Social Costs of Air Pollution and Fossil Fuel Use – A...

99 Sosiale og økonomiske studier Social and Economic Studies

Knut Einar Rosendahl (ed.)

Social Costs of Air Pollution andFossil Fuel Use– A Macroeconomic Approach

Statistisk sentralbyrå • Statistics NorwayOslo − Kongsvinger

Sosiale ogøkonomiske studier

Social andEconomic Studies

Serien Sosiale og økonomiske studier omfatter nye forskningsbidrag –monografier og redigerte arbeider – på de områder Statistisk sentralbyråhar forskningsvirksomhet. Analysemetoder og temavalg vil variere, menhovedsakelig vil arbeidene være av anvendt og kvantitativ natur med vektpå utnytting av SSBs data i analyser for samfunnsplanleggingsformål og tilallmenn forståelse av sosial og økonomisk utvikling.

The series Social and Economic Studies consists of hitherto unpublishedstudies in economics, demography and other areas of research in StatisticsNorway. Although the studies will vary in analytical methods and in sub-ject matter, they tend to be applied studies based on quantitative analysisof the data sources of Statistics Norway. The research programmes fromwhich the studies originate typically emphasize the development of toolsfor social and economic planning.

Statistics Norway, June 1998When using material from this publication, pleasegive Statistics Norway as your source.

ISBN 82-537-4542-7ISSN 0801-3845

Emnegruppe01.06 Miljøøkonomi og -indikator

EmneordFossile brenslerHelseefekterLikevektsmodellerLuftforurensningSamfunnsøkonomiske kostnaderVeitrafikkØkonomi-miljø-modeller

Design: Enzo Finger DesignTrykk: Falch Hurtigtrykk

3

Abstract

Knut Einar Rosendahl (ed.)

Social Costs of Air Pollution and Fossil Fuel Use– A Macroeconomic Approach

Social and Economic Studies 99 • Statistics Norway 1998

Economic activity and environmental conditions are related to each other in several ways.Production and consumption may pollute the environment, and at the same time the stateof the environment may affect the production capacity of the economy. Thus, it followsthat studying social costs of air pollution should be handled within an integrated model.Moreover, air pollution mostly stems from the use of fossil fuels, which also brings aboutother non-environmental externalities, particularly in the transport sector. It is thereforetopical to include these externalities in a full social costs evaluation.

In this book we are concerned with social costs on a national level, although the environ-mental effects are evaluated on a more local level. We apply a general equilibrium modelof the Norwegian economy, which is extended to integrate environmental and non-environmental effects of fossil fuel use. Moreover, the model includes feedback effectsfrom the environment to the economy. In four independent studies, selected environ-mental and non-environmental externalities are analysed within this model. These arematerial damages, crop damages and health damages from air pollution, and finally healthdamages from traffic accidents.

Keywords: Air pollution, fossil fuel use, integrated economy-environment model, roadtraffic, social costs.

Acknowledgement: We acknowledge the support given by the Ministry of Environment.

4

Sammendrag

Knut Einar Rosendahl (red.)

Samfunnsøkonomiske kostnader av luftforurensning ogfossile brensler– En makroøkonomisk tilnærming

Sosiale og økonomiske studier 99 • Statistisk sentralbyrå 1998

Økonomisk aktivitet og miljøforhold er knyttet til hverandre på flere måter. Produksjon ogkonsum kan forurense miljøet, samtidig som miljøtilstanden kan påvirke produksjons-kapasiteten i økonomien. Det er derfor viktig å studere samfunnsøkonomiske kostnader avluftforurensning i en integrert modell. Samtidig skyldes luftforurensning i hovedsak bruk avfossile brensler, som også medfører andre eksternaliteter, spesielt i transportsektoren. Deter derfor hensiktsmessig å inkludere disse eksternalitetene i en samlet evaluering av desamfunnsøkonomiske kostnadene.

Denne boka konsentrerer seg om samfunnsøkonomiske kostnader på et nasjonalt nivå,selv om miljøeffektene analyseres på et lokalt nivå. Vi benytter en generell likevektsmodellfor den norske økonomien, som er utvidet til å inkludere miljøeffekter og andre effekter avfossile brensler. Modellen inneholder også tilbakevirkende effekter fra miljøet tiløkonomien. I fire uavhengige studier blir utvalgte miljø- og andre eksternaliteter analysertved hjelp av denne modellen.

I kapittel 3 studeres korrosjonskostnader på bygningsmaterialer og biler som følge avluftforurensning. Basert på norske data for luftforurensning, materialbeholdning ogvedlikeholdspriser, benyttes dose-respons funksjoner til å analysere vedlikeholdskostnaderknyttet til nasjonale utslipp av SO2. Beregningene for Oslo blir utført ved bruk av enspredningsmodell for luftforurensning, og bygningsregisteret GAB. For andre deler avNorge blir mer generelle metoder anvendt. Til tross for lave utslipp av SO2 i Norge (i 1994),indikerer beregningene at årlige vedlikeholdskostnader som følge av denne forurensningener omtrent 200 millioner kroner, hvorav en tredel rammer Oslo. Når disse resultatene blirimplementert i den integrerte modellen, øker de samfunnsøkonomiske kostnadene tilnesten 300 millioner kroner. Dette skyldes en høyere brukerpris på kapital, som fører til atkapitalnivået faller. Dermed avtar den økonomiske veksten.

Kapittel 4 presenterer beregninger av avlingsskader som skyldes bakkenær ozon i et år(1992) med høye ozon-nivåer i Norge. Kjennskap til ozon-eksponeringen i løpet avvekstsesongen (AOT40) fås på basis av spredningsmodeller og målestasjoner. Basert pågeografiske data om plantearealer og avlinger, beregnes tap av hvete, potet og gress (fradyrket eng). Siden jordbrukssektoren er svært regulert i Norge, er skyggeprisen på

5

avlingene avhengig av hvordan myndighetene reagerer. To ulike beregninger blir derforutført. I den ene antas det at avlingstapet kompenseres ved økt import. De direktekostnadene er da rundt 200 millioner kroner. Når disse resultatene implementeres i denintegrerte modellen, blir de totale kostnadene nesten doblet. I den andre beregningenantas det at den innenlandske ressursinnsatsen økes for å opprettholde produksjonsnivået.I dette tilfellet blir de direkte kostnadene ca. 500 millioner kroner, mens de totalekostnadene øker til over 1,2 milliarder kroner. Forklaringen på denne store økningen er atressurser blir trukket vekk fra andre og mer produktive sektorer i økonomien.

Kapittel 5 analyserer samfunnsøkonomiske kostnader av helseskader knyttet til luft-forurensning. Den internasjonale litteraturen om dose-respons funksjoner blir gjennomgått,og det blir dokumentert hvordan disse funksjonene kan bli brukt til å analysereøkonomiske virkninger av luftforurensning i Norge. Ved å benytte denne informasjonen bliren egen beregning av helseeffekter og samfunnsøkonomiske kostnader av luftforurensninggjennomført for Oslo. Dette er basert på sammenhenger mellom utslipp og konsentrasjonav partikler (PM10) og NO2, framkommet ved hjelp av en spredningsmodell. De totalesamfunnsøkonomiske kostnader beregnes til 1,7 milliarder kroner. 90 prosent av dissekostnadene er imidlertid knyttet til verdsetting av ikke-produktive effekter (dvs. fram-skyndet dødelighet og kronisk sykdom). Videre er bare 1 prosent knyttet til tilbakevirkendeeffekter på økonomien (dvs. 10 prosent av de produktive effektene). Disse effektene erderfor ikke spesielt viktige for helseskader, i motseting til hva analysene i kapittel 3 og 4konkluderer med.

I det siste kapitlet studeres eksternaliteter knyttet til trafikkulykker. Norske studier avsammenhengen mellom trafikkulykker og drivstofforbruk (og andre forklaringsfaktorer),samt detaljert kunnskap om ulykkeskostnader, blir brukt til å modellere samfunns-økonomiske kostnader av drivstofforbruk. Virkninger av trafikkulykker på arbeidstilbudetog offentlige utgifter, som følge av dødsfall og personskader, blir analysert. Sammen-hengene er videre implementert i den integrerte modellen. Det vises at framskrivninger avBNP i 2020 blir noe redusert, nærmere bestemt med 0,34 prosent, når tilbakevirkningenefra trafikkulykker blir tatt hensyn til. Dette skyldes at trafikkvolumet forventes å økeframover, noe som medfører flere ulykker og dermed en mindre arbeidsstokk enn veduendret ulykkesfrekvens. Innføring av en CO2-avgift som stabiliserer utslippene viser segvidere å være mindre kostbar for økonomien når tilbakevirkningene tas hensyn til. BNP blirredusert med 0,44 prosent i 2020, sammenlignet med 0,47 prosent når tilbakevirkningeneignoreres.

Emneord: Fossile brensler, helseeffekter, likevektsmodeller, luftforurensning, samfunns-økonomiske kostnader, veitrafikk, økomomi-miljø modeller.

Prosjektstøtte: Miljøverndepartementet har gitt finansiell støtte til prosjektet.

Social and Economics Studies 99 Social Costs of Air Pollution

7

Contents1. Introduction .................................................................................................. 91.1. Motivation ..................................................................................................................91.2. ntegrated analyses ....................................................................................................101.3. An integrated economy-environment model .............................................................121.4. Valuing environmental damages and other externalities............................................131.5. Outline of the book...................................................................................................14

2. An integrated economy-environment model (Knut Einar Rosendahl)......... 172.1. MSG-EE: An applied general equilibrium model ........................................................172.2. MSG-EE with feedback effects from the environment ...............................................19

3. Corrosion costs of building materials and cars in Norway (Solveig Glomsrød,Odd Godal Jan Fr. Henriksen, Svein E. Haagenrud and Torstein Skancke) ..................... 23

3.1. Introduction ..............................................................................................................233.2. Dose-response and lifetime functions for some materials ..........................................243.3. Air quality .................................................................................................................273.4. Stock of materials at risk ...........................................................................................293.5. Maintenance costs ....................................................................................................323.6. Marginal corrosion costs of SO2 emissions ................................................................353.7. Macroeconomic effects of material corrosion............................................................383.8. Change since 1985 ...................................................................................................403.9. Uncertain factors ......................................................................................................413.10.Conclusion................................................................................................................42

4. Social costs of crop damage from ground-level ozone (Kjetil Tørseth,Knut Einar Rosendahl, Anett C. Hansen, Henning Høie and Leiv Mortensen) ........... 45

4.1. Introduction ..............................................................................................................454.2. Ozone exposure and crop damage.............................................................................474.3. Economic analyses of crop damage............................................................................544.4. Conclusion.................................................................................................................65

5. Health effects of air pollution and impacts on economic activity(Knut Einar Rosendahl)...................................................................................... 67

5.1. Introduction ..............................................................................................................675.2. Health effects of particulates.....................................................................................725.3. Health effects of nitrogen dioxide (NO2) ....................................................................845.4. Health effects of sulphur dioxide (SO2) ......................................................................885.5. Health effects of ozone (O3) ......................................................................................895.6. Population exposure to air pollution in Oslo ..............................................................915.7. Public health effects and social costs of air pollution in Oslo......................................935.8. Conclusion................................................................................................................99

Social Costs of Air Pollution Social and Economics Studies 99

8

6. Modelling impacts of traffic injuries on labour supply and publichealth expenditures (Solveig Glomsrød, Runa Nesbakken andMorten Aaserud)............................................................................................ 101

6.1. Introduction.............................................................................................................1016.2. Data sources ............................................................................................................1026.3. The model framework .............................................................................................1036.4. Traffic accidents as a function of fossil fuel consumption and other variables ...........1046.5. Labour supply reductions due to traffic accidents.....................................................1076.6. Public health sector costs .........................................................................................1106.7. Simulations ..............................................................................................................1126.8. Conclusions .............................................................................................................115

References........................................................................................................ 117

AppendicesA. Appendix to chapter 3: Tables.................................................................................129B. Appendix to chapter 5: Overview of dose-response functions for health effects ......140C. Appendix to chapter 6: The relation between traffic volume, traffic density and

traffic injuries ..........................................................................................................143

Recent publications in the series Social and Economic Studies ................. 146


9

1.1. MotivationThere has been a growing awareness overthe last decades that economic activity insome respects leads to extensive negativeexternalities on environmental resources,implying a suboptimal deterioration of theenvironment. This has called for govern-mental actions to bring the economy on amore optimal path. Traditionally,economists have favoured market-basedinstruments like Pigouvian taxes (Pigou1932), i.e., the polluter must pay a taxcorresponding to the marginal damageinflicted on others.1 Natural scientists, onthe other hand, have usually advocatedcommand and control policies, whichhave often been adopted by policymakers, too. Irrespective of instrumentchoice, in order to make right decisionsone has to know the actual social costsassociated with an environmental exter-

* Thanks to Torstein Bye and Nils Martin Stølen forvaluable comments on earlier drafts, and to MonaIrene Hansen for valuable research assistance relatedto all the four analyses in this book. Thanks to PeterThomas for translating earlier versions (inNorwegian) of chapters 3, 4 and 5. As the chaptershave been edited since, the editor is resposible forboth the content and the language.1 In a seminal paper, Coase (1960) attacks thePigouvian tradition by emphasizing property rightsaspects.

nality. Then these costs may be comparedwith the costs of control. In this study wepresent calculations of the social costs ofcertain environmental externalities, aswell as other externalities related to theuse of fossil fuels.

Current economic activity and the state ofthe environment are in many ways tightlyconnected. As pointed to above, produc-tion and consumption of goods andservices may cause pollution, e.g., relatedto the use of energy. The evolution of theenvironmental quality therefore dependson the economic development. Simul-taneously, pollution is responsible forhuman and non-human damages, whichto some degree is detrimental to theresource base of economic activity. Hence,the economic development may be ham-pered if the pollution levels come out ofcontrol.

These interactions favour integratedanalyses of economic and environmentalaspects. This point is emphasised in ourstudy of social costs of environmentalexternalities. Air pollution causes, e.g.,various health effects, material corrosionand crop damages, which in turn reducethe actual supply of labour, increase the

1. Introduction*


10

user cost of capital and decrease agricul-tural productivity. These effects havemacroeconomic implications which maybe considerable. Hence, the social costs ofair pollution may be miscalculated if thesemacroeconomic feedback effects areignored.

Nevertheless, whereas the environmentalimpacts of economic activity are wellcomprehended, the opposite links arerarely taken into account in studies ofenvironmental damages.2 Major studiesconducted for the European Commission(EC 1995) and the US Department ofEnergy (ORNL/RFF3 1994) analyseexternal costs of energy productionthoroughly using partially integratedanalyses, but do not consider themacroeconomic impacts pointed to above.

The environmental damages discussed inthis book are all related to air pollution,which for the most part stems from theuse of fossil fuels. At the same time, thereare other important externalities relatedto fossil fuels, particularly in the transportsector (e.g., accidents, noise andcongestion). Thus, it may be argued thatan integrated analysis of air pollutionshould also focus on these non-environmental externalities, at least whenit comes to policy recommendations.Moreover, several of these externalitieshave detrimental effects on the resourcebase of economic activity, just like theenvironmental externalities. E.g., bothtraffic accidents and transport noise mayhave negative consequences on theefficient supply of labour. Hence, incalculating social costs of transport-related

2 Bergh (1993) and Rosendahl (1997) are two theore-tical exceptions.3 Oak Ridge National Laboratory and Resources forthe Future.

externalities, one should take a macro-economic approach.

This book is not aiming at including allenvironmental externalities, not to say allexternalities from fossil fuel use. Wepresent studies of four selected extern-alities, three of them are environmentalexternalities and the last one is related totraffic accidents. Moreover, even withinthe specific environmental areas we focuson, there are at all probabilities severaleffects that are ignored. The reason is thatenvironmental impacts are a complexmatter, so that the current scientificknowledge is insufficient to calculate thetotal social costs of environmental dam-ages. Thus, the four externalities analysedin this book are not selected because theyare the most important ones, but ratherbecause of the applicable information thatexists for these externalities. This is animportant point when interpreting theresults in this book.

1.2. Integrated analysesIntegrated analyses have become apopular scientific method, e.g. in thestudies of climate change. By integratedanalyses is meant bringing togetheranalyses of various parts of a jointproblem into one simultaneous analysis.In this book we shall restrict ourselves todiscuss such analyses related to socialcosts of local and regional environmentalexternalities. In order to calculate thesecosts in a credible way it is necessary tointegrate analyses of natural science andeconomics. Natural science may provideinformation about the natural links,whereas economics may provide infor-mation about the social costs of certainenvironmental damages. As the naturallinks are particularly complex, lack ofscientific knowledge has for long time puta restraint on valuing environmental


11

externalities. Thus, earlier analyses haveto some degree been based on expertjudgements4 and control costs5, whichhave a more questionably scientificfoundation, or on various valuationstudies of, e.g., clean air, where thespecific impacts are skipped.6

The rationale for using integrated analysesas indicated above, has increased con-siderably the last decade. New researchhas managed to estimate quantitativerelationships between particularly airpollution and various human and non-human damages. These associations arecommonly referred to as dose-responsefunctions. Whereas expert judgements,control costs and valuation methods leavelittle information about the characteristicsof the damages, dose-response functionshelp identifying the specific impacts, e.g.,hospital admissions and reduced lifetimeof various materials. These functions havebeen used by the two major studiesmentioned above (EC (1995) and ORNL/RFF (1994)) to calculate the direct exter-nal impacts of energy production. Further-more, the dose-response functions makequantification of feedbacks to the econo-mic resource base possible. Hence, theyare natural links in a fully integratedeconomy-environment model.

Integrated analyses of environmentalexternalities, using dose-responsefunctions, clearly call for a disaggregatedapproach. First, the level of emissions ofvarious pollutants depends on the choice

4 E.g., the social costs of health damage in Alfsen et al.(1992).5 The social costs in Hohmeyer (1988) and PACE(1990) were partly based on control costs.6 The most common valuation methods are Contigentvaluation method (CVM) and hedonic approachmethod (see Brookshire et al. (1982) for a comparisonof these methods).

of energy use, the choice of combustiontechnology and substitution possibilities,which vary between different sectors ofthe economy. Second, the costs ofenvironmental externalities vary withrespect to both space and time. Forinstance, health damages from a certainemission of particulate matter are clearlyhigher in the middle of the day in a largecity than at night or in the countryside.Thus, an integrated model for our purposeshould be disaggregated both on theeconomic and the environmental part.

An important justification for applyingdose-response functions is their trans-parency. However, Stirling (1996) claimsthat this methodology may not come upwith even approximately correct numbers.There are several reasons for this. First,there is a number of uncertainties relatedto the dose-response functions applied;both to the interpretation of the originalstudy and to the transferability of theresults to other locations. However, thisuncertainty is partly reduced as the num-ber of original studies grows, and aconsensus view is reached. Second, asmentioned above there will always be achance of overlooking important associ-ations which for some reason have notbeen demonstrated. Thus, there is anunderlying risk of underestimating thetotal impacts of pollution. Third, given thephysical information, an economic valua-tion will necessarily have to rely on somevalue judgements, like how to appraiserisk, distributional aspects and non-economic impacts in general. However,this problem applies to all methods thatintend to calculate social costs of environ-mental externalities (see section 1.4).


12

1.3. An integrated economy-environment model

Although this book presents four separatestudies, they all apply the same integratedeconomy-environment model. This modelis an extended version of MSG-EE (seeAlfsen et al. 1996), which is an appliedgeneral equilibrium model for energy andenvironmental analyses of the Norwegianeconomy, with inter alia a detailed model-ling of the transport sector. In a submodelMSG-EE calculates the national emissionsof several air pollutants. The extension ofMSG-EE is more or less based on resultsfrom the four studies presented in thisbook. Both MSG-EE and the extendedversion is further outlined in chapter 2 ofthis book. Below we give a brief descrip-tion of how the economy and the environ-ment are connected within the model.

The extended model is illustrated in figure1.1, where the shaded area is the originalMSG-EE model. Economic activity isdetermined by inter alia the size of theresource base (labour and capital stocketc.) and other input variables. The sizeand allocation of economic activitydetermine, through the use of fossil fuelsfor transport, heating and industrialprocesses, the national emissions of thevarious pollutants. In the extended modelthe national emissions are partlydistributed on various geographicallocations (main cities etc.), and then theambient concentrations of differentpollutants are determined for theselocations. Dose-response functions, asdescribed in section 1.2, are then used tocalculate the human and non-humandamages of air pollution. Finally, thesedamages affect the resource base of theeconomy and other input variables. Thus,we have a simultaneous economy-environ-ment model.

Similarly, economic activity and thetransport level are tightly connected, andthe extended model calculates thenational road traffic volume. This andother variables determine the extent ofnon-environmental traffic externalities,which in turn affect the basis of the eco-nomy. Again, the circle is closed, and thetraffic externalities (which in this book arerestricted to accidents) and the economicactivity are determined simultaneously.

The new information about social costsobtained with this analysis compared tomost other externality analyses mayoriginate from two effects. To see this,consider a marginal increase in theemissions of a specific pollutant. Throughthe concentration and dose-responsefunctions, this increased emission bringsabout some damages that are valued atfixed prices in traditional analyses. In ourmodel, on the other hand, the costs of thedamages also depend on the effects oneconomic activity, i.e., how the economicequilibrium is changed on the marginthrough the changes in input variables. Aswill be seen in some of the chapters of thisbook, the resulting costs may differ signifi-cantly from the direct costs (from smallincreases to a doubling of the costs).

The other effect is of less importance, butshould be included for the sake of comp-leteness. As the economic equilibrium ischanged, the total emissions are changed,too, and in the end we arrive at anequilibrium where all the links in figure1.1 are fulfilled. Since economic activity isnegatively affected by emissions asindicated above, and emissions are anincreasing function of economic activity, amarginal increase in emissions has anegative feedback effect on totalemissions. Thus, this effect dampens thesocial costs of emissions somewhat.


13

However, as the economy after all is veryinelastic with respect to emissions, andthe elasticity of emissions with respect toeconomic activity presumably is nothigher than one, this effect turns out to benegligible.

1.4. Valuing environmental damagesand other externalities

It is useful to separate the valuation ofenvironmental damages and otherexternalities into market and non-marketeffects. This is illustrated in figure 1.1.Some damages, which affect elements ofthe economy, are treated within themodel, which chooses the right valuationas well as the feedback effects on theeconomy. This could, e.g., be corrosion ofbuilding materials. Other effects, which donot (merely) have impacts on theeconomy, are valued in a subsequentmodel. This could, e.g., be reduced qualityof life related to increased morbidity ormortality (which of course may haveeconomic impacts, too). This separationprovides that the externalities are treatedconsistently and transparently.

In most studies of environmental extern-alities (e.g. EC (1995)) the valuation of aspecific damage is made without separa-

ting market from non-market effects ofthe damage. For health damages oneeither chooses results from a willingnessto pay (WTP) study (or other contigentvaluation studies), or uses results basedon a cost of illness (COI) approach, whichintends to measure the lost earnings andmedical costs. As WTP estimates aregenerally assumed to capture the entirewelfare cost of the damage, i.e., includingthe COI estimates, the latter estimates areusually corrected for by a factor of 2. Thisis based on the results of some empiricalstudies of specific morbidity endpoints(see the discussion by US EnvironmentalProtection Agency in EPA (1995)). How-ever, as this relationship may differ signi-ficantly between different health dam-ages, this should not be done withoutcaution. Moreover, treating COI as aportion of WTP may be wrong in cal-culating social costs in countries likeNorway, where the economic losses ofbeing ill is mainly born by the govern-ment. Thus, the two estimates may ratherbe partly additive.

Valuation methods of non-market effectshave been subject to a lot of criticism. Onemain reason is that objective valuationsof, e.g., increased mortality or biological

Figure 1.1. An integrated economy-environmental model

Road traffic volume (RT)

Economic activity (Y) Emissions (Ej)of pollutant j

Ambient concentrations(Cj) of pollutant j

Trafficexternalities (TEk)-Traffic accidents--

Resource base and otherinput variables (Ri)-Labour stock-Depreciation rate of capital-Public expenditure-Producitivity change

Valuation ofnon-market effects (Vk)

Human and non-humandamages (Dk)-Health damage-Material corrosion-Crop damage-

MSG-EE


14

diversity may not be feasible. Ideally thevaluation should therefore be placed onthe decision-makers.7 Moreover, severalstudies have pointed to major weaknessesof the existing valuation methods.8 Asplacing the valuation on the decision-makers may not be practically feasible inall respects, the valuation estimates maybe used as indicative numbers which areexposed to alterations. In any case thephysical non-market effects should bepointed out.

1.5. Outline of the bookThis book presents four separate works onthe social costs of externalitities from fos-sil fuel use in a macroeconomic frame-work; three of them are concerned withenvironmental externalities, whereas thelast one is concerned with externalitiesfrom traffic accidents. In the following abrief outline of each chapter is presented.9

Chapter 3, by Glomsrød, Godal, Hen-riksen, Haagenrud and Skancke, dealswith corrosion costs of building materialsand cars due to air pollution. Based onNorwegian data on air pollution, materialstocks and maintenance prices, they applydose-response functions to analyse main-tenance costs due to national emissions ofSO2. The calculations for Oslo are carriedout with the aid of a dispersion model forair pollution, and the GAB buildingregister. For other parts of Norway moregeneral methods have been used. Despitesmall emissions of SO2 in Norway (in

7 Nyborg (1996) discusses the information require-ments that are needed to succeed in this attempt.8 Kahneman and Knetsch (1992) point to someimportant problems with contigent valuation methods(CVM). This is further analysed by Halvorsen (1996),using data from a Norwegian CVM survey. Herfindings largely support the criticism.9 Alfsen and Rosendahl (1996) give a shortpresentation of the work behind chapter 3, 5 and 6.

1994), the calculations indicate that theannual maintenance costs due to thispollution is about Nkr 200 million, ofwhich one third falls on Oslo. When thesefindings are put into the model illustratedin section 1.3, the social costs increase toalmost Nkr 300 million. This is due to ahigher user cost of capital, which impliesthat the desired capital stock decreases.Thus, the economic growth is dampened.

Chapter 4, by Tørseth, Rosendahl,Hansen, Høie and Mortensen, presentscalculations of crop damages from groundlevel ozone in a year (1992) with highozone levels in Norway. Information onozone exposure during the growth seasons(AOT40) is found on the basis of dis-persion models and measuring sites.Based on geographical data on crop areasand yields, total loss of wheat, potato andmeadow is calculated. As the agriculturalsector is very regulated in Norway, theshadow prices of the crops depend onhow the government responds. Thus, twosets of calculations are carried out. In onecalculation, it is assumed that the yieldlosses are compensated for by increasedimports. Then total direct costs are foundto be around Nkr 200 million. Whenintegrating these links into the modelabove, the total social costs almost double.In the other calculation, it is assumed thatthe domestic resource use is increased inorder to maintain the production level. Inthis case the direct costs are about Nkr550 million, whereas the total costs foundby using the integrated model is morethan Nkr 1.2 billion. The explanation forthis big increase is that resources aredrawn away from other, and moreproductive, sectors of the economy.

Chapter 5, by Rosendahl, analyses socialcosts of health damages due to air pol-lution. The international literature on


15

dose-response functions are examined,and it is documented how these functionscan be applied to analyse economic im-pacts of air pollution in Norway. Usingthis information, a specific calculation ofannual health effects and social costs oflocal air pollution is carried out for Oslo.This is based on relationships betweenemissions and concentrations for particul-ate matter (PM10) and NO2, established bya dispersion model. The total social costsare found to be about Nkr 1.7 billion.However, 90 per cent of these costs aredue to valuations of non-market effects(i.e. premature mortality and chronicillness), which may be viewed as parti-cularly debatable as stated above. More-over, only 1 per cent is attributed to thefeedback effects on the economy (i.e., 10per cent of the market effects). Thus, asopposed to the preceding chapters, thiseffect does not seem to be very importantfor health damages.

Finally, chapter 6, by Glomsrød, Nes-bakken and Aaserud, considers extern-alities related to traffic accidents.Norwegian studies on the associationbetween accidents and fuel consumption(and other factors), and a socialaccounting system for accident costs, areused to model the social costs of fuelconsumption related to traffic accidents.Impacts of accidents on labour supply andpublic expenditure through deaths andinjuries are analysed. The links are furtherimplemented in the model illustrated insection 1.3. It is shown that projections ofGDP in 2020 are slightly reduced, i.e. by0.34 per cent, when the feedback effectsof traffic accidents are taken into account.This is due to a projected increase intraffic volume, implying more accidentsand thus a smaller labour stock than inthe case of unchanged frequency ofaccidents. Moreover, introducing a CO2

tax to stabilise emissions is found to beless expensive when these feedbacks areaccounted for. GDP is reduced by 0.44 percent in 2020, compared to 0.47 per centwhen the feedbacks are ignored.


17

In this chapter we give a description of theintegrated economy-environment modelthat is used in the four studies presentedin this book. The core of this model is anapplied general equilibrium model for theNorwegian economy called MSG-EE. Thismodel is briefly outlined in section 2.1,emphasizing features that are importantfor the analyses in the following chapters.A more thoroughly description is given inAlfsen et al. (1996). Then in section 2.2we describe a version of MSG-EE wherethe economic model is extended to inc-lude links to and from the environment.Figure 1.1 in the preceding chapter givesan illustration of the integrated model,where the shaded area covers the originalMSG-EE model.

2.1. MSG-EE: An applied generalequilibrium model10

MSG-EE (Multi-Sectoral-Growth – Energyand Environment) has been developed byStatistics Norway for energy and environ-mental analyses of the Norwegian econo-my.11 Both the choice of industries,

10 This section is to a large extent based on Alfsen etal. (1996).11 MSG-EE is a special version of the fifth officialgeneration of the MSG model, originally worked outby Leif Johansen (Johansen 1960). MSG-EE has been

commodities and input factors in themodel reflect the kind of use of the model.Thus, MSG-EE offers interesting studies ofe.g. environmental effects of both variouslevels and compositions of economicactivity.

As energy and environmental issues havea long-term perspective, MSG-EE is basedon the theory of economic growth. Thus,increases in the primary input factors(e.g., capital stock and an exogenouslabour supply) are the main determinantsof the economic development, togetherwith exogenous changes in productivity,see figure 1.1. Producer and consumerbehaviour are explicitly modelled basedon optimisation principles. Parameters inthe utility and production functions are toa large extent based on estimation resultsfrom Norway, which are based on datafrom the National Accounts for the periodfrom 1960 to 1989 (see chapter 3 inAlfsen et al. (1996)).

MSG-EE is a fairly disaggregated model,both with respect to commodities and

used in a wide range of energy and environmentalstudies, e.g. Glomsrød et al. (1992), Aasness et al.(1996) and Moum (1992).

2. An integrated economy-environment model

Knut Einar Rosendahl

Social Costs of Air Pollution Social and Economic Studies 99

18

industries.12 As the sectors are not equallyefficient, this disaggregated industrystructure means that the sector com-position also affects the aggregateproduction level. Moreover, the modelincludes a detailed description of themarkets for energy and transport. Thedisaggregated approach with emphasis onenvironmentally important sectors is aclear advantage when studying environ-mental issues, as the emission intensitiesdiffer greatly between industries andcommodities. Thus, changes in emissionscan occur through changes in the inputdemand as well as changes in the industrystructure. However, this requires that thesubstitution possibilities are well known,both within an industry and betweenvarious sectors of the economy.

The production structure for the indu-stries in MSG-EE is illustrated in figure2.1. At the top level there are five inputfactors, i.e., capital (other than transportequipment) (K), other materials (V),labour (L), transport (T) and engergy (U):

(2.1)Y f K L V T K F U E FT T U= { , , , ( , ), ( , )}

These factors are determined according toa constant returns to scale flexible techno-logy. The capital stock is a sector specificLeontief aggregate of eight capital goods,which again are Leontief aggregates of allthe basic commodities in the model. Othermaterial inputs are also Leontief aggre-gates of these commodities.

Transport is divided into five types oftransport services, i.e., transport by road,air, rail, sea and post and telecommuni-cation. Each of these services may be

12 MSG-EE specifies 47 commodities, and the numberof industries is 33.

purchased in the market from acorresponding transport sector. In ad-dition a significant share of road transportand some sea transport are produceddirectly by the industries themselves (owntransport). The volume of own transportis approximated by the use of transportcapital (KT) and transport fuels (FT). Theamount of own transport in a sector islinked to the amount of commercial tran-sport services by fixed coefficients. As railtransport and post and telecommunicationare relatively clean transport technologies,a shift between the five transport sectorsin favour of these will contribute toreduced emissions. However, due to datalimitations, the compositition of transportservices within the industries is exo-genous. Still, changes in industry structuremay lead to substitution effects at themacro level.

As transport fuels are modelled as inputfactors to the transport services, oilproducts used for transport are excludedfrom the energy aggregate U at the toplevel of the production function (seeequation (2.1)) and figure 2.1. The energyaggregate is used for stationary com-bustion, and is divided into electricity (E)and fuel for heating purposes (FU) accor-ding to a CES production function withconstant returns to scale.

There are several household groups in themodel. At the top level, each group allo-cates total consumption expenditure on15 consumption goods. At the next levelconsumption of transport services isdivided into private and public transport.Private transport is further divided intopetrol and car maintenance, and the stockof cars, whereas public transport isallocated into five transport services.Energy is an aggregate of electricity andfuels (energy demand functions are based


19

on econometric studies in Norway). Thus,at the bottom line we end up with 22consumption activities. We see that thechoice of activities is clearly relevant forstudies of environmental problems. Eachof the consumption activities consists of aLeontief aggregate of all the basiccommodities. There is no intertemporalbehaviour among the households in themodel, and total consumption expenditureis assumed to ensure full capacityutilisation in the economy.

In MSG-EE the government receives bothdirect and indirect taxes (or offer sub-sidies). The indirect taxes and subsidiesvary across sectors and commodities, andaffect prices and incomes. A carbon tax isspecifically modelled. Moreover, employers’contribution to social security and Nation-al Insurance is also included. In additiongovernmental production is exogenouslyspecified on health care and three othersectors, and the model distinguishesbetween local and central services.

In a long run equilibrium domesticproducer prices are assumed to equal totalunit costs. As the production functionshave constant returns to scale, unit costsare independent of the scale of produc-tion. Thus, the domestic producer pricesare only functions of so called primarycost components, which include the wagerate, the user cost of capital, importprices, technological change, indirect taxrates and prices of public services. Boththe wage rate and the user cost of capitaldiffer between sectors.

These two cost components are by natureendogenous. The same apply to the tradesurplus and the capital stock. However, asthe model is not intertemporal, in order toclose the model, either the wage rate orthe trade surplus have to be exogenous,

and either the shadow price of capital orthe capital stock have to be exogenous.This choice is left to the model user. In theanalyses in this book the trade surplus andthe shadow price of capital have beenchosen as exogenous variables. Accordingto Alfsen et al. (1996), this closure rulehas “been frequently used in normativepolicy studies of welfare and resourceallocation” as “one wants to exclude wel-fare gains that are financed by increasingforeign debt.”

MSG-EE includes several subroutines, andone of them calculates the nationalemissions of 8 air pollutants based on theuse of fossil fuels and material inputs inthe various sectors of the economy (seefigure 1.1). For our purpose, emissions ofparticulate matter, NOx and SO2 areparticularly relevant. 6 different emissionsources are identified for each of theproduction sectors and the private house-holds. Four of them are related to tran-sport combustion (FT in equation (2.1) forthe production sectors) and one is relatedto stationary combustion (FU in equation(2.1)). The final source covers the remain-ing emissions, which are mainly fromindustrial processes (connected to V inequation (2.1)). The emission calculationsare based on exogenous coefficients foreach source in each sector. The coeffici-ents are generally linked to certaineconomic variables in the model, and maychange over time due to expected changesin emission intensities.

2.2. MSG-EE with feedback effectsfrom the environment

The extensions in this version of MSG-EEare more or less based on results from thefour studies presented in this book, andwe will not anticipate these results here.However, we will give a formal descrip-tion of the general links that are used, as


20

illustrated in figure 1.1 in the introductorychapter. First, we formalise theconnections within the original MSG-EE,i.e. without feedback effects from theenvironment, with emphasis on variablesthat are important in this study. Aspointed out in section 2.1, the economicdevelopment (Y) depends on the develop-ment of the resource base and other inputfactors (Ri), jointly denoted R:

(2.2) Y = Y (R)

Whereas the labour stock growth isexogenous, the growth in capital stockdepends on the user cost of capital, whichis a function of inter alia the shadow priceof capital and the depreciation rate. More-over, productivity changes and publicexpenditures are other exogenous inputfactors to MSG-EE. The size and structureof economic activity determine, mainlythrough the use of fossil fuels, the nation-al emissions (E s,e

j) of the 8 pollutants (j),distributed on sector (s) and source (e):

(2.3) Es,ej = Es,e

j (Y)

Statistics Norway collects and calculatesemission data for each municipality inNorway, and these emissions are alsodistributed on pollutants, sectors andsources, in the same manner as thenational emissions. Thus, using fixedcoefficients for each emission source ineach economic sector, calculated in thebase year, the extended version of themodel distributes national emissions onvarious geographical locations (main citiesetc.) in a fairly detailed way. Then, basedon dispersion models and/or measuringsites, the ambient concentrations (Cj) of 4different pollutants are determined for thesame locations:

(2.4) Cj = Cj (Es,ej)

The concentrations of air pollutants leadto various human and non-humandamages (Dk), such as health damages,material corrosion and crop damages:

(2.5) Dk = Dk (Cj)

These associations are based on dose-response functions, which were discussedin chapter 1. The functions are usuallylinear. Some of these damages affectcentral input factors to the economy, suchas the labour stock and the depreciationrate of capital:

(2.6) Ri = Ri (Dk)

These functions are also generallyassumed to be linear, and are based onvarious national statistics. Thus, sum-marising equations (2.2) to (2.6) we get:

(2.7) Y = Y{R[D(C(E(Y)))]}

where the variables must be viewed asvectors. That is, we have a simultaneouseconomy-environment model.

Similarly, the detailed transport modellingof MSG-EE gives a good foundation forcalculating the road traffic volume (RT):

(2.8) RT = RT (Y)

which is a main determinant of severalnon-environmental externalities from roadtraffic (TEk):

(2.9) TEk = TEk (RT)

In this book we only focus on traffic acci-dents. As for the environmental damages,these non-environmental externalities alsoaffect the input of economic activity, suchas the labour stock and public expendi-tures:


21

(2.10) Ri = Ri (TEk)

Again, the circle is closed, and equation(2.7) may be extended to includeequations (2.8) to (2.10):

(2.11) Y = Y{R[D(C(E(Y))),TE (RT(Y))]}

Thus, in this model economic activity,environmental conditions and trafficaccidents are determined simultaneously.

When the input factors of the economyare affected by environmental or traffic-related externalities, prices will change,too, and the structure of the economychanges. Consider e.g. that the laboursupply is reduced due to increased sickleaves, either because of air pollution ortraffic accidents. Then labour becomes ascarcer resource, and the wage rises. Thisimplies that employers will hire feweremployees, so that the labour marketclears. Each industry will generallybecome less labour intensive, and theindustry structure will change. Labourintensive industries will experience highercost increases than other industries, andwill in general diminish. However,demand conditions and the selection ofother input factors in production are alsocrucial, and the final outcome has to befound from the model. As the productionof investment products also faces costincreases, the accumulation of capitaldeclines, so that future productioncapacity is altered, too, even if future sickleaves are not taken into account.

To calculate social costs of externalities byemploying this integrated model, we focuson changes in the present value of GDP inaddition to the valuation of non-marketeffects. Using GDP only as a measure ofeconomic costs may however give a biased

result, at least for two reasons in this case.First, when air pollution causes e.g.increased material corrosion and hospitaladmissions, more economic resources areused for maintenance and health care.However, compared to a situation withoutair pollution, the value added from thisresource use is zero, and should besubtracted from GDP in calculations ofsocial costs. Second, economic welfare isnot a function of production, but ofconsumption. Thus, if investments areincreased today at the expense ofconsumption, GDP will rise in the future,but economic welfare is not necessarilyhigher. This depends on the marginalutility of consumption today and in thefuture, and on the relevant discount rate.

Figure 2.1. Production structure in the MSG-EEmodel


22

Thus, changes in the present value ofconsumption or, even better, moneymetric utility, is a better indicator foreconomic welfare. Whereas the first pointis easily handled within the model, thesecond point is not because the model isnot intertemporal.13 Thus, the relevantdiscount rate is unknown (see howeverthe study by Aasness et al. (1996) usingresults from MSG-EE).

13 In the latest version of MSG (MSG-6), the model isintertemporal (see e.g. Bye (1996)).


23

3.1. IntroductionAir pollution causes increased corrosion ofbuilding materials and motor vehicles.This entails higher maintenance outlaysand increases the user cost of capital. Newknowledge and new methodology nowmake it possible to compute these costs insome detail in Norway. In this study wedo this for the year 1994.

The study is based on the use of geogra-phical information systems (GIS), data onlocal air pollution and distribution ofmaterials at risk. Internationally estab-lished relations between air pollution anddegradation of various materials are alsoemployed. Full use is made of GIS forOslo. Estimates for the rest of the countryare done by extrapolation adjusted forpollution levels and stocks of materials.The project quantifies both direct main-

14 This study was commissioned by the State PollutionControl Authority (SFT) and carried out jointly by theNorwegian Institute for Air Research (NILU), theNORGIT Centre and Statistics Norway. The articklehas earlier been published in Norwegian in Glomsrødet. al. 1996.15 Statistics Norway16 CICERO (Statistics Norway at the time of the study).17 Norwegian Institute for Air Research18 NORGIT-center

tenance costs and the feedback-effects ofsuch costs in the economy as a wholewhen building capital becomes moreexpensive for enterprises and households(indirect costs). The study is a followingup on Glomsrød and Rosland (1988), whomade similar calculations for the year1985.

In later years new and improved descrip-tions of the link between concentration ofair pollution and the decomposition ratefor various materials have been developedinternationally. Moreover, new data onbuilding materials have emerged whichenable more precise computation ofmaterial stocks at risk.

In section 3.2 we present the quantitativerelations between concentrations of airpollution and materials degradation. Airpollution levels and the volume ofmaterials involved are described in section3.3 and 3.4 respectively, while in section3.5 corrosion rates and total maintenancecosts resulting from air pollution arecomputed. Section 3.6 explains themarginal costs of increased SO2emissions. The effects these have for thenational economy are elucidated insection 3.7.

3. Corrosion costs of buildingmaterials and cars in Norway14

Solveig Glomsrød15, Odd Godal16, Jan Fr. Henriksen17,Svein E. Haagenrud17 and Torstein Skancke18


24

3.2. Dose-response and lifetimefunctions for some materials

3.2.1. Current knowledge of dose-response functions

Dose-response functions (see e.g. Lipfert(1987)) describe the physical/chemicalrelations between materials degradationand exposure to pollution. When calcula-ting corrosion damage these must betranslated to capital degradation in eco-nomic terms. The usual approach is to seta criterion for how far corrosion can pro-ceed before maintenance or replacementof a building component has to be carriedout. Use of dose-response functionsenables us to calculate to which extent thelifetime of building elements is affected byincreased pollution levels. The dose-response function is thus transformed intoa damage function.

In the past decade numerous corrosionstudies have been carried out with respectto dose-response and damage functions,material stocks and exposure conditions,see e.g. Haagenrud and Henriksen (1995).With respect to dose-response functions,three studies are particularly prominent:

Lipfert (1987) has performed a synopticstatistical analysis of environmental andcorrosion measurements for importantmetals covered in eight international testprogrammes from up to 72 field stations.Lipfert has carried out a similar survey ofcalcareous stone materials. Dose-responsefunctions are also given for types of paintcoatings.

Two studies carried out by Henriksen etal. (1981) and Haagenrud et al. (1984)contain highly important basic data forNorway in terms of dose-response func-tions for metals. Good statistical analysesare available for two Norwegian towns

(Sarpsborg and Fredrikstad), but moredetailed and synoptic analyses of all datasets have yet to be carried out. The most extensive and best documenteddatabase for dose-response functions isthe ECE-ICP base. The 8 year researchprogram on which it is based, is not yetcompleted, but preliminary results areavailable. Equations for corrosion develop-ment over time have not been developed.However, the ECE-ICP base containsdescriptions of degradation as a functionof SO2, O3 and H+ within a geographicalarea covering the greater part of Europe.It also encompasses considerably morematerials than previous surveys.

Examination of the dose-response func-tions shows that fairly reliable functionsexist for many important buildingmaterials such as metals, painted metal,calcareous stone and the like. The func-tions contain terms describing the effect ofSO2, and where relevant also O3, H

+ con-centration in precipitation and climatevariables expressed as time of wetness(TOW). Time of wetness is defined as thepart of the year with relative humidityhigher than 80 per cent and temperaturehigher than 0°C.

3.2.2. Lifetime functions formaterials

When damage functions are elaborated,account is taken of how far degradationcan proceed before maintenance orreplacement is necessary. In practice thereis a large difference between standardexposure tests and substantive effects onbuildings. It is assumed that maintenanceor replacement is only based on the stateof the materials, and not on other factorssuch as economic value. Damage functionscan be determined directly by fieldinspection through visual description of


25

the state of wear and tear and actual dam-age to buildings, or indirectly by recordingmaintenance performed at regularintervals. When the optimal interval formaintenance or replacement is deter-mined, the damage function is usuallytermed the lifetime function.

Lifetime functions are as a rule dominatedby the most aggressive pollutant. Severalstudies have developed lifetime functionsfor building materials. A comprehensivestatistical sample of different houses invarious pollution areas has been analysedby Kucera et al. (1993). This study,known as the MOBAK study, is the mostcomprehensive of its type and containsresults from Prague, Stockholm and theNorwegian town Sarpsborg. Based on thisstudy, results have been extrapolated tothe national level in Sweden (Andersson1994), and to the European level (Cowelland ApSimon 1994). Lifetimes and main-tenance intervals as a function of variousSO2 levels are available for many buildingmaterials. Using extrapolation techniques,Andersson (1994) has also introducedacid precipitation sensitivity (H+) in thesefunctions when calculating material costsin Sweden.

Thus, lifetime functions may be arrived ateither directly from inspection of buildingsor from dose-response functions. In thelatter case degradation (D) is describedusing linear dose-response functions inc-luding pollution parameters as a deg-radation factor. The general formula usedin our calculations is:

(3.1) D1 = a1⋅SO2 + b1,

or

(3.2) D2 = a2⋅TOW⋅SO2⋅O3

+ b⋅Rain⋅H+ + c2

where a, b and c are constants, SO2 and O3

concentrations are measured in µg/m3, H+

concentrations in mg/l, and Rain ismeasured in meter precipitation per year.Degradation is here measured in thicknessreduction per year.

To arrive at lifetime functions, we notethat lifetime (L) is inversely proportionalto degradation. For most materials a life-time function of the following type isemployed (based on the first dose-response function, equation 3.1):

(3.3) L1 = 1/[a⋅10-3 ⋅SO2 + b⋅10-3 ] = 1000/[a⋅SO2 + b]

These are taken directly from Anderson(1994). However, for zinc and copper thedose-response functions from the ECEproject are employed (i.e., equation 3.2),and the following lifetime function isarrived at:

(3.4) L2 = m/[a⋅TOW⋅SO2⋅O3

+ b⋅Rain⋅H+ + c]

where m is reduction in thickness inmicrometer (µm) before maintenance orreplacement is recommended. Table 3.1shows the selected or derived lifetimefunctions for 14 materials that are used inthis study. In addition, lifetime functionsexist for 3 other materials that are ex-cluded because of lack of material stockdata.

Regarding zink, for galvanised sheets andwire where the mean thickness of zinc is30µm, the premise has been that repaint-ing should be carried out after m=20µmhas corroded, while replacement shouldtake place when all zinc (m=30µm) hasgone. For galvanised profiles with a meanthickness of 80µm, painting should takeplace when m=60µm has corroded.


26

The corresponding equation for copper isdetermined by how much a copper sheetcan corrode and still be in functionalorder. Copper sheets for roofing andfrontages are currently 0.5-0.7mm thick.Due to the unevenness of corrosion, andimpaired strength resulting from corro-sion, replacement is recommended when0.1mm (m=100µm) of the sheet has beencorroded.

Concrete has long been by far the mostused construction material and is there-fore of major economic significance.Concrete breaks down more rapidly in anindustrial and urban atmosphere than inan unpolluted atmosphere. Since thereasons for this are very difficult to clarify,it has not been possible to arrive at gooddose-response functions or damagefunctions. This is because concrete is ahighly complex and complicated material.It is porous, contains a number of addi-tives, and the water/cement mix is initself of great significance. Severalenvironmental variables and othermechanisms influence degradation. Theenvironmental factors are carbonation,temperature fluctuations, dampness,

chlorides (sea salt, road salt), atmosphericpollutants (SO2 and NO2) and solarradiation. Where reinforced concrete isconcerned, carbonation is of greatestinterest, i.e. the reaction between CO2 andconcrete.

Despite the lack of good dose-responsefunctions, as concrete constitutes a sub-stantial share of the materials in allbuilding categories in the survey (seetable A2 in appendix A), it has beenconsidered more important to includeconcrete in the calculations than to omit iton grounds of uncertainty. The lifetime ofconcrete is specified as follows for back-ground and corrosive atmosphere respec-tively (Heinz et al. 1995). Maintenance/lifetime in the background atmosphere(defined as SO2 concentrations below 10µg/m3) is assumed to be 20-80 years, andin corrosive atmosphere (SO2 concen-trations above 10 µg/m3) 10-70 years.This averages out to 50 and 40 yearsrespectively. Hence in our context wehave chosen to use a stepwise lifetimefunction (see table 3.1).

Table 3.1. Lifetime functions for materials at risk

Material name Lifetime function, year

Galvanised steel sheet, replacement L = 30 / (0.51 + 0.0015⋅TOW⋅SO2⋅O3 + 2.82⋅H+⋅Rain)Galvanised steel sheet, maintenance L = 20 / (0.51 + 0.0015⋅TOW⋅SO2⋅O3 + 2.82⋅H+⋅Rain)Galvanised steel wire L = 30 / (0.51 + 0.0015⋅TOW⋅SO2⋅O3 + 2.82⋅H+⋅Rain)Galvanises steel profile L = 60 / (0.51 + 0.0015⋅TOW⋅SO2⋅O3 + 2.82⋅H+⋅Rain)Copper roofing L = 100 / (0.54 + 0.00031⋅SO2⋅O3 + 4.58⋅H+⋅Rain)Strip-lacquered aluminium L = 1000 / (0.107⋅SO2 + 32.6)Strip-lacquered galvanised steel L = 1000 / (0.155⋅SO2 + 38.6)Painted galvanised steel L = 1000 / (0.803⋅SO2 + 84.5)Limestone/Cement plaster L = 1000 / (0.124⋅SO2 + 15.7)Painted plaster L = 1000 / (0.278⋅SO2 + 19.9)Felt roofing L = 1000 / (0.327⋅SO2 + 48.9)Painted/Stained wood L = 1000 / (1.03⋅SO2 + 91.4)Brick IF (SO2<10 then 70 years else 65 years)Concrete IF (SO2<10 then 50 years else 40 years)

Sources: Anderson (1994), ECE-ICP base and Heinz et al. (1995).


27

Brick is also a complicated material that isporous and contains many differentingredients in varying mixes. A number ofdegradation mechanisms are present, and,as in the case of concrete, it is difficult todetermine dose-response or damage func-tions. In the same way as for concrete, theGerman study by Heinz et al. (1995) hascarried out practical studies of lifetime,which is generally longer than for con-crete. Using the same method as for con-crete we arrive at the lifetime function intable 3.1.

3.3. Air qualityCalculating material corrosion due to airpollution requires a quantitative descrip-tion of air quality. The concentration of anair pollutant depends not only on theemission level in that area but also onemissions in other areas combined withmeteorological variables. The best methodfor establishing the concentration at alocation is to measure it. Since there arepractical constraints on measuring concen-tration at all locations, one is dependenton calculating concentrations away frommeasuring stations to obtain a goodpicture of the extent to which, forexample, building materials are subjectedto air pollution.

3.3.1. Modelling pollution in theOslo area

The level of pollution in Oslo is modelledusing NILU's (Norwegian Institute of AirResearch) dispersion model AirQuisModels based on the emission databaseAirQuis Emissions (Grønskei and Walker1993). The model calculates the concen-tration of SO2 and NO2 in grid comprising44x36 (=1 548) squares of 500x500 m2,over a selected time period, whereaccount is taken of emissions from heatingand vehicle traffic as well as factors suchas wind and temperature. A regional

contribution is also included which isbased on measurements outside Oslo. Ourcalculation of material corrosion costs wasdone for the year 1994. Measurementsand calculations from previous studiesprovide the basis for emission data in eachsquare: Data for NO2 emissions are basedon 1991 figures (Gram 1994) and for SO2on 1979 figures (Gram 1982). The SO2data were subsequently adjusted in 1987(without compiling new basic data) basedon known energy consumption in Osloand on known reductions in industrialpoint sources in the preceding few years.

When calculating the lifetime of materials,grid values must be transformed toaverage annual concentrations (annualmeans). In order to minimise uncertain-ties attached to transformation, thedispersion estimates must represent themean distribution of pollutants over theyear. This was done by selecting ascenario where meteorological conditionsare representative for the year. With abasis in a reliable dispersion estimate, thetransformation to annual values based onthe results from two measuring stations inthe centre of Oslo (i.e., Johannes Brunsgate and Hausmanns gate) will producegood annual values for all squares in thecalculation. The mapping of individualsources is poorer for the SO2 databasethan for the NO2 database, whereas thetotal values for SO2 emissions and NO2emissions are approximately equallyaccurate. Hence in the present work weemploy total SO2 emissions, while NO2emissions are distributed on the individualsources heating, vehicle traffic and back-ground, and distributed on the grid on thebasis of each group's contribution in thesquare.

Formation of ozone in the troposphere,i.e. ground-level ozone, is a complex


28

process containing several combinationsof chemical reactions. The outcome ofthese reactions is ozone formed in thepresence of hydrocarbons with NO2 as acatalyst. In the presence of a large surplusof NO, ozone is reduced to oxygen. Thereis a large surplus of NO in areas of highvehicle traffic density. Hence in practiceozone levels are lower in town centresand higher in the areas surroundingtowns. Kucera et al. (1995) present resultsof a four-year comparative measurementin several towns between NO2 and ozonelevels at the same sites. They find that thefollowing equation gives the best descrip-tion of this relation:

(3.5) O3 = 60.5 exp-0.014⋅NO2

with concentrations measured in µg/m3.This equation has been used to estimatean ozone value in each square in Oslo,based on the corresponding grid values forNO2.

We assume that the variables rain, time ofwetness (TOW) and acidity of precipi-tation (H+) remain constant across variousparts of Oslo. For these variables it hasbeen decided to use data from a UN-ECEstation in the centre of Oslo (i.e., Haus-mannsgate) as a mean for the period1987-1993 (see table A1 in appendix A).

3.3.2. Data for the rest of thecountry

In estimating material corrosion costs forthe rest of the country, we have chosen tosingle out 15 towns and urban areas (inaddition to Oslo). These are locationswhere either high SO2 concentrations or alarge building stock are present, andhence where a high proportion of the totalmaterial corrosion costs are likely to beincurred. The rest of the country isdivided into three groups (see table 3.2).

For each area data for the annual meanSO2 and ozone concentration and valuesfor pH, precipitation and time of wetness(TOW) are required. These parametersare shown in table A1 in appendix A.

As opposed to the pollution data for Oslo,we only use one representative value forthe other towns. Thus, the calculations forthese towns are not as precise as the cal-culations for Oslo. However, the variationin air pollution is probably higher withinOslo than within most other towns. More-over, an earlier study by Glomsrød andRosland (1988) indicated that one third ofthe Norwegian material corrosion costsare incurred in the capital.

The values for SO2 are annual meanstaken from NILU's monitoring programmefor the SFT (Norwegian Pollution ControlAuthority) for towns and urban areas(Hagen 1994) and the SFT's own monitor-ing programme in the industrial regionGrenland. Ozone and pH in precipitationare based on data from SFT's monitoringprogramme for long-range air pollution(SFT 1994). Data for precipitation aretaken from The Norwegian MeteorologicalInstitute's tables for the past 30 years'mean (Aune 1993; Førland 1993), andtime of wetness data are recorded byNILU's own stations which measuretemperature and relative humidity (Ofstad1995). Where measurement data are notavailable, data are generated by extra-polation from the closest available datasets.

The three remaining groups are urbanareas19 in the southern part of Norway(i.e., from Sør-Trøndelag and south-wards), urban areas in the northern part

19 Urban areas with more than 2 000 inhabitants excl.those already included in the analysis.


29

of Norway, and the rest of the country(mainly rural). In the north the values forsulphur dioxide, ozone, rain, time ofwetness and acidity of precipitation forTromsø were used for all towns and urbanareas. In the south, 3 µg/m3 for SO2 and55 µg/m3 for O3 were chosen for all townsand urban areas. For rain, time of wetnessand acidity of precipitation the arithmeticmean values from the 15 selected townsand urban areas in Southern Norway wereused. Account has not been taken of thefact that precipitation is higher in westernNorway, as this is only significant forcalculations for zinc and copper. For therest of the country, an SO2 concentrationof 1 µg/m3 (which is equal to the back-ground concentration – see the nextsection) is chosen. The other data aresimilar as those for urban areas in thesouth.

3.3.3. Local and long-range contri-butions to pollution

The long-range contribution to SO2 con-centration in Norway is at present about 1µg/m3. The remainder is due largely tolocal combustion of petroleum products.Only in the vicinity of certain industrialenterprises will process emissionsdominate (e.g., in Sarpsborg and Skien).

Ozone formation is as stated more com-plex and the local contribution is difficultto estimate. Several calculations havebeen made to quantify ozone levels with-out (local and long-range) contributionsfrom air pollutants, and 40 µg/m3 appearsto be a figure acceptable to a number ofresearchers (Bojkov 1986). However,most of the difference between this back-ground level and the actual ozone concen-trations in Norway is due to long-rangecontributions. Still, in lack of otherestimates we will regard the exceeding of40 µg/m3 as a local contribution. This is

clearly a major simplification. However, aswill be seen in section 3.6, ignoring localcontributions of ozone reduces the locallyimposed costs by only 1.4 per cent. Thus,pollution data on SO2 are clearly the mostcrucial.

For NO2, which affects the ozone levels,background concentrations in Norway are3-4 µg/m3. The remainder is attributableto local emission sources where vehicletraffic dominates over heating. Thepercentage distribution between vehicletraffic and heating will vary from squareto square in Oslo and from town to townin Norway.

For pH in precipitation the local contri-bution is minimal, and in the calculationsall acidification has been assumed to berelated to long-range pollutants.

3.4. Stock of materials at risk

3.4.1. MethodThere exist no complete statistics onoutdoor use of materials in Norway. Weare therefore consigned to estimate thestock of various materials at risk. Severalmethods are available for this purpose.Glomsrød and Rosland (1988) tried tocalculate the total amount of buildingmaterials based on the volume of dome-stic production over a given periodadjusted for imports and exports. Sincethen, information has emerged that makesother types of analysis possible.

In this study we start out from investiga-tions that have been made on the use ofmaterials in varous buildings and infra-structure. An important contribution hasbeen the results of the comprehensiveMOBAK study undertaken in Stockholm,Prague and the Norwegian town Sarps-borg (Kucera et al. 1993). Thorough


30

fieldwork was carried out on surveyingthe condition and material-mix of varioustypes of buildings. One of the conclusionswas that the mix of materials for varioustypes of buildings in Sarpsborg and Stock-holm was almost identical whereas adifferent mix was found in Prague. Oneexplanation given for this result was thatthe Nordic countries share a similararchitectural tradition. This may in turnbe due to the relatively uniform availa-bility of various types of buildingmaterials and the similarity of climate.Hence it was considered that the mix ofmaterials in buildings in Sarpsborg couldbe extrapolated to the rest of Norwaywithout major margins of error.

A number of materials not classified asbuildings are subjected to corrosioncaused by air pollution. Part of this stockof materials can be classified under theterm infrastructure. We have includedcontributions from wire fencing aroundbuildings, high-voltage transmission linesfor electricity and railways, lampposts,traffic signs and the motor vehiclepopulation under this term. The material-mix for these items is also based on theMOBAK study. Table A2 in appendix Ashows average size and material-mix foreach building category.

Now, the following procedure shows howwe will calculate the stock of materials atrisk, where point 2 is based on the resultsof the MOBAK study:

1. Derive total area of building materialsfor various building categories andinfrastructure (dwellings,manufacturing, agriculture, officebuildings, service buildings and motorvehicles) in each location.

2. Make use of the recorded mix ofmaterials for an average building ofeach main type.

In the following sections we will describehow we have derived at the figures out-lined in point 1.

3.4.2. Stock of materials at risk inOslo

In Oslo we have used the buildingregister, GAB (the Norwegian acronymfor: Properties, Addresses, Buildings), toobtain information on point 1 above. GABrecords all buildings exceeding 15 sq.m.throughout the country. The registerincludes detailed information on thegeographical location as well as thebuilding's purpose. The register has madeit possible to distribute the building stockon a grid covering Oslo whose square sizematches that of the pollution parameters(500m x 500m). Within each square wehave information on the number ofbuildings distributed by purpose.

Retrieving information from this registeris highly resource-demanding. Hence wehave had to confine ourselves to Oslo andthen employ various extrapolationtechniques to arrive at national estimates.As mentioned above, Oslo was a naturalchoice since previous estimates (Glomsrødand Rosland 1988) showed that a third ofthe national material costs are incurred inOslo, where both the concentration ofbuildings and air pollutants are high.

3.4.3. Stock of materials at riskoutside Oslo

In regard to geographical distributionoutside Oslo, we follow the same divisionas in section 3.3.2. In order to extrapolatefrom Oslo to other regions, we have toknow what proportions of the varioustypes of building are to be found in Oslo


31

compared with the other municipalities.Such information can be retrieved fromvarious statistics containing municipaldata. Stocks of material in the variousmunicipalities are then estimated asfollows.

Dwellings, which include all privatehousehold buildings, are divided intosmall houses and apartment blocks, eachshowing a different geographical distri-bution. Apartment blocks are more heavilyrepresented in towns than in peripheraldistricts. They will therefore, relativelyspeaking, incur higher corrosion costs.Figures from the Population and HousingCensus 1990 (Statistics Norway 1992a)show the number of different types ofhousing distributed by municipality. TheSurvey of Housing Conditions 1988(Statistics Norway 1990) shows theaverage size of each dwelling type. Usingthese statistics we have arrived at amunicipal distribution of the housingcapital.

For distribution of materials in industrialbuildings (mining/manufacturing), figureshave been taken from ManufacturingStatistics 1992 (Statistics Norway 1994a)stating the fire-insurance value ofindustrial buildings at municipal level.Our assumption is that the distribution ofmaterial stocks follows the distribution offire-insurance values of the buildings.For office/commercial/transport, hoteland restaurant, and public and privateservices, the choice of distribution is bynumber of employees in the respectivetrades as stated in the Population andHousing Census 1990 (Statistics Norway1992a).

In primary industries, i.e. agriculture,forestry, fishing etc., farm buildingspredominate. We have therefore taken a

basis in the Census of Agriculture andForestry for 1989 (Statistics Norway1992b) which states the floor area ofoutbuildings (1000 sq.m.).

Buildings that are not specified in the GABregister (called other buildings), aremainly attached to small houses. Hencethese buildings have been assigned thesame geographical distribution as smallhouses.

For infrastructure we have calculated theaverage stock of materials per capita anddistributed the result on towns/regions inrelation to population. For Oslo we havecalculated the total stock of materialsusing the population figure and thendistributed the result on the grid inproportion to the building area. In thisway the city centre, which contains arelatively large amount of infrastructureper capita, is assigned a larger share ofmaterials. Since it has not been possible toapply such considerations to the rest ofthe country, distribution of material stockshas been carried out directly based onpopulation.

Distribution of material stocks for all typesof buildings in the areas ‘urban south’,‘urban north’ and the ‘rest of the country’has been done using population figures.The resulting keys used to distributebuilding capital among regions are set outin table A3 in appendix A. Table 3.3 insection 3.5.4 shows the stocks of variousbuilding categories for the whole country.Based on this information, distribution ofmaterial stocks in the various regions andon various types of buildings are calcula-ted (see tables A4 and A5 in appendix A).


32

3.5. Maintenance costs

3.5.1. AssumptionsSubstantial maintenance is carried out onexterior surfaces of buildings since thematerials making up the frontage undergoa continuous process of degradation. Thisprocess can, for our purposes, be split intotwo main components: Natural degrada-tion which takes place irrespective of airquality, and degradation due to emissionof pollutants into the atmosphere.

The owner of the building is him/herselfin a position to decide when maintenanceis to be carried out. A reasonable assum-ption is that owners wish to minimisetotal outlays on maintenance of thebuilding over time and will thereforeactively adapt maintenance frequency tomaintenance needs. We also assume thatowners know when maintenance shouldbe carried out with a view to minimisingcosts.

When calculating costs of corrosion ofbuilding materials it is not enough to con-sider the actual cost of maintenance work.The maintenance that would have beenrequired in a “clean” immediate environ-ment with natural corrosion (backgroundcorrosion) must be subtracted to obtain acorrect estimate of material costs due toair pollution.

Our calculations incorporate two mainassumptions about agents' informationand patterns of behaviour. First, we haveassumed that the owner of the buildingknows when maintenance of the buildingshould be performed in order to minimisecosts in the long term and that he/sheperforms the work at this point. Second,as mentioned earlier, we disregard othermotives behind replacement or mainten-ance work than material degradation. If

the agent performs maintenance moreoften than required by the criterion ofphysical wear and tear, e.g. because ofaesthetic reasons or because the economiclifetime is shorter than the physical, airpollution may not alter agents' mainten-ance frequency. If this frequency is higherthan optimal maintenance frequencybased on technical criteria, air pollutionwill not result in increased costs for theowner. On the other hand, for ownerswith an initially optimal maintenancefrequency who do not adapt this frequen-cy to the increased wear and tear causedby air pollution, the long-term costs mayexceed our calculations. Information fromthe construction industry suggests thatagents are more concerned with adaptingbuilding maintenance to budget con-straints than to maintenance needs. Thesum total of these sources of uncertaintyresulting from lack of information gives noclear indication as to its effect on costs.

3.5.2. Model for calculating costsWe will define the material corrosion costsdue to national air pollution (Kp) as thedifference between the total corrosioncosts (Kt) and the corrosion costs thatwould occur in the background atmos-phere (Kb):

(3.6) Kp = Kt - Kb

Moreover, we can state the corrosion costsof a given material j as an expression ofthe total stocks of the material (Mj),maintenance costs per square metre of thematerial (Cj), and the lifetime of thematerial (Lj). The lifetime of the variousmaterials differ between various environ-ments as discussed in section 3.2.2 (seetable 3.1). Thus, we have the followingtwo expressions for corrosion costs ofmaterial j in, respectively, the background


33

atmosphere (b) and the actual atmosphere(t):

(3.7) KM C

Ljb j j

jb=⋅

(3.8) KM C

Ljt j j

jt=⋅

The numerator indicates the cost ofcutting the lifetime of the material by oneyear. Inserting (3.7) and (3.8) into (3.6)we obtain

(3.9) K M CL Lj

pj j

jt

jb= ⋅ ⋅ −

1 1

This expression may be summed over allmaterials, and over all regions.

3.5.3. Maintenance pricesFor most materials in the analysis main-tenance is in the form of cleaning andrepainting. Only galvanised steel wire,roofing felt and copper roofing areinvariably replaced. For galvanised steelsheeting we assume that 50 per cent ismaintained and 50 per cent replaced. Theprices of maintenance and replacement ofmaterials depend inter alia on the extentof damage, building size and shape,design, material quality and choice ofcontractor. The data, based on infor-mation provided by the industry, arepresented in table A6 in appendix A.

Part of the building maintenance workdone to remedy material corrosion isperformed by the owner in person, and insuch cases the maintenance work is not aservice traded in the market. Still, weassume that the market price is the bestestimate to use here, too.

3.5.4. Cost calculationsFrom sections 3.3 and 3.4 we have dataon, respectively, air quality and stocks ofeach individual material distributed bybuilding type. These data are given foreach of the 1 584 squares of Oslo as wellas for the other 18 towns/areas. Actuallifetime and lifetime in backgroundatmosphere of each material in eachgeographical location is then calculated asdescribed in section 3.2.20 By applying themaintenance prices shown above, materialcorrosion costs are calculated for eachmaterial and geographical location usingequation (3.9). The results are set out intable 3.2 to table 3.4, where the totalcosts are distributed on region, buildingtype and material type, respectively. More

20 With 1 584 squares, 9 building types and 14 typesof material, the total costs in Oslo is aggregated from199 584 separate cost units.

Table 3.2. Material corrosion costs by region in 1994. Mill. 1995-Nkr

Region Costs Per cent

Halden 1.8 0.9Sarpsborg 5.4 2.7Fredrikstad 2.4 1.2Moss 1.6 0.8Bærum 7.1 3.6Asker 2.6 1.3Oslo 74.4 37.6Drammen 3.8 1.9Porsgrunn 2.1 1.1Skien 13.2 6.7Bamble 0.5 0.2Kristiansand 2.1 1.1Stavanger 8.6 4.4Bergen 22.6 11.4Trondheim 9.2 4.7Tromsø 0.8 0.4

Urban-south 34.8 17.6Urban-north 3.6 1.8Rest of the country 1.3 0.7

Total 197.7 100.0


34

detailed results are shown in tables A7and A8 in appendix A.

The geographical distribution of the costsshows that Oslo incurs more than one-third of the costs even though only 12 percent of the stock of materials at risk islocated here. This is due to a highconcentration of SO2. On the other hand,the category ‘rest of the country’ which

has about 40 per cent of the stock ofmaterials, carries less than one per cent ofthe costs because of the low pollutionlevel in this category (the SO2 concen-tration is set equal to the backgroundconcentration, and so the costs are onlydue to ozone concentration). The distri-bution of costs on building types showsthat dwellings (small houses and apart-ment blocks) incur almost half the costs,

Table 3.4. Total material stocks, material corrosion costs and costs per square meter, 1994.Mill. 1995-Nkr

Material type Total materialstock, mill. m2

Percent

Totalcosts

Percent

Average cost,Nkr/m2

Galvanised steel, untreatedsheet maint. 15.3 1.8 11.2 5.7 0.74sheet repl. 15.3 1.8 13.7 6.9 0.90wire 5.2 0.6 1.9 1.0 0.36profile 3.0 0.4 1.6 0.8 0.52

Galvanised steel, treatedstr.-lacq. 64.9 7.6 5.0 2.5 0.08painted 67.0 7.8 51.1 25.8 0.76

Aluminium, str.-lacq. 28.9 3.4 1.5 0.8 0.05Copper 1.3 0.2 0.3 0.2 0.25Wood, stained/painted 269.9 31.5 53.6 27.1 0.20Plaster, untreated 28.4 3.3 5.0 2.5 0.18Plaster, painted 71.9 8.4 22.3 11.3 0.31Concrete 97.1 11.3 14.2 7.2 0.15Roofing felt 82.4 9.6 14.6 7.4 0.18Brick 105.1 12.3 1.6 0.8 0.02

Total 855.6 100.0 197.7 100.0 0.23

Table 3.3. Stocks and corrosion costs of building types for the whole country, 1994. Mill. 1995-Nkr

Building Area,mill. m2

Percent

Corrosioncosts

Percent

Small house 282.1 33.0 42.7 21.6Apart.block 107.1 12.5 47.1 23.8Manufact. 24.5 2.9 4.8 2.4Office 191.2 22.3 46.4 23.5Hotel 1.7 0.2 0.5 0.2Service 27.6 3.2 6.7 3.4Agriculture 128.0 15.0 10.8 5.5Other 50.4 5.9 8.8 4.5Infrastr. 43.1 5.0 30.0 15.2

Total 855.6 100.0 197.7 100.0


35

whereas e.g. the manufacturing buildingsonly incurs 3 per cent of the costs.

From the distribution of costs on materialswe see that metals' share of total costs islarger than metals' share of total materialstocks. The reason is that metals are themost sensitive materials to SO2 pollution.This is particularly reflected in the cate-gory ‘infrastructure’ which accounts for15 per cent of the costs but only 5 percent of the material stock. A quarter of thecosts can be ascribed to painted or stainedwood. This is not a sensitive material, butthe stock at risk is large. Brick and con-crete incur costs in only three towns (seetable A7 in appendix A). This is becausethe lifetime functions for these materialsare assumed to be stepwise functionswhich are only triggered where SO2

concentration is higher than 10 µg/m3.

3.6. Marginal corrosion costs of SO2emissions

In public decision-making related to thisissue it will be useful to know themarginal social costs of more emissionsdue to increased material corrosion, or inother other words the marginal gains ofreducing emissions. Then this and otherenvironmental gains may be compared tothe marginal costs of implementing anemission reduction initiative. In thischapter we therefore calculate themarginal costs of SO2 emissions due tomaterial corrosion.

Systems for monitoring air quality anddevelopment of dispersion models showthat for SO2 the link between emissionand concentration may be well describedby a linear relationship. Furthermore, asthe main part of SO2 concentrations arecurrently caused by local emissions, weassume that, except for the backgroundlevel, SO2 concentration in a municipality

depends exclusively on emissions in thismunicipality. Thus, we do not considertransport of emissions between munici-palities. This is a reasonable simplificationfor most of the areas we study. Moreover,we shall assume that the contribution toSO2 concentration of a tonne of SO2

emitted in a municipality is independentof the source of the emission. Thisassumption is tenable if particularly largechanges in point sources are disregarded.A percentage increase in emissions willthen produce the same percentageincrease in concentration adjusted for thebackground level.

A corresponding relation between ozoneconcentrations and NOx emissions is farmore complicated, since NOx is involvedboth in the formation and decompositionof O3. While decomposition predominatesin town centres, ozone formation is therule in the outskirts of towns away fromthe main traffic arteries. Since reliablequantitative descriptions of these effectsare not available, we have been unable tocalculate marginal costs for nitrogenemissions.

It is easily seen from equation (3.9) thatfor 7 of the 14 materials in table 3.1,which constitute 77 per cent of the totalcosts, the relation between SO2 concen-tration and costs is linear. This is becausematerial lifetime is inversely proportionalto the concentration of SO2, and costs areinversely proportional to lifetime. Forconcrete and brick, which constitute 8 percent of the costs, the stepwise functionsimply that the marginal costs are zeroexcept at the point where concentrationspass 10 µg/m3, where marginal costs areinfinite. Of course, this is a pure simplifi-cation, and we will assume in this sectionthat the marginal costs are equal to theaverage costs in each region. This is


36

equivalent to assuming that the derivedcosts were obtained from a linear functionof the concentration. For the last 5 mat-erials the multiplicative ozone concen-tration term in the lifetime functionscomplicates the matter as the marginalcosts of SO2 concentrations depend on thelevel of the ozone concentrations. In ourcalculations we shall assume that theozone concentrations are fixed at thecurrent level.21 Then the relation betweenSO2 concentrations and costs is linear forthese materials, too.

The easiest way to calculate marginalcosts of SO2 emissions is now to split thetotal corrosion costs for each region intotwo components, where the first com-ponent is the corrosion costs with SO2

concentrations at its background level.These costs are for some materials equalto zero, for others solely due to ozoneconcentrations. Then the second costcomponent is the contribution from SO2

concentrations when ozone concen-trations are at the current level. From thediscussion above we established that thiscost component is proportional to theconcentration levels (in excess of thebackground levels). Since emissions areproportional to these same concentrationlevels, the second cost component can bedirectly divided by the emission levels foreach region in order to calculate marginalcosts, i.e. the marginal costs equal theaverage costs of this component.22

21 If we had rather fixed the ozone concentrations atthe background level, the marginal costs of SO2

emissions would have decreased by 2-3 per cent.22 Calculating the second cost component for Oslo iscomplicated by the fact that concentration levels forSO2 and ozone vary across the grids. We have ratherused the central value in the concentration intervalsshown in table A1 in appendix A, i.e. 14.5 and 34µg/m3 for SO2 and ozone, respectively. If we hadchosen the most conservative estimates, i.e. thehighest value for ozone and the lowest value for SO2,

SO2 emissions distributed at municipallevel by emission source are available(Statistics Norway 1995a). The figures, setout in table 3.5, are exclusive of emissionsfrom international air traffic above 1000metres and sea traffic outside the Nor-wegian economic zone. Since emissionfigures distributed by municipality do notexist for 1994, national 1994 figures foreach source are used combined withmunicipality distribution from 1992.

Except for Oslo and ‘Rest of the country’the costs of SO2 emissions are between 97and 100 per cent of the total corrosioncosts (compare table 3.2 with table 3.5).For Oslo the costs of SO2 emissions areactually higher than the total costs. Thereason is that the concentration level ofozone in Oslo is below its backgroundlevel, so that the costs due to ozone isnegative in the capital. In ‘Rest of thecountry’ the concentration level of SO2

was assumed to equal its backgroundlevel, which means that the whole cost isattributable to the ozone concentration.

The resulting marginal costs, also pre-sented in table 3.5, show wide regionalvariations. However, the figures should betaken by caution as we have not con-sidered transport of emission between theareas. This is apparent when comparingthe results for e.g. Porsgrunn and Skien,which are neighbouring towns. Thefigures indicate that a significant amountof emissions in Porsgrunn are transportedto Skien. The highest marginal costs are inareas where total costs are high comparedto emissions, as for example in Bergenand Oslo. These are towns with a highdensity of buildings and infrastructure. On

the second cost component would have decreased bymerely 0.5 per cent. Moreover, we do not considerthat marginal costs of SO2 emissions actually varyacross the grids.


37

the other side, the lowest marginal costsare found outside the towns where thematerial density is low.

In Statistics Norway (1995a) theemissions in each area are distributed onsix separate sources (i.e., 4 mobilesources, stationary emission and processemission). Thus, as we assumed that themarginal effects of emissions from varioussources in a municipality are equal, thecosts displayed in table 3.5 can also bedistributed on the various emissionsources. Then, by summing over thewhole country, the total and average costof emissions from one specific source on anational level can be easily calculated.This is shown in table 3.6, together withthe national SO2 emissions distributed bysource. We emphasise that we ignoredifferences in average costs within aspecific area. In e.g. Oslo the average cost

of 70.9 Nkr/kg is assumed to apply to allemission sources.

The highest average cost is incurred bymobile road-traffic sources where a tonneof SO2 inflicts damage worth about Nkr11 700. The process industry contributesjust under half of the total SO2 emissionsin Norway, but incurs less than a quarterof the costs. Thus, the average cost isrelatively low, i.e. about Nkr 2 400. Thereason is that this industry is often locatedin sparsely populated areas where theconcentration of buildings is low. A largeshare of SO2 emissions from ships comesin ‘Rest of the country’, where the costsare zero. However, due to large emissionsin the harbour of Oslo and other bigtowns, the average cost of emissions fromthis source is not as small as for theprocess industry.

Table 3.5. SO2 emissions, total and marginal costs of SO2 emissions by region, 1994. Mill. 1995-Nkr

SO2 emissionstonnes

Costs of SO2

emissionsMarginal costs

Nkr/kg SO2

Halden 68 1.8 25.9Sarpsborg 1 669 5.4 3.2Fredrikstad 885 2.4 2.7Moss 664 1.5 2.3Bærum 119 7.1 59.7Asker 62 2.6 41.1Oslo 1 051 74.5 70.9Drammen 74 3.8 51.6Porsgrunn 556 2.1 3.7Skien 229 13.1 57.4Bamble 24 0.5 19.6Kristiansand 1 008 2.1 2.1Stavanger 277 8.5 30.6Bergen 263 22.3 84.8Trondheim 674 9.1 13.5Tromsø 108 0.8 7.3

Urban-south 17 886 34.0 1.9Urban-north 5 798 3.5 0.6Rest of the country 13 120 0.0 0.0

Total/average 44 535 194.9 4.4


38

3.7. Macroeconomic effects ofmaterial corrosion

We have thus far seen how air pollutionreduces the lifetime of various materialsand how this increases the costs ofbuilding-stock maintenance. Buildings arean important economic factor, both asdwellings for households and as an inputin the production of goods and servicesfor enterprises. Although the estimatedextra maintenance costs are not heavy inthe national economic context, theindirect effect of changes in corrosioncosts for economic adjustment may besubstantial compared to the estimatedextra maintenance costs.

Air pollution increases the cost of buildingmaintenance and thus the user cost ofcapital rises. This implies that agents mayreduce their use of buildings and increasetheir use of other factors to adjust to theprice changes. Investment in new buildingcapital is reduced and the total buildingstock scaled down in the long term. Sincethe economy's growth potential dependsin part on the availability of capital, areduction in the stock of building capitalwill lead to reduced economic growthcompared with a situation with no airpollution. Thus, there are allocation

effects in the economy which must beanalysed in addition to the direct main-tenance cost.

In the short term agents will not changetheir use of buildings even though itbecomes more expensive to maintainthem, since adjustment costs may be high.However, analysed within a long-termmodel, it is reasonable to assume thatenterprises' and households' adjustment interms of building choice is optimal andflexible.

These allocation effects cannot beobserved. The only way to calculate themis to carry out controlled experiments inmodels that simulate economic activity.Several models exist that simulate theNorwegian economy, each being designedto study various problems. Hence thechoice of model must be based on theproblem to be elucidated. We have chosenthe MSG-EE model (Alfsen et al. 1996)which is designed to analyse questionsassociated with energy, transport and theenvironment (emissions to air). Themodel is outlined in chapter 2 of thisbook. The user cost of capital dependsinter alia on material corrosion, which inthe model have been linked to the

Table 3.6. SO2 emissions and material corrosion costs distributed by SO2-source, 1994

Stationary Mobile sources Process Total

Road andconstr.mach.

Rail Aircraft Ships

SO2 emission- Tonnes 7 932 3 745 76 132 13 418 19 232 44 535- Per cent 17.8 8.4 0.2 0.3 30.1 43.2 100.0

Corrosion costs- Mill. 1995-Nkr 63.1 43.7 0.4 0.6 40.6 46.6 194.9- Per cent 32.3 22.4 0.2 0.3 20.8 23.9 100.0

Average cost(Nkr/kg) 7.9 11.7 5.1 4.4 3.0 2.4 4.4


39

pollution level following the coursedocumented in this chapter.23 It is used aweighted average of the SO2 concen-trations in the various areas included inthis work. The quantitative relationshipbetween the user cost and SO2 emissionswill vary for the various economic sectorsowing to differing material-mix andgeographical distribution. We havedivided the sectors into four; primaryindustry, manufacturing industry, services(incl. public services) and dwellings. Thenthe costs for different building typesdisplayed in table 3.3 are distributed onthese four groups of sectors, and furtherdistributed on the individual sectorsaccording to building capital values atend-1994. The costs for infrastructure aredistributed on each sector according to carcapital values.24 The resulting figures areshown in table 3.7. We see that corrosioncosts per building capital value are highestfor dwellings, but the costs per capitalvalue are even higher for infrastructure.The building capital of the manufacturing

23 Because of the difficulties in distributing thecorrosion costs on SO2 and ozone concentrations, forsimplicity all costs are attributed to SO2 concen-trations in this section. This will probably give anegligible bias on the allocation costs.24 As no final figures were available for the stock ofbuilding and transport capital as at end-1994, thefigures presented are based on national accountsfigures from 1991 adjusted for capital depreciationand gross investment (Statistics Norway 1993a,1995b).

industry, on the other hand, is hit quitemodestly compared to the value of theirbuilding capital.

The model estimates changes in grossdomestic product and other variablesprior to and after inclusion of materialcosts from air pollution. Since emissionsto the atmosphere influence the user costof buildings and infrastructure, and thusthe level of investment, economic activitywill be affected several years after theemissions take place. Future allocationeffects resulting from an increase incurrent emissions are discounted by 7 percent per year. The effects prove to begreatest in the years immediately follow-ing the emission increase; after about 10years there is no longer any measurableeffect as economic agents gradually moveback to the reference path of theeconomy.

We will use the discounted changes ingross domestic product as an indicator ofthe allocation costs. Then we find that themarginal allocation cost at the nationallevel caused by SO2 emissions is 2.1 1995-Nkr per kg, which is about one-half of themarginal maintenance costs (see table3.5). Assuming constant marginal costs,the total calculated allocation cost comesto Nkr 93 million. The total economiccosts equal the sum of maintenance costsand the allocation costs.

Table 3.7. Capital stock and material corrosion costs by sector group, end-1994. Mill. 1995-Nkr

Sector group Capital value Share Corrosion costs Share

Building capitalPrimary industry 94 082 4.9 10.8 5.5Manufacturing industry 331 365 17.2 4.9 2.5Services 838 232 43.5 55.2 27.9Dwellings 638 805 33.1 100.4 50.8

Infrastructure 26 265 1.4 26.6 13.4

Total 1 928 748 100.0 197.7 100.0


40

Allocation costs constitute about 50 percent of maintenance costs. The modelcannot compute regional effects but,based on a reasonable assumption ofproportionality between allocation costsand maintenance costs, the total marginalcosts in each region can be estimated bymultiplying the figures in table 3.5 by1.47. For Oslo the marginal cost includingthe allocation cost then comes to aboutNkr 105 per kg of SO2.

3.8. Change since 1985A calculation of material costs resultingfrom SO2 emission in Norway in 1985found that maintenance costs totalled_220 million 1985-Nkr (Glomsrød andRosland 1988). Applying the housingconstruction cost index this gives a figureof 348 million 1995-Nkr. When comparingthis result with ours, two importantchanges must be taken into account. First,SO2 emissions have been reduced by60 per cent in the period 1985-1994.Second, we have included far morematerials in the present work. These twoeffects pull in opposite directions in terms

of costs, as our results show a slightreduction in maintenance costs.

If the level of pollution had remainedunchanged since 1985 (111 600 tonnes ofSO2 emissions), maintenance costs due toSO2 emissions calculated by today'smethods would be 489 million 1995-Nkrper year, i.e., maintenance expenditure onbuildings have been reduced by294 million Nkr in the period 1985-1994(see table 3.8).

Table 3.9 shows how the total costs, i.e.,including the allocation costs, havechanged since 1985. Here we haveincluded costs from ozone concentrations,too, which we have assumed to beunchanged since 1985. The allocationcosts are further assumed to be propor-tional to the direct maintenance costs. Itshould be noted that the saving iscalculated for the year 1994 and not forthe entire period 1985 to 1994. Theannual saving prior to 1994 was smallerinasmuch as SO2 emissions were largerthan in 1994. Total costs in 1994 were

Table 3.8. Changes in SO2 emissions and maintenance costs 1985-1994


Road andconst.mach.

Rail Aircraft Ships

1994SO2 emissions (1 000 tonnes) 7.9 3.7 0.1 0.1 13.4 19.2 44.5Costs (Mill. 1995-Nkr/yr) 63 44 0.4 0.6 41 47 195Marginal costs (Nkr/kg) 7.9 11.7 5.1 4.4 3.0 2.4 4.4

1985SO2 emissions (1 000 tonnes)1 31.4 6.1 0.2 0.4 26.4 47.2 111.6Costs (Mill. 1995-Nkr/yr)2 250 71 1.1 1.6 80 114 489

1985-1994Emission reduction 1985-1994(1 000 tonnes) 23.5 2.3 0.1 0.2 12.9 28.0 67.1Cost reduction 1985-1994(Mill. 1995-Nkr/yr) 187 27 0.7 1.0 39 68 2941 The distribution of mobile emissions in 1985 is not known and is calculated using the same weights as in 1994.2 Using present calculation method.


41

Nkr 431 million lower than in 1985 as aresult of the reduction in SO2 emissions.

3.9. Uncertain factorsSome uncertainties attach to the calcula-tions presented, and we will brieflydescribe the most important ones. Theseare, however, difficult to quantify.

Regarding the chemical-physical relationbetween air pollution and material corro-sion (dose-response functions), a mainproblem is that damage functions onlyexist for a small selection of materials.Recent decades have seen the develop-ment and use of a number of new buildingmaterials for which dose-response func-tions are not available. Developing suchfunctions is very time-consuming since thematerials must be exposed over a longperiod before the results can be processed.The ongoing test programme under theauspices of UN-ECE, e.g. Haagenrud andHenriksen (1995), extends over eightyears. Good descriptions are unavailableparticularly for cement, tiles, andconcrete. These materials are employed soextensively that the economic implicationsmay be substantial. Air pollution is knownto have a harmful effect on othermaterials too, e.g. asphalt, rubber,textiles, sealing and electrical contactmaterials. The relations have so far notbeen quantified. The absence of dose-response functions for a number ofmaterials suggests that our estimate formaterial costs is on the low side.

The key to calculating stocks of materialsin the present report has been the MOBAKstudy from Sarpsborg which involvedextensive fieldwork inspection ofbuildings. The results from that studyhave been used in the present report formaterial stock calculations for the entirecountry. The uncertainty lies in how farSarpsborg is representative in terms ofmaterial-mix for the rest of the country.The MOBAK study shows close corre-spondence between material use inSarpsborg and Stockholm. Hence we canbe fairly confident that it provides a gooddescription of material use throughoutScandinavia. Another uncertainty arisesfrom the lack of material stock calcula-tions for a number of building types. Thesector ‘Production and distribution ofelectrical power’ accounts for a substantialstock of building capital (9 per cent of thetotal in value terms). Although large partsof this capital are located in areas of lowpollution, the sector should have receivedspecial treatment in the study.

Costs of corrosion and weathering ofbuildings of historical value have not beencalculated in the present study. A mone-tary value would have to be given to theindividual cultural monument in order totake these costs into account.

Traditionally material corrosion has beenassociated exclusively with sulphurpollution because sulphur has been thepredominant component of pollution inrecent decades. Since recent years haveseen a shift away from sulphur emissionstowards ozone and nitrogen pollution,researchers have postulated that thesepollution components also influencematerial corrosion, both alone and insynergy effect with each other. In thepresent report ozone has been incor-porated as a variable for five materials.

Table 3.9. Comparison of yearly costs due to material corrosion. Mill. 1995-Nkr

Costs Saving

1985 1994 1985-1994

Maintenance costs 491 198 294Allocation costs 230 93 137Total costs 721 291 431


42

Ozone also has (so far unquantified)effects on other materials. Other pollutionparameters also influence the costs ofmaintaining building capital. A pertinentexample in this context is soiling.Deposition of dust and soot particlesdiminishes buildings' economic lifetimeand aesthetic quality. Today considerableresources are spent on exterior cleaning ofbuildings, especially window-cleaning.

The allocation cost is calculated using amodel that simulates the entire nationaleconomy. Such models will of coursesimplify reality and therefore produceunceratin results. The allocation costsfurther depend on maintenance costs, andhence, the uncertainties attached to thesecalculations will influence the accuracy ofthe allocation cost.

The calculation of marginal costs is basedon maintenance costs and a number ofadditional assumptions regardingemissions and dispersion of SO2. Thesecalculations are therefore more uncertainthan the total maintenance costs calcula-tions.

3.10. ConclusionEven with the uncertainties discussedabove, the analyses have been improvedin several ways compared to a few yearsago. Thanks to the ECE materials project,more realistic dose-response functions areavailable today for a greater number ofimportant materials than previously.Moreover, it has now been documentedthat NO2 and O3 have a stronger corrosiveeffect on many materials than earlierthought. Substantial improvements in thecalculation base have also been achievedby introducing a GIS-based modellingtool, which have been linked up to theNorwegian building register. For Oslo thislinkage means that information is now

available on building and pollutionconditions in a grid of 500 x 500 m2

covering the whole of Oslo. The calcula-tions can be further improved by imple-menting these tools in other towns aswell. The MOBAK study, giving infor-mation about the mean distribution ofvarious materials on various buildingcategories, has also improved the validityof the estimates. Finally, the economicmodelling apparatus used to estimateallocation costs also represents animprovement on the one previously inuse.

To sum up the results, maintenance costsascribable to material corrosion ofbuildings and motor vehicles that iscaused by air pollution have beencalculated at about 198 million 1995-Nkr.In addition allocation costs in theeconomy were estimated at 93 million1995-Nkr. We cannot assign the results aprobability distribution because the basicdata do not permit this. However, sincemany types of materials are not dealt within the report the estimates will in allprobability be on the low side in relationto the real costs inflicted by air pollutantson the building capital and the motorvehicle population of Norway.

The marginal-cost calculations reflect thatthe smaller the distance between anemission source and a building the greaterdamage the emission will inflict on thebuilding. Hence emissions in areas wherethere is a high concentration of buildingswill cause relatively heavy damage. Thismeans that emissions from sources such asvehicle traffic will generally cause greaterdamage than industrial emissions, asindustrial emissions in Norway are oftenlocated in smaller towns and villages. Thehighest marginal costs were found inBergen and Oslo, where the emission of


43

1 kg of SO2 inflicts damages valued atNkr 85 and 71, respectively (excludingallocation costs).

Of the pollution parameters included inthe calculation, SO2 is by far the mostdominant cost factor. Since only 15 percent of costs are related to materials thatare affected by O3 and H+, changes inthese parameters will only produce smallchanges in costs. On the other hand, wefound that the large reduction inNorwegian SO2 emissions in the period1985-1994 has resulted in an annualsaving of 431 million 1995-Nkr. Thus,reducing emissions of air pollutants areclearly not merely associated witheconomic costs.


45

4.1. IntroductionControlled experiments have foundsignificant evidence of a negative relation-ship between exposure to ground-levelozone and the yields of several types ofvegetation. On a European level consensusdose-response functions (or damagefunctions) have been formulated forwheat and potato (UN-ECE 1994), where-as several Norwegian studies have foundsignificant dose-response functions forvarious grass species used on cultivatedmeadowland (see e.g. Mortensen (1995)).

The harmful effect of a given ozone doseis influenced by various climate factorssuch as temperature, daylight conditionsand humidity as well as soil moisture.This, in addition to variation in ozonesensitivity between types within a speciesmakes it difficult to establish reliabledamage functions. As the number of

25 This study was commissioned by the State PollutionControl Authority (SFT) and carried out jointly by theNorwegian Institute for Air Research (NILU), theSærheim Research Station and Statistics Norway. Thearticle has earlier been published in Norwegian inTørseth et al. (1997).26 Statistics Norway.27 Særheim Research Station.28 Norwegian Institute for Air Research.

experiments are limited so far, theparameters in the dose-response functionsmay therefore be considerably alteredwhen more, and more sophisticated,studies are conducted. Despite the greatuncertainty the recommended damagefunctions will nevertheless give someindications about the damages.

In this study we calculate the cropdamages from ground-level ozone basedon current knowledge. Moreover, theeconomic losses related to these damagesare analysed. The calculations areconducted for the year 1992, when theozone concentrations in Norway wereparticularly high. Thus, the results areprobably not representative for an averageyear. On the other hand, potential dam-ages to several agricultural plants are notconsidered because reliable, controlledexperiments have not been conducted.Damages to tree species from ground-levelozone are neither included although somedose-response functions have beenestimated (see, however, Tørseth et al.(1997)).

The contributors to ground-level ozoneare emissions of NOx and VOC. However,the formation of ozone is a complex

4. Social costs of crop damagefrom ground-level ozone25

Anett C. Hansen26, Henning Høie26, Leiv Mortensen27,Knut Einar Rosendahl26 and Kjetil Tørseth28


46

process, and close to large emissionsources of NOx the concentration of ozonemay actually be reduced (see chapter 3 inthis book for a more thoroughly descrip-tion). In Norway ozone concentrations aremainly determined by foreign emissions,which are transported in the air to Nor-way. Thus, the crop damages calculated inthis study is to little extent affected bydomestic actions.

So far the experiments indicate thatAOT40, i.e. accumulated exposure toozone concentrations exceeding 40 ppb(parts per billion) or 80 µg/m3, may be agood indicator when it comes to cropdamage from ground-level ozone. AOT40is here measured during the growingseason, which may differ betweenlocations and type of vegetation. InTørseth et al. (1996) levels of AOT40 forthe period 1989 to 1993 have beencalculated on the basis of measurementdata recorded at Norwegian monitoringstations. These calculations were based onthe recommendations from a UN-ECEmeeting in Bern in 1993 (UN-ECE 1994),which inter alia recommended criticallevels of AOT40. The question of how therecommendations can be adapted toNordic conditions was considered byTørseth et al. Levels of AOT40 for varioustypes of vegetation in the Nordic regionwere mapped.

The results from Tørseth et al. (1996),which will be described below, are useddirectly in our calculations of crop dam-ages. An important finding is that thedecision of growing season, and par-ticularly the start of the latter, is highlysignificant when calculating AOT40 levelsand identifying areas where the criticallevels are exceeded. This is because ozoneconcentrations are highest in spring andearly summer. At a UN-ECE working

meeting in Kuopio, Finland in 1996 (UN-ECE 1996), the recommendations at UN-ECE (1994) were revised based on newexperience. These changes may substan-tially affect the calculations of crop dam-ages in Norway. However, the presentstudy does not take account of the newrecommendations, but points out the mostimportant changes.

Detailed information exists on areas,yields and position of various agriculturalplants and meadow in Norway. Togetherwith the calculated exposure doses andthe available dose-response functions, thishas provided a basis for rough estimatesof crop yield losses that may result fromexposure to ozone. Moreover, land areawhere critical ozone levels, as identifiedby UN-ECE (1994), are exceeded may becalculated, too. However, as indicatedabove, much uncertainty attaches to theresulting figures. Still, it is not possible toconstruct reasonable uncertainty intervals.

The social costs of a crop yield loss aredifficult to analyse since the farmingindustry is heavily regulated and largelysheltered from imports. The costs willdepend on, among other things, theauthorities' policy aims for this sector. Apolicy aim to maintain a certain pro-duction potential in the interest of foodsecurity has different implications than anaim to maintain a certain level of activity(e.g. in terms of employment) in supportof regional policy objectives. In the firstcase a crop yield loss and consequentproduction shortfall has to be compen-sated for by raising activity levels in thesector. The direct cost is then largelyreflected in production costs in agricul-ture, and the domestic producer price plusany subsidies will be a good indicator. Inthe second case a crop yield loss can becompensated for by increasing imports,


47

and prices on the world market will be agood indicator.

The rest of the national economy isaffected when crop yields per hectare arereduced as a result of ground-level ozone.Here too the effects are dependent onwhich scenario is most realistic. Inasmuchas far more resources are needed toachieve a particular production value inagriculture than in other industries,attracting resources to agriculture willentail substantial social costs over andabove the direct value of the loss of yield.

The uncertainty of the economic cal-culations compounds the uncertaintyinvolved in calculating the physical ozonedamage. The economic uncertainties arerelated both to choice of shadow price forthe individual type of crop (i.e. the directsocial cost per unit volume), and to theanalysis of the macroeconomic effects.The social costs must therefore be seen asrough estimates. However, the calcula-tions do provide (based on currentknowledge) some indications of whatground-level ozone can cost Norway inthe form of crop yield reductions in yearswith very high ozone exposures.

In the next chapter the physical damage ofozone exposure is discussed. Then, inchapter 4.3 the economic analysis ispresented.

4.2. Ozone exposure and cropdamage

4.2.1. Methods for estimating cropyield losses

Damage functions for various types ofvegetationOzone sensitivity varies widely amongdifferent types of vegetation. Of the

approximately 50 species examined inNorway (Mortensen 1992, Mortensen andNilsen 1992, Mortensen 1993, 1994a,1994b), birch, timothy and clover can becharacterised as highly ozone-sensitive.Dose-response curves for ozone effects onplant growth and yield are available foronly a fairly small number of species atpresent. An overview is shown in table4.1, where the fuctions are expressed inrelative yield. By relative yield (y) ismeant actual yield divided by yieldwithout ozone exposure, and converted topercentage points. Wheat is the mostthoroughly examined (UN-ECE 1994), andthe critical ozone level for wheat has beenrecommended for other agriculturalplants. The potato seems to have asensitivity matching that of wheat. Forcultivated meadowland where timothyand clover are important grass species,Norwegian tests have ascertained asomewhat lower sensitivity than for wheat(Mortensen 1995).

The damage functions are based on theassumption that there exists a linearrelation between ozone exposure andyield loss. The relations have been derivedempirically from a limited testing.However, at low exposures in controlledexperiments, relations between dose andresponse have been difficult to discern,and no significant correlation have beenestablished. For that reason UN-ECE(1994) recommended that the damagefunctions are only applied for AOT40values that according to the functionimply yield losses exceeding 10 per cent(see table 4.1). That is, 10 per cent ischosen as the smallest statistically reliablereduction, and the corresponding AOT40value is identified as the critical level. Thismeans that we actually use functions thatare discontinuous around this criticallevel. The damage functions in table 4.1


48

show that an AOT40 value of 5 300 ppb hfor wheat and potatoes, and 6 700 ppb hfor meadow causes a 10 per cent loss ofyield, and these values are accordinglycritical levels.

The timing of the start of the growthperiod for a given type of plant is cruciallysignificant for whether the ozone dose willexceed the critical level. Since April, Mayand June are generally the months withthe highest AOT40 values (normallyaccounting for more than 50 per cent ofoverall exposure in the growing season),an early growth start will result in higherexposure. The growing season starts laterin Scandinavia than in the rest of Europe.Thus, the recommended choices of seasonfrom UN-ECE (1994) are ill suited in ourcase. Using growth periods more relevantto Nordic conditions reduces the exposuredoses considerably. However, even whenthis is done, exposure doses may exceedthe recommended critical levels forvegetation in southern Norway. Grassproduction in low-lying areas of southernNorway is the most heavily exposed(growth start 15-25 April). Cereals(germination 10-20 May) are less exposedto ozone, whereas potatoes, which can beextremely sensitive (wide differencesbetween different types), do not startgrowing properly until about 1 June andare therefore exposed to a relatively lowAOT40 dose through the growing season.

Vegetation in the mountains of southernNorway will also in general be protectedagainst high ozone exposure by the lategrowth start (June).

Accumulated exposure doses for ozoneare determined as follows. For agriculturalcrops the growing season is defined as theperiod from germination of wheat to itsmaturation. Normal values for two weeksafter sowing to two weeks beforeharvesting in the various regions areemployed as a Nordic adjustment of thegrowing season. Only daylight hours areincluded in calculations of AOT40. Forcultivated meadow we have, as a Nordicapproximation, defined the growingseason as those days whose normal meantemperature exceeds 5C. Only daylighthours are included in calculations ofAOT40. The growing seasons employedare shown in table 4.2.

Measurements over a wheat field insouthern Sweden have shown that ozoneuptake leads to a considerable reductionin ozone concentrations closest to groundlevel (Pleijel and Grennfelt 1995). Hencecalculating the exposure dose on the basisof measurements taken 2-3 metres aboveground level could result in an over-estimation of 20 per cent in relation to thereference height used in dose-responseexperiments, which is 1.1 metres.

Table 4.1. Overview of damage functions for various types of vegetation, where y is relative yield and x is ozone exposure (AOT40 measured in ppm h). (1 ppm h = 1000 ppb h)

Type of vegetation Damage function Area of application(critical level)

Wheat y = 99.6 - 1.804x(UN-ECE, 1994)

x>5.3(5 300 ppb h)

Potatoes y = 99.6 - 1.804x(UN-ECE, 1994)

x>5.3(5 300 ppb h)

Meadow y = 100 - 1.5x(Mortensen, 1995)

x>6.7(6 700 ppb h)


49

A recommendation from the UN-ECEworking meeting in 1996 (UN-ECE 1996)was that present critical level criteria arenot suited to estimating regional cropyield losses. This is primarily due touncertainties related to the degree towhich ozone is absorbed by vegetationand to other factors which may affectplants' sensitivity to ozone. There areindications that interacting effects mayentail increased sensitivity to ozoneexposure. On the other hand drought-induced stress could cause plants' leafpores to become smaller, thereby reducingozone uptake. Knowledge of theseprocesses is limited, but could beincorporated in the guidelines when moreinformation becomes available. Extendingthe critical level criterion to embrace suchfactors is referred to as level 2, whilesurveys in accordance with UN-ECE(1994) and the new recommendationsfrom UN-ECE (1996) are designated level1. The UN-ECE meeting in 1996 alsoagreed upon that the critical level forwheat should be reduced, so that 5 percent yield loss becomes the smallestsignificant reduction. This would meanthat the critical level of AOT40 werehalved. This change would significantlyincrease the exceeded area, and thus thecalculated crop damage. However, takingthe other factors above into account, too,it is unclear whether the estimated cropdamages are underestimated or not.

Use of measurements for calculating ozoneexposureGround-level ozone is measured at about15 monitoring sites in Norway withfinance provided by the State PollutionControl Authority through a separatemonitoring programme under the StateProgramme for Pollution Monitoring (SFT1994). In addition 19 monitoring sitesexist in the rest of Scandinavia. The

locations of the sites are shown in figure4.1, and the sites' coordinates, altitudeand appropriate growing seasons areshown in table 4.2.

Calculating accumulated ozone-exposuresover the whole country requires much interms of the monitoring sites' represen-tativeness, the monitoring network'sdensity and data completeness. Therepresentativeness of the individual sitesvaries somewhat and has not beenquantified, above all in terms of AOT40values. First, sites affected by local NOemissions will not be representative forlarger regions. The monitoring sites atNordmoen (outside Oslo), and to a lesserdegree at Søgne (nearby Kristiansand)and Svanvik (close to the Russian border),are affected by local NO emissions, whichentails lower observed ozone concen-trations.

Second, as ozone is taken up by plantsand is reacting with a large number ofsurfaces, ozone concentration at groundlevel is often considerably lower than atan altitude of 50 - 100 metres. Because ofthis, high valley sides and ridges are morevulnerable to high AOT40 values than arelarger continuous forest areas and low-lying areas. Thus, regional exposure dosesfor various types of land cannot bedetermined precisely based on currentozone monitoring. The measurements dohowever give an indication of regionaldistribution and typical exposure dosesand show geographical areas wherevarious critical levels may be exceeded.

Measurements for the year 1992 from the34 monitoring sites in Scandinavia areused (Lövblad et al. 1996). Most of theseare EMEP (European Monitoring andEvaluation Programme) sites, but othermeasurements are also employed. Because


50

a number of sites provided low coverageor were not representative, the number ofsites was reduced to 24. Exposure dosesfor varying length of growing season arecalculated from the measurement results.Inclusion of the remaining Scandinaviansites substantially improves the databasefor regional surveys, particularly along theborders and in the north of the country.

Accumulated exposure doses (AOT40) arecalculated as the sum of the differencesbetween hourly mean concentrations and40 ppb for each hour where ozone

concentration exceeds 40 ppb. Exposuredoses calculated for the various planttypes and growth periods are shown intable 4.1. For agricultural crops onlyhours of global light influx in excess of50 W/m2 are included in the calculations.Global light influx is calculated as afunction of the sun's meridian and noaccount is taken of cloud cover.

Since the AOT40 concept states a sum ofmeasurement values in excess of 40 ppb,any absence of data will lead to greatuncertainties in the calculated result. Thisis especially critical if no measurementsare taken during episodes of high concen-trations. Several considerations can beapplied to reduce the uncertainties. Datacoverage should be as good as possible forall parts of the growing season. Sites withpoor data coverage should therefore beomitted from the calculations. However,setting an excessively high criterion fordata coverage will reduce the number ofsites, which creates problems for theoverall survey. In 1992 data coverageexceeded 90 per cent at most monitoringsites. Sites offering data coverage between80 and 90 per cent are, in the event,included after assessment and comparisonwith sites in the vicinity. Sites with lessthan 80 per cent data coverage are notincluded. To correct for absence of data, itis assumed that the absent data show thesame distribution as the observed dataand correction is obtained by dividing thecalculated exposure dose (AOT40) byrelative data coverage.

Calculated exposure doses at monitoringsites are interpolated to a grid by kriginginterpolation, which is a statistical methodfor estimating unknown values fromnearby observations. The method wasoriginally developed for geostatic pur-poses (Matheron 1963, Journel and

Figure 4.1. Location of Scandinavian moni-toring sites used to survey ozone exposure

Svanvik

Jergul

Pallas

Osen

Kårvatn

Vindeln

Høylandet

Tustervatn

Ammarnäs

Esrange

Oulanka

Raja-Jooseppi

Langesund

Jeløya

Haukenes

BirkenesSøgne

Valle

Voss

Norra Kvill

Velen

Rörvik

Ulborg

Nordmoen

Utö

Aspvreten

Fredriksborg

Lille Valby

Vavihill

Sännen

Virolahti

Ähtari

Tvärminne

Prestebakke


51

Huijbregts 1981), but has in recent yearsalso been used to describe regionaldistribution of, for instance, trans-boundary air pollution (Schaug et al.1993). Kriging interpolation is based onthe assumption that the variance betweenthe observed values is a function ofdistance and direction, and weighting isdetermined by variogram analysis.Exposure doses of ground-level ozonebased on the AOT40 principle are difficultto interpolate owing to wide variance inthe calculated values. Moreover, becausethe number of sites is on the small side,the grid values calculated are roughapproximations and relatively uncertain. Agrid comprising 50 km squares is usedbased on division of the ordinary grid useby the EMEP (150x150 km2).

A comparison between exposure dosescalculated both on the basis of monitoringdata and with the aid of EMEP's oxidantmodel (Lövblad et al. 1996) showsrelatively good correspondence insouthern Norway, southern Sweden andin southern Finland. In Denmark themodel overestimated exposure by about50 per cent compared with the exposurebased on monitoring data. In the morenortherly parts of Scandinavia the modelshowed substantially lower exposure thanthat based on measurements.

Areas and yields of ozone-sensitivevegetationIn order to link ozone exposure to dam-ages on various types of vegetation, areasand yields of the following types of landwere distributed on the EMEP grid(50x50km2): spring-sown wheat, autumn-sown wheat, cultivated meadow,extensively cultivated meadow (i.e.,surface-cultivated and fertilised pasture),and potatoes. Data on agricultural areasare taken from information provided by

farm operators as a basis for annualgovernmental production subsidies (as at31 July 1994). For the country as a wholeapplications for production subsidiescovered an estimated 94.3 per cent ofgrain-growing land, 97.4 per cent ofmeadow and 98.7 per cent of land usedfor potato production. The data for farmareas are basically distributed by munici-pality. Yields are Normal Year Yields 1994calculated by the Norwegian AgriculturalEconomics Research Institute. No distinc-tion is drawn between yields of spring-and autumn-sown wheat. Information onyields on surface-cultivated and fertilisedpasture is unavailable, and is assumed tobe two-thirds of the yield from cultivatedmeadow.

Normal annual yields are calculated atcounty level and all municipalities in thecounty are each assigned an identicalyield per hectare. Moreover, for thepurpose of distribution on the EMEP grid,the municipalities' arable areas as well asproduction are distributed in proportionto the distribution of the entire area of themunicipality. In reality variations willoccur both among municipalities within acounty, and within the municipality.These sources of error contributes some-what to overstated agricultural productionfor higher-lying, and generally the north-erly, EMEP grid squares, and correspond-ingly understated production levels forlower-lying and southerly squares.

Wheat is more dependent on warmtemperatures than any other cereal plantgrown in Norway, and all wheat pro-duction is concentrated in the areasaround the Oslo Fjord. About 70 per centof Norwegian wheat is grown in thecounties of Akershus, Østfold and Vest-fold. Potato cultivation is also concen-trated in south-east Norway, with about


52

45 per cent of the country's total pro-duction taking place in the neighbouringcounties Hedmark and Oppland. Grassproduction is associated with husbandryof livestock feeding on rough grazingssuch as cattle, sheep and goats. Theseproductions predominate in all farmingareas apart from in the grain-growingareas around the Oslo Fjord.

4.2.2. Results and discussionThe study by Tørseth et al. (1996) for theyears 1989-1993 found that in most yearsexposures range from 5 000 to 10 000 ppbh in southern Norway and from 2 000 to5 000 ppb h in northern Norway in theperiod April - September. In the growingseason the exposure doses are around thecritical levels for cultivated meadow andgrain. The highest exposure doses in thefive years period were with a few excep-tions observed in 1992, which is the yearwe focus on here. The high exposures inthat year were due to meteorological

conditions which resulted in effectivetransport of ozone and precursors fromthe Continent to Scandinavia during thespring and early summer. However, at themonitoring site Tustervatn in the northernpart of Norway, AOT40 values wereactually lowest in 1992. For a completediscussion of the results of the survey, seeTørseth et al. (1996).

Table 4.2 shows growing seasons, datacoverage and calculated exposure dosesfor ozone in 1992 for agricultural plantsand cultivated meadow at all monitoringsites. The determination of growingseason was explained above, and thisvaries between locations and plants. Forcomparison, calculated AOT40 levelsusing the growing season generallyrecommended by UN-ECE (1994) foragricultural plants in Europe, are alsoshown in the table. However, these resultsare not used in the further calculations asthis choice of growing season is thought to

Table 4.2. Growing seasons, data coverage and calculated exposure doses for ozone for agriculturalplants and cultivated meadow in 1992

Agricultural plants(Nordic adjustment)

Cultivated meadow(Nordic adjustment)

Agricult. plants(UN-ECE)*

Monitoringsite

Alti-tude

Growingseason

Datacover-age (%)

AOT40(ppb h)

Growingseason

Datacover-age (%)

AOT40(ppb h)

Datacover-age (%)

AOT40(ppb h)

Birkenes 190 15/5-15/8 91 10 102 15/4-1/11 93 11 990 92 11 052Haukenes 20 15/5-15/8 – – 15/4-1/11 – – – –Høylandet 60 1/6-1/9 – – 1/5-1/10 – – – –Jeløya 3 15/5-15/8 82 6 567 15/4-1/11 87 7 215 96 5 991Jergul 255 1/6-1/9 73 – 1/6-15/9 74 – 99 2 643Klyve 60 15/5-15/8 94 5 872 15/4-1/11 97 7 606 95 6 620Kårvatn 210 1/6-1/9 99 1 968 1/5-15/10 99 8 830 98 8 863Langesund 5 15/5-15/8 96 5 080 15/4-1/11 95 6 041 96 5 352Nordmoen 200 15/5-15/8 – – 15/4-1/11 69 – – –Osen 440 15/5-15/8 100 4 721 15/4-15/10 95 5 867 96 5 278Prestebakke 160 15/5-15/8 – – 15/4-1/11 – – – –Svanvik 30 1/6-1/9 – – 1/6-1/9 – – – –Søgne 15 15/5-15/8 – – 15/4-1/11 – – – –Tustervatn 439 1/6-1/9 100 387 15/5-15/9 100 1 101 100 1 225Valle 250 15/5-15/8 91 6 127 1/5-15/10 – – 91 6 423Voss 500 15/5-15/8 100 5 490 1/5-15/10 100 5 663 100 5 630

* Growing season: 1 May - 1 August


53

be inadequate for Nordic conditions (seeabove). Figure 4.2 and figure 4.3 showregional patterns in the distribution ofAOT40 for growing seasons for wheat andmeadow respectively.

Table 4.2 and figure 4.2 show that foragricultural plants the highest doses wereobserved in southern Scandinavia. AtBirkenes at the south coast of Norway theAOT40 value was in excess of 10 000 ppbh in 1992, which according to the damagefunction in table 4.1 should indicate ayield reduction of almost 20 per cent. The

critical level of 5 300 ppb h is exceededsouth of 60°N (around Bergen). As shownin table 4.2, using the growing seasonrecommended by UN-ECE (1994)increases the AOT40 levels significantly,and the critical level is exceeded in thewhole of southern Norway (i.e. south ofTrondheim). For cultivated meadow theresults indicate that the critical level isexceeded in most of southern Norway.The highest AOT40 value was againmeasured at Birkenes, at about 12 000ppb h.

Figure 4.2. Distribution of AOT40 (ppb h) for growing seasons for wheat (Nordic adjust-ment), 1992

Figure 4.3. Distribution of AOT40 (ppb h) for growing seasons for cultivated meadow (Nordic adjustment), 1992


54

Table 4.3 shows the areas of exceedancefor the different types of vegetation in1992. The areas are calculated by assum-ing that the entire area of a vegetationtype in each EMEP square showedexceedance if the AOT40 value in thesquare was higher than the critical level.From the table we see that the criticallevels were exceeded for a large share ofagricultural land in 1992, particularly forwheat where the area of exceedanceconstituted 88 per cent of all wheat fields.If we had rather used the growing seasonrecommended by UN-ECE (1994), thearea of exceedance would have increasedto a share of more than 99 per cent. Interms of absolute numbers the biggestexceedance was in the case of meadow, asmeadow clearly is the most widespreadtype of land. For both meadow andpotatoes the area of exceedance equalledabout 59 per cent in 1992.

Based on calculated AOT40 values and thedamage fuctions set out in table 4.1, yieldlosses may be determined. The calcula-tions were based on the assumption thatnormal annual yield is affected by ozoneexposure. Hence, we get the followingexpression for yield loss:

yield loss = (normal annual yield/relativeyield) - normal annual yield

The biggest yield loss was found in grassproduction where the weight loss is aboutone order of magnitude higher than forwheat and potatoes respectively. How-ever, this is due to the large production ofgrass compared to other plants. In termsof relative numbers the yield losses con-stituted around 10 per cent for all threecrops. The UN-ECE (1994) recommenda-tions would have increased the calculatedyield loss in wheat production to 14 percent.

4.3. Economic analyses of cropdamage

We now go on to analyse the social costsassociated with the yield losses resultingfrom ground-level ozone in Norway (seetable 4.3). Reduced yield entails economicloss since a given set of resourcesproduces less return than it would havedone in the absence of ozone exposure.The pollution reduces the productivity ofthe resources, i.e. the relation betweenproduction level and quantity of inputs.

In this presentation we start out from apotential situation of no ozone exposure,and discuss the effects resulting from anoccurrence of crop damage. Although theyield reductions are calculated for a yearof particularly high ozone exposures(1992), we regard the problem of ground-level ozone as a long-term one. Crop dam-age resulting from excessive ozone

Table 4.3. Areas of exceedance and potential yield loss for various vegetation types in Norway, 1992

Vegetation type Area of exceedance Yield loss

km2 per cent 1 000 tonnes per cent

Wheat* 597 88 30.7 11Potatoes 97 59 32.4 7Total meadow** 3 329 59 332 8

* includes spring-sown and autumn-sown wheat** includes cultivated meadow, surface-cultivated meadow and fertilised pasture


55

concentrations is not considered primarilyas a short-lived shock to which agricultureis vulnerable, but as a lasting problemwhich the farmer takes into account in hisadjustment on a par with climaticconditions. The economic analyses aretherefore rooted in long-term perspective.However, since ozone exposure variessomewhat from year to year, and wasparticularly high in 1992, we also discusshow a short-term perspective affects thecosts incurred.

Calculating real social costs of reducedproductivity in agriculture is particularlydifficult inasmuch as this sector is highlyregulated and largely sheltered fromimports. Hence it is important to ascertaina relevant price for various crop types.This price, often termed the shadow price,depends inter alia on the policy aims setby the authorities for the agriculturalsector. This and other methodologicalquestions are discussed in detail in thefollowing subsection. Then, section 4.3.2.presents results for the direct costs ofexposure to ground-level ozone, i.e. theshadow price multiplied by yield loss. Insection 4.3.3 it is shown how the yield losscan affect the rest of the economy, eitherthrough reduced supply of domesticallyproduced foodstuffs, or because theagricultural sector attracts more inputs inorder to maintain production levels. Theseeffects are analysed using a macro-economic model.

4.3.1. Methods for analysing cropdamage

Aims and guidelines for Norwegianagricultural policyThere are several reasons why the agri-culture sector in Norway is regulated andlargely sheltered against imports. Pro-position no. 8 to the Storting from the

Ministry of Agriculture (1992) sets outguidelines for agriculture policy, andessentially expresses the authorities'current policy aims for agriculture.29

Several factors are given weight, and inthe context of the present study we focuson two of them. The first is long-termfood security. The proposition states that“importance should be given to main-taining production potential on thelimited geographical resources available”,whereas a continual high level of self-sufficiency is regarded as less important(chap. 3.1.3). A palpable indication of thisis the authorities' switch from production-dependent to area-dependent plantproduction subsidies. This switch ofemphasis entails that lasting changes inextraneous factors affecting productivityin the sector may lead to policy shifts topreserve agricultural land. Such a factorcould be changes in ozone exposure.

The second factor we focus on, and whichis given weight by the authorities, is theagriculture sector's great significance forsafeguarding population settlement andemployment in outlying areas.30 In thiscontext activity levels in agriculture, in thefirst instance the number of personsemployed in the sector, are important.However, the Ministry of Agricultureemphasises (1992) the need for a regionalpolicy strategy focusing on a broadereconomic base, and the need to baseactivities in agriculture to a greater degreeon market potential that offers a broaderand more varied basis for employment.

29 Although the Storting's treatment of the propositionresulted in minor changes in the adopted recommen-dation, wide disagreement exists between someparties in the Storting.30 The agricultural sector and its employment effectson other economic activity account for more than onehalf of jobs in about one fourth of the country'smunicipalities (Ministry of Agriculture 1992).


56

This may entail that reduced productivityas a result of extraneous conditions insome areas may contribute to a reductionin activity levels in agriculture at theexpense of other economic activity.Putting the emphasis on regional policyconditions may thereby give rise to otherand in some cases contrary conclusions tothose resulting from emphasis on foodsecurity.

While production potential and activitylevels are both presented as important, itis conceivable that one of these policyaims may already be amply fulfilled in thecurrent situation. However, there is nobasis for saying which of the two thismight be. With this in mind, we examinetwo different scenarios.

In scenario 1, called “Maintenance ofproduction potential”, we assume that theauthorities respond to reduced productivityin agriculture by adjusting agricultural landuse to ensure that production potentialremains unchanged. We also assume thatthe potential is exploited to the samedegree as prior to the productivity fall, sothat actual production also remainsunchanged. This requires an increase inresource use, or activity levels, in the sectorto offset reduced productivity.

In scenario 2, called “Maintenance ofactivity levels”, we assume that theauthorities opt for unchanged activitylevels in agriculture if productivity isreduced. By “activity” we mean the overalluse of inputs, including employment(number of person-hours worked) andreal capital.31 Hence in this scenario agri-cultural production is reduced in propor-

31 In reality, the emphasis should be on employment.However, as the factor shares are fairly constant, wechoose to focus on overall use of inputs for simplicity.

tion to crop damage. It was suggestedabove that activity levels in agriculturecould fall as a result of lower productivity,at the same time as other economicactivity levels were stimulated. Since thescope of this mechanism is highly uncer-tain, and compensatory activity wouldprobably require some governmentsupport in the same way as agriculture,we choose to disregard this factor.

The authorities seek to achieve the policyaims set for agriculture via various types ofsubsidies, and through prices set in theMain Agreement for Agriculture combinedwith high tariff rates. In plant productionthe subsidies are largely per unit of land,and not per unit of production as previous-ly (Agricultural Budget Commission 1996).This is related to the goal of long-term foodsecurity and environment-friendly produc-tion. In the case of livestock productsproduction subsidies persist, with anindirect bearing on grass production. Price-fixing through the Main Agreement forAgriculture also entails indirect support tothe agriculture sector, inasmuch as prices inmany cases are far higher than worldmarket prices. This arrangement works inpractice as production support. Part of thedifference between the producer price andthe world market price for grain is coveredby the authorities in writing down the priceof Norwegian grain supplied by the Nor-wegian National Grain Administration(Ministry of Agriculture 1994). The bulk ofthe difference however is borne by theconsumer. The various support schemes inagriculture make it difficult to uncover theactual social cost of crop yield lossesbecause the market price does not ingeneral reflect the real shadow price of thecrop. Hence we are compelled to take abasis in the policy goals for agriculturementioned above in order to analyse thesocial cost of crop damage.


57

Valuing wheat and potatoesThe domestic price of wheat is generallyfar higher than the world market price.Whereas the Norwegian producer pricehas hovered around Nkr 2.50 per kg in thepast five years (tending to edge down)(Agricultural Budget Commission 1995;1996), the import price has averagedabout Nkr 1 per kg. (Statistics Norway1996a).32 Although the price differencesmay conceal quality differences anddifferent purchase times, this probablyplays a small role for wheat. Hence theprice difference, which is supported byhigh tariff rates and state price support isonly assumed to reflect a value intrinsic toNorwegian production. The domesticprice is fixed in the Main Agreement forAgriculture, and wheat production isconsequently determined by the individualfarm operator, based on profitabilityconsiderations (given other supportschemes such as agricultural land areasubsidies).33 Imports of wheat and othertypes of grain cover just over half the totalgrain consumption in Norway (StatisticsNorway 1994b).

Hence the shadow price of wheat is verydependent on the policy goal set forwheat production. In scenario 1, whereproduction levels are in focus, the dome-stic price best reflects the direct cost ofcrop yield reductions, inasmuch as theauthorities in this case wish to increaseinputs in agriculture in response to cropdamage. The shadow price for 1992 wasNkr 2.67 per kg (see table 4.4), and is

32 The past year (1996), however, has seen asubstantial increase in the world market price ofwheat.33 In order to avoid overproduction, producers arecharged if production exceeds a maximum level. Theprices quoted are therefore the prices paid, and notpre-agreed prices (Agricultural Budget Commission1996).

taken from the Agricultural BudgetCommission (1995). Since other supportschemes, first and foremost agriculturalland area subsidies, contribute to profitableoperation of the individual holding, the realsocial cost is higher than the stated price.Land area support varies between regionsand the size of the land area in question,and the rates for 1996/97 are estimated atNkr 1 620-3 500 per hectare of grain(Norwegian National Grain Administration1996). According to Statistics Norway(1996b), average yield of wheat per hec-tare in 1995 was 4 990 kg. In other wordsthe implicit support per quantity of yieldaverages between Nkr 0.32 and 0.70 perkg. Inasmuch as the support is generallyhighest for farmland producing a smallyield per hectare, support per kg of wheatvaries widely. Compared with the producerprice for 1992 referred to above, the landarea subsidy increases the farmer's incomefrom wheat production by an average ofabout one-fifth. This is captured by themacroeconomic analyses.

In scenario 2, where activity in the sector isin focus, production levels play a smallerrole. In this case reduced production will becounterbalanced by increased imports ofwheat, so that the world market price is therelevant shadow price. This applies even ifthe consumer has to pay a higherdomestic price due to the customs tariff.According to external trade statistics(Statistics Norway 1996a), the averageprice for imported wheat (excl. tariff) wasNkr 0.83 per kg in 1992, i.e. less than one-third of the domestic price. In the yearssince, some increase in import prices andfalling domestic prices have substantiallynarrowed the price difference.34

34 In 1995 import prices were about half the domesticprice (Agricultural Budget Commission 1996;Statistics Norway 1996a).


58

The producer price on potatoes is notregulated to the same degree as for typesof grain. The price varies both over timeand among regions, and with respect toquality. The average producer price can becalculated by dividing total sales incomeby total sales production (AgriculturalBudget Commission 1995), giving Nkr1.39 per kg of potatoes in 1992, which isemployed as the shadow price in scenario1 (i.e. unchanged production). The landarea subsidy for potatoes is estimated atNkr 1 860-1 970 per hectare for 1996/97in southern Norway (Norwegian NationalGrain Administration 1996). Average yieldper hectare in 1995 was 22 240 kg (Statis-tics Norway 1996b), such that implicitsupport per quantity of crop is Nkr 0.08-0.09 per kg. Thus for farmers in southernNorway the land area subsidy constitutesa far smaller component of income thanthe potato price.

Moreover, the domestic potato price is notvery different from the world market price.According to Erdal and Nersten (1996) theprice (cif) of Dutch and Danish potatoesaveraged, respectively, 23 and 11 per centless than the Norwegian producer price inthe period 1990-1995. The price differ-ences vary however from month to month,and in some cases the Norwegian price islowest. Concurrently external trade stati-stics (Statistics Norway 1996a) show thatthe average price for imported potatoes(excl. tariff) was Nkr 1.74 per kg in 1992,i.e. 25 per cent higher than the domesticprice. However, this is because imports,

which in 1992 amounted to 10 per cent ofNorwegian production (Statistics Norway(1996a); Agricultural Budget Commission(1995)), largely take place outside the highseason for harvesting Norwegian potatoes.Thus, imports take place when prices arehigh both in Norway and abroad.35 In 199560 per cent of imports were from Cyprusand Canada (Statistics Norway 1996c). It istherefore wrong to conclude that Nkr 1.74per kg is the relevant shadow price forpotatoes in scenario 2 (i.e. unchangedactivity level). In the event of a shortfall inNorwegian production the import price atthe time of the harvest must be used. Sinceno such explicit price exists, we use the sameprice as in scenario 1, i.e. the domesticprice. This is probably on the high side.

Valuing grassGrass is overall the most valuable plantproduct in Norway. It differs from wheatand potato in that it is not a final product,but is used as an input in livestockproduction. Furthermore, the farmer isusually self-sufficient in grass. This isbecause grass (both processed and unpro-cessed) has a high volume-to-weight ratiomaking it expensive to transport andstore. It is therefore difficult to find arelevant market price for grass. Sales areminimal, and this does not reflect thevalue of grass. Since feed requirementsare met by an appropriate mix of roughfodder (i.e. grass) and concentrates,unexpected production changes can be

35 This is because potatoes are less durable than grain.

Table 4.4. Shadow price for various types of crop in 1992 in various scenarios. 1992-Nkr per kg

Scenario 1Maintenance of production potential

Scenario 2Maintenance of activity level

Wheat 2.67 0.83Potato 1.39 1.39Grass 1.19 0.37


59

remedied by regulating the use of concen-trates. Concentrates are produced from,among other items, barley and oats whichare to some extent imported.

Livestock production is regulated to aneven greater degree than plant production,and an explicit goal for Norway is self-sufficiency in livestock products. Moreover,for the farmer the returns on livestockhusbandry are generally higher than onplant production. Hence a reduction ingrass production as a result of ground-levelozone will not affect the volume of live-stock production since the farmer has tocover the yield loss so as to meet his fodderrequirements as previously. If a yield loss isanticipated, he can do this in several ways.In the first place the farmer can use moreconcentrates than previously, as is custo-mary in the case of sudden yield losses.However, excessive use of concentrates isinadvisable, making this alternativeproblematic. The other possibility open tothe farmer is to increase his pasture. Thiscan be done either at the expense of otherplant production, or by increasing totalland use. In the first case it is natural toassume that land used for growing barleywould be displaced. Alongside grass, barleyis the most important plant product inNorway (in value terms)36 and, as men-tioned, is used in the production of con-centrates. It is both propitious andrelatively common to grow barleytogether with grass for rotation purposessince barley gives good protection tonewly germinated grass-seed.

Against this background we calculate theshadow price in two different ways. In the

36 Incomes from barley production accounted for 45-52 per cent of total incomes from grain production inthe years 1992-95 (Agricultural Budget Commission1995, 1996).

first calculation we assume that the costsper hectare of producing barley and grassare virtually identical. This is because bothtypes of plant are grown over much of thecountry, and because they are oftengrown in conjunction. Hence the wideregional cost differences do not give riseto large distortions. The second calcula-tion assumes that the value per unit offodder is identical for barley and grass,since both are mainly used to feed live-stock (barley, however, has a lowerprotein content than grass and must firstbe processed into concentrates). In eachcase the shadow price of barley is used toestimate the shadow price of grass. Theprice of barley will however depend onwhich scenario we are studying.

Statistics Norway (1996b) provided dataon yield per hectare of barley and culti-vated meadow for the years 1988-1995.The ratio varies somewhat from year toyear, but the average scaling factor in theperiod is 0.46. In other words the cost perkg of grass using the first method ofcalculation is assumed to be 0.46 timesthe cost per kg of barley. A unit of fodderfor milk production is equivalent to 1.0 kgof barley and 1.9 kg of grass (StatisticsNorway 1994b). Hence the value per kg ofgrass in the second calculation is assumedto be 0.53 times the value per kg ofbarley. We see that the two approachesgive roughly the same ratio of cultivatedgrass to barley, and choose to employ theaverage of the two (= 0.495).

In scenario 1 the authorities wish to main-tain domestic agricultural production. Thisentails increasing the total resource inputin agriculture, including land use, as aresult of crops being exposed to ground-level ozone. As in the case of wheat, thedomestic producer price of barley isemployed. According to Agricultural


60

Budget Commission (1995), the averageprice paid for barley in 1992 was Nkr 2.40pr kg. This gives a shadow price for grassof Nkr 1.19 per kg. Here too, however, thesocial cost will be higher inasmuch as theland-area subsidy contributes to profitableproduction of both barley and grass forthe individual farmer. According to theNorwegian National Grain Administration(1996), land-area support for roughfodder is estimated to range from Nkr 1270 to Nkr 5 890 per hectare in 1996/97,whereas support for grain is put atNkr 1 620 to Nkr 3 500 per hectare.Inasmuch as the yield per hectare isgenerally smaller than for wheat, at thesame time as the price is lower, land-areasubsidies constitute a larger share of thefarmer's income in the case of barleyproduction than wheat production. Withour procedure this means that the realshadow price in table 4.4 is relativelyspeaking understated to an even greaterextent for grass than for wheat.

In comparison, the average price of con-centrates in 1992 was Nkr 3.36 per kg,according to the Agricultural BudgetCommission (1995). If the unit value ofrough fodder and concentrates is iden-tical, this corresponds to a price of Nkr1.89 per kg of grass, i.e. 60 per centhigher than the shadow price used by us.This suggests that it is more costly for thefarmer to increase the share of concen-trates than to cultivate more grassland atthe expense of barley, which supports ourassumption. However, in the event of asudden occurrence of crop damage, thelatter measure is infeasible.

In scenario 2 activity levels in agricultureare maintained at the same level as in theabsence of crop damage, so that mead-owland is increased at the expense ofbarley-growing land (see over). Reduced

production of barley must then becompensated for by increased imports ofbarley (or other plants) for production ofconcentrates to maintain livestock pro-duction. Hence in this case the shadowprice of barley is the import price. FromStatistics Norway (1996a) we find thatregistered imports of barley in 1992 (and1991) were so low that average importprices are uninformative. We thereforeemploy instead the average for corre-sponding prices in 1990 and 1993, whichwere respectively Nkr 0.80 and 0.70 perkg. The shadow price for grass in thisscenario is then 0.37 per kg, i.e. aboutone-third of the price in scenario 1.

Macroeconomic analysesWhen agricultural production is affectedby ground-level ozone, the rest of theNorwegian economy is also affected. Acorrect estimate of the social costs of cropyield loss is only obtained when theseeffects are included. The effects are highlydependent on which scenario is in focus. Ifthe authorities wish to maintain agri-cultural production at the same level as inthe absence of ozone exposure (scenario1), the agricultural sector must besupplied with labour and other resourcesfrom the rest of the economy. As a resultproduction in other sectors will bereduced. Since substantially moreresources are required to achieve a givenproduction value in agriculture than inother sectors, this resource allocation willlead to an extra social cost over and abovethe direct value of the yield loss. Since theauthorities wish to maintain the agricul-tural sector despite low productivity, theremust be an intrinsic value, above thatwhich shows in the accounts, linked eitherto production (scenario 1) or activitylevels (scenario 2) in agriculture. Changedprices and income conditions will alsoaffect the economy.


61

If the authorities choose to let activitylevels (resource use) in agriculture remainunchanged (scenario 2), production in thesector will be reduced in the event ofozone exposure. Hence in this case therewill be no transfer of resources from othersectors. However, since incomes inagriculture are reduced (because supportremains unchanged) and the supply ofdomestic agricultural commoditiesdiminishes, the rest of the economy isnegatively affected in this scenario too.

In order to analyse these macroeconomiceffects we apply the general equilibriummodel MSG-EE (Alfsen et al. 1996), whichis outlined in chapter 2 of this book. Inthis macro model resource allocationbetween various sectors may be studied.This is particularly important in ouranalysis because of the productivitydifferences mentioned above. However,modelling the agricultural sector withinthe macro model is straightforward. Thesector by and large produces one type ofcommodity, called agricultural products,which can be perceived as a kind ofaverage commodity in the sector. Thiscommodity, together with imports ofagricultural products, is demanded forvarious applications in the model. A largeportion of agricultural products arefurther processed in a separate sectordevoted to processing agricultural and fishproducts which also by and large producesone type of commodity. Another portionof agricultural products is used as inputsin the agricultural sector itself. Thisapplies particularly to feed for livestockproduction. While some agriculturalproducts go directly to food consumption,a larger share of consumption is via thesector that processes the products. Asmaller share of agricultural products isexported. Production value in theagricultural sector is determined outside

the model, since it is largely regulated bythe authorities and is independent ofeconomic factors. However, input use inthe model changes somewhat withvariations in factor prices. Imports ofagricultural products are also endogenous,and depend on demand for agriculturalproducts and on domestic production.

Because of the model's structure it isdifficult to distinguish the individual cropyield losses from other types of productionloss in the agricultural sector. Since bothproduction structure and application ofthe products can vary somewhat, par-ticularly where plant contra livestockproduction is concerned, the analysis maygive a distorted picture of changes at thesector level. The model's most importantcontribution is however linked to the maineconomic aggregates such as GDP andprivate consumption. Here the modelsupplies valuable information.

Crop damage is modelled in such a waythat productivity in agriculture is reduced;in other words the ratio between pro-duction value (at constant prices) andlevel of inputs diminishes. In scenario 1the production value in agriculture is fixedsuch that it is at the same level both withand without crop damage. This entailsthat at given factor prices the use of allinputs rises proportionately if damageoccurs. In scenario 2 the production valuein agriculture is determined in such a waythat the difference in level between asituation with crop damage and onewithout crop damage corresponds exactlyto the value of the damage. This opens theway for higher imports in the case ofozone damage. Inasmuch as grass is nottraded in the market, and hence does notshow directly in the accounts, it may atfirst glance seem unreasonable to changethe production value due to damage to


62

meadowland. However, we have assumed(see section 4.3.1) that the farmer desiresan appropriate mix of rough fodder andconcentrates, so that ozone damage leadsto an increase of meadowland at theexpense of areas used for barleyproduction. As a result trading in barley isreduced, and production value declines.At given factor prices, the use of inputs inscenario 2 is identical regardless ofwhether the damage occurs or not.

State agricultural subsidies are assumed tobe in proportion to activity levels in thesector. This is because a large share of thesupport is in the form of land-areasubsidies, which are independent ofproduction volume. However, the marketregime for grain means that the supportvia the Government Grain Corporationdepends on the volume of production. Onthe other hand it is conceivable that theauthorities will wish to compensate forthe farmers' income loss by raising therates of support. Our supposition entailsthat scenario 1 increases the support inproportion with the increased resourcesused when ozone damage occurs. Inscenario 2 the level of support remainsunchanged because resource use isassumed to remain unchanged. Thegovernmental overall net income isassumed to be unchanged, and the pay-roll tax rate is chosen as the endogenousvariable.

4.3.2. Direct costs of crop damageThe direct costs of crop damage can becalculated by multiplying the yield lossesset out in table 4.3 with the shadow pricesin table 4.4 (adjusted to 1995-Nkr). Asexplained above, the costs depend on howthe authorities respond to crop damage,and we therefore distinguish betweenscenario 1 and scenario 2.

It will be seen from the table that the directcosts are very dependent on the authorities'response, which in turn depends on whatpolicy aims the authorities have for agricul-ture in Norway. If the authorities choose toallow activity levels in agriculture to remainunchanged (scenario 2), the direct costs areabout Nkr 200 million. This is because theyield loss is compensated for by relativelycheap imports. In this case costs of reducedgrass production make up about 64 percent of total direct costs, while reducedwheat and potato production accounts for13 and 22 per cent respectively. If theauthorities choose instead to raise activitylevels in order to maintain domestic pro-duction (scenario 1), costs come to aboutNkr 550 million. This is what it costs toproduce the concrete yield loss in Norway.In this case the costs of reduced grassproduction make up 76 per cent of totaldirect costs, while reduced wheat andpotato production now account for 16 and19 per cent respectively. As mentioned insection 4.3.1, no account is taken of landarea support that comes in addition to theproducer prices. The real costs aretherefore somewhat higher than Nkr 550million. However, this is captured in themacroeconomic analyses in the nextsection.

In 1992 the production value in agricul-ture was 27.2 billion 1995-Nkr (StatisticsNorway 1995c). Since production ismainly measured by the same prices asstated for scenario 1, we find that theyield loss corresponded to 2.0 per cent ofthe potential agricultural production (i.e.in the absence of crop damage). This isattributable to yield loss of respectively11, 8 and 7 per cent of the potential yieldof wheat, grass and potatoes in 1992 (seetable 4.3). Gross product, defined asproduction value less the value ofpurchased input products, is a better


63

measure of value-added in the sector, as itshows the value increase produced bylabour and real capital in the sector. In1992 the agriculture sector's gross productcame to 13 billion 1995-Nkr (StatisticsNorway 1996b).

If yield loss is viewed in a short-termperspective, i.e. as a one-time occurrence,scenario 1 becomes less relevant. This canto a degree be defended since 1992 was ayear of high ozone exposure. As a one-time occurrence the yield loss wouldprobably be compensated for by higherimports so that the result in scenario 2 isrepresentative for the costs. However,since the farmer will in this case increasethe share of concentrates, which aredearer per unit of feed than rough fodder,the costs will probably be somewhathigher than illustrated by scenario 2.Further additional costs, such as adjust-ment costs, will also be associated with asudden yield reduction of this kind.Moreover, income lost in agriculture willdistort national income distribution.

4.3.3. Macroeconomic effects ofcrop damage

The macroeconomic effects depend asmentioned on how the authorities respond,which again depends on which policy aimsare emphasised for agriculture. In the mainwe allow changes in gross domestic pro-duct (GDP) to be estimates for the total

social costs of crop damage. GDP equals thesum of gross product in all sectors, and istherefore a measure of total value-added inthe economy. However, as shown earlier,the GDP concept may give a distortedpicture of value-added in agriculture,inasmuch as the product prices on which itis based may differ from the real shadowprices. Gross product in agriculture istherefore dealt with separately. The resultsare shown in table 4.6.

It is seen in scenario 1, “Maintenance ofproduction potential”, that GDP is reducedby the considerable margin of Nkr 1.2billion, which is more than double thedirect costs in table 4.5. Almost threequarters of the loss in GDP is incurred insectors other than the agriculture sector,since resource use in agriculture has to beincreased to compensate for the yield loss.As a result resources are withdrawn fromother economic activity which (in theaccounting context) are far more produc-tive. Given the authorities' choice ofresponse in this scenario, the results maybe interpreted to mean that agriculturalproduction has far greater real value thanthe GDP concept indicates. Even if agricul-tural production is maintained, grossproduct in the sector falls by 3-4 per cent.This is because farmers have to purchasemore inputs in order to maintain the sameproduction level, thereby reducing value-added in the agriculture sector.

Table 4.5. Direct costs of crop damage resulting from ground-level ozone, 1992. Mill. 1995-Nkr1

Scenario 1Maintenance of production potential


Mill. 1995-Nkr Share (per cent) Mill. 1995-Nkr Share (per cent)

Wheat 87 16 27 13Potato 48 9 48 23Grass 420 76 131 64Overall total 555 100 206 100

1 The consumer price index is used to index current prices to 1995-Nkr (Statistics Norway 1996b).


64

The resource allocation is illustrated bythe fact that employment in agricultureincreases by almost 3.5 million person-hours in this scenario. The relativeincrease roughly equals the relativedecline in productivity. Increased resourceuse also means more inputs, consistinginter alia of agricultural commodities (e.g.concentrates). This explains why importsof agricultural commodities increase evenwhen production is maintained. Such anoutcome may seem unreasonable in thisscenario, and a better approach couldhave been to allow these imports to beunchanged. Agricultural production wouldthen have attracted even more domesticresources, and the fall in GDP would havebeen even greater.

Private consumption of foodstuffs isslightly reduced in this scenario, partlybecause agricultural production demandshigher inputs of processed agriculturalcommodities. Prices of such commoditiesare consequently somewhat higher. Con-currently reduced GDP leads to lowerincomes which is reflected in lower dem-and for all normal goods. Governmentsubsidies to agriculture rise by about Nkr

150 million. This builds on our assump-tion of proportionality between activitylevels and size of support.

In scenario 2, “Maintenance of activitylevels”, GDP is reduced by just over Nkr700 million. The biggest reduction is inthe agricultural sector, where gross pro-duction is reduced by 6 per cent. This isbecause production value is reduced byabout 2.3 per cent, at the same time asuse of inputs shows negligible change.37

The relative reduction of value-added, orgross product, is therefore quite large.Compared with the direct costs of aroundNkr 200 million in table 4.5, the decline ofabout Nkr 550 million in the agriculturesector's gross product seems large. This isbecause the shadow prices in scenario 2are largely based on world market prices,while gross product is measured by dome-stic prices. We interpret this scenario suchthat in the event of marginal changes inactivity and production, activity levels inagriculture have a greater intrinsic value

37 Actually employment rises by 200 000 person-hours, whereas other inputs are somewhat reduced(e.g. real capital).

Table 4.6. Changes in macroeconomic aggregates in connection with crop damage resulting from ground-level ozone, 1992. Mill. 1995-Nkr

Scenario 1Maintenance of production

potential


Mill. Nkr Per cent Mill. Nkr Per cent

Gross domestic product (GDP) -1 236 -0.2 -730 -0.1 Gross product in agriculture -333 -3.6 -554 -5.8 GDP excl. agriculture -903 -0.1 -176 0.0

Gross production value in agriculture1 0 0 -555 -2.3Imports of agricultural commodities 122 1.2 486 4.8Private consumption of foodstuffs -41 -0.1 -1 0.0Person-hours worked in agriculture (1 000 hrs) 3 369 2.4 223 0.2Government transfers to agriculture 157 2.4 0 01 The model's reference path gives a somewhat smaller production value (and gross product) in agriculture (to be exact 23.1 bn1995-Nkr) than the amount mentioned in section 4.3.2 (i.e. 27.2 bn 1995-Nkr). Hence the yield loss measures 2.3 per cent ofproduction rather than 2.0.


65

than production (beyond the world marketprice). In this case it may therefore beasserted that the reduction in gross productoverstates the real cost of reduced value-added in the sector by about Nkr 350million (i.e., the difference betweenreduced gross product in agriculture intable 4.6 and the direct cost in table 4.5).The real social costs in this scenario areaccordingly about Nkr 380 million, i.e. farlower than the reduction in GDP.

Gross product in other sectors is reducedby Nkr 150-200 million, despite the factthat the agriculture sector does not layclaim to more resources. There are severalreasons for this. First, the yield reductionleads to reduced incomes for farmers(government subsidies are unchanged inthis scenario), so that total demand forgoods in the economy is reduced. Second,imports of agricultural commodities riseby about 5 per cent due to yield loss. As aresult the balance of trade is disrupted,and other imports have to be restricted tocompensate (in the event, exports must beincreased). This entails restrictions for therest of the economy.

Private consumption of foodstuffs remainsroughly unchanged despite the fact thatincomes are reduced as a result of thedrop in GDP. This is because food pricesfall by 0.1 per cent since a larger share ofagricultural products are imported.

A summary of the economic analyses sug-gests that the overall social costs are highlydependent on how the authorities respondto reduced plant production resulting fromozone exposure, see table 4.7. If theauthorities choose to let activity levels inagriculture remain unchanged, and tocompensate for the yield loss by higherimports, the costs come to about Nkr 400million, according to our calculations. If,

instead, the authorities choose to increaseactivity levels in agriculture so that pro-duction levels remain unchanged, we findthat costs are of the order of Nkr 1.2billion, i.e., three times as large. Moreover,note the importance of studying theindirect macroeconomic effects. In scenario1 these effects actually constitute a largershare of total costs than do the direct costs.In scenario 2 the macroeconomic costs alsoconstitute a considerable share.

4.4. ConclusionIn this study we have made an attempt onthe basis of current knowledge to elucidatewhat consequences ground-level ozonemay have for farm crops in years featuringvery high ozone concentrations in Norway.Before summing up our conclusions, someof the main uncertainties in the calculationsare set out. It is important to have these inmind when considering our results.

First, the assessments are based on theyear 1992 when exposure doses ofground-level ozone in Norway were parti-cularly high. It is not clear to what degreethe results are relevant for more normalyears. Second, the damage functionsemployed are in general highly provisionalsince documentation through crop experi-mentation is very limited with the possibleexception of wheat. Wide variation inozone-sensitivity between plant typeswithin a species makes it difficult to arriveat correct damage functions. Moreover,

Table 4.7. Total social costs of crop damage resulting from ground-level ozone, 1992.Mill. 1995-Nkr

Scenario 1Maintenance of

productionpotential

Scenario 2Maintenance of

activity level

Direct costs 555 206Indirect costs 681 176Total costs 1 236 382


66

we know that various climatic factors im-pact on the ozone effect, but this has yetto be documented quantitatively. Third,the representativeness of monitoring sitesfor ground-level ozone has not beenquantified, and the monitoring network isrelatively sparse. Ozone gradients overvegetation that are affected by vegetationtype, topography and meteorologicalconditions also make it complicated tocalculate correct exposure doses to whichvegetation is exposed. Fourth, theauthorities have several policy aims foragriculture, and it is therefore unclearhow they would respond to reduced yieldsover a long period. Moreover, as the agri-cultural sector is highly regulated, it isdifficult to bring to light the real economicvalue of crop yields. Finally, macroecono-mic models attempt to describe howplayers in the economy act, a processwhich is naturally extremely complex.

Thus, these factors indicate that as ourknowledge increases the estimates for cropdamage will probably change substantially.Our provisional results indicate that in1992 ground-level ozone entailed a loss ofabout 30 thousand tonnes of wheat, andabout the same volume of potatoes. Asregards cultivated meadow, ozoneexposure led to a loss of just over 300thousand tonnes of grass, according to theresults. These losses correspond to about10 per cent of a normal annual yield. Thesize of the losses is related to the fact thatalmost 90 per cent of the total wheat-growing area experienced ozone exposuresin excess of recommended critical levels,while the corresponding shares for potatoesand cultivated meadow were just under60 per cent.

The further calculations indicate that theeconomic cost of yield loss caused byground-level ozone in 1992 was in the

region of Nkr 400 - 1 200 million. The widevariation is due to two alternative assump-tions regarding the authorities' policy aimsfor the agricultural sector. The lowest costemerges when the focus is on activity levelsin the sector. The yield loss in this caseentails reduced production, which is com-pensated for by increasing imports ofagricultural products. The highest costemerges when maintenance of productionlevels is regarded as especially important.In this case the ozone exposure entailswithdrawal of resources from other sectorsto the agriculture sector in order tocompensate for the crop damage. Since farmore resources are needed to attain a givenproduction value in agriculture than inother sectors, the total costs of the yieldloss are extra large in this case.

In both cases the direct costs make uponly about half of the total costs. Thisdemonstrates the importance of analysingmacroeconomic impacts in such calcula-tions in order to study spillover effects inthe economy. In the case of unchangedproduction the total costs are as much as130 per cent higher than the direct costsowing to productivity differences amongthe various sectors.

Because of the substantial uncertaintyassociated both with physical andeconomic calculations, the social costsmust be viewed as rough estimates.However, they provide a good indicationthat long-range pollution in the form ofground-level ozone gives rise to majorcosts for Norway in years of high ozoneconcentrations. This underscores theauthorities' view that it is “necessary toreduce the environmental burden inflictedon agricultural land by, for example, airpollution” (Ministry of Agriculture 1992,chapter 3.1.2).


67

5.1. Introduction

5.1.1. BackgroundIt has long been clear that exposure tovarious polluting gases and particulatesmay give rise to, or exacerbate, a varietyof respiratory diseases. Greater uncer-tainty however has attached to the level ofrisk of incurring such ailments. Recentresearch have brought increased know-ledge in this field. Numerous epidemio-logical studies have pinpointed so-calleddose-response functions, i.e. statisticalrelationships between air pollution levelsand morbidity and mortality. These func-tions can be used to compute associationsbetween economic activity, pollution andhealth effects. Some health effects alsohave a feedback-effect on the economy inthat increased morbidity leads to work-force reductions and lower workerproductivity.

This study examines the internationalliterature dealing with associationsbetween air pollution and health effects,and shows how such associations can beutilised to estimate the feedback-effects ofpollution on economic activity. Moreover,

38 The article has earlier been published in Norwegianin Rosendahl (1996).

concrete calculations of health effects inOslo resulting from the current pollutionsituation are presented together withassociated costs. Alongside the dose-response functions figuring in the inter-national literature, use is made offunctions for the concentration of airpollutants (particulates and NO2) in Osloat various emission levels. These functionshave been constructed by the NorwegianInstitute for Air Research (NILU) at therequest of Statistics Norway (Walker1997).

Previous Norwegian studies of health costsof pollution were based largely on theNorwegian Pollution Control Authority's(SFT's) analyses of pollution abatementmeasures for the capital Oslo (SFT 1987)and the small neighbouring towns Sarps-borg and Fredrikstad (SFT 1988)39. Theseanalyses were in turn based on epidemio-logical studies by Lave and Seskin (1970),and were indirectly linked to numbers ofpersons exposed to SO2 concentrations inexcess of the SFT's recommended thresholdvalues. The cost estimates were based oncontingent-valuation studies, and costs

39 See for instance Brendemoen et al. (1992).

5. Health effects of air pollutionand impacts on economicactivity38

Knut Einar Rosendahl


68

associated with other pollutants wererelated to SO2 costs by a panel of experts.

There are several reasons why it is nowtime to update health costs in Norway.First, the focus on SO2 in the studiescarried out in the seventies has beencalled into question; internationallyattention is now primarily directed atvarious particulate categories and ozonesince these components seem to correlatebest with severe health effects. Based onthis, Hall et al. (1992) have calculated thehealth benefits of cleaner air in the LosAngeles area. They found that the gainachieved by reducing pollution levels tonational standards would amount to USD6.4 billion and USD 2.7 billion forparticulates and ozone respectively (bestestimate). Other studies (e.g. EC 1995,ORNL/RFF 1994) also find health costs tobe greatest for particulates and thereafterozone. However, it should be realised thatthese components are indicators for acomplex pollution situation and thatdirect causal relationships are difficult toprove. Hence the possibility that othercomponents, such as SO2 and NO2, are justas harmful as particulates and ozonecannot be disregarded.

Second, there is little reason to focusparticularly on specific threshold valuesfor air pollution in a population; suchvalues are specific to individuals, and sofar it has not proved possible to documentany lower level of concentration for healtheffects that is applicable to all individuals.This is especially relevant for particulateswhere the results from a number ofepidemiological studies have prompted anexpert group attached to the EuropeanOffice of the World Health Organisation –in a new report on guidelines for airquality, WHO (1995) – to no longerrecommend guidelines for particulates.

Third, recent literature (see for instanceOstro (1993)) has provided a broaderbasis for identifying dose-responsefunctions that can be utilised directly inanalyses of health effects. Unlike in earlierNorwegian studies it is now possible tocalculate effects of various pollution levelson mortality, lost labour productivity andpublic health expenditure etc. Hence it isalso possible to distinguish betweenpurely economic costs and other costswhich must be valued separately (e.g. thevalue of reduced mortality).

Recent years' increasing awareness ofassociations between air pollution andhealth effects has been accompanied bymajor international efforts to computeexternal costs of various types of electri-city production. Two wide-rangingprojects carried out by the US Departmentof Energy (ORNL/RFF 1994) and the EUCommission (EC 1995) are particularlysignificant. Dose-response functions werecentral to their analyses. These studies,together with Ostro (1993,1994), a studyfrom the US Environmental ProtectionAgency (EPA 1995), and draft reports bythe National Institute of Public Health inNorway (1995), form much of the back-ground information for the present study.

5.1.2. Dose-response functionsThere are two main methods foruncovering links between air pollutionand health effects. One involves clinicalstudies where volunteers (or animals) areplaced in a chamber containing a highconcentration of a certain airbornepollutant and subsequently examined forvarious health effects, e.g. impairedpulmonary function. While such studiescan be used to demonstrate biologicaleffects, they throw no light on the fre-quency or risk of various health effects ina normal pollution situation. One reason


69

for this is that the situation in a chamberdiffers greatly from the real situation in anurban atmosphere. There are alsoconstraints on the types and extent ofhealth effects that lend themselves toexperimentation in a chamber (c.f.sickness absence and mortality).

The second method is epidemiologicalstudies where an area is examined forstatistical relationships between airpollution levels and the incidence ofcertain health effects in a sample of thepopulation. These studies can be used toformulate quantitative associationsbetween air pollution and health effects,i.e. dose-response functions. All functionspresented in this study stem from suchstudies, and they can in general beexpressed in the following form:

(5.1) ∆H = F(∆C) * P

where ∆H is the change in frequency orrisk of a specific health effect (i.e. mor-tality, sickness absence), ∆C is change inconcentration of a certain atmosphericpollutant (e.g. mean annual concentrationof particulates), and function F denotesthe estimated quantitative relationbetween these two. P is the number ofpersons living in the area concerned40. Inmost cases the estimated relation can beexpressed as a linear function. The actualrelation will of course be more complex.

The health damage that arises may haveeffects on economic activity, e.g. viaincreased sickness absence. This can ingeneral be expressed as

(5.2) ∆E = G(∆H)

40 In some functions the left-hand side in (5.1) is∆H / H, and P is removed from the right-hand side.

where ∆Ε denotes change in the level of aspecific economic variable, and function Gdenotes the quantitative relationshipbetween this and the resultant healthimpairment. Whereas the functions inequation (5.1) stem from internationalstudies, the functions in equation (5.2)are constructed on the basis of variousNorwegian statistics and various choicesof assumptions.

By combining these two functions weobtain the following general relationshipbetween the concentration of air pollutionand economic activity41:

(5.3) ∆E = G(F(∆C)*P) = J(∆C,P)

We distinguish between acute and chronicdose-response functions, or effects. Byacute effects we mean health effects thatoccur after short-term exposure to airpollution. In this case most epidemio-logical studies utilise average concen-trations over one or more days as anindicator of the pollution level. Chroniceffects refer to health effects that occurafter long-term exposure, i.e., one orseveral years. These are only referred forparticulate matter.

Pioneering studies of dose-responserelationships between air pollution andhealth were carried out by Lave andSeskin (1970, 1972). Since then a varietyof similar studies have been carried outwhich will be discussed more fully in theforthcoming sections. Since the end of the1980s both the number and quality ofsuch studies have made substantial advan-ces. International confidence in their fin-dings has grown, partly because in manycases they support each other to a surpris-

41 In some functions the left-hand side in (5.3) is∆Ε / E, and P is removed from the right-hand side.


70

ingly large degree. It is also interesting tonote that the studies have been performedin very different areas in terms of pollu-tion levels, climate etc.

The quality of the studies has improved inseveral ways. First, more and more studiesare using time series data, particularly toreveal acute effects that arise with short-term changes in pollution levels. Theadvantage of using time series data is thatpopulation-specific factors which alsohave a bearing on health are fairly con-stant: these include socioeconomic factors,smoking habits, state of health etc. Themost important remaining factors are theenvironmental and meteorological ones.

Second: an important criticism that can belevelled at earlier studies is the varyingdegree to which they managed to take thevarious pollution components intoaccount. The studies by Lave and Seskinhave for example been interpreted todemonstrate the existence of a link be-tween exposure to SO2 and mortality.However, they used an aggregatedpollution index, and subsequent studiesindicate that the strongest association isbetween particulate concentration andmortality (and other health effects). EC(1995) comment that in several studiesSO2 effects apparently disappear whenparticulates are measured correctly. Onthe other hand, a number of recentstudies exist where this is not the case(see section 5.4). Most epidemiologicalstudies now estimate with respect toseveral pollutants simultaneously. Somestudies find separate effects of particulatesand NO2 respectively (e.g. Braun-Fahrländer et al. 1992), implying that itought to be possible to add them together.In other cases it shows difficult to findseparate effects of different pollutants asthese are often strongly correlated

(Calthrop and Maddison 1996). One studymay find significant relationships betweenparticulates and a health impairment whileanother study finds significant relationsbetween NO2 and the same healthimpairment. Interpretation then becomesmore problematic. The interaction betweendifferent components is altogethercomplex. According to EC (1995) this isprimarily a problem for particulates, NO2

and SO2, whereas ozone effects canreasonably be considered as additive inrelation to the other components. Thus,awareness of the problem of double-counting must be maintained. There is,however, little danger of overlappingbetween the associations we use in thecalculation of social costs in this study.

The growing number of studies allows amore discriminating choice to be made.Ostro (1995) explains the criteria that hadto be applied to make use of a study inconnection with ORNL/RFF (1994): alldose-response relations available in thespecialist literature were studied, includ-ing relations linked to mortality and mor-bidity. The relations were assessed on thebasis of criteria such as generalisation (forapplication in other towns) and signi-ficance. The same procedure has beenapplied in other applications, such as EC(1995). Hence the dose-response func-tions taken from these studies have beenthoroughly assessed in internationallyrecognised studies. Other functionsutilised in our studies have also beenthoroughly assessed by recognisedspecialist bodies, such as EPA (1995).

Although relatively few epidemiologicalstudies have found significant associationsbetween NO2 and health effects, there areseveral reasons to warn against toningdown the importance of NO2. First, themix of air pollutants in Norway differs in


71

some respects from the mix in the USAand on the continent where epidemiolo-gical studies have largely been carried out.Moreover, several Finnish studies indicatethat NO2 may have serious consequencesin a climate resembling that in Norway(see section 5.3). Another importantreason why NO2 has been consideredharmful is health effects observed inclinical studies and in studies of indoorpollution.

The difficulties associated with demon-strating quantitative relationshipsbetween air pollution and health effects(see for instance Ostro (1994)) are worthemphasising in this context. It is conceiv-able that measurement problems explainwhy effects of outdoor NO2 are frequentlyunidentifiable. If effects primarily arise asa result of short-term exposure to parti-cularly high concentrations (e.g. close to aheavily trafficked road), the effects will behighly dependent on individual exposurewhich may be difficult to measure. Thesame applies if the health effects arise as aresult of several years' exposure.

A pertinent question is whether resultsfrom epidemiological studies carried out inone area can be applied in other areas. Thisis a relevant concern since the greatmajority of studies referred to in the litera-ture have been carried out in the USA.Commenting on this, Ostro (1995)recommends that further studies of asimilar nature be carried out in Europe. Healso states that the perceived uncertaintiesshould not prevent the results observed inthe USA from being transferred toEuropean cities with a view to estimatingsocial costs of pollution in Europe. He basesthis view on the fact that the studiesutilised for instance in ORNL/RFF (1994)and EC (1995) have made allowance fordisturbing factors in their estimations, and

that they are mutually supportive(especially for mortality). Moreover, theyhave been carried out in widely differingclimatic areas. A number of studies carriedout in Europe and other parts of the worldin recent years are also worth noting. Wecite studies in for example Finland,Germany, Switzerland, Spain, the UnitedKingdom, Chile and China that essentiallyconfirm the US results.

Pearce (1995) addresses the uncertaintyaspect in his valuation analysis, and assertsthat any uncertainty involved in the trans-fer of dose-response functions is of a minornature. EC (1995) also touches on thisuncertainty, but the authors are content tostate that the functions describing a bio-logical response, such as mortality, are to ahigher degree transferrable to other townsthan functions which also describe a socialresponse, such as hospital admissions.However, this is not to say that EC (1995)omits the latter type of functions.

When applying dose-response functions inthe present study the effects are accumul-ated over a year and expressed as theeffect of changes to mean annual concen-trations. This is common practice insimilar applications (EC (1995); ORNL/RFF (1994)). It presupposes that associa-tions between air pollution and healthdamage are approximately linear, and thatthey are relevant for the pollution level atwhich the functions are applied. The firstcondition is met in the sense that theoriginal studies largely found linear orapproximately linear functions thatcorrespond significantly with the obser-vations42. There are, however, grounds for

42 A number of the functions are linear on a percent-age basis; within the area of application these func-tions are approximately linear inasmuch as thepollution level only explains a small share of theparticular health effect.


72

suspecting that the actual relationshipmay be more complex than this. Apossible objection is that acute healthdamage may only arise with episodes ofhigh pollution over several days, and thatpollution levels are otherwise of littlesignificance. Whether or not this is correctis somewhat unclear. However, sinceapproximately linear functions seem tocorrespond well with the observations it isreasonable to use them. Moreover, thereis reason to believe that episodes of heavypollution are strongly correlated with theannual mean concentration. As for thesecond condition and the question ofthreshold values, these are discussed foreach individual component in the follow-ing sections. It has already been pointedout that in the case of particulates thedose-response functions also seem toapply at very low concentrations.

We present dose-response functions forparticulates, NO2, ozone and SO2. In theapplication for Oslo, however, we concen-trate on the two first-mentioned. Episodesof high ozone concentrations in Norwayare mainly due to foreign emissions,although Norwegian emissions of NOx areof some significance (see Simpson et al.1996). SO2 concentrations are relativelylow in Oslo, and it is unclear whether theyhave a separate health effect beyond thatof particulates. Since the present study isespecially concerned with effects oneconomic activity, the selection of dose-response functions presented will by nomeans be complete (see Aunan (1995) fora similar review of the literature from asomewhat different angle). However wealso focus on serious health effects whichare not necessarily significant for econo-mic activity, such as higher mortalityamong the elderly and chronically ill. Inthat connection we discuss to some extenteconomic valuation of such effects.

Although a number of quantitativeassociations between pollution levels andhealth effects have been demonstrated,there is no reason to believe that all theground has been covered (see e.g. Ostro(1994)). As mentioned above, associationsmay in many cases be difficult to find. Thefew studies that have so far found linksbetween long-term exposure and varioushealth effects suggest that a substantialvolume of effects remains to be clarified. Inother words, using quantitative relations tocalculate costs of pollution will in allprobability yield an underestimate.

In the next sections dose-response func-tions based on international literature ispresented for the four pollutants. More-over, we analyse the economic impactsfollowing from health effects induced byair pollution. Threshold values are dis-cussed in the end of each section. Insection 5.6 the population exposure to airpollution in Oslo is discussed, and wepresent annual health effects and socialcosts in Oslo based on the dose-responsefunctions.

5.2. Health effects of particulatesThe term particulates includes a numberof chemical compounds, among them sul-phates and nitrates. However, size is whatis considered most important in a normalpollution situation43. Particulates with adiameter less than 10µm (PM10) aremainly considered harmful to the lowerrespiratory tract. In the upper respiratorytract larger particulates may also be harm-ful. PM10 further comprises coarse parti-culates and fine particulates, the latterwith a diameter below 2.5µm (PM2.5).

43 If the particulates contain toxic heavy metals owingto emissions from a particular industrial source, thechemical composition may be just as hazardous assize.


73

These are thought to be of the greatestsignificance for health since they pene-trate deeper into the respiratory tract.Recently discussion has turned onwhether PM1 or even PM0.1 is mostimportant. Particulates are carriers ofother pollutants, among them SO2, andsuch combinations are also consideredimportant. The main sources of fineparticulates are combustion processes,especially related to vehicle traffic,residential heating and, in some towns,port traffic. The sources of coarseparticulates are largely mechanicallygenerated, e.g. by use of studded tires.Studies of dose-response relations linkedto particulates have been carried out usingvarious choices of target variables such asPM2.5, PM10, TSP (total suspended par-ticles) and sulphates. In most applicationsof dose-response functions the variousvariables are converted to PM10, on theassumption of fixed relations between thevarious particulate sizes.44 This is a simpli-fication inasmuch as composition mayvary from town to town.

In the following we distinguish betweenacute and chronic effects. Each healtheffect will be dealt with in a separatesubsection. Appendix B contains a tablesummarising the dose-response functionstaken from the literature and a tableshowing the impact of the health effectson economic activity, etc.

5.2.1. Acute effects

Premature mortalityStudies that have found a significantlyhigher risk of death on days (or in periods)

44 The relations between PM2.5 and PM10, and betweenPM10 and TSP, are assumed to be 0.65 and 0.55,respectively (ORNL/RFF 1994). These, according tothe Norwegian Institute for Air Research, are also fairconversion rates for Oslo.

of high particulate concentration (see forinstance Plagiannakos and Parker (1988)and Schwartz (1993b)) are now in doublefigures. Moreover, the results from most ofthese studies are strikingly uniform, despitethe fact that the underlying observationshave been made at widely different levelsof concentration, climatic conditions andmix of other pollutants. Some of the studiesbuild on time series data, while othersbuild on cross-sectional data.

Based on a number of these studies, Ostro(1993) concludes that the consensus dose-response relationship between mortalityand short-term change in particulateconcentration is:

(5.4) ∆D / D = 9.6*10-4 * ∆PM10

where D denotes the number of deaths inthe period, while ∆PM10 denotes change inPM10 concentration (period mean, µg/m3).In this context, period means one or moredays, since the studies focused on short-term changes in concentration. The lowerand upper limits for the coefficient arerespectively 6.3*10-4 and 1.3*10-3. In otherwords an increase of 10µg/m3 in PM10

concentration is associated with an increaseof about 1 per cent in the number of deaths.Pearce (1995) states that a linear relationappears to be soundly documented. Hebases this on the fairly high level ofconsistency shown by the studies despitetheir having been carried out at differentconcentration levels. A new study fromChile (Ostro et al. 1995) confirms thedose-response function mentioned,indicating that this function applies inlarge measure universally.

In some of the studies it is possible toisolate various causes of death, and itturns out that a link exists between deathsfrom respiratory ailments and increases in


74

particulate concentration. This accordswith the results for inter alia hospitaladmissions in section 5.2.1.

The degree to which increases in parti-culate concentration cause an increase inthe mortality rate in that year is uncertain.The reason is uncertainty as to how longthose affected would otherwise haveremained alive. Rowe et al. (1995) referto a study by Schwartz and Dockery(1992) which found that the risk ofmortality at increased concentrations ofPM10 is more than 70 times higher forindividuals above 65 than for individualsbelow 65. However, this figure must beviewed in the light of the higher generalmortality in the oldest age group. TheNational Institute of Public Health (1995)refers to a study in the Czech Republicindicating that deaths due to pulmonarydisease among children aged one monthto one year rise by 58 per cent with a 10µg/m3 increase in PM10 concentration.Clearly this particular group has a longresidual lifetime. All in all the resultsindicate that it is primarily the chronicallyill and the elderly who are affected, butthat residual lifetime is not insignificant(Schwartz and Dockery 1992).

This must however be collated with recentresults for the relationship between long-term particulate exposure and mortalitywhich are discussed in section 5.2.2.Although the uncertainty here is larger,they at any rate indicate that particulatepollution (short-term or long-term) leadsto several years' loss of lifetime. Pearce(1995) also refers to a study by Cropperand Simon (1994) which states that pro-bable residual lifetime is 10 to 15 years inthe case of deaths resulting from pollution.

The fact that particulates include compo-nents with various chemical and physical

characteristics, and that particulatesgenerally carry other pollutants, hasprompted discussion about real causalrelations. Perceptions variously focus onthe importance of fine particulates(PM2.5), acid aerosols (sulphates/nitrates)or SO2. A study by Pope et al. (1992) isespecially pertinent in this respect. Thisstudy was carried out in the winter half-year in Utah Valley where there was littleSO2 and ozone in the atmosphere, andwhere no measurable acidity was attachedto particulates. The result for mortalityfrom this study was somewhat higher thanthe interval for the dose-response functionin equation (5.4).45 This and similarstudies (e.g. Fairley 1990) indicate that acausal relationship exists between (at anyrate) particulates, measured in size, andmortality. Ostro (1995) also asserts thatthe numerous studies showing consistentrelations between PM10 (and TSP) andmortality (and morbidity) across climate,season, other pollutants and population,strongly suggest that the causal com-ponent is closely correlated with PM10

(and TSP), although many of them do notinclude acid aerosols. However, a causalconnection can never be proven byepidemiological studies.

Ostro (1993) states that most criteria forcausality, as established by Hill (1965),are satisfied by the dose-response functionin equation (5.4). This entails, briefly, thatthe relation between particulates andmortality is consistent, that it is disease-specific (linked to respiratory diseasesamong others), that the relation generatesa dose-response curve, and that it accordswith observed effects on other healthparameters. The only criterion which

45 The increased mortality is due to cardiovasculardisease, pneumonia and chronic obstructive respira-tory disease.


75

according to Ostro is not satisfied is thebiological mechanism between pollutionand mortality. However, theories exist(e.g., Seaton et al. (1995)). Ostro (1993)specifies that not all the criteria have to bemet in order to establish a strong indi-cation of a causal relation.

It is reasonable to believe that the effectson mortality discussed in this section havenegligible impact on economic activitythrough a reduced work force. Personswho die due to a short-term increase inpollution levels are mainly elderly andchronically ill and are outside the work-force at the outset. The prime exceptionhere is sick children who could haverecovered if they had been able to growup. However, this group represents amarginal share of deaths. Hence moststudies which attempt to calculate thecosts of increased mortality opt for anestimate for the value of a statistical life.Effects on economic productivity are notconsidered. However, in section 5.2.2such effects are discussed for chroniceffects on mortality risks.

Sickness absence and reduced labourproductivityRelatively few studies have estimated therelationship between air pollution andsickness absence or reduced labour pro-ductivity. Most relevant studies in thisfield are those carried out by Bart Ostroon an extensive annual US health survey(120 000 persons in 84 urban districts).The first study utilised data material from1976 (Ostro 1983). Significant relationswere observed between particulate con-centration and the number of restrictedactivity days (RAD). In a later study aspecification test was carried out on the1976 result for the years 1976-1981(Ostro 1987). This study confirmed theresults from the first study in all years.

Based on these results, Ostro (1994)himself arrives at the following relation-ship between PM10 concentration and RAD(best estimate), which is the one used inall the cost studies referred to in theintroduction.

(5.5) ∆RAD = 0.058 * ∆PM10 * P

RAD denotes the number of restrictedactivity days per year, and P is the numberof persons exposed. Here PM10 denotesmean annual concentration.

The relationship applies to all adults. Thefunction states that a one-unit increase inPM10 concentration entails an increase inrestricted activity days of 0.058 days perperson per year.

Subsequent studies have tested for per-sons in employment (Ostro 1990), and forseveral pollutants simultaneously (Ostroand Rothschild 1989). Similar resultswere found. A recent study on sick-leavesin Oslo by Hansen and Selte (1997) alsoconfirms the results by Ostro (1987). Thisis particularly interesting for the presentstudy, as we calculate annual healtheffects in Oslo. Moreover, it should beadded that the function is also supportedby the numerous studies that find arelation between particulate concentrationand various types of morbidity. Since it isbased on the largest data material, thefunction referred to in equation (5.5) hasbeen chosen in applied studies (see forinstance EC (1995) and ORNL/RFF(1994)).

According to ORNL/RFF (1994), 62 percent of all RAD are bed-disability days andwork-loss days, while the remainder aredescribed as minor restricted activity days(MRAD). This breakdown features in thehealth study on which Ostro's study is


76

based, and is used in the application of hisresults. ORNL/RFF have valued the cost ofone MRAD at just over 1/4 of wages. Thisis based on a study of willingness to pay,implying that the valuation probably alsocovers some welfare loss. We thereforeassume that one MRAD results in an esti-mated 10 per cent reduction in labourproductivity. Other RAD of course reducelabour productivity by 100 per cent.Hence an average day of restricted activitycorresponds to a 66 per cent reduction inproductivity. We further assume that RADare evenly spread across workdays andleisure days. In the latter case this has noeffect on the supply of labour. If eachperson has 0.058 more RAD per year witha one-unit increase in PM10 concentration,the average daily increase in RAD is0.058÷365=0.00016. Since an averageRAD reduces productivity by 66 per cent,the average daily reduction in productivitywill equal 0.00016*0.66=0.00011. Thiswill apply regardless of whether a personis employed or not, so an average person-hour in the economy is correspondinglyreduced by a factor of 0.00011 by a one-unit increase in concentration. This givesthe following function for change in thesupply of labour resulting from changedparticulate concentration:

(5.6) ∆L / L = - 1.1*10-4 * ∆PM10 * P/TP

L denotes the supply of labour in terms ofperson-hours worked throughout theeconomy, and P/TP is the share of thetotal population exposed to this change inconcentration. The function states that ifPM10 concentration increases by 10 µg/m3,labour productivity diminishes by anaverage of 1‰ per person-hour.

A much utilised study of the occurrence ofrespiratory symptoms is Krupnick et al.(1990). The study focuses on symptoms in

the upper and lower respiratory tract.According to Ostro (1994) and Rowe et al.(1995) the results lead to the followingfunction:

(5.7) ∆RS = 0.18 * ∆PM10 * P

where RS denotes the number of dayswith respiratory symptoms per year, and Psignifies the number of persons exposed.PM10 signifies the mean annual concen-tration measured in µg/m3.

The function states that an increase of0.18 days in respiratory symptoms perperson per year will accompany a one-unitincrease in annual concentration of PM10.It is however conceivable that a day withrespiratory symptoms may partiallyoverlap a day of restricted activity (RAD -see equation (5.5)). Hence we deduct thenumber of RAD from the function for thenumber of days with respiratorysymptoms resulting from changed PM10

concentration. Eskeland (1995) alsomakes use of the dose-response functionsfor RAD and respiratory symptoms, butwithout deducting any overlapping. Hefurther assumes that a day with res-piratory symptoms entails a 6 per centreduction in labour productivity. Since hedoes not believe in overlapping betweenrespiratory symptoms and RAD, 6 per centcannot accommodate more extensivesymptoms that would lead to greaterrestrictions on activity. The basis for thissupposition is unclear. Following the sameprocedure as for RAD, we obtain thefollowing function:

(5.8) ∆L / L = - 2.1*10-5 * ∆PM10 * P/TP

This function states that an increase of 10µg/m3 in PM10 concentration results in areduction of labour productivity averaging0.2‰ per person-hour in the economy.


77

Although Krupnick et al. (1990) is theonly study to have focused on respiratorysymptoms in general, many studies havefor example focused on asthma attacks,the common cold etc. A Swiss study byBraun-Fahrländer et al. (1992) foundinter alia significant links between PM10

concentration and the duration of variousrespiratory symptoms in children. Wehave not included these functions becausethey probably overlap the function fromKrupnick et al. They do contribute how-ever to support the function we utilise.

Hospital admissionsMany studies have focused on therelationship between particulate concen-tration and the number of respiratoryhospital admissions. Some of the studiesaddress specific respiratory diseases whileothers take in the entire group. To ourknowledge the only two studies that focuson respiratory admissions in general arePlagiannakos and Parker (1988) which isapplied by ORNL/RFF (1994), and Pope(1991) which is used by Rowe et al.(1995) and Ostro (1994). The two studiesdiffer greatly: Plagiannakos and Parkersuggest a relationship that is 8-9 timeslarger than that found by Pope. Thereason for this is not clear. The study byPope (1991) is particularly interesting forNorwegian conditions since it was carriedout in Utah Valley in the winter half-yearwith low concentrations of SO2 and ozone(cf. the discussion under mortality). Onthe other hand the the number of respira-tory hospital admissions per capita isabout three times greater in Norway (cf.Statistics Norway 1995d) than in the areastudied by Pope. If this is because thethreshold for admission is lower inNorway's welfare state, it suggests thatPope's results will underestimate the effectin Norway.

EC (1995) employ the number of hospitaladmissions related to respiratory infec-tions and chronic obstructive pulmonarydisease (COPD) respectively as theirtarget variable. If we assume that theincrease in respiratory hospital admissionsbreaks down to equal percentages onvarious respiratory diseases, we find bycomparing with Norwegian statistics onpatients (Statistics Norway 1995d) thatthese results fit in well with Pope's result(1991). A Spanish study by Sunyer et al.(1991) focused on hospital admissionsrelated to COPD. By rough and readyconversion of the results from this study,we find a relationship that is about threetimes larger than Pope's result (1991).This result is confirmed by other studies.

For our purpose we opt for the result fromPope (1991), multiplied by a factor of 3.This appears to be an acceptable estimatebased on the various results and assess-ments commented on above. We obtainthe following relation between PM10

concentration and respiratory hospitaladmissions:

(5.9) ∆RHA = 3.6*10-5 * ∆PM10 * P

where RHA denotes the number ofrespiratory hospital admissions per year,and P is the number of persons exposed.PM10 signifies the mean annual concen-tration measured in µg/m3. The functionstates that a one-unit increase in the meanannual concentration of PM10 results in anincrease in hospital admissions of 3.6 peryear in a population of 100 000. A naturalchoice of uncertainty interval here wouldbe the results of Pope (1991), andPlagiannakos and Parker (1988),respectively.

We can now utilise the dose-responsefunction to calculate the relationship


78

between change in PM10 concentrationand change in public health expenditure.An implicit assumption is that the changein number of respiratory admissionsresulting from a changed pollutionsituation changes the level of publichealth expenditure. The alternative wouldbe to assume that it leads to changes inhospital queues. In the short term thelatter alternative is probably mostrealistic, while in the long term there isreason to believe that the level of publichealth expenditure will be affected by thedemand for hospital services.

Since the studies referred to above do notfocus on the duration of hospitaladmissions, it is natural to assume thatthese do not differ from respiratoryadmissions in general. In 1993 bed-daysper admission averaged 5.96 (8.78) forpersons admitted with diseases of therespiratory organs in Norway (figures forOslo in parentheses) (Statistics Norway1995d). The figure is lower for respiratoryinfections, and higher for pulmonarydiseases. We opt to keep to the averagefigure. Net operating expenditure per bed-day at Norwegian hospitals in 1994 wasNkr 3 187 (Statistics Norway 1993b).46 Inaddition there is investment expenditurewhich we put at 5 per cent of operatingexpenditure. We obtain the followingfunctions (figures for Oslo in paren-theses):

(5.10) ∆BD = 2.1*10-4 (3.2*10-4) * ∆PM10 * P

(5.11) ∆PHE = 0.76 (1.12) * ∆PM10 * P

where BD is the number of bed-days andPHE is public health expendituremeasured in 1994 Nkr.

46 Wage costs account for 71 per cent of this.

5.2.2. Chronic effectsIt is difficult to measure chronic healtheffects of pollution. However, in recentyears studies have been carried out in thisfield indicating that costs associated withchronic effects are especially heavy. Eventhough the uncertainty is larger for theseeffects, it is important to throw light ontheir impacts, where they are available.

Mortality riskA cohort study by Dockery et al. (1993)examined the effect on mortality of long-term exposure to various pollutants basedon historical data for six towns. Theymonitored 8 000 adults, and controlledfor individual risk factors such as smoking.They found a significantly higher risk ofmortality at higher concentrations ofparticulates (PM10 and PM2.5). For PM10

they found that an increase of 1 µg/m3 inmean concentration over several yearswas associated with an increase in themortality rate of about 8.5*10-3, i.e. 0.85per cent (95 per cent coefficient interval:2.7*10-3 - 1.4*10-2). A similar cohort studyby Pope at al. (1995) found results ofalmost the same order of magnitude. TheWHO (1995) refers to these two studiesand writes that a combination of theirresults indicates a relative risk of 1.10 atan increase of 10 µg/m3 in concentrationof PM2.5. Converting to PM10 gives thefollowing dose-response function (seefootnote 44):

(5.12) ∆D / D = 6.5*10-3 * ∆PM10

Whether it is in fact the average concen-tration over several years or an accumul-ation of episodic effects over a long periodthat is the deciding factor so far remainsunclear.

It is important to use these coefficientscorrectly. They cannot be used directly to


79

estimate the effect on the overall mortalityrate in the population in the long term.The coefficients must be interprested inthe light of the mortality rate in variousage groups, i.e. the risk of dying within acertain period (e.g. a year) upon reachinga certain age. The function in equation(5.12) states that this risk increases by anaverage of 0.65 per cent for every unitincrease in PM10 concentration. How thisis reflected in the overall mortality ratedepends inter alia on the distribution ofdeaths on age groups.47 These result canat all events safely be stated to indicate aconsiderably larger effect of chronicexposure than that found with short-termvariations in particulate concentration(see section 5.2.1). In the first place thecoefficient is about 7 times larger. Second,there is no question here of smallreductions in lifetime inasmuch as themortality rate rises.

Dockery et al. (1993) point out that theresults may be influenced by concen-trations present in earlier periods (i.e.prior to 1974) which could not be takeninto account. Since pollution levels at thattime were higher (hence also the differ-ence between concentrations in differenttowns), the effect may to some extenthave been overstated. Against this back-ground EC (1995) choose not to makedirect use of this estimate in their cal-culations, but present it to indicate thatmortality resulting from chronic exposuremay be a substantial factor. The WHO(1995) also writes that further cohortstudies of this type are needed beforeconfident conclusions can be drawn about

47 In an extreme population where everyone died onreaching the age of 80, this type of increased riskwould have virtually zero effect. Everyone would stillreach 80 years of age.

particulates' long-term effects onmortality.

Despite the uncertainty related to thiseffect on mortality risk, we want toillustrate how this effect may influence thework force of the economy. In the studyby Dockery et al. (1993) persons weredivided into 5-year groups (from 25 to 74years), and distinctions were drawn forinstance between those who were exposedto dust or gases at work and those whowere not. The article does not make clearwhether or not the risk factor variedamong the age-groups. However, theresults do show that increased risk wasprimarily associated with death from lungcancer (8.4 per cent) and heart andpulmonary diseases (53.1 per cent). Forthe remainder of the deaths as a wholerisk was unaffected by particulate concen-tration. We therefore choose risk factorsfor these causes of death corresponding tocoefficients of 1.1*10-2 per unit of PM10,and assume that the relative risk increaseis identical in all age-groups. In this waywe avoid problems associated with causesof death that are typical for younger age-groups, such as road accidents. Usingstatistics on causes of death in 1993(Statistics Norway 1995e), change inmortality at different ages can be cal-culated, both for Norway and Oslo. More-over, the effect this has on the workforcecan be calculated by assuming thataffected persons in the age-group 25-64years are average workers with a retire-ment age of 65 years. This calculationdoes not take account of the loss of labourin the period these persons are sick due tolung cancer or heart and pulmonarydisease. Hence the total effect on employ-ment will probably be substantially under-stated.


80

In 1993, 464 (53) persons in the agegroup 25-64 years died as a result of “amalignant tumour in the respiratoryorgans and the organs in the thoraciccavity” in Norway (Oslo figure in paren-theses) (Statistics Norway 1995e). All inall this category comprised 3.6 per cent ofall deaths in 1993 (against 8.4 per cent inthe study by Dockery et al.). If we assumethat the average age in each 5-year groupis halfway through the 5 years, we obtainan average age of 57.2 years for these 464(53) persons, i.e. they had 7.8 years left toretirement. Similarly we find that 2 106(223) persons in this age group died from“diseases of the circulatory organs orrespiratory organs”. This group accountedfor a total of 56.4 per cent of all deaths in1993 (53.1 per cent in Dockery et al.).The average age of these 2 106 (233)deceased persons was 56.3 years, i.e. 8.7years prior to assumed retirement age.Combining these two groups gives 2 570(276) deceased with an average of 8.5years to assumed retirement age. Bydividing the number of person-years in theeconomy in 1993 (2.87 billion - StatisticsNorway (1994c)) by the population in theage range 20-70 years at year-end (2.70million - Statistics Norway (1995f)), weobtain an average of 1 060 hours workedper person between 20 and 70 years in1993. This is probably a good approxi-mation for the age group upon which weare focusing (mainly 55-65 years).

We can now calculate by way ofillustration the effect on the workforce ofchanges in long-term concentration ofPM10. The above statistics indicate anannual loss of 23.2 (2.5) million person-hours owing to deaths in these categoriesin Norway (Oslo). The above dose-response function suggested a 1.1 per centincrease in risk per unit increase in con-centration of PM10. This gives the follow-

ing function for Norway (figures for Oslostill in parenthesis), which must be regar-ded as highly illustrative:

(5.13) ∆L = -2.6*105 (2.8*104) * ∆PM10*P/TP

where L is the number of person-hoursworked in the entire economy per yearand P/TP denotes the share of thepopulation exposed to changed PM10

concentration (e.g., in Oslo P=TP whenthe figure in parenthesis is used). PM10 is along-term mean. The next section showshow this function can be converted to afunction of the mean annual concen-tration of PM10.

The impact on the workforce of such aneffect is naturally less serious than thedirect health effect, namely a higher riskof death. It is difficult to quantify this interms of increased mortality. However,using mortality tables (Statistics Norway1995f) we can calculate the effect of sucha risk increase on life expectancy in thepopulation. If we apply the estimatespresented in WHO (1995), i.e. a riskincrease of 0.65 per cent per unit increasein PM10 concentration, to the age group 25years and over, we find the followingdose-response function which again mustbe regarded as illustrative (the relation wecomputed is approximately linear):

(5.14) ∆LEW = -0.059 * ∆PM10

(5.15) ∆LEM = -0.064 * ∆PM10

where LE is lifetime expectancy forwomen and men measured in number ofyears. Here too PM10 is a long-term mean.Thus, life expectancy for women falls by0.6 years at an increase of 10 µg/m3 inPM10 concentration, whereas the equiva-lent reduction for men is 0.65 years. This


81

fits in well with calculations done in theNetherlands, according to WHO (1995).The WHO refer to a study which, applyingthe same risk increase to Dutch men,found a reduction of 1.1 years in lifetimeexpectancy with a 10 µg/m3 increase inPM2.5 concentration. This corresponds toabout 15 µg/m3 of PM10 (see footnote 44).It should be stressed that this effect ishighly uncertain, and should only beregarded as indicative.

Chronic pulmonary diseasesFour studies from different areas - Abbeyet al. (1993), Schwartz (1993a), Xu andWang (1993) and Portney and Mullahy(1990) - have analysed the relationshipbetween particulate concentration and theoccurrence of various types of chronicobstructive pulmonary disease (COPD -incl. inter alia chronic bronchitis, asthmaand emphysema). What is surprising isthat all four find that the risk increases byabout 0.6 per cent with an increase of 1µg/m3 in long-term concentration of TSP.The relationships are significant. Conver-ting to PM10 concentration (see footnote44), we obtain the following associationbetween the long-term concentration ofPM10 and the relative change inoccurrences of COPD:

(5.16) ∆COPD / COPD= 1.1*10-2 * ∆PM10

That is, the frequency of COPD in thepopulation increases by 1.1 per cent perunit increase in the long-term concen-tration of PM10. Although there is someuncertainty attached to the conversionbetween TSP and PM10, this is reduced bythe availability of four different studies.PM10 is assumed to be a better healthindicator than TSP.

The study by Abbey et al. (1993) isespecially interesting since it is a timeseries study over 10 years in which asample of 4 000 persons were monitoredand examined for chronic pulmonarydiseases. The authors state that there wasa significant association between parti-culate concentration over the 10-yearperiod and the risk of developing chronicpulmonary diseases (COPD) in general.The same applied to chronic bronchitis inparticular, but not to asthma (both belongin the COPD group). In a follow-up to thisstudy Abbey et al. (1995) find similarresults and, in addition, a significantrelationship between particulate concen-tration and the risk of developing asthma.

In Rowe et al. (1995) and Ostro (1994),the result from Abbey et al. (1995) isconverted to a linear relationship applyingon an annual basis. This is done in light ofthe fact that the study was carried outover a ten-year period. The effect ofincreased concentration in any year isassumed to be one tenth of the effectobserved over the entire ten-year period.They arrive at the following relationship:

(5.17) ∆COPD = 6.1*10-5 * ∆PM10 * P

Here ∆COPD denotes change in thenumber of persons in the population witha chronic pulmonary disease, while P isthe size of the exposed population. Asstated, PM10 is the annual mean in thisfunction. It is natural to envisage some lagin this relationship inasmuch as the resultswere observed over a ten-year period.Hence a more correct function would beone where PM10 denotes mean concen-tration over a ten-year period at the sametime as the function still indicates annualeffects. It follows from the above cal-culation that the incidence of COPD in thepopulation is at the outset stipulated at


82

5-6 per cent. This probably fits in wellwith Norwegian conditions.

It is difficult to use equation (5.17) tocalculate effects on economic variablessuch as sickness absence and hospitaladmissions. It is simpler to apply equation(5.16), which moreover showed a highdegree of consistency across the variousstudies. We therefore construct linearrelations for various economic variablesbased on available statistics for Norwegianconditions. For hospital admissions weassume that the relative increase in thenumber of cases of COPD is reflected in aproportional increase in the number ofbed-days. In Norway in 1993 there were28.3 bed-days with the diagnosis COPDper 1 000 head of population (StatisticsNorway 1995d). Assuming that a 1.1 percent increase in risk accompanies a per-manent increase of 1 µg/m3 in the concen-tration of PM10, we obtain (the factor forOslo - where the average number of bed-days was 35 per cent higher - is in paren-theses):

(5.18) ∆BD = 3.1*10-4 (4.2*10-4) *∆PM10 * P

Here BD denotes the number of bed-daysat Norwegian hospitals per year, whilePM10 now denotes a long-term mean.

The number of bed-days in a given yearthus depends on PM10 concentrations inprevious years. In order to convert from along-term mean (chronic exposure) to anannual mean, we assume that PM10 con-centration in a given year has an effect onthe number of bed-days in the 10 ensuingyears and that this effect is identical for all10 years. This is of course a simplification,but should be viewed in light of the lackof knowledge about how a chronic diseasearises. Our choice reflects the study by

Abbey at al. (1993) which covered a 10-year period. The change in the number ofbed-days resulting from new cases ofCOPD now becomes a function of changedPM10 concentration in the preceding 10years. By utilising the same cost data as insection 5.2.1, we obtain the followingfunction for change in public healthexpenditure (the factor for Oslo is inparentheses):

(5.19) ∆PHE=1.1(1.5)*∆[i=∑

0

9

(PM10)-i /10] * P

As previously, PHE denotes public healthexpenditure measured in 1994 Nkr, while(PM10)-i signifies the mean annual concen-tration of PM10 i years ago. Hence theexpression in square brackets is a 10-yearmean.

We also assume that the relative increasein incidence of COPD results in a similarincrease in sickness absence, rehabilitationand disability pensioners with this dia-gnosis. Here we have drawn on figuresfrom the National Insurance Admini-stration for disbursements and number ofpersons in these categories for thediagnosis COPD. The figures refer to1993/1994. The effect of increasedsickness absence and rehabilitation on thenumber of hours worked in the economyis calculated by dividing the total dis-bursement by the average hourly wage.48

There is a risk of some understatementhere since the data from the NationalInsurance Administration may be incom-

48 Disbursements for sickness absence with thediagnosis COPD was Nkr 8.6 million in 1994, and forrehabilitation Nkr 47.6 million, according to datafrom the National Insurance Administration. How-ever, uncertainty attaches to the data. The averagehourly wage in 1994 was Nkr 135.1 per hour.


83

plete. Where sickness absence is con-cerned the statistics only cover absencebeyond the first 14 days. For disabilitypensioners the disbursement is at theoutset far lower than the average annualwage, and is also determined by factorssuch as responsibility for dependants etc.We therefore choose to multiply thenumber of disability pensioners by theaverage number of hours worked perperson in the age range 20-70 years. Thisfigure was 1 061 in 1993.49 We now havethe following relation between long-termconcentration of PM10 and the workforce:

(5.20) ∆L = -6.9*102 * ∆[i=∑

0

9

(PM10)-i /10] * P/TP

(sickness absence)

(5.21) ∆L = -3.8*103 * ∆[i=∑

0

9

(PM10)-i /10] * P/TP

(rehabilitation)

(5.22) ∆L = -6.9*104 * ∆[i=∑

0

9

(PM10)-i /10] * P/TP

(disability)

L denotes the number of person-hoursworked in the economy, while P/TPsignifies the share of the Norwegianpopulation that is exposed. (PM10)-i

denotes the mean annual concentration of

49 The total number of person-hours worked in theeconomy in 1993 was 2.87 billion (Statistics Norway1994c), while the number of persons between 20 and70 years totalled 2.70 million at year-end (StatisticsNorway 1995f). The number of disability pensionerswith the diagnosis COPD was 5 984.

PM10 i years ago, so the expression insquare brackets is a 10-year mean.

There is little risk of overlap between thefunctions in this section and the functionsin sections 5.2.1. The effects on the labourforce in this section do not include short-term absences (i.e., less than two weeks),whereas the functions in section 5.2.1 werebased on short-term studies and thereforedo not capture longer absences (i.e., morethan two weeks). Moreover, the impacts onpublic health expenditures in this sectionregards the effects of more people beingchronicly ill, whereas the interpretation ofthe effects in section 5.2.1 is presumablythat chronicly ill persons are being sent tohospital more often than without pollution.

5.2.3. Threshold values forparticulates

The question of threshold levels is animportant one when discussing therelevance of transferring dose-responsefunctions to Norway. Concentrations inNorwegian towns are somewhat lowcompared with American cities wheremost of the studies have been undertaken.Internationally however the trend ismoving more and more away fromspecific threshold levels in a population.The reason is that threshold levels are atbase specific to the individual, and so farit has not been possible to establish anyminimum value that is applicable to allindividuals. Moreover, significant relation-ships have been found between particulateconcentration and mortality at concen-trations well below previously establishedthreshold levels. This has inter aliaprompted an expert group attached to theEuropean Office of the World HealthOrganisation (WHO 1995) to state in anew report on guidelines for air qualitythat it will no longer recommend guide-lines for particulates. The WHO also write


84

that long-term effects appear to arise atvery low concentrations, specifically10 µg/m3 of PM2.5 (corresponding to about15 µg/m3 of PM10, which is well below thelevel in Oslo - see section 5.6).

Another interesting point is that the slopecoefficient in the dose-response functionsdoes not seem to diminish withdiminishing concentration. A simple plotof average concentration and coefficientestimates from the 9 mortality studiesreferred in Ostro (1993) (see section5.2.1) actually indicates a negative slope.Average concentrations in these studieswere in the area 37-80 µg/m3. Moreover,a fairly recent study by Dockery et al.(1992) found coefficient estimates 1.5times the consensus estimate in Ostro(1993). This was in areas featuring anaverage PM10 level of 28 and 30 µg/m3

respectively. This picture is in contrastwith the previous thought that theincidence of health damage is a convexfunction of concentration at low pollutionlevels (gradual threshold), and thereafterapproximately linear. Moreover, EC(1995) also points to indications that themarginal effect is greatest at low concen-trations, whereas WHO (1995) assertsthat in the case of particulates the effect isapproximately linear up to 200 µg/m3,and only then does the marginal effectrecede.

Dockery et al's (1993) study of mortalityresulting from long-term exposure foundmean annual concentrations of PM10

between 18 and 47 µg/m3. Hence townswith the lowest particulate pollutionshowed lower PM10 concentrations thanfor example Oslo. This means that if theresults of Dockery et al. can be generalisedsuccessfully, the dose-response function isclearly relevant for Norwegian towns.

In the study of restricted activity days(Ostro 1987) in section 5.2.1 the meanconcentration of PM10 was about43 µg/m3. and this average was based on84 urban areas with a wide range ofconcentration levels (probably alsocovering the level in Norwegian towns).

The mean concentration in Abbey et al'sstudy of chronic pulmonary diseases(1993) is not specified. However, 13 percent were stated to live in areas featuringconcentrations below about 33 µg/m3 ofPM10. This and other figures may suggestthat the mean concentration was rela-tively high (probably about 50 µg/m3).Collated with the WHO assessment above,namely that long-term effects also seem toarise at very low concentrations, we non-etheless find it relevant to utilise theresults from Abbey at al. (and otherstudies) for Norwegian towns.

5.3. Health effects of nitrogendioxide (NO2)

Nitrogen dioxide (NO2) is a gas formed bycombustion of fossil fuels. The principalsource in Norwegian towns is vehicletraffic. In some locations space heating,manufacturing or shipping may also beimportant emission sources.

Compared to particulates, relatively fewepidemiological studies have found signifi-cant relationships between outdoor NO2

concentrations and health effects. How-ever, as mentioned in the introduction,there are several reasons why the impor-tance of NO2 should not be toned down.First, numerous clinical studies havedemonstrated health effects. However,since such experiments differ in manyways from normal outdoor conditions it isdifficult to draw quantitative conclusions.Similarly, many studies have demon-strated relationships between indoor NO2


85

concentrations related to use of gasstoves, and various respiratory diseases.Here too, however, it is not clear to whatextent these relations can be converted foruse with outdoor concentrations. None ofthe applied studies we referred to earlierutilise dose-response functions from in-door NO2 concentrations. Only two of thestudies, ORNL/RFF (1994) and Ostro(1994), apply a function for outdoor con-centrations, and this is related to respira-tory symptoms. Certain studies, however,do find effects of NO2, but not of parti-culates. In this study we describe andapply a study from Finland related toasthma attacks.

Ostro (1994) mentions three possiblemethodological reasons why little has beenfound in the way of effects of NO2 concen-trations outdoors. These are 1) problems ofmeasuring NO2 concentrations outdoors, 2)effects of NO2 only arise at high concen-trations, and 3) chronic effects of NO2 areimportant, not acute. If point 2 is correctthen it is primarily brief exposures to parti-cularly high concentrations that have abearing on acute health effects, for instancealong a road carrying heavy traffic. In sucha case individual exposure is particularlyimportant, and this may be difficult tomeasure. Moreover, the resultant healtheffects will show a smaller degree ofcorrelation with daily mean concentrationswhich on the whole are used in epidemio-logical studies.

Appendix B includes a table summarisingthe dose-response functions taken fromthe literature, and a table showing howthese functions have been used in regardto economic activity etc.


Hospital admissionsA Finnish study by Pönkä (1991) found apositive association between hospitaladmissions due to asthma attacks andexposure to NO2. This study is particularlyinteresting for Norwegian conditionsbecause it was carried out in a neigh-bouring Nordic country in an area with acold climate. An increase in NO2 concen-tration from 28 to 45 µg/m3 was accom-panied by a 29 per cent increase inhospital admissions. We assume that thepercentage increase in admissions isidentical at all concentration levels (i.e. a1.5 per cent increase per unit increase inNO2 concentration), and that the relation-ship between admissions and number ofbed-days is constant. We can then derive alinear function based on Norwegianpatient statistics (Statistics Norway1995d) which approximates to the under-lying function. We find that for every1 000 head of population in Norway(Oslo) in 1993 there were 8.7 (9) bed-days with the diagnosis bronchial asthma.This gives the following function forchange in number of bed-days (the factorfor Oslo is in parentheses):

(5.23) ∆BD = 1.3*10-4 (1.4*10-4) * ∆NO2

* P

where BD denotes number of bed-days peryear and P signifies the number of personsexposed. NO2 denotes mean annual con-centration measured in µg/m3.

If we apply the cost per bed-day at Nor-wegian hospitals (see section 5.2.1) weobtain the following function for changein public health expenditure (the factorfor Oslo in parentheses):

(5.24) ∆PHE = 0.47 (0.48) * ∆NO2 * P


86

where PHE denotes health expenditure in1994 Nkr.

Compared with equivalent results forparticulates in section 5.2.1, whichapplied to all types of respiratory hospitaladmissions, this figure seems fairly high(NO2 concentration in Oslo is about twiceas high as PM10 concentration). A reasonfor this may be that the combination ofNO2 and low temperature has a largeimpact on health effects such as asthmaattacks. Rossi et al. (1993), anotherFinnish study, found a significant relationbetween NO2 concentration and asthmaattacks, even though the mean value wasas low as 13 µg/m3.

A German study by Schwartz et al. (1991)found a significant relation between NO2

concentration and the number of cases ofcroup. An increase from 10 to 70 µg/m3

was associated with an increase of 27 percent in incidence of this ailment,according to the National Institute ofPublic Health (1995). Applying the sameassumptions as above, this corresponds to0.4 per cent greater risk of croup per unitincrease in NO2 concentration. Accordingto the background material from StatisticsNorway (1995d) there were 5 323 bed-days with the diagnosis “acute laryngitisand tracheitis”, including croup. Thisimplies 1.23 bed-days per 1 000 head ofpopulation. In the same way and with thesame assumptions as above, we can nowcalculate the effect of NO2 pollution onthe number of bed-days and public healthexpenditure:

(5.25) ∆BD = 4.9*10-6 * ∆NO2 * P

(5.26) ∆PHE = 0.019 * ∆NO2 * P

where BD is the number of bed-days peryear, P is the number of persons exposed

and PHE is public health expenditure in1994 Nkr. NO2 denotes mean annual con-centration. We see that hospital costsattributable to increased incidence ofcroup are substantially lower than corre-sponding costs resulting from increasedasthma attacks.

A number of studies have demonstratedan association between indoor NO2 con-centrations and respiratory diseases.Hasselblad et al. (1992) carried out ametaanalysis of existing studies . Theyfound that an increase of 30 µg/m3 wasaccompanied by a 20 per cent increase indiseases of the lower respiratory tract.Whether or not it is appropriate to convertthese results to functions for outdoorconcentrations is not clear. Until more isknown about the relative importance ofoutdoor contra indoor concentrations, it isdifficult to make use of these results.

Respiratory symptomsORNL/RFF (1994) and Ostro (1994) haveused results from Schwartz and Zeger(1990) for the relation between NO2

concentration and frequency of phlegmproduction. In Ostro the result isconverted to the following function:

(5.27) ∆RS = 5.1*10-3 * ∆NO2 * P

RS denotes the number of days withrespiratory symptoms per year, while P isthe number of persons exposed. In thisfunction NO2 signifies the average hourlymaximum (daily), measured in µg/m3.According to the Norwegian Institute forAir Research the relationship between thehourly maximum and daily mean variesfrom 1.2 to 2.8 in Oslo, with 1.8 as the‘best’ estimate. Some uncertainty attachesto whether this relationship variessystematically with pollution levels. If not,there is reason to believe that the mean


87

daily concentration can be used as anindicator for the one-hour maximum. Ifwe assume this to be the case, and applythe conversion factor 1.8, we obtain thefollowing function:

(5.28) ∆RS = 9.2*10-3 * ∆NO2 * P

where NO2 now denotes the mean annualconcentration (the transition from dailymean to annual mean is assumed to applyas previously).

A shortcoming of the study by Schwartzand Zeger (1990) is that particulates werenot controlled for. Hence there is a riskthat this result may overlap the result insection 5.2.1. However, other studiessupport the relation between outdoor NO2

concentration and respiratory diseases. ASwiss study by Braun-Fahrländer et al.(1992) found for instance that theduration of respiratory symptoms inchildren rose with increasing exposure toNO2, even when particulates were takeninto account. As mentioned in thepreceding section, a Finnish study byRossi et al. (1993) found significantrelations between NO2 concentration andasthma attacks. Since it is currentlydifficult to convert these results to a dose-response function we let them support theabove function instead.

What this entails for labour productivity issomewhat unclear. Ostro (1994) is notconcerned with valuation in his analysis,while ORNL/RFF (1994) omit to valuephlegm production for lack of relevantstudies. We choose to follow the sameprocedure as in section 5.2.1, i.e. toestimate the productivity loss resultingfrom one day of respiratory symptoms at6 per cent (based on Eskeland (1995)).The relative change in the supply of

effective labour resulting from increasedNO2 concentration is then calculated to:

(5.29) ∆L / L = - 1.5*10-6 * ∆NO2 * P/TP

L is the number of person-hours workedper year in the entire economy, and P/TPsignifies the share of the Norwegianpopulation that is exposed. Although themean annual concentration of NO2 isnormally higher than for PM10, the effectof NO2 here is clearly lower than found forparticulates in section 5.2.1.

5.3.2. Threshold levels for NO2

The question of threshold levels is moreproblematic for NO2 than for particulates.The comparative lack of studies to turn tomakes it difficult to draw conclusions.WHO (1995) consequently continue torecommend air quality guidelines for NO2.Their guideline annual mean is 40 to 50µg/m3. By way of comparison theNorwegian Pollution Control Authority(SFT 1992) employs a six-month mean of50 µg/m3. This is slightly above the meanannual concentration in larger Norwegiantowns. The guidelines are not intended tobe absolute threshold levels. A certainsafety margin is included at the same timeas no guarantee is given that particularlysensitive individuals will not experiencehealth effects at concentrations below thethreshold levels. It must also be kept inmind that mean annual concentrationsvary within the individual town, and thatthe short-term guidelines may be sub-stantially exceeded.

A pertinent question is whether the airquality criteria for NO2 will in time betightened in the same way as for parti-culates. The criteria for particulates havebeen adjusted downwards in the light ofincreased knowledge, and if method-ological problems explain why there are


88

currently fewer studies to turn to for NO2,it is not improbable that in time thecriteria for NO2 will also be tightened.

The results from two Finnish studies,Pönkä (1991) and Rossi et al. (1993)which were discussed above, are worthnoting. In the first-mentioned study asignificant relation was found betweenNO2 concentration and hospitaladmissions due to asthma attacks atconcentrations in the interval 28-45µg/m3. This is well below the guidelinesfrom the WHO and the NorwegianPollution Control Authority. Rossi et al.(1993) found a significant relationbetween NO2 concentration and asthmaattacks at a mean concentration as low as13 µg/m3. This is less than a third of themean annual concentration in Oslo, andcorresponds to the background level insouth-east Norway. An interesting featureshared by these two studies is that theywere undertaken in a cold climate(temperature was of course taken intoaccount in the regressions). They oughttherefore to be particularly relevant forNorwegian conditions, and indicate thatthe dose-response functions may applywell below the guidelines recommendedby the WHO and the Norwegian PollutionControl Authority.

5.4. Health effects of sulphur dioxide(SO2)

Sulphur dioxide (SO2) is formed bycombustion of oil and coal, and theemissions are due to the sulphur contentof these fuels. SO2 has historically beenviewed as an important pollutant in thecontext of health effects. However, thesignificance of SO2 has been toned downin the past decade. There are two impor-tant reasons for this. First, the importanceof SO2 is believed to have been somewhatoverstated in relation to other pollutants,

in the first instance particulates. Second,with the substantial reduction in SO2

concentrations in most areas the healthproblem is now mainly confined tospecific local areas with heavy sulphuremissions from manufacturing industry(SFT 1992). However, SO2 may have asubstantial health effect in its own right insuch locations. SO2 emissions also lead tothe formation of sulphates belonging inthe group of small particulates (PM2.5).

Although the effect of SO2 disappears inmany studies when particulates arecorrectly measured (EC 1995), somestudies continue to find health effects ofSO2 when particulates are included(ORNL/RFF 1994). Below we presentsome dose-response functions described inthe literature. It is however unclear towhat extent these effects can be includedin addition to the effects of particulates.


MortalityORNL/RFF (1994) and Ostro (1994)describe several European studies thathave found significant links betweenconcentration of SO2 and mortality.Common to all of them is that they wereundertaken in areas with relatively highconcentrations, all exceeding an averageof 60 µg/m3. This is considerably higherthan the mean annual concentration inOslo. Ostro constructs a dose-responsefunction based on a Greek study byHatzakis et al. (1986):

(5.30) ∆D / D = 4.8*10-4 * ∆SO2

where D denotes number of deaths peryear, while SO2 denotes an annual meanmeasured in µg/m3.


89

Several of the studies described by Ostrofind no relation between particulates andmortality. This confirms the supposition inthe introduction that particulates (orother pollutants) are primarily indicatorsof a complex pollution situation. Hence itwould be advisable not to add togetherthe mortality effects of particulates andSO2.

Respiratory symptomsORNL/RFF (1994) and Ostro (1994) alsogive an account of studies of various typesof respiratory symptoms. One of them(Schwartz et al. 1988) finds significantrelations between concentration of SO2

and frequency of chest pains. Ostroderives the following dose-responsefunction:

(5.31) ∆RS = 1.0*10-2 * ∆SO2 * P

where RS denotes the number of dayswith respiratory symptoms per year, whileP is the number of persons exposed. SO2

denotes mean annual concentration.

In the same way as in sections 5.2.1 wecan derive the effect on the workforce:

(5.32) ∆L / L = - 1.64*10-6 * ∆SO2 * P/TP

where L denotes number of person-hoursworked and P/TP the exposed share of thepopulation.

5.4.2. Threshold levels for SO2

Studies that have found significantassociations between SO2 concentrationand health effects have been carried out inareas with high concentrations. Againstthis background the WHO (1995) recom-mends a mean annual guideline of 50µg/m3, which is substantially higher thanthe level in Oslo. The Norwegian PollutionControl Authority (SFT 1992) employs a

six-month mean of 40 µg/m3. Against thisbackground it is uncertain to what degreethe dose-response functions for SO2 arerelevant in Norwegian towns (with theexception of particular areas). It ishowever conceivable that health damagemay also be associated with SO2 inepisodes of high SO2 concentration.

5.5. Health effects of ozone (O3)Ozone (O3) is a so-called secondary airpollution component formed by reactionsbetween NOx, methane and VOCs (volatileorganic compounds). As mentioned insection 5.1 ozone is the component re-ceiving the greatest international attentionalongside particulates. In Norway, how-ever, ozone concentration is very largelydetermined by long-range pollution fromelsewhere in Europe. Increased emissionsof NOx in Norwegian towns lead for themost part to somewhat lower local ozoneconcentrations, but to somewhat higherconcentrations at regional level and insuburban areas.50 However, the relation-ship between local and regional effects arenot well understood yet. Furthermore, theoverall indication is that economic activityin Norway is of little significance forhealth effects caused by ozone. Thus,discussion of ozone is limited.

According to EC (1995) the effects ofozone are regarded as additive in relationto particulates, NO2 and SO2. This isbecause correlations between ozone andthe other components are generally weak.

50 A report from The Norwegian MeteorologicalInstitute, Simpson et al. (1996), calculates theregional effects on ozone concentrations of Norwegianemissions of NOx and VOCs. It finds that whileNorwegian NOx emissions have a certain regionaleffect on ozone concentrations, Norwegian VOCemissions have little regional effect.


90


MortalityEC (1995) refers to a study by Kinney etal. (1994) which found the followingsignificant relation between ozone con-centration and mortality:

(5.33) ∆D / D = 7.5*10-5 * ∆O3

where D denotes the number of deathsper year, and O3 is the average daily one-hour maximum measured in µg/m3. Roweet al. (1995) refer to two studies whichfind approximately the same risk increase.Rowe et al. choose however to halve thecoefficient since several studies havefound a weaker or no relationship bet-ween ozone concentrations and mortality.Moreover, a pertinent question is whetherthese studies, which were all carried outin the distinctive Los Angeles area, arerelevant for Norwegian conditions.

Reduced labour productivityTwo studies by, respectively, Ostro andRothschild (1989) and Portney andMullahy (1986) have found relationsbetween concentration of ozone andminor restricted activity days (MRAD).The latter study (described in Ostro(1994)) found an effect that was twice aslarge as that found in the first-mentionedstudy (recounted in EC (1995) and Roweet al. (1995)). We present a dose-responsefunction with a coefficient estimate equalto the average of the two studies:

(5.34) ∆MRAD = 0.012 * ∆O3 * P

MRAD denotes the number of minorrestricted activity days per year, and P thenumber of persons exposed. O3 signifiesthe annual average of daily one-hourmaxima, measured in µg/m3.

Using the procedure described in section5.2.1, we obtain the following function forthe effect on the workforce:

(5.35) ∆L / L = - 3.4*10-6 * ∆O3 * P/TP

L is the number of person-hours worked inthe economy per year, while P/TP denotesthe exposed portion of the population.

Respiratory symptomsIn section 5.2.1 reference was made to astudy by Krupnick et al. (1990) whichfound a significant relation between con-centration of particulates and respiratorysymptoms. The same study found a similarrelation for ozone concentration. EC(1995) sets out the following function:

(5.36) ∆RS = 0.026 * ∆O3 * P

where RS denotes the number of dayswith respiratory symptoms per year, and Pdenotes the number of persons exposed.O3 signifies the annual average of dailyone-hour maxima, measured in µg/m3.Several other studies have found effects ofozone on respiratory symptoms, especiallyasthma.

Following the same procedure as insection 5.2.1, including deduction ofMRAD as in the previous section, we findthe following effect on the workforce:

(5.37) ∆L / L = - 2.3*10-6 * ∆O3 * P/TP

L is the number of person-hours worked inthe economy per year, while P/TP denotesthe exposed portion of the population.

Hospital admissionsSeveral studies have found significantrelations between ozone concentrationand hospital admissions. Ostro (1994)gives an account of a study by Thurston et


91

al. (1992) which gives the following dose-response function:

(5.38) ∆RHA = 3.9*10-6 * ∆O3 * P

where RHA denotes the number of res-piratory hospital admissions per year, Pthe number of exposed persons, while O3

remains the annual average of daily one-hour maxima. EC (1995) describes severalstudies that have found significant rela-tions between ozone concentration andhospital admissions resulting from varioustypes of respiratory diseases. Thesestudies indicate that the function inequation (5.38) may be an underestimate.

Based on the above function we constructthe following function for costs in thepublic health system (same procedure asin section 5.2.1 with figures for Oslo inparenthesis):

(5.39) ∆PHE = 0.08 (0.12) * ∆O3 * P

where PHE denotes public health expen-diture measured in 1994 Nkr.

5.5.2. Threshold levels for ozoneThe WHO (1995) recommends an eight-hour guideline of 120 µg/m3 for ozone,but writes that this guideline gives noguarantee against all acute health effectsof ozone. In comparison the NorwegianPollution Control Authority (SFT 1992)employs an equivalent guideline of80 µg/m3. According to the SFT, the airquality criteria for ozone (one-hour mean)are exceeded about 3 per cent of the timein southern Norway and 1 per cent of thetime in northern Norway. Hence there isreason to believe that the dose-responsefunctions found for ozone are alsorelevant for Norway.

5.6. Population exposure to airpollution in Oslo

The Norwegian Institute for Air Research(NILU) has elaborated a detailed dispersionmodel, called EPISODE, for air pollution inOslo. Given variation in emissions fromdifferent sources, it determines populationexposure to inter alia particulates and NO2

in squares of 1 km2. NILU was commis-sioned by Statistics Norway to use thismodel to calculate a function for thepopulation-weighted mean annualconcentration of PM10 and NO2 in Oslo(Walker 1997). The function depends onchanges in local emissions from vehicletraffic, other local emissions (largely spaceheating) and background concentrationdue to foreign and Norwegian emissionsoutside Oslo. Mean annual concentration iscomputed for each square and is thenweighted with the population in the square.This provides a detailed description ofpopulation exposure in Oslo compared withsimilar international studies. By way ofcomparison EC (1995) employ squares of100 km2, while Ostro (1994) uses squaresof 25 km2 for Jakarta and calls it asubstantial improvement on other studies.

In view of the fact that a large portion ofOslo's population works in the city centre(where the pollution is heaviest) and liveson the outskirts, a possible objection tothe method is that it partially underesti-mates the actual population exposure.Moreover, dispersion of air pollution is acomplicated process and considerableuncertainty attaches to the functions.

Walker (1997) presents the followingfunction for population-weighted meanannual concentration of PM2.5 in Oslo:

(5.40) PM2.5 = 3.9 * IT + 5.3 * IO

+ 5.9 * IB


92

Ij denotes indexes (Ij=1 in 1992) forrespectively vehicle traffic emissions (T),other emissions (O) and backgroundconcentration (B) of PM2.5. From equation(5.40) we see that in 1992 mean concen-tration in Oslo was calculated at 15.1µg/m3, and 61 per cent of this concen-tration was due to local emissions withinthe municipality. Walker makes clear thatsince it is not possible in such calculationsto take account of the high concentrationsalong roads carrying heavy traffic, theabove function will underestimate theeffect of vehicle traffic emissions.

Walker (1997) presents scaling factorsbetween the PM10 and PM2.5 concen-trations for the summer and winter half-year respectively. These are based onobservations at measuring stations inOslo. For vehicle traffic emissions there isa considerable seasonal difference wherethis scaling factor is concerned, i.e. 3.0and 1.2 respectively for the winter andsummer half-year, according to Walker.The wide seasonal variation is due to useof studded tires which whirl up asphaltdust consisting partly of PM10. Since thiscontribution primarily depends on thenumber of kilometres driven, and not onexhaust emissions, we opt for a scalingfactor between PM10 and PM2.5 of 1.2 forthe whole year, and include an index forkilometres driven. This is particularlyimportant if it is wished to split the effectof exhaust emissions from the effect ofwear and tear by studded tires. For otheremissions the factor is 1.1 for the wholeyear, according to Walker. The back-ground concentration of PM10 is stated tobe 8.7 µg/m3. This gives the followingfunction for PM10 concentration in Oslowhere the indexes IT and IO refer toemissions of PM10, while IKM is the index

for distance driven (1992 remains thebasis year):51

(5.41) PM10 = 4.7 * IT + 5.8 * IO+ 4.0 * IKM + 8.7

Hence in 1992 the mean annual concen-tration in Oslo was 23.2 µg/m3, of which14.5 µg/m3 (63 per cent) was due to localcontributions in Oslo. Of the local contri-bution vehicle traffic accounts for 60 percent, while other emissions account for40 per cent. The contribution of vehicletraffic can be further divided into 54 percent from exhaust emissions and 46 percent resulting from the use of studdedtires. The latter distribution must howeverbe viewed with caution inasmuch as it isestimated by the author and not byWalker (1997).

In contrast to particulates, NO2 concen-tration is not a linear function of localemissions. The concentration is deter-mined by atmospheric reactions betweenlocal emissions of NOx and the regionalcontribution of NOx and O3 (backgroundcontribution). Walker (1997) has cal-culated the population-weighted meanannual concentration of NO2 for variationsin vehicle traffic emissions, otheremissions and background contribution.52

This gives a calculated average meanannual NO2 concentration of 46.5 µg/m3

in Oslo in 1992. A 40 per cent reductionin local emissions reduces concentrationby 7.7 µg/m3, i.e. 17 per cent, while acorresponding reduction of the regionalcontribution reduces concentration by all

51 Total contribution from road traffic is stated byWalker to be 8.7 µg/m3. Of this, 1.2*3.9=4.7 µg/m3

is assumed to stem from exhaust emissions, while theremainder (4.0 µg/m3) is assumed to stem fromasphalt dust.52 This is done for indexes (equal to 1 in 1992) whichvary stepwise from 0.6 to 1.4 with intervals of 0.1.


93

of 30 per cent. What the effect of biggerreductions might be is unclear. In regardto changes within the +/- 30 per centrange a linear function is found which isapproximately identical to the discreetfunction calculated by Walker (1997):53

(5.42) NO2 = 14.6 * IT + 2.1 * IO

+ 32.3 * IB - 3.1

The indexes here refer to NOx emissions.The large difference in the first twoweights is partly related to the fact thatNOx emissions from vehicle traffic in Osloin 1992 were 3-4 times bigger than thesum total of other emissions. For NO2 aswell, Walker makes it clear that thecalculations understate the relative effectof vehicle traffic emissions which areclearly the dominating local contributor toNO2 pollution in Oslo (87 per cent of thelocal contribution on a marginal basis).It is uncertain how large a portion of thebackground contribution is due to Nor-wegian emissions from outside Oslo.Hence in this calculation we confineourselves to local emissions in Oslo.

5.7. Public health effects and socialcosts of air pollution in Oslo

5.7.1. Annual health effectsIn this section we present calculations ofhealth effects of air pollution in Oslo,based on the functions described earlier inthe study. As mentioned in section 5.1,there are problems involved in trans-ferring such relations from one area toanother where people and physicalsurroundings are different. However, thishas become common practice in manystudies, such as EC (1995) for Europe andORNL/RFF (1994) and EPA (1995) forthe USA. By virtue of their compass and

53 R2>0.99

the bodies that commissioned them, thesestudies carry much weight internationally,and for this reason the present study is tosome extent based on their choice offunctions. Moreover, two studies from theWorld Bank, Ostro (1994) and Eskeland(1995), have used dose-response func-tions to calculate health effects of airpollution in respectively Jakarta (Indo-nesia) and Santiago (Chile). They discussthe problem of transferring functions todeveloping countries and consider theprocedure justifiable. Compared withthese two studies the calculations presen-ted in the present chapter must be said topossess greater credibility.

It is worth noting that the ‘predecessor’ toORNL/RFF (1994) and EC (1995), viz.PACE (1990), did not make use of dose-response functions. Instead it used theresult from a study by ECO (1987) whichfound a relation between increasedemissions of particulates and higher riskof mortality in urban and sparselypopulated areas respectively. Such anapproach is naturally somewhat lessprecise than using the dose-responsefunctions described above so long as onehas serviceable models to calculate therelation between emissions and con-centration levels. Hence the results in thenew generation of applications must besaid to be considerably more reliable thanwas previously the case.

In the previous section it emerged thatlocal emissions in Oslo contributed 14.5µg/m3 (population weighted) to theconcentration of PM10 in 1992. We nowapply this contribution to the dose-response functions set out in section 5.2and calculate the total annual healtheffects of local air pollution in Oslo. Thisentails that we are not looking at theeffect of the regional contribution to the


94

level of pollution in Oslo. The calculationsrefer to a typical meteorological year withemission data for 1992, but sincepollution levels have been relatively stablein recent years, the results are also rele-vant for the situation today. We wouldpoint out that uncertainty attaches to thecalculations and that the results must beviewed as indications (cf. the discussionsin section 5.2). Similar calculations havebeen made for example by the WHO(1995) for a notional town. The results forOslo are presented in table 5.1 (roundedoff to the nearest one or two significantdigits).54

The results indicate that in about 90 fata-lities in Oslo, equivalent to 1.4 per cent ofthe total, death was hastened by episodesof high local air pollution. As mentionedearlier the size of the lifetime reduction isunclear. By way of comparison, Pearce

54 Data on for example population, deaths andemployment (Statistics Norway 1994c,1995f) are usedin the calculations. Some of the data are given for1993 rather than 1992. Further, a full person-year isassumed to contain about 1 700 person-hours.

and Crowards (1996) calculates that7 000 persons die each year in Englandand Wales as a result of episodes ofparticulate concentration. This calculationwas made for the urban population whichin his calculation accounts for 44 per cent.Bown (1994) makes a similar calculationbut employs a higher figure for urbanpopulation (68 per cent). Her result is10 000 deaths each year.

The calculations also indicate that eachyear about 400 new persons arediagnosed as suffering from chronicpulmonary disease resulting frompolluting emissions in Oslo. This is on thereasonable assumption that the level ofconcentration in 1992 is representative ofprevious years. In the longer term it alsoprobably causes an increase of about 100persons (not an additional 100 each year)in the number of recipients of disabilitybenefit. This corresponds to an annualdisappearance of about 70 person-yearsfrom the labour market. In addition, afurther 330 person-years may disappear inOslo as a result of reduced productivity

Table 5.1. Total annual health effects and impacts on economic activity of local particulate pollution in Oslo, 1992

Annual effects

Acute effects:Number of premature deaths 90

Person-years lost- reduced productivity 70- short-term sickness absence 260

Bed-days in hospital (respiratory ailments) 2 100Public hospital expenditure (mill. 1994-Nkr) 7

Chronic effects:*New cases of chronic pulmonary disease 400

Person-years lost- long-term sickness absence/rehabilitation/disability 70

Bed-days in hospital (respiratory ailments) 2 800Public hospital expenditure (mill. 1994-Nkr) 11

Reduced life expectancy (years) - highly uncertain 0.9

* Calculations of chronic effects presuppose that local particulate pollution has been stable for several years


95

and short-term sickness absence. Theperson-years lost owing to short-termsickness absence are equivalent to bet-ween 5 and 10 per cent of total sicknessabsence lasting less than 14 days in Oslo.

Moreover, particulate emissions in Oslo(including asphalt dust) cause about5 000 hospital bed-days per year, i.e.about 12 per cent of all bed-days relatedto respiratory diseases. This correspondsto expenditure of about Nkr 18 million.Almost half the bed-days are due to acutepollution episodes.

If we calculate the effect on mortality oflong-term particulate pollution, we findthat life expectancy for women and men,respectively, in Oslo is about 0.86 and0.93 years less than it would have been inthe absence of local pollution. It is notclear whether this is a uniform effectacross the population or whether it forexample entails that 10 per cent of thepopulation has its life expectancy reducedby about 9 years. Another consequence isthe further loss of 200 person-years peryear in Oslo for people who die beforereaching the age of 65. We shouldhowever reiterate the clarification by theWHO (1995) that further cohort studiesare needed in order to draw firm con-clusions. Calculations of the reduction inlife expectancy, and the effects of suchreduction, must therefore be regarded asrough and ready indications.

Section 5.6 showed the difficulty of estab-lishing the size of the increase in NO2-concentration caused by total local NOx

emissions. We found that a 40 per centreduction in emissions would entail areduction of 7.7 µg/m3 in NO2 concen-tration. If, for simplicity's sake, we extra-polate this to a reduction of 100 per cent,we find that the local contribution to the

concentration is 19.25 µg/m3. This isprobably an underestimate inasmuch asthe function seems to be somewhat con-cave. We nonetheless apply this concen-tration contribution to the dose-responsefunctions in section 5.3. The uncertaintyhere is just as great as for particulates.The results are shown in table 5.2.

We note that the effects of NO2 pollutionare apparently far smaller than corre-sponding effects of particulates. Asalready mentioned, it is not clear to whatdegree this is due to measuring problems.According to the results the greatesthealth effect of NO2 pollution is hospitaladmissions as a result of asthma attacks.This leads to more than 1 000 bed-days athospital in Oslo each year, and costs thegovernment more than Nkr 4 million. Inaddition there are person-hours lost dueto absence from work, either because ofparents' or children's hospital admissions.These are not included in the calculations.NO2 pollution also leads to reduced labourproductivity, quantified to the equivalentof about 5 person-years.

5.7.2. Social costs of air pollutionWhat, then, are the social costs of thehealth effects discussed above? The costsmay be divided into pure economic costsand costs due to reduced quality of life.The purely economic costs refer primarilyto the direct value of lost person-years and

Table 5.2. Total annual health effects and impacts on economic activity of local NO2

pollution in Oslo, 1992

Annualeffects

Acute effects:Bed-days in hospital (respiratory ailments) 1 300Public hospital expenditure (mill. 1994-Nkr) 4.5Person-years lost - reduced productivity 5


96

of public hospital expenditure. The valueof lost person-years can either be cal-culated with the aid of average hourlywages or average hourly labour costs(including employers' social securitycontributions). Hourly labour costs areprobably the most relevant since theyreflect the value of person-years forenterprises. In 1994 this value wasNkr 178.5 per hour while the averagehourly wage was Nkr 135.1.55

In addition to the direct costs there areindirect effects on the economy through theworkforce becoming a scarcer resource. Byemploying the general equilibrium modelMSG-EE for the Norwegian economy (seechapter 2), we can calculate the total costsof the health effects. These are defined asthe reduction in GDP plus the increase inpublic health expenditure. The model hasbeen expanded to include the relationspresented in this study. Thus, increasedpollution leads to reduced employment andhigher public expenditure. The calculationscapture secondary effects in the economydue to a scarcer supply of labour. Thiscauses changes in the inputs compositionthroughout the economy. Hence industrystructure also undergoes change. Animportant effect is that a reduced supply oflabour reduces production of capital goodsso that growth in real capital slows down.Furthermore, higher public expenditureentails some withdrawal of resources fromthe private sector, which is assumed to bemore productive than the public sector.These allocation effects are in addition tothe direct productive costs and are oftenomitted in similar cost calculations (e.g EC(1995) and ORNL/RFF (1994)).

55 Since we are interested in the health costs in a yearwith a typical pollution situation, and specify costs in1994-kroner, we use figures for 1994 instead of 1992.

However, the most important health cost ofair pollution is associated with thereduction in the quality of life resultingfrom higher risk of mortality and increasedincidence of disease. Hence the total costlargely depends on how quality-of-lifereductions are valued. This is a difficultquestion to which no objective answer canbe postulated. Hence the physical effectsshould be allowed to speak for themselveswhen such information is sufficient. Some-times, however, it would be preferable tomake calculations based on concretevaluations to exemplify how large the costsmay be. In our study this is done for thetwo major health effects, i.e., increasedmortality and chronic pulmonary disease.The valuations are based on variouswillingness to pay (WTP) studies.

For mortality we employ the estimateused by the Norwegian public admini-stration to value a statistical life. This wasdevised by the Institute of TransportEconomics (ITE) in the context of roadaccidents, and is equivalent to 10.5million 1993-kroner (Elvik 1993). It isopen to question whether this is relevantfor deaths resulting from pollution inas-much as the residual life expectancy maybe lower than the residual life expectancyin the case of road accidents. The estimatefrom ITE only expresses the welfare effectand is therefore relatively low comparedwith corresponding internationalestimates. For instance Pearce (1995)employs an estimate of GBP 1.5 million(i.e. about Nkr 15 million) and calls itconservative, while EPA (1995) uses anaverage estimate of USD 3.5 million (i.e.about Nkr 25 million).56 Both these studiesfocus on deaths hastened by episodes of

56 This is based on estimates of $3.4 million forpersons over 65 years and estimates of $4.5 millionfor persons below 65 years.


97

air pollution. Another pertinent questionis whether the value of a statistical life isunderestimated when certain groups ofthe population (the elderly and sick) aremore at risk than others. We do not wishourselves to decide the value of a statis-tical life in this study, and point out thatour choice is merely intended to beillustrative.

The same reservation may be applied toour choice of valuation in regard to thedevelopment of chronic pulmonarydisease. Here we opted for an estimateused in the same connection by Rowe etal. (1995), which builds on study ofwillingness to pay (WTP). Their value isUSD 210 000 (i.e. about Nkr 1.5 million).In comparison, the EPA (1995) chooses avalue of USD 240 000, which is also basedon WTP. It may be objected that theseestimates overlap with the pure economiccosts related to more disabilities andincreased public expenditures. Thisdepends on to what degree the WTPfigure includes the consideration ofreduced income in addition to reducedquality of life. As the estimates areobtained from American studies, wherecompensation is lower than in Norway,there is reason to believe that the incomesituation is taken into account. However,the pure economic costs in our resultsonly cover about 10 per cent of the WTP,so the overlap is relatively low.

Given these assumptions the total healthcost of air pollution in Oslo can beestimated. We have summated costslinked to the effects set out in table 5.1and table 5.2. The results are shown intable 5.3. As mentioned, costs linked tomortality and chronic pulmonary diseasesare to be regarded as illustrative. Sincethe value of reduced quality of life due toincreased morbidity is not fully captured,

the results are not intended to cover allhealth costs, or indeed other environ-mental costs.

Calculations performed using the MSG-EEmodel indicate that the total pureeconomic costs are about Nkr 163 million.At the same time the direct economiccosts may be calculated to about Nkr 147million, so that the allocation costs areNkr 16 million. This means that theindirect effect of a reduced supply oflabour and increased public expenditureamounts to about 11 per cent of the directcosts.57 In fact the immediate effectrelated to structural changes in theeconomy places a slight damper on thedirect cost. This is due to shifts in theeconomy from sectors with low produc-tivity to sectors with high productivity.The more long-term effect, on the otherhand, which is related to reduced rates ofinvestment, contributes to raising the totalcost. These costs are discounted at a rateof 7 per cent.

As seen from table 5.3, the total socialcosts of Nkr 1.7 billion depend heavily onhow increased mortality and increasedincidence of chronic pulmonary diseasesare valued. With the valuation estimateswe have employed, the purely economiccosts are only around 10 per cent of thetotal costs. If we had applied the highestvaluation estimates referred to above, thetotal costs would have been more thanNkr 3 billion. Thus, there are major uncer-tainties both with regard to the annualhealth effects and the valuations of theseeffects. The most important pollutant inour calculation is clearly particulates, as

57 Had we valued the direct cost of lost person-yearsby using average hourly wages, the direct costs wouldhave been 115 million Nkr, so that the allocation costwould have been 48 million Nkr, i.e., 42 per cent ofthe direct costs.


98

less than 0.5 per cent of the total costsstem from health effects of NO2 pollution.This is mainly because the quality of lifereductions are associated with healtheffects resulting from particulatepollution.

Based on the discussion in section 5.6 weknow that 60 per cent of PM10 concen-tration stems from vehicle traffic, and thatalmost half the contribution by vehicletraffic is due to the use of studded tires.Hence these shares can be used to cal-culate directly the health effects and socialcosts attributable to various particulatesources. For example, annual social costsassociated with use of studded tiresamount to about Nkr 480 million. How-ever, it is necessary to discuss whetherPM10 really is the particulate to focus on.As mentioned in section 5.2, there is apossibility that PM2.5 is a better indicator,and that the gravest health effectsobserved in connection with PM10

pollution may primarily be associated withPM2.5. If this is the case, use of studdedtires is less hazardous than suggestedabove, even though it probably alsocontribute somewhat to the PM2,5 concen-tration. On the other hand, if PM10 con-centration is less harmful than indicatedhere, PM2.5 concentration is more harmful

such that the total cost is about the same.This is due to the relationship betweenPM2.5 and PM10 which is about the same inOslo as in most towns where epidemio-logical studies have been carried out.

We have also calculated the marginal costof increasing emissions of particulates(PM10) by one tonne in Oslo in 1995.58

This increases concentration in 1995 andentails a reduction in the supply of labouretc., in 1995 and to some extent in theyears ahead (owing to the long-term effectof pollution). Here too the contraction ofGDP for the years after 1995 is discountedat a rate of 7 per cent per year. The samediscount rate applies to the valuation ofnew chronic pulmonary diseases arisingafter a period of several years. Table 5.4shows the marginal cost. It also shows themarginal cost per litre of petrol and dieselbased on figures for average emissions perlitre for the entire vehicle population. Wenote that costs related to use of diesel areparticularly high. This is because dieselengines emit on average more then 10

58 Owing to the large uncertainty attached to healtheffects of NO2 exposure, we have chosen not tocalculate any marginal cost for NOx emissions. Such acalculation would probably greatly underestimate thereal marginal cost.

Table 5.3. Total annual social costs associated with health effects of local air pollution in Oslo, 1992.Mill. 1994-Nkr

Total costs Share of total costsPer cent

Person-years lost 125 7.2Public expenditure 22 1.3Allocation costs 16 0.9

Total (pure) economic costs 163 9.4

Increased mortality 936 54.2Extra chronic pulmonary diseases 629 36.4

Total costs of reduced quality of life 1 565 90.6

Total social costs 1 728 100


99

times as many particulates per litre of fuelas petrol engines. In addition we havecalculated the social cost of using studdedtires, due to asphalt dust. This is based ondata on annual traffic volume in Oslo(Vegdirektoratet 1996) and the assump-tion that 75 per cent of vehicles in Oslouse studded tires in the winter half-year.The average cost of using studded tires isthen about Nkr 0.40 per kilometre, whilecosts for light vehicles are somewhatlower and for heavy vehicles somewhathigher.

A reservation is made here about theimportance of PM10 versus PM2.5. If PM2.5 isthe pollutant to focus on, the cost perkilometre will be lower, while the cost perlitre of petrol and diesel will be higher. Itis also worth mentioning that if parti-culates are above all an indicator of apollution situation where componentssuch as NO2, SO2 etc., are more or lessequally harmful to health (see discussionsearlier in this chapter), the cost estimatesfor petrol and diesel will tend to converge.Moreover, the same applies if the PM10

concentration is partly caused byemissions of other components like NOx orSO2 (in the form of nitrate and sulphateparticles). However, according to theNorwegian Institute for Air Research, thisis not of importance in Norwegian towns.

Here too the marginal cost does not coverall social costs of health damages, norindeed other environmental costs. If othervaluation estimates are desired, e.g. formortality, the shares shown in table 5.3may be used to calculate marginal costs.We find for example that pure economiccosts make up about 10 per cent of thetotal, i.e. about Nkr 190 per kilogram ofPM10.

We have so far disregarded costsassociated with increased mortalityresulting from long-term particulatepollution. As mentioned this is due to thevery great uncertainty attached to theseestimates. Moreover, it is difficult, if notimpossible, to value reductions in lifetimeexpectancy. We have however studiedwhat effect this may have on the economyin terms of people in the workforce dyingbefore retirement age. The marginal costper kilogram of PM10 in terms of purelyeconomic effects increases in this case byabout Nkr 30, i.e. by 15 per cent. How-ever, the sickness period prior to death,which may well be of even greater import-ance, is not included here.

5.8. ConclusionThis study has described a number ofassociations between air pollution andvarious health effects as demonstrated byinternational studies. We have also shownhow these effects may impact on econo-mic activity via changes in sicknessabsence and public health expenditure.We closed with calculations of healtheffects in Oslo and endeavoured toquantify social costs associated with sucheffects.

Table 5.4. Social health costs associated with particulate pollution in 1995. 1994-Nkr

Marginal cost

PM10 2 020 per kg

Petrol- light vehicles 0.54 per litre- medium vehicles 0.54 per litre- heavy vehicles 0.15 per litre

Diesel- light vehicles 6.6 per litre- medium vehicles 7.3 per litre- heavy vehicles 4.0 per litre

Use of studded tires 0.41 per km


100

In recent years international literature haspresented an increasing number ofepidemiological studies of the relationsbetween air pollution and various healtheffects. The relation between short-termchanges in particulate concentration andthe number of deaths in the population isespecially well documented, and asurprisingly high degree of concordance isfound between the results observed invarious towns. Fewer studies are currentlyavailable for other health effects, butresults so far give a good indication ofwhat may be expected in the way ofhealth damage at various pollution levels.An interesting feature of several recentstudies is that associations betweenparticulate pollution and health effectsappear to arise at relatively low concen-trations. This has prompted an expertgroup at the WHO (1995) to cease recom-mending guidelines for particulates.Hence the results observed in other townsare probably also relevant for Norwegiantowns.

The calculations for Oslo illustrate that airpollution has a substantial effect on thestate of health of the population, and thatthis may entail a major cost for society.This cost is partially linked to pureeconomic effects which are calculated atabout Nkr 160 million per year, but thebiggest costs are probably associated withreduced quality of life. We have inter aliafound that around 90 persons may dieprematurely each year in Oslo as a resultof air pollution, and that about 400 per-sons may be diagnosed as suffering fromchronic pulmonary disease each year. Wehave also seen that the long-term conse-quences of air pollution in terms ofreduced life expectancy may be even moreserious.

Although some uncertainty attaches to thefunctions used, it is important to specifythat the greatest uncertainty refers toitems for which functions have yet to beestablished. This applies above all to thelong-term effects of air pollution, of whichwe still have inadequate knowledge. Wehave discussed certain relations of thistype in the study, based on the few studiesthat have been carried out. Air pollutionhas also been demonstrated to affect thebody's immune system, indicating that therisk of falling ill is increasing (NationalInstitute of Public Health 1995). However,this effect is difficult to quantify inasmuchas the initial stages of a disease mayextend over a long period if the immunesystem is impaired. Results indicating thatincreased risk of cancer accompaniesincreased air pollution are also a part ofthis picture (see for example Törnkvistand Ehrenberg (1992) on cancer riskassociated with PAH). There is also a realdanger that the health effects of NO2 aresubstantially understated in this study.

These uncertainties entail that calcula-tions of environmental costs based onavailable dose-response functions onlycapture parts of the cost. This is also a keypoint in EC (1995). Moreover, whencalculating social costs of health damagesin Oslo we have not valued all costsassociated with reduced quality of life.However, some physical health effects arestated, enabling costs to be calculatedbased on various choices of ‘price tickets’.This lends greater transparency to theresult, and required variables can beextracted. The study shows that theapproach chosen for valuing the reductionin quality of life that results from highermortality risk and increased incidence ofdisease is crucial for the size of costsinvolved.


101

6.1. IntroductionA simple model framework for assessingenvironmental costs and other externalcosts in road traffic in Norway waspresented in Alfsen et al. (1992) andBrendemoen et al. (1992). They cal-culated several important costs associatedwith fossil fuel consumption, in order toestablish some systematic informationabout local benefits from carbon emissioncontrol. In Brendemoen et al. (1992) itcame out that traffic related costs weredominant among external marginal costsof fossil fuel consumption in Norway. Thishas called for a further effort in investi-gating traffic related externalities, in thesame manner as for the environmentalexternalities elaborated in chapter 3, 4and 5 in this book. Recently, the data onparticularly traffic accidents and relatedcosts in Norway have been improvedconsiderably. This study elaborates on therelations between fuel use by roadvehicles, traffic accidents and the laboursupply within a CGE61 model framework.The approach is simultaneous, in thesense that economic growth increases

59 Statistics Norway.60 Fellesekspedisjonen for Medisinsk Informasjon(Statistics Norway at the time of the study).61 Computable General Equlibrium.

road traffic volumes causing personinjuries with feedbacks to labour supplyand public health sector costs, which inturn affects the economic growth.

The welfare loss associated withindividual suffering from reduced healthstandard is not included in this study.Instead, by intention, we focus on impactswhich can be traced along flows of goodsand services. The valuation of sufferingand limited physical capabilities of thevictims is not left out because this cost isassumed to be insignificant. Haukeland(1991) provides considerable informationon how traffic injuries affect life quality ofthose who survive. This is furtherevaluated in economic terms by Elvik(1993). Our guideline here is, however, todescribe the link between the scale oftraffic injuries and the national economyas pictured by GDP, via the effect on theavailable labour resources and the drain onpublic health expenditures. Our approach isa somewhat narrow human capitalapproach aiming at catching some relevantimpacts for general policy simulations.

For this study, the question whether costsare internal or external to the driver is notof primary concern. We only focus on how

6. Modelling impacts of trafficinjuries on labour supply andpublic health expenditures

Solveig Glomsrød59, Runa Nesbakken59 and Morten Aaserud60


102

traffic injuries affect the availableresources to the society in terms of laboursupply and the subsequent impact onproduction and income generation. Thequestion of how costs are allocated andwhich incentives they generate is notlooked into. However, the extensive publichealth and social security system inNorway makes medical costs almostcompletely external to the car driver. Thecost of traffic accidents identified in ourstudy has the strength, compared to morepartial studies, that it includes the indirecteffects throughout the economy of setting,in the case of Norway, some 20 000 man-years out of function every year due totraffic injuries. Also, the public hospitalcosts are evaluated including the dead-weight loss due to distortive taxes, whichis dealt with in the CGE-model. These areimprovements compared to Alfsen et al.(1992) and Brendemoen et al. (1992),where the costs were calculated in a linearsub-model of the CGE-model (based on theCGE-forecasted consumption of gasolineand diesel for various economic policies).

We start by giving a description of theavailable sources of information on trafficaccidents and person injuries. Then, insection 6.3 the model framework is setout. In section 6.4 we discuss theassociation between traffic accidents andgasoline/diesel consumption (and othervariables). The reductions in the labourforce, contemporary and permanent, aredescribed in section 6.5. Section 6.6contains information about public costsrelated to traffic injuries. Finally, wepresent scenario simulations illustratingthe feedback effects of traffic accidentswithin a CGE modelling framework.

6.2. Data sourcesA statistical study of the associationbetween the incidence and severity of

person injuries in road accidents, andpotentially explanatory variables at anaggregate level has been carried out byFridstrøm and Bjørnskau (1989). Thecombined cross-section and time-seriesstudy is based on monthly data for 18counties in Norway over the years 1974-1986. Road casualties are explained byfuel consumption (instrument for drivingdistance), traffic density, road capital,precipitation, daylight, road maintenanceand several variables related to drivers’behaviour and traffic control. Trafficaccidents with material damage only arenot included in the study.

Hagen (1993) establishes a socialaccounting system for the costs of trafficaccidents, i.e. an accounting of variouscost elements like hospital costs, treat-ment costs etc. The purpose of theaccounting system is to improve estimatesof total costs generated by trafficaccidents and to facilitate annual up-dating. Among sources of input to thisinformation and accounting system, webriefly mention below those which inparticular have been relevant for our CGEstudy of costs due to traffic injuries.

Haukeland (1991) surveyed some welfareimpacts of person injury in trafficaccidents. This is a thorough study of thehealth and living conditions of the injured1-5 years after the accident took place.The impact on relatives of the injured orother persons was not assessed, except forthe value of some supporting services tothe injured. For injured children, however,the impact on family members is included.In our study we have applied the infor-mation on typical reduced ability inworking situations to assess the impact onlabour productivity. Also, a survey byElvik (1988) provides information on thereduction in the labour supply associated


103

with temporary absence from work. InHagen (1993) there are data on sickleaves related to traffic accidents.

In our model, the cost of labour produc-tivity loss is spread out among productionsectors, via the reduced supply of labourand rising real wage.

6.3. The model frameworkThe interlinkage between generaleconomic activity, traffic injuries andlabour supply is modelled in the followingway, see figure 6.1: Aggregated use ofgasoline (B)62 and autodiesel (D) in theeconomy is calculated by the computablegeneral equilibrium model MSG-EE(Alfsen et al. 1996) for the Norwegianeconomy. A brief description of MSG-EE isgiven in chapter 2 of this book. Then, atraffic injuries module within the MSG-EEmodel framework, called FEEDBACK,relates the use of transport fuels to thetraffic volume (KM):

(6.1) KM = KM(B, D E),

The rate of energy saving (E) in vehicles isincluded to take account of the fact thatmore kilometres will be driven per unitfuel in the future. Then, the number oftraffic injuries (S) is modelled as afunction of the traffic volume, trafficdensity (CON) and a trend variable (T):

(6.2) S = S(KM, CON T).,

The traffic density is determined by trafficvolume (vehicle kilometres) in relation tothe extension of the road net. The roadnet measured in kilometres is assumed tofollow the trend regressed on historical

62 All variables presented in this section relate to thecurrent period t, thus the index t is dropped from thevariable notations.

data from 1966-1993 and data from theRoad Plan for Norway 1994-1997. Insection 6.4 we elaborate further equations(6.1) and (6.2).

Rather than focusing on the total effectsof traffic injuries, we are interested inchanges due to an increase or decrease inthe number of injuries compared to a baseyear. Thus, we will relate all changesbelow to the number of injuries divergingfrom the base year level (∆S). In theFEEDBACK module we have focused ontwo main types of feedbacks from trafficinjuries to the macro model:

• Changes in labour supply• Changes in public expenditure

Figure 6.1. Modelling impacts of traffic injuries in MSG-EE

Laboursupply

Technologicalprogress

Rates of

Man- hours loss Public expenditures onhealth traffic injured

Allocation

return to capital

Fiscal demand

Traffic injuries

Traffic volume

of resources

MSG-EE

GDP

FEEDBACK


104

Changes in labour supply in the economy(∆L) due to changes in traffic injuriesconsist of present changes (∆LC) due toshort term sick leaves and less produc-tivity for traffic injured back at work, andpermanent changes (∆LP) due to fatalitiesand disabilities. Both are functions l of thechange in traffic injuries (∆S):

(6.3) ∆ ∆ ∆ ∆ ∆L = L L l S l SC P c p+ = +( ) ( )

The relations for present and permanentchanges in labour supply are discussed insection 6.5.

In the macro model the labour supply isexogenous and based on populationforecasts. In our study we assume thatthese population forecasts reflect deathrates and employment rates in the baseyear. When the number of traffic injuriesin our model deviate from the base yearlevel, labour supply is adjusted accor-dingly. Thus, the socio-demographic fore-casts of labour supply (L) is adjusted by:

(6.4) L L L* = + ∆

which define the updated labour supply L*

in the macro model.

Public expenditures are generally exo-genous, but we allow expenditures due totraffic injuries to be sensitive to changes indemand. The public resources allocated tocapital ( ),GK

H labour ( )GLH and materials

( )GMH used for treatment of traffic injuries

in the health sector are assumedproportional to the number of injuries:

(6.5) ∆ ∆G g S i K L MiH

iH= =( ) , , ,

Overall public expenditures ( )GiH are

then adjusted for changes in public

resources allocated to treatment of trafficinjuries in the health sector:

(6.6) G G G i K L MiH

iH

iH* , , ,= + =∆

Section 6.6 elaborates these functionsfurther.

6.4. Traffic accidents as a function offossil fuel consumption andother variables

The study by Fridstrøm and Bjørnskau(1989) provides information for relatingthe number of person injuries in trafficaccidents to fuel consumption as aninstrument for driving distance. Gasolinefuelled cars generate 85-90 per cent of thetraffic volume measured in vehicle kilo-metres, although they represent only 60per cent of the total road transport fueluse. It turned out in their estimates that a10 per cent increase in driving distance ofgasoline fuelled cars leads to a 8-9 percent rise in the number of traffic injuries,other factors held constant. All types ofperson injuries, including fatal accidents,are equally affected.

Contrary to what is frequently argued, forinstance by Barker et al. (1993), butconsistent with other studies (see below),Fridstrøm and Bjørnskau found that moreavailable roads and less crowded trafficincrease the number of person injuries. Astudy of the impact on traffic accidents ofbuilding parallel roads in Denmarksupports this result (Vejdirektoratet1979), showing that when new roads takeover traffic from the existing roads, thenumber of accidents on the old roadsdecreases less than proportional to trafficvolume. Thus, high traffic density mayseem to have a beneficial effect on thenumber of accidents with person injuries.The reason may be found in lower speedand possibly more alert drivers. However,


105

the results are ambiguous concerning theimpact of traffic density on accident ratesin general. Similar to Fridstrøm and Bjørn-skau, Zlatoper (1987) found a significantnegative effect of rural traffic density onmotor vehicle deaths in a cross-sectionalstudy for the US. Urban traffic density onthe other hand, had a positive althoughstatistically insignificant effect. Later,Zlatoper (1991) found a positive effect oftraffic density on motor vehicle deathrates. Vitaliano and Held (1991) found nosignificant evidence that traffic densityaffects accident rates in New York State.However, they included accidents withproperty damage in their sample, whichmakes comparison difficult, since it ispossible that accidents with propertydamage respond differently to changes intraffic density than do accidents withperson injuries.

The relation between traffic density andaccident rates is crucial for the deter-mination of external costs related to roadtraffic. Even privately financed insurancecosts are exposed to externalities, if trafficflow affects the rate of accidents (Newbery1988). In his study of road user charges,Newbury applies an assumption that theratio between accident rates of marginaland average vehicle kilometres (VKT) is1.25, which means that the risk ofaccident is higher the more cars are on theroaad. This estimate is a compromisebetween Vickery’s (1969, cited by New-bury 1988) ratio of 1.5, with evidencefrom California freeway driving, and thepractice applied by Department of Tran-sport in Britain based on results showingno significant externality taking place, i.e.a ratio of 1. On the other hand, the statis-tical evidence from Norway in Fridstrøm,and Bjørnskau (1989) (see above) pointsto a positive externality, since increasingtraffic volume reduces the average risk per

kilometre of an accident with personinjury. This effect applies particularly todeaths among pedestrians and bicyclists,but also on fatal accidents in general.

Below, we state formally the abovementioned effects for Norway asimplemented in our model. Other vari-ables than fuel use and traffic density areassumed to be constant (road capital,exposure of pedestrians, climatic vari-ables, demographic variables). Equation(6.7) describes the number of injured (St)in year t as a function of the total drivingdistance in kilometres by gasoline cars(KMB

t), diesel vehicles (KMDt), a traffic

density (congestion) index (CONt) and atrend term (eεt). The derivation of thefunctional form and the values of theelasticities are documented in appendix C.St includes all accidents with person injurywhere one or more vehicles are involved(also bicycles). Costs where only bicyclesare involved are negligible, however.Relative changes in St are proportional torelative changes in explanatory variables.

(6.7)

tt

tB

tD tS K e KM KM

CONCON

= ⋅ ⋅ ⋅ ⋅

⋅ε α σ

β

( ) ( )0

The trend term (eε⋅t) is included to takecare of other variables than fuel consump-tion and congestion, which might affectthe frequency of injuries, but which is notspecified in our study. According to Frid-strøm and Bjørnskau (1989), there was aconsiderable downward shift in risk ofperson injury for a given traffic volume inthe period 1974-1986. The probability ofan accident with person injury would havebeen 50-60 per cent higher than observedin 1986 if it had not been affected byother variables than fuel consumptionsince 1974. To the extent that congestion


106

has been increasing over time, part of thiseffect is taken care of by CONt in ourmodel. However, substantial downwardshift in the probability of person injuryfollowed other particular events in theestimation period. The most importantone was the introduction of seat beltswhich was estimated to explain 20 percent of the observed risk decline.

In the future, several events might affectthe risk of injury. Although the effect ofintroducing seat belts is no longer relevant(the frequency of seat belts reached 90per cent in 1986), there is further techno-logical potential for lowering the risk ofinjury (airbag etc.). In the opposite direc-tion works an increasing share of elderlydrivers in the future (possibly with ahigher risk of accident) and the tendencytowards an increase in exposure ofbicyclists. A recent study (Borger andFrøysadal 1993) found that the transportwork by bicyclists (15 years and older) in1992 was 150 per cent above the 1987-level. Thus the exposure of unprotectedparticipants in the traffic show a risingtrend. The model user may test assump-tions on shifting behaviour through thetrend variable ε. It is, however, importantto remember that the downward shift inrisk may be attended with considerablecosts. These costs may take differentforms, like investment in infrastructure toseparate cars from pedestrians andbicyclists, the time cost of pedestrians/drivers, or the stress involved in takingcare of children’s or old people’s security.

Equations (6.8) and (6.9) transformgasoline and diesel consumption tokilometres driven. MO

B and MOD are base

year mileage per tonnes of gasoline anddiesel, respectively. Bt and Dt are con-sumption of gasoline and diesel, and θ

and ω are fuel specific annual rates ofenergy efficiency increases in gasoline anddiesel vehicles, respectively.

(6.8) KM M B etB B

tt= 0

θ

(6.9) KM M D etD D

tt= 0

ω

Equation (6.10) defines the traffic densityvariable CONt as the relation betweentotal traffic, measured as vehicle kilo-metres (KMt

B+ KMtD), and total length of

the road system (ROADSt).

(6.10) CONKM KM

ROADSttB

tD

t=

+( )

In our calculations, ROADSt is exogenousand determined as an extrapolation basedon historical undertakings in road construc-tion measured in kilometre roads. Anexponentially decreasing function is fittedto historical data from 1966 to 1993 andnumbers for 1994-1997 from the Nor-wegian Road Plan, see appendix C. Thedecreasing growth rate of available roadlength may reflect that the rate of invest-ment is falling and/or that investments perkilometre road are rising. Alternatively, wecould have specified ROADS endogenouslyas a function of economic activity.

The possible effect on the number oftraffic injuries of improved road standard,measured as increasing investments perkilometre road is not included in thismodel. The impact of increased roadcapital per kilometre came out ambiguousin the statistical study by Friedstrøm andBjørnskau (1989). At the county level,higher road capital significantly reducedthe number of injured persons, but onstate roads, the effect was opposite (alsosignificant). However, increasing road


107

maintenance had an unambiguous (nega-tive) impact, reducing the number ofinjured. This effect is not included in ourmodel.

6.5. Labour supply reductions due totraffic accidents

Each year a number of persons are injuredin car accidents. By expectation, a certainshare (69 per cent) of traffic accidentvictims are members of the labour forceand on average each working personrepresents 0.8 man-years due to someamount of part time working (StatisticsNorway 1992c). Some of these are per-manently excluded from work due todeath or reduced capabilities from roadaccidents. Those cases are discussed insection 6.5.2 below. Others are temporaryabsent from work or less productive in aperiod after returning to their jobs. This isdiscussed in section 6.5.1.

In the FEEDBACK module, the laboursupply variable is affected when thenumber of accidents deviates from thebase year level. Thus we implicitly assumethat labour supply along the referencebaseline path in the CGE model does notreflect any foreseen changes in trafficaccident rates.

Figure 6.2 sketches the various elementswhich are included in the injury/laboursupply-module. Non-fatal injuries affectthe labour supply in several ways. Thefirst year after the accident, there isabsence from work due to sick leaves forown injury or children’s injuries. Also,injured returning to work have somewhatreduced productivity. After the first year,those injured who are not going to fullyrecover, are classified as being 100 percent or 50 per cent disabled. Theremaining persons are assumed to be lessproductive at work than before the injury.

Productivity loss is assumed to take placeuntil 10 years after the accident. Sincemost victims are young people, disabilityas well as fatalities on average lead to lossof 38 man-years. These elements arediscussed further in the following sections.

6.5.1. Temporary reduction inlabour supply

By temporary reduction in the labourforce we mean sick leaves during the firstyear after the accident, and productivitylosses up to the 10th year after theaccident. After 10 years, the productivityimpact is assumed to be negligible. Thus,the change in labour supply due totemporary reduction in year t Lt

T( )∆ isdepending on the number of accidents(St) in the 10 previous years, and in thebase year (S0).

Figure 6.2. Impacts of traffic accidents on labour supply

Man-year loss first year

Man-year loss the 2nd year and later

Absence,own injury

Absence,children’s

injury

Productivityloss for traffic

injured, atwork

Productivityloss, 9 years

Man-year loss 38 yearsMan-year loss

Dead in trafficaccidents

accidentsInjured in traffic

50 percentdisabled,38 years

disabled,38 years

100 percent


108

(6.11)∆L S S

S S S S

tT

t

t t

= −

+ − + −= −∑ β

σ λττ τ0

90

0 0

( )

( ) ( )

The first term on the right hand side ofequation (6.11) sums up the change inproductivity for employed workers whowere injured τ years earlier, but still sufferin year t. All changes are measured inman-years. The next element reflectsabsence due to sick leaves the same yearas the accident takes place. This is relatedto own injury. Moreover, in some casesmembers of the labour force stay awayfrom work to care for other injured.Among these cases the available informa-tion only allows for including absence dueto taking care of injured children in ourstudy. This is reflected in the last elementof equation (6.11). We assume allparameters to be constant in our modelsimulations. sections 6.5.1 provide thebackground information for the parameterestimates.

Productivity lossWhen productivity of labour is lowered bya long term impact of traffic injury, theefficiency per working hour should bereduced, not the quantity. To simplify,however, we transform the efficiency lossto a reduced amount of man-years of thehomogenous labour which is available inthe model.

The parameters βτ in equation (6.11) isthe average loss of work effort (measuredin man-years) per injured person τ yearsafter the accident (τ= 0,...,9). βτ are fixedcoefficients estimated from data providedby Haukeland (1991). The study surveysthe share of injured adults who sufferfrom the injury up to 5 years after theaccident. Haukeland’s sample containedinjured with serious and heavy injuries, aswell as minor head or neck injuries.

Although his sample only included up to 5year old injuries, there is reason to believethat even older injuries might causeproductivity losses. Based on hisobservations, we have linearized therelation between the proportion of injuredthat are still suffering, and the number ofyears after the accident. In our model, weassume this relation to be valid up to 9years after the accident.

To estimate βτ we first identify thenumber of man-years (MY) performed byindividuals who were injured τ years agoin a traffic accident, and stiller suffer:

(6.12) MY a AS b SL a INV= ⋅ − ⋅ − ⋅τ τ τ

The injured adults still suffering (AS) areassumed to participate in the labour forcewith the same frequency (a) as theaverage of the working population. Thisworking intensity is the product of twocomponents. One component is theproportion of persons expected to beemployed (i.e., 0.69 according to StatisticsNorway (1992c)), and the other compo-nent is the average number of man-yearsper employed person (i.e., 0.80 accordingto Statistics Norway (1992c)), which islower than 1 due to part time work. Thus,we find that a=0.55. To adjust for thosewho are suffering, but on sick leave, wesubtract the number of persons on sickleave (SL), assigning a work participationrate of b= 0.80 since they by definitionare employed. We also subtract theexpected work amount of those who aredisabled, who by assumption are assignedthe average work participation rate a. Foreach cohort τ of injured adults who stillsuffer, a certain share (Rτ) experiencereduced capacity, implying a productivity


109

loss of 10 per cent on average.63 Hence theman-year loss per traffic injured due toreduced productivity can be written as:

(6.13)

ττ τ τ τβ

τ

=⋅ ⋅ ⋅ − ⋅ − ⋅

=

R 0.1 (a AS b SL a INV )

S ( 0 )

0

(6.14)

ττ τ τβ

τ

=⋅ ⋅ ⋅ − ⋅

=

R 0.1 (a AS a INV )

S

( 1 ,.... 9 )0

,

Then βτ can be estimated from Hauke-land’s (1991) data. Equation (6.13)represents productivity loss due toaccidents in the same year, taking accountof the injured still on sick leave. In thefollowing years, all have either returnedto the working force or become classifiedas disabled, and equation (6.14) is simpli-fied accordingly. Probably some fractionof the injured (although not disabled)suffers more than 10 years after theaccident. A possible effect on this group isnot included in our study.

Absence from work due to own andchildren’s injuryThe average loss of man-years per injuredassociated with own and children’s injuryare σ and λ respectively (equation 6.11)).The coefficient for own injury is based onHagen (1993), and found to be 0.04.Man-year loss associated with children’sinjury is estimated by Elvik (1988) for1986, with information also from Lereim(1984) who studied the temporaryabsence from work related to trafficaccidents in the city of Trondheim. The

63 Rτ corresponds to Haukeland’s (1991) proportion ofinjured with reduced functional capacity.

coefficient for children’s injury isestimated to be 0.005.

6.5.2. Labour supply reduction dueto fatalities and disability

Reduction due to fatalitiesThe effect of fatal traffic accidents in thismodel is limited to the impact of losingthe labour potential of an average personkilled in the accident. No value of life-considerations are made. The change inlabour supply due to fatal traffic accidents

( )∆LtF is given by:

(6.15) ∆ tF

t 0L (S S )= − −η

where the coefficient η is determined by:

(6.16) η = a FAS

0

0

FA0 is the number of fatalities in the baseyear (1991), counting 332 persons (Bil ogvei 1992), whereas the parameter a(=0.55) was discussed above. Thus, thelabour supply reduction is calculated to be186 man-years. The majority of those whoare killed by traffic injuries are relativelyyoung people. Elvik (1988) provides anage distribution for men and womenkilled in traffic accidents. Based on thisage distribution and an upper age limit inthe labour force of 74 years, it turns outthat the members of the labour forcekilled by traffic injuries in average lose 38active working years. The upper age limitof the labour force might seem high.However, the average number of man-years carried out by a person in the workforce (a) is taking account of the lowwork participation rates of elderly people.


110

Reduction due to disabilityDisability also incur long term changes inlabour supply ( )∆Lt

D and this is treated in asimilar way to fatalities. However, disabilityoccurs after an initial year on sick leave:

(6.17) ∆ tD

t 0L (S - S )= − −µ 1

(6.18) µ =a INV

S0

1

INV1 is the number of persons who aredeclared disabled one year after anaccident. We assume that those who havenot returned to their work one year afterthe accident are transferred to job trainingor social pension. (The first year of sickleaves after the accident, including man-years of persons who later face disability,is dealt with in section 6.5.1)

According to Hagen (1993), 2.7 per centof the traffic injured were on sick leave formore than a year. It is assumed that 75per cent of these individuals later receivedpensions for their living (738 persons in1991). Some of these have first been injobtraining programs. We assume here,though, that job trainees have not contri-buted to productive activities. FromHaukeland (1991) we know that 63 percent of the disabled are classified as 100per cent disabled, and the rest areclassified as 50 per cent disabled. More-over, we still assume that the injuredpersons on average worked 0.55 man-years before the accident. The averageman-year loss per injury (µ) is then foundto be 0.009 per year. It is assumed thatdisability lasts the whole lifetime. Thus, byassuming the same age distribution fordisabled as for people killed in trafficaccidents (Elvik 1988), we conclude thatthis yearly effect applies for 38 years onaverage.

Adjusting the labour supplyWe assume the age distribution amonginjured and the share of deaths or dis-abled in traffic injuries to be constant.Hence the impact on the labour supply perinjured will be constant throughout thesimulation period. As the time horizon inour model simulations is shorter than 38years, the man-year loss per killed anddisabled is treated as a permanent shift inthe labour stock. The labour supply is thuschanged according to the followingequation:

(6.19) ∆ ∆ ∆ ∆tP

tF

tD

t-1PL = L + L + L

where ∆LtP is the permanent reduction in

labour supply due to feedbacks fromtraffic accidents (in excess of the accidentlevel in the base year).

6.5.3. Annual effects on laboursupply of traffic accidents

Table 6.1 depicts some key variables ofman-year loss related to traffic accidentsbased on the associations discussed in thissection. In 1990, 33 900 persons wereinjured, of these 332 fatally. Man-yearlosses from the fatal accidents, includingfuture losses, were about 7 000. More-over, 13 000 man-years were lost due todisability. Productivity loss accounts forabout the same reduction in labour supplyas sick leaves (1 400 man-years). Theresulting estimate of total man-year lossover the whole time horizon of injured isabout 23 000.

6.6. Public health sector costsTraffic accidents demand considerableresources through medical treatment andlegal actions. When the number of acci-dents decline, these resources can be usedfor other purposes. In our model, we let thelevel of government expenditures in the


111

health sector deviate from the exogenousreference path when the number of trafficaccidents deviates from the base year level.The level of all other services from thepublic health sector is assumed to remainat the level of the reference path.In addition to allow for more privateconsumption, the whole economy mightgain some efficiency by transferringresources from public to private sector. Asestimated by Brendemoen and Vennemo(1993), the marginal cost of public fundsin Norway is about 1.6, i.e. 1 Nkr used inthe public sector reduces the income inthe private sector by 1.6 Nkr.

Elvik (1988) calculated the total medicaltreatment cost of traffic injuries to be 229million Nkr in 1986, including 6 million inmedical care in the case of fatalities. Thecomplete surveys of the number of trafficaccident do not go back to 1986. How-ever, so does the police registration oftraffic accidents. To find the level ofhealth costs in our base year (1991), weassume that the cost developed propor-tionally to the police registered number oftraffic accidents over the years from 1986to 1991, adjusting for inflation which was30 per cent from 1986 to 1991 (StatisticsNorway 1996b). The number of policeregistered accidents fell from 12 458 in1986 to 12 035 in 1991 (Bil og vei 1992).The health sector expenses thus amounted

to 289 million Nkr in 1991. These costsare mainly hospital expenses. With 36 450injured persons in 1991, the averagemedical cost per injured was 7 900 Nkr.

Hagen (1993) provides a revised estimatefor 1991. This survey includes moreelements of medical care than Elvik(1988). Table 6.2 presents the mainelements in the cost assessment. Theestimated hospital costs are about thesame level as estimated by Elvik (1988),but they make up only 60 per cent of totalpublic health sector expenditures. For an

Table 6.1. Key variables related to traffic accidents in 1990

Personsaffected

Man-yearloss

Deaths in traffic accidents 332 7 254Traffic injured 33 900Sick leaves the first year after the accident (own injury) 1 350Absence from work due to children’s injury in traffic accidents 167Productivity loss for former traffic injured (total for ten years after the accidents) 1 34650 per cent disabled 272 2 888100 per cent disabled 477 10 146Total 23 151

Table 6.2. Public health sector expenses in 1991 due to traffic accidents1. Million 1991-Nkr

Public sector:Hospital stay 334Nursing homes 146Policlinic consultations and treatment 42Ambulance transport 3Primary health sector consultations (and physiotherapy) 18

Total public sector costs 543

Private Sector:Nursing homes 36Policlinic consultations and treatment 11Primary health sector consultations 4Medicines 203

Total private sector costs 2541 Hospital stay is 100 per cent publicly financed, while theprivate sector covers 20 per cent of costs in other healthinstitutions.Source: Hagen (1993).


112

overview, table 6.2 also presents theprivate sector costs, although they are notincluded in the FEEDBACK module.

In this study we focus on the public sectorcosts. Table 6.3 presents the distributionof variable hospital costs on various inputvariables. In addition comes capital costs.Due to the nature of the public sectoraccounting systems, data on capital costsare not readily available. However, a lowestimate of capital costs is included inorder not to completely overlook thesereal costs. The capital costs are set to5 per cent of variable costs, which isregarded as a low estimate by the StateAgency for Buildings and Property (asreferred by Hagen (1993)).

We assume that the variable medical costsrelated to road traffic accidents are allo-cated on input factors roughly propor-tional to the general allocation of inputfactors in the hospital sector (see table6.3). Thus, we assume that 70 per cent ofvariable costs are labour costs; the other30 per cent are intermediates.

The activity in the public health sector isbasically exogenously determined withinthe model, in the sense that the variousfactor inputs are fixed. When trafficaccidents deviate from the base year level,the use of each input factor in the publichealth sector is adjusted according to thedistribution displayed above:

(6.20) ∆ itH

iH

t 0G (S S ) i K L M= ⋅ − =ω , , ,

GiH is the specific cost related to input of

factor i in public hospitals, which isproportional to the number of injuredpersons in traffic accidents (see equation(6.5) in section 6.3).

6.7. SimulationsIn this section we first describe how theinclusion of the traffic module into theCGE model MSG-EE changes the referencescenario. Then we report from simulationsof a CO2-tax scenario, with and withoutthe feedback from traffic injuries onlabour supply and public health sectorcosts.

6.7.1. The scenariosSimulations with two model versions havebeen carried out. Model version 1 is MSG-EE itself. Model version 2 is MSG-EE withthe traffic injury module, adjusting laboursupply and public health expenditures asexplained above. For each of the modelswe first simulate a reference path whereexogenous variables including the realvalues of the CO2 taxes are kept constantat their 1993 levels throughout thesimulation period 1988-2020. Hence, thereference scenarios REF1 and REF2 onlydiffers when it comes to modelling of thefeedbacks from traffic injuries. We then

Table 6.3. Variable hospital costs 1991. Per cent

Wages 71Equipment 2Maintenance 1Other costs 24Transfers 1

Source: Statistics Norway (1993b)

Table 6.4. Scenarios

MSG-EEwithoutfeedbacks

MSG-EEwithfeedbacks

Reference- constant 1993-level CO2 tax

REF1 REF2

Alternative- increasing CO2 tax

TAX1 TAX2


113

simulate within both model versions theeffect of introducing an increasing CO2

tax. These scenarios, TAX1 and TAX2, arediscussed in the next section.

Table 6.5 shows the development in thetwo reference scenarios. In REF1, basedon demographic forecasts, labour supplygrows by 0.265 per cent per year over thesimulation period (1988-2020). AnnualGDP growth is 1.65 per cent in the sameperiod. This growth is absorbed byincreased private consumption and a risein government expenditures, while netexport and gross investments areunchanged. Factor neutral technologicalprogress is assumed to be 0.5 per cent peryear in all the production sectors exceptfor road, air and domestic sea transport.Based on a study by the NorwegianInstitute for Transport Economics (Thune-Larsen 1991) annual autonomous fuelefficiency improvement for road transportusing gasoline and autodiesel is assumedto be 1.0 and 0.6 per cent respectively,while the rates of improved energyefficiency for air and domestic sea tran-sport are 0.9 and 1.3 per cent. Assump-tions for exogenous variables are furtherdocumented in Johnsen et al. (1996). Theassumptions are roughly based on theassumptions in the “KLØKT” analysis ofclimate policy problems on a nationalscale (Moum 1992).

As production and consumption increasein the REF1 scenario, the total fuel use inroad transport also increases. The annualgrowth of 1 per cent in transport fuelscorresponds to an annual growth of 2 and1.6 per cent in kilometres driven bygasoline and diesel vehicles respectively.The direct effect of the growth in roadtransport activity is an increase in thenumber of persons injured or killed intraffic accidents. However, the trafficgrowth also has an opposite smaller effecton the injuries through increasedcongestion. In total, the number of personinjuries rises by 52 per cent over thesimulation period in the REF1 scenario, toreach 47 831 injured per year in 2020.64

In the REF2 scenario the increase ininjuries and fatal accidents leads to a fallin labour supply as a first order effect. Thenegative feedback on labour supply willslightly slow down the economic growthand, ceteris paribus, the increase in fueluse and accidents in road transportation.An opposite second order effect onaccidents occurs when the decrease inlabour supply increases the wage rate toclear the labour market. The substitution

64 The number of persons injured or killed in 1988(i.e., 31 464) is based on the estimated number of33 900 persons injured or killed in 1990 (Haukeland1991) and adjusted proportionally to the change inpolice registered traffic accidents from 1988 to 1990(Bil og vei 1992).

Table 6.5. Main variables, reference paths

LEVEL1988

REF1

Annual growth1988-2020. Per cent

REF2

Dev. from REF1,level 2020. Per cent

GDP (billion 1988-Nkr) 583 1.7 -0.34Labour supply (million man-hours) 3 019 0.3 -0.29Fuel use in road transport (1000 metric tonnes) 2 867 1.0 -0.35Persons injured in traffic accidents 31 464 1.3 -0.24


114

effect of this change in relative prices is anincreased demand for other input factors,among them transport fuels, accompaniedby an increase in the number of trafficaccidents.

The increase in traffic accidents from thebase year level also leads to a rise inpublic expenditures on labour, capital,energy and materials needed for treat-ment of traffic victims in the health sector.With a given amount of productiveresources in the economy as a whole,these incremental needs in the healthsector has to be covered by resourceswithdrawn from other sectors. Our resultscover the efficiency loss when distortingtaxes reallocate resources from private topublic sectors. However, the effect turnsout to be small.

When coming to terms with all of theseeffects, the number of persons killed orinjured in traffic accidents in the REF2scenario compared to the REF1 scenarioturns out to be lowered by 114 in 2020due to the feedback effects in theeconomy, whereas the effect on the laboursupply is a 0.29 per cent decline in 2020.Although not insignificant, the declinetaking place in the period 1988-2020corresponds to only one year expectedincrease in the labour supply during the

period. The reduced labour supply andincreased public expenditures result in aGDP level in REF2 which is 0.34 per cent,or about 3.3 billion 1988-Nkr, lower thanthe REF1 level in 2020. The effect viareduced labour supply is far stronger thanthe effect via increased demand forresources in the public health sector. REF2may be considered as a reference pathwhich is an alternative to the “official”path from the simulation with MSG-EEwithout any feedbacks.

6.7.2. The impacts of a CO2 taxIn the reference paths REF1 and REF2, thereal value of the CO2 tax is kept constantat the 1993 level throughout the simulationperiod, i.e. the nominal value increases by3.5 per cent per year. Below we reportsome results from simulations with anincreasing CO2 tax designed to stabilise thecarbon dioxide emissions in the REF1scenario on a 1989 level around year 2020.See table 6.6.

The scenarios TAX1 and TAX2 show theimpact of increasing the nominal value ofthe CO2 tax by 10 per cent per year from1994 to 2000, and then by 7 per cent peryear from 2001 to 2020. The differencebetween the reference scenarios and theCO2 tax scenarios for some main variablesin 2020 are studied.

Table 6.6. Main variables, tax simulations

Deviation 2020. Per cent

TAX1 vs. REF1 TAX2 vs. REF2

CO2 tax 198 198GDP -0.47 -0.44Labour supply 0 0.02Labour use, public health sector 0 -0.03Materials/energy use, public health sector 0 -0.04Capital use, public health sector 0 -0.02Fuel use in road transport -4.94 -4.92Person injuries, road traffic -3.37 -3.35


115

In model version 1, with no feedbacksfrom traffic accidents to the economy, theintroduction of an increasing CO2 taxreduces GDP by 0.47 per cent in 2020.The tax increases the cost of fuel use andthus implies a restriction on economicactivity. For a further discussion of theimpact of increased CO2 taxes in MSG-EE,see Johnsen et al. (1996). It should benoted that the tax does not affect thelabour supply in this model version. Whensimulating the impact of the CO2 tax withmodel 2, the cost of carbon control comesout somewhat less expensive by 2020 thanwith the model 1 alternative. The taxreduces the fuel use and thus the trafficvolume, leading to fewer accidents andincreased labour supply. In total the taxreduces GDP by 0.44 per cent in 2020.Thus, the GDP reduction is about0.3 billion 1988-Nkr smaller than in theModel 1 case.

6.8. ConclusionsIt should be stressed that the simulationsin this study are of illustrative nature.Although the scale of the effects on thelabour force is fairly well documented,there are, of course, uncertaintiesassociated with the data and the valuesused in the tentative calculations.However, the simulations indicate thatclimate policies in the long run might notbe as costly – measured by GDP loss – asshown by model simulations withoutnegative feedback effects to the economyfrom the use of fossil fuels. The abovesimulations only include two of thesefeedbacks. Other effects have beenpresented in other chapters of this book,and several others could be mentioned(e.g., congestion and noise). It should alsobe remembered that welfare gains fromreduction in the number of person injuriesis not included in this study.


117

Aasness, J., T. Bye and H.T. Mysen(1996): Welfare Effects of Emission Taxesin Norway, Energy Economics 18, 335-346.

Abbey, D.E., M.D. Lebowitz, P.K. Mils, F.F.Petersen, W.L. Beeson and R.J. Burchette(1993): Long-Term Ambient Concen-trations of Particulates and Developmentof Chronic Disease in a Cohort of Non-smoking California Residents, InhalationToxicology 7, 19-34.

Abbey, D.E., M.D. Lebowitz, P.K.Mills, F.Petersen, W.L. Beeson and R.J. Burchette(1995): Long-Term Ambient Concen-trations of Total Suspended Particulates,Ozone and Sulfur Dioxide and RespiratorySymptoms in a Non-Smoking Population,Archives of Environmental Health 48, 33-46.

Agricultural Budget Commission (1995):Jordbrukets totalregnskap 1993 og 1994.Jordbrukets totalbudsjett 1995 (Aggregateaccounts for agriculture 1993 and 1994.Aggregate budget for agriculture 1995),Budsjettnemnda for jordbruket(Agricultural Budget Commission), Oslo.

Agricultural Budget Commission (1996):Jordbrukets totalregnskap 1994 og 1995.

Jordbrukets totalbudsjett 1996 (Aggregateaccounts for agriculture 1994 and 1995.Aggregate budget for agriculture 1996),Budsjettnemnda for jordbruket(Agricultural Budget Commission), Oslo.

Alfsen, K.H., A. Brendemoen and S.Glomsrød (1992): Benefits of climatepolicies: Some tentative calculations,Discussion Papers 69, Statistics Norway.

Alfsen, K.H., T. Bye and E. Holmøy (eds.)(1996): MSG-EE: An Applied GeneralEquilibrium Model for Energy andEnvironmental Analyses, Social andEconomic Studies 96, Statistics Norway.

Alfsen, K.H. and K.E. Rosendahl (1996):Economic Damage of Air Pollution,Documents 96/17, Statistics Norway.

Andersson, B (1994): Corrosion damagecosts caused by sulphur dioxide emissions,Swedish Environmental Audits. (Swedishtext)

Aunan, K. (1995): Exposure-ResponeFunctions for Health Effects of AirPollutants Based on EpidemiologicalFindings, Report 1995:8, CICERO, Oslo.

References


118

Aune, B. (1993): Temperaturnormaler;normalperioden 1961-1990, (Temperaturenormals; normal period 1961-1990),DNMI-rapport Klima 2/93, The Nor-wegian Meteorological Institute, Oslo.

Barker, T., N. Johnstone and T. O'Shea(1993): The CEC Carbon/Energy Tax andsecondary Transport- related Benefits.Energy-Environment-Economy Modelling,Discussion Paper 5. Dept. of AppliedEconomics, University of Cambrigde, UK.

Bergh, J. van den (1993): A Frameworkfor Modelling Economy-Environment-Development Relationships Based onDynamic Carrying Capacity and Sustain-able Development Feedback, Enviro-nmental and Resource Economics 3, 4, 395-412.

Bil og vei (1992): Bil og vei, statistikk1992 (Car and road, statistics 1992),Opplysningsrådet for veitrafikken, NorskVeiforening, Oslo.

Bojkov, R.D. (1986) Surface ozone duringthe second half of the nineteenth century,Journal of climate and applied meteorology25, 343-352.

Borger, A. and E. Frøysadal (1993): Dennasjonale sykkelundersøkelsen (The nation-al bicycle study), Report 217/1993,Institute of Transport Economics (TØI),Oslo.

Bown, W. (1994): Dying From Too MuchDust, New Scientist, March 12, 12-13.

Braun-Fahrländer, C., U. Ackermann-Liebrich, J. Schwartz, H.P. Gnehm, M.Rutishauser and H.U. Wanner (1992): AirPollution and Respiratory Symptoms inPreschool Children, American Review ofRespiratory Disease 145, 42-47.

Brendemoen, A. and H. Vennemo (1993):The marginal cost of funds in the presenceof external effects, Discussion Papers 99,Statistics Norway.

Brendemoen, A., S. Glomsrød, and M.Aaserud, (1992): Miljøkostnader imakroperspektiv (Environmental costs in amacro perspective), Reports 92/17,Statistics Norway.

Brookshire, D., M. Thayer, W. Schulze andR. D’Arge (1982): Valuing Public Goods: AComparison of Survey and HedonicApproaches, American Economic Review72, 165-177.

Bye, B. (1996): Environmental TaxReform and Producer Foresight: AnIntertemporal Computable GeneralEquilibrium Analysis, Discussion Papers185, Statistics Norway.

Calthrop, E. and D. Maddison (1996): TheDose-Response Function Approach toModelling the Health Effects of AirPollution, Energy Policy 24, 599-607.

Coase, R. (1960): The Problem of SocialCosts, The Journal of Law and Economics3, 1-44.

Cowell, D. and H. ApSimon (1994):Estimating the Cost of Damage toBuildings by Acidifying Pollution inEurope. 1. UN ECE Workshop onEconomic Evaluation of Damage Causedby Acidifying Pollutants, London, May1994.

Cropper, M. and N.B. Simon (1994): AreWe Valuing Risks to Life Correctly?,mimeo, World Bank, Washington DC.

Dockery, D.W., C.A. Pope, X. Xu, J.D.Spengler, J.D. Ware, M.E. Fay, B.G. Ferris


119

Jr. and F. E. Speizer (1993): AnAssociation Between Air Pollution andMortality in Six US Cities, New EnglandJournal of Medicine 329, 1753-59.

Dockery, D.W., J. Schwartz and J.D.Spengler (1992): Air Pollution and DailyMortality: Associations with Particulatesand Acid Aerosols, Environmental Research59, 362-373.

EC (European Commission) (1995):ExternE: Externalities of Energy, DG XII(Joule Programme), Volume 1-6.

ECO (1987): Generic Coal Study: Quantifi-cation and Valuation of EnvironmentalImpacts, Report to Bonneville PowerAdministration, 1987.

Elvik, R. (1988): Ulykkeskostnader ogverdien av unngåtte ulykker (The loss ofwelfare from road accidents), TØI-notat877/1988, Institute of TransportEconomics (TØI), Oslo.

Elvik, R. (1993): Økonomisk verdsetting avvelferdstap ved trafikkulykker (Economicvaluation of reduced risk of traffic acci-dents), Rapport 203/1993, Institute ofTransport Economics (TØI), Oslo.

EPA (U.S. Environmental ProtectionAgency) (1995): Human Health BenefitsFrom Sulfate Reductions Under Title IV ofthe 1990 Clean Air Act Amendments,Final Report, November 10, 1995.

Erdal, L. and N.K. Nersten (1996): Analyseav norske landbruksvarers beskyttelsesnivå.Grøntsektoren (Analysis of Norwegianprotection levels for agricultural com-modities. Vegetable sector), Notat 1996/5,Norwegian Agricultural EconomicsResearch Institute.

Eskeland, G. (1995): “The Costs andBenefits of an Air Pollution ControlStrategy for Santiago, Chile” in Chile:Managing Environmental Problems:Economic Analysis of Selected Issues, Report13061-CH, World Bank, Washington DC.

Fairley, D. (1990): The Relationship ofDaily Mortality to Suspended Particulatesin Santa Clara County, 1980-1986,Environmental Health Perspective 89, 159-168.

Fridstrøm, L. and T. Bjørnskau (1989):Trafikkulykkenes drivkrefter. En analyse avulykkestallenes variasjon i tid og rom (Thedeterminants of road traffic accidents inNorway. A combined cross-sectional andtime-series analysis), Rapport 39/1989,Institute of Transport Economics (TØI),Oslo.

Glomsrød, S. and A. Rosland (1988):Luftforurensning og materialskader:Samfunnsøkonomiske kostnader (Airpollution and material damage: Economiccosts), Reports 88/31, Statistics Norway.

Glomsrød, S., H. Vennemo and T. Johnsen(1992): Stabilization of Emissions of CO2:A computable General EqulibriumAssessment, Scandinavian Journal ofEconomics 94, 1, 53-69.

Glomsrød, S. O. Godal, J.F. Henriksen,S.E. Haagenrud and T. Skancke (1996):Luftforurensninger – Effekter og verdier.Materialkostnader på bygninger og biler iNorge. SFT-reports 96:07, NorwegianPollution Control Authority, Oslo.

Gram, F. (1982): Utslipp av luftforurens-ninger i Oslo området 1979 (Emission ofatmospheric pollutants in the Oslo area1979), NILU OR 10/82, Norwegian


120

Institute of Air Research, Lillestrøm,Norway.

Gram, F. (1994): Utslipp av NOx ogNMVOC i Oslo 1991. Nirogenoksider ogoksidanter i byområder; Dannelse ogeffekter (Emission of NOx and NMVOC inOslo 1991. Nitrogen oxides and oxidantsin urban areas. Formation and effects),NILU OR 79/94, Norwegian Institute ofAir Research, Lillestrøm, Norway.

Grønskei, K.E. and S.E. Walker (1993)Evaluation of a model for hourly spatialconcentrations distributions, Atmosphericenvironment 273, 105-120.

Haagenrud, S.E. and J.F. Henriksen(1995): Material corrosion caused by airpollutants - with emphasis on dose-responserelations, NILU OR 74/94, NorwegianInstitute of Air Research, Lillestrøm,Norway.

Haagenrud, S.E., J.F.Henriksen and F.Gram (1984): Basisundersøkelse iSarpsborg og Fredrikstad 1981-1983 (Basisstudy in Sarpsborg and Fredrikstad 1981-1983) Final report. Interim report B:Corrosion and environment, NILU OR28/84, Norwegian Institute of AirResearch, Lillestrøm, Norway.

Hagen, K.E. (1993): Samfunnsøkonomiskregnskapssystem for trafikkulykker ogtrafikksikkerhetstiltak (Sosio-economicaccounting system for traffic accidentcountermeasures), Rapport 182/1993,Institute of Transport Economics (TØI),Oslo.

Hagen. L.O. (1994): Rutineovervåking avluftforurensninger. April 1993-mars 1994(Routine monitoring of air pollutants.April 1993 - March 1994), NILU OR

46/94, Norwegian Institute of AirResearch, Lillestrøm, Norway.

Hall, J., A. Winer, M. Kleinman, F.Lurmann, V. Brajer and S. Colome (1992):Valuing the Health Benefits of Clean Air,Science 255, 812-817.

Halvorsen, B. (1996): Ordering Effects inContigent Valuation Surveys. Willingnessto Pay for Reduced Health Damage fromAir Pollution, Environmental and ResourceEconomics 8, 4, 484-499.

Hansen, A. and H. Selte (1997): AirPollution and Sick-leaves - is there aConnection? A Case Study using AirPollution Data from Oslo, DiscussionPapers 197, Statistics Norway.

Hasselblad, V., D. Eddy and D. Kotchmar(1992): Synthesis of EnvironmentalEvidence: Nitrogen Dioxide EpidemiologyStudies, Journal of Air and WasteManagement Association 42, 662-671.

Hatzakis, A., K. Katsoouyanni, A.Kalandidi, N. Day and D. Trichopoulos(1986): Short-term Effects of Air Pollutionon Mortality in Athens, InternationalJournal of Epidemiology 15.

Haukeland, J.V. (1991): Velferdstap vedtrafikkulykker (Welfare impacts of injuriesby traffic accidents), Rapport 92/1991,Institute of Transport Economics (TØI),Oslo.

Heinz, I.B. Isecke and M. Weltchev(1995): Volkwirtschaftliche Verlustedurch UmveltverschmutzungsbedingteMaterialschäden in BundesrepublikDeutschland. Vorschungsbericht10103110-02 UBA. SP 91074.


121

Henriksen, J.F., Haagenrud, S.E. andF.Gram (1981): Impact of mandatory useof oil with a low sulphur content onatmospheric corrosion damage, NILU OR17/81, Norwegian Institute of AirResearch, Lillestrøm, Norway (Norwegiantext).

Hill, A.B. (1965): The Environmental andDisease: Association or Causation?, ProcRoyal Soc Med 58, 295-300.

Hohmeyer, O. (1988): Social Costs ofEnergy, Springer-Verlag, Berlin.

Johansen, L. (1960): A multi-sectoral studyof economic growth, North HollandPublishing Company, Amsterdam.

Johnsen, T. A., B. M. Larsen and H. T.Mysen (1996): Economic Impacts of a CO2

Tax, in K. Alfsen et al. (1996) (eds.):MSG-EE: An Applied General EquilibriumModel for Energy and EnvironmentalAnalyses, Social and Economic Studies 96,Statistics Norway.

Journel, A.G. and C.J. Huibregts (1981):Mining Geostatistics, Academic Press,London.

Kahneman, D. and J.I. Knetsch (1992):Valuing Public Goods: The Purchase ofMoral Satisfaction, Journal of Environ-mental Economics and Management 22,57-70.

Kinney, P.L., K. Ito and G.D. Thurston(1994): Daily Mortality and Air Pollutionin Los Angeles: An Update (Abstract),American Review of Respiratory andCritical Care Medicine; Supplement: p.A660. (Reference incomplete - see EC(1994))

Krupnick, A.J., W. Harrington and B.Ostro (1990): Ambient Ozone and AcuteHealth Effects: Evidence from Daily Data,Journal of Environmental Economics andManagement 18, 1-18.

Kucera, V., J.F.Henriksen, D. Knotkovaand Ch. Sjöström (1993): Model forcalculation of corrosion costs caused byair pollution and its application in threecities. Progress in understanding andprevention of corrosion, 10th EuropeanCorrosion Congress, Barcelona, July 1993.

Kucera, V., J.Tidblad, J. Henriksen, A.Bartonova and A.A. Mikhailov (1995):Statistical analyses of 4 year materialsexposures and acceptable deteriorationand pollution levels, Swedish CorrosionInstitute, Stockholm.

Lave, L.B. and E.P.Seskin (1970): AirPollution and Health. The QuantitativeEffect, with an Estimate of the DollarBenefits of Pollution Abatement, isConsidered, Science 169, 723-733.

Lave, L.B. and E.P.Seskin (1972): AirPollution, Climate, and Home Heating:Their Effects on U.S. Mortality Rates,American Journal of Public Health 62,909-916.

Lereim, I. (1984): Traffic accidents andtheir consequences. A study on injured roadusers treated at the regional hospital ofTrondheim. Tapir, Trondheim.

Lipfert, F. W. (1987) Effects of acidicdeposition on the atmosphericdeterioration of materials, MaterialPerformance 26, 7, 12-19.

Lövblad, G., P.I. Grennfelt, L. Kärenlampi,T. Laurila, L. Mortensen, K. Ojanperä, H.Pleijel, A. Semb, D. Simpson, L. Skärby, J.P.


122

Tuovinen and K. Tørseth (1996): Ozoneexposure mapping in the Nordic countries,Report from a Nordic cooperation projectfinanced by the Nordic Council ofMinisters.

Matheron, G. (1963): Principles of geo-statistics, Economic Geology 58, 1246-1266.

Ministry of Agriculture (1992): Landbruk iutvikling. Om retningslinjer for landbruks-politikken og opplegget for jordbruksopp-gjørene m.v. (Agriculture on the move.Guidelines for agriculture policy and theprogramme for income settlements in theagriculture sector), St.prp. nr. 8 (1992-93).

Ministry of Agriculture (1994): Omomdanning av Statkorns forretnings-virksomhet til statseid aksjeselskap ogeiermessig samordning (Conversion ofStatkorn into a state-owned joint stockcompany and coordination of ownership),St.prp. nr. 1. Tillegg nr. 8 (1994-95).

Mortensen, L.M. (1992): Effects of ozoneon growth of seven grass and one cloverspecies, Acta Agriculture Scandinavia (Sect.B, Soil and Plant Sci.) 42, 235-239.

Mortensen, L.M. (1993): Effects of ozoneon growth of several subalpine plantspecies, Norwegian Journal of AgriculturalScience 7, 129-138.

Mortensen, L.M. (1994a): Further studieson the effects of ozone concentration ongrowth of subalpine plant species, Nor-wegian Journal of Agricultural Science 8, 91-97.

Mortensen, L.M. (1994b): The influence ofcarbon dioxide or ozone concentration ongrowth and assimilate partitioning inseedlings of nine conifers, Acta Agriculture

Scandinavia (Sect. B, Soil and Plant Sci.)44, 157-163.

Mortensen, L.M. (1995): Air pollution,effects and values. Effects on crops, forest andother vegetation (dose-response), SFT Report95:21, Norwegian Pollution ControlAuthority, Oslo.

Mortensen, L.M. and J. Nilsen (1992):Effects of ozone and temperature ongrowth of several wild plant species,Norwegian Journal of Agricultural Science 6,195-204.

Moum, K. (ed.) (1992): Klima, økonomiog tiltak (KLØKT) (Climate, economy andcontrol policies), Reports 92/3, StatisticsNorway.

National Institute of Public Health (1995):Helseskader forårsaket av luftforurensning- en kunnskapsstatus av dose-respons-forhold (Health effects caused by airpollution - a state of knowledge aboutdose-response relationships), draft.

Newbery, D.M. (1988): Road UserCharges in Britain, The Economic Journal98, 161-176.

Norwegian National Grain Administration(1996): Produksjonstillegg i jordbruket(Agricultural production subsidies) LJ-001.

Nyborg, K. (1996): Information Requir-ements for Environmental Policy Making,Ph.D. thesis, University of Oslo.

Ofstad, T. (1995): NILUs feltstasjoner forkorrosjon. Miljø- og korrosjonsmålinger1993 (NILU's field stations for corrosion.Environment and corrosion measurements1993), NILU OR 8/95, NorwegianInstitute of Air Research, Lillestrøm,Norway.


123

ORNL/RFF (Oak Ridge National Labora-tory og Resources for the Future) (1994):External Costs and Benefits of Fuel Cycles, AStudy by the U.S. Department of Energyand the Commission of the EuropeanCommunities, Oak Ridge, TN.

Ostro, B. (1983): The Effects of AirPollution on Work Loss and Morbidity,Journal of Environmental Economics andManagement 10, 371-382.

Ostro, B. (1987): Air Pollution andMorbidity Revisited: A Specification Test,Journal of Environmental Economics andManagement 14, 87-98.

Ostro, B. (1990): Associations BetweenMorbidity and Alternative Measures ofParitculate Matter, Risk Analysis 10, 421-427.

Ostro, B. (1993): The Association of AirPollution and Mortality: Examining theCase for Inference, Archives of Environ-mental Health 48, 336-342.

Ostro, B. (1994): Estimating the HealthEffects of Air Pollutants. A Method withan Application to Jakarta, Working Paper1301, PRDPE Division, World Bank,Washington DC.

Ostro, B. (1995): Addressing Uncertaintiesin the Quantitative Estimation of AirPollution Health Effects, Paper presentedat the Workshop on the External Costs ofEnergy, EU/IEA/OECD, 30. -31. Jan.1995.

Ostro, B. and S. Rothschild (1989): AirPollution and Acute Respiratory Mor-bidity: An Observational Study of MultiplePollutants, Environmental Research 50,238-247.

Ostro, B., J.M. Sanchez C., C. Aranda andG. Eskeland (1995): Air Pollution andMortality - Results from Santiago, Chile,Working Paper 1453, PRDPE Division,World Bank, Washington DC.

PACE (1990): Environmental Costs ofElectricity, Oceana Publications, New York.

Pearce, D. (1995): The Development ofExternality Adders in the United Kingdom,Paper presented at the Workshop on theExternal Costs of Energy, EU/IEA/OECD,30. -31. Jan. 1995.

Pearce, D. and T. Crowards (1996): Parti-culate Matter and Human Health in theUnited Kingdom, Energy Policy 24, 609-620.

Pigou, A. (1932): The Economics of Wel-fare, 4th ed., London.

Plagiannakos, T. and J. Parker (1988): AnAssessment of Air Pollution Effects onHuman Health in Ontario, Ontario Hydro,March.

Pleijel, H. and P. Grennfelt (1995):Estimation of yield loss from ozone - theproblem of the height gradient above afield crop, EMEP Workshop on the Controlof Photochemical Oxidants over Europe,24-27 October 1995, St. Gallen, Switzer-land.

Pope, C.A. III (1991): Respiratory HospitalAdmissions Associated with PM10 Pollutionin Utah, Salt Lake and Cache Valleys,Archives of Environmental Health 46, 90-97.

Pope, C.A. III, J. Schwarz and M.R.Ransom (1992): Daily Mortality and PM10

Pollution in Utah, Salt Lake, and Cache


124

Valleys, Archives of Environmental Health47, 211-217.

Pope, C.A. III, M.J. Thun, M.M.Namboodiri, D.W. Dockery, J.S. Evans,F.E. Speizer and C.W. Heath Jr. (1995):Particulate Air Pollution as a Predictor ofMortality in a Prospective Study of U.S.Adults, American Journal of Respiratoryand Critical Care Medicine 151, 669-674.

Portney, P. and J. Mullahy (1986): UrbanAir Quality and Acute Respiratory Disease,Journal of Urban Economics 20, 21-38.

Portney, P. and J. Mullahy (1990): UrbanAir Quality and Chronic RespiratoryDisease, Regional Science and UrbanEconomics 20, 407-418.

Pönkä, A. (1991): Asthma and Low LevelAir Pollution in Helsinki, Archives ofEnvironmental Health 46, 262-270.

Rosendahl, K.E. (1996): Helseeffekter avluftforurensning og virkninger påøkonomisk aktivitet, Reports 96/8,Statistics Norway.

Rosendahl, K.E. (1997): Does ImprovedEnvironmental Policy Enhance EconomicGrowth?, Environmental and ResourceEconomics 9, 3, 341-364.

Rossi, O.V.J., V.L. Kinnula, J. Tienari andE. Huhti (1993): Association of SevereAsthma Attacks with Weather, Pollen, andAir Pollution, Thorax 48, 244-248.

Rowe, R., L. Chestnut and C. Lang (1995):The New York Environmental ExternalitiesCost Study: Summary of Approach andResults, Paper presented at the Workshopon the External Costs of Energy,EU/IEA/OECD, 30-31. Jan. 1995.

Schaug, J., T. Iversen and U. Pedersen(1993): Comparison of measurements andmodel results for airborne sulphur andnitrogen components with kriging,Atmospheric Environment 6, 831-844.

Schwartz, J. (1993a): Particulate AirPollution and Chronic RespiratoryDisease, Environmental Research 62, 7-13.

Schwartz, J. (1993b): Air Pollution andDaily Mortality in Birmingham, Alabama,American Journal of Epidemiology 137,1136-1147.

Schwartz, J., V. Hasselblad and H. Pitcher(1988): Air Pollution and Morbidity: AFurther Analysis of the Los AngelesStudent Nurses Data, Journal of AirPollution Control Association 38, 158-162.

Schwartz, J. and S. Zeger (1990): PassiveSmoking, Air Pollution and Acute Respira-tory Symptoms in a Diary Study ofStudent Nurses, American Review ofRespiratory Disease 141, 62-67.

Schwartz, J., C. Spix, H.E. Wichmann andE. Malin (1991): Air Pollution and AcuteRespiratory Illness in Five GermanCommunities, Environmental Research 56,1-14.

Schwartz, J. and D.W. Dockery (1992):Increased Mortality in PhiladelphiaAssociated with Daily Air PollutionConcentrations, American Review ofRespiratory Disease 145, 600-604.

Seaton, A., W. MacNee, K. Donaldson andD. Godden (1995): Particulate AirPollution and Acute Health Effects, TheLancet 345, 176-178.

SFT (1987): Ytterligere reduksjon avluftforurensningen i Oslo (Further


125

reductions of air pollution in Oslo),Norwegian Pollution Control Authority,Oslo.

SFT (1988): Reduksjon av luftforurens-ning og vegtrafikkstøy i Sarpsborg-/Fredrikstadområdet (Reduction of airpollution and road traffic noise in theSarpsborg/Fredrikstad area), NorwegianPollution Control Authority, Oslo.

SFT (1992): Virkninger av luftforurensningpå helse og miljø. Anbefalte grenseverdier(Effects of air pollution on health and theenvironment. Recommended guidelines),Rapport 92:16, Norwegian PollutionControl Authority, Oslo.

SFT (1994): Monitoring of long-rangepolluted air and precipitation. Annual report1993, State Programme for PollutionMonitoring, Report 583/94, NorwegianPollution Control Authority, Oslo.

Simpson, D., H. Styve and E. Berge(1996): Ozone Concentrations in Norwayand Europe. Modelling the Contributionfrom NOx and NMVOC Emissions, SFT-report 96:01, Norwegian Pollution ControlAuthority, Oslo.

Statistics Norway (1990): Husholdnings-undersøkelsen 1988 (Survey of HousingConditions 1988), Table 6, OfficialStatististics of Norway NOS B 892.

Statistics Norway (1992a): Boforholds-undersøkelsen 1990 (Population andHousing Census 1990), Table 5, 13 and30, Official Statististics of Norway NOS B892.

Statistics Norway (1992b): Landbruks-teljing 1989: Hefte 3, investeringar-maskinar-bygningar (Census of Agricultureand Forestry 1989: Book 3, investments-

machinery-buildings), Table 3.10, OfficialStatististics of Norway NOS C 52.

Statistics Norway (1992c): Arbeidsmarked-statistikk 1991 (Labour Market Statistics1991), Official Statististics of Norway NOSC 20.

Statistics Norway (1993a): Nasjonalregn-skapsstatistikk 1991 (National AccountsStatistics 1991), Official Statististics ofNorway NOS C 83.

Statistics Norway (1993b): Helse-institu-sjoner 1991 (Health institutions 1991),Official Statististics of Norway NOS C 81.

Statistics Norway (1994a): Industri-statistikk 1992 (Manufacturing Statistics1992), Official Statististics of NorwayNOS C 162.

Statistics Norway (1994b): Jordbruks-statistikk 1993 (Agricultural statistics1993), Official Statististics of NorwayNOS C193.

Statistics Norway (1994c): Statistisk årbok1994 (Statistical Yearbook 1994), OfficialStatististics of Norway NOS C 161.

Statistics Norway (1995a): Naturressurserog miljø 1995 (National Resources andEnvironment 1995), Statistical Analyses 6.

Statistics Norway (1995b): Bygge-kostnadsindeks for boliger (Building costindex for dwellings).

Statistics Norway (1995c): Økonomiskeanalyser 6/95.

Statistics Norway (1995d): Pasient-statistikk 1993 (Patient Statistics 1993),Official Statististics of Norway NOS C 231.


126

Statistics Norway (1995e): Dødsårsaker1993 (Causes of Death 1993), OfficialStatististics of Norway NOS C 246.

Statistics Norway (1995f): Befolknings-statistikk 1994 (Population Statistics1994), Official Statististics of Norway NOSC 244.

Statistics Norway (1996a): Statistics fromthe Information Office for External Trade,Oslo (see also Statistics Norway (1996c)and similar publications from previousyears).

Statistics Norway (1996b): StatisticalYearbook 1996, Official Statististics ofNorway NOS C321.

Statistics Norway (1996c): Utenrikshandel1995 (External Trade 1995), OfficialStatististics of Norway NOS C316.

Stirling, A. (1996): Limits to the Value ofExternal Costs, paper presented at theEuroforum Conference on ‘The Com-parative Costs of Energy’, Paris, 19-20June 1996.

Sunyer, J., J.M. Antó, C. Murillo and M.Saez (1991): Effects of Urban AirPollution on Emergency Room Admissionsfor Chronic Obstructive PulmonaryDisease, American Journal of Epidemiology134, 277-286.

Thune-Larsen, H. (1991): Teknologiskeperspektiver for energieffektivitet ogklimagassutslipp i transport 1985-2025(Outlook for energy efficiency andgreenhouse gas emissions in transport1985-2025), TØI-notat 0991/1991,Institute of Transport Economics (TØI),Oslo.

Thurston, G.P., K. Ito, P. Kinney and M.Lippmann (1992): A Multi-Year Study ofAir Pollution and Respiratory HospitalAdmissions in Three New York StateMetropolitan Areas: Results for 1988 and1989 Summer, Journal of Exposure AnalEnvironmental Epidemiology 2, 429-450.

Tørseth, K., L.M. Mortensen and A.G.Hjellbrekke (1996): Mapping of ground-level ozone applying critical levels based onaccumulated dose in excess of 40 ppb, NILUOR 12/96, Norwegian Institute of AirResearch, Lillestrøm, Norway.

Tørseth, K., K.E. Rosendahl, A.C. Hansen,H. Høie and L. Mortensen (1997): CropDamage from Ground Level Ozone, SFT-reports 97:04, Norwegian PollutionControl Authority, Oslo.

Tørseth, K., K.E. Rosendahl, A.C. Hansen,H. Høie and L. Mortensen (1997): Avlings-tap som følge av bakkenært ozon, SFT-rëports 97:02, Norwegian PollutionControl Authority, Oslo.

Törnqvist, M. and L. Ehrenberg (1992):Risk Assessment of Urban Air Pollution,Pharmacogenetics 2, 297-303.

UN-ECE (1994): Critical levels for ozone, aUN-ECE workshop report. Bern, Switzer-land, 1-4 November, 1993. Ed. by J. Fuhrer& B. Achermann. Liebefeld-Bern, FederalStation for Agricultural Chemistry andEnvironmental Hygiene (Schriftenreihe derFAC, 16).

UN-ECE (1996): Critical levels for Ozone inEurope: Testing and Finalizing the Con-cepts, a UN-ECE workshop report, Kuopio,Finland, 15-17 April 1996. Ed. by L.Kärenlampi and L. Skärby. University ofKuopio, Department of Ecology andEnvironmental Science.


127

Vegdirektoratet (1996): Norsk veg- ogvegtrafikkplan 1998-2007 (Norwegianplan of road and road transport 1998-2007), Veileder 4, Vegdirektoratet, Oslo.

Vejdirektoratet (1979): Undersøgelse avuheldsmønsteret på vejstrekninger der erblevet aflastet for trafik (Study of theaccident pattern on roads that are relievedof traffic), Økonomisk-statistisk avdeling,Vejdirektoratet, Copenhagen.

Vickery, W. (1969): Congestion Theoryand Transport Investment, AmericanEconomic Review 59, 251-260.Vitaliano, D. F. and J. Held (1991): Roadaccident external effects: an empiricalassessment, Applied Economics 23, 373-378.

Walker, S.E. (1997): Beregning avpersonvektet årsmiddelkonsentrasjon i Osloav PM2,5, PM10 og NO2 (Calculatingpopulation-weighted annual meanconcentration of PM2,5, PM10 and NO2 inOslo), NILU OR 3/97, Norwegian Instituteof Air Research, Lillestrøm, Norway.

WHO (1995): Update and Revision of theAir Quality Guidelines for Europe(unedited), World Health Organization(WHO), Regional Office for Europe.

Xu, X. and L. Wang (1993): Association ofIndoor and Outdoor Particulate Level withChronic Respiratory Illness, AmericanReview of Respiratory Disease 148, 1516-1522.

Zlatoper, T.J. (1987): Factors AffectingMotor Vehicles Deaths in the U.S.A.: SomeCross-Sectional Evidence, AppliedEconomics 19, 753-761.

Zlatoper, T.J. (1991): The Determinationof Motor Vehicles Deaths in the UnitedStates, Accidents and Prevents 25, 431-436.


129

Table A1. Air quality and climate parameters by region

Region SO2, µg/m3 O3, µg/m3 H+, mg/l Regn m/yr TOW

Halden 5 55 0.040 0.80 0.38Sarpsborg 16 50 0.032 0.88 0.39

Fredrikstad 7 50 0.032 0.79 0.38

Moss 5 55 0.032 0.81 0.40

Bærum 6 45 0.032 0.82 0.39

Asker 5 45 0.032 0.94 0.41

Oslo 5-24 17-51 0.025 0.60 0.32

Drammen 6 43 0.032 0.95 0.34

Porsgrunn 5 55 0.032 0.92 0.33

Skien 14 55 0.032 0.85 0.33

Bamble 3 53 0.032 0.87 0.33

Kristiansand 3 54 0.040 1.38 0.49

Stavanger 6 60 0.032 1.25 0.70

Bergen 7 62 0.020 2.25 0.53

Trondheim 5 55 0.008 0.93 0.41

Tromsø 2 55 0.010 1.03 0.24

Urban south 3 55 0.031 1.03 0.42

Urban north 2 55 0.010 1.03 0.24

Rest of the country 1 55 0.031 1.03 0.42

Appendix A

Appendix to chapter 3: Tables


130

Table A2. Material-mix per type of building from the MOBAK study. Per cent

Galvanised steel

Meanarea

Untreated Treated Aluminium Weath-ering

Copper

Building type in GAB register m2/building

Sh.repl

Sh.maint

Wire Profile Str.-lacq. Painted Untreat* B.l. steel*

Dwellings Small houses (incl. garages) 470 0.6 0.6 0.0 0.0 2.8 1.7 0.1 1.1 0.9 0.1

Block/condominium 2470 1.6 1.6 0.0 10.3 3.6 0.2 4.0 2.7 0.2

Mining/manufacturing 2450 2.0 2.0 0.0 0.0 21.4 2.7 0.4 6.2 4.5 0.4

Office/Comm./Transport Wholesale & retail, for- warding, warehouse, gar.

875 2.8 2.8 0.0 0.0 7.0 5.1 0.3 4.0 2.8 0.2

Office & admin.buiold.,other build.

875 2.8 2.8 0.0 0.0 7.0 5.1 0.3 4.0 2.8 0.2

Hotels & restaurants 875 2.8 2.8 0.0 0.0 7.0 5.1 0.3 4.0 2.8 0.2

Public and private services 875 2.8 2.8 0.0 0.0 7.0 5.1 0.3 4.0 2.8 0.2

Agriculture, Forestry, Fisheryetc.

785 1.6 1.6 0.0 0.0 8.7 3.6 0.1 3.7 2.6 0.0

Other buildings 470 2.0 2.0 0.0 0.0 5.0 3.6 0.2 2.8 2.0 0.2

Infrastructure (per capita) 9.9 0.0 0.0 12.1 7.1 0.0 80.8 0.0 0.0 0.0 0.0

Woodstained/painted

Plaster

Untr. Painted

Con-crete

Roofingfelt

Tiles Stone* Glass* Asbesto-cem.*

Other Total

Dwellings Small houses (incl. garages) 42.5 2.5 5.7 13.2 5.1 11.6 1.5 3.9 5.2 0.9 99.9

Block/condominium 6.6 8.3 23.5 12.6 5.2 8.2 1.6 7.8 1.3 0.8 100.0

Mining/manufacturing 3.6 2.1 2.9 5.9 21.1 8.6 1.5 4.7 3.5 6.6 100.0

Office/Comm./Transport Wholesale & retail, for- warding, warehouse, gar.

18.9 1.3 7.5 8.0 19.0 12.3 1.5 5.3 0.4 0.8 100.0

Office & admin.buiold., other build.

18.9 1.3 7.5 8.0 19.0 12.3 1.5 5.3 0.4 0.8 100.0

Hotels & restaurants 18.9 1.3 7.5 8.0 19.0 12.3 1.5 5.3 0.4 0.8 100.0

Public and private services 18.9 1.3 7.5 8.0 19.0 12.3 1.5 5.3 0.4 0.8 100.0

Agriculture, Forestry, Fisheryetc.

34.2 1.1 0.1 4.9 0.2 9.6 1.0 11.1 9.0 7.0 100.0

Other buildings 23.5 7.0 7.6 13.9 9.2 11.2 1.0 5.3 1.5 2.0 100.0

Infrastructure (per capita) 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 100.0

* These materials are not included in the cost calculations as damage functions are lacking.Abbreviations: Sh.repl = sheeting, replacement, Sh. maint = sheeting, maintenance, str.-lacq. = strip-lacquered, untr. = untreated,Cem. = cement.


131

Table A3. Geographical distribution of building categories. Per cent

Region Small houses/other

Appartm.blocks

Manufac-turing

Office/services

Hotel &rest.

Agri-culture

Infra-structure

Halden 0.7 0.5 1.2 0.5 0.5 0.6 0.6

Sarpsborg 0.3 0.5 1.6 0.2 0.2 0.0 1.1

Fredrikstad 0.6 0.8 1.5 0.5 0.5 0.1 0.6

Moss 0.6 0.7 2.2 0.5 0.6 0.1 0.6

Bærum 1.9 2.8 0.9 3.0 1.4 0.2 2.2

Asker 1.0 0.6 0.3 1.4 0.6 0.2 1.0

Oslo 5.3 46.6 10.6 13.7 11.3 0.1 11.0

Drammen 1.1 2.0 1.6 1.2 1.0 0.2 1.2

Porsgrunn 0.8 0.6 6.2 0.6 0.6 0.1 0.7

Skien 1.2 0.9 1.2 1.0 0.9 0.5 1.1

Bamble 0.4 0.1 3.9 0.2 0.4 0.1 0.3

Kristiansand 1.4 2.1 1.4 1.7 1.6 0.1 1.6

Stavanger 2.5 1.9 2.4 2.3 2.4 0.2 2.4

Bergen 4.1 9.8 2.6 5.9 5.5 0.3 5.1

Trondheim 2.7 5.9 2.3 3.9 3.7 0.7 3.3

Tromsø 1.3 0.8 0.7 1.5 1.7 0.2 1.2

Urban-south 25.1 7.9 20.1 20.9 22.6 32.5 22.3

Urban-north 5.5 1.7 4.4 4.5 4.9 7.1 4.9

Rest of the country 43.8 13.8 35.1 36.4 39.4 56.8 38.9

Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0


132

Table A4. Stocks of materials by area. 1 000m2

Galvanied steel Alumi- Copper

Untreated Treated nium

Area Sh.maint Sh.repl Wire Profile Str.-lacq Painted str.-lacq.

Halden 89 89 31 18 410 397 175 8

Sarpsborg 42 42 56 33 240 460 91 6

Fredrikstad 81 81 32 19 373 382 156 9

Moss 80 80 30 18 396 362 159 9

Bærum 314 314 113 66 1 166 1 395 546 30

Asker 136 136 52 30 464 624 225 13

Oslo 2 059 2 059 575 335 9 693 8 127 4 150 220

Drammen 168 168 63 37 725 769 316 18

Porsgrunn 115 115 38 22 701 470 254 16

Skien 144 144 59 34 599 694 265 14

Bamble 56 56 17 10 374 215 130 8

Kristiansand 208 208 81 48 837 972 376 21

Stavanger 291 291 123 72 1 129 1 420 512 30

Bergen 740 740 265 154 2 986 3 296 1 352 75

Trondheim 493 493 171 100 2 006 2 157 905 48

Tromsø 172 172 65 38 604 783 288 16

Urban-south 3 404 3 404 1 163 678 14 238 15 006 6 394 260

Urban-north 742 742 254 148 3 106 3 273 1 395 57

Rest of the ountry 5 943 5 943 2 031 1 184 24 860 26 201 11 164 454

Total 15 278 15 278 5 219 3 044 64 908 67 002 28 853 1 311

Per cent 1.8 1.8 0.6 0.4 7.6 7.8 3.4 0.2

Alumi-nium

str.-lacq.


133

3

Table A4. Cont.

Area Woodstained/painted

Plaster

Untreat Painted

Con-crete

Roofingfelt

Brick Total Percent

Halden 1 642 170 412 590 504 633 5 170 0.6

Sarpsborg 580 102 261 290 299 281 2 784 0.3

Fredrikstad 1 334 184 481 582 538 573 4 825 0.6

Moss 1 193 171 446 538 548 538 4 568 0.5

Bærum 4 591 626 1 788 2 005 1 957 2 042 16 953 2.0

Asker 2 248 231 650 874 877 925 7 483 0.9

Oslo 18 314 5 986 17 505 12 902 11 211 11 510 104 645 12.2

Drammen 2 481 392 1 074 1 148 1 037 1 119 9 515 1.1

Porsgrunn 1 649 220 519 734 889 769 6 510 0.8

Skien 2 705 293 743 1 023 866 1 051 8 634 1.0

Bamble 768 85 179 319 456 362 3 032 0.4

Kristiansand 3 157 448 1 242 1 398 1 311 1 393 11 701 1.4

Stavanger 5 101 587 1 569 2 067 1 919 2 084 17 195 2.0

Bergen 9 889 1 721 4 918 4 837 4 450 4 698 40 121 4.7

Trondheim 6 740 1 087 3 080 3 143 2 932 3 128 26 482 3.1

Tromsø 2 889 300 832 1 127 1 118 1 184 9 589 1.1

Urban-south 69 036 5 335 12 205 21 423 17 355 24 555 194 456 22.7

Urban-north 15 058 1 164 2 662 4 673 3 785 5 356 42 415 5.0

Rest of the ountry 120 540 9 316 21 311 37 405 30 302 42 874 339 526 39.7

Total 269 914 28 418 71 875 97 078 82 352 105 073 855 603 100.0

Per cent 31.5 3.3 8.4 11.3 9.6 12.3 100.0


134

Table A5. Stocks of materials by type of building for the whole country. 1 000 m2

Galvanised steel Alumi- Copper


Building Sh.maint Sh.repl Wire Profile Str.-lacq. Painted str.-lacq

Small house 1 775 1 775 0 0 9 036 5 487 3 550 322

Apart.block 1 939 1 939 0 0 12 882 4 503 5 003 250

Manufact. 607 607 0 0 6 658 840 1 929 125

Office 6 021 6 021 0 0 15 054 10 968 8 602 431

Hotel 53 53 0 0 133 97 76 4

Service 871 871 0 0 2 177 1 586 1 244 63

Agriculture 2 860 2 860 0 0 16 085 6 654 6 838 0

Other 1 152 1 152 0 0 2 883 2 074 1 611 117

Infrastr. 0 0 5 219 3 044 0 34 793 0 0

Total 15 278 15 278 5 219 3 044 64 908 67 002 28 853 1 311

Woodstained/painted

Plaster

Untreat Painted

Con-crete

Roofingfelt

Brick Total Per cent

Small house 137 165 8 070 18 396 42 601 16 460 37 438 282 077 33.0

Apart.block 8 255 10 381 29 392 15 759 6 504 10 256 107 061 12.5

Manufact. 1 120 653 902 1 836 6 565 2 676 24 516 2.9

Office 40 644 2 796 16 129 17 204 40 859 26 451 191 178 22.3

Hotel 360 25 143 152 362 235 1 694 0.2

Service 5 877 404 2 333 2 487 5 908 3 825 27 645 3.2

Agriculture 63 249 2 057 204 9 034 398 17 743 127 981 15.0

Other 13 245 4 033 4 376 8 005 5 297 6 450 50 395 5.9

Infrastr. 0 0 0 0 0 0 43 056 5.0

Total 269 914 28 418 71 875 97 078 82 352 105 073 855 603 100.0

Alumi-nium

str.-lacq.


135

Table A6. Maintenance prices

Price (Nkr/m2)

Excl. VAT Incl.VAT1

Type of material Treatment Assumption Min Max Average Average

Galvanised steel sheet Maintenance Cleaning + 2 coats of paint 100 200 150 183Galvanised steel sheet Replacement 250 300 275 336Galvanised steel wire Replacement 90 120 105 128Galvanised steel profile Maintenance Cleaning + 2 coats of paint 250 350 300 366Limestone/Cementplaster

Replacement 3-tiered new plaster incl. carveand scaffolding

300 400 350 427

Painted plaster Maintenance Repair + 2 coats of originalpaint

200 300 250 305

Copper roofing Replacement Incl. new felt 400 500 450 549Strip-lacqueredaluminium

Maintenance Cleaning + 2 coats of paint 100 200 150 183

Strip-lacqueredgalvanised steel

Maintenance Cleaning + 2 coats of paint 100 200 150 183

Painted galvanised steel Maintenance Sandblasting + 3 coats of paint 250 350 300 366Roofing felt Replacement New covering, two layers 120 200 160 195Painted/stained wood Maintenance Cleaning + 2 coats of paint 60 100 80 98Brick Maintenance Repair, resealing incl.

scaffolding200 400 300 366

Concrete Maintenance Repair and painting incl.scaffolding

350 700 525 641

1 22 per cent in 1994. Our cost calculations are based on prices incl. VAT.


136

Table A7. Material corrosion costs by region, 1994. 1 000 Nkr



Region Sh.maint Sh.repl Wire Profile Str.-lacq. Painted str.-lacq.

Halden 109 134 18 15 47 467 14 3

Sarpsborg 172 210 107 89 102 2 029 27 7

Fredrikstad 131 161 24 20 63 673 18 5

Moss 103 126 18 15 45 425 12 4

Bærum 387 473 65 54 165 2 049 53 12

Asker 141 173 25 21 53 733 18 4

Oslo 2 421 2 959 313 261 2 277 19 243 671 100

Drammen 171 209 30 25 103 1 131 31 7

Porsgrunn 122 150 19 16 80 553 20 6

Skien 476 581 91 76 221 2 652 68 17

Bamble 30 37 4 4 21 126 5 2

Kristiansand 171 209 31 26 47 571 15 4

Stavanger 894 1 092 177 147 160 2 086 50 16

Bergen 2 120 2 591 354 295 508 5 813 159 50

Trondheim 652 797 105 88 228 2 535 71 19

Tromsø 40 48 7 6 17 230 6 2

Urban-south 2 424 2 962 386 322 808 8 820 250 55

Urban-north 171 209 27 23 88 962 27 7

Rest of the country 508 621 81 67 0 0 0 12

Total 11 243 13 742 1 883 1 569 5 033 51 100 1 515 332

Per cent 5.7 6.9 1.0 0.8 2.5 25.8 0.8 0.2

Alumi-nium

str.lacq.


137

Table A7. Cont.

Region Woodstained/painted

Plaster

Untreat Painted

Con-crete

Roofingfelt


Halden 660 36 140 0 129 0 1 771 0.9

Sarpsborg 874 81 332 929 287 113 5 360 2.7

Fredrikstad 805 58 245 0 206 0 2 410 1.2

Moss 480 36 151 0 140 0 1 555 0.8

Bærum 2 307 166 758 0 625 0 7 114 3.6

Asker 904 49 220 0 224 0 2 565 1.3

Oslo 14 365 2 611 12 207 10 011 5 876 1 087 74 401 37.6

Drammen 1 247 104 455 0 331 0 3 843 1.9

Porsgrunn 663 46 176 0 227 0 2 077 1.1

Skien 3 535 202 819 3 276 718 423 13 153 6.7

Bamble 154 9 30 0 58 0 481 0.2

Kristiansand 635 47 211 0 167 0 2 135 1.1

Stavanger 2 564 156 665 0 612 0 8 620 4.4

Bergen 5 965 547 2 502 0 1 704 0 22 609 11.4

Trondheim 2 710 230 1 045 0 749 0 9 229 4.7

Tromsø 290 16 71 0 71 0 804 0.4

Urban-south 13 880 565 2 070 0 2 215 0 34 758 17.6

Urban-north 1 514 62 226 0 242 0 3 557 1.8

Rest of the ountry 0 0 0 0 0 0 1 288 0.7

Total 53 552 5 020 22 322 14 216 14 581 1 623 197 732 100.0

Per cent 27.1 2.5 11.3 7.2 7.4 0.8 100.0


138

Table A8. Material corrosion costs by type of building, 1994 (1 000 Nkr)



Building Sh.maint Sh. repl Wire Profile Str.-lacq. Painted str.lacq.

Small house 1 177 1 439 0 0 502 3 161 136 63

Apart.block 2 326 2 843 0 0 2 041 7 393 547 97

Manufact. 475 580 0 0 546 714 109 31

Office 4 779 5 841 0 0 1 175 8 867 463 103

Hotel 41 50 0 0 11 87 5 1

Services 656 801 0 0 167 1 263 66 14

Agriculture 1 020 1 247 0 0 420 1 799 123 0

Other 770 941 0 0 169 1 264 65 23

Infrastr. 0 0 1 883 1 569 0 26 552 0 0

Total 11 243 13 742 1 883 1 569 5 033 51 100 1 515 332

Per cent 5.7 6.9 1.0 0.8 2.5 25.8 0.8 0.2

Woodstained/painted

Plaster

Untreat. Painted

Con-crete

Roofingfelt


Small house 27 030 838 3 058 2 869 2 060 317 42 650 21.6

Apart.block 4 636 3 071 13 923 7 307 2 319 597 47 101 23.8

Manufact. 326 100 221 387 1 212 71 4 771 2.4

Office 11 240 407 3 762 2 148 7 175 415 46 375 23.5

Hotel 110 4 37 42 70 8 465 0.2

Services 1 600 58 536 474 1 021 91 6 748 3.4

Agriculture 5 850 100 16 163 23 40 10 803 5.5

Other 2 760 443 769 827 701 84 8 815 4.5

Infrastr. 0 0 0 0 0 0 30 005 15.2

Total 53 552 5 020 22 322 14 216 14 581 1 623 197 732 100.0

Per cent 27.1 2.5 11.3 7.2 7.4 0.8 100.0


139

Table A9. Material corrosion costs distributed by SO2-source and region, 1994 (1 000 Nkr)


Road andconstr.mach.

Rail Aircraft Ships

Halden 1 176 516 13 0 48 - 1 753

Sarpsborg 674 124 1 0 6 4 548 5 352

Fredrikstad 956 103 0 0 14 1 326 2 399

Moss 1 341 34 0 0 7 156 1 538

Bærum 2 113 4 333 2 274 371 - 7 092

Asker 827 1 675 15 0 38 - 2 555

Oslo 27 603 16 410 158 10 30 301 - 74 481

Drammen 1 256 1 905 12 0 664 - 3 837

Porsgrunn 414 93 0 0 41 1 509 2 057

Skien 2 042 1 458 7 16 103 9 501 13 128

Bamble 96 268 - - 108 - 473

Kristiansand 99 72 0 2 60 1 851 2 084

Stavanger 1 815 1 674 9 52 4 925 - 8 475

Bergen 10 194 9 380 37 199 2 498 - 22 307

Trondheim 3 799 652 46 0 915 3 710 9 123

Tromsø 463 227 - 6 86 - 782

Urban-south 7 624 4 485 67 13 406 21 424 34 020

Urban-north 559 327 20 5 14 2 539 3 463

Rest of the country - - - - - - -

Total 63 051 43 735 386 577 40 606 46 565 194 920


140

Table B1 and table B2 below contain an overview of the dose-response functions takenfrom international literature and applied in this study. The functions build onepidemiological studies in various parts of the world. Uncertainty intervals are statedwhere they exist. They are based either on comparison of several studies (e.g. formortality), or on standard deviations in the study/studies undertaken. Hence they arenot uncertainty intervals in the strict sense, and the uncertainty is compounded in theapplications in Norway. Construction of such uncertainty intervals for Norway is notpossible at present.

Table B2. Dose-response functions for NO2

Change in health effect Coefficient estimates

(per change in mean annualconc. of NO2 (µg/m3))

Source

(source of uncertainty intervalsin parentheses)

Acute effects:

N1. Risk of hospital admissionsbecause of asthma

1.5 per cent Pönkä (1991)

N2. Risk of hospital admissions dueto croup

0.4 per cent Schwartz et al. (1991)

N3. No. of days with respiratorysymptoms per person per year

0.009[0.0013 ; 0.005]

Schwartz and Zeger (1990)(ORNL/RFF (1994))

Appendix B

Appendix to chapter 5: Overview of dose-responsefunctions for health effects

Table B1. Dose-response functions for particulates

Change in health effect Coefficient estimates(per change in mean annualconc. of PM10 (µg/m3))

Source(source of uncertainty intervals inparentheses)

Acute effects:P1.Premature mortality per 1 000

deaths0.96[0.63 ; 1.30]

Ostro (1993) - consensus estimate

P2.Restricted activity days perperson per year

0.058[0.036 ; 0.090]

Ostro (1987) (ORNL/FF (1994))

P3.No. of days with respiratorysymptoms per person per year

0.18[0.09 ; 0.27]

Krupnick et al. (1990) (Ostro (1994))

P4.Annual respiratory hospital ad-missions per 100 000 persons

3.6[1.2 ; 10.2]

Based on various studies (Pope (1991);Plagiannakos and Parker (1988), Sunyeret al. (1991))

Chronic effects:P5.Mortality risk (associated with

dying at a particular age)0.65 per cent[0.46 ; 0.91]

WHO (1995) - combination of Pope et al.(1995) and Dockery et al. (1993)

P6.Risk of chronic obstructivepulmonary disease (COPD)

1.1 per cent[0.5 ; 1.7]

Abbey et al. (1993) (Rowe et al. (1995))


141

Table B3 and table B4 give an overview of functions we have arrived at on the basis ofthe dose-response functions in and, respectively. They are primarily functions associatedwith economic activity. The coefficient estimates apply to Norway in general, withestimates for Oslo in parentheses.

Table B3. Functions for change in economic activity etc., associated with changed particulateconcentration

Change in activity etc. Coefficient estimates(per change in mean annualconc. of PM10 (µg/m3))

Source(refers to dose-responsefunctions in table B1)

Acute effects:Changed supply of labour (short-term sicknessabsence and reduced productivity)

-0.011 per cent(of total supply of labour)

P2

Changed supply of labour (reduced productivity) -0.0021 per cent(of total supply of labour)

P3

No. of bed-days in hospital per 100 000population per year

21(32)

P4

Public hospital expenditure per capita per year Nkr 0.76(Nkr 1.12)

P4

Chronic effects:Life expectancy (women) -0.059 P5Life expectancy (men) -0.064 P5No. of bed-days in hospital per 100 000population per year

31(42)

P6

Public hospital expenditure per capita per year Nkr 1.1(Nkr 1.5)

P6

Changed supply of labour; person-hours (long-term sickness absence) -6.9*102*α P6Changed supply of labour; person-hours(rehabilitation) -3.8*103*α P6Changed supply of labour; person-hours(disabled) -6.9*104*α P6

α = share of Norwegian population that is exposed (α=0.11 for Oslo).


142

Table B4. Functions for change in economic activity etc., associated with changed NO2

concentration

Change in activity etc. Coefficient estimates(per change in mean annualconc. of NO2 (µg/m3))

Source(refers to dose-responsefunctions in table B2)

Acute effects:Changed supply of labour (reduced productivity) -0.0015 per cent

(of total supply of labour)N3

No. of bed-days in hospital per 100 000population per year (asthma)

13(14)

N1

Public hospital expenditure per capita per year(asthma)

Nkr 0.47(Nkr 0.48)

N1

No. of bed-days in hospital per100 000population per year (croup) 0.5 N2Public hospital expenditure per capita per year(croup) Nkr 0.02 N2


143

Equation (6.7) in chapter 6 describes the impact of traffic volume and traffic density onthe number of person injuries in road traffic, St. The parameters in this equation arebased on information provided by Fridstrøm and Bjørnskau (1989). Below we show howparameters for our model are developed from their estimates.

Fridstrøm and Bjørnskau (1989) estimated the impact of different variables on thenumber of person injuries. The estimations were based on microeconomic data fromdifferent Norwegian counties and different months. As the relation estimated were onrelative form with constant elasticities and there were only minor differences betweencounties, we use the relation on a national level by scaling up.

Fridstrøm and Bjørnskau estimated the following equation

(C1) ln ln ln ln lnS = k + B + BUSKM + ( CON - CON )t t t t t 0α σ β

whereSt = number of persons injuredBt = consumption of gasoline - applied as proxy for total driving distance in

kilometres by gasoline and automotive diesel fuelled vehicles (except buses)BUSKMt= driving distance (kilometres) for busesCONt = traffic density (congestion) indicator.kt = vector of other variables (climatic and demographic variables, consumption

of alcoholic and a scale constant).

α, σ and β are the parameters estimated by Fridstrøm and Bjørnskau (1989). Theestimated values are:

α = 0.812σ = 0.136β = 0.284

The indicator CONt is expressed by

(C2) tt

tCON

BROADS

=

where

ROADSt = total kilometres of national and county roads.

Appendix C

Appendix to chapter 6: The relation between trafficvolume, traffic density and traffic injuries


144

The road length seems to increase over time, but at a decreasing rate. Thus we havefitted the following curve to historical data on road lengths and the expected increase inroad length 1994-1997 according to the Norwegian Road Plan:

(C3) t 1965ROADS = 58495 + 8769.4 (t - t )ln

The relation (C1) can be written in the form

(C4) t t t tt

0S = K B BUSKM

CON

CON⋅ ⋅ ⋅

α σβ

where ln Kt = kt

As gasoline consumption is used as proxy for the traffic activity in Fridstrøm andBjørnskau's study, α is interpreted as the elasticity of number of person injuries withrespect to total road traffic activity except bus traffic. We therefore assume that no othervariables than gasoline describes the effect of traffic activity by diesel vehicles (otherthan buses). Thus, we rewrite relation (C4) as follows

(C5) ( )S K e KM KM BUSKMCON

CONtt

tB

tDEB

tt= ⋅ ⋅ + ⋅ ⋅

⋅ε α σ

β

0

where (c2) is changed to

(C6)( )

CONKM KM

ROADSttB

tD

t

=+

KMBt = kilometres driven by gasoline vehicles

KMDEBt = kilometres driven by diesel vehicles except buses

KMDt = kilometres driven by all diesel vehicles

We have also split Kt from relation (C4) into K⋅eεt, which represents the impact of climateand demography variables and consumption of alcoholic, as well as a scale constant. K iskept at its base year level during the simulations. ε is a parameter representing trends inthe constant variables, preliminary set equal to zero in our simulations.

We want to relate the number of injuries to use of gasoline and diesel, variables whichare determined in the CGE model (MSG-EE). We do this by explicitly taking into accountthe technological change affecting average mileage per fuel use over time.

Assuming annual rates of energy saving of θ and ω for gasoline and diesel, respectively,and a constant relation between KMt

DEB and BUSKMt given by γ/(1-γ), the relationsbetween vehicle kilometres and fuel use can be written as:


145

(C7) KM M B etB B

tt= 0

θ

(C8) tDEB

tD

0D

ttKM = KM = M D eγ γ ω⋅ ⋅

(C9) t tD

0D

ttBUSKM = (1- ) KM = (1- ) M D eγ γ ω⋅ ⋅

θ = rate of technological change, i.e. increase in fuel efficiency, for gasoline carsω = rate of technological change, i.e. increase in fuel efficiency, for diesel vehiclesM0

B = kilometres per litre gasoline in the base yearM0

D = kilometres per litre diesel in the base yearDt = total diesel consumption

Thus equations (C5) - (C9) give the relations between total use of diesel and gasolineand the number of persons injured.


146

74 H. Valen, B. Aardal, G. Vogt: Endringog kontinuitet Stortingsvalget 1989.1990. 172s. 100 kr. ISBN 82-537-2963-4

75 O. Aukrust: Økonomisk forskning ogdebatt. Economic research and debate.Utvalgte artikler 1942 1989. 1990.383s. 125 kr. ISBN 82-537-2984-7

76 G. Haraldsen, H. Kitterød: Døgnetrundt. Tidsbruk og tidsorganisering1970 90. Tidsnyttingsundersøkelsene.1992. 185s. 189 kr. ISBN 82-537-3639-8

77 J. E. Lystad: Norsk hotellnæring 19501990. 1992. 174s. 115 kr. ISBN 82-537-3677-0

78 O. Ljones, B. Moen, L. Østby (red.):Mennesker og modeller: Livsløp ogkryssløp. 1992. 336s. 165 kr. ISBN 82-537-3699-1

79 I. Gabrielsen: Det norske skatte-systemet 1992. The Norwegian TaxSystem. 1992. 175s. 115 kr. ISBN 82-537-3728-9

80 Einar Bowitz: Offentlige stønader tilhusholdninger En økonometriskundersøkelse og modellanalyse. 1992-119s. 100 kr. ISBN 82-537-3785-8

81 Svein Blom, Turid Noack og LarsØstby: Giftermål og barn – bedre sentenn aldri? 1993-167s. 115 kr. ISBN82-537-3808-0

82 Rolf Aaberge, Tom Wennemo:Inntektsulikhet og inntektsmobilitet i

Norge 1986-1990. 1993-46s. 90 kr.ISBN 82-537-3911-7

83 Ingvild Svendsen: Empirical Tests ofthe Formation of Expectations – ASurvey of Methods and Results. 1993-52s. 75 kr. ISBN 82-537-3948-6

84 Bjørn E. Naug: En økonometriskanalyse av utviklingen i importandel-ene for industrivarer 1968-1990. AnEconometric Analysis of the Develop-ment of Manufacturing Import Shares1968-1990. 1994-78s. 95 kr. ISBN 82-537-3955-9

85 Einar Bowitz og Ådne Cappelen: Pris-dannelse og faktoretterspørsel i –norske næringer. Price Formation andFactor Demand in NorwegianIndustries. 1994-177s. 125 kr. ISBN82-537-4024-7.

86 Klaus Mohn: Modelling RegionalProducer Behaviour − A Survey.Modellering av regional produsentatferd− En litteraturoversikt. 1994-71s. 95kr. ISBN 82-537-4042-5.

87 Knut A. Magnussen: Old-Age Pensions,Retirement Behaviour and PersonalSaving. A Discussion of the Literature.Alderspensjon, pensjoneringsatferd ogprivat sparing. En diskusjon av littera-turen. 1994-69s. 95 kr. ISBN 82-537-4050-6.

88 Klaus Mohn, Lasse S. Stambøl og KnutØ. Sørensen: Regional analyse avarbeidsmarked og demografi – Driv-krefter og utviklingstrekk belyst vedmodellsystemet REGARD. Regional

De sist utgitte i serien Sosiale og økonomiske studierRecent publications in the series Social and Economic Studies

Social and Economic Studies 99 Social Costs of Air Pollution

147

Analysis of Labour Market and Demo-graphy with the Model – REGARD.1994-165s. 125 kr. ISBN 82-537-4082-4.

89 Nils Martin Stølen: Wage Formationand the Macroeconomic Functioningof the Norwegian Labour Market.Lønnsdannelse og den makroøkono-miske funksjonsmåten til det norskearbeidsmarkedet. 1995-306s. 180 kr.ISBN 82-537-4141-3.

90 Øystein Kravdal: SociodemographicStudies of Fertility and Divorce inNorway with Emphasis on theImportance of Economic Factors.Sosiodemografiske studier avfruktbarhet og skilsmisse i Norge medvekt på betydningen av økonomiskefaktorer. 1994-267s. 155 kr. ISBN82-537-4088-3.

91 Tom Kornstad: Empirical Life CycleModels of Labour Supply andConsumption. Empiriske livsløps-modeller for arbeidstilbud og konsum.1995-115s. 110 kr. ISBN82-537-4166-9.

92 H.C. Bjørnland: Trends, Cycles andMeasures of Persistence in theNorwegian Economy. Trender,konjunktursvingninger og varighet avsjokk i norsk økonomi. 1995. 109s.110 kr. ISBN 82-537-4220-7

93 Å. Cappelen, R. Choudhury, T. Eika:Petroleumsvirksomheten og norskøkonomi 1973 1993. The Oil Industryand the Norwegian Economy 19731993. 1996. 128s. 110 kr. ISBN 82-537-4287-8

94 K.O. Aarbu, B. Lian: Skattereformenog delingsmodellen: En empirisk

analyse. The Norwegian tax reform andthe capital income imputation method:An empirical analysis. 1996. 94s. 95kr. ISBN 82-537-4297-5

95 T.J. Klette, A. Mathiassen: Vekst ogfall blant norske industribedrifter: Omnyetablering, nedlegging og omstillingGrowth and turnover amongNorwegian manufacturing plants.1996. 112s. 110 kr. ISBN 82-537-4298-3

96 K.H. Alfsen, T. Bye, E. Holmøy (eds.):MSG EE: An Applied GeneralEquilibrium Model for Energy andEnvironmental Analyses. MSG EE: Enanvendt generell likevektsmodell forenergi og miljøanalyser. 1996. 171s.125 kr. ISBN 82-537-4342-4

97 A. Barstad: Store byer, liten velferd?Om segregasjon og ulikhet i norskestorbyer. Big Cities, Little Welfare?Segregation and Inequality inNorwegian Cities. 1997. 153s. 125 kr.ISBN 82-537-4402-1

98 T.O. Thoresen: Mikrosimulering ipraksis. Analyser av endring ioffentlige overføringer tilbarnefamilier. Tax Benefit Model inUse. Analysing Changes in the PublicPolicy towards Families with Children.1998. 102s. 135 kr. ISBN 82-537-4527-3

99 K.E. Rosendahl (ed.): Social Costs ofAir Pollution and Fossil Fuel Use – AMacroeconomic Approach. Samfunns-økonomiske kostnader av luftforurens-ning og fossile brensler – En makro-økonomisk tilnærming. 1998. 147s.135 kr. ISBN 82-537-4542-7

Social Costs of Air Pollution and Fossil Fuel Use – A...

Documents

Transcript of Social Costs of Air Pollution and Fossil Fuel Use – A...