``The Effects of Institutional Development on Entry Mode Selection``

``The Effects of Institutional Development on Entry Mode Selection``

University of Groningen

Faculty of Economics

MSc International Economics and Business

Student: Jasper Schuring

Student ID: 1323377

E-Mail: jasper_schuring@hotmail.com

Title: ``The Effects of Institutional Development on Entry Mode Selection`` Month and Year: 13-08-2007

Supervisor: Dr. Desislava Dikova

TABLE OF CONTENT

Abstract ……… 3

Introduction ………... 4-6 Literature Review and Hypotheses………... 6-21

Entry Mode and Control………... 6

Entry Mode and Resource Commitment……….. 7

Entry Mode and Dissemination Risk……… 7

The New Institutional Economics……… 8-10

TCT……….. 10-12

Institutional Development……… 12-16

Institutional Development in CEE………... 14-16

Firm-specific Assets……… 16-21

R&D intensity……….. 17-18 International & Regional experience……….. 18-21

Methodology ………... 22-37

Data Collection ………... 22-24

Measures……….. ………... 24-30

Data Analysis ……… ………. 30-37

Results & Conclusions………. 37-38

Limitations & Recommendations……… … 39-42

Appendix ………….……… 43-55

ABSTRACT

The choice of entry mode into a foreign market has a major impact on the success of a firm’s international operations. The options we discuss in this paper include joint ventures and wholly-owned subsidiaries. Each entry mode is characterized by varying levels of control, resource commitment and dissemination risk. Several other factors, which can be classified as transaction costs, influence the choice of a foreign market entry mode. In this paper we specifically examine the independent and joint effects of institutional development and firm-specific assets (e.g. R&D) on the entry mode choice of Western European firms in Central and Eastern Europe.

INTRODUCTION

Within the context of our current trend of globalization firms are increasingly challenged by foreign market opportunities. The choice to enter such market is probably one of the crucial strategic decisions that a firm will encounter in its process. A firm’s initial entry mode choice is difficult to change without considerable loss of time and money (Root, 1987). In face of high switching costs, it’s therefore not surprising that firms and scholars perceive these entry modes choices as a critical part of the international business field and literature.

Faced with foreign market opportunities there are multiple ways in which a firm can enter such market. Options available to a firm include exporting, licensing, joint ventures and wholly-owned subsidiaries. Each entry mode has implications on the level of control that a firm will experience. Furthermore, every entry mode requires different resource commitments, which increase with the amount of control required. Both these varying levels of resource commitments and control influence the degree of risk a firm will perceive when entering a foreign market (Contractor, 1984; Gatignon and Anderson, 1998; Hennart, 1989). It is generally recognized that as a firm moves on the continuum from a non-integrated (e.g., exporting) to a fully integrated one (e.g., wholly-owned subsidiary), control, resource commitments and risk all increase (Agarwal and Ramaswami, 1992; Anderson and Gatignon, 1986; Barkema et al., 1996; Erramilli, 1991). All this suggest that the choice of foreign market entry should be based on a trade-off between risk and return (Anderson and Gatignon, 1986; Chi and McGuire, 1996). An MNE is expected to choose the entry mode that offers the highest risk-adjusted return on its investment (Hennart, 1989; Madhok, 1997).

Obviously the optimal entry mode has to be chosen in a dynamic environment of variables. All these variables are influencing the potential benefits from varying control, resource and risk levels. Broadly speaking, firms in the entry process have to deal with various transaction costs, which are for example influenced by firm-specific assets.

Williamson, 1991). Therefore, many scholars used this framework to explain choices between joint ventures and wholly-owned type of entry modes.

Although many studies in this field led to useful contributions in understanding the relations of various variables in the entry mode process, a gap in the literature is still prevalent. A particular research area that deserves more attention is the issue of how the joint-effects of predictors influence the firm’s entry mode choice. The importance of examining the joint-effects derives from the fact that they may explain firm behaviour that cannot be captured by the predictors’ main effects on the entry mode choice (Agarwal and Ramaswami, 1992). By analyzing the joint-effects we hope to get a more complete picture at the way certain predictors interact and influence the entry mode choice. In this paper we will use the transaction cost theory to specifically analyze the moderating effects of the institutional development level on firm-specific assets. Institutional development will be our main focus, since this predictor has often been mentioned as the main barrier of entry into new markets. Furthermore, the uncertainties and risks embodied in the contextual environment are usually beyond the control of the firm. The firm-specific assets, however, can still be controlled by the firm and are supposed to generate the sustainable competitive advantage in the future. Within this context it’s especially interesting to see how institutional development, which is uncontrollable, is moderating and influencing the main effects of firm-specific assets in the entry mode selection process.

Specifically, the current thesis aims to provide an answer to the following research questions:

(1). How is the institutional development level shaping the choice between joint ventures and wholly-owned type of entry modes? (2). Are firm-specific assets influencing the entry mode selection decision, and in what way? (3). What is the joint effect of the institutional development level, given the relevant firm-specific assets, on entry mode choices?

light of firm-specific assets of Western European firms we hope to get more insight on the entry mode choice of firms into CEE.

The remainder of this paper is organized into four parts. The first part provides a review of the relevant literature covering this topic including the development of several hypotheses. The second part will cover questions concerning the methodology, while the last two sections provide a conclusion followed by a discussion of several important limitations and recommendations.

LITERATURE REVIEW

As mentioned previously, the choice of a foreign market entry mode should be based on a trade-off between risks and returns. A firm is expected to choose the entry mode that offers the highest risk-adjusted return (Agarwal and Ramaswami, 1992). This return is partly dependant on the level of control, resource commitment and dissemination risk that a firm is experiencing, which varies among different types of entry modes. In much of the international business literature the main focus is on two distinct modes of entry into a foreign market; entering into a joint venture or setting up a wholly-owned subsidiary. We will now discuss the varying levels of control, resource commitment and dissemination risk for both joint ventures and wholly-owned subsidiaries.

Entry Mode and Control

Entry Mode and Resource Commitment

Each entry mode also requires different resource commitments (Vernon, 1983). An important implication is that certain dedicated assets cannot be redeployed to alternative uses without cost. These assets can be either tangible or intangible of nature. Important to note is that this resource commitment is partly dependent on the firm’s resource availability, which refers to the financial and managerial capacity of a firm for serving a particular foreign market (Agarwal and Ramaswami, 1992). When looking at a wholly-owned subsidiary, we see that the MNC bears all the responsibility and costs over the revenue generating assets when entering a foreign market. This obviously results in a high resource commitment level. The level is lower for joint ventures, because the resource commitment level is dependent on the percentage of ownership and resource sharing between the venture partners. The resource commitment level is important, because it can function as an exit barrier which limits the flexibility of the firm (Harrigan, 1981). When resource commitment levels are high a MNC can only exit the foreign market at substantial cost.

Entry Mode and Dissemination Risk

Before we can explain the interaction between transaction costs and varying control, resource commitment and dissemination risk levels for joint-ventures and wholly-owned subsidiaries, we have to show the framework in which transaction costs are being shaped. This is where ``The New Institutional Economics`` of Williamson (2002) comes into play. Below we will provide a brief description of the literature with a visual summary in Figure 1.

The New Institutional Economics

The top level is the social embeddedness level. This is where traditions, mores, norms and customs are located. Religion plays a huge role at this level. Level 1 is generally taken as given, because institutions at this level change very slowly, on the order of centuries or millennia. A lot of these informal institutions have mainly spontaneous origins. Given these evolutionary origins, they are adopted and subsequently display a great deal of inertia, some because they are functional and others because of symbolic values to true believers. The resulting institutions have therefore a long-lasting grip on the way a society conducts itself (Williamson, 2000).

(Coase, 1960). Therefore, it was needed to go beyond the rules of the game (property) to include the play of the game (contract) (Williamson, 2000).

Before we bring in and show how the institutional development level and firm-specific assets might influence the decision to enter by means of joint ventures or wholly-owned type of entry modes, we will first summarize and state the assumptions on which the transaction cost theory is based.

TCT

Williamson (1975) assumes, for expositional convenience, that at the beginning, there were markets. The existence of the firm is explained by market failures. The firm is preferred as a governance structure when the costs of carrying out certain exchange transactions in the open market are greater than organizing these transactions within the firm. A key feature of the firm is its allocation of resources through authority relations rather than a price mechanism (Coase, 1937).

T h e tw o k ey b eh av io u r al a ssu mp ti o n s o f th e T C th eo r y are o pp o r tu n is m and bounded rationality (Williamson, 1991a). The former is interpreted as self-interest seeking with guile, while the latter implies that behaviour is ``intendedly rational, but only

limitedly so`` (Simon, 1976). Therefore, all complex contracts are unavoidable incomplete

because of bounded rationality and hence are subject to hazards of opportunism. Contractual incompleteness poses added problems when paired with the capacity of conscious foresight and opportunism, which can manifests itself as moral hazard, adverse selection, sub goal pursuit, shirking and other types of strategic behaviour. An appropriate governance structure would economize on bounded rationality and safeguard transactions against opportunistic behaviour (Tsang, 2000).

From the TC perspective, the main objective of a firm is to economize on transaction cost through choosing appropriate governance structures for handling its transactions; referred to as the make-or-buy decision (Tsang, 2000).

For TC theory, the governance structure chosen is primarily determined by the extent of asset specificity involved in the transaction concerned. Asset specificity refers to ``the degree to which an asset can be redeployed to alternative uses and by alternative users without sacrifice of productive value`` (Williamson, 1988). Simply put, if the assets are highly specific to the parties of a transaction, a small-numbers exchange condition arises and leads to the potential of serious opportunism. In this case, the hierarchy is preferred to the market as a governance structure (Tsang, 2000).

Since we now have a clear view of the important characteristics of the transaction cost theory and the framework in which it is being shaped, it is time to see how the institutional development level and the firm-specific assets might influence the decision to enter by means of joint ventures or wholly-owned type of entry modes. We will first review the institutional development level in the section below.

Institutional Development

That institutions affect economic performance is hardly controversial. Yet, few Western economists understand the institutional requirements that are vital in creating efficient markets, since they simply take institutions for granted. However, entailed in efficient markets are both a set of political and economic institutions that provide for a credible commitment and low transaction costs. These institutions form the foundation of efficient factor and product markets that underlie current economic growth. Before we show how institutions can lower the cost of measuring and enforcing contracts, we will first go over some variables that determine the costliness of transacting in exchange.

There are three main variables determine the costliness of transacting in exchange. First is the cost of measuring the valuable attributes of the services and goods or the performance of agents in exchange. Measurement consists of defining the physical dimensions of the rights exchanged (weight, etc.) along with the property rights dimensions of the exchange (rights defining uses, etc.). The enormous resources that societies allocate to organizations and enforcement would be redundant in a world where measurement costs were zero. However, since such costs are exceptionally high and, as a result, the rights are imperfectly specified, the costs of transacting and the uncertainty associated with transacting rise dramatically (North, 1992).

The third variable is enforcement. In a world of perfect enforcement, ideally, a third party would impartially and costlessly evaluate disputes and give compensation to the injured party when contracts are violated. Cheating, shirking and opportunism would never pay in such an environment. However, such a world does not exist. In fact, the creation of a relatively impartial judicial system that enforces agreements has been a critical stumbling block in the path of economic development. In the Western world, the advancement of legal systems, courts and a relatively impartial body of judicial enforcement has played a major role in the development of a complex system of contracting that could extend over time and space: a critical requirement of a world of specialization (North, 1992).

Efficient markets in this case, are the result of institutions that provide for low-cost measurement and enforcement of contracts. This is accomplished by rules that encourage adaptive efficiency, which must cover bankruptcy laws and provide incentives to promote decentralized decision-making and effective competitive markets, as well as provide low-cost measurement of property rights (North, 1992). Since formal rules are only one part of the institutional framework, they must be complemented by informal constraints that can work out innumerable exchange problems not entirely covered by formal rules and that have tenacious survival ability. Their effectiveness is dependent on the effectiveness of enforcement (North, 1992). The ability to enforce agreements across space and time is, therefore, the essential foundation of efficient markets. Such enforcement would appear to be an easy requirement to fulfill. The only thing one’s needs is an effective, impartial system of laws and courts for the enforcement of formal rules, for the "correct" societal sanctions to enforce norms of behavior, and for strong normative personal standards of integrity and honesty to support self-imposed standards of behavior. The definition and enforcement of efficient property rights, however, is dependent on the polity (North, 1992).

defection rise dramatically. They rise because the gains from stealing, shirking, and cheating rise as well as the costs of monitoring and measuring performance. Let’s now see how the institutional development level in CEE, under these circumstances, is influencing the cost of measuring and enforcing contracts and the way that it is shaping the choice between joint ventures and wholly-owned type of entry modes.

Institutional Development in CEE

Central and Eastern Europe is a region where very significant institutional changes took place during the 1990s. The passage from a planned to a market oriented economy induced major reforms in domestic institutions. The abandoning of the centralized system by these transition countries, and their increasing willingness to join the EU motivated their governments to undertake large reforms at the institutional level. Institutional reforms were a chief condition imposed by the EU on new candidate countries. In order to permit the good functioning of the enlarged union, candidate states were required to adopt changes in their legislation, increase the efficiency of the judicial system and reduce corruption (Cheptea, 2007).

Unlike in the case of trade policy, however, progress that has been made in improving the functioning of CEE institutions (e.g., progress in the privatization process or in reducing corruption) has been fairly moderate and varies significantly across countries. Hence, it allows us to see the impact of different institutional development levels on entry mode selection across the various CEE transition countries.

therefore be attended by whole series of features such as provisions for information disclosure and penalties. Additional mechanisms such as the use of hierarchy to effect coordination and decide on disputes, are introduced when transactions are removed from the market and placed under unified ownership. This tendency, however, will be reduced by the development of high-quality national CEE institutions, since they have a major transaction cost reducing effect.

Consequently, CEE countries with higher institutional development levels due to institutional reforms are more likely to be able to work out the contractual hazards and thereby lower transaction cost. They are more likely to economize on bounded rationality and safeguard transactions against opportunistic behaviour. It also makes them deal better with measurement issues of valuable attributes of goods and services or the performance of agents in exchange, impersonal exchange and enforcement. Their adaptively efficient rules provide incentives for the acquisition of knowledge and learning, induce innovation, encourage risk taking and creative activity. In general, their effectiveness of enforcement is improved. Therefore, Western European firms entering these markets are more likely to prefer equity joint ventures with local CEE partners over wholly-owned subsidiaries. The dissemination cost savings will have to outweigh the higher bureaucratic cost and resource commitment levels for wholly-owned subsidiaries before these transactions are removed from the market and placed under unified ownership. Since these countries with higher institutional development levels not only have the formal rules of the game right, but also the governance structure, the costs of measuring and enforcing agreements will probably be lower which reduces the need for full control under wholly-owned subsidiaries.

H1: A higher institutional development level will be positively associated with the likelihood of entering in an equity joint venture with a local CEE partner.

Let’s now see how firm-specific assets are shaping the decision between equity joint ventures and wholly-owned subsidiaries.

Firm-specific Assets

Researchers in international business have long argued that MNEs doing business abroad face costs arising from unfamiliarity of the environment, from cultural, political, and economic differences, and from the need for coordination across geographical distance (Hymer, 1976; Kindleberger, 1969). Additional cost can arise such as: 1, cost directly associated with spatial distance (e.g., transportation and coordination); 2, firm-specific costs based on a firm’s unfamiliarity with the local environment; 3, costs resulting from the host country environment (e.g., economic nationalism and lack of legitimacy of foreign firms); 4, costs from the home country environment (e.g., sales restrictions to certain countries) (Zaheer, 1995). Whatever its source, this liability of foreignness implies that foreign firms will have a lower profitability and probability of survival compared to local firms (Zaheer, 1995). To overcome this so called ``liability of foreignness`` and compete successfully against local firms, it has been argued that MNEs should provide their subunits with some firm-specific advantage (Buckley & Casson, 1976; Caves, 1982; Dunning, 1977; Hennart, 1982). Foreign firms can only compete with host country firms in their own markets, by possessing and transferring superior assets and skills that can earn economic rents that are high enough to counter the higher cost of servicing these markets (Agarwal and Ramaswami, 1992). Throughout the international business literature; R&D intensity, international experience and regional experience have been identified as the most important firm-specific advantages that could lead to a sustainable competitive advantage and overcome the higher cost and risk associated with operating in a foreign market. We will now discuss each in turn:

R&D intensity

transaction-specific and vulnerable for opportunistic behaviour. Opportunism in this situation might arise, because a joint venture partner may try to secretly learn and internalize skills of the other partner more than what was stipulated in the joint venture agreement. Furthermore, when tacit knowledge is transferred from one firm to another, it is very hard for either party to know ex ante what the cost and the value of the transaction will be (Hennart, 1988). Because of bounded rationality, it is problematic for the parties concerned to estimate ex ante the value of the technology and to monitor ex post one another’s opportunistic behaviour (Hennart, 1988). Under this setting it is problematic to assess till what degree both parties fully performed their duties. Since the TC theory focuses on how the rent pool is distributed among firms involved, opportunistic behaviour is an attempt to obtain a larger share than that to which a firm is entitled (Tsang, 2000). TC considerations are to guard against such behaviour ex ante by choosing a suitable governance structure.

A firm unnecessarily exposing its critical resources may provide its partner firm with an advantage in the future, while in the meantime reducing its own capacity to develop a sustainable competitive advantage (Collis, 1991). Therefore, Williamson (1981) argued that skills acquired in a learning-by-doing fashion need to be embedded in a protective governance structure. Within this context wholly-owned subsidiaries will better protect a firm’s tacit knowledge and strategic resources compared to equity joint ventures with local CEE partners (Luo, 2001). Dunning (1980) argues that an internalized entry mode protects firm-specific assets better and creates a maximum return from firm-specific advantages. Preceding the discussion, we can assume that the amount and tacitness of proprietary knowledge or their perceived economic rents, will be positively associated with the need to maintain integrated control (Luo, 2001). Therefore, firms with higher levels of R&D intensity are more likely to internalize the transaction by entering with a wholly-owned subsidiary, since the potential dissemination cost savings will outweigh the higher bureaucratic cost and resource commitment level. The higher control of a wholly-owned subsidiary is therefore likely to be preferred over the flexibility of an equity joint venture with a local CEE partner. Thus: H2a: The greater the level of R&D intensity of the parent, the higher the likelihood of entering with a wholly-owned subsidiary.

have formal rules and governance structures in place, the costs of measuring and enforcing agreements would probably be lower for R&D intensive MNEs in equity joint ventures with local CEE partners. Because better safeguard measures further reduce the potential dissemination costs threat, it makes it a less likely and attractive option to maintain full control under wholly-owned subsidiaries, while incurring a higher bureaucratic cost and resource commitment level. So, if the technology can protected by a patent and the transfer process can be clearly specified in the contract, the related transaction costs will be lower and the equity joint venture is the more appropriate governance structure (Tsang, 2000). Thus: H2b: A higher institutional development level will have a positive moderating effect on the tendency of R&D intensive MNEs to enter in equity joint ventures with local CEE partners.

International and Regional experience

A firm’s level of international and regional experience has shown to influence entry mode choices. Firms without foreign market experience are likely to have greater problems in managing foreign operations. Overstating potential risks, while understating the potential returns of operating in a foreign market is common under these circumstances (Agarwal and Ramaswami, 1992). Furthermore, firms with lower levels of international and regional experience are considered less likely to have the know-how to manage subjectively and monitor appropriately (Anderson and Gatignon, 1986). It follows, therefore, that firms with little or no experience will try to limit their risk exposure (Chang, 1995). In this circumstance, selecting the joint venture mode is preferable because it not only reduces the firm’s resource commitment and risk-taking but also facilitates learning through cooperation and interaction with local firms (Barkema et al., 1996). In this context, a local partner’s country-specific knowledge is of strategic importance to foreign companies with little international and regional experience (Inkpen and Beamish, 1997).

managing inputs resulting in a more detailed and more accurate perception of foreign risks and returns (Agarwal, 1994). Finally, according to TCE, firms with more international and regional experience develop organizational capabilities that enable them to make greater commitments to a foreign investment (Johanson and Vahlne, 1977). As firms with greater experience face fewer local knowledge disadvantages, the need for local partner to ease up liabilities of foreignness decreases and the desire for full ownership increases (Dikova and Witteloostuijn, 2007).Within this context, MNEs with significant international and regional experience will prefer wholly-owned subsidiaries, since the rational for partnering with local firms is reduced (Luo, 1997) and to ensure maximum return from their own distinctive experience. Thus:

H3a: The higher the level of international experience of the parent, the stronger the tendency of entering with a wholly-owned subsidiary.

H3b: The higher the level of regional experience of the parent, the stronger the tendency of entering with a wholly-owned subsidiary.

Other similar research papers, because of the supra-national characteristic of international and regional experience, usually do not predict a moderating effect of the institutional environment on the influence that such MNEs’ experience has on subsequent entry mode choices (Dikova and Witteloostuijn, 2007). Since the parent firm with experience decides how much of its strategic assets are to be transferred to the subsidiary, only those transferred will be exposed to externalities. The firm only needs to transfer technology to be able to integrate the subsidiary within its network or to guarantee production quality (e.g. know how and related knowledge), which effect on entry modes given institutional development has been previously discussed in the R&D intensity section. Therefore, we are less concerned with asset specificity and dissemination risk levels under high international and regional experience levels, and will mainly deal with control and resource commitment levels in which institutional development could play a critical role.

environment in some CEE countries is characterized by ineffectiveness; a lack of a clear regulatory framework, a weak functioning property rights system and corruption have been mentioned as the main problems in this area. Western European MNEs with plans to enter this CEE region will, therefore, face additional costs and constraints in order to comply and survive under this ineffective climate.

Subsidiaries established in such an environment would have a hard time to survive. In general, investments in CEE are perceived as risky due to the lower development level of host market institutions, characterized by ineffectiveness and uncertainty (Meyer, 2001). In addition, Western European MNEs often lack personal networks, face unclear regulatory networks, weak intellectual and property rights regimes and widespread corruption which contributes to the costs of doing business abroad (Luo and Peng, 1999). In addition to the transaction costs associated with overcoming the liability of foreignness, joint ventures and wholly-owned subsidiaries will both face the additional cost of managing their subsidiary operations under these ineffective institutional settings. Since the transition economies are at different stages of transformation, the effects of institutional development will differ per CEE country, depending on the specific level of ineffectiveness. Consequently, its effect on ownership modes can be explored for different levels of institutional development.

assets will choose for joint ventures in countries with under-development institutions, while wholly-owned subsidiaries will be seen more often in countries with higher institutional development levels.

As mentioned previously, we now have a firm’s international and regional experience on the one hand, which leads to the development of organizational capabilities resulting into fewer local knowledge disadvantages and a reduction in the need for a local partner to ease up the liabilities of foreignness. On the other hand, institutional development lowers the need for joint venture modes to limit an MNE’s exposure by reducing its resource commitment and by increasing its ability to exit from the market quickly to avoid a substantial loss should the environment worsen. Moreover, local partners are not needed anymore for their knowledge, experience, and business networks to cultivate a better relationship with governmental authorities, which are often crucial in emerging economies with low institutional development levels (Luo, 2001). All together, a local partner in a cooperative entry mode can support stability of venture activities and reduce vulnerability towards contextual risks, which becomes superfluous under high institutional development levels (Collis, 1991; Root, 1994). Therefore, MNE’s with high experience levels will use wholly-owned subsidiaries to ensure a maximum return from their own distinctive experience, which will be further stimulated by higher institutional development levels in this CEE region. Thus:

H4a: A higher institutional development level will have a positive stimulating effect on the tendency of firms with high levels of international experience to enter with wholly-owned subsidiaries.

METHODOLOGY1

Data collection

To test the above hypotheses, data collected by Dikova (2005) was used. She conducted an international mail survey among international companies from the EU that had invested in ten transition economies in the CEE region. In compliance with the majority of empirical studies that use a stake of 10 per cent and above in a foreign enterprise as a minimum to qualify as a foreign direct investment (Benito and Gripsrud, 1992; Padmanabhan and Cho, 1999; Larimo, 2003), she initially selected from the AMADEUS dataset all registered companies based in the then fifteen member states of the EU that had at least a ten per cent ownership stake in a branch/subsidiary located in any of the following countries between 1992-2002: Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia and Slovenia. These countries were chosen for this study because they were, and still are, in various stages of the transformation from centrally planned to market economies. Therefore, this set of countries offers a unique opportunity to test the extent to which our theoretical arguments and hypotheses hold true. Additionally, as a by-product, we can evaluate to which extent our logic, which has been largely used in the context of western-style market economies, is applicable to transition economies. In total, initially 2,798 questionnaires were mailed: 35 were returned as non-deliverable, which compressed the sample size to 2763 questionnaires. She received 209 usable questionnaires, representing an overall response rate of 7.5 per cent.2 _{As a result of 55 missing data, we could use 154 of them in the current study.} To test the representativity of the sample, we collected data from the Amadeus data base on the firm size variable (number of employees worldwide) as well as available information about the percentage of ownership for the MNEs in the sample in order to compare and validate the results based on primary data with secondary sources. We initially had 209 firms in our primary data sample. In a first step, we find out whether the firms in the sample are MNEs with available subsidiary reports from 1992-2002 through AMADEUS. Firms are also not accepted if they do not provide a consolidated company report. We performed statistical tests to compare the primary data with these pieces of secondary –source information.

1 _{The methodology and data concerning this paper is partly borrowed from Dr. D. Dikova, who used it in her}

``Studies on Foreign Direct Investments in Central and Eastern Europe: Establishment Modes, Performance and Strategies of Western European Enterprises``, pp 36-44, 2005.

2 _{International mail surveys aiming at an industrial population have a history of a very low response rates:}

sample t-tests showed that the differences in means between the survey-collected information and the Amadeus data were insignificant for the employees variable (t=.914, df= 83 and

p=.362). The tests on percentage of ownership levels revealed similar results (t=1.103, df= 98

and p=.271). In addition, following Uhlenbruck and DeCastro (2000), we determined a reliability coefficient of .96 for 79 of the firms in the sample that had data in Amadeus recorded from 1992-2002 as to both variables.To obtain this coefficient, we used the general form of the Spearman- Brown prophecy formula and incorporated the correlations of .91 (employees worldwide) and .90 (percentage ownership) between the archival data and the survey information.3 _{So, all tests conducted confirm that the survey sample is representative} for the larger population.

Finally, although not all our measures are based on the single-respondents questionnaire, we also checked the primary data for potential common-method variance. Podsakoff and Organ (1996) argue that if the variables in a study all load on one factor or if there is one factor that explains the majority of the variance, then common-method variance may be a problem. A Harman’s one-factor test is a common technique to check for common-method variance and has been used in many similar research papers such as Luo (2001). In this test all variables from the questionnaire are entered into a factor analysis. When no single factor emerges or no general factor accounts for the majority of covariance in the variables, severe common-method variance is likely to be absent. We performed a factor analysis for the dependent and independent variables from the survey that were used in this study (e.g. percentage ownership, techints, mneyears, ceeyears, globempl, mktconcn and ownership restrictions). Prior to performing the principal component analysis (PCA), the suitability of data for factor analysis was assessed. Inspection of the correlation matrix revealed the presence of only one correlation coefficient above .3 (0.465). The Kaiser-Meyer-Oklin value was .424, below the recommended value of .6 (Kaiser, 1970, 1974) which means that even though the Bartlett’s Test of Sphericity (Barlett, 1954) reached statistical significance (ρ=0.000), the data showed little support of the factorability of the correlation matrix. The principal component analysis revealed the presence of three components with an eigenvalue exceeding 1, explaining 22.58,

3 _{See appendix for Paired-sample t-tests and correlations.}

Source: http://www2.chass.ncsu.edu/garson/pa765/reliab.htm Spearman-Brown prophecy coefficient:

rSB1 = (k* rij)/[1 + (k-1)* rij)] where,

rSB1 = the Spearman-Brown split-half reliability

rij = the Pearson correlation between forms i and j

17.87 and 16.1 per cent of the variance respectively.4 An inspection of the screeplot further supported our finding by revealing a clear break after the third component. Using Catell’s (1966) scree test, we decided to retain a three-factor solution with the largest factor explaining only 22.58 per cent of the variance. Therefore, it appears that the dataset based on the survey does not suffer from common-method variance. The use of archival data to measure several other variables further reduces this threat.

Measures

The dependent variable is the EU firm’s latest Ownership type choice into a CEE country, which distinguishes between an equity joint venture and a wholly-owned operation. This measure is obtained by separation from the survey data, were subsidiaries from 95 till 100 per cent ownership represent the wholly-owned entry mode and the remaining firms belonging to equity joint ventures. The ownership type was determined by the actual percentage of foreign ownership in the CEE subsidiary. So, ownership mode is captured by a dummy variable, which takes the value of 0 in the case of an equity joint venture and 1 in the case of a wholly-owned entry.

Our hypotheses relate to four independent variables. First, R&D intensity (techints) is operationalized. Primary survey data was obtained by asking the respondents a five-point Likert-type of question as to the percentage of sales spent on R&D (ranging from very low to very high), because it was believed that the surveyed sample of managers would be unlikely to answer adequately or at all questions regarding a monetary estimation of the annual R&D budget.

Second, International experience was initially a composite measure constructed by asking the respondents to indicate the number of countries and the number of years they have been doing business abroad. We performed a factor analysis to assure that the pair of items obtained for international experience converges on one factor. Prior to performing the principal component analysis (PCA), the suitability of data for factor analysis was assessed. Inspection of the correlation matrix revealed the presence of a correlation coefficient above .3 (0.611). The Kaiser-Meyer-Oklin value was .5, close to the recommended value of .6 (Kaiser, 1970, 1974) and the Bartlett’s Test of Sphericity (Barlett, 1954) reached statistical significance (ρ=0.000), supporting the factorability of the correlation matrix. The principal component analysis revealed the presence of one component with an eigenvalue exceeding 1,

explaining 80.55 per cent of the variance respectively.5 An inspection of the screeplot further supported our finding by revealing a clear break after the first component. Using Catell’s (1966) scree test, it was decided to retain one component. However, in our binomial logistic regression model the single measure indicating the number of years of doing business abroad (mneyears) performed better than the composite measure obtained from the factor analysis, we therefore decided to include the latter in the model.

Third, following the same logic, Regional experience was initially a composite measure constructed by asking the respondents to indicate the number of countries and the number of years they have been doing business in CEE. We performed a factor analysis to assure that the pair of items obtained for regional experience converges on one factor. Prior to performing the principal component analysis (PCA), the suitability of data for factor analysis was assessed. Inspection of the correlation matrix revealed the presence of a correlation coefficient above .3 (0.543). The Kaiser-Meyer-Oklin value was .5, close to the recommended value of .6 (Kaiser, 1970, 1974) and the Bartlett’s Test of Sphericity (Barlett, 1954) reached statistical significance (ρ=0.000), supporting the factorability of the correlation matrix. The principal component analysis revealed the presence of one component with an eigenvalue exceeding 1, explaining 77.16 per cent of the variance respectively.6 An inspection of the screeplot further supported our finding by revealing a clear break after the first component. Using Catell’s (1966) scree test, it was decided to retain one component. However, in our binomial logistic regression model the single measure indicating the number of years of doing business in CEE (ceeyears) again performed better than the composite measure obtained from the factor analysis, we therefore as mentioned above decided to include the latter in the model.

Fourth, Institutional development is operationalized. This secondary institutional development variable is a composite measure obtained from the World Bank’s biennial Governance Research Indicators, providing a score on items such as; voice and accountability, political stability/no violence, government effectiveness, regulatory quality, rule of law and control of corruption (-2.5= poor / 2.5= excellent; http://info.worldbank.org/governance/kkz2005/tables.asp). The choice of a source for our institutional measure was determined by two factors; extensiveness and accuracy. The World Bank’s index covers one of the broadest ranges of institutional issues, and is updated once in

two years.7 Our measure is the sum of the six separate scores; low values demonstrate institutional under-development, and high values reveal the opposite. The Cronbach alpha coefficient of 0.957 indicates internal consistency of the scale, since it is above the recommended cut-off of 0.7.8 Due to a good fit of the institutional development scale based on World Bank data (instdevelopwb) in our binomial logistic regression model, we decided to use it as our main variable for institutional development levels.

Finally, consistent with previous research, we included five control variables. First, we control for cultural differences. Previous studies argue that the greater the perceived distance between the home and host country in terms of culture, economic systems, and business practices, the more likely it is that MNC’s will shy away from direct investments in favour of licensing or joint venture agreements (Anderson and Coughlan, 1987; Davidson, 1980; Green and Cunningham, 1975; Johanson and Vahlne, 1977; Kobrin, 1983; Stopford and Wells, 1972). This is because the latter institutional modes enhance MNCs’ flexibility to withdraw from the host market should they be unable to comfortably acclimatize themselves to the unfamiliar setting (Chan Kim and Hwang, 1992). Faced with the uncertainty that arises from the unknown, a MNC may be unwilling to commit substantial resources to a foreign operation since such a commitment would substantially reduce the MNC’s ability to exit without cost if the host market should prove unattractive (C.W.L. Hill et al., 1990). In contrast to the integration costs of an acquisition, joint ventures frequently serve the purpose of reducing risks by sharing equity and assigning management tasks to local partners who may be better able to manage the labour force and relationships with suppliers, buyers, and the government in their respective countries (Root, 1987). Local firms are expected to have better knowledge of the culture, polity, and the functioning of the local market (Root, 1987), which they can supply at a lower cost than the entering firm (Beamish and Banks, 1987). Therefore, joint venture forms of entry provide an effective means of gaining local market knowledge and ease a firm’s entry into unfamiliar cultural settings, making it an attractive option when entering a cultural distant market (Dunning, 1993; Erramilli, 1991). Hence, firms from culturally distant countries will attach greater information-acquisition and integration costs to the management of acquisitions relative to joint ventures than firms from culturally similar

7 _{The World Bank’s governance indicators cover even years from 1996-2002. Therefore, all investments prior to}

1996 were linked to the estimates for 1996 (such investments represent 8% of the sample) and odd year entries were rounded up to even years.

8 _{See appendix for complete results on the Cronbach alpha coefficient for the institutional development scale}

countries. Therefore, we expect the use of acquisitions by foreign firms to be dissuaded, the more distant the culture of the country of origin, making MNEs opt for entry modes that involve relatively low resource commitments (joint ventures)(Erramilli and Roa, 1993). Cultural difference is measured following Kogut and Singh’s (1988) formula, based on Hofstede’s (2001) updated national culture scores (0= low / 100= high; http://www.geert-hofstede.com/hofstede_dimensions.php). To date, Hofstede’s study is the only one providing cultural distance indices for the CEE nations central in this study. Kogut and Singh (1988) defined national cultural distance as the degree to which cultural norms in one country differ from those in another country. Using Hofstede’s indices, they formed a composite index based on the deviation along each of the four cultural dimensions (i.e., power distance, uncertainty avoidance, masculinity/femininity, and individualism) between home and host country. The deviations were corrected for differences in the variances of each dimension and then arithmetically averaged. Algebraically, they built the following index:

(

)

[

]

∑

=

−

=

4 1 2

4

/

/

i

Vi

Iih

Iij

CDj

_,

Where Iij stands for the index for the ith cultural dimension and jth country, Vi is the variance of the index of the ith dimension, h indicates the home country, and CDj is the cultural difference of the jth country from the home market.

opportunity to establish long-term market presence (Agarwal and Ramaswami, 1992). Therefore, firms are likely to be selective and favour investments in more attractive markets (Agarwal and Ramaswami, 1992; Terpstra and Yu, 1988). Agarwal and Ramaswami (1992) found that when it came to choosing between joint ventures and sole ventures, one of the few variables that pointed toward sole ventures was high market potential. Typically, ownership levels are thought to increase with increasing host market size (Erramilli, 1996). The size of the host market (hostgdp) in our study was measured as GDP in billions of US dollars (current prices) for each year of entry in a CEE country. (http://www.imf.org/external/pubs/ft/weo/2006/02/data/weoselgr.aspx)

The third relevant control variable is Firm size. When firms want to succeed in entering foreign markets and compete with host country firms, extra resources are needed to absorb the higher cost of marketing, enforcing contracts and patents, and for achieving higher economies of scale (Hood and Young, 1979). This absorptive cost capacity is party reflected in the size of the firm (Buckley and Casson, 1976). Stated differently, firms characterized by larger size have a greater capacity to commit resources and absorb risk (Buckley and Casson, 1976; Shan, 1991). This capacity to commit resources and absorb risk basically means that a larger firm can deal with market failures at a lower cost compared to smaller firms (Jones, 1987). Larger firms can buffer themselves from specific transaction cost requirements (Agarwal, 1994). It has also been noted that large firms may be more opportunistic and relatively invulnerable to market failures, due to abundant slack resources and a stronger ability to retaliate against incursions (Osborn and Baugh, 1990). All the above mentioned arguments are backed up by empirical evidence. Research showed that larger firms are more likely to enter foreign markets in general and choose wholly-owned subsidiaries in particular, due to their stronger capacity to commit resources, absorb risk and lower transaction cost (Buckley and Casson, 1976; Cho, 1985; Caves and Mehra, 1986; Yu and Ito, 1988; Terpstra and Yu, 1988; Kimura, 1989). We control for Firm size through theapproximate number of MNEs’ employees worldwide (globempl).

social, demographic, and regulatory factors produce a situation of intense competition, be that on the basis of price, marketing expenditures, or investments (C.W.L. Hill et al. 1990). Such conditions are usually the result of a large number of competitors in the market and require a quick response from the firm. Insofar as resource commitments limit a MNC’s ability to adapt to changing market circumstances without incurring substantial sunk costs, a MNC can be theorized to favour joint ventures involving low resource commitments when competitive pressures in the host market are intense (C.W.L. Hill et al. 1990). Similar support was found in an article from Chan Kim and Hwang (1992), who argued that firms would do well to avoid internal organization when the intensity of competition in the host market is high; as such markets tend to be less profitable and therefore do not justify heavy resource commitments. The industrial organization literature predicts that the smaller the number of firms, the lower the level of competition intensity (Caves and Porter, 1977). This is because where there are a small number of competitors, with only one or a few of them dominant, competition is likely to be orderly, and the market leaders set the industry standards for the smaller firms to follow (Y. Pan et al. 1999). Each firm in this scenario enjoys a bigger slice of the pie in this concentrated industry. In contrast, in an industry where there are numerous competitors and none of them can exercise regulatory or leading roles (i.e., a fragmented industry), competition is more intense, and each firms ends up with a smaller piece of the pie (Y. Pan et al. 1999). Hence, the greater the intensity of competition in the host market, the more MNC’s will favour entry modes involving low resource commitments (joint ventures). Here, we asked the respondents to assess the intensity of local competition on a five-point Likert-type of scale. This variable is named Host market concentration (mktconcn). High scores reveal high fragmentation (a maximum of 5 for many competitors) and low scores point to high concentration (a minimum of 1 for few competitors).

Greenwood, 1997; Gatignon and Anderson, 1988). Where legal restrictions exist, firms tend to seek legitimacy, as well as efficiency, by utilizing less integrated modes of entry (Delios and Beamish, 1999). Thus, institutional theory tends to suggest that a firm's ability to exploit or enhance its capabilities may vary across institutional contexts in different national environments such that: firms entering countries with few legal restrictions on mode of entry tend to use wholly owned modes while firms entering countries with many legal restrictions on mode of entry tend to use joint venture modes (Brouthers, 2002). We used primary data from the survey to control for any legal restrictions on the level of ownership that foreign companies were allowed to have in the country’s domestic firms. Ownership restrictions are captured by a dummy variable (restrict) taking the value of 1 if ``legal restrictions`` where present, and 0 if there were not any limitations.

Data analysis

To explore the influence of the independent and control variables on the likelihood of either a joint venture or wholly-owned type of entry mode, we conducted a binomial logistic regression analysis which has been used in most recent-entry mode investigations (e.g., Gatignon and Anderson, 1988; Agarwal and Ramaswami, 1992; Davidson and McFertridge, 1985; Kogut and Singh, 1988; Kim and Hwang, 1992). In short, this statistical method was applied because of the ability of logistic regression techniques to incorporate a wide range of diagnostics, the dichotomous characteristic of the dependent, and the mix of continuous and categorical independent variables we use (Hair et al., 1995). Logistic regression allows us to test for the impact of each variable when the other variables are also present (what we call ``simultaneous`` testing). Thus, we can determine if each of the key variables is important in determining the shape of CEE entry when all four key variables are analyzed simultaneously (Brouthers, Brouthers and Nakos, 1998).

Since our data set is composed of continuous, categorical, single-scale and multiple-scale constructs, all continuous variables were converted to a standardized z-score, prior to the analysis (e.g. Ztechints). Because we are using a large number of interaction terms involving one variable in our model, the likelihood of serious multicollinearity problems exist. We, therefore, rescaled the original variables using procedures recommended by Aiken and West (1991). All continuous variables were `` mean-centered`` (by subtracting the corresponding variable mean from each value) prior to forming the moderators in order to avoid problems with multicollinearity (e.g. Cinstdevelopwb * Ctechints)9. After having produced a correlation matrix including all dependent, independent and control variables in our model, we can suggest that multicollinearity is not a likely threat to the substantive conclusion drawn from the parameter estimates based on rescaled mean-centered scores. Thus, we will not hesitate to include all main and interaction effects simultaneously in one equation, since the highest correlation of 0.465 is well below the cut off level of r =.9, after which high correlation between independent variables results in serious multicollinearity problems.10

Binomial logistic regression is used to obtain the maximum likelihood estimates of all the parameters along with their p values. The regression coefficients estimate the impact of the independent (or control) variables on the probability that the ownership mode is of the wholly-owned type (which carries the value of 1). The joint venture form will be used as the base mode. The utility of this option is assigned a value of zero and the utility of a wholly-owned subsidiary is estimated and interpreted with reference to the zero value. The size of the various coefficients indicates the extent to which the variables contribute to the utility of choosing that option beyond their contribution to the utility of choosing the default option. A positive coefficient means that the independent variable tends to increase the probability that the wholly-owned mode will be chosen, a negative coefficient the opposite. The model chi-square tests the null hypothesis that all the estimated coefficients

( )

β_S except β₀ are zero. This test will show us whether we are able to reject the null hypothesis from which we can conclude that our set of independent variables improves the prediction of the probability of a wholly-owned entry occurrence. 11

9 _{Note that the dummy variable - (ownership restriction) is not mean-centered or standardized (Jaccard and}

Turrisi, 2003), but entered as a categorical variable in SPSS.

10 _{See appendix for correlation matrix including all variables to check for multicollinearity: Table 1.}

11 _{Since the numbers of entries per CEE host country are too small to test separately, all CEE host countries were}

The model can be expressed as:

Probability of choosing a wholly-owned subsidiary

( )

_Y 1/

(

1 _e z

)

,

P = + −

where Y is the dependent variable, and Z is the linear combination of the independent, interaction and control variables. That is,

, 2 2 1 1 0 X X ... nXn Z =β +β +β + β

where β₀ is the intercept term, X₁,X₂,...X_n are the explanatory variables (denoting the

independent, interaction and control variables), andβ₁,β₂,...β_n are the corresponding regression coefficients.

We conducted the analysis by creating four models: Model 1 estimates the main effects of our predictor variables on the likelihood of a wholly-owned entry mode, and Models 2 to 4 separately add the moderating effects of institutional development on the predictors of MNEs’ firm-specific assets.12

The parameters were estimated using maximum likelihood, employing the BINARY LOGISTIC REGRESSION procedure of the SPSS (version 14.0) statistical package. Before we could make any inferences about the models we had to establish a baseline line model, which is labelled block 0 in SPSS. From the ``Case Processing Summary`` table we could determine how many cases were missing and were included in the analysis. In our model 154 out of the 209 cases were included in the model, which means that 55 cases were missing. One of the assumptions for a binomial logistic regression model is that the sample size is large enough; otherwise a solution might fail to converge. The rule of thumb in this situation

one-way between-groups ANOVA. If one country had a significantly different entry mode than the other countries then we would expect to find significant differences in the tests. Based on the analysis of variance, we were able to conclude that there were no significant differences (F=0.747 and Sig. 0.665) between entry modes for the CEE host countries, while meeting the homogeneity of variance assumption (Sig. for Levene’s test 0.116). From this we were able to see that combining all the target countries would not adversely affect our results (Brouthers, Brouthers and Nakos, 1998). See appendix for full results on the one-way between-groups ANOVA.

12 _{In the Appendix we give an interpretation of regression coefficients for the ``main effects” and ``moderator}

is that you need at least 10 cases for every variable you include in your model. Since the amount of variables in our model varies from 9 to 12, we concluded that a sample size of 154 is large enough. We also checked if variable outliers were genuine scores or errors. In the case of genuine scores we decided to leave them in the sample, since changing them to a less extreme value would distort original values which were within the range and excluding them would lead to a further reduction in the sample size. The next section of the output which is relevant for the analysis is the table headed ``Block 0``, it shows the results of the analysis without any of our independent variables used in the model. This serves as a baseline later for comparing the model with our predictor variables included. In the ``Classification`` table the overall percentage of correctly classified cases is 72.7 per cent. In this case SPSS classified (guessed) that all cases were of the wholly-owned kind, simply due to the fact that the sample is highly lopsided towards wholly-owned entry modes. We hope that later, when our set of predictor variables is entered, we will be able to improve the accuracy of these predictions. 13 We also need to check the Log likelihood of the baseline model: A “likelihood" is a probability, specifically the probability that the observed values of the dependent may be predicted from the observed values of the independents. Like any probability, the likelihood varies from 0 to 1. The log likelihood (LL) is its log and varies from 0 to minus infinity (it is negative because the log of any number less than 1 is negative). LL is calculated through

iteration, using maximum likelihood estimation (MLE). Log likelihood is the basis for tests of

a logistic model. The likelihood ratio is a function of log likelihood. Because -2LL has approximately a chi-square distribution, -2LL can be used for assessing the significance of logistic regression. The -2LL statistic is the likelihood ratio, also called goodness of fit. It reflects the significance of the unexplained variance in the dependent. In SPSS output, this statistic is found in the "-2 Log Likelihood" column of the "Iteration History" table. The likelihood ratio is not used directly in significance testing, but it is the basis for the likelihood ratio test, which is the test of the difference between two likelihood ratios (two -2LL's). It test the significance of the difference between the likelihood ratio (-2LL) of our model minus the likelihood ratio for a reduced model. When the reduced model is the baseline model with the constant only, as in our case, the likelihood ratio test tests the significance of our four models as a whole. The degrees of freedom in this test equal the number of terms in the model minus 1 (for the constant). Model chi-square measures the improvement in fit that the explanatory

variables make compared to the null model. When the probability (model chi-square) is less than .05, we reject the null hypothesis that knowing the independents makes no difference in predicting the dependent in the logistic regression. 14 In general, as the model becomes better, -2LL will decrease in magnitude. Therefore, we hope that later, when our four models are being tested the -2LL is lower than the baseline model, which is 180.473.15

Let’s now have a look at the first model, which included the standardized independent and control variables. The ``Omnibus Tests of Model Coefficients`` gives us an overall indication of how well the models performs, over and above the results obtained for Block 0, with none of the predictors entered into the model. This is referred to as a ``goodness of fit`` test. For this set of results we want a highly significant value (the Sig. value should be less than .05). In the first model the value is .003. Therefore, the model (with our set of variables used as predictors) is better than SPSS’s original guess shown in Block 0. The chi-square value is 24.879 with 9 degrees of freedom and is based on the lower -2LL of 155.595 (baseline 180.473). The results shown in the table headed ``Hosmer and Lemeshow Test`` also support our model as being worthwhile. This test, which SPSS states is the most reliable test of model fit available in SPSS, is interpreted very differently from the omnibus test discussed above. For the Hosmer-Lemeshow Goodness of Fit Test poor fit is indicated by a significance value less than .05. In our first model the chi-square value for the Hosmer-Lemeshow Test is 3.230 with a significance level of .919. This value is larger than .05, therefore indicating support for the model. The table headed ``Model Summary`` gives us another piece of information about the usefulness of the model. The Cox & Snell R Square and the Nagelkerke R Square value provide an indication of the amount of variation in the dependent variable explained by the model (from a minimum value of 0 to a maximum of approximately 1). In the first model the two values are .149 and .216, suggesting that between 14.9 per cent and 21.6 per cent of the variability is explained by this set of variables. The next table in the output to consider is the ``Classification table``. This provides us with an indication of how well the model is able to predict the correct category (joint venture/wholly owned) for each case. We can compare this with the Classification table shown for Block 0, to see how much improvement there is when the predictor variables are included in the first model. The first model correctly classified 73.4 per cent of cases overall (sensitivity 92 per cent, specificity 23.8 per cent), an improvement

14 _{Source: http://www2.chass.ncsu.edu/garson/PA765/logistic.htm}

over the 72.7 per cent in Block 0. 16 The ``Variables in the Equation`` table gives us information about the contribution or importance of each of our predictor variables. The test that is used here is known as the Wald test. We scan down the column labelled Sig. looking for values less than .10, .05, .01, 0.001. These are the variables that contribute significantly to the predictive ability of the model. Of the five control variables only ownership restriction (restrict) is significant at the ρ<.10 level (.067), since its B value is negative it means that the presence of ownership restrictions will result in a decreased probability of choosing for a wholly-owned type of entry mode. The other four control variables are not significant, but except for Zcultdiffks do have the proposed B signs. Most of the independent variables behaved as expected. Ztechints is significant at the ρ<.05 level (.018) and Zmneyears is significant at the ρ<.01 level (.007), both B values are of the positive sign which means that higher levels of technological intensity and international experience lead to an increased probability of choosing for a wholly-owned type of entry mode. Zceeyears was also significant at the ρ<.01 level (.003), however, contrary to our expectations had a B value that was of a negative sign, which means that higher levels of regional experience are associated with a decreased probability of choosing for a wholly-owned type of entry mode. Zinstdevelopwb had the expected negative B value, which is of the right sign but was rejected at a high ρ-level (.728). The constant (intercept), however, was significant at ρ<.001 (.000) with a positive B value, which means that wholly-owned entry modes are the default option. 17

Let’s now have a look at the second model, which included the standardized independent and control variables and the Cinstdevelopwb * Ctechints interaction. The ``Omnibus Tests of Model Coefficients`` shows a Sig. value of .006, with a chi-square value of 24.890 with 10 degrees of freedom and is based on the lower -2LL of 155.583 (baseline 180.473). The results shown in the table headed ``Hosmer and Lemeshow Test`` also support our model as being worthwhile. In our second model the chi-square value for the Hosmer-Lemeshow Test is 2.811 with a significance level of .946. This value is larger than .05, therefore indicating support for the model. In the second model The Cox & Snell R Square and the Nagelkerke R Square values are .149 and .216, suggesting that between 14.9 per cent and 21.6 per cent of the variability is explained by this set of variables. The next table in the output to consider is

the ``Classification table``. The second model correctly classified 73.4 per cent of cases overall (sensitivity 92 per cent, specificity 23.8 per cent), an improvement over the 72.7 per cent in Block 0. 18 The ``Variables in the Equation`` table shows that the interaction between Cinstdevelopwb * Ctechints has a negative B value as expected, but is rejected at a very high ρ-level (.915). This explains the overall fit of the second model compared to the baseline model Block 0, but its poor improvement compared to the first model.19

Let’s now have a look at the third model, which included the standardized independent and control variables and the Cinstdevelopwb * Cmneyears interaction. The ``Omnibus Tests of Model Coefficients`` shows a Sig. value of .001, with a chi-square value of 29.892 with 10 degrees of freedom and is based on the lower -2LL of 150.581 (baseline 180.473). The results shown in the table headed ``Hosmer and Lemeshow Test`` also support our model as being worthwhile. In our third model the chi-square value for the Hosmer-Lemeshow Test is 7.407 with a significance level of .493. This value is larger than .05, therefore indicating support for the model. In the third model The Cox & Snell R Square and the Nagelkerke R Square values are .176 and .256, suggesting that between 17.6 per cent and 25.6 per cent of the variability is explained by this set of variables. The next table in the output to consider is the ``Classification table``. The third model correctly classified 76.0 per cent of cases overall (sensitivity 92.9 per cent, specificity 31.0 per cent), an improvement over the 72.7 per cent in Block 0. 20 The ``Variables in the Equation`` table shows that as expected the interaction between Cinstdevelopwb * Cmneyears has a positive B value and is significant at ρ<.05 (.037). From this we can conclude that a unit increase in the institutional development level results into an increase of the odds for firms with average levels of international experience to choose for wholly owned type of entry modes. This explains the overall better fit of the third model compared to the baseline model Block 0 and its improvement compared to the first model.21

Finally, we have a look at the fourth model, which included the standardized independent and control variables and the Cinstdevelopwb * Cceeyears interaction. The ``Omnibus Tests of Model Coefficients`` shows a Sig. value of .002, with a chi-square value of 27.941 with 10

degrees of freedom and is based on the lower -2LL of 152.532 (baseline 180.473). The results shown in the table headed ``Hosmer and Lemeshow Test`` also support our model as being worthwhile. In our fourth model the chi-square value for the Hosmer-Lemeshow Test is 4.852 with a significance level of .773. This value is larger than .05, therefore indicating support for the model. In the fourth model The Cox & Snell R Square and the Nagelkerke R Square values are .166 and .240, suggesting that between 16.6 per cent and 24.0 per cent of the variability is explained by this set of variables. The next table in the output to consider is the ``Classification table``. The fourth model correctly classified 76.0 per cent of cases overall (sensitivity 93.8 per cent, specificity 28.6 per cent), an improvement over the 72.7 per cent in Block 0. 22 _{The ``Variables in the Equation`` table shows that contrary to our expectations the} interaction between Cinstdevelopwb * Cceeyears has a negative B value and is significant at ρ<.10 (.090). From this we can conclude that a unit increase in the institutional development level actually results into an increase of the odds for firms with average levels of regional experience to choose for equity joint ventures with local CEE partners. This explains the overall better fit of the fourth model compared to the baseline model Block 0 and its improvement compared to the first model.23

RESULTS & CONCLUSIONS

Based on the previous analysis we found that the hypotheses related to Ztechints and Zmneyears were indeed accepted (H2a & H3a), which made us conclude that higher levels of technological intensity and international experience lead to an increased probability of choosing for a wholly-owned type of entry mode. The hypothesis related to Zceeyears was rejected (H3b), which meant that contrary to our expectations higher levels of regional experience were associated with an increased probability of choosing for equity joint ventures with local CEE partners. Regional experience had, therefore, a positive influence on the likelihood of entry via joint ventures, while International experience showed the opposite effect. A plausible explanation could be that more familiarity with the way business is conducted in the region comes with the ability to trust a local partner, hence the preference for joint venturing as opposed to the doing-it-alone option (preferred in the case of abundant international experience) (Dikova and Witteloostuijn, 2007). The hypothesis concerning Zinstdevelopwb was also rejected (H1). It was of the proposed negative sign, however not