In this section a number of recent empirical findings and of models of INs are going to be discussed. Most of the empirical evidence comes from the PhD thesis of Catherine (2007). In this thesis a database of about 7,600 collaborative agreements in biotechnology and covering the period 1973–1999 has been constructed. The INs studied here have three types of nodes – DBFs, LDFs, and PRIs. The agreements belong to two generations, depending on the technologies on which they are based. Agreements of the first generation are based on recombinant DNA and on monoclonal antibodies; agreements of the second generation are based on genomics. Furthermore, within each generation agreements are classified as either R&D- or market based. As it was previously pointed out, both of these generations are contained in what is often called third generation biotechnology. In other words, they are both based on the progress of molecular biology.
Figures 4.1–4.4 show the variation of the number of agreements during the period studied both for the whole set and for each generation. These results
provide a confirmation for the existence of a technology-based life cycle for INs.
For both generations R&D agreements rise first and then fall. R&D agreements dominate in the early phases of the life cycle and market-based agreements in the more mature ones. The life cycle of the first generation seems to be in decline and very few new R&D agreements are taking place within it. Agreements of the second generation start in the early 1990s and already show signs of saturation.
However, when the two generations are combined, the total number of agreements keeps rising for the whole period (Fig. 4.4).
73 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 0
10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230
Years
Number of Agreements
Fig. 4.1. Number of agreements in the first generation of biotechnology
R2 = 0,8739 R2 = 0,9149
0 10 20 30 40 50 60 70 80
73 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
Years
Number of Agreements
Marketing agreements R&D agreements
Fig. 4.2. R&D and marketing agreements in the first generation of biotechnology
R2 = 0,9874
R2 = 0,8866
0 50 100 150 200 250
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
Years
Number of Agreements
Marketing agreements R&D agreements
Fig. 4.3. R&D and marketing agreements in the second generation of biotechnology
R2 = 0,9832
0 100 200 300 400 500 600 700 800 900 1000
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
Years
Number of Agreements
Fig. 4.4. Total number of agreements in biotechnology, 1973–1999
0 5 10 15 20 25 30 35
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Years
Centralization index
R2 = 0, 9198
Fig. 4.5. Evolution of network density for the first biotechnology generation
0 2 4 6 8 10 12 14 16
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
Years
Centralization Index
R2 = 0, 8969
Fig. 4.6. Evolution of network density for the second generation of biotechnology
0 5 10 15 20 25 30
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
Years
Centralization Index
R2 = 0, 9277
Fig. 4.7. Evolution of network density for the two biotechnology generations combined Figures 4.5–4.7 show the change in the density of the networks for each generation (Figs. 4.5 and 4.6) and for the two generations combined (Fig. 4.7). In all three cases the density falls first and then rises. However, we can notice that the period of fall is much shorter in the second than in the first generation. Our findings here contrast with those of Rojikanners and Hagedoorn (2006), who found that network density has always been rising in the period since 1975, but confirm those of Orsenigo et al. (2001). We cannot immediately our results because the other measures were obtained by means of different techniques. However, we are able to provide an interpretation of the reasons for which density can be expected to behave in the way we observe. This behaviour of network density can be explained by our previous analysis of scale-free networks, according to which density was expected to fall during periods of environmental discontinuity and to rise during the subsequent period of environmental stabilization. Of course, in this context the environment is the socio-economic one and not the purely physical or biological. In the case of biotechnology INs the network dimension which gave rise to the discontinuity is knowledge, and in particular the emergence of molecular biology and of its industrial applications. We can consider molecular biology as a new paradigm, forcing firms in the biotechnology-based sectors (pharmaceuticals, agrochemicals etc) to radically update their knowledge base (Grabowski and Vernon 1994; Saviotti et al. 2003, 2005; Nesta and Saviotti 2005, 2006). According to this view, a rising density should imply that the new paradigm is being absorbed by incumbent firms and that the science/entrepreneurial phase is being followed by a more market-based one. We can notice that the period during which density falls is shorter for the second than for the first generation of biotechnology INs. This seems to imply that the second generation was less of a discontinuity than the first, not an illogical interpretation given that the second generation is based on genomics, which can be
considered a refinement of molecular biology, of which it uses the basic conceptual apparatus. Extending our interpretation to other sectors, we can expect density to rise for more mature ones such as automobiles or packing machinery, thus confirming the results of Dyer and Nobeoka (2000) and of Lorenzoni and Lipparini (1999). If our interpretation is correct, network density can be a very important variable, capable of indicating the extent of knowledge discontinuity arising at given times during the evolution of industrial activities, for example, when new technological paradigms emerge. However, such a phenomenon clearly requires further study.
Another important property of INs is the centrality of the different types of actors involved. Tables 4.1 and 4.2 show the evolution of centrality for DBFs, PRIs and LDFs during the period studied. We can see a clear pattern emerging in both generations of INs, with DBFs being the most central actor in the early phases and LDFs becoming the most central actor in the late phases. PRIs have initially a high degree centrality but become very marginal as the technology moves towards maturity. The fall of DBFs and PRIs centrality becomes even more evident when betweenness centrality is used.
Table 4.1. Centrality of the different actors (DBFs, LDFs, PRIs) involved in the first generation of biotechnology alliances
1976–1984 1985–1992 1993–1999 First generation
biotechnology DBFs LDFs PRIs DBFs LDFs PRIs DBFs LDFs PRIs Average number of
agreement 5.67 3.20 2.63 6.20 10.79 2.61 6.12 15.50 2.05 Average N-degree
centrality 5.84 3.30 2.71 1.71 2.97 0.72 1.41 3.58 0.47 Median N-degree
centrality 3.09 2.06 2.06 1.10 1.38 0.55 0.92 1.73 0.23 Average betweenness
centrality 4.95 2.82 0.85 0.71 1.38 0.20 0.40 1.40 0.10 Median betweenness
centrality 1.17 0 0 0.20 0.53 0 0.07 0.18 0
Table 4.2. Centrality of the different actors (DBFs, LDFs, PRIs) involved in the second generation of biotechnology alliances
1985–1992 1993–1999 Second generation
biotechnology DBFs LDFs PRIs DBFs LDFs PRIs
Average number of
agreement 5.98 5.90 5.48 12.50 25.13 9.27
Average N-degree
centrality 2.48 2.45 2.27 1.68 3.38 1.25
Median N-degree
centrality 2.08 1.66 1.66 1.21 1.55 0.81
Average betweenness
centrality 2.70 2.21 2.69 0.41 0.97 0.13
Median betweenness
centrality 1.91 1.36 2.14 0.18 0.15 0.07
In order to interpret adequately these results it is important to establish clearly the meaning of centrality. In general we can expect a more central actor to be more important in a network. The gradual shift of average N-degree centrality from DBFs and PRIs towards LDFs as a technology ages would seem to confirm this interpretation. We expect that as a technology matures exploitation will become more important than exploration. Since LDFs are the only actor in these networks having the resources required for exploitation, it seems logical that they become more central as the technology matures. Likewise, since we expect both DBFs and PRIs to have a competitive advantage in the early phases of a technology life cycle, it is logical for their centrality to be higher in these phases and to fall as the technology moves towards maturity. Incidentally, these results indicate that although recent trends in knowledge production seem to make boundaries of institutions fuzzy and their functions overlap, DBFs, PRIs and LDFs still conserve different roles. However, in spite of this apparent support for the interpretation of centrality as indicator of the relative importance of an organization in its economic environment, this interpretation is limited. For example, it is not clear that in spite of their growing centrality pharmaceutical DBFs are in a good position. Although they have substantially increased their R&D expenditures and participated in INs, large pharmaceutical firms cannot find enough new molecules to sustain their past strategy of identifying blockbusters (Hopkins et al. 2007). Apparently the new biotechnology has introduced radically new concepts but does not seem to be able to provide a lease of life for the blockbuster strategy. Perhaps the strategy itself is doomed and biotechnology is leading to a new strategy for the pharmaceutical industry.
Further doubts about the meaning of centrality arise if we take into account the results of the different measures used, such as N-degree or betweenness centrality.
Although all the measures tend to confirm the earlier results, they give absolute and relative values of centrality. Clearly although centrality seems to be an important variable to map the dynamics of INs, further research is required in order to clarify its meaning.
By combining centrality and density measures we can see that periods of falling density tend to coincide with periods in which DBFs and PRIs have a relatively high centrality while periods of growing density tend to coincide with a relatively high centrality of DBFs. This finding seems logical given that both falling network density and a relatively high centrality of DBFs and PRIs are expected to occur immediately after a knowledge discontinuity while a high centrality of LDFs is more likely to occur when a new technology begins to mature.
Some recent models of INs shed light on the impact of knowledge generation and utilization on network dynamics. Gilbert et al. (2001) have developed an agent-based model of INs. Although their model is not specific to biotechnology, it provides several interesting insights about the dynamics of INs. In the model, firms innovate by changing their Kene, an expression which is used to indicate the individual agent knowledge base. The Kene comprises a set of units of knowledge. Each unit is defined by three parameters which represent the scientific, technological or business domain, the ability to perform a given application in this domain and the expertise level with which such applications can
be performed. Firms innovate by means of innovation hypotheses, which are derived from a subset of the existing kene. The kene can be transformed into a product by means of a standard mapping function which combines the firm’s existing knowledge with an innovation hypothesis. Firms try to improve their overall competitiveness by innovating. They do so by improving their knowledge base through adaptation to user needs, by incremental or radical learning and by cooperation or networking. When they form partnerships, firms can adopt a conservative strategy by choosing a partner having similar capabilities, or a progressive strategy by choosing a partner with a different capability set. In this model networks are ‘normal’ agents. The network can create innovations in addition to those of its members and it can distribute rewards to the members. The profits for each member will be the sum of individual and of network profits, which would explain the advantage of being in a network as opposed to proceeding alone. Simulations of this model show that INs are a viable form of industrial organization. The model predicts that both the number of actors and the number of networks will increase in the course of time. The model also leads to predictions about the relative merits of progressive and conservative strategies, about the expected network connectivity and about the evolution of industrial concentration.
In a recent paper Pyka and Saviotti (2005) developed a model of INs in biotechnology. The model includes two types of actors, DBFs and LDFs. The former have technological competencies and the latter only so-called economic competencies. Of course, LDFs do have technological competencies, but they are obsolete. The only valuable competencies they are left with initially are other competencies required to produce final products, such as financial, marketing etc.
They are in fact complementary assets (Teece 1986). The model starts with the objective of comparing the two strategies of collaboration and of going it alone.
Given the starting conditions it predicts that collaboration will be more effective.
However, beyond this simple prediction the model gives interesting insights about network dynamics. For example, it predicts that collaboration and the formation of INs will continue in the long run even if by collaborating LDFs learn the new technological competencies. This can occur because of the changing role of DBFs, which shifts from that of translators of the new knowledge for LDFs to that of explorers of a knowledge space which LDFs can in principle know, but which expands at too fast a pace. Also, the model predicts correctly that while in the initial phases the collaboration will be mainly between DBFs and LDFs, in subsequent phases collaboration between different DBFs will become frequent.
The model is oversimplified and it does not contain subsequent generations of biotechnology. Thus, it misses an important source of continued network formation. However, an alternative mechanism of continued network formation is here provided by the extremely high rate of growth and by the increasing complexity of new knowledge, which makes it impossible for any LDFs to explore it all, even when it was in principle capable of doing it. Shifting to the role of explorers allows DBFs to keep entering and giving rise to new INs. In other words, INs could be an advantageous form of industrial organization both due to knowledge discontinuities, that is to the qualitative difference between old and
new knowledge, and to the rate at which new knowledge is created, which could exceed the capabilities of LDFs. At this point the relative weight of DBFs entry and of the succession of two generations of biotechnology constitutes an interesting question which requires further work.