Atp.nist.gov/Eao/Ir-7323/Refer.htm the Determinants of Success

¶ … atp.nist.gov/eao/ir-7323/Refer.htm

the Determinants of Success in R&D Alliances

Innovation is an increasingly important dimension of competition in technology intensive industries. In seeking innovation, individual firms often find that external knowledge and research partners are critical to success. Innovation is often the result of synthesizing or "bridging" ideas from different knowledge domains (Hargadon and Sutton, 2000; Burt, 2004). Therefore, firms increasingly enter into research and development (R&D) alliances with other firms to combine complementary knowledge in the pursuit of new innovative technologies. Indeed, many governments around the world support cooperative research activities in the expectation that collaborating firms will successfully develop new technologies that will improve economic competitiveness.

Unfortunately, while R&D alliances have become a popular mechanism to pursue innovation, prior research suggests that alliances have failure rates of around 50% (Kogut, 1989; Alliance Analyst, 1998; Kale et al., 2002). R&D alliances are presumably even more challenging than other types of alliances because collaborators must simultaneously share knowledge while trying to prevent undesired knowledge spillovers (Hamel, 1991; Oxley and Sampson, 2004). The free exchange of knowledge by partners is critical for ideas and knowledge to be recombined in such a way as to produce innovations (Hargadon and Sutton, 2000). However, various factors -- attributes of partners, governance arrangements, communication processes -- may inhibit the exchange of complementary knowledge in an alliance, thereby decreasing the probability of innovation. Moreover, even when R&D alliance partners are able to simulate the free flow of knowledge that can occur within a firm, the innovation process itself is characterized by a high degree of uncertainty, which makes success extremely difficult to predict.

Understanding how firms can enhance the probability of success in R&D alliances is an important question for both firms and governments. Researching the determinants of knowledge sharing and innovative success in R&D alliances is especially challenging since innovation processes are inherently uncertain. Despite these challenges, numerous scholars have examined a variety of factors that may influence the performance outcomes of R&D alliances.

This study seeks to provide a more complete understanding of the factors that influence success in R&D alliances. The perspective adopted in this study is that success in an R&D alliance is more likely to occur when the firms initiating the alliance: (a) partner with other firms that possess relevant complementary knowledge, and (b) effectively share and combine that complementary knowledge. Having the requisite knowledge within the alliance team and having the necessary processes in place to exchange that knowledge is critical to producing technical innovations. Effective "knowledge management" is therefore a fundamental driver of success in R&D alliances. Alliance success depends broadly on three categories of factors that influence the exchange of complementary knowledge: alliance "design" decisions made during the alliance formation stage; alliance "management" decisions made during the alliance execution stage; and "luck," that is, the playing out of random events under uncertainty. During the alliance formation stage, the designers of the alliance seek to identify the "win-win" opportunity for potential alliance members and recruit potential members based on a cost-benefit analysis of what each potential member might contribute to the alliance objective. In this cost-benefit analysis, if alliance designers properly optimize their decision, then each member's contribution to the alliance is balanced against the burden that it imposes on the alliance. As such, in an ex-post analysis, the analyst would not expect to see differential results in actual performance outcomes related to alliance design characteristics.

In real life, however, not all ex-ante factors may have been fully optimized. Not all alliance design decisions and alliance management decisions results are necessarily fully "optimal" in an equilibrium sense. For example, if ex-ante design decisions on the number of firms to include in an alliance, or whether to include competitor firms, were optimal, then we would not expect to see these alliance characteristics to be correlated with alliance outcomes in an ex-post analysis. But such alliance design and management decisions are in fact made by actors in a context of imperfect information, uncertainty, and learning, so we can expect that these decisions are less than perfect, and therefore, we assess whether some decisions turn out to be "less optimal" than others in an ex-post analysis.

To examine how alliance-design and alliance-management factors influence R&D alliance success, we use a unique survey dataset that includes 397 firms in 142 R&D alliances. These R&D alliances received funding from the Advanced Technology Program, a U.S. federal government program that supports innovation and early-stage technology in U.S. industry. The data were collected by the Advanced Technology Program.

In our analysis of the determinants of R&D alliance success, we go beyond prior studies in several ways. First, we focus exclusively on R&D alliances. Many prior studies include all types of alliances (e.g., marketing, manufacturing, R&D, etc.) rather than focusing specifically on R&D alliances. R&D alliances are different from other types of alliances, especially in their focus on knowledge sharing and innovation, so analyses that pool different types of alliances are difficult to interpret for R&D alliances. Second, we develop our analysis by drawing upon multiple theoretical perspectives from economics, organization theory, and strategic management. Prior studies of R&D alliances focus on a narrow set of factors that may influence R&D alliance success, typically relying on a particular theoretical lens rather than drawing on a broad set of theoretical perspectives. While using a single theoretical perspective has the advantage of allowing for deeper theoretical insight, it has the disadvantage of excluding many factors that may be important for empirical understanding. Third, we employ multiple measures of alliance success. Most prior studies have inadequate measures of R&D alliance success. Firm-level measures such as firm profitability or stock price are only remotely related to R&D alliance performance outcomes. Alliance survival is a poor measure of R&D alliance success since most R&D alliances are designed to last for a limited time period. Survey-based perceptual measures capture the degree to which an alliance has achieved broad and diverse goals, but may also be subject to a variety of response biases. Weaknesses associated with each type of performance measure suggest that a study of performance outcomes of R&D alliances would ideally include a combination of objective and subjective measures. We utilize multiple measures of performance outcomes at the firm-level, including a perceptual measure (subjective assessment of overall value to the firm), a patent measure (patent applications filed by the firm), and a financial measure (revenues or cost savings realized by the firm from commercialization of technology). Finally, we employ firm-specific measures of alliance outcomes from multiple members of an alliance. Most prior studies do not have alliance-wide measures of performance outcomes, that is, they do not have data from different members in the alliance. Different firms in an alliance have different objectives, and benefits to each firm may vary depending on a variety of factors. In order to better understand alliance "success" and factors that influence success, we use data from more than one alliance partner for 121 out of the 142 R&D alliances represented in the sample. We find substantial variance within an alliance in terms of individual partner firms' assessment of the success of the alliance. For example, in 16% of the alliances, one alliance partner rated the alliance as "successful" or "very successful" in delivering value to the firm, while another alliance partner rated the alliance as "unsuccessful" or "very unsuccessful" in generating value. Although there is positive correlation in performance outcomes among alliance partners, our analysis indicates that alliance success is an individual firm-level phenomenon, so data gathered from only one partner cannot generalize to the "alliance" level.

We examine alliance design factors that are expected to influence alliance success. We consider alliance structure characteristics such as the number of partners, type of partners (e.g., presence of competitors), and geographic proximity of partners. We also consider firm-level attributes such as the firm's prior experience with alliances in general or with specific alliance partners, and the firm's existing stock of R&D knowledge and capabilities. These alliance design factors (alliance structure characteristics and firm-level attributes) are largely established at the time of alliance formation, and reflect the decisions made by the alliance designers.

We also examine alliance management factors that are expected to influence alliance success. Alliances that are able to establish effective governance arrangements and institute processes that build trust are expected to be more likely to share knowledge and achieve innovation success. Alliances that facilitate communication among partners effectively are also expected to be more likely to achieve innovation success. Alliance partner commitment and effort devoted to the alliance project, as measured by technical personnel resources allocated, is also expected to relate to alliance success. These alliance management factors develop during the course of the project, that is, in the process of alliance execution.

In summary, we examine the relative importance of alliance design factors and alliance management factors in determining R&D alliance success. We also examine whether partners in an R&D alliance realize similar, or dissimilar, benefits from participation in the alliance. Finally, applying insights from our analyses, we explore what both firms and governments might do to increase the likelihood of success in R&D alliances.

Return to Table of Contents or go to next section of interim report.

Date created: August 29, 2006

Last updated: September 11, 2006

NISTIR 7323 - the Determinants of Success in R&D Alliances

Part 2 - Theoretical Perspectives on R&D Alliance Success

The theoretical perspectives underpinning hypotheses tested in this study are derived from both our review of prior literature on R&D alliance success, and from our exploratory interviews with participants in R&D alliances that received funding support from the Advanced Technology Program (ATP) at the National Institute of Standards and Technology (NIST). We conducted semi-structured interviews that focused on the question: What are the factors that contribute to, or inhibit, alliance success? The participants in these interviews consistently identified factors that related to knowledge sharing in the alliance, which in turn affected success in achieving technical objectives, generating research outcomes, and commercialization of technology (Dyer and Powell, 2001). Drawing on prior research and on these interviews, we developed the theoretical logic and hypotheses presented in the following sections.

Alliance Design Factors: Alliance Structure

Number of Alliance Partners

In alliance design, firms that initiate an alliance aim to optimize the number of partners to involve in the alliance. Additional partners may bring additional knowledge and resources to the alliance, but each additional partner also brings additional transaction and coordination costs (Gulati and Singh, 1998). Each additional partner firm must be included in negotiations regarding the goals of the collaboration, protection of intellectual property, control and ownership of research output, how to share knowledge and collaborate in R&D, etc. Adding more partners to an alliance may also hinder knowledge sharing by increasing the risk of unintended knowledge leakage (Oxley and Sampson, 2004). The greater the number of partners in an alliance, the more reluctant individual firms may be to share knowledge, fearing greater potential for unintended knowledge spillovers when more firms have access to the knowledge.

With each additional partner, the number of alliance partners increases linearly as N, but the number of dyadic relationships increases quadratically as N (N-1)/2. With two firms there is one relationship to manage; with three firms there are three relationships; with four firms, six relationships, and so on. In our interviews with R&D alliance participants, many observed that knowledge sharing and coordination was more difficult with more members. As one participant stated, "The more people you have, the more people you have to coordinate. It gets unwieldy at some point." (Dyer and Powell, 2001, p.14). Above some threshold number of partners, transaction and coordination costs become significant, and concerns about knowledge leakage inhibit the ability of alliance partners to share knowledge, which is critical to R&D alliance success. In the cost-benefit calculation for deciding how many partners to include in an alliance, if alliance designers err in optimizing the number of partners to involve, we expect that the tendency is to underestimate transaction and coordination costs. Hence, we expect that alliances with a greater number of partners will have lower performance outcomes.

Hypothesis 1: The greater the number of R&D alliance partners, the lower the performance outcomes of the R&D alliance.

Presence of Competitors

Prior research suggests that R&D alliances are fraught with risks because firms must simultaneously share knowledge and technology, as well as protect knowledge (Hamel, 1991; Oxley and Sampson, 2004). Firms must find the right balance between maintaining open knowledge exchange to further the technological goals of the alliance while also preventing unintended leakage of knowledge. Preventing opportunism within R&D alliances is a prime concern, and especially challenging for a number of reasons.

First, joint R&D often requires high levels of investment in complementary assets or knowledge by the participants. When a firm performs part of a research project, the knowledge it gains may be useless unless combined with the work of partner firms with complementary knowledge. This creates potential for opportunistic behavior on the part of partner firms that possess the complementary assets or knowledge. In effect, these knowledge assets are "transaction-specific" assets, and in this transaction relationship there is significant potential for opportunism (Klein et al., 1978; Williamson, 1985).

Second, R&D alliances are characterized by a high degree of uncertainty regarding both inputs and outputs. Monitoring inputs and "effort" on the part of one's partner is extremely difficult. R&D alliance tasks are largely intellectual in nature and, therefore, third party monitoring is inefficient. Under these conditions, effective self-monitoring (Demsetz, 1988) is required because it is impossible to really know whether an alliance partner is truly sharing its most relevant knowledge. In short, the high degree of uncertainty regarding inputs and outputs provides numerous opportunities for opportunistic behavior on the part of alliance partners.

Finally, there are significant information asymmetries among partners in R&D collaborations. Each firm brings different knowledge to the table and may be reluctant to share information due to the desire to prevent unintended knowledge spillovers. Once technological information is revealed, the receiver of knowledge has no incentive to pay for the information. Thus, a primary challenge in R&D alliances is to figure out how to openly share knowledge that is relevant to the alliance objectives while preventing undesirable knowledge spillovers.

These challenges to knowledge-sharing in R&D alliances are exacerbated in the case of competitor collaborations where partners are ultimately engaged in a zero-sum game in the marketplace (one partner's commercial success ultimately has a negative impact on the commercial success of another partner). For example, the decision regarding the extent to which a partner should fully collaborate (sending the most high caliber researchers, sharing proprietary knowledge, etc.) may be characterized by a Prisoner's Dilemma game, where despite the fact that the two firms would be better off by jointly cooperating, both firms individually have the incentive not to share skills and information. Hamel's (1991) detailed examination of nine alliances revealed that firms typically try to internalize their partner's skills while protecting their own. As one manager in his study observed, "[Our partner] tries to suck us dry of technology ideas they can use in their own products. Whatever they learn from us, they'll use against us worldwide." (Hamel, 1991, p.87). In our interviews, an R&D alliance manager stated, "Having direct competitors in the [alliance] definitely inhibited information sharing. I don't take guys to the [alliance] meetings if they talk too much; sometimes I have to say to them 'That's enough. You are talking too much.'" (Dyer and Powell 2001, p.13).

Even if the outcome of research effort is not characterized as a zero-sum game, R&D alliances with direct competitors could lead to a collective reduction of research efforts. Katz (1986) demonstrates the possibility that when firms cooperate in cost-reducing R&D, but compete in product markets, the firms might collaborate to conduct less R&D to lessen the severity of competition in product markets. Branstetter and Sakakibara (2002) empirically examine Japanese government-sponsored R&D consortia and find that the research productivity of participating firms is lower when the degree of product market competition among participants is higher.

Hypothesis 2: If an R&D alliance involves firms that are direct competitors in product markets, then the performance outcomes of the R&D alliance are lower.

Geographic Distance between Alliance Partners

Prior research shows that geographic proximity plays an important role in facilitating interaction and knowledge-sharing between collaborating firms (Saxenian, 1994; Dyer, 1996; Almeida and Kogut, 1999). For example, Dyer (1996) finds a strong relationship between geographic proximity of automaker and supplier facilities, and the extent to which the firms engage in face-to-face interaction. He also finds that greater face-to-face interaction between supplier-customer engineers leads to fewer defects and higher overall product quality. Geographic distance presumably increases the cost of frequent face-to-face communication, thereby decreasing knowledge sharing and reducing coordination effectiveness (especially when tasks are highly interdependent).

Hypothesis 3: The greater the geographic distance between R&D alliance partners, the lower the performance outcomes of the R&D alliance.

Alliance Design Factors: Firm Attributes

General and Partner-Specific Alliance Experience

Prior research generally suggests that firms with greater partnering experience develop "relational capabilities" that enhance their ability to extract value from subsequent alliances (Anand and Khanna, 2000; Kale et al., 2002). Partner experience can be either "general alliance experience" or "partner-specific alliance experience." Whereas the former refers to experience gained from all prior alliances, the latter refers to prior alliance experience with a specific partner. Firms that engage repeatedly in an activity are able to draw inferences from their experiences, and store and retrieve such inferred learning for use in subsequent engagements in the activity (Levitt and March, 1988). In the alliance context, firms with substantial experience in alliances often have dedicated personnel charged with capturing, codifying, and communicating best practices in managing alliances. Similarly, when firms have repeated alliances with specific partners, the partnering firms may be induced to invest in relation-specific assets that reduce transaction and coordination costs. Moreover, learning accumulated through partner-specific experience may lead to the emergence of stable and efficient inter-organizational knowledge-sharing routines (Dyer and Singh, 1998; Zollo et al., 2002).

Most studies have found results consistent with the expectation that general and partner-specific experience lead to superior alliance performance. However, a recent study by Hoang and Rothaermel (2005) calls into question the relationship between partner-specific experience and R&D alliance performance. In their study of R&D alliances for new drug development, they find that general partnering experience has a positive effect on performance for small biotechnology firms, but not for their larger pharmaceutical firm partners. They suggest that larger firms have already acquired significant experience and capabilities at alliances, so there is little difference in capabilities of these firms. Interestingly, contrary to their expectations, they find that partner-specific experience has a weakly negative effect on successful drug development. Hoang and Rothaermel suggest that the reason for this counterintuitive finding may be that partners inappropriately generalize from their prior experience with that partner -- but the next drug development project is not like the last one. The logic for this conclusion is similar to that in Haleblian and Finkelstein's (1999) study of acquisition and performance, which concludes that firms with a moderate amount of experience in acquisitions may be less successful in subsequent acquisitions if the new acquisition is significantly different in nature from prior acquisitions. Another plausible reason for this result may be that for firms pursuing innovation objectives in alliances, repeated transactions with the same partner -- while promoting efficiency -- may not result in novelty. Thus, previous findings of a positive relationship between partner-specific experience and alliance performance in general may not hold for R&D alliances in particular. Our study offers the opportunity to examine the effects of both general and partner-specific experience on R&D performance, using multiple performance measures.

Hypothesis 4a: The greater the general alliance experience of the firm, the better the performance outcomes of the R&D alliance for the firm.

Hypothesis 4b: The greater the partner-specific experience of the firm with its alliance partners, the better the performance outcomes of the R&D alliance for the firm.

R&D Capability

Generally, we expect that the overall R&D capability and total stock of knowledge of an R&D organization would have a positive impact on performance outcomes of an R&D alliance that the organization participates in. A proxy measure for an organization's overall R&D capability and stock of knowledge is the total number of R&D personnel at the organization. To realize benefits from participating in alliances, firms must identify, assimilate, and commercialize useful knowledge developed through collaboration. A firm's ability to take advantage of externally generated knowledge -- its absorptive capacity -- depends upon the stock of related knowledge accumulated by the firm (Cohen and Levinthal, 1989; 1990). Relatively few empirical studies have examined the impact of absorptive capacity on the benefits that firms obtain from participation in alliances. Mowery et al. (1996) find that pre-alliance technological overlap with alliance partners enhances a firm's absorption of technological capabilities, but a firm's R&D intensity has no effect. We also expect that economies of scale and scope in R&D characterize the R&D capability of a firm. We expect that "knowledge spillovers" between personnel within a firm enhance the R&D capability of the firm. Within a firm, R&D personnel working on any given project are able to learn from other technical personnel at the firm working on different projects. We hypothesize that firms with a greater R&D capability and stock of knowledge, as proxied by total R&D employment, are able to benefit more from R&D alliances, and therefore have better alliance performance outcomes.

Hypothesis 5: The greater the total R&D employment of the firm, the better the performance outcomes of the R&D alliance for the firm.

Alliance Management Factors

Number of Technical Personnel

A key management decision for a firm participating in an R&D alliance is to determine the number of technical personnel to allocate to the effort. The level of technical personnel resources that a firm devotes to an alliance project affects both the direct R&D output and the R&D learning benefits that the firm can expect to receive from participating in the alliance. For the firm, the allocation of R&D personnel is firstly related to direct innovation outputs as a result of the effort, and secondly related to R&D absorptive capacity and R&D learning (Cohen and Levinthal, 1989).

In regard to absorptive capacity, Lane and Lubatkin (1998) compare firm-level and firm dyad-specific measures of absorptive capacity and find that the latter is better in explaining learning outcomes from alliances. Thus, prior research indicates that alliance-specific measures of absorptive capacity, i.e. pre-alliance technological overlap (Mowery et al., 1996), or similarity of knowledge stock and management practices (Lane and Lubatkin, 1998), accounts for alliance benefits better than traditional firm-level measures of absorptive capacity, i.e., R&D spending or R&D intensity.

Prior research has emphasized that gatekeeping or boundary-spanning roles are important for absorbing external knowledge (Cohen and Levinthal, 1990). The more people that a firm positions at the "gate," the more receptive the firms become to external or alliance knowledge. By involving more R&D personnel in an alliance collaboration, a firm is able to transfer more individual and interpersonal knowledge, and is able to widen the conduit for knowledge flows from the R&D alliance to itself. Agrawal (2006) examines the impact of total amount of time that professors, graduate students, and research scientists work in collaboration or close communication with firms that license university inventions, and finds that as total collaboration time increases, both the likelihood of commercialization and the degree of commercialization success increases. Thus, we expect that firms that allocate more R&D personnel resources to an R&D alliance effort will receive greater benefits from the alliance, both in innovation outcomes and learning outcomes.

Hypothesis 6: The greater the number of technical personnel allocated to an R&D alliance by the firm, the better the performance outcomes of the R&D alliance for the firm.

Frequency of Communication

Prior studies suggest that when firms collaborate on complex problems, they are more likely to be successful if they develop processes that facilitate frequent communication (Clark and Fuji-moto, 1991; Dyer, 1996). Frequent communication results in greater knowledge-sharing between alliance partners, which increases the likelihood of success in collaborative efforts. Mohr and Nevin (1990) define communication as the process by which partner firms in an alliance transmit information, coordinate activities, prompt participatory decision-making, and encourage commitment and loyalty to the alliance. Some research suggests that partners that develop relation-specific know-how through frequent communication are less likely to misunderstand or misinterpret information (Nishiguchi, 1994; Clark and Fujimoto, 1991). More efficient communication and coordination should result in better performance. With greater complexity of collaborative tasks, direct face-to-face communication and work interaction is believed to be of greater importance relative to other forms of communication, such as email, telephone, or even video conference. Face-to-face interaction is described as having high knowledge carrying capacity because it presents immediate feedback opportunities and makes use of both visual and audio modes of communication (Daft and Lengl, 1986; Dyer, 1996).

Interestingly, the relationship between communication and performance has rarely been empirically tested in R&D alliances, and, based upon our review of the R&D alliance literature, the relationship between face-to-face interaction and performance has never been tested. Our hypothesis is that more frequent communication and interaction will result in greater knowledge sharing and better performance outcomes in R&D alliances.