Energy planning under import restrictions

518

European Journal of Operational Research72 (1994) 518-528 North-Holland

Theory and Methodology

Energy planning under import restrictions Ga~lar Giiven Department of Industrial Engineering, Middle East Technical University, 06531 Ankara, Turkey Abstract: Changing energy prices will normally have important long range implications for an oil-importing country with a persisting trade deficit. The availability of relatively expensive local energy can complicate the issue further since developing such resources may save foreign exchange but may be inefficient in the long run. This paper describes two nonlinear programming models for energy policy analysis in which foreign trade is considered within a one-sector representation of the economy. A putty-clay technology is assumed as it is more representative of non-mature market economies which have to deal with such problems. The models are capable of addresing energy issues in general as demonstrated by results for Turkey. Keywords: Energy; Economics; Planning

I. Introduction Since the introduction of computable models, the general equilibrium framework has been firmly established in economic policy analysis. A considerable number of energy models also employ this approach to represent the energy-economy i n t e r d e p e n d e n c e (Kavrako~lu, 1987; Bergman and Lundgren, 1990). The adoption of general equilibrium modelling in energy studies was hastened by sudden and sharp rises in world energy prices that took place in nineteen-seventies, when it became painfully clear that economic activity could very much depend on energy prices. While equilibrium models are conceptually rich and have strong economic underpinnings, they depend foremost on neoclassical assumptions, which are more appropriate for an economy in which market forces can function without much hindrance. If an energy importing

Correspondence to: Dr. Cs.a~larGiiven, Department of Industrial Engineering, Middle East Technical University, 06531 Ankara, Turkey.

country finds it difficult to readjust to higher prices, energy policies will most likely be influenced by the availability of foreign exchange and the ability to borrow foreign capital. The situation can be complicated by the presence of local energy resources which may be of economic significance only if higher energy prices continue to rise. In the case of Turkey for example, where abundant energy reserves do not exist, an important problem is to decide whether or not, or to what extent to develop existing resources in the form of several hydro schemes and low quality lignite deposits. Development of such indigenous resources may seem wasteful in the short run, but may help to reduce energy imports in the future. The effects of such decisions can be assessed only by considering the interdependence between foreign exchange expenditures for energy, and for the rest of the economy under varying scenarios. This kind of analysis normally requires a model in which the energy sector is represented with sufficient detail, whereas the rest of the economy can have a more aggregated specification. Models of the type of Manne's ETA-Macro

0377-2217/94/$07.00 © 1994 - ElsevierScienceB.V. All rights reserved SSDI 0377-2217(92)00117-3

~. Giiven / Energy planning under import restrictions

(Manne et al., 1981) combine a detailed energy model with a one-sector model of economic growth in which different forms of energy are intermediate inputs to economy-wide production, and an equilibrium solution is computed by utility maximization (Alan and Seong, 1983). A one-sector specification is probably more suitable for producing long-term projections which are compatible with the long lifetimes of energy investments, than a more disaggregated but necessarily shorter term representation. Furthermore, recognizing only a single consumer allows an optimization setup which makes it straightforward to obtain solutions with standard algorithms. Finally data requirements are modest and the economic model is therefore relatively easy to estimate. This paper reports on two energy-economy models which allow at varying degrees of detail the inclusion of foreign trade in energy planning. The introduction of a foreign sector that can display the impact of an inflated energy bill on the rest of the economy allows more realistic representations for countries which have to rely on foreign capital inflows, such as those moving from a planned to a market economy. Both models are compact, easy to use and reasonably adaptable. A one-sector representation and an optimization framework are adopted with nonlinearities in the constraint sets as well as the objective functions. Model estimation is rather demanding owing to the functional forms adopted for the production functions. This type of difficulty is persistent in economy-wide planning as trade-offs between adequate econometric estimation and adequate functional specification cannot be removed. In the applications reported here, some of the parameter values were obtained from existing literature while others were estimated by benchmarking to a base year, thereby calibrating the models using Turkish data. This is not fully satisfactory but is a common practice in energy planning where long horizons necessarily diminish the importance of reliable estimations, and models are built more as tools of scenario analysis than as forecasting devices. Nonetheless the models were at least partially validated by starting the planning horizon from 1983 which allowed for some after-the-fact comparisons. The models are described in the following two sections where emphasis is on the economy submodel in each case. The energy submodels are

519

larger in size and take on a country specific activity-analysis framework like several other linear programming models that have appeared in the literature. These submodels are therefore described in the Appendix in outline only. An alphabetical list of the variables is also given there.

2. Model I

Oil-poor economies which rely on foreign capital inflows or have to service heavy external debts in order to sustain their industries may find it especially difficult to adjust to higher energy prices. In the case of a sharp increase in the import bill the flow of foreign exchange earnings can acquire added importance which, more often than not, may in turn require uninterrupted imports of energy, capital goods, and intermediates like raw materials and semi-finished products. In some cases the imports of these intermediate goods are nearly as crucial to the functioning of the economy as investment goods and oil are. A further complication may be added by the presence of indigenous resources of borderline value which must be reassessed within the framework of new energy policies. The present model takes these points into consideration; it recognizes imports and exports explicitly and using an economy-wide production function, relates the gross output of the economy to the input factors of capital, labour, intermediate goods and energy. The objective is to compute intertemporal equilibrium solutions over a planning horizon of T periods, and solutions are obtained by maximizing the following discounted utility function which is consistent with the conventional theory of consumption (Henderson and Quandt, 1980, Chapter 3): T

max ~

At

log Ct,

(1)

t=l

where C t stands for consumption. Each period is represented by a year t for which gross production is defined as the sum of the gross domestic product, the payments for imported intermediates (INT t) and payments for e n e r g y (ECt). Hence we have Yt = GDP, + INT, + EC t .

(2)

520

~. Giiven / Energyplanning under import restrictions

GDP satisfies the conventional identity, GDP t =

C t + I D t + I F t + X t - M t,

(3)

in which investments are differentiated according to whether they are made with domestic (ID t) or imported (IFt) capital goods and where X t and M t stand respectively for exports and imports. In the literature imports are sometimes classified as competitive or noncompetitive, implying that they are either substitutes or complements of the domestic commodity. The present treatment takes a middle course depending on the import category. There are four specific categories; capital goods, nonenergy intermediates, fuels and consumer goods. Hence total imports are: M t = IFt + INTt + EFt +

CG t

(4)

where EF t stands for payments for imported fuels such as oil, gas and coal, and CG t for imports of consumption goods. The first three import categories in (4) are essentially treated as imperfect substitutes of other factors by way of their contribution to the gross output of the economy which is handled by a multilevel production function. Consumption can be obtained from the above expressions as: Ct = Yt - I D t - X t

+

EFt + C G t - ECt.

(5)

The two energy cost variables ECt and EFt are determined by the energy submodel. Production is assumed to take place according to a putty-clay technology. That is, the economywide production function allows substitution among inputs only for the increments in these factors, whereas surviving stocks are assumed to remain unchanged. This lessens the effect of the substitution assumptions inherent in a neoclassical function and provides a more realistic representation of an economy which, although responsive to changes in relative prices, is nevertheless burdened by institutional restraints. Accordingly, newly made additions to stocks of capital and labour are assumed to be substitutable, subject to a specified elasticity, and thus constitute a valueadded aggregate. Energy inputs of different types similarly constitute a second aggregate. These two and imported intermediates are then put together in a constant elasticity-of-substitution (CES) production function. That is, incremental

production in period t resulting from factors input since the previous period is given by

+/3tINTNtp 1/p + (1 - at - fit)

[ENt ~" PNtV" SN~s] Pl

(6) Note here that all variables end with an N, indicating that they represent incremental values newly added to existing stocks. After experimenting with several alternatives this functional form was chosen as a compromise between a multilevel CES representation and ease of model calibration. Aggregation of the primary inputs is according to a nested Cobb-Douglas function. At the first level, two varieties of capital stock, as formed by domestically produced investment goods (KDN t) and imports ( K E N t) are recognized, and then combined at the second level with labour (LNt). Two types of investment capital, domestic and foreign are thus recognized in keeping with the disaggregation of imports. Energy likewise, is assumed to be a composite of electricity (EN,), petroleum products including natural gas (PN t) and solid fuels (SNt). A constant elasticity of substitution exists among the primary and the energy aggregates and the intermediate goods imports (INTNt); it is given by or= 1 / ( 1 - p ) . Total production for year t is: Yt = YN/+ AkYt_1,

(7)

that is, the sum of the outputs of new and surviving stocks from the previous period. The constant A is the ratio of net investments to gross which reflects the rate of decline of stocks due to retirement and k is the period length in years. Similar equations establish the relation between the inherited and the new components of each input factor as follows: 1NPUTt = INPUTNt + XkINPUTt-1,

(8)

where INPUT obviously stands for KD, KF, L, 1NT, E, P and S. Capital increments during a k-year period are calculated as accumulated investments belonging

~. Giiven / Energyplanning under import restrictions

to the previous and the current representative years; K D N t = ½k(ID t + IDt_ 0 , KFN, = ½k(IF, + IF, ,).

(9) (10)

Terminal year investments are fixed at a fraction of the capital stock so as to reduce end of horizon distortions: ID, = gd. KD,,

(11)

IF, = gf . KF,,

(12)

521

individual energy activities into a vector of energy inputs for each period in a linear programming format. By the same token, [B t] is the cost matrix for energy activities and equation (18) provides the energy and imported fuel costs in the LP model. The rest of the energy submodel comprises technological relations among the energy activities such as primary to secondary energy conversion and bounds on domestic energy resources. These can be written again symbolically as,

[ G ] [ z , , z 2 . . . . . z r ] > [b] where the constants gd and gf imply the assumed post terminal growth rates. The foreign exchange constraint is as follows:

X, + F, + Wt> M t

(13)

where Wt, representing factor incomes from abroad, can be fixed exogeneously. F t represents foreign capital inflows and can either be set exogeneously or limited to within a proportion of the G D P as F t _
(14)

Exports and consumer goods imports can similarly be bounded from above and from below as follows: (b,GDP~ _2GDP,,

(15)

r h G D P t < C G t < ~2GDP,.

(16)

These ad hoc constraints would not normally be needed if the economy were able to follow the optimal time path anticipated by utility maximization. The p a r a m e t e r s in these constraints will therefore reflect the level of distortion that is thought to prevail in an imperfect market. It turns out nevertheless that some constraints remain nonbinding in the solutions. The link between the economy and the energy submodels is established by energy demand and cost relations which we indicate here symbolically as:

[E,, P,, S,] = [ A t ] [ z t ] ,

(17)

[EC,, EFt] = [B,l[z,].

(18)

In the above, the vector [z t] represents energy activities and [At] , the associated technology matrix which features in the energy submodel. Equation (17) represents therefore, the mapping of

(19)

where [G] is a matrix of coefficients and [b], a vector of constants. More information on the energy submodels is given in the Appendix. Expressions (1)-(19), written out for t = 1. . . . . T, complete the model which expresses the two-way linkage between energy and the economy taking into account foreign trade constraints.

3. Model II

The information provided by a one-sector representation of the economy is of necessity limited and it is conceivable that a less elaborate formulation might perform equally well and require less effort. Furthermore if the economy is actually less flexible than assumed in the first model, such a simplification could also enhance model accuracy. The second model follows this line and treats foreign exchange restrictions by essentially projecting today's economic structure into the future (Gtiven, 1982). This is done by restricting the imports of energy and nonenergy goods to within limits calculated as linear functions of the GDP. An economy-wide production function relates gross output to inputs of capital, labour and three types of energy. Gross output is defined as the sum of G D P plus the interindustry payments for energy which are now calculated in domestic and foreign currency components separately: Yt = GDPt + ECDt + ECFt.

(20)

G D P for year t is defined by the accounting identity,

GDPt=C,+It+Xt-M

t.

(21)

522

~. Giiven / Energy planning under import restrictions

Here M t represents merchandise imports and X t is the sum of merchandise and net service exports. The difference (22)

R, = M, - X t

would then represent the resource gap equal to the sum of the net factor incomes from abroad and the current account gap. This representation requires only two types of primary factors, capital and labour, and the three energy factors as defined previously. The production function is separable and as before, it is assumed as though producers first determine their demand for primary factors and subsequently the required cost minimising energy mix. The functional form is simplified as follows:

YNt = ')/t[ 6, [KNSkLNtl] ° + (1 - 3,)[EN:ePNtPSN ss] o] ,/o.

(23)

Equations (8) are now to be written for K, L, E, P and S only, and (9)-(12) are replaced by the following: KNt = lk( I' +/t-l),

(24)

I t = gK t .

(25)

Foreign trade is modelled through imports and a sustainable resource gap; exports do not feature in the model explicitly. Total imports are calculated as the sum of nonenergy and energy imports; the latter is obtained from the energy submodel and the former is assumed to be a linear function of the GDP. Thus we have: )kit = ~'t + v t G D P t +

ECFt"

(26)

Constraints concerning the balance of foreign payments are of two types. The first of these prevents total imports from exceeding a certain proportion of the GDP: M, _<¢tGDVt,

(27)

and the second one imposes an upper limit on the resource gap as follows: R t <

t.ttGDPt .

(28)

This looks rather like imposing hard and fast rules on imports when compared to Model I. If however it is thought that the existing economic structure is less substitution rich than is implied

by the CES functions, then constraint (26) would be justified. Costs computed by the energy submodel can be symbolically represented this time as [ECDt, ECFt] = [Bt][z,]

(29)

while (17) and (19) remain unchanged.

4. Results and comments

Although the approach is applicable to other countries, the exact specification of these models reflects the particular issues faced in Turkey. In this section therefore, we briefly provide some background information first and then a number of results for Turkey. These however do not provide a complete analysis of the issues and should rather be viewed as demonstrative of model use. Energy policies in Turkey have generally followed the broader economic policies adopted without much change by successive governments until about 1979. These policies have mostly been inward-looking and have tried to build up an industrial basis by fostering import substitution through controlled trade and low interest rates. The energy sector has been predominantly stateowned or controlled and emphasis has been on developing indigeneous resources mostly in the form of hundreds of hydro schemes and many low quality lignite deposits. The economic viability of some of these public sector projects has inevitably been confounded by other socio-political benefits that these projects were thought to posses. The distortions that these policies may have caused were undoubtedly aggravated by frantic efforts to shield the consumer from rising energy costs and Turkey found herself almost bankrupt after the second oil crisis. This brought about a fundamental change of course in 1979 and the market has since been liberalized to a large measure; the exchange rate and the interest rate have been allowed to float and price controls and import restrictions have been lifted. This requires a new perspective on Turkey's energy reserves and provides the grounds for the neoclassical framework adopted in this study. However it also restricts the period over which the models can be validated.

~. Giiven / Energy planning under import restrictions Table 1 Model I projections (first figure) vs actual values (second figure) (in 109 Turkish Liras)

GDP Investment Exports Totalimports Imports of fuels Imports of Intermediate goods

1983

1986

1989

11062 11532 2590 2230 1087 1580 2057 2124 888 824 619 670

13989 13868 2860 3110 1698 1715 2657 2554 464 461 1053 1119

16450 15600 4030 3550 2581 2674 3786 3625 681 654 1512 1767

Energy consumption: oil and gas solid fuels

a

a

electricity b

17.15 17.54 15.25 17.56 27.89 27.89

17.34 19.62 17.48 22.37 38.61 37.66

20.86 22.52 22.63 25.23 51.28 49.37

a Million tons of oil equivalent (MTOE), net of electricity generation. b Billions of kilowatt-hours (TWh).

The two models were calibrated using 1983 data for Turkey. This year was selected for two reasons: new policies would have time to take effect and a consistent data base would be available. Under the assumptions of a base case scenario a number of runs were first made in which the first three periods were assigned shorter durations initially and the planning horizon was started from 1981 so that some retrospective comparison between model predictions and observed values for the years 1983, 1986 and 1989 might be made. These comparisons are summarized for the first model in Table 1. A study of the table indicates general agreement between model predictions and observed values, although in some instances the agreement is not that close. This is partly due to shifts caused by policy changes or temporary swings observed in the economy. For example a new exchange rate policy and the provision of strong incentives after 1979 caused a surge in exports which was beyond normal expectations. A similar case in point is the relative upsurge in energy consumption observed after 1986 when the effects of economic stabilization policies launched in 1979-1980 were finally abated. A long term energy model is hard

523

put to capture such fluctuations exactly, except through exogeneously specified scenarios, and the performance of the model may therefore be accepted as reasonable over the long run, although improvements would be possible with better data. Once calibration was over, the period length was fixed at five years and the planning horizon started from 1986, with 1983 as the base year. Projections are therefore reported here for every fifth year from 1986 to 2006, although a longer time horizon was used in the computations. All monetary values are in terms of 1983 Turkish Liras or US dollars and the exchange rate remains fixed at 230 T L / $ in the absence of inflation. The models are relatively compact with around 300 structural constraints for a six period formulation with only one nonlinear constraint per period. The solutions were obtained using the nonlinear optimization program MINOS (Murtagh and Saunders, 1980), which converged without much difficulty. A summary of the energy activities defined is given in the Appendix. The discussion presented here of model projections is primarily intended for illustrting model performance rather than a complete analysis of Turkey's energy policies. Except for Table 3 in which the two models are compared, all experiments and discussions will be based on the output of Model I. The exogeneously specified variables making up the base case scenario are summarized in Table 2. The projections were essentially based on government reports. They include energy Table 2 Base case assumptions 1986

1991

1996

2001

2006

15 100 57

20 100 47

20 100 57

25 100 70

30 100 70

Prices: oil a gas b coal c

Growth rates: labour force productivity

0.01 0.01

0.01 0.01

0.01 0.01

0.015 0.0125

0.015 0.015

Parameters: Cost and utility discount rate: 10%/year. U p p e r limit on foreign capital inflows: 4% of GDP. U p p e r limit on exports: 20% of GDP. U p p e r limit on consumer goods imports: 3% of GDP. Workers' remittances: $2800 millions/year. a S/barrel.

h $/1000m3. c S/ton.

~. Giiven / Energy planning under import restrictions

524

Table 3 Macroeconomic results from the two models under the base case a. 1986

1991

1996

2001

GDP

13989 20029 27255 36111 34873 13980 2004O 26420 5775 9302 Investment 2800 4223 2770 4212 6080 9810 4731 7222 Exports 1698 3466 6472 9317 Imports 2657 4917 Oil&gas b 19.01 26.03 34.79 42.39 19.10 27.15 33.16 40.00 Solids b 20.98 27.22 39.14 49.30 20.12 26.35 38.75 50.15 Electricity c 38.61 63.95 108.45 138.65 110.27 140.72 36.62 62.34

2006 49607 47904 10500 11705 9921 12556 55.23 49.96 66.21 63.45 171.13 160.75

a The first figure is from Model I and the second figure wherever applicable, is from Model II (109TL). b MTOE. c TWh.

prices, which are in turn based on World Bank forecasts; the labour force and productivity growth rates and other parameters. In addition to these the scenario incorporates two other types of assumptions; a timetable for ongoing and scheduled energy projects and bounds on incremental changes that take place in energy activities. Imposition of such bounds reflects technological bottlenecks which have to be observed. A summary of the results obtained from the models under base case conditions are listed in Table 3 and additionally in Table 4 for Model I only. Inspection of Table 3 shows reasonably close agreement between the two models and this is more by design than chance. The models differ in Table 4 Import composition for the base case (106U5 $)

Total imports Capital goods % Fuels % Intermediates % Consumer goods %

1986

1991

1996

2001

2006

11553

21379

28137

40507

54590

4348 37.6 2018 17.5

6304 29.5 3897 18.2

8485 30.2 5384 19.1

11393 28.1 7625 18.8

12070 22.1 11481 21.0

4580 39.6

8565 40.0

13083 46.5

18163 44.8

24714 45.3

608 5.3

2612 12.2

1185 4.2

3326 8.2

6325 11.6

complexity and so does the calibration procedure. An identical set of benchmarking assumptions cannot be used which might have made comparisons more meaningful. Instead the models were calibrated separately under the same scenario assumptions. Noticeable disparity between the outputs is observed for later periods, with Model II yielding lower values for most variables except gross investments. This reflects the greater degree of flexibility and scope for substitution recognized by Model I compared to Model II, which adheres more rigidly to the present trade structure. Effects of this rigidity were also evident in the sensitivity tests. When energy prices were sharply increased, Model II ran into infeasibility and limits on foreign borrowing had to be relaxed. It must also be pointed out, on the other hand, that this model may be capable of producing more realistic projections in the short term. Choice will clearly depend on the case. In Table 3 Model I predicts an average GDP growth rate of about 6.5% per year which declines to 5.8% in 2001 and then recovers. Exports an imports grow at a somewhat faster rate reflecting an elasticity above unity. The growth rate of energy consumption is also higher than the GDP growth rate in the early periods but gradually falls behind after 2001. This indicates substitution of other inputs for energy as energy becomes relatively more expensive. The best year for most variables is 1991; a slowdown in growth is observed afterwards, which starts leveling only in 2001. The strong performance of the early periods is attributable in part to the drop in energy prices in 1986 from their 1983 values at which time fuel imports claimed more than half of all export earnings. By the year 2001, on the other hand, most of the cheaper reserves of indigenous energy will have been depleted. The model predicts an increase in investments during this period in an effort to substitute capital for energy. The unfavourable effects of expensive energy are partially reduced in 2006 as substitution effects begin to prevail. The growth in oil and gas consumption predicted by the model is commensurate with official forecasts of the government (Aybar, 1990). Predictions of solid fuel demand are slightly lower than forecasts, but in the case of electricity consumption model results for later periods are noticeably lower than forecasts. In the final period

~. Giiven / Energy planning under import restrictions

for example, the government forecast is about 30% higher than the projected level. According to the model electricity demand grows quickly at about 10% per year at the beginning of the planning horizon but this is in part due to ongoing and scheduled investments that were planned following the oil shocks of the seventies. Most of these are scheduled for commissioning by the mid 1990's. From then on the growth rate of electricity declines to about 5% in 2001. This is clearly a result of substitution away from electricity in production to other fuels and also to nonenergy factors such as capital and to a lesser extent, imported intermediates. These effects can be discerned from Table 4 where the composition of imports is listed. It seems that the model gives more credit to the substitution capacity of the economy than is given by government forecasts. Since demand projections in Turkey are normally based on the extrapolation of present trends, the predictions of a neoclassical model would most likely disagree in any case with such projections over the long term. Several experiments were conducted by modifying the base case scenario. Some changes concerned model parameters such as those defining the capacity of the economy for exports or for obtaining foreign credit. Modifying these parameters produced predictable changes in the results but did not cause significant shifts in base case energy policies. Changes in the discount rate and the substitution elasticity do lead to more significant alterations but there is no benchmark at this time against which different results can be compared. Scenarios incorporating different energy price projections are more interesting and we briefly report on these next. World oil price forecasts adopted in the base case scenario originate from government reports

Table 5 Results for the high energy cost scenario (units as in Table 1) 1986

1991

1996

2001

2006

GDP 13986 19980 26126 34472 47992 Investment 2800 4137 5293 8556 8490 Exports 1684 3276 5225 6894 9460 Imports 2644 4726 6920 8923 12003 Oil&gas 18.71 24.70 31.01 36.94 45.96 Solid fuels 20.87 26.25 37.22 44.72 61.83 Electricity 38.74 65.01 111.38 142.45 170.58

525

Table 6 Import composition for the high cost scenario (106US $) Total imports Capital goods % Fuels % Intermediates % Consumer goods %

1986

1991

1996

2001

2006

11497

20548

30089

38798

52185

4347 37.8 1971 17.1

6224 30.3 3215 15.6

8100 26.9 8031 26.7

9595 24.7 10187 26.3

11229 21.5 13546 26.0

4571 39.8

8503 41.4

12822 42.6

17517 451

23784 45.6

608 5.3

2606 12.7

1136 3.8

1499 3,9

3626 6.9

and are rather conservative. Summary results of a high energy price scenario are given in Tables 5 and 6. It was assumed in this experiment that oil, gas and coal prices would have risen by 1996, 2001 and 2006 to 50, 75 and 100% respectively, above their 1986 levels. A comparison of the base case (Table 3) and the high energy cost (Table 5) scenarios reveals about a 5% decline in G D P in 1996 and 2001 due to higher energy prices. Export capability of the economy and therefore total imports are also reduced and a sharp drop is observed in consumer goods imports. The decline in G D P for 2006 on the other hand is around 3% only and is indicative of the start of a recovery. The interpretation must be that higher prices cause reduced energy use and hence reduced production at first, but that in time the economy begins to adjust as more energy efficient capacity comes on line. A positive income effect may also be present in 2006 as the reduced energy bill begins to allow a higher proportion of capital goods and intermediate imports as observed in Table 6. This effect may be partly responsible for the recovery observed in 2006. Higher prices induce a shift away from imported oil and a faster rate of development for domestic resources. This results in increased use of electricity in the third and fourth periods and the energy submodel shows that the extra generation comes from new hydros. Results also indicate a higher level of lignite consumption, but the economy copes with higher prices over the long term by adjusting production technology and fostering conservation.

~. Giiven / Energy planning under import restrictions

526

Table 7 Remaining capacity under base case prices (%) a

Hydro 1 Hydro 2 Hydro 3 Lignite 1 Lignite 2 Lignite 3

a

1986

1991

1996

2001

2006

85 30 79 83 88 88 85 84 100 100 100 100

26 0 50 63 88 88 63 48 100 100 100 100

26 0 50 65 68 68 34 8 95 95 96 96

0 0 15 28 68 68 30 4 64 55 93 93

0 0 0 0 68 68 0 0 46 0 90 90

The first set of results corresponds to the case of restricted, and the second set to that of unrestricted growth.

The precise effect of higher energy prices on domestic resource use can be better understood by relaxing the base case limits imposed on the development of these resources. The models recognize three categories of hydros in order of increasing unit cost and similarly three categories of lignite reserves. Total capacity can be expressed as the annual energy that can be generated in the case of hydros, and as total extractable energy in the case of lignites. Tables 7 and 8 summarize the results of four cases in the form of remaining, i.e. unused, resource expressed as a percentage of the total availability of each category. Two sets of results are given in each table, the first corresponding to the original base case assumption of restricted development and the second to a lifting of these restrictions. It is evident from Table 7 that not all domestic resources are economically attractive under the Table 8 Remaining capacity under high energy prices (%) (see footnote in Table 7) Hydro 1 Hydro 2 Hydro 3 Lignite 1 Lignite 2 Lignite 3

86 28 79 83 88 88 85 84 100 100 100 100

26 0 49 52 88 88 63 48 100 100 100 100

14 0 49 52 68 68 34 8 95 95 96 96

0 0 6 10 68 68 30 4 64 62 93 79

0 0 0 0 68 68 0 0 46 0 85 71

base case assumptions. In fact most of the moderately expensive (category 2) and almost all of the expensive (category 3) lignite resources are left unused and the model relies instead on increased coal and gas imports. Gas fired power generation is similarly preferred to the development of expensive hydro schemes in the later periods. The tables duely point out an increase in domestic resource use when world prices are high, but the increase is not as much as might be expected for some categories especially in the case of restricted development. On the other hand lifting these restrictions gives rise to increased usage which takes place mostly before 2006 indicating that some of these resources can acquire economic significance only if they can be developed at short notice. The conclusion must be that detailed analysis at the project level is necessary for a final evaluation of some of these resources. Overall results indicate that Turkey must remain increasingly dependent on imported energy but securing these imports may not be as important as generating foreign exchange. It is clear that in the absence of new information, not all domestic energy resources will be of economic value in the foreseeable future and Turkey cannot rely on these to the extent that is still thought possible. The problem must be to improve import capability by restructuring towards a more export oriented economy rather than to increase self reliance in energy. Despite an aggregated representation of the economy the models can capture some of the interdependency between growth and energy policies. Improvements would be possible if a more flexible specification were adopted which allows different elasticities of substitution among input factors. This calls for a more complicated functional form which was in fact used in the earlier versions of the models, but later abandoned because of estimation problems. It seems however that a one-sector framework cannot be pushed much more than this and further improvements would only be possible with a price endogeneous multisector formulation in which the micro theory of the consumer is combined with the macro theory of the open economy model. This may also force abandoning of the optimization framework for an equilibrium seeking one with the concomitant difficulties of estimation and solution which would be greatly increased

~. Gfiven / Energy planning under import restrictions

and would probably force an unwanted degree of aggregation in the energy activities.

Appendix Table of variables Variables which appear in the text are defined below in alphabetical order, except those which are named by appending the letter N to other variable names. Subscripts have been suppressed for clarity. Consumption. C: CG: Imports of consumption goods. E: Electricity consumption. EC: Total energy cost. ECD: Domestic money component of EC. ECF: Foreign exchange component of EC. EF: Cost of fuel imports. F: Capital inflows from abroad. GDP: Gross domestic product. I: Gross investments. ID: Domestic money component of I. IF: Foreign exchange component of I. INT: Imports of intermediate goods excluding fuels. K: Capital stock. KD: Domestic money component of K. KF: Foreign exchange component of K. L: Labour stock. M: Total imports. P: Oil and gas consumption. R: Current account deficit. S: Solid fuel consumption. W: Factor incomes from abroad. X: Total exports. Y: Gross production. z: Vector of energy activities.

The energy submodels Energy activities include oil and coal, both imported and domestic, and imported gas. Domestic lignite is available in three categories aggregated from scattered deposits on a cost per calorie basis. Noncommercial energy is also included at a restricted level. Electricity can be produced from imported oil, gas, and coal as well as from domestic lignite. Turkey has substantial hydro resources at over 300 sites which are here

527

grouped into three categories depending on unit cost and plant factor. All constraints are linear and fall into one of six categories. Each category can be illustrated by an example as follows. Factor demand constraints ensure that each of the three energy factors are provided in sufficient quantity in each period. In the case of electricity for example, we have Et<

Y'~ET,/t t ,J

where ETii t denotes electricity generated in year t from primary resource i (i = oil, coal, hydro, etc.) at cost category j. Fuel demand constraints ensure sufficient supply of each fuel through production or importation. For example, the first of the following constraints, LIG/t

>

eiETijt,

LIGjt > E SWijt -t- e i E ETijt (i = lignite), J

J

J

makes sure that enough lignite coal of category j must be made available for use in power stations burning this type of lignite, and the second constraint limits the consumption of lignite for end ue and electricity generation to within limits of total availability shown on the left. The constants e i are efficiency coefficients. Resource availability constraints limit the use of domestic fuels over the planning horizon by the quantity of proven reserves. For example, k ~ LIGjt _< (reserve limit)j. t

Similar constraints are included for sustainable power generation based on domestic lignite and hydro resources: EWij t <_ (upper limit)ij (i = hydro or lignite),

where j denotes the category. Continuity constraints regulate energy use and capacity retirement; for example: ETijt = E T i j , t - 1 + k E T N i j t - k ETN~j.t_m

where ETNij t is the new generation electricity activity i t coming on line, m is the useful lifetime of this type of generating capacity in terms of periods, and k is the period length in years. Finally there are the accounting rows which calculate energy and imported fuel costs.

528

~. Giiven / Energy planning under import restrictions

Production f u n c t i o n parameters Table A.1 Parameter values sl = 0.55 sk = 0.45 skd = 0.70 skf = 0.30 se = 0.30 sp = 0.45 ss = 0.25 A = 0.97 = 0.998 3' = 8.32 a = 0.899 /3 = 0.098 tr = 0.30

action model", Computers & Operations Research 10/1, 9-22. Aybar, E. (1990), "Preliminary results for energy planning", Report, Ministry of Energy and Natural Resources, Department of Planning and Finance, Ankara. Bergman, L., and Lundgren, S. (1990), "General equilibrium approach to energy policy analysis in Sweden", in: L. Bergman, D.W. Jorgenson and E. Zalai (eds.), General Equilibrium Modeling and Economic Policy Analysis, Basil Blackwell, Oxford, 351-382. Giiven, (~. (1982), "Mathematical models for planning energy investments", Unpublished Thesis, Middle East Technical University, Ankara. Henderson, J,M., and Quandt, R.E. (1980), Microeconomic Theory, A Mathematical Approach, third edition, McGraw-Hill Kogakusha Ltd., New York. Kavrako~lu, I. (1987), "Energy models", European Journal of Operational Research 28, 121-131. Manne, A.S., Richels, R.G., and Weyant, J.P. (1979), "Energy policy modelling, a survey", Operations Research 27/1,

1-36. References Ahn, B., and Seong, B. (1983), "Computational investigation of a model linking algorithm for the energy-economy inter-

Manne, A., Condap, R., and Preckel, P. (1981), "ETA-Macro: A User's Guide", Report EPRI EA-1724, Electric Power Research Institute, CA. Murtagh, B.A,, and Saunders, M.A. (1980), " M I N O S / A U G MENTED User's Manual", Department of Operations Research, Stanford University, Stanford, CA.