The four types of organizational learning and their impact on innovation.

Subramanian, Annapoornima M.

Introduction

Evolutionary economics perceives organizations as biological organisms that struggle and compete in hostile environments (Nelson and Winter, 1982). Survival in a dynamically accelerated marketplace requires these organizations to be innovative, and learning becomes an important process through which organizations generate valuable innovation. In highly technical industries, a growing element of a firm's innovation strategy involves developing competency in both science and technology. Researchers belonging to one stream of literature that embraces competency in science have provided evidence that basic science research stimulates technological innovation (Jaffe and Trajtenberg, 1996, Henderson and Cockburn, 1994, Zucker and Darby, 1996). The general tenet of this literature is that scientific ideas arising from basic science are helpful in generating valuable innovation. For example, Jaffe (1989) finds a positive relationship between university research expenditure and local patenting rates.

The other stream of research embracing technological experience claims that it is the tacit knowledge gained through the process of technological innovation which enhances the value of innovation. The adaptive process of learning (Cohen and Levinthal, 1990), trial and error learning (Van de Ven, 1992, March and Olsen, 1976), learning by doing (Von Hippel and Tyre, 1995) and technological trajectories (Rosenberg, 1982) are good examples of this branch of research. Scholars have also empirically tested the importance of firms experimenting with novel, emerging and pioneering technologies so as to come up with breakthrough inventions (Ahuja and Lampert, 2001).

The recent findings of Gittelman and Kogut (2003) connect these two streams by concluding that the competency of firms in generating scientific ideas does not lead them to create valuable innovation. While scientific research can become useful inputs for both scientific and technology output, it is through application of science to technological innovation process that firms can generate valuable innovation. They further show that high impact innovations build heavily upon scientific literature. Therefore, it is imperative for firms to bridge the gap between science and technology to generate valuable innovation. While their research posits that it is not science itself but the application of science that enhances the value of innovation, our study tests this relationship.

In testing the relationship we build on the organizational learning model proposed by Raelin (1997), which describes the four types of learning that are fundamental to knowledge building. They are (i) conceptualization (ii) experimentation (iii) experience and (iv) reflection. We argue that each type of learning has its independent influence on the value of innovation, and that the relevance between each type of learning depends on the interplay between theory and practice. Theory is referred to the science underlying the technological innovation and practice is related to the process of technological innovation. Conceptualization learning is the extent to which firms develop scientific competency (by theorizing) necessary for technological innovation. Experimentation learning is the extent to which firms experiment their scientific knowledge in the technological innovation process. Experience is the extent to which firms develop competency by learning through innovating (by practicing). Reflection learning is the extent to which firms develop competency to relate their practical technological experience back to science and draw lessons from it in order to improve the value of innovation. Experimentation and reflection signify the competency of firms in bridging theory and practice.

The influence of the four types of learning on the value of innovation is tested using panel data of 127 biotechnology firms and their patents issued between 1980-2001. The dependent variable viz. the value of a firm's innovation is gauged using the forward citation counts of its patents. Conceptualization learning is captured using the number of scientific papers published by the firm. Experimentation learning is measured using the references to non-patent literature made by the firm in its pioneering patents. Reflection learning is measured using the number of non-patent references in nonpioneering patents. Since the number of patents issued to a firm represents the firm's hands on experience in the innovation process, it is used in gauging the experience learning.

[FIGURE 1 OMITTED]

Learning and the Value of Innovation

This section develops the research model in Figure 1, which represents the four types of learning viz. conceptualization, experimentation, reflection and experience that are essential for firms to generate valuable innovation. Conceptualization and experience correspond to learning emphasized by the two streams of research mentioned in the introduction section. Reflection and experimentation signify dimensions that bridge the gap between scientific and technological domains.

Conceptualization Learning

Conceptualization provides a common language that helps in illuminating and describing actions (Raelin, 1997). In essence, conceptualization is trying to understand why events occur, i.e. the acquisition of knowledge related to know-why (Kim, 1993). Conceptualization occurs at all levels. At the individual level, people as cognitive creatures think about the world before experiencing it. In conducting their thought processes, people order their world with beliefs about cause and effect relationships, which is guided by socialization, education, etc. In highly technological innovation, conceptual learning at the firm level means developing competency in science that forms the foundation for innovation. Scientific theories provide a basis of why phenomena occurs, thus providing a source for predicting untried experiments. It also provides a base to tackle new and different problems in different contexts. Scientific theories help researchers in actively experimenting and reflecting upon their practical innovation. Scientific research also provides a basis in understanding the causal relationship between technological components before experimenting, thus reducing the combinatorial search pace of inventors by eliminating fruitless avenues (Nelson, 1982).

In R&D process, even though firms do not get everything right the first time, they do attempt to anticipate and correct as many problems before embarking on the R&D process. This process is also termed as learning before doing (Pisano, 1994), for which knowledge gained through scientific theories is a reliable source. The notion that scientific research stimulates valuable technological innovation is long established as far back as Adam Smith (Fleming and Sorenson, 1996). A series of studies by Zucker et al. (1998, 2001 and 2002) identifies the significance of star scientists (mostly from academic science) on the innovation performance of firms, again reiterating the importance of conceptual learning. Science intensity has been observed to be an important predictor of value of innovation (Gittelman and Kogut, 2003). Firms that are competent in conceptualization have better chance of increasing the scientific intensity of their innovations and hence their value. Innovation that involves tightly coupled components requires adept skills in background theory so as to ease the knowledge search process (Fleming and Sorenson, 2004). It is through developing competency in science that firms can embark on such rugged landscaped innovation that are of high value.

While the above arguments explicate the immediate benefits of conceptualization for firms, science competency can potentially hamper a firm's ability to generate valuable innovation. One of the widely criticized reason is that conceptualization can be too abstract for it to be translated into practice. Strongly grounded theoretical knowledge and assumptions can become a trap for a firm's innovative capacity by preventing it from experimenting new avenues. Excellent scientific ability may have strong negative impact on the value of innovation if a smooth knowledge transfer between science and innovation is not carried out (Gittelman and Kogut, 2003). Thus, while conceptualization provides a strong basis for improved technological innovation, it can also result in rigid cognitive maps that can have negative impact on the value of innovation. These arguments suggest that:

Hypothesis 1: The value of innovation of a firm is related to its conceptualization learning in a curvilinear (inverted u-shaped) manner.

Experimentation Learning

While conceptualization provides a foundation for understanding events in practice, too much reliance on conceptual learning has serious limitations in explaining real-world phenomena. According to Polyani (1966), most of the real life problems are not sufficiently coherent to be conceptualized as a theory. Hence, it is mandatory that learning accrued through conceptualization is tested in practice. The process of trying out abstract theories in practice is called as experimentation (Raelin, 1997). Quite often experimentation plays a vital role in resolving the dissonance between theory and practice. Argyris and Schon (1974) refer to this dissonance as the inconsistency between 'espoused theory' and 'theory in use'. Espoused theory is the abstract understanding obtained through conceptualization. Espoused theory may not be directly applicable in practice and it is through experimentation and alteration that they are changed into theory in use. Hence, conceptual learning is converted into valuable knowledge in practice through experimentation.

In technological innovation, scientific ideas are not ready-made inputs for valuable innovation (Gittelman and Kogut, 2003).While scientific research can be tailored to be useful inputs for both scientific output and technology output; invertors of firms tend to utilize their scientific competency in just producing more scientific publication so as to gain reputation in the community. But it has been shown that innovations that draw extensively on scientific ideas are valuable. Hence, apart from using the scientific competency for basic research, firms need to be efficient in transferring their scientific competency into workable ideas for innovation. A smooth transfer of knowledge from espoused scientific theories to valuable innovation requires firms to be skilled in experimentation learning. In fact, experimentation can help firms to overcome the learning traps in other types of learning. One such trap is the familiarity trap that affects firms that totally rely on learning from technological innovation experience, making them ineffective in managing disruptive innovations (Levinthal and March, 2003, Christensen, 1997). Through experimentation of scientific ideas, firms can bring about novel innovations that are more valuable.

Despite the immediate benefits of experimentation, over experimentation can also obstruct firms in creating valuable innovation. Experimentation by a firm results in exploration of new ideas that the firm is not familiar with. This results in knowledge search along unfamiliar trajectories (Fleming and Sorenson, 2004). Experimentation with unfamiliar ideas often comes with the risks of confusion and information overload that might impair the ability of firms in creating valuable innovation (Levinthal and March, 2003). These arguments suggest that

Hypothesis 2 : The value of innovation of a firm is related to its experimentation learning in a curvilinear (inverted u-shaped) manner.

Experience Learning

The extent of performance improvement through learning by repetition of task was first demonstrated in production environment (Yelle, 1979). Similar argument was extended to technological domain (Stalk et al., 1990). Cohen and Levinthal (1990) recognize this as cycle of learning. They explain that experience with a technology results in enhanced competence, which in turn fosters usage and eventually increases experience again. Hence through learning from experience, firms become competent in the respective technological domain that enhances the value of their innovation. But when firms rely extensively on learning from experience, it can also trap them in temporal, spatial and failure myopia that eventually deteriorates the value of their innovation (Levinthal and March, 1993). Since a number of studies have reiterated the benefits and drawbacks of learning by doing (Hippel and Tyre, 1995), we offer the following hypothesis as a validation of the argument.

Hypothesis 3: The value of innovation of a firm is related to its experience learning in a curvilinear (inverted u-shaped) manner.

Reflection Learning

The process of inquiry into practical experiences in order to relate it to cohesive theory for constructing meaning and explaining is called as reflection (Raelin, 1997). Viljoen et al. (1990) claim that many skilled practitioners are unable to develop a cohesive theory and explanation of their work. However, practitioners who are good at reflection learning are capable of reasoning out what they have performed and observed which can in turn shape their actions for better outcome (Raelin, 1997).

In technological innovation, it is imperative that firms are able to relate their innovation process back to theory for three reasons: (a) legitimacy, (b) absorptive capacity and (c) enhancement. First, although tacit experience helps the firms in generating valuable innovation, it is equally important that firms are also able to externalize their tacit knowledge and demonstrate the legitimacy of their innovation (Nonaka and Takeuchi, 1995). Scientific theories and jargons are legitimate tools that help in the process of externalization. Second, reflection learning can help firms to overcome maturity trap and develop capability to work on emerging technologies. Maturity trap is the tendency to work on established technologies that are well known and understood rather than working on emerging technologies that are relatively new (Ahuja and Lampert, 2001). Since emerging technologies are recently developed innovations, if firms have the capability in relating these innovations back to science, they have better absorptive capacity to work on such emerging technologies and develop them further. Third, when firms are able to relate their technological innovation experience back to science, they can overcome propinquity trap and use their knowledge in established scientific theories to enhance their innovation. Propinquity trap is the propensity to search for solutions in the neighborhood of existing solutions (Ahuja and Lampert, 2001). Firms relying only on experience tend to become a prey for propinquity trap. When firms are good in reflecting technology innovation experience back to science, their scope of knowledge search is widened. Thus, reflection learning helps firms in using their scientific knowledge for improving emerging technologies more effectively.

While reflection learning has the above benefits, it also comes with a cost that can prevent firms from coming up with valuable innovation. As explained by Argyris and Schon (1974), there exists dissonance between 'espoused theory' and 'theory in use'. Assumptions underlying espoused theories need to be challenged and updated through trial and error so as to make it fit for applying in practical situations. In the absence of such an endeavor, reflection learning can misguide inventors by affecting their decision through their tacit theoretical knowledge. These arguments suggest that

Hypothesis 4: The value of innovation of a firm is related to its reflection learning in a curvilinear (inverted u-shaped) manner.

Research Methodology

Sample and Data Selection

Patenting activities of global biotechnology firms are used in testing the hypotheses. Restricting our scope to patent data has several limitations such as (a) not all companies have same propensity to patent (b) firms can limit their patents only to most successful innovations etc. In spite of the above limitations patent data has been widely used in testing the importance of science to innovation (Fleming and Sorenson, 2004, Gittelman and Kogut, 2003, Zucker et al., 2002). The data collection was done in two phases. In the first phase, firms from global biotechnology industry were identified using Recap (Recombinant Capital) database. Recap database provides a comprehensive list of biotech companies worldwide along with their alliance, valuation and clinical trials information. For our study we chose biotech firms from all over the world that have formed alliance with at least one university during the year 1990-2000. Since academic science is one of the main attractions for these firms in forming alliance with university, a sample from this set was perceived to be most appropriate for our study that relates science and innovation. The alliances were restricted to 1990-2000, because this was the period that the industry witnessed maximum number of university alliances. 265 firms were identified in the first phase. In the second phase, we collected the patent data of these firms issued yearly, between 1980-2001. Patent related information was collected from NUS patent database, which contains all US patents issued between 1976-2004. Corresponding year wise financial data of these firms were also collected from Compustat global. Filtering those firms that did not have patent or financial data, our final sample size was 127 firms. To assess if the value of the innovation is driven by network externalities due to university alliances, robustness check was conducted to test the effect of the independent variables on dependent variable, controlling for the alliances of these firms. The results show that restricting our scope to these of firms with alliance need not affect the generalizability of the findings.

Variable Definition

Value of the Innovation

Value of the innovation was measured using the number of citation that a patent receives following its grant date. According to patent law each patent must cite previous patents that relate closely to its technology. Research demonstrates that the number of citation that a patent receives is related to its technological importance and economic value (Trajtenberg, 1990 and Albert et al, 1991, Hall, Jaffe and Trajtenberg, 2000).

Independent Variable

The independent variables and some of the control variables described below are based on the focal firm's patenting history in the period before the year that the dependent variable was observed. Thus the dependent variables and independent variables are based on different sets of patents.

Conceptualization Learning

For this variable we needed to develop a measure that taps into the extent to which firms are competent in scientific theories underlying technological innovation. We computed this using the count of scientific publications made by a firm in a year. Publication count is used in prior research to measure scientific competency of firms (Gittelman and Kogut, 2003). Web of science that provides access to peer reviewed scientific publication was used in measuring this variable.

Experimentation Learning

For this variable we needed to develop a measure that captures the extent to which firms are competent in experimenting scientific theories for their novel technological innovation. We based this measure on the number of non-patent references made by a firm in its pioneering patents. Pioneering patents are those that do not cite any other patents (Ahuja and Lampert, 2001). Since they do not cite any patents, it indicates that they have no discernable technological antecedents. It has been observed by Fleming and Sorenson (2004) that 69% of non patent references are from peer reviewed scientific journals. Hence when a firm uses science in coming up with pioneering patents, it represents its capability in experimenting science to come up with a new technological innovation that is original.

Experience Learning

For this variable we needed to develop a measure that gauges the extent to which firm learns from its prior experience. Number of patents issued to a firm represents the firms' practical hands on experience with the innovation process. Hence this was used as a proxy in capturing firms' experience learning. Similar to other independent variables, this variable was also measured based on the number of patents issued to a firm in the year previous to that of observed dependent variable patent.

Reflection Learning

For this variable we needed to develop a measure that evaluates the extent to which firms are competent in relating their technological innovation experience back to science and draw lessons from it. We based this measure on the number of non-patented references in non-pioneering patents of a firm. Non-pioneering technologies are those that have technological antecedents (Ahuja and Lampert, 2001). These patents have prior technological lineage and they have citations to those patents on which they are built on. Since these technologies have been in practice for sometime, if a firm uses science to come up with such patents, it represents the firm's capability to relate the practical innovation experience back to science for further improving the technology. Hence this was used as a proxy to measure reflection learning.

Control Variables

We included other control variables both at firm level as well as patent level. These variables include R&D intensity, firm size as measured by log of employees, firm age, total number of patents issued to a firm in the focal year, number of claims and age of the patents. In all the models we included the unobserved heterogeneity control variable, value of the patent in the previous year. Since frequency of patenting and citation can vary across technology class defines by United States Patent and Trademark Office (USTPO), we used dummy variables to reflect each class that cover biotechnology sector. For each observation, these dummy variables reflect a firm's participation or non-participation in that class in the observed year. Control variables for total number of patents, number of claims, age of the patent and technology class are based on time period t, while the other control variables are based on time period Table-1.

Model Specification and Results

Since the dependent variable is forward citation count, count model would be more appropriate for our study. Poisson model is a frequently used count model (Gittelman and Kogut, 2003). But since patent data can exhibit over dispersion, we used negative binomial model that is best suited for estimating over dispersed parameter and for providing correct standard errors (Cameron and Trivedi, 1998).

Table-1 and Table-2 provides descriptive statistics and correlation among the variables. The standard deviations reveal that there is huge difference between firms in the variables of interest. The Model-1 in Table-3 presents the results for all the control variables. Every model reported in Table-3 includes dummy variables for each technology class, but their coefficients are not reported. All the control variables turned out to be significant in model-1 except for firm size, R&D intensity and number of claims. Model-2 includes variables of the four hypothesized effects without the squared terms. Conceptualization and reflection learning coefficients are insignificant, but experimentation learning has significant positive influence on value of innovation. However, experience learning turned out to have negative impact on the value of innovation. Except for the significance level of few variables, most of the control variables had similar influence on the dependent variable as that of Model1. Model-3 includes all the four squared terms to test the possibility of curvilinearity. The coefficients of both conceptualization learning and conceptualization learning squared are insignificant suggesting that conceptual learning has no relationship with the value of innovation. Hence hypothesis-1 is rejected. But experimentation, reflection learning and both their squared terms are significant with the predicted sign, thus providing support for hypotheses 2 and 4. However, the coefficient of experience learning is negatively significant and its squared term is positively significant suggesting a u-shaped curvilinear relationship, while the hypothesis predicted inverted u-shaped relationship with the value of the innovation. Thus hypothesis-3 is rejected.

Since the sample firms for our study were chosen from Recap database that had formed alliance with at least one university, few may contend that the above relationships hold true because of network externality effects caused by the alliances. In order to check the robustness of our results, we tested the hypothesized relationship controlling for the alliance of these firms in Model-4. Apart from including the number of alliances that the focal firm had with universities, we also controlled for the number of alliances that the focal firm had with other firms. The results after controlling for the alliances are quite similar to that of model-3, suggesting that the hypothesized relationship can be generalized. Model-5 presents the hypothesized relationships after dropping the conceptualization learning variable that turned out to be consistently insignificant. The results of Model-5 are consistent with that of model-3 and 4 except for a change in significance level of experience learning coefficient.

The above results provide support for predictions in hypotheses 2 and 4, but do not provide evidence in support of hypotheses 1 and 3.

Discussion and Conclusion

The results demonstrate that conceptualization learning does not have any significant influence on the value of innovation. The result reiterates the conclusion drawn by Gittelman and Kogut (2003) that the competency of firms in generating good science does not guarantee them valuable innovation. Present study follows their contention that only through skillful application of science in their innovation process can the firms transform their scientific capability into valuable technological innovation. We contribute to the innovation literature by furthering their work and showing the learning process through which an effective transformation of knowledge from scientific to technological domain can be carried out. The strong evidence of the inverted u-shaped relationships of experimentation and reflection learning with the value of innovation in fact suggest the potential and the limits of the bridging process between scientific ideas and technological innovation. These results provide support to our argument that through experimentation and reflection learning, firms can overcome the myopia of learning in good science and effectively bridge the gap that exists between scientific and technological domain. However, in the long run, experimentation and reflection learning can be detrimental for the value of innovation.

In contrast, the effect of experience learning on valuable innovation exhibits a u-shaped relationship. Experience learning appears to be a drawback initially, most probably because of the trial and error practice in the innovation process leading to new mistakes and problems. However, in the long run when firms increasingly gain more experience in problem solving, experience learning can be very useful for generating valuable innovation.

Therefore, in the bridging process between science and technological innovation, firms need to maintain a balance among the different types of learning. This reiterates the importance of ambidexterity in the organizational learning literature (He and Wong, 2004). Our study also extends the work of Ahuja and Lampert (2001) that explains how organizations need to try out novel, pioneering and emerging technologies in order to overcome learning traps. While their study highlights the importance of innovating on such technologies, our study sheds light on the competence required by the firms to develop these valuable innovations.

The present paper has several important implications for managerial practice. Firstly, rather than focusing on a single domain of scientific or technological knowledge, firms need to nurture inventors who are competent in bridging theory and practice. Secondly, even if some firms do not have the luxury of investing in basic research, these firms can encourage their inventors to be closely connected to the open science community to benefit from knowledge spillover that can be applied to their innovation process. Thirdly, firms need to align the decision about organizational learning with their technology strategy. In uncertain situations firms tend to work on familiar technologies rather than risking themselves for breakthrough innovations. Under such circumstances, experience learning can be consistently beneficial. On the contrary, firms attempting for radical or disruptive innovations that require deviation from some of the fundamental concepts can spend more resources on experimentation learning. Similarly emerging technologies require firms to be competent in reflection learning.

Limitations and Future Research

This research is subject to the following limitations: First is the limitation pertaining to patent data which is described in the methodology section. Second, in measuring the application of science, count of all nonpatent references were taken into consideration. A more appropriate measure would have been to consider only citations to scientific publications. However, this limitation is to some extent mitigated by the observation of Fleming and Sorenson (2004). They show that majority of the non-patent references are citations to scientific publications.

The proposed learning model for creating valuable innovation has created a few avenues for future research. One possible research topic is to understand the detailed process through which experimentation and reflection learning are used respectively in transferring scientific knowledge to technological innovation. A recent study that may fall under this branch of research is the work by Fleming and Sorenson (2004). Their research has shown that science enhances technological innovation by facilitating the inventors in their knowledge search process. Furthermore, they explain that the extent to which science enhances the effectiveness of search varies systematically across applications, depending on the coupling of technological components involved in the innovation. Certainly, future studies can explore other means through which the experimentation and reflection learning can be used in transforming scientific knowledge to valuable technological innovation. This would include avenues such as investigating the extent to which experimentation and reflection learning can help firms to span both geographical and technological boundary of knowledge spillover. Since scientific community follows the idea of open science, firms that are competent in applying science to innovation may not be constrained by the technical distance and the geographical location in benefiting from knowledge spillover. Similarly, competency in the application of science to innovation can beget the advantage of cross pollination of ideas, which is helpful for valuable innovation.

References

Ahuja, G and Lampert, C. M. (2001), "Entrepreneurship in the large corporation: A longitudinal study of how established firms create breakthrough inventions." Strategic Management Journal, Vol. 22, pp. 521-543.

Albert, M., Avery, D., Narin, F. and McAllister, P. (1991), "Direct validation of citation counts as indicators of industrial important patents." Research Policy, Vol. 20, pp. 251-259.

Argyris, C. and Schon, D. A. (1974), Theory in practice: Increasing professional effectiveness, San Francisco, CA: Jossey-Bass.

Cameron, A. C. and Trivedi, P. (1998), Regression analysis of count data, Cambridge University Press, Cambridge, U.K.

Cohen, W. and Levinthal, D. (1990), "Absorptive capacity: A new perspective on learning and innovation." Administrative Science Quarterly, Vol. 35, pp. 128-152.

Fleming, L. and Sorenson, O. (2004), "Science as a map in technological search." Strategic Management Journal, Vol. 25, pp. 909-928.

Gittelman, M and Kogut, B. (2003), "Does science lead to valuable knowledge: Biotechnology firms and the evolutionary logic of citation patterns?" Management Science, Vol. 49, pp. 366-382.

Hall, B., Jaffe, A. and Trajtenberg, M. (2000), "Market value and patent citations: a first look," NBER working paper 7741.

He, Z.L. and Wong, P.K. (2004), "Exploration vs. Exploitation: An empirical test of the ambidexterity hypothesis." Organization Science, Vol. 15, pp. 81-494.

Henderson, R. and Cockburn, I. (1994), "Measuring competence?" Exploring firm effects in drug discovery. Strategic Management Journal, Vol. 15, pp. 63-84.

Jaffe, A. and Trajtenberg, M. (1996), "Flows of knowledge from universities and federal labs: Modeling the flow of patent citations over time and across institutional and geographic boundaries." Proceedings of the National Academy Sciences, Vol. 93, pp. 12671-12677.

Jaffe, A. (1989), "Real effects of academic research." American Economic Review, Vol. 79, pp. 957-970.

Kim, D. H. (1993), "The link between individual and organizational learning." Sloan Management Review, Vol. 35, pp. 37-50.

Levinthal, D.A. and March, J.G. (1993), "The myopia of learning." Strategic Management Journal, Vol. 14, pp. 95-112.

March, J. G. and Olsen, J.P. (1976), "Organizational learning and the ambiguity of the past, in Ambiguity and choice in organizations," Bergen: Universitetsforlaget, pp. 54-69.

Nelson, R. and Winter, S. (1982), An evolutionary theory of economic change, Harvard University Press, Cambridge.

Nonaka, I and Takeuchi. (1995), The knowledge creating company, Oxford University Press, New York

Pisano, G. P. (1994), "Knowledge, Integration and the locus of learning: An empirical analysis of process development." Strategic Management Journal, Vol. 15, pp. 85-100.

Polanyi, M. (1966), The tacit dimension, Garden City, NY.

Raelin, J. A. (1997), "A Model of work based learning." Organization Science, Vol. 8, pp. 563-578.

Rosenberg, N. (1982), Inside the black box, Cambridge University Press, Cambridge.

Stalk, G. and Hout, J. M. (1990), Competing against time, The Free Press, New York, NY

Trajtenberg, M. (1990), "A penny for your quotes: Patent citations on the value of innovations." Rand Journal of Economics, Vol. 21, pp. 172-187.

Van de Ven, A. H. and Polley, D. (1992), "Learning While Innovating." Organization Science, Vol. 3, pp. 92-116

Viljoen, J., Holt, D. and Petzall, S. (1990), "The MBA experience: Participants entry level conception of management." Management Education and Development, Vol. 21, pp. 1-12.

Von Hippel, E. and Tyre, M. (1995), "How the learning by doing is done: Problem identification in novel process equipment." Research Policy, Vol. 24, pp. 1-12.

Yelle, L.E. (1979), "The learning curve: Historical review and comprehensive survey." Decision Science, Vol. 10, pp. 302-208.

Zucker, L.G., Darby, M.R. and Armstrong, J. (2002), "Commercializing knowledge: University science, knowledge capture and firm performance in biotechnology." Management Science, Vol. 48, pp. 138-153.

Zucker, L.G. and Darby, M. R. (2001), "Capturing technological opportunity via Japan's star scientists: Evidence from Japanese firms' biotech patents and products." Journal of Technology Transfer, Vol. 26, pp. 37-58.

Zucker, L, Darby, M. and Brewer, M. (1998), "Intellectual human capital and the birth of U.S. biotechnology enterprises." American Economic Review, Vol. 88, pp. 290-306.

Zucker, L. and Darby, M. Stuart. (1996), "Scientists and institutional transformation: Patterns of invention and innovation in the formation of the biotechnology industry." Proceedings of the National Academy Sciences, Vol. 93, pp. 709-716.

Annapoornima M. Subramanian, Division of Engineering & Technology Management, National University of Singapore, Block E3A, # 04-11, 7, Engineering Drive 1, Singapore-117574, E-mail: anna.poornima@gmail.com

Table 1: Descriptive Statistics

Variables Mean Std. Deviation

Value of Innovation (FwdCitat t) 34.39 94.96
Conceptualization_Learning t-1 54.14 208.14
Experimentation_Learning t-1 15.80 46.98
Experience_Learning t-1 5.68 8.87
Reflection_Learning t-1 146.09 95.02
Firm_age t-1 10.39 5.86
Firm_Size t-1 0.74 1.35
R&D_Intensity t-1 94.18 136.76
Patent_Age t 6.54 3.58
No_of_Claims t 128.33 256.450
Total_Patent t 6.58 10.349

Table 2: Correlations

 Variables 1 2 3

1 Value of Innovation 1 -.017 .025
 (FwdCitat t)

2 Conceptualization_ -.017 1 .075
 Learning t-1

3 Experimentation_ .025 .075 1
4 Learning t-1 .147 (**) .100 (*) .665 (**)
 Experience_
 Learning t-1

5 Reflection_Learning .044 .049 .546 (**)
6 t-1 FwdCitat t-1 .532 (**) -.016 .075

7 Firm_age t-1 -.055 .247 (**) .174 (**)

8 Firm_Size t-1 .101 .322 (**) .065
9 R&D_Intensity t-1 .057 .353 (**) .023
10 Patent_Age t .272 (**) .069 -.147 (**)

11 No_of_Claims t .188 (**) .062 .537 (**)
12 Total_Patent t .296 (**) .088 (*) .558 (**)

 Variables 4 5 6

1 Value of Innovation .147 (**) .044 .532 (**)
 (FwdCitat t)

2 Conceptualization_ .100 (*) .049 -.016
 Learning t-1

3 Experimentation_ .665 (**) .546 (**) .075
4 Learning t-1 1 .837 (**) .348 (**)
 Experience_
 Learning t-1

5 Reflection_Learning .837 (**) 1 .150 (**)
6 t-1 FwdCitat t-1 .348 (**) .150 (**) 1

7 Firm_age t-1 .223 (**) .309 (**) -.015

8 Firm_Size t-1 .165 (**) .160 (**) .127 (*)
9 R&D_Intensity t-1 .090 .128 (*) .084
10 Patent_Age t -.142 (**) -.219 (**) .224 (**)

11 No_of_Claims t .714 (**) .564 (**) .129 (**)
12 Total_Patent t .828 (**) .733 (**) .216 (**)

 Variables 7 8 9

1 Value of Innovation -.055 .101 .057
 (FwdCitat t)

2 Conceptualization_ .247 (**) .322 (**) .353 (**)
 Learning t-1

3 Experimentation_ .174 (**) .065 .023
4 Learning t-1 .223 (**) .165 (**) .090
 Experience_
 Learning t-1

5 Reflection_Learning .309 (**) .160 (**) .128 (*)
6 t-1 FwdCitat t-1 -.015 .127 (*) .084

7 Firm_age t-1 1 .041 .071

8 Firm_Size t-1 .041 1 .642 (**)
9 R&D_Intensity t-1 .071 .642 (**) 1
10 Patent_Age t -.298 (**) .022 .000

11 No_of_Claims t .166 (**) .141 (**) 0.017
12 Total_Patent t .193 (**) .164 (**) 0.084

 Variables 10 11 12

1 Value of Innovation .272 (**) .188(**) .296 (**)
 (FwdCitat t)

2 Conceptualization_ .069 .062 .088 (*)
 Learning t-1

3 Experimentation_ -.147 (**) .537 (**) .558 (**)
4 Learning t-1 -.142 (**) .714 (**) .828 (**)
 Experience_
 Learning t-1

5 Reflection_Learning -.219 (**) .564 (**) .733 (**)
6 t-1 FwdCitat t-1 .224 (**) .129 (**) .216 (**)

7 Firm_age t-1 -.298 (**) .166 (**) .193 (**)

8 Firm_Size t-1 .022 .141 (**) .164 (**)
9 R&D_Intensity t-1 .000 .017 .084
10 Patent_Age t 1 -.143 (**) -.111 (**)

11 No_of_Claims t -.143 (**) 1 .869 (**)
12 Total_Patent t -.111 (**) .869 (**) 1

** Correlation is significant at the 0.01 level

* Correlation is significant at the 0.05 level

Table 3: Negative binomial regression of the impact of conceptual,
experimentation, reflection and experience learning on the value
of innovation

Variables Model-1 Model-2

Constant -2.5414 *** -2.2701 ***
 [0.4895] [0.4975]

Conceptualization_Learning t-1 0.0016
 [0.0022]

Conceptualization_Learning_Squared t-1

Experimentation_Learning t-1 0.0014 +
 [0.0010]

Experimentation_Learning_Squared t-1

Experience_Learning t-1 -0.0357 *
 [0.0164]

Experience_Learning_Squared t-1

Reflection_Learning t-1 0.0001
 [0.0004]

Reflection_Learning_Squared t-1

FwdCitat t-1 0.0021 * 0.0029 **
 [0.0010] [0.0011]

Firm_age t-1 0.0500 + 0.0472 +
 [0.0263] [0.0266]

Firm_Size t-1 -0.0305 -0.0543
 [0.0982] [0.1045]

R&D_Intensity t-1 0.0011 0.001
 [0.0008] [0.0008]

Patent_Age t 0.3463 *** 0.3213 ***
 [0.0395] [0.0406]

No_of_Claims t 0.0002 0.0000
 [0.0003] [0.0003]

Total_Patent t 0.0131 + 0.0359 **
 [0.0075] [0.0125]

No_Uni_Alliance t-1

No_Firm_Alliance t-1

Log-likelihood -768.3144 -762.8886

Variables Model-3 Model-4

Constant -2.5280 *** -2.6253 ***
 [0.5016] [0.5027]

Conceptualization_Learning t-1 0.0062 0.0057
 [0.0045] [0.0044]

Conceptualization_Learning_Squared t-1 -0.0000 0.0000
 [0.0000] [0.0000]

Experimentation_Learning t-1 0.0064 ** 0.0074 **
 [0.0025] [0.0026]

Experimentation_Learning_Squared t-1 -0.0001 * -0.0001 **
 [0.0000] [0.0000]

Experience_Learning t-1 -0.0651 ** -0.0608 **
 [0.0253] [0.0248]

Experience_Learning_Squared t-1 0.0005 * 0.0005 *
 [0.0003] [0.0003]

Reflection_Learning t-1 0.0011 * 0.0011 *
 [0.0006] [0.0005]

Reflection_Learning_Squared t-1 -0.0000 * -0.0000 **
 [0.0000] [0.0000]

FwdCitat t-1 0.0030 * 0.0028 *
 [0.0012] [0.0012]

Firm_age t-1 0.0512 + 0.0536 *
 [0.0271] [0.0269]

Firm_Size t-1 -0.0825 -0.0514
 [0.1079] [0.1099]

R&D_Intensity t-1 0.001 0.0011
 [0.0009] [0.0009]

Patent_Age t 0.3417 *** 0.3562 ***
 [0.0423] [0.0432]

No_of_Claims t -0.0000 -0.0001
 [0.0003] [0.0003]

Total_Patent t 0.0362 *** 0.0367 ***
 [0.0108] [0.0106]

No_Uni_Alliance t-1 -0.1477
 [0.1293]

No_Firm_Alliance t-1 -0.0601
 [0.0559]

Log-likelihood -759.4357 -758.2276

Variables Model-5

Constant -2.6653 ***
 [0.5080]

Conceptualization_Learning t-1

Conceptualization_Learning_Squared t-1

Experimentation_Learning t-1 0.0070 **
 [0.0026]

Experimentation_Learning_Squared t-1 -0.0001 **
 [0.0000]

Experience_Learning t-1 -0.0557 *
 [0.0248]

Experience_Learning_Squared t-1 0.0004 *
 [0.0003]

Reflection_Learning t-1 0.0011 *
 [0.0006]

Reflection_Learning_Squared t-1 -0.0000 *
 [0.0000]

FwdCitat t-1 0.0026 *
 [0.0012]

Firm_age t-1 0.0596 *
 [0.0265]

Firm_Size t-1 -0.0113
 [0.1056]

R&D_Intensity t-1 0.001
 [0.0009]

Patent_Age t 0.3636 ***
 [0.0432]

No_of_Claims t -0.0001
 [0.0003]

Total_Patent t 0.0391 ***
 [0.0110]

No_Uni_Alliance t-1 -0.1741
 [0.1289]

No_Firm_Alliance t-1 -0.0535
 [0.0566]

Log-likelihood -759.0583

+ p<0.1, * p<0.05, ** p<0.01, *** p<0.001. Standard error is
provided in the parentheses.

Technology class dummy variables were included.

Single tailed t-test was used for all hypothesized variables, two
tailed t-test was conducted for control variables.