Post-regulatory reform productivity gains in Japan's electricity industry

Abstract

Following two periods of regulatory reforms in 1996 and in 2000, there has been rapid progress in restructuring Japan's electricity industry. The purpose of this study is to analyse the effect of the regulatory reforms on the technical efficiencies of electric power companies. Our findings show that first-period reforms implemented in 1996–1999 were able to decrease costs by 7.5%, while second-period reforms, during the period of 2000 to 2002, effectively cut costs by 11.8% with respect to the base costs before regulatory reforms.

Notes

¹ One alternative would be to estimate the system of equations using a random effects model. Results from the random effects model could be used to compute cost efficiency scores for each year.

² By estimating the model using data on the generation, transmission and distribution sectors of the electric utility industry, the theory underlying the model suggests that the generation, transportation and distribution of electric power is characterized by the same technology. This may be a strict assumption. Otherwise, we could estimate the model for only one sector of the industry.

³ We only consider the first-order term of time trend t to avoid the multicollinearity problem (see Evans and Heckman, 1983 for details).

⁴ There has so far been little agreement on how to define capital costs. Here, we also use the definition of Hall–Jorgenson-type capital, following previous studies of electricity such as Nemoto et al. (1993) and Ida and Kuwahara (2004). Consequently, we obtain the same results as shown in Table 3.

⁵ The reason we do not adopt the definition of Gilsdolf (1994) is that the difference between companies becomes too large due to the multiplier effect.

⁶ At this point, we have carried out the Chow test to investigate whether two regulatory reforms have caused structural changes. The result was that the first regulatory reform is statistically significant at the 10% level, while the second is statistically significant at the 5% level.

⁷ We have examined the properties of the functional form of Equation 1. The concavity conditions and monotonicity conditions with respect to input prices hold for all observations. The monotonicity conditions with respect to outputs hold for all observations but Y ₂ of Hokuriku (1990–2002) and Shikoku (1990–2000), which cover extremely small regions approximate zero.

⁸ Otherwise, it would be possible to adopt efficient two-stage least squares (2SLS) or maximum likelihood (LM) methods (cf. Hatanaka, 1974). We have confirmed here that all methods guaranteed almost the same results.