- Email: [email protected]
Energy and economic growth
Energy and economic growth A case study of Pakistan
T. Riaz
The paper investigates the relationship between energy consumption and economic growth with special reference to Pakistan. It then establishes future energy needs of the country and presents a long-run energy plan to meet these requirements. The energy-economy relationship is explained. The analysis shows that energy shortages can be extremely harmfulfor a developing country like Pakistan. However, the energy problem is one of shortages of capital andforeign exchange which can be dealt with by prudent planning of remittances from the Middle East. Ke_vwwd.r
Economic
growth;
Energy
consumption;
Since 1973, oil and subsequently all energy prices have experienced rapid changes, which in turn have resulted in supply-side disturbances for the energy importing economies. During the 1970s and early 1980s energy prices mostly rose, which in most cases led to an upward shift in the supply curve, a fall in national output and a corresponding rise in the rate of inflation. More recently, oil and subsequently most energy prices have fallen, leading to a downward shift in the supply curve, a rise in output and a corresponding fall in the rate of inflation. The last 15 years, therefore, have been a period of disequilibrium, structural adjustment and adoption of strategies to cope with the previously less recognized phenomena of stagflation. This paper sets out first to establish an empirical relationship between energy consumption and economic growth with specific reference to Pakistan. It then estimates energy demand for Pakistan and presents a plan that is consistent with planned social and economic progress and which is unconstrained by energy availability. The article develops an integrated energy sector plan which is partially linked to the rest of the economy. Finally, an explicit energy-economy
The author is with the School of Economics, Newcastleupon-Tyne Polytechnic, Northumberland Building, Northumberland Road, Newcastle-upon-Tyne, NE1 8ST, UK. Final manuscript
received 28 August
0140-9883/87/030195-IO
$03.00
1986.
0
1987 Butterworth
Interrelationship
interactive framework is adapted to assess the interlinkage impact and to formulate policy.
Energy and economic growth Energy has played a vital role in taking mankind from an early basic existence to the present supersonic era. In the early stages of economic development, most economies were agricultural and depended mainly on human labour. During industrialization energy was used intensively to produce artifacts which combined with human labour raised productivity, output and living standards. The urbanization phase, which closely corresponded with the industrialization process, intensified energy consumption. The history of economic development thus shows a parallel development in energy consumption in more and more convenient forms. There is an obvious reason for this economic growth depends on increasing use of factors of production (energy being one of the factors) as well as their efficient use; at the same time, it also depends on the effective demand for goods and services. The empirical evidence on the energy-growth relationship can be analysed in a number of ways: (i) by constructing a scattergram of per capita energy consumption and GNP for a group of countries for a specified time; (ii) by finding energy-output ratios over time and space; and
& Co (Publishers)
Ltd
195
Energy and economic growth: T. Riaz (iii) by establishing single and multivariate relationships over time.
statistical
Posner r/ UI [ 131 have estimated a number of energyoutput ratios for three groups of countries. They have found that such ratios show a trend of rising at the early stages of economic development declining slightly with industrial maturity and then settle down at a fairly high level. The increase in the ratio is explained in terms of industrialization, urbanization and fuels substitution. The relative decline and stability results from improvements in energy use efficiencies and economic transition to a service-based economy. The energy-output ratio for Pakistan has increased from 1.05 in 1965 to 1.45 in 1980, reflecting structural changes in the economic system. This ratio is likely to go up again before it shows any decline or stability. A single or multi-variable statistical analysis can be relatively more precise for a single country. The estimated energy-GNP and GNP-energy relationships for Pakistan for 1960-80 are shown below. log E, = -0.3679 + 1.2274 log GNP, (0.0264) SE (40.42) t ratio R2 = 0.991 R2 = 0.990
(1)
log GNP, = 0.3824 + 0.8080 logE, (0.0774) SE (46.42) t ratio R 2 = 0.992 w2 = 0.990
(2)
= commercial energy consumption = gross national product = standard error = time period = R square =adjusted R square in parentheses are t ratios
The estimated energy consumption as a function of gross national product, Equation (l), is statistically significant, shows a negative intercept and positive slope which is consistent with the hypothesis that at an early stage of economic development each country relies heavily on non-commercial sources of energy: however, with time commercial fuels are substituted for the non-commercial and increasingly used during the industrialization and urbanization process. The estimated coefficient of 1.2274 is the income elasticity of energy consumption, which means that, if ceferis parihus, gross national product increases by 1%, then commercial energy consumption increases by 1.23%. The inverse relation between gross domestic product and energy consumption, ie Equation (2), is also
196
log GNP, = 1.1464 + 0.3836 log L, + 0.4125 log K, + 0.3994 log E, (3) (0.1213) (0.1825) (0.0298) SE t ratio (2.89) (3.18) (3.29) R2 = 0.957 R’ = 0.949 F= 111.62 DW = 1.695 where
where E GNP SE t R R2 Figures
statistically significant. The estimated parameter 0.8080 represents the commercial energy elasticity of GNP which shows that if ceteris parihus. energy consumption increases by 1% then Pakistan’s GNP will increase by 0.8%. The estimates, ie Equations (1) and (2), indicate the interdependence of the two variables. These estimates, however, must be interpreted with caution, as the energy consumption figures do not include non-commercial energy which is used extensively in Pakistan, and the conversion of different fuels into a common denominator does not take account of fuel efficiency in different uses. In order to find the partial contribution of each factor of production to economic growth in Pakistan, it was decided to estimate an aggregate production function with special emphasis on energy input. Because of limited data availability it was possible to estimate such a function only for the manufacturing sector of the economy. The best estimated function is given below:’
GNP L K E
= = = =
gross national product manufacturing employment manufacturing capital stock commercial energy consumption
t. SE. R2 and 7i2 are as before. The generalized least square method has been used for estimation purposes and the results given in Equation (3) have been obtained from the transformed equation, which was obtained using the generalized differences process. All estimated coefficients are economically and statistically significant. The explanatory variables jointly explain all except about 0.043% of the variation in the dependent variable log Y. The estimated coefficients are elasticities with respect to each factor of production included in Equation (3). The sum of these elasticities is 1.195, which indicates increasing returns to scale in the manufacturing sector. But it should not
‘The estimates are based on data from the period 1959/60 to 197X/79 and were provided by the Institute of Development Economics, Islamabad. Pakistan. It is realized that the estimated function possibly presents a numberofproblems, eg theerror term may not be random or have normal distribution; the independent variables may not be random or independent of error term; and there may be specification error and the assumptions to perfect competition and profit maximization may not be satisfied. The justification for these estimations is given in Riaz [14].
ENERGY
ECONOMICS
July 1987
Energy and economic growth: T. Riaz assume2 to imply increasing return for the economy as a whole. The commercial energy elasticity of manufacturing output has been estimated as 0.3994. This means, if, ceteris paribus, commercial energy consumption increases by 1% then manufacturing output will increase by almost 0.4%. The explanation of other elasticities is similar. From the above analysis, it is clear that economic growth and energy consumption are closely related and that the economic development process is an energy intensive process. Hence a basic inference may be drawn that the social and economic structure of an economy could be substantially and regressively altered by large constraints upon energy use.3 An extreme version of this argument will imply that energy savings in the developing countries cannot be achieved unless GNP is proportionally reduced pari passu with energy consumption. Thus, in order to avoid stagnation or even decline in GNP, these countries must avoid energy supply shortages, which can be caused by controlled energy prices, delays in the development of feasible technical options or shortages of foreign exchange. First it is essential to estimate future energy demand, which is consistent with the societal progress target of the country and only then may one plan for adequate supplies.
Energy forecast for Pakistan A number of energy forecasts4 have been based on the relationship as in Equation (1) or a single coefficient such as energy-output ratio. While such empirical findings and their interpretation are interesting, one cannot assume that there is a stable ‘world line’ to forecast energy consumption over time. Each country’s energy demand is determined by individual factors such as climate, orientation of the economy, efficiency of industrial and household conversions and energy prices. Thus, the forecasts based on the aggregate analysis cannot be taken more than as just a rough estimate. Hence, a need for a more refined approach. Energy demand behaviour is reflected in two sequent decisions - first in the purchase of a specific fuelburning durable good and second in the frequency and intensity of its use. The stock of fuel-burning durable goods is fixed in the short run, but is variable in the ‘Such an assumption could have serious implications for the analysis conducted below. 3Because of this, I am opposed to energy demand management by the government. %ee for instance Starr and Field [I73 who have adopted the following approach. First they forecast employment levels using population forecasts, then using forecasted level of einployment. They forecast GNP and then use this forecasted GNP in an energy/GNP empirical relationship to project future energy demand.
ENERGY
ECONOMICS
July 1987
long run. To take account of these features and economic situations in Pakistan, a combination of behavioural and fixed coefficients type models have been developed and estimated to predict future energy demand for the country. The adapted model is a partial adjustment type which takes account of the consumer’s habit formation and the vital role of fuel burning durable goods. (i= 1, 2, 3 . . . N)
rlc = Iy + BXit + U,’ x-r,_,=E,(y--Y,_,)+U,
(4) (5)
where c is the ideal (or desired) but unobserved level of the dependent variable (ie fuel demand) at time t. Xi, are the independent variables (eg own price, income/output, cross prices) which determine the level of each fuel demand. x is the actual observed level of the dependent variable (ie fuel consumption) at time t and /. is an adjustment coefficient which indicates the rate of adjustment of Y, to v and its value lies between zero and 1. The value of Y: in Equation (4) is not known but it is assumed that the consumer wishes to move from x to r;*. However, he is only partially successful during one time period. The reasons why x does not adjust to I;* in one period are many and may include economic, technological and institutional constraints as well as habit formation.5 The relation (5) provides a link between x andY,*, solving for ,Iv shows: nr;*=I:+(n-l)k;_,-U i = (OZ+ fiXi, + U,) substituting
(6)
Y: in Equation
Y;=crI,+/?LX,+(l
-A)x_,
(6) gives: +e,
(7)
where e,=1U,+ Following equations
U, the above, is obtained:
the following
set of demand
where xj, stands for the quantity demanded of fuel i (i = electricity, gas, oil and coal) in sector iti = agricultural, industrial and domestic/commercial) in time period 1. Pij, stands as the unit price of fuel i in sector j in time period I. Xkj, stands for independent variables i(k = income/output, number of customers) in sector j in time period t, and Y,_ 1 is one year lagged demand. The transport sector and the power industry’s demand for fuels has been estimated separately using an input-output type framework. The transport sector demand for petroleum products is based on the 5For a good discussion on habit formation Balestra [I].
see Grillinches
[8] and
197
number of vehicles and estimated fuel consumption per vehicle per year.” The power industry demand for fuels has been estimated using a fuel efficiency factor of different thermal plants and their annual output, which in turn has been determined by an optimal energy plan. The non-commercial energy demand is explained by the following model.
NCE, = F( r,, P,, 7;)
(9)
where NCE represents per head non-commercial energy consumption; Y stands for per head real income; P stands for per unit real prices of commercial energy; Tstands for time related social and economic progress. The subscript t indicates time period in years. The set of demand equations as set out in Equation (8) and in log linear form have been estimated using the generalized least squares method. The data consist of 20 annual observations. The prices used as explanatory variables are the real (ie deflated) average prices. The income and output variable has been deflated using the GNP deflator. The sources of data are as follows: quantity demanded of each fuel, price of oil and natural gas, Ministry of Petroleum and National Research [6, 71; price of electricity and number of customers, Water and Power Development Authority [18, 191; price of coal, Malih and Siddique [ 1 I]; deflator, GNP/GDP and population, Pakistan Economic Surre~q [S]; number of gas customers, Pukistan Stuti.sticu/ Year Book [4]. The estimated relations are shown in Tables l-5. ‘The transport sector does not consume any natural gas and the consumption of coal and electricity is strictly confined to railways. The railways demand for coal and electricity and the aviation demand for petroleum products are based on a simple trend. Remaining demand estimates were arrived at by developing a dilTerence equation such as: X,=/(X,_ ,. DX,. AX,) where D is dropout rate and A stands for new vehicles. To develop time series data for buses. tractors, cars and light vehicles. and then using the following fuel consumption figures, High-speed diesel consumption,‘tractor/year = 18.83 tons Htgh-speed diesel consumption/bus/year = 10.69 tons Motor spirit consumption/car/year = 1.237 tons Motor spirit consumption/2 &/wheels;year = 0.412 tons These estimates are given by the Pakistan Planning Commision [12]. ‘A number of statistical problems faced in estimation or prior to the estimation were resolved as follows. The problem ofjoint determination ofdependent and independent variables was resolved simply as energy prices in Pakistan are set by pubhc regulation rather than market forces. Furthermore refuge is taken under the partial equilibrium blanket, accepting the view that a vast system of simultaneous equations has its place in economic theory, but it may not contribute much to the cause of empirical research. The problem or identification is serious especially in the presence of cross-prtce variables. For such a model there exists no valid criterion to test for identification. The structural parameters of our model, however, satisfy ordinary statistical tests. The problem of proJect collinearity was not faced as the matrix of the independent variables (ie xx) was found to be non-singular. The generalized least square has been used todeal with themilderform ofmulticollinearityand autocorrelation.
198
Table
1. Estimated
Explanator)
demand models for electricity.
variables
Price of electricity Prrcc of natural
gas
Price of oil Price of coal Real income per capita Real GDP
(sectoral)
Quantity previous
demanded year
Number
of customers
Class of customers Residential commercial Industry
Agriculture
-0.117 (0.042) NS
-0.08X (0.03 1) NU
0.02 I (0.012) NS 0.416 (0.151) NU
- 0.092 (0.023) -0.01 7 (0.006) NS NS NU 0.59x
0.032 (0.0 I 2) N L’ NL
(0.128)
0.403 (0.065)
0.614 (0.301) 0.301 (0.976) 0.976 1.95 0.210
0.632 (0.213) 0.178 (0.055) 0.9x5 1.x9 0.269
in 0.718 (0.235) 0.250 (0.109) 0.95 2.15 0.165
R2 DW P
Norc. All prices are Rupees per gigajoule (Rs.‘GJ). All money unit have been deflated. All variables were transformed into natural lo form except the number ol customers whose partial impact wa estimated separately. NS = statistically non-significant at 5% levc and was dropped from estimation. NU =either ‘Not used’ o ‘not included’ in the sector. R2 stands for adjusted R square. 1 stands for estimated Hildreth Lu coefficients. The figures in paren theses are standard errors of the estimates.
The estimates,’ in most cases, are economically am statistically significant and provide a reasonable ex planation of energy demand in Pakistan. At the root of all energy projections lies a judgemen on the probable level of output of goods and services o a future economy which is a quantitative corollary of; perceived future level of economic welfare and develop ment. The future perception of economic growth Table 2. Estimated demand models for oil.
Explanatory
variables
Class of customers Residential/ commercial Industry
Price of electricity Price of natural gas
NS 0.132 (0.054) -0.106 (0.033)
Price of oil Price of coal Real income per capita Real GDP
(sectoral)
Quantity demanded the previous year RZ DW P
Nolc: See footnotes
in
ts63 (0.093) NU
-0.219 (0.102) NU NU
0.23 I (0.016)
0.426 (0.205)
0.538
0.761
(0.227) 0.928 2.01
(0.320) 0.952 2.11
0.28
IO Table
ENERGY
Agriculture
NS 0.093 (0.015) -0.159 (0.022) NS NU
0.205
NS
NU
'0.600'
(0.175) 0.900 I .95 0.143
1.
ECONOMICS
July 1985
Energy
and economic
growth:
T. Ria:
is also realistic, must have some relation to the historical growth rates of employment, productivity and output. Thus in developing energy projections for Pakistan it is assumed that the country will follow its historical growth path. Thus the changes in GNP. GNI and sectoral output have been obtained by establishing time path predictions to 2005. The trend growth rates are GNP = 6.3%; industrial output = 7.5%; agricultural output = 2.9% and population growth = 2.5 to 2.9%. The population forecast is taken from the census series. The number of customers is based on planned targets. The energy prices are assumed to move to their internal levels and then start to decline, reaching their 1974 level in 1990 and then these prices will rise again reaching their 1980 level in 1992 and then maintain their values in real terms. The energy projections based on the estimated models and the future perceptions as outlined above are shown in Table 6. The projections show an average growth rate of 7.5% over the planning horizon. The consumption of commercial fuels will increase (7.74%) at a greater rate compared with non-commercial energy (1.7 I O/u).This reflects on the energy-intensive industrialization, urbanization and substitution of commercial for noncommercial fuels. However, it must be pointed out that in spite of this rapid increase in energy consumption, per capita energy consumption in Pakistan in 2005 will still be about 37.4 GJ, which is far below per capita consumption of the advanced industrialized countries. The tracking performance of the model has been checked by calculating the mean percentage errors of the dependent variables. It is found to fall in the range 0.4-3.25. These magnitudes reflect that the overall performance of the model is not unreasonable. But it must be noted that the developed projections are conditional and are subject to such uncertainties as the nature of random error, the accuracy of estimated coefficients and forecasted independent variables etc.
which Table 3. Estimated demand models for natural gas. Explanatory
Class of customers Residential/ commercial Industry
variables
NS -0.115 (0.014) 0.047 (0.020)
Price of electricity Price of natural gas Price of oil Price of coal Real income per capita Real GDP Quantity previous Number
(sectoral)
NS -0.113 (0.036) 0.102 (0.042) NS NU
ys891 (0.216) NU
0.673 (0.227) 0.48 1 (0.116) 0.217 (0.075) 0.988
demanded in the 0.369 (0.107) year 0.179 of customers (0.058) 0.962 2.14 0.12
RZ DW P
Norm: Natural to Table 1.
1.98 0.192
gas is not used in the agricultural
sector. See footnotes
Table 4. Estimated demand models for coal.
Explanatory
Class of customers Residential/ Industry commercial
variables
Price of electricity Price of natural gas Price of oil Price of coal Real income Real GDP
per capita (sectoral)
Quantity demanded the previous year RI DW
in
P
NS NS 0.013 (0.002) -0.103 (0.049) -0.029 (0.012) NU
NS NS 0.152 (0.084) -0.120 (0.041) NU 0.340 (0.161) 0.212 (0.069) 0.952 I .98 0.105
0.519 (0.269) 0.886 2.13 0.091
Note,: Coal is not used in the agricultural
Table
I
sector.
See footnotes
to
1.
Energy sector plan Table 5. Estimated demand model for noo-commercial
Explanatory
variable
Estimated coefficients
Constant
2.105
(I ,982) Real mcome
per capita
Real average
price of commercial
Time R2 DW Nore: See footnotes
ENERGY
energy.
lo Table
ECONOMICS
energy
-0.104 (0.048) 0.092 (0.020) -0.047 (0.058) 0.938 1.25
1
July 1987
Having set the national energy targets consistent with historical growth rates and future national aspirations, a national energy plan must be developed to satisfy these energy targets. Such a plan must relate to the rest of the economy, so that the role of energy both as a driving force and as a constraint in the economic development process may be analysed. The development of an integrated energy-economy model was found to be too ambitious, because of its costs and the impossibility of data collection. In consequence, a partial equilibrium model for the Pakistan energy sector was developed and linked with
199
Energy
and economic
growrh:
T. Riaz
Table 6. Energy projection (MGJ).
Year Fuels
1980
Electricity Natural gas Oil Coal Total
108.13 120.34 189.56 44.38 461.41
1985 166.37 193.8 1 241.93 55.30 657.21
1990 255.98 312.13 308.77 67.29 944.17
1995
2ooo
2005
393.82 502.69 366.73 79.92 1343.16
6.60 809.59 435.55 89.54 1940.68
932.4 1303.85 5 17.30 105.32 2X58.88
Power industry demand Coal Oil Natural Total
gas
0.018 0.077 54.50 54.59
0.018 0.077 70.85 70.94
0.319 0.077 89.91 90.31
0.319 0.077 122.00 122.79
2.139 0.077 930.00 932.22
5.993 0.077 1500.00 1506.07
Total commercial 516.00
728.15
non-commercial
359.36
394.82
Total energy
875.36
1122.97
energy
1034.48
1465.95
2872.90
4364.95
469.62
508.41
549.05
1935.57
3381.31
4914.00
Total
model of the Pakistan economy.’ While this approach may appear cruder, it does reflect the decentralized nature of economic reality both in the analysis and in decision-making. The energy-economy interdependence has been established as follows. The energy model has been developed by accepting shadow prices of economic resources (ie capital, labour, foreign exchange etc) and other relevant macro variables from the global macro model. The shadow prices of manpower and foreign exchange have been used in estimating the costs of energy options. These costs, of course, exclude transfer payments. The shadow price of capital has been used as a discount factor. The GNP/GDP growth rates have been used in generating demand projections for each fuel. The basic solution of the energy plan has been obtained with this information incorporated. Then, ‘potential savings’ resulting from the efficient decisions in the energy sector were incorporated into the input-output coefficients of the macro model and it was resolved and results observed. In concluding, it must be accepted that the linkage approach developed has been able to establish only downward linkage successfully. This means that there remains the possibility of social welfare loss. The energy model developed is a generalized linear programming transport model of the KoopmansHitchcock-Kantorovich form and can be written an existing
sThis is a multi-sector, dynamic model of the Pakistan economy. It was developed by A. MacEwan [IO] for the Pakistan economy. The same model has been revised and updated by the author (see Riaz
C161).
200
43 1.66 1466.14
mathematically
as such:9
Min CTX sub to
(10
AX 2 b
(11
x20
(12
where C is a cost vector. A is a ‘constraint’ matrix. b is 2 ‘constraint’ vector and X is a vector of energy capacit! and flows which is to be solved. The model finds a set ofcapacities, outputs and tradt level (ie X) through the energy system (ie A) thal minimizes the total costs (ie C) of energy supplies ovel time to the nation, subject to the constraints (ie b) 01 satisfying specified energy targets, capacity, resource: and trade limitations. The optimization has beer performed by minimizing the present value of tota: costs over the entire planning period. The basic optimal plan based on economic costs and other relevant data as shown in Table 7 is presented in Table 8. The optimal plan favours the development of biogas. coal, gas, oil, and electricity (hydro and gas-based: capacity. This reflects on their comparative advantages. The synthetic oil/gas industries have not been chosen because of their relative comparative costs. The base-load demand for electricity is shared between hydro, nuclear and gas steam plants. The medium-load
‘The model consists ofan objective function (ie minimization of total costs) and seven sets of constraints (ie demand-supply, capacity, firewood situation, import-export limits, maximum resources, maximum capacity, and finally seasonal hydro constraints). It took 0.11 minutes use ofan IBM 370 machine to find an optimal solution. The full algebraic details of the model are given in Riaz [15].
ENERGY
ECONOMICS
July 1987
Energy
and economic
growth: T. Riaz
Table 7. Data used in the computation of the model. Costs (Rs) Fuels and plants Non-commercial Bio-gas Firewood Cylinder gas Kerosene
CCjGJ
VCJGT
32.55 21.94 25.11 19.01
1.95 3.75 7.10 0.22
20.01
11.38
TCjGJ
Existing capacities GJ x lo6
Resources GJ x IO6
Availability factor GJ x lo6
630/year 37/year
0.9 0.9
Efficiency factor GJ x 10”
fuels
Coal Coal production Coal imports Coal exports Oil Oil production Synthetic oil production Oil imports Oil exports Natural gas Gas production Synthetic gas production Gas imports Gas exports Electricity Coal power Oil power Gas steam power Gas turbine power Hydro, run of river Hydro storage Nuclear conventional Nuclear breeder Solar
44.34
11293 48.30 41.05 14.8 21.18 111.30
1780
7.22 13.34
0.70 61.05 51.89 114.9
38.03 111.30
24559
1.22 13.34
0.70 61.05 51.89
311.00 299.00 281.00 209.40 279.80 348.00 800.00 819.00 230.52
31.22 35.54 33.32 30.54 31.10 38.88 44.44 44.44 10.32
0.473 1.987 35.980 23.785 8.407 23.974 4.322 _
316.2l/year
0.65 0.65 0.65 0.65 0.65 0.65 0.70 0.70 0.65
0.30 0.30 0.30 0.20
0.38 0.38 0.10
Notes: Most of the data have been taken from official sources but some have been taken from the open literature especially that which relates to new technologies such as synthetic oil/gas, nuclear breeder and solar etc. For source see Planning Commission 1123. CC = capital costs; VC = variable costs; TC = trade costs; all costs are in 1980 prices. Reserves stand for proven deposits. Oil, gas and coal import and export prices decline after 1982 steadily up to 1992 when these regain their 1980 level again,
demand is satisfied from gas and coal plants whereas the peak load is met by operating oil, coal and gas turbine plants. The trade pattern of optimal solution also reflects the relative availability of fossil fuels in Pakistan. The total cost of the plan amounts to Rs 338.25 billion, whereas the capital costs amount to Rs 173.2 billion and the total foreign exchange needs are about Rs 183.55 billion. The capital and foreign exchange requirements stated in annual terms show that the energy sector will require Rs 6.93 billion in the form of capital funds and Rs 7.34 billion in the form of foreign exchange. The Pakistan energy problem thus turns out to be not a problem of energy resources or technology but of foreign exchange and capital availability. If the implementation of the energy sector plan (which is consistent with economic growth and the shadow..prices of economic resources) did not affect economic scarcities elsewhere in the economy then the
ENERGY ECONOMICS
July 1987
optimal solution is part of a general equilibrium which ensures mazimization of social welfare. However, in view of the large capital and foreign exchange requirements, the economic scarcities are quite likely to be affected. In order to take account of this possibility, the energy model has been solved” using different sets of shadow prices of the basic economic resources (ie capital and foreign exchange). Then the percentage difference between the most expensive and the least expensive plan has been calculated. Then this difference is assumed to represent ‘a potential saving’ resulting from a better choice of energy options. The cost differences (or potential savings) are 36% between the most and least expensive plan. This can be
“‘The changes in the structure of the optimal solution resulting from the changes in the shadow prices ofcapital and foreign exchange are discussed in Riaz [14]. The capital and foreign exchange prices were allowed to vary up to 100% of the level used in the basic model.
201
Table 8. Basic solution: capacity, output and energy trade (MGJ).
Planning periods 1981-85
I98690
199-95
19% moo
ZOO-2Ot
418.99 0.0 22.07 325.500 60.00 0.0 0.0 0.0 0.0 0.0 161.56 10.270 214.01 0.0 0.0 0.0
36.00 0.0 1x.00 0.0 55.0 0.0 0.0 0.0 0.0 75.61 0.0 0.0 70.08 0.0 0.0 0.0
9.5 0.0 6.0 0.0 150. I2 0.0 0.0 0.0 0.0 120.69 0.0 0.0 81.92 0.0 0.0 0.0
9.5 0.0 0.3 0.0 1045.x9 0.0 0.0 0.0 0.0 700.36 0.0 0.0 0.0 0.0 0.0 0.0
11.0 0.0 24.0 0.0 720.65 0.0 0.0 0.0 0.0 375.27 0.0 0.0 0.0 2OO.00 0.0 0.0
0.0 0.0 452.93 0.0 132.62
0.0 1376.16 0.0 0.0 181.55
0.0 1688.75 0.0 0.0 246.40
0.0 2005.7 0.0 0.0 7 14.35
0.0 2382.12 0.0 0.0 982.70
1885.45 0.0 249.29 1531.66 1326.6 0.0 0.0
2066.2 0.0 07.32 0.0 1801.1 0.0 0.0
2253.2 0.0 369.62 0.0 2733.0 0.0 0.0
2445.07 0.0 429.78 0.0 8645.05 0.0 0.0
New capacity Bio-gas plant New forests Coal Oil Natural gas Synthetic gas Synthetic oil Coal power Oil power Gas steam power Gas turbine power Hydro. run of river Hydro storage Nuclear conventional Nuclear breeder Solar Trade Coal imports Oil imports Oil exports Gas imports Gas exports Output Biogas Firewood Coal Oil Gas Synthetic gas Synthetic oil
seen in Table 9 which shows the total discounted costs of six alternative optimal plans. The potential saving of 36% was then incorporated as an increased efficiency in the input- output coefflcients of the energy industries in the macro model. With this efficiency (ie 36%) incorporated. the macro model optimization was repeated. (This is exactly what would have happened if the model was an energyeconomy integrated model.) The new optimal solution did not show any major changes, except that the changes in shadow prices are less than 5% and the
Table 9. Total discounted costs of six alternative
Total discounted
plans (Rs x IO”).
A
B
C
D
E
F
338.25
330.1
300.2
405.3
297.9
312.6
COSIS
Norm,: Plan A shows the cost of basic solution. Plan B shows total costs based on commercial prices. Plan C shows costs based on $ = 15.63 Rs and r = 12.5 %. Plan D shows costs based on $ = 25.0 Rs and r = 12.5 %. Plan E costs are based on $ = 15.63 Rs and r = 20%. Plan F costs are based on $ = 25.0 Rs and r = 20%.
202
2643.65 0.0 507.48 0.0 13706.60 0.0 0.0
value of the maximand (ie national consumption) i: changed less than 2.1% and the effects on production levels are of similar magnitude. These relatively mino changes in the shadow prices and national income car be explained by the large labour earnings from thf Middle East, which have led to an increase in nationa income, capital and foreign exchange availability. Since 1950, Pakistan’s GNP has grown by ar average growth rate of 5%, which has led to an increase: in personal per capita income and personal consump. tion. This increased real income and consumption i: the result of greater industrialization and economic productivity, which are directly associated with the expanded energy use throughout the economy. I continuity of this trend is desirable as a nationai objective and one wishes to avoid energy constraint! on social and economic progress, then Pakistan should adopt energy targets as specified earlier and to achieve these targets the energy plan must be implemented which favours the development of domestic resources of coal, natural gas, oil and hydro and nuclear electricity. It also favours the development of the biogas industry. Since new energy supply options have
ENERGY
ECONOMICS
July 1987
historically taken we cannot count ever, continuous ment, engineering pursued to assure
30 to 50 years to become significant on solar or breeder reactors. Howdiligent efforts in research, developand system integration should be availability to future generations.
Energy+xonomy approach
interaction: an alternative
The two way linkage approach between the energy sector and the rest of the economy, as developed in the last section, is not completely satisfactory, because of different methodological and aggregation levels employed in the two models. Here an alternative approach has been adopted to assess the impact of energy-economy interaction on the level ofincome and energy prices and demand. As of 1980 the value of primary energy inputs in Pakistan has not exceeded 6% of GNP.” The net oil import bill in 1980, stood at Rs 8.921 billion, which is 16.36% of total imports and 30% of total exports in the same year. However, it is also important to note that the net oil import bill of 1980 is only 49% of the labour remittances from the Middle East, which are a direct consequence of the increase in oil prices since 1973. Energy sector investment in Pakistan has been less than 15% of capital formation. These statistics are not conclusive in determining the significance of the energy sector in relation to the rest of the economy. Hogan and Manne have developed a simple single output, two inputs (ie energy and nonimpact of energy) model’Z to show the possible energyeconomy interactions. Using a price/marginal productivity relationship the authors show the importance of energy sector in the national economy: (AF/AE).(E/Y)=(Pe.E)/Y
(13)
where the left hand side of Equation (13) describes the energy elasticity of output with other factors input
” In 1980 the value of primary inputs was Rs 5.05 % of’GNP. “The model is based on two production function of the form
11.653billion,
accounting
identities
which is and
being assumed constant. The right-hand side of Equation (13) shows the value share of energy input as a proportion of total output. The implication is that if energy consumption changes by 1% it leads to n% (n = Pe . E/Y) change of GNP. If one assumes N to remain constant over a wide range of E then, (Y’Y,)=(EIEo)
where Y stands for GNP and zero subscript indicates base year values. The application of (14) allows us to make some inference about the possible impact of energy shortages on the level of GNP. In 1979-80, the value of energy input share as a proportion of Pakistan’s GNP (ie n = Pe . E/Y) was equal to 0.055. This means that a 50% reduction in energy consumption would lead to a mere 3.9% fall in GNP. The assumption of constant n (ie n = Pe. E/Y) is strong and can only be justified if the price elasticity (or substitution is equal to one, which of course is not very likely to happen). Therefore, if we asume that n is not a constant and it is likely to increase to 0.1 because of a rise in oil and other energy prices. In that case, a 50% reduction in the energy consumption can lead to a fall of 6.7% in GNP. This 10% energy sector share can be considered as an upper limit in view of falling energy prices in recent years. Having made some rough estimates to the possible impact of energy shortages on the level of GNP, what is left is to estimate the possible impact of high energy prices on energy demand over time. The long-run price elasticities of demand for commercial fuel can be derived from Tables l-4 and are shown in Table IO. The long-run price elasticities of demand for commercial fuel for different sectors of the economy range from 0.152 to 0.66. However, there are only two values which are higher than 0.5, the rest of the values are fairly small. The estimated demand function for commercial energy in Pakistan is: log E, = 0.875 + 0.661 log GNP, - 0.206 log P, + 0.270 log E,_ 1
one
standard
Y = Pe.E + Px.X
(9
Y = Pe.E + GNP
(ii)
Y = F(E.X)
(14)
error
R2 = 0.97
(0.123)
(0.101)
R-2 =0.965
(0.092)
DW = 2.125
(iii) Table IO. Long-run price elasticities of demand for commercial
The national output, ie Equation (iii), is maximized, subject to factors availability as such. Max
F(E, X)-Pe.E-Px.X
(iv)
where P denotes price, E stands for energy and X for non-energy. Yis no energy sector output. gee Hogan and Manne [9] for details.
ENERGY
ECONOMICS
July 1987
Sectoral demand
Electricity
Natural
Residential/commercial Industrial
0.415 0.238
Agriculture
0.239
0.182 0.218 -
gas
fuels,
Oil
Coal
0.229 0.665 0.547
0.2 14 0.152 -
203
EnergJl and economic growth:
T. Riaz
It gives a long-run price elasticity of 0.28. With this price elasticity, the possible price increase of 35% this is the possible difference between the most expensive and the least expensive energy plan - can lead to a fall of 9.8% in the energy demand. But if we assume that the long-run price elasticity is equal to 0.5, in that case a 35% price increase can lead to a reduction in demand of about 17%. The demand variations of this magnitude are well within the forecasting errors. In conclusion, it is evident that the balance of payment difficulties created by the high oil import bill have been compensated for by the opening up of the Middle Eastern markets to Pakistani goods and labour. With these opportunities Pakistan is perfectly capable of avoiding any energy constraints and thus maintaining her historical growth rate, which is essential for social economic progress.
6
10 11
12 13
14
References P. Balestra, The Demandfor Natural Gas in the United States, North-Holland, Amsterdam, 1967. L. G. Brookes, ‘More on the output elasticity of energy consumption’, Journal of Industrial Economics, Vol 21, No 1, 1973, pp 83-92. E. Cook, Man, Energy, Society, W. H. Freeman, San Francisco, 1976, 192. Government of Pakistan, Pakistan Statistical Year Book, various issues. Government of Pakistan, Pakistan Economic Surve) 1980-81, Ministry of Finance, Islamabad, 1980.
104
15
16
17
18 19
Government of Pakistan, Pakistan Energ), Book, Min istry of Petroleum and Natural Resources, Islamabad 197&80. Government of Pakistan, Pakistan Oil and Gas Pricing Ministry of Petroleum and Natural Resources, 1980. Z. Grilliches, ‘Distributed lags: a survey’. Econometrica Vol 35, January 1976, pp 16-35. W. Hogan and A. S. Manne. ‘Energy economy inter. action: the fable of the elephant and the rabbit’, in R S. Pindyck, ed, Advances in the Econorniu of Energ) and Resources, JAI Press, Connecticut. USA. 1978. A. Macewan, Development Alternative in Pakistan Harvard University Press, Cambridge. MA. 1971. Malih and Siddique, ‘Coal mining industry in the energy perspective of Pakistan’, Pakistan Mineral Develop. ment Corporation, Pakistan, 198 I. Pakistan Planning Commission, Proforma P.C.I. Power and Fuel Projects, 1968-l 972. M. Posner et al, Energ!, Economics: Growth, RcJsourcc and Policies, Cambridge University Press, Cambridge, UK, 1980. T. Riaz, Pakistan: The Energ), Sector: A Study irl Sectoral Planning, Ferozsons, Lahore. 1984. T. Riaz, ‘Energy policy formation for Pakistan: an optimization approach’, Energ!, Economics, Vol 3, No 3, July 1981, pp 191-197. T. Riaz, A Multi-Sector Dynamic Model,for the Pakistan Economy, discussion paper, Economics Department,
Newcastle Polytechnic, 1978. C. Starr and S. Field, ‘Economic growth. employment and energy’, Enfrgy Policy, Vol 7, No 1, March 1979, pp 2-22. Water and Power Development Authority, WAPDA Financial Statistics, various issues, Pakistan. Water and Power Development Authority, A Handbook qf Data on WAPDA, Pakistan.
ENERGY
ECONOMICS
Julv 1987
T. Riaz
The paper investigates the relationship between energy consumption and economic growth with special reference to Pakistan. It then establishes future energy needs of the country and presents a long-run energy plan to meet these requirements. The energy-economy relationship is explained. The analysis shows that energy shortages can be extremely harmfulfor a developing country like Pakistan. However, the energy problem is one of shortages of capital andforeign exchange which can be dealt with by prudent planning of remittances from the Middle East. Ke_vwwd.r
Economic
growth;
Energy
consumption;
Since 1973, oil and subsequently all energy prices have experienced rapid changes, which in turn have resulted in supply-side disturbances for the energy importing economies. During the 1970s and early 1980s energy prices mostly rose, which in most cases led to an upward shift in the supply curve, a fall in national output and a corresponding rise in the rate of inflation. More recently, oil and subsequently most energy prices have fallen, leading to a downward shift in the supply curve, a rise in output and a corresponding fall in the rate of inflation. The last 15 years, therefore, have been a period of disequilibrium, structural adjustment and adoption of strategies to cope with the previously less recognized phenomena of stagflation. This paper sets out first to establish an empirical relationship between energy consumption and economic growth with specific reference to Pakistan. It then estimates energy demand for Pakistan and presents a plan that is consistent with planned social and economic progress and which is unconstrained by energy availability. The article develops an integrated energy sector plan which is partially linked to the rest of the economy. Finally, an explicit energy-economy
The author is with the School of Economics, Newcastleupon-Tyne Polytechnic, Northumberland Building, Northumberland Road, Newcastle-upon-Tyne, NE1 8ST, UK. Final manuscript
received 28 August
0140-9883/87/030195-IO
$03.00
1986.
0
1987 Butterworth
Interrelationship
interactive framework is adapted to assess the interlinkage impact and to formulate policy.
Energy and economic growth Energy has played a vital role in taking mankind from an early basic existence to the present supersonic era. In the early stages of economic development, most economies were agricultural and depended mainly on human labour. During industrialization energy was used intensively to produce artifacts which combined with human labour raised productivity, output and living standards. The urbanization phase, which closely corresponded with the industrialization process, intensified energy consumption. The history of economic development thus shows a parallel development in energy consumption in more and more convenient forms. There is an obvious reason for this economic growth depends on increasing use of factors of production (energy being one of the factors) as well as their efficient use; at the same time, it also depends on the effective demand for goods and services. The empirical evidence on the energy-growth relationship can be analysed in a number of ways: (i) by constructing a scattergram of per capita energy consumption and GNP for a group of countries for a specified time; (ii) by finding energy-output ratios over time and space; and
& Co (Publishers)
Ltd
195
Energy and economic growth: T. Riaz (iii) by establishing single and multivariate relationships over time.
statistical
Posner r/ UI [ 131 have estimated a number of energyoutput ratios for three groups of countries. They have found that such ratios show a trend of rising at the early stages of economic development declining slightly with industrial maturity and then settle down at a fairly high level. The increase in the ratio is explained in terms of industrialization, urbanization and fuels substitution. The relative decline and stability results from improvements in energy use efficiencies and economic transition to a service-based economy. The energy-output ratio for Pakistan has increased from 1.05 in 1965 to 1.45 in 1980, reflecting structural changes in the economic system. This ratio is likely to go up again before it shows any decline or stability. A single or multi-variable statistical analysis can be relatively more precise for a single country. The estimated energy-GNP and GNP-energy relationships for Pakistan for 1960-80 are shown below. log E, = -0.3679 + 1.2274 log GNP, (0.0264) SE (40.42) t ratio R2 = 0.991 R2 = 0.990
(1)
log GNP, = 0.3824 + 0.8080 logE, (0.0774) SE (46.42) t ratio R 2 = 0.992 w2 = 0.990
(2)
= commercial energy consumption = gross national product = standard error = time period = R square =adjusted R square in parentheses are t ratios
The estimated energy consumption as a function of gross national product, Equation (l), is statistically significant, shows a negative intercept and positive slope which is consistent with the hypothesis that at an early stage of economic development each country relies heavily on non-commercial sources of energy: however, with time commercial fuels are substituted for the non-commercial and increasingly used during the industrialization and urbanization process. The estimated coefficient of 1.2274 is the income elasticity of energy consumption, which means that, if ceferis parihus, gross national product increases by 1%, then commercial energy consumption increases by 1.23%. The inverse relation between gross domestic product and energy consumption, ie Equation (2), is also
196
log GNP, = 1.1464 + 0.3836 log L, + 0.4125 log K, + 0.3994 log E, (3) (0.1213) (0.1825) (0.0298) SE t ratio (2.89) (3.18) (3.29) R2 = 0.957 R’ = 0.949 F= 111.62 DW = 1.695 where
where E GNP SE t R R2 Figures
statistically significant. The estimated parameter 0.8080 represents the commercial energy elasticity of GNP which shows that if ceteris parihus. energy consumption increases by 1% then Pakistan’s GNP will increase by 0.8%. The estimates, ie Equations (1) and (2), indicate the interdependence of the two variables. These estimates, however, must be interpreted with caution, as the energy consumption figures do not include non-commercial energy which is used extensively in Pakistan, and the conversion of different fuels into a common denominator does not take account of fuel efficiency in different uses. In order to find the partial contribution of each factor of production to economic growth in Pakistan, it was decided to estimate an aggregate production function with special emphasis on energy input. Because of limited data availability it was possible to estimate such a function only for the manufacturing sector of the economy. The best estimated function is given below:’
GNP L K E
= = = =
gross national product manufacturing employment manufacturing capital stock commercial energy consumption
t. SE. R2 and 7i2 are as before. The generalized least square method has been used for estimation purposes and the results given in Equation (3) have been obtained from the transformed equation, which was obtained using the generalized differences process. All estimated coefficients are economically and statistically significant. The explanatory variables jointly explain all except about 0.043% of the variation in the dependent variable log Y. The estimated coefficients are elasticities with respect to each factor of production included in Equation (3). The sum of these elasticities is 1.195, which indicates increasing returns to scale in the manufacturing sector. But it should not
‘The estimates are based on data from the period 1959/60 to 197X/79 and were provided by the Institute of Development Economics, Islamabad. Pakistan. It is realized that the estimated function possibly presents a numberofproblems, eg theerror term may not be random or have normal distribution; the independent variables may not be random or independent of error term; and there may be specification error and the assumptions to perfect competition and profit maximization may not be satisfied. The justification for these estimations is given in Riaz [14].
ENERGY
ECONOMICS
July 1987
Energy and economic growth: T. Riaz assume2 to imply increasing return for the economy as a whole. The commercial energy elasticity of manufacturing output has been estimated as 0.3994. This means, if, ceteris paribus, commercial energy consumption increases by 1% then manufacturing output will increase by almost 0.4%. The explanation of other elasticities is similar. From the above analysis, it is clear that economic growth and energy consumption are closely related and that the economic development process is an energy intensive process. Hence a basic inference may be drawn that the social and economic structure of an economy could be substantially and regressively altered by large constraints upon energy use.3 An extreme version of this argument will imply that energy savings in the developing countries cannot be achieved unless GNP is proportionally reduced pari passu with energy consumption. Thus, in order to avoid stagnation or even decline in GNP, these countries must avoid energy supply shortages, which can be caused by controlled energy prices, delays in the development of feasible technical options or shortages of foreign exchange. First it is essential to estimate future energy demand, which is consistent with the societal progress target of the country and only then may one plan for adequate supplies.
Energy forecast for Pakistan A number of energy forecasts4 have been based on the relationship as in Equation (1) or a single coefficient such as energy-output ratio. While such empirical findings and their interpretation are interesting, one cannot assume that there is a stable ‘world line’ to forecast energy consumption over time. Each country’s energy demand is determined by individual factors such as climate, orientation of the economy, efficiency of industrial and household conversions and energy prices. Thus, the forecasts based on the aggregate analysis cannot be taken more than as just a rough estimate. Hence, a need for a more refined approach. Energy demand behaviour is reflected in two sequent decisions - first in the purchase of a specific fuelburning durable good and second in the frequency and intensity of its use. The stock of fuel-burning durable goods is fixed in the short run, but is variable in the ‘Such an assumption could have serious implications for the analysis conducted below. 3Because of this, I am opposed to energy demand management by the government. %ee for instance Starr and Field [I73 who have adopted the following approach. First they forecast employment levels using population forecasts, then using forecasted level of einployment. They forecast GNP and then use this forecasted GNP in an energy/GNP empirical relationship to project future energy demand.
ENERGY
ECONOMICS
July 1987
long run. To take account of these features and economic situations in Pakistan, a combination of behavioural and fixed coefficients type models have been developed and estimated to predict future energy demand for the country. The adapted model is a partial adjustment type which takes account of the consumer’s habit formation and the vital role of fuel burning durable goods. (i= 1, 2, 3 . . . N)
rlc = Iy + BXit + U,’ x-r,_,=E,(y--Y,_,)+U,
(4) (5)
where c is the ideal (or desired) but unobserved level of the dependent variable (ie fuel demand) at time t. Xi, are the independent variables (eg own price, income/output, cross prices) which determine the level of each fuel demand. x is the actual observed level of the dependent variable (ie fuel consumption) at time t and /. is an adjustment coefficient which indicates the rate of adjustment of Y, to v and its value lies between zero and 1. The value of Y: in Equation (4) is not known but it is assumed that the consumer wishes to move from x to r;*. However, he is only partially successful during one time period. The reasons why x does not adjust to I;* in one period are many and may include economic, technological and institutional constraints as well as habit formation.5 The relation (5) provides a link between x andY,*, solving for ,Iv shows: nr;*=I:+(n-l)k;_,-U i = (OZ+ fiXi, + U,) substituting
(6)
Y: in Equation
Y;=crI,+/?LX,+(l
-A)x_,
(6) gives: +e,
(7)
where e,=1U,+ Following equations
U, the above, is obtained:
the following
set of demand
where xj, stands for the quantity demanded of fuel i (i = electricity, gas, oil and coal) in sector iti = agricultural, industrial and domestic/commercial) in time period 1. Pij, stands as the unit price of fuel i in sector j in time period I. Xkj, stands for independent variables i(k = income/output, number of customers) in sector j in time period t, and Y,_ 1 is one year lagged demand. The transport sector and the power industry’s demand for fuels has been estimated separately using an input-output type framework. The transport sector demand for petroleum products is based on the 5For a good discussion on habit formation Balestra [I].
see Grillinches
[8] and
197
number of vehicles and estimated fuel consumption per vehicle per year.” The power industry demand for fuels has been estimated using a fuel efficiency factor of different thermal plants and their annual output, which in turn has been determined by an optimal energy plan. The non-commercial energy demand is explained by the following model.
NCE, = F( r,, P,, 7;)
(9)
where NCE represents per head non-commercial energy consumption; Y stands for per head real income; P stands for per unit real prices of commercial energy; Tstands for time related social and economic progress. The subscript t indicates time period in years. The set of demand equations as set out in Equation (8) and in log linear form have been estimated using the generalized least squares method. The data consist of 20 annual observations. The prices used as explanatory variables are the real (ie deflated) average prices. The income and output variable has been deflated using the GNP deflator. The sources of data are as follows: quantity demanded of each fuel, price of oil and natural gas, Ministry of Petroleum and National Research [6, 71; price of electricity and number of customers, Water and Power Development Authority [18, 191; price of coal, Malih and Siddique [ 1 I]; deflator, GNP/GDP and population, Pakistan Economic Surre~q [S]; number of gas customers, Pukistan Stuti.sticu/ Year Book [4]. The estimated relations are shown in Tables l-5. ‘The transport sector does not consume any natural gas and the consumption of coal and electricity is strictly confined to railways. The railways demand for coal and electricity and the aviation demand for petroleum products are based on a simple trend. Remaining demand estimates were arrived at by developing a dilTerence equation such as: X,=/(X,_ ,. DX,. AX,) where D is dropout rate and A stands for new vehicles. To develop time series data for buses. tractors, cars and light vehicles. and then using the following fuel consumption figures, High-speed diesel consumption,‘tractor/year = 18.83 tons Htgh-speed diesel consumption/bus/year = 10.69 tons Motor spirit consumption/car/year = 1.237 tons Motor spirit consumption/2 &/wheels;year = 0.412 tons These estimates are given by the Pakistan Planning Commision [12]. ‘A number of statistical problems faced in estimation or prior to the estimation were resolved as follows. The problem ofjoint determination ofdependent and independent variables was resolved simply as energy prices in Pakistan are set by pubhc regulation rather than market forces. Furthermore refuge is taken under the partial equilibrium blanket, accepting the view that a vast system of simultaneous equations has its place in economic theory, but it may not contribute much to the cause of empirical research. The problem or identification is serious especially in the presence of cross-prtce variables. For such a model there exists no valid criterion to test for identification. The structural parameters of our model, however, satisfy ordinary statistical tests. The problem of proJect collinearity was not faced as the matrix of the independent variables (ie xx) was found to be non-singular. The generalized least square has been used todeal with themilderform ofmulticollinearityand autocorrelation.
198
Table
1. Estimated
Explanator)
demand models for electricity.
variables
Price of electricity Prrcc of natural
gas
Price of oil Price of coal Real income per capita Real GDP
(sectoral)
Quantity previous
demanded year
Number
of customers
Class of customers Residential commercial Industry
Agriculture
-0.117 (0.042) NS
-0.08X (0.03 1) NU
0.02 I (0.012) NS 0.416 (0.151) NU
- 0.092 (0.023) -0.01 7 (0.006) NS NS NU 0.59x
0.032 (0.0 I 2) N L’ NL
(0.128)
0.403 (0.065)
0.614 (0.301) 0.301 (0.976) 0.976 1.95 0.210
0.632 (0.213) 0.178 (0.055) 0.9x5 1.x9 0.269
in 0.718 (0.235) 0.250 (0.109) 0.95 2.15 0.165
R2 DW P
Norc. All prices are Rupees per gigajoule (Rs.‘GJ). All money unit have been deflated. All variables were transformed into natural lo form except the number ol customers whose partial impact wa estimated separately. NS = statistically non-significant at 5% levc and was dropped from estimation. NU =either ‘Not used’ o ‘not included’ in the sector. R2 stands for adjusted R square. 1 stands for estimated Hildreth Lu coefficients. The figures in paren theses are standard errors of the estimates.
The estimates,’ in most cases, are economically am statistically significant and provide a reasonable ex planation of energy demand in Pakistan. At the root of all energy projections lies a judgemen on the probable level of output of goods and services o a future economy which is a quantitative corollary of; perceived future level of economic welfare and develop ment. The future perception of economic growth Table 2. Estimated demand models for oil.
Explanatory
variables
Class of customers Residential/ commercial Industry
Price of electricity Price of natural gas
NS 0.132 (0.054) -0.106 (0.033)
Price of oil Price of coal Real income per capita Real GDP
(sectoral)
Quantity demanded the previous year RZ DW P
Nolc: See footnotes
in
ts63 (0.093) NU
-0.219 (0.102) NU NU
0.23 I (0.016)
0.426 (0.205)
0.538
0.761
(0.227) 0.928 2.01
(0.320) 0.952 2.11
0.28
IO Table
ENERGY
Agriculture
NS 0.093 (0.015) -0.159 (0.022) NS NU
0.205
NS
NU
'0.600'
(0.175) 0.900 I .95 0.143
1.
ECONOMICS
July 1985
Energy
and economic
growth:
T. Ria:
is also realistic, must have some relation to the historical growth rates of employment, productivity and output. Thus in developing energy projections for Pakistan it is assumed that the country will follow its historical growth path. Thus the changes in GNP. GNI and sectoral output have been obtained by establishing time path predictions to 2005. The trend growth rates are GNP = 6.3%; industrial output = 7.5%; agricultural output = 2.9% and population growth = 2.5 to 2.9%. The population forecast is taken from the census series. The number of customers is based on planned targets. The energy prices are assumed to move to their internal levels and then start to decline, reaching their 1974 level in 1990 and then these prices will rise again reaching their 1980 level in 1992 and then maintain their values in real terms. The energy projections based on the estimated models and the future perceptions as outlined above are shown in Table 6. The projections show an average growth rate of 7.5% over the planning horizon. The consumption of commercial fuels will increase (7.74%) at a greater rate compared with non-commercial energy (1.7 I O/u).This reflects on the energy-intensive industrialization, urbanization and substitution of commercial for noncommercial fuels. However, it must be pointed out that in spite of this rapid increase in energy consumption, per capita energy consumption in Pakistan in 2005 will still be about 37.4 GJ, which is far below per capita consumption of the advanced industrialized countries. The tracking performance of the model has been checked by calculating the mean percentage errors of the dependent variables. It is found to fall in the range 0.4-3.25. These magnitudes reflect that the overall performance of the model is not unreasonable. But it must be noted that the developed projections are conditional and are subject to such uncertainties as the nature of random error, the accuracy of estimated coefficients and forecasted independent variables etc.
which Table 3. Estimated demand models for natural gas. Explanatory
Class of customers Residential/ commercial Industry
variables
NS -0.115 (0.014) 0.047 (0.020)
Price of electricity Price of natural gas Price of oil Price of coal Real income per capita Real GDP Quantity previous Number
(sectoral)
NS -0.113 (0.036) 0.102 (0.042) NS NU
ys891 (0.216) NU
0.673 (0.227) 0.48 1 (0.116) 0.217 (0.075) 0.988
demanded in the 0.369 (0.107) year 0.179 of customers (0.058) 0.962 2.14 0.12
RZ DW P
Norm: Natural to Table 1.
1.98 0.192
gas is not used in the agricultural
sector. See footnotes
Table 4. Estimated demand models for coal.
Explanatory
Class of customers Residential/ Industry commercial
variables
Price of electricity Price of natural gas Price of oil Price of coal Real income Real GDP
per capita (sectoral)
Quantity demanded the previous year RI DW
in
P
NS NS 0.013 (0.002) -0.103 (0.049) -0.029 (0.012) NU
NS NS 0.152 (0.084) -0.120 (0.041) NU 0.340 (0.161) 0.212 (0.069) 0.952 I .98 0.105
0.519 (0.269) 0.886 2.13 0.091
Note,: Coal is not used in the agricultural
Table
I
sector.
See footnotes
to
1.
Energy sector plan Table 5. Estimated demand model for noo-commercial
Explanatory
variable
Estimated coefficients
Constant
2.105
(I ,982) Real mcome
per capita
Real average
price of commercial
Time R2 DW Nore: See footnotes
ENERGY
energy.
lo Table
ECONOMICS
energy
-0.104 (0.048) 0.092 (0.020) -0.047 (0.058) 0.938 1.25
1
July 1987
Having set the national energy targets consistent with historical growth rates and future national aspirations, a national energy plan must be developed to satisfy these energy targets. Such a plan must relate to the rest of the economy, so that the role of energy both as a driving force and as a constraint in the economic development process may be analysed. The development of an integrated energy-economy model was found to be too ambitious, because of its costs and the impossibility of data collection. In consequence, a partial equilibrium model for the Pakistan energy sector was developed and linked with
199
Energy
and economic
growrh:
T. Riaz
Table 6. Energy projection (MGJ).
Year Fuels
1980
Electricity Natural gas Oil Coal Total
108.13 120.34 189.56 44.38 461.41
1985 166.37 193.8 1 241.93 55.30 657.21
1990 255.98 312.13 308.77 67.29 944.17
1995
2ooo
2005
393.82 502.69 366.73 79.92 1343.16
6.60 809.59 435.55 89.54 1940.68
932.4 1303.85 5 17.30 105.32 2X58.88
Power industry demand Coal Oil Natural Total
gas
0.018 0.077 54.50 54.59
0.018 0.077 70.85 70.94
0.319 0.077 89.91 90.31
0.319 0.077 122.00 122.79
2.139 0.077 930.00 932.22
5.993 0.077 1500.00 1506.07
Total commercial 516.00
728.15
non-commercial
359.36
394.82
Total energy
875.36
1122.97
energy
1034.48
1465.95
2872.90
4364.95
469.62
508.41
549.05
1935.57
3381.31
4914.00
Total
model of the Pakistan economy.’ While this approach may appear cruder, it does reflect the decentralized nature of economic reality both in the analysis and in decision-making. The energy-economy interdependence has been established as follows. The energy model has been developed by accepting shadow prices of economic resources (ie capital, labour, foreign exchange etc) and other relevant macro variables from the global macro model. The shadow prices of manpower and foreign exchange have been used in estimating the costs of energy options. These costs, of course, exclude transfer payments. The shadow price of capital has been used as a discount factor. The GNP/GDP growth rates have been used in generating demand projections for each fuel. The basic solution of the energy plan has been obtained with this information incorporated. Then, ‘potential savings’ resulting from the efficient decisions in the energy sector were incorporated into the input-output coefficients of the macro model and it was resolved and results observed. In concluding, it must be accepted that the linkage approach developed has been able to establish only downward linkage successfully. This means that there remains the possibility of social welfare loss. The energy model developed is a generalized linear programming transport model of the KoopmansHitchcock-Kantorovich form and can be written an existing
sThis is a multi-sector, dynamic model of the Pakistan economy. It was developed by A. MacEwan [IO] for the Pakistan economy. The same model has been revised and updated by the author (see Riaz
C161).
200
43 1.66 1466.14
mathematically
as such:9
Min CTX sub to
(10
AX 2 b
(11
x20
(12
where C is a cost vector. A is a ‘constraint’ matrix. b is 2 ‘constraint’ vector and X is a vector of energy capacit! and flows which is to be solved. The model finds a set ofcapacities, outputs and tradt level (ie X) through the energy system (ie A) thal minimizes the total costs (ie C) of energy supplies ovel time to the nation, subject to the constraints (ie b) 01 satisfying specified energy targets, capacity, resource: and trade limitations. The optimization has beer performed by minimizing the present value of tota: costs over the entire planning period. The basic optimal plan based on economic costs and other relevant data as shown in Table 7 is presented in Table 8. The optimal plan favours the development of biogas. coal, gas, oil, and electricity (hydro and gas-based: capacity. This reflects on their comparative advantages. The synthetic oil/gas industries have not been chosen because of their relative comparative costs. The base-load demand for electricity is shared between hydro, nuclear and gas steam plants. The medium-load
‘The model consists ofan objective function (ie minimization of total costs) and seven sets of constraints (ie demand-supply, capacity, firewood situation, import-export limits, maximum resources, maximum capacity, and finally seasonal hydro constraints). It took 0.11 minutes use ofan IBM 370 machine to find an optimal solution. The full algebraic details of the model are given in Riaz [15].
ENERGY
ECONOMICS
July 1987
Energy
and economic
growth: T. Riaz
Table 7. Data used in the computation of the model. Costs (Rs) Fuels and plants Non-commercial Bio-gas Firewood Cylinder gas Kerosene
CCjGJ
VCJGT
32.55 21.94 25.11 19.01
1.95 3.75 7.10 0.22
20.01
11.38
TCjGJ
Existing capacities GJ x lo6
Resources GJ x IO6
Availability factor GJ x lo6
630/year 37/year
0.9 0.9
Efficiency factor GJ x 10”
fuels
Coal Coal production Coal imports Coal exports Oil Oil production Synthetic oil production Oil imports Oil exports Natural gas Gas production Synthetic gas production Gas imports Gas exports Electricity Coal power Oil power Gas steam power Gas turbine power Hydro, run of river Hydro storage Nuclear conventional Nuclear breeder Solar
44.34
11293 48.30 41.05 14.8 21.18 111.30
1780
7.22 13.34
0.70 61.05 51.89 114.9
38.03 111.30
24559
1.22 13.34
0.70 61.05 51.89
311.00 299.00 281.00 209.40 279.80 348.00 800.00 819.00 230.52
31.22 35.54 33.32 30.54 31.10 38.88 44.44 44.44 10.32
0.473 1.987 35.980 23.785 8.407 23.974 4.322 _
316.2l/year
0.65 0.65 0.65 0.65 0.65 0.65 0.70 0.70 0.65
0.30 0.30 0.30 0.20
0.38 0.38 0.10
Notes: Most of the data have been taken from official sources but some have been taken from the open literature especially that which relates to new technologies such as synthetic oil/gas, nuclear breeder and solar etc. For source see Planning Commission 1123. CC = capital costs; VC = variable costs; TC = trade costs; all costs are in 1980 prices. Reserves stand for proven deposits. Oil, gas and coal import and export prices decline after 1982 steadily up to 1992 when these regain their 1980 level again,
demand is satisfied from gas and coal plants whereas the peak load is met by operating oil, coal and gas turbine plants. The trade pattern of optimal solution also reflects the relative availability of fossil fuels in Pakistan. The total cost of the plan amounts to Rs 338.25 billion, whereas the capital costs amount to Rs 173.2 billion and the total foreign exchange needs are about Rs 183.55 billion. The capital and foreign exchange requirements stated in annual terms show that the energy sector will require Rs 6.93 billion in the form of capital funds and Rs 7.34 billion in the form of foreign exchange. The Pakistan energy problem thus turns out to be not a problem of energy resources or technology but of foreign exchange and capital availability. If the implementation of the energy sector plan (which is consistent with economic growth and the shadow..prices of economic resources) did not affect economic scarcities elsewhere in the economy then the
ENERGY ECONOMICS
July 1987
optimal solution is part of a general equilibrium which ensures mazimization of social welfare. However, in view of the large capital and foreign exchange requirements, the economic scarcities are quite likely to be affected. In order to take account of this possibility, the energy model has been solved” using different sets of shadow prices of the basic economic resources (ie capital and foreign exchange). Then the percentage difference between the most expensive and the least expensive plan has been calculated. Then this difference is assumed to represent ‘a potential saving’ resulting from a better choice of energy options. The cost differences (or potential savings) are 36% between the most and least expensive plan. This can be
“‘The changes in the structure of the optimal solution resulting from the changes in the shadow prices ofcapital and foreign exchange are discussed in Riaz [14]. The capital and foreign exchange prices were allowed to vary up to 100% of the level used in the basic model.
201
Table 8. Basic solution: capacity, output and energy trade (MGJ).
Planning periods 1981-85
I98690
199-95
19% moo
ZOO-2Ot
418.99 0.0 22.07 325.500 60.00 0.0 0.0 0.0 0.0 0.0 161.56 10.270 214.01 0.0 0.0 0.0
36.00 0.0 1x.00 0.0 55.0 0.0 0.0 0.0 0.0 75.61 0.0 0.0 70.08 0.0 0.0 0.0
9.5 0.0 6.0 0.0 150. I2 0.0 0.0 0.0 0.0 120.69 0.0 0.0 81.92 0.0 0.0 0.0
9.5 0.0 0.3 0.0 1045.x9 0.0 0.0 0.0 0.0 700.36 0.0 0.0 0.0 0.0 0.0 0.0
11.0 0.0 24.0 0.0 720.65 0.0 0.0 0.0 0.0 375.27 0.0 0.0 0.0 2OO.00 0.0 0.0
0.0 0.0 452.93 0.0 132.62
0.0 1376.16 0.0 0.0 181.55
0.0 1688.75 0.0 0.0 246.40
0.0 2005.7 0.0 0.0 7 14.35
0.0 2382.12 0.0 0.0 982.70
1885.45 0.0 249.29 1531.66 1326.6 0.0 0.0
2066.2 0.0 07.32 0.0 1801.1 0.0 0.0
2253.2 0.0 369.62 0.0 2733.0 0.0 0.0
2445.07 0.0 429.78 0.0 8645.05 0.0 0.0
New capacity Bio-gas plant New forests Coal Oil Natural gas Synthetic gas Synthetic oil Coal power Oil power Gas steam power Gas turbine power Hydro. run of river Hydro storage Nuclear conventional Nuclear breeder Solar Trade Coal imports Oil imports Oil exports Gas imports Gas exports Output Biogas Firewood Coal Oil Gas Synthetic gas Synthetic oil
seen in Table 9 which shows the total discounted costs of six alternative optimal plans. The potential saving of 36% was then incorporated as an increased efficiency in the input- output coefflcients of the energy industries in the macro model. With this efficiency (ie 36%) incorporated. the macro model optimization was repeated. (This is exactly what would have happened if the model was an energyeconomy integrated model.) The new optimal solution did not show any major changes, except that the changes in shadow prices are less than 5% and the
Table 9. Total discounted costs of six alternative
Total discounted
plans (Rs x IO”).
A
B
C
D
E
F
338.25
330.1
300.2
405.3
297.9
312.6
COSIS
Norm,: Plan A shows the cost of basic solution. Plan B shows total costs based on commercial prices. Plan C shows costs based on $ = 15.63 Rs and r = 12.5 %. Plan D shows costs based on $ = 25.0 Rs and r = 12.5 %. Plan E costs are based on $ = 15.63 Rs and r = 20%. Plan F costs are based on $ = 25.0 Rs and r = 20%.
202
2643.65 0.0 507.48 0.0 13706.60 0.0 0.0
value of the maximand (ie national consumption) i: changed less than 2.1% and the effects on production levels are of similar magnitude. These relatively mino changes in the shadow prices and national income car be explained by the large labour earnings from thf Middle East, which have led to an increase in nationa income, capital and foreign exchange availability. Since 1950, Pakistan’s GNP has grown by ar average growth rate of 5%, which has led to an increase: in personal per capita income and personal consump. tion. This increased real income and consumption i: the result of greater industrialization and economic productivity, which are directly associated with the expanded energy use throughout the economy. I continuity of this trend is desirable as a nationai objective and one wishes to avoid energy constraint! on social and economic progress, then Pakistan should adopt energy targets as specified earlier and to achieve these targets the energy plan must be implemented which favours the development of domestic resources of coal, natural gas, oil and hydro and nuclear electricity. It also favours the development of the biogas industry. Since new energy supply options have
ENERGY
ECONOMICS
July 1987
historically taken we cannot count ever, continuous ment, engineering pursued to assure
30 to 50 years to become significant on solar or breeder reactors. Howdiligent efforts in research, developand system integration should be availability to future generations.
Energy+xonomy approach
interaction: an alternative
The two way linkage approach between the energy sector and the rest of the economy, as developed in the last section, is not completely satisfactory, because of different methodological and aggregation levels employed in the two models. Here an alternative approach has been adopted to assess the impact of energy-economy interaction on the level ofincome and energy prices and demand. As of 1980 the value of primary energy inputs in Pakistan has not exceeded 6% of GNP.” The net oil import bill in 1980, stood at Rs 8.921 billion, which is 16.36% of total imports and 30% of total exports in the same year. However, it is also important to note that the net oil import bill of 1980 is only 49% of the labour remittances from the Middle East, which are a direct consequence of the increase in oil prices since 1973. Energy sector investment in Pakistan has been less than 15% of capital formation. These statistics are not conclusive in determining the significance of the energy sector in relation to the rest of the economy. Hogan and Manne have developed a simple single output, two inputs (ie energy and nonimpact of energy) model’Z to show the possible energyeconomy interactions. Using a price/marginal productivity relationship the authors show the importance of energy sector in the national economy: (AF/AE).(E/Y)=(Pe.E)/Y
(13)
where the left hand side of Equation (13) describes the energy elasticity of output with other factors input
” In 1980 the value of primary inputs was Rs 5.05 % of’GNP. “The model is based on two production function of the form
11.653billion,
accounting
identities
which is and
being assumed constant. The right-hand side of Equation (13) shows the value share of energy input as a proportion of total output. The implication is that if energy consumption changes by 1% it leads to n% (n = Pe . E/Y) change of GNP. If one assumes N to remain constant over a wide range of E then, (Y’Y,)=(EIEo)
where Y stands for GNP and zero subscript indicates base year values. The application of (14) allows us to make some inference about the possible impact of energy shortages on the level of GNP. In 1979-80, the value of energy input share as a proportion of Pakistan’s GNP (ie n = Pe . E/Y) was equal to 0.055. This means that a 50% reduction in energy consumption would lead to a mere 3.9% fall in GNP. The assumption of constant n (ie n = Pe. E/Y) is strong and can only be justified if the price elasticity (or substitution is equal to one, which of course is not very likely to happen). Therefore, if we asume that n is not a constant and it is likely to increase to 0.1 because of a rise in oil and other energy prices. In that case, a 50% reduction in the energy consumption can lead to a fall of 6.7% in GNP. This 10% energy sector share can be considered as an upper limit in view of falling energy prices in recent years. Having made some rough estimates to the possible impact of energy shortages on the level of GNP, what is left is to estimate the possible impact of high energy prices on energy demand over time. The long-run price elasticities of demand for commercial fuel can be derived from Tables l-4 and are shown in Table IO. The long-run price elasticities of demand for commercial fuel for different sectors of the economy range from 0.152 to 0.66. However, there are only two values which are higher than 0.5, the rest of the values are fairly small. The estimated demand function for commercial energy in Pakistan is: log E, = 0.875 + 0.661 log GNP, - 0.206 log P, + 0.270 log E,_ 1
one
standard
Y = Pe.E + Px.X
(9
Y = Pe.E + GNP
(ii)
Y = F(E.X)
(14)
error
R2 = 0.97
(0.123)
(0.101)
R-2 =0.965
(0.092)
DW = 2.125
(iii) Table IO. Long-run price elasticities of demand for commercial
The national output, ie Equation (iii), is maximized, subject to factors availability as such. Max
F(E, X)-Pe.E-Px.X
(iv)
where P denotes price, E stands for energy and X for non-energy. Yis no energy sector output. gee Hogan and Manne [9] for details.
ENERGY
ECONOMICS
July 1987
Sectoral demand
Electricity
Natural
Residential/commercial Industrial
0.415 0.238
Agriculture
0.239
0.182 0.218 -
gas
fuels,
Oil
Coal
0.229 0.665 0.547
0.2 14 0.152 -
203
EnergJl and economic growth:
T. Riaz
It gives a long-run price elasticity of 0.28. With this price elasticity, the possible price increase of 35% this is the possible difference between the most expensive and the least expensive energy plan - can lead to a fall of 9.8% in the energy demand. But if we assume that the long-run price elasticity is equal to 0.5, in that case a 35% price increase can lead to a reduction in demand of about 17%. The demand variations of this magnitude are well within the forecasting errors. In conclusion, it is evident that the balance of payment difficulties created by the high oil import bill have been compensated for by the opening up of the Middle Eastern markets to Pakistani goods and labour. With these opportunities Pakistan is perfectly capable of avoiding any energy constraints and thus maintaining her historical growth rate, which is essential for social economic progress.
6
10 11
12 13
14
References P. Balestra, The Demandfor Natural Gas in the United States, North-Holland, Amsterdam, 1967. L. G. Brookes, ‘More on the output elasticity of energy consumption’, Journal of Industrial Economics, Vol 21, No 1, 1973, pp 83-92. E. Cook, Man, Energy, Society, W. H. Freeman, San Francisco, 1976, 192. Government of Pakistan, Pakistan Statistical Year Book, various issues. Government of Pakistan, Pakistan Economic Surve) 1980-81, Ministry of Finance, Islamabad, 1980.
104
15
16
17
18 19
Government of Pakistan, Pakistan Energ), Book, Min istry of Petroleum and Natural Resources, Islamabad 197&80. Government of Pakistan, Pakistan Oil and Gas Pricing Ministry of Petroleum and Natural Resources, 1980. Z. Grilliches, ‘Distributed lags: a survey’. Econometrica Vol 35, January 1976, pp 16-35. W. Hogan and A. S. Manne. ‘Energy economy inter. action: the fable of the elephant and the rabbit’, in R S. Pindyck, ed, Advances in the Econorniu of Energ) and Resources, JAI Press, Connecticut. USA. 1978. A. Macewan, Development Alternative in Pakistan Harvard University Press, Cambridge. MA. 1971. Malih and Siddique, ‘Coal mining industry in the energy perspective of Pakistan’, Pakistan Mineral Develop. ment Corporation, Pakistan, 198 I. Pakistan Planning Commission, Proforma P.C.I. Power and Fuel Projects, 1968-l 972. M. Posner et al, Energ!, Economics: Growth, RcJsourcc and Policies, Cambridge University Press, Cambridge, UK, 1980. T. Riaz, Pakistan: The Energ), Sector: A Study irl Sectoral Planning, Ferozsons, Lahore. 1984. T. Riaz, ‘Energy policy formation for Pakistan: an optimization approach’, Energ!, Economics, Vol 3, No 3, July 1981, pp 191-197. T. Riaz, A Multi-Sector Dynamic Model,for the Pakistan Economy, discussion paper, Economics Department,
Newcastle Polytechnic, 1978. C. Starr and S. Field, ‘Economic growth. employment and energy’, Enfrgy Policy, Vol 7, No 1, March 1979, pp 2-22. Water and Power Development Authority, WAPDA Financial Statistics, various issues, Pakistan. Water and Power Development Authority, A Handbook qf Data on WAPDA, Pakistan.
ENERGY
ECONOMICS
Julv 1987