LIBRARY OF THE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN 510.84 T&63c no.Gl-70* AUG 31976 1 he person charging this material is re- sponsible for its return to the library from which it was withdrawn on or before the Latest Date stamped below. Theft, mutilation, and underlining of books are reasons for disciplinary action and may result in dismissal from the University. UNIVERSITY OF ILLINOIS LIBRARY AT URBANA-CHAMPAIGN a .ninctl! jaw. - e apr 8 m kr PHOTO REPRODUCTION OCT 2 2 WO PHOTO REPRODUCTION NOV 16^ gHjrowEwooucrw* oti o i ffn 'BOfO RE OCT EHOIO i?EPJWini.rT ITT tjn i ■ i \ y i 2 7 REC'B ,? gfEROOU 3 REDD L161 — O-1096 Digitized by the Internet Archive in 2012 with funding from University of Illinois Urbana-Champaign http://archive.org/details/energypollutione68folk ENGINt UNIVERSIiy OF iLLi URBANA, ILLINOIS enter for Advanced Computation UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN URBANA. ILLINOIS 61801 CAC Document No. 68 AN ENERGY, POLLUTION AND EMPLOYMENT POLICY MODEL By Hugh Folk and Bruce Hannon February 10, 1973 BIBLIOGRAPHIC DATA SHEET 1. Report No. UIUC-CAC-DN-73-68 3. Recipient's Accession No. 4. Title and Subtitle An Energy, Pollution and Employment Policy Model 5- Repott Date February 1973 7. Author(s) Hugh Folk and Bruce Harmon 8. Performing Organization Rept. No CAC Doc. No. 68 9. Performing Organization Name and Address Center for Advanced Computation University of Illinois at Urbana-Champaign Urbana, Illinois 6l801 10. Project/Task/Work Unit No. 11. Contract /Grant No. NSF GI-35179X 12. Sponsoring Organization Name and Address National Science Foundation 1800 G Street Washington, D. C. 20301 13. Type of Report & Period Covered Research 14. 15. Supplementary Notes 16. Abstracts A large national linear input-output policy model is being developed at the Center for Advanced Computation in the University of Illinois at Urbana-Champaign. The model contains detailed economic activities showing industry demand generated by expenditure categories so that national budgets or individual "life styles" or scenarios can be evaluated. The 3^7 industry input-output model produces estimates of total demand by industry through manipulation of an activities vector. For example, energy-output matrices in 3^7 sector detail show demand for coal, crude oil, refined petroleum, electricity, natural gas implied by scenarios. Other matrices estimate employment effects and ten specific pollution components. Thus, employment, energy and pollution consequences of a shift in expenditures, either on a national or individual basis, can be estimated. The model is easily edited to adapt it to specific applications by non- programmers, is remotely accessible through the ARPA Network, and uses innovative and efficient computational facilities. 17. Key Words and Document Analysis. 17a. Descriptors Energy Policy Model (EPM) Industry Input-Output Model Energy Output Matrices Employment Effects Pollution ARPA Network 17b. Identifiers/Open-Ended Terms 17c. COSATI Field/Group 18. Availability Statement No restriction on distribution. Available from National Technical Information Service, Springfield, Virginia 22151 FORM NTIS-3S (REV. 3-72) 19. Security Class (This Report) UNCLASSIFIED. 20. Security Class (This Page UNCLASSIFIED 21. No. of Pages 25 22. Price USCOMM-DC M952-P72 CAC Document No. 68 AN ENERGY, POLLUTION AND EMPLOYMENT POLICY MODEL by Hugh Folk and Bruce Hannon Center for Advanced Computation University of Illinois at Urbana- Champaign Urbana, Illinois, 61801 February 10, 1973 The research was supported in part by the National Science Foundation, Grant No. GI-35179X. ABSTRACT A large national linear input-output policy model is "being developed at the Center for Advanced Computation in the University of Illinois at Urbana, Champaign. The model contains detailed economic activities showing industry demand generated by expenditure categories so that national budgets or individual "life styles" or scenarios can be evaluated. The 367 industry input-output model produces estimates of total demand by industry through manipulation of an activities vector. For example, energy-output matrices in 367 sector detail show demand for coal, crude oil, refined petroleum, electricity, natural gas implied by scenarios. Other matrices estimate employment effects and ten specific pollution components. Thus, employment, energy and pollution consequences of a shift in expenditures, either on a national or individual basis, can he estimated. The model is easily edited to adapt it to specific applications by non-programmers, is remotely accessible through the ARPA Net- work, and uses innovative and efficient computational facilities. / ACKNOWLEDGEMENTS We gratefully acknowledge the assistance of Roger Bezdek, Clark Bullard, David Healy, Robert Herendeen, Al Meyers, Sue Nakagaraa, Toni Prevedell and Janet Spoonamore in preparing this report. TABLE OF CONTENTS Page Introduction 1 I. The Model 2 II. The Modeling System 6 III. Forecasting 8 IV. Preliminary Applications of the Model 11 References 18 LIST OF FIGURES Figure Page 1. Energy, Employment, Pollution Policy Model Under Development "by Energy Research Group at Center for Advanced Computation, University of Illinois k 2. Causal Ordering of Experts 9 3. Total Energy and Employment (Direct & Indirect) Per Dollar Delivered to Final Demand, 1963. Source: Energy Employment Policy Model; CAC, Feb. 1973 • 12 k. Changes in Total Energy and Employment Requirements for a One Billion Dollar Increase in Final Demand from the Noted Industry, Proportionately Absorbed from all Other Industries, 1963. Source: Energy Employment Policy Model-, CAC, Feb. 1973- . • lU 5. Changes in Total Energy and Employment Requirements for a 10$ Increase in Final Demand from the Noted Industry, Proportionately Absorbed from all Other Industries, 1963. Source: Energy- Employment Policy Model-, CAC, Feb. 1973 15 6. Changes in Total Energy and Employment Requirements for a 10$ Increase in Final Demand from the Noted Industry, Pro- portionately Absorbed from all Other Industries, 1963 . An Enlargement of Center Portion of Figure 2. Source: Energy- Employment Policy Model*, CAC, Feb. 1973 l6 INTRODUCTION Energy policy is public policy relating to the production and consump- tion of energy. Because energy is widely used in production, energy policy cannot be divorced from economic policy in general. The model and modeling system presented in this paper are intended to permit examination of the energy consequences of alternative economic policies and the economic consequences of alternative energy policies. The interactions of energy, employment, capital, pollution, and other natural resource requirements are extremely complex. Natural gas shortages have already led to employment cutbacks. Emission standards imposed for pollution control have caused inter-fuel substitution and employment effects and require additional capital expenditures. The task of modeling such a complex system is staggering, and especially so given the fact that it is not simply a matter of combining models for energy, employment, pollution, and capital sectors which have already been developed and tested for the separate sectors. I. The Model The energy policy model (EPM) proposed here grows out of earlier work at the Center for Advanced Computation of the University of Illinois in which Hannon attempted to measure the energy consequences of shifts from non- returnable to returnable beverage containers [1], Folk attempted to measure the employment consequences of the same shift [2] and Herendeen [3] examined the different energy prices of various industries. The problem of determining the impact of a change in spending on a program or commodity is ubiquitous, but relatively little is written about methods. Converting expenditures in program (or object of expenditure) terms to final demand by industry is time-consuming and often arbitrary. The use of input-output theory to convert the final demand pattern by industry to total demand by industry is straightforward, but not trivial. Conversion of total demand by industry to the derivation of employment, energy, capital, pollution, or natural resource requirements from total demand by industry is usually quite difficult. Bezdek and his associates at CAC developed in ERGWORKS a system which accepted a specified vector in final demand by 218 expenditure types, converts it through use of a matrix to a vector of final demands by industry (80 industries), then generates employment for about 200 occupations [h]. This model has been used for occupational forecasts for 1975 and I98O and to test the feasibility of the Urban Coalition's Counterbudget [5]. Babcock [6] developed an Illinois employment model which extends Bezdek 1 s model and can be used for any regional extension of the mode. Parallel to the manpower demand models there are population labor force, migration, enrollment, and occupational mobility models under development which together constitute the STEP I model, a complete supply and demand model for occupational forecasting. The energy, capital, pollution, and other natural resource sectors are formally parallel to the employment models and the general schema for the model is shown in figure 1. The major parts of the model are the "scenario" which is the set of assumptions specifying an economic and legal environment for the model and the stochastic growth model which provides feasible consumption, investment and government expenditure totals for each year to be forecast. The scenario specifies a vector of expenditures by detailed program for each year, q, . The vector q. has m elements representating m different expenditure programs. This number can be indefinitely large. The number used in the 80 industry version of the model is currently 218 activities. The q vector is converted to expenditures for final demand by industry by a program- industry matrix P for each year with as many rows as industries and as many columns as there are programs. A program column of P, represents the distribution of expenditures of final demand by industry on that particular program. The programs need not include only empirical estimates of new programs, but program vectors for new, proposed, or hypothetical programs can be included. The P matrix is one place in the model in which technological change must be forecast. Thus, if Y is final demand by industry, then Y t ■ p t\ « Total direct and indirect output by industry is given by the vector x , and x t = Vt + y t (2) in which A is the input-output direct coefficients matrix. Matrix A^ is another part of the model in which technological change must be forecast. PROGRAM INDUSTRY MATRIX X STOCHASTIC GROWTH MODEL • *\ 1 T GNP BY MAJOR EXPENDITURE CATEGORY VECTOR ir SCENARIO GNP BY DETAILED EXPENDITURE j PROGRAM VECTOR GNP BY INDUSTRY FINAL DEMAND VECTOR X TOTAL DEMAND COEFFICIENTS MATRIX TOTAL ENERGY REQUIREMENTS BY TYPE ENERGY INTENSITY VECTORS OCCUPATIONAL EMPLOYMENT REQUIREMENTS OCCUPATION INTENSITY VECTORS x- x- POLLUTION OUTPUTS BY TYPE POLLUTION- INTENSITY VECTORS X CAPITAL REQUIREMENTS CAPITAL- INTENSITY VECTOR X TOTAI DEMAND VECTOR Figure 1. Energy, Employment, Pollution Policy Model Under Development by Energy Research Group at Center for Advanced Computation, University of Illinoi: It will be possible to consider alternative input vectors, or disaggregate one or more industries to finer detail, or formulate this part of the model as a programming model. Solving (2) for x , we obtain x t = (I - A)" 1 y t (3) Total energy, employment, capital, and pollution requirements can be obtained by premultiplying x, by row vectors which estimate intensity (re- quirements per unit of total output). Thus, for energy, total energy re- quirements for energy type i are E u ■ W 1 - V" 1 p t% 00 J- v. in which e is the energy intensity vector for the i energy type. Similar estimates can be developed analogously for pollution types, occupational employment, capital and natural resource types. If e., is used as a diagonal matrix instead of as a row vector, then it ' energy requirements are obtained as a column vector which presents the energy required in each industry as a consequence of the specified final demand. The intensity vectors represent a third part of the model in which technological change must be forecast. It is known that labor productivity changes considerably more rapidly than input-output coefficients and that energy input-output coefficients have tended to change more than others ["J], So far in our work, only energy and total employment intensity vectors corresponding to the 1963 input-output tables have been completed. Work continues on 1967 energy and employment vectors, and these will be updated to current years. Our research concentrates on updating the energy and employ- ment intensity vectors and not on the input-output matrices. We expect some of the pollution vectors will be derived from work now underway at other research centers. Ultimately, forecasts for each of the components of the model will be made. II. The Modeling System For a policy model to be useful, it must be generally accessible and inexpensive. For this reason we have designed MEASURE (Mathematical and Economic Analysis System for Use in Research and Evaluation) as a computer system on the ARPANET. A user with a few hundred dollars for computer time, a computer terminal, and some time should be able to adapt the EPM (or one of the other models in MEASURE, such as STEP I) to his specific needs (such as simulation of another forecasting model). Today, of course, the use of large data bases of diverse provenance for impact studies or forecasting presents the researcher with serious problems. When government authorities examine the consequences of a proposal (such as the SST, or oil import restrictions or gas price changes) time is of the essence. There is usually too little time to do any serious data collection, and the use of other people's models and data bases is prevented by problems of accessibility, cost, and lack of documentation. As a result, even if government authorities want reliable, state-of-the-art forecasts, they are unable to get them. Very often the models which exist are not conformable with the problem under investigation. Levels of aggregation and assumptions differ, and it is difficult to make appropriate changes in short order. MEASURE has been designed to meet these problems. It uses virtual memory machines for editing and reporting generation and very large and fast computers such as ILLIAC IV and a 360/91 for numerical processing, superior mathematical algorithms , and general modeling and command languages to reduce computational costs by an order of magnitude or more over comparable conventional systems. The ARPANET provides cheap communications and will soon be publicly accessible. There will be extensive on-line documentation of the system, data and models. We can foresee individual researchers preparing complete forecasts in a few days or -weeks which are comparable or superior in every way to those which today cost tens of thousands of dollars and require months of work. We hope that this capability will assist in the development of a more open and more even debate on questions of policy than is possible now. We do not expect there to be a revolution in which politicians hand over the decision-making responsibilities to a cabal of quantitative modellers, but we do expect that modeling of this sort, which is now limited to govern- ment agencies, large corporations, and adventitious or far-sighted academic researchers, can soon be carried out by a much larger community of users. 8 III. Forecasting Estimating past coefficients of the EFM and updating them to the present is difficult enough, but forecasting the model will be even harder. In addition to forecasts of the coefficients in the component matrices of the models, feasible and internally consistent scenarios must be developed. Forecasting the model and developing scenarios will require the use of a large number of persons with highly specialized expertise. Developing a set of consistent final demand by program vectors for each year over a fore- cast period, for instance, will require participation of political, economic, marketing, and technical experts and the use of a substantial battery of subsidiary and specialized models to make sure that the patterns of final demand specified are feasible (could be produced) and are consistent (for instance, provide enough energy consumption for the energy using goods to be produced). It is not now easy to incorporate a large number of expert judgment (or "implicit models") into formal models because of the difficulties of up- dating expert opinions and resolving differences between experts by conferencing. We hope to begin such work in one of more industries by identifying a small panel of experts and having them identify the information inputs upon which their predictions (of a matrix row or column, for instance) are contingent. The causal ordering of experts can then be identified as is shown in Figure 2. The scenario is shown as a program vector of m components, (q) and expert A's prediction of coefficient 1 is influenced by the values taken by components 1 and 2 of the program vector and by the estimate of coefficient 2. Expert B's estimate of coefficient 2 is influenced directly by component 3 of the scenario, but is also influenced by the prevailing estimate of coefficient 1. Thus expert A is influenced directly or indirectly by three of the components. SCENARIO < MODEL /\ 'Expert A Coefficient 2 Coefficient 1 Expert B * 1 ExpertC H Coeff| cient3r - lExpert p Coefficient a Figure 2. Causal Ordering of Experts 10 In contrast, Expert C's prediction of coefficient 3 is influenced only by- component k of the scenario, and expert D's prediction of coefficient k is influenced only by the prediction of coefficient 3« In practice, the causal ordering of experts can be expected to become complex. Each component of the model should be covered by more than one expert and they can be expected to disagree and must conference. Experts who are early in the causal ordering (such as c) must react before those down the tree are called upon. A computer network system such as MEASURE appears to provide a convenient way of mobilizing expert opinion and keeping a forecasting model updated as expectations change. There are many problems in using live experts in forecasting, but fore- casts of the future are necessarily subjective and depend on expert judgment even if they use econometrics, trend projection, bird flight, the inspection of the entrails of sacrificial victims or other quantitative methods. The identification and incorporation of experts familiar with their obligations into the forecasting system seems to us the best way to approach the problem of forecasting in highly disaggregated systems and, at the same time, identify- ing and documenting the components of the forecast in a way that good forecasters can be identified and retained and the forecasting community can learn from its past mistakes. 11 IV. Preliminary Applications of the Model While the system is not complete, it has been possible to use the model for some simulations of 1963 • The energy and employment intensities of the 360 industries are presented in Fig. 2. While a large proportion of the industries are centrally clustered, there are some very energy intensive industries (asphalt coatings and asphalt paving, cement, primary aluminum, building paper, and chemicals) and some very labor intensive industries (hospitals, hotels, credit agencies). The pattern shown in Fig. 3 represents the energy and labor requirements of an additional dollar delivered to final demand. It represents, for a consumer, the direct and indirect effect on energy and employment of the expendi- ture of one dollar. It does not include any multiplier effects of the expendi- ture. It is, therefore, inappropriate for use in an impact analysis. The effects of an increase of a $1 billion expenditure delivered to final demand on each industry offset by a $1 billion decrease in final demand distri- buted among all other industries in proportion to their share of total GNP (less the industry in question) is shown in Fig. if. This represents a possible reallocation, but not a terribly interesting one. Yet another way of considering the problem is by examining the effects of 10 percent proportionate growth in each industry, with an offsetting decrease prorated among the other industries in proportion to their share of GNP (Fig. 5 and 6). First quadrant industries are primarily agricultural; second quadrant industries are basic material production, construction and fabrication oriented; third quadrant industries are service oriented with a high degree of technology and high wages; fourth quadrant industries are service oriented without a great degree of special labor saving technology and with low wages. 12 r ! » Asphalt Coatings •Asphalt Paving Cement * Primary Aluminum . Building Paper „ Chemicals '1\ o r-l * >> -P •H •3* ra e 0) "0 « Misc. Chem. * Syn. Rubber « Steel . » Brick , Prim. Zinc Struct. Clay, n.e.c. "Alum. Polling % Drwg. x Plastics Paper Bd.„ „ Ce11 " Flbers Pet. Ref,» * x Ll me Fert. Mine, *«• chem - n ' e ' - Paper Mills, .Fertilizers _ , ... , , Iron Forge Pulp Mills « .Glass Containers Carbon . Gas « Alum . Cast . ' »Air trans. «Prim.Non-Fer. « *, « , Motor vehlc ighway Construction lectric Motors •Cotton. Oil Nursery > Prod. Coffee* «gg * Butter , -, "Credit agencies xCheese « Cotton » Dairy Farm Prod. » Ag. services " l^°P?od ft. xMilT tobacco Drying -Hotels?*™^ }.G Ttoctors 4.3 * «Commun. \ "Banks •Barber Shops * ••♦- f - • - - - f ♦— - • O.n i2.j 17.? 21.5 25.3 Employment Intensity, 10" 5 Jobs Per Dollar to Final Demand f— 30. 1 Figure 3. Total Energy and Employment (Direct and Indirect) Per Dollar Delivered to Final Demand, 1963- Source: Energy Employment Policy Model j CAC, February 1973- 13 Thus, Fig. 5 and 6 are addressed more to the policymaker concerned about the question of growth. In general, Fig. 5 and 6 are similar to Fig. h, i.e., most industries remain in the same quadrants but their relative positions have changed. The magnitudes in Fig. 5 and 6 reflect the relative dependence of the U. S. society in 1963 on each of its industries. For example, a 10$ increase in delivery to final demand by motor vehicles would have required 12 a direct and indirect energy increase of 3*+ x 10 BTU and a decrease in employment of 10U,000 jobs (direct and indirect). A 10$ increase in deliveries of postal services to final demand would have reduced energy 12 consumption by about k x 10 BTU and increased employment about 36,000 in 1963. Some intermediate products deliver little to final demand such as steel and primary aluminum. A proportional increase in final demand tends to de- flate the artificial importance of the intermediate products seen in Fig. k. The problem with the approach used in Figures k, 5 and 6 is that the gain in delivery to final demand is absorbed proportionately from all other industries. Quite to the contrary, the product of an industry competes with only a few other products, e.g., aluminum with steel and wood as structural members, steel with glass and plastic as food containers. If one industry's gain were at the expense of a few competitors, the complexion of Fig. k, 5 and 6 would change. Suppose for instance that a one billion dollar gain in primary aluminum deliveries were obtained at the expense of an identical loss in steel deliveries. Then from Fig. 2 energy use would increase about 116 x 10 1 BTU and employment would decrease by 15,000 jobs. A similar increase in primary aluminum deliveries at the proportional expense of all other 12 industries would produce an increased use of energy of 322 x 10 BTU and a loss of 65,000 jobs (Fig. 1+). Ik * Asphalt Coatings Paving Mixtures x Cement 3 a. a) x Primary Alum. a Building Paper x Chemicals « Misc. Chem. "Syn. Rubber "Steel _, __, Clay Prod. x * Bricks «Zinc "Alum. Foiling * Plastics Ref. Pet.. ^Pej ^J^* Fert. Mine., Ag. Chem. xAlum. Cast. xPrim. NonFer.x" *x IB „ i x Coffee, * _ ** * * Real Es"_ xCommun. V a ^ t^ i -i ^Doctors Q 3 ,0vn. Dwell. Small Guns > -0. 13? -O.G'VJ -0.044 «di*rtains« Ice "Cotton, oil x Eggs" Butter Cheese „ . * * Milk " Cotton Per. Ser. "Ag- Serv. ■ "Fresh Fruits* Tobac - Dry- Dairy Farms "» »Movies*Retail Trade Fee. Ser. "Barber _ Credit * Tobacco xHotels „ Hospitals -+- -+- -(- 0.000 0.0*4 . 0&(J . ! 3? 0. i76 0.220 Change in Employment, 10 Jobs Figure h. Changes in Total Energy and Employment Requirements for a One Billion Dollar Increase in Final Demand from the Noted Industry, Proportionately Absorbed from all Other Industries, 1963- Source: Energy Employment Policy Model; CAC, February 1973- a PC 5 c d 15 * Petroleum Refining x Chemicals x Gas Utll. * Motor Vehicles x Highway Constr. » Elec. Util. Water «Air transport. Misc. Chem. Bldg. Constr.g Trans.- Uti],. Constr.* Steel#|? lotor freight * new constr. "Highway "Water % Sanitary leggg "e djtgranap. ^ ^^ ^^ Ale oh. Bev Pfo.3er.* fAir -TVeq Cigar Auto Rep. * Batiks x Communicatibns ; Doctors ^ duC ?»9St' Hc raft Barbers'" ce . Pers. Serv. *Non- \ . , /t , x profit Apparel (Pur. ) •Insurance Carriers Rec. xHospitals % 'Real Estate x Wholesale Trade Own Dwellings (-.1+2,-191+) Retail Trade (.35,167) + ■+- -+- -0. 114 -0.076 -0.033 0.000 0.033 0.076 0.114 0. 152 0.190 Change in Employment, 10 Jobs Figure 5- Changes in Total Energy and Employment Requirements for a 10$ Increase in Final Demand from the Noted Industry, Proportionately Absorbed from all other Industries, 196 3- Source: Energy Employment Policy Model; CAC, February 1973- 16 3 cr