| 摘要 | 1. Research Purpose
Since 1887 when the first electric light was lit up in the Kyongbok palace, Korea has experienced the rapid growth of electric generation. In 2004, Korea generated 367,581 GWh with 64 million kW power plants. In the process of developing, world oil crises led to Korea's pursuit of diversification of energy source such as nuclear power and Korea has exerted to overcome monopolistic problem of KEPCO (Korea Electric Power Corporation) by switching the power generation sector into competitive economic system.
Major energy source for generating electric power has diversified into coal, oil, LNG, water and nuclear but Korea has suffered from the change in circumstances of world's energy market. In addition, the choice of energy has been constrained by the large-scale investment in power plants and equipments under the long-term demand forecast, so that the fuel choice of electric power has been different from the economic choice, cost minimization or profit maximization of private companies.
To investigate fuel-substitution of power generation sector in Korea, this study adopted and modified the linear logit cost share model that has been utilized for fuel-substitution study in industry sector. So far, the distinct aspect of power generation sector and insufficient data of generation cost have been obstacles in econometric analysis of power generation sector in Korea. Thus, this study is supposed to be first attempt to estimate input demand and price elasticity of power generation sector.
2. Summary
The translog model has been popular in investigating fuel substitution such as Pindyck (1979a, 1979b), Shin and Ryu (1988), etc. However, the translog was also known that sometimes it generates the theoretically inconsistent result. Among other alternative models to overcome the inconsistence, the linear logit cost share model is most widely adopted recently. Considine (1989) investigated the fuel substitution by static logit model and Jones (1995) adopted a dynamic logit cost share model for the study of fuel substitution in US industry sector. Park (2003) and Park (2005) also analyzed the fuel substitution in industry and manufacture sectors of Korea.
This report employed and modified the minimum logit chi-square model by changing the characteristics of choice set in fuel for power generation. The minimum logit chi-square which is a member of linear probability models, is a weighted least square method originally developed for the case when many observations are available in each observation points (Maddala 1983). The estimate of minimum logit chi-square is consistent and asymptotically normally distributed with asymptotic variance matrix equivalent to that of maximum likelihood estimate.
The modified logit cost share model describes the log of cost share ratio by single fuel. Thus, the modified model estimates the cost share individually rather than a system of cost share. This fact makes the estimation easy since nuclear or coal undertakes the base load which produces certain amount of electricity at any input price.
For comparison, this study estimated fuel cost share functions of Korea's manufacture sector from previous logit cost share model, modified individual cost share model and modified cost share system. Modified individual cost share model or cost share system does not have the identification problem, which does not require normalization process. Price elasticities from the parameter estimates showed that all elasticities are inelastic with the magnitude in order of individual cost share, cost share system and previous logit cost share. Own price elasticities were all negative and city gas was most elastic than other fuels.
In the next chapter, was reported the estimation result of modified cost share model with fuel consumption in power generation sector. Due to Insufficient data series and unclear price data, only modified individual cost share model was applicable to analysis of fuel substitution in power generation. Each cost share function showed high of 0.90~0.98. Total product of electricity was not statistically significant in Heavy oil and LNG but most important variable in coal and nuclear which are for basis load. Especially, almost all price variables were statistically insignificant in nuclear equation but the total electricity variable had the highest t-value. For the price elasticity, all own price elasticities except nuclear and coal were negative, and inelastic except oil. Oil and LNG were shown as substitutes and oil price had bigger effect on LNG demand than LNG price on oil demand. As a result, consumption share of coal increased by 3.7%p on 10% increase of total elasticity, but nuclear decreased by 6.6%p during 1990~2000, which implies that the elasticity for basis load increased by coal power plants more than nuclear power.
3. Research Results & Policy Suggestion
Due to distinct feature of power generation sector, fuel substitution of the sector in Korea has been rarely investigated. For example; power plants are large-scale equipment business under a matter of policy rather than economy; fuels for power generation are in order of basis and peak load and the fuel price for peak load has been always higher than that of basis load; data for price and fuel amount is not serially sufficient for estimation.
Under the consideration of circumstance in power generation sector, this study attempted to adopt and modify previous fuel substitution model. From a typical linear logit cost share model, an individual cost share model with binary choice was developed, which satisfies demand theory of nonnegativity and homogeneity, and can be extended to a system by employing symmetry condition. Individual cost share model can be applied to the case when a system model is not applicable due to the insufficient data.
To improve the analysis of the fuel substitution of power generation, the next study should consider many limitations of this study. Extended and more elaborated data is the most important among them. Estimating the serial correlation will be another study utilizing individual cost share model as well as dynamic system model. |
