Entry Mode Choices of Dutch Manufacturing SMEs: A Transaction Cost Perspective and the Moderating Effect of Entrepreneurial Orientation

(1)

Words: 11.513

Entry Mode Choices of Dutch Manufacturing SMEs:

A Transaction Cost Perspective and the Moderating Effect of

Entrepreneurial Orientation

Author:

I.T. Schaafsma

Supervisor:

Dr. I. Kalinic

Co-assessor:

Prof. dr. S. Beugelsdijk

September 2013

University of Groningen

Faculty of Economics and Business

(2)

2 ABSTRACT:

(3)

3 This thesis is the conclusion of my Masters in International Business and Management. For the past year and a half I have worked on this project. Whilst I was collecting the data for this research I lost two family members, which put things into perspective. I would like to thank my parents and girlfriend for their support during this period.

Last but not least I would like to thank dr. Igor Kalinic for having the patience in guiding me through this process.

I hope this thesis is an easy and pleasurable read!

In loving memory of:

(4)

4

List of Tables and Figures

Tables:

Table 1 Discriptives & Correlations...29 Table 2 Results of Logistic Regression...30 Tables in Appendixes:

Table 3 Linearity of the logit...45 Table 4 Collinearity Diagnostics...45 Table 5 CFA Pattern matrix...46 Figures:

(6)

6

1. Introduction:

Firms that seek to increase demand for their goods and/or services might consider adopting an export strategy (Aulakh & Kotabe, 1997). However, whilst preparing for expansion through internationalization such firms face a plethora of choices. Apart from assessing the suitability of their products or services to a foreign market (and customers) it is the choice between available market entry modes that is arguably one of the strategically most important decisions to be made. Not the least because a wrong entry mode decision can turn out to be very costly. Changing export channels, after initially making the wrong channel choice, will often result in disproportionally high switching costs. Especially for Small and Medium Sized Enterprises (SMEs), this could well result in a financial burden that could threaten the

sustainability of the entire firm. So, they better get it right, first time round (Brouthers & Brouthers, 2003).

The subject of entry mode choice has received ample attention in academic research into exporting practices, however most researchers focus on large firms. And only few scholars have researched entry mode choices made by SMEs (Brouthers & Nakos 2004).

The reason for examining SMEs separately is because they have been found to react

differently to their environment compared to large firms (Shuman & Seeger, 1986). SMEs are found to be different from multinational enterprises (MNEs) in their managerial style;

ownership; and level of independence (Coviello & McAuley, 1999). Additionally, due to their typically limited resources, they make different international strategic choices compared to MNEs (Zacharis, 1997; (Erramilli & D’Souza, 1993)).

(7)

7 Transaction Cost theory (TC theory) captures the external effects on entry mode decisions and was introduced by Williamson (1975, 1985). It takes into account the specificity of the assets engaged in the transaction, external uncertainty, internal uncertainty and the

frequency/volume of transactions, and their effects on entry mode choice.

The concept of Entrepreneurial Orientation (EO) has primarily been investigated in the research field of entrepreneurship. It captures the internal aspects of the firm and its key decision makers (Zhou et al.2010). EO looks into the behavior and culture of a firm which is captured by the innovative, proactive and risk taking nature of the firm.

This thesis will investigate whether the internal aspects of EO can be a fitting addition to the SME TC - Entry Mode model as was designed by Brouthers and Nakos (2004). The proposed model is shown in Figure 1.

After seeing figure 1, we hypothesize that EO moderates the relationship between TC theory variables and the chosen entry mode. By examining the influence of EO on the relationship between TC theory and chosen Entry Mode we hope to:

 identify and add new insights to the field of TC theory – Entry mode research;

 improve the predictive values of entry mode modeling;

 and help Dutch SME’s make better entry mode decisions.

We will contribute to the literature in two ways:

First by adding Entrepreneurial Orientation as a moderating variable to the relationship between TC theory and chosen entry mode. EO has received considerable attention in the entrepreneurship literature but has not been researched in the context of choice of export channels and/or entry modes. EO is a function of three constructs that influence firm behavior: proactiveness; innovativeness; and risk taking .

(8)

8 because the volume of those transactions are still of an influence (Erramilli & Rao., 1993; Klein & Roth, 1990).

In chapter 2 a short literature review of the variables included and the hypotheses will be presented. In chapter 3 we will show how the variables have been operationalized. In chapter 4 we will discuss the methods we used and chapter 5 presents the results of the binary logistic regression. In chapter 6 we will discuss the results and the limitations of this research. In chapter 7 the conclusions and suggestions for future research are given.

(9)

9

2. Literature review and Research Hypotheses:

The key hypothesis in this thesis is that EO moderates the TC – entry mode choice

relationship. In explaining entry mode choice, TC-theory has a history of mixed results. We hope to find that Entrepreneurial Orientation is an important factor influencing the

relationship between TC and entry mode decisions. And that it therefore improves the

predictive value of existing models. EO is proven to reflect the internal aspects of the firm and of its key decision makers.

This chapter will first discuss the past literature on entry modes. Subsequently it will investigate TC theory and its hypotheses. And it will conclude with discussing EO and its hypothesized moderating effect on the TC-entry mode relationship.

2.1. Entry modes:

Making the correct channel choice is important to the sustainability of every firm that considers to internationalize or to engage in exports. It’s even more relevant to SMEs because they typically lack the resources to compensate for making a wrong decision in such situations (Brouthers, et al., 2003). Even if an SME would survive a wrong entry mode decision it is very difficult to change or correct it afterwards. A wrong entry mode decision is therefore likely to generate prolonged negative consequences for an internationalizing firm (Brouthers & Hennart, 2007).

In research conducted by (Anderson & Gatignon, 1986) as much as 17 different entry modes were identified. These seventeen entry modes can be grouped into three main ways through which a firm might enter a foreign market. These main modes are:

 Wholly owned subsidiary;

 Intermediate options or Joint Venture;

 Independent Distributor/Agent.

These entry modes can be placed in a spectrum with at one end firms opting for a wholly owned subsidiary (WOS), and therefore performing all distribution and marketing tasks in a foreign market by themselves .And, at the other end of the spectrum, firms opting not to execute any of these tasks by themselves, but contracting them to independent distributors or agents. Distributors will take title to the product upon order or upon delivery and will

(10)

10 option. In this option firms make, for example, use of commission agents, or enter into

distribution joint-ventures with other firms (Anderson & Coughlan, 1987; Aulakh & Kotabe, 1997; Klein, et al., 1990).

According to Brouthers & Hennart (2007) ,the above stated categorization can be placed alongside a continuum of increasing commitment, risk and control. Starting with contracting and ending with a WOS. Where maximum control, commitment, and risk are desired by the firm it would typically opt for a WOS.

The second perspective in the literature, which firstly was articulated by Hennart (1988), splits entry modes into two separate categories: contracts and equity (Brouthers & Hennart, 2007). In this perspective JV’s and WOSs are placed in the equity section. This

categorization is based upon the way suppliers are paid. The fundamental characteristic of the equity category is that JV and WOS pay their suppliers ex post , with profits arising out of the venture. Whereas in the category of contracts suppliers are paid ex ante, upon order or

delivery of the goods or services.

With regard to TC theory we are mostly interested in the level of control, commitment and risk arranged alongside a continuum.

2.2. Transaction Cost Theory:

(11)

11

Literature:

Transaction cost theory was first discussed by Coase (1937) who stated that firms and markets are alternative governance structures that differ in their transaction costs. Coase specifically proposed that under certain conditions, the costs of doing business in a market may exceed the costs of doing business internally. Coase’s article however did not provide a way to measure transaction costs directly and it was not until Williamson (1975) that research on this topic really took off. Williamson (1975, 1981, 1985) was the first to operationalize and refine transaction cost theory making it a researchable phenomenon. Williamson suggested that by taking the observable dimensions of transactions, i.e. uncertainty, asset specificity, and transaction frequency, one can develop testable hypotheses by associating the relative efficiency of different governance structures (Geyskens, et al., 2006).

Since then research on the topic has been covered in multiple fields, namely in economics, law, organization, marketing, finance, sociology, accounting and operations management (Geyskens, et al., 2006). Due to the fact that multiple research fields use transaction costs to investigate a broad range of exchange related issues, there is a lot of analytical diversity. These however can be grouped into four contextual domains: (1) Vertical integration, (2) vertical inter-organizational relationships, (3) horizontal inter-organizational relationships, (4) tests of TC theory’s assumptions (Rindfleisch & Heide, 1997). For the scope of this research however we will only discuss the TC- entry mode research field which falls within the contextual domain of vertical integration (Canabal & White 3, 2008).

In the light of this field, the basic rationale for TC theory is that firms need to create

governance structures that will minimize costs and inefficiencies associated with entering and operating in foreign markets (Canabal & White 3, 2008). Gatignon & Anderson (1988) started with looking at the effects of TC theory on total versus partial ownership of a foreign subsidiary in exporting US MNCs. They established that full ownership was more likely when asset specificity and internal uncertainty were high and external uncertainty was low.

Eramilli & Roa (1993) extended the application of the transaction cost framework to service firms. They found that service firms favor shared control when asset specificity is low which is moderated by firm size, county risk, and degree of seperability of production and

(12)

12 successfully linked TC theory and entry mode choice to performance. In their research they discuss that firm entry mode choices, that ‘fit’ with what TC theory predicted, will show better performance than firms whose entry mode choices had no ‘fit’. In 2002 Nakos and Brouthers were the first to research SME entry modes. Using Dunning’s eclectic framework they were able to predict SME entry mode selection in Central Eastern European markets. Building on this Brouthers and Nakos (2004) were able to identify that transaction cost theory also might influence SME entry mode choice. They debated that only few researchers’ studied SME mode choice and that TC have been overlooked entirely. They argue that entry mode decisions are one of the most important strategic decision small firms can make. They found that TC theory did a good job in explaining SME entry mode choice and that SMEs fitting TC theory predicted mode choices perform significantly better than firms which did not have that fit. However the variance explained by their model was 20 %. This indicates there are

additional variables at play.

Overall the literature on transaction cost theory has been extensive and its variables and measures have been discussed thoroughly. Transaction costs have been studied across multiple fields and a lot of different exchange related issues have been addressed. Only recently has it successfully been applied to SME entry mode research. This section will now continue with discussing the different attributes that comprise TC theory.

Asset specificity:

(13)

13 When asset specificity is high, firms are more concerned about protecting their technology and knowledge. High asset specificity may well be the core competence of which

dissemination would have an adverse effect on the performance of the internationalizing firm. Klein et al., (1990) state that firms tend to internalize foreign operations in order to gain control and reduce the risk of dissemination.

Brouthers and Hennart (2007) identified that R&D and/or advertising intensity most commonly have been used as a measure for asset specificity. But they also identified four other perceptual measures of asset specificity in entry mode research. These include: physical (service) asset specificity, human asset specificity (Brouthers & Brouthers 2003; Erramilli & Rao 1993; Klein, Frazier & Roth, 1990; Palenzuela & Bobillo, 1999), technology asset specificity (Taylor et al. 1998; (Palenzula & Bobillo, 1999) and dedicated asset specificity ( Brouthers & Brouthers 2003; Brouthers et al., 2003). In their meta analysis Zhou et al 2004 established that the majority of TC based entry mode studies tend to find mixed significant relationships between TC variables and entry mode choice.

Despite that Brouthers and Nakos (2004) did establish that SMEs prefer hierarchical modes of entry when asset specificity is high.

Therefore:

H1A: SMEs will tend to prefer hybrid modes of entry when asset specificity is low, but tend to

prefer hierarchical modes of entry when asset specificity is high.

External Uncertainty:

(14)

14 maintain flexible in a volatile environment and if the situation requires it can switch between agents or exit the market.

Brouthers and Hennart (2007) found wide variation in the way the two most commonly used constructs for external uncertainty - country risk and cultural distance - were measured. Where some used the Euromoney country risk index (Delios & Beamish, 1999) others used the Frost & Sullivan Country Risk Guide (Contractor & Kundu, 1998). Other measures include perceived market potential (Brouthers, 2002), perceived political and economical stability (Brouthers, 2002; Brouthers & Brouthers, 2003; Brouthers, et al., 2003; Kim & Hwang, 1992; Luo, 2001), and Industry concentration (Kim & Hwang, 1992).

Zhao et al.’s (2004) meta anlysis found that country risk negatively impacts the likelyhood of a firm choosing an hierarchical mode of entry. In SME research however this is much less clear. Burgel and Murray (2000) for example did not find a significant relationship between country risk and entry mode choice. Where Brouther and Nakos (2004) did find that external uncertainty negatively impacts the likelyhood of a firm choosing an hierarchical mode of entry.

H1B: SMEs will tend to prefer hierarchical modes of entry when external uncertainty is low,

but tend to prefer hybrid modes of entry when external uncertainty is high.

Internal Uncertainty:

Internal or internal uncertainty arises from the inability of a company to predict the behavior of individuals in a foreign country. This type of uncertainty may lead to opportunistic behavior which involves cheating, disruption of information, dodging responsibilities and other forms of false behavior (Williamson, 1985). To minimize false behavior a firm must develop control mechanisms. (Gatignon & Anderson, 1988) (Klein, et al., 1990). One of the control mechanisms is internal control which can be obtained by adopting a hierarchical entry mode. However,hHierarchical modes give firms the right but not the means to exert control (Anderson & Gatignon, 1986). Anderson & Gatignon (1986) state that controlling foreign operations is a special ability firms acquire over time.

(15)

15 or third parties control the foreign operations which reduces control related problems arising from internal uncertainty (Gatignon & Anderson, 1988) (Delios & Beamish, 1999).

There is a wide variety of measures for internal uncertainty. Research into MNEs focuses on experienced based measures like number of years of world wide experience (Contractor & Kundu, 1998); total number of foreign investments (Delios & Beamish, 1999),;ratio of foreign to total number of investments (Contractor & Kundu, 1998); number of years of presence in host country (Hennart, 1991),; or number of country specific ventures (Luo, 2001). Despite the various measures for MNEs Zhao et al. (Zhao, et al., 2004) established that experienced MNEs prefer hierarchical modes of entry. These experience measures however are indirect (Brouthers & Hennart, 2007). Some researchers examining SMEs used non-experienced based measures (Brouthers, 2002) (Brouthers & Brouthers, 2003) (Brouthers, et al., 2003). They examined, with mixed results, the perceived difficulty in partner selection and perceived ability to enforce, monitor and control contractual arrangements.

Brouthers and Nakos (2004) debate that internal uncertainty is very important for SMEs because they rely more on the experience of management and have less developed

management teams than MNEs. Internal uncertainty may discourage SMEs from making a hierarchical foreign market entry for a couple of reasons. SMEs don’t have the ability to send their own people to guide the transactions for a long period of time (Lu & Beamisch,

2001).And SMEs also have less international experience and therefore have not developed systems and processes for foreign control (Zacharakis, 1997).

H1C: SMEs will tend to prefer hybrid modes of entry when they have not developed internal

control mechanisms, but will tend to prefer hierarchical modes of entry when they possess internal control mechanisms.

Frequency:

(16)

16 In some studies frequency is used as a control variable (eg. (John & Weitz, 1988); (Klein, et al., 1990). They see frequency as a dichotomous phenomenon which can control by

examining only recurring exchanges. Brouthers and Nakos (2004) debate that frequency is an important determinant of TC’s. But in entry mode studies ‘transactions are considered

continuous, thus precluding the need for a separate measure of frequency’ (Brouthers & Nakos 2004. P.230). In following this argument frequency of transactions cannot be measured because it is considered as continuous. However it ignores the aspect of volume which can be distilled from Williamson (1985, p 60.) . ‘’the cost of specialized governance structures will be easier to recover for large transaction of a recurring kind’’. This justifies the inclusion of Frequency to the TC model to be measured as volume of turnover originating from most important export market.

H1D: SMEs will tend to prefer hybrid modes of entry when frequency of transactions is low,

but tend to prefer hierarchical modes of entry when frequency of transactions is high

Conceptually and by measurement we can see that there is much consensus on the TC

variables. Several studies build upon the Anderson (1985, 1988) and Williamson (1975, 1985) papers. Even though the concepts and measures are more or less agreed upon, there is a large variety in the results. Most studies cannot corroborate all the aspects of TC theory and are mostly only partially consistent with Williamson’s (1975) TC theory (Carter & Hodgeson, 2006).

2.3. Entrepreneurial Orientation:

Entrepreneurial Orientation (EO) was first touched upon by Miller(1983) who stated that it was a function of an organization’s exhibition of risk taking, innovativeness, and

(17)

17 EO has two schools of thought that operationalize EO differently. On the one end there is Covin and Slevin (1989) which conceptualize EO as a function of three dimensions: Proactiveness; Innovativeness; and Risk taking. Because these dimensions show high intercorrelations most studies combine them into one single factor (e.g. Walter, et al., 2006.; Lee, et al., 2001; Wiklund and Shepherd, 2003). On the other hand there is Lumpkin & Dess (1996) who conceptualize EO as a function of 5 dimensions that represent independent predictors. The two added dimensions are competitive aggressiveness and autonomy. In both schools the definition and operationalization of the first three dimensions are the same. Where:

1. Innovativeness refers to actively engaging in creativity and experimentation through the introduction of new products as well as R&D in new processes (Covin & Slevin).

2. Risk taking refers to bold actions by venturing into the unknown, borrowing heavily,

and/or committing significant recourses to ventures in uncertain environments (Covin & Slevin, 1989).

3. Proactiveness is an opportunity seeking forward looking perspective characterized by the

introduction of new products and services ahead of the competition and acting in anticipation of future demand (Covin & Slevin, 1989).

The two additional dimensions are defined by Lumpkin and Dess (1996) :

4. Competitive aggressiveness is the intensity of a firm’s effort to outperform rivals and it’s

characterized by strong offensive posture or aggressive responses to competitive threats (Lumpkin & Dess, 1996).

5. Autonomy refers to independent action undertaken by entrepreneurial leaders or teams

directed at bringing about new ventures and seeing it to fruition (Lumpkin & Dess, 1996).

(18)

18 used the three dimensions, establishing that these dimensions correlated substantially and thus justifying the view of EO as a unidimensional concept.

Zhou, et al., (2010) established that firms with EO have the ability to learn better and upgrade their capabilities quicker than competitors that do not exhibit EO in their firm. When

institutional distance increases firms will need to generate information about these differences that can be transformed into effective policy (Chan & Makino, 2007). Firms that exhibit EO have proven to be better able to adopt these learnings into workable solutions in a shorter time frame than low EO firms. Therefore hierarchical export channels in firms with high EO will benefit because it helps exporters overcome differences in goals, objectives, norms and institutional practices, providing them with effective ability to generate, disseminate and respond to demands of the export market (He, et al., 2013).

Zhou et al., (2010) argue that entrepreneurial actions and resources can be usefully combined to provide new capabilities that contribute to competitive advantage in international markets: “Entrepreneurial proclivity (EO) is a market-oriented organizational learning culture, which provides the global vision and norms that shape decision maker’s attitudes toward proactive, risk taking and innovative behavior’’ (p.886, Zhou, et al., 2010). We therefore suggest that firms facing a lot of uncertainty and institutional distance, and that have EO, will be more likely to adopt hierarchical export structures. The opposite can also be argued in that hybrid modes are prefered in firms exhibiting low EO. This can be argued through their inabillity to quickly establish new (learning) capabilities which they then substitue through hybrid export structures. In weak EO firms, the need for information is much higher because without it they can not adapt product offerings to the needs of the foreign market (He, et al., 2013).

H2: EO moderates the relationship between TC theory and entry mode choice in a way that

high EO within firms will increase the propensity that such firms choose hierarchical channels.

(19)

19

3. Variable Construction:

In the previous chapter the definitions and the status quo of current literature were elaborated upon. In order for us to execute the analysis the variables have to be operationalized. This chapter will first discuss the methods of data collection that have been used. Subsequently we will discuss the selection of the variables and how we have operationalized them. We’ll then conclude with a short paragraph on how we prepared the data for statistical testing.

3.1. Dependent Variable

Hybrid or Hierarchical entry modes. To test the first hypotheses the dependent variable entry

mode choice was selected. We conceptualized it as a dichotomous variable. Respondents were asked to answer questions about their most important export market. As in the He et al., (2013) study, hierarchical channels were assigned a value of 1 and included two types of measures: ‘’We have a wholly owned sales subsidiary’’ and ‘’We serve it directly from the Netherlands, using company personnel’’. Hybrid channels were assigned a value of 2 and included 3 types of measures: ‘’We are involved in a joint venture with another company to handle sales in this market’’, ‘’We use commission agents’’ and ‘’We sell to a merchant distributor who takes title to our product and contacts buyers directly’’.

The choice for a dichotomous variable can be justified in two ways. First, when interpreting Williamson’s (1985) words we see that TC theory concerns the choice between markets and hierarchical governance forms, this suggests a dichotomous variable. Second, previous entry mode studies, including SME studies (Brouthers & Nakos, 2004), tend to code entry modes in this way.

3.2. Predictor Variables

(20)

20

Asset specificity:

Past studies (eg. Hennart, 1991) on TC theory used R&D intensity as a proxy to measure asset specificity. When specificity of assets is high firms choose for Wholly Owned Subsidiaries (WOS). This is the case because high investments increase the risk surrounding a transaction. Therefore firms desire increased control and opt for WOS. R&D is measured using a single question asking the percentage of total turnover invested in R&D activities.

Rindfleish and Heide (1997) suggest that R&D has a focus on physical capital and does not capture the human capital surrounding asset specificity. Physical capital includes financial, production and research and development resources. Where human capital arises from special training for personnel (Williamson 1985).Human capital investments are measured using four 7- point Likert scale questions asking about the ease/difficulty of getting to know the,

industry, market and product (Cronbachs Alpha 0.666).

Internal Uncertainty:

Internal Uncertainty was measured using a single item 7- point Likert scale question. It asked about the ease or difficulty of measuring the performance of individuals with whom the firm may cooperate. This measure was adopted from Shervani et al. (2007). (Shervani, et al., 2007). In other researches Cultural distance and International experience are used as a measure of internal uncertainty. These measures are however susceptible to other variables and have proven to be unreliable in a meta analytical review conducted by Zhoa et al (2004).

External Uncertainty:

(21)

21 variables with a 1 when a firm enters a high risk country, and with a 0 when it is in any other ; as suggested by Erramilli and Rao (1993).

Additionally we used a scale adopted from Shervani et al. (2007) who in their turn adopted it from John and Weitz (1988). To measure ‘perceived volatility in the firm’s environment’ we use four items, all with a 7-point Likert scale (Cronbach’s Alpha 0.666).

Frequency:

As mentioned earlier on in Chapter 2, there is no need to measure the amount of recurring transactions because they are considered to be continuous (Brouthers & Nakos 2004).

However we do need to look at the volume of those transactions because it is an integral part to the operationalization of Frequency. Because one part of the variable cannot be measured, using the logic of past research, we assume it is present. The other part, volume, still can be measured.

Frequency of transactions is measured through one single item. It asked the respondents about the percentage of total yearly export sales in the most important export market. This is a widely accepted measure in the field. ( eg. Klein, et al., 1990; He, et al., 2013).

3.3. Moderator

EO: Substantial amounts of research have been conducted into EO and its measures. Some researchers built their own measurement models, however most used or modified the scales designed by Khandwalla(1977), or Miller (1983). Knight (1997) established that the scale designed by Covin and Slevin (1986,1989), based on Khandwalla (1977) and Miller & Friezen (1982), was applicable for measuring EO outside the USA. We adopted nine questions , all on a 7-point Likert scale, that measure three dimensions of EO, i.e.

proactiveness; risk taking; and innovativeness. In following Rauch et al’s( 2009) suggestion a unidimensional construct for EO was created. This was done through summing each of the nine items and dividing the sum by nine (Cronbach’s Alpha= .84)

3.4. Control Variables

(22)

22 mode choice. Therefore we follow Brouthers et. Al. (2003) in their argument that these

variables need to be included because it is likely that firms do not base their entry mode choice solely on TC theory variables. The control variables selected for this research are: firm size; cultural distance; industry type; market experience; market size; international

experience, and export channel experience.

Firm size is a frequently used control variable (eg.Brouthers, 2002). Large firms tend to have

more resources at their disposal and can therefore fund hierarchical entry modes easier than smaller firms (Contractor, 1984). They do not perceive the risk accompanying the entry mode decision in the same way. For instance a young start-up firm has little resources at its

disposal. When it makes a wrong entry decision it could be detrimental for its survival. Therefore a small firm is more likely to choose hybrid entry modes and large firms are more likely to choose hierarchical entry modes. Consistent with Brouthers and Nakos (2004) and Gatignon and Anderson (1988) we used the number of employees as a proxy for firm size.

Experience is a function of three other variables that are not central to TC theory but that do

influence the TC entry mode choice relationship. When a firm has a lot of experience it is assumed that it has developed systems and processes for controlling subsidiaries (Brouthers & Nakos, 2004). Also Luo and Peng (1999 p. 270) suggest that “experience is a prime source of learning in organizations.” The three variables measuring experience are Market

Experience; International Experience; and Export Channel Experience. Market Experience is measured through the number of years of exporting towards the most important export

market. International Experience is measured by number of countries the firm has exported to. Export Channel Experience which is measured by asking the respondents to select the number of times they used the same export structure in different countries.

Market Size and Growth has been found to be of influence on entry mode decisions. Previous

(23)

23 was used for market growth. All data for the Market Size and Growth variables were obtained from the IMF (International Financial Statistics 2012).

Cultural Distance has often been used as (control) variable in entry mode literature ( eg. Zhoa

et al., 2007; He et al., 2013). Mostly it has been used as a measure of internal uncertainty in combination with International Experience. Zhoa et al’s meta-analytic review established that results were mixed and suggests other measures for internal uncertainty to be taken. We do however think this variable is of influence on the entry mode decision. Kogut and Sing’s formula is used for calculating the cultural distance between the Netherlands and the host market. The farther away a country is on the cultural distance scale, using Kogut & Singh’s (1988)index, the more risk and uncertainty is associated with entering that particular market. The index is based on the four cultural dimensions as provided by Hofstede (1980) . The equation for the cultural distance index is:

Where :

CDj- is cultural distance of country j from the Netherlands

Iij – is the cultural dimension of country j

Iid- is the cultural dimension of the Netherlands

Vi – is the variance of cultural dimension .

The cultural dimension scores were collected from Geert Hofstede’s website (www.geerthofstede.nl) .

Industry Dummies are used to control for industry specific effects. In this research 22 different

(24)

24

4. Methodology

Before this study can start to test whether EO moderates the TC – entry mode relationship the sample needs to be checked according to the assumptions of binary logistic regression. First the method and data collection are discussed in section 4.1. In 4.2 the sample and its descriptives are elaborated upon. In section 4.3 the assumptions are being discussed and binary logistic regression is justified.

4.1 Method of Data Collection

Most of the data in this thesis are primary, except for the predictor variable External

uncertainty; for the control variable Cultural Distance; and for the control variables that made use of Gross Domestic Product (GDP). The survey data was collected through sending two surveys, over a 5 month period in 2012, to participating firms. The time between each round was one month. The companies were selected through Bureau van Dijk’s Orbis Database. This database contains all registered companies in the Netherlands.

Data for the independent variable External Uncertainty was collected by using secondary data sources. We used two sources, one for the categorization of the countries as provided by Goodnow and Hansz (1972), and the Euromoney country risk index 2012 for an up to date list of high risk countries.

For our control variable, cultural distance, we also used secondary data sources. We used Geert Hofstede’s (2012) personal website which provides a cultural dimension index for the scores needed to calculate the distance between the Netherlands and each target country.The data for the control variables that made use of GDP, were collected from the IMF websi

4.2 Sample

This section will first discuss how the sample of 179 manufacturing SMEs was collected in sub section 4.2.1. In sub section 4.2.2 the descriptive statistics will be elaborated upon.

4.2.1 Sample Construction

(25)

25 manufacturing and not service firms is because these two types react differently to TC

variables (Brouthers & Brouthers, 2003).

A sample of 6330 SMEs was drawn from the ORBIS database, maintained by the Dutch chamber of commerce. The selected firms are independent and owner managed. We only considered firms with a number of FTEs between 10 and 250. This is consistent with the European commission’s (2009) definition of Small and Medium Sized enterprises. We also only considered firms that had the NACE industry code ranging from 10-17 and 19-32. Of the 6330 firms, we randomly selected 1070. All the SMEs, in this sample of 1070 firms, were then contacted by phone and asked to participate in our study.

4.2.2 Descriptive Statistics

After the first mailing, 276 firms responded which were contacted prior to the mailing by phone. The second round of mailing left us with 209 firms that filled out both surveys. Of the 209 responses 24 were not usable (18 used multiple export structures, 4 had too few FTEs, 1 employed too many FTEs, and 1 did not export). This left us of with an original sample of 185 firms. In this thesis we only make a distinction between Hierarchical and Hybrid entry modes. 122 firms indicated that they used hierarchical export structures and 63 hybrid choices. Table 1 gives an overview of the means and standard deviations of the other variables.

4.3 Empirical analysis.

In principle one can always run a logistical regression analysis regardless of the assumptions. However for the results to be valid one must make sure that the assumptions are met. This section will continue with discussing in section 4.3.1 the assumptions of logistic regression and discuss the general assumptions of multiple regression in section 4.3.2. Section 4.3.1 firstly addresses the empirical assumptions and then discussen the non empirical assumptions which can be answered with yes or no. Section 4.3.2 will discuss the general issues of

(26)

26

4.3.1 Logistic Assumptions

Burns & Burns (2008) provided us with 5 assumptions (listed in figure 2). Not all of these assumptions are empirical, meaning that an answer of yes and no suffices.

Figure 2: The assumptions of logistical regression (Burns & Burns ,2008)

Assumption one, linearity of the logit, is presented in table 1 in Appendix A. This assumption can be tested empirically through the Box-Tidwell approach (Hosmer & Lemeshow, 1989), which is mostly used in the literature (Tebachnick & Fidell, 2001). Tebachnick & Fidell (2001, p. 522) state that in this approach ‘’ terms are added to the logistic regression model which are composed of the interactions between each predictor and its natural logarithm.’’ Tebachnick & Fidell (2001)suggest that when one of the interactions, of a predictor variable with its natural logarithm, is significant the assumption is violated. The violating variables then need to be transformed. From this analysis we see that the interaction of the variable Internal Uncertainty with its natural logartitm is significant meaning that this variable is violating the assumption of non-linear relationship. We transformed the variable using a square root transformation and after transformation it was normally distributed, as suggested by Tebachnick & Fidell (2001).

Assumptions 2-5 are non empirical assumptions which can either be valid or invalid.

Assumption 2 requires the dependent variable to be a dichotomy, meaning that it must have 2 categories to which the data can belong. In this case it is 1(Hierarchical) and 0 (Hybrid). Assumption 3 is always met. It states that the predictors can be both linear and non linear distributed. Assumption 4 states that there are 2 groups to which a case can belong to. Assumption 5 (assumption of 50 cases per predictor.)is slightly violated. However , other

 Logistic regression does not assume a linear relationship between the dependent and independent variables.

 The dependent variable must be a dichotomy (2 categories).

 The independent variables need not be interval, nor normally distributed, nor linearly related, nor of equal variance within each group.

 The categories (groups) must be mutually exclusive and exhaustive; a case can only be in one group and every case must be a member of one of the groups

(27)

27 authors (Field, 2010)state that 30 cases per predictor are enough so it’s a rather arbitrary assumption. We therefore don’t foresee any troubles with the size of our sample.

4.3.2 General Assumptions

The assumptions that apply to multiple regressions also apply to logistic regression. These assumptions regard multi-collinearity and the absence of outliers.

High correlations between the predictor variables influence the reliability of the model. We therefore have to check our correlations table, in table 1, for significant values of > 0.7 as suggested by Palant (2007). When checking table 1 we see that two control variables correlate significantly at the 0.72 level. This indicates that we need to further examine wether there is multicollinearity in our data.

Because we are working with interactions we expect high multicollinearity between the predictors and the interaction terms. In order to solve this problem we have mean centered the predictors as suggested by Aiken & West (1991). This prevents multicollinearity of the predictors with the interaction variables. To be safe we did another correlation test presented in Appendix A table 2. None of the VIF score’s exceed 5 which is suggested as cut off point by Kutner, Nachtsheim & Neter (2004) indicating there is no multicollinearity. The VIF scores of the two control variables, which we earlier identified to be problematic, also do not exceed the cutoff point.

(28)

28

4.4 Binary Logistic Regression Analysis

This section will continue by elaborating on the models for logistic regression.

Because our dependent variable is a dichotomy of either: 1 Hierarchical entry mode choice, or 0 Hybrid entry mode choice, we will perform a binary logistic regression in SPSS. In this binary logistic regression a positive significant regression coefficient implies that the predictor variable increases the likelihood for hierarchical entry modes. The model can be expressed as:

Where Y is the outcome variable and Z the linear combination of the predictor variables. Considering the fact that we are trying to find a moderating effect of EO on the TC- entry mode relation the results are given in 9 models. The first model specifies how the control variables influence the probability of a hierarchical mode choice, the second includes TC variables, the third includes all the predictors and EO. Models 4-8 include the interactions of EO with the TC variables and model 9 represents the full model.

For the first three models equation (2) is to be applied.

(2)

Where is the intercept, are the coefficients and X the control variables and EO.

For models 4-9 equation 3 is to be used which is the same as model 2 but with the interaction effects of EO with the TC variables.

(3)

Where is the intercept, are the regression coefficients that represent each of the selected variables, X1...Xn are the control and predictor variables, and where is the interaction effect.

(29)

29

For an interactive logistic model with multiple predictors, X and Z, and a product term, XZ, the exponent of the logistic coefficient for X equals a multiplicative factor by which the predicted odds change given a 1-unit increase in X when Z is 0.

(30)

30

Table 1 Discriptives and Correlations

M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 1. Mode Choice ,670 ,471 1 2. Firm Size 55,000 48,195 -,024 1 3. Cultural Distance 1,777 ,777 -,246** ,177* 1 4. Export market experience 18,687 14,588 ,083 ,186* -,085 1 5.International experience 22,078 22,580 -,286**

,389**

,303**

,248**

1 6. Export channel experience 11,575 15,527 -,220** ,237** ,315** ,201** ,718**

1 7. Food Industry ,101 ,302 ,076 ,269** -,113 ,062 ,131 ,075 1 8. Fabricated metal ,207 ,406 ,094 -,101 -,113 -,009 -,174* -,084 -,171* 1 9. Machinery ,223 ,418 -,166* ,102 ,154* ,041 ,153* ,093 -,179* -,274** 1 10. GDP Growth 2,733 1,901 -,040 ,149* ,435** ,047 ,249** ,232** -,130 ,061 ,162* 1 11. Real GDP ,000 ,000 -,171* ,025 -,041 ,044 ,305** ,250** -,126 -,015 ,041 ,108 1 12. GDP per Capita 41150,740 14021,037 ,261** -,183* -,726** ,113 -,164* -,082 ,090 ,078 -,219** -,451** ,032 1 13. EO -,012 ,991 -,156* ,125 ,138 -,167* ,150* ,157* -,104 -,048 ,037 ,068 ,113 -,113 1 14. RnD ,000 11,433 -,084 ,008 ,002 -,120 -,043 ,021 -,099 -,049 -,036 ,059 ,164* ,002 ,232** 1 15. Assetspecificity ,000 ,884 -,121 ,108 ,077 ,119 ,116 ,098 -,033 -,105 ,112 ,137 ,098 -,103 ,137 ,068 1 16. External Uncertainty ,000 1,013 ,055 -,099 ,025 ,001 -,144 -,106 -,044 -,007 ,029 -,028 -,158* -,032 -,134 -,015 ,052 1 17. Frequency ,000 23,346 ,229** ,070 -,090 ,071 -,111 -,080 -,002 -,072 -,031 ,013 ,043 ,108 ,112 ,076 ,070 ,019 1 18. Internal Uncertainty 19,922 12,412 ,118 ,063 ,061 ,105 -,095 ,056 -,098 ,021 -,053 ,039 -,094 ,000 -,050 ,074 ,030 ,191* -,066 1 19. EO*RnD 2,617 8,183 ,117 ,059 ,050 -,035 -,110 -,137 -,071 ,052 -,064 ,073 -,071 -,086 -,166* ,291** ,092 ,052 ,031 ,157* 1 20. EO* Assetspecificty ,119 ,946 -,003 ,078 ,008 -,095 -,055 -,042 ,029 -,023 ,021 ,055 ,019 -,044 ,228** ,062 -,067 -,055 ,090 -,082 -,013 1 21. EO* ExtU -,134 1,017 ,067 ,051 -,035 ,137 ,042 ,020 ,076 ,112 -,080 ,077 ,076 ,050 -,048 ,037 -,059 -,172* ,211** -,058 -,078 ,116 1 22. EO* Frequency 2,579 23,695 ,104 ,059 -,120 -,050 -,106 -,129 ,076 -,019 ,005 -,009 -,078 ,000 -,102 ,022 ,095 ,209** -,239** ,102 ,244** ,012 -,217** 1 23. EO* IntU -,607 13,100 -,098 ,071 -,023 ,188* ,062 ,036 ,067 ,024 -,020 -,032 ,111 ,068 ,076 ,106 -,083 -,055 ,098 -,062 -,021 -,070 ,258** -,069 1

(31)

31

Table 2 Result of Logistic Regression

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9

Control Variables B Wald B Wald B Wald B Wald B Wald B Wald B Wald B Wald B Wald

Firm Size ,006 1,566 ,004 ,757 ,005 ,873 ,004 ,525 ,005 ,866 ,005 ,877 ,004 ,622 ,004 ,823 ,004 0,5

Cultural Distance -,291 ,491 -,388 ,825 -,375 ,747 -,390 ,785 -,374 ,744 -,376 ,751 -,296 ,434 -,305 ,488 -,290 0,417

Export market exp ,016 1,333 ,012 ,693 ,009 ,370 ,010 ,454 ,009 ,370 ,009 ,355 ,011 ,547 ,018 1,246 ,021 1,602

International Experience -,027 4,771 -,020 2,394 -,019 2,192 -,021 2,411 -,019 2,152 -,019 2,199 -,020 2,246 -,020 2,422 -,025 3,381

Export channel exp -,010 ,363 -,014 ,681 -,013 ,526 -,011 ,378 -,013 ,526 -,013 ,523 -,011 ,408 -,016 ,806 -,013 0,505

Food Industry ,421 ,313 ,459 ,363 ,348 ,205 ,519 ,435 ,346 ,201 ,347 ,203 ,381 ,242 ,503 ,398 ,778 0,92 Fabricated metal -,062 ,016 ,025 ,002 ,017 ,001 -,011 ,000 ,017 ,001 ,013 ,001 ,028 ,003 -,035 ,005 -,183 0,115 Machinery -,483 1,179 -,429 ,843 -,444 ,892 -,383 ,656 -,444 ,892 -,443 ,889 -,407 ,726 -,446 ,852 -,348 0,492 GDP growth ,284 4,133 ,283 3,799 ,277 3,516 ,289 3,565 ,277 3,486 ,277 3,487 ,277 3,297 ,260 3,165 ,264 3,093 GDP per Capita ,044 4,271 ,036 2,709 ,035 2,598 ,035 2,800 ,038 2,594 ,035 2,599 ,038 2,853 ,035 2,596 ,039 2,875 Real GDP -,011 2,182 -,012 2,384 -,012 2,319 -,011 1,929 -,012 2,319 -,012 2,323 -,011 1,959 -,009 1,425 -,008 0,92 Independent Variables RnD -,025 1,938 -,021 1,418 -,032 2,663 -,021 1,418 -,022 1,420 -,023 1,583 -,017 ,843 -,034 2,565 Assetspecificity -,286 1,752 -,264 1,445 -,312 1,893 -,263 1,439 -,263 1,433 -,312 1,896 -,346 2,270 -,452 3,6 External Uncertainty ,034 ,029 ,010 ,003 -,003 ,000 ,010 ,002 ,012 ,004 -,033 ,025 -,023 ,012 -,034 0,027 Frequency ,029** 6,832 ,031** 7,163 ,030** 6,818 ,031** 7,151 ,031** 6,927 ,032** 8,320 ,033** 7,940 ,031** 7,504 Internal Uncertainty ,025 2,192 ,024 2,134 ,023 1,821 ,025 2,110 ,024 2,138 ,024 1,961 ,028 2,538 ,027 2,051 EO -,178 ,701 -,099 ,193 -,180 ,650 -,177 ,684 -,074 ,107 -,239 1,104 -,021 0,007 Interaction Variables EO* RnD ,040 2,030 ,047 2,06 EO * Asset specificty ,006 ,001 -,127 0,282 EO * External Uncertainty ,015 ,006 ,284 1,43 EO * Frequency ,016 1,962 ,013 0,922 EO * Internal Uncertainty -,036 3,968 -,050* 5,633 Constant -,522 ,118 -,324 ,043 -,319 ,040 -,510 ,100 -,322 ,041 -,310 ,038 -,674 ,167 -,659 ,170 -,894 0,291 X2 37,462 49,657 50,364 52,261 50,365 50,37 52,317 54,873 60,139

X2 change from model 1 12,195

X2 change from model 2 ,707

X2 change from model 3 2,250 ,001 ,006 1,952 4,509 9,774

Nagelkerke R2 ,263 ,337 ,341 ,354 ,341 ,341 ,353 ,367 ,397

2log likelihood 189,475a 177,280 176,572 174,322 176,571 176,567 174,620 172,063 166,798

Predicted % 75,40% 72,10% 73,70% 74,90% 73,30% 73,70% 72,60% 76,50% 75,40%

(32)

32

5. Results:

In this chapter the results of the binary logistic regressions are presented in section 5.1. We will start by describing the results of model 1 and continue doing that until the last model, model 9. Section 5.2 discusses figure 3 which illustrates an interaction effect. In section 5.3 we will elaborate on ‘the goodness of fit’ of all of the models.

5.1. Logistic regression models

In table 2. The results of the logistic regression are presented in 9 models. Model 1 is our base model and was statistically significant (p<0.01). The control variables explain about 26 % of the variance in the outcome variable, entry mode choice. The control variables International experience, GDP growth and GDP per capita were significant at the p<0.05 level.

‘International experience’ has a significantly negative effect on mode choice. Indicating that the when international experience increases by one unit (1 unit is 1 extra country) the

likelihood of an hierarchical entry mode choice decreases by .027 or approximately 3%. GDP per capita has a significant positive coefficient (0.044) and has an odds ratio of 0.039 indicating that a 1 unit increase (of 1000 euro’s), will increase the propensity of choosing a hierarchical entry mode with 3.9%. GDP growth has a positive coefficient (0.284) and the odds ratio1.328 indicates that when GDP growth increases by one unit (1 percent) the likelihood of choosing a hierarchical export channel will increase by approximately 32%.

In model 2 the TC variables are added to the control variables. This model is also significant and the variables explain 33.7% of the variance in the model. International Experience and GDP growth are not significant anymore. GDP per capita remains significant at the p<0.05 level. Frequency is highly significant at the p<0.01 level with a coefficient of 0.029 and odds ratio of 1.029. This means that when Frequency increases by one the likelihood of an

hierarchical export channel choice will increase by approximately 3%. Hypothesis 1D (SMEs

will tend to prefer hybrid modes of entry when frequency of transactions is low, but tend to prefer hierarchical modes of entry when frequency of transactions is high) can therefore be

(33)

33 specificity. The insignificance of the TC variables therefore do not allow for corroboration hypotheses 1A,B and C.

Model 3 includes the EO variable. The model explains for 34.4% of the variance which is a slight improvement. EO is not significant and has a negative coefficient of -0.158 and odds ratio of 0.837. This would indicate that if EO would increase by 1 the likelihood of choosing a hybrid channel would increase by 16.3%. Frequency and GDP growth remain significant at the p>0.01 and p>0.1 level. The significance of Frequency provides support for hypothesis 1D and insignificance of the other TC variables does not allow for the corroboration of

hypotheses 1A, 1B and 1C.

Models 4 adds the first interaction of EO with RnD. This model explains for 35.4 % of the variance in the dependent variable, entry mode choice. GDP growth and Frequency remain significant with coefficients of 0.289 and 0.03 respectively. All of the other variables are not significant.

Model 5 adds the second interaction of EO with Asset Specificity. This model explains for 34.1 % of the variance. GDP growth and Frequency remain significant with coefficients of 0.277and 0.031respectively. All of the other variables are not significant.

Model 6 adds the second interaction of EO with External Uncertainty. This model explains for 34.1 % of the variance. GDP growth and Frequency remain significant with coefficients of 0.277 and 0.031 respectively. All of the other variables are not significant.

Model 7 adds the second interaction of EO with Frequency. This model explains for 35.3 % of the variance. GDP growth and Frequency remain significant with coefficients of 0.277and 0.032 respectively. All of the other variables are not significant.

(34)

34

5.2. Interaction effect

The graph in figure 3 displays the moderating effect of Internal Uncertainty by EO. This graph must be interpreted with caution however due to the fact that the separate coefficients of EO, Internal Uncertainty, and the constant, are not statistically significant in all of the models. Hypothesis 2 therefore cannot be confirmed.

Figure 3 Graph of moderation effect Internal Uncertainty by EO

When Internal Uncertainty is at its sample mean , a 1 unit increase in EO results in the predicted odds of choosing an hierarchical entry mode, changing by a multiplicative factor of .409 (odds ratio of the constant). When EO increases by 1 unit from its mean, then the

predicted odds of choosing an hierarchical entry mode is (.409)*(.979)=0.4 indicating a 60% increased likelihood for choosing a hybrid entry mode.( Holding Internal Uncertainty constant at its sample mean).

-6 -4 -2 0 2 4 6

Low Int Uncert High Int Uncert

(35)

35

5.3. Goodness of fit

The overall fit of the models can be tested via two methods. The first method, as proposed by Hosmer & Lemeshow (1989), looks at the Nagelkerke R squared statistic. This statistic must be insignificant at the 0.05 level (Menard, 1995) in order for the model to be well fitting. The second method looks at the significance of the difference in Chi squared statistic of the models. If there is a significant improvement in the difference of the model compared to the base model, the model fits well (Menard, 1995).

(36)

36

6. Discussion & Limitations

This chapter will discuss the results of the binary logistic regression. First the results of the moderating effect will be discussed in section 6.1. The findings will be elaborated upon and the limitations will also be presented. Secondly TC variables will be discussed in sections 6.2-6.5 first by elaborating upon the results and then by discussing the limitations.

6.1. Entrepreneurial Orientation

To our surprise EO was not significant in any of our models. Hypothesis 2, EO moderates the

relationship between TC theory and entry mode choice in a way that high EO within firms will increase the propensity that such firms choose hierarchical channels, therefore cannot

corroborated. One argument, for not finding entrepreneurial oriented firms, can be found when looking at the dataset. 115 respondents, of a total of 179, had their most important export market in neighboring countries (Germany and Belgium). Traditional firms

internationalize incrementally before moving to psychically more distant countries(Johanson & Valhne, 1977). This indicates that Dutch manufacturing firms behave more like traditional firms when they export. Additionally EO is most apparent in situations where the

environment changes rapidly and very competitive. EO firms are the ones that can effectively operate in environments that are rapidly changing and competitive and hostile (Kraus, et al., 2012). The majority of respondents in our data set originate from, and export to, countries that have non hostile, and non rapid changing environments, and are psychically and physically close. We have to conclude that our data has an insufficient amount of more psychically and physically distant markets and that could be a reason for not finding a significant value for EO.

We structured EO as a unidimensional construct. This means that not all of the attributes of EO have to be scored high upon in order to get a high average. When one scores 7 on proactiveness and 6 on innovativeness but only a 2 on risk taking the average will be 5 indicating a high EO. But Miller (1981) stated that only firms that score high on all attributes will be considered entrepreneurial. Therefore our way of structuring EO might have

(37)

37 directly or positively affect a dependent variable. Evidence for this line of reasoning can be seen in models 4 to 8 where the interaction coeffiecients are positive for all the TC variables except for Internal uncertainty, which has a negative coefficient. In order for us to be more conclusive, in choosing the correct method of operationalization, we did a confirmatory factor analysis.

When looking at the confirmatory factor analysis in table 5(Appendix A) we see that the variables associated with risk perfectly load on the factor. This provides support for viewing EO as a function of three variables that vary independent from each other. However when we look at Innovativeness and Proactiveness we see that they cross load on each other’s factors. This provides support for a unidimensional constructs with variables that co-vary. When looking at table 5(Appendix A) it is therefore not possible to provide conclusive support for either of the two methods of operationalization of EO.

6.2. Asset specificity

Asset specificity was hypothesized to have a positive relation to entry mode choice in that when asset specificity would increase by 1-unit, the likelihood of choosing a hierarchical export channel would also increase. This however could not be established within this data set. The coefficients for, as well RnD as Asset specificity, were negative and insignificant. Zhao et al (2004) established in their meta analytical review on entry modes, that R&D as a proxy for Asset Specificity does not distinguish the investments made for a specific market entry but only for the firms entire operations. They suggested that future research should ask for R&D specifically devoted to a foreign country instead of overall R&D intensity. A measure of R&D that specifically focuses on a foreign country indicates a projects susceptibility more accurately to the expropriation hazards in a host country.

Our perceptual measure of Asset Specificity did a better job and was marginally significant in model 9. However in all the other models it was insignificant. This measure was added in order to capture the human capital surrounding transactions. This insignificance might originate from the fact that manufacturing firms don’t have as much human capital surrounding transactions as service firms do (Brouthers & Brouthers 2003).

(38)

38

6.3. Internal Uncertainty

Internal Uncertainty was hypothesized to have a positive relationship to entry mode choice in that a 1-unit increase would result in the increase of the likelihood of a hierarchical export channel to be chosen. This however could not be established in this data set. The coefficient was positive and insignificant. Our single perceptual measure might not have captured the entire aspect of internal uncertainty. Other studies used two additional items to directly measure internal uncertainty. We were only able to have one perceptual measure in our survey. The reason for using a perceptual measure instead of the experienced based measures is because it directly measures the uncertainty surrounding the transaction (Brouthers & Hennart, 2007).

Because we only had one direct measure of internal uncertainty we added the indirect

measures as control variables. These control variables asked about the experience of the firm with exporting, internationalization and channel. All these variables proved to be non

significant contributors to the models. A combination of the standardized residuals of the control variables however did a better job. When we compute Internal Uncertainty as a function of the three experience variables (Cronbachs alpha .66) the variable is significant at the .05 and.1 level. The Cronbachs alpha however is below the .7 cutoff point as suggested by Nunnaly (1978) so our choice for the direct measure might have been better.

(39)

39

6.4. Frequency

Frequency was hypothesized to have a positive relationship to entry mode choice, in that a 1 percentage increase in turnover, originating from most important export market, would increase the likelihood of an hierarchical entry mode. Frequency was the only variable of the TC framework that was significant in our models. In operationalizing the variable we did not include questions regarding frequency of transactions to our survey. This is in compliance with the reasoning of Brouthers & Nakos (2004) in that frequency of transactions in entry mode research cannot be measured due to the recurring nature of the transactions.

We used a measure for the percentage of turnover originating from the most important export market. This is consistent with Klein et al (1990) however some (eg Brouthers and Hennart, 2007) debate that such a measure does not capture the frequency of transactions as intended by Williamson (1975,1985). The argument of Brouthers & Hennart (2007) ignores the aspect of volume which can be distilled from Williamson (1985, p 60.) . ‘’the cost of specialized governance structures will be easier to recover for large transaction of a recurring kind’’. This thesis therefore provides support for using volume as a proxy to measure the TC variable Frequency .

6.5. External Uncertainty

(40)

40

7. Conclusion & Suggestions for future research

This research set out to show that EO moderates the TC entry mode choice relation. In order to achieve this, hypotheses were tested, using 179 companies in our sample in a binary logistic regression.

We constructed nine models of which the first three models evaluated the control variables, predictor variables and moderating variable. We found International Experience only to be significant in the first model together with GDP per Capita. GDP Growth proved to be a significant and positive control variable across all models. The TC variable frequency proved to be highly significant and positive indicating that an increase in channel volume means that the chance of a hierarchical entry mode choice will increase. This finding is an important contribution to the TC entry mode research because it points out that previous research (Brouthers & Nakos, 2004; Rindfleish & Heide, 1997; Shervani et al, 2007) ignored volume as proxy for frequency.

In models 4-9 the interaction effects between EO and the TC variables were tested. The results from these tests are inconclusive. There is only one significant interaction effect but because the EO and Internal Uncertainty variables are not significant in that model we cannot state that there is a moderating effect.

Future research must focus on companies that export to hostile and benign environments because those companies are the most likely to have an Entrepreneurial Orientation (Kraus, et al., 2012). Another suggestion is to let the EO constructs of innovativeness, proactiveness, and risk taking, vary independently from each other. The positive and negative beta’s of the interaction effects suggest this. The confirmatory factor analysis in table 5 (Appendix A) also provides partial support for this argument.

(41)

41 This thesis set out to contribute to the literature in two ways:

First by adding Entrepreneurial Orientation as a moderating variable to the relationship between TC theory and chosen entry mode. Even though we could not provide conclusive results this research does suggest that EO could be a moderating variable on the TC entry mode model. The significant interaction of EO with Internal Uncertainty provides an insight which future research can investigate. We suggest that the variables of EO should vary independently from each other rather than taking EO as a multidimensional construct. The reason for this is because the positive and negative coefficients of the interaction effects suggest that the TC variables react to different parts of EO.

(42)

42

Bibliography

Aiken, L. S. & West, S. G., 1991. Multiple regression: Testing and interpreting interactions. Thousand Oaks: Sage.

Anderson, E., 1985. The sales person as outside agent or employee: A transaction cost analysis.

Marketing Science, 84(August), pp. 234-254.

Anderson, E., 1988. Transaction costs as determinants of opportunism in integrated and independent sales forces. Economic Behavior and Organization, 9(May), pp. 247-264.

Anderson, E. & Coughlan, A. T., 1987. International market entry and expansion via independent or integrated channels or distribution. Journal of Marketing, 51(january), pp. 71-82.

Anderson, E. & Coughlan, A. T., 1987. International market entry and expansion vie independent or integrated channels of distribution. Journal of Marketing, 51(January), pp. 71-82.

Anderson, E. & Gatignon, H., 1986. Modes of foreign entry: A transaction cost analysis and propositions. Journal of International Business Studies, 17(3), pp. 1-26.

Anderson, E. & Schmittlein, D., 1984. Integration of the sales force: An emperical examination. Rand

Journal of Economics, 15(Autumn), pp. 385-395.

Aulakh, P. & Kotabe, M., 1997. Antecedents and performance implications of channel integration in foreign markets. Journal of International Business Studies, 28(1), pp. 145-175.

Balakrishnan, S. & Wernerfelt, B., 1986. Technical change, competition and vertical integration.

Strategic Management Journal, 7(July/August), pp. 347-359.

Barkema, H. G. & Vermeulen, F., 1997. hat differences in the cultural backgrounds of partners are detrimental for international joint ventures?. ournal of international business studies, pp. 845-864. Barkema, H. G. & Vermeulen, F., 1998. International expansion through start up or acquisition: A learning perspective. Acadamy of Management Journal, 41(1), pp. 7-26.

Brouthers, 2002. Institutional, cultural and transaction cost influences on entry mode choice and performance. Journal of International Business Studies, 33(2), pp. 203-221.

Brouthers, K. & Brouthers, L., 2003. Why service manufacturing entry mode choices differ: the influence of transaction cost factors, risk and trust. Journal of Management Studies, pp. 1179-1204. Brouthers, K., Brouthers, L. & Werner, S., 2003. Transaction cost-enhanced entry mode choices and firm performance. Strategic Management Journal, Volume 24, pp. 1239-1248.

Brouthers, K. & Hennart, J., 2007. Boundires of the Firm: Insights From International Entry Mode Research. Journal of Management, 33(May), pp. 395-423.

(43)

43 Bucklin, L. P. & Sengupta, S., 1993. Organizing succesfull co-marketing alliances. Journal of Marketing, 57(April), pp. 32-46.

Burns, R. B. & Burns, R. A., 2008. Business Research Methods and Statistics using SPSS. 1 ed. London: Sage Publications Ltd.

Cahill, D., 1996. Entrepreneurial orientation or pioneer advantage. Acadamy of Management Review, 21(3), pp. 603-605.

Canabal, A. & White 3, G., 2008. Entry mode research: Past and Future. International Business

Review, Volume 17, pp. 267-284.

Carter, R. & Hodgeson, G. R., 2006. The impact of imperical tests of transaction cost economics.

Strategic Management Journal, 27(Febuary), pp. 461-476.

Chan, C. & Makino, S., 2007. Legitimacy and multi-level institutional environments: Implications for foreign subsidiary ownership structure. Journal of International Business Studies, Volume 38, pp. 621-638.

Coase, R., 1937. The nature of the firm. Economica, Volume 4, pp. 386-405.

Cohen, J. & Cohen, P., 1983. Applied multiple regression/correlation for the behavioral sciences. 2 ed. Hillsdale, NJ: Lawrence Erlbaum Associates.

Contractor, F. J., 1984. Choosing between direct investment and licensing: Theoretical considerations and empirical tests. Journal of International Business Studies, pp. 167-188.

Contractor, F. & Kundu, S., 1998. Modal choice in a world of alliances: analyzing organizational forms in the international hotel sector. Journal of International Business Studies, Volume 29, pp. 325-358. Coviello, N. & McAuley, A., 1999. nternationalisation and the smaller firm: a review of contemporary empirical research. management international review, pp. 223-256.

Covin, J., Green, K. & Slevin, D., 2006. Strategic process effects on the entrepreneurial orientation - sales growth rate relationship. Entrepreneurship Theory and Practice, 30(1), pp. 57-81.

Covin, J. & Slevin, D., 1989. Strategic management of small firms in benign environments. Strategic

Management Journal, 10(1), pp. 75-87.

CPB, 2009.

Davidson, W. H., 1982. Global strategic management. New York: John Wiley.

de Clercq, D., Sapienza, H. & Crijns, H., 2005. The Internationalization of Small and Medium-Sized Firms. Small Business Economics, Volume 24, pp. 409-419.

Delios, A. & Beamish, P., 1999. Ownership strategy of Japanese Firms: Transactional, institutional, and experience influences. Strategic Management Journal, Volume 20, pp. 915-933.

Erramilli, M. & D’Souza, 1993. Venturing into foreign markets: the case of the small service firm.

Entry Mode Choices of Dutch Manufacturing SMEs: A Transaction Cost Perspective and the Moderating Effect of Entrepreneurial Orientation