International Joint Venture Survival

(1)

International Joint Venture Survival

An Empirical Study on the Effects of Multi-Dimensional Distances Between Partners on Joint Venture Duration

By M.W. van de Ridder

Date: 22-06-2011 Supervisor: dr. G. de Jong

(2)

ABSTRACT

(3)

1. INTRODUCTION

In time, the organization of productive processes has seen many forms. International Strategic Alliances (ISAs) are a form in which separate businesses from different nationalities commit resources to achieve a better outcome than they would by operating at arm’s length. The rationale of cooperating in ISAs instead of operating as individual firms can be found in overcoming changes in national and global market conditions, changes in consumer demands, and lack of resource capabilities (Culpan, 2009). Despite these advantages, the success rate of ISAs shows that there are disadvantages to this form of organizing as well. According to Hofstede, G.J. (2010), the average rate of success lies near 50 percent. Choosing for cooperation with partners from different nations presents both opportunities for success as well as failure. Differences between partners may inhibit strategic alliances from attaining their full potential or could even lead to termination of the alliance.

(6)

problem of dealing with the unfamiliar altogether. The success of an IJV, in this light, could then be said to fundamentally depend on the differences between the partners that are involved in an IJV.

When we view these differences as a form of distance observed between the partners, the goal of this study is to analyze the extent to which these distances affect the lifetime duration, or the probability of survival, of IJVs. The research question dealt with in this study therefore is:

To What Extent does the probability of Survival of International Joint Ventures depend on Cultural, Economic and Physical Distances between its Partners?

The primary aim of this research is to add to the field of research concerning IJV’s, by focusing on the influence of distance between alliance partners, measured by Cultural, Economic and Physical distance. Secondly, the preparations necessary for the empirical analysis requires several additions to the database developed by the efforts of Jagersma & Bell (1992), Boersma (1999) and Kleijn (2008). The changes in the database will allow for time-varying variables for future research. Thirdly, the results of this research are expected to have important consequences for managers in IJVs.

(7)

2. THEORETICAL FOUNDATION

2.1 Definition of Concepts

Before a discussion of the literature on the different dimensions of distance between alliance partners can take place, several concepts relevant to this study need to be identified and defined. This section will define those concepts that are relevant to this study.

International Strategic Alliances

Throughout this dissertation, the concept of International Strategic Alliances (ISAs), is used to refer to a form of organization of productive processes by partners originating from different nations. By definition, an alliance exists of at least two partners. An alliance is set up by its members because they believe they can gain more by cooperating or collaborating than they would by operating individually. Strategic alliances often involve technology transfer, e.g. access to knowledge and expertise, economic specialization, shared expenses and shared risk (Aaker, 1984; Mowery, Oxley & Silverman, 1996; Pitts & Lei, 2006; Wu, Ho & Chang, 2009)

Gulati (1998) defined a strategic alliance as a group of firms entering into voluntary arrangements that involve exchange, sharing, or co-development of products, technologies, or services. Furthermore, Culpan (2002, p.37) noted two distinctive properties of a strategic alliance: long-term commitment by partners and contribution to strategic performance of partnering firm(s) to distinguish it from other types of business arrangements (i.e., market transactions). Similarly, Hitt, Ireland & Hoskisson (2005, p.271) defined a strategic alliance as “a cooperative strategy in which firms combine some of their resources and capabilities to create a competitive advantage”.

International Joint Ventures

(8)

operating on a larger scale, gaining access to assets of a group of individuals that complement each other, sharing the risks and costs that individuals face and shaping the scope of activities performed by a group to form a basis of competition (Koh & Venkatraman, 1991). These benefits of organizing productive processes can be said to lead to two main reasons for the formation of IJVs: (1) to gain market access and (2) to jointly perform R&D (ibid.). The authors argue that the latter can be further split up into faster access to new technologies, faster access to new markets, benefits from economies of scale in joint research, sources of knowledge outside of the firm and sharing risks that are beyond the scope of a single organization. If these goals can only be achieved or achieved better when firms cooperate, then the rationale for the existence of IJVs becomes evident. One way to look at this is by seeing the firm as an individual and the IJV as a firm: roughly the same arguments hold for organizing the productive processes of the first into to the latter.

The concept of IJV typically refers to a creation of a new business entity by two or more parent firms through investing equities and assigning members to the board of directors of this new firm. Following the definition of Hitt et al. (2005), the definition of the concept of IJVs that is used throughout this thesis will identify a joint venture when “two or more firms create a legally independent company to share some of their resources and capabilities to develop a competitive advantage” (Hitt et al., 2005, p.271).

Culture

(9)

Distance, (2), Individualism, (3) Masculinity, (4) Uncertainty Avoidance Index, and (5) Long Term Orientation, are discussed more thoroughly in the Literature Review section of this dissertation, after which their hypothesized effects on duration are formulated.

Duration

The concept of duration used in this thesis will refer to the number of years between the start-up and termination of an IJV or the number of years that the IJV was observed to have survived within the observation period.

Although duration is often used as a performance measure for IJVs (Harrigan, 1986; Kogut, 1988b; Geringer & Hébert, 1991; Barkema, Shenkar, Vermeulen & Bell, 1997; Koh & Venkatraman, 1991) it will not be considered as such in this study. The reasons for this are as follows. First, the effectiveness and validity of considering duration as a measure of IJV success depends on what the intended length of the alliance is. Gulati (1998, p.307) states that “many successful alliances terminate because they are predestined to do so by the parent firms at the very outset”. Furthermore, in some cases the IJV is merely a transitional form of arrangement that the partners plan to dissolve as soon as the IJV partners have met their objectives or new information becomes available that renders staying in the IJV less appealing than outside options. Such outside options might include an acquisition or divesture of the IJV (Kogut, 1991; Bleeke & Ernst, 1991; Balakrishnan and Koza, 1993).

(10)

2.2 Theory and Hypotheses

2.2.1 Cultural Distance

As defined in the previous section, the concept of culture proposed by Hofstede can be broken down into 5 dimensions (Hofstede, 2001). A discussion on each dimension, along with their hypothesized effect on the probability of survival is discussed in the following paragraphs.1 As Shenkar (2001) states: “At the strategic phase, cultural differences may be a basis for synergy while at the operational phase they may erode the applicability of the parent's competencies.” In other words, it might be beneficial for IJV partners to select partners that have different culturally determined perspectives, while at the same time those differences in perspectives might be detrimental to the actual operation of the IJV. For each cultural dimension described below, this trade-off is incorporated in their hypothesized effects on the probability of survival of the IJV.

Power Distance

The power distance is measured by the extent to which less powerful members accept and expect that power will be distributed unequally in institutions and society (Hofstede, 2001). The mechanism through which power distance could affect the duration of an IJV is the difficulty that subordinates face in approaching their bosses (ibid.). A large difference between partners with respect to this culturally determined distance could lead to irreconcilable frictions between hierarchical levels within the IJV and may shorten the duration of the IJV. Still, it is conceivable that behavior on the work floor with respect to hierarchy is affected by the experience gained in previous settings, such as family situations or schools (Hofstede, 2001). It is not farfetched to expect that great differences in expectations about the hierarchal relationships within an IJV can create irreconcilable frictions between and within hierarchal levels, as those expectations are culturally determined and rooted deep in the projections of employees (ibid.). On the other hand, friction can also be expected

(11)

to occur in IJVs where both partners come from cultures wherein the power distances are similar. In the case where there are two IJV partners, both rooted in cultures that have large hierarchical distances, there most likely will be a struggle for leadership. As Hofstede, G.J. (2010) states: “There is always one partner that is supposed to give in to the other”. Starting an IJV implicitly means starting off with two bosses, and the struggle for control might cause irreconcilable friction and even termination of the IJV. In case both partners come from cultures with small hierarchical distances, negotiating the balance of control and establishing a leader might prove easier, as sharing authority in the context of an alliance is less seen as a socially unacceptable action Hofstede, G.J. (2010). However, even in the case of both IJV partners coming from small power distance cultures, there is a downside of their similarity. Following Brewster (1995) and O’Grady and Lane (1996), the adjustment process between partners that are similar might be just as difficult as adjustment to dissimilar partners. This is because subtle differences in culturally determined hierarchical distances are not anticipated when partners are relatively similar (Hofstede, G.J., 2010). This in turn makes the IJV vulnerable to friction. The ideal difference in Power Distance would than be an intermediate one. This leads to the following hypothesized effect of Power Distance:

Hypothesis 1a: Power Distance Difference has an Inverse U-shaped Effect on the probability of Survival of IJVs, with the optimal value in the middle of the range.

Individualism

(12)

which is both unique and whole, and by an emphasis on the individual's interests and attributes, rather than on the interest of the group(s) to which individuals may belong. Conversely, in a collectivist culture, a person is seen as whole only when considered in terms of a group affiliation (Harrison, 1993). Barkema & Vermeulen (1997) and Schut and Frederikslust (2004) find that the Netherlands has one of the highest scores on the dimension of Individualism. The irreconcilable friction between members of IJV Partners that greatly differ in the degree of individualism could affect the duration of IJVs. The way in which the corporate culture of the IJV promotes individualistic or collectivistic values is determined by the national culture of the IJV Partners. Moreover, the amount of friction between members of the IJV Partners caused by a difference in the culturally determined degree of individualism could culminate in the termination of the IJV (Hofstede, 2001). On the other hand, a slight difference in this dimension could hold benefits for the IJV, more so than were there no difference at all. The social cohesion of the newly formed IJV could be strengthened if one partner has a slightly more collectivistic perspective. The individual efforts of employees could be boosted when one partner has a slightly more individualistic approach. In this case, there could be synergy between partners that have slightly different culturally determined perspectives on the degree of individualism. The effects of distance between partners measured by the cultural dimension of individualism are hypothesized as follows:

Hypothesis 1b: Individualism Difference has an Inverse U-shaped Effect on the probability of Survival of IJVs, with the optimal value in the middle of the range.

Masculinity

(13)

feminine cultures, in which modesty and caring are the more likely predominant values (Hofstede, 2001). Great differences with respect to the degree of masculinity of the employees of IJVs, as a result of the influences of the culturally determined environment the IJVs are exposed to and influenced by, can result in frictions in the operation of the IJV. On the other hand, an intermediate difference in the degree of masculinity of partners could be beneficial for the IJV. First, as mentioned before, the adjustment to a relatively similar culture can be just as hard as the adjustment to a dissimilar one. Second, there rises a potential for synergy between partners that have a slight difference in the degree of masculinity. Just like in the case of an intermediate difference in power distance perspectives, the social cohesion that is promoted more by partners from feminine cultures combined with the drive for success and competition that is promoted more by partners from more masculine cultures could lead to a stronger IJV. Combining the expected friction between partners caused by a difference in the cultural dimension of Masculinity with the expected benefits for partners with an intermediate difference, the hypothesized role of Masculinity is formulated as follows:

Hypothesis 1c: Masculinity Difference has an Inverse U-shaped Effect on the probability of survival of the IJV, with the optimal value in the middle of the range.

Uncertainty Avoidance Index

(14)

are more tolerant of opinions different from what they are used to and try to have as few rules as possible (ibid.). Logically, IJVs that comprise partners that are very different with respect to uncertainty avoidance are likely to exhibit a different corporate culture and different degrees of bureaucracy and formalization. This may cause frictions between individuals that have different culturally determined uncertainty avoidance preferences, which could affect the duration of the IJV. On the other hand, as markets in which IJVs operate can change constantly, so does the degree of uncertainty that needs to be dealt with by the IJV. It might be beneficial for IJV partners to differ slightly in their perspective towards uncertainty. First, the same argument that holds for all cultural dimensions holds here as well: the adjustment to a relative similar culture can be just as hard as adjusting to a relatively dissimilar one. Furthermore, an intermediate difference between IJV partners in the degree of uncertainty avoidance might protect the IJV from unnecessary risk-taking by partners from an uncertainty accepting culture, as the uncertainty-avoiding partner will push for more rules and regulations. It may also boost the IJV’s performance as the uncertainty accepting partner will push for taking risks that the uncertainty-avoiding partner would not have taken, making the IJV more competitive. In theory, an intermediate difference in the degree of uncertainty avoidance exhibited by IJV partners may lead to a more balanced company. This leads to the following hypothesis:

Hypothesis 1d: Uncertainty Avoidance Difference has an Inverse U-shaped Effect on the probability of survival of the IJV, with the optimal value in the middle of the range.

Long Term Orientation

(15)

With respect to the duration of IJVs, the background of partners might create conflicts on the basis of their culturally determined orientation towards life. If a culture is characterized by a short-term orientation towards life, then partners from that culture most likely will put an emphasis on finding results that fit into their expectations of life based on the past and the present, while long term oriented partners will focus on the future.

One direct implication of a difference in orientations could be found in the role of time horizons, as discussed by Zaheer & Harris (2006). The importance of prior ties, a shared past, reputation, whether emphasis lies on achieving short or long term goals and even the definition of short or long terms are captured by the role of time horizons in IJVS. The role of time horizons in the choices made in managing an IJV could be argued to stem from culturally determined perspectives of IJV partners.

The different culturally determined orientations about life, which come together in the interactions in the lifetime of the IJV, are expected to affect the duration of the IJV in the following manner.

Hypothesis 1e: Long Term Orientation Difference has a Negative Effect on the probability of survival of the IJV.

2.2.2 Economic Distance

Besides culture, other dimensions of distance may affect the duration of IJVs as well. The state of the economy of the originating countries of the IJV partners might change the expectations partners have about each other, and of the IJV relationship, during the course of the IJV, which in turn might influence the duration of the IJV.

(16)

spillovers to occur, using this line of reasoning results in the expectation that economic distance has an influence on the spillovers between alliance partners.

Another argument in favor of including the economic background of alliance partners is that, in the course of time, changes in this type of distance can affect the benefits partners derive from staying in the IJV. More specifically, the state of the home economy of partners is expected to affect the importance of the benefits derived from the IJV.

Considering what this means in the case in which the partners’ commitment of resources to the IJV does not change, but the importance of the returns on those commitments does: this might have consequences for the fit between the partners, as the importance of the IJV to the partners has shifted (Douma, Bildebeek, Idenburg & Looise, 2000). Furthermore, if the committed resources are highly relationship specific, the chances of a hold up problem might increase (Ariño & Reuer, 2004). To see how hold up might work in this case, suppose that one of the IJV partners derives bargaining power from the specificity of the committed resources by the rest of the alliance. The committed resources are less valuable outside of the alliance because they are relationship specific. A change for the worse in the national economic environment of a specific partner that has bargaining power derived from the specific nature of the committed resources over the rest of the IJV might induce incentives for contract renegotiations. From the perspective of a partner that sees the national economic conditions faced by the other partner or partners taking a turn for the worst, the fear of contract renegotiations, i.e. being held up, might increase the chances on suboptimal commitment of resources or termination of the IJV. It is not farfetched to come to the expectation that the possible issues caused by a change in the economic background of the alliance partners affect the duration of IJVs. It is for these reasons that the hypothesized effect of economic distance between IJV partners is as follows.

(17)

2.2.3 Physical Distance

The previously described dimensions of distance all pertain to culturally and economic determined factors. However, the physical distance between IJV partners cannot be ignored. The greater this distance, the higher the costs involved in moving physical resources from partners to the IJV. Even when no physical resources are involved in the IJV structure, the physical distance between partners can still be a factor of influence on the operationalization of the IJV through the transaction costs associated with time zone differences. For the purpose of clarity, the seemingly related elements of physical distance and their hypothesized effect on IJV duration will be discussed separately in this paragraph.

Geographic Distance

The transaction costs associated with geographic costs are straightforward. The farther two economic agents are removed from each other, the higher the cost of trade. Measuring the effects of geographical distance on trade often requires the use of the gravity theory of trade, which states that international trade is positively influenced by economic size of a country, but negatively affected by the physical distance between countries (Ghemawat, 2001). Using a gravity model for international trade, Frankel & Rose (2000) found that a 1 percent increase in physical distance was associated with a 1.1 percent decline in international trade. One of the benefits of IJV might be to eliminate the effects of the geographic distance. However, the perceived distance between partners might outweigh these benefits as the geographic distance increases. Therefore, the following effect of geographic distance on the probability of IJV survival is hypothesized:

Hypothesis 3a: The Geographical Distance between IJV partners has a negative effect on the probability of survival of the IJV

Time Zones

(18)

in time zone differences also increases complexity and costs of communicating in real time.

With the use of a panel data analysis of the evolution of the time zone effect over time, Stein & Daude also found that Time Zone Difference effects are increasing. In other words, the findings of their study suggest that time zone differences have become an increasingly negative effect on the location of FDI and international trade. With respect to IJVs, the hypothesized effect of time zone differences is as follows:

Hypothesis 3b: Distance between Alliance Partners measured by the Time-Zone Difference has a Negative Effect on Duration

3. METHODOLOGY

The purpose of this study is to analyze the effects of several types of distances between partners on the probability of survival of IJVs. This makes the choice for a survival analysis a logical step, since that method is designed for the analysis of time until an event or time between events (Helsen & Schmittlein, 1993).

The Cox (1972) proportional hazards model is a multivariate technique for analyzing the effect of variables on survival. One of its advantages is that it does not require a specification of the survivor function, and yet the effects of the covariates in the model are parameterized to alter the case for which all covariates are zero, which is called the baseline survivor function (Tabachnick & Fidell, 2007; Cleves, Gould, Gutierrez & Marchenko, 2008). In the case of the Cox (1972) proportional hazards regression model, the covariates are assumed to multiplicatively shift the baseline survivor function. To understand how this works, consider the following equation:

!

h(t | x

_j

) = h

0

(t)exp(x

j

ß

x

)

(3.1)

(19)

example, 1/day then after a day we would expect a failure. The baseline hazard function, h0, is a function of the hazard that every subject faces, modified by xj. This

model is proportionate because the hazard that subject j faces is multiplicatively proportional to the baseline hazard. The functional form exp() is chosen to prevent the hazard for subject j from turning negative. One advantage of this model is that h0 can

be left unestimated. In other words, the model makes no assumptions about the shape of the hazard over time. What is assumed is that, whatever the general shape of the baseline hazard, it is the same for everyone. One subject’s hazard is a multiplicative replica of another’s (Cleves et al., 2008). For the interpretation of the hazard rate for the purpose of IJV survival, consider that the higher the rate at which terminations are expected to occur, the higher the probability that eventually the IJV will succumb to this hazard. So in a way, the probability that an IJV will survive and the probability that the IJV will succumb to the hazard it faces are two sides of the same coin. Therefore, although the Cox (1972) proportional hazards model has the hazard ratio for subject j as its dependent variable, the analysis can still result in a better understanding of the probability of survival.

Another advantage of the Cox (1972) model is that allows continuous covariates in the analysis, as well as non-linear effects of those covariates. This suits the needs of this study well. Furthermore, others have used this technique for similar studies on IJV survival (e.g. Blodgett, 1992; Barkema & Vermeulen, 1997; Lu & Hébert, 2005; Lu & Xu, 2007, Kleijn, 2008; Meschi & Riccio, 2008).

Stratification

Based on the fact that Kleijn (2008), also using the Cox (1972) proportional hazards model on the same database, found that the two control variables ‘industrialization’ and ‘firm size’ mattered the most in his regressions (i.e. had the most significant influences on the duration of the IJV), these dummies are selected for stratification of the analysis. The reason for stratification is as follows.

(20)

!

h(t | x

_j

) = h

₀₁

(t)exp(x

_j

ß

_x

)

, if subject j is in group 1,

!

h(t | x

_j

) = h

₀₂

(t)exp(x

_j

ß

_x

)

, if subject j is in group 2, (3.2)

!

h(t | x

_j

) = h

₀₃

(t)exp(x

_j

ß

_x

)

!

h(t | x

_j

) = h

₀₄

(t)exp(x

_j

ß

_x

)

In this case, the baseline hazard is allowed to differ according to which group the subject j belongs, but the coefficients, ßx are constrained to be the same. The groups

are formed by the stratifying dummies IND and SIZE, which represent the values for ‘industrialization’ and ‘firm size’ from Kleijn’s (2008) database, respectively. See Chapter 4: Data Description for a description of the stratification process.

Tied Failures

The Cox (1972) proportional hazards regression is in fact a series of conditional logistic analyses for which the regression coefficients are constrained to be the same in each analysis and for which we know that in each analysis, only one observation fails (Cleves et al., 2008). This requires an ordering of the failure times of the subjects, IJVs in the case of this study, and the consequence is that after each subsequent analysis, there would be one less observation in the following analysis: the already terminated subject. Depending on the chosen analysis time, which are years in this study, there can be multiple failures after the same amount of analysis time. These are referred to as tied failures. Dealing with tied failures means dealing with the question of how we can calculate the probability that multiple subjects fail at what appears to be the exact same time (Cleves et al., 2008). Regarding this study, this question can be rephrased as follows: how can we calculate the probability that multiple IJVs are terminated after the same number of years? The IJVs are more likely to have failed at different times within the same year and the problem actually is that the order in which they did is unknown. However, there are several methods for approximating the possibility of failure when failures are tied.

(21)

available when using the exact methods. Moreover, the Efron (1977) approximation comes very close to that of the more precise exact methods in a comparison of all four approximations (Breslow, Efron, Exact Marginal, and Exact Partial) on all models in this study, which can be found in Appendix A2.

Censoring

One of the problems with survival analysis is the loss of a subject during the observation period, without the failure event happening. In this study, only right-censored observations that have survived up until the end of the observation period are used, and they might pose a problem for the efficiency of the empirical analysis. However, since the censoring only occurs at the beginning and end of the observation period, delayed entry and right censoring due to termination of the observation period, the censored cases are still informative for this study (Allison, 1995). This type of censoring is easily dealt by most statistical software packages2, and does not harm the results of the empirical analysis (Allison, 1995; Cleves et al., 2008)

Models

Using the variables in Table 3.1, four models are analyzed. As a first step, the regression results using only linear effects for the Cultural Distance dimensions are presented in Table 5.1.

In the second step, the results for the models, which incorporate the hypothesized inverse U-shaped effects, are presented. In the first model, the effect of the Cultural Distance measured by the first four Hofstede dimensions will be analyzed, without stratification. Long Term orientation is left out of these models because of its interference with the model and insignificance of its effect after a series of both forward and backward stepwise selection tests. In the second model, the analysis of the first model is extended by stratifying by IND and SIZE, representing the state of industrialization of the country in which the IJV is located and the size of the partners.

(22)

The nature of these stratification variables, as well as their sources is described in section 4.1. In the third model the Cultural Distance variable Long Term Orientation, the Economic Distance variable ED and the Physical Distance variables Time zone Difference and Geographic Distance will be added. The fourth model extends the third model by stratifying the analysis by IND and SIZE.

Table 3.1 Variables in the Cox (1972) Proportional Hazards Model

Variable Name Description Prediction

Dependent

Hazard Rate The ratio of the hazard rate of subject j to the baseline _{hazard rate.}

Independent Variables

Cultural Distance PDI

Cultural Distance between the Dutch and Non-Dutch firm measured by the Hofstede dimension of Power Distance

Inverse U-shaped effect IDV

Cultural Distance between the Dutch and Non-Dutch firm measured by the Hofstede dimension of Individuality

Inverse U-shaped effect MAS Cultural Distance between the Dutch and Non-Dutch firm measured by the Hofstede dimension of

Masculinity

Inverse U-shaped effect UAI Cultural Distance between the Dutch and Non-Dutch firm measured by the Hofstede dimension of

Uncertainty Avoidance

Inverse U-shaped effect LTO

Cultural Distance between the Dutch and Non-Dutch firm measured by the Hofstede dimension of Long Term Orientation

Negative effect Economic Distance

ED Economic Distance between the Dutch and Non-Dutch firm measured by the difference in GDP per capita in nominal US dollars

Negative effect Physical Distance

TZD Physical Distance between the Dutch and Non-Dutch firm measured by the Time Zone Difference between the partners’ nations’ capitals

Negative effect GD Physical Distance between the Dutch and Non-Dutch _{firm measured by the Geodesic Distance} Negative effect Stratification by:

SIZE Dummy Variable for which 0 means both partners are _{small and 1 means that at least one is large} IND Dummy Variable for which 0 means that the IJV is located in an industrialized country and 1 means it is

(23)

4. DATA DESCRIPTION

4.1 Sources

This section contains a description of the sources and nature of the data used for the empirical analysis in this study. The data partly comes from a contribution from Kleijn (2008) to the existing datasets of Jagerma & Bell (1992) and Boersma (1999) and partly from my own contributions to the dataset of Kleijn (2008).

Duration & Cultural Distance

The data on duration of IJVs comes from the dataset of Jagersma & Bell (1992). Boersma (1999) and Kleijn (2008). The data is the fruition of their efforts in analyzing articles in ‘Het Financiële Dagblad’ reporting the start-ups and terminations of IJVs with at least one Dutch and one non-Dutch partner. The original dataset of Jagersma & Bell (1992) did not include data on terminations of IJVs. Kleijn (2008) expanded the original database by adding termination information on the IJVs over the period 1990-2006. Furthermore, Kleijn updated the values for the Hofstede dimensions of cultural distance.

In this study, the empirical analysis of two of the four models incorporates the values of the Hofstede dimension Long Term orientation in the Kleijn (2008) database. Furthermore, the dataset used in Kleijn (2008) is expanded to include time-varying covariates. The transformation process is explained in Appendix A1.

Economic Distance

For economic distance, the difference in GDP per capita between the Dutch partner and the first partner in nominal US dollar is used. The source of the data is the IMF economic outlook retrieved from the IMF Data Mapper. (IMF)

Physical Distance

(24)

That database contains numerous variables relating to distance and trade. The bilateral distance is used in this study. Because the names of non-Dutch partners are not included in the database of Kleijn (2008) it is impossible to know the location of the partners’ headquarters and consequently the precise distance between the partners is unknown. Because the nationality of the largest partner is sufficiently represented in the database, the geodesic distance between the Dutch and the non-Dutch partner can be used to approach the actual physical distance between the IJV partners. To improve the approximation of the real distance, a weighted bilateral distance measure is used. The distance between two countries is calculated based on bilateral distances between the largest cities of those two countries, those inter-city distances being weighted by the share of the city on the overall country’s population. The distance formula used is a generalized mean of city-to-city bilateral distances developed by Head and Meyer (2002).

The difference in time zones is also based on the location of the Dutch and largest partner’s nation’s capital. The source of the data is the Greenwich Mean Time web page (GMT).

Stratification: Industrialization and Firm Size

In Kleijn (2008) the control dummy ‘industrialization’ is used to indicate whether the IJV is located in an industrialized country or not. Kleijn states that the original sources of this dummy are ‘Het Financiéle Dagblad’ and Margreet Boersma. Kleijn also uses the control dummy ‘firm size’, which is used to indicate whether partners are small or at least one is large. The criterion for a large firm is that the firm must be in the Fortune 500. The original sources for this dummy are ‘het Financiéle Dagblad’ and Fortune. For this study, the values for both ‘industrialization’ and ‘firm size’ are taken from the Kleijn (2008) database and used to stratify the analysis.

(25)

Table 4.1 Measures of Variables

4.2 Descriptive Statistics

A summary of key descriptive statistics can be found in table 4.1 All but the dummy variables are continuous variables. Noticeable are the great range of Masculinity in the sample, the missing values of the Long-Term orientation dimension, and the great range in GDP per capita. For all but Masculinity, the means of the Cultural Distance dimensions lie much lower than the center of their range.

Measured by Measured _as/in

Cultural

Power Distance Absolute difference between Partners Index Individualism Absolute difference between Partners Index Masculinity Absolute difference between Partners Index Uncertainty Avoidance Absolute difference between Partners Index Long-term Orientation Absolute difference between Partners Index Economic

GDP per Capita Absolute difference between Partners Nominal US _Dollars Physical

Geographic Absolute difference between Partners Kilometers Time-Zone Absolute difference between Partners Hours

Industrialized Dummy

(26)

Table 4.1 Key Descriptive Statistics of Variables

Source: Kleijn (2008) database

N Range Mean Std. Deviation Data on Distances Cultural Power Distance 720 0-66 19,09 16,59 Individualism 720 0-68 27,11 21,62 Masculinity 720 0-96 44,38 16,63 Uncertainty Avoidance 720 0-59 20,78 12,43 Long-term Orientation 542 0-74 28,93 22,09 Economic GDP per Capita 695 0 – 48.608 13.024,2 11.546,55 Physical Geographic 680 0 – 18.384 4976,83 4049,52 Time-Zone 666 0 -11 3,733 3,42 Industrialized (Dummy) 0;1 0,386 0,487 Size (Dummy) 0;1 0,4941 0,500 Data on Survival Year of Start-Up 697 1985-2006 1995 5,37

(27)

5. RESULTS

5.1 Survival Analysis Results

The first step of the analyses of the four models is presented in Table 5.1; here only the linear terms are used in the regression. The columns show the name and category of the variables, the hypothesized sign of the effects of those variables, and the estimated coefficients for each of the models. The estimated coefficients are the log hazard ratios.

Table 5.1 Regression Results Using Only Linear Effects of Cultural Distance

Sign M1 M2 M3 M4 Cultural distance PDI + -0.0002 (0.011) 0.016 (0.014) -0.034 (0.044) 0.010 (0.033) IDV + -0.016** _(0.008) -0.014* _(0.008) -0.023 _(0.020) -0.024 _(0.021) MAS + 0.009 0.006 0.016 (0.007) 0.003 (0.012) 0.016 (0.012) UAI + 0.00008 (0.010) -0.008 (0.012) 0.013 (0.022) -0.002 (0.019) LTO + 0.010 (0.016) -0.006 (0.014) Economic Distance ED + 0.00002 (0.00003) 0.00004 (0.00004) Physical Distance TZD + 0.112 _(0.140) 0.019 _(0.155) GD + -0.0001 (0.0001) -0.00008 (0.0001) Stratification by:

Size No Yes No Yes

Ind No Yes No Yes

Model Statistics

-2log likelihood 1347.49 798.52 624.61 399.38

P > chi-squared 0.0108 0.1342 0.0322 0.2476

Subjects (by id) 702 702 470 470

Terminations 110 110 54 54

(28)

In M1, only the Individuality Difference’s estimated coefficient is significant, although its sign is opposite of the hypothesized one. The P > chi square of M4 indicates that this model is not very well specified in this manner.

The second step is to examine the validity of the hypothesized inverse U-shaped effects of the first four Cultural Distance dimensions. This is done by employing the quadratic terms in the models. The key results of the Cox (1982) proportional hazards regression analysis are as follows. Hypothesis 1a states that the Cultural Distance measured by Hofstede’s Power Distance has an inverse U-shaped effect on the probability of survival. Models 1, 3 and 4 support this part of the hypothesis. Both terms of the effect, PDI and PDI2, are significant at the 5% level. However, the signs from estimation are exactly opposite from the hypothesized signs. The quadratic relationship has a minimum within the range of PDI in all models. This means that the effect of Power Distance on the probability of survival turns within the range, decreasing at first and increasing thereafter. In terms of its effect on the probability of survival, the other side of the coin of the hazard rate, this means that the Power Distance Difference has less influence around a score of 30 than elsewhere.

Hypothesis 1b states that the Cultural Distance measured by the Hofstede dimension Individuality has an inverse U-shaped effect on the probability of survival. Although the signs are as they were hypothesized, the estimates of the regression analysis are not significant. Therefore, hypothesis 1b is rejected.

(29)

Table 5.2 Cox (1972) Proportional Hazards Regression Results for all Models Sign M1 M2 M3 M4 Cultural distance PDI + -0,059** (0,030) -0,052 (0,035) -0,392** (0,147) -0,438** (0,203) PDI2 - 0,001** (0,0005) 0,001** (0,0005) 0,005** (0,002) 0,008** (0,003) IDV + _(0,027)0,013 _(0,034)0,025 _(0,077)0,066 _(0,132)0,140 IDV2 - -0,0004 (0,0004) -0,0005 (0,0005) -0,001 (0,001) -0,003 (0,002) MAS + 0,0530** (0,024) 0,089** (0,037) 0,073* (0,042) 0,108** (0,053) MAS2 - -0,0005** (0,0002) -0,0008** (0,0004) -0,001** (0,001) -0,002* (0,053) UAI + _(0,035)-0,007 _(0,041)0,007 _(0,104)0,010 _(0,139)0,056 UAI2 - _(0,0006)0,0006 _(0,0007)0,0003 _(0,003)0,003 _(0,004)0,002 LTO + 0,063* (0,036) 0,065 (0,052) Economic Distance ED + 4,3E-4 (3,3E-4) 6,2E-4 (4,2E-4) Physical Distance TZD + _(0,136)-0,064 _(0,155)-0,205 GD + _(1,2E-3)2E-4 _(0,0001)0,0001 Stratification by:

Size No Yes No Yes

Ind No Yes No Yes

Model Statistics

-2log likelihood 1336,86 787,23608 614,077 386,877

P > chi-squared 0,0026 0,0146 0,0069 0,0298

Subjects (by id) 702 702 470 470

Terminations 110 110 54 54

*** significant at the 1% level ** significant at the 5% level * significant at the 10% level

(30)

Hypothesis 1d states that the Cultural Distance measured by the Hofstede dimension Uncertainty Avoidance has a quadratic effect on the probability of survival. This hypothesis is fully rejected as the terms of the quadratic effect are neither significant nor do they have the expected sign. Furthermore, the extreme of the quadratic effect lies outside of the range of the Uncertainty Avoidance dimension. This means that within the range of Uncertainty Avoidance scores, there is no turning point.

Hypothesis 2 states that the Economic Distance as measured by the difference in GDP per capita between the two partners has a negative effect on the probability of survival. Because the variable is insignificant in Models 3 and 4, this hypothesis is rejected, although it does have the expected sign.

Hypothesis 3a and b state that the Physical Distance as measured by the Time Zone Difference and the Geographic Distance have negative effects on the probability of survival. Neither hypothesis is supported by the results of analysis of Models 3 and 4. The coefficient estimated for the Time Zone Difference does not have the expected sign. The coefficient for the Geographical Distance does have the expected sign.

Concerning the overall models: the high -2log likelihood statistics and a p-value lower than 0.05 for the Pearson Chi-squared test mean that the models are not wrongly specified.

5.2 Diagnostic Analyses

Tests based on re-estimation

There are several diagnostic tests available for the Cox proportional hazards model that deal with model misspecification. Cleves et al. recommend to start with a link test. This test is widely used to test the specification of Cox models (e.g. : Gold et al., 2001; Mølstad, 2007; Dussault, Ejnar Hansen, & Mikhailov, 2005). The link test is used to verify that adding variables to the model will add little or no explanatory power (Cleves et al.,2008).

(31)

estimates ßx from the standard Cox model and then estimates ß1 and ß2 from a second-round model

!

LRH = ß

1

(xˆ

ß

x

) + ß

2

(xˆ

ß

x

)

2 (5.1)

Where LRH stands for the log-relative hazard, which rearranges the exp() part from equation (3.1). Under the assumption that xßx is the correct specification, ß1 = 1 and

ß2 = 0. Thus one tests that ß2 = 0” (Cleves et al., 2008, p.198)

The relevant predictors for the models as well as the performance of the models in the link tests are described in Table 5.2.

Another way of testing the proportional-hazards assumption is by letting the analysis time interact with the covariates in the model. When the effects of these interacted variables are not different from zero, the proportional hazards assumption is verified. This is because the assumption states that the effects of the covariates do not change with time other than in the way already specified in the model (Cleves et al., 2008) The results of this analysis time interaction test (ATI) can also be found in Table 5.2.

Table 5.2 Results of Diagnostic Tests based on Re-estimation of the Models

M1 M2 M3 M4

Link Test

p-value 0.718 0,885 0,051 0,084

Performance Passed Passed Passed Passed

ATI- Test

P-value 0.859 0.557 0.9035 0.9214

Performance Passed Passed Passed Passed

Test based on Schoenfeld (1982) Residuals

(32)

and time, means that the proportional-hazards assumption does not hold. The results of the test based on Schoenfeld (1982) residuals provides no evidence that the specification of the models are in violation with the proportional-hazards assumption. The global tests of the models yield 0,8049, 0.4299, 0,5526, and 0,807 for the four models, respectively. More detailed results can be found in the table of Appendix A5.

6 CONCLUSION

The goal of this study was to examine the relationship between Cultural, Economic and Physical Distance between partners on the probability of survival of their IJV. The effects for four of the five Cultural Distance dimensions (all but Long Term Orientation) on the probability of survival are hypothesized to have an inverse U-shape, with an extreme within their ranges. For Power Distance and Masculinity this quadratic effect is found using a Cox (1972) proportional hazards regression model for the survival analysis of 752 IJVs with at least one Dutch and one non-Dutch partner over the observation period 1990-2006. These quadratic effects, with its turning point within the ranges of Masculinity supports the idea that firms face trade-offs concerning the benefits of entering an IJV and the drawbacks of having to deal with the problems of Cultural Distance. Empirical analysis on the database of 752 IJVs with one Dutch and one non-Dutch partner suggests that, concerning the Cultural Distance dimension Power Distance, the optimal point for probability of survival seems to lie around 30 (ranging from 23.62 to 36.01 in the models; see Appendix A3). The exact opposite goes for the Cultural Distance dimension Masculinity. The quadratic effect on the hazard rate suggested from the empirical analysis means that intermediate values yield the lowest probability of survival, with the worst score somewhere around 40 (ranging from 27.81 to 54.88 in the models; see Appendix A3). This means that very little difference or very much difference yield higher probability of survival than intermediate differences between partners concerning the Cultural Distance dimension Masculinity.

(33)

the Cultural Dimension Long Term Orientation was found to have a borderline significant result in only one of the four models. Thus, the hypothesis stating its effect was rejected. The Economic and Physical Distance dimensions proved insignificant in the models used in this study.

This study adds to the field of IJV survival by introducing an inverse U-shaped relationship between survival and four dimensions Cultural Distance based on theory and supported by the results of empirical analysis using a Cox (1972) proportional hazards regression on a dataset of 752 IJVs with one Dutch and one non-Dutch partner. Furthermore, it required a transformation of the existing Kleijn (2008) database, making the implementation of time-varying covariates possible for future studies.

7 Limitations & Further Research

7.1 Limitations

The limitations of this study for the larger part pertain to the nature and quality of the data used in the empirical analysis. Concerning the limiting nature of an index to proxy for a concept as intangible as culture, the Hofstede dimensions might not perfectly reflect culture as it intends to. This has consequences for any analysis using these dimensions and consequently for the results. Secondly, the quality of the data available for analysis of the Physical Distance dimensions might have influenced the results of the analysis in this study. For both geographic distance and the time zone difference between partners in an IJV, it is imperative to know the exact location of the partners within each nation. Although we know that in this database at least one partner is Dutch, and the Netherlands is a relatively small country with only one time zone, some nations, like Russia, have more than one time zone in their country. This makes a precise analysis difficult and may distort the results.

7.2 Further Research

(34)

survival. An interesting venue in the field of IJV survival analysis would be to investigate the possibilities of time-dependence of covariates in a Cox (1972) proportional hazards regression model survival analysis by introducing time-varying coefficients as well. The interpretation of the variables with significant time-varying coefficients would be that the effects of those variables change with the amount of time the IJV has existed. By updating, upgrading and adding new data to the existing database, the way could be paved for a better understanding of IJV survival.

REFERENCES

Aaker, D. (1984). Strategic Market Management. NY, NY: John Wiley & Sons, Inc. Allison, P. (1995). Survival Analysis using SAS - A Practical Guide. Cary, NC: SAS

Institute.

Arino, A., & Reuer, J. (2004). Designing and Renegotiating Strategic Alliances Contracts. Acadamy of Management Executive , 18 (3), 37-48.

Balakrishnan, S., & Koza, M. (1993). Information Asymmetry, Adverse Selection and Joint Ventures: Theory and Evidence. Journal of Economic Behavior and Organization , 20, 99-117.

Barkema, G., & Vermeulen, F. (1997). What Differences in the Cultural Backgrounds of Partners Are Detrimental for International Joint Ventures? Journal of

International Business Studies , 28.

Barkema, H., Shenkar, O., Vermeulen, F., & Bell, J. (1997). Working Abroad, Working with Others:How Firms learn to operate International Joint Ventures. The Academy of Management Journal , 40 (2), 426-442.

Bleeke, J., & Ernst, D. (1991). The Way to Win in Cross Border Alliances. Harvard Business Review , 69 (6), 127-135.

Blodgett, L. (1992). Factors in the Instability of International Joint Ventures: An Event History Analysis. Strategic Management Journal , 13 (6), 203-221. Boersma, M. (1999). Internationale Joint Ventures; een empirische analyse, deel 2.

Groningen: Unpublished paper.

Breslow, N. (1974). Covariance Analysis of Censored Survival Data. Biometrics , 30, 89-99.

(35)

Conley, T., & Ligon, E. (2002). Economic Distance, Spillovers, and Cross Country Comparisons. Journal of Economic Growth , 7, 157-187.

Cox, D. (1972). Regression Models and Life-Tables. Journal of the Royal Statistical Society, Series B , 34, 187-220.

Culpan, R. (2002). Global Business Alliances: Theory and Practise. Westport, CT: Quorum Books.

Culpan, R. (2009). A Fresh Look at Strategic Alliances: Research Issues and Future Directions (Vol. 1). Int. J. Strategic Business Alliances.

Douma, M., Bildebeek, J., Idenburg, P., & Looise, J. (2000). Strategic Alliances: Managing the Dynamics of Fit". Long Range Planning , 33, 579-598. Dussault, D., Ejnar Hansen, M., & Mikhailov, M. (2005). The Significance of

Economy in the Russian Bi-Lateral Treaty Process. Communist and Post-Communist Studies , 38 (1), 121-130.

Efron, B. (1977). The Efficiency of Cox's Likelihood Function for Censored Data. Journal of the American Statistical Association , 72, 557-565.

Frankel, J., & Rose, A. (2000). An Estimate of the Effect of Currency Unions on Trade and Growth. NBER Working Paper No. 7857.

Geringer, J., & Hébert, L. (1991). Measuring Performance of International Joint Ventures. Journal of International Business Studies , 22 (2), 249-263.

Ghemawat, P. (2001). Distance Still Matters: The Hard Reality of Global Expansion. Harvard Business Review , 79 (8), 137-147.

Gold, E., Bromberger, J., Crawford, S., Samuels, S., Greendale, G., Harlow, S., et al. (2001). Factors Associated With the age at Natural Menopauze in a Multiethnic Sample of Midlife Women. American Journal Epidemiol , 153, 865-874. Gulati, R. (1998). Alliances and Networks. Strategic Management Journal , 19 (4),

293-317.

Harrigan, K. (1986). Stragegies for Joint Ventures. Lexington, MA: Lexington Books. Harrison, G. (1993). Reliance on Accounting performance Measures in Superior

Evaluative Style - The Influence of National Culture and Personality. Accounting, Organizations and Society , 18 (4), 319-339.

(36)

Helsen, K., & Schmittlein, D. (1993). Analyzing Duration Times in Marketing: Evidence for the Effectiveness of Hazard Rate Models (Vol. 11). Marketing Science.

Hitt, M., Ireland, R., & Hoskisson, R. (2005). Strategic Management:

Competitiveness and Globalization Concepts (6th ed.). Mason, OH: Thomson, South-Western.

Hofstede, G. (1980). Culture's Consequences: International Difference in Work-Related Values. Beverly Hills, California: Sage.

Hofstede, G. (1984). National Cultures and Corporate Cultures. Belmont, California: Wadsworth.

Hofstede, G. (1989). Organising for Cultural Diversity. European Management Journal , 7, 390-397.

Hofstede, G. (1991). Cultures and Organizations: Software for the Mind. Berkshire, England: McGraw-Hill.

Hofstede, G. (2001). Culture's Consequenc: Comparing Values, Behaviors, Institutions, and Organizations across Nations (2nd ed.). Thousand Oak, California: Sage Publications, Inc.

Hofstede, G. J. (2010). Why do international alliances fail? Some insights

from culture and human social biology. In: Ulijn, J; Duysters,G. & Meijer, E. (Ed.) Strategic alliances, mergers and acquisitions: The influence of culture on successful cooperation. Northampton, MA:

Edward Elgar. Jagersma, P., & Bell, J. (1992). Internationale Joint Ventures: Een Empirische Analyse. Economisch Statistische Berichten , 77 (3884), 1064-1068. Kleijn, N. (2008).The Effect of Equity Control and National Cultural Distance on the

Performance of International Joint Ventures: An Empirical Study of

International Joint Ventures with at least one Dutch Partner. Master Thesis, Rijksuniversiteit Groningen.

Koh, J., & Venkatraman, N. (1991). Joint Venture Formations and Stock Market Reactions: an Assessment in the Intormation Technology Sector. Academy of Management Journal , 34 (4), 869-892.

(37)

Kogut, B. (1991). Joint Ventures and the Option to Expand and Acquire. Management Science , 37 (1), 19-33.

Kogut, B., & Singh, H. (1988). the Effect of National Culture on the Choice of the Entry Mode. Journal of International Business Studies , 19 (3), 411-432. Koh, J., & Venkatraman, N. (1991). Joint Venture Formations and Stock Market

reactions: an Assessment in the Information Technology Sector. Academy of Management Journal , 34 (4), 869-892.

Kumar, V., & Reinartz, W. (2000). On the Profitability of Long-Life Customers in a Non-Contractual Setting: An Empirical investigation and implications for Marketing. Journal of Marketing , 64 (4), 17-35.

Lu, J., & Hébert, L. (2005). Equity Control and the Survival of International Joint Ventures: a Contingency Approach. Journal of Business Research , 58, 736-745.

Lu, J., & Xu, D. (2007). Technological Knowledge, Product Relatedness, and Parent Control; The Effect on IJV Survival. Journal of Business Research , 60 (11), 1166-1176.

Mølstad, P. (2007). Coronary Heart Disease in Diabetics: Prognostic Implications and Results of Interventions. Scandinavian Cardiovascular Journal , 41 (6), 357-362.

Meschi, P.-X., & Riccio, E. (2008). Country Risk, National Cultural Differences between Partners and Survival of International Joint Ventures in Brazil. International Business Review , 17, 250-266.

Mowery, D., Oxley, J., & Silverman, B. (1996). Strategic Alliances and Interfere Knowledge Transfer. Strategic Management Journal , 17, 77-91.

Newman, K., & Nollen, S. (1996). Culture Congruence: the Fit Between Management Practises and National Culture. Journal of International Business Studies . Pangarkar, N. (2003). Determinants of Alliance Duration in Uncertain Environments:

The Case of the Biotechnology Sector. Long Range Planning , 36 (3), 269-284. Park, S., & Ungson, G. (1997). The Effect of National Culture, Organizational

Complementarity, and Economic Motivation on Joint Venture Dissolution. The Academy of Management Journal , 40 (2), 279-307.

(38)

Pitts, R., & Lei, D. (2006). Strategic Management: Building and Sustaining Competitive Advantage. OH: Thomson South-Western.

Schoenfeld, D. (1982). Partial Residuals for the Proportional Hazards Regression Model. Biometrika , 69, 239-241.

Schut, G., & Frederikslust, R. (2004). Shareholder Wealth Effects of Joint Venture Strategies. Multinational Finance Journal , 8 (3/4), 211-225.

Shenkar, O. (2001) ‘Cultural distance revisited: towards a more rigorous conceptualisation and measurement of cultural distance’, Journal of International Business Studies, 32 (3), 519–535.

Slocum, J., & Lei, D. (1993). Designing Global Strategic Alliances: Integrating Cultural and Economic Factors. In G. Huber, & W. Glick, Organizational Change and Redesign (pp. 295-322). New York: Oxford University Press. Stein, E., & Daude, C. (2007). Longitude Matters: Time Zones and the Location of

FDI. Journal of International Economics , 71, 96-112.

Tabachnick, B., & Fidell, L. (2007). Using Multivariate Statistics (5th ed.). Boston: Pearson/Allyn and Bacon.

Wu, T.-F., Ho, Y.-L., & Chang, Y.-H. (2009). Establishing A Strategic Alliance for Taiwanese Rural Wineries. Asian Journal of Management and Humanity Sciences , 4 (2-3), 147-160.

Yip, G. (1992). Total Global Strategy. Englewood Cliffs, NJ: Prentice Hall.

Zaheer, A., & Harris, J. (2006). Interorganizational Trust. In O. Shenkar, & J. Reuer, Handbook of Strategic Alliances (pp. 169-197). Thousand Oaks, CA: Sage. Online Resources:

CEPII. (n.d.). Geodesic Distances. From CEPII Research Center: http://www.cepii.fr/anglaisgraph/bdd/distances.htm

GMT. (n.d.). GMT: Greenwich Mean Time - World Time / Time in every Time Zone. From http://wwp.greenwichmeantime.com

(39)

APPENDIX

A1 Dataset transformation

In order to perform the empirical analysis using time-varying covariates, the dataset of Kleijn (2008) needed transformation. The dataset needs to allow for the implementation of data that changes over the life span of a subject. For convenience of the reader, this transformation is illustrated here by means of condensed illustrations of the formats of the Kleijn (2008) database and the new database that the incorporates time-varying data of the ‘Economic Distance’ variable used in this study. The values are arbitrary.

Table A1.1 Duration and Cultural Distances in the format of the Kleijn (2008) Database

Docnr. Diff. PDI IDV MAS UAI LTO

1 3 44 65 43 43 34

2 5 66 54 32 12 14

3 1 23 26 2 34 64

(40)

sensible analysis time for the new database used in this study as well. The new format is shown in Table A1.2.

Table A1.2 Format of the Transformed Database

ID Dur. Year PDI IDV MAS UAI LTO ED

1 3 1 44 65 43 43 34 3437.14 1 2 44 65 43 43 34 3510.74 1 3 44 65 43 43 34 2579.84 2 5 1 66 54 32 12 14 4342.53 2 2 66 54 32 12 14 6101.05 2 3 66 54 32 12 14 3518.26 2 4 66 54 32 12 14 3359.72 2 5 66 54 32 12 14 5097.98 3 1 1 23 26 2 34 64 283.51

Again, each IJV has data on its duration and the variables used in the study. However, the duration of the IJV is now split up into years: the analysis time used in this study. In other words, for the number of years the IJV has existed, the data now reflects the values of the covariates per year.

(41)

A2 Tied Failures: Comparison of Four Approximation Techniques

The four different models in this study are used to compare the estimates using four different approximation techniques for handling tied failures in the Cox (1972) proportional hazards regression model.

Table A2.1 Model 1

Variable Breslow Efron Exact Marginal Exact Partial

PDI -.05825474 -.05891394 -.05891833 -.05933116 PDI2 .00096171 .00097138 .00097143 .00097842 IND .01308243 .01321394 .01321423 .01321804 IND2 -.00036196 -.00036416 -.00036415 -.00036548 MAS .05248232 .05302775 .05303341 .05340288 MAS2 _-.00052312 _-.00052891 _-.00052897 _-.00053269 UAI -.00658637 -.00694994 -.006958 -.00695907 UAI2 .00060623 .00062108 .00062131 .0006227

Table A2.2 Model 2 (Stratified by IND and SIZE)

(42)

Table A2.3 Model 3

PDI -.3898655 -.3918531 -.3918752 -.3971067 PDI2 .0054105 0054411 .0054414 .0055079 IND .0659776 .0664291 .0664308 .0666041 IND2 -.001275 -.0012832 -.0012833 -.0012866 MAS .0728202 0731367 07314 .0740996 MAS2 -.0013083 -.0013149 -.0013149 -.0013305 UAI .009283 .0098319 .009837 0100867 UAI2 .0031818 .0634158 .0031887 .0032277 LTO .0632012 .0634158 .0634193 .0642507 ED .000043 .000043 .000043 .0000439 TZD -.0633764 -.0638699 -.0638748 -.0647742 GD .0000198 .00002 .00002 .0000203

Table A2.4 Model 4 (Stratified by IND and SIZE)

(43)

As can be seen in the four tables, the estimated coefficients from regressions using the different approximations for handling tied failures lie very close to each other. The Efron approximation technique clearly is the righ choice for this stuy, as its estimates lie closer to the more precise ‘exact’ methods, while still supporting Schoenfeld (1982) residuals to be retrieved after regression.

A3 Extremes of the Quadratic Effects per Model.

Type M1 M2 M3 M4

PDI min 30.32** 23.62* 36.01** 28.41**

IDV max 18.14 24.97 25.88 25.78

MAS max 50.13** 54.88* 27.81* 32.46*

UAI min 5.59 -10.35 -1.54 -11.66

** Both terms are significant at the 5% level * Only one term is significant at the 5% level

(44)

A4 Regression Results with Hazard Ratios

Summary of Survival Analysis Results with Hazard Ratios

Exp. M1 M2 M3 M4 Cultural distance PDI >1 0,943** _(0,028) 0,949 _(0,034) 0,675** _(0,147) 0,645** _(0,203) PDI2 >1 1,001** _(0,0005) 1,001** _(0,0005) 1,005** _(0,002) 1,007** _(0,003) IDV >1 1,013 _(0,027) 1,025 _(,035) 0 1,069 _(0,077) 1,150 _(0,132) IDV2 >1 1,000 (0,0004) 1,000 (0,0005) 0 ,999 (0,001) ,997 (0,002) MAS >1 1,054** _(0,025) 1,092** _(0,040) 1,076* _(0,042) 1,113** _(0,053) MAS2 >1 ,999** _(0,0003) 0,999** _(0,0004) 0,999** _(0,001) ,998* _(0,053) UAI >1 0,993 _(0,034) 1,007 _(0,041) 1,010 _(0,104) 1,057 _(0,139) UAI2 >1 1,001 _(0,0006) 1,000 _(0,0007) 1,003 _(0,003) 1,002 _(0,004) LTO >1 1,065* _(0,036) 1,066 _(0,052) Economic Distance ED >1 1,000 _(3,3E-4) 1,000 _(4,2E-4) Physical Distance TZD >1 0,938 _(0,136) 0,814 _(0,155) GD >1 1,000 _(1,2E-3) 1,000 _(0,0001) Stratification by:

Size No Yes No Yes

Ind No Yes No Yes

Model Statistics

-2log likelihood 1336.86 787.23608 614.077 386.877

P model 0.0026 0.0146 0,0069 0.0298

Subjects (by id) 702 702 470 470

Failures 110 110 54 54

(45)

International Joint Venture Survival