The Choice Between Contract and Equity Alliances and Their Effect on Financial Performance

The Choice Between Contract and

Equity Alliances and Their Effect on

Financial Performance

By

Jelle Vonk

Master Thesis Strategy & Innovation

Supervisor: Pedro de Faria

1

Abstract

This thesis focuses on the differences between contract and equity alliances, and tries to find out when companies choose for either of these governance structures. In addition, I have performed exploratory statistical research to see whether there is a possible effect of alliances on financial performance. I make use of a portfolio approach with a total of 487 firm-year observations originating from 65 different companies with a timeframe of 1998 until 2010. Statistical analyses are used to test the set hypotheses for significance.

2

Acknowledgements

Without the helping hand of a number of people I would not have been able to write this thesis the way it turned out to be. The insights gained from other people were of great value and I therefore highly appreciate them.

Firstly, my supervisor Pedro de Faria who has helped me with all the day-to-day problems that I faced. Always answering my emails within 30 minutes and meeting up the next day if I wanted to, I felt that Pedro was highly committed to this project. Thanks for your useful feedback and great ideas to structure my thesis. Because of this I am very grateful that you assisted me so well in completing this thesis.

Secondly, I would like to thank Brenda Bos for her contributions. Brenda has helped me work with the SDC database and how to use the portfolio approach. Although the creation of my database took a long time and was often tedious, I want to say many thanks for helping me with this essential part of my thesis.

Thirdly, my good friends Tim Kassenberg and Arjen van Veen. I have spent countless hours in the University Library with Tim. Thanks for the useful talks we had about both our theses, while drinking, often much needed, cups of coffee. And Arjen, a well appreciated roommate, thanks for walking into my room for the often insightful discussions we had, both on-topic and off-topic.

Finally, I would also like to thank Marleen Olthoff for all her support while I was writing my thesis.

4

1. Introduction

1.1 Introduction

Alliances are a popular method to combine complementary resources between companies (Eisenhardt and Schoonhoven, 1996), share costs and risks (Hagedoorn, 1993) or co-develop new products (Tsang, 1998). Although different structures for cooperating with outside partners exist (e.g. acquisitions, mergers, licensing etc), alliances are a popular way for inter-firm cooperation and started to soar in the 1970’s and 1980’s (Hagedoorn and Schakenraad, 1994; Gulati, 1995; Ghemawat et al., 1986; Glaister and Buckley, 1994; Hergert and Morris, 1988; Lavie, 2007; Sampson, 2007). Due to the increased number of alliances, scientific literature regarding alliances has also increased (Lavie, 2007; Jiang et al., 2010). I use the definition of Das and Teng (2003) to define strategic alliances, which are inter-firm cooperative arrangements aimed at pursuing mutual strategic objectives.

The motivations behind strategic alliances can be looked upon from two broad perspectives, Resource Based View (RBV) and Transaction Cost Theory (TCT) (Yasuda, 2005). RBV states that companies are packages of resources, and if resources that are needed cannot be purchased via the market an alliance would be an option to obtain those resources. TCT focuses on minimizing transaction costs, and strategic alliances offer an opportunity to avoid costly internal development, but also to avoid market purchases with high transaction costs. Alliances are, then, an option in between market and internal transactions. Hamel, Doz and Prahalad (1989) recognize the need for alliances mostly from the fact that it takes too much resources to enter new markets and develop new products, and that for companies alone that is not doable anymore.

Despite the popularity of alliances, many scholars agree that alliances have a high failure rate, much higher than internal ventures or corporate buyouts (Park and Ungson, 2001; Porter, 1987; Harrigan, 1988; Parkhe, 1993; Das and Teng, 2003). Possible explanations are opportunistic behavior, bad coordination or alliances not reaching their set goals. These explanations are, however, often anecdotal, ad-hoc and do not form a complete story (Park and Ungson, 2001). Although I have included an exploratory part about the effect of both governance structures on company performance, this part is complementary and not the main focus of my thesis.

The main theme of this study is, however, focused on a comparison between contract and equity alliances. I describe what both types of alliances are, how they relate to each other and what their differences are. Next to that, I perform exploratory statistical tests to see whether there might be a difference in financial performance effects between the two structures. I also give attention to the factors that might have an influence on the choice between either contract or equity alliances. For instance, the diversity between partners is an important explanation. This can be caused by a different home nation, industry, or the number of participants. Including R&D into this study is important, as this might also be an important factor in the choice between contract and equity alliances.

I use a portfolio approach to structure the collected data and to perform the statistical analyses on (Lavie, 2007; Lavie and Rosenkopf, 2006; Jiang, Tao and Santoro, 2010). This approach looks at alliance portfolios of companies as an interdependent whole, instead of looking at each alliance as an ad-hoc decision. In the portfolio approach firm-year observations are created, which are always a combination of data from a specific company and a specific year.

5

The findings of this study are presented in the results section and explained into detail in the discussion section. The main findings of this study are concerned with the degree of diversity in an alliance that is caused by cross border, multilateral and different industry participants in an alliance. In this study I present that all three of these sources of diversity seem to be influencing the choice of alliance governance structure. In line with previous literature, I also find that contract alliances are more popular than equity alliances. However, in case R&D is involved, companies engage significantly more into equity alliances than when there is no R&D involved.

In the next chapter I will explain the relevance of this thesis.

1.2 Relevance

Performing research on alliance performance is relevant for a few reasons.

First, alliances are popular among companies and it is important to see why that is the case, but also to check if alliances actually have any effect on the performance of companies. Many researchers have consequently labeled many alliances as failures (Beamish, 1985; Das and Teng, 2000). Despite this finding, companies are still forming alliances. This could either mean that researchers label alliances in the wrong way, or that companies ignore findings of researchers, or do not have knowledge about their findings. It then makes sense to research the financial impact of alliances to see whether there is actually an effect or not.

Second, companies seem eager to form alliances but it remains a difficult question what kind of governance structure they should use. Although many scholars have looked at motivations to choose for the different governance structures (Pisano, 1989; Chen, 2004; Gulati, 1995; Osborn and Baughn, 1990), so far I have not found a scholar that has specifically focused on motivations to choose for contract and equity alliances. Also, as far as I know there is no scientific literature yet about the impact of alliance governance structures on financial performance.

2. Literature Review

Before introducing details about hypotheses and methodology it is necessary to frame my problem within current literature. To give a complete picture of current strategic alliance literature I will discuss formation issues, R&D alliances, alliance output, and the different contract types of alliances.

2.1 Alliance Formation

Before an alliance is actually formed companies need to evaluate what types of activities they want to perform with an alliance, and see what their goals are with that alliance. Finding a partner that fits within these goals is the next challenge. The types of different alliances are endless, and different motives exist to choose for all of them. Examples of alliances include joint ventures, joint production, joint R&D, joint bidding, contracted R&D, co-marketing, product bundling , licensing, code-sharing, manufacturing agreements, and many more (Das and Teng, 2003; Oxley, 1997).

6

pushes towards the choice of new partners. This indicates that companies’ networks are dynamic, and that, based on earlier alliance experience, new alliances will be formed, whether with similar companies or with different companies. Previous alliance experience is often mentioned to be a determinant for the type of alliance that is chosen (Hoang and Rothaermel, 2005; Sampson, 2004).

Motivations why firms want to engage in strategic alliances can be explained from various perspectives, but the main two theories are the Resource Based View (RBV) and Transaction Cost Theory (TCT).

Williamson (1975) introduced TCT, which argues that decisions made within an organization are based on minimizing the sum of transaction costs and production costs. Firms will internalize (e.g. internal development or mergers and acquisitions) when costs of transactions in the market are considered to have a higher price than internal production costs. When a firm decides to use the market it will only bear the transaction costs, but not the production costs. This will be the preferred option when transaction costs are lower and production costs higher (Das and Teng, 2000). Alliances are in between internalization and a market approach, as they combine elements of both. In case that an alliance offers the lowest transaction costs, for example through economies of scale and scope, it will be the preferred method to perform a certain transaction with.

The RBV motivation behind alliances stems from the fact that firms do not always possess the right resources and capabilities to undertake something on their own (Glaister and Buckley, 1996). Examples include knowledge about new markets or technologies, or capabilities in efficient production processes. Eisenhardt and Schoonhoven (1996) show that alliances are most likely to be formed by companies that are in dire need of resources and capabilities, or companies that have excess resources and capabilities, and thus want to share them and thereby earn a profit. Although the RBV typically focuses on the internal resources and capabilities of a firm, it is just that what a firms needs to know about its internal position to combine that with an outward view, and see what the firm needs from external sources. Ramanathan et al. (1997) brings it all together in stating that when inputs that are needed by a firm are possessed by another firm, a strategic alliance is the favored option when those inputs are inseparable from other assets that the owner possesses.

Two principles are essential to TCT, which are opportunism and bounded rationality (Williamson, 1991). Opportunism is the ‘self interest-seeking with guile’ and bounded rationality sees behavior as ‘intendedly rational, but only limitedly so’ (Simon, 1976). Agreements made for strategic alliances will always be incomplete and bounded by rationality of firms, and will therefore always be subject to possible opportunistic behavior of partners. To deal with this issue, TCT argues that the governance structure of an alliance is important, as it will determine how partners deal with their bounded rationality and how they monitor behavior, and thereby limit opportunistic behavior. In governance structures with a larger hierarchy, the chance of partners behaving opportunistically will be smaller. RBV focuses on bounded rationality in combination with the organizational capabilities of a firm instead, and that the capabilities of a firm decide the actions of that company (Tsang, 2000). Basically, RBV states that the actions of a firm are based on its embedded organizational routines that it has created (Nelson and Winter, 1982). These capabilities might be highly tacit in nature, and because of that highly valuable, as they are difficult to replicate.

7

long term value creations, such as the creation of knowledge. RBV sees the entrepreneur as a strategist that makes decisions based on creating value by innovative behavior and creating value by doing something new and different (Carland et al., 1984), whereas TCT views entrepreneurs as managers that solely base their decisions on transaction costs.

Another important feature of TCT and the chosen governance structure is asset specificity, or the ‘degree to which an asset can be redeployed to alternative uses and by alternative users without sacrifice of productive value’ (Williamson, 1988). When asset specificity is high, the various numbers of different transactions which can be done with the asset are limited, and therefore subject to potential opportunism. Value of a highly specific asset decreases when it is used in a different setting than its original, mainly because the essential supporting functions have disappeared (Tsang, 2000). In this case, a more hierarchical governance structure is preferred according to TCT. A larger hierarchy will lead to an increase of control by people with authority, and thereby limit the chance of opportunistic behavior, as this will be recognized better (Tsang, 2000). RBV deals with asset specificity in a different way. Whereas TCT focuses on a specific asset involved in the transaction, RBV focuses not only on the asset in question for a specific transaction, but also considers implications of a transaction for other firm resources. Also, RBV looks at a resource in the current firm situation instead of its specificity in combination with other resources.

So far I have established that TCT and RBV have a different view on what firms are and how they are being managed. Although they differ greatly in their views, both of them have explanatory power in the matter of why strategic alliances exist. TCT in particular has been thoroughly discussed in current literature, whereas RBV has received less attention. Williamson (1975) is one of the first authors to discuss TCT in his paper, and although it was criticized for being too simplistic, many authors have developed his ideas further (Tsang, 2000). Kogut (1989), for example, argues that following TCT logic, strategic alliances are most likely to be formed when there is ‘high uncertainty over specifying and monitoring performance, in addition to a high degree of asset specificity’. A hybrid form of governance, as Williamson (1975) calls strategic alliances, is in between market transactions and internal transactions. In an alliance specifying and monitoring performance is easier than in a market transaction. The high degree of asset specificity precludes market transactions, this because the value of such an asset is bounded to a set of circumstances and thus not suitable for market transactions. The choice for a construction in the middle, between market and internal transactions, is then a logical option.

Hennart (1988) looks at TCT theories regarding the creation of strategic alliances a bit differently, as he discusses that for strategic alliances to exist there must be an inefficiency in markets for intermediate inputs. When difficulties arise in pricing or transfer of inputs, cooperating in strategic alliances is a good way of internalizing some of the problems to get more control over them. Beamish and Banks (1987) argue that when a strategic alliance is formed when both partners express long term commitment and trust towards each other, chances of opportunism and a small-numbers condition can be reduced. Tsang (2000) argues that the theories of Hennart (1988) and Beamish and Banks (1987) are related, as both consider inefficiencies in the market and link them to performance monitoring problems. Because of those inefficiencies, monitoring costs are likely to go up. Strategic alliances, then, offer a solution in the form of internalizing some transactions by which, especially in combination with trust among the partners, the monitoring costs of transactions are likely to go down. Monitoring tacit knowledge flows is especially difficult, as valuating these flows is nearly impossible (Hennart, 1988). Because of this, it is difficult to assess if all partners have lived up to expectations. In a situation with higher hierarchy and control than in a market transaction, with trust among partners, firms are less likely to act opportunistically and monitoring of flows will be less needed, and thus less costly.

8

over M&A’s in two conditions (Das and Teng, 2000). First, a company might not need all resources of another firm, but in an M&A is obliged to buy those assets they do not value. Second, a degree of asset specificity is often present. When, if so, the acquirer wishes to sell off unneeded resources this is likely to be difficult (Ramanathan et al., 1997).

Tsang (1998) discusses five different motives for firms, from an RBV perspective, to choose for a strategic alliance.

First, the creation of rents. Although in a highly competitive market rents are often close to zero, by being able to create a specific set of resources and capabilities rents can increase. It is often a combination of certain resources that can generate rents, instead of focusing on a single resource. Partners in a strategic alliance do not necessarily have to be market leaders, but combining their strengths might deliver a strong set of resources. When these resources are imperfectly imitable and immobile, chances are higher that the competitive advantage created is sustainable. However, although rents might be sustainable, when the relationship between partners deteriorates an alliance might not earn superior rents. Especially joint ventures are known to be notoriously unstable (Blodgett, 1992; Gomes-Casseres, 1987).

Second, expansion of the resource usage. Exploiting resources to their full capacity will yield the largest rents. This especially when taken into account the large fixed R&D investments that companies need to make in order to develop new products. When a product is developed, a relatively short period of time is available to earn the investment back, and therefore the resource needs to be exploited to the fullest before the product becomes obsolete. To do this, firms have to often look beyond the borders of the industry they are active in, and thus may require other skills a firm is not yet equipped with (Tsang, 1998). Other problems include the ‘Penrose’ problem, which means that the capacity of a firm is limited by the existing managerial personnel (Marris, 1963). Managerial personnel can be hired to expand, but will first need to be integrated in the organizational culture and managerial pool, which will take time. Limited access to financial resources is also an important aspect, as a single firm might not be willing to make large investments on its own. In order to solve the latter two problems, strategic alliances can be a good option. It must be noted that these problems are more apparent for smaller firms than for larger firms, as larger firms have larger resources available. However, from an RBV perspective, forming alliances to get access to other firms’ resources is a major theme.

Third, diversification of resource usage. In essence, a way to reduce risk. Because of high initial investments of new product development and uncertain market conditions after release, companies can choose to cooperate with other firms to reduce the risk they take with such an investment (Tsang, 1998). By doing so, alliances are a great tool to hedge risk as neither one of the partners has to carry the entire risk (Porter and Fuller, 1986).

Fourth, imitation of resources. In the latter motive regarding resource usage a firm is interested in sharing its own resources and thereby sharing risk. In case of imitation of resources, the flow of resources is the opposite as a partner does not enter an alliance to share resources but wishes to learn to imitate resources from its partner. Alliances are well suited to facilitate transfer of skills, technology or exploitation of synergies (Harrigan, 1985). However, this is especially true for long term commitments between parties. Also, if one party decides to imitate resources of the other while not giving the other something for that in return, stability of the alliance might be in danger. In case both parties are in an alliance with imitation objectives, stability issues become more apparent (Tsang, 1998).

9

customer and supplier loyalty, and supplier’s trust are all likely to decrease. Third, disentangling the business unit from its parent company is a difficult operation. To solve these problems, for example, a joint venture with a 50/50 stake can be a solution for the first year, in which all the before mentioned problems can be dealt with, and after a few years the acquirer takes over the other half of shares (Tsang, 1998).

Next to the latter two theories, other strategic intensions exist to choose for an alliance. One might be that companies form alliances to improve their competitive position vis-à-vis their competitors (Kogut, 1988). Although the motivations behind strategic intent and transaction cost theory differ, both explanations are compatible and give insight into why firms engage in strategic alliances. Glaister and Buckley (1996) specifically research the motivations behind alliances and come to a similar conclusion. Although their research focuses on joint ventures, it shows that the main reason to form an alliance is to improve the firm’s competitive position and by doing that, maximize profits.

2.2 R&D Alliances

R&D alliances are mostly used to collaborate to be able to respond to rapidly changing technologies and product developments (Sampson, 2004). Because of this, R&D alliances are a popular method for companies to acquire and leverage technological capabilities (Oxley and Sampson, 2004). Also, they offer firms a chance to gain knowledge about their competitors, such as partner strategies and technology, competitive benchmarking, identifying key personnel, acquiring codified knowledge, and exploration of partners’ tacit knowledge (Oxley and Sampson, 2004). R&D alliances can exist in a lot of different ways. They can be equity based, such as a joint-venture, or non-equity based, such as an R&D contract. Also, R&D alliances can exist in a ‘pure’ form but can also occur in combination with a marketing or production element to it. In earlier days R&D was a part of companies that they used to shield off to protect their own inventions, but in today’s economy firms need each other to be able to compete (Narula and Hagedoorn, 1999).

R&D alliances offer an alternative to internal development and buying from the market. Alliances can work without the internal bureaucracy of a firm itself, and might therefore have higher innovative capabilities. R&D alliances also offer more coordination and control than a complete market contract (Williamson, 1991; Oxley, 1997). Although forming an R&D alliance might have cost benefits because of the sharing of product development, transaction costs in R&D alliances might be high. Due to difficulties in assessing value and exchange of knowledge, monitoring and control, transaction costs might increase (Gulati, 1995). In effect, alliances that have an R&D component to them are likely to have higher transaction costs than alliances that do not. Especially in R&D alliances the sharing of knowledge is important. This, however, does put partners in a potentially risky situation, as a company puts itself at risk of losing its appropriation of certain knowledge (Eden, 2008). Selecting the right partner is, then, critical, especially in R&D alliances. This because a company needs a partner they can trust, otherwise the transfer of knowledge that is so important in R&D alliances might not occur. Previous alliance experience is a factor that might create a sense of trust, which is supported by Eden (2008) who found that alliances are preferably formed with either friend firms they can trust, or stranger firms that have little knowledge about them.

Especially in case of R&D alliances, the choice of governance structure is important. Due to the highly tacit nature of information that flows through an R&D alliance, a structure that supports this flow is important. In the following chapter I will explain more about the different types of contract types, what their characteristics are and explain formation of strategic alliances from different perspectives.

2.3 Governance Structures

10

agreements. Examples of equity based alliances include joint ventures and equity investments (Chen, 2004; Das and Teng, 1996).

Das and Teng (1996) define two different types of risks that influence a firm’s choice of contract type; performance risk and relational risk. Performance risk is the risk that an alliance will not meet its objectives, and relational risk is the risk that partners in an alliance do not work well together. When relational risk is perceived higher than performance risk, an equity alliance will be preferred. In an equity alliance, better control mechanisms are available to check partner’s behavior. Thus, when a company expects its partner to behave opportunistically and perceives this as a higher risk than the chance that performance objectives will not be met, equity alliances are preferred. However, when performance risk is perceived higher as relational risk, contract based alliances are preferred. In a contract based alliance, flexibility of partners is higher, especially in terms of exit possibilities. Thus, when companies perceive the risk that performance objectives will not be met and want to keep their flexibility to exit an alliance as higher than the relational risk, contract based alliances are preferred.

The choice for either forms is often based on the perceived risk that partners in an alliance behave opportunistically. By doing this they will aim to maximize their own profits on the expense of other partners (Das and Teng, 1996). Extra costs will be made in an alliance when partners are behaving opportunistically, as negotiation and monitoring costs will go up (Hennart, 1988). Thus, because of these increasing transaction costs the need for an alliance structure that reduces the risk of opportunistic behavior also increases. Pisano and Teece (1989) talk about a preferred ‘mutual hostage’ situation in which both partners do not benefit from behaving opportunistically, since there will be a decrease in value of both their equity investment (Pisano, 1989). This situation is most likely to occur when partners are shared owners of an entity, or investment. In the case of a non-equity based alliance this is not the case, and therefore not the preferred option when chances of opportunistic behavior are high and costs of this behavior are considered high. In that case an equity based alliance has preference (Gulati, 1995). Osborn and Baughn (1990) add to this that technological intensity is also an important factor to account for, as high technological intensity will increase uncertainty and thus risk, which in turn will increase transaction costs. Therefore, in alliances with high technological intensity equity based alliances are preferred. Pisano (1989) too argues that for R&D alliances the preferred governance structure is equity based, although alliances without ‘pure’ R&D but that are ‘mixed’ with other activities (e.g. marketing, manufacturing) are even more likely to be equity based. Oxley (1997) confirmed this in her later study. Firms choose more hierarchical governance structures when chances of contracting hazards, she defines them as weak property rights, are larger. This finding is also supported by Gulati (1998), who perceives a joint venture to have the closest resemblance to the hierarchical control features of a normal organization in which hierarchy is high, whereas alliances without equity sharing have less hierarchical controls.

In a multilateral alliance (more than two participants) these hierarchical features could be beneficial. Diversity increases when the number of participants increases, and that is likely to cause a more complicated communication channel (Chen, 2001). This might increase conflict between partners, which in turn will reduce the effectiveness of an alliance (Gulati, 1995). A separate entity in case of an equity alliance might mediate these communicational problems. Both Chen (2001) and Gulati (1995) cannot find a direct relationship between the number of participants and equity alliances. However, Chen (2001) does find a moderating effect in the relationship between environment and the choice of alliance structure. ‘Multilateral alliances’ interacts strongly with the complexity of an environment, and complexity of the environment in turn influences the choice of alliance structure.

11

performance risk. When looking at alliances, however, the partners would have the local knowledge already so that is not likely to be the source of risk then. It would much rather be depending on a stable relationship and smooth communication, which would account to more relational risk than performance risk. Either way, engaging in a separate entity in an equity alliance could be beneficial for smoothening communication channels between partners. Gulati (1995) also looks at alliances with partners from different nations and concludes that they are indeed more likely to be equity based than contract based. Lavie (2007) looks at the performance of alliance portfolios and concludes that the performance of companies is negatively affected by the number of foreign partners in their portfolio. Possibly, this effect is different in case of a joint venture with foreign partners but Lavie (2007) does not focus specifically on this difference.

Diversity can also arise between companies engaged in an alliance that are not from the same industry. As explained earlier, this higher diversity might lead to more conflict and therefore less smooth communication. An equity alliance will be the preferred choice for such a circumstance. However, Oxley (1997) does not find a significant relationship between same industry participants and the choice of governance structure. It might be that the diversity caused by partners that are not from the same industry creates less conflict and therefore companies choose not to engage in an equity alliance, or that companies do not see the potential conflict and do not base their choice of governance structure on this aspect. Oxley (1997) also uses four digit SIC codes to establish the industry, whereas two digit SIC codes might be more appropriate. Using two digit codes might change outcomes because it looks at bigger differences between industries, and therefore ignores smaller differences between companies in the same ‘overlapping’ industry.

Many scholars find that when R&D activities are involved partners are more likely to choose for an equity based alliance rather than a contract based agreement (Osborn and Baughn, 1990; Das and Teng, 1996; Gulati, 1995; Pisano, 1989). An equity alliance has a couple of features that are highly suitable for the purpose of R&D. First, in an R&D alliance often tacit knowledge flows are intense. The environment in an equity alliance offers easier knowledge flows, but also higher controllability and therefore a better protected environment (Badaracco, 1990; Hennart, 1988). Especially in R&D alliances there is a likelihood of opportunistic behavior. In an equity alliance more effective control mechanisms are possible, which monitor progress of strategic goals and thereby prevent partners from manipulating its function or exploiting other parties (Gulati, 1995).

Flexibility in alliances is also an important aspect, especially in high-tech industries, as fast product development and continuously changing strategies of companies make it a larger risk to have an equity based alliance (Osborn and Baughn, 1990). It must be noted that this finding is based on high technology industries, and does not differentiate in the focus of an alliance (e.g. whether it is R&D, marketing, manufacturing etc.). Narula and Hagendoorn (1999), however, do find that non-equity R&D alliances are increasingly used over equity-based alliances. Despite the fact that many scholars find that (tacit) knowledge sharing is more difficult in a non-equity based alliance (Badaracco, 1990; Hennart, 1988).

12

performance more intensively. Monitoring performance, including tacit knowledge flows, is difficult and costly (Hennart, 1988). However, when trust increases, and contractual safeguards decrease, monitoring costs are likely to go down as partners trust each other more that they will perform adequately. By experiencing increased trust, alliances are then likely to limit their transaction costs (Gulati, 1995). Put differently, trust can substitute for hierarchical control systems, and is hereby likely to decrease transaction costs (Bradach and Eccles, 1989).

Another factor influencing the choice of contract type is the external environment an alliance will be in. Hagedoorn (1993) and Hladik (1985) find a relationship between the technology-intensity of an industry with the occurrence of strategic alliances with an R&D component. Harrigan (1985) adds here that in rapidly changing technological industries the preferred option are contract alliances, as they are somewhat less formal and offer higher flexibility. When an industry reaches a more mature state more formal strategic alliances, such as an equity alliance, are preferred. In essence, it is a paradigm between higher flexibility in contract alliances or higher control in equity alliances. Osborn and Baughn (1990) argue similarly, that alliances in R&D intensive sectors have a high need for flexibility, and that alliances in industries that have low R&D intensity have a higher need for control. Latter theories relate strongly to the description of the pre-paradigmatic and paradigmatic phase (Teece, 1988). In the pre-paradigmatic phase a firm finds itself in a competition among designs, which are all distinctly different from one another. In this phase it is crucial for companies to be flexible, as there is no clarity yet regarding the final design of a product. When alliances are formed in this phase, I would assume they are more likely to be based on contractual agreements. In this period characterized by uncertainty a trial and error process takes place, out of which eventually one dominant design will emerge. Next is the paradigmatic phase, in which the dominant design is set and firms focus less on product design. Instead focus shifts to price, and thereby scale and learning processes become more important (Teece, 1988). In this phase, innovation can still take place but they are more likely to be incremental innovations. Innovations in production, however, are taking place to reduce costs as much as possible. In this phase control is important, and therefore equity joint ventures are likely to be suited well to this phase. A longitudinal study starting in a highly dynamic market and slowly going into a mature direction would be best to study the previous mentioned hypotheses. As I will not address these issues in a quantitative study, it would be an interesting future research opportunity.

To my knowledge, no other scholars have tested the performance effects of both governance structures on company performance. Although many scholars signal that equity based alliances are better suited for joint R&D purposes, I have not found research articles that either confirm or reject this statement. Although differences between the governance structures have been thoroughly discussed, differences in output effects, however, have not.

2.4 Output of Alliances

Now that I have discussed formation of alliances and the different types of alliances, it is also important to look at the different outputs of alliances. Because of the different types of goals of alliances (e.g. they can be either financial or strategic in nature), looking at the different ways to assess output of alliances is an important aspect. The assessment of alliances can happen at two different levels, which are either the company level or alliance level. Although many studies have been performed to research both alliance and company performance, there is not much agreement between different authors on how to best capture this performance (Das and Teng, 2003; Lunnan and Haugland, 2008; Arino, 2003).

13

and not performance of the alliance. Second, operational measures include duration, termination and stability (Lunnan and Haugland, 2008). They add that when an alliance is terminated before the initial end-date, it does not necessarily mean that the alliance performance was not satisfactory. Termination might actually mean a successful end of an alliance (Inkpen and Beamish, 1997). Third, the effectiveness measures are the most commonly used in scientific literature. They include the overall satisfaction of partners about the alliance’s performance, and the fulfillment of strategic goals (Venkatraman and Ramanujam, 1986). These are based on initial and emergent goals, as well as common and private alliance goals. Although this measure is highly subjective, Arino (2003) indicates that when multiple indicators are used they may give reliable results.

Partners in alliances are bound to have diverging objectives. Das and Teng (2003) recognize three different types of objectives: compatible, conflicting, and same or similar objectives. When objectives are similar they are likely to be achieved simultaneously, and thus will not be any problem to both partners. When objectives are compatible they are not entirely the same, the exact definition might be different, but objectives are likely to be achieved almost simultaneously. For example, this could be sales level and market share objectives, they are not exactly the same but do have tendency to be highly correlated. Conflicting objectives are not compatible with each other and will not be achieved simultaneously (Das and Teng, 2003). Firms in an alliance then have common goals, for example to introduce a new product within a certain time period, or private goals, that might include financial performance goals. Anderson (1990) argues that the best method to assess alliance performance is to combine different measurements, and thereby look from different perspectives.

It is apparent that there are numerous ways of assessing alliance performance, either with a focus on the alliance itself or the effects on parent companies. Some researchers focus on operational and effectiveness measures (Zollo, Reuer and Singh, 2002; Chen, 2004; Hoang and Rothaermel, 2005; Robson, Katsikeas and Bello, 2008). Others choose to focus on objective measures, such as ROA, profitability, R&D spending, patents, or ROI (Schakenraad and Hagendoorn, 1994; Baum, Calabrese and Silverman, 2000; Shrader, 2001; Porrini, 2004; Lin, Yang and Demirkan, 2007; Sampson, 2007; Jiang and Li, 2008).

14

2.5 Hypotheses

A gap in literature exists regarding the effect that both equity and contract based alliances have on company performance. Many scholars chose to look at the performance of the alliance itself, but only few scholars have focused on the effect of strategic alliances on company performance. Das and Teng (2003) argue that strategic alliances ultimately could lead to a competitive advantage and thereby superior company performance. The difference between equity and contract based alliances is an interesting one, as many scholars argue that equity based alliance have an advantage over contract based alliances when alliance circumstances are more uncertain, or when knowledge sharing is highly important, for instance in an R&D alliance. Despite these findings, research has shown that most strategic alliances are contract based, which, in theory, might be less beneficial for company performance.

H1a: Strategic alliances are more likely to be contract based than equity based.

I hypothesize that contract alliances are more common than equity based alliances. With contract alliances there is no establishment of a new business entity. This makes initial effort and initial investment costs typically lower for contract alliances, and that makes this type of alliance easier to establish. Operating costs are also likely to be lower for contract based alliances, as equity alliances are more complex to administer and control (Narula and Hagendoorn, 1999). Another drawback of equity based alliances is that they are less flexible than contract based alliances, mostly because of the larger effort that must be put in to create a new entity, and the larger initial investment and exit costs. An advantage of equity alliances is that opportunistic behavior might be better dealt with (Narula and Hagendoorn, 1999; Oxley and Sampson, 2004; Osborn and Baughn, 1990; Gulati, 1995). In contrast, Haugland (1999) proposes that in contract alliances ‘relational contracting’ will counteract opportunistic behavior. This through coordination mechanisms as reciprocity norms, trust, and social capital embedded in multiplex exchanges and social interactions (Haugland, 1999).

Contract alliances are a hybrid form in between market exchanges and a hierarchy, and they are likely to be formed when transaction costs are too high for market exchanges but too little for internalizing the transactions (Williamson, 1985). I hypothesize that in most cases the costs of creating a hierarchy are considered too high, and therefore that contract alliances are most common.

H1b: Companies engaging in R&D alliances have a higher % equity based alliances than companies that do not engage in R&D alliances.

Equity alliances have a few distinct advantages over contract alliances which are especially useful for R&D alliances. R&D alliances are likely to have a high degree of tacit knowledge transfer, and because of that it is difficult to monitor in- and output of partners into an alliance. An equity based alliance offers better control mechanisms, and could be better for protecting alliance participants from opportunistic behavior (Gulati, 1995). Badaracco (1990) and Hennart (1988) argue similarly by saying that sharing (tacit) knowledge is more difficult in a contract based alliance.

15

proposed advantages of equity alliances in case R&D is involved, companies that engage in R&D alliances have a higher percentage of equity alliances than firms that do not engage in R&D.

H2a: Companies engaging in alliances that are characterized by a high % of cross boarder participants are more likely to form equity based alliances instead of contract based alliances. H2b: Companies engaging in alliances that are characterized by a higher number of average alliance partners are more likely to form equity based alliances instead of contract based alliances.

H2c: Companies engaging in alliances that are characterized by a low % of partners from the same industry are more likely to form equity based alliances instead of contract based alliances.

This hypothesis stems from the overlapping hypothesis that when there is a larger diversity between partners, the risk that the alliance will not work out will be higher. Oxley (1997) finds that when contracting hazards are greater, the likelihood of choosing a more hierarchical governance structure will be greater. Equity alliances, and especially joint ventures, are more similar to an internal firm structure and therefore offer a better chance to protect knowledge, and monitor different flows of in- and output of partners of an alliance. The chance that cooperation between partners will be less smooth can be assumed to be higher when alliance participants do not share similar working methods, culture, religion etc. Das and Teng (1996) labeled this ‘relational risk’, and when this risk is considered high the chance of participants choosing for an equity alliance is higher. I created three categories for this hypothesis, this in order to see what kind of effect the different aspects that might increase relational risk have on the division between contract and equity alliances. Although I have coupled these three hypothesis as they do have similarities in reasoning, below I have explained each hypothesis more into detail.

A high percentage of cross boarder alliances will increase the diversity in an alliance, as it is likely that companies from different countries have different business habits, deal with different market structures, have different legal systems and so on. These differences will ask adaptation from all partners, and will complicate communication channels (Chen, 2001). Related to cross boarder diversity is the international risk that comes with it (Miller, 1992; Das and Teng, 1996), which will be higher in case a company engages in more cross boarder alliances. Because of this, the risk that an alliance will not perform up to standards, either on a relational level or performance wise, will be higher. In this case an equity alliance will be preferred, as it offers an environment that is characterized by a higher controllability.

Another source of diversity is the number of different partners in an alliance. A higher number of participants will result in a more diverse environment in which there might be increased conflict. I expect that companies see the distinct advantages of equity alliances for this purpose, as an equity alliance offers an environment with a higher controllability. Both Gulati (1995) and Chen (2001) do not find a direct effect between percentage of cross boarder alliances and the number of equity alliances. However, Chen (2001) does find an indirect relationship. I expect, following Chen (2001), that equity alliances, direct or indirect, do offer specific advantages, and therefore that companies engaging in alliances with a high percentage of cross boarder participants are likely to form more equity based alliances.

16

H3a: Companies engaging in contract and equity alliances have a better financial performance than companies engaging only in contract alliances.

H3b: Companies engaging in R&D contract and R&D equity alliances have a better performance than companies engaging only in contract R&D alliances.

The motivations to choose for a particular governance structure are interesting to study. The next step is to see what the effects are of the chosen governance structures on companies’ financial performance. Although this part is complementary and highly exploratory, I did include it because it will give thought for future research. I will explore these hypotheses in a simplistic manner, as a more robust analysis would include panel data analysis that includes control variables. By comparing averages of financial performance data, it might be that companies engaged in contract and equity alliances do have a higher financial performance than companies that only engage in contract alliances. However, a causal link cannot be established with these tests. Unfortunately I do not have the time and data available to perform more robust analyses. Nevertheless, these tests will yield some exploratory results that can be used for future research directions.

Looking at the risk of contracting hazard (Oxley, 1997), opportunistic behavior of partners (Hennart, 1988) and the ‘mutual hostage’ situation that Pisano and Teece (1989) describe, one would expect that companies engaging in equity alliances have a better financial performance than firms that do not. However, firms do have to have good reason to engage in equity alliances. When their goals with an alliance would better fit a contract alliance, it would be too costly and take too much effort to establish an equity alliance, which in turn would not result in better financial performance. For example, an equity alliance is suitable for an alliance in which there are tacit knowledge flows. If, however, an alliance deals solely with explicit knowledge a simpler contract alliance would be favored over an equity alliance. In line with this reasoning, Narula and Hagendoorn (1999) indicate that the contract based alliance is becoming more popular. Looking at Das and Teng’s (1996) theory regarding performance risk and relational risk, Narula and Hagendoorn’s (1999) finding that contract based alliances are becoming more popular might indicate that firms perceive the performance risk higher than the relational risk and therefore choose contract based alliances. However, the fact that companies choose for a particular governance structure does not mean that this will give them the highest company performance boost.

17

3. Methodology

3.2 Data Sources

The main source for alliance data is Securities Data Corporation (SDC), a division of Thomson Financial. Although SDC collects data on a wide range of company related information, I have only used the alliance data offered. It is one of the largest alliance databases around, with the largest range of sectors available (Schilling, 2009). As the SDC alliance database does not offer data on companies’ financial performance, I have used Orbis (Bureau van Dijk) to complement the SDC data with. Orbis contains comprehensive information on companies worldwide.

3.3 Description of Sample

The aim was to have data on 50 companies to perform statistical analyses on. Because of the complexity of collecting data from various industries, I chose to perform analyses on one industry.

The software industry (SIC 7371-7376) is a good fit for my research, as the industry is very dynamic, it is characterized by fast developments and characterized by a lot of alliance activity. Of all public companies in this industry in 1990, somewhat over 10% of companies engaged in alliances. In 2001 this figure had already risen to 95%. The average alliance portfolio size has also increased in size dramatically, from an average of 5 alliances in 1990 to an average of 30 in 2001 (Lavie, 2007). These developments characterize why the software industry fits the portfolio approach well. Although this data is somewhat outdated already, it still represents the major changes that occurred in the number of alliances formed by companies.The software industry is also an industry that is characterized and shaped by complementarities and network effects (Gao and Lyer, 2009). Often companies are dependent on each other to obtain components that use particular technologies that these companies do not have in-house, and therefore have to acquire. These assets combined have much greater value than the individual assets, and therefore alliances are popular within the software industry. Another important factor is data availability. With a lot of companies being publicly traded most data on financial performance, employees, revenue and R&D intensity is available. Although the US accounts for half of the software industry, in my sample I have included companies globally to create more variability.

The timeframe I am working with is from 2003 until 2010, and alliance data is gathered from 1998 until 2010. This timeframe is selected because of the availability of data, since SDC does not contain that much information on alliances before 1990. Also, Orbis offers complete company data from 2003 onwards. All companies were selected from the software top 100 list of 20111.

To construct my sample, I make use of a portfolio approach. Per company and per year a portfolio of alliances is created, one observation is then always a company name, year, and all the collected variables. Making use of alliance portfolios makes sense because it creates a perspective from which alliances are not considered to be single isolated events but rather an interdependent series of events (Lavie, 2007). Next to this, I cannot see the impact of a single alliance on company performance, as there is no data available on appropriation of profits from individual alliances (Khanna et al, 1998). Because of this, looking at a portfolio of alliances is more convenient, and it is actually possible to gather the needed data. And finally, it makes sense because most studies have looked at single alliances and their stability, longevity or managerial assessments, instead of focusing on financial performance measures (Lavie, 2007).

The starting point for gathering data was the Orbis database, as availability of the ROA and control variables data was essential for the statistical tests to be performed. Orbis offers company financial data from 2003 until 2010 so I therefore chose to use this timeframe. Using the Orbis database and the

18

Software Top 100 list the largest software related companies were selected. Although I worked my way down the list, and thus the largest companies with alliances present were selected, the eventual sample is quite balanced in terms of company size. The largest company has a total revenue of 67,383 mln dollars and the smallest company has a total revenue of 1,563 mln dollars. Although this last company is still of relevant size, the absolute size differences in the sample are still quite large which is good for the diversity of the sample. Companies were included into the sample when they had at least the ROA data available, with a maximum of two values missing, and the other control variables available, also with a maximum of two values missing. All companies that did not meet the criteria were deleted. I also deleted all subsidiaries as they were missing financial data, but also to avoid interference with alliances of their parent companies. In the end I created a data file containing all information retrieved from Orbis.

After establishing the companies in my sample I gathered the needed data from SDC. The first step was to retrieve all information per company for all years (1998 until 2010). Companies that did not have at least five alliances within the timeframe were deleted from the sample. Then I created one data file for each company containing data from all years. The ‘unresolved’ datasheets (output of SDC due to incorrect values) were inserted into the ‘resolved’ datasheets. Although there were incorrect values now inserted in the dataset, there were no real problems with this data so it is usable. All of the data files now had to be transcribed to another format to be usable. At this moment I had 70 data files with sheets per year from 1998 until 2010, if data for those years was available. The next step was to see which variables were needed and how they should be calculated to be usable for testing. For creating the portfolios I use a pooled time-series analysis (Lavie, 2007), which means that alliances formed in the last five years are included to ensure that possible effects of earlier formed alliances are also taken into account. For example, when testing the effect of formed alliances on the dependent variable ‘financial performance’ of 2003, all alliances from 1998 until 2003 are taken into account. Although not every alliance will last five years, for simplicity reasons I assume they do. Otherwise it would be highly extra complicated. This approach is also taken by various other researchers (Stuart, 2000; Lavie, 2007; Hagedoorn and Schakenraad, 1994).

I created a general datasheet containing all the calculations and results for all years within the timeframe, and for the years 2003-2010 one line was created that contained all the variables that were usable for statistical testing. After this I inserted the variables from Orbis into this datasheet as well. After that I inserted this general datasheet in all 70 data files, and also inserted a sheet that retrieved all usable variables from that general datasheet. At that point alliance portfolios for 70 companies were created and these had to be manually put together to form one large dataset. Because SDC is known to be not entirely accurate (Schilling, 2009), I checked basic company facts for accurateness such as year of incorporation, SIC codes and possible mergers or acquisitions.

Eventually I gathered data for 70 different companies from the software industry. Of these 70 companies 65 have usable data for the general analyses. Of these 65 companies, 20 have usable data on R&D alliances. The sample consists of a total of 486 firm-year observations for the general analyses, of which 104 firm-year observations can be used for the R&D related hypotheses. All the firm-year observations that had no alliance data were deleted. When only few other variables were missing, I kept them in the dataset.

3.4 Construction of Variables

19

To be able to statistically test the hypotheses it is important to make the distinction between contract and equity alliances. To do this, I make use of the ‘joint venture’ variable from SDC for all hypotheses. I reckon joint ventures capture most of the equity alliance activity. Next to that, there is no other way in SDC to make this distinction so it is the best option.

The variables I included are:

Year Revenue

Return on Assets (ROA) R&D expenses Total Alliances Nation

R&D Alliances Similar SIC code Percentage of R&D alliances Age of firm Number of Employees

The following variables are SDC binary variables, being either a 0 or 1. Because the portfolio approach does not look at single alliances but at multiple alliances, these variables are inserted as a percentage in each firm-year observation.

Number of participants Joint venture Technology transfer Funding agreement Supply agreement Exploration Strategic alliance Exclusive licensing Spinout flag Equity transfer Same nation Equity stake purchase

Royalties Disclosed value

20

4. Results

In this section I will report on the results from the statistical tests that I have performed for each separate hypothesis.

Hypothesis 1

H1a: Strategic alliances are more likely to be contract based than equity based.

% Equity - Complete Dataset Firm-Year

Observations Mean Std. Deviation Minimum Maximum Variance

486 8,92% 18,096 0 100 327,469

This hypothesis looks at the dataset as a whole with all firm-year observations included (N=487), the average percentage of equity alliances for all firm-year observations is 9% (SD=18,096). It seems that, indeed, contract alliances are more popular than equity alliances, therefore H1a is accepted. It must be noted here that the 8,92% refers to the average percentage of equity alliances for firm-year observations, and this does not mean that 8,92% of all alliances in the sample are equity alliances. Percentages range from 0% and 100% equity alliances. It is important to stress that these figures are based on the averages of the firm-year observations, and not on individual alliances. Thus, the average percentage of equity alliances in company portfolios was 8,92%. The percentage of equity alliances on an individual alliance level was not calculated here. These results are based on the software industry, and results might be different for other industries.

H1b: Companies engaging in R&D alliances have a higher % equity based alliances than companies that do not engage in R&D alliances.

% Equity Engagement in R&D alliances

Firm-Year

Observations Mean

Std.

Deviation Minimum Maximum Variance

Yes 104 13,48% 18,748 0 100 3,5

No 382 7,69% 17,756 0 100 3,2

Related-Samples Wilcoxon Signed Sig.

Rank Test (sig. 0.05) 0.001

21 Hypothesis 2

H2a: Companies engaging in alliances that are characterized by a high % of cross boarder participants are more likely to form equity based alliances instead of contract based alliances.

% Equity % Cross boarder participants

Firm-Year

Observations Mean

Std.

Deviation Minimum Maximum Variance

Low % 275 6% 16,582 0 100 2,7

High % 211 12,66% 19,382 0 75 3,8

Related-Samples Wilcoxon Signed Sig.

Rank Test (sig. 0.05) 0.000

Two groups were created for this hypothesis, the boundary is set at the median of the percentage of cross boarder participants that were present in the firm-year observations. This resulted in a group of firm-year observations with a low percentage of cross boarder participants (‘Low%’, N=275) and a group of firm-year observations with a high percentage of cross boarder participants (‘High%’,

N=211). The mean of the ‘low %’ group is 6% equity alliances (SD=16,582) and of the ‘high %’

group is 13% equity alliances (SD=19,382). As this is quite a substantial difference, I have performed a statistical test to see whether these means are significantly different. The variables are ratio, not normally distributed and the group variable is created from the same sample, therefore a related-samples Wilcoxon signed rank test is applicable. As we can see P<0,05 and therefore H2a is accepted. Based on these results we can see that companies appear to see a higher risk involved with cross boarder alliances, and therefore choose to engage in more equity alliances. This is in line with Gulati (1995), who also finds that cross boarder alliances are more likely to be equity based than domestic alliances.

H2b: Companies engaging in alliances that are characterized by a higher number of average alliance partners are more likely to form equity based alliances instead of contract based alliances.

% Equity Avg. number of participants

Firm-Year

Observations Mean

Std.

Deviation Minimum Maximum Variance

2 309 4,18% 11,224 0 75 1,3

More than 2 177 17,26% 23,994 0 100 5,8

Related-Samples Wilcoxon Signed Sig.

Rank Test (sig. 0.05) 0.000

22

data analysis or a regression analysis, should be used in future research to control for other influences. The software industry is very suitable for this type of research, as it offers a high variability between the different alliances. To see, however, whether results are the same in other (less rapidly developing) industries should be tested in future research.

H2c: Companies engaging in alliances that are characterized by a low % of partners from the same industry are more likely to form equity based alliances instead of contract based alliances.

% Equity Similar SIC %

Firm-Year

Observations Mean

Std.

Deviation Minimum Maximum Variance

Low % 246 9,86% 18,126 0 100 3,33

High % 240 7,98% 18,083 0 100 3,33

Related-Samples Wilcoxon Signed Sig.

Rank Test (sig. 0.05) 0.007

I have created two different groups based on the median percentage of partners that had similar SIC codes. This resulted in a group with firm-year observations that had a low percentage of similar industry partners (‘Low%, N=246) and a group with firm-year observations with a high percentage of same industry partners (‘High%’, N=240). The mean of the ‘low%’ group is 10% (SD=18,126)and that of the ‘high %’ group 8% (SD=18,083). Because there is a slight difference in these percentages, further statistical analysis was needed to establish a significant difference. Because the data was not normally distributed, the variables were ratio and the data had the same sample source, I had to use a non-parametric related pairs test. The related-samples Wilcoxon signed rank test results in P<0,05 and therefore H2c is accepted. This is a different finding compared to Oxley (1997), and should therefore be investigated further. It might be that this finding is specific for the software industry, and because of that Oxley (1997) did not conclude the same. Oxley (1997), however, did use more complex statistical analyses including control variables that might have filtered out other influential variables.

Hypothesis 3

The outcomes of the following hypotheses are highly exploratory. I am not controlling for other variables’ influence, the lag in effect on financial performance is not controlled for, and therefore I cannot draw conclusive results for these hypotheses. By comparing average ROA data, it might be that equity alliances do have a contributing effect on financial performance. However, a causal link cannot be established with these exploratory tests. To get a better overall picture, a panel data analysis should be performed to exclude the effects of other variables on financial performance. This is, however, due to the insufficient number of firm-year observations, not possible on this dataset. Thus, the following hypotheses are highly exploratory of nature, but might be used for interesting future research opportunities.

H3a: Companies engaging in contract and equity alliances have a better financial performance than companies engaging only in contract alliances.

% ROA Firm-Year

Observations Mean

Std.

Deviation Minimum Maximum Variance

Contract 293 8,84% 10,43 -96,29 40,91 108,853

Contract and Equity 193 11,4% 8,9 -18,28 32,72 79,23

Related-Samples Wilcoxon Signed Sig.

23

It would be better to make a direct distinction between contract and equity alliances. However, there are no firm-year observations available in the dataset that only include equity alliances. Because of this I cannot group into firm-year observations with only equity alliances. Instead I chose to group the variables into firm-year observations with only contract (N=293), or contract and equity alliances (N=193). The mean ROA of the contract variable was 9% (SD=10,43) , which is lower than that of the contract and equity variable, with an ROA of 11% (SD=8,9). The variables are interval, not normally distributed and the group variables originate from the same sample, therefore I have to use a non-parametric paired test. Although there is a difference in means, the related-samples Wilcoxon signed rank test resulted in a P>0.05 and therefore H3a is rejected. Using a standard non-parametric test without control variables is, however, too simplistic for this purpose. Because of this, these results are not reliable and panel-data analysis with a larger dataset should be performed to get more reliable results.

H3b: Companies engaging in R&D contract and R&D equity alliances have a better performance than companies engaging only in contract R&D alliances.

% ROA Firm-Year

Observations Mean

Std.

Deviation Minimum Maximum Variance

Contract R&D 75 10,14% 8,33 -10,27 29,05 69,31

Contract + Equity R&D 29 13,31% 10,07 -2,78 32,72 101,329

Related-Samples Wilcoxon Signed Sig.

Rank Test (sig. 0.05) 0.581