution disclosure 94, 97 property ownership 24 sulphur dioxide allowance-trading system 22 STATSGO 64–5 US Forestry Service 267 US National Spatial Data Infrastructure Website 45 utility industries 135 valuation, environmental eﬀects of transport 229–32 value functions, benefit transfer 66–7 value-added tax (VAT) 14 variable taxes 239 variance-mean ratio (VMR) 55–6

301

vector data 47–8, 49 vector overlays 64 vehicle choice/mileage 15 vehicle ownership 226 vehicle ownership taxes 234–7, 239 vehicle size 223 vehicle-use eﬀects 209 vertical monopolies, monopsony chains 179 Vietnam, pollution control 95, 99, 102, 103 viewsheds 51, 52 virtual reality (VR) software 52 virtual reality GIS (VRGIS) 51–3 virtual regulation 110–11, 113–14 visibly hazardous pollutants 106 visual display of data 74 volatile organic compounds emissions 218 voluntary disclosure 93, 98 Wales crop yields 58 recreation demand values 71, 72 waste management facilities 20 wastewater treatment costs 17–18 water grant program, US 18 weighting functions 51 weights, choice of 27 welfare 127–8 welfare comparison, permits/taxes 173 welfare gains, environmental improvements 32 welfare relationship with environmental improvements diﬀerential evaluation of environment 23–4 market responses to environmental changes 24 welfare-maximizing transport 225 wetlands, studies of 56, 58–9 wildlife habitats forests as 265 studies of 65 willingness to pay (WTP) 23–4, 25, 52–3 environmental quality 109–10 impact of distance 58–60 wine-aging problem 257

302

Yearbook of environmental and resource economics

World Bank indicators 112 pollution control strategies 95

zero emission vehicles 228–9 Zhenjiang, China 102, 103

The International Yearbook of Environmental And Resource Economics 2006 2007: A Survey of Current Issues (New Horizons in Environmental Economics)

The International Yearbook of Environmental And Resource Economics 2005 2006: A Survey of Current Issues (New Horizons in Environmental Economics)

The International Yearbook of Environmental and Resource Economics 2004 2005: A Survey of Current Issues (New Horizons in Environmental Economics)

Recent Advances in Environmental Economics (New Horizons in Environmental Economics)

Resource and Environmental Economics

Resource and Environmental Economics

Natural Resource and Environmental Economics

Principles of Environmental and Natural Resource Economics

Economic Valuation of River Systems (New Horizons in Environmental Economics)

Dictionary of Environmental Economics

Dictionary of environmental economics

Economics Of Environmental Conservation

Principles of environmental economics

Dictionary of Environmental Economics

Evolutionary Economics and Environmental Policy: Survival of the Greenest (New Horizons in Istitutional and Evolutionary Economics)

Environmental and natural resource economics 5th edition

Natural Resource and Environmental Economics (3rd Edition)

Applications Of Simulation Methods In Environmental And Resource Economics

Environmental Economics and the International Economy

Global Climate Change: The Science, Economics, and Politics (New Horizons in Environmental Economics)

Choice Experiments Informing Environmental Policy: A European Perspective (New Horizons in Environmental Economics)

Environmental & Natural Resource Economics, 9th Edition

Environmental Economics, Experimental Methods

The Evolution of Markets for Water: Theory And Practice in Australia (New Horizons in Environmental Economics)

Environmental and Natural Resources Economics

Environmental and Natural Resources Economics

Principles of Environmental Economics: Economics, Ecology and Public Policy

Whaling Diplomacy: Defining Issues In International Environmental Law (New Horizons in Environmental Law Series)

A Survey of Ecological Economics

Distributional Impacts of Climate Change and Disasters: Concepts and Cases (New Horizons in Environmental Economics)

Waste and Environmental Policy (Routledge Explorations in Environmental Economics)

The International Yearbook of Environmental And Resource Economics 2006 2007: A Survey of Current Issues (New Horizons in Environmental Economics)

The International Yearbook of Environmental And Resource Economics 2005 2006: A Survey of Current Issues (New Horizons in Environmental Economics)

The International Yearbook of Environmental and Resource Economics 2004 2005: A Survey of Current Issues (New Horizons in Environmental Economics)

Recent Advances in Environmental Economics (New Horizons in Environmental Economics)

Resource and Environmental Economics

Resource and Environmental Economics

Natural Resource and Environmental Economics

Principles of Environmental and Natural Resource Economics

Economic Valuation of River Systems (New Horizons in Environmental Economics)

Dictionary of Environmental Economics

Dictionary of environmental economics

Economics Of Environmental Conservation

Principles of environmental economics

Dictionary of Environmental Economics

Evolutionary Economics and Environmental Policy: Survival of the Greenest (New Horizons in Istitutional and Evolutionary Economics)

Environmental and natural resource economics 5th edition

Natural Resource and Environmental Economics (3rd Edition)

Applications Of Simulation Methods In Environmental And Resource Economics

Environmental Economics and the International Economy

Global Climate Change: The Science, Economics, and Politics (New Horizons in Environmental Economics)

Choice Experiments Informing Environmental Policy: A European Perspective (New Horizons in Environmental Economics)

Environmental & Natural Resource Economics, 9th Edition

Environmental Economics, Experimental Methods

The Evolution of Markets for Water: Theory And Practice in Australia (New Horizons in Environmental Economics)

Environmental and Natural Resources Economics

Environmental and Natural Resources Economics

Principles of Environmental Economics: Economics, Ecology and Public Policy

Whaling Diplomacy: Defining Issues In International Environmental Law (New Horizons in Environmental Law Series)

A Survey of Ecological Economics

Distributional Impacts of Climate Change and Disasters: Concepts and Cases (New Horizons in Environmental Economics)

Waste and Environmental Policy (Routledge Explorations in Environmental Economics)

Recommend Documents