Estimating sector-specific data of the electric power industry to analyze 1 the effect of the liberalization: the case of Japan

7 Input-output tables are employed to analyze the liberalization of electricity industries around the world. However, 8 the input-output tables do not have sectoral data of the electricity industry in many countries. Electricity industries 9 that consist of an electricity generation, transmission, and retail (and/or distribution) sector have experienced the 10 different degrees of liberalization; the generation and retail sector have experienced liberalization in many 11 countries or regions, while the transmission sector has not faced liberalization. In this study, we estimate the 12 sectoral data of the electricity industry in an input-output table by using annual reports of the electric power 13 companies and also defined the relationship among the sectors. Finally, we find that the data which are analyzed 14 by previous methods can distort policy decisions and that especially the definition of the sectors' intersection in 15 the industry is critical. 16


21
Electricity industries consist of several sectors; an electricity generation, transmission, and retail (and/or 22 distribution) sector. 23 24 Many previous studies on liberalization employ a partial equilibrium paradigm to analyze the effects of decreasing 25 electricity prices and of introducing competition (e.g. Zhang et al. (2008), Hattori and Tsutsui (2004)), however, 26 a general equilibrium might be a more suitable framework because the liberalization have influences not only on 27 customers of electricity but also on suppliers to electricity industry. A few studies have analyzed electricity 28 industries in a general equilibrium framework. Kunneke and Voogt (1997) analyzed an electric power industry in 29 the Netherlands, who employed a dynamic CGE model to estimate a welfare improvement of liberalization in the 30 Dutch electricity market. Hosoe (2006) analyzed a regulatory reform for the electric power industry in Japan, who 31 simulated the effects of the regulatory reform using the CGE model. 32 The input-output tables used in the previous studies do not have sectoral data of the electricity industry; the 33 data of all sectors are included/aggregated in the electricity industry-wide data. The previous studies defined the 34 sectoral data by simplifying each sector in the electric power industry. Hosoe (2006), Akkemik and Oguz (2011), 35 and Hwang and Lee (2015) assumed that the inputs which should have been input to the electric power industry 36 are thrown only into the electricity generation sector, and the outputs of the electricity generation sector are sent 37 to the electricity transmission sector (Figure 1). 38 Two problems are raised. 1) The inputs to the electricity industry are duplicated; all of the inputs to the whole 39 electricity industry is thrown into only the generation sector, , and ∑ =1 + is thrown into the 40 transmission sector. ∑ =1 + + is defined as the output of the transmission sector, 41 ∑ =1 + + + is the output of the retail sector. As a consequence, ∑ =1 is counted 42 three times.

Studies about the effects of the liberalization
2) The output of the retail sector will be larger than the one of the transmission sector and the 43 generation sector regardless of the real size of the outputs; even if the output of the generation sector is the larger 44 than the transmission and the retail sector, the effects on the retail sector is the largest sector. Thus, the effects of 45 the liberalization on the retail sector might be evaluated larger than the reform on the generation sector because 46 merely the retail sector obtain all the inputs of the electricity industry and the outputs from the generation and 47 transmission sector (Figure 2). It is a wrong flow as a social accounting. 48 1.2 Outline of this study 49 In the following sections, we estimate the sectoral data of the electricity industry in an input-output table by using 50 annual reports of the electric power companies, and finally find that the data which is estimated by previous 51 methods is possible to distort policy decisions 1 . The estimation is demonstrated by using the input-output table in 52 Japan, but this result would be found to be beneficial to other countries. Figure 3 depicts the results that we would 53 like to obtain. We suppose that each sector (generation, transmission, and retail) has its sector-specific inputs.  where is a unit matrix, ′ is a diagonal matrix of import coefficients, and A is a matrix of input coefficients.  92 We employed the latest input-output table in Japan, belonging to the year 2015 (we termed this table 2015-table). 93 In the original input-output table, there is one electricity sector ('electric power industry') in a row, while there are 94 two sectors in a column (thermal and other generation). The size of the table is 509 × 391. We reaggregated the 95 sectors of the electric power industry into three new sectors (generation, transmission, and retail) using the cost 96 tables of the electric power companies, thus, the size of the modified table was 511 × 392. Finally, we summarized 97 the table into a square matrix of size 191 × 191 to conduct the input-output analyses. To disaggregate the input-98 output table, we referred to "the electric utility operating expenses schedule" provided in the annual reports of the 99 electric power companies, which have to be mandatorily disclosed. 100 101 The Ministry of Internal Affairs and Communications (2019) provides the definition and the estimation method of 102 the "Electric power business" in the 2015-table. The estimation method is stated: "We allocate the electricity 103 generation costs using 'the electric utility operating expenses schedule', and also allocate other costs (transmission, 104 substation, distribution, selling, general and administrative expenses)". The electric utility operating expenses 105 schedules are published in the annual reports of the incumbent electric power companies (we termed as "EPCOs"). 106 The electric utility operating expenses schedules, which all the EPCOs are obliged to submit, are printed in the 107 annual reports 2 . The correspondence between the input-output table and electric utility operating expenses 108 schedule has not been clearly described by the Ministry of Internal Affairs and Communications (2019). 109 Furthermore, as the inputs of the input-output tables are activity-based, while those of the expenses schedules are 110 occurrence-based, there is no match between those. Thus, we reaggregated the information of the electric power 111 industry in the input-output table into three sectors based on their proportions using the electric utility operating 112 expenses schedule. In the electric utility operating expenses schedules, there are fifteen sectors, namely 113 hydropower generation, steam power generation, nuclear power generation, internal combustion power generation, 114 alternative energy generation, purchased power from other zones, purchased power from other company, 115 transmission, substation, distribution, selling costs, outage facility, loan facility, general and administrative 116 expenses, and others. In this study, we defined the correspondence between the sectors in the electric utility 117 operating expenses schedules and an input-output table, as described in Table 1.

133
We reaggregated the new input-output table, and then calculated the inverse matrix and backward / forward 134 linkages using the newly defined sectors of the reaggregated input-output table. We compared our results with 135 those calculated using previous research methods. 136 137 We calculated the input weight factors, , that were calculated by matching 509 items in the original input-output 138 table and 214 items in the electric utility operating expenses schedule.  The electricity generation sector consumes all the inputs in the "Mining" industry, and most of the inputs in 143 the "Petroleum and coal products" industry. The electricity transmission sector consumes a majority of the inputs 144 in the "Construction" and "Real estate" industries. A certain amount of inputs in the "Finance and insurance" and 145 "Information and communications" industries is sent to the electricity retail sector. 146

The inverse matrix coefficients and forward/backward linkages of the sectors
147 Figure 6 describes the inverse matrix coefficients of industries that are generated by the new electricity sectors. 148 The electricity generation sector generates the largest inverse matrix coefficients in "Electricity, gas, and heat 149 supply" industry, followed by "Business service", "Transport and postal services", and "Petroleum and coal 150 products" industries. The largest inverse matrix coefficients that are generated by the electricity transmission sector 151 is "Business services", followed by "Transport and communications", "Finance and insurance", "Commerce", 152 "Construction", and "Real estate". The electricity retail sector generates the largest inverse matrix coefficients in 153 "Business services", followed by "Transport and communications", "Finance and insurance", and "Commerce". 154 The inverse matrix coefficients of the "Electricity, gas, and heat supply" and "Petroleum and coal products" 155 industry is almost entirely generated by the electricity generation sector. 156

The inverse matrix coefficients and forward/backward linkages of the sectors 157
The forward and backward linkages are shown in Figure 7. 158 The forward linkages indicate the degree of sensitivity of the industries. Figure 7 shows that all the newly 159 added sectors are < 1, which implies that all sectors are less sensitive than the average sensitivity of all the 160 industries. 161 The backward linkages indicate how much influence each industry has on the others. 162 The backward linkages of the electricity transmission and electricity retail sectors are < 1, which implies that 163 the impacts of these industries are less than the average. The electricity generation sector has a backward linkage 164 > 1, indicating that the impact on the other industries is greater than the average. 165 Based on the examples shown in Figure 5, it can be observed that the electricity generation sector is close to 166 the manufacturing industries such as the automobile industry, while the power transmission and electricity retail 167 sectors are close to the stand-alone industries, such as agriculture or electricity.  170 We assumed that the goods and services would be directly input to the power generation, transmission, and 171 electricity retail sectors. In contrast, previous studies such as Hosoe (2006) and Hwang-Lee (2015) assumed that 172 the inputs to the electric power industry are provided only to the electricity generation sector. The outputs of the 173 electricity generation sector are then sent to the electricity transmission sector, and finally the outputs of the 174 electricity transmission sector are forwarded to the electricity retail sector. 175 We compared the inverse matrix coefficients calculated by using our method and those by using method of 176 previous studies.

Comparison of the inverse matrix coefficients
177 Figure 8 shows that the inverse matrix coefficients of the electricity generation sector on the "Mining", 178 "Petroleum and coal products", and "Electricity, gas and heat supply" industries are larger in our method, as 179 compared to the previous method. In contrast, those on the "Business services" industry are smaller using our 180 method, as compared to the previous method.
181 Figure 9 shows that the electricity transmission sector has a large inverse matrix coefficient on the electricity 182 generation sector when using the previous method, while there is very small coefficient using our method. This is 183 because we assumed that there are no inter-sectoral transactions (based on this assumption, the electricity retail 184 sector has very small coefficients on the electricity generation or electricity transmission sectors). 185 The electricity transmission sector has inverse matrix coefficients on the "Petroleum and coal products" and 186 "Electricity, gas and heat supply" industries when using the previous method, while our method does not have 187 coefficients on any of these industries. The coefficient on the "Business services" industry is larger when using 188 our method, as compared to the previous method.
189 Figure 10 shows that the electricity retail sector produces large inverse matrix coefficients on the electricity 190 generation and electricity transmission sectors when using the previous method. In contrast, there are very small 191 coefficients when using our method due to the different definition of the inter-sectoral transaction. The large 192 inverse matrix coefficients in the "Petroleum and coal products" and "Electricity, gas and heat supply" industries 193 by our method exist by the previous method is same as described above. 194 The inverse matrix coefficients on the "Business services" industry are similar to those previously seen when 195 using our method and the previous method.  Figure 11 shows that the influence of the electricity transmission and electricity retail sectors calculated using 201 the previous method are larger than that obtained using our method, while the influence of the electricity generation 202 sector calculated using the previous method is smaller. This figure also shows that all the backward linkages, 203 except those of the electricity transmission and electricity retail sectors, obtained using the previous method are 204 smaller as compared to our method.
205 Figure 12 shows that the influences that the electricity generation, electricity transmission, and electricity 206 retail sectors, and the "Mining" industry receive are larger when using the previous method, as compared to our 207 method. 208 These results indicate these industries are too sensitive when the previous method are used. We assumed that 209 the electricity sectors receive the inputs and sell the outputs directly, while the previous method assumed that all 210 the inputs are received by the electricity generation sector, and all the electricity is sold by the electricity retail 211 sector.
212 Figure 13 shows the differences of the forward and backward linkages calculated between using the previous 213 and our methods. 214 This shows that the electricity generation is more influential on the other industries, while the electricity 215 transmission and electricity retail sectors are less influential when using our method as compared to the previous 216 method, and also shows that all of the electricity sectors are less sensitive when using our method. Furthermore, 217 the forward and backward linkages of most industries other than electricity sectors are smaller when calculated 218 using the previous method, as compared to our method.  223 We separated the electric power industry into the electricity generation, electricity transmission, and electricity 224 retail sectors in the input-output table, and analyzed the inverse matrix coefficients, forward linkages, and 225 backward linkages. Prior to separating the electric power industry in the input-output tables, we could only estimate 226 the electric power industry as a whole; however, after separation, we could estimate the sectoral inverse matrix 227 coefficients and backward / forward linkages. 228 We showed that the inverse matrix coefficients and forward / backward linkages largely differed among the