Technology Trend Analysis of Japanese Green Vehicle Powertrains Technology Using Patent Citation Data

Abstract

As automobiles are major contributors to greenhouse gas emissions, the technological shift towards vehicle powertrain systems is an attempt to lower problems such as emissions of carbon dioxide and nitrogen oxides. Patent data are the most reliable measure of business performance for applied research and development activities when investigating knowledge domains or technology evolution. This is the first study on Japanese patent citation data of the green vehicle powertrains technology industry, using the social network analysis method, which emphasizes centrality estimates and community detection. This study not only elucidates the knowledge by visualizing flow patterns but also provides a precious and congregative method for verifying important patents under the International Patent Classification system and grasping the trend of the new technology industry. This study detects leading companies, not only in terms of the number of patents but also the importance of the patents. The empirical result shows that the International Patent Classification (IPC) class that starts with “B60K”, which includes hybrid electric vehicle (HEV) and battery electric vehicle (BEV), is more likely to be the technology trend in the green vehicle powertrains industry.

1. Introduction

Natural resource-based and environmentally sensitive industries include, for example, agriproducts, forest products, minerals and mining, oil and petrochemicals, power generation, transportation, and automobiles, which are important sources of urban pollution. Many of these are mature industries that consume scarce resources and are highly polluting. They are ecologically unsustainable and need fundamental restructuring through technological innovation [1]. Conversely, sustainable green practices embrace the use of eco-friendly design, raw materials, packaging, distribution and even reuse/retreatment after the useful life of a product. It describes practices throughout the manufacturing process that are not harmful to the environment. These practices include recycling, conserving the environment, managing and reducing wastes, complying with regulation, and controlling pollution [2]. In fact, it is argued that a firm’s long-term profitability and existence are best served by balancing them with social and environmental goals [3].

In recent years, against the backdrop of the rapid growth of companies in Asian countries and regions, the Japanese government has promoted measures to appropriately protect and utilize intellectual property in order to enhance the international competitiveness of domestic industries. Academic institutions such as universities and companies are devoting great efforts to the investigation and management of intellectual property [4]. Furthermore, as the importance of analysis tools for accurately grasping the patent application and citation status of competitors is increasing, patent data are the most reliable measure of business performance for applied research and development activities when investigating knowledge domains or technology evolution [5]. In a macro sense, patent analysis has often been employed to generate economic indicators that gauge the linkage between technology development and economic growth, estimate technological knowledge flows and their impact on productivity, or compare innovative performance in an international context. At the micro level, patent analysis has been used to evaluate the competitiveness of firms, develop technology plans, prioritize Research & Development (R&D) investment, or monitor technological change in firms [6].

As patents are the most common way in the automotive industry to protect intellectual property [7], the authors of Ref. [5] created a Japanese patents dataset of the vehicle powertrain systems for HEVs, BEVs, and fuel cell electric vehicles (FCEVs). They proposed a method of combining IPC and keywords to define “green” patents in vehicle powertrain systems, based on data of patent applications to the Japan Patent Office as recorded in EPO’s PATSTAT (2021 Spring Edition, see https://www.epo.org/searching-for-patents/business/patstat.html; accessed on 1 November 2022) database during the years 2010 to 2019. Furthermore, when analyzing patents, it is necessary to consider the social situation of each country, including the language background. Therefore, the authors of Ref. [5] collected patent description documents (abstracts and titles) not only written in English but also in Japanese, so that the database has a high level of completeness. This dataset is accessible on Mendeley Data [8]. The present study is an extension of the study reported in Ref. [5].

In the literature, patent citations are informative of links between patented innovations as well as indicative of the quality of patents or innovations [9]. Many studies have been proposed to calculate the degree of relevance and importance between documents by analyzing citation information [7,10]. For this purpose, a number of techniques may be used to manipulate and analyze patent citation data. Among these techniques, the most frequently adopted tool is social network analysis (SNA), which is a quantitative technique derived from graph theory that facilitates the analysis of interactions (edges) between actors (nodes). Network analysis shows the relationships among patents as a visual network, thereby assisting the analyzer to intuitively comprehend the overall structure of a patent database. In the context of patent analysis, individual patents account for nodes and the relationships among patents represent edges in the network. The intrinsic connectivity between patents is made cartographical by visualizing the locations of individual patents and linkage patterns among patents. In this way, it becomes possible to view the overall landscape on a global scale and from different perspectives [6], and the positions of applicants within citation networks are useful to know to explain the behavior of the applicants in the marketplace [10].

When assessing patent solutions, some of the solutions that are particularly important from the point of view of the company’s development are not patented, in order not to disclose production or technological details. However, the empirical results reported in [11] show that, in contrast to American respondents’ reports that patents are less important than other major mechanisms in protecting product innovations, Japanese respondents report patents to be about as effective as any other mechanism for protecting product innovations. Secrecy, for example, is prominent in the appropriability strategies of the U.S. but not Japanese firms. Patents are the most important channel for information flow in Japan. Regarding Japanese patents, there have not been enough reports on research of citations because of the delay in developing a database of citation relationships [4]. Previous research on the importance of patents has verified the relationship between important patents and the number of citations in a specific technology [6]. However, besides simply counting how many times a particular patent was cited forward, other factors such as whether that patent was cited by a rival company, or even by a rival company abroad, and whether the patent holds a central position in or transforms a community group should also be considered when analyzing patent citation data. Thus, a more precious and congregative method for verifying important patents and grasping the trend of the new technology industry is necessary.

Overall, the innovation literature lacks comprehensive analyses of both automotive suppliers and their intellectual property strategies against the background of the technological change towards alternative powertrain systems. In response, the present study was aimed at addressing the following research question: By investigating the landscape of the patent citation data, is there a method for looking deeply into the features of a given technology industry, such as finding out who the main powerful players are, what their intellectual property strategies are, and what the strengths and weaknesses of the industry of a given country are?

The main contributions of the present study are as follows: We provide a fresh perspective on “green” technologies which can produce ecological efficiencies. We collected citation relationships in Japanese green vehicle powertrains technology and propose a congregative method to evaluate the importance of one patent and one patent class, emphasizing a visual exploration of the patent citation landscape. This method can also be used to forecast the development tendency of green vehicle powertrains technology.

2. Materials and Methods

2.1. Schematic of Research Procedure

Figure 1 shows a schematic of our research procedure.

Our research hypothesis lies in the consideration of the impact of centrality and modularity on the measurement of the patent’s relative importance. To measure the patent’s relative importance, we first discuss the potential variables that are relevant to the patent classification and further propose a statistical learning model that has a good performance. The empirical results are then used to derive useful information on investigating and forecasting technology trends.

2.2. Method for Defining “Green” Vehicle Powertrains Patents

The technological shift towards electric vehicles is an attempt to lower transportation-related problems, such as emissions of carbon dioxide, nitrogen oxides, and particulate matter [7], especially toward BEVs and FCEVs, which are categorized as zero-emissions vehicles [12]. In [7], an elaborate search strategy using IPC classes and keywords was introduced that revealed all relevant patents and clearly differentiated the four powertrain technologies, which are internal combustion engine vehicles (ICEVs), HEVs, BEVs, and FCEVs. The results indicated that a technological shift towards alternative powertrain systems will be a major trend that will substantially affect the automotive supply industry, especially the strong ties between car manufacturers and incumbent suppliers. The authors of Ref. [13] applied a combined search strategy of IPCs and keywords as well as ‘patent families’ and ‘priority dates’ to construct their global patents dataset. Note that for the environmental innovations overlap, they took advantage of subgroup-level IPC codes, as these codes can distinguish green innovations from nongreen ones. The IPC green inventory adopted by Ref. [14] was used, which is a combination of the WIPO’s IPC Green Inventory and the OECD’s list of environmentally sound technologies (EST). In Ref. [5], the authors defined “green” patents in the vehicle powertrains field, based on data of patent applications to the Japan Patent Office as recorded in EPO’s PATSTAT 2021 Spring Edition database during the years 2010 to 2019. Then, the authors searched the titles and abstracts of patents using IPC class codes and keywords. IPC class codes for green patents were supplied by the IPC GREEN INVENTORY (see https://www.wipo.int/classifications/ipc/green-inventory/home; accessed on 1 November 2022); keywords are taken from previous research [7,13,15]. The authors also searched the titles and abstracts of patents written in Japanese. A comparison table of keywords and firm names between English and Japanese can be found in Ref. [8]. The main IPC codes and keywords are summarized in

Table 1. Data search strategy to define “green” patents.

IPC Class Symbol	Keywords
B60K*	(“automobile*” OR “vehicle*” OR “car*”) AND (“hybrid vehicle*” OR “hybrid electric vehicle*” OR “hybrid propulsion” OR “hybrid electric” OR “hybrid car*” OR “plug-in hybrid vehicle” OR “charge-in hybrid vehicle” OR “hybrid automobile*” OR “hybrid electric car*”) AND (“electric vehicle*” OR “electric car” OR “electric automobile*”) AND (“battery” or “batteries”) AND (“fuel cell*”)
F02*
F16H*
B60W*
B60L*
H02K*
H01M*

. In this paper, we collected these patents’ forward citation data through 2020 from the PATSTAT 2021 Spring Edition database. In the end, 605 pairs of patent citation data were collected.

2.3. SNA

Recent developments in the SNA field have resulted in software tools for empirical analysis and visualization, such as Gephi and R, which can facilitate the analysis and interpretation of patent statistics, e.g., patent applications [9], patent citations [10,16,17], joint patent applications, and patent license transfers [18,19]. In the present research, we used the SNA tool Gephi 0.9.2, (https://gephi.org/; accessed on 1 January 2023) for networks and graphs.

2.3.1. Evaluation of Import Patents Using Hyperlink-Induced Topic Search (HITS) Model

The network structure of a hyperlinked environment can be a rich source of information about the content of the environment, provided we have effective means for understanding it [20]. The author of Ref. [20] proposed HITS, which is a link analysis algorithm that evaluates the importance of web pages. They proposed a link-based model for the conferral of authority and showed how it leads to a method that consistently identifies relevant authoritative web pages for broad search topics. The main idea of the HITS algorithm is that the number of web pages referenced and the number of other sites that link to it are used to calculate a page’s authority and hub values, respectively. This method can be described in Figure 2 and has been used to rank the importance of journals and websites.

Simply put, pages with high authority values are pages with good content based on citations, and pages with high hub values are pages with many links to high-authority pages. In this study, patents with high authority values were regarded as important.

2.3.2. Community Detection

Social, technological, and information systems can often be described in terms of complex networks having topologies of interconnected nodes combining organization and randomness [21]. Community detection requires the partitioning of a network into communities of densely connected nodes [22]. Modularity has been used to compare the quality of the partitions obtained via different methods for detecting communities, but it also as an objective function in optimization [23]. In the present study, we detect network communities using an algorithm proposed in Ref. [22] for optimizing modularity, which allows us to study networks of unprecedented size. The method consists of two phases. First, it looks for “small” communities by optimizing modularity locally. Second, it aggregates nodes of the same community and builds a new network whose nodes are the communities. These steps are repeated iteratively until a maximum of modularity is attained.

Part of the algorithm’s efficiency stems from the fact that the gain in modularity $\Delta Q$ obtained by moving an isolated node i into a community C can easily be computed by [22]

$$\Delta Q=\left[\frac{{{\displaystyle \sum }}_{in}+2k_{i,in}}{2m}-{\left(\frac{{{\displaystyle \sum }}_{tot}+k_i}{2m}\right)}^2\right]-\left[\frac{{\sum }_{in}}{2m}-{\left(\frac{{\sum }_{tot}}{2m}\right)}^2-{\left(\frac{k_i}{2m}\right)}^2\right],$$

(1)

where ${\sum }_{in}$ is the sum of the weights of the links inside C (here the weight indicates the number of communications between two nodes), ${\sum }_{tot}$ is the sum of the weights of the links incident to the nodes in C, $k_i$ is the sum of the weights of the links incident to node i, $k_{i,in}$ is the sum of the weights of the links from node i to nodes in C, and m is the sum of the weights of all the links in the network. A similar expression is used in order to evaluate the change of modularity when i is removed from its community. In practice, one therefore evaluates the change of modularity by removing i from its community and then by moving it into a neighbor community [22].

2.4. Other Variables

We created two dummy variables for every patent. One variable equals 1 if it was cited by rival firms and is 0 otherwise, and the other variable equals 1 if it was cited by rival firms abroad and is 0 otherwise. We summarized how many times one patent was cited forward. We also collected information on the locations (longitudes and latitudes) of both the citing and cited firms in Google Maps, to allow visual exploration of the patent citation behavior.

2.5. Data Summary

The study variables are summarized in

Table 2. Summary of study variables.

IPC Class Symbol	vehicles_Classification	Total Number of Patents Was Cited Forward	Representative Cited Patent Holding Firms	Counterpart Patent Citing Firms	Times of Being Cited by Rival Firms	Times of Being Cited by Rival Firms Abroad
B60K*	HEV BEV	48	TOYOTA MOTOR CORP HONDA MOTOR CO LTD MITSUBISHI MOTORS CORP…	MITSUBISHI MOTORS CORP PORSCHE AG ZHEJIANG GEELY HOLODING GROUP…	83	22
F02*	HEV	30	HITACHI AUTOMOTIVE SYSTEMS LTD MITSUBISHI MOTORS CORP HONDA MOTOR CO LTD…	HONDA MOTOR CO LTD   NISSAN MOTOR PORSCHE AG…	38	3
F16H*	HEV	14	MAZDA MOTOR NISSAN MOTOR…	HONDA MOTOR CO LTD NISSAN MOTOR…	17	2
B60W*	HEV	37	MITSUBISHI MOTORS CORP   TOYOTA MOTOR CORP SUZUKI MOTOR CORP …	ZHEJIANG GEELY HOLODING GROUP MITSUBISHI MOTORS CORP BIT HUACHUANG ELECTRIC VEHICLE TECHNOLOGY CO LTD…	70	14
B60L*	HEV BEV FCEV	69	SUMITOMO WIRING SYSTEMS MITSUBISHI MOTORS CORP   HONDA MOTOR CO LTD…	ZHUHAI YINTONG NEW POWER TECHNOLOGY CO LTD ZHEJIANG GEELY HOLODING GROUP BELENOS CLEAN POWER HOLDING AG …	113	30
H01M*	BEV	27	TATSUNO CORP   SUMITOMO ELECTRIC INDUSTRIES TATSUNO CORP…	LONGYUAN POWER TECHNOLOGY & ENG CO LTD DENSO CORP   BIT HUACHUANG ELECTRIC VEHICLE TECHNOLOGY CO LTD…	106

.

2.6. Examination of Calculation on Patent IPC Importance Considering All Study Variables

The authors of Ref. [4] examined a scale for measuring the importance of patents using citation information for Japanese patents. While following the conventional evaluation method based on the number of citations, their results show that patents have high HITS values, which are measures of the quality of the cited literature, have the highest importance, and self-citation is important in evaluating important patents. The authors of Ref. [24] described the development of a generic approach for detecting and visualizing emerging trends and transient patterns in scientific literature. They adapted a burst-detection algorithm [20] to identify emergent research-front concepts. They also used a betweenness centrality [25] metric to highlight potential pivotal points of paradigm shift over time. Their conclusions were that research-front terms are informative cluster labels and betweenness centrality metrics identify semantically valid pivotal points. However, in short, there remains insufficient research about the methodology for evaluating the importance of one patent and one patent class, and our research will contribute to closing this gap.

We propose a multinomial logistic regression model to calculate the patent’s relative importance by considering the variables described in the previous sections. The analyzing framework of the multinomial logistic regression model, which was originally proposed by [26,27], is a useful tool for the purposes of both classification and prediction. Let $P$ be the probability that a patent is in IPC class $k$, where $k\in G:=\left\{\mathrm{B}60\mathrm{K},\mathrm{B}60\mathrm{L},\mathrm{B}60\mathrm{W},\mathrm{F}02,\mathrm{F}16\mathrm{H},\mathrm{H}01\mathrm{M}\right\}$. Using logistic curves, the multinomial logistic regression model is defined as

$$P\left(\mathrm{IPC}=k|x \right)=\frac{\exp \left({\beta }_{k0}+{\beta }_{k1}\cdot x_{VEC}+{\beta }_{k2}\cdot x_{MOD}+{\beta }_{k3}\cdot x_{AUT}+{\beta }_{k4}\cdot x_{CIA}+{\beta }_{k5}\cdot x_{RIV}+{\beta }_{k6}\cdot x_{ABR}\right)}{{{\displaystyle \sum }}_k\exp \left({\beta }_{k0}+{\beta }_{k1}\cdot x_{VEC}+{\beta }_{k2}\cdot x_{MOD}+{\beta }_{k3}\cdot x_{AUT}+{\beta }_{k4}\cdot x_{CIA}+{\beta }_{k5}\cdot x_{RIV}+{\beta }_{k6}\cdot x_{ABR}\right)}, \forall k\in G$$

(2)

where the notation is defined in

Table 3. Multinomial logistic regression model notation.

$x_{VEC}$	The vehicles classification
$x_{MOD}$	The modularity value
$x_{AUT}$	The authority value
$x_{CIA}$	Times was cited forward
$x_{RIV}$	$x_{RIV}=1$ if the patent is cited by rival firms. Otherwise, $x_{RIV}=0$
$x_{ABR}$	$x_{ABR}=1$ if the patent is cited by rival firms abroad. Otherwise, $x_{ABR}=0$
${\beta }_{k0},\dots ,{\beta }_{k6}$	Coefficients to be estimated

.

Equation (1) can be solved using the method of maximum likelihood. One can easily implement the model using R packages such as “nnet” or others that handle generalized linear models. However, not all of the coefficients are uniquely identifiable. We therefore use the IPC class “B60K” as the reference class in order to avoid the non-uniqueness issue. Note that although the estimates of coefficients change if we use a different reference class, the estimate values of response variable $P$ are unique. With “B60K” as the reference class, Equation (1) can be reformulated as

$$\ln \left(\frac{P\left(\mathrm{IPC}=j|x \right)}{P\left(\mathrm{IPC}=\mathrm{B}60\mathrm{K}|x \right)}\right)={\beta }_{j0}+{\beta }_{j1}\cdot x_{VEC}+{\beta }_{j2}\cdot x_{MOD}+{\beta }_{j3}\cdot x_{AUT}+{\beta }_{j4}\cdot x_{CIA}+{\beta }_{j5}\cdot x_{RIV}+{\beta }_{j6}\cdot x_{ABR},$$

(3)

where $j\in G\setminus \left\{\mathrm{B}60\mathrm{K}\right\}=\left\{\mathrm{B}60\mathrm{L},\mathrm{B}60\mathrm{W},\mathrm{F}02,\mathrm{F}16\mathrm{H},\mathrm{H}01\mathrm{M}\right\}$. Furthermore, the probability for a patent ID being assigned to IPC class $G$ satisfies

$$P\left(\mathrm{IPC}=\mathrm{B}60\mathrm{K}|x \right)=\frac{1}{1+{{\displaystyle \sum }}_j\exp \left(\ln \left(\frac{P\left(\mathrm{IPC}=j|x \right)}{P\left(\mathrm{IPC}=\mathrm{B}60\mathrm{K}|x \right)}\right)\right)},$$

(4)

$$P\left(\mathrm{IPC}=j|x \right)=\frac{\exp \left(\ln \left(\frac{P\left(\mathrm{IPC}=j|x \right)}{P\left(\mathrm{IPC}=\mathrm{B}60\mathrm{K}|x \right)}\right)\right)}{1+{{\displaystyle \sum }}_j\exp \left(\ln \left(\frac{P\left(\mathrm{IPC}=j|x \right)}{P\left(\mathrm{IPC}=\mathrm{B}60\mathrm{K}|x \right)}\right)\right)}, \forall j.$$

(5)

This is equivalent to

$$1=P\left(\mathrm{IPC}=\mathrm{B}60\mathrm{K}|x \right)+P\left(\mathrm{IPC}=\mathrm{B}60\mathrm{L}|x \right)+P\left(\mathrm{IPC}=\mathrm{B}60\mathrm{W}|x \right)+P\left(\mathrm{IPC}=\mathrm{F}02|x \right)+P\left(\mathrm{IPC}=\mathrm{F}16\mathrm{H}|x \right)+P\left(\mathrm{IPC}=\mathrm{H}01\mathrm{M}|x \right).$$

(6)

3. Results

3.1. Visualization of Community Detection and HITS Model on Patent Citation Network

We labeled patents using the names of the companies holding them. Clusters were detected using the algorithm in Section 2.3.2. A patent network provides valuable insights into the holistic nature of a subject set [6]. For instance, Figure 3 provides an overall network view in terms of community and centrality. From this figure, we can find there are some main clusters in our sample. The biggest cluster is colored in violet and includes many famous Japanese automotive manufacturers, such as TOYOTA MOTOR CORP HONDA MOTOR CO LTD, NISSAN MOTOR, and MITSUBISHI MOTORS. The nearby black cluster is the MITSUBISHI MOTORS-NISSAN MOTOR-TOYOTA MOTOR cluster. Then the light blue cluster is the DENSO CORP-SUMITOMO ELECTRIC-MITSUBISHI HEAVY IND LTD cluster. The light green cluster comprises HONDA MOTORS, MAZDA MOTOR, NISSAN MOTOR, MITSUBISHI MOTORS, and HITACHI AUTOMOTIVE SYSTEMS. The dark green cluster is made up of SHARP KK-NEC CORP and SUZUKI MOTOR. There is also a MITSUBISHI ELECTRIC-NISSAN MOTOR cluster, which is marked in beige. Furthermore, we can observe an orange SUMITOMO-ELECTRIC INDUSTRIES-YAZAKI CORP-SANYO ELECTRIC CO cluster and a magenta OVONIC BATTERY CO INC-PANASONIC CORP-CHUGOKU ELECTRIC POWER-SAXA INC cluster. Firms in the same cluster cite each other’s patents frequently. We can also find companies that often cite their own patents, so that there are multiple occurrences of a single firm holding a patent in one cluster.

Next, we expand the size of the nodes which have high authority values. From Figure 3, TOYOTA MOTOR has the biggest cluster in the sense that TOYOTA MOTOR has the highest authority value, meaning that TOYOTA MOTOR has many important and valuable patents in the green vehicle powertrains technology field. Other companies with high authority values are SANYO ELECTRIC CO, MAZDA MOTOR, MITSUBISHI MOTORS, and HONDA MOTOR.

3.2. Visualization of Geographically Mapped Patent-Holding Firms

Figure 4 visualizes the locations of the firms; note that data collection was described in Section 2. As shown, the Japanese green vehicle powertrains technology patent-holding firms are concentrated in four regions of Honshu, Japan, with most companies in Japan’s capital economic circle around Tokyo, followed by the Nagoya area, the Kinki region, and finally the Hiroshima area.

3.3. Visualization of Geographically Mapped Patent Citation Network

With the location information collected from Google Maps, we were able to trace patent citation knowledge flows (see Figure 5). As shown, the Japanese green vehicle powertrains technology patents were mostly cited by other domestic manufacturers, being only rarely cited by firms abroad. Among the patents cited by firms abroad, almost all were cited by Chinese manufacturers or a Chinese university (Tsinghua University), with the remaining two cited by two European companies.

3.4. Empirical Results

The results using Equation (1) (equivalent to the model of Equations (2)–(4)) are summarized in

Table 4. Results.

Coefficients	B60L		B60W		F02		F16H		H01M
	$\mathsf{\beta }\mathbf{\left(}\mathit{S}\mathbf{.}\mathit{E}\mathbf{.}\mathbf{\right)}$	$\mathbf{exp}\mathbf{\left(}\mathsf{\beta }\right)$	$\mathsf{\beta }\mathbf{\left(}\mathit{S}\mathbf{.}\mathit{E}\mathbf{.}\mathbf{\right)}$	$\mathbf{exp}\mathbf{\left(}\mathsf{\beta }\mathbf{\right)}$	$\mathsf{\beta }\mathbf{\left(}\mathit{S}\mathbf{.}\mathit{E}\mathbf{.}\mathbf{\right)}$	$\mathbf{exp}\mathbf{\left(}\mathsf{\beta }\mathbf{\right)}$	$\mathsf{\beta }\mathbf{\left(}\mathit{S}\mathbf{.}\mathit{E}\mathbf{.}\mathbf{\right)}$	$\mathbf{exp}\mathbf{\left(}\mathsf{\beta }\mathbf{\right)}$	$\mathsf{\beta }\mathbf{\left(}\mathit{S}\mathbf{.}\mathit{E}\mathbf{.}\mathbf{\right)}$	$\mathbf{exp}\mathbf{\left(}\mathsf{\beta }\mathbf{\right)}$
Intercept	20.3165 * (0.2894)		−8.8619 * (0.1749)		−6.1653 * (0.2295)		−7.5880 * (0.4710)		−8.7321 * (0.5419)
VEC = HEV	−21.1684 * (0.1862)	0.0000	8.8301 * (0.1749)	6836.9080	5.4243 * (0.2295)	226.8456	5.6839 * (0.4710)	294.0882	−11.1670 * (0.0000)	0.0000
VEC = BEV	−18.8067 * (0.2585)	0.0000	−11.4300 * (0.0000)	0.0000	−8.8347 * (0.0000)	0.0001	−9.8746 * (0.0000)	0.0001	10.1222 * (0.5419)	24888.9362
MOD	−0.0153 (0.0134)	0.9848	−0.0060 (0.0118)	0.9940	−0.0404 † (0.0169)	0.9604	0.0136 (0.0211)	1.0137	−0.0589 ‡ (0.0207)	0.9428
AUT	0.7927 (0.5378)	2.2093	0.0218 (0.4935)	1.0220	−0.8164 (0.8174)	0.4420	−7.7165 (7.4462)	0.0004	1.2058 (0.8318)	3.3394
CIA	−0.0837 (0.0512)	0.9197	−0.1341 ‡ (0.0514)	0.8745	0.0279 (0.0686)	1.0283	−0.0663 (0.0975)	0.9358	0.0918 (0.0650)	1.0961
RIV = 1	0.0686 (0.4550)	1.0711	−0.0100 (0.3896)	0.9901	0.0598 (0.5195)	1.0616	−0.3745 (0.7170)	0.6876	0.8219 (1.0826)	2.2749
ABR = 1	0.2400 (0.3474)	1.2713	−0.1840 (0.3286)	0.8319	−1.6105 ‡ (0.6133)	0.1998	−0.5720 (0.6126)	0.5644	−1.1938 † (0.5359)	0.3031

*: $p<0.001$, ‡: $p<0.01$, †: $p<0.001$.

.

We report both the estimates of coefficients $\beta =\left({\beta }_{k0},\dots ,{\beta }_{k6}\right), \forall k$ and their exponentials. Since the left-hand side of Equation (2) represents the relative odds of being in IPC class $j$ versus in “B60K”, the exponentials of $\beta$ are odds ratios, which are relative risk ratios for a unit change in the response variable. For example, the relative log odds of being in IPC class “H01M” versus in “B60K” will decrease by 1.1938 if the patent is no longer cited by rival firms abroad, and in such a case, the odds ratio (or relative risk ratio) of switching from ABR = 0 to 1 is 0.3031 for being in the IPC class “H01M” versus in “B60K”.

Considering that a change in reference class leads to a difference in estimates of coefficients $\beta$ (and therefore a difference of statistical significance of the coefficients), we then checked the overall performance of the proposed method. (Recall that the estimate values of response variable $P$ in Equation (1) are unique no matter which reference class is used.) We tested the goodness of fit using Pearson’s ${\chi }^2$ test (${\chi }^2=467$, $p<0.001$) and calculated three pseudo $R^2$s (Cox and Snell’s $R^2=0.628$, Nagelkerke’s $R^2=0.6513$, McFadden’s $R^2=0.297$), which all gave values implying that our model has pretty good performance.

Using Equations (3) and (4), we calculated the probability for a patent ID being assigned to an IPC class. Partial results are shown in

Table 5. Partial results for the probability of a patent ID being assigned to an IPC class.

ID	B60K	B60L	B60W	F02	F16H	H01M	Sum
408834179	0.1992	0.5590	0.1235	0.0828	0.0355	0.0000	1.000
380340449	0.1927	0.1236	0.1590	0.0658	0.0386	0.4203	1.000
378843943	0.2883	0.2768	0.2603	0.1027	0.0482	0.0237	1.000
378840504	0.1858	0.1454	0.1056	0.0437	0.0230	0.4964	1.000
378391336	0.1782	0.1873	0.1745	0.0633	0.0424	0.3543	1.000
	77 rows are omitted
328971327	0.2518	0.1417	0.3407	0.1905	0.0753	0.0000	1.000
328686600	0.2796	0.2017	0.2961	0.1515	0.0603	0.0107	1.000
328027904	0.0823	0.5237	0.0000	0.0000	0.0000	0.3940	1.000
320698373	0.1028	0.4281	0.0270	0.0147	0.0056	0.4218	1.000
317652676	0.1954	0.5244	0.1602	0.0631	0.0160	0.0408	1.000
Average	0.2318	0.2285	0.2119	0.0828	0.0397	0.2053	1.000

and the full results can be found in the Supplementary Materials.

Note that the sum over the IPC classes in each row of Table 5 equals 1 because of Equation (5). For the patent ID = 408834179, its probability of being assigned to class “B60L” is 0.559, based on the study variables. The average probabilities for each IPC class help us understand the trend in technology: the larger the value, the more likely it is that the trend corresponding to that IPC class will be realized. As the results show, IPC class “B60K”, which includes HEV and BEV, is more likely to be the technology trend in the green vehicle powertrains industry.

4. Conclusions

The green technology orientation at the national economic policy level is a source of change in the competitive landscape. By targeting environmental technologies for development and seeding R&D into these technologies, concrete measures are taken in Japan to support this type of innovation, for example, work towards sustainable manufacturing which utilizes recycling resources and reduces resources at industry level; deployment of zero emission social infrastructures, e.g., zero-emission-type coal fired power generation with efficient coaling/carbon capture and storage, and distribution and diversification of energy sources using IT technology at the level of infrastructures; and realization of sustainable consumption and lifestyle, for instance, by selling functionalities not goods [28]. In 2020, the Government of Japan declared its “Green Growth Strategy through Achieving Carbon Neutrality in 2050” and launched a long-term strategy to create a “virtuous cycle of economy and environment”. Japanese companies possess many technologies that contribute to decarbonization, but it is essential to expand investment in environmental technology development, create new industries related to this technology, and select companies for support. In addition, the use of intellectual property and intangible assets related to environmental technology and the reconstruction of intellectual property strategies are also necessary. Patents in particular are observed to play a more central role in diffusing information across rivals in Japan and appear to be a key reason for greater industry R&D spillovers [11].

This research mainly focuses on the research question: By investigating the landscape of patent citation data, how can we capture the features of a given technology industry? To answer this question, we studied Japanese patent citation data of the green vehicle powertrains technology industry, using a method combining SNA and statistical empirical analysis. This study contributed to closing some research gaps such as developing a database of citation relationships on Japanese patents and proposing a precious and congregative algorithm to evaluate the importance of one patent and one patent IPC class in a given technology domain. This kind of method can also be applied to analyzing the trend in other technology domains. We visualized some important SNA indexes such as centrality estimates and community groups, detected leading companies, and traced the knowledge flow from Japan. We concluded the strengths and weakness of the technology field and, finally, gave some suggestions on how to overcome the weakness of the Japanese green vehicle powertrains technology industry.

Our main results are summarized as follows:

1. This study has provided a perspective on “green” technologies which can produce ecological efficiencies. This study was an extension study of Ref. [5], in which the authors searched IPC class codes and keywords of titles and abstracts and finally defined the number of “green” patents in vehicle powertrain systems. We collected these patents’ forward citation data until 2020 from the PATSTAT 2021 Spring Edition database. In total, we collected 605 pairs of patent citation data.

2. This study emphasized visual exploration of the patent citation landscape, highlighting some SNA indexes such as HITS centrality and community group. The visualization results show that there are some main clusters in our sample. The biggest cluster is TOYOTA MOTOR CORP group and then the HONDA MOTOR CO LTD group, etc. Firms in the same cluster cite each other’s patents frequently. We also found that companies often cite their own patents, so that there were multiple occurrences of a single firm holding a patent in one cluster. TOYOTA MOTOR has the highest authority value, meaning that TOYOTA MOTOR has many important and valuable patents. Other companies with high authority values are SANYO ELECTRIC CO, MAZDA MOTOR, MITSUBISHI MOTORS, and HONDA MOTOR.

3. This study tried to visualize the patent citation flow data; the resulting figures show that the Japanese green vehicle powertrains technology patent-holding firms are concentrated in four regions of Honshu, Japan, being mostly in Japan’s capital economic circle around Tokyo, followed by the Nagoya area, the Kinki region, and finally the Hiroshima area. Regarding the main leading companies detected in the previous paragraph, their patents were mostly cited by other domestic manufacturers, being rarely cited by firms abroad. From these findings, we can conclude the features or strengths of the Japanese green vehicle powertrains technology: there are mature industrial centers, and leading companies are playing a prominent and substantial role in strategic partnering in green vehicle powertrains technologies. However, we can also find the weakness that there is a lack of transnationality from the fact that the patents are rarely cited abroad, especially outside of Asian countries.

4. The trend toward “decarbonization” is highly noticeable in the transportation sector, and “electrification” of motor vehicles on a global scale as led by EV is a big, rising tide [29]. This research proposed a congregative method, which uses multinomial logistic regression model to evaluate the importance of one patent or one patent class by considering such factors as times one patent was cited forward, whether it was cited by a rival company or even a rival company abroad, and whether it holds a central position in or transformed a community group. The empirical results showed that the IPC class that starts with “B60K”, which includes HEV and BEV, is more likely to be the technology trend in the green vehicle powertrains industry.

Building on our analysis of the technology trend in Japanese green vehicle powertrains, we suggest that in order to expedite technological progress and innovation, Japan should spawn a greater tendency to encourage Japanese companies to promote greater information sharing across rival companies and collaborate with overseas companies to help spread Japanese knowledge and technology across borders. Furthermore, in order to capture the global market, Japan should establish goals to compete and beat Europe and China, who are running ahead of Japan in terms of price competitiveness and supply records [29], and promote international standardization.

There is another salient feature in this study, which is that the companies that possess core technology, in our paper, are companies with high centralities and are quite famous and major companies such as Toyota and Honda, which indicates that Japanese manufacturing small and medium enterprises (SMEs) are lacking pivotal technology patents; thus, we suggest Japan should offer preferential policies to support SMEs, which are also acknowledged as the backbone to any economy [2].

There are also some limitations in our research. The main limitation of this study is that the sample size is relatively small, so in future research, we plan to expand our research data to include green technology data of other countries such as the U.S. and China, since it is an established fact that the U.S. and China have the most patent applications on green technology in the world, and the numbers of applications are increasing rapidly as they fiercely compete with each other. Finally, making comparisons among rival companies from multiple countries is also under consideration.