Country-level predictors of the prevalence of Born
Globals
Exploring the relation between domestic market factors and
Born Globals
MSc Thesis by
Marthijn Voerman
Department of Economics and Business
Rijksuniversiteit Groningen
Academic Supervisors: Drs. A. Visscher
Dr. S.R. Gubbi
Information
Author Marthijn Voerman
Address Gorthoek 8
Abstract
This master thesis explores the relation between domestic market factors and Born Globals (BGs), which are firms that internationalise from or near inception. This particular type of firm is chosen because the EU expects this type of firm to be highly effective in the recovery of the economic and labour market in the aftermath of the global financial crisis. However, even though hopes are high when it comes to BGs, little is known regarding in which circumstances they are most likely to appear. Therefore it is hard to develop policies to stimulate BGs. This thesis is an initial attempt to narrow this research gap, by exploring which domestic market factors have an influence on the prevalence of BGs in a country. These factors have been subdivided into enablers, factors that facilitate firms’ internationalisation; and deficiencies, shortcomings in the domestic market that push firms towards internationalisation. We found that the following enablers were found to have a positive influence on the prevalence of BGs: the level of knowledge intensity in the domestic market, as well as membership of the Eurozone and the strength of the regulatory environment. A country’s innovative capacity surprisingly has a negative influence. From the deficiencies the level of competition in the domestic market and the income level proved to be positively related to the prevalence of BGs, whereas domestic market size, market growth and the level of corporate tax rate have a negative impact. Finally, we found that the level of infrastructure and the prevalence of foreign ownership in the domestic market do not play a significant role. Overall we found that domestic market deficiencies explain more of the variance in the prevalence of BGs than domestic market enablers, which could indicate that internationalisation is a necessity for many firms, in order to compensate for deficiencies in the domestic market.
Key words:
Table of Contents
Information iii
Abstract iv
List of tables vii
List of figures vii
List of abbreviations viii
1. INTRODUCTION 1
2. THEORY AND HYPOTHESES 3
2.1 Development of Internationalisation Theories 3
2.2 Born Globals 4
2.3 Reasons for Early Internationalisation 5
2.4 Firm-Level Enablers of Early Internationalisation 6 2.5 Domestic Market-Level determinants of Early Internationalisation 9
2.6 Conceptual Model 17 3. EMPIRICAL ANALYSIS 18 3.1 Sample Design 18 3.2 Statistical Technique 19 3.3 Variables 19 3.4 Statistical Model 26 3.5 Modelling Procedure 27 3.6 Estimation Method 27
3.7 Evaluation of Method Assumptions 27
4. EMPIRICAL RESULTS 29 4.1 Homoskedacity 29 4.2 Endogeneity 29 4.3 Multicollinearity 30 4.4 Normality 31 4.5 Results of Hypotheses 32 4.6 Robustness Check 35
6. LIMITATIONS 40
7. RECOMMENDATIONS FOR FUTURE RESEARCH 41
REFERENCES 43
APPENDIX
A. Countries included in the sample 51
B. Components of the Global Innovations Index score 52
C. Outlier test – excluded values 53
D. Heteroskedacity test – Breusch Pagan 54
E. Multicollinearity test – VIF test 55
List of tables
1.1 BGs in the EU 1
3.1 Sample criteria 18
3.2 Overview of variables 25
4.1 Descriptive statistics 32
4.2 Results of regression analysis 34
4.3 Answers to hypotheses 35
A.1 Countries included in the sample 51
C.1 Excluded outliers 53
E.1 VIF test results 55
F.1 Robustness check – omitted dummies 56
F.2 Robustness check – sample cut 57
List of figures
2.1 Conceptual model 17
List of abbreviations
BG Born Global
EMU European Monetary Union EOS Executive Opinion Survey EU European Union
GCI Global Competitiveness Index GDP Gross Domestic Product
GEM Global Entrepreneurship Monitor IB International Business
MNE Multinational Enterprise NV New Venture
1. INTRODUCTION
In the aftermath of the global financial crisis the European Union (EU) is attempting to facilitate rapid economic and labour market recovery. In order to do so, it relies on firms that have the potential to grow rapidly, thereby creating the jobs required for economic recovery. On type of firm the EU is looking at is the Born Global (BG). There are many different definitions for BGs, such as “business organisations that, from inception, seek to derive significant competitive advantage from the use of resources and the scale of output in multiple countries” (Oviatt and McDougall, 1994) or “entrepreneurial start-ups that, from or near their founding, seek to derive a substantial proportion of their revenue from the sale of products in international markets” (Knight and Cavusgil, 2004). The EU classifies BGs as firms that generate at least 25% of their revenues abroad within 3,5 years (Eurofound, 2012). The EU looks at these firms because they are expected to internationalise and grow rapidly, thereby creating new jobs. Several scholars found evidence for this relation (e.g. Acs and Armington, 2006; Wright et al, 2015). However, as of yet, there is little understanding of the circumstances under which they are most likely to appear (e.g. Knight and Liesch, 2016). This will be the focus of this research. The table below depicts the share of BGs of the total number of young firms in the EU (Eurofound, 2012). As can be seen, the share varies greatly, from 12,89% in Finland to 41,84% in Luxemburg, while on first sight no clear pattern can be detected.
Table 1.1: BGs in the EU Source: Eurostat, 2012
If the EU wants to achieve recovery from the crisis through the establishment of BGs, it needs to understand where these differences come from. In this paper we will conduct an exploratory research, attempting to identify which domestic market factors affect the share of BGs, and whether this effect is positive or negative. We will identify these factors by conducting a thorough literature review, after which we will test their
0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 40.00% 45.00%
influence with a multiple regression analysis. This will enhance our understanding of BGs, and of the environment in which they are most likely to appear in particular. This knowledge can serve as a stepping-stone for future research, allowing scholars to focus on a set of relevant domestic market factors. Moreover, it can serve as a starting point for European policy making to foster BGs, and allow national governments to create an environment in which BGs are more likely to appear. Finally, it could give entrepreneurs with international aspirations insights as to which environment is most suitable for pursuing these aspirations.
2. THEORY AND HYPOTHESES
In this section we will present our literature review. We start with summarising the different internationalisation theories, after which we will look at Born Globals in particular. Next, we will look at the firm-level enablers of early internationalisation, which will give us further understanding of BGs. Finally, we will summarise the literature regarding the influence of domestic market factors on early internationalisation, based on which we will formulate our hypotheses. These will be summarised in our conceptual model, which provides an overview of the analysis that we will conduct.
2.1 Development of Internationalisation Theories
One of the earlier theories regarding the internationalisation process of the firm is the Uppsala model from Johanson and Vahlne (1977). They studied the internationalisation process of Swedish firms, and found that they typically internationalise incrementally; it frequently started with ad hoc exports, followed by formalisation by using agents in the foreign market. As sales grow, these agents would be replaced by the firm’s own sales organisation, and as sales continued to grow, the firm would start producing in the foreign market to overcome trade barriers. They also found that the choice of markets to enter would change over time; initially firms would enter markets that are psychically close, and over time they would start entering more psychically distant markets.
Hedlund and Kverneland (1985) investigated the expansion of Swedish firms to Japan. They found that diminishing differences between industrialised nations, and firms’ improved ability to handle the complexity of international business led to more direct and rapid expansion of firms than the above-named Uppsala model. This can be illustrated by the fact that more than half of the firms in the sample skipped the establishment of a sales subsidiary, and moved directly from having a sales agent to establishing a manufacturing subsidiary.
element that drives the process of internationalisation. Firms are likely to internationalise when important partners are located in/expanding to other markets, and when these partners have a strong position in these markets. The degree of internationalisation is expected to depend on the share of the firm’s network that is located abroad. The choice of market depends on the location of the firm’s partners, and if the firm has no relevant partners, it will choose markets in which it is relatively easy to establish a network.
These three studies show that firms internationalise incrementally, albeit with an increasing speed due to changing environment and improved firm processes and understanding of internationalisation, leading to a lower commitment threshold and more firms being able internationalise more rapidly (Knight and Liesch, 2016).
Rennie found an interesting dichotomy when studying the internationalisation patterns of Australian SMEs. On the one hand there are those who establish a strong position on the domestic market and then gradually start internationalising, following the pattern described before. But on the other hand there are those who compete on a global level from the moment of inception, without establishing a dominant position in the domestic market, nor gradually expanding to (psychically) nearby markets (Rennie, 1993). The firms in the second category were branded “Born Globals” (BGs). These firms managed to successfully compete with large multinationals, and achieved high growth rates. Similar firms, with internationalisation patterns deviating from the Uppsala pattern were found in other studies (Knight & Cavusgil, 1996; Oviatt & McDougall, 1994), confirming that the incremental growth model does not hold for all firms. Prior research has found three main reasons for the appearance of BGs, being new market conditions, advances in technology and managerial change (Kudina et al, 2008). The so-called Born Global is the type of firm on which this research will be focused, since it is expected to be able to deliver the kind of growth the EU needs to boost its labour market and economy (Eurofound, 2012). In the next paragraph we will look further into existing literature about BGs.
2.2 Born Globals
As mentioned before, in 1993 Rennie was the first to identify a new type of firm, which deviates from the traditional internationalisation pattern, and named it Born Global. Even though more than 20 years have past ever since, scholars have not agreed on a single definition for BGs (Aspelund et al, 2007; Knight and Liesch, 2016).
seek to derive significant competitive advantages from the sale of output in multiple countries” (Knight and Cavusgil, 2004; Oviatt and McDougall, 1994; Rennie, 1993). However, few firms internationalise from inception, and most of them need a couple of years to start doing so. Taking this into account, scholars have come up with multiple ways of operationalising the definition of BGs. Rennie defined them as firms who start internationalising within 2 years, and who have a minimum share of 75% of foreign sales as a percentage of total sales (Rennie, 1993). Knight and Cavusgil (1996) defined them as firms that internationalise within two years, and achieve at least 25% of their sales abroad within ten years. Madsen et al (2000) defined them as firms that earn at least 25% of their sales abroad within three years. McDougall et al (2003) had only one condition, being that firms start internationalising within six years. The most widely accepted definition, which is also being used by the EU, entails firms who are maximally 3,5 years old and earn at least 25% of their sales abroad (Eurofound, 2012).
2.3 Reasons for Early Internationalisation
The reasons for new ventures to internationalise in an early stage, hence becoming a BG, vary widely. Kudina et al (2008) analysed high-tech BGs in Sillicon Fen, UK and found that their main reason to go global was the inadequacy or even non-existence of the domestic UK market. On the other hand, their home market offered a strong knowledge base, allowing companies to develop a competitive advantage based on sophisticated techonology. Another reason for these firms to go global was to serve global customers, which dominate the high-tech industry. Finally, the companies were often operating in niche industries in which they wanted to set the global standard, which prompted them to expand rapidly. Interestingly, they found that British BGs did not face the choice whether to go overseas or not: they had to go global, otherwise they simply would not have survived.
1997). Since these three drivers, and the first two in particular, depend on relatively new developments, this also explains the increase in the number of BGs in the past two decades (Knight and Cavusgil, 2004).
2.4 Firm-Level Enablers of Early Internationalisation
In this section we will give a résumé of the existing literature regarding the firm-level factors that influence a firm’s propensity to become a BG. These factors will not be included in our research, since many scholars have already investigated BGs on the firm level, and our focus lays on the country-level factors. However, these firm-level factors will allow us to gain a better understanding of BGs, and will serve as a stepping-stone for the formulation of the country-level factors.
2.4.1 Management
Since we are looking at very young firms, management often mainly consists of the firms’ founders. The characteristics of these founders strongly influence whether young firms internationalise, and at what speed they do so (e.g. Acedo & Jones, 2007; Gilbert et al, 2006; Madsen and Servais, 1997; McDougall et al, 2003). The founders of those new ventures that become BGs are found to have more industrial and international experience than their counterparts who do not become BGs (Acedo & Jones, 2007; Aspelund et al, 2007; McDougall et al, 2003). Their experience in International Business (IB) allows these founders to see more opportunities and to reduce the perceived distance to other markets, thereby reducing the perceived risk of entering these markets. Moreover, they can use their existing networks and market knowledge to facilitate market entry (Aspelund et al, 2007). Foreign market knowledge is generally accepted to lead to faster internationalisation (Aspelund et al, 2007; Zhou, 2007). However, whereas some scholars believe the possession of this knowledge is the main driver for founders to pursue international expansion (e.g. Acedo & Jones, 2007; Aspelund et al, 2007; McDougall et al, 2003), Zhou (2007) believes it is the founders’ entrepreneurial proclivity rather than previous experiences. This proclivity does not only lead to a pursuit of internationalisation, but also to the acquisition of knowledge, which is essential for BGs to compete on a global level (Knight and Cavusgil, 2004).
2.4.2 Organisational resources and competencies
Madsen et al, 2000). This is in line with the findings of Rammer and Schmiele (2008), who found that German firms with a technological advantage resulting from owning patents or trademarks have a higher propensity to internationalise in an early stage. This knowledge also allows them to create products, which are very hard to imitate. Imitation would cost competitors years of development, allowing the BG to remain ahead of competition (Kudina et at, 2008). Interestingly, Kudina et al (2008) found that the amount of R&D spending did not affect performance, but the allocation did: the acquisition of new technology in foreign markets was found to have a positive impact. This could indicate another incentive for going global, albeit on the input side.
Besides knowledge, the competitive advantage of BGs can also be found in their production capacity, which is either more efficient or cheaper than that of competitors (Fan & Phan, 2007). However, since these firms are still young, they tend to lack substantial human and financial resources, as well as equipment, plant and other physical resources. Therefore, it is more likely that they will rely upon leveraging intangible knowledge-based assets rather than e.g. production capacity (Knight and Cavusgil, 2004). Due to this scarcity of resources BGs emphasise controlling assets rather than owning them, which in turn allows them to grow faster, and to change more rapidly to changing customer needs. As a result, they depend on supplementary competencies from other firms (Madsen and Servais, 1997; Oviatt and McDougall, 1994).
2.4.3 Strategy
BGs are found to pursue a more aggressive strategy than those New Ventures (NVs) who do not become a BG. Not surprisingly, they put more emphasis on expansion, for which they use more distribution channels. They compete on the basis of differentiation rather than costs, and typically have a niche focus (Aspelund et al, 2007; Madsen and Servais, 1997). Although it is often claimed that BGs can mainly be found in high-tech industries, Madsen and Servais (1997) show that they can be found in a wide array of industries, from tech to trade. Competition in the industries in which BGs operate is mostly based on differentiation and the firms’ ability to be flexible and to adapt to ever changing needs of the market (Kudina, 2008). BGs typically emphasise product innovation, quality, service and marketing (McDougall et al, 2003).
of the management. According to Johanson and Vahlne (2009) firms will start internationalising through their network. Therefore they will opt for markets in which their partners already have a strong position. If the firm has no partners abroad, it might choose any possible market. Madsen and Servais (1997) found that high-tech BGs focus their sales and marketing activities on lead markets, since they depend on access to lead customers and the newest technological developments. BGs with a focus on trade will opt for markets in which their present customers operate. Baum et al (2001) found that BGs often pursue niche strategies and first-mover advantages, which force them to internationalise rapidly.
2.4.4 Network relationships
Research has shown that it is hard for firms to innovate in isolation without being engaged in collaborative activities (Freel and Harrison, 2006). BGs in particular strongly depend on knowledge and innovation (Acedo & Jones, 2007, Kudina et al, 2008) hence have a lot to gain from participating in networks. This is confirmed by several studies (e.g. Dyer and Hatch, 2006; Gilmore et al, 2006), which show that the sharing of information facilitates new product development, which in turn enables BGs to compete in international markets.
Business networks comprise relationships with clients and suppliers, as well as industrial and governmental associations (Welch et al, 1998). Oviatt and McDougall (1994) claim that one of the main predictors for NVs to become BGs is the access to such an international network. The efforts in this network are not only restricted to sales activities, but are extended by procurement and production activities abroad, which create more ties in other markets, allowing for faster and more successful market entry. Having ties in other markets has been proven to be an antecedent of early internationalisation in several studies (e.g. Freeman et al, 2012; Johnson, 2004).
expected to be the mechanism behind the acquisition of knowledge abroad, which is mentioned in the previous paragraph. This opinion is shared by Johanson and Vahlne (2009), who found that New Ventures (NVs) are typically relatively small due to their short existence, and therefore do not have the capacity yet to internalise all the (creation of) knowledge. Therefore they rely on the establishment of networks to gain access to new knowledge. This does not only hold for knowledge; as mentioned before, BGs tend to rely on supplementary competences sourced from other firms in general (Madsen and Servais, 1997).
Johanson and Vahlne (2009) argue that a firm’s international network can prompt internationalisation for the following reasons. First, the firm may become aware of opportunities in other markets through partners in its network. Partners are believed to rely on the same knowledge base, making the partner’s knowledge base and the accompanying opportunities in other markets highly relevant. Second, partners may request a firm to follow, a request that can be adhered to in order to show commitment to the relationship. Once the decision to internationalise has been made, strong ties with foreign distributors can facilitate market entry. This is not limited to logistics, but also entails provision of market intelligence, cultivating market segments, and gaining access to local networks (Knight and Cavusgil, 2004).
2.5 Domestic Market-Level Determinants of Early Internationalisation
its profits. Second, domestic market factors may enable firms to gain access to resources and capabilities, allowing them to attain competitive advantages. Based on this dichotomy, the factors will be subdivided in domestic market deficiencies and domestic market enablers.
Domestic market deficiencies
First we will present the various domestic market deficiencies, which have been identified in the existing literature. The factors constitute push factors, forcing firms to internationalise.
2.5.1 Domestic market competition
Past research has shown that firms who face strong rivalry in their domestic market tend to adopt more aggressive strategies, and perform better on an international level (Porter, 1997; Zhao and Zou, 2002). These findings are in line with the spil lover effects, which occur when firms face many other competing firms in its proximity (Capello and Fagian, 2005). In their report, the World Economic Forum states that a healthy market condition is important for the development op productive firms, since it ensures that the most efficient firms are those that thrive (Schwab, 2016). Freeman et al (2012) analysed Australian exporting SMEs from regional and metropolitan areas, to see whether local competitive rivalry affected the firms' ability to develop export-related resources and capabilities. On a national level the regional level of competitive rivalry was found not to have a significant impact, since the firms were operating on an international level. On an international scale however, the authors expect the level of competitive rivalry in the domestic market to have a significant positive influence, and mark this as a gap in the existing literature which should be filled in future research.
Rammer and Schmiele (2008) found that having a low number of competitors in the domestic market is positively related to increasing exports of new products, and negatively to increasing exports of old products. Johnson (2004) found that avoiding fierce domestic competition is a determinant for early internationalisation, albeit with limited explanatory power. This is in line with the findings of Oesterle (1997), who found that intensive competition in the domestic market drives margins down, providing an incentive to firms to start internationalising.
Following the arguments stated above, we expect that competition in the domestic market will be positively related to the share of BGs per country. The reason for this could have different natures; competition might force firms to adopt more aggressive strategies to stay ahead of competition, lead to spill over effects, stimulate innovation or decrease margins. One could argue this factor to be an enabler rather than a deficiency in the market since it might provide firms with e.g. spill over effects. However, since it
forces firms to become more aggressive, decreases their margins and to innovate, in this
context it will be categorised as a deficiency.
H1: Competitive rivalry in the domestic market is positively related to the share of BGs in total NVs.
2.5.2 Market size
Several studies (e.g. Crick and Jones, 2000; Kudina et al, 2008; Oesterle, 1997; Schwab, 2016) have shown that being located in a small domestic market makes it more likely for NVs to start internationalising in order to compensate for the lack of demand in the domestic market. Hence they state that the smaller the home market, the more likely it is that NVs will start internationalising.
Evans et al (2008) found that American firms tend to take longer to exploit the opportunities in their domestic market before internationalising than their British counterparts, because they perceived their domestic market as having more opportunities than the foreign one. This is in line with Johnson's (2004) findings, who compared early internationalisation of high-tech start-ups in the USA and UK. In the latter country the small domestic market was found to be an important driver for internationalisation, whereas only one out of 52 American companies indicated it to play a role.
The existing literature agrees on the expectation that firms from smaller markets will internationalise faster hence have a higher probability of becoming BGs. This research will follow this expectation.
H2: Domestic market size is negatively related to the share of BGs of total NVs.
2.5.3 Lack of market growth
do not display growth, tend to start internationalising earlier on. Aspelund et al (2007) expect the growth rate of the domestic market to influence the propensity of firms to internationalise, for the same reason as Evans et al.
H3: Domestic market growth is negatively related to the share of BGs of total NVs.
2.5.4 Average income – purchasing power
Oesterle (1997) found that a domestic market with a relatively low purchasing power can constitute a push for firms to start internationalising, in order to look for a market for their products elsewhere. The lower the purchasing power, the less suitable the market is for new innovations, the earlier NVs will start internationalising.
H4: The level of average income in the domestic market is negatively related to the share of BGs of total NVs.
2.5.5 Corporate tax rates
Another domestic market deficiency that could push firms towards internationalisation is high corporate tax rates in the domestic market. It is generally acknowledged that corporate tax rates in host markets affect the internationalisation strategy of firms, and the choice of location in particular (e.g. Devereux & Freeman, 1995; Devereux & Griffith, 1998; Hebous et al, 2011). A presence in countries with lower tax rates allows firms to decrease their effective tax rate through cross-border profit shifting (Haufler & Schjelderup, 2000). Devereux and Freeman (1995) found that the difference in tax rates has no significant influence on the division of investment between domestic and foreign markets.
Since the corporate tax rate in the host country is considered a significant indicator for the location choice of internationalisation, it is possible that the tax rate in the domestic market also plays a role. To the best of our knowledge, the domestic corporate tax rate has not been identified as a push factor for internationalisation of yet. However, since firms from a country with a high corporate tax rate potentially have more to gain from expanding to a country with a low tax rate, we expect that NVs from countries with a high corporate tax rate will start internationalising in an earlier stage.
Domestic market enablers
In this section we will present the domestic market enablers. These are factors that allow firms to internationalise, by enabling them to compete on an international level.
2.5.6 Level of infrastructure
There are several studies that indicate that the propensity of firms to internationalise is affected by the level of infrastructure in its location. North and Smallbone (2000) for example found that firms in locations with poorly developed infrastructure and services have less propensity to internationalise than firms in more developed locations. This is in line with the findings of Mariotti and Piscitello (2001), who found that the firm's location and its infrastructure and accessible resources are strong predictors of its propensity to internationalise early on.
According to Westhead et al (2004) Infrastructure can provide firms with several benefits, such as favourable supply conditions, access to financial institutions, technology partners, specialised labour and, very important in this context, export-related resources and capabilities. This is confirmed by Freeman et al (2012), who found that the level of export-related infrastructure is positively related to firms' resources and capabilities. They acknowledge that different types of exporting firms require different types of infrastructure, e.g. firms from the high-tech- and heavy machinery industry. However, they state that it is very likely that there is some degree of cross-over, e.g. in communication infrastructure.
Finally Hart and McGuiness (2003) state that transport and logistics are the main
challenge for internationally operating firms, for which they obviously strongly depend on the level of infrastructure in a country (Hart and McGuinness, 2003).
In general the existing literature agrees on infrastructure being one of the enablers of early internationalisation. This assumption will be followed in this research.
H6: The level of infrastructure in the domestic market is positively related to the share of BGs in total NVs.
2.5.7 Knowledge intensity
firm needs to acquire new information, which might not be included in its existing knowledge base (Ghoshal, 1987). This information includes experiential knowledge about foreign business practices and international norms, as well as general experience with organizing competition on an international level (Eriksson et al, 1997). Scholars following the traditional internationalisation theories, such as the Uppsala Model, argue that this knowledge can only be created by operating abroad, and that as this knowledge increases, the commitment to the respective market will also increase (Johansson & Vahlne, 1977; 1990; Eriksson et al, 1997).
Except for one article, written by Kudina et al (2008), no literature could be found linking the knowledge in the domestic market to firms internationalising. Kudina et al (2008) found that high-tech BGs in the Greater Cambridge Area Cluster in the UK conduct almost all their R&D activities in the domestic market, thereby making optimal use of the strong knowledge base in their home market, with highly renowned institutions such as the University of Cambridge. This allowed them to gain a knowledge-based competitive advantage, on which BGs often rely as we have mentioned before (Acedo & Jones, 2007; Knight and Cavusgil, 2004; Kudina et al, 2008).
So even though Kudina et al link the knowledge intensity to a different kind of knowledge than e.g. Johansson & Vahlne, they do show that a strong knowledge base in the domestic market may allow firms to achieve knowledge-based competitive advantages, on which BGs often rely.
Moreover, we also believe that a strong knowledge base, rooting partially in education, can provide a basis for the type of knowledge required for internationalising, e.g. language education, knowledge about different markets and knowledge about internationalisation in general, thereby reducing the barriers to internationalisation. Several studies (e.g. Autio et al, 2000 : Erkko et al, 2000) found that the knowledge intensity within a firm leads to faster internationalisation. However, in order for firms to make use of knowledge, it needs to internalise or gain access to it. We expect that being embedded in a home market with high knowledge intensity can provide NVs with both the knowledge required for developing knowledge-based competitive advantages and the basis for the knowledge required for internationalising. As a result, we expect that NVs from countries with a strong knowledge base will have a higher probability of becoming BGs.
2.5.8 Prevalence of foreign ownership
Johnson (2004) found that having international contacts was positively related to early internationalisation. When firms are located nearby subsidiaries of foreign firms, they are more conscious of and responsive to opportunities in foreign markets. This also increases their understanding of the standards required for doing business on an international level, and the number of inquiries they get from foreign firms (Fernhaber et al, 2008). Aspelund et al (2007) expect the openness of the economy to have a positive impact on the propensity of firms to internationalise, for the same reasons mentioned by Johnson.
The presence of foreign firms in a firm’s environment will provide access to contacts and information in foreign markets, making it possible to identify possibilities in these markets. The more foreign firms are located in a firm’s domestic markets, the more possibilities in other markets it could perceive.
H8: prevalence of foreign ownership in the domestic market is positively related to the share of BGs in total NVs.
2.5.9 Membership of the Eurozone
Membership of the Eurozone is widely accepted to have a positive impact on the internationalisation of firms. In his study of export behaviour of German firms before and after the establishment of the Economic and Monetary Union (EMU), Bagci (2012) found that it increased the likelihood of SMEs to internationalise. The main reason he identified was that the EMU had omitted the exchange rate risk, a risk that is hard to counter for small firms since they do not have the mass to hedge, nor do they have the knowledge to do so. This risk prevented especially small and young firms from exporting, therefore the effects of the EMU were found to be stronger for these firms. By omitting the exchange rate risk, the EMU decreased the costs and risks of internationalising, and by doing so allowed them to internationalise at an earlier stage (Bagci, 2012; Baldwin et al, 2008; Esteve-Pérez et al, 2011).
The existing literature seems to agree on the positive impact of membership of the Eurozone on the propensity to start internationalising. We expect to come to the same conclusion.
H9: membership of the Eurozone is positively related to the share of BGs in total NVs.
2.5.10 Regulatory environment
In their literature study Sousa et al (2008) found that the regulatory environment is one of the determinants which is most cited when it comes to the export propensity of firms. However, they refer to the regulatory environment in the target market, and relate to e.g. the protection of the exporting firm’s patents. Aspelund et al (2007) expect the regulatory environment in the home market to have an impact as well, since a strong regulatory environment will protect the intellectual property of innovating firms, thereby decreasing the risks attached to innovating and stimulating innovation. As mentioned before, BGs depend on knowledge and innovation, in order to be able to compete on international markets. This expectation is confirmed by an earlier study of Gilbert et al (2006), who found that firms located in regions with strong legal protection tend to apply for more patents.
A strong regulatory environment allows firms to innovate without having to fear that their intellectual property is being imitated. As a result, they will have more possibilities to establish knowledge-based competitive advantages than their counterparts who are based in a weaker regulatory environment. We expect this head start to result in these firms internationalising in an earlier stage, since they are more likely to have the capacity to outperform foreign firms.
H10: the level of the regulatory environment in the domestic market is positively related to the share of BGs in total NVs.
2.5.11 Innovative capacity of the economy
knowledge and human resources, and the legal protection of these patents (Gilbert et al, 2006).
Since BGs often rely on innovations to outperform their competitors in other markets, it is quite likely that their innovative capacity affects their decision to internationalise. We expect that being embedded in an innovative domestic market will increase these firms’ innovativeness, thereby increasing their likelihood to become BGs.
H11: The level of innovative capacity in the domestic market is positively related to the share of BGs of total NVs.
Below we will summarise the hypotheses formulated above in our conceptual model, providing an overview of the research that we will conduct.
2.6 Conceptual Model
The hypotheses formulated in this chapter compose the conceptual model displayed below. The domestic market characteristics explain the relation between the domestic market and the share of BGs of total NVs in the respective country. The domestic market characteristics are subdivided into two groups; the deficiencies in the domestic market, and the factors that enable early internationalisation. We control for the three years in our model, thereby controlling for external factors that influence the share of BGs worldwide.
3. EMPIRICAL ANALYSIS
After presenting the literature review, hypotheses and our conceptual model, which will guide our research, this chapter will present the methods and data that will be used. First, the data, sample and sampling design will be explained. Next, the measurements used for the dependent and independent variables will be explained, and finally a detailed description of our methodological approach and our statistical model will be provided.
3.1 Sample Design
Our research is based on a question raised in a report from the European Commission, hence it would be logical to restrict the sample to the EU member states. However, since this would leave us with an insufficient amount of data for a solid regression analysis with relatively little variance, we have decided to expand the sample to all the countries in the world, filtered for the following criteria:
Sample criteria
1. Region Worldwide
2. Data availability (DV) BGs as percentage of total number of NVs 3. Data availability (IV) Data regarding the 11 IVs
4. Years 2014, 2013 or 2012
Table 3.1.: Sample criteria Source: own results
years; some only had the share of BGs of total NVs for one or two years. In the end, we have 203 observations.
3.2 Statistical Technique
In order to test the relation between the Dependent Variable (DV) and the Independent Variables (IVs) we will run a multiple linear regression analysis, which is suitable for a model with a continuous DV, and several explanatory variables. Doing so will provide us with insights as to whether the DV and the explanatory variables are related, and with estimates of their correlation. The model will be fitted using the Ordinary Least Squares (OLS) approach. By doing so, the difference between the values predicted by the model and the values observed is being minimalized.
A multiple linear regression comes with several key assumptions, for all of which we will conduct tests in a later stage: Homoscedacity, Endogeneity, Multicollinearity and Normality.
We will not lag our IVs hence we will measure the relation of the share of BGs with the domestic market factors of that same year. We opted to do so because we expect that our factors will have an immediate effect on the behaviour of entrepreneurs. To give an illustration, if an economy is growing we deem it unlikely that an entrepreneur will not respond to this, and assume him to anticipate immediately.
In the next paragraph we will present our approach for constructing the variables for our model, followed by our statistical model and a number of tests to check whether the model is appropriate.
3.3 Variables
In this paragraph we will present the approach that has been used to construct each single variable, and clarify why the respective approach has been chosen. All of the variables relate to observational data.
3.3.1 Dependent variable
The dependent variable in this research is the share of Born Globals of the total amount of New Ventures in a country, expressed in percentage. As mentioned before, the most widely accepted definition of BGs is “business organizations that from inception seek to derive significant competitive advantages from the sale of output in multiple countries” (e.g. Knight and Cavusgil, 2004; Oviatt and McDougall, 1994; Rennie, 1993). We will adopt this definition in our research as well.
(GEM), who has labelled it as “International Orientation early-stage Entrepreneurial Activity”.
Therefore we will adopt their measurement, being “Percentage of TEA who indicate that at least 25% of the customers come from other countries”. TEA stands for “total early-stage entrepreneurial activity”, and represents the percentage of 18-64 year old population who are either a nascent entrepreneur or owner-manager of a new business, hence in the stage before the start of a new firm or the stage directly after the start of the new firm (GEM, 2016).
As mentioned before, we use data from up to three years, resulting in three values per country. In order to make sure that we have enough observations for a robust regression analysis, we will treat each combination of country and year as an individual observation, rather than calculating and using the average of the three values. This means we will have 132 observations.
3.3.2 Independent variables
For our model we will use 11 IVs, subdivided into two categories, being domestic market deficiencies and domestic market enablers. The first resembles factors that might push firms towards early internationalisation, the latter factors that enables early internationalisation.
3.3.3 Domestic market deficiencies
3.3.3.1 Domestic competition
For measuring the intensity of the domestic competition we use the database of the World Economic Forum. They included the following question into their annual Executive Opinion Survey: “In your country, how intense is competition inn the local markets (1 = not intense at all; 7 = extremely intense)?” (Schwab & Sala-i-Martín, 2016). We use the country-level score for this question, and adopt the coherent score on a scale of 1 to 7.
3.3.3.2 Domestic market size
3.3.3.3 Market growth
To the best of our knowledge, the relation between country-level market growth and the propensity of early internationalisation has not been tested in a quantitative way as of yet. Evans et al (2008) looked at the retail industry in the UK and USA and adopted a qualitative approach to estimate the (perceived) market growth and its impact on early internationalisation. Since it is virtually impossible to collect such data on every industry for all the countries in our sample, we will adopt the annual increase in GDP as a proxy. Although this metric might neglect industry-level differences, we consider this an appropriate metric since it shows us whether or not an economy is growing. Moreover, its calculation is uniform all over the globe, providing us with solid, comparable data (Dobbs et al, 2015; OECD, 2005). For this variable we have used the data provided by the World Bank (2016). The data is constructed as percentage wise GDP growth in the respective year and country.
3.3.3.4 Average Income
Following general practice (e.g. Babones, 2005; Caseli, 2005) for our calculation of the income level of a country we have opted for the gross domestic product (GDP) per capita based on purchasing power parity (PPP). The GDP is the sum of gross value added by all producers in an economy plus product taxed and minus subsidies not included in the value of products. This value is then converted to international dollars using purchasing power rates, which allows for cross-country comparisons. An international dollar has the same purchasing power as one U.S. dollar has in the United States. The data we use are in current international dollars (World Bank, 2016).
Next, we have standardised the obtained values in order to make sure that the regression output can be interpreted properly, and to ensure that all variables contribute evenly when added together (UCLA, 2016).
3.3.3.5 Corporate Tax Rate
3.3.4 Domestic market enablers
3.3.4.1 Level of infrastructure
Several studies have linked the level of infrastructure in the domestic market to the propensity of firms to start internationalising using a qualitative approach. North and Smallbone (2000) used a survey to compare different areas within England. Freeman et al (2012) interviewed managers and trade advisors to investigate differences between different Australian territories. In both cases, the perceived level of infrastructure was used as a measure for the level of infrastructural development.
Following their approach, we will rely on qualitative data presented in the Global Competitiveness Report of the World Economic Forum (Schwab and Sala-i-Martín, 2016). They included the following question in their Executive Opinion Survey: “How do you assess the general state of infrastructure (e.g. transport, communications, and energy) in your country? (1 = extremely underdeveloped – among the worst in the world; 7 = extensive and efficient – among the best in the world)”. We have adopted the scores given in the report, on a continuous scale from 1 to 7.
3.3.4.2 Knowledge intensity
To the best of our knowledge, the knowledge intensity level in the domestic market has not been linked to internationalisation as of yet, whereas many scholars have linked knowledge intensity on the firm level to its propensity to internationalise (e.g. Autio et al, 2000; Erkko et al, 2000).
For the construction of this variable, we have adopted the measurement of the Global Competitiveness Report of the World Economic Forum (Schwab & Sala-i-Martín, 2016). They have constructed the level of higher education and training by calculating the aggregate score of eight indicators, being: secondary education enrolment rate; Tertiary education enrolment rate; Quality of the education system; Quality of math and science education; Quality of management schools; Local availability of specialised training services; Extent of staff training.
The aggregated score is given on a scale from 1 to 7, whereas 7 is the highest possible score. This measurement has been adopted as such.
3.3.4.3 Prevalence of foreign ownership
To the best of our knowledge, prevalence of foreign ownership in the domestic market has not been linked to the appearance of BGs in a quantitative manner as of yet.
For our analysis we rely on the data provided by the World Economic Forum, which included the following question into their Executive Opinion Survey: “In your country, how prevalent is foreign ownership of companies? (1 = extremely rare; 7 = extremely prevalent)” (Schwab and Sala-i-Martín, 2016). We have adopted the variable in the same fashion as the World Economic Forum constructed it, on a scale ranging from one to seven.
Although this measurement of foreign ownership is based on the perceptions of the respondents and therefore not very precise, it is the only indicator available for the prevalence of foreign ownership in all the countries in our sample.
3.3.4.4 Membership of the Eurozone
This variable will be constructed as a dummy variable. For every country and corresponding year membership of the European Monetary Union will be labelled “1”, whereas not being a member has been labelled “0”.
3.3.4.5 Regulatory environment
To the best of our knowledge, the regulatory environment in the domestic market has not been linked to the prevalence of BGs in the existing literature. For our measure of the regulatory environment, we have adopted the score that the World Economic Forum has allocated to property rights in the Global Competitiveness Index (Schwab and Sala-i-Martín, 2016). For assessing a country’s protection of property rights, respondents had to answer the following question: “In your country, to what extent are property rights, including financial assets, protected?”. This question is answered on a scale from one to seven, whereas seven is the highest score. The final score is the average of all the given answers per country. We have adopted the score as such, hence on a continuous scale from one to seven.
3.3.4.6 Innovative capacity
In the existing literature innovative capacity has only been linked to internationalisation on the firm level, not on a country level. However, several scholars do expect that the level of innovative capacity in the domestic market could be an antecedent for internationalisation (Love and Roper, 2015; Oesterle, 1997; Wright et al, 2015).
Index, 2016). They create an annual ranking based on an aggregated score composed of a number of subcategories, an overview of which can be found in appendix B. This aggregated score is given on a scale of zero to 100, where 100 is the highest possible score. We will adopt these scores for our model.
3.3.5 Control variables
In our model we will control for global events influencing the share of BGs by adding dummy variables for the three years. If this variable has a strong positive or negative impact on the share of BGs, this could indicate a global event that goes beyond the domestic market factors, such as a global crisis, or a negative spiral due to e.g. war. We have considered, but opted not to control for geographic location (e.g. continent). We opted to do so because this location is likely to be an antecedent for our IVs. E.g. a country in Europe is
Label Variable Definition Expected impact
Y Share of BGs Share of BGs as a percentage of
the total number of NVs
i Country Country of observation
j Year Year of observation
Domestic market deficiencies
Comp Domestic
competition
The intensity of competition in the domestic market
+
Size Market size GDP plus value of imports
minus value of exports
-
Growth Market growth Percentage growth of GDP -
Income Income rate GDP per capita (PPP) –
standardised
- Taxrate Corporate tax rates Standard corporate tax rate
per country.
+ Domestic market enablers
Infra Infrastructure Perceived level of
infrastructural development + Knowledge Knowledge intensity Level of higher education and
training
+ Forown Foreign ownership Perceived prevalence of
foreign ownership – foreign companies (partially) owning companies in the domestic market
+
EMU Membership EMU Having the Euro as a currency + Regulatory Regulatory
environment
Protection of property rights + Innovative Innovative capacity Score on the Global Innovation
Index
3.4 Statistical Model
Using the variables mentioned before, we will now construct the statistical models for our multiple regression analyses. We will run four tests, using four different models. Please note, that in the equations 𝛼 represents the intersect, 𝛽 stands for the estimated coefficient of the related variable, and 𝜀 is the error term. 𝑌 is the dependent variable, being the share of BGs. Finally, 𝑖𝑗 indicates the country and year of the observation.
1) First, we will run a test containing only the dummy variables: 𝑌𝑖𝑗= 𝛼 + 𝛽1𝑌2012𝑖𝑗+ 𝛽2𝑌2013𝑖𝑗+ 𝛽3𝑌2014𝑖𝑗+ 𝜀
2) Next, we will test for the influence of domestic market deficiencies on the share of BGs.
𝑌𝑖𝑗 = 𝛼 + 𝛽1𝑌2012𝑖𝑗+ 𝛽2𝑌2013𝑖𝑗+ 𝛽3𝑌2014𝑖𝑗+ 𝛽4𝐶𝑜𝑚𝑝𝑖𝑗+ 𝛽5𝑆𝑖𝑧𝑒𝑖𝑗 + 𝛽6𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝑗+ 𝛽7𝐼𝑛𝑐𝑜𝑚𝑒𝑖𝑗+ 𝛽8𝑇𝑎𝑥𝑟𝑎𝑡𝑒𝑖𝑗+ 𝜀
3) In the third equation we will test the relation between domestic market enablers and the share of BGs.
𝑌𝑖𝑗= 𝛼 + 𝛽1𝑌2012𝑖𝑗+ 𝛽2𝑌2013𝑖𝑗+ 𝛽3𝑌2014𝑖𝑗+ 𝛽4𝐼𝑛𝑓𝑟𝑎𝑠𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒𝑖𝑗 + 𝛽5𝐾𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒𝑖𝑗+ 𝛽6𝐹𝑜𝑟𝑜𝑤𝑛𝑖𝑗+ 𝛽7𝐸𝑀𝑈𝑖𝑗+ 𝛽8𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑜𝑟𝑦𝑖𝑗 + 𝛽9𝐼𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑣𝑒𝑖𝑗+ 𝜀
4) Finally, we will construct a model including both domestic market deficiencies and enablers in order to assess the overall explanatory power of our variables:
𝑌𝑖𝑗 = 𝛼 + 𝛽1𝑌2012𝑖𝑗+ 𝛽2𝑌2013𝑖𝑗+ 𝛽3𝑌2014𝑖𝑗+ 𝛽4𝐶𝑜𝑚𝑝𝑖𝑗+ 𝛽5𝑆𝑖𝑧𝑒𝑖𝑗 + 𝛽6𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝑗+ 𝛽7𝐼𝑛𝑐𝑜𝑚𝑒𝑖𝑗+ 𝛽8𝑇𝑎𝑥𝑟𝑎𝑡𝑒𝑖𝑗
3.5 Modelling Procedure
After collecting the data from various sources, we further processed the data using Microsoft Excel. Here we composed our final sample, which we will use in all our analyses. This final sample consists of 70 countries, leading to data limitations since a number of countries has been omitted, which might lead to deviating results.
3.6 Estimation Method
For testing our model we will use interval and ratio data, as well as nominal data, which we will construct as dummy variables, thereby allowing us to use a linear regression. However, since in some cases we use observations from different years from one country, it is likely that our values are not fully independent, since there might be within- country influences that are not included in our model. We will conduct several tests for this to assess whether our model is robust, hence whether our approach is viable. For our research we will use Stata 14.
3.7 Evaluation of Method Assumptions
calculation we found 16 outliers, meaning that 92,2% of the values are within the range. Finally, we tested whether we should exclude the outliers from our model using the DFBETA value, representing the change in the coefficient when the observation is being deleted, hence the relative effect of a single observation on each coefficient. The critical value of DFBETA is 2/N2, which in our case is 0.14 (Hair et al, 2010). When looking at
the DFBETA values of observations with critical values for the studentised residuals, we found that all of them had critical DFBETA values for at least one variable. Therefore, we opted to exclude all the outliers from our model. By doing so, we managed to increase the R2 of our model from .3940 to .5318, hence we improved the explanatory power of
the model. After excluding these 16 observations, 187 observations remain. The outliers that we have excluded can be found in appendix C.1.
4. EMPIRICAL RESULTS
We assume the conditions for our analysis as described in the previous section. In this chapter we will discuss the importance of the four assumptions and the outcomes of the corresponding tests. Next, we will test the previously formulated hypotheses and explain their results in detail. Finally, we will carry out several robustness checks. 4.1 Homoskedasticity
OLS models are based on the assumption that the level of variance in regression errors for all observations is equal, a situation we call homoskedacity (Field, 2009). If this assumption is violated, and the error variance is unequal throughout the model, we speak of heteroskedacity. If heteroskedacity is present, OLS models attaches more weight to observations with large error variances than on those that have small error variances. As a result, we can obtain biased estimates of the variance of each of the estimated parameters (Pindyck & Rubinfeld, 1991). However, heteroskedacity would not automatically imply that coefficients in our model are biased, for the OLS is still linear, and an unbiased estimator. Instead, it means that OLS is not the best approach with the smallest variance. The logic behind OLS is that it minimises squared errors, therefore the main problem with heteroskedacity is that values attached to potential errors have more weight than others, which affects the regression line (Hill, Griffiths and Lim, 2012).
In order to test for heteroskedacity we have conducted a Breusch-Pagan test. Since we obtained a p-value of 0,0076, we rejected H0 hence we can conclude that there is
heteroskedacity in our sample. In order to deal with this, we will adjust our model to account for heteroskedacity in Stata using the “robust” command as suggested by Stock and Watson (2003). By doing so, we slightly increased our R2 from 0.5318 to 0.564.
4.2 Endogeneity
indicating negative first-order autocorrelation. A value of 2 means that there is no autocorrelation. Since our value is below two, we will conduct a test for positive autocorrelation, which means that the error of a certain sign is followed by an error of the same sign (e.g. a positive error is followed by a positive error). We find that our test statistic (0.0037146) is smaller than dL(14, 187): 1.61448, which means we reject Ho,
hence we have autocorrelation in our model.
Since we use panel data, this is to be expected. Following Stock and Watson (2003), we used the Hausman test to examine whether we should use fixed or random effects. The resulting p-value is 0.0318, hence we reject H0 and should opt for fixed effects. Fixed
effects regression controls for omitted variables that differ between cases but are constant over time.
However, as Williams (2015) noted, fixed effects can only be applied if the unobserved variables are time-invariant (e.g. gender, location). In our case it is very likely that there are unobservable variables that are not time-invariant, e.g. sanctions in the case of Russia, which were introduced in 2014 but do not affect 2012 and 2013. Other examples are natural disasters, political turmoil, national crises or conflicts. If unobserved variables are likely to play a role, it is impossible to out rule omitted variable bias (Blumenstock, 2016). Moreover, fixed effects would exclude countries with only one variable, which in our case would exclude only developing countries from the sample, thereby biasing the outcome towards developed countries.
Therefore we have opted to neglect the outcome of the Hausman test and to use OLS instead of fixed effects. As a result, our estimates may contain some bias, which we should take into account when interpreting them (Blumenstock, 2016; Williams, 2015). 4.3 Multicollinearity
Multicollinearity, also called collinearity or intercorrelation, occurs when independent variables are correlated with one another. If this occurs, the estimated regression coefficients of the correlating variables tend to have large sampling errors. Multicollinearity has two consequences. First, because the coefficients have large variability, the sample coefficient could be very different from the actual population parameter, and could even have opposite signs. Second, the coefficients will have small t-statistics, leading to the assumption that there is no linear relation between the independent and dependent variable while there actually is one (Keller, 2012).
be considered critical do not exist. However, a generally accepted level at which one should start worrying is 10 (Field, 2009; Williams, 2015). Another indicator, which provides more evidence for multicollinearity, is the tolerance statistic. This indicator is calculated by 1/VIF, and is considered critical when below 0,1.
As can be seen in appendix E, none of the VIF values exceeds the threshold of 10, nor does one of the tolerance statistics have a value below 0.10. Therefore we can conclude that we have no significant multicollinearity in our model.
4.4 Normality
The fourth and final assumption we will test for is normality. This assumption is important in OLS, and means that the error terms are normally distributed along the corresponding means. Violating the normality assumption could lead to biased p-values, thereby affecting the significance resulting in erroneous interpretations of the model (Hill, Griffiths and Lim, 2012).
In order to test for normality, we used the Skewness and Kurtosis test in Stata, which is a Jarque-Bera test adjusted for small samples. The resulting p-value for skewness is 0,1173 and for kurtosis we obtained 0,9411. The overall value for our model, adjusted for the sample size, is 0,2882. Neither value allows us to reject the H0 of normality hence
4.5 Results of Hypotheses
In this section we will present the summary statistics, and the output of the regression analysis. Using these, the previously defined hypotheses will be answered. The table below presents the number of observations per variable, the means, standard deviations, and minimum- and maximum value.
(1)
(2)
(3)
(4)
(5)
VARIABLES
N
mean
sd
min
max
Y2012
187
0.332
0.472
0
1
Y2013
187
0.337
0.474
0
1
Y2014
187
0.332
0.472
0
1
ShareofBGs
187
0.143
0.0969
0
0.430
Comp
187
5.212
0.535
3.424
6.371
Size
187
4.265
1.142
1.738
7
Growth
187
0.0277
0.0276 -0.0730 0.103
Income
Taxrate
187
187
-1.12e-09
0.256
0.0632
1.000
-1.387
0.100
3.159
0.400
Infra
187
4.552
1.089
2.040
6.540
Knowledge
187
4.670
0.892
2.533
6.265
Forown
187
4.782
0.790
2.138
6.476
EMU
187
0.241
0.429
0
1
Regulatory
187
4.238
0.858
2.789
6.098
Innovative
187
4.192
0.822
2.306
6.011
Table 4.1.: Descriptive statistics Source: own results
On the next page table 4.2has been displayed, presenting the outcome of our regression models. We will refer to this output table when answering our hypotheses. For all of our models the share of BGs of total NVs is the dependent variable. The first value given for each variable is the estimated coefficient, indicating the variable’s relation with the DV. The value below, in parentheses, is the robust standard error. The stars next to the coefficient indicate whether the variable is significant, and at which level. Since our goal is to identify which variables have a significant relation with the DV, and whether this relation is positive or negative, we are less interested in the actual strength of the relations. Moreover, as mentioned earlier on, our coefficients may be biased due to an omitted-variable bias and endogeneity (Blumenstock, 2016).
With the second model we try to answer hypotheses 1 till 5, thereby analysing the influence of domestic market deficiencies on the share of BGs. As the regression output shows, all 5 variables have a significant impact on the DV. As expected, the level of competition is positively related to the share of BGs, albeit with a low coefficient (0.0246). Also following our expectations, both size and growth are negatively related to the DV. But whereas size is weakly related to the DV, with a coefficient of -0.0254, the relation between growth and the DV is significantly stronger (-0.430). Contrary to our expectations, the corporate tax rate seems to be negatively related to the share of BGs (-0.403). Finally, the relation between the income level and the DV is positive (0.0467), whereas we expected it to be negative. This model has an R2 value of 0.527, hence it
explains 52,7% of the variance in the DV. However, since we only look at one aspect that might explain the share of BGs, being the domestic market factors, we do not expect nor aspire to achieve a very high R2 value since there are numerous other factors that are
not included in our models. The outcomes of model 2 confirm hypotheses 1, 2 and 3, and show that hypotheses 4 and 5 were not confirmed.
In our third model we test the relation between domestic market enablers, potentially stimulating companies to become BGs, and the DV. The level of infrastructure turns out not to be significant in explaining the DV. The level of knowledge on the other hand is positively related to the DV on the 99% significance level, albeit with a very small coefficient (0.0438). Just like the level of infrastructure, the prevalence of foreign ownership in the domestic market is not significantly related to the share of BGs. As we expected, both membership of the EMU and the strength of the regulatory environment are significantly positively related to the DV (0.0643 and 0.0454) at the 99% significance level. Contrary to our expectations, the innovative capacity proves to have a significant negative influence (-0.0421) at the 99% significance level. This model’s R2 is far lower
than that of model 2, with a value of 0.381. This could indicate that domestic market deficiencies are a better predictor of the share of BGs than the enablers. Model 3 confirms hypotheses 7,9 and 10.
