How the Partner Asymmetries affect Innovation Quantity?

4.2.1 Descriptive Statistics

The sample size of this study is 506 dyad R&D alliances: 102 in 1981-1990, 339 in 1991-2000

and 65 in 2001-2010. Among 506 dyad alliances, 148 are AB type, 98 are BB type, and 260 are BP type. The mean of technological heterogeneity is 3.25; the mean of network resource asymmetry is 5.03; the mean of number of co-patents following the R&D alliances is 8.01. The descriptive statistics with means, standard deviations and correlations of variables regarding research in technology discrepancy are depicted in Table 4.

Table 4. Results of Correlation Analysis

+ p < .10; * p < .05; ** p < .01; *** p < .0001

4.2.2 Regression Results

Table 5 presents the results of negative binomial regression. The second model reports the effects of alliance types (contract or joint venture), prior cooperation experience, time dummy, partner types included and time to market as controls. This model served as a baseline from which the analysis proceeded. From model 3 to 6, we introduced technological heterogeneity to assess the possibility of its linear and nonlinear effects on innovation quantity. However, we did not find any significant relationships about them.

In model 7, we introduced network resource asymmetry to assess the possibility of its linear effects on innovation quantity, and we found a significant positive relationship between them. Then, we introduced network resource asymmetry and its squared term to assess the possibility of its nonlinear effects on innovation, and a significant downward curve correlation (inversed U-shape) between network resource asymmetry and number of co-patent following the alliance was observed (β1= 0.12, p<0.01; β2= -0.01, p<0.05) (Model 9 and Figure 9). The regression equation is as follows:

Number of co-patents = 1.46 + 0.45 (Alliance Type) – 0.16 (Prior Cooperation Experience) + 0.06 (Time of Contract [1991-2000 vs. 1981-1990]) – 0.26 (Time of Contract [2001-2010 vs. 1981-1990]) + 0.11 (Partner Type [BB vs. AB]) – 0.04 (Partner Type [BP vs. AB]) – 0.02 (Time to Market) + 0.12 (Network resource asymmetry) – 0.01 (Network resource asymmetry)² + e.

Even though the moderating effect of “Time to market” on the above linear relation between network resource asymmetry and innovation quantity was not observed (model 8), the results of subgroup analysis indicate that “Time to market” weakens the previous positive linear relationship when the alliances belong to contract rather than joint venture (β1= 0.17, p<0.05; β2= -0.02, p<0.1) (Model 11 and Figure 10) and when the alliances were made by a biotechnology firm and a

pharmaceutical firm (BP type) (β1= 0.09, p<0.1; β2= -0.01, p<0.1) (Model 12 and Figure 11).

Table 5. Results of Negative Binomial Regression (n=506)

+ p < .10; * p < .05; ** p < .01; *** p < .001

Figure 9. The relationship between network resource asymmetry and innovation quantity (full sample)

Figure 10. The moderating effect of time to market of product on the relationship between network resource asymmetry and innovation quantity (Contract)

Figure 11. The moderating effect of time to market of product on the relationship between network resource asymmetry and innovation quantity (BP)

Chapter 5: Discussion 5.1 Partner Asymmetries and Innovation Speed

The findings of this research revealed new mazes about partner asymmetries features acting on innovation. Our hypotheses were developed on the basis of organizational learning theory, social exchange theory and practical ratiocination. Overall, it appears that our hypotheses could be explained partly by established theories and practices. In this section, we analyze our empirical results, and discuss their implications.

The statistical results revealed a significant inverse U-shaped relationship between technological heterogeneity and innovation speed. These results support hypothesis H1. In other words, technological heterogeneity of alliance might be beneficial and detrimental for innovation speed, both insufficient or excessive technological heterogeneities result in slow innovation; only appropriate technological heterogeneity benefits innovative efficiency.

According to our argument for H1 from social exchange perspective, technological heterogeneity contributes the interaction of knowledge and technology between partners due to the generation of reciprocity, which accelerate the innovation. From organizational learning perspective, even though making alliances with high technological heterogeneity will create opportunities for a distant search, recombination inefficiency would be generated quite often due to the difficulties of integrating various types of technology, and it delays the innovation. On the other hand, making alliances with low technological heterogeneity will have more opportunity to do local search and to refine their quality of technology, which benefits the speed of innovation. For biopharmaceutical R&D alliances, our empirical study revealed that the middle level of technological heterogeneity comes together with appropriate search and learning will lead to faster innovative performance.

In biopharmaceutical cases, there is no direct influence of technological heterogeneity on innovation speed. Several companies made alliances with lower technological heterogeneity gained

sooner innovation than that with higher technological heterogeneity. Geron (a biotech company) is a typical example. It allied with Pharmacia (a pharmaceutical company) which had different technology in 1996, and then allied with Johns Hopkins University, which has similar technology, in 1997. Finally, the first alliance took more than two years to get the first co-patent, while the second alliance got the first co-patent within one. In contrast, several companies made alliances with higher technological heterogeneity gained sooner innovation than that with lower technological heterogeneity. ImClone System allied with the University of North Carolina which has similar technology in 1988, and then with Merck which has different technology in 1990.

Eventually, the previous alliance took 12 years to get the first co-patent, while the later alliance got its first co-patent after 8 years. Therefore, firms are not advised to seek partners with high or low technological heterogeneity in order to increase the speed of innovation. Instead, appropriate heterogeneity and other factors are important.

According to our statistical results, we did not find a significant inversed U-shaped non-linear relation between network resource asymmetry and innovation speed, so H3 was not supported.

However, a positive linear relationship between network resource asymmetry and innovation speed exists when the R&D alliances were created during 1991-2000, and when the partner type belongs to BB type. In other words, network resource asymmetry of alliance has an important influence on innovation speed (efficiency) under specific contingencies, and the higher network resource asymmetry, the higher innovation speed.

According to the organizational learning perspective, larger network resource asymmetry (“non-matched dyad”) is helpful for biopharmaceutical R&D alliance, because they combine high technology depth with scope. Our statistic results confirmed this ratiocination, because the interplay of depth and breadth within resource asymmetry “non-matched dyad” helps to increase the innovation speed. Although we also argued that too much network resource asymmetry might

harmful for innovation speed due to the perceived unfairness on social exchange perspectives, and proposed an inverse U-shaped nonlinear relationship between network resource asymmetry and innovation speed. However, in many biopharmaceutical R&D alliance cases, our results still indicate that “non-matched dyad” is better than “matched dyad” for innovation speed. Undeniably, both organizational learning and social exchange perspectives are useful explanations for overall conditions, since we did not find a significant positive linear relation between network resource asymmetry and innovation speed for all cases.

In biopharmaceutical cases, there is no direct influence of network resource asymmetry on innovation speed. Several companies formed alliances with larger network resource asymmetry gained innovation more quickly than that with smaller network resource asymmetry. A typical example is the biotech company NPS Pharmaceutical which with allied with the pharmaceutical company Pfizer which has 25.63 times as many alliances as it did in 1987, and then allied with Brigham and Women’s Hospital which has only 1.63 times as many of alliances as it did in 1993.

Finally, the previous alliance took 5 years to gain first co-patents later on, while the later alliance took 6 years to get it. In contrast, several companies made alliances with smaller network resource asymmetry gained innovation more quickly than that with larger network resource asymmetry. For example, the pharmaceutical company Depomed allied with the Bristol-Myers Squibb which had 10.08 times as many alliances as it did in 1996, and then allied with the biotech company Biovail which had 1.23 times as many alliances as it did in 2002. Consequently, the previous alliance took 8 years to gained first co-patents, while the later alliance took only 2 years. Hence, firms could make alliances with both “matched” and “non-matched” partners to benefit from faster innovation.

In H5, we proposed that an alliance’s product time to market negatively moderates the relationships between technological heterogeneity and innovation speed, and the relationship was supported. As figure 6, higher technological heterogeneity results in better innovation speed,

particularly for those later-stage products, because the slope of later product is larger than that of earlier product. Due to this, we know that the beneficial effects of technological heterogeneity on innovation speed become more obvious when the target product is closer to market. In other worlds, for those R&D alliances focusing on later-stage products, selecting partners with higher heterogeneous technology benefits their future innovation speed.

As for the moderating effect of time to market on the relationship between network resource asymmetry and innovation speed, the statistic results did not appear a significant effect so that H7 was not supported overall. Only for R&D alliances happened during 1991-2000 and for BB partner type alliances, the negative moderating effects exist. Both figure 7 and figure 8 show that network resource asymmetry is beneficial for the innovation speed of later-stage products but is detrimental for that of earlier-stage products under specific conditions (1991-2000 alliances and BB type alliances). Therefore, H7 was partial supported, which means that the network resource asymmetry is more demand for reaching innovation when the collaborating product is in the later stages. In other worlds, for those R&D alliances focusing on later-stage products, selecting partners with higher network resource asymmetry benefits their future innovation speed.

Further, we conducted the interaction analysis between technological heterogeneity and network resource asymmetry on innovation speed. Our finding shows significant counteracting effects on innovation speed. As Model 13 in table 3 shows, both heterogeneous technology and network resource asymmetry increase the speed of innovation, but when both are high, the previous benefits would be weakened or disappear.

5.2 Partner Asymmetries and Innovation Quantity

Our hypotheses were developed on the basis of organizational learning theory, social exchange theory and practical ratiocination. It appears that some but not all of the hypotheses could be partially explained by established theories and practical points of view. In this section, we analyze our empirical results and discuss their implications.

According to our statistical results, we did not find a significant linear or inversed U-shaped non-linear relation between technological heterogeneity and innovation quantity, so H2 was not supported. In other words, technological heterogeneity of alliance has no important influence on innovation quantity.

According to the argument for H2 from social exchange perspective, technological heterogeneity contributes the interaction of knowledge and technology between partners due to the generation of reciprocity, which increases the innovation quantity. From organizational learning perspective, even though making alliances with high technological heterogeneity will create opportunities for a distant search, recombination inefficiency would be generated quite often due to the difficulties of integrating various types of technology, and it reduces the innovation quantity.

However, above argument could only used to interpret the linkage between technological heterogeneity and innovation speed, rather than innovation quantity, which means technological heterogeneity makes innovation more efficiency, but it is not much helpful on the long-term innovation quantity. On the other hand, making alliances with higher technological heterogeneity will not have more opportunity to get a lot of innovation output. Undeniably, both social exchange perspectives and organizational learning are still useful explanations for overall conditions, since we did not find significant positive or negative linear relations between technological heterogeneity and innovation quantity for all cases.

According to the statistical results, we found a significant inverse U-shaped non-linear relation

between network resource asymmetry and quantity of co-patent following the alliance, which supports hypothesis 4. R&D alliances with moderate network resource asymmetry benefit more from innovation than do alliances with very low or very high levels of network resource asymmetry.

In other words, both too low or too high network resource asymmetries result in poor innovation.

The larger network resource asymmetry (“non-matched dyad”) is helpful for biopharmaceutical R&D alliance from organizational learning perspective, because alliances with large network resource asymmetry have combination of high technology depth and scope immediately. However, too much “non-matched dyad” is worse than “matched dyad” for innovation due to the perceived unfairness, based on the social exchange theory. Our statistic results partially confirmed this finding, because either asymmetry or non-asymmetry can increase the innovation.

Therefore, both organizational learning and social exchange perspectives are useful explanations.

This result is inconsistent with several previous studies of the asymmetry within alliances. For example, Veugelers and Kesteloot (1996) explored the asymmetry of size, R&D capability and production issues. They argued that the asymmetry between partners will influence the incentives to form a joint venture through their impact on the payoffs of own development. In addition, the larger the size of asymmetry, the larger (smaller) the big (small) firm's development profits. With lower asymmetries, profits in all scenarios are affected negatively (positively) for the big (small) firm.

In biopharmaceutical cases, several companies that formed alliances with larger network resource asymmetry gained more innovation than that with smaller network resource asymmetry. A typical example is the biotech company NPS Pharmaceutical. It allied the pharmaceutical company Pfizer which had 25.63 times as many alliances as in 1987, and then allied with Brigham and Women’s Hospital which has only 1.63 times as many alliances as in 1993. Finally, the previous alliance gained three co-patents later on, while the later alliance received only one. In contrast, several companies made alliances with smaller network resource asymmetry gained more

innovation than that with larger network resource asymmetry. For example, the pharmaceutical company Microcide allied with a pharmaceutical company, Ortho-McNeil, which had 3.43 times as many alliances as it did in 1995, and then allied with the pharmaceutical company, Pfizer, which had 29.29 times as many alliances as it did in 1996. Consequently, the previous alliance gained seven co-patents, while the later alliance got only two.

In Hypothesis 6, we proposed that time to market moderates the linear relationship between technological heterogeneity and quantity of innovative performance, however, this relationship was not observed. So Hypothesis 6 was not supported. In Hypothesis 8, we proposed that time to market moderates the linear relationships between network resource asymmetry and quantity of innovative performance, so that the correlations are affected by the clinical development stage of products.

Although there is no significant moderating effect of time to market overall, the results of subgroup analysis indicate that time to market weakens the positive relationship between network resource asymmetry and innovation quantity when the alliance was created by contract without financial investment. Figure 11, show that network resource asymmetry is beneficial for the innovation quantity of later-stage products (negative slope) but is a little bit detrimental for that of earlier-stage products (positive slope) under specific conditions (contract type alliances). Similarly, for BP type R&D alliances, the beneficial effects of network resource asymmetry on innovation quantity become more obvious (larger slope) when the target product is closer to market. In other worlds, for those R&D alliances focusing on later-stage products, selecting partners with higher network resource asymmetry benefits their future innovation quantity (figure 12). Therefore, H8 was partial supported.

Several researchers have explored the role of the stage of new product development on the outcomes in biopharmaceutical industry. Frahm et al. (2007) argued that the success of a new product depends on the stage of product discovery pipeline. They proposed that firms have to adapt

divergent strategies and behaviors on products in different stages. Some previous studies suggested that later stage biopharmaceutical products have higher potential for innovation and success than earlier stage products (Vanderbyl & Kobelak, 2006). However, our results further indicate that R&D alliances focusing on early products gain more innovation than late products; while the network resource asymmetry helps partners receive more co-patents for later products. In other words, the stage of new product development moderates the relationship between network resource asymmetry and quantity of innovative performance, especially for those products closer to market.

Chapter 6: Conclusion 6.1 Summary

Over the last two decades, biotechnology has changed the way in which large pharmaceutical firms obtain critical R&D capabilities through alliances with biotechnology firms. Due to the immature but complex nature of biotechnology, knowledge transfer in biotechnology R&D often entails technological uncertainty. Therefore, the exchange of knowledge in biotechnology requires stronger governance structures. With the development of bio-pharmaceutical technology, R&D alliances provide companies with another way to integrate resources, knowledge and technologies, create more research and business ideas, and facilitate innovation. Despite many studies on partner selection of R&D alliance, less research has been done on topics of technological heterogeneity and network resource asymmetry. This study explored the relationships among these factors and presents the empirical results. We have developed a research framework and developed hypotheses which were tested by quantitative analysis approach using secondary data. The results confirm that the effects of technological heterogeneity on innovation speed like an inverse U-shaped (non-linear) relationship, while the effects of network resource asymmetry on innovation speed is positive (linear) under specific conditions. Appropriate asymmetries of technology and network resource are suggested to be strategies for partner selection in order to get better innovation speed and quantity.

Moreover, considering various factors and conditions prior to making decision is helpful for innovation of R&D alliance.

6.2 Contributions

The dissertation contributes to the literature in many ways. First, based on organizational learning theory (economic perspective) and social exchange theory (social perspective) these two lenses, as well as practical logics, we designed an integrated framework to explain a complex phenomenon, presented our arguments, developed our hypotheses, and clarify the effects of technological heterogeneity and network resource asymmetry on innovation speed and quantity.

Second, the unit of analysis should emphasize the selection of an alliance partner. Dyadic approach is a better than a firm-specific one, since selecting optimal partner merely relies on the perspective from only one of partners is inadequate; while the dyadic approach for partner asymmetry study is more objective. Following this trend, we look at the technological heterogeneity and network resource asymmetry in dyadic approach instead of using one partner’s resource as a research object.

Third, in terms of the measurement of alliance performance, most previous studies rely on financial indicators, market value, and patents of distinct partners rather than on the performance of the alliance itself (e.g. Lunnan & Haugland, 2008; Gulati et al., 2009; Lin et al., 2009; Jiang et al, 2010). However, the measurement of individual performance might not reflect real outcomes of the alliance. The second study uses “innovation of target alliance” as the construct of performance of alliance in order to measure the outcomes more precisely. Further, in terms of innovation performance, we used not only innovation quantity (effect) in the second study, we introduced the concept of “time” by using innovation speed as performance of innovation. Innovation speed represents the efficiency of the innovation, which is a kind of dynamic performance.

Forth, we often see alliances composed of both biotechnology firms and pharmaceutical firms or of both universities and biotechnology firms. In fact, beyond inter-firm alliance, the R&D cooperation has been made by universities or academic institutions and bio-pharmaceutical firms.

Since this study covers academic institutions, biotechnology firms, and pharmaceutical firms, the finding will broaden the choices for diversified organizations in this industry. The results of this study contribute to the industry’s future decision making of partner selection.

Finally, this study discusses the effects from multiple perspectives, at the industrial (partner type), organizational (prior cooperation experience and timing of alliance), and product levels (the

Finally, this study discusses the effects from multiple perspectives, at the industrial (partner type), organizational (prior cooperation experience and timing of alliance), and product levels (the