Changes in the Axes of Convergence of Innovation Management Research

The multidisciplinary character of the theories that support research in the discipline of Innovation Management (IM), the growing importance being attributed to the increasingly rigorous approach to IM studies by academics, and the impact of IM on the competitive advantage of firms are just some of the indicators demonstrating the relevance of this discipline in the broader field of management. These developments explain why a quantitative analysis of IM studies based on bibliometric techniques is particularly opportune. The aim of this paper is to analyze the dynamics of the intellectual structure of IM research throughout the last 20 years, to find out the main convergence axis within the field. The analysis of the intellectual structure shows that there are four convergence axes during the said period: (1) the study of how to manage innovation from the Strategic Management and Business Administration ambit; (2) New Products Development; (3) The importance of Organizational Learning and Knowledge Management for IM; and (4) The importance of the technological change, supply of technology, innovation process and innovation model. The dynamic analysis of all this, shows that in the 90’s there was a predominance of the axis (1) and (2). Subsequently, the predominance changed during the first decade of the XXI century, because the axis (2) remained dominating, but axis (4) began to appear strongly. Finally, it’s evident the strengthening of the authority of axis (4) in the intellectual structure of the IM research during the last period of study.


Methodology
We chose to use articles published in journals, because these can be considered 'certified knowledge.' This is the term commonly used to describe knowledge that has been submitted to the critical review of fellow researchers and has succeeded in gaining their approval. Research articles play a fundamental role in the said certification process (Callon, Courtial, & Penan, 1993). The use of citations from articles in journals, moreover, is a standard practice that enhances the reliability of results (Ramos-Rodríguez & Ruíz-Navarro, 2004).
To obtain a representative collection of the IM research articles, we decided to retrieve all the articles published in the Social Science Citation Index (SSCI) from 1956 through 2012 with the sequence of characters "IM" in their titles, keywords or abstracts. The reasoning behind this choice can be summarized as follows: (1) by their nature, all the published articles address Social Science issues, which saves us the arduous task of sifting through other databases in search of articles relating to the discipline that concerns us, as well as help us avoiding other publications corresponding to Sciences we are not interested to include in the present study; (2) it is highly regarded by researchers in the field; (3) its entire contents have the type required for citation analysis techniques. Once the set of articles was retrieved from the database (387 in total, from now on, citation sample), we then created a file with all the references cited in the said articles. There are, however, certain inconsistencies in the coding used in the database. Since the bibliometric software (Note 1) employed in this study recognizes only exactly coinciding strings of characters, a manual normalization process is required in order to guarantee accuracy, especially in the spelling of authors' names, the journals in which the articles appear, and the first edition of each book cited.
After analyzing the production of citations per year, we were able to identify the emergence of the discipline in 1990. Then, a deepest study was performed from 1990 through 2012.
Another file with the references corresponding to the period of 1990-2012 was then retrieved from the one which included 1956 to 2012. This file is the citation sample for that stage . Then, the cited references were determined with Bibexcel, and the inconsistencies were normalized. The cited references file was used later to count the appearance frequency. The frequency distribution obtained shows a classification of the cited documents arranged by its influence degree. After that, three sub-periods were distinguished just to identify the changes of influence throughout the time: 1990-1999, 2000-2008, and 2009-2012. Then, and specifically it's what was done in this research, the relations among the most influential works are analyzed, in order to group them into homogeneous blocks. A co-citation analysis was performed to that end (Fernandez-Alles and Ramos-Rodríguez, 2009;Gmür, 2003;McCain, 1990;Ramos-Rodríguez, 2004;Small, 1973;Shafique, 2013;White and Griffith, 1981;White and McCain, 1998).
The co-citation analysis is based on the count of the frequency with which appears cited any couple of documents corresponding to the most influential ones which were identified before. These counts are then ordered in a square and symmetrical matrix, with a dimension equal to the number of the cited documents that were chosen and with an undefined main diagonal (known as gross co-citation matrix).
Subsequently, the gross co-citation matrix is converted into another matrix which reflects similarity or normalized distance between each couple of documents. Most of the works performing co-citation analysis of authors use the Pearson correlation coefficient as a measure of normalized similarity (McCain, 1990). In this paper this criterion has been followed, although the study in this case is not about authors, but about documents.
Because of the large number of cited references which usually is handled in this kind of co-citation studies, to perform the procedure described before, it's necessary to establish a citation threshold. So, the references with a citation frequency higher or equal to that threshold are the chosen ones to perform the analysis. In the present paper, the correlation matrices contained the measures of proximity among works which had a citation threshold higher to 3,42 % from the citation sample. Table 1 shows the number of articles in the citation sample corresponding to each period, the citation threshold already established, which is the same for all the periods, and the number of references of each case. With the values which were used for the co-citation analysis, the necessary matrices for the application of traditional multivariate analysis techniques were obtained. In this occasion, and following the methodological recommendations of White and Griffith (1981), McCain (1990), White and McCain (1998) and Ramos-Rodríguez (2004), a factor analysis with a varimax rotation to the correlation matrices was carried out.
Despite of the application of the said techniques, a graphical representation of the underlying structure corresponding to the correlation matrices was also performed, to complement the study. The graph of each matrix with a correlation coefficient superior to 0,7 was represented. The final outcome of it all consists of the identification of components which reflect the homogeneous research themes and permit to interpret the results of the factor analysis in a better way

Results and Discussion
Considering that the emergence of the discipline was at the beginning of the 90's, firstly, the results gotten from the analysis of the whole period  are shown then.
Later, the same analysis was performed for each sub-period which had been established: one comprising the decade of the 90's, another including the first years of the 21st Century, and another with the last four years until 2012. All this was done to detect changes in the intellectual base of the field throughout the time.

The Intellectual Structure of IM research during the Period of 1990-2012
In this section, the proximity relations of contents among the co-citation profiles of the 51 most cited documents (cited by the 380 articles which constitute the citation sample of this period, corresponding to a citation threshold of 3,42 %), are analyzed by using Pearson correlation coefficient.
As shown in Table 2 for the period 1990-2012, seven factors with eigenvalues higher than one were extracted, which together explain the 86.575% of the total variance. As it is shown in Table 3, the documents with higher factor loading in the first factor (explained variance 43, 770 %) are Utterback (1994), Dosi (1982), Teece (1986) and Porter (1980). Those documents belong to the topic of Product Development, Technology, Innovation and Strategic Management.
Within this same factor, but with a negative loading, stands out the work of Linton and Thongpapanl (2004).
The documents corresponding to this factor can be located on the graph of the intellectual structure of the discipline (Figure 1), just in the center of it, colored in green.
The fourth factor explains an 8,052 % of the total variance. It seems to be a factor where clearly Knowledge Management predominates. Surely, it's because of the importance that Knowledge Management has for IM. Two main works stand out in this case, which at the same time are classics, above all within the topic of Knowledge Management: Nonaka (1994), and Takeuchi and Nonaka (1995). Nonaka (1994) can be located in the right-hand part of the graph (Figure 1), colored in green, where the documents corresponding to the first factor are predominant. This is because it has important factor loadings in that component as well. On the other hand, the relation between this fourth factor and the first one is more than evident, because there are several documents which correspond to the first factor, nevertheless, they have considerable loadings in the fourth factor and are related to Nonaka (1994), as it can be seen in Figure 1. They are: Barney (1991), Prahalad and Hamel (1990), and Kogut and Zander (1992).
The fifth factor explains a 5,380 % of the total variance; 78,667 % along with the other factors already commented. Two main documents belong to this component: Senge (1990), with a positive loading, about the topic of Organizational Learning, in which he establishes five competent technologies that keep strong the learning organizations; and with a negative loading, the work of Adams, Bessant and Phelps (2006), which is about the innovation process, and where the author propose a frame to evaluate the activity of innovation, with the certainty that innovation can be measured.
The sixth factor explains only a 5,118%, and 83,785% along with the others. Both works belong to the same author: Yin (1994 and, with positive loadings. Both documents are hard to classify, because of the theme they treat: the future of the case study method in evaluative research. Finally, the seventh factor explains the 2,790 % of the total variance. It's made up by a work; Pavitt (1984), about Technological Change. Eisenhardt (1989) can be analyzed as a part of this factor too (because of its loading), although it belongs to the second factor. In this work the author studies the Theory of the Agency, and recommends an agency perspective to study the problems in organization with a cooperative structure.
Von Hippel (2005) is a document that has an important loading within this factor, although it's less than 0,5. It's a book about democratizing innovation.

Dynamic Analysis: Changes in the Intellectual Structure of IM
Even though the study of the intellectual structure of IM during the last 22 years is interesting, a dynamic vision of its evolution, and transformation might contribute with the results of the current research in a very positive way. So, in the following sections we replicate the previous analysis to three correlative sub-periods.
Taking into consideration that the number of works which have been published and constitute the citation sample is not the same in each sub-period, the citation threshold of 3, 42 % was taken as the criterion to include a certain document in the analysis. Every document which had a quantity of citations, in relative terms, greater than the threshold (including it), was then included in the study. Research: 1990Research: -1999 The correlation matrix which resulted from the co-citation profiles of the 38 documents cited by the 49 articles for this period was used to perform an extraction of the main components. These components needed a varimax rotation, and eigenvalues higher than 1 to be taken into consideration. As it is shown in    Lindblom (1959) ,962 Hayes and Wheelwright (1984) ,958
The fourth factor, which explains an 8,998% of the total variance, is constituted by the following works: Schumpeter (1934) with a positive loading, which belongs to the economic field, and Simon (1976), which belongs to the same field, but in this case with a negative loading.

The Intellectual Structure of IM Research: 2000-2008
The correlation matrix which resulted from the co-citation profiles of the 69 documents cited by the 152 articles for this period was used to perform an extraction of the main components. These components needed a varimax rotation, and eigenvalues higher than 1 to be taken into consideration. As it is shown in Table 6, the result consists of seven factors which explain the 87,900 % of the total variance. As it is shown in Table 7, the first factor which explains more than 38 % of the total variance is composed by a series of works with positive loadings such as Dosi (1982), Teece (1986), Von Hippel (2005), Utterback (1994) and Schumpeter (1934); the first four are about Technologies, Innovation Process and Product Development, and the last one, about Economics. These documents have been explained previously. Christensen and Raynor (2003) stands out within the topic of Competitiveness and Technology Management in the industry. In his international bestseller "The Innovator's Dilemma", the authors expose the paradox behind the failure of many industry leaders: by placing too much focus on pleasing their most profitable customers, these firms actually paved the way for their own demise by ignoring the disruptive technologies that aggressively evolved to displace them. Dosi (1982) ,969 Teece (1986) ,959 Utterback (1994) ,956 Von Hippel (2005) ,953 Schumpeter (1934) ,947 Christensen and Raynor (2003) ,933 March (1991) ,924 Henderson and Clark (1990) ,915 Tushman and Anderson (1986) ,912 Christensen (1997) ,910 Cohen and Levinthal (1990) ,885 Nelson and Winter (1982) ,872 www.ccsenet.org/ijbm Porter (1980) ,861 Abernathy and Utterback (1978) ,860 Song (1997) -,817 Cooper and Kleinschmidt (1987) -,812 Ottum and Moore (1997) -,803 Griffin (1997) -,797
Within this same factor, but with negative loadings, there are some documents related in general to the topic of Product Development and Marketing: Urban and Hauser (1993), and Wind and Mahajan (1997). There are located in the top central part of the graph (Figure 3), colored in red.
The fourth factor explains an 8,970 % of the total variance, and a 70,259 % along with the other three. It seems to be a factor consecrated to the study of Strategic Management and Product Development. Examples of this are the works of Yin (1994), Eisenhardt (1989), Von Hippel (1986), andLeonardbarton (1992), which were already explained.
Some of the elements of this fourth factor can be visualized on the graph (Figure 3), in the top left-hand area, colored in green.
The fifth factor explains a 7,318 % of the total variance, and it seems very easy to characterize, considering that in this case the topics are homogeneous. On one side, appear Kogut and Zander (1992), and Nonaka (1994), which were explained previously and belong to the Knowledge Management field. On the other side, it's found Prahalad and Hamel (1990), that was already explained and corresponding to the topic of Strategic Management; Mintzberg (1979), whose author performs a research on the process of strategy formation, deliberate strategies, and emergent ones; and Chandler (1972), which studies the changing strategy and the structure of large industrial firms in USA.
Some of the references of this factor are shown on the graph (Figure 3). Prahalad and Hamel (1990) is one of them, located in the bottom right-hand part, just on a vertex of the triangle, colored in blue; and Mintzberg (1979), located in the center and in the left-hand part of the graph, colored in green. Surely, this last location is like that because the said document has a considerable negative loading in the first factor.
The sixth factor explains a 5,613%, and has just three works. Chiesa (1996) is one of them, about internationalization of R+D and with a positive loading. This document is on the graph (Figure 3) in the bottom left-hand part, colored in green. Rothwell (1994) is another example, but with a negative loading, and is about the periods of the innovation process, generations, and introduces topic of the fifth-generation Innovation Process. This document is also on the graph, in the top right-hand part, colored in blue.
The seventh factor explains a 4,797% of the total variance and is composed by some documents: firstly with positive loadings, Senge (1990), and in second place, Cooper and Kleinschmidt (1993) corresponding to three documents about Product Development. With a negative loading stands out Kohli y Jaworski (1990), in which the authors synthesize extant knowledge on the Marketing concept and provide a foundation for future research by clarifying the construct's domain. This document can be seen in Figure 3, in the top central part of the graph, colored in red.
Rotation Method: Varimax with Kaiser Normalization.
a Rotation converged in 11 iterations.
Within the ambit of the Strategic Management there are some documents: Henderson and Cockburn (1994) in which the authors study the theme of heterogeneous organizational 'competence' in competition, in the context of pharmaceutical research; Dyer and Singh (1998), which suggests that a firm's critical resources may span firm's boundaries and may be embedded in interfirm resources and routines; Barney (1991), about the firm's resources and sustained competitive advantage; and Williamson (1985), also related to the economic field.
Within the topic of Organizational Learning stand out some works: Levinthal and March (1993), that examines the ways in which organizations approach the problems of the learning process through simplification and specialization; Powell, Koput and Smith-Doerr (1996), whose authors argue that when the knowledge base of an industry is both complex and expanding and the sources of expertise are widely dispersed, the locus of innovation will be found in networks of learning, rather than in individual firms; March (1991), explained previously;and Zollo and Winter (2002), which performs a research on the mechanisms through which www.ccsenet.org/ijbm International Journal of Business and Management Vol. 11, No. 5; organizations develop dynamic capabilities, defined as routinized activities directed to the development and adaptation of operating routines.
On the other hand, within the topic of Innovation, mainly about mergers, Knowledge Management and Innovation Process Management, stands out Henderson and Clark (1990); Ahuja (2000), with the elaboration of a theoretical frame about interfirm networks in Innovation; Kogut and Zander (1992), which was already explained; and Benner and Tushman (2003), where a contingency view of process management's influence on both technological innovation and organizational adaptation is developed.
These elements can be visualized on the graph corresponding to Figure 4 in the central part, colored in yellow, except for Zollo and Winter (2002) which is in the bottom part, colored in blue, together with Teece, Pisano and Shuen (1997), and Helfat and Peteraf (2003), these documents belong to this very factor, but with lower loadings.
There are three important works in the same factor, but with a negative loading. One of them belongs to the topic of Product Development and Marketing, another to Diffusion of Innovation, and the third to Management.
The second factor explains the 17,838% of the total variance, and a 44,474% along with the first one. The first document to be commented is Eisenhardt and Graebner (2007), about the opportunities and challenges to building theory from case studies, then there are a group of works that have been explained previously: Yin (2003) about the method of case study; Rothwell (1994), about the Innovation Process; Tushman and Anderson (1986), about Technology; Chesbrough (2003), Chesbrough and Crowther (2006), and Yin (1994), about Open Innovation, IM Models, and Case Study; Von Hippel (1986 and, the first about Product Development and the second about Innovation Process; and Eisenhardt (1989), clearly from the ambit of Management.
There are other documents within this factor which also stand out and haven't been explained before: Piller and Walcher (2006) which treats the topic of generating competitive ideas through certain tools as a novel method to integrate the users with the development of the new products; Chesbrough (2007), and Prahogo and Sohal (2006). Some of the mentioned references can be seen in Figure 4, in the top part of the graph, colored in grey, green, blue, and black.
The third factor explains the 17,478 % of the total variance, and a 61,952 % along with the rest of factors. It's a component very easy to characterize because most of the documents that compose it can be classified as studies related to the topic of Product Development: with positive loadings stand out Ernst (2002), with an empirical review of success factors of new product development; Brown and Eisenhardt (1995), that was already explained; Langerak, Hultink, and Robben (2004), which represents two documents, one of them published in the R&D Management Journal, about the role of pre-development activities related to the orientation and performance of the market; and the other published in the Journal of Product IM, about the importance of the activities of New Products Development for the companies with a market orientation, both documents belong to the same research; Montoyaweiss Calantone (1994), explained previously; Lovelace, Shapiro and Weingart (2001), about cross-functional new product teams' innovativeness and constraint adherence; Henard and Szymanski (2001), with the conduction of a meta-analysis of the new product performance literature; Griffin (1997), andDougherty (1992), both of them were explained before.
In this very factor, but with a negative loading, appears the work of Hidalgo y Albors (2008). The objective of these Spanish researchers is to conduct a detailed review of the scope, trends, and most important actors (organizations, companies, government, consulters, academy, etc.) in the development and use of methods to manage innovation in a knowledge-driven economy. In this research they identify the main IM techniques which better improve competitiveness through Knowledge Management. This research study, is based on a survey at the European level, and concludes that a knowledge-driven economy affects the innovation process and its approach. This reference can be classified as one of those which deal with the topic of Knowledge Management and Innovation.
The fourth factor, which explains the 9,444% of the total variance, is composed by works related to the topic of Innovation, and Innovation Process. Oke (2007) performs a research on the different types of innovation that are predominant in companies in the United Kingdom services sector, the degree of innovativeness, the practices associated with the pursuit of innovation and their relationship with company performance; Damanpour (1991), that was explained previously; and Adams, Bessant and Phelps (2006), where a frame to evaluate the innovation activity is proposed, with the convincement that IM can be measured.
Within the ambit of Strategic Management and Management, are found Tidd (2001) and Wernerfelt (1984). The latter explores the usefulness of analyzing firms from the resource side rather than the product side (Barney, www.ccsenet.org/ijbm International Journal of Business and Management Vol. 11, No. 5;. Related to the field of Organizational Learning, Hurley and Hult (1998) present a conceptual framework for incorporating constructs that pertain to innovation in market orientation research. And about structure of firms, Chandy and Tellis (1998), provide an alternate view to the one which defends that firm size is the key organizational predictor of radical product innovation. Only two documents of this factor can be seen on the graph (Figure 4), in the top left-hand part, colored in rose.
The fifth factor explains an 8,514% of the total variance, and is composed by five documents. , a document which updates the standing of the Technology IM journals as a specific domain (Linton, 2006). Biemans, Griffin and Moenaert (2007), with the performance of a bibliometric analysis using all references in articles published in JPIM during its 20 years of existence. Linton and Thongpapanl (2004) with a bibliometric study, which was explained previously. If a document which uses network analysis, and two other more, that use bibliometric techniques, appear here as important documents, it's proper to suggest that the utilization of this kind of tools has increased lately within the study of IM.
Then stand out Takeuchi and Nonaka (1995) with a work which corresponds to the field of Knowledge Management, and Utterback and Abernathy (1975), with a research about Innovation of Product and Process. Nonaka (1994) is a document that must be mentioned in this context too; a paradigm to administrate dynamical aspects of the processes of creation of organizational knowledge is proposed in it. Despite of the reality that this work corresponds to the first factor, since its higher loading is located there (0,626), the loading it has in the factor five (0,625), according to the opinion of the author, is as important as the first one. Figure 4 shows that some elements of this factor are located in the central left-hand part of the graph, colored in a dark blue, almost black: ; Biemans, Griffin and Moenaert (2007); Linton and Thongpapanl (2004)). There are others which are in the bottom right-hand part of the graph, colored in yellow (Takeuchi and Nonaka (1995); Utterback and Abernathy (1975)), it's an area where Nonaka (1994) and several other documents that correspond to the first factor, are also located.
The sixth factor explains the 4,176%. It might be considered a component related to innovation measurement, and with the economic field. Only Dosi (1988) can be visualized on the graph (Figure 4) in the bottom right-hand part of it, colored in yellow.
The seventh factor explains a 2,407% and is constituted only by one document: Rogers (1995) which has the Theory of Diffusion of Innovation as the main topic.
There are other works that correspond to a different factor, but also have a considerable loading in the seventh factor, although not higher than 0,5: Utterback and Abernathy (1975), about Innovation of Product and Process; Cooper (1994), about New Product Development, and Woodman, Sawyer and , about Creativity.
The eighth factor explains the 2,077% of the total variance. Similar to the last one, is made up only by a document: Tidd, Bessant and Pavitt (2005), with which the authors develop a theoretical frame to better understand Creativity within a complex social scenario.
There are other works that correspond to a different factor, but also have a considerable loading in this seventh factor, although not higher than 0,5: Cooper, Edgett and Kleinschmidt (1999), which corresponds to the ambit of Management; Porter (1985), from the ambit of Strategic Management, and Von Hippel and Katz (2002), about Product Development.
The ninth factor explains the 1,816% of the total variance and only has the contribution of Hidalgo and Albors (2008), which has already been explained in the third factor, and is about Knowledge Management and Innovation.
Just as it happened with the other two components, there are other works that correspond to a different factor, but also have a considerable loading in the ninth factor, although not higher than 0,5: Adams, Bessant and Phelps (2006), about the Innovation Process; Hamel and Prahalad (1994), related to Strategic Management; and Cohen and Levinthal (1989), about Organizational Learning and I+D Management.
The tenth and last factor explains the 1,762%, and is also represented by only a document, which even correspond to a different factor. It's Abernathy and Utterback (1978), about the Industrial Innovation Patterns. The authors of this work propose a new model that answers some issues related to the way in which Innovation changes in a company, while this company grows up and matures.
There are other works that correspond to a different factor, but also have a considerable loading in the tenth factor, although not higher than 0,5: Chandy and Tellis (1998), about Entrepreneurial Structure; Tushman and

Conclusions
The theme of IM, and its constitution as a discipline or field of study, has been studied in several occasions, trying to establish its origins and, in general, its development. (Alba et al., 2006;Nieto Antolin, 2003;Prahogo & Sohal, 2004b;Xu et al., 2002;Xu et al., 2007;Yang et al., 2003). Usually, when something like this happens with a discipline, a series of researches begin to appear with a quantitative approach, that seek to analyze the literature generated by the academics, with the intention to assess the state of the issue, identify the achieved progress, and propose research agendas for the future (see, for example, Busenitz et al., 2003;Fernandez-Alles and Ramos-Rodríguez, 2009;Shane, 2000;Landström, 2001;Shafique, 2013). Researches like these exist within the ambit of IM too, and have been already mentioned in the present paper. Starting from the limitations of the said researches, is that it's been possible to carry out the present one and finally arrive at the following conclusions, during the last 22 years the intellectual structure of the research in IM has had in general, some axes of convergence: (1) The study of how to manage innovation from the Strategic Management and Business Administration ambit; (2) New Products Development; (3) The importance of Organizational Learning and Knowledge Management for IM; and, (4) The importance of the technological change, technology supplying, innovation process and innovation model.
The dynamic analysis of all this, shows that in the 90's there was a predominance of the axis 1 and 2. Subsequently, the predominance changed during the first decade of the XXI century, because the axis 2 remained dominating, but axis 4 began to appear strongly. Finally, it's evident the strengthening of the authority of axis 4 in the intellectual structure of the research of IM during the last period of study.
In order to deepen the results of the present paper, in future research the author pretends to widen the citation sample including more datasets, extend the temporal horizon of analysis, use more terms related to IM to do the search on the Internet, increase the number of documents used to identify the intellectual base (through a reduction of the citation threshold), and improve the general interpretation by making a good use of the explanatory power of the typical methods of Social Network Analysis, for calculations of reachability, closeness, betweenness, centrality, and the identification of sub-groups, such as cliques, N-cliques, and the calculation of Lambda Set.
Finally, corresponds to recognize the usefulness of this kind of studies to assess the state of the art in an objective way, as a complement, but never a substitute of the traditional methods of literature review. In fact, they constitute a good tool for identifying relevant authors, works and journals for the researchers of a certain discipline, as well as for showing the relations thereof. Therefore, these studies might be useful as a guide for researchers, so that they can be able to identify the relevant literature within a certain subject, building up its intellectual map permitting in this way to have an objective vision of the field through the behavior of its own authors.