A dynamic analysis of Japanese energy policies

Energy Policy 1994 22 (7) 595-594

A dynamic analysis of Japanese energy policies Their impact on fuel switching and conservation

F C Perkins In 1973, the Japanese economy was highly dependent on imported oil. After the 1973 oil price shock vigorous policy action was undertaken by the Japanese government in an attempt to reduce Japan's energy consumption, and in particulat, its dependence on imported oil. This study attempts to determine how flexibly the Japanese economy has responded to the dramatic changes in relative and absolute fuel prices which occurred since 1973; and to what extent government energy policies may have reinforced (or undermined) market signals and facilitated fuel switching and conservation. To analyse the price and income elasticities of energy component demand, a dynamic approach to energy demand modelling is employed. The study reveals that Japanese energy users have responded quite flexibly to the two major oil price rises in the 1970s. The probable impact of energy policies on this response is, however, somewhat problematic. Keywords: Energy economics; Energy policies; Japan

In 1973 95% of Japanese energy requirements were met by imports. Approximately 78% of these imports were oil, of which 80% came from the politically volatile Middle East. The oil price rises in the early and late 1970s sent major shock waves through Japan's highly energy intensive, export oriented manufacturing sector, and hence the economy as a whole. As a consequence, vigorous policy action was undertaken by the Japanese The author is with the Economics Program, National Centre for Development Studies, Research School of Pacific and Asian Studies, Australian National University, Canberra, Australia 0200.

0301-4215/94/07 0595-13 © 1994 Butterworth Heineman Ltd

government in an attempt to reduce Japan's energy consumption, and in particular, its dependence on imported oil. These policies included both direct attempts to influence the amounts and types of fuels used throughout the economy, and industry structure policies aimed at facilitating a shift out of energy intensive industries. The purpose of this study is to determine how flexibly the Japanese economy has responded to the dramatic changes in relative and absolute fuel prices which have occurred since 1973; and to what extent government energy and industry policies have reinforced (or undermined) market signals and facilitated fuel switching and conservation. These questions are important for a number of reasons. Given Japan's position as the world's largest energy importing country, shifts in Japanese energy demand in response to price movements and policy intervention have a significant impact on world energy markets. Hence, improved estimates of the price and income elasticities of demand for various fuels will be of interest to energy exporting countries. Furthermore, the question of the efficacy of energy policies is of interest to energy consuming countries, many of which have made their own, often less coherent, attempts at energy policy formulation in the last two decades. Throughout the paper, the term 'energy policy' is used to refer not only to policies designed to directly influence energy consumption and switching, but also to relevant industrial restructuring policies which have indirectly had these impacts. To analyse the price and income elasticities of energy component demand, a dynamic approach to energy demand modelling is employed. 1 This was developed because it was likely that the static approach traditionally used in empirical studies of this nature would fail to capture adequately the dynamic short-run responses of energy users to changes in relative prices.

595

Japanese energy policies: F C Perkins

The first section of the paper briefly examines shifts in Japanese energy policy since the Second World War. The second outlines the two stage modelling procedure employed. The third section gives the estimation results derived from the interfuel substitution and aggregate energy demand models. The results obtained in the third section are used to test the hypothesis that Japan's energy policies have influenced energy consumption levels and patterns. They are also employed to examine the extent to which Japan has responded flexibly to changes in fuel prices. The final section draws conclusions regarding the flexibility of Japanese energy demand and the effectiveness of Japan's energy policies. The study reveals that Japanese energy users have responded reasonably flexibly to the two major oil price rises in the 1970s. The probable impact of energy policies on this response is, however, somewhat problematic.

Government energy policies regarding fuel switching and conservation Japanese energy policy regimes: 1960-87 The following section briefly outlines the three major government energy policy regimes which Japan has experienced since the Second World War. 2 Period 1 : 1960-73 This period was characterized by government encouragement of the local oil refining industry, which was heavily protected and regulated, under the Petroleum Industry Law, 1962. It also promoted the rapid growth of energy intensive, export orientated, heavy industry (chemicals, iron and steel, aluminium etc). The increasing price competitiveness of imported oil compared to local coal, led the Ministry of International Trade and Industry (MITI) to reduce its support for the domestic coal industry. Nevertheless, the government enforced a system of protection for the coal industry, which enabled it to sell its coal well above the international price, and required major coal consumers to purchase a designated proportion of their coal demand from local mines. This was the period of Japan's most rapid post-war growth. GNP and energy use were rising rapidly and the GNP elasticity of energy demand exceeded unity. Period 2:1973-79 The fourfold increase in yen denominated crude oil import prices between June 1973 and June 1974 forced an urgent rethinking of the priorities of energy policy and the appropriateness of Japan's industrial structure. In 1973 approximately 78% of all Japan's energy needs were met by imported oil, and of this, 80% came from

596

the Middle East. A further 17% of energy needs were fulfilled by imported coal, gas or nuclear fuels. 3 Furthermore, because of the pattern of industrial growth in the 1960s and early 1970s, Japan's industrial structure was heavily skewed towards energy intensive basic industries. These were directly or indirectly export orientated. 4 Hence the rapid rise in oil prices threatened to severely undermine the viability of Japan's industrial sector and external trade position. Laws were enacted in the period 1973-75 to enable the government to control energy prices and supply. Legislation also required the stockpiling of oil supplies by importers, refiners and distributors. Substantial government funding was allocated to subsidize private sector research and development into energy conservation and alternative energy sources. The basic thrust of energy policy was altered in a policy document released in 1975 by the Ministerial Council on General Energy Policy 5 headed by the prime minister. This indicated that in future, top priority would be given to security of energy supplies, to ensure the continued viability of the national economy. Previously this objective had been balanced by the desire to obtain energy at the most competitive prices. Security of supply was to be achieved by reducing dependence on oil and substituting non-oil energy sources; securing a stable oil supply; pursuing energy conservation; and undertaking research and development into new energy sources. 6 In the long term, reduced dependence on imported oil, particularly Middle East oil, was believed to be the best means of improving the security of energy supplies. 7 This document set the basic direction of energy policy until at least late 1986. These policies were implemented by a range of measures including interest rate subsidies, cash grants, legal regulations, administrative direction and indicative, or rather consensus, planning via MITI's long-term energy supply and demand outlooks. The latter outlooks are formulated after exhaustive consensus negotiations between MITI and energy users and producers. They became a quite powerful policy tool in encouraging energy conservation and fuel switching in the 1970s and early 1980s. ~ A vigorous energy policy regime was therefore in place prior to the 1979 oil crisis. The main objectives of these policies were to encourage energy (particularly oil) conservation by promoting an increase in the efficiency of energy use, and a reduction in the energy intensity of economic activity and a switching from oil to alternative fuels and energy sources. In this way it was hoped that energy, and particularly oil, demand could be cut without sacrificing long-term economic growth. If successful, these policies would be expected to reduce the GNP elasticity of energy demand and,

Energy Policy 1994 Volume 22 Number 7

Japanese energy policies: F C Perkins

during periods of rising oil prices, raise the absolute value of the own-price elasticity of oil demand.

Period 3:1979-90 The doubling of crude oil import prices between March and December 1979 (from ¥18 692/kl to ¥39 158/kl), and their steady drift up to peak of ¥57 030/kl by November 1982, resulted in a renewed crisis in the Japanese economy. Unlike the situation after the 1973 oil price rises, in 1979 oil lost its competitiveness compared to fuels like coal and liquid natural gas in electricity generation and a range of industrial users. The policy response was a considerable tightening of existing energy policies. In 1979 and 1980 laws were introduced to promote energy conservation and fuel switching (the Law Concerning the Rational Use of Energy and the Alternative Energy Law). These laws included a range of prescriptive and suasive measures, as well as administrative mechanisms to achieve these objectives. In addition, a package of tax concessions and low interest loans were provided as financial incentives to obtain compliance and to ensure that the viability of firms was not jeopardized. 9 At the same time, several industry restructuring plans were implemented to shift capital and labour out of industries like aluminium smelting, which had lost their international competitiveness because of rising energy prices. These subsidies were worth ¥450 billion (US$4.3 billion at 1994 exchange rates) in 1984 alone. They were financed under the Oil Special Account, by a ¥640/kl (US$6.13/kl) excise on crude oil imports and a 4.7% sales tax on petroleum. Incentives to electricity generators worth ¥78 billion (US$747 million), were also paid in 1984 from the electricity account to encourage them to switch to non-oil generation, l0 The basic objective of the incentives for fuel switching was to increase the absolute values of the (negative) own-price elasticity of demand for oil as well as the (positive) cross-price elasticities of demand for oil and alternative fuels. As a consequence, it was hoped that a given increase in oil prices would stimulate a larger decrease in demand for oil and a greater increase in demand for alternative fuels than would otherwise have been the case. In the case of the incentives offered to implement tightened energy conservation policies, the objective of the government was to further reduce the income elasticity of demand for oil and increase the absolute value of the (negative) price elasticity of demand for aggregate energy. If these were achieved, a given increase in GNP would require a smaller amount of additional energy consumption, and a given increase in the aggregate price of energy would stimulate more switching into other factors of production (labour, capital and raw materials).

Energy Policy 1994 !/~lume 22 Number 7

Despite the dramatic fall in yen denominated oil prices, starting in 1986, the Japanese government remained committed to energy conservation and fuel switching. Both the government and the business sector anticipated medium- to long-term supply constraints and volatility in the oil market. This expectation was borne out by the continued instability in the Middle East, culminating in the Gulf War. Hence, all the policies developed after the 1979 oil crisis to promote fuel switching and energy conservation remained in place at the beginning of the 1990s. The impact, if any, of energy policies is very difficult to isolate from the normal market responses of energy users, which would have been made anyway in the absence of energy policies. Nevertheless, considerable success has been achieved in securing a switch from oil to alternative fuels. The share of oil in total energy consumption dropped from its high of 78% in 1973 to a low of 56.6% by 1986 and then stayed virtually constant, at 57%, until 1988. The reduction in oil's share has mainly been taken up by nuclear power and natural gas. The share of nuclear power in total energy consumption rose from only 0.6% in 1973 to 9% by 1988, while that of natural gas increased from 1.5% to 9.6% over the same period. 11 There has also been a significant slowdown in the growth of total energy consumption since 1973. This grew only 7% between 1973 and 1986, while real GNP rose more than 50% over the same period. However, as a result of falling energy prices after 1986, energy consumption jumped 10.8% between 1986 and 1988, the same rate as economic growth over this period. 12

Model specification In order to assess the flexibility of the response of energy users and to determine the impact of government energy policies on their behaviour, a partial equilibrium approach was employed to produce estimates of the income and own- and cross-price elasticities of demand of Japan's main energy sources. The first stage employs the cost function dual of an energy submodel of an implicit KLEM (capital, labour, energy and materials) production function, to estimate the elasticities of substitution, and cross- and own-price elasticities of demand of the five major fuel types examined. The second stage of the model provides estimates of the price and income elasticities of demand for aggregate energy in Japan. The methodology takes as its starting point an approach developed by Fuss and Berndt and Wood 13 which was based on developments in duality theory and flexible form production functions. 14 Similar approaches have been employed to analyse various aspects of Australian energy demand in recent years.15

597

Japanese energy policies." F C Perkins

The major problem with the approach taken in these studies is that they do not take adequate account of the dynamics of energy demand. In most energy demand modelling it is accepted that, because of adjustment costs, demand for different fuel types will not adjust instantaneously to changes in relative fuel prices. This is because most energy users are tied in to specific fuel using, capital equipment, which cannot be instantaneously replaced when relative fuel prices change. The inclusion of lagged dependent and independent variables among the regressors in the fuel share and aggregate energy demand equations recognizes the existence of a partial adjustment mechanism to relative price changes. In addition, the impact of Japanese energy policies is assessed by examining changes in price and income elasticities of aggregate energy demand. Hence, in the second stage of the model, the level of aggregate energy consumption is estimated as a dynamic function of the level of economic activity, lagged energy consumption, aggregate energy prices, and climatic conditions. This provides estimates of the income and price elasticities of aggregate energy demand during the various policy regimes since 1960. A detailed presentation of the methodology employed, the dynamic specification of the interfuel substitution model, and the results obtained are outlined in a paper by Perkins. t6 The f u e l share model The static version of the commonly used estimation system is derived from a unit cost function for fuel. This can be represented in an arbitrary flexible form such as a translog cost function, the dual to the translog production function developed by Christensen et al. 17 Imposing the simplifying constraint of constant returns to scale on the more general formulation gives:

InPE= lne% +

i=1 ~ °tilnPEi + 2i=11 ~ i=1 ~ °tiilnPEiln PEJ (1)

where: PE

= the aggregate price of energy index, which is the unit cost of energy to the optimizing user PEi' PEj = price of fuel i or j, with i , j = 1 ..... n

This assumption would appear reasonable for a mature economy like Japan, which is unlikely to have significant, unexploited scale economies. Using Shephard's lemma) 8 cost minimizing behaviour by energy users would imply that the demand functions for individual fuels, in terms of their shares of aggregate energy cost

598

can be represented as: 11

s i = % +,2=1%lneE,

i , j = 1..... n

(2)

where the share of total energy expenditure spent on fuel i Xi -- the cost minimizing input of fuel i Otii,O~ii = the energy own- and cross-price coefficients Si

= PEiXi/~PEiXi,

The analysis can be extended to include estimation of any fuel biased technological change by including a linear time trend, t, in each cost share equation as a proxy for technological progress. ~9 The system of equations in (2) can then be estimated subject to the usual constant returns to scale and Slutsky symmetry constraints, for the five major fuel types and energy sources: electricity (E), liquid natural gas (G), domestic and imported steaming coal (D and M) and oil products (0). 20 For simplicity, these energy sources are hereafter referred to as fuels, though of course electricity is not a primary fuel. The estimated coefficients of the fuel share equations, O~ii and aij, can be used directly to calculate the own and cross-price elasticities of demand of the different fuel types and the elasticity of substitution between the fuel types. Berndt and Wood 2j show that the own and cross-price elasticities of fuel demand can be derived from the following expressions, respectively: "flEPii = [(O~ii + S2i - Si)/S2i]Si "l~Epii =

[((~/iq- s is/)/stsi]s

(3)

i

Given the separability assumption, if the assumption of constant aggregate energy demand is relaxed, total individual own- and cross-price elasticities can be represented as a function of the fuels' own and crossprice elasticities, "qEeij, the price elasticity of demand for aggregate energy, xlTEEand the share of the fuel in total energy consumption, S.:./ 22 "qrij = "qEP!j + Si'qTeE

i j = E, G, O, M, O

(4)

Measures of partial elasticity of substitution derived by Morishima, 23 and discussed by Blackorby and Russell 24 can also be estimated from the price elasticities of demand. Blackorby and Russell 25 show that the Morishima elasticities of substitution (MES) can be estimated from the factors own and cross price elasticities of demand: MESii = Xleeii- "qEeii

i j = E, G, O, M, D

(5)

Energy Policy 1994 Volume 22 Number 7

Japanese energy policies." F C Perkins

Dynamic specification A static specification of the fuel share model seems likely to introduce estimation problems. This is because energy demand is to a large extent tied to the fuel burning and energy use characteristics of existing fixed assets in the productive and household sectors. The response of energy demand to fuel price changes is therefore limited by the rate at which energy using equipment is replaced or retrofitted. Hence, there would be substantial adjustment costs involved in responding immediately to any given relative fuel price change, and employing the cost minimizing energy mix, and its corresponding fixed capital mix, in each period. 26 These adjustment costs, rather than random errors, seem likely to be the major reason for the divergence of actual and cost minimizing fuel shares in any given period. There may also be dynamic adjustments taking place associated with the revision of consumers' expectations about future energy prices. It therefore seems plausible that correct specification of the fuel share equations will require that dynamic behaviour be specifically incorporated. Failure to do so is likely to result in autocorrelation of the fuel share equation residuals, and hence bias in the estimates of the price elasticity coefficients in the fuel share equations. This was a problem encountered by Cox. 27 As other earlier users of this technique fail to report diagnostic statistics such as Durbin-Watson, there is a possibility that they encountered similar problems. Non-dynamic equations tested by the author for the current data set exhibited marked autocorrelation of residuals (see Tables 1 and 2). The simplest method of introducing dynamic specification into energy demand equations is via a standard partial adjustment mechanism. However, the conditions this specification imposes on the coefficient estimates are too restrictive. In particular, the ratio of the short- to long-run price elasticities for all fuel types are required to equal the constant speed of adjustment parameter. 28 The dynamic specification used is one of a range proposed by Wickens and Breusch, 29 when the major focus of interest is the long-run behaviour of the model. A long-run perspective is more appropriate when examining energy market flexibility and the impact of government energy policies, because of the constraints on rapid shifts in the structure of fuel consumption discussed above. Several alternative dynamic specifications were estimated, involving manipulations of the formulations developed by Wickens and Breusch 3° for single dynamic equations. Extending the result to a system of simultaneous equations and manipulating it, the reduced form of this dynamic system 31 can be applied to the fuel share model. The estimating system is then:

Energy Policy 1994 Volume 22 Number 7

n=l

m

m

S t =t?olnetAii -~,AS, B . + Y~ AlnP, E + T + V, jO d J jO -J !

(6)

where: AS,:/= S,.j - St_j.p and A l n P i - lnP,4q, S, is the (1 × n - l ) vector of shares of each of the fuels, less the one on which prices were normalized, Xn-lP, is the (n-1 X n - l ) matrix of normalized logs of prices of the n fuels, less the one on which prices were normalized, Xm AS,_j is the (m X n - l ) matrix of first differenced lags of fuel shares, ZmAlnPt_j is the (m × n - l ) matrix of first differenced logs of fuel prices, ZA i is the (n-1 × n - l ) matrix of long-run (own- and cross-price) fuel share price multipliers, T is the trend vector which provides a proxy for technological change, and V, is a vector of error terms. 32

The aggregate energy demand model The second stage of the model involves the estimation of an aggregate energy demand equation. This is employed to analyse the changes in the income and price elasticity of demand for aggregate energy which have occurred over the period since 1960. These provide a further measure of the flexibility of the Japanese energy market, and the effectiveness of government energy policies. The government's stated objective was to reduce the income elasticity of energy demand, so that energy conservation could be achieved without jeopardizing economic growth. 33 A rise in the absolute value of the price elasticity of aggregate energy demand would indicate an increase in the capacity and willingness of the Japanese economy to substitute capital, labour and materials for energy, and would therefore produce the same result as a fall in the income elasticity of energy demand. The total demand for energy was modelled using several versions of a single dynamic equation. As in the case of the interfuel substitution model, it is unlikely that all of the adjustments of energy demand to changes in aggregate energy prices and GNP will be made within the quarter in which these changes occur. Hence alternative dynamic specifications were postulated, drawing on Wickens and Breusch, 34 and Hendry and Mizon. 35 Several alternative specifications outlined by Wickens and Breusch were tested. The estimating equation which was found to best fit the behaviour of energy consumers, is given in (7) below. As in the case of the interfuel substitution model, they are based on an assumption of a partial adjustment in each period of actual energy consumption, Et, to the hypothetical, optimum level of energy consumption, Et*.36 The formulation found most appropriate was: W

r

V

E, = O XlxiXi,- * ~. crjAE,_i - ~ . i=0

j=i+ 1

S

( X %)2tX,_ i + Oe,

i = 0 .j=i+ 1

(7)

599

Japanese energy policies: F C Perkins Table 1. Estimated coefficients of the interfuel substitution cost share function, 1960-877 Coefficient

Dynamic model 1 1960-87

1960-74

(1)

(2)

~OG

0.175 na 0.119 (5.457) -0.133 (-8.943) na na na na -0.213 na 0.03 na fla

0.053 na 0.340 (0.762) -0.094 (~3.639) na na na na -0.244 na 0.191 na na

~OM

na

na

nED

0.095 5.523

~EG

na

-0.097 -0.639 na

~EM

na

na

~DM

na

na

~DG

na

na

~GM

na

na

Constant G

-1.657 na 0.216 (3.401) -0.034 (-0.853) na

0.833 na -0.110 -0.541) 0.472 (1.156) na

Constant M

na

na

0.001 na 0,001 (4.266) na

0.005 na -0.003 -0.850) na

-0.002 (-23.282) na

-0.002 -3.286) na

~oo

~EE ~DD ~GG ~MM ~OE ~OD

Constant O Constant E Constant D

Trend O Trend E Trend G Trend D Trend M

Dynamic model2 1974-87 (3)

Non-dynamic 1979-87 (4)

Model l 1960-87

Model2 1975-87

0.175 na 0,212 (15.727) -0.002 (-1.164) 0.045 (4.7821) 0.006 (2.073) ~).170 na -0.005 na -0.017 na 0.001 na -0.001 (0.332) -0.030 (-6.307) -0.012 (-t0.401) 0.006 (3.505) -0.001 (-0.230) 0.002 (0.788) 0.057 na -0.051 (-1.306) 0.003 (0.806) -0.007 (-0.507) -0.002 (-1.164) -0.004 na 0.002 (3.936) 0.001 (13.324) -0.0001 (4.01 l) 0.0004 (12.446)

0.139 na 0.210 (14.618) -0.003 (-3.009) 0.044 (4.486) 0.007 (2.449) -0.145 na 0.024 na -0.018 na -0.0002 na -0.028 (0.442) -0.024 (-6.152) -0.013 (-11.592) 0.006 (13.784) -0.002 (-0.690) 0.0002 (0.066) 1.084 na -0.068 (-1.564) 0.001 (0.021) -0.014 (-0.888) -0.003 (-1.486) -0.005 na 0.001 ( 1.071 ) 0.002 (11.523) 0.001 (-2.828) 0.0005 (5.426)

0.072 na 0.117 (4.415) 0.083 (1.069) na na na na -0.053 na -0.019 na na na na na -0.064 (-0.762) na na na na na na na na na na 0.350 na 0.332 (5.104) 0.318 (I.679) na na na na 0.0029 na 0.0001 (0.378) na na -0.003 (-12.132) na na

0.186 na 0.222 (21.014) 0.010 (5.845) 0.055 (7.601) 0.009 (6.427) -0.159 na -0.006 na -0.015 na -0.009 na -0.009 (-4.208) -0.047 (-15.199) -0.007 (-4.271) -0.001 (-1.223) 0.006 (3.049) 0.002 (0.653) -0.029 na -0.083 (-3.494) 0.030 (7.813) 0.032 (4.321) -0.008 (-2.884) -0.003 na 0.002 (8.002) 0.001 (18.321) 0.00002 (0.776) 0.0003 (15.053)

a o = oil products; E = electricity; G = liquid natural gas; D and M = domestic and imported steaming coal. Figures in paremtheses refer to t-statis-

tics.

w h e r e AEt_ i = Et. i - Et_i_l, first differenced lagged aggre-

gate energy consumption AXt_i = Xt_ i - Xt.i. I and xl2"E")/i are the long-run energy demand multipliers. The Xit terms are natural logs of the independent variables, PEt the quarterly aggregate energy price index, T E M P t the degrees centigrade by which the average Tokyo daytime temperature for the quarter diverged from 20°C and YR,

600

quarterly real Japanese GNP. The term e t is a randomly distributed error term, with E ( e t) = O. The estimated coefficients for Equation (7) are given in Table 3. The energy submodel outlined in Equation (1) could be used to generate estimates of P E , the cost of the optimal energy mix, at each data point, for use as an instrumental variable for the price of energy, in the

Energy Policy 1994 Volume 22 Number 7

Japanese energy policies: F C Perkins

Table 2. Diagnostic tests (data from Table 1). a Rz

Durbin-Watson test Statistic Lower Upper (5%) limit limit

Fuel share equations Dynamic specification 1 (1960-87), 3rd order lags Electricity 0.927 Domestic coal 0.92 Non-dynamic specification 1 (1960-87) Electricity 0.657 0.336 Domestic coal 0.713 0.190 Dynamic specification Imported coal Electricity Liquid natural gas Domestic coal

1.374 1.374

1.542 1.542 1.542 1.542

Breusch-Pagan test (heteroscedasticity) t-statistic t-statistic Trend Predicted coefficient coefficient

t-statistic Critical Predicted value (5%) coefficient 2

23.29 37.96 53.3 1.12 2.41 39.8

-1.374 -0.351

-2.139" -1.371

-2.196" 0.721

1.99 1.99

-7.03 25.46 4.18 12.74

0.316 -1.604 -1.266 -0.581

0.277 -2.07 0.481 0.119

0.228 -1.617 0.255 ~).094

2.03 2.03 2.03 2.03

1.768 1.768

2 (1973-87), 3rd order lags 0.989 0.973 0.994 0.972

Non-dynamic specification 2 (1973-87) Imported coal 0.952 1.189 Electricity 0.937 1.949 Liquid natural gas 0.965 0.892 Domestic coal 0.702 1.463

Breusch-Godfrey test (autocorrelation) X~ ×2 Critical (lst) (2nd) value (5%)

10.18 14.44 7.22 2.83

15.38 15.38 15.38 15.38

1.776 1.776 1.776 1.776

aFails to reject the null hypothesis at the 2% significance level (critical value, 2.38).

aggregate energy demand model. An almost equivalent approach, adopted in this study, is to form a Tornqvist index of the aggregate price of energy, for use in the aggregate energy demand model. This is an exact aggregator function for a translog of the form given in Equation (1).37 Estimation results Estimation results of the energy submodel. Time series

data on quarterly fuel consumption levels and prices from January 1960 to December 1987 were provided by the Energy Data and Modelling Centre, at the Institute of Energy Economics, Tokyo38 (see Appendix 1 in Perkins 39 for the Japanese energy consumption and price data used). Data were tested for stationarity using Dickey-Fuller and augmented Dickey-Fuller tests. 4° Virtually all the series were found to be non-stationary, but the static versions of the estimating regressions were found to be cointegrating, using Engle-Granger and Hansen tests. 4~ Because two of the fuels, liquid natural gas and imported steaming coal, only became significant in the Japanese energy market in the mid-1970s, after the rapid rise of oil prices, two separate systems of fuel share equations were estimated. The first system, which included only oil products, electricity and domestic steaming coal, was estimated for the whole period 1960 (first quarter) to 1987 (fourth quarter) and for the period of the first policy regime, 1960 (first quarter) to 1973 (fourth quarter). The second system, which included all

Energy Policy 1994 Volume 22 Number 7

of the five major fuels (oil products, electricity, domestic and imported steaming coal and liquid natural gas), was estimated over the period of the second and third policy regimes 1975 (first quarter) to 1987 (fourth quarter), and the subperiod of the third policy regime, 1979 (fourth quarter) to 1987 (fourth quarter). The coefficients estimated from the system of fuel share equations (6), for both the two fuel share models, are given in Table 1 and the diagnostics are outlined in Table 2. The dynamic fuel share models appear to considerably reduce the auto-correlation problem experienced with the models employing a static specification. Diagnostic tests were run on the individual equations in the dynamic systems, and for comparative purposes, the static systems. 42 Table 2 indicates that Breusch-Godfrey tests of the residuals of the estimating equation systems which included a dynamic adjustment mechanism fail to reject the hypothesis of an absence of autocorrelation at the 95% significance level (except in one case where it was rejected only at the 90% significance level). On the other hand, the Durbin-Watson statistics of the static models' equations indicated the presence of a significant level of autocorrelation. The dynamic specification also appears to provide a better overall fit to the fuel consumption data than do the non-dynamic specifications, as indicated by the R 2 values obtained. Breusch-Pagan tests for heteroscedasticity of the dynamic models failed to reject the hypothesis of homoscedasticity at the 95% significance level (with again one exception, accepted at 90%).

601

Japanese energy policies: F C Perkins Table 3. Determinants of aggregate energy demand, Japan, 1960-877 Dependent variable, aggregate energy demand, Et 1960:1a-1987:4 b 1960:1a-1973:4 b Variable Full Final Full Final model model model model Elasticities E(-1) E(-2) E(-3) E(-4) E(-5) PE PE(- 1) PE(-2) PE(-3) PE(-4) PE(-5)

YR YR(-1 ) YR(-2) YR(-3)

YR(-4) YR(--5)

TEMP TEMP(- 1) TEMP(-2) TEMP(-3) TEMP(-4) TEMP(-5)

0.98 (8.88) -0.13 (-0.89) -0.30 (-2.21 ) 0.65 (5.24) -0.22 (-2.27) -0.84 (-22.75) 0.97 (8.95) -0.34 (-2.35) -0.05 (-0.38) 0.48 (3.8) -0.27 (-3.40) 2.84 (2.55) -1.78 (-1.08) 0.52 (0.31) 1.01 (0.61 ) -0.30 (-0.17) -2.55 (-2.08) 0.002 (0.18) -0.02 (-1.04) 0.03 (4.76) -0.002 (-0.33) -0.01 (0.40) 0.04 (2.69)

0.86 (15.44)

-0.24 (-4.50) 0.56 (5.81 ) -0.19 (-2.13) -0.86 (-26.66) 0.09 (12.42) -0.23 (-4.40)

0.38 (4.80) -0.24 (-3.29) 2.23 (5.84)

--2.50 (-6.56) 0.01 (1.76)

0.03 (6.97)

0.02 (5.45)

0.87 (3.50) 0.03 (0.08) -0.15 (-0.53) 0.06 (0.20) 0.09 (0.38) -1.62 (-9.45) 2.06 (4.22) -1.53 (-2.20) 1.09 ( 1.85) -0.71 (-1.38) 0.37 (1.15) 0.21 (0.23) 0.24 (0.23) -1.63 (-1.47) 1.61 ( 1.38) -0.92 (-0.78) -0.71 (-0.80) 0.04 (1.22) -0.03 (-0.77) 0.10 (3.80) 0.04 (1.15) 0.002 (0.05) 0.08 (1.68)

1.00 (54.93)

-1.29 (-10.93) 1.83 (6.46) -1.06 (-2.86) 0.92 (2.69) -0.59 (-2.81 )

0.81 (2.38)

-1.04 (-3.33) 0.005 (1.74) -0.08 (-3.23) 0.03 (23.12)

0.08 (3.61)

1974:1a-1987:4 b Full Final model model

1980:1a-1987:4 b Full Final model model

0.27 (0.60) -0.41 (-0.83) 0.86 (2.12) -0.18 (-0.41 ) 0.36 (1.39) -0.73 (-4.98) 0.24 (0.54) -0.46 (-0.90) 1.08 (2.84) -0. l0 (-0.21) -0.19 (-0.58) 1.28 (0.20) -0.105 (-0.11) -4.45 (-0.55) 8.23 ( 1.04) 9.69 (1.10) -16.47 (-2.84) 0.05 (1.60) 0.10 (2.23) 0.09 (2.22) 0.07 (1.69) 0.0002 (0.01) -0.03 (-1.03)

1.38 (4.36) -0.46 (-0.88) 0.23 (0.52) 0.10 (-0.42) -0.10 (-0.39) ,0.65 (-9.76) 0.60 (2.40) 0.06 (0.17) -0.18 (-0.60) 0.34 (1.14) -0.35 (-1.44) --9.85 (-0.97) -7.48 (-0.77) 19.84 (2.35) -8.86 (-0.80) 14.66 (1.24) -11.28 (-1.82) 0.13 (2.22) 0.03 (0.80) 0.07 ( 1.43) 0.07 (1.36) -0.09 (-1.88) 0.06 (2.05)

0.83 (20.32)

(3.27)

~).77 (-20.36) 0.84 (20.96)

-0.08 (-2.08) 3.90 (2.44)

-3.82 (-3.22)

0.02 (4.02) 0.02 (4.82)

1.26 (15.93) -0.26 (-3.37)

-0.15

-0.61 (-11.75) 0.57 (13.49)

--9.71 (--2.59)

8.83 (2.61)

0.14 (2.11 )

0.03 (6.90)

-0.04 (-2.06) 0.03 (6.18)

a E ( - l ) . . . (-4) = lagged aggregate energy demand; PE, PE(-1) - •. (-4) = current and lagged aggregate energy price index; TEMP, T E M P ( - I ) . . . (-4) = degrees centigrade by which the Tokyo day-time temperature diverged from 20°C (current and lagged); YR= quarterly real Japanese GNP.

The non-dynamic specification produced estimates with a similar sign and scale for most of the own- and cross-price share coefficients, but the estimates produced by the dynamic specifications generally indicated greater elasticity. This is compatible with the expectation that the coefficients derived from the dynamic specification, representing longer-term share elasticities, should generally be larger than elasticities estimated from the static model.

602

As would be expected, technical progress and the changing industrial structure were biased in favour of using oil in the period up to 1973, but against it on average over the period 1973-87. This anti-oil bias increased slightly after 1979. Technical progress was generally biased in favour of using electricity in the post-1973 period, as was the case in most developed countries. Also in line with expectations, there was a bias against using domestic coal up to 1979, and in favour of imported coal and liquid natural gas.

Energy Policy 1994 Volume 22 Number 7

Japanese energy policies: F C Perkins

The Morishima elasticities of substitution of the five fuels, were derived from the fuel share model using the relationship in Equation (5). These are given in Table 4. Once again, the elasticities are evaluated over the three policy subperiods and the average for the period 1960-87. The positive elasticities of substitution indicated that most of the fuels were substitutes. Their elasticities tended to increase in the period 1979-87, compared to earlier periods, indicating fuel users became increasingly willing to switch between fuels. The long-term own- and cross-price elasticities of demand of the various fuels (total energy demand assumed constant) are given in Table 5. These were derived from the relationship in Equation (3), using the fuel share model coefficients (Table 1) and data on market shares. Given the lack of significance of own- and cross-price fuel share coefficients in the period 1960 to 1974, they were omitted from Tables 3 and 6. It is interesting that both dynamic and non-dynamic models for this period produced estimated coefficients with low significance. This could well be due to a combination of the strong growth in the Japanese economy, the high positive income elasticity of energy demand (Table 7) and the generally low energy prices of this period. Together these factors probably made energy consumers relatively unresponsive to the small changes in relative fuel prices which occurred at this time. The estimates of the elasticities of demand of individual fuels given in Table 5 assume that the total demand for energy in relation to other factors of production remained constant over the period under consideration. However, this is unrealistic, since the steep rise in the price of all fuels, as experienced in the post-1973 period, obviously raised the price of aggregate energy relative to other factors of production such as capital, labour and raw materials. It could therefore be expected that there would have been a substitution out of energy and into other production factors. By virtue of the separability assumption, the total individual fuel price elasticities can be estimated as a function of the own-price elasticities of the various fuels, given in Table 5, the price elasticity of demand for aggregate energy, and the share of the fuel in total energy consumption. This relationship is outlined in Equation (4). The estimates of Japanese price elasticity of demand for aggregate energy, are derived from the aggregate energy demand model, the results of which are given in Table 3. The total own-price elasticity of the fuels which were derived when energy demand was assumed variable are given in Table 6. The coefficients in this table and Tables 1-5 follow the convention that the price elasticity Eij shows the impact of a change in the price of fuel j on the demand for fuel i.

Energy Policy 1994 Volume 22 Number 7

The total price elasticities for oil and electricity, and the cross-price elasticities were significantly more elastic, once the effect of substitution between energy and other factors and energy conservation was taken into account. In addition, in most cases the absolute value of the total own-price elasticities of demand of the individual fuels rose after 1979, indicating a greater capacity and willingness to shift between fuels in the 1980s than previously. All own-price elasticities of demand of the major fuels were found to be negative. The absolute value of the own-price elasticity of demand for oil rose modestly after 1979 to -0.48 from -0.45 over the whole period from 1960 to 1987. This may indicate some increase in the willingness of consumers to reduce oil consumption for a given price increase. The small rise in the absolute value of the price elasticity of oil demand post-1979 may also be at least partially attributable to energy policies implemented by the government. Certainly, the imposition of levies on imported crude and retail petroleum products in the late 1970s would have been more effective in reducing oil demand, as a result of this increase in its own-price demand elasticity. As a result of market responses, and (possibly) government intervention, oil consumption dropped from 195.4 trillion kilocalories in 1979 (first quarter) to 151.7 trillion kilocalories in 1983 (first quarter). 43 Against the trend of the other energy sources, the absolute value of the own-price elasticity of demand for electricity declined considerably over the period, from -0.72 between 1960 and 1987, to -0.404, after 1979. This trend can be explained by the tendency evident throughout Western countries for electricity to be increasingly used in applications for which there are few acceptable substitute energy forms. These uses include the operation of industrial equipment and household electrical appliances. With the move away from energy intensive to human capital intensive industries in Japan this tendency has been particularly marked. 45 However, this decline in the own price elasticity of demand of electricity ceased after 1974. This may well have been the result of the emphasis on energy conservation, rapidly rising electricity tariffs after 1974 and expectations of further rises in the future. The own-price elasticity of liquid natural gas increased significantly over the period 1979-87, from -0.04 to -0.22. The low elasticity of demand estimated for the earlier period can probably be explained by the introduction of LNG as a new fuel during this period, its early competitiveness and the rapid increase in its market share, despite real price increases. Domestic coal had a very high own-price elasticity in the 1960 to 1973 period, -3.95. This unusually high elasticity can most probably be explained by the nonmarket influence of the domestic coal price fixing and

603

Japanese energypolicies: F C Perkins Table 4. M o r i s h i m a elasticities of substitution: d y n a m i c fuel s h a r e model? Elasticities

Eoo

EEE EGG EDD

EMM EoE EOG EOD EOM EEO EEG EED EEM E~o Ec;E EGD EGM

EDO EDE ED~; EDM EMO EME EMG EMD

Average 1960-87 na na na na na O. 147 na 1.351 na 0.192 na 2.905 na na na na na

4.064 4.109 na na na na na na

Average 1975-87

Average 1979-87

na na na na na 0.248 0.221 0.195 0.657 0.180 ~3.024 0.542 -1.330 0.011 ~3.004 ~).004 0.338 1.145 1.155 1.142 2.062 0.065 0.040 0.101 0.510

na na na lla na 0.361 0.308 2.508 0.601 0.227 0.203 - 1.561 ~).929 0.192 0.198 O. 198 0.265 1.268 1.151 1.173 1.929 0.156 0.132 0.160 0.639

~O = oil products; E = electricity; G = liquid natural gas; D and M = domestic and imported steaming coal.

Table 5. L o n g - t e r m price elasticity of d e m a n d for fuels in interfuel substitution model: total e n e r g y d e m a n d constant, a Elasticities

Average 1960-87

Average 1974-87

Average 1979-87

Eoo

~3.150 -0.198 na --3.908 na ~).005 0.000 0.155 0.000 ~3.003 0.000 0.201 0.000 na na na na 1.201 2.707 0.000 0.000 na na na na

~). 164 ~).071 ~3.004 --1.144 ~).057 0.109 0.007 0.001 0.008 0.084 --0.009 0.011 ~).017 0.057 ~).095 -0.002 0.044 0.031 0.471 ~).007 0.453 0.493 -- 1.401 0.333 0.918

~).254 ~J.068 ~). 184 --l. 189 ~3.148 0.159 0.008 0.078 0.008 0.107 0.014 ~).038 ~).015 0.053 0.135 ~).017 0.012 2.254 --1.628 ~0.075 0.491 0.347 ~0.997 0.081 0.740

EEE E6G

EDD EMM EOE EOG EOD EOM EEo EF~G EEt) EEM

Ec;o EGE EGI) E~;g EDo EDE EOG EoM EMo EME EM~ EMD

aE = electricity; G = L N G ; D a n d M = domestic and imported steaming coal; O = oil products.

604

pro rata allocation system, which was in force during this period. The own-price elasticity of domestic coal dropped to more normal levels during the period after 1974, -1.15, and then rose again slightly after 1979, to -1.197. This increase could be explained by the strong growth in the market share of the close substitute, imported steaming coal, at the expense of the considerably more expensive domestic coal, during the late 1970s and early 1980s. The energy component price elasticities estimated in the current study are compared with those estimated for industrial sector energy users, using the interfuel substitution model in Ontario (Fuss), Australia (Tumovsky et al) and the USA (Halvorsen). 45 The results given in Table 7 reveal that the current study's estimates are within the range of those estimated for these countries. The main exception is the price elasticity for oil products, which were found to be lower in this study. The main reason for this is that in this study total energy demand is examined, including the demand for oil by the transport sector, while the other studies listed only consider the demand for energy in the industrial sector. Industrial sector users which use oil for heating are usually able to switch more readily between oil and the many viable substitutes for oil, for heating purposes (gas, coal and other solid fuels). We would therefore expect the price elasticity of demand for oil products in the industrial sector to be higher than that of the overall economy. The own price elasticity of LNG was also somewhat lower than that found in other studies. This could probably be explained by the fact that LNG was a very new and cost effective fuel in Japan during the years covered by this study. Hence technical factors related to the speed of conversion to LNG using plant, rather than the relatively small price movements which occurred, were the main influences on demand. We could expect that in these circumstances its price elasticity of demand would be lower than in countries where gas is a long established fuel. Aggregate energy demand model The final stage of the model involved estimation of an aggregate energy demand Equation (7). This provided estimates of the changes in the income elasticity of demand for energy and price elasticity of aggregate energy demand over the period since 1960, as a further measure of the effectiveness of government energy policies. The results for each of the major policy periods are outlined in Table 3 and diagnostics are given in Table 8. The price elasticity of demand for aggregate energy shifted substantially over the period 1960-87. The average price elasticity for the whole period 1960-87 was ~).84. Pindyck 46 also estimated the long-run own-price

Energy Policy 1994 Volume22 Number 7

Japanese energypolicies: F C Perkins Table 6. Total price elasticities of demand: total energy demand variable? Elasticity

Average 1960-87

Average 1975-87

Average 1979--87

Eoo

-0.449 -0.720 na -3.947 na -0.528 0.000 0.117 0.000 -0.302 0.000 0.162 0.000 na na na na 0.902 2.185 0.000 0.000 na na na na

-0.477 -0.476 -0.041 - 1.154 -0.062 -0.296 -0.030 -0.009 0.003 -0.229 -0.045 0.002 -0.022 -0.256 -0.500 -0.012 0.039 --0.283 0.066 -0,044 0.448 0.179 -1.806 0.297 0.908

-0.480 -0.404 -0.219 - 1.197 -0.153 -0.177 -0.027 0.070 0.003 -0.119 -0.021 -0.046 -0.021 -0.173 --0.201 -0.024 0.007 2.028 -- 1.964 -0.110 0.486 0.121 -1.332 0.046 0.732

EE~

EGG EDo E~tM

EoE Eo~; Eoo EoM EEO EEG EEo EEM

EGO EGE E~t) E~m

Et)o EDE EDG EOM EMO EME EM¢;

E~to

aO = oil products; E = electricity; G = liquid natural gas; D and M =

domestic and imported steaming coal.

Table 7. A comparison of the price elasticities of interfuel substitution models: Japan, Australia, USA and Canada? Perkins(a)

Ess Eoo EEE

Ec;G Eso EsE Esc~ Eos Eoc Eo~~ EEs EEo EE¢; Eas EGO E~;E

-1.20 -0.48 --0.40 -0.22 2.03 -1.96 -0.11 0.07 -0.18 -0.03 -0.05 -0.12 -0.02 -0.02 -0.17 -0.20

TFU(b)

Halvorsen(c)

Fuss(d)

-0.75 -0.99 -0.31 -1.45 0.46 0.35 -0.06 1.16 -0.33 0.16 0.31 -0.12 0.11 -0.59 0.68 0.37

(-2.53, -0.66) (--4.3,-1.15) (--1.10,-0.12) (-2.13, -0.43) (-0.00, 1.92) (-0.12, 3.55) (-1.55, 1,27) (0.00, 1,73) (-0.38, 3,65) (0.47, 2.14) (-0.04, 0.98) (-0.09, 0.64) (-0.02, 0.52) (-0.25, 0.95) (0.16, 1.25) (0.05, 1.48)

-1.41 -1.22 -0.52 -1.21 0.30 0.09 0.71 0.32 0.17 0.27 0.77 0.04 0.04 0.85 0.20 0.02

aS = solid fuel, E = electricity, G= natural gas, O = oil products. For Perkins, S = domestically produced coal, G = LNG. Sources: (a) Dynamic model 2, period, 1979-87; (b) M Turnovsky, M Folie and A Ulph, 'Factor substitutability in Australian manufacturing with emphasis on energy inputs', The Economic Record, Vol 58, March, 1982, pp 61-72; (c) R Halvorsen, 'Energy substitution in US manufacturing', Review of Economics and Statistics, Vol 59, 1977, pp 381-388, range of 10 US industries; (d) M A Fuss, 'Demand for energy in Canadian manufacturing', Journal of Econometrics, Vol 5, 1977, pp 89-116, evaluated at the mean level for Ontario.

Energy Poli~y 1994 Volume22 Number 7

elasticity of energy in Japan to be -0.84. By comparison, Turnovsky et a147 estimated the own-price elasticity of energy for Australia to be -0.68 over the period 1946-75. In Table 3 the apparent decline in the price elasticity of demand of energy to -0.77 in 1973-79 and its further decline to -0.62 in 1979-87 tends to indicate a reduction in the flexibility or ability of Japanese energy users to respond to changes in energy prices in this latter period. On the other hand, since this period includes the years of declining energy prices, after 1985-86, the fall in the price elasticity of energy demand may actually reflect the success of energy conservation policies. That is, with declining prices, buyers were less willing to increase consumption than they had been willing to decrease energy consumption while prices were rising. The short period since energy prices fell till the end of the data set preclude a definitive analysis of this phenomenon at this point. Movements in the income elasticity of energy demand showed wide fluctuations over the period 1960-87. The average short-run income elasticity of demand was 2.13 over the whole period, but was negative during the period following the second oil crisis. This may indicate that energy conservation and industrial restructuring policies, reinforcing the impact of flexible price signals, could have had an effect in reducing the income elasticity of energy consumption. The success of industrial restructuring may in fact partially explain the observed reduction in aggregate energy price elasticities. This is because the knowledge intensive industries, like electronics, into which Japan has diversified, may have less flexibility and incentive than the previously dominant energy intensive industries to switch out of the relatively lower levels of (mainly electric) energy which they do consume, and into other factors like capital and labour, for a given change in energy prices. Conclusions The absolute values of own- and cross-price elasticities of most of the individual fuels increased after 1979. This indicates a greater flexibility and willingness by energy users to switch between fuels, particularly between oil and alternative fuels. This was in line with incentives provided by Japanese energy policies, and as such may indicate that they had the desired effect on energy consumption patterns. In addition, the national income elasticity of demand for aggregate energy fell after 1973 and continued to decline after 1979, even becoming negative. This occurred despite the fall in oil prices in the mid-1980s. Hence, a given increase in national income required a smaller level of increased energy consumption after 1973 than before.

605

Japanese energy policies: F C Perkins Table 8. Diagnostic tests.

R2 Durbin-Watson statistic Sum (residual)2 F test ratio - E sum (residual) 2 (final/full model) Critical value - 5% significance) Breusch-Godfrey autocorrelation test statistic Critical valuex 2 - 1% significance) Breusch-Pagan heteroscedasticity test statistic Critical value X2 - 5 % significance - 10% significance

1960:1-1987:4 Full Final model model

1960:1-1973:4 Full Final model model

1974:1-1987:4 Full Final model model

1980:1-1987:4 Full Final model model

0.997 1.99 0.0300

0.999 1.87 0.0010

0.999 3.09 0.0005

0.999 2.72 0.0014

0.996 1.69 0.0327

0.917

0.999 2.48 0.0022

0.444 1.48

0.993 2.03 0.0055

0.088 2.43

0.999 1.93 0.0043

0.321 3.96

2.77

0.936

3.565

1.608

2.859

6.825"

6.131

5.656

2.738

39.87

43.05

8.57

11.52

1.601

2.088

2.56

8.26

0.010

0.075

5.562

1.299

0.0005

0.004

1.334

(/.003

3.841 6.635

3.841 6.635

3.841 6.635

3.841 6.635

3.841 6.635

3.841 6.635

3.841 6.635

3.841 6.635

a 10% significance level.

After 1979 increases in GNP were achieved at the same time as reductions in the level of energy consumption, in several years. This reduction in the energy intensity of the Japanese economy can be at least partially explained by the significant restructuring of the Japanese economy, out of energy intensive industries and into knowledge and human capital intensive industry. 48 Energy conservation measures throughout industry and in the household sector would also have contributed to this result. While rational responses to market signals may well explain these phenomena, government energy conservation and industrial restructuring policies may also have played a role. On the other hand, the price elasticity of demand for aggregate energy, while negative, actually appeared to decline in absolute terms over this period. Such reduced price elasticity of demand could be interpreted as a decreased willingness or ability to shift out of energy and into other factors of production, though as discussed above, other interpretations are possible. This occurred despite the continuing and even heightened priority put on energy conservation by Japanese energy policies over this period. The estimated movements in price, income and substitution elasticities are also of considerable interest as an indication of the extent to which adjustment in energy using behaviour is feasible in a relatively flexible market economy, which is a major importer of energy. To determine these effectively, a dynamic modelling approach appears to be useful. The author spent two years at the Institute of Energy Economics in Tokyo and wishes to thank Kenichi Matsui, Director of the Energy

606

Data Modelling Centre, for his assistance in making these data available. The author also wishes to thank Dr Trevor Breusch and Dr Will Martin for their helpful advice on earlier drafts.

ISee F C Perkins, A Dynamic" Approach to Estimating lnter-Juel Substitution: Japanese Energy Demand, Working Paper No 91/6, National Centre for Development Studies, Australian National University, Canberra, 1991. 2A more detailed discussion is included in P N Nemetz, I Veminsky and P Verninsky, 'Japan's energy strategy at the crossroads', Pacific Affairs, Vol 57, No 4, 1985, pp 533-576; F C Perkins, The Impact of

Japanese Energy Policies on Inter-fuel Substitution." A Dynamie Approach, Working Paper No 89/8, National Centre for Development Studies, Australian National University, Canberra, 1989; F C Perkins,

Japan's Energy Policies." Their Evolution and Implementation, Institute of Energy Economics, Tokyo, November, 1985; T Tomitate,

Japan's Post-War Energy Policies, Cambridge University Press, Cambridge, 1988; and M Masuda, A History of the Oil Industry in Japan, Vols 1 and 2, mimeo, Oxford Institute of Energy Studies, June 1985. 3See Ministry of International Trade and Industry, Energy in Japan." Facts and Figures, various issues, Tokyo, 1980-87. 4See Ministry of International Trade and Industry, The Twenty First Centuo' Energy Vision: Entering the Multiple Energy Era, Tokyo, November, 1986; and F C Perkins, The Restructuring of Japan's Basic Industries: Its" lmpaet on Japan-Australia Resources Trade, Institute of Energy Economics, Japan, March 1986. 5See Ministerial Council on General Energy Policy, Basic Direction of General Energy Policy, A policy document of The Ministry of International Trade and Industry, mimeo, Tokyo, 1975. 6Funding for research and development for energy conservation and oil alternative energy sources is described in New Energy Development Organisation, NEDO Annual Report, Tokyo, various years including 1985. 7See op cit, Ref 2, Tomitate. SSee op cit, Ref 2, Perkins. 9Japan Energy Conservation Centre, Energy Conservation in Japan, Tokyo, 1985, 1986, 1987. U~See Japan Electric Power Information System, Electric Power Industry in Japan, Tokyo, 1987 and Japan International Electric Research Exchange Council, Long-term Outlook for Supply and

Energy Policy 1994 Volume 22 Number 7

Japanese energy policies: F C Perkins Demand of Energy and Electric Power, Tokyo, 1984. HSee Institute of Energy Economics, Japan, Energy in Japan, various issues, Tokyo, 1980-88; Energy Balances in Japan, various years, Tokyo, 1980-87 and Energy Economics, various issues, Tokyo, 1970-87. 120p tit, Refs 4 and 11. ~3See M A Fuss, 'Demand for energy in Canadian manufacturing', Journal of Econometrics, Vol 5, 1977, pp 89-116 and E R Berndt and D O Wood, 'Technology, prices and the derived demand for energy', The Review of Economics and Statistics, Vol 57, No 3, 1975, pp 259-268. ~4See W E Diewert, 'An application of Shephard duality theorem: a generalised Leontief production function', Journal of Political Economy, Vol 79, No 3, 1971, pp 481-507; W E Diewert, 'Exact and superlative index numbers', Journal of Econometrics, No 4, 1976, pp 115-146; and L R Christensen, D W Jorgenson and L J Lau, 'Transcendental logarithmic production frontiers', The Review of Economics and Statistics, Vol 55, February 1973, pp 28-45. t-SSee for example M Turnovsky, M Folie and A Ulph, 'Factor substitutability in Australian manufacturing with emphasis on energy inputs', The Economic: Record, Vol 58, March, 1982, pp 61-72 and M M Tumovsky and W A Donnelly, 'Energy substitution, separability and technical progress in the Australian iron and steel industry', Journal of Business and Economic Statistics, Vol 2, No 1, 1984, pp 54--63; and Australian Department of Primary Industries and Energy, Forecasts of Energy Demand and Supply, Australia, 1986-87 to 1999-2000, Australian Government Publishing Service, Canberra, 1987. t60p cit, Ref 1. I70p cit, Ref 14. 18See R W Shephard, Cost and Production Functions, Princeton University Press, Princeton, 1953. ]gSee R C Duncan and H P Binswanger, 'Energy sources, substitutability and biases in Australia', Australian Economic' Papers, Vol 15, 1976, pp 289-301. 200p cit, Ref 1. 210p cit, Ref 13. 221bid. 23See M Morishima, 'A few suggestions on the theory of elasticity' (in Japanese), Keizai Hyoron [Economic Review], Vol 16, 1967, pp 144-150. 24See C Blackorby and R R Russell, 'The partial elasticity of substitution: symmetry, constancy, separability and its relationship to Hicks and Allen elasticities', Review of Economic Studies, Vol 48, 1981, pp 147-158; and 'Will the real elasticity of substitution please stand up? (A comparison of the Allen/Uzawa and Morishima elasticities)', American Economic Review, Vol 7, No 4, 1989, pp 882-888. 250p cit, Ref 24, 1989. 26See E R Bemdt, C J Morrison and G Watkins, 'Dynamic models of energy demands: an assessment and comparison', in E R Berndt and B C Field, eds, Modelling and Measuring Natural Resource Substitution, MIT University Press, Cambridge, MA, 1981; H S Houthakker and L D Taylor, Consumer Demand in the United States, Harvard University Press, Cambridge, MA, 1966; and E A Hudson and D W Jorgenson, 'US energy policy and economic growth (1975-2000)', Bell Journal of Economics and Management Science, Autumn 1974, pp 461-514.

27personal communication with A. Cox, Australian Bureau of Agricultural and Resource Economics, Canberra, June 1989. 2SSee H S Houthakker and L D Taylor, Consumer Demand in the United States, Harvard University Press, Cambridge, MA, 1966. 29See M R Wickens and T S Breusch, 'Dynamic specification, the long-run and the estimation of transformed regression models', The Economic Journal, Vol 98, 1988, pp 189-205. 3°lbid. 31The reduced form of this system of equations is given in ibid, p 195. 321bid and op cit, Ref 1. 330p cit. Ref 6. 340p cit, Ref 29. 35See D Hendry and G E Mizon, 'Serial correlation as a convenient simplification, not a nuisance: a comment on the demand for money by the Bank of England', Economic Journal, Vol 88, 1978,549-563. 36The results of tests for autocorrelation in the presence of lagged dependent variables (Breusch-Godfrey) and misspecification are given in the section on estimation. 37See op cit, Ref 13, Fuss, and W E Diewert, 'Exact and superlative index numbers', Journal of Econometrics, No 4, 1976, pp 115-146. 38The data were made available as a result of the kind assistance of Kenichi Matsui, Director of the Energy Data Modelling Centre. 39See op cit, Ref 2, Perkins. 4°See W Fuller, Introduction to Statistical Time Series, John Wiley, New York, 1976 and D A Dickey and W A Fuller, 'Distribution of the estimators for autoregressive time series with a unit root', Journal of the American Statistical Association, Vol 74, 1979, pp 427-431. 41See R F Engle and C W J Granger, 'Co-integration and error correction representation, estimation and testing', Econometrica, Vol 55, No 2, 1987, 251-276 and B E Hansen, A Powerful, Simple Test for Co-integration using Cochrane-Orcutt, University of Rochester Working Paper No 132, May 1990 and J G MacKinnon, Critical Values for Cointegration Tests, Working paper, Queen's University, Canada, January, 1990. The use of these tests is described in more detail in op tit, Ref 1. 42See J J Beggs, 'Diagnostic testing in applied econometrics', The Economic Record, Vol 64, No 185, 1988, pp 81-101. 43See op cit, Ref 2, Perkins; Table 3. 44See op cit, Ref 4, Perkins. 45See op cit, Ref 13, Fuss; Ref 15, Turnovsky et al; and R Halvorsen, 'Energy substitution in US manufacturing', Review of Economics and Statistics, Vo159, 1977, pp 381-388. 46See R S Pindyck, 'Inter-fuel substitution and the industrial demand for energy: an international comparison', Review (~["Economics and Statistics, Vol 61, 1979, pp 169-179. 470p cit, Ref 15, Turnovsky et al. 48See op cit, Ref 4, Perkins. 49See R G D Allen, Mathematical Analysis jor Economists, Macmillan, London, 1956 and H Uzawa, ~Production functions with constant elasticities of substitution', Review of Economic Studies, Vol 29, 1962, pp 291-299. 5°Op cit, Ref 15. 51Ibid. 521bid.

Appendix The results of the estimation of the coefficients from the equation system (6), for the two fuel share models are given in Table 1 and the diagnostics are outlined in Table 2. The Morishima elasticities of substitution of the five fuels were derived from the dynamic fuel share models using the relationship in Equation (5). These are given in Table 4. Unlike Allen-Uzawa elasticities,49 no Morishima elasticities

Energy Policy 1994 Volume 22 Number 7

can be derived for own-elasticities of substitution. There is no theoretical or intuitive rationale for the existence of an elasticity of substitution of a factor, in this case fuel, for itself. In most cases, the interfuel elasticity of substitution of the major fuel types were higher over the subperiod 1979-87 than over the whole period 1974-87. This apparent increase in the elasticity of interfuel substitution included that of oil

for all alternative fuels. As such it appears to indicate an increase in the flexibility of the Japanese economy to switch between fuels and in particular out of oil, over the period since 1979. Such fuel switching out of oil and into alternative fuels was heavily emphasized by Japanese government energy policies during this period, but may also have been explained by consumers' long-term expectations regarding oil prices.

617