Input-output analysis and pollutant emissions in France.

This paper deals with the principle of pollutant emissions defined by Leontief in 1971, based on a fixed coefficient model. I have tested the plausibility of this model by attempting to replicate data on French emissions of S|O.sub.2~ and N|O.sub.x~ by combustion and processes.

INTRODUCTION

Input-output

analysis was widely applied to environmental problems during the 1970s, but the technique was more or less abandoned at the beginning of the 1980s. This date coincides with the economic crisis sparked by the second oil shock, which pushed environmental problems into the background. Environmental concerns came again to the fore in the middle of the 1980s, chiefly in connection with the greenhouse effect and the hole in the ozone layer. A prerequisite for the study of the effects of pollution is an in-depth knowledge of emissions. There has been renewed interest in input-output analysis, evidenced by the research work devoted to it, which focuses on:

- the inclusion of environmental expenditures in input-output tables,

- the economic impact of environmental protection policies,

- the emission of pollutants.

This study will be devoted to the third subject. Knowledge of emissions is an essential step in any modeling technique that deals with the transportation or impacts of pollutants. Making forecasts about the emission of air pollutants can be viewed as a first step towards the development of a more complete strategy for future environmental problems.

I shall therefore review the principle of pollutant emissions defined by Leontief in 1971, based on a fixed-coefficients model, and test its validity with regard to S|O.sub.2~ and N|O.sub.x~ emissions in French industry.

By necessity this approach will tend to treat technical issues in a simplified and artificial manner and should ideally be complemented with more detailed studies of technical topics.

THE INPUT-OUTPUT MODEL IN THEORY AND IN PRACTICE

Input-output - or interindustrial - analysis is based on input-output tables in the national accounts. I will first review the basic principles of the method. I will then illustrate why it is necessary to adapt the French National Accounts input-output table in order to test the pollutant emission principle in input-output tables extended to nature (Huriot, 1980).

Basic Principles of Input-Output Analysis

Recall the main assumptions and equations of the basic model of input-output analysis. Leontief's elementary model is based on a "commodity by commodity" table which supposes that there is a one-to-one correspondence between commodities and industries. The model uses fixed technical coefficients so that the value of each input of an industry is strictly proportional to the value of the output of this industry.

Let us suppose that there are I industries i, j = 1.....I.

Let |X.sub.i~ be the output of industry i, |X.sub.ij~ the input of commodity i used by industry j, |Y.sub.i~ the final demand of commodity i and |a.sub.ij~ = |X.sub.ij~/|X.sub.j~ the fixed technical coefficient defined above.

Leontief's model is based on the system:

|X.sub.i~ = ||summation~.sub.j~ |a.sub.ij~|X.sub.j~ + |Y.sub.i~

If X and Y are respectively the column vector of outputs and the column vector of final demands, if A is the square matrix of the technical coefficients and I the unit matrix with the same dimensions as A, the above system becomes:

X - AX = Y

and its solution is the well-known Leontief equation:

X = |(I-A).sup.-1~Y

The application of the model to the French case requires some adaptation of the available data.

Adaptation of National Accounts to the Theoretical Framework

The French National Accounts input-output table is not directly adaptable to the matrix solution system of input-output analysis, chiefly because it is not square and therefore a problem of evaluation arises. The table is not square because there are empty rows (services and public sector) and because there is no equivalent for a column in a row (hypothetical unit sector). The revenue-expenditure balance of the input-output table is written as follows:

Production + Imports + VAT + Customs duties + Commercial margins

=

Intermediate consumption + Final consumption + Gross fixed capital formation + |delta~ stocks + Exports

In order to satisfactorily perform the traditional input-output analysis, a balance of the following type has to be reached:

Production + Imports

=

Intermediate consumption + Final consumption + Gross fixed capital formation + |delta~ stocks + Exports

To accomplish this we used tables derived from the NOUBA database(1) (Input-Output Tables without VAT) that were processed according to a method advocated by INSEE(2) (issue to appear in the INSEE METHODES series).

With the general theory discussed and the data defined I now turn to look at a pollutant emission input-output model which can be applied to France's industrial sector.

A POLLUTANT EMISSION INPUT-OUTPUT MODEL FOR FRENCH INDUSTRY

In recent years, input-output models of atmospheric pollutant emissions have been estimated in many countries across Europe (for example, R.I.M. in Netherlands, D.I.V.A. in France(3)). The expansion of such models makes vitally important the need to study emissions coefficients upon which the analysis is based.

The Input-Output Approach Extended to Pollution

Pollutant emissions in these models are based on a simple assumption: each industry generates emissions in fixed proportion to its output. Let R be the matrix of these pollution coefficients. Each coefficient |r.sub.pj~ represents the output of pollutant p generated by output unit of industry j (amount of pollutant p in physical unit / output of industry j in currency unit). The approach applied to industrial production can be extended to final consumption (Leontief, 1971). This matrix coincides perfectly with the principle of the input-output model:

E = R |(I-A).sup.-1~ Y

where E is the pollutant vector

R is the pollution coefficient matrix.

Many studies of emission projections have been based on this equation (Forsund and Stram, 1974). By analogy, the fixed technical coefficient assumption is applied to the emission |r.sub.pj~. We propose to test this assumption in relation to French industry.

Industry Emissions of S|O.sub.2~ and N|O.sub.x~ in 1985

Basic data come from the European Community's information on air borne emissions. The purpose of the CORINAIR(4) is to assess the emissions of certain pollutants in different EEC countries. We have used the data, arranged according to the CITEPA(5) nomenclature, to draw up the following table:

Table 1. S|O.sub.2~ and N|O.sub.x~ emissions in France - 1985
(Kt)
                                   S|O.sub.2~   N|O.sub.x~
Industry + agriculture(*)             365          115
Energy transformation(*)              111           16
Urban heating(*)                       71           15
Electricity generation(*)             436          159
Processes(**)                         194          111
* Emissions produced by combustion
** Emissions produced by manufacturing processes and not by
combustion, e.g. nitric acid production.
Source: From CITEPA, Etudes documentaires no. 96

Based on these data, disaggregated in sectors, an emission coefficient matrix can be constructed as follows.

There are actually two matrices of emissions coefficients. Two types of emissions coefficient have been developed to account for emissions produced by combustion and by manufacturing processes. The principle for determining the basic coefficient is nevertheless the same:

|Mathematical Expression Omitted~

|Mathematical Expression Omitted~

|r.sup.c~ = combustion emission coefficient

|r.sup.pr~ = process emission coefficient

p = S|o.sub.2~, N|O.sub.x~

We now have the practical means of testing the fixed emission coefficient assumption made in most empirical work.

TEST OF THE FIXED EMISSION COEFFICIENT ASSUMPTION

To test the assumption of fixed emission coefficients I have calculated emissions over a known period (1985-1989), using emission coefficient matrices. The results over this very short projection period lead us to determine the reason why the coefficients were so obviously not fixed.

Emission Projections from 1986 to 1989

The period of reference is restricted both by the available emissions data and by the Input-Output Tables published by INSEE. If we assume that the emissions coefficients determined in 1985 are fixed and if we know industrial production (by sector) for the period 1986 to 1989, we can obtain the calculated emissions for the period (Table 2).

Table 2. Calculated emissions (in Kt) of S|O.sub.2~ and
N|O.sub.x~ by industry, France, 1986-1989
               S|O.sub.2~                N|O.sub.x~
         combustion   processes    combustion   processes
1986       1070         192           310         109
1987       1091         185           307         108
1988       1123         189           325         112
1989       1150         188           334         113

We will now compare these projections with the actual emissions estimated by CITEPA(6), which serve as a reference (called the "inventory CITEPA").

Test of the Fixed Emission Coefficient Assumption

Figures 1 and 2 provide a parallel representation, for combustion only, of S|O.sub.2~ and N|O.sub.x~ emissions for all the industrial sectors obtained by the input-output analysis method and those published by CITEPA. The divergence between the two curves shows that even over a short period the fixed coefficient assumption is too strong and leads to an incorrect estimation of emissions.

The comparison between process emissions (Figures 3 and 4) give the same result, although there is some uncertainty about the reliability of the CITEPA figures. It would appear that the CITEPA estimates have not taken into account the recovery of French industry over the period 1988-1989. This would explain the drop in the S|O.sub.2~ emissions curve and the steady N|O.sub.x~ emissions curve, since no new technology could have led to a reduction in process emissions during this period.

It seems difficult to come to any conclusion on the validity of the fixed emission coefficient assumption where processes (|r.sup.pr~) are concerned. On the other hand, it has to be admitted that the assumption is too strong where combustion emissions (|r.sup.c~) are concerned. There are several possible explanations for this.

Elements Indicating Non-Fixity of the Emission Coefficient

A number of analyses have used fixed emission coefficients. Cumberland (1974) considers fixed coefficients to be a "restriction" of the method. However he admits that it does not apply if information is available for adjusting the coefficients. S|O.sub.2~ and N|O.sub.x~ are mainly related to combustion of fossil fuels. Stationary combustion sources are the largest source of S|O.sub.2~ emissions.

Three hypotheses may be put forward to explain the non-fixity of the coefficients:

Technologies and Legislation

These two elements directly influence the emission coefficient. Technologies for the control of emissions are available and improvements have been achieved in coal-cleaning processes, the fuel combustion process and post-combustion cleaning of exhaust gases.

Legislation in Member States and in the European Community as a whole has been adopted to limit S|O.sub.2~ and N|O.sub.x~ emissions (sulphur content of fuels, emission limits). Within the Community a major breakthrough on stationary sources has been achieved with the adoption of the council directive on the limitation of emissions of certain pollutants into the air from large combustion installations above 50 MW thermal (88/609/EEC). But this directive was translated into French law only in 1990.

With respect to this study, however, these two elements cannot explain the difference in projections. There have been no particular technologies and no new specification concerning energy products during this period.

Energy Intensity

If one assumes that the emission coefficient for combustion is fixed, then energy intensity (energy consumed / output of industry) must consequently be taken as fixed as well. However, energy intensity in industry decreased between 1985 and 1989.

A decrease in energy intensity contributes to environmental protection. With this in mind, the Commission to the Council on "Energy and the Environment" has proposed a coherent global programme on energy efficiency improvements and energy conservation covering both the generation and use sectors (S.A.V.E. programme).

The Structure of the Energy Balance

The coefficient established in 1985 for each sector is the weighted average of the emission coefficient specific to each fuel multiplied by the energy consumed, divided by the output of the sector.

If one assumes that the emission coefficient is fixed, the relative proportion of each fuel must be taken as constant. However, the trend in the structure of the industrial energy balance (excluding the energy branches) shows a relative reduction in pollutant products (petroleum products and solid fuels) and an increase in the share of electricity.

The switch from high to low energy emitting fuels is an efficient energy measure to reduce emissions into the air, especially for S|O.sub.2~. In comparison with solid fuels or oil, natural gas is the cleanest fossil fuel which does not produce any ash, dust or smoke and only negligible amounts of S|O.sub.2~. Gas combustion only produces significant N|O.sub.x~ emissions which are lower than N|O.sub.x~ emissions from coal or oil and which can be controlled technologically.

For the European Commission on "Energy and the Environment" the increased use of renewable energies and less environmentally damaging fuels, such as natural gas or biomass-based fuels, low-sulphur oil and coal, and electricity produced from non-fossil fuel generating plants are important elements of an environmentally friendly energy strategy.

There are many variables that may explain the non-fixity of the emissions coefficients. This does not mean that input-output analysis cannot be used as a means of calculating emissions. The analysis needs to be adapted; TABULAR DATA OMITTED perhaps coefficients should be proportional not to the sector output but to intermediate consumptions, which are sources of emissions (Breuil-Houllier, 1992).

Until now emissions from industries due to stationary combustion were associated with the output of the industry. In fact, it is possible to associate these emissions with the use of fuel oils, and process emissions with demand for intermediate materials (Table 4).

Emissions factors can be calibrated in a base year, and projected by taking into account the effects of many factors including planned and implemented environmental control policies like emission standards, limits on sulphur content of heating oils, and direct regulation of emissions from specified firms.

Table 4. Emission Sources Indicators
types of source       indicators
1) stationary         fuel consumption
2) process            raw materials, intermediate goods

CONCLUSION

Input-output analysis applied to the environment is again enjoying a certain amount of popularity. Nevertheless, little work has been done on pollutant emissions themselves. As we have pointed out, this is not a static area, as many changing factors influence the rate and magnitude of pollutant emissions.

Using a fixed coefficient of emissions by industry does not closely replicate existing data on actual emissions. This does not mean that fixed coefficients analysis is not useful, just that care needs to be taken in choosing the choice of the indicator. We have demonstrated that for combustion emissions, using the output of industry as the indicator is not sensible and that other factors (i.e. intermediate materials) need to be considered.

The framework described above is still partial, with no treatment of other environmental problems like water pollution or generation of hazardous waste. Without a multi-pollutant framework, the danger of transferring air pollution problems to other areas exists.

The year 1992 is the start of a period of rapidly changing regulations concerning pollution and of generalized use of pollution control devices, both of which affect emission coefficient stability. These elements, together with those traditionally connected with emissions, i.e. economic activity and energy consumption, need to be included in a procedure designed to update emissions coefficients.

* This work is included in a larger study under the direction of Professor J.M Huriot, Universite de Bourgogne, France.

1. The NOUBA (nouvelle base) database includes data on: 1) margin contents on intermediate consumption; 2) margin contents on final demand; 3) intermediate consumption in imported goods; 4) final demand in imported goods; and 5) input-output tables without VAT.

2. Institut National de la Statistique et des Etudes Economiques. In order to obtain a square matrix, INSEE modifies the original input-output table as follows:

1) public goods are integrated in the final demand of public administration;

2) territorial correction is abolished; and

3) the "unite fictive" branch is transferred to final demand of private administration.

3. "One of the most important responses by economists to environmental problems has been to extend the application of input-output models to the examination of relationships between economic activity and the emission of pollutant materials". (Cumberland-Stram, 1976, page 368, quoted by Forsund, 1985).

4. CORINE: Coordination of Information on the Environment - EEC; CORINAIR: the air-related section of CORINE.

5. Centre Interprofessionel Technique d'Etude de la Pollution Atmospherique, 3 rue H. Heine, 75016 Paris.

6. CITEPA bases its calculation on emissions estimates as follows: M/AN = A/AN * M/A where A is activity, e.g. the production of electric power in MWh or coal consumption in tons, or sulphuric acid production, etc. and M/A is the weight of pollutant emitted per unit of activity in Kg/t, for example. The emission factor takes into account the means employed to reduce emissions.

REFERENCES

Breuil, J.M. and C. Houllier (1992), "Inventaire d'emission prospectif: une approche technique ou economique," Econometrie de l'environnement, 33eme colloque international de l'A.E.A.

C.E.E. (1990), "proposition de la Commission sur l'energie et l'environnement", mars, n|degrees~329 (document dossier), 19 p.

Cumberland, J.H. (1974), "Energy, environment and social science research," Energy and environment-Methods to analyse the long term relationship, OCDE, Paris, 1-50.

Forsund, F.R. (1985), "Input-output models, national economic models and the environment", chap. 8, pp. 325..341, in: Handbook of natural resource and energy economics, vol. 1, edited by A.V. Kneese and J.L. Sweeny, Elsevier Science Publishers B.V.

Forsund, F.R. and S. Stram (1974), "Industrial structure, growth and residual flows," The management of water quality and the environment, J.G. Rothenberg et I.G. Heggie (eds), London, Mac Millan, pp. 21..69.

Huriot, J.M. (1980), "Economie et nature. Essai sur l'elargissement de l'analyse entrees-sorties," Collections de l'I.M.E., Dijon, n|degrees~23.

Leontief, W. (1966), Input-output economics, New Oxford University Press.

Leontief, W. (1971), "L'environnement et la structure economique," Analyse et prevision, mars, 253-276.