Choosing an appropriate alliance governance mode: The role of institutional, cultural and geographical distance in international research & development (R&D) collaborations

Abstract

We identify a variety of R&D alliance modes in a knowledge-intensive industry (e.g., Pharmaceuticals), and classify them into four ordered categories which go beyond the traditional binary equity vs non-equity alliance classification. This enriches our understanding of alliance governance structures and broadens the application of alliance modes in what is today a more complicated international R&D collaboration setting. We then explore national, industry and firm factors that determine the selection of an appropriate R&D alliance governance mode, using a sample of 237 international alliance deals. The likelihood of using a more-integrated alliance governance mode decreases as the difference or “distance” between nations of the partner firms increases in terms of human capital and cultural distance. On the other hand, a greater geographic and institutional difference is positively associated with the selection of more integrated alliance governance modes. Furthermore, firms in the research stage are more likely to use a more-integrated governance mode, as opposed to firms in the development stage. These findings advance research on alliance governance structure. They reveal the factors affecting the R&D alliance governance mode choice.

Abstract

Nous identifions une variété de formes d'alliances de R&D dans un secteur à forte intensité de connaissances (par exemple, le secteur pharmaceutique), et nous les classons en quatre catégories ordonnées qui vont au-delà de la traditionnelle classification binaire : alliances avec engagement capitalistique vs alliance sans engagement capitalistique. Cela enrichit notre compréhension des structures de gouvernance des alliances et élargit l'application des formes d’alliance dans ce qui est aujourd'hui un contexte plus complexe de collaboration internationale en R&D. Nous étudions ensuite les facteurs nationaux, industriels et relatifs à l’entreprise qui déterminent la sélection d'un mode de gouvernance approprié d'une alliance en R&D en utilisant un échantillon de 237 opérations d’alliances internationales. La probabilité d'utiliser un mode de gouvernance plus intégré pour une alliance diminue à mesure que la différence ou la "distance" entre les nations des entreprises partenaires augmente en termes de capital humain et de distance culturelle. D'autre part, une plus grande différence géographique et institutionnelle est positivement associée à la sélection de modes de gouvernance plus intégrés pour les alliances. En outre, les entreprises au stade de la recherche sont plus susceptibles d'utiliser un mode de gouvernance plus intégré que les entreprises au stade du développement. Ces résultats font progresser la recherche sur la structure de gouvernance des alliances. Ils révèlent les facteurs influant le choix du mode de gouvernance des alliances en R&D.

Abstract

Identificamos una variedad de modos de alianzas I+D en una industria intensiva en conocimiento (ej. productos farmacéuticos), y los clasificamos en cuatro categorías ordenadas que van más allá de la clasificación binaria de alianzas de valores participativos versus valores no participativos. Esto enriquece nuestro entendimiento de las estructuras de gobernanza de las alianzas y amplía la aplicación de los modos de alianza en lo que es hoy un escenario más complicado de colaboración internacional de I+D. Después exploramos los factores nacionales, de industria y de empresa que determinan la selección del modo de gobernanza de una alianza I+D usando una muestra de 237 ofertas de alianzas internacionales. La probabilidad de usar un modo de alianza de gobernanza más integrada disminuye en la medida que la diferencia o “distancia” entre las naciones de las empresas asociadas aumenten en términos del capital humano y distancia cultural. Por otra parte, una mayor diferencia geográfica e institucional es asociada positivamente con la selección de modos de alianza de gobernanza integrada. Adicionalmente, las empresas en etapa de investigación son más propensas a usar un modo más integrado de gobernanza, al contrario que empresas en etapa de desarrollo. Estos hallazgos avanzan la investigación en estructura de gobernanza de alianza. Estos muestran que los factores que afectan la selección del modo de gobernanza de la alianza de I+D.

Abstract

Nós identificamos uma variedade dos modos de aliança em P&D em uma indústria intensiva em conhecimento (por exemplo, produtos farmacêuticos), e os classificamos em quatro categorias ordinais que vão além da tradicional classificação binária de aliança de capital próprio versus capital não próprio. Isso enriquece a nossa compreensão sobre as estruturas de governaça em alianças e amplia a aplicação de modos de aliança no que hoje se apresenta como um ambiente de colaboração internacional em P&D mais complicado. Então, nós exploramos os fatores país, indústria e firma na definição da escolha de um modo de governança de aliança de P&D apropriado, utilizando uma amostra de 237 acordos de aliança internacionais. A probabilidade de usar um modo de governança de aliança mais integrado diminui à medida que a diferença ou "distância" entre as nações das empresas parceiras aumenta em termos de capital humano e de distância cultural. Por outro lado, uma maior diferença geográfica e institucional está positivamente associada com a seleção dos modos de governança de aliança mais integrados. Além disso, as empresas em fase de pesquisa são mais propensas a usar um modo de governança mais integrado, ao contrário de empresas em fase de desenvolvimento. Esses achados avançam a pesquisa sobre a estrutura de governança de alianças. Eles revelam os fatores que afetam a escolha do modo de governança de aliança em P&D.

Abstract

我们在知识密集型产业 (例如制药业) 发现了各种各样的R&D联盟模式, 并且将它们分成四个有序的类别, 这超越了传统二元股权和非股权联盟的分类。这丰富了在当今更为复杂的国际R&D 合作环境下我们对联盟治理结构的理解, 拓宽了联盟模式的应用。我们然后使用237个国际联盟交易的样本, 探索了决定选择一个适合的R&D 联盟治理模式的国家、产业和公司因素。当合作伙伴公司的国家差异或“距离”在人力资本和文化距离上增加时, 使用一个更为集成化的联盟治理模式的可能性会减少。另一方面, 更大的地理和制度差异与选择更加综合的联盟治理模式之间呈正相关。此外, 公司在研究阶段更有可能使用更为集成的治理模式, 而处在发展阶段的公司则截然相反。这些发现推进了联盟治理结构的研究。它们揭示了影响R&D 联盟治理模式选择的因素。

APPENDIX

Methodology for Constructing the DV – DOI

An optimum alliance governance choice – along a continuum between arms-length transactions and a fully integrated mode (i.e., EJV) (Contractor & Lorange, 2002; Gulati & Singh, 1998; Vrande et al., 2009), minimizes uncertainties and maximizes benefits. Below, we identify diverse agreement-based governance modes used in the pharmaceutical industry, and rank order them into distinct categories with ascending inter-partner integration. In a subsequent section of this article, we add EJVs to the right-hand side of the spectrum, and then explore the determinants of the governance choice in each international alliance.

Classifying alliance governance types using cluster analysis

This empirical study constructs a DV we label as the “Degree of Overall Integration” in the governance of the international alliance. In reading and analyzing international alliance R&D agreements in the pharmaceutical sector, we identified different types of tasks and provisions that constitute the overall bundle of an agreement. These include

Contractual provisions (further details in the box below)

• i Asset Purchase (AP)

• ii Contract Development (CD)

• iii Contract Research (CR)

• iv Cross-Licensing (CrL)

• v Passive Equity Purchase (E)

• vi Joint Development (JD)

• vii Joint Research (JR)

• viii License (L)

• ix Loan (Lo)

• x Manufacturing (M)

• xi Supply(S)

Box The Components of International Alliance Agreements in the Pharmaceutical Field

A: Alliances that are contractual and do not involve EJVs

(1) Asset Purchase (AP): One company acquires legal control of one or more physical assets such as manufacturing plants, equipment, finished or work-in-progress product inventories, laboratory supplies, animals and so on.

(2) Contract Development (CD): One party sponsors clinical trials at the other company; a pharmaceutical company sponsors clinical trials at a small biotech, which completes all development (conducts clinical trials on its own).

(3) Contract Research (CR): In a research agreement, a sponsoring party engages another party to perform basic research services in the discovery and/or lead stages of an R&D project.

(4) Cross-Licensing (CrL): One party obtains a license to the intellectual property of the other party in exchange for granting a license of its own intellectual property.

(5) Passive Equity Purchase (E): An agreement in which one company issues shares of its stock to the other company, either in exchange for cash or loan amounts. Many agreements utilize minor equity purchases as part of the upfront or continuing compensation to the other company.

(6) Joint Development (JD): Both parties participate in and share the costs and risks of clinical development and/or commercial expenses.

(7) Joint Research (JR): Both parties participate in the basic research program. They may exchange data, information and materials necessary for each party to perform its obligations under the research plan. And either party may supply the other party with proprietary materials for use in the research program.

(8) License (L): One party obtains a license under the other party’s intellectual property to research, develop, make, use, sell or market or promote a product or technology.

(9) Loan (Lo): Repayment of the loan may be in the form of cash or equity from the borrowing company. A loan can be used to fund studies or research. In return, the lending partner can receive repayment upon any milestone achievement in a clinical or regulatory stage.

(10) Manufacturing (M): In a manufacturing agreement, one party manufactures a product, usually a compound, for use by the other company in clinical development or pre-commercialization.

(11) Supply (S): In a supply agreement, the company will make or have made a product for use or sale by the client company. The major difference between supply agreement and manufacturing is that a supply agreement usually contains delivery/distribution of products or lead compounds for clinical development trials as well as active pharmaceutical ingredients.

B: Alliances that are EJVs

(12) Equity Joint Venture (EJV): Company A and company B (or more parties) create a new separate legal entity which is jointly staffed and operated by the principals.

Equity investment

• i Equity Joint Venture (EJV)

Depending on the mix of the above twelve provisions or “ingredients” chosen for an alliance, we then classified non-EJV alliances (approximately 85% of the sample cases) along two dimensions which were then used for the cluster analysis.

(Dimension 1): Degree of Interaction: The degree of workflow/task interdependence between alliance partners, after Thompson (1967):

• Pooled task: Tasks that are performed independently but the allies are interdependent in economic or financial terms; examples are Loan and Passive Equity Purchase; no-way interaction

• Sequential task: The output of one task is an input for the other partner. In other words, the interaction between the partners is unidirectional (i.e., one way); examples are licensing, contract research and contract development; one-way interaction

• Reciprocal task: The output of a task is inputted simultaneously to both allies or is a jointly performed task. The interaction between the partners is therefore bilateral or joint; examples are cross-licensing and joint research/development agreement; two-way interaction

(Dimension 2): Degree of Complexity: The degree of alliance contract complexity that stipulates resource allocation, adjustment and adaptation of ongoing tasks (i.e., Number of alliance components in an alliance, and the number of pages of the alliance agreement)

The Degree of Complexity of an R&D alliance deal can be measured by the number of deal components in an alliance. For example, an agreement that includes licensing plus joint research can be coded with a value of two. In addition, we also considered the number of pages to capture the complexity of an alliance deal, following Hagedoorn and Hesen (2009). The number of pages is a crude index but has been used in previous studies because there is a presumed correlation between contract complexity and the number of pages. For example, if we compare a pure licensing agreement with a licensing plus option agreement, the latter is more complex and contains more details. A licensing plus option contains not only royalties but an additional contingent future financial reward which could be a lump-sum or a claim on future earnings. Thus it will tend to have more pages in the agreement.

Putting the above two dimensions together, non-equity-based alliance governance modes can be classified into three clusters; (1), (2) (A and B) and (3) using the “K-Means Clustering Procedure.” Figure 4 illustrates the increasing degree of overall integration rising from Clusters (1) to (2) to (3).

Recall that the cluster analysis was done only on the non-EJV, or agreement-based alliance subsample. This resulted in three groups of non-EJV alliances, for the DV “Degree of Overall Integration” (1) Low-Integrated, (2) Moderately Integrated, (3) High-Integrated. To this, on the right hand side of Figure 3, we add a fourth group, namely (4) EJVs. This conforms to three decades of alliance studies that conclude that when the partners create a separate JV firm, jointly staffed and operated with personnel from both partners and often with a more substantial financial commitment than a contractual alliance, the degree of overall integration is the highest.

Since each alliance agreement is a unique mix of disparate provisions, the traditional certitudes of older theories need to be modified and adapted. No longer can we use just “markets vs hierarchies” or even a “market, vs quasi-integration (EJV), vs hierarchy” categorization as in earlier TCE literature, because complex agreement provisions enable incentive alignment and controls without the use of equity. The role of more complex agreement provision as an incentive alignment is where TCE has been missing. Using the two clustering dimensions of degree of partner interaction and alliance structural complexity enables a more nuanced classification. For example, it is sometimes better for firms to add a joint collaboration provision that incentivizes knowledge sharing activities through a “joint steering committee” that constrains partner’s opportunism (e.g., technological leakage) than to share knowledge simply through a single licensing agreement. This is the case where partnering firms increase interaction through a “joint steering committee,” but also this increases alliance contractual/structural complexity by adding more specifically an “operationalization of joint steering committee” provision in the alliance contract. In a similar vein, earlier studies using the KBV suggested the EJV as an ideal mode of knowledge creation and transfer due to greater organizationally embedded control mechanisms (e.g., joint management board) (Oxley & Wada, 2009). But this can also effectively occur in complex non-equity contracts whose joint board provisions result in a high degree of partner interaction.

Cluster Analysis Procedure

Coding of alliance types (by degree of interaction)

Based on these 12 different types of alliance, we scored the degree of interaction where no-way is coded as 1, one-way as 2 and two-way as 3.

• No-way (1): Agreements that include Asset Purchasing, Loan and Passive Equity Purchase

• One-way (2): Agreements that include Manufacturing, Supply, License, Contract Research and Contract Development

• Two-way (3): Agreements that include Joint Research, Joint Development and Cross-licensing

If an alliance agreement contains multiple components such as a license as well as joint research, then we summed the interaction score between license (one-way: coded as 2) and joint research (two-way: coded as 3) to make a total score (i.e., 5) for the degree of interaction.

Coding of alliance types (by complexity of agreement)

We coded the degree of complexity by counting the number of alliance components in each alliance agreement (e.g., if an alliance agreement that contains three different components such as equity, license and joint research is coded as “3”), and the number of pages of the alliance agreement (after Hagedoorn & Hesen, 2009).

Cluster analysis (for classifying non-EJV alliances) in terms of rising levels of overall integration between partners

Of 237 alliances in the sample, 208 are non-EJV while 29 are EJVs. The cluster analysis was restricted to the 208 non-EJV alliances. Based on three items (i.e., the degree of interaction, the number of alliance components and the number of pages), we performed a K-means cluster analysis since it allows us to minimize the variance within each cluster, and this is more robust than any other hierarchical method in terms of presence outliers and errors in the distance measures (Salter & Olson, 2001). Initially we selected four clusters as a starting point. But later since the “overall degree of integration” rises generally in the “northeasterly” direction in Figure 4, we combined two clusters labeled 2A and 2B into one cluster as representing a moderately integrated alliance governance mode. And then we checked correlations among items. There was a very high correlation (i.e., 0.90) between the degree of complexity measured by the number of alliance components and the degree of interaction. This is possible because alliance partners are more likely to interact as the number of alliance deal components increases. Given this fact, we decided to use the number of pages as one of the dimensions of cluster analysis.

Table A1 provides ANOVA statistics. And the followings show the number of cases in each cluster (see also Figure A1 Scatter Plot (A): Four-Cluster Method).

Table A1 ANOVA statistics

Figure A1

Scatter plot for 208 non-EJV alliances.

Cluster 1 (Low-Integration): 92

Cluster 2A (Moderately-Integrated): 67

Cluster 2B (Moderately-Integrated): 14

Cluster 3 (High-Integration): 35

Although we interested in four-classification of alliance governance modes (i.e., Clusters 1, 2A, 2B and 3) we also use three-cluster method in order to test robustness of our (four) classification. Under three-cluster method, we were able to get three non-equity-based alliance clusters as follows.

Cluster 1: 124

Cluster 2: 67

Cluster 3: 17

As can be seen from Figure A1 Scatter Plot (B), there is no significant distinction between Clusters 2 and 3 in terms of “Degree of Overall Integration” which undermines the goal of our study. In addition, the empirical results based on the three-cluster method do not support our hypotheses. Therefore we use the four-cluster method for our empirical test.