Patent Analysis on the Structure and Patterns of Competition across
Information and Communications Technologies:
Lotka-Volterra Equation Approach
Sungjoo Leea, Jeewon Choib, Hyunmyung Chob, Yongtae Parkb,*
a
Department of Industrial & Information Systems Engineering, Ajou University, Republic of Korea
b
Department of Industrial Engineering, Seoul National University, Republic of Korea
Accepted 14 August 2008
Abstract
This research suggests a new approach to incorporating both the positive and negative effects of innovations when analysing interactions between information and communications technologies (ICTs). To this end, the U.S. Patent and Trademark Office (USPTO) database was used and Lotka-Volterra equations applied. This research is an exploratory one, investigating how the equations can be applied to patent data. In particular, the equations were used to analyse competition patterns between ICTs and among ICT-related industries, and also to divide ICTs into several technology groups based on their competition patterns. The research findings will help to understand the nature of the ICT sectors and are expected to have numerous implications for strategic planning in the ICT-related industries.
Keywords: Information and communication technology, technological competition, patent analysis, lotka-volterra equations
1. Introduction1
In the contemporary knowledge economy, the ICT (information and communication technology) sector is increasing its trend share of economic activity. The outlook for the sector is improving, and many businesses seek to use ICT to sustain their competitive advantages (Sohal et al., 2001). ICTs can play pivotal roles in both national growth and firm competitiveness and, accordingly, the ICT sector has received increasing interests from academicians and practitioners alike, with many studies being conducted, especially into understanding its technological innovations and diffusions. One of the most common approaches to studying technological innovations and diffusions is patent analysis. Patents have been regarded as abundant sources of knowledge about technical progresses and innovative activities (Ernst, 2003). From the varied information contained in patent documents, patent citations are defined as the number of citations of a patent in subsequent patent, which reflect the impact of technological innovation and the pervasiveness of technological information (Narin, 1994). Citation analysis has long been applied to understand linkages between industries, nations or technologies in terms of technological innovations and knowledge flows. The analysis can be similarly used for investigating linkages among ICTs and is especially useful in the ICT sector, where technologies are closely connected with each other. Actually, recently highlighted studies have focused on
* Corresponding author. E-mail: [email protected]
identifying the relationships among ICTs by examining citation analyses using ICT sector patent data (Sorenson et al., 2006). These studies have provided us with valuable lessons about the ripple effects of technological innovation on other technologies (Jaffe and Tranjtengerg, 1999).
However, citation analysis assumes that knowledge flows tend to boost innovations and thus the main concerns have been about the positive effects of a certain technological innovation on others. Quite naturally, most previous studies investigating the patterns of technological innovation and the interrelated diffusion process between technological innovations have tended to seek for positive outcomes, and to overlook negative, inhibitory effects. The development of a specific technology may also act to arrest the development of other technologies, firstly due to substitution possibilities between technologies and secondly in terms of competition for financial resources. Where technologies can be substituted for each other, the growth of one may result in the decline of alternative technologies. In addition, within a limited R&D budget, spending on one technological improvement may be at the cost of other technologies losing the opportunity for growth. Thus, a new approach incorporating both positive and negative effects of innovation is needed when analysing interactions between technologies, especially in the ICT sector which is characterised by rapidly changing and simultaneously converging and diverging technologies. To consider both positive and negative perspectives synthetically, this study adopts Lotka-Volterra equations as the primary methodological approach. The equations are one of the most famous competition diffusion models, but there have been few attempts to apply the equations to patent data, which is one of the most well-known sources for diffusion studies.
Therefore, the purpose of this research, exploring the possibilities of applying Lotka-Volterra equations to patent analysis, is three-fold. Firstly, we intend to examine the competitive relations among ICTs by incorporating both the positive and negative effects of one ICT technological innovation into another. Secondly, we propose going into detail to uncover some core ICT technologies judged by their role in the whole sector, analyse the relations between ICT-related industries. Finally we are to classify ICTs into different groups according to their competitive diffusion processes. To this end, the U.S. Patent and Trademark Office (USPTO) database was used and, in particular, Lotka-Volterra equations were applied to ICT sector patents granted between 1963 and 2005, resulting in a competition matrix describing pair-wise competitive relations between ICTs. The matrix was then employed as an input for the next-step analyses.
The contribution of this research is as follows. Firstly, it enables us to consider various aspects of technological growth such as competition and substitution, since it considers the growth of a technology as a process of technological co-evolution. Secondly, the process enlarges our view of technological relations, by taking into account of both positive effects and negative effects. This view especially supports better decision-making for R&D investment, as most existing studies have focused only on the positive effects of one technology to others, failing to discriminate technologies with only positive effects from those with both positive effects and negative effects. Finally, in addition to our methodological contributions, our research findings regarding the ICT sector offer a practical contribution by increasing understanding of the nature of the sector, and produce numerous implications for policy-making and strategic planning in ICTs.
Other major parts of this article is organised as follows. In Section 2, the basis of citation analysis and the theoretical background and operational applications of Lotka-Volterra equations are presented. In Section 3, the overall research framework and the main themes are explained. In Section 4, the proposed research themes are analysed and related implications are provided after the contents of the database and the process of manipulating the raw data are described. Section 5 offers concluding remarks and future research issues.
2. Related studies 2.1 Citation analysis
Patents contain very specific information about technologies and thus have been studied for a long time. They are regarded as offering valuable data about technology innovation and trends and various approaches have been developed for their analysis. Patent citation analysis is one of the most commonly used approaches, especially as it is a demonstrative approach that tries to explain current technological situations in terms of technology diffusion or technology development, and focuses on the comparative or dynamic analysis of R&D outputs across corporations, industries and nations. It is a bibliographic method that leads to diverse information being obtained from the relationships between citing of patents and patent citations. The basic assumption behind citation analysis is that this frequency of citation can be used as a proxy measure to estimate the degree of the subsequent technological effects of the patented invention (Karki, 1997). The USPTO firstly started to collect citation data in 1975, and as the accumulated data has since reached a critical mass for analysis, various studies have been conducted based on the data, which can be classified into three categories according to their data processing methods and analysis contents.
The first measures patents’ strength and the technological strength of a firm possessing the patents. In those studies, it is assumed that the more often a patent is cited, the more ‘leading edge’ and central to a particular technology it is likely to be. Trajtenberg (1990) used the number of patents, weighted by citation frequencies, to measure the value of innovations, while Lanjouw and Schankerman (1999) have suggested that patent citations can be used as an indicator to measure the quality of the patent. More recent studies focus on the firm-level analysis: Hall et al. (2001) have insisted that the economic value of firms can be estimated based on patent citations, while Shane and Klock (1997) have shown that the valuation criteria of a firm should include a valuation of its intangible assets using patent citations.
The second category of studies aims to gain information about technology activities. The time gap between patents being granted and their subsequent citation in other patent applications gives a hint as to the technological cycle time involved, and patterns of knowledge flows or knowledge spill-over can also be observed through patent citation analysis. Several researchers have attempted to analyse the diffusion of R&D among industries or nations (Jaffe and Tranjtengerg, 1999) by examining the knowledge flows revealed by patent technological flows (Scherer, 1981). In a similar way, patent information has been widely used to analyse linkages between technologies, degrees of technological influence and the impact of new technologies, as well as the structure of knowledge networks between industries or nations based on the citing-cited relationship.
In the final category, studies have used citation data to analyse the similarity of technologies, which can be especially useful for firms needing to gain information about potential R&D collaborators or competitors. The common application is to measure the degree of technology overlap based on citations (Mowery, 1998), while another application in this category is to classify patents based on their co-citation information and then to suggest new patent classification schemes like the International Patent Classification (IPC) or the United States Patent Classification (USPC) (Lai and Wu, 2005).
Likewise, citation analysis has been applied to many research areas, as these are simple to analyse and easy to understand. However, in spite of its usefulness, the analysis method also has several inherent limitations. Firstly, citation analysis tends to focus only on individual patents, identifying their technological characteristics and importance. Where it is based on relations between two particular patents, it may make it difficult to understand the overall relations of all relevant patents. Most existing studies cannot help but rely on aggregated
values of bi-patent relations to understand relationships among firms, industries or nations, which may not adequately reflect reality. Secondly, there is a danger that citation relations cannot represent the true nature of relations between technologies because the citation relations are determined by the limited knowledge of inventors or patent agents. So, even for the same patent, citation relations may differ in terms of the scope of knowledge made accessible, rather than by the inherent nature of the technology being examined. Actually, Jaffe et al. (1993) found that patent citations more frequently occur between the patents held by geographically co-located firms and universities. In a similar vein, not all patents are referenced to make an invention and not all patents relating to the invention are cited in applications, which may cause a problem of ‘omission and error’. Finally, the treatment of timing is very important in citation analysis. Since old patents are exposed to public for a longer time than newer ones, their citation frequencies tend to be higher. In some technologies, patents may be cited more frequently in the earlier stage of their life cycles, while others may be referenced more in the later life cycle stages, or continuously throughout their life cycles. Those factors need to be considered if citation analysis is to yield accurate results.
2.2 Lotka-Volterra equations
Analysing technological interaction requires a model that generates intuitive understanding of interacting mechanisms. Many first-order models exist, with most emphasising the substitutive relations between successive technologies (Kumar and Kumar, 1992; Morris and Pratt, 2003; Young, 1993). Although they are often quite effective, they model only the invading competitor whose population is increasing, and ignore the declining competitor whose population in decreasing, thus failing to model the whole of the process driving competition (Morris and Pratt, 2003). To model the diffusion process of competitive technologies simultaneously, a technology competition model is suggested, which is similar to a substitution model but indicates relations more thoroughly.
Competition diffusion models primarily rely on evolutionary game theory where it is assumed that evolution really originates in the results of co-evolution, whether biological or technological evolution. Among them, one of the most famous models is the Lotka-Volterra equations. This model constitutes the simplest description of two interacting populations (Lotka, 1956), and has several advantages over other models. Firstly, although quite simple, the model can fully describe well-known economic principles such as competitive effects and the law of increasing returns. Secondly, it can account for the effects of group behaviours, as well as those of individual behaviour as emphasised in previous diffusion theories, by modelling both the internal and external effects on the populations of two different species. Finally, it can be applied to illustrate not only a predator-prey system, but also various other types of competition systems (Modis, 1999), and some essential interaction features can be studied by this model (Arato, 2003). It is thus judged a very practical model and studies are increasingly adopting it to illustrate reality (Lee et al., 2004).
To understand the Lotka-Volterra equations, we first suppose that two species S1 and S2 are each living alone in a limited environment. When both compete together in the same environment, they would increase in numbers according to the following equations.
2 1 2 2 2 2 2 1 2 1 1 1 1 1 ) ( ) ( ) ( ) ( N N b N a r dt t dN N N b N a r dt t dN − − = − − = (1)
where i
N : the populations of Si (I = 1,2) i
r : the rate constants of Si (I = 1,2) i
a : the coefficient of its own growth of Si (I = 1,2) i
b: the coefficient of competition of Si (I = 1,2)
The Lotka-Volterra equations were initially suggested to describe the interaction of biological species competing for the same resources (Lotka, 1956), and then applied to display parasitic and symbiotic relations or emerging and declining competitors, allowing for an intuitive understanding of the factors that drive co-evolution (Bazykin, 1998). Later interesting analogies between biological ecology and technological development were drawn, and several researchers started to use the equations to model competing technologies, since when the model has attracted considerable attention in the technology diffusion areas (Bharagava, 1989; Porter et al., 1991). When applied to the areas, the Lotka-Volterra model enables one to examine the technological diffusion process though competition, and also to investigate relations between two technologies (Modis, 1997). The term ‘competition’ generally implies a confrontation between two entities (in this case, technologies), and its meaning is apt to invoke thoughts of only negative effects. However in this paper, interactions between technologies are viewed in a wider sense and do not represent ‘competition’ in the limited, negative sense of the word (Pistorius and Utterback, 1997). Thus different technologies can interact with each other in a variety of modes - competition, predator-prey, mutualism, commensalism, amensalism and neutralism as well as just in simple ‘competition’ relationships (Modis, 1999). We expect this multi-mode competition approach, that allows for various permutations of positive and negative interactions between interacting technologies rather than a knowledge-flow approach, to yield a much richer setting for assessing the interaction and the subsequent plotting of offensive and defensive strategies than via a simple citing-cited approach.
Here, we make two essential assumptions to apply Lotka-Volterra equations to patent analysis. The first one is that two technologies in the analysis should be related to each other, and competing for the limited resources. This assumption is quite reasonable when we restrict our focus to technology in the ICT sector, since the argument can be made that technologies in the same industry sector are connected to each other directly by substitution relations or indirectly by complementary relations. The second one is about the two-competitor system, which requires that only two technologies should be in the competition system. Though the growth or decline of a certain technology may be affected not only by the competitor within the system but also other technologies outside the system, some essential features of the interactions between technologies can be explored using this model, and it can serve as a stepping stone towards understanding more realistic but mathematically still less tractable models of competition system (Arato, 2003). Actually, this type of analysis facilitates the understanding of ICTs, giving such information as how technologies have co-evolved in terms of technological growth or diffusion. It also gives second-hand information about competitive relations among technologies. As a consequence, we concluded it is worth applying Lotka-Volterra equations to patent analysis, even given that some assumptions may seem unrealistic.
3. Research framework
The overall research framework is described in Figure 1.
Figure 1. Overall research framework.
The analysis draws on patent documents granted by the USPTO, which uses USPC codes to classify patents according to their technology characteristics. In the first step, the USPC codes and patents to the ICT sector are selected. The OECD has suggested five industry fields in the ICT sector, and by matching the IPC (International Patent Classification) codes that correspond to each industry field to USPC codes, we could finally identify 32 USPC codes in the ICT sector. At the second step, patent documents relevant to the selected USPC codes were extracted from the USPTO database. The third step assigned a competition type to each pair of technologies represented by the USPC codes, applying Lotka-Volterra equations to the growth of patents numbers. Based on the competition patterns among the 32 technologies, we derived the competition structure of ICTs as a technological competition matrix. This matrix was used to construct the network and clusters of ICTs in the fourth and final step: specifically, core technologies in the network of ICTs were identified in the first module, competitive interactions among nine industries in the ICT sectors analysed in the second module, and the whole sector of ICTs were classified into several groups based on their competition patterns in the third module, providing some strategic implications for the management of ICTs.
4. Analysis and results 4.1 Data
The OECD definition of the ICT sector developed in 1998 consists of five categories including: telecommunications; consumer electronics; computers and office machinery; measuring and control instruments and equipment; and electronic components. For defining ICT-related patents, the OECD (2003) has adopted an approach that identifies a list of IPC codes associated with ICT-related patents (see Appendix A). In this definition, three categories from the OECD definitions are retained, and the other two (measuring and control instruments and equipment and electronic components) integrated into one category named other ICT. By matching IPC codes included in the above definition to USPC, we could
identify 37 USPC classes of ICT-related patents. However, the classification system of USPC has been updated several times, and some new classes and subclasses have been generated, and therefore data for some of the 37 classes were unavailable (especially for the early 1960s) and to avoid this problem, only 32 classes of the 37 classes were used for the analysis (see Appendix B). We identified the annual numbers of patents for each class from 1963 to 2005, the cumulative numbers of which were finally applied to the analysis. Historically, the first ICT revolution occurred in the early 1980s with the appearance of PCs, and considering this it might seem that only patents dating after that would be meaningful for our analysis. However, in terms of investigating the process of technology growth and diffusion, the introduction and growth periods provided much information about the technology as it grew towards maturity that ought also be considered, and so we decided to collect all patent data back to 1963. During the data collection we found that class 438 (semiconductor device manufacturing: process) contained the largest number of patents (61,507), while class 706 (data processing: artificial intelligence) showed the least number, at only 3,585.
4.2 Pair-wise competition analysis
Estimation of Lotka-Volterra equations: Because the basic Lotka-Volterra equations are continuous forms, the transformation into a discrete form is necessary to perform the empirical study. The basic system of equations (1) is closely related to the following logistic equations (2) for two competing species, working in discrete time intervals (Leslie, 1958).
) ( ) ( 1 ) ( ) 1 ( ) ( ) ( 1 ) ( ) 1 ( 1 2 2 2 2 2 2 2 1 1 1 1 1 1 t N t N t N t N t N t N t N t N γ α λ γ α λ + + = + + + = + (2)
Here, λi and αi are the logistic parameters for the species Si, while the coefficient of γi expresses the magnitude of the effect which each species has on the rate of increase of the other. These equations (2) are discrete forms, but non-linear, and the non-linear results can be varied by changing initial values. To solver the problem, the equations (2) can be transformed into the following linear equations (3) by taking the reciprocal, and the equations (4) express the relationships between the equations (1) and (3) (Lee et al., 2004).
2 1 1 1 1 1 1 1 2 2 2 2 2 2 ( ) 1 1 ( 1) ( ) ( ) ( ) 1 1 ( 1) ( ) ( ) N t c d e N t N t N t N t c d e N t N t N t = + + + = + + + (3) 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 ln ln ln , , 1 1 ln ln ln , , 1 1 c d e d d a b d d c d e d d a b d d γ γ = = = − − = = = − − (4)
Construction of the technological competition matrix: Theoretically we can acquire the relationships between two competing technologies from the signs of b1 and b2, the coefficients
of competition in the equations (1). If b1 has positive value, it means that technology 2 has arrested the growth of technology 1. As we use equations (3) for the estimation (rather than equations (1)), we can infer the signs of b1 and b2from the relations between the parameters in
1 1 2 2
(ln ) /(d d −1)((lnd ) /(d −1)) always has a positive value in the range of
1 0, 0 1 ( or 0, 0 2 2 )
d ≥ d ≠ d ≥ d ≠ . Thus, we can also use the signs of e1 and e2 to identify
competition types, as shown in Table 1.
Table 1. Competition types (Lee et al., 2004).
1 e (b1) e2 (b2) Types + + pure competition – (+) + (–) predator-prey – – mutualism – (0) 0 (–) commensalisms + (0) 0 (+) amensalism 0 0 neutralism
Figure 2. Technological competition matrix.
For each pair of technologies, we estimated parameters e1 and e2 repetitively, with results as shown in Figure 2(a). The matrix was then transformed into a simplified form to contain only information about the sign of the parameter. Values of 1 for positive parameters, 0 for insignificant parameters and -1 for negative parameters were assigned as shown in Figure 2(b), named as the technological competition matrix.
4.3 Competition structures and patterns
After the technological competition matrix was completed, we then applied three modules to investigate competition structures and patterns in the ICT sector.
4.3.1 Competition patterns among technologies
The aim of the first module is to examine the competition structures and patterns among the technologies, and to find major distinguished technologies. Network analysis using UCINET 6 (a specialised software system for network analysis) (Borgatti et al., 2002) was employed for this purpose, with the technological competition matrix as the input for the analysis (see Figure 2). If an n by n matrix, representing the degree of relationships between n entities, is submitted as an input for the analysis, the relations among n entities are visualised in a network form. The links between entities are generated only when the degree of their relations meet the criteria. The system also enables various analyses to be conducted to
identify the role of each entity in the developed network. In this module, the possible values for the relationships are only -1, 0 and 1, and the criteria value of 1 was used to visualise the links in the positive interaction network, and -1 in the negative interaction network. As a result, two kinds of networks could be developed, as shown in Figure 3. Coloured spots in the figure mean the technologies mainly affected by other technologies, while rectangles mean those mainly affecting other technologies.
Figure 3. Competition patterns among ICTs.
From the network, we can identify four types of influential technologies. The first and second are technologies that have been affected positively or negatively by other technologies on their own diffusion process, while the other two are technologies that have exercised positive or negative effects on the diffusion of other technologies. The five most remarkable technologies in each type (in terms of the degree of effect measured by the number of other technologies ‘affected’ or ‘affecting’) are listed in Table 2.
Table 2. Influential technologies in the ICT sector.
Interaction types USPC classes
Affecting 349, 700, 705, 345, 701 Positively Affected 349, 353, 342, 707, 235 Affecting 353, 342, 235, 367, 361 Negatively Affected 701, 345, 713, 705, 386
Note: The order is based on the degree of influence.
One interesting finding was that technologies in classes 353 (Optics: image projectors), 342 (Communications: directive radio wave systems and devices), 705 (Data processing: financial, business practice, management, or cost/price determination), 345 (Computer graphics processing and selective visual display systems) and 701 (Data processing: vehicles, navigation, and relative location) have prey-predator relations with most other technologies in the ICT sector. As it turns out, technological growths in the first two classes were inclined to impede the growth of most other ICTs, but to be facilitated by them, as those two classes correspond to both types of ‘negatively affecting’ and ‘positively affected’.
Both classes 353 and 342 (see text underlined from Table 2) are related to technologies in emerging industries and are thus based on existing technologies. 353 technologies are the core technologies in digital TV and broadcasting industries. 342 technologies are emerging technologies involved in the introduction of RFID (Radio Frequency Identification), which can be applied to or incorporated into a product, (or a person) and uses radio waves for identification purposes, and is frequently applied in distribution and supply chain
management. These new applications have replaced a whole way of life based on older technologies, which can explain the results of analysis. The technologies in the other three classes are both ‘positively affecting’ and ‘negatively affected’ (see text in bold from Table 2), and it can therefore be supposed that their technological growth has accelerated the growth of most other ICTs but has been inhibited by them. These are all established technologies (such as software solutions, telematics and PCs) which have been long-developed, with continuing incremental innovations, and can be regarded as the first generation of ICTs. While they have in part been substituted by newer radical innovations, these technologies are still basis of the other ICTs, which again matches the analysis results. The other unique technologies were those in class 349 (Liquid crystal cells, elements and systems) (see text in Italic from Table 2), which indicate the innovation patterns of ‘positively affecting’ and ‘positively affected’. Liquid crystal technologies are well known as being closely connected with other technologies in the production of a large number of IT-related products or services, resulting in the mutualism between technologies in class 349 and those in most other ICT sector classes. 4.3.2 Competition patterns among industries
The purpose of the second module is to investigate the competition patterns among industries (i.e. an aggregation of several technological fields). More meaningful lessons can be expected if the object of analysis is changed from the individual technology level to the industry level. To the end, we referenced nine next-generation industries as defined by the Korean Ministry of Information and Communications. Each of the 32 USPC classes was assigned to one of these nine industries - the final concordance results are shown in Table 3, together with the abbreviations (in brackets) used for convenience in the following text.
Table 3. Industries and technologies.
No Industry USPC classes
1 Mobile telecommunications (M) 340, 342, 343, 379, 455
2 Digital TV and broadcasting (B) 348, 349, 353, 367, 375, 381, 382, 386
3 Home networking (H) 370
4 IT SoC (So) 365, 438, 711
5 Personal computer (PC) 345, 361, 707, 708, 710, 714
6 Embedded software (SW) 235, 341, 712
7 Software solution and digital contents (SD) 705, 713
8 Telematics (T) 701
9 Intelligent service robot (R) 318, 700, 706
Taking the average of the technological competition matrix values by industry, we get an industrial competition matrix that shows the patterns of technological innovations and competitions between industries. Using the industrial competition matrix as an input for the network analysis, again, we established two kinds of networks revealing interactions among industries’ technological diffusion processes, one for positive interactions and the other for negative interactions, as shown in Figure 4. In this case, however, the mean values for all patents in each industry were used to measure the degree of relationships between the industries, and so the values in the industrial competition matrix determine the value that most clearly represented the relations between entities for both a positive network and a negative network. Setting the cut-off value as α for positive relations means that the links with a relational value bigger than α are visualised as positive interactions, while those with a relational value of less than -α are visualised as negative relations.
Figure 4. Competition patterns among industries.
The network allows several distinct patterns of competition among the industries examined. For instance, all industries B, PC and R have tended to move independently of the other six industries, implying that technological innovations in those three industries have neither exercised influence on nor been influenced by the other industries. By contrast, industries H, SD and T, each with more than 4 links on average, are shown as having been closely correlated with other sectors. Another finding is that there are no links with the same signs at both ends, which means ICTs do not exercise positive effects or negative effects interactively at the sector level across several groups of technologies.
As far as individual industry details are concerned, industries M, So and SW show prey-predator interactions with industries T, H, and SD. In the network, the growths of M, So and SW have been positively influenced by the growth of T, H and SD, while the opposite was also valid, in that the growths of M, So and SW have held back the growth of T, H and SD. Thus each of the M, So and SW industries is in a ‘positively affected’ and ‘negatively affecting’ relationship with each of the T, H and SD industries, which originate from their characteristics. Generally, in order for T, H and SD to develop, technologies from M, So and SW are essential, thus M, So and SW ‘positively affect’ T, H and SD. For the same reason, R&D budgets initially allocated to T, H, and SD have tended to be re-allocated to M, So and SW, demonstrating the ‘negatively affecting’ relationship.
Nevertheless, the interaction patterns of industry H are somewhat different from industries T and SD. H is mostly related to networking technology, and as such technologies are the very basis of the ICT industry, the H sector has positively affected most other sectors, including M, SW and So. However, the two sectors T and SD appear to deviate from our expectations, in that their growth has been negatively affected by H. We concluded that as SD comprises software solutions and digital contents, the growth of network technologies leads to technological changes in SD, impeding its stable diffusion. In the same way, T industry technologies, which specifically relate to global positioning system technology integrated with computers and mobile communications technology in automotive navigation systems, may be fundamentally affected by SD industry changes, accounting for the results in the figure. In contrast, H is positively affected by SD and T, which is probably due to the fact that high quality digital contents and software solutions encourage more people to use networks to share them, attracting more interest to the sector and facilitating more active network technology innovation. In the case of home networking technologies in particular, digital contents and software solutions might be more critical for their growth, explaining the positive effects of SD on H. In case of the effects of T on H, since T, which includes mobile technologies, helps the development of the networking technologies in H. On the other hand, SD and T positively affect and are negatively affected by all the other industries linked to them. They are closer to applications-focused industries, which do not require such high
technologies as other industries, and thus are more easily influenced by, but also support the development of, other industries.
The interactions between industries B and T were also interesting. No other industries were positively related to B, but its technology has impacted negatively on the growth of T, demonstrating amensalism relations between the two industries. Telematics are the synthesis of telecommunication and informatics, related to the broad industry connecting computers with telecommunications systems. Quite naturally, B and T share many technologies in common, with most T industry technologies being covered by the technologies in B. However, T is a basic technology sector on which many other industry sectors can prosper. For example, recent interest in a navigation systems application requires advanced image processing or satellite broadcasting technologies, which are strongly relating to B. This new application is based on T industry technologies and supported by B industry technologies, and has increased the R&D investment in the B sector, but restricted the potential for the technological growth in T. Consequently, technological development in B might reduce the possibilities of technological development in T. For other industry interactions, several interpretations of competition patterns were possible.
4.3.3. Classification of technology groups
The final module was employed to develop a new classification scheme for ICTs. Although competition structures among ICTs have already been examined in the first module, where detailed information about individual technologies can be gained, it was not easy to grasp the interactive patterns at first sight. The purpose of this module is to classify ICTs into groups according to their innovation patterns, and to investigate interrelationships between the classified groups. We used SPSS 12.0 to execute a clustering analysis on the technological competition matrix. The results of hierarchical clustering analysis indicated that it would be reasonable to organise six clusters, and k-means clustering analysis was followed to divide the 32 ICTs into six groups. These groups and the relevant USPC classes are presented in Table 4.
Table 4. Results of clustering analysis.
Group USPC classes No
1 345, 349, 700, 701, 705, 707 6 2 235, 341, 342, 353, 361, 367 6 3 318, 340, 343, 365, 379, 381, 708 7 4 386, 706, 710, 713 4 5 370, 375, 438, 712, 714 5 6 348, 382, 455, 711 4
Little similarity between ICTs was discovered in terms of contents since our clustering criteria were not based on contents similarity (such as co-citations or co-classifications) but on competition patterns. The result of clustering analysis therefore implies that competition patterns might be irrelevant to technological contents. To understand each group’s competition patterns, further network analysis was conducted on the aggregated competition matrix whose values were produced by average values in the technological competition matrix by cluster. As in the second module, the degree of relationships in the matrix took values from -1 to 1. Changing the criteria value by 0.1 from 0, we determined the cut-off value that could best describe the relations among clusters, and using the value allowed two kinds of networks to be developed describing interaction patterns among the clusters, as shown in Figure 5.
Figure 5. Competition patterns among groups.
The networks enabled us to understand the characteristics of each group from the point of view of their interaction types. For instance, technologies in 386 (Television signal processing for dynamic recording or reproducing) belong to group 4. According to the matrix analysis, group 4 impacted positively on the groups 2, 3 and 5, but was influenced negatively by group 5. This is due to the fact that 386 class technologies tend to share their technological innovations and developments with those in groups 2, 3 and 5, and also to be controlled by technologies in the group 5, such as 370 (multiplex communications), 375 (pulse or digital communications), 438 (semiconductor device manufacturing: process), 712 (electrical computers and digital processing systems: processing architectures and instruction processing), and 714 (error detection and correction, fault detection and recovery) classes. Table 5 summarises the characteristics of the six groups.
Table 5. Competition patterns among technology groups. Interaction type
Group
Affecting Affected 1 Fairly positively (3,0) Fairly negatively (0,3) 2 Very negatively (0,4) Fully positively (5,0) 3 Fairly negatively (1,4) Fairly positively (4,1) 4 Fairly positively (3,0) Slightly negatively (0,1)
5 Mixed (2,1) Mixed (2,3)
6 Mixed (2,2) Slightly negatively (0,2)
Note: (i,j): The notation of i stands for the number of positive link and j stands for the number of negative link.
From Table 5, we can see that Group 1 and Group 4 show similar interaction patterns, both characterised as positively affecting and negatively affected relations. Technologies in the two groups are mostly related to data processing, which supports the growth of the technologies in other groups. Compared to Group 4, Group 1 technologies are more negatively affected by other technologies as they include comparatively conventional data processing technologies, which can be more easily replaced by technologies from other groups. Group 2 and Group 3 are similar, having negatively affecting and positively affected relations with other groups. Technologies in those groups, which are related to generating, storing and retrieving coded data such as radio waves, audio signals and images, are mostly emerging technologies, involving huge R&D investments and with high possibilities of technology substitution. However, compared to Group 2, where the impact is rather extreme,
Group 3 (which includes electricity and electrical element technologies) coexists relatively easily with other technologies, and thus shows less negative effects towards other groups. Finally, Group 5 technologies relate to semi-conductors and their manufacture, and their positive impact is quite difficult to generalise. However, those technologies have been developed for a quite long time, with some having been substituted by others, causing negatively affected relations to other groups. Group 6 involves television, image, telecommunications and memory technologies, all of which are basic ICT technologies, and are thus continuously developed and utilised. These technologies positively affect those of other groups, but at the same time may cause technology substitutions, negatively affecting other groups. They also are affected negatively by other groups, especially when they are involved in radical innovation by other groups, causing a great deal of technology changes. 4.4 Discussions and implications
There have been several attempts to apply patents data to analysing ICTs to understand their nature and to forecast future trends. The most common approach is to analyse impacts based on patent citations (Sonrenson et al., 2006). Shin and Park (2007) developed a Korean national ICT network based on patent citations to provide a way of focusing more efficiently on key technologies and to open various strategic opportunities for national ICT development. Taking a different approach, Corrocher et al. (2007) identified 122 ICT applications using patent abstracts, which were then classified into two groups of high- and low-opportunity ICT applications. On the other hand, Choi et al. (2007) calculated the cross-impact of ICTs represented by USPCs, based on co-classification information. From these relationships, unidirectional and bidirectional impacts between ICTs were identified, although only static data was used and the types of relationships between ICTs were not identified. In addition to these previous studies, our research involves another new approach to understanding the nature of ICTs. From the parameter estimations of Lotka-Volterra equations using patent data, we have developed a technological competition matrix reflecting pair-wise interactions between ICTs. This matrix helps us to predict the possible influences of technological growths or innovations in one technology on the growth of other technologies, on the assumption that the co-evolution patterns of the past would tend to last in the future.
In particular, the technological competition matrix in the first module may make a significant contribution to the methodology of look into possible technology interaction patterns. Specifically, this analysis firstly can help decide investment priorities. If other factors are the same, a technology that both gives and gets positive effects during its co-evolution process should be prioritised. Secondly, it is possible to forecast the results of technology investment in the context of co-evolution. Though this analysis is a simple trend analysis, the analysis result could be a potential basis for a causal relationship like a cross-impact analysis. Finally, the analysis result can be used to judge the stage of the technology in its life cycle. If a technology is affected negatively by most other technologies, and the number of its patent applications is decreasing, it is likely that the technology is in its declining stage, and has been substituted by other emerging technologies. In this way, the first module will support ICT-related strategic decision-making, facilitating understanding of the co-evolutionary nature of ICTs and enabling more effective R&D investment.
Likewise, the second module helps to reveal patterns of interactions among industries rather than individual technologies, based on their diffusion processes, and thus governments may gain some strategic implications for ICT policy-making or investment decision-making. For example, a policy that a specific industry sector should be intensively fostered as a strategic industry can be discussed in terms of how the policy implementation resulted in the accumulation of current technology knowledge measured by the number of patents in other sectors, or in terms of other policy implementation outputs. Another possible application of
this research is a comparison analysis among nations. By producing a co-evolution network by nation, we can understand how nations’ ICT development patterns differ. In addition, when analysing their technology policies and investment principles together, we will be able to discover the factors behind such differences, which can support a better policy.
Finally in the third module, 32 ICTs were classified into six groups in accordance with their interactive diffusion patterns. By taking out internal and/or external factors that are common to technologies in the group, it will be possible to deduce the underlying causes that drive the competition. This module can help decision-making and policy-making in two ways. First, it can be utilised to create a new classification scheme based on technology co-evolution. Second, it can be applied to technology clustering for effective technology management. Those applications can be useful when technologies need to be evaluated by their relations with other technologies.
5. Conclusions
Since the advent of the techno-economic paradigm, technological innovation has driven economic growth and anchored the direction of social transitions. Particularly in this paradigm shift, ICTs have played a leading role, which has made their strategic importance more apparent. As the process of innovation becomes more complex, the cycle of innovation becomes shorter, and market demand in the ICT sector more volatile, a number of ICTs are emerging, growing and declining over very short time-spans through the mechanism of competition. And thus the notion of technological competition has recently been recognised as an important theme in ICTs studies.
Responding to these needs, this article deals with competition issues in the ICT sector. We have proposed a new method for investigating competition patterns and structures among ICTs, as well as among ICT-related industries, attempting to analyse the differences across the industries. We have also tried to classify ICTs into several groups in terms of structural and interactive competition patterns. To that end, Lotka-Volterra equations were adopted as a main methodology, with the numbers of patents granted in the ICT sector used as a proxy measure of technological growth. By applying patent data to the Lotka-Volterra equations, not only the positive effects of knowledge spill-over among industries, but also the negative effects caused by competition between technologies for set R&D resources, have been considered. This research may be one of the earliest attempts to use this well-known model and popularly used data to analyse technological competition and understand technological characteristics. Overall, the idiosyncratic differences between both technologies and industries were made obvious by their competition patterns, and 32 ICTs were divided into six groups competing with each other. By conducting this kind of analysis and interpreting the results, it is possible to understand the technological characteristics of the ICT sector. The competition patterns and structures can then be considered for ICT policy-making or in ICT-related R&D investment decision-making processes.
Despite those substantial contributions, this article has some limitations. First, it estimates the relationship models between ICTs in a pair-wise manner. However, the growth or decline of any one ICT can be influenced by not only one but by several other ICTs. Omitting the effects of other technologies can result in biased estimations about ICTs’ relationships. In addition, the indirect effects of one technology on others (quite usual within the ICT sectors), cannot be captured by pair-wise analysis. Though we believe that the pair-wise analysis can provide the basis of a useful understanding of technological relations in ICT sectors, other approaches that could estimate both direct and indirect effects, or the relations among several technologies simultaneously, need to be addressed in future research. For example, Structural Equation Modelling (SEM) is one possibility to capture both direct and indirect effects of one
technology on others, while co-integration analysis would be another possibility to model the co-evolutionary process of technologies. It is also possible that Lotka-Volterra equations could be developed to model more than two technologies. However, the applicability of these alternative approaches needs further investigation. Secondly, the network we obtained was too complex to allow us to make out specific interactions between technologies, since the degree of competition was left out of account in the analysis. Although the types of interaction between two competing technologies were expressed by three discrete values (-1, 0 and 1), with the sign of the parameter representing the effects of competing technology, not only its sign but also its absolute value needs to be considered for enriched results. Finally, this study is merely descriptive in nature, and needs additional work from future research before it can be used as the basis for deriving policy implications.
References
Arato, M. (2003) A famous nonlinear stochastic equation: Lotka-Volterra model with diffusion. Mathematical and Computer Modelling, 38(7/9), 709-726.
Bazykin, A. (1998) Nonlinear dynamics of interacting populations. World Scientific, Singapore.
Bharagava, S.C. (1989) Generalized Lotka-Volterra equations and the mechanism of technological substitution. Technological Forecasting and Social Change, 35(4), 319-326. Borgatti, S.P., Everett, M.G., Freeman, L.C. (2002) UCINET for Windows: Software for
social network analysis. Analytical Technologies, Harvard, MA.
Choi, C., Kim, S., Park, Y. (2007) A patent-based cross impact analysis for quantitative estimation of technological impact: The case of information and communication technology. Technological Forecasting and Social Change, 74(8), 1296-1314.
Corrocher, N., Malerba, F., Montobbio, F. (2007) Schumpeterian patterns of innovative activity in the ICT field. Research Policy, 36(3), 418-432.
Ernst, H. (2003) Patent information for strategic technology management. World Patent Information, 25(3), 233-242.
Hall, B., Jaffe, A.B., Trajtenberg, M. (2001) The NBER patent citation file: Lessons, insights and methodological tools. NBER Working Paper, (October), 84-98.
Jaffe, A.B., Tranjtengerg, M. (1999) International knowledge flows: Evidence from patent citations. Economics of Innovation and New Technology, 8(1/2), 105-136.
Jaffe, A.B., Trajtenberg, M., Henderson, R. (1993) Geographic localization of knowledge spillovers as evidenced by patent citations. Quarterly Journal of Economics, 108(3), 577-598.
Karki, M. (1997) Patent citation analysis: A policy analysis tool. World Patent Information, 19(4), 269-271.
Kumar, U., Kumar, V. (1992) Technological innovation diffusion: The proliferation of substitution models and easing the user’s dilemma. IEEE Transactions on Engineering Management, 39(2), 158-168.
Lai, K., Wu, S. (2005) Using the patent co-citation approach to establish a new patent classification system. Information Processing and Management: An International Journal, 41(2), 313-330.
Lanjouw, J.O., Schankerman, M. (1999) The quality of ideas: Measuring innovation with multiple indicators. National Bureau for Economic Research, Cambridge, MA.
Lanjouw, J.O., Schankerman, M. (2001) Characteristics of patent litigation: A window on competition. RAND Journal of Economics, 32(1), 129-151.
Lee, D., Kim, J., Park, S. (2004) An estimation of demand function of Korean mobile phone market and its dynamic competition analysis using Lotka-Volterra model. Telecommunications Review, 14(2), 303-313.
Leslie, P. (1958) A Stochastic model for studying the properties of certain biological systems by numerical methods. Biometrika, 45(1-2), 16-31.
Lotka, A.J. (1956) Elements of Mathematical Biology. Dover Publications, New York.
Modis, T. (1997) Genetic re-engineering of corporations. Technological Forecasting and Social Change, 56(2), 107-118.
Modis, T. (1999) Technological forecasting at the stock market. Technological Forecasting and Social Change, 62(3), 173-202.
Morris, S., Pratt, D. (2003) Analysis of the Lotka-Volterra competition equations as a technological substitution model. Technological Forecasting and Social Change, 70(2), 103-133.
Mowery, D. (1998) Collaborative R&D: How effective is it. Issues in Science and Technology, Fall, 37–44.
Narin, F. (1994) Patent bibliometrics. Scientometics, 30(1), 147-155. OECD (2003) OECD Communications Outlook. OECD, Paris.
Pistorius, C.W.I., Utterback, J.M. (1997) Multi-mode interaction among technologies. Research Policy, 26(1), 67-84.
Porter, A.L., Roper, A.T., Mason, T.W., Rossini, F.A., Banks, J. (1991) Forecasting and management of technology. Wiley, New York.
Scherer, F.M. (1981) Using linked patent and R&D data to measure inter-industry technology flows. In Z. Griliches (Ed.), R&D, Patents, And Productivity. University of Chicago Press for NBER, Chicago, pp. 417–464.
Shane, H., Klock, M. (1997) The relation between patent citations and Tobin's Q in the semiconductor industry. Review of Quantitative Finance and Accounting, 9(2), 131-146. Sohal, A., Moss, S., Ng, L. (2001) Comparing IT success in manufacturing and service
industries. International Journal of Operations and Production Management, 21(1-2), 30-45.
Shin, J., Park, Y. (2007) Building the national ICT frontier: The case of Korea. Information Economics and Policy, 19(2), 249-277.
Sorenson, O., Rivkin, J., Fleming, L. (2006) Complexity, networks and knowledge flow. Research Policy, 35(7), 994-1017.
Trajtenberg, M. (1990) A penny for your quotes: Patent citations and the value of innovations. RAND Journal of Economics, 21 (1), 172-187.
Young, P. (1993) Technological growth curves, a competition of forecasting models. Technological Forecasting and Social Change, 44(4), 375-389.
Appendices
Appendix A. ICTs and related IPC codes
Class Corresponding IPC codes
Telecommunication G01S, G08C, G09C, H01P, H01Q, H01S3/(025, 243, 063, 067, 085, 0933, 0941, 103, 133, 18, 19, 25), H1S5, H03B, H03C, H03D, H03H, H03M, H04B, H04J, H04K, H04L, H04M, H04Q Consumer electronics G11B, H03F, H03G, H03J, H04H, H04N, H04R, H04S Computers, office machinery B07C, B41J, B41K, G02F, G03G, G05F, G06, G07, G09G, G10L, G11C, H03K, H03L Other ICT G01B, G01C, G01D, G01F, G01G, G01H, G01J, G01K, G01L, G01M, G01N, G01P, G01R, G01V, G01W, G02B6, G05B, G08G, G09B, H01B11, H01J(11/, 13/, 15/, 17/, 19/, 21/, 23/, 25/, 27/, 29/, 31/, 33/, 40/, 41/, 43/, 45/), H01L
Appendix B. UPSC classes and ICT-related patents
Class Description
235 Registers
318 Electricity: motive power systems 340 Communications: electrical 341 Coded data generation or conversion
342 Communications: directive radio wave systems and devices (e.g., radar, radio navigation) 343 Communications: radio wave antennas
345 Computer graphics processing and selective visual display systems 348 Television
349 Liquid crystal cells, elements and systems 353 Optics: image projectors
361 Electricity: electrical systems and devices 365 Static information storage and retrieval
367 Communications, electrical: acoustic wave systems and devices 370 Multiplex communications
375 Pulse or digital communications 379 Telephonic communications
381 Electrical audio signal processing systems and devices 382 Image analysis
386 Television signal processing for dynamic recording or reproducing 438 Semiconductor device manufacturing: process
455 Telecommunications
700 Data processing: generic control systems or specific applications 701 Data processing: vehicles, navigation, and relative location
705 Data processing: financial, business practice, management, or cost/price determination 706 Data processing: artificial intelligence
707 Data processing: database and file management, data structures, or document processing 708 Electrical computers: arithmetic processing and calculating
710 Electrical computers and digital data processing systems: input/output 711 Electrical computers and digital processing systems: memory
712 Electrical computers and digital processing systems: processing architectures and instruction processing (e.g., processors)
713 Electrical computers and digital processing systems: support 714 Error detection/correction and fault detection/recovery
