Costs of partner search and selection in strategic alliances

Geringer (1991) and Al-Khalifa and Peterson (1999) identify task- and partner-related factors in analyzing the partner selection process in international joint ventures. More recently, Cummings and Holmberg (2012) specify task-related, learning-related, partnering-related and risk-related critical success factors that are supposed to be crucial for decision-making in a dynamic alliance partner selection framework. A review and synthesis of the factors outlined in these studies led us to differentiate between task-related factors comprising aspects that refer to the task at hand and the expected learning outcomes, on the one hand, and company-related factors including aspects that concern the attributes of potential partner firms and the relationship between them, on the other hand.

We follow Hoetker and Mellewigt (2009, p. 1027, footnote 2) and focus on an existing alliance as a selected governance form in which the involved parties interact and implement governance mechanisms to reduce exchange hazards and achieve a desirable outcome. We view search and selection costs as transaction costs that occur at the beginning of the alliance management process due to a firm’s needs for safeguarding, coordination and contingency planning. The collection and evaluation of information on potential partners reduces the risk of cooperation and coordination problems from the outset of the relationship (Blumberg 2001).

A knowledge-based argument can lead to the same hypothesis as the TCE-based logic that we apply. If a company aims to establish an alliance to seize opportunities to learn and combine capabilities, it may even intensify its search and selection processes in the alliance formation stage to find an appropriate partner. The broader the intended alliance scope is, the higher the complexity of the tasks that are to be performed as well as the number and intensity of interactions that are required between partner firms to transfer and combine knowledge in order to generate innovative outcomes (Novak and Eppinger 2001). Firms tend to prefer tighter control modes, for instance, in terms of more thorough search and selection processes in the alliance formation stage because they facilitate the anticipated integration and coordination of joint activities and prevent conflicts from the outset of the alliance (Erramilli and Rao 1993). An alliance-initiating company’s higher need for control need not be an outcome of its fear of partner opportunism but can also reveal its desire to stimulate a joint identity with a potential alliance partner that facilitates the intended knowledge transfer between partner firms (Kogut and Zander 1996). Although the underlying reasons differ, both transaction cost economics and the knowledge-based perspective imply that a broader alliance scope leads to more costly search and selection efforts in the alliance formation stage. We thank an anonymous reviewer for bringing this point to our attention.

The dataset has been used for the following previous publications: (Mellewigt and Das 2010; Hoetker and Mellewigt 2009; Reuer et al. 2006). The dependent variable search and selection costs has not been used in previous studies. In addition, in this study we have no omitted variables bias (F(3, 72) = 2.5; Prob > F = 0.1).

The results of this additional model, which are compared to the results referring to our hypotheses reported in Table 3, Model 4, are available from the authors upon request. We are grateful for an anonymous reviewer’s thorough comment referring to common method bias. We are aware of the fact that the reported “plug-in solution to the omitted variables problem” bears considerable pitfalls (see, for instance, Chenhall and Moers 2007 for more details). Instrumental variables estimation may represent another technical solution. For our four independent variables, we would need at least four theoretically sound instrumental variables that must be appropriately strong (Bascle 2008). Unfortunately, we do not have this opportunity due to the limitations of our dataset and cannot collect additional data ex post. Moreover, we must provide a solution that is both technically and theoretically sound. The question of whether the chosen instrumental variables are appropriate is always an issue of debate and may raise severe theoretical concerns. Furthermore, our sample is limited in its size. Instruments from 2SLS are biased in small samples. The consistency of using the instrumental variables approach is hence at risk. We thus prefer a technically inferior solution that at least allows us to provide a theoretical rationale that is rooted in the perspectives that we use to explain our hypotheses.

We used STATA 9, especially the sureg command and the small, corr and dfk options for small sample sizes to estimate the simultaneous equations that serve an exploratory purpose and are additionally shown in Table 3. SUR models are not identical to subgroup analyses that require post hoc-tests, which examine whether the differences between coefficients across models are significant. The interpretation of the results based on the SUR models is based on the coefficients and the respective p-values (for studies using SUR models, see, for example, Kennedy and Fiss 2009; Kim et al. 2011; Weigelt 2009).

An alternative explanation is that when companies are planning a broad alliance scope, they exclude many potential partner firms from the outset. Only large firms that are highly visible due to their reputations in the marketplace can provide a broad range of functions to be covered in an alliance (Castellucci and Ertug 2010). Because their number is limited, search and selection are relatively straightforward and cost-effective.