Born local or born global : assessing the geographical boundaries of e-commerce companies

(1)

1

Born Local or Born Global

Assessing the geographical boundaries of e-commerce companies

Lennard Kooy

MSc. in Business Studies – International Management track Student number: 10342222

1st Supervisor: Erik Dirksen MSC 2nd Supervisor: Dr. Arno Kourula

(2)

2 Contents 1.0 Abstract ... 3 2.0 Introduction ... 4 2.1 Research Questions ... 5 2.2 Hypotheses ... 6 3.0 Literature review ... 7

3.1 Overview of the existing literature ... 7

3.2 The playing ground of E-commerce... 13

3.3 Reasons Why Companies are Adopting E-commerce ... 16

3.4 Barriers to E-Commerce... 18

3.5 Theory and practise ... 19

4.0 Methodology and structure ... 20

4.1 Research Approach ... 20

4.2 Data Collection Method ... 20

4.3 Data Preparation ... 27

5.0 Analysis / results ... 29

5.1 Foreign market entry and speed of entry ... 29

5.2 Amount of sales from foreign markets ... 35

5.3 The profitability of international markets ... 39

6.0 Discussion ... 43

6.1 Going abroad ... 43

6.2 Foreign presence ... 45

6.3 Foreign profit... 47

6.4 Limitations and future research ... 49

7.0 Conclusion ... 51

8.0 Reference list ... 52

9.0 Appendices ... 57

Appendix 1: List of internet retail companies ... 57

(3)

3

1.0 Abstract

The internet has inarguably changed the landscape of (international) retail. Theoretical constructs such as the ‘Born Global Firm’ and ‘The International New Venture’ are created to explain the influence of forces like the internet on company’s internationalization strategies and successes. Those constructs are often used by researchers and authors to argue that the internet has made it possible for companies to operate international from their founding, or significantly sooner in their lifecycle, as well as being more profitable than companies who don’t leverage the internet.

In this research, two groups, of American internet retailers and non-internet retailers, are compared to assess whether internet retailers indeed go abroad more often, significantly sooner, whether their international presence is more significant and whether they are indeed more profitable than non-internet retailers. The findings suggest that internet retailers actually don’t go abroad more often than their ‘traditional’ counterparts, however when they do go abroad, they do that sooner in their lifecycle. Furthermore, internet retailers don’t have a significantly larger presence in foreign countries than non-internet retailers and are actually less profitable in their international endeavours.

These findings implicate that the theory suggesting that internet on its own serves as an advantage for retailers is at least arguable and that the best model for a company is still very much dependent on other factors as well. Future research is needed to assess whether these findings are also applicable to other geographical areas and whether these findings will hold in the future, seeing that the internet is becoming a more significant part of people’s lives every day.

(4)

4

2.0 Introduction

Selling things or offering services ‘online’ is nothing new. The first online shopping system was introduced in 1979 by Michael Aldrich and the first working applications which actually serviced people through the internet go back as far as the early 80’s, when devices as the Minitel were introduced (Markoff 2005). Using the world wide web with browsers to sell things as we know it today was made available when the world wide web was opened up for commercial use in 1991. Internet ‘giants’ of today, such as Amazon and eBay, first opened up their shop in the mid 90’s and the e-commerce business grew explosively from that point on. Online Business to Consumer sales passed the 1.000 billion dollar barrier globally the first time in 2012 and is reported to hit 1.500 billion dollars worldwide this year (2014). In North America alone the markets stands at 431 billion dollar in 2013 and is expected to grow with about 11% a year for the coming years (eMarketer 2014).

Breaking down the physical barrier for selling things, whether they are goods or services, has opened up a new way of reaching customers, resulting in new, different forms of market entry, at other stages in a company’s lifecycle than what used to be custom in a world that was limited by physical interaction. These changes have been quite extensively described by several authors, resulting in new theoretical constructs such as the ‘International New Venture’ and the ‘Born Global Firm’(Oviatt and McDougall 1992; Knight and Cavusgil 2004).

These theoretical constructs describe how an increasing globalizing world makes it possible for companies to conduct business internationally or globally from the very beginning of a company’s lifecycle. Especially the ‘Born Global’ concept emphasizes the role of technology in this change, and the concept of ‘International New Ventures’ has been adjusted to incorporate this construct’s as well (Zahra, Matherne and Carleton 2003).

(5)

5 While both theoretical constructs are not exactly the same, which will be discussed in the literature review, according to both the Born Global and International New Venture concepts, especially the internet has made it possible for companies to transcend physical barriers to sell products (services and goods) to customers and are therefore not subject to geographical boundaries any longer, or at least are a lot less restricted by them.

While the theory behind this the reasoning of transcending physical barriers for sales through the internet seems logical, not much research has been done to empirically validate the theory. In recent history, some companies that heavily rely on the internet (such as internet retailers as eBay and Groupon), have had a hard time successfully expanding their business outside their home market, North America, and have only succeeded in recent years or didn’t became successful at all (Wang 2010; Robles 2012). In Europe some similar situations are to be seen at for example Dutch e-commerce companies as Wehkamp.nl and Bol.com (Weblog 2013). Transcending geographical boundaries still seems to be more difficult for some of these companies relying on the internet than theory would suggest.

This thesis will research whether the observations about geographical boundaries for e-commerce companies are deceptions, or that those geographical boundaries still seem apparent, even for technological intensive firms that in theory should be able transcend these boundaries. This research will do this by comparing a set of publicly traded internet retail companies on the NASDAQ and set of publicly traded traditional (non-internet) retail companies situated on the NASDAQ.

2.1 Research Questions

More specifically, this research will try to answer the questions whether internet retail companies in fact go abroad more often and significantly sooner (i.e. sooner in their company lifecycle) than the ‘traditional’ retailers and whether their proportion of sales outside their

(6)

6 domestic market is larger than that of traditional retail companies. The latter is an interesting case on itself since it is always hard to assess when a sale should be qualified as ‘outside the domestic market’. E.g. a foreign customer can purchase something in the US and take it back to their domestic country, is that a foreign sale or not? In this thesis, the qualification that is used by the NASDAQ and COMPUSTAT database will be used, namely that a sale is foreign when it is either exported to a foreign country or sold by a foreign vendor (without them having to export it to the company’s domestic country) (COMPUSTAT 2002).

Furthermore, this research will try to answer whether the decision to go abroad has actually benefited the companies who had done so in terms of increased profit (assessed by operating income).

While the hypotheses in this research seek to validate the theory that is briefly discussed in this introduction and will be discussed more in depth in the literature review, namely that internet retailers operate beyond boundaries with more ease than non-internet retailers, this research argues that it could well be that this is not the case and that internet retailers are still subject to similar geographical limitations as non-internet retailers.

2.2 Hypotheses

To answer the research questions as outlined in paragraph 2.1, the following hypotheses will be answered by conducting statistical analyses on the two research groups, the internet retailers and the non-internet retailers:

1.a. The % of internet retail companies going abroad is significantly higher than that of non-internet retail companies going abroad.

1.b. When retailers go abroad, internet retail companies conduct sales outside of their foreign markets significantly sooner in their lifecycle than non-internet retail companies.

(7)

7 2. Overall, the % of net sales coming from foreign markets for internet retail companies

is significantly higher than that of non-internet retail companies.

3. The % of net profit for internet retail companies, gained from international markets, is significantly higher than for non-internet retail companies.

3.0 Literature review

3.1 Overview of the existing literature

As said in the introduction the emerging phenomenon of the internet as a sales platform has not gone unnoticed by researchers. Especially quite a lot of attention has recently been paid to the benefits of the internet to Small Medium Enterprises (SME’s); how SME’s can leverage the internet, how it could accelerate their growth and how it affects their strategies (see for example the articles of Moen and Servais 2002; Bell and Loane 2010; Onetti, Zucchella, Jones and McDougall-Covin 2012). Not very surprisingly, these articles come to the conclusion that the internet could aid SME’s in accelerating growth when the right strategic choices are made and that technologies, like the internet, provide companies with new possibilities, such as extra sales channels, advertising, communication etc. However, these articles focus especially on SME’s and their strategic choices and possibilities, rather than looking at their situation during their founding or the situations for larger companies. More importantly these articles use cases and theory to illustrate their findings, rather than making actual quantitative empirical observations.

Already emphasized in the introduction is another stream of literature that focuses more on the ‘founding’ phase of a company. This stream of literature has come up with the concepts of ‘Born Globals’ and ‘International New Ventures’. The concept of ‘International New Ventures’ was introduced by Benjamin M. Oviatt and Patricia Phillips McDougall in

(8)

8 their famous essay ‘Toward a Theory of International New Ventures’. In that particular article Oviatt and McDougall describe what sets these new types of firms apart from others and describe elements that these firms should possess in order to become a successful ‘International New Venture’ (Oviatt and McDougall 1992). Since the publication of the Oviatt and McDougall article, the interest in the concept of ‘International New Ventures’ has grown, and the impact of the article of Oviatt and McDougall is widely recognized (Autio 2004).

While an important fundament for the theory that is describing companies that are international orientated from birth, the elements that Oviatt and McDougall describe are somewhat dated and don’t really capture the rise of technological intensity that the world have witnessed over the past two decades. Authors as Knight and Cavusgil build on the same construct but put more emphasis on the capabilities required for firms to achieve this internationalization from birth (Knight and Cavusgil 2004). In their concept of ‘Born Globals’, Knight and Cavusgil describe how this theory is predominantly based on two sources, namely the effects of the increasing globalization and the rise of certain technologies, such as, and especially, the internet. As Knight and Cavusgil put it:

“Electronic interconnectedness in particular is driving the emergence of a borderless global economy. Information technology and the Internet are liberating forces, permanently altering the landscape of international trade” (Knight and Cavusgil 2004, p. 137).

Other authors, such as Zahra, Matherne and Carleton have extended the concept of Oviatt and McDougall to also integrate this relatively new concept of technological resources. They find that technological networks and technological reputations are predictors of the speed and degree of internationalization, partly thus recognizing the value of for example the internet (a technological network) in the concept of ‘International New Ventures’ (Zahra, Matherne and Carleton 2003).

(9)

9 The authors of ‘International new ventures: revisiting the influences behind the 'born-global' firm’ are only partly in agreement with the theory of the ‘Born Global firm’. Fan and Phan argue that those ‘Born Global’ firms in fact are very similar, if not totally similar to ‘traditional’ firms in terms of internationalization and argue that the decision to go abroad relatively early is influenced by the size of a company’s home market, its production capacity and cultural and economic factors. Fan and Phan however do agree with authors as Knight and Cavusgil, Zahra, Matherne and Carleton that a technological advantage or intensively could lead to a different, faster internationalization pattern (Fan and Phan 2007). Again however, Fan and Phan build on theory, rather than on empirical findings.

While often used as interchangeable, for example by Fan and Phan, Dave Crick points out that there are some differences between the two theoretical constructs of ‘Born Global’ and ‘The International New Venture’. According to Crick, the word “global” in ‘Born Global’ suggests a company has a presence in at least the world's triad regions. ‘International New Ventures’, however, may just have and international presence (thus in foreign countries), but nothing in the construct advocates where that should be. Crick argues that ‘International New Ventures’ tend to have a more regional focus than companies that are classified as ‘Born Global’ (Crick 2009).

This research however, looks at internationalization, whether that is classified as global or more regional is out of the scope of this research. Thus while the distinction between the two constructs is good to note, for the sole purpose of this research, they will be seen as similar. In that light, both these streams of literature, ‘The International New Venture’ and the ‘Born Global firm’, have predominantly focused on the effects, such as globalization, technology rise and among those the internet, on firms in general. Following the findings of those frameworks, one might presume that companies whose business model is explicitly build around the internet and use the internet as their main sales channel (in this research

(10)

10 called e-commerce companies or internet retailers) have most to gain from these recent developments.

However, this is not specifically addressed in those articles. In an article of Singh and Kundu, the eclectic paradigm or Ownership, Location and Internationalization (OLI) framework of Dunning is extended to try to come up with a framework, incorporating the theories as the Network Theory Approach, the Resource Based View (RBV) and the Transaction Cost Perspective, in which they do try to explain the growth of e-commerce companies specifically, setting them apart from ‘traditional’ companies. The new theoretical construct that Singh and Kundu create is called N-OLI, where the N stands for Network since they find that cyber presence is best explained by the addition of network structures, alliances and relationships in addition to extending the original three areas (Singh and Kundu 2002).

Building on their N-OLI construct, Singh and Kundu come to the conclusion that technological complexity provides cost effective and efficient access to resources, saying that the more companies rely on technologies such as the internet, the more advantage they will have in accessing certain resources (like markets). However, to take advantage of these resources, e-commerce companies need dynamic capabilities that can help them leverage their resources and generate value-creating strategies (Singh and Kundu 2002).

The research of Singh and Kundu provides some answers to the question on how e-commerce companies internationalize, the research however is already 12 years old, and as they imply in their ‘future research’ sector, the list of N-OLI advantages that e-commerce companies have in comparison to ‘traditional’ companies, which they propose in their article, is not exhaustive. In addition, Singh and Kundu don’t emphasize in their article how the internationalization strategies and speed of e-commerce companies differentiate specifically from ‘traditional’ companies in terms of when they go abroad and whether e-commerce companies have more success in going abroad, they just argue that they differ.

(11)

11 In line with the framework of Singh and Kundu, Luo, Hongxin Zhao and Du investigated what explains the internationalization speed of e-commerce companies and whether this differentiates with non-e-commerce companies. Luo, Zhao and Du argue that there is a difference in market entry strategies and that speedy foreign market entry by e-commerce companies is positively influenced by the top management team's international experience, and innovative and marketing capabilities. However Luo, Zhao and Du just accept the fact that there is a difference in market entry moment between e-commerce companies and non-e-commerce companies, referring to sources that build on a theoretical framework rather than on empirical research (Luo, Hongxin Zhao and Du 2005).

Two other researchers who come to a similar conclusion are Mats Forsgren and Peter Hagstrom in their article ‘Ignorant and impatient internationalization? The Uppsala model and internationalization patters for Internet-related firms’. Forsgren and Hagstrom compare the internationalization of internet-related firms with that of a traditional model called the Uppsala model. The Uppsala model argues that companies gradually gain experience in their domestic market before entering a foreign market. Forsgren and Hagstrom find that a similar incremental behaviour is not found with internet-related firms and that those firms tend to internationalize in a much faster pace than companies that do follow the Upssala model (Forsgren and Hagstrom 2007). Forsgren and Hagstrom conducted their research on a set of case studies that they compare to a theoretical framework and present their findings as a conceptual model rather than as empirical findings. Interesting thus would be to see whether their findings and model also hold up in comparison to actual findings of non-internet related firms.

Closely related to the subject of speed and performance of firms, is the literature about entry and establishment mode. Because e-commerce companies are argued to be able to make use of a more exporting nature, from the founding of their company, and thus conduct

(12)

12 business abroad without having an actual equity presence, entry and establishment mode literature is of concern for this research as well. Literature about entry and establishment mode nowadays is extensive. Authors as Benite and Gripsrud, Chang and Rhee and Dikova and van Witteloostuijn have extensively described the predictors and effects of foreign direct investments (FDI) (Benite and Gripsrud 1992; Dikova and van Witteloostuijn 2009; Chang and Rhee 2011).

The impact of e-commerce on entry mode however is still underexposed. An attempt is made in the 2004 article of Ekeldo and Sivakumar. Like Singh and Kundu, their article is rooted on the Resource Based View model. They find that some concepts in marketing must be modified to comprehend the effects of digitalization in order to be able to accurately explain the impact of e-commerce on entry modes. More specifically, they find that the internet has made it possible to increasingly centralize certain activities, such as marketing, order taking and billing. However, Ekeldo and Sivakumar fail to come up with direct effects of commerce on specific entry mode choices and entry mode moments, such as whether e-commerce companies actually do use an exporting entry mode more often than traditional companies or go abroad relatively sooner. They do argue that the rise of e-commerce will lead to the blurring of country barriers and argue that a global marketing strategy will take the place of entry choices. Referring to the ‘Born Global’ concept of Knight, Ekeldo and Sivakumar also stress the fact that companies can become global from the start of their venture, because online transactions can traverse national borders and thus indicate towards the fact that those types of businesses should theoretically be able to conduct sales internationally sooner in their lifecylce (Ekeldo and Sivakumar 2004).

Fact is however that even if the blurring of country barriers will become more present in the future, a point that is certainly not without debate, e-commerce companies currently still do have to make choices in when and how to enter, just as ‘traditional’ firms. The

(13)

13 argument can be made that they come across these choices relatively less frequently than ‘traditional’ firms because often they have a more exporting nature due to their electronic distribution channel, i.e. the world wide web (Javalgi and Ramsey 2001).

However, as described in the introduction, the decision to conduct business abroad seems, also for e-commerce companies, not a given, but a distinctive choice. In another way, this is also exemplified by the streak of acquisitions companies as Google, Amazon and eBay have been conducting the last couple of years (Costine 2014). While technologies such as the internet, according to for example Knight and Cavusgil, Singh and Kundu and Keldo and Sivakumar would enable companies to have a global presence a lot easier than traditional firms, e-commerce companies still make use of mergers and acquisitions (M&A), even within their own discipline (buying companies that have similar activities in different geographical areas). This is also the finding of Lilach Nachum and Srilata Zaheer in their article ‘The persistence of distance? The impact of technology on MNE motivations for foreign investment’ (2005). Nachum and Zaheer describe why MNE’s still conduct FDI while technologies, such as the internet, make it possible to de business at a distance. They find that (geographical) knowledge seeking and efficiency seeking, two concepts explained by Dunning in his eclectic paradigm, still apply to technological and information intensive industries (Nachum and Zaheer 2005).

3.2 The playing ground of E-commerce

Although the debate about whether e-commerce companies follow different rules when internationalizing is thus still ongoing and will be researched in this thesis, theoretically, and according to the existing literature, it seems that the existence of e-commerce has led to the expansion or alteration of companies’ boundaries to some extent one way or the other. Especially in the last decade it has seemed to become more and more difficult to, even

(14)

14 theoretically, determine the geographical boundaries of companies, due to the existence of e-commerce. This possible transcending of geographical boundaries has brought a lot of potential opportunities, and challenges, for companies (which will be discussed hereafter), but also for more macro factors playing in the world economy.

One of these challenges of the theoretical boundless nature of e-commerce business is that of legislation. Because various laws are created to regulate a given national geographical region, it is hard to decide which laws apply to which part of the business when applying these laws on e-commerce companies. In theory, e-commerce companies can operate beyond boundaries, and are thus subject to multiple national boundaries where different laws apply. Now, the law serves the purpose of both promoting order and predictability. Predicting the impact that a certain law will have on business operations is part of day-to-day activity in commerce. National boundaries are usually served by courts, which try to ensure that various rules are enhanced on a wider array referred to as jurisdiction, but this is only possible when the transacting parties reside in the same region governed by similar laws. When transacting parties are located in different geographical boundaries, the rules become more complicated (Diana 2008).

By transcending these boundaries, the emergence of e-commerce has threatened the logic of traditional rules of jurisdiction and legislation. For example, a resident of the Netherlands can sell services or goods to a resident of France via the internet, through a site hosted on a server located in Germany and paid from a bank located in the United States, without any physical contact between the parties involved. If a dispute between the parties involved erupted it’s very hard to decide which laws apply and why. These occurrences have forced legislators to come up with various statutory solutions. One example is the federal anti-cybersquatting law that has been enacted whereby legal action can be taken wherever the host

(15)

15 server is located even when the domain owner resides in another geographical region (Longenecker 2012).

While there are thus some challenges that arise with this new form of business, e-commerce also gives economies great opportunities. An important one is the reach of the potentially global market. The world population is increasing as time passes by, however the growth is not proportional across all counties. Some countries’ growth is rapid, others slow and in other countries it is stagnant. For example, according to a report released by International Institute for Applied Systems Analysis, China’s population is expected to grow from 1.35 Billion to around 1.48 billion in the coming 30 years (Schachter 2009). However, in some countries located in Europe, there is the expectation that the population will decline.

Such developments can trigger companies, for example located in Europe, to start looking at other potential growth markets, such as China. Other parts in the world, for example in Africa, are also seeing an increased number of individuals using the internet. Currently, the global e-commerce market stands somewhere between 1.200 and 1.500 billion dollar (eMarketer 2014). Longenecker asserts that there is a prediction of the market growing at a rate of 17% annually and for example China’s burgeoning e-commerce market is estimated to deliver $1 billion in terms of sales on a daily basis in the future (Longenecker 2012).

Economies participating in e-commerce theoretically could thus have the ability of increasing their market share if they venture in these countries, without actually physically having to go there. This could make e-commerce very attractive as a speedy way of expanding a business and the export of a country.

Theoretically, it does not require lengthy procedures of establishing physical businesses in these countries, whereby various procedures must be followed, something which could potentially result to high costs. Global organizations and various governments are also

(16)

16 encouraging e-commerce between less developed nations and wealthier nations, because often, the transactions are beneficial for both, especially when a part of the earnings from the selling company is reinvested in the country where the purchase took place, but more importantly, knowledge is transferred. Therefore, e-commerce has the ability of providing new ways of diversifying national economies, better jobs for the young population and enhances new possibilities of participating in global markets (Schachter 2009).

3.3 Reasons Why Companies are Adopting E-commerce

There are several reasons that prompt organizations to get involved with e-commerce ventures. Among those reasons, are the macro-economic reasons as discussed in the previous section (3.2). More specifically, there are various micro-economic factors that potentially make e-commerce an interesting way of conducting business. One of those is that the market seems to be evolving constantly; for example, mobile commerce was initiated only a few years ago, and has already captured 20% of the total e-commerce sales worldwide and is growing with almost 70% a year. As smart phones and tablets with internet are becoming more and more common, this is likely to greatly benefit the e-commerce market (Brohan 2013).

Another reason for engaging in e-commerce is the ability to conduct business 24/7. Online stores are never closed, and this brings a lot of convenience to shoppers and to the stores. Customers can purchase whatever they want, whenever they want it and from wherever they are. E-commerce also enhances accessibility and convenience on the side of consumers. Consumers can get their services or goods from any vendor irrespective of their geographical location and without actually having to go there physically.

Potentially, companies could also save a lot of costs when they operate online stores as opposed to brick and mortar stores. There is potentially saving on rent, construction and

(17)

17 display units among others. On the other hand there are costs for running an internet shop that don’t exist for brick and mortar stores such as server costs, web development and maintenance and potentially extra storage costs due to larger inventories (e.g. compare Amazon to a traditional book store as Barnes & Noble). However, the costs for internet shops seem to be far lower in comparison to the costs for traditional brick and mortar shops. Companies also could potentially increase their profits when they adopt e-commerce, for example by being able to serve more individuals (because generally there are no waiting times, physical restrictions to registers etc.) (Diana 2008). In addition, the internet offers the possibility to streamline and scale parts of a company’s business, that weren’t possible before, such as billing, supply chain management, project management and logistics (Sharma 2009).

To other companies, e-commerce is used for personalization aspects. An example of a company that has taken advantage of this aspect is Amazon.com. The company personalizes its consumers’ experience so that once an individual becomes a registered user; the company can move closer and have one-to-one conversation, offering the customer personalized deals, relevant products and even engaging the customer whenever he/she seems to have questions. This aspect helps in creating consumer loyalty since the company is giving the customer what he/she really wants, and by using technological constructs only available in online environments, this leads to differentiation to ‘traditional’ commerce that don’t have those tools at their disposal (Sharma 2009).

Apart from all these potential advantages of using the internet, there are indications that companies that engage in exporting entry modes (whether that is using the internet or not) into foreign markets early in their lifecycle have a higher profitability overall than companies who don’t (Moen and Servais 2002). These indications can thus potentially stimulate companies to engage in international ventures earlier, potentially using e-commerce as their sales channel.

(18)

18

3.4 Barriers to E-Commerce

Of course there are also barriers for companies to e-commerce. Since its inception, several bottle-necks to its growth have been experienced. Among the challenges, predominantly faced in the earlier stages, is Personal Computer (PC) penetration in most parts of the world. Not very surprisingly, there seems to be a direct relationship between e-commerce and PC penetration (eMarketer 2013). Income earnings usually determine who has the ability of owning a PC, which is a major consideration in developing countries. For example, in China, nowadays a low-end PC might cost only US$ 450, but this still represents an amount similar to two months wage for an average user. This means that many people in such countries have been locked out in participating in e-commerce for at least the first period of this novel market (Thanasankit 2003).

While the costs of computers keep going down and the average income is steadily going up in these markets, the full potential market share is not yet realized, and this currently still acts as a stumbling block to the growth of e-commerce. Internet infrastructure in developing countries also acts as a barrier to e-commerce. It is evident that PC prices are declining significantly, but with limited internet availability, this will not boost e-commerce growth, due to the obvious fact that internet tis required to actually engage in e-commerce. Inadequate infrastructure inhibits internet connection to prospective e-commerce participants. Countries in Africa, Asia and in a minor degree South America seem to be most affected with regard to this issue (eMarketer 2013).

Pahladsingh asserts that cultural perspective also plays a vital role in spearheading e-commerce. It affects e-commerce businesses as much as it affects traditional multi-national corporations. Some cultures are more used to personal contact during sales, negotiating and building relationships. These kind of things influence the way that business is conducted and thus will influence whether customers are comfortable shopping online or not. In addition,

(19)

19 almost 80% of web content is published in English. Such occurrences prompt business owners to translate their websites into different languages and obviously, there are costs incurred of transferring content to other languages (Pahladsingh 2006). These predicaments are interesting for (internet) retailers and there is still a lot of ground to cover there.

Lack of trust has also emerged as a troubling barrier to e-commerce. It is upon businesses to build trusting relationships with its customers. With traditional retail sales, sales representatives have the opportunity to actually get to know the customer through physical contact. As discussed earlier the internet also provides ways of personalization, but this still works differently than actual physical contact.

In line with this, is the issue of anonymity, of the customer but more importantly, of the supplier. There are numerous examples of fraudulent websites that fail to deliver the products they ‘sold’ to customers. These cases might lead to trust issues for the entire sector (Schachter 2009). With this emerging phenomenon, e-commerce companies are trying to use various mechanisms to regain the customers trust. SSL certificates, which ensure safe connections, secure shopping seals, customer ratings and feedback among others, are used to retain or regain consumers trust (Schachter 2009).

3.5 Theory and practise

All these opportunities and drawbacks have led to companies either using or not using the internet to conduct their sales. This research doesn’t aim to advocate the one over the other. However, interesting dynamics come to play when comparing companies that do and don’t use the internet for their sales.

As discussed, there are a lot of authors, and theoretical constructs made by those authors, that stress the possibilities that internet retail brings to internationalization speed and successfulness of those endeavours. A lot of these constructs rely on theory, that in light of all

(20)

20 the findings and potential possibilities of the internet make a lot of sense. However, to really validate the theory, and test whether the observations that somewhat make those theories questionable are true, this research will seek to empirically test these theoretical constructs by conducting a database study, comparing internet retailers and non-internet retailers. The setup of the research is described in the next chapter.

4.0 Methodology and structure

4.1 Research Approach

To find answers on the research questions and test the hypotheses, this research decided to use a quantitative approach. Many of the theoretical constructs as discussed in the literature review are either build on other theoretical constructs or drawn from implications of case studies. To gain more quantitative empirical insights in the research questions, two datasets (later combined in to one) were used: one data set with internet retailers (e-commerce companies) and one data set with traditional, non-internet, retailers. Statistical analyses were later done to compare the results between the two datasets.

4.2 Data Collection Method

Out of the two datasets, the first dataset contains 65 internet retail companies that have the United States as their home market and are publicly traded on the National Association of Securities Dealers Automated Quotations (NASDAQ) and thus report their annual key financials.

To come to this selection, first the US database of Internet Retailer was used to see which of the companies located in that database were also located on the NASDAQ, which was a total of 98 (Internet Retailer 2013).

(21)

21 Then the unique stock symbols (commonly called Ticker or TIC symbol) of all those companies were obtained. Subsequently the COMPUSTAT historical segment database (this is a database of the Wharton Research University in Pennsylvania which has data of all the key fundamentals of all the companies located on the NASDAQ) was used to retrieve the required information per company. The following variables were obtained:

- Global Company Key - Segment Annual Fundamentals - Operating Profit (Loss)

- Net Sales

- Data Date - Segment Annual Fundamentals - Geographic Segment Type

- Segment Name - Company Name - Ticker Symbol

Where the ‘Global Company Key’ is a unique key for that company, the ‘Operating Profit’ is the bottom line result for a company in that year in that segment, the ‘Net Sales’ are the sales for a company in that year in that segment, the ‘Data Date’ is the date where the Net Sales/Operating Profit apply to (e.g. 20131231 for 2013), the ‘Geographical Segment Type’ is the (numerical) identifier of the geographical segment (e.g. domestic being “2” and foreign being “3”), the ‘Segment Name’ is the name of the geographical segment (e.g. “United States” or “China”), the ‘Company Name’ is the name of the company where the data applies to and the ‘Ticker Symbol’ is the unique stock symbol, or TIC symbol, for that particular company.

For every year, for every ‘Segment Name’ that was applicable for a company, the data was obtained. Resulting in for example the results given in Table 1, for ‘Net Sales’ for Blue Nile inc. (a company that sells diamonds and jewellery through the internet):

(22)

22 Table 1: BLUE NILE INC. Filtered on ‘Net Sales’, ‘Data Date’, ‘Segment Name’ and ‘Company Name’

Net Sales (in million $)

Data Date - Segment Annual

Fundamentals Segment Name Company Name

243.27 20061231 United States BLUE NILE INC

8.317 20061231 Other Countries BLUE NILE INC

Source: COMPUSTAT Historical Segment database

The COMPUSTAT database returned usable data for 95 companies out of the e-commerce dataset (for the other 3 no valid data was found). The reason for the 3 missing companies could have multiple causes, such as the data not being correctly filed with the NASDAQ or not correctly transferred to COMPUSTAT database.

From the data for the 95 companies all the duplicate values were taken out (in the database a lot of values occur twice, for instance if a correction had been made) and the ‘Segment Name’ variable was made consistent (e.g. changing all the North America, US etc. in to United States).

After that, the companies that didn’t have the United States as their home market were taken out by looking at the geographical segment and the name of the geographical segment.

(23)

23 If “2” (domestic) didn’t match United States, the company was taken out. This to minimalize the effects of geographical factors between the companies (to make the comparison between internet retail (e-commerce) and non-internet retail companies as clear as possible, other factors should have to be controlled as much as possible). After this, 65 internet retail companies that are located on the NASDAQ and have the United States as their domestic market remained.

To make the comparison between internet retailers and non-internet retailers as clean as possible, 65 non-internet retailers, which were/are situated on the NASDAQ and have the US as their domestic market were selected for the non-internet retail database.

To come to that selection, the entire database of NASDAQ companies was taken as a starting point. First the following sectors were chosen: consumer services, consumer durables and consumer non-durables to end up with retailers. As the NASDAQ doesn’t have a specific sector or industry for internet retailers, these sectors potentially included internet retail companies as well. To take these out, first the internet retail companies were added to the entire list and given a dummy-variable number of 0, while all the other companies got a number of 1. Then all the duplicates (duplicate TIC symbols) were removed to filter out the internet retail companies from the non-internet retailers set, which distinguished all the internet retail companies that were in the original list of consumer services, consumer durables and consumer non-durables companies from the rest (value of 0). After this, the internet retail companies were taken out again to come up with a clean traditional, non-internet, retailers set. Of these companies (a total of 987), the key financials were also obtained through the COMPUSTAT historical segment database.

This brings up the question why the set of internet retailers couldn’t be obtained from within the list of all the retail companies as well, because apparently there were some duplicates in there. The total amount of duplicate internet retailers removed was 12, which

(24)

24 means that the other 54 were situated in other segments (as mentioned, internet retail doesn’t have its own segment in the NASDAQ), which is the reason why the method of retrieving them from the Internet Retailer database was used.

The COMPUSTAT historical segment database initially returned data for 951 non-internet retail companies. First, the same procedure was followed as with the non-internet retail companies to take out the companies that didn’t have the United States as their home market and remove all duplicate values. After this, there were 819 non-internet retail companies left. To also control for time variables, the years wherein the internet retail companies and the non-internet retail companies first reported sales were aligned as much as possible. This was done by first making a frequency table of the years in which the internet retail companies first reported sales. The frequency table and the statistics are given in Table 2 and Table 3. Table 2: Statistics of frequency table of first reported sales year internet retail companies

N Valid 65

Missing 0

Mean (sales year) 2001.25

Median (sales year) 2000.00

Std. Deviation 5.887

Percentiles 25 1995.00

50 2000.00

75 2006.00

Table 3: Frequencies first reported sales year internet retail companies

Year Frequency Percent

Valid 1995 17 26.2 1996 6 9.2 1998 5 7.7 1999 3 4.6 2000 6 9.2 2002 4 6.2 2004 3 4.6

(25)

25 2005 4 6.2 2006 2 3.1 2007 1 1.5 2009 3 4.6 2010 4 6.2 2011 7 10.8 Total 65 100.0

To align with the first reported sales year frequencies of internet retail companies the following steps were taken: all the companies that had sales reported before 1995 were removed by looking for sales in the timeframe before 1995, hereby filtering out all the companies that reported sales before this timeframe (since the starting year of a company is not given in the COMPUSTAT database, this is the closest approximation to make, and as it aligns with what was done for the internet retail companies, it should give an accurate comparison). After this 445 non-internet retailers remained.

Subsequently an effort was made to choose the same amount of companies per year as seen in the frequency table (Table 3) for internet retail companies to control for time variables (e.g. companies starting foreign sales in 2000 or 2001 might have had a harder time going abroad than companies founded in 2010 due to macro-economic factors). However not for every year the same amount of companies were available (e.g. for 1995 only 10 non-internet retail companies were available). E.g. for 1995 this could be explained by the fact that a new technology brought a boom to the founding of internet retail companies, which led to a relatively high amount of new internet retail companies being founded in those years, for other years it might just be coincidence. In the end, an effort was made to come up with frequencies that approached the frequencies of internet retail companies as close as possible, which resulted in the frequency table and statistics given in Table 4 and Table 5 for non-internet retailers.

(26)

26 Table 4: Statistics of frequency table of first reported sales year non-internet retail companies

N Valid 65

Missing 0

Mean (sales year) 2000.86

Median (sales year) 1998.00

Std. Deviation 5.129

Percentiles 25 1997.50

50 1998.00

75 2005.00

Table 5: Frequencies first reported sales year non-internet retail companies

Year Frequency non internet retailers Percent non internet retailers Frequency internet retailers Percent internet retailers Valid 1995 10 15.4 17 26.2 1996 3 4.6 6 9.2 1997 3 4.6 0 0 1998 18 27.7 5 7.7 1999 3 4.6 3 4.6 2000 3 4.6 6 9.2 2001 1 1.5 0 0 2002 2 3.1 4 6.2 2003 1 1.5 0 0 2004 3 4.6 3 4.6 2005 3 4.6 4 6.2 2006 2 3.1 2 3.1 2007 2 3.1 1 1.5 2008 4 6.2 0 0 2009 1 1.5 3 4.6 2010 2 3.1 4 6.2 2011 4 6.2 7 10.8 Total 65 100 65 100

Important concluding remark on the two data sets is the following: the companies in the internet retail dataset have the internet as their main sales channel, and the non-internet retailers don’t. This however, in no way, means that the companies in the non-internet retail dataset have made no endeavours into internet retail whatsoever or the other way around, that

(27)

27 internet retailers have no physical store presence anywhere. Both can be true, and both are likely to be true in multiple cases. The distinction between the two datasets lays in where the emphasis is; the Internet Retailer database includes companies that have at least more than 50% from their sales originating from online sales (Internet Retailer 2013).

4.3 Data Preparation

After the data collection, several variables were added to both datasets to be able to conduct the necessary analyses. First a dummy-variable was made called ‘Ecommerce’ where the value of 1 indicates that the company is an internet retail company and a value of 0 that a company is a non-internet retailer. After this the two datasets were bundled into one dataset.

Subsequently, in the new bundled dataset more variables were added. First the ‘Data Date – Segment Annual Fundamentals’ variable was recomputed into a (new) ‘Rounded date’ variable that resembles the first four figures of the ‘Data Date – Segment Annual Fundamentals’ variable, making it e.g. 1997 instead of 19971231.

Then a variable named ‘Year 1995 = 0’ was computed where the ‘Rounded date’ is set so that 1995 is 0, 1996 is 1 etc. Furthermore, a variable was created (‘Year Start = 0’) that gives the year in which a company first started reporting sales as 0 (by making a pivot table of every company and sorting it on the lowest data in which a company conducted sales). E.g. if a company first started reporting sales in 2005, 2005 will be 0, 2006 will be 1 etc. The same thing was done for foreign sales, computing a variable (‘Year Foreign Sales’) that has a value of “9999” when the line applies to domestic sales, and the value of the ‘Year Start = 0” when the line included foreign sales.

For the amount of foreign sales, a variable was computed called ‘Foreign Sales’ that was used for pivot tables and simply gives the value of ‘Net Sales’ when the applicable line is applicable to foreign sales. Furthermore a variable was computed that gives the total amount

(28)

28 of sales in that year named ‘Sum of Year Sales’. This was done by making a pivot table that gives the sum of all the sales in a year. This is because all the lines are either foreign or domestic and can thus apply to the same year (e.g. domestic sales in 1998, foreign sales in 1998). With this variable the total sales (foreign and domestic) for a particular year are given.

Next, a dummy-variable that simply tells whether the line is applicable to foreign sales (‘Foreign Sales yes or no’) was computed, which takes a value of either 1 or 0. Then a variable called ‘% of Foreign Sales’ was made, which divides the ‘Foreign Sales’ by the ‘Sum of Year Sales’ to give the percentage of foreign sales in comparison to the total sales.

To assess the profitability, two variables were made. The first variable, called ‘Foreign Profit’, is comparable to ‘Foreign Sales’, only for profit instead of sales. Furthermore a variable was made called ‘% of Foreign Profit’ which divides the sum of ‘Foreign Profit’ in a year (there can be multiple lines for foreign, e.g. Europe and Asia) by the sum of ‘Foreign Sales’ in a year, which thus assesses the total foreign profit in a year in comparison to the total foreign sales in a year of a particular company.

Lastly a variable called “Merge Variable’ was made, which is a combination of the ‘Company Name’ and the ‘Rounded Date’, giving the ability to sort on a particular year of a particular company, including both domestic and foreign sales.

Everything as described above was done in Excel. Thus the treatment of the data in a preparatory stage was done in Excel. Once the data was ready for statistical analysis, the data was brought over the SPSS. However, for some analyses, new forms of data still had to be imported in to SPSS from Excel during the analyses. I.e. Excel gives the opportunity to make ‘pivot’ tables from data, which, for example, gives the possibility to make a table sort on the minimum, maximum or sum of a certain variable, hereby creating new variables such as the minimum year in which foreign sales were conducted. If such a pivot table is used, this is described in the analysis. The results of the analyses are discussed in the following chapter.

(29)

29

5.0 Analysis / results

With all the variables in place and two clean datasets, it became possible to test the hypothesis and gain insights on the research questions. As mentioned, the statistical analyses were done in SPSS. The following paragraphs describe the analyses made and the results of these analyses.

5.1 Foreign market entry and speed of entry

The first hypotheses that were tested were:

1.a. The % of internet retail companies going abroad is significantly higher than that of non-internet retail companies going abroad

1.b. When retailers go abroad, internet retail companies conduct sales outside of their foreign markets significantly sooner in their lifecycle than non-internet retail companies

To test this hypotheses the following variables were used: ‘Foreign Sales Yes or No’, which simply has a value of 1 for yes and 0 for no, ‘Year Foreign Sales’, which as discussed is able to have a value between 0 (a company started doing sales abroad at the moment that it started reporting sales as a company, also referred to as ‘Born Global’) and 19 (if a company started reporting sales in 1995 and the line under examination is for data applicable to 2013 and resembles sales in a foreign market). However, for domestic sale lines in the dataset, the value of this variable is ‘9999’ (which is excluded in the analysis). The last variable used was ‘Ecommerce’ (with a value of 0 or 1) to distinguish between internet retailers and non-internet retailers.

First, to test hypothesis 1.a., an assessment was made how many companies from both groups (internet retailers and non-internet retailers) actually went abroad at some point (this because it is very possible that a company actually never started conducting sales abroad).

(30)

30 This is easily seen in SPSS by looking at a crosstab analysis. This crosstab analysis sorts a pivot table made in Excel, which shows for every company the maximum value of ‘Foreign Sales Yes or No’ (1 meaning that the company did foreign sales at some point, 0 meaning that it didn’t) by ‘Ecommerce’. The results are shown in Table 6.

Table 6: Frequencies ‘Foreign Sales Yes or No’ sorted by ‘Ecommerce’

Foreign sales yes or no

Total 0 1 Non-Ecommerce = 0 Ecommerce = 1 0 24 41 65 1 30 35 65 Total 54 76 130

Source: COMPUSTAT Historical Segment database, Thesis selection

The above table indicates that for 24 non-internet retailers and 30 internet retailers, no foreign sales were found. This is a difference of 14.4%. To test if this is a significant difference, a Chi-Square test was used. The output of this test is given in Table 7:

Table 7: Chi-Square test ‘Foreign Sales Yes or No’ sorted by ‘Ecommerce’

Value df Asymp. Sig. (2-sided) Exact Sig. (2-sided) Exact Sig. (1-sided) Pearson Chi-Square 1.140a 1 .286 Continuity Correctionb .792 1 .374 Likelihood Ratio 1.142 1 .285

Fisher's Exact Test .374 .187

Linear-by-Linear Association

1.132 1 .287

N of Valid Cases 130

Pearson Chi Square gives a value of 1.140 and the significance is .286. Because the hypothesis is one sided, the actual significance is .286 divided by 2: .143 meaning that the result is not statistical significant (under a 95% significance assumption) We must thus reject the hypothesis 1.a; the % of internet retail companies going abroad is significantly higher than that of non-internet retail companies going abroad.

(31)

31 To test hypothesis 1.b; when retailers go abroad, internet retail companies conduct sales outside of their foreign markets significantly sooner in their lifecycle than non-internet retail companies, first the normality of the distribution of the variable ‘Year Foreign Sales’ had to be assessed to determine whether a parametric or non-parametrical test should be used.

The distribution of this variable was determined by looking at the variable ‘Year foreign Sales’ for both the research groups (non-internet retailers and retailers), where companies who never went abroad were taken out. First an indication on the normality could be found by looking at a histogram of the variable (Figure 1). Here is already to be seen that the distribution, especially of internet retailers, most likely doesn’t follow a normal distribution.

Figure 1: Histogram showing ‘Year Foreign Sales’ sorted by ‘Ecommerce’. Source: COMPUSTAT Historical Segment database, Thesis selection.

To verify whether this indication was true, a more in depth analysis was made. In this case, the amount of elements for both internet retailers and non-internet retailers is less than 2000. Thus Wilk can be best used to test the normality. The significance of the

(32)

Shapiro-32 Wilk test is for both groups .000 (Table 8), meaning that the distribution is most likely not normal. To ensure this conclusion the Z values (calculated by dividing the test statistic by the standard error) of the Skewness and the Kurtosis were computed (Table 9). For non-internet retailers, the Z values are 1.15 for the Skewness and -8.60 for the Kurtosis. For the internet retailers, the Z values are 1.03 for the Skewness and -4.33 for the Kurtosis. This means that both the Z values for the Skewness are in the acceptable range between -1.96 and 1.96; however the Z values for the Kurtosis, both, are well outside that range, concluding that a normal distribution could not be assumed and used.

Table 8: Test of normality ‘Year Foreign Sales sorted by ‘Ecommerce’

Non-Ecommerce = 0 Ecommerce = 1

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

Year foreign sales 0 .090 1326 .000 .955 1326 .000

1 .089 376 .000 .954 376 .000

Table 9: Skewness and Kurtosis ‘Year Foreign Sales sorted by ‘Ecommerce’

Non-Ecommerce = 0

Ecommerce = 1 Statistic Std. Error

Year foreign sales 0 Mean 8.31 .143

Skewness .077 .067

Kurtosis -1.152 .134

1 Mean 7.33 .252

Skewness .129 .126

Kurtosis -1.086 .251

An attempt was made to transform the variable ‘Year Foreign Sales’ into a variable that did have a normal distribution. First a variable was computed called ‘SQRT Year Foreign Sales’ which took the square root of the variable ‘Year Foreign Sales’. In addition a variable was computed called ‘LN Year Foreign Sales’ which took the Log10 of the variable ‘Year Foreign

(33)

33 Sales’. Both transformations didn’t lead to a normal distribution; graphs to exemplify that are given in Figure 2 and Figure 3.

Figure 2: Histogram showing ‘SQRT Year Foreign Sales’ sorted by ‘Ecommerce’. Source: COMPUSTAT Historical Segment database, Thesis selection.

Figure 3: Histogram showing ‘LN Year Foreign Sales’ sorted by ‘Ecommerce’. Source: COMPUSTAT Historical Segment database, Thesis selection.

(34)

34 When looking at the graphs and the data, some explanation can be found for the data not being normally distributed. From both groups, the amounts of observations per year are quite similar. Obviously, with a normal distribution the lower and higher values should have relatively less observations. The relatively same amount of observations causes a flat distribution and thus a high Kurtosis, instead of a normal distribution.

Because a normal distribution couldn’t be assumed or computed, a non-parametrical test was conducted to compare the speed of internationalization between the internet retail group and the non-internet retail group. With two independent samples, the Mann-Whitney U test was conducted on the variable ‘Year Foreign Sales’, grouped by ‘Ecommerce’ to compare the Median of the two groups. The results are given in Table 10 and Table 11.

Table 10: Ranks ‘Year Foreign Sales’ sorted by ‘Ecommerce’

Non-Ecommerce = 0

Ecommerce = 1 N Mean Rank Sum of Ranks

Year foreign sales 0 1326 871.66 1155816.00

1 376 780.42 293437.00

Total 1702

Source: COMPUSTAT Historical Segment database, Thesis selection Table 11: Mann-Whitney U test statistic ‘Year Foreign Sales’ sorted by ‘Ecommerce’

Year foreign sales

Mann-Whitney U 222561.000

Wilcoxon W 293437.000

Z -3.182

Asymp. Sig. (2-tailed) .001

The effect size (R) equals Z divided by the square root of the total amount of observations (N). In this case -3.182 divided by SQRT(1702), which comes to a R of -0.08. This means there is a small effect. In addition the small effect is significant, because the 2-tailed

(35)

35 significance is .001 and in this case the hypothesis tested is again one sided (which means the value to look at is .001 divided by 2, 0.0005), which is smaller than .005.

Looking at the medians of the two groups (see Table 12), there is to be seen that the median of non-internet retailers is higher than the median of internet retailers (8.00 vs 7.00). This in combination with the results from the Mann-Whitney U test leads to the conclusion that internet retailers do go abroad significantly sooner than non-internet retailers, when they go abroad at all. The effect size and difference, however, is small.

Table 12: Mean and median of ‘Year Foreign Sales’ sorted by ‘Ecommerce’

Year foreign sales Ecommerce = 0 Ecommerce = 1

N Valid 1326 376

Missing 801 586

Mean 8.31 7.33

Median 8.00 7.00

Std. Deviation 5.192 4.877

With these results, hypothesis 1.b; when retailers go abroad, internet retail companies conduct sales outside of their foreign markets significantly sooner in their lifecycle than non-internet retail companies doesn’t have to be rejected.

5.2 Amount of sales from foreign markets

The second research question of the research was the following: do internet retail companies have a larger share of their net sales emanating from foreign countries than non-internet retail companies? As discussed, sales originating from foreign countries entail sales exported to foreign countries or made by a foreign vendor. Sales can be internet sales or non-internet sales, the comparison will be on a more macro level between companies and where their main focus lies (internet retail or non-internet retail) and thus not on individual sales. To find an answer to this research question, the following hypothesis is tested:

(36)

36 2. Overall, the % of net sales coming from foreign markets for internet retail companies is significantly higher than that of non-internet retail companies.

The following variables were used to test this hypothesis: ‘Merge Variable’, which gives a value like ‘AMAZON.COM INC1995’, ‘% of Foreign Sales’, which gives a value of a % of the ‘Sum of Net Sales’ when that line applied to foreign sales and lastly the variable ‘Ecommerce’. By making a pivot table, per year, per company, it became visible what the percentage of Foreign Sales was for that year.

To find the difference between internet retail companies and non-internet retail companies, again the normality of the distribution had to be tested in order to determine whether a parametric or non-parametrical test should be used. The same steps were undertaken as was done to test hypothesis 1.b. The results were as follows.

The Z value for the Skewness and Kurtosis were 12.4 and -0.41 for non-internet retailers and 19.59 and 21.72 for internet-retailers, which means that normality couldn’t be assumed even remotely. In addition, Shapiro-Wilk gave a significance of 0.000, indicating non-normality. The distribution is given in Figure 4.

(37)

37 Figure 4: Histogram showing ‘% of Foreign Sales sorted by ‘Ecommerce’. Source: COMPUSTAT Historical Segment database, Thesis selection.

Looking at Figure 4, the non-normality can be explained by the relative large amount of ‘0’ values that are present in both groups. This means that for relatively many lines of data / years, companies didn’t conduct sales abroad. With this as a given, the transformation of the variable ‘% of Foreign Sales’ to a variable that will be normally distributed is not likely to bear much fruit. However, an effort was made.

As with the testing of hypothesis 1.b. a square root variable was computed and a LN10 variable was computed. The Z values for the square root variable were 4.06 and -7.72 for non-internet retailers and 8.83 and -2.85 for non-internet retailers respectively. For the LN10 variable the Z values ended up at -13.71 and 14.88 for non-internet retailers and -11.68 and 12.22 for internet retailers. This means, again, a non-parametric test had to be conducted to compare the variable ‘% of Foreign Sales’ between the two groups of non-internet retailers and internet retailers.

(38)

38 Like for hypothesis 1.b, and because there are two independent sample groups, Mann-Whitney U was used to compare the medians of the two groups. The results are given in Table 13 and Table 14.

Table 13: Ranks ‘% Foreign Sales’ sorted by ‘Ecommerce’

Non-Ecommerce = 0

Ecommerce = 1 N Mean Rank Sum of Ranks

Sum of % of Foreign sales 0 797 764.10 608991.50

1 591 600.63 354974.50

Total 1388

Source: COMPUSTAT Historical Segment database, Thesis selection Table 14: Mann-Whitney U test statistic ‘% Foreign Sales’ sorted by ‘Ecommerce’

Sum of % of Foreign sales Mann-Whitney U 180038.500 Wilcoxon W 354974.500 Z -7.944

Asymp. Sig. (2-tailed) .000

The effect size ‘R’ in this case is 7.944 divided by SQRT(1388), which results in a R of -0.21. This means there is a small till medium effect. In addition this effect is significant, because the 2-tailed Asymptotic Significance is 0.000, which is well below the 0.05 benchmark. Table 15 gives the medians for the two groups.

Table 15: Mean and median of ‘% Foreign Sales’ sorted by ‘Ecommerce’

% of Foreign sales Ecommerce = 0 Ecommerce = 1

N Valid 797 591

Missing 0 0

Mean .264938 .121684

Median .117135 0.000000

Std. Deviation .3249936 .1954225

(39)

39 The median for the non-internet retailers is 11.71% while the median for internet retailers is 0%. With the results as described above, it can be concluded that non-internet retailers have a larger % of their net sales originating from foreign markets. With this, the second hypothesis, Overall, the % of net sales coming from foreign markets for internet retail companies is significantly higher than that of non-internet retail companies, has to be rejected.

5.3 The profitability of international markets

The last hypothesis that was tested is:

3. The % of net profit for internet retail companies, gained from international markets, is significantly higher than for non-internet retail companies.

In contrast with hypothesis 2, to test this hypothesis, this research only looked at lines containing foreign sales, thus excluding lines with domestic sales. This is because the aim for the research question / hypothesis is to assess whether it is true that when an internet retail company goes abroad, it is relatively more profitably than non-internet retailers, as argued in theory by Singh and Kundu and Keldo and Sivakumar (Singh and Kundu 2002; Keldo and Sivakumar 2004).

To test this hypothesis, the following variables were used: ‘Merge Variable’, ‘% Foreign Profit’, which as discussed in section 4.3 can take any value in the form of any percentage (e.g. it technically could also have a value of -200% when the investments in foreign markets are larger than the foreign sales of the company in that year) and finally the independent variable ‘Ecommerce’.

At this point an important note should be made: for 39 of the total 130 companies in the dataset, the variable ‘Operating Profit’ was missing for all years. The primary focus of the COMPUSTAT Historical Segment database is Net Sales; other variables are not as complete,

(40)

40 due to companies not having them filed with the NASDAQ, the NASDAQ not having them filed completely or the COMPUSTAT database not having them filed completely or not at all (most likely scenario). In addition to the 39 companies that didn’t have any profitability numbers, for 32 companies the profitability numbers of foreign markets were missing. When these and the companies that didn’t conduct sales abroad at all were taken out (by filtering out all the companies that had an invalid value for ‘% Foreign Profit’), just 28 companies with usable data remained, of which 16 non-internet retailers and 12 internet retailers. This thus drastically reduced the research groups in size, which will have an effect on the statistical significance of any analysis.

With this in mind, the normality of the distribution of the ‘% Foreign Profit’ was the first thing to assess. By just looking at the data, it became apparent that there were a few, and one very big in particular, outliers. The Boxplot is given in Figure 5.

Figure 5: Boxplot of ‘% Foreign Profit’ sorted by ‘Ecommerce’. Source: COMPUSTAT Historical Segment database, Thesis selection.

(41)

41 The biggest outlier (marked as ‘*1’) comes from a company called Netflix. Netflix is an internet retail service company that offers media (movies, series etc.) through streaming on the internet to its customers. Early in this decennium, Netflix heavily invested in internationalizing the company, expanding it in to Europe, Canada and South America (Garrahan 2014). Those heavy investments led to the outlier in Figure 5, which applies to the year 2010 where Netflix invested over $12 million in international expansion and earned only $3.6 million. A year later (2011), Netflix invested over $106 million in international expansion and earned $82.85 million, which is shown in outlier ‘*3’. This is an interesting case on its self, however not in the scope of this research.

Other outliers included similar cases, for Netflix or for other companies like Groupon and Egain. Because these values are accurate and valid, removing the outliers wasn’t preferable. But even when an attempt was made to remove the extreme outliers to come to a normal distribution, this was unsuccessful. Even after removing the 5 biggest negative outliers and the 1 most positive outlier the Z scores were still -3.66 and 6.56 for non-internet retailers and 1.52 and 3.95 respectively.

Again the transformations square root and Log10 were made. Because the variable ‘% of Foreign Profit’ can have a negative value, the transformation was made in the form of square root / Log10 of ‘% Foreign Profit’ plus the most negative value, in this case 3.3503 (this because making a square root or LN transformation from negative values will not give valid results). The Z values for the newly computed SQRT variables were -5.83 and 37.55 for the Skewness and the Kurtosis of non-internet retailers and -32.15 and 143.56 for internet retailers. The Z values for the newly computed variable for the LN transformation were 7.86 for Skewness and 40.65 for Kurtosis and -40.65 for Skewness and 200.6 for Kurtosis for non-internet retailers and non-internet retailers respectively. These high values of the Kurtosis can be explained by the large amount of outliers in combination with the relatively small amount of