economics

The economics dissertation below has been submitted to us by a student in order to help you with your studies.

Are joint-ventures and their parent firms more closely related in terms of skill-relatedness than in terms of value-chain?

Abstract

The distance between joint-ventures and their parent firms is a fairly new topic of research in the field of diversification. In the process of determining the actual parent firm of a joint-venture out of all alternative industries, it turned our that both vertical- and skill-relatedness proved to be significant. The results indicate further that skill-relatedness is more predictive in this process than vertical relatedness and that joint-ventures are more likely to have parents that have skill overlap to their primary activity than industries that do not. These results hold for the entire sample and a subsample of manufacturing firms. Another finding is that joint-ventures and their parents tend to be more closely related in skills than their parents active in the joint-venture. Suggesting that joint-venture are a mechanism to reduce cognitive distance and increase the absorptive capacity of the new knowledge being transferred.

1 Introduction

Diversification and relatedness between firms has been widely researched and finds their theoretical foundation in Coase (1937); Penrose (1959) and others. Coase (1937) can be regarded as one of the first to address the transaction costs theory, while Penrose (1959) addressed the resource based view of the firm. Both theories form the basis of many empirical research to understand more about diversification behavior of firms. These theories therefore form a basis for further research in diversification and especially in the distance between joint-ventures and their parents for this paper.

Output produced by one industry often form the basis of production in other industries. It makes economic sense to integrate these activities into already existing activities to improve efficiency, make the company less dependable on their primary activity and expand the company, in order to achieve growth. Fan and Lang (2000) found this already before in their research about diversification. Firms tend to have secondary segments that are related in terms of in-output. This vertical integration of activities is most likely to occur when facing high market transaction costs.

Neffke and Henning (2010) also investigated diversification behavior of firms using in-output relatedness. Their research however introduced a new measure of skill-relatedness, which turned out to be dominant in predicting diversification behavior of firms. Conform the resource based view, regarding human capital as the prime asset of the firm, their research firstly compared observed job switchers against the predicted job switchers between certain industries. Individuals gain, certain specific skills during their working life and can only redeploy this knowledge (“know-how”) in other industries which posses some degree of knowledge overlap. If this is not the case, switching will only hurt the individual, since he or she will be not valued for all acquired skills during his working life. This measure of skill-relatedness between industries proved to outperform in-output relatedness and supported the resource based view in diversification activities of firms.

While diversification behavior and relatedness is widely researched, this is not the case for the relationship between joint-ventures and their parents. A joint-venture is a separate legal entity (Harrigan, 1988) and has at least two parents, who are in joint-control and reliable for their equity share in the joint-venture. This paper will discuss, why a joint-venture might be preferred over alternatives and how this might influence the distance between joint-ventures and their parent firms. This will provide new insights in the relationship between joint-ventures and their parent firms.

In order to investigate the relationship between joint-ventures and their parent firms and testing for dominant mode in this relationship, this paper used a sample of 237 German joint-ventures between 2005-2011 and constructed an average vertical relatedness according to Fan and Lang (2000) for German industries between 2005-2007. It than included the skill-relatedness of Neffke and Henning (2010) based on Swedish labor switchers between 2004-2007. These data allowed us to make an overview of all joint-ventures and all industries in which it can have their parent's. We then tagged the actual parent firms of the joint-ventures with a one and all other industries with a zero, allowing us to run a logit regression with the actual parent firms as our dependent variables. The findings indicate, that the resource based view is the dominant mode in explaining distance between joint-ventures and their parent firms. This provides more insight in the importance of close relatedness in terms of skills over other forms of relatedness.

In the following chapter, we start with a theoretical framework about diversification and argue that the resource based view can be regarded as the most important in diversification moves. We will then discuss the alternative diversification possibilities and the limitations of diversification. This will be followed by a discussion of the most important strategic motives on joint-ventures choice. These strategic motives and the theoretical framework will be summarized at the end of chapter 2 and we will explain how this all will relate on the distance between joint-ventures and their parents. In chapter 3 we will discuss empirical evidence on diversification and joint-ventures and their implications on our research. Then we will follow with a data discussion and our method of research in chapter 4 and 5. The results and out findings will be presented in chapter 6 and the final chapter will discuss our outcomes, provides some more insight in the distance between both parents active in the joint-venture, limitations, policy implications and future research.

2 Theory on diversification: an introduction

This paper investigates the distance between joint-ventures and their parents. Before we can have a look at this relation, we start with a theoretical framework about diversification and the implications of this framework for our research. Thereafter, we will explain more about diversification motives and the problems and limitations firms might face when diversifying. We then summarize all this and discuss how this all relate to our expected findings on joint-ventures and the relatedness with their parents.

2.1 Theoretical framework behind diversification strategies

A theoretical framework behind diversification strategies will be discussed in the coming chapter. This theoretical framework provides more insights in the diversification motives of firms. After these motives have been discussed, we can discuss their influence on the distance between joint-ventures and their parents at the end of chapter 2.

2.1.1. Transaction cost theory

The transaction cost view is a theory of Coase (1937) and Williamson (1975, 1985) and addresses the view that economizing is the core problem of economic organizations. The core of these problems in organizational context lay in the assumptions of incomplete information and self interest seeking firms / people. Incomplete information in contracts implies that it is impossible for individuals and firms to predict each future event, therefore all contracts are incomplete and exposed to uncertainty of future situation not foreseen by firms and individuals. If these future states / conditions change, the incentives for the individuals and firms involved might also change. In other words, there is room for self-interest of individuals. In the transaction cost theory, these assumptions of bounded rationality and self-interest seeking are paired and as a result there is room for fraud or guile of economic agents. Economic agents are driven by self-interest and the transaction costs theory allows these agents to deceive, disguise and confuse in order to maximize their self interest. Opportunistic behavior and moral hazard are thus included in the theory of transaction costs.

These assumptions are the basis for the theory of transaction cost and have some consequences, especially when it comes to contract modes and thus joint-ventures. Due to bounded rationality and opportunistic behavior of economic agents, all contracts are incomplete (Williamson, 2006). This means, economic agents have an incentive to behave to their own optimal ex post outcome if situations change which cannot be contracted. The second assumption is contract as promise (Williamson, 2006). This assumes that economic agents will fulfill contracts as promised. However, this will not be obtained if these agents are given opportunistic opportunities. “The transaction costs analysis entails an examination of the comparative costs of planning, adapting, and monitoring task completion under alternative governance structures” (Williamson, 2006, p. 58). The transaction will become the basic unit of analysis and minimizing transaction cost will result in the most efficient governance structure. Transactions differ in three ways from each other; (1) frequency at which transactions recur; (2) level of uncertainty to which they are subjected; (3) level of asset specificity involved.

Since asset specificity is of crucial importance, we elaborate some more about the characteristics of asset specificity. “Asset specificity has reference to the degree to which an asset can be redeployed to alternative uses and by alternative users without sacrifice of productive value” (Williamson, 2006, p. 59). This asset specificity becomes of importance in the context of incomplete contracts, while asset specificity can take different forms; (1) physical asset specificity; (2) site specificity; (3) dedicated asset specificity and (4) human asset specificity.

The complexity of a transaction is therefore highly dependent on the asset specificity (k) of the asset and investments in that asset. A supplier can for example use a general purpose technology with low asset specificity (k=0) or it might invest in a specialized technology with high asset specificity (k=1). High asset specificity is likely to involve high bilateral dependency between the parties in the transaction. Since the parties involved in the contract become vulnerable of each other, switching is difficult and costly option due to the mutual dependency and the investments done in specific assets. The buyers cannot easily turn to an alternative supplier and the current supplier is highly dependable on the demand of its current buyer. Therefore the higher the asset specificity, the more likely it become that higher contract costs have to be faced. Both parties have more incentives to devise safeguards to protect the investment in the transaction if asset specificity is high. However, if there is low asset specificity (k=0) and we thus have a general purpose asset, contract are easily monitored and market transactions will be preferred.

Back to the diversification decision, minimizing transaction costs is regarded of crucial importance for the choice in governance mode. This implies that firms choose between a wholly owned subsidiary, a simple market transaction or a hybrid made, as a joint-venture for example. This trade-off between a joint-venture and other governance modes has been widely researched. Hennart (1991) for example found that; Japanese firms start joint-ventures with U.S. counterparts to combine intermediate inputs when they are subjected to high market transactions costs. This paper uses a relatedness in terms of in-output and can therefore measure the distance in terms of the use of intermediate products between industries. The influence on joint-ventures and partner distance will be discussed at the end of chapter 2. At this point of the paper, it is however important to understand that high relatedness in the use of intermediate products is likely to be caused due to high transaction costs. This would imply that if diversification has a high level of relatedness in value-chain and are thus closely relatedness in terms of vertical relatedness, this is most likely caused by high transaction costs and supports the transaction costs view of diversification..

2.1.2. Knowledge and resource based view

In the resource based view, knowledge (“know-how”) is regarded as the most important production factor within the firm. The origin of the resource based view goes back to the work of Penrose (1959), who inspired the discussion of the resource based view of the firm and the importance of resources to achieve firm growth. Penrose stated that: “the firm is a collection of productive resources (human and non-human) under administrative coordination and authoritative communication that produces goods and services for sale in the market for a profit” (Penrose, 1959, p. xvii).” “The administrative coordination and authorities' communication define the boundaries of the firm” (Penrose, 1959, p. xvii). The firm specific human resources are regarded as the most important of all resources within the firm. Without these human knowledge, there can be no operating firm. As a result, the firm cannot make decisions, long-term planning, run operations and it can certainly not make any expansions.

From this point of view Penrose (1959) indentified two major causes of firm growth. First of all, causes external to the firm and secondly those causes that are internal to the firm. “External causes for firm growth, as capital constraints, cannot be fully understood without an examination of the nature of the firm itself” (Penrose, 1959, p. 532). We may therefore conclude that firm growth is endogenous to the firm; this is a result of two reasons mentioned by Penrose (1959). In order to execute plans and strategic action, human capital is required. After completion of the project/action, managerial resources will be released with increased knowledge. These resources gained experience and knowledge during the time of the expansion and can be redeployed at alternative use after the time of the expansion. The redeployed individuals with an increased knowledge and skills might improve efficiency and organization of the firm, but might also be able to development new or specialized services. Depending on the expansion, individuals involved might also gain ‘unique' knowledge of their experience; this is particularly true for certain forms of tacit knowledge, which are more difficult to transmit.

The theory of firm growth of Penrose (1959) has been regarded as one of the earliest contributions to the resources based view of the firm, stressing the importance of knowledge as the key production factor within a firm. The drive of firms for growth, is a drive for new knowledge that is not accessible to the firm before their diversification. However, the motives and goals of each diversification differ and so do the resources possessed by each firm in a diversification. These differences and similarities in knowledge are of crucial importance in the resource based view, where acquiring new knowledge is the ultimate goal for achieving growth. Acquiring knowledge comes with certain problems; the “fundamental paradox” of knowledge and the difficulty arising from transferring tacit knowledge are two of those problems. In the fundamental paradox of information it is extreme difficult to determine the value of the knowledge for the buyer of the knowledge, which causes high contract costs. Since it is impossible for the buyer of knowledge to estimate ex ante the characteristics of what is being bought. On the other hand, if the seller of the knowledge provides this information, he will be revealing important information and transferring his “know-how” free of charge (Arrow, 1959).

If the targeted knowledge, is a certain “know-how” which cannot be patented and protected against spillovers to competitive firms and other industries it become far more difficult. Certain types of knowledge cannot be put on paper and granted a patent. Firm's experiences in manufacturing, distribution, and country-specific knowledge, knowledge of markets, customers and especially high educated employees cannot be patented but are of crucial importance of a firm's success in the resource based view. “This type of knowledge that cannot embody specifications, designs and drawings, but instead is embedded in the individual is called ‘tacit knowledge'”. (Polanyi, 1959; Hennart, 1988, p. 366). These individual characteristics of experience and social nature make transfer, coordination and spread of knowledge between firms, extreme complex and difficult (Lam, 2006). The transfer and spread of this tacit knowledge is one of the difficulties when facing diversification decisions. The transfer and spread of this tacit knowledge can be done in different alliance forms, which will be discussed later in this paper. However, for now, it is important to know that diversification is undertaking to gain new knowledge, which must be for same part related to the knowledge of the firm. This is the case since the new resources must be redeployed at alternative use after a project, which might be a joint-venture for example. As for distance in diversification, higher skill-relatedness and thus diversification activities that are more closely related in skills stresses the importance of the resource based view.

2.1.3. Portfolio management theory

A third and final theory behind diversification motives is the portfolio theory of Markowitz (1952). Diversification decisions of firms are important decisions taken by firm's management in order to maximize the expected returns of their portfolio of investments. These investors are the shareholders of the firm and have a claim on the residual value of the company assets, when debt has been paid. In order to maximize this expected return of the firm outstanding shares, the law of large numbers will ensure that the actual yield of the portfolio will be almost the same as the expected yield. In any case, holding a diversified portfolio would be preferred over all non-diversified portfolios (Markowitz, 1952). Increasing variance in your portfolio would mean an increase in the number of projects, since each project would be successful / unsuccessful at a certain probability, which is referred to as risk. Holding a large variety, in other words, betting on more than one horse, increases your probability on having a winning project. The portfolio management theory suggests that diversification tends to take place in activities that are unrelated to the primary activity of the firm. If this is the case, diversification activities (such as a joint-venture) would be unrelated to the primary activity of the firm. There would be a large distance between the firm and its diversification activities, while transaction costs and the resource based view are stressing the importance diversification in more closely related activities, although for different motives.

2.2 Different diversification alternatives

In all theories discussed, the main driver for diversification is in order to achieve growth. Either, by minimizing transaction costs in the transaction costs economy or by diversification of risk, which increases the probability of a winning innovation. In all these theories is explained how they might influence the distance between diversification activities. Is there however any limit to firm growth in their challenge to innovate and to expand?

According to Penrose (1959) there is no limit on the size of a firm, however the growth of the firm has some limits it can reach. In the Hercules Powder Company case study Penrose claimed:

“Growth is governed by a creative and dynamic interaction between a firm's productive resources and its market opportunities. Available resources limit expansion; unused resources (including technological and entrepreneurial) stimulate and largely determine the direction of expansion. While product demand may exert a predominant short-term influence, over the long term any distinction between ‘supply' and ‘demand' determinants of growth becomes arbitrary” (Penrose, 1959, p.1)

How does this reflect to diversification strategies? Penrose (1959) distinguished between different areas of diversification. The firm can be divided into different productive activities, that consist of machines, processes, skills and materials, all closely and complementary associated in the production process, which Penrose (1959) calls the production/technology base. The firm now faces the decision to diversify into a new market using the existing technology base. It might prefer entering an existing market using a new technology base, which is referred to as horizontal/complementary expansion. The last scenario would be to enter a new market using a new technology base. As described above, the ability of a firm to expand and grow is limited by its internal resources, from which human resources is regarded as the most important. Diversification increases the creative and dynamic interaction of a firm and its resources.

All these forms of diversification have implications on the expected distance between the diversification activities and thus joint-ventures and our research. Entering a new market using a new technology would probably have a larger distance in terms of skills from its primary activity than entering a new market with an existing technology. In this latest case, the technology and specific knowledge can be partially redeployed at alternative use, while this is not the case in the first alternative.

The main implication from Penrose (1959) famous work is that firms diversify in order to achieve growth. According to Penrose (1959) the resource based view of the firm is the dominant view in order to achieve this growth by diversification. This would suggest that the distance between diversification activities would be more closely related in terms of skills and less closely in vertical relatedness, used as a measure for the transaction costs theory. If diversification is undertaken in order to diversify risk, conform the portfolio management theory diversification activities would not be related at all.

2.3 Limits on diversification and diversification distance?

There are different diversification forms as discussed in the previous chapter. It is important to understand that firm growth is limited by its human capital (Penrose, 1959). A firm should therefore carefully choose its diversification activities. A clear understanding of these limits and where these limits depend on is extremely important to understand the distance between firms diversification activities.

Since this implicitly answers the question, to what extent firms diversify and is there a limit on the distance between partners and their diversification activity? Cohen and Levinthal (1990) discuss the ‘absorptive capacity' of a firm, which indicates: “the ability of a firm to recognize the value of new, external information, assimilate it, and apply it to commercial ends, which is critical to its innovative capacity” (Cohen and Levinthal (1990), p. 128). This absorptive capacity puts limits on the commercialization of new knowledge and boundaries on diversification. Cohen and Levinthal (1990) assume that a firm's absorptive capacity and the individual absorptive capacities of its employers are largely a function of the firm's level of prior related knowledge. Earlier research suggest that absorptive capacity might be a byproduct of a firm's R&D investments and others suggest that firms can also invest directly in absorptive capacity while investing in specialized education/training. The key to absorptive capacity is that organizations needs prior related knowledge to assimilate and use new knowledge for exploitation. This is very important for the resource based view in our paper, since this implies that diversification activities of firms should be related in terms of skills. Since, the higher the prior knowledge in ones memory, the higher their ability to acquire new knowledge and the ability to recall and use that knowledge. What is often the case in organizations and especially expected in joint-ventures is the transfer of learning skills across bodies of knowledge that are organized and expressed in similar ways. Mowery et al. (1996) indicated that joint-ventures are the most efficient alliance form for transferring tacit knowledge, which could certainly human specific skills. As a consequence, experience or performance on one learning task may influence and improve performance on some subsequent learning task (Ellis, 1965). Cohen and Levinthal (1990) make two important assumptions about knowledge, important for diversification strategies. “Firstly, knowledge is cumulative and secondly, learning performance is greatest when the object of learning is related to what is already known” (Cohen and Levinthal, 1990, p. 131). This implicit that learning is more difficult in novel domains, in other words radical exploration of new ideas, products, technologies and standards. Diversification might offer an advantage, since with diversification comes a wider knowledge base and as a results an increasing probability that the new knowledge is already / partially known to the organization.

The absorptive capacity of an organization however, does not only exist off the aggregated absorptive capacity of its individuals, but also on the ability to exploit this knowledge. Cohen and Levinthal (1990) mention there is a trade-off between high levels of absorptive capacity of an organization and the ability to exploit this. They describe this as a trade-off between inward-looking (specialization) versus outward-looking (diversify) trade-off, where excessive dominance by one or the other will be suboptimal. Exploitation can best been seen as specialization of old familiar ideas and certainties in organizational learning, while exploration can best be described as the invention of new technologies, standards, products or ideas in an organization. Cohen and Levinthal (1990) discuss also the importance for innovation of close relationship with both buyers and suppliers, suggesting a vertical relatedness would be beneficial for innovation performance. In the trade-off described, Cohen and Levinthal (1990) suggest that to the keep an effective, creative utilization of new knowledge a portion of prior knowledge should be closely related with a the firm new knowledge, and another part should be fairly diverse, although still related. If this is the case, firm diversification activities should be closely related in terms of skills supporting the resource based view of the firm.

Why is it important to have both creative utilization and a portion of prior knowledge is best described by March (1991), who distinguishes between exploration and exploitation. Returns of exploration are systematically less certain than those of exploitation (March, 1991), this might influence the choice for diversification for the long term however, exploration has long run positive return although this outcome is certainly not always the case in the short run. Exploration activities therefore capture much more risk taking, uncertainty, variation, flexibility, discovery and innovation than exploitation. Exploitation is more focused on production, choice, efficiency, marketing, costs and benefits (March, 1991).

The importance of exploration is best described in a model of mutual learning in an closed organization and its personnel in it (March, 1991). The organization is regarded as a storage of knowledge (consisting of procedures, norms, rules and tacit assets) and the organization, accumulate knowledge over time by learning from their personnel. “Individuals (personnel) however, are socializing the organizational beliefs, which are diffused to individuals through various forms of instruction, indoctrination, and exemplification” (March, 1991, p. 74). This mutual learning approach between organizations and individuals has implications for the choice between exploitation and exploration in organizations and has therefore consequences for the short-run and long-run incentives. In this model of mutual learning organizational code is affected by the beliefs of their personnel, the other way around, the individuals are influenced by the organizational code / norm. Important to know is, that individuals can not influence each other, the influence each other through the organizational code. What will happen in this closed model? In this organization, each adjustment in beliefs is served to eliminate the difference between the organizational code and the individual beliefs. If the individuals over time become more knowledgeable about the code, they become also more homogeneous with respect to knowledge and in the end will find an equilibrium. In this equilibrium the individual's beliefs share the same organizational code. It is therefore important to keep a portion of new knowledge in order to increase the organizational code.

March (1991) also describes a second model, evaluating the role of personnel turnover in the organization and turbulence environment are considered. The length of service of an individual in an organization has a positive effect on the knowledge of the individual and therefore also a positive effect on the average knowledge of the individuals. A recruit therefore has a negative effect on the average knowledge of the individuals. The role of turnover on the organization knowledge is more complicated and is a problem of learning rates versus turnover rates. As described in model of mutual learning the strength of the recruit is, the diversity in knowledge, since the recruit posses on average less knowledge than the individual it replaces. Long serving individuals, on average know more, but their knowledge is already reflected in the organizational code over time and therefore they are less likely to contribute to the organizations knowledge base. Now consider environmental turbulence to the organization, this can be the case of processes involving lags in adjustment rates. Consider an organization without personnel turnover, in this organization the beliefs reflected by the individual and these beliefs do not change, although the environment is changing. After some time the organizational code is systematically degraded through changes in reality and a much lower equilibrium is reached. Organizations with a moderate personnel turnover however, are resistant to these environmental shocks and adjust to the new knowledge of the recruits (diversified knowledge).

March (1991) extent this model of competitive ecology in a model to compete for scarce resources and opportunities. Assuming the performance of a firm is a measure of the average value (x) and some measure of variability (v), which are normally distributed. An increase in both will increase the probability to gain competitive advantages over competitors. In this part there consist a trade-off between an increase in the mean and the variance. Which supports earlier literature, that diversification is undertaken to gain excess to new knowledge to some extent, but is expected to be related to prior knowledge of the firm. March (1991) conclude that exploration firms compete far more on variance than exploitation firms.

2.4 Implications and differences between the theories discussed

The main difference between the management portfolio theory and the resource based view and the transaction costs view is that the management portfolio expect that diversification tends to take place in unrelated industries, while this is not the case for the other two theories, although at different level of relatedness. The resource based view stresses the importance of knowledge gain and the benefits of this new knowledge in diversification. Transaction costs theory however focuses more on the cost side of the transaction.Leaving the transaction costs as basic unit of analysis to determine an appropriate alliance form, which will minimizes the transaction costs of the firm. According to Wang (2007), a firm should focus on maximizing the transaction value with its partner through exploiting and developing its resources. In choosing the most efficient alliance form a firm will evaluate the costs and benefits of all alternatives. Both theories are therefore highly valuable in choosing the appropriate alliance form. Where the focus of the transaction costs of the alliance is on monitoring, renegotiation, uncertainty and asset specificity and can this provide an overview of the most efficient form of alliance in terms of costs. The resource based view focus on the acquisition of the targeted knowledge of the firm. Which alliance results in optimal acquisition of this knowledge, regardless the cost of the acquisition. The potential benefits of an acquisition / joint-venture / agreement should be in the improvement of the firm resources.

Diversification tends to be unrelated according to the portfolio management theory, while this is not the case for the transactions costs and resource based view. The resource based view regards human capital as the prima asset of the firm. This paper uses a measure of skill overlap between industries to measure skill relatedness as a proxy for the resource based view. If firms are subject to high market transaction costs firms would choose for diversification in common use of intermediate products, using a measure of vertical integration based on input-output tables as a proxy for the transaction costs theory of diversification. If minimizing transaction costs is the main driver behind in diversification, then we expect vertical integration will be more important in diversification due to the use of new, less related technologies and products that could be integrated in the value-chain. If the resource based view is the main driver behind diversification, then we expect that the level of skill overlap is dominant and diversification is more closely related in terms of skills.

Earlier research resulted in many of our inspiration and theoretical foundation for this paper. The main interest of this paper is focused on a particular form of diversification, namely joint-ventures. All previous literature contribute to diversification and has some implications on expected distance between joint-ventures and their partner. The theoretical framework was therefore important to discuss at the start of this paper. We shift our focus now on the joint-venture and how this all relate to diversification.

This would imply that we might expect higher skill-relatedness between partners and joint-ventures active in manufacturing industries, who are not characterized by large R&D intensity are intuitively more focused on exploitation. On the other hand, the hotels and restaurant industries for example (SIC 55-60) are intuitively less likely to diversify vertical in the value chain, while the skills used in both industries are expected to be quite similar between these industries. While for manufacturing industries for chairs and seats (NACE Rev1.1 3611) and a manufacturer of sports goods (NACE Rev1.1 3640) would intuitively use very different skills, while some of the input can come from the same industry and the thus would have a higher relatedness in terms of in-ouput. For these reasons, we made a subsample of joint-ventures active in manufacturing industries to see if we can find different findings for the manufacturing industry.

2.5 What is a joint-venture?

In the previous part of the literature review the most important theories of diversification are discussed and how the influence the distance between different diversification activities. As explained before, this paper is interest in a certain sort of distance in diversification activities, namely the distance between joint-ventures and their parents. It is therefore necessary to shift our focus to joint-ventures and explain more about joint-ventures, which are best described as a business agreement, where two or more owners also known as “parents” create a separate entity (Harrigan, 1988). These owners (parents) setup a joint-venture for a partnership and co-operation. Important in this corporation is that all firms or “parents” of the joint-venture remain their own entity, the newly founded entity has its own liability and is separated from its founders in terms of legal responsibilities. The founders of the joint-ventures face (limited) liability, up to the height of the equity share invested in the joint-venture. In the following part, different diversification forms are discussed which are alternatives to joint-ventures.

2.6 What are alternatives alliance forms?

When explaining distance between a joint-venture and their parents it is important to have better knowledge about what causes a certain firm to start a joint-venture over alternatives. This will provide more insight in the motives and it might influence the expected distance between our the joint-ventures and their parents. In this paper we will consider an acquisition (Balakrishman and Koza, 1991; Hennart and Reddy, 1997), a license agreement, joint development agreement, R&D contract and partnership (Mowery et al., 1996) as most common used alternative alliances available.

The most important difference between an acquisition and a joint-venture is the integration of its acquired asset. For a joint-venture, the newly acquired asset is of shared interest with the other partner. This is not the case in an acquisition, where the acquired firm becomes a wholly owned subsidiary. However, if only certain assets of the target firm are of interest a joint-venture might offer a better alternative. This would suggest that joint-ventures are a setup of specific targeted knowledge of products and more are closely related to their parents than the alternative partner available. If this is indeed the case, joint-ventures and parents would be more closely related in terms of both skills and value-chain.

Major difference in other forms of alliances such as partnerships and contracts are; the incentives of the firms, controlling power, risk of self-interesting seeking agents and the ability to transfer ‘tacit' knowledge between the two separate organizations. In alliances as a contract form and partnerships for example, both firms will maximize their own utility, this is not the case in a joint-venture where the joint utility will maximize the residual share available for both firms. As Mowery et al. (1996) described joint-ventures are the most efficient form to transfer tacit knowledge, this would imply that we can find that joint-venture activities share a level of skill relatedness and overlap in human capital in order to access valuable and complementary resources to the firm. If joint-ventures are indeed the most efficient alliance form to transfer knowledge, we would expect that joint-ventures and their parents are closely related in skills.

2.7 Theory and motives preferring joint-ventures in strategic management

What motivates a firm to start a joint-venture over the alternatives, as an acquisition, a license or R&D contract for example according to strategic management? As discussed earlier in this chapter, joint-ventures have two main characteristics. Firstly, the relationship between the parent(s) and the joint-venture is an equity relation and secondly, the control is shared with the other parent(s). In strategic management, joint-ventures are regarded to achieve five main objectives: (1) the advantage of economies of scale; (2) diversifying risk; (3) overcoming entry barriers into new markets; (4) Pooling complementary resources (knowledge or investments); (5) reducing political risk when entering a new market (Pfeffer and Nowak, 1976; Harrigan, 1988 and Hennart, 1988). In the following part, these five incentives and their influence on the expected distance between joint-ventures and their parents according to our theories are discussed one-by-one.

The advantage of economies of scale

In an increasing global environment, costs reduction and cost focus are of increasingly importance to compete. Hennart (1988) calls this the drive of firms to reach minimum efficient scale (MES). Efficiency gains can come from economies of scale in the production for example. But it also possible to setup. A joint-venture supplying components with a general purpose technology to several of their parents would be an example of a joint-venture achieving minimum efficient scale. It is however difficult to test this strategic motive, since diversification is a move into a new activity for a firm, while scale advantages are setup for existing activities. It might however be the case, that firm start a joint-venture with other firms to reach advantages of scale for other than their primary activity. This is extreme difficult to measure, since we cannot a measure relatedness within an industry, but measure between industries. This research does provide some insight in the economies of scale for the distance between partners active in the joint-venture, where a large share of partners is active in the same four digit industry code

Diversifying risk

Choosing for a joint-venture in with another firm, will make the parent(s) less vulnerable of the results of their core activities. Conform the portfolio theory (Markowitz, 1952) investors should diversify in such a way, that they should maximize expected returns of their portfolio of investments. The law of large numbers will ensure that the actual yield of the portfolio will be almost the same as the expected yield. In any case, holding a diversified portfolio would be preferred over all non-diversified portfolios (Markowitz, 1952). An increase in portfolio variance would mean an increase in the number of project, since each project would be successful / unsuccessful at a certain probability, which is referred to as risk. Holding a large variety, in other words, betting on more than one horse, increases your probability on having a winning project. As March (1991) described the risk of investments is higher in joint-ventures of exploration nature than in investments of exploitation. Since exploration involves products new to the world in macro perspective, there is no information about needs, market and others. However, exploration is important for every firm to keep up with economic developments (March, 1991). Exploration is the key to economic growth in an endogenous growth model and thus firms would certainly pursue some level of totally knowledge, while this new knowledge should also exhibit some level of overlap with their current activities in order to exploit the new knowledge with success.

If firm's primary reason for starting a joint-venture is to diversify risk, joint-ventures are more likely to be located into unrelated activities from the parent firm conform the portfolio theory. This would also imply, that joint-ventures are undertaken in order explore rather than exploit.

Overcoming entry barriers into new markets

The ambition of firms to compete on international level might result in a joint-venture startup in foreign and new market for the parent(s). This drive to act on all global markets is very costly and difficult, since each market has its own characteristics. Joint-venture choice for a local distributer is expected to be preferred, especially if the local firm has significant more knowledge about the market and consumers. This strategy allows the multinational firms to maximize the market involved in and minimize the investment costs for each market. This strategy stresses the importance of gaining knowledge possessed by the other firm. Acquiring this knowledge might be more costly or involve more risk of government expropriation. This paper focuses on German joint-ventures and therefore this strategic motive is out of the scope of this research.

Pooling complementaryresources (knowledge or investments)

Much empirical research has been done on pooling resources in joint-ventures. Regarding this motive, firms setup a joint-venture to have access to the competitive advantages of other firms (Hennart, 1991). This complementary idea of resources can be seen in terms of technology (patents), assets, regional knowledge, legal advantages, production processes or in terms of capital. If an investment is regarded too large for a specific firm to undertake on its own, a partner firm would offer a solution for joint investments in new technologies for example. Pooling resources is a one of the key interests of this paper, using a skill-relatedness measure and in-output tables we would expect to find evidence that support the complementary resource strategy to diversify. If this is the case, joint-ventures are more closely related to their parents in terms of both value-chain and skills than all alternative industries. Furthermore, we expect that pooling resources of skills would be more important than in value-chain conform the resource based view. If the resource based view is indeed the dominant mode in explaining distance between joint-ventures and their parents, we expect to find that our joint-ventures activities are more closely in terms of skills than in value-chain with their parents.

Reducing political risk

The following motive comes from Hennart (1988). Multinationals entering new foreign markets might prefer to choice for a joint-venture with a local partner firm if political risk is higher. In this strategic motivation a wholly-owned subsidiary would be treated differently in the market than a joint-venture with a local firm. Especially for markets where international property right are insufficient protected and government involvement is high, a joint-venture with a local partner might be desirable. The strategic motive of international partner choice due to political risk is out of the scope of this paper since our sample consist of joint-ventures of German origin.

2.8 What are the implications of the theoretical framework on the distance between joint-ventures and their parent firm's?

In chapter 2 we started with a theoretical framework for diversification. After identifying theoretical frameworks to diversify, we discussed their implications on diversification and the limits of diversification as a mechanism of firm growth. Having a clear picture of diversification, allowed us to give an introduction in joint-ventures and alternative alliance forms available to firms. From here it was only a small step to discuss the strategic motives for joint-venture choice and explain how this might influence our expected distance between joint-ventures and their parents activities. In this part, we will discuss how these strategic motives related to the different theories and what this might imply for our expected findings. In order to have a clear overview of all type of alliances and the expected relatedness, we made an overview in table 1 and will now discuss this table in more detail.

2.8.1 Transaction costs view

The transaction costs theory focus on minimizing the transaction costs and when looking at a joint-venture, this would be most efficient at intermediate level of asset specificity. When firms setup a joint-venture in a different industries, according to the transaction costs view their focus should be combine in-outputs as efficient possible. Since the in-output of both industries are not the same, we speak about a symbiotic or vertical relatedness. It is also referred to as divergent alliances, since these alliance do not share the same technologies of products, organizing them in a joint-venture might be the best solutions.

If our sample exhibit a higher vertical relatedness than skill-relatedness, it would support the transaction view as the dominant view for diversification. However, this would only be the case if our sample exist the same amount of vertical and horizontal alliances. Since, we test the overall sample in terms of skills relatedness and vertical relatedness, it might be the case that about 70% of our sample is undertaken for horizontal alliance purposes. It is however, impossible to make a clear distinction between these sort of alliance having only the primary NACE code of them.

Furthermore, also exploitation alliances are best measured using in-output relatedness. These alliances are an extension of existing technologies and this can be in efficiency of input and outputs of production for example. If the transaction costs theory is main driver for the distance between joint-ventures and their parents, we expected to find higher relatedness in terms of value-chain than skills supporting the transaction costs view as the dominant mode.

2.8.2 The resource based view

Human capital as the prime asset of firms and the drive of firms to acquire new knowledge, which is for some part familiar to the firm would be characterized as a horizontal alliance. As described by Cohen and Levinthal (1990), the key to absorptive capacity is, that firms need prior related knowledge to assimilate and use new knowledge for exploitation. This suggest that joint-ventures and parents are closely related in terms of skills, supporting the resource based view as the dominant view.

Horizontal alliance and convergent alliances are all alliance forms, that increase the knowledge of the firm and are undertaken to get access to new technologies, markets or products that are somewhat similar and easily integrated. This integration is of crucial importance, since the new skills acquired must be redeployed at alternative and efficient use for a successful diversification. This search for complementary knowledge and convergent alliances, that increase the knowledge base of a firm would suggest that joint-venture activities of firms are more closely related in terms of skills than in terms of vertical relatedness. If our sample is more closely related in terms skills than value chain, it suggest that joint-venture activities are driven by the resource based view of the firm.

2.8.3 Portfolio management theory

The portfolio management theory assumes that diversification tends to take place to diversify risk. This would suggest that we can find almost no relation between the partner and the joint-venture activities and we have only divergent and exploration alliances. These alliance forms are systematically less certain and focus on totally new knowledge, which are not related to the partners. In our opinion, the absorptive capacity of these alliance forms are too high to put the actual acquired knowledge to alternative use and therefore we do not expect to find joint-ventures that are unrelated to their parents in terms of value-chain and skills.

Table 1. Overview of different alliance strategies and expected relatedness

Characteristics

Expected relatedness between partners

E xpected relatedness between parent's and JV

Vertical alliance (symbiotic / vertical / mutual interdependence)

Alliances that share the same relatedness in the value-chain, the common use of input and output of each other industries. (Fan and Lang, 2000; Pfferer and Nowak, 1976).

Relatedness in symbiotic (IO) expected to be larger than in skills.

High relatedness in terms of IO-tables linkage between partners.

High relatedness between partner and joint-venture in terms of IO linkage.

Horizontal alliance (complementary interdependence / competitive interdependence)

Alliance that will diversify horizontal in the value chain to get access to new technologies, markets or products that are somewhat similar and easily integrated to compete on similar markets (Pffefer and Nowak, 1976). The competition for scarce resources in support of the RBV.

Higher relatedness in terms of skills than value-chain, in order to compete for resources as is their main purpose. (Pffefer and Nowak, 1976)

Higher relatedness in terms of skills than IO-tables expected.

Convergent alliances

Alliances with significant transfer of knowledge and technological capabilities (Mowery et al., 1996)

Higher skill-relatedness would be expected, due to focus on RBV

Higher skill-relatedness would be expected, due to focus on RBV

Divergent alliances

Declining technological overlap, in other words, alliances for accessing rather than acquiring capabilities (Mowery et al., 1996)

Higher relatedness in terms of IO-tables than skill.

Higher relatedness in terms of IO-tables than skill.

Exploitation alliance (convergent alliances)

Alliance focusing on an extension of existing technologies (Nooteboom et al.,2007) this can be in efficiency of input and/or outputs (production) for example.

Higher relatedness in value-chain expected.

Higher relatedness in value-chain expected.

Exploration alliance (divergent alliances)

Alliance focuses on the exploration of new technologies, products, ideas or standards, breaking with the existing dominant design, norms, rules and activities. (March, 1991 and Nooteboom et al., 2007)

Higher skill-relatedness than IO relatedness, since exploration is focused on resources.

Higher skill-relatedness than IO relatedness, since exploration is focused on resources.

3 Previous empirical research discussion

In this part of the paper the empirical evidence about diversification is discussed. Starting with diversification strategies and the evidence on joint-venture choice over alternative alliance forms. We then discuss the empirical evidence on diversification and the distance between joint-ventures and their parents. Finally, we summarize the most important results of earlier research in a table and present our hypotheses.

3.1 Empirical evidence over the decision to start a joint-venture over alternative alliance forms? What are implications for this paper?

The importance of firms to diversify and the drive to gain access to new knowledge in order to grow is stressed earlier in this paper. The different alliance forms available to a firm in order to diversify have also been discussed. However, what does the empirical evidence state about the decision to start a joint-venture over alternative alliance forms. Hennart and Reddy (1997) investigated the determinants for the choice between two alternative methods, namely joint-venture and merger/acquisition for pooling similar resources and complementary assets. They built further on earlier research done by Balakrishman and Koza (1991, 1993), who suggest that a joint-venture is preferred if the potential target and the acquirer belong to different industries, which would cause higher transaction costs. The argument in favor of the transaction costs theory is that heterogeneous assets of different industries hold different information about quality, performance and value, which is certainly no common knowledge to firms of different industries. This will result in higher transaction cost and joint-ventures should be preferred. Joint-ventures should be preferred over acquisitions when firms combining assets have little knowledge of each other's business. When acquirer and target are active in the same industry, transaction costs would expected to be lower and a merger might be preferred. Hennart (1988) found another reason which could explain joint-venture choice over acquisition. A firm will favor joint-venture if the target asset is heavily commingled with other unneeded assets within the target firm. Hennart and Reddy (1997) tested these earlier findings on a sample of Japanese manufacturing entries in the United States. “Their results indicate that joint-venture are preferred over acquisition when the desired asset is ‘indigestible', i.e. when they are commingled with non desired assets because the U.S. firms owning them is large and not divisionalized (Hennart and Reddy, 1997, p.2)”. Also, joint-ventures are preferred over acquisition if Japanese investors have no prior knowledge/experience of the American market, when the Japanese and American firm produce the same product and when the industry is characterized by intermediate levels of growth. The most important findings of Hennart and Reddy (1997) is the evidence on experience, where no experience increased the likelihood of a joint-venture between Japanese investors and American firms. Since experience is a learning-by-doing process (Polanyi, 1959), American firms holding tacit knowledge about markets, industry and regulations have a knowledge advantage over Japanese investors. Pooling resources is found to be more likely if Japanese investors have no prior knowledge / experience in the market. This is in support of the resource based view, since experience can be regarded as a form of specific human capital, which should be fairly related to the diversification activity.

As for other alternatives, Kogut (1988) argued that joint-ventures are the most efficient form to transfer tacit knowledge between organizations, since with other forms of transfer, the knowledge is being transferred at organizationally embedded. Mowery et al. (1996) tested which alliance form would be the most efficient in the transfer of tacit knowledge. Since tacit knowledge is extremely difficult to capture, Mowery et al. (1996) assume this tacit knowledge can be best represented by patents. These patents flows between industries then are regarded to be closely related to the tacit knowledge flows between firms. Mowery et al. (1996) investigated these patterns of inter-firm knowledge transfer and with the use of a patent citation measure. Furthermore, they made a distinction between equity joint-ventures and other forms of alliances to investigate for difference between alliance forms.

Mowery et al. (1996) use the cross-citation rate (Citations to Firmj patents in Firmi's patents / Total citations in Firmi's patents), which provides a relative measure of the degree to which Firmi's technology-based capabilities are acquired from Firmj. In other words, it indicates the technological ‘overlap' between the two firms.

A successful outcome of the transfer of tacit knowledge is far from certain and is regarded to depend on the firm's ability to absorb new capabilities and knowledge. This absorptive capacity discussed in chapter 2, requires that a firm has specific knowledge, skills and expertise that complement the acquired tacit knowledge of its alliance partner in order to be successful. R&D intensity is often used as a proxy for the absorptive capabilities of a firm, however Mowery et al. (1996), use the pre-alliance cross citation rate to indicate the absorptive capability among partners. This covers the technological overlap among the partners before the alliances and their capability to absorb each other tacit knowledge.

Furthermore, a control sample is created of non allied firms by generating random pairings of firms. This will allow the authors to compare change in citation patterns of alliance partners with those of a similar sample of non allying firms.

The results exhibit no significant positive relation for relative measure of cross-citation between alliance partners. This absence of a pattern in change of cross-citation ratios between alliances can be explained, by the presence of ‘convergent' and ‘divergent' alliances which might offset the effect of each other. Convergent alliances are alliances where partners share each other capabilities in a joint-venture. Where a divergent alliance, don not use capabilities of both partners but rather use rather different technologies and knowledge, these divergent alliance have low level of overlap and certainly exhibit no increase of cross-citation.

Focusing on the sample with that exhibits technological ‘convergence' (cross-citation ratio after alliance - before alliance > 0), a significant positive impact can be found for joint-ventures over contract alliances. Also the measure for the absorptive capacity of a firm, the cross-citation rate prior the alliance, is positively and significant. This support the theory, that firm with higher level of absorptive capacity tends to have higher level of knowledge transfers from their partner.

Overall joint-ventures are seen as the most efficient form to transfer tacit knowledge, especially if only a part of the knowledge of the other firm is targeted. If joint-ventures are indeed the most efficient form to transfer tacit knowledge, the resource based view would be expected the dominant mode in explaining distance between joint-ventures and their parents. It would thus outperform value-chain relatedness and support earlier evidence of Montgomery (1994) and Neffke and Henning (2010) who found that when predicting diversification, the resource based view is dominant. This research is however focused on a particular type of diversification and distance, namely joint-ventures and the distance to their parents.

3.2 What is the empirical evidence about diversification strategies and partner choice in joint-venture?

Chapter 3.1 identifies when joint-ventures ought to be preferred over alternative alliances. This chapter will discuss in more details the results of empirical studies focusing on relatedness patterns between industries and between joint-venture partners. In chapter 3.2.1. the empirical evidence on diversification and relatedness will be discussed and thereafter the evidence on partner choice will follow.

3.2.1. Empirical evidence about diversification

Fan and Lang (2000), also found similar results; firms are more likely to own segments within the same two-digit SIC industry. This SIC measure is also better capable of capturing complementary relatedness than vertical relatedness (Fan and Lang, 2000). Fan and Lang (2000) also found that firms are more likely to own secondary segments that are complementary with their primary segments and firms are more likely to own secondary segments that are related to their primary segment and indicate that firms are more likely to own secondary segments that are vertically related to their primary segments. Lemelin (1982) found that firms are more likely to diversify within Porter's (1976) classification, than across these classifications. Using IO-tables proved to be a quite successful measure for vertical relatedness in their research. These findings suggest that firms are more likely to diversify into related industries than unrelated industries, which has also be found by Chang (1996). Neffke and Henning (2010) argue that prediction of diversification can be done on basis of labor flows, indicating that diversification tend to follow a resource interdependence. Their reveal skill relatedness index indicates that diversifications tend to have a higher skill-relatedness compared to the overall distribution. While Teece et al. (1994) found that diversifying firms tends to keep a constant level of relatedness between their core activity and neighbor activities. All this research agrees that there is some level of relatedness between diversification moves of firms. Since joint-ventures are a form of diversification, this paper expects that joint-ventures choose partners that are to some extent related to their core business and this relatedness is higher in terms of skills than in value chain relatedness.

3.2.2. Empirical evidence on partner choice in joint-ventures

According to Pfeffer and Nowak (1976), joint-ventures tend to follow patterns of resource interdependence, although these results are for a small set of 166 joint-ventures and based on a broad two-digit SIC measure. In Pfeffer and Nowak (1976) there is however no support for the in-output relation and transaction interdependence. What they did found, was that more highly concentrated industries, which is a proxy for market power, exhibit more purchase interdependence. The more highly concentrated the industry, the more technology intensity related to joint-venture activity, which would indicate that reducing risk of R&D projects would motivate firms to start joint-ventures in the R&D sector. Using a cross-citation rate of patents, Mowery et al. (1996) indicate that joint-ventures are the most efficient alliance form for transferring tacit knowledge. Their findings indicated that joint-ventures tend to have higher levels of cross-citation rates compared to alternative alliances forms. This would imply that we can find higher levels of relatedness between partners than their alternatives.

Concerning the trade-off of diversification described in by Cohen and Levinthal (1990) and March (1991); Nooteboom (2007) found evidence in support for the inverted U-shaped effect of cognitive distance on innovation performance. Moreover, they found that firms engaging in radical, exploratory alliances have higher positive effect of cognitive distance than firms that engage in exploitative alliances. This evidence suggests that we can find some level of relatedness, since diversification into unrelated industries would not be beneficial for the exploitation.

In order to have an overview of all findings discussed, a summary is presented in table 2. A distinction has been made between horizontal and vertical relatedness in partner choice, since both forms of relatedness are best captured using different measures.

Table 2 Summary of results on diversification strategies of firms and implications for this paper

Alliance form

Findings

Measure used

Vertical relatedness

Secondary segments are related to primary segments

IO-tables (Fan and Lang, 2000)

Relation between marketing and distribution is significant on basis of IO-tables.

IO-tables (Lemelin, 1982)

- pairs of goods sold (proxy)

The more concentrated the industry, the more vertical relatedness is observed.

IO-tables (Pfeffer and Nowak, 1976)

- purchase interdependence

Horizontal relatedness

Relatedness within industries combinations in which a diversification takes place is generally higher

Revealed skill relatedness (Neffke and Henning, 2009)

Firms are more likely to own segments within the same two digit industry

SIC industry classification (Fan and Lang, 2000; Pfeffer and Nowak, 1976)

Diversification into new activities tend be related to existing technologies

Teece et al. (1994) joint occurrence of industries as measure of relatedness

Firms are more likely to own segments that are more closely related

Porter's (1976) classification

Overall findings:

* SIC-based measures seems to capture complementary better than it captures vertical relatedness (Fan and Lang, 2000)

* Their seems to be a trade-off between the benefits of diversification and specialization, as may be concluded from the work of Nooteboom et al. (2007), who found evidence for the inverted U-shaped effect of cognitive distance on innovation performance.

* Diversified firms keep a constant level of relatedness between other activities (Teece et al., 1994).

Joint-venture findings on RBV:

* Joint-ventures tend to have higher levels of cross-citation ratios compared to alternative alliance forms (Mowery et al., 1996).

3.3 What are the general findings and how does this influence this paper and our hypotheses?

The theory and empirics discussed agree, that there is some level of relatedness in diversification moves. However, their opinions on this relation differ. Partly, this is caused by the difference in sort of alliances and because of the different measures used. Neffke and Henning (2010) found, that industries where diversification takes place, exhibit a higher relatedness in terms of skills than other industries.

Our first interest is if we can observe difference in explanatory power between skill-relatedness and in terms of value chain linkage, with the use of in-output tables to capture vertical relatedness. As Kogut (1988) stated; joint-ventures are the most efficient form of transferring tacit knowledge, we expect our hypothesis 1:

Joint-ventures and their parent firms are more closely related in terms of skill-relatedness than in terms of value-chain.

As a proxy for knowledge similarities between industries, we use the skill-relatedness index conducted by Neffke and Henning (2010). This paper is however also interested in the diversification strategies in different industries as explained earlier. We expected that manufacturing industries, characterized by large turnovers and low margins would be intuitively more likely to diversify vertical in the value chain. While this is different for hotels and restaurants for example. We therefore expect to find difference among industries and especially manufacturing industries;

Joint-ventures active in manufacturing industries and their parent firms are more closely related in terms of skill-relatedness than in terms of value-chain.

3.4 What is the relevance of this paper contribution to diversification strategies?

The aim of this paper is to provide more insight in explaining the distance between joint-ventures and their parents. Our findings might than explain how joint-ventures are related with their parents and which of our measures is dominant in predicting this distance. This will give us more understanding of relatedness in diversification and the absorptive capacity of firms. Firms need a prior share of knowledge to be related to their activities in order to assimilate and use new knowledge for exploitation. If this is true, the results should exhibit skill relatedness in joint-ventures and their parents. As Porter (1987) explained that opportunities for skill-transfer and skill-sharing are regarded as one of the most important determinants for synergies between firms and activities.

These findings might place also some new perspectives on joint-venture motives when we compare this distance with the distance between both parents active in the joint-venture. A joint-venture might be a mechanism to reduce the cognitive distance between both parents in order to transfer knowledge more easily.

4 Data

This thesis uses data from Zephyr, a database specialized in joint-venture data. In total 434 joint-ventures were selected from 2005-2011 of these joint-ventures, 197 were removed from the sample for diverse reasons , one for example; a missing NACE (European Nomenclature générale des Activités économiques dans les Communautés Européennes Rev 1.1) code to indicate the joint-venture or parent firm's primary activity. This process resulted in a sample of 237 joint-ventures of German origin and their partners. Next to these German joint-ventures, their partners and NACE code information, data is used from earlier research done by Neffke and Henning (2010) and industry in-output tables from Eurostat 2005-2007 are used, to conduct an average IO-relatedness between 2005-2007 for German industries. The Neffke and Henning (2010) industry skill-relatedness measure will be used to investigate the distance between the joint-ventures and their partner in terms of overlap in human capital. Due to the importance of our explanatory variables value-chain relatedness and skill-relatedness, the following part will explain the setup and importance of both variables.

Value-chain relatedness

The value chain industry relatedness measure is based on earlier research done by Fan and Lang (2000) and is constructed as follows;

We constructed an average industry in-output table from the Eurostat database from 2005-2007 for Germany. From this average 2005-2007 average industry in-output table, the vertical relatedness is conducted using the method of Fan and Lang (2000) in the following way; dividing the euro output value of industry i used in industry j by total output of industry j(aij), resulted in vij and represents the euro value of industry i's output required to produce 1 euro's worth of industry j's output. This process is also done for aji to get vji, however conversely of aij. We then take the average the average of the two input requirement coefficients to obtain the vertical relatedness coefficient of industries i and j, Vij = ½ (vij + vji). Vij can be intuitively interpreted as a proxy for the opportunity for vertical integration between industries i and j.

Appendix table A1 provides an example of value-chain relatedness of the food products and beverage industry and two other industries, illustrating how the vertical relatedness coefficients are conducted for this research. Using in-ouput tables of German industries, this research was limited since the in-ouput statistics where only available on a 2-digit NACE level. This limitation forced us to assume that the measure of relatedness conducted on a 2-digit NACE level would be similar on 4-digit level.

Skill-relatedness

Skill-relatedness is a measure for similarities in human capital between industries. Where human capital can be characterized by certain skills an individual posses, that are gained by learning-by-doing (Polanyi, 1959) and experience. These skills are to some extent, specific to a certain job, since individuals establish a certain job specifics skills during their working life. However, not all of these skills are job specific and in many cases can be partially redeployed to alternative jobs. The degree to which this specific human capital can be redeployed in other industries, depends on the level of overlap in human capital between these industries. If the specific human capital of the individual cannot be redeployed, his previously acquired knowledge will become “obsolete” and that will negatively affect his benefits as financial compensation. Since this paper is interested in partner choice for joint-ventures into related or unrelated alternatives, human capital is therefore an important determinant. The level to which skills can be redeployed in alternative industries, might therefore be highly important when considering partner choice in joint-venture.

Neffke and Henning used a sample of 9 million official registers of Sweden between 2004-2007 and in those years, about 4.5 million individuals were active in the labor market, where approximately 280,000 of them changed jobs. Their focus was on the specific human capital, therefore Neffke and Henning (2010) left managers and below wage earners out their sample of observed labor switchers. These two groups are individuals with more general than specific skills and can therefore far more easily be redeployed.

Having a sample of observed job switcher, Neffke and Henning (2010) conduct a baseline to compare their observed labor switcher against what would have been expected, using some basic industry characteristics as total employment, employment growth, average wages, industry size and industry growth comparing the predicted and observed labor flows between industries, result in the “relative excess of labor flows” (Nefkke and Henning, p. 15)

The relative excess of labor flows is determined by dividing the observed flow of individuals (Fijobs­), that move between industry i and j by the predicted flow of individuals that move between these two industries, based on the industry-level variables mentioned above. This process resulted in a measure of skill-relatedness as shown in equation one.

If the observed flow of individuals equals the predicted flow of individuals between industries, in other words SRij equals 1, than the is no relative excess flow of labor and these industries are thus unrelated in terms of skills. Values larger than 1, indicate a relative excess of labor flows and thus skill-relatedness, while values under 1 indicate skill-dissimilarities. To reduce skewness of SRij, Neffke and Henning (2010) transform their measure into RSR using the following equation;

5 Research method: a joint-venture perspective

To estimate the predictive power of the skill-relatedness index and vertical relatedness in partners choice for joint-ventures, this paper faced some problems, that forced use to make some assumptions. In real world, one of the parents is probably leading in setting up the joint-venture and therefore in choosing a partner or their partners. Another possibility might be, that the partners are already working together and start a joint-ventures to expend and improve their co-operation. In both possible scenarios it not clear which of the parent firms, took initiative in choosing a partner and which firm had a “target” role in joining the initiator in the setup of the joint-venture.

This forced us to have a look at the partner choice decision from a different perspective. Regarding the joint-venture as the initiator in the process of setting up the joint-venture. Start from a joint-venture (j) perspective allowed us to assume that the joint-venture is actually choosing her parents in one of the 408 possible industries available on 4-digit level, that are left after removing all industries were no skill-relatedness and vertical relatedness information was available, this is discussed in more detail in the footnote of chapter 4. These partner industries can be represented by a set of industries Xp = {x1, x2, x3, …,xn } that contains all industries in the economy available for the joint-venture to choice its partners. Let us assume that industry i, Xpis the primary activity of the joint-venture j. We can now describe the situation in which a joint-venture j can choose their partner of industry i using the following three arrays:

Cf ­­contains all possible partner industries from the primary activity of the joint-venture. Rf is a vector of skill-relatedness indices corresponding to the industry combinations listed in Cf. The same vector is conducted for the IO-relatedness. Finally, the vector df tags the partner choice that are actually being chosen with an one and zero elsewhere. This database shows the actual relatedness in partner choice from a joint-venture perspective compared to the alternative partners available no chosen. Comparing these samples will provide us with an indication of predictive power of our measures for partner choice in joint-venture.

Testing hypothesis 1 and 2, the importance of skill-relatedness and IO-relatedness on partner choice from joint-venture perspective and having a binary dependent variable allows us to use a simple logit regression using df as our dependent variable.

6 Results

In this chapter the results will be presented and discussed. Before we actually have a look at our regression results, some descriptive statistics will be analyzed. In table 3, the summary statistics of our dependent and independent variable are summarized.

Looking at the Skewness and Kurtosis, especially the SR is highly skewed to the right and therefore transformed in a more reliable measure of RSR, which is calculated by the transformation in equation (2). Running a Skewness and Kurtosis test for normality indicated that both vertical relatedness and RSR are still not normally distributed, however the transformation proved to be an improvement for the RSR and its distribution. Furthermore, we standardized vertical relatedness and skill-relatedness in order to compare both measures more easily and used these measures in all our models.

Table 3 . Descriptive statistics

Vertical relatedness (vc)

Vc_t 2

SR

RSR

RSR_t 2

d f

Number of obs.

95,784

95,784

95,784

95,784

95,784

95,784

Mean

0.012

0.000

1.123

-0.643

0.000

0.003

Standard dev

0.023

1.000

4.966

0.597

1.000

0.059

Standard error

0.000

0.003

0.016

0.002

0.003

0.000

Min

0.000

-0.496

0.000

-1.000

-0.596

0.000

Max

0.275

11.200

371.119

0.995

2.741769

1.000

Variance

0.001

1.001

24.657

0.357

1.000

0.003

Skewness

5.468

5.468

20.642

1.350

1.345

16.949

Kurtosis

40.199

40.199

907.864

3.261

3.261

288.258

This table reports the number of observations, means, standard errors, min and max of vertical relatedness and skill-relatedness coefficients as defined in chapter 4 (Data).

In table 4 an overview is presented of some descriptive statistics for the actual partner choice and the alternative options. What immediate becomes clear is that for comparing between means of both vertical relatedness and skill-relatedness, is that the mean of the actual partner choice is significantly higher than the mean of the alternative choices available for the joint-venture. These descriptive indicate, that vertical relatedness and skill-relatedness are important measures when choosing partner from a joint-venture perspective. The mean of both measurements is clearly significantly higher for the actual partner choice indicating that diversifying risk, mentioned in the strategic motives is not supported according with these descriptive statistics. Since this would suggest, that joint-ventures and their partner are more unrelated activities compared to alternatives. These descriptive statistics indicate that there is indeed a significant degree of relatedness between joint-ventures and their partners. This supports earlier findings of Mowery et al. (1996), Lemelin (1982), Teece et al. (1982), although their research was not focused on joint-ventures, but much broader. According to their findings, diversification tend to be more likely to take place into related activities.

Table 4 . Descriptive statistics: df =0 versus d­f =1 (alternative options not chosen versus actual partner choice)

Df=0

Df=1

H0 : mean (df=0) - mean (df=1) = 0

V c _t

V c _t

Ha : diff < 0

Number of observations

95,454

330

0.000

Mean

- 0 .00 5

1.38 5

Ha : different means

Standard error of the mean

0.003

0.154

0.000

RSR _t

RSR _t

Ha : diff < 0

Number of observations

95,454

330

0.000

Mean

- 0 .004

1.213

Ha : different means

Standard error of the mean

0.003

0.070

0.000

Before, the logit regression results are discussed, in table A2 of the appendix the correlation table is presented. The correlation between the two independent variables is about 0.23, which would indicate no severe multicolinearity between the independent variables, testing for this with variance inflation factor, supported these finding with a value of just over 1, indicating almost the absence of multicolinearity.

Additional insight in skill-relatedness

Having a more in-depth look at the relatedness measure for skills and its composition, we found that of our 330 observations of parent firms, only 240 of them exhibit actual relatedness in terms of skills and thus 90 observations have no skill-relatedness at all, which equals 80 joint-ventures in this case. Of the 408 possible industries a firm can choose to diversify, 220 of them have no skill-relatedness at all.

6.1 Regression results

To answer the main hypothesis of this paper, we moved to a multivariate setting including vertical relatedness, skill-relatedness and using our parent choice dummy (df) as dependent variable in a logit regression. In the first column, parent choice is regressed on skill-relatedness only. In the second column this is done on vertical relatedness and in the third column both skill-relatedness and vertical relatedness are included. Then, in the fourth model, we use both vertical and skill relatedness with an interaction term and finally in column four and five, we test the model including industry dummies for the joint-venture industry on 2-digit and 4-digit level. In all models we used the transformed and normalized measures for vertical- and skill-relatedness. Next to the coefficients, the size effect of the explanatory variables is included. This is the change in probability for the explanatory variable in the column (vc / RSR) associated with changing the explanatory variable (vc / RSR) from ½ standard deviation below its mean value to ½ standard deviation above its mean value. In other words, it measures the effect on the unconditional probability of a joint-venture and their parent, if the mean value increases with one standard deviation. The increase in diversification probability is striking, especially in the case of skill-relatedness. Using only a skill-relatedness and moving from ½ standard deviation below the mean to ½ deviation above the mean, will increase the unconditional probability with almost 50%, while this is only 25% for vertical relatedness in column two. Including both measures in one model, resulted in the same findings as of the effect size for both skill-relatedness and vertical relatedness. Indicating supporting our hypothesis 1, partner choice is better explained in terms of skill-relatedness than value-chain.

Interaction effect

In column 4, we used the standardized measures of skill-relatedness and value chain relatedness and included an interaction term. The interaction term turned out to be significant at a 5% level, however interpretation of interaction terms in logit regression is extremely difficult, since the interaction effect always follows an S-shaped patterns when plotted against predicted probability (Ai and Norton, 2003). This means that the interaction term is always positive and always negative for some observation, making it extreme difficult to understand the real relationship, especially when only using two independent variables which will not allow us to compute the cross difference.

Dummy variables

Since the previous models used no control variables for industry size for example, what might be an important characteristic for the decision in partner choice, we included joint-venture industry dummies on 2-digit and 4-digit level to check for omitted variables. The evidence indicate that using dummies variables does not significantly improve the model. Since the LR-test examines if leaving all dummies out of the model reduces the fit of the model in column four, we found no evidence for this at any of our significance levels.

These results are however not surprising, since the dataset is constructed in such a way, that each joint-venture industry is tagged with one 1 and some with two, it is clear that these differences are already captured in our existing model. Including dummies for all alternative industries as done in appendix table A3, will give us a completely different result. Indicating that dummies for alternative industries would significantly improve the model of column one compared to column two of the appendix table A3. Note that this is only for a reduced sample size of 29,169 observation. Skill-relatedness and vertical relatedness are both significant. This latest result indicate that when including dummies for all alternative industries for each joint-ventures improved the reliability and controlled for unobserved heterogeneity between alternative industries.

Table 5. Logit regression of parent choice from joint-venture perspective

Variable

(1)

Point estimate

Effect size

(2)

Point estimate

Effect size

(3)

Point estimate

Effect size

(4)

Point estimate

Effect size

(5)

Point estimate

(6)

Point estimate

RSR_t

0.874***
(0.046)

0.0019

0.740***

(0.048)

0.0015

0.713***

(0.050)

0.0015

0.730***

(0.051)

0.751***

(0.051)

vc_t

0.380***

(0.019)

0.0011

0.260***

(0.021)

0.0005

0.147**

(0.062)

0.0003

0.203***
(0.060)

0.190***
(0.060)

vc_t * RSR_t

0.057*

(0.028)

0.0002

0.040

(0.026)

0.047

(0.027)

JV ind. dummies on 2-digit level

No

No

No

No

Yes

No

JV ind. dummies on 4-digit level

No

No

No

No

No

Yes

Constant

-6.165***

(0.079)

-5.837***

(0.061)

-6.193***

(0.078)

-6.177***

(0.077)

-6.350***

(0.323)

-5.602***

(0.602)

LR score

24.36 (39)

49.90 (110)

P-value of LR

0.968

1.0000

Log-likelihood

-2,016.07

-2,084.61

-1,961.61

-1,959.06

-1,959.06

-1,934.11

Nobs

95,784

95,784

95,784

95,784

95,784

95,784

Significance levels: ***, p<0.01; **, p<0.025; *,p<0.05. All relatedness variables are continuous numeric variables described in chapter 4 and 5. Effect size indicates the increase in the actual parent choice of the joint-venture probability, associated with moving up half a standard deviation in the sample for the relatedness measures (vc and RSR).

6.2 Results on manufacturing joint-ventures

The interest of our second hypothesis was if manufacturing joint-ventures might exhibit are more closely relatedness in terms of vertical relatedness than skill-relatedness. Removing all joint-ventures, that are not active in a manufacturing industry. Left us with a sample of 81 joint-ventures active in manufacturing industries (NACE codes 15-40) and their parent firms, 112 in total, who can be active in one of the 408 alternative industries. Note, that these parent firms, do not have to be active in manufacturing business but can be active in one of the 408 alternative industries. As presented in table 7, the mean of vertical relatedness is not higher in manufacturing industries, which indicates that the average relatedness in terms of value chain is lower for manufacturing industries. While the mean of RSR is also lower for joint-ventures active in manufacturing industries, we use again our logit regression to see if vertical relatedness is better in explaining parent choice in joint-ventures than skill-relatedness.

Table 7 . Descriptive statistics: df =0 versus d­f =1 (alternative options not chosen versus actual parent of the joint-venture) in the manufacturing industry.

Df=0

Df=1

H0 : mean (df=0) - mean (df=1) = 0

V c _t

V c _t

Ha : diff < 0

Number of observations

33,276

112

0.000

Mean

- 0 .08 6

1.176

Ha : different means

Standard error of the mean

0.005

0.279

0.000

RSR _t

RSR _t

Ha : diff < 0

Number of observations

33,276

112

0.000

Mean

- 0 .058

1.060

Ha : different means

Standard error of the mean

0.005

0.122

0.000

As is done in table 5 for the entire sample size, size effects, interaction and dummies are estimates in the models presented in table 8. Due to the decrease in sample size, we also have had to adjust for the unconditional probability when comparing the results. As found in all other models, skill-relatedness outperformed vertical relatedness in all regressions and again the effect size of skill-relatedness is about twice as large than for vertical relatedness. Skill-relatedness will increase the unconditional probability with about 53%, while this is increase is only 29% for vertical relatedness.

The interaction effect and the dummies included, are not significant and do not add any explanatory value to the models. All coefficients for vertical relatedness and skill-relatedness are significant, which is a results found for all models again and does indicate that both measures are actually important when choosing a parent firm from a joint-venture perspective. Not surprisingly, skill-relatedness however outperformed vertical relatedness as found earlier.

Table 8 . Logit regression of paent choice from a joint-venture perspective active in manufacturing industries

Variable

(1)

Point estimate

Effect size

(2)

Point estimate

Effect size

(3)

Point estimate

Effect size

(4)

Point estimate

Effect size

(5)

Point estimate

(6)

Point estimate

RSR_t

0.799***
(0.076)

0.0018

0.680***
(0.079)

0.0015

0.680***

(0.082)

0.0015

0.703***

(0.083)

0.727***

(0.085)

vc_t

0.347***
(0.032)

0.0010

0.231***
(0.036)

0.0005

0.231***

(0.077)

0.0005

0.249***

(0.078)

0.239***

(0.078)

vc_t * RSR_t

0.000
(0.036)

0.0000

-0.005
(0.036)

-0.001
(0.036)

JV ind. dummies on 2-digit level

No

No

Yes

No

JV ind. dummies on 4-digit level

No

No

No

Yes

Constant

-6.059***

(0.124)

-5.807***
(0.102)

-6.067***
(0.123)

-6.067***
(0.123)

-5.973***

(0.327)

-5.992***

(0.328)

LR-score

7.19

13.53

P-value of LR-score

0.981

1.000

Log-likelihood

-696.20

-717.29

-681.11

-681.11

- 677.51

-674.34

Nobs

33,388

33,388

33,388

33,388

33,388

33,388

Significance levels: ***, p<0.01; **, p<0.025; *,p<0.05. All relatedness variables are continuous numeric variables described in chapter 4 and 5. Effect size indicates the increase in the actual parent choice of the joint-venture probability, associated with moving up half a standard deviation in the sample for the relatedness measures (vc and RSR).

6.3 Additional research and an introduction in new possibilities

As stated earlier in this paper, in real life partner choice might actually be a process of one leading firm and a target firm forming a joint-venture. Since this topic is very interested for further research, we transformed the data in such a way that the partner, first in the list of Zephyr, can choose a partner firm for their joint-venture. Again this setup test the probability of the actual partner choice against the alternatives. The main descriptive results are presented in table 8 and what immediate becomes clear that the mean of both vertical relatedness and skill-relatedness are significantly larger between actual partners than all other alternatives.

Table 9 . Descriptive statistics: df =0 versus d­f =1 for the distance between partner 1 and 2 in the joint-venture.

Df=0

Df=1

H0 : mean (df=0) - mean (df=1) = 0

Vc _t

Vc _t

Ha : diff < 0

Number of observations

69,573

171

0.000

Mean

- 0 .00 4

1.537

Ha : different means

Standard error of the mean

0.004

0.211

0.000

RSR _t

RSR _t

Ha : diff < 0

Number of observations

69,573

171

0.000

Mean

- 0 .002

1.361

Ha : different means

Standard error of the mean

0.004

0.098

0.000

Using a simple logit regression including skill-relatedness and vertical relatedness and estimating the effect size, as is done in column 2 of the third model in table 4 and 6, turned out that skill-relatedness is again more important than vertical relatedness when explaining partner choice. Comparing the unconditional probability for partner 1

With an effect size of 0.1% for skill-relatedness and 0% for vertical relatedness. The probability of choosing 1 out of 409 industries that are available (all NACE industries minus industries without skill-relatedness and vertical relatedness), the chance increases with about 42% when the SR value is increased 1 standard deviation above the mean. When using both measures in the model, as is done in column 3, the results indicate that again skills-relatedness outperform vertical relatedness since the vertical relatedness effect diminishes.

Table 1 0 . Logit regression of partner choice from partner 1 perspective

Variable

(1)

Point estimate

Effect size

(2)

Point estimate

Effect size

(3)

Point estimate

Effect size

RSR_t

0.937***

(0.064)

0.0013

0.787***

(0.066)

0.0011

Vc_t

0.405***

(0.026)

0.0008

0.261***

(0.029)

0.0003

Constant

-6.606***
(0.115)

-6.209***

(0.086)

-6.625***

(0.114)

Log-likelihood

-1,083.35

-1,127.30

-1,053.45

Nobs

69,744

69,744

69,744

Significance levels: ***, p<0.01; **, p<0.025; *,p<0.05. All relatedness variables are continuous numeric variables described in chapter 4 and 5. Effect size indicates the increase in partner choice probability associated with moving up half a standard deviation in the sample for the relatedness measures (vc and RSR).

Comparing the distance between joint-ventures and their parents

The results indicate, that skill-relatedness and thus the resource based view is again the dominant mode. It is now interesting to compare the distance between the joint-venture and its parent's with the distance between both parents in the joint-venture. This will give more insight in were the joint-venture activity is located compared to their parents relatedness. We therefore used the original dataset with 330 joint-ventures and removed all joint-ventures with more than two parents and all joint-ventures with combinations wherefore no skill-relatedness was available, this left us with 250 joint-ventures with a total of 500 parent firms. The main interest is now to provide more insight in were these joint-venture activities are located and how this distance is compared with the distance between the parent firms. We therefore made an overview in table 11 and 12 and will discuss the main findings.

Table 1 1 . Results distance between parents versus parent - joint-venture.

Joint-venture and parent 1

Joint-venture and parent 2

Parent 1 and parent 2

Both parents with the joint-venture

Nobs that share same industry classification code

96

114

85

59

Total

250

250

250

250

From out sample of 250 joint-ventures is becomes clear that about 40% of the relationships share the same industry classification. Note, that for each joint-venture we have 3 different relationships. Namely, the relationship between the two parents active in the joint-venture and we have the distance between the joint-venture with both their parents, which are indicated as parent 1 and parent 2. These results indicate that our sample of joint-ventures have at least 1 parent active in the same industry for about 42% of the cases. While, parents with share the same industry code on 34% of the cases.

In the original database, these observations with the same industries codes were removed since we did not have an measure of skill-relatedness within an industry, but one between industries. Based on the assumption that skills in one industry can partially be redeployed in another industry, it is easy to see that skills in one industry can easily be used by another firm active in the same industry. These firms might have a smaller cognitive distance, while the posses the absorptive capacity to exploit this new knowledge of other firms, since the acquired skills can be redeployed far more easily.

Another possibility might be, that these firms start joint-ventures with parents active in the same industries for scale advantages. Assuming that firms active in the same 4-digit industry classification share the same production inputs and outputs. Choosing to setup a joint-venture with a partner active in the same industry would then provide both parents with scale advantages, this would probably suggest that their joint-venture would be active in the same industry as their parents, which is the case for about 24% of our sample.

Are these parent firms more closely related in skills than with their joint-venture? To compare this we subtracted the distance between the joint-venture and parent 1 from the distance parent 1 and parent 2. We also subtracted the distance between joint-venture and parent 2 form the distance between parent 1 and 2. We then removed all joint-ventures were for no skill-relatedness for both differences was available. This left us with 165 joint-ventures and their difference in distance between the parents and the joint-venture and both parents. The results for these difference are summarized in table 12.

The results indicate that for both vertical and skill relatedness the majority has at least one parent firm that is more closely related in terms of skills and value-chain. The parents tend to more closely related in terms of value-chain than skills, since almost 33% of the 165 joint-ventures has parents that are more closely related than the joint-venture to both parent firms.

For skill relatedness, the parent firms tend to be less closely related than the joint-venture and both parents, since 29% of the observations have none of the parents that is more closely related than the distance between the joint-venture and both parents. While only 15% are more closely related in terms of skills between the parents versus joint-venture and parents.

Table 1 2 . Distance between parents versus parent - joint-venture.

Parent-parent SR >= parent-JV SR of:

Two or more parents

One parent

None of the parents

Total

Parent-parent VC rel. >= parent JV VC rel . of :

Two or more parents

13

26

16

55

One parent

6

63

27

96

None of the parents

6

3

5

14

Total

25

92

48

165

7 Discussion and Conclusion

Our empirical analyses indicated that partner choice in joint-venture tend to be related in terms of both vertical relatedness and skills, indicating that both vertical and horizontal alliance are formed. This is also in support with some of our strategic motives as scale advantages and pooling resources. The empirical results indicated further, that in partner choice from a joint-venture perspective, skill-relatedness has stronger predictive power than vertical relatedness, stressing the importance of pooling complementary resources. The effect of both measures are significant, however skill-relatedness outperform vertical relatedness in the overall sample and in the manufacturing industries sample. Indicating that the resource-based view of the firm is the dominant mode for partner choice in joint-venture and stressing the importance of horizontal alliances. When moving up from ½ standard deviation below the mean to ½ standard deviation above the mean, we found support that probability with skill-relatedness increases with 50%, while this is about 25% for vertical relatedness for the overall sample. Overall, the effect size of skill-relatedness is about twice as large than vertical relatedness, supporting our hypothesis 1. Joint-ventures and their parent firms are more closely related in terms of skill-relatedness than in terms of value-chain.

Partner choice from in joint-ventures is more closely related in terms of skill-relatedness than in terms of value-chain.While we found no support for hypothesis 2, joint-ventures active in manufacturing industries and their parent firms are more closely related in terms of skill-relatedness than in terms of value-chain.It turned out that skill-relatedness outperformed vertical relatedness in all regression results. When looking at partner choice for joint-ventures, firms tend to choice partners with vertical and skills relatedness, where skills relatedness is more important than vertical relatedness.

Additional insights

Additional research has been done from the partner perspective, although no information about the initiator of the parents of the joint-venture was available. This paper still had a look at the distance between the partners, when owning a joint-venture, that was limited to only two parents. Here, again skill-relatedness outperformed vertical relatedness. Another interesting result found in table 11 is that joint-ventures tend to have at least one parent active in the same industry code for 50% of the cases. While this is smaller for the relation between both parents in active in the joint-venture. This would imply that joint-ventures and their parents are more closely related than their parents in terms of skills, which is supported by our findings in table 12. Where the distance between the joint-venture and their parents is overall larger than the distance between the two parents. These results would suggest that joint-ventures are undertaken to reduce the cognitive distance between the two parents in order to improve the absorptive capacity of the parent firms, who are more closely related to the joint-venture, the knowledge center.

Policy implications

The policy implications for managers, responsible for diversification and partner choice in joint-venture is that the must be aware of the importance of skill-relatedness, when choosing a partner. Human capital is regarded as one of the most important assets of the firm and diversification moves tend to strongly depend on the overlap in skills. As the results indicated, a joint-venture can be seen as more efficient form to transfer knowledge since it reduced the cognitive distance in terms of skills between two firms and therefore their will be a higher absorptive capacity. Since the joint-venture is more closely related to their parent firms than the two parent firms are to each other, knowledge can be more easily redeployed in the parent firm is it comes from the joint-venture. It is important to understand we make no predictions over performance, only over the distance between the joint-venture and their parents. This would be an interesting expansion of our research.

The increase of new knowledge does increase the variability of the firm and increases the probability to gain competitive advantages over competitors and make these firms more resistant to external shocks as described by March (1991).

Limitations and recommendations for future research

The main limitation certainly lays in the assumption that the joint-venture is the initiator in the process of partner choice, which is in real life not a realistic case. Who is the initiator of the joint-venture does not change the distance between the joint-venture and their partner. It does however, change the average relatedness of the alternatives, since the initial industry would be different (assuming the initiative partner and joint-ventures are not active in the same industry). Additional information about the joint-ventures and their parents, would be highly beneficial to do further research on this topic. One's it is clear, who is the actual initiator of the joint-venture, extra information about performance would be interesting to use in further research. This would give more information about the performance of the firms if they are more closely related in terms of skills, what would be a proxy for cognitive distance and its influence on diversification performance.

Another limitation is that our information is based on both Swedish and German data. We therefore assumed that skill-relatedness in Sweden is similar in Germany, which might not be the case. For further research, a selection of Swedish joint-ventures and their in-output relatedness might exhibit the same results with more reliability. In further research we would encourage the use of control variables as industry size or performance, since this might attract diversification moves of other firms in their research for profits.

A final improvement would to use in-output relatedness based on a 4-digit level, although this is not available, this would again improve the reliability of our results. At the moment, actual partner choice in industry 3611 (Manufacturing of chairs and seats) has the same vertical relatedness as the alternative industry 3640 (Manufacturing of sports goods). Their might will be a difference between these industries, as found for many 4-digit skill-relatedness industries. Nevertheless, our empirical results found support for our claim that skill-relatedness is better in explaining partner choice in joint-ventures than vertical relatedness.


More from UK Essays