address supply-side issues directly, though there is clearly a connection between the
supply and demand sides through the price mechanism. World energy prices are
treated as exogenous in this study.
Energy demand analysis
Energy demand models can be grouped into two types: econometric models and process models.
Econometric models typically take aggregated energy consumption data and seek to establish ex post relationships with key macro-economic variables (such as incomes, output levels, interest rates) and energy prices. The results of econometric modelling of Japan's energy demand, and estimates of the main factors determining levels of energy demand in Japan, are summarised later in this chapter. Econometric techniques are limited to short-run functional forms as the estimation depends on an absence of underlying structural change. Structural breaks can be identified with these techniques, but robust results are limited to periods between such breaks. The
structural decomposition analysis described later in this chapter is more appropriate for analysing energy issues with Japan's rapidly changing economic structure.
According to Capros and Samoulidis (1988 p 43), process models use disaggregated data to combine engineering approaches to energy consumption with
a priori analysis of consumer decisions based on expectations from economic theory. Engineering approaches assess the technical energy efficiency of a process solely in terms of the volume of energy consumed and volume of output generated. Process models examine the mechanisms of change in fuel substitution and fuel conservation within industries in the medium to long term. This approach is applied to the main energy-intensive manufacturing industries in Chapter 6.
Energy-economy interactions
The third strand of energy analysis involves modelling which integrates energy demand into the broader economy. These models are of four main types: at the broadest level of aggregation is analysis of the energy coefficient, total energy consumption per unit of economic output. The three other types are more
disaggregated: techniques using production functions; structural decomposition analysis; and input-output approaches.
One of the main motivations for this study is to understand the changing pattem of Japan's aggregate energy intensity (its energy coefficient) and in particular to assess the apparent decoupling of aggregate energy demand from GDP observed through the 1970s and early 1980s. The overall trends in the energy
coefficient through time for some industrialised economies were introduced in the previous chapter. A number of researchers have examined these trends for
development paths and signs of commonality between different economies. The results of this work is summarised later in this chapter.
The production function modelhng process typically uses KLEM functions incorporating capital (K), labour (L), energy (E) and other material inputs (M). It is useful in long-run analysis to link energy demand and economic growth, and these techniques allow derivation of the aggregate elasticities (both own-price and cross-price) of the four factors. There has been some sophisticated modelling of energy in the Japanese economy using these techniques and the results are summarised later in this chapter.
Structural decomposition analysis focuses on the mechanisms of medium to long-run change in the energy-economy interaction and is a useful complement to the short-run functional forms that can be estimated with econometric techniques. The methodologies take changes in key indices, such as the energy coefficient, and decompose the changes into components, quantifying the contribution of each component. Components can be both sub-periods and the different parts of the whole economy. The extent of structural change between 1973 and 1991, especially within the manufacturing sector, is examined in Chapter 4. Results suggest that structural change has been the norm rather than the exception in this period, limiting the scope of econometric estimation to shorter sub-periods within this time frame. Structural decomposition analysis is combined with input-output techniques in Chapter 5 to yield a comprehensive disaggregation of energy
intensity shifts in the Japanese economy in the 1980s.
Input-output approaches use disaggregated output and energy consumption data to model the relations between sectors of the economy, both statically and through time. Input-output models are mechanical, in that they do not incorporate behavioural adjustments and are hmited by the enforced exogeneity of
technological coefficients (Capros and Samoulidis 1988). Change in technological coefficients can be tracked between periods where consistent data is available. Proops et al describe the approach as:
.... flexible and encompassing. It can be used for theoretical and empirical analysis on a disaggregated and aggregated level for systematic enquiry of static and dynamic micro-economic and macro-economic problems.
(Proops et al 1993 p 98). Input-output approaches to energy-economic systems analysis are
appropriate when seeking to decompose changes in aggregate statistics, like the energy coefficient, by time periods or economic sectors. Input-output techniques allow analysis of not only direct energy consumption but indirect energy
consumption as well. Using this approach, changes in the mix of inputs and the consequent level of indirect energy consumption embodied in the inputs can also be analysed.
Under the broad strands of energy demand models and energy-economy interaction, this thesis makes use of the six approaches to analysis of energy demand introduced above:
i) energy demand modelling using econometric techniques;
ii) energy demand modeUing using micro-level process techniques;
iii) energy-economy interactions modeUed through changes in the aggregate energy coefficient;
iv) production functions incorporating energy as an input; v) structural decomposition analysis; and
vi) input-output techniques incorporating energy.
Results for some of the approaches are readily available from the energy economics literature, and this study applies the other approaches to Japanese data for the first time. Results for Japan available from published research are limited to