Deregulation, market competition, and innovation of utilities: Evidence from Japanese electric sector

Energy Policy 111 (2017) 403–413

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Deregulation, market competition, and innovation of utilities: Evidence from Japanese electric sector Nan Wanga, a b

⁎,1

MARK

, Gento Mogib,2

Systems Analysis Group, Research Institute of Innovative Technology for the Earth, Japan Department of Technology Management for Innovation, School of Engineering, the University of Tokyo, Japan

A R T I C L E I N F O

A B S T R A C T

Keywords: Deregulation Utility innovation R & D expenditure Patent

This paper explores the determinants of electric utility innovation, and examines the impact of electric sector deregulation and market competition on it using a balanced panel database of nine Japanese utilities from 1978 to 2014. Both input (R & D expenditure) and output (patent application number and patent quality) aspects of innovations are examined. The empirical results indicate that deregulation and market competition decrease the former, but increase the latter. These results are followed by a discussion on why this scenario occurs. The results also suggest that, after deregulation, utilities focus more on short-term, business-oriented R & D projects. Hence, we call for governments to support long-term, public-oriented, and environmental research in the electric sector.

1. Introduction Over the last two decades, deregulation has been implemented in Japanese electric sector in order to stimulate competition, increase efficiency, and reduce electricity prices, following the global deregulation trend. Deregulation policies were adapted to the overall economic reforms meant to activate the Japanese economy. Along with gas deregulation scheduled in 2017, electric deregulation has been changing the make-up of the industry. Electricity retailing in Japan was fully deregulated in April 2016. Consequently, gas, oil, and telecommunication companies immediately entered electricity retailing. Thus, fierce competition and restructuring are expected to activate the electric industry. 1.1. Japanese electric sector deregulation and the impacts on efficiency Japanese electricity reform was gradually implemented based on the experiences of the EU and the US. However, the crucial importance of electricity to society makes deregulation a long and difficult process fraught with political interference, opinion conflicts, and policy uncertainties. The confidence in market mechanism and competition motivates the government to open access to generation and retail markets, despite strong resistance from incumbent utilities. The generation market opened in 1995, while full deregulation of retail markets took more than 15 years, from 1999 to 2016. However, sectoral restructuring has not been able to proceed, because researchers argue that

the functional separation of generation, transmission, and distribution will increase the cost of the industry (Nemoto and Goto, 2004; Goto et al., 2013). Wang and Mogi (2017) also provided a detailed description of the process of Japanese electricity deregulation. Table 1 illustrates the main measures taken during the process. Electric sector deregulation, in theory, should produce an increased alignment of managerial incentives with firm financial performance, ultimately promoting a more efficient use of resources. Indeed, most studies on the economic consequences of deregulation in the Japanese electric sector generally show consistent efficiency gains and improvements in productivity (Goto and Sueyoshi, 2009; Goto and Tsutsui, 2008). A large body of literatureis also focused on evaluating the effectiveness of Japanese electricity reforms. Hattori and Tsutsui (2004) elaborated the relationship of deregulation and electricity price using OECD panel data. Kaino (2005) evaluated the impacts of electricity and gas reforms based on firm-level financial statistics. His analysis revealed that deregulation leads to a reduction in capital investment and labor expenditure of the electric companies, which, in turn, results in reduction of total cost and increasing efficiency. Nakano and Managi (2008) also examined the efficiency of electric companies with Luenberger indicator using the DEA approach. They showed that deregulation increases efficiency, but may also lead to investment uncertainty and blackouts. Deregulation resulted in important structural changes in the electric sector, along with technical efficiency improvements.

⁎

Corresponding author. E-mail addresses: [email protected] (N. Wang), [email protected] (G. Mogi). Address: 9-2 Kizugawadai, Kizugawa-shi, Kyoto 619-0292, Japan. 2 Address: 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656 Japan. 1

http://dx.doi.org/10.1016/j.enpol.2017.09.044 Received 11 April 2017; Received in revised form 12 September 2017; Accepted 24 September 2017 0301-4215/ © 2017 Elsevier Ltd. All rights reserved.

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

collaborative research. However, "the 0.2% commitment" policy was left invalid after April 2016 due to the financial deterioration and strong motivation to cut the cost of the utilities. We may expect R & D funds from electric utilities to CRIEPI to decline in the following years. As the behavior of research institutes is different from utilities, CRIEPI is excluded as a sample in this research. This work examines the impact of deregulation on innovation in electric utilities with Japanese data. It contributes to extant literature in two aspects. First, to our knowledge, this is the first analysis that investigates the effect of regulatory reform on electric utilities innovation. It econometrically measures both innovation input and output. Most previous studies focused on either input or output, and thus, could not provide an overall assessment of the impact of deregulation on firm innovation. Second, in Japan, this topic has been scarcely investigated. Only Hattori (2005) is known to have reported initial observations on R & D investment and patent activities within the Japanese electric sector. Due to lack of empirical analysis, the impacts of Japanese deregulation and competition on electric utilities innovation remains unclear. However, we aim to address this gap by investigating the impact of deregulation policy and competition in both generation and retail markets with respect to utility innovation behaviors. The results of this study could have important policy implications in the ongoing deregulation of the Japanese electric sector. This paper proceeds as follows: In Section 2, we build hypotheses based on economic theory and literature reviews. Section 3 outlines the research methodology and model specifications. Models are established to estimate the impacts of deregulation andmarket competitions. Section 4 explains data and variables. Section 5 reports the results of the analysis. Finally, Section 6 draws the conclusion of this study, along with a discussion on policy implications.

Table 1 Main measures of electricity sector deregulation in Japan. Year

Deregulation measures

1995 1999 2003

Generation market entry liberalization, opening access for IPPs. Partial retail market entry liberalization (capacity over 2000 kW). Establishment of the wholesale power exchange market: Japan Electric Power Exchange (JPEX). Partial retail market entry liberalization,(capacity over 500 kW). Partial retail market entry liberalization (capacity over 50 kW). Establishing the independent grid regulator: Organization for Cross-regional Coordination of Transmission Operation, Japan. Retail deregulation for residential users and low power users. Legally unbinding of the transmission and distribution sector. Removal of price regulation in residential sector electricity retailing.

2004 2005 2015 2016 2018–2020

Joskow (2006), in his keynote speech at the 2004 International Industrial Organization Conference, noted that: “Research in industrial organization and related public policy prescription has placed too much emphasis on static efficiency gain or loss and not enough emphasis on the factors influencing the rate and direction of product and process innovation which are likely to have much larger consumer welfare effects.” Most previous studies in Japan only focus on the benefit of static efficiency brought on by reforms. However, in the long run, innovation must be the source of continued efficiency and productivity improvements. Thus, the impacts of deregulation on innovation within the electric sector should not be neglected.

1.2. Innovation in the electric sector

2. Economic theory and literature review

The electric and energy industries, despite their crucial importance to economy and society, exhibit low levels of R & D intensity (GEA, 2012). The report on science and technology by Statistics Bureau of Japan also provides an overview of R & D intensity of all Japanese industries for 2014.3 In a comparison of R & D intensity (R & D expenditure divided by total sales) of each industry, we find that the electric and gas utilities (0.19%) and the oil and coal industries (0.19%) have one of the lowest concentrations of R & D activity, though slightly higher than the broadcast industry (0.10%). Researchers have raised concerns regarding the “unintended consequences” of deregulation since the beginning (Dooley, 1998). Numerous studies also reported post-deregulation R & D decline (GAO, 1996; Bell and Schneider, 1996; Bell and Seden, 1998; Margolis and Kammen, 1999). Through examining activities of companies related to the electric industry under deregulation in the US and the EU, recent scholarship, however, argues that static efficiency improvements may come at the expense of dynamic efficiency and overall R & D intensity (Sanyal and Cohen, 2009; Sterlacchini, 2012; Kim et al., 2012). Studies have concluded that deregulation reduces R & D outlays, leaving profound implications for the future reliability of electricity systems (Joskow, 2006). In Japan, the Central Research Institute of Electric Power Industry (CRIEPI) serves as the primary research institute of the electric sector. The commitment that each electric utility should fund it with 0.2% of its operating revenue helps maintain CRIEPI research despite R & D funding cuts during deregulation. This fund is included in the overall R & D expenditure of each utility. Even though more than 90% of the research fund is directly obtained from electric utilities, CRIEPI's research activities are relatively independent from the electric utilities’. For instance, joint research between CRIEPI and electric utilities is still very low, accounting for less than 10% among CRIEPI's total 3

2.1. Market structure, competition, and firm innovation What kind of market structure promotes rapid technology progress? This question can be traced back to “Theory of Economic Development” by Joseph Schumpeter in 1911. In the book “Capitalism, Socialism and Democracy" published in 1943, he further developed his theory that large firms with market power accelerate the rate of innovation. In that book, Schumpeter notes that “a market involving large firms with a considerable degree of market power is the price that society must pay for rapid technological progress.” He argues that monopolies favor innovation because they face less market uncertainty and have larger and stable cash flow to fund innovation activities. Thus, Schumpeter suggests that monopolies have a stronger incentive to innovate. According to the Solow's growth model, technology advancement is crucial to economic growth. How to balance the social gains from Schumpeter's innovation and social loss from high monopoly price is a recurrent topic of regulation economics. However, even though a large variety of empirical tests of the Schumpeter hypothesis have been implemented, it is still controversial. Adolf and Gardiner (1932) argued that the R & D in large firms might be less efficient because of agency problems; large incumbent companies may be resistant to radical innovation due to organizational inertia. Arrows (1962) claimed that competition pressure is the main driving force of innovation. A large number of studies focus on uncovering the relationship between competition and innovation (Kamien and Schwartz, 1975; Cohen and Levin, 1989; Gilbert, 2006). However, the findings are always diverse and sometimes conflicting. More recently, Aghion et al. (2005) suggested that product market competition and innovation follows an inverted-U shape based on the Schumpeter and agency models. The authors used the UK industry data (17 industry from 1973 to 1994) to support their results. Thus, it is difficult to find strong theoretical support to describe the behaviors of firms under transition from a regulated and protected market to a competitive and liberalized one.

The report can be found at http://www.stat.go.jp/data/kagaku/.

404

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

2.2. Deregulation and electric sector firm innovation

faster commercial application. Third, after deregulation, electric utilities manage R & D activities with more efficiency due to a tighter R & D budget. In this sense, an increase in the number of patents issued should be attributed to higher productivity of research efforts. These three explanations support the expectancy of a rise in patent production following deregulation. However, patent production cannot be immune to decreasing R & D input from utilities. This negative impact will decrease patent application due to less R & D effort on innovation. The 1990s saw a steep fall in utility patenting. This trend was the result of the impact of decreased R & D input overwhelming positive impact, such as increased R & D efficiency (Jamasb and Pollitt, 2015). Fortunately, Japanese electric utilities have not experienced this “collapse” until now. Jacquier-Roux and Bourgeois (2002) examined the effect of deregulation on patenting activities of large utilities and equipment suppliers in the electric and petroleum industry. They concluded that both industries increase patenting despite the evidence of less R & D investment. Furthermore, Nesta et al. (2014) found positive correlation between deregulation and patent count using information on renewable energy policies, competition, and green patents for OECD countries since the late 1970s. Their results also imply that promoting policies for renewable energy is more effective in a more competitive market. Hence, we hypothesize,

In the electric sector, innovation behaviors of firms can be altered due to regulatory change and policy-induced market structure change in both power generation and power selling markets. Before deregulation, the Japanese electric market was immune to competition. Nine regional vertically-integrated monopolies generated and supplied electricity to residential, commercial, and industrial users under the rate of return (ROR)-based price regulation. It is assumed that deregulation affects utility R & D behavior through two channels: First, market entry liberalization increases R & D risks and market uncertainties, which may reduce possible returns from R & D input; and second, by competition pressure through the market mechanism. Theoretically, the effect of competition on a firm's R & D is ambiguous. On one hand, a firm may increase its research spending, especially on projects that will increase profits or directly related to gaining market share. On the other hand, for short-term cost reduction, a firm may cut off R & D investment if it does not contribute to gaining a competitive advantage. Economic theory suggests that competition and profit incentive should somehow provide a firm efficiency gain and cost saving that can be passed to consumers through lowered prices. Thus, in a competitive environment, in order to gain a competitive advantage, R & D activities should play a more crucial role than that in a regulated environment. However, we should not neglect the public aspect of the electric utilities in the regulated environment, where in R & D, focused on grid efficiency, environmental issues, and energy efficiency, does not necessarily relate to increasing a firm's financial benefit. In an ROR regulated market, the R & D investment of the electric sector utility has been regarded as part of the ordinary cost, which is transferred to consumers. In this way, the risk of R & D is also transferred to consumers simultaneously. In addition, expectations of the spillover effect of R & D further limited incentives to pursue direct returns from R & D due to the network industry's nature. After deregulation, firms are no more obligated to act for the interest of the public welfare or for the overall industry. They may be less inclined to carry out research programs that go far beyond their immediate business needs. Thus, electricity transforms into a commodity, rather than public service, after deregulation. Consequently, short-term cost saving and increase in profitability can be achieved through limiting R & D expenditures. Hence, we hypothesize,

Hypothesis 2: The impact of deregulation on utility innovation output is positive.

2.3. Government R & D effort and utility innovation Recent literature highlights that high interdependence exists between public policy and private sector R & D behavior. Government intervention in electric sector R & D investment is required to allocate sufficient resources for innovation. There are two policy tools available for the government to encourage private R & D. The first is to provide favorable tax treatment for firms undertaking R & D. The second is to directly subsidize private R & D projects. According to Becker and Pain (2003), weak macroeconomic growth and declining government efforts are the main reasons for the drop in private sector R & D investment in the 1990s in the UK. Margolis and Kammen (1999) elaborated the government's crucial role in supporting R & D in the energy sector in their analysis of R & D input and patent data in the US energy sector. Jamasb (2007) showed that the cost reduction effect of learning-byresearch is stronger than that of learning-by-doing. Thus, public support for electric sector R & D is a strong policy instrument to promote innovation. Jamasb and Pollitt (2015) also emphasized the effectiveness of regulatory interventions to stimulate innovation. They noted that the UK's Low Carbon Networks Fund successfully created a new institutional arrangement aimed at improving energy sector social technology. Government funding can encourage firms to engage in R & D projects. Empirical results also show that government funding accelerated the completion of business R & D projects, expanded their scale and scope, and encouraged firms to conduct more challenging research (OECD, 2006). From extant literature on the US and the EU, one phenomenon is obvious: during deregulation processes, government energy sector R & D budget decreases together with the private sector. Hence, we hypothesize,

Hypothesis 1: The impact of deregulation on utility innovation input is negative. Under a deregulated environment, a tighter control of the investors and shareholders is also expected. Thus, utilities’ managers tend to reduce investment in long-term, high-risk, and public-oriented R & D projects. Instead, they focus on short-term, cost-reduction, and business-oriented results (Jamsb and Pollitt, 2008). Munari and Sobero (2002) claimed that deregulation may push management to reconsider the scope of R & D projects undertaken. They are likely to focus on projects closely linked to the needs of the core business. As the result of changed research priority, deregulation may also affect R & D outcomes. The shift from a regulated market to a competitive market significantly influences the firms’ patent behavior. Deregulation alters patent production in three ways. First, it is likely that, under a regulated environment, utility R & D pays less attention to control mechanisms against information leakage and know-how spill over, as their target is to maximize social returns to R & D activities. On the contrary, after deregulation, the company has no obligations to act for the interest of public welfare, and would focus on maximization of firm profit. Therefore, its patent applications will increase in order to give it a competitive advantage. Second, the increase in patenting may also reflect a shift in research portfolio towards more applied work. Utilities tend to prioritize research projects that offer more direct and

Hypothesis 3: Government R & D funding encourages utilities to invest in R & D projects. Hypothesis 4: Government R & D funding promotes innovation output of electric utilities.

2.4. Firm size and utility innovation Following the Schumpeter hypothesis, the size of internal funds and 405

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

financial accessibility will increase with firm size. Even though crossindustry studies suggest that this hypothesis may not be consistent (Kamien and Schwartz, 1975; Cohen and Levin, 1989), studies on the electric sector suggest that R & D spending positively correlates to firm size. However, the choice between sales and assets as a proxy for size has not been determined. Sanyal and Cohen (2009) found that R & D expenditure elasticity with respect to real operation revenue was 1.039 in their analysis of 195 firms between 1990 and 2000. Kim et al. (2012) found expenditure elasticity with respect to total asset to be within 1.118–1.382 in their analysis of 70 electricity-generating utilities across 15 OECD countries from 1990 to 2008. Salies (2010) tested the Schumpeter hypothesis with panel data of 22 European electric utilities from 1980 through 2007, and found the elasticity of size to be 1.064. Finally, Murari and Soberero (2002) examined the effect of privatization on R & D and patent activities for 35 European companies in 11 industries. They found that R & D expenditure and intensity positively correlate to size and leverage, but negatively correlate to privatization. Furthermore, they found that firm size and privatization positively correlate to firm level patent activities. Hence, we hypothesize,

tariffs. Thus, utility R & D projects have been under severe scrutiny. Finally, the macroeconomic conditions and the demand-pull effect on R & D outlays can affect innovation behaviors of electric utilities. Sanyal and Cohen (2009) used gross state product to test the effect of increasing demand on R & D expenditure, while Kim et al. (2012) used GDP growth rate to control macroeconomic conditions. In this research, the regional GDP growth rate is included as an indicator of the demandpull effect. 3. Methodology 3.1. Basic model Our methodology begins with the basic model expressed by Eq. (1), which estimates the overall effect of deregulation on firm-level R & D expenditure.

Ln (RDi, t ) = α + βDereg +

m

∑r=1

δr Fi, t +

n

∑k =1 θk Mi,t + εi,t

(1)

The dependent variable is the natural logarithm of R & D expenditure, where i denotes the firm and t denotes the year. α is the constant. Dereg is the deregulation dummy that captures the effect of deregulation. Fi, t is the vector of variables that describes the character of utilities, and Mi, t stands for the vector of macroeconomic variables. Unlike the UK, Japanese electric utilities are all private companies. Thus, ownership shift (privatization) is not considered in this model. Japanese electric utilities follow very similar organization, business model, and ownership (regional monopoly, vertically integrated, and privately owned). Thus, firm character variable vector is simplified as two variables: the firm size and the nuclear share of the total generation. The macroeconomic control vector is simplified as two variables: government spending on energy R & D that represent government emphasis on energy R & D, which may stimulate the private sector R & D. The Tobit, Poisson, negative binomial, and logit models are usually applied to estimate patent counts and citations, especially for datasets with many observations that are censored to zero. Considering the dataset, each Japanese electric utility patent number is strictly positive during the examined period. Thus, log transformation is applied to estimate the patent count. The estimation follows Eq. (2).

Hypothesis 5: larger firms invest more in R & D. Hypothesis 6: Larger firms are more active in patent application and high-quality patents.

2.5. Other factors related to firm innovation According to Mokyr (1992), financial constraint (leverage, liquidity, and cash-flow) is another possible explanatory variable to estimate firm-level innovation activities in most industries following the Schumpeter hypothesis. However, this argument may not hold in a regulated energy sector. More recently, Costa-Campi et al. (2014) analyzed the incentives and barriers related to knowledge access and market structure in the energy sector. The results suggest that financial barriers are not a determinant in explaining R & D investment in the energy sector. Hence, the financial barrier is also checked in this analysis using the net income of each utility. However, financial constraint is finally dropped in the regressions due to its insignificance. The technology used for power generation affects utility R & D and innovation. Sanyal (2007) found positive correlation between the “fossil fuel in the total generation” and the firm environmental R & D spending. Sanyal and Cohen (2009) also reported similar results—that generation mix will affect utility R & D investment. Sterlacchini (2012) indicated that, in the electricity sector, some technological trajectories are more R & D-intensive than others; they can significantly affect the research expenditures of electric utilities. Salies (2010) implied that, for a firm like EdF (Électricité de France), which produces electricity essentially from nuclear and hydro plants, these variables will have a positive effect on R & D. In Japan, nuclear power was an especially dominant research priority in the electric sector before the 2011 Fukushima crisis. Furthermore, nuclear R & D projects are usually supported by utility consortium as a joint research with support from public agencies and government funds. Thus, nuclear power generation share is adopted as an indicator of the impact of generation mix on utility R & D expenditure. A utility with a larger share of nuclear power generation will invest more in joint nuclear R & D projects. In this research, we also include the share of nuclear power in power generation as a control variable. The Fukushima crisis greatly altered R & D trends in the electric sector. The impact of Fukushima is thus included in our analysis, and R & D expenditure has been automatically transferred to consumers based on ROR price regulation. Deregulation is supposed to cancel this mechanism; however, utility R & D expenditure is still included in the cost calculation. Nevertheless, after the Fukushima crisis, there has been growing concern regarding the continuous rise in electricity

Lnpatenti, t / Acitation i, t = α + βDreg + +

m

∑r =1

δr Fi, t

n

∑k =1 θk Mi,t + εi,t

(2)

As Margolis and Kammen (1999) indicated, there is a positive correlation between R & D expenditure and patent output. We may assume that R & D efforts will positively affect patent applications. We include the impact from innovation input to innovation output in the firm character variables. Thus, firm character variable vector is simplified as two variables in the analysis: the firm size and the R & D intensity. In fact, it is possible to exclude R & D input in the explanatory variables as Munari and Sobero (2002) indicated. However, this may greatly decrease the estimation significance. The macroeconomic vector is simplified as government spending on energy R & D. 3.2. Extended model We assume that deregulation alters the behavior of firm R & D by way of two channels: the effect of entry liberalization and the effect of the competition pressure through market mechanism. Both effects are supposed to exist in the generation, distribution, and retailing sectors. However, the “standard textbook reform” has not been fully implemented in Japan. Thus, the electricity distribution remains regulated. In the extended model, the firm and the macroeconomic variables are the same with those in the basic model while Dereg is replaced with the description variables in order to examine the separate impacts of entry liberalization of power generation, power retailing, and impacts 406

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

4. Data and variables 4.1. Dependent variables The Japanese electricity market is controlled by 10 companies: Chugoku Electric Power Company, Chubu Electric Power, Hokuriku Electric Power Company, Hokkaido Electric Power Company, Kyushu Electric Power, Kansai Electric Power Company, Okinawa Electric Power Company, Tokyo Electric Power Company, Tohoku Electric Power Company, and Shikoku Electric Power Company. These electricity utilities are so-called general electric utilities, which are regional monopolies and have been vertically integrated since the 1950s. Our database excludes Okinawa Electric Power Company due to its short history, small size, and relatively independent status in the electricity market compared to the other companies in mainland Japan. The utilities included in our sample account for more than 75% of total power generation, 99% of transmission and distribution, and 95% retail. The database thus consists of a balanced panel of nine electricity utilities from 1978 to 2014. The first dependent variable is R & D expenditure, expressed as the natural logarithm of the R & D input. This panel is believed to be relevant to the analysis because little R & D seems to occur in the smaller firms or new players in the same industry (Sterlacchini, 2012). Most importantly, R & D behaviors of incumbent utilities can represent the overall R & D behavior of the private sector in the electric sector. R & D expenditures are compiled with the annual report of each utility; they are deflated at the 2005 constant price using CPI provided by the Statistics Bureau of Japan. In this study, utility innovation output is measured by two dependent variables: patent quantity (the number of patent applications of a utility) and patent quality (patent average citations of a utility). The application year, rather than the granted year of patent, is adopted in our model, since the former better captures the actual time of innovation (Cornaggiaa et al., 2015). Furthermore, use of patent average citation also compensates for the bias that patent citation may include. In the years where in utilities acquire more (fewer) patents, total citations of the utility may increase (decrease) simply because there are more patents granted or because of the increased (decreased) existence of citing patents, which, however, does not reflect a change in patent quality. Thus, the total patent may not be a good indicator of patent quality. Moreover the average patent citation will change only when the rate of change in citation is larger/ smaller than the rate of change in the number of patents. Thus, we use both as indicators to measure the effect of deregulation on patent quality. In this analysis, the patent application count of all nine utilities, from e1978 to 2011,is included. The data for the patent count is compiled from IIP-DB (The Institute of Intellectual Property Patent Database) by Goto and Motohashi (2007). The database includes patent application data, patent registration data, application data, rights holder data, citation information, and inventor data. IIP-DB covers 12,706,640 patent applications from 1964 to 2012. The patent application count from 1978 to 2011 is extracted, as there is a 1.5 years lag between the patent application and its publication. According to Popp (2006), patent-granted count is mainly used for the diffusion of technologies. According to Fujii and Managi (2016), the advantage of using patent application to evaluate the firm-level R & D is that it reflects investors’ R & D strategy more accurately than patent-granted count. As this research focuses more on strategic R & D change during deregulation, rather than technology diffusion, patent application data count is applied for analysis over patent-granted count. Similarly, the average patent citations are also compiled from IIPDB using the number of backward citation received. However, it is obvious that old patents have an increased chance to be cited than newer patents. To avoid this bias, only citations within five years after being granted are considered in our analysis. Thus, time coverage of the

Fig. 1. R & D expenditure of J-Power and Hokkaido Electric.

of competition from both markets.

3.3. Does deregulation really change utility R & D behavior? To answer whether deregulation is responsible for the decline in R & D spending, the difference in the difference model is usually a good choice. However, it is usually extremely difficult to find a good control group in most regulated industries due to the small number of samples. While the gas industry could be a good control group, it was partly deregulated even earlier than the electric sector, and thus, cannot serve as the comparative group. Even though an econometric approach is unavailable, it is still possible to observe the impact of retail deregulation on R & D expenditure by comparing general utilities (vertically integrated) with wholesale utilities (utilities only evolve in electricity generation), as retail deregulation should not affect the latter. Fig. 1 compares the R & D expenditure of Hokkaido Electric Power Company and Electric Power Development Company(J-Power, the largest wholesale utility in Japan). The former cut off its R & D investment after retail deregulation, while the latter increased its R & D investments. We may thus infer that the gap after the retail deregulation in 2000 is partially due to retail deregulation policy and the competitive pressure from the retailing market. Fig. 2 implies the share of Japanese electric utilities among total patent applications in Japan. The rise in patent application of electric utilities after the electric deregulation is evident, especially after retail deregulation.

Fig. 2. Share of electric utilities among total patent application.

407

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

Table 2 Variable description and statistics. Variable Dependent Variable Ln(RD) Ln(patent) Acitation Deregulation and Competition Variables Deregulation dummy Generation entry liberalization Retail entry liberalization Generation competition Retail competition Firm Variables Nuclear share Ln(RD intensity) Ln(assets) Macroeconomic Variables Ln(govRD) Fukushima dummy GDP growth

Description

Data source

Min/Max

Mean(S.D.)

Natural logarithm of real R & D expenditure (constant price 2005) Nature logarithm of patent application count Average patent citation count (within 5 years)

Annual report IIP-DB IIP-DB

6.885/11.093 0/7.057 0/3.333

8.891 (0.983) 3.766 (1.346) 0.831 (0.251)

Overall indicator of deregulationa Dummy of generation sector entry liberalizationb Sales of deregulated retail market of each utility/ Total sales of electricity of each utility Share of non-utility electricity generated in each region*100 Regional market share of PPS*100

METI METI Annual Report

0 /1 0/2 0/0 0.622

0.541 (0.499) 0.868 (0.878) 0.206 (0.270)

FEPC,METI METI

24.601/32.399 0/9.8

27.456 (2.322) 0.555 (1.312)

Share of nuclear generation capacity in total power generation capacity Natural logarithm of firm R & D intensity, R & D intensity = R & D expenditure/revenue Natural logarithm of firm assets

FEPC Annual report

0 /0 0.320 − 6.113/ − 4.056

0.170 (0.094) − 5.016 (0.444)

Annual report

13.044/16.533

14.848 (0.794)

IEA

12.401/13.145

12.882 (0.139)

Cabinet Office

0/1 − 3.7/6.4

0.108 (0.311) 2.119 (2.372)

Nature logarithm of real government energy R & D expenditure (constant price 2005) Dummy indicator for Fukushima crisisc Percentage of regional GDP growth rate*100

a

Before 1995, deregulation dummy = 0, after1995, deregulation dummy = 1. Before 1995, Generation deregulation = 0, after opening access to generation sector in 1995, Generation deregulation = 1, after setting up wholesale market in 2003, Generation deregulation = 2. c Before 2011, Fukushima dummy = 0, after 2011, Fukushima dummy = 1. b

competition is often the greatest (Salies, 2010). 4) Retailing market competition indicator (the market share of PPSs in the regional retailing market)—the greater share of PPSs stands for the greater competitive threats to the incumbent utilities.

average citations is from 1978 to 2006. 4.2. Explanatory variables Table 2 summarizes the description of dependent variables and explanatory variables. Japanese electric utilities share very similar organizational structure. Thus, the heterogeneity of firms is mainly controlled by their firm size and the share of nuclear power in the total generation of the utility.

There are other indicators to measure market competition, such as the number of firms, market share, concentration ratio, HerfindahlHirschman index, Hannah and Kay index, and Lerner index. However, due to data limitation, market share is used to proxy market competition in this work. All deregulation variables are compiled from the Ministry of Economy, Trade and Industry (METI),the database of Federation of Electric Power Companies of Japan (FEPC), and the annual reports of each utility.

4.2.1. Deregulation and competition In order to capture the overall effect of deregulation, this research follows the approach of Sanyal and Cohen (2009) and Salies (2010). The overall impact is captured through a deregulation dummy that equals to 1 since the start of deregulation in 1995. To see the separate impact of each process of the deregulation, Kim et al. (2012) provide an approach that separates the process of the deregulation as “Entry liberalization, vertical integration, and privatization". Each process is represented by dummy indicator. However, their results are polluted by multicollinearity, thus they report separated estimation of each measure. This research also follows the same idea to separately estimate the deregulation impact on R & D input through five indicators. In Japan, the process of privatization, vertical unbundling, and horizontal splitting cannot be observed. The indicators of entry liberalization and market competition are built with:

4.2.2. Government incentive, firm size, macroeconomic condition, and technological incentive To estimate R & D input, firm revenue is used as a proxy for firm size, which has been widely applied in extant literature (Sanyal and Cohen, 2009; Munari and Sobrero, 2002). Firm revenue is also in the natural logarithm form. As mentioned in Section 2, generation technology may also affect R & D investment. The share of nuclear power in power generation is applied as a proxy of generation technology, since nuclear technology is the first R & D investment priority during the examined period. The government energy R & D spending is adopted as a proxy for government efforts that stimulate private R & D. Regional GDP growth rate is also adopted as a controlling variable, which aims at controlling the macroeconomic conditions during the examined period. For innovation output, as discussed in Section 3, firm assets and R & D intensity are chosen as explanatory variables to avoid endogeneity problems. Firm assets indicate the impact of firm size, and R & D intensity stands for the R & D investment effort of each utility. Similarto the case of R & D input, the government energy R & D spending is adopted as a proxy for government incentive to stimulate private R & D output. R & D intensity and firm asset are calculated by/ compiled with the annual report of each utility. Nuclear share and generation competition indicator are calculated with the database of the FEPC. Government energy R & D expenditure is obtained from the International Energy

1) Power generation market entry liberalization indicator—0 when generation market is regulated, 1 when generation market is deregulated after 1995, and 2 when a wholesale market is established after 2003. 2) Retailing market entry liberalization indicator—since retail entry liberalization is implemented gradually, the indicator is defined asthe share of deregulated market in the regional retailing market. 3) Generation market competition indicator—the generation market competition indicator is defined as the share of non-utility generation (electricity self-generation and electricity generated by IPPs) in theregional power generation market. Generation competition pressure will alter the behavior of R & D, since a large portion of R & D is usuallyconnected to electricity generation where 408

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

Table 3 Correlation of explanatory variables for the estimation of R & D expenditure. Variables

Generation entry liberalization

Generation entry liberalization Retail entry liberalization Generation competition Retail competition Nuclear share Ln(assets) Ln(govRD) Fukushima dummy GDP growth

1.000

Retail entry liberalization

Generation competition

0.901

1.000

0.783

0.842

1.000

0.523 0.411 0.228 0.091 0.449 − 0.545

0.588 0.351 0.175 − 0.090 0.533 − 0.457

0.542 0.213 0.072 − 0.301 0.616 − 0.328

Retail competition

Nuclear share

1.000 0.523 0.438 − 0.155 0.497 − 0.268

1.000 0.429 0.153 0.163 − 0.312

Agency (IEA) database. Regional GDP growth rate is from the Cabinet Office, Government of Japan.

Ln(assets)

Ln(govRD)

1.000 0.173 0.093 − 0.204

1.000 − 0.305 − 0.282

Fukushima Dummy

GDP growth

1.000 − 0.227

1.000

Table 5 Results of Hausman test.

4.2.3. Endogeneity and model selection It is possible that firm size may introduce an endogeneity problem. However, firm size of utility is more likely to respond to domestic electricity demand due to the demand matching obligation of the utilities, even though this obligation is cancelled by deregulation process. As the electric sector is significantly different from other technologyintensive industries like IT and communication, it is reasonable to assume that the size factor is exogenous. As explained in Section 2, a deregulation dummy is applied to capture the overall deregulation effect in the basic model. In the extended model, deregulation is estimated with policy variables, such as generation deregulation policy, retail deregulation policy, and market competition variables, including generation competition variable and retail competitionvariable. Variables on deregulation policy and competition are highly correlated as Table 3 indicates. It is difficult to estimate the independent effect of each variable simultaneously. For R & D output, a similar approach is adopted. Other explanatory variables include Ln(asset) to capture the firm size effect, Ln(RD intensity) to capture the effect of R & D input, and Ln(govRD) to capture government R & D incentive. Since there will be a time lag between the R & D input and output, further regression is examined with a lag effect. Extent literature suggests that there is a time lag between R & D input and patent applications, but only a one-year lag is consistently significant in different models (Wang and Hagedoom, 2014; Dang and Motohashi, 2015). Furthermore, using more lagged variables will decrease the sample significantly. For these reasons, in the regression, patent count is estimated with one-year lag in R & D input and incases without lag in R & D input. Table 4 indicates the correlation within explanatory variables. We implement Hausman test to select the proper model. The statistical results of Hausman test is illustrated in Table 5.

Chi2(2) Prob > chi2 Model

R & D expenditure

Patent application

Patent average citation

7.07 0.019 Fix effect

2.67 0.445 Random effect

2.27 0.321 Random effect

5. Estimation results 5.1. The effect of deregulation on R & D input In order to capture the overall impact of deregulation, we examine the coefficients of deregulation dummy. The coefficient of deregulation dummy is negative and significant in Model 1 (− 0.505), Model 2 (− 0.495), and Model 3 (− 0.477). These coefficients are consistent with Hypothesis 1, which implies that the overall impact of deregulation on R & D expenditure is negative. The effects of generation entry liberalization and retailing entry liberalization are negative and significant as illustrated by Model 4 (− 0.317) and Model 5 (− 0.873). These results are in line with the previous analysis that reveals negative impact of deregulation. Furthermore, the coefficients of market competition (generation market competition and retailing market competition) are also negative and significant based on the results of Model 6 (− 0.108) and Model 7 (− 0.138). Comparing the absolute value of coefficients, we also find that the impacts of retail market entry liberalization/ competition on R & D investment are stronger than those of generation market competition. The elasticity of R & D spending with respect to the size of the electric utilities and government energy R & D is positive and significant. Thus Hypotheses 3 and 5 are confirmed by the regression results. In line with the Schumpeterhypothesis, size advantage in R & D proved to be an important factor that affects the resources allocated to technology and innovation. Government R & D efforts also tend to encourage private R & D investment based on regression results. This

Table 4 The correlation of explanatory variables for the estimation of patent data. Variables

Generation entry liberalization

Generation entry liberalization Retail entry liberalization Generation competition Retail competition Ln(RD intensity) Ln(assets) Ln(govRD)

1.000 0.884 0.748 0.538 − 0.191 0.222 0.249

Retail entry liberalization

Generation competition

1.000 0.800 0.612 − 0.288 0.162 0.065

1.000 0.485 − 0.447 0.003 − 0.156

409

Retail competition

1.000 − 0.135 0.405 − 0.013

Ln(RD intensity)

1.000 0.368 0.208

Ln(assets)

Ln(govRD)

1.000 0.225

1.000

Energy Policy 111 (2017) 403–413

− 0.138*** (0.014) 0.565* (0.315) 1.157*** (0.083) 0.213* (0.132) − 0.417*** (0.058) 0.039*** (0.007) − 10.896*** (1.494) 0.853 333

expresses the positive role of the government's technological innovation policy and investment. The share of nuclear power in power generation also positively affects the R & D expenditure. A strong impact from a technology mix of energy generation can be observed. The results also indicate the demand-pull effect is significantly positive; However, the coefficient value is relatively small. As we discussed before, the deregulation variables have high correlation. Model 8 estimates all the deregulation variables simultaneously. The overall regression results are reported in Table 6. 5.2. The effect of deregulation on R & D output 5.2.1. Patent quantity The R & D input level positively correlated with the patent application count, which can be proved by the coefficients of Ln(RD intensity). The coefficient of deregulation dummy is positive and significant in Model1 (0.238), Model2 (0.191), Model3 (0.247), Model 9 (0.204), Model 10 (0.206), and Model11 (0.206). These coefficients are consistent with Hypothesis 2, which implies that deregulation positively correlates with patent application of electric utilities. The effect of generation entry liberalization and retailing entry liberalization are positive and significant, as illustrated by Model 4 (0.226), Model 5 (1.010), Model 12 (0.183), and Model 13 (0.737). These results are also in line with the previous analysis that the impact of deregulation is positive. The coefficients of market competition are positive and significant, as illustrated by Model 6 (0.073), Model 7 (0.273), Model 14 (0.061), and Model 15 (0.232). Similar to innovation input, it is reasonable to infer that the impacts of retail market entry liberalization/competition on the patent application are much stronger than those of generation market competition. Hypotheses 4 and 6 are also supported by the regression results. The coefficients of patent application with respect to the assets of the electric utilities and government energy R & D are positive and significant. Positive and elastic government R & D efforts coefficient further suggests that government support encourages private sector patent application. However, in the case of lagged R & D, the impacts of government funding become insignificant. Models 8 and 16 capture the impacts all the deregulation variables simultaneously. Table 7 shows the overall regression results on patent count without considering the lag effect of R & D input. Table 8 shows the regression results on the patent application count with one-year lag of R & D input level. 5.2.2. Patent quality The coefficient of deregulation dummy is positive and significant. The effect of the process of deregulation and market competition are in line with the results in patent quantity. These coefficients are consistent with Hypothesis 2, which implies that deregulation positively correlates with electric utility patent quality. Hypothesis 6 is also supported by regression results. The coefficient of patent quality with respect to the assets of the electric utilities is positive and significant. The R & D input level positively correlates with patent quality. The variable of government R & D fund is dropped due to its insignificance. It seems that government R & D efforts are less relevant with utility patent quality. Stronger impacts from retail sector are found during comparison with the generation sector. Table 9 shows the regression results on average patent citations. To summarize, deregulation positively affects utility patent quality. 6. Conclusions and policy implication This paper explores the determinants that drive utility innovation. It investigates how deregulation affects R & D input and output of incumbent Japanese electric utilities. We seek to contribute to the understanding of the impact of deregulation in terms of private utility innovation behavior. In this study, the overall impact of deregulation, as well as the

* p = 0.1. ** p = 0.05. *** p = 0.01.

0.868*** (0.300) 1.353*** (0.082) 0.352*** (0.123) − 0.571*** (0.050) 0.027*** (0.007) − 15.617*** (1.453) 0.899 333 2.263*** (0.199) 1.080*** (0.022) 0.506*** (0.133) − 0.548*** (0.060) 0.028*** (0.008) − 13.777*** (1.711) 0.912 333

1.132*** (0.284) 1.232*** (0.061) 0.435*** (0.117) − 0.555*** (0.050) 0.026*** (0.007) − 14.943*** (1.436) 0.906 333

1.242*** (0.280) 1.297*** (0.073) 0.183* (0.114) − 0.454*** (0.047) 0.025*** (0.006) − 12.678*** (1.295) 0.914 333

1.048*** (0.294) 1.141*** (0.074) 0.075 (0.122) − 0.411*** (0.051) 0.031*** (0.007) − 9.309*** (1.377) 0.910 333

0.869*** (0.271) 0.891*** (0.069) 0.027 (0.115) − 0.264*** (0.051) 0.026*** (0.007) − 8.887 (1.417) 0.910 333

− 0.121** (0.422) − 0.0.147 (0.115) − 0.063*** (0.013) − 0.071*** (0.015) 0.658* (0.282) 1.129*** (0.080) 0.018 (0.111) − 0.259*** (0.053) 0.022*** (0.006) − 6.235*** (1.711) 0.914 333 − 0.317*** (0.021) − 0.495*** (0.041) − 0.505*** (0.043)

Deregulation dummy Generation entry liberalization Retail entry liberalization Generation competition Retail competition Nuclear share Ln(assets) Ln(govRD) Fukushima dummy GDP growth Constant R-square Observations

− 0.477*** (0.040)

Model4 FE Model2 FE Model1 OLS

Basic model

Explanatory variable

Table 6 Regression results of innovation input.

Dependent variable : ln(RD)

Model3 RE

Extended model

Model5 FE

− 0.873*** (0.066)

Model6 FE

− 0.108*** (0.007)

Model8 FE Model7 FE

N. Wang, G. Mogi

410

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

Table 7 Regression results of innovation output (patent quantity) (without R & D lag). Dependent variable : ln(patent) Basic model Explanatory variable Deregulation dummy Generation entry liberalization Retail entry liberalization Generation competition Retail competition Ln(RD intensity) Ln(assets) Ln(govRD) Constant R-square Observations

Model1 OLS 0.238

**

(0.104)

Extended model Model2 FE *

0.191 (0.110)

Model3 RE **

0.247

Model4 RE

Model5 RE

Model6 RE

Model7 RE

(0.100) − 0.201 (0.160)

0.226** (0.065) 1.010*** (0.225) 0.073*** (0.028) 0.585*** (0.114) 1.178*** (0.065) 1.511*** (0.400) − 29.373*** (5.388) 0.630 306

0.668*** (0.136) 1.969*** (0.611) 1.016** (0.509) − 33.571*** (6.802) 0.614 306

Model8 RE

0.658*** (0.130) 1.208*** (0.148) 1.447** (0.382) − 28.583*** (5.362) 0.629 306

0.790*** (0.140) 1.364** (0.623) 1.251** (0.506) − 27.635*** (6.621) 0.637 306

0.9143*** (0.144) 0.827 (0.637) 1.765*** (0.520) − 26.134*** (6.246) 0.627 306

0.838*** (0.157) 2.220** (0.552) 1.162*** (0.509) − 39.982*** (5.578) 0.609 306

0.273*** (0.056) 0.873*** (0.138) 1.372** (0.574) 1.574*** (0.505) − 31.412*** (5.719) 0.645 306

1.164*** (0.441) − 0.063 (0.052) 0.156** (0.065) 0.533*** (0.200) 0.913** (0.366) 1.563*** (0.440) − 25.652*** (6.91) 0.655 306

* p = 0.1. ** p = 0.05. *** p = 0.01.

3) Utilities are less motivated to own patents even if they contribute to collaborative research with other companies under regulation, as they are the only beneficiaries. However, utilities are motivated to apply for patents as free-riders may begin to exist in a deregulated market; 4) Utilities have changed their research priorities to short-term, business- or consumer-oriented research projects that can increase patenting. However, in the long run, the declining R & D investment may eventually lead to a reduction in patenting and innovation, which has been observed in the US (Sanyal and Cohen, 2009) and the UK (Jamasb and Pollitt, 2011). Furthermore, it is worth mentioning that the impact of retail market competition is stronger than that of generation sector competition.

separate effects of generation deregulation, retail deregulation, generation competition, and retail competition, are examined based on firm-level empirical evidence. The results are two-fold. On one hand, in line with related extant literature (Costa- Campi et al., 2014; Dooley, 1998; Kim et al., 2012; Jamasb and Pollitt, 2015; Salies, 2010; Sterlacchini, 2012), we find that electric deregulation has a negative impact on utility R & D input. On the other hand, we found evidence that deregulation increases firm patent applications and patent average citations of incumbent utilities. Thus, we inferred that deregulation could boost utility R & D productivity as Fig. 3 illustrates, acting as a short-term benefit of deregulation. Four reasons could explain the increase in innovation productivity: 1) The existence of a certain degree of inefficiency in utilities with respect to R & D before deregulation; 2) The commercialization of electric utilities—growing competition encourages utilities to increase patenting in order to gain competitive advantage;

We infer that government energy R & D funding can promote private sector innovation input. However, the impact is less certain on the innovation output. Popp et al. (2011) found that technology push (R & D expenditure), rather than market pull (installed capacity of the

Table 8 Regression results of innovation output (patent quantity) (one year R & D lag). Dependent variable : ln(patent) Basic model Explanatory variable Deregulation dummy Generation entry liberalization Retail entry liberalization Generation competition Retail competition Ln(RD intensity) (lag1) Ln(assets) Ln(govRD) Constant R-square Observations

Extended model

Model9 OLS

Model10 FE

Model11 RE

**

*

**

0.204

(0.100)

0.206 (0.109)

0.206

Model12 RE

Model13 RE

Model14 RE

Model15 RE

Model16 RE

0.232*** (0.050) 0.745*** (0.128) 1.015*** (0.162) 0.964** (0.424) − 19.161*** (5.673) 0.627 297

− 0.025 (0.133) 0.820* (0.452) − 0.059 (0.047) 0.166*** (0.064) 0.732*** (0.145) 0.974*** (0.176) 0.958** (0.474) − 17.081*** (6.54) 0.632 297

(0.952) 0.183*** (0.054) 0.737*** (0.182) 0.061** (0.025)

0.542*** (0.111) 1.166*** (0.637) 0.494 (0.481) − 16.222*** (6.363) 0.612 297

0.597*** (0.130) 1.170* (0.655) 0.458 (0.516) − 15.549* (8.633) 0.612 297

0.585*** (0.124) 1.159*** (0.146) 0.469 (0.460) − 15.610*** (6.227) 0.612 297

0.658*** (0.126) 1.121*** (0.157) 0.507 (0.437) − 15.225** (5.958) 0.619 297

* p = 0.1. ** p = 0.05. *** p = 0.01.

411

0.733*** (0.129) 1.095*** (0.148) 0.793* (0.425) − 18.168*** (5.724) 0.623 297

0.712*** (0.137) 1.161*** (0.146) 0.804* (0.432) − 20.877*** (5.794) 0.612 297

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

Table 9 Regression results of average patent citation (one year R & D lag). Dependent variable : acitation Basic model Explanatory variable Deregulation dummy Generation entry liberalization Retail entry liberalization Generation competition Retail competition Ln(RD intensity) (lag1) Ln(assets) Constant R-square Observations

Model1 OLS ***

0.460 (0.047)

Extended model Model2 FE ***

0.324 (0.727)

Model3 RE

Model4 RE

Model5 RE

Mode6 RE

Model7 RE

0.441 (0.042)

0.266*** (0.033)

0.177*** (0.058) 0.691*** (0.142) 0.072*** (0.015)

0.221*** (0.055) 0.036 (0.325) 1.195* (0.642) 0.373 252

0.114 (0.090) 0.423** (0.167) − 5.008* (2.769) 0.381 252

Model8 RE

***

0.211*** (0.065) 0.033 (0.225) 1.215* (0.622) 0.366 252

0.271*** (0.058)

0.293*** (0.063)

0.043 (0.034) 1.384** (0.679) 0.307 252

0.067* (0.036) 1.230* (0.734) 0.213 252

0.337*** (0.066) 0.074** (0.036) − 0.479 (0.746) 0.231 252

0.210*** (0.060) 0.285*** (0.065) 0.048 (0.040) 1.506* (0.791) 0.198 252

0.392* (0.227) − 0.005 (0.024) 0.032 (0.072) 0.270*** (0.058) 0.041 (0.036) 1.392 (0.905) 0.320 252

* p = 0.1. ** p = 0.05. *** p = 0.01.

Table 10 Change in private sector R & D expenditure by technology type from 1994 to 2005 (compiled by author with data from Central Electric Power Council). Category Overall R & D expenditure Public-oriented

Cost-saving

Research topic

Change

Nuclear Environmental technology Energy efficiency Power system Power generation efficiency Information technology other

− − − − − − − −

41.9% 50.8% 46.5% 42.7% 63.5% 18.1% 23.3% 27.7%

oriented and based on cost-saving, such as power generation efficiency and information technology, are less affected. Public-oriented research may hardly provide any short-term incentives to utilities in a deregulated market. However, environment-related technology, energy efficiency, and power system are among the top priorities in Japanese energy policy. In order to achieve government targets in the deregulated market, further market and institutional reforms are thus required (Newbery, 2012). The results do not necessarily imply that utilities should simply increase or decrease R & D investment. It is impossible to determine if current R & D investment is above or below the “optimal level” (if it does exist). Furthermore, it might be reasonable to suspect that declining R & D input from utilities may indicate that the main players of innovation shift toward electric equipment industry or other new entrants. However, Sanyal and Ghosh (2013) found a decline in upstream innovation due to the restructuring of the US electric sector. Meanwhile, new entrants exhibit little interest or capability in R & D. Competitive strategies adopted by new entrants mostly focus on two topics: efficiency in order to reduce cost and increase margins and differentiation in contracts (Jamasb and Pollitt, 2008). The declining R & D effort may be detrimental to the reliability and dynamic efficiency of the electricity system, especially when more renewable energy has been incorporated. It could also damage innovation maintenance when introducing smart grids and dealing with environmental concerns. Gugler et al. (2013) stated that there is an inherent trade-off between static and dynamic efficiency in high sunk-cost

Fig. 3. Utility R & D productivity (1978–2011).

renewable energy) encourages renewable energy technology innovation. More recently, Lindnab and Soderholm (2016) measured policyinduced innovation using wind energy patent data in Europe, which indicates a positive impact of government efforts on renewable energy. Thus, government R & D effort toward renewable energy has often been emphasized along with the renewable energy deployment policy, such as feed-in tariff policy. Innovation in the electricity industry exhibits strong path dependence. Therefore, at incumbent utilities, we find incremental innovation, which builds high barriers for radical change. Markard and Truffer (2006) argued that market liberalization induced a shift from incremental, technology-oriented innovation to more radical, customer-oriented product, and organizational innovations. They concluded that market liberalization is a drive for the overall level of innovation activities as competition challenges incumbent electric utilities and newcomers. Even though their results were mainly based on EU utilities surveys, similar events took place in Japan during the deregulation process. Except for the decrease in the amount of total R & D investment, Table 10 indicates the change in R & D expenditure breakups after deregulation. We note that the input of public-oriented R & D projects, including nuclear power, environmental technology, energy efficiency, and power system, is greatly reduced. The projects that are business412

Energy Policy 111 (2017) 403–413

N. Wang, G. Mogi

Goto, M., Inoue, T., Sueyoshi, T., 2013. Structural reform of Japanese electric power industry: separation between generation and transmission & distribution. Energy Policy 56, 186–200. Goto, M., Sueyoshi, T., 2009. Productivity growth and deregulation of Japanese electricity distribution. Energy Policy 37, 3130–3138. Goto, M., Tsutsui, M., 2008. Technical efficiency and impacts of deregulation: an analysis of three functions in U.S. electric power utilities during the period from 1992 through 2000. Energy Econ. 30, 15–38. Gugler, K., Rammerstorfer, M., Schimitt, S., 2013. Ownership unbundling and investment in electricity markets –A cross country study. Energy Econ. 40, 702–713. Hattori, T., 2005. Liberalization and R & D in Japanese electricity industry- an initial observation of patent data. In: Proceedings of the 28th Annual International Association for Energy Economics (IAEE) Conference. Hattori, T., Tsutsui, M., 2004. Economic impact of regulatory reforms in the electricity supply industry: a panel data analysis for OECD countries. Energy Policy 32, 823–832. Jacquier-Roux, V., Bourgeois, B., 2002. New networks of technology creation in energy industries: reassessment of the roles of equipment suppliers and operators. Technol. Anal. Strateg. Manag. 14, 399–417. Jamasb, T., 2007. Technical change theory and learning curve: patens of progress in energy technologies. Energy J. 28, 51–71. Jamasb, T., Pollitt, M.G., 2008. Liberalisation and R & D in network industries: the case of the electricity industry. Res. Policy 37, 995–1008. Jamasb, T., Pollitt, M.G., 2011. Electricity sector liberalisation and innovation: an analysis of the UK's patenting activities. Res. Policy 40, 309–324. Jamasb, T., Pollitt, M.G., 2015. Why and how to subsidise energy R+D: lessons from the collapse and recovery of electricity innovation in the UK. Energy Policy 83, 197–205. Joskow, P.L., 2006. Markets for power in the United States: an interim assessment. Energy J. 27, 1–36. Kaino, K., 2005. Analysis of policy impacts on the electric power industry and gas industry. RIETI-Discussion Paper05-J-034. Kamien, M.I., Schwartz, N.Z., 1975. Market structure and innovation: a survey. J. Econ. Lit. 13, 1–37. Kim, J., Kim, Y., Flacher, D., 2012. R & D investment of electricity- generation firms following industry restructuring. Energy Policy 48, 103–117. Lindnab, A., Soderholm, P., 2016. Wind energy and green economy in Europe: Measuring policy-induced innovation using patent data. Appl. Energy 179, 1351–1359. Margolis, R.M., Kammen, D.M., 1999. Evidence of under-investment in energy R & D in the United States and the impact of Federal policy. Energy Policy 27, 575–584. Markard, J., Truffer, B., 2006. Innovation processes in large technical systems: market liberalization as a drive for radical change? Res. Policy 35, 609–625. Mokyr, J., 1992. The Lever of Riches: technological Creativity and Economic Progress. Oxford University Press, Oxford. Munari, F.,Sobrero, M., 2002. The effects of privatization on R & D investments and patent productivity. FEEM Working Paper No. 64. Nakano, M., Managi, S., 2008. Regulatory reforms and productivity: an empirical analysis of the Japanese electricity industry. Energy Policy 36, 201–209. Nemoto, J., Goto, M., 2004. Technological externalities and economies of vertical integration in the electric utility industry. Int. J. Ind. Organ. 22, 67–81. Nesta, I., Vona, F., Nicolli, F., 2014. Environmental policies, competition and innovation in renewable energy. J. Environ. Econ. Manag. 67, 396–411. Newbery, D.M., 2012. Reform competitive electricity markets to meet environmental targets. Econ. Energy Environ. Policy 1, 69–100. Newbery, D.M., 2016. Towards a green energy economy? The EU Energy Union's transition to a low-carbon zero subsidy electricity system-Lessons from the UK's Electricity Market Reform. Appl. Energy 179, 1321–1330. OECD, 2006. Government R & D funding and company behavior: measuring behavioral additionality. OECD Publishing, Paris. Popp, D., 2006. Internationalinnovation and diffusion of air pollution control technologies: the effectsof NOX and SO2 regulation in the US, Japan, and Germany. J. Environ. Econ. Manag. 51, 46–71. Popp, D., Hascic, I., Medhi, N., 2011. Technology and the diffusion of renewable energy. Energy Econ. 33, 648–662. Salies, E., 2010. A test of the Schumpeter hypothesis in a panel of European electric utilities. In: Innovation, Economic Growth and the Firm. Edward Elgar Publishing, Northampton. Sanyal, P., 2007. The effect of deregulation on environmental research by electric utilities. J. Regul. Econ. 31, 335–353. Sanyal, P., Cohen, L.R., 2009. Powering progress: restructuring, competition, and R & D in the U.S. electricity utility industry. Energy J. 30, 41–79. Sanyal, P., Ghosh, S., 2013. Product market competition and upstream innovation: evidence from the US electricity market deregulation. Rev. Econ. Stat. 95 (1), 237–254. Sterlacchini, A., 2012. Energy R & D in private and state-owned utilities: and analysis of the major world electric companies. Energy Policy 41, 494–506. Wang, N., Hagedoom, J., 2014. The lag structure of the relationship between patenting and internal R & D revisited. Res. Policy 43, 1275–1285. Wang, N., Mogi, G., 2017. Industrial and residential electricity demand dynamics in Japan: How did price and income elasticities evolve from 1989 to 2014? Energy Policy 106, 233–243.

network industry. This analysis was based on a dynamic panel regression with the data of 16 European countries between 1998 and 2008. We also emphasize that the trade-offs between static and dynamic efficiencies also exist from the perspective of firm innovation. Less incentive for electric sector capital investment may be preferable considering Japan's current situation. However, less motivation in technological R & D investment will, in long term, negatively affect the electric sector's efficiency. The most efficient way to deal with such “unintended consequences” continues to be an issue of further policy design during the implementation of deregulation. For example, the collapse in the UK energy sector R & D provided good lessons on the importance of government interference when market failure occurs (Jamasb and Pollitt, 2015). In terms of environmental regulation and governance policy, it will become increasingly demanding to achieve environmental and renewable energy targets in a deregulated market. According to Newbery (2016), the pathway to a low-carbon future requires adequate and efficient allocation of funding for R & D, along with public support to mitigate the adverse impact of deregulation. Our empirical results support this idea. In a deregulated electric sector, the government's role becomes even more important, especially to fund and maintain research on long-term, public-oriented, and environment-related projects. Acknowledgement We are grateful to Mr. Kaneboshi (Kansai Electric Power Company), Dr. Hattori (CREIPI), and Dr. Nagano (CREIPI) for their valuable advice. We also thank Prof. Matsumura (the University of Tokyo) and Mr. Toyoda (the Institute of Energy Economics, Japan) for their constructive comments. We also thank the feedback from 40th IAEE International Conference in Singapore. References Adolf, B., Gardiner, M., 1932. The Modern Corporation and Private Property. Transaction Publishers, New Brunswick. Aghion, P., Bloom, N., Blunbell, R., Griffith, R., Howitt, P., 2005. Competition and innovation: an inverted-U relationship. Q. J. Econ. 120, 701–728. Arrows, K.J., 1962. Economic welfare and allocation of resources for innovation. In: Nelson, R. (Ed.), The Rate and Direction of Inventive Activity: Economic and Social Factors. Princeton UniversityPress, Princeton. Becker, B., Pain, N., 2003. What determines industrial R & D in the UK? National Institute of Economic and Social Research, London. Bell, R., Seden, W., 1998. Utility restructuring and the transformation of industry sponsored R & D. Electr. J. 11, 32–39. Bell, R.A., Schneider, T.R., 1996. Balkanization and the future of electricity R & D. Electr. J. 12, 87–98. Cohen, W.M., Levin, R.C., 1989. Empirical studies of innovation and market structure. Handbook of Industrial Organization 2 Elsevier, Amsterdam. Cornaggiaa, J., Mao, Y., Tian, X., Wolfed, B., 2015. Does banking competition affect innovation? J. Financ. Econ. 115, 189–209. Costa-Campi, M.T., Duch-Brown, N., Garcia-Quevedo, J., 2014. R & D drivers and obstacles to innovation in the energy industry. Energy Econ. 46, 20–30. Dang, J., Motohashi, K., 2015. Patent statistics: a good indicator for innovation in China?Patent subsidy program impacts on patent quality. China Econ. Rev. 35, 137–155. Dooley, J.J., 1998. Unintended consequence: energy R & D in a deregulated market. Energy Policy 26, 547–555. Fujii, H., Managi, S., 2016. Research and development strategy for environmental technology in Japan: a comparative study of the private and public sector. Technol. Forecast. Social. Change 112, 293–302. GAO, 1996. Changes in Electricity-related R & D Funding. General Accounting Office GAO/RCED-96-203. GEA, 2012. Global Energy Assessment: Towards a Sustainable Future. Cambridge University Press, Cambridge. Gilbert, R., 2006. Looking for Mr. Schumpeter: Where are we in the competition-innovation debate. Innovation Policy and the Economy 6 MIT Press, Cambridge. Goto, A., Motohashi, K., 2007. Construction of a Japanese patent database and a first look at Japanese patenting activities. Res. Policy 36 (9), 1431–1442.

413