Introduction

Brouthers, Chen, Sali, and Shaheer (this issue) argue that “technological advancements and improved global connectivity are providing businesses novel ways to enter foreign countries” (p. 22- all page references are to the online version), and that “these new forms of foreign country involvement”, namely capital access, innovation outposts, virtual presence, and managed ecosystems, “have seen little discussion in the traditional entry mode literature” (p. 4). Because it is dominated by transaction cost theory (TCT), they deem this literature unable to explain those modes which “do not solely serve as a vehicle for transaction cost minimization”, but “can be better explained through the application of theories focusing on (1) deployment of existing capabilities or access to new capabilities (Teece, 2014) and (2) firm ‘involvement’ instead of investment into foreign markets (Liesch, Buckley, Simonin, & Knight, 2012)” (p. 2). To remedy this, Brouthers et al. propose a new model based on whether the entry mode is undertaken for exploration or exploitation and on its degree of embeddedness.

Brouthers et al. are quite right about the need to bring to the fore important contributions to the foreign entry mode literature that may not have received the attention they deserve. Their article prompts a careful analysis of what is actually new in these entry modes, and what can be explained by extant theory. Perhaps most importantly, the criticisms the authors lodge against the TCT of entry modes provide an opportunity to update and clarify it.

In the next section I address some of the criticisms Brouthers et al. make of the TCT of foreign entry modes. I then describe and discuss their four supposedly new entry modes, and argue that they are neither new nor do they challenge a TCT-based theory of entry modes. In contrast to their exploitation–exploration–embeddedness (EEE) framework, which they acknowledge is not a theory but a conceptualization, I show that the extant TCT-based theory of foreign entry modes has made testable predictions as to when these new entry modes are optimal. I therefore see no reason to discard existing theories in favor of the authors’ EEE.

The Transaction Cost Theory of Foreign Entry Modes

Brouthers et al.’s Critique of the Transaction Cost Theory of Foreign Entry Modes

In a section entitled ‘Entry mode research’ meant to provide support for their EEE model, Brouthers et al. pinpoint two weaknesses of the extant foreign entry mode literature. First, they write that, despite the development of new perspectives such as institutional and network theory,

TCT continues to dominate our thinking to the extent that foreign subsidiaries are often viewed as the preferred organizational structure because of their capacity to protect from misappropriation firm advantages to be exploited (Hennart, 1991), while much less attention is paid to modes of entry in regard to capability augmentation (Cantwell & Mudambi, 2005; Frost, Birkinshaw, & Ensign, 2002) (p. 3).

Second, extant theory has, according to them, focused too much on the minimization of transaction costs and not enough on transfer of capabilities. They write that “although the FDI [foreign direct investment] portion of entry mode studies frames ownership as a means of containing transaction costs (Brouthers, 2013), much of the non-equity mode literature tends to focus on efficient transfer of capabilities instead of control of operations as the main driver of mode choices (Erramilli, Agarwal, & Dev, 2002)” (p. 4). New modes of involvement such as “accessing foreign financing without further entry” (p. 4), and using outposts to access foreign knowledge and the internet to service markets from a distance should be “conceptualized as vehicles for transferring (and sourcing) knowledge, instead of a governance structure for the exchange of intermediate goods or complementary assets in a specific transaction” (p. 4).

What should be made of these criticism? Before we consider the authors’ allegedly new entry modes and the EEE model they use to categorize them, it is useful to provide some updates and clarifications on the TCT of entry modes.

The Transaction Cost Theory of Foreign Entry Modes

Foreign market entry modes are governance structures to acquire complementary resources and exploit firm capabilities through sales to intermediate and/or final customers.1 It is generally acknowledged that entry modes have two main dimensions, location and governance, which are conceptually distinct (Dunning, 1993; Hennart & Slangen, 2015). The first dimension is location. In some cases, production is most efficiently located at home, with the product shipped to customers in the foreign country, the service provider traveling from the home office to the foreign customer, or the customer coming to the service provider. In other cases, producing the product or service in the target foreign country is optimal. The second dimension is governance. A firm must decide whether it wants to undertake an activity with its own employees or to contract it out to independent parties. For example, it chooses how it wants to exploit its intangibles, whether to rent or sell them to other firms in the market for intellectual property, or instead embed them in products and services to be sold on product markets. In that case, a firm will either manufacture the goods at home and export them to the foreign market, or, if locating production in the foreign market is optimal, operate a foreign plant with its own employees.

There are many alternative ways by which firms can organize their foreign activities, but they all have location and governance dimensions. For example a firm that produces at home and serves foreign markets through exports can manage sales through its own employees or through independent parties, agents, or distributors (the governance dimension). In the first case, employees can be located at HQ or in foreign sales subsidiaries. In the second, agents or distributors can be located in the exporter’s home country, or in the targeted foreign countries (Hennart, 1991). The two dimensions are seen as separate because it does not follow that, if the optimal locus of production is abroad, the firm must necessarily operate there with its own employees, as it could license or franchise a foreign producer.

Transaction cost theorists see governance, i.e., the choice between hierarchical, market, and hybrid modes of organization, as a key dimension of entry modes because they believe that governance affects outcomes. TCT is a theory of optimal governance. Hennart (1993) argues that there are only two generic methods available to organize interdependencies, the price system and hierarchy. These two generic methods of organization provide different incentives, and hence generate different outcomes. The price system rewards agents based on their output, while hierarchy functions through the control of behavior. Institutions use a mix of market and hierarchy, with markets relying predominantly, but not exclusively, on the price system, and firms mostly, but not exclusively, on hierarchy. Hybrid institutions, such as franchising, feature a more even mix of behavior and output controls. Control of behavior is achieved in firms by the employment contract, by which agents agree, in exchange for a fixed salary, to obey the orders of the boss. Activities organized through employment contracts, i.e., by employees, are inside the firm, while those linked to the firm through spot markets or external contracts are outside (Hennart, 1993).

Each method of organization has its own limitations. Using the price system will be problematic when output is difficult to measure and/or when it is not protected by strong property rights. Control of behavior will be inefficient when it is difficult to observe, or if it is imperfectly related to performance. Consequently, whether it is optimal to organize international interdependencies mainly through the price system (the market), mainly through hierarchy (within the firm), or through a mix of the two (hybrids) will hinge on the characteristics of the interdependency.2 Because the context matters, it is quite common to see firms entering foreign markets with diverse entry modes (e.g., using distributors in some countries and sales subsidiaries in others), and indeed even choosing within a given foreign country different entry modes for different consumer segments.3

Comparing equity with non-equity entry modes is the bread and butter of transaction cost theorists. While it is true that many IB scholars have focused on the choice between shared and full ownership modes, many of them have also looked at the relative efficiency of equity versus non-equity ways of exploiting knowledge (e.g., Arora & Fosfuri, 2000; Davidson & McFetridge, 1985) and of accessing it (e.g., Davies, 1977; Hennart, 1989); of exploiting reputation (e.g., Contractor & Kundu, 1998; Fladmoe-Linquist & Jacque, 1995); of acquiring raw materials and natural resources (e.g., Casson & Associates, 1986; Hennart, 1988; Stuckey, 1983); and of accessing finance (Hennart, 1994a, b). If, along with Liesch et al. (2012), we call equity forms of market entry ‘investment’ and non-equity ones ‘involvement’, then the transaction cost approach is one that systematically compares the two.

TCT predicts the choice of optimal governance to organize a specific interdependence in a particular context. It does not predict the process by which governance decisions are made. It posits that, given sufficient competition, the optimal arrangement predicted will be the one observed, as wrong decisions will be undone or the firm will go bankrupt (Williamson, 1985). On page 2, Brouthers et al. write that the four modes of entry they analyze “do not solely serve as a vehicle for transaction cost minimization”, implying that, for transaction cost theorists, this is the criterion for entry mode choice. This is not the case. TCT implies that the optimum governance structure is the one that generates the highest net gain, and that can sometimes be the one with the highest transaction costs. For example, managers can choose between using salaried salespeople and contracting with outside sales representatives paid on commission. Paying commissions requires identifying which salespersons have made the sale, adjudicating their competing claims, and verifying that the orders have been paid. This can involve higher transaction costs than paying salespeople a fixed salary independent of their sales performance. Yet, the mode with higher transaction costs will be chosen when the additional revenues brought about by higher output incentives more than make up for the costs of implementing them (Anderson & Schmittlein, 1984; Hennart, 2015).

Brouthers et al. also argue that a transaction cost-based explanation differs from one emphasizing the efficient transfer of capabilities. On page 4, they write “although the FDI portion of entry mode studies frames ownership as a means of containing transaction costs (Brouthers, 2013), much of the non-equity mode literature tends to focus on the efficient transfer of capabilities instead of control of operations as the main driver of mode choice.” In fact, TCT analyzes which governance mode is optimal to  transfer capabilities, including technological capabilities. Transaction cost theorists explore whether transferring knowledge (a capability) within a firm will generate greater net gains than transferring it through a licensing contract or some other non-equity arrangement. Because TCT focuses on interdependencies, the same theoretical arguments explain both knowledge exploration and knowledge acquisition: if knowledge is tacit, and its transfer subject to high market transaction costs, then it will be difficult to both exploit and acquire. Hennart (1989), for example, discusses the difficulties experienced by Algeria in obtaining tacit knowledge through turnkey contracts, and why that knowledge would have been more efficiently transferred through equity, because hierarchical control better aligns the incentives of knowledge senders and receivers.

Lastly, transaction costs theorists acknowledge that foreign entry and expansion involves bundling the MNE’s intangibles with local complementary resources. Consequently, the optimal entry mode is not unilaterally decided by the foreign direct investor based on a trade-off between its need for control and its desire to minimize risk, as in Anderson and Gatignon (1986), but is the one that allows for the most efficient bundling of the resources necessary for foreign market entry. To predict the optimal mode of entry, one must therefore simultaneously look at the transactional characteristics of all the resources being bundled, the intangibles brought by the foreign direct investor, and the complementary resources required for their exploitation. When the markets for local complementary assets in a given foreign host country are more efficient than those for the intangibles held by the foreign direct investor, the latter will operate in the host country with a wholly-owned subsidiary. In the opposite case, there will be no foreign direct investment, as the local owner of complementary assets will acquire intangibles on international markets, and will remain the sole owner of the business. When intangibles and local complementary resources are equally difficult to transact on markets, a joint venture between the foreign and the local firm will be optimal (Hennart, 2009). This approach highlights the fact that most entry modes involve both foreign resource acquisition (exploration for Brouthers et al.) and resource exploitation. To do a greenfield entry, intended to exploit its intangibles, a foreign direct investor must search for and access local complementary resources on the four markets available (those for inputs necessary to build complementary resources, for the services of such resources, for the assets that yield such resources, or for the local firm(s) in which these resources are embedded), a task which Brouthers et al. would describe as exploration.

Brouthers et al.’s New Foreign Entry Modes

Capital Access

The first of the four entry modes introduced by Brouthers et al. is capital access. They state that “capital access entry modes refer to foreign market entries for seeking and gaining access to new financial resources while at the same time undertaking little, if any, other activities in the foreign market” (p. 5). Brouthers et al. concede that this is “not a completely novel mode of entry”, but argue that “recent transitions in financial regulations, emergence of international venture capital firms, and ease of financial flows across countries have made this task easier and more common” (p. 5). They note that access has taken the form of

initial public offerings (IPO), seasoned equity offerings (SEO), cross-listings in foreign stock markets, bank loans, foreign bond issues, engagement with private equity firms, international investment syndicates, foreign venture capital (VCs), sovereign wealth funds (SWFs), as well as several informal equity capital channels…. For example, companies such as Alibaba and Aramco sought IPOs in the US without operating or having physical assets in the “host” country….Similarly, many start-ups and other young ventures receive funding from foreign VCs while staying in their home countries (Humphery-Jenner & Suchard, 2013; Liu & Maula, 2015) (p. 5).

These entry modes which “have not been widely discussed in the entry mode literature” (p. 4) constitute “a low-cost, low control mode of entry” because they “do not rely on equity investment and have limited reference to control” (p. 15). They do not require strong embeddedness in the foreign country providing the finance (see the authors’ Table 2).

This description conflates, in my view, two separate types of foreign market entry. The first is venture capital (VC) firms going into a foreign country to provide funding to local firms, for example US VCs taking equity in Chinese start-ups, or syndicating with local VC firms to do so. Of the 21 articles referenced by Brouthers et al. as discussing capital access, 13 deal with this type of entry. Since the business of VCs is to take stakes in local start-ups, the issue of “undertaking little, if any, other activities in the foreign market” does not arise here, as it does in the case of foreign manufacturing, extracting, or service firms with no activities in the country where they source finance (e.g., Alibaba listing in the US). Foreign investment by VC firms is not particularly novel, and is well explained by the TCT of entry modes. VC firms provide finance – as well as management skills and legitimacy – to the firms in which they take a stake; they enter the markets in which there is a demand for what they have to offer, subject to the well-known limitations imposed by the liability of foreignness (Carneiro, Moreira, & Sheng, 2022; Guler & Guillen, 2010). The logic of their foreign entry is similar to that of banks, consulting firms, and other service providers.

A more interesting case is that of firms entering a foreign country to raise capital, even though they have no operations there. Pagano, Röell and Zechner (2002) show that firms list abroad when they need capital but cannot obtain it efficiently in their own country. In the late 1980s, the bulk of foreign listing was carried out by European high-tech firms listing in the US. More recently, their place has been taken by Israeli and Chinese firms (US China Economic and Security Review Commission, 2021). By mid-2021, there were around 400 Chinese firms with US listings (Economist, 2021). How do we make sense of this?

A few IB scholars have already looked at this phenomenon, and at the choice between sourcing foreign debt and equity, which Brouthers et al. suggest should be investigated. These scholars point out that the phenomenon of firms floating shares in a foreign country without having any other activity there is not new, and that it was common before WW1. These firms, which have been called “free-standing” firms, were headquartered in the main financial centers of the time – London, Paris, Brussels, and Amsterdam – but had no other activity in those places (Casson, 1994; Hennart, 1994b; Jones, 2005; Wilkins & Schroeter, 1998). Because in most of them HQ exercised strategic control on foreign operations – see the evidence in Hennart (1994b) – they were foreign direct investors. Corley (1998) estimates that this form of foreign direct investment (FDI) made up by 1914 close to half the total value of FDI worldwide. Aramayo Franke, for instance, a major Bolivian tin mining company, incorporated itself in London in 1906 and floated shares there to obtain capital for the development of its Bolivian tin mines (Contreras, 1993).4 Free-standing firms are still around. An example is Guyana Goldfields, a Canadian firm listed on the Toronto Stock Exchange, whose only business is mining gold in Guyana.5

As mentioned earlier, capital access can take many forms: firms located in countries with scarce capital but with potentially profitable projects can obtain loans from foreign financial institutions, sell bonds to foreigners, be fully or partially acquired by foreign VCs or MNEs, and register in a foreign country and sell shares there – the case of free-standing firms. Brouthers et al.’s EEE model is silent on the determinants of that choice, but there is a TCT-based literature that throws light on this (Hennart, 1994a, b).

The starting point is the realization that the market transfer of funds between those with excess liquidity and those with projects needing financing incurs transaction costs. Funds are lent today in the expectation that they will be repaid later with interest. Because borrowed funds are fungible, they can often be used for unintended purposes. Incentives to lenders and borrowers are not symmetrical, since borrowers capture the gains of profitable projects net of interest payments, while lenders are stuck with the losses if projects fail. Hence, borrowers have incentives to take excessive risks. Lenders can use three non-mutually exclusive strategies to protect themselves. They can ask borrowers to pledge a collateral. They can stick to projects with which they are familiar and to borrowers with a good track record. They can control the use of funds through lending covenants. When lenders cannot implement these strategies, for example, because they do not have information on borrowers or when the assets acquired with the loan do not provide good collateral, the best solution is to vertically integrate into the project, i.e., lenders will become project equity owners. This has two main advantages: first, equity owners have real-time control over the behavior of borrowers and can direct them to minimize the chances of default; second, this solution aligns the risk preferences of lenders and project managers. On the other hand, salaried project managers are likely to exert less effort than independent entrepreneurs and to opt for overly conservative strategies (Hennart, 1994a; Williamson, 1988).

We would therefore expect projects with assets which can provide good collateral, those in established industries, and those led by entrepreneurs well-known to lenders, to be able to use debt financing (e.g., bank debt or bonds sold to the public), while those which offer poor collateral, or which use new and/or untested technologies, or are led by entrepreneurs with poor or no track record (the case of many high-tech start-ups) will have to rely on equity financing: they will sell equity shares and/or invite VC participation. A second implication is that obtaining financing is facilitated by being located physically close to investors who know you and your business.

Another important characteristic of financial transactions is intermediation. Financial transactions are intermediated when a third party, bank, trading company, or manufacturing firm stands between fund supplier and fund user. Intermediation is efficient when there are significant differences between the average size and time structure of savings and investments. Interaction between the level of equity and that of intermediation yields four cases: (1) when loans cannot be obtained but there is no benefit to intermediation, firms located in capital-poor countries will sell equity shares to investors in capital-rich countries – the foreign listing and free-standing firm cases; (2) when loans cannot be had but intermediation is efficient, VCs and banks in capital-rich countries will take equity in firms in capital-poor countries; on the other hand, when the loan market is reasonably efficient, (3) the optimal solution will be for firms in capital poor countries to sell bonds to the public of capital-rich countries when there is no need for intermediation; and (4) to borrow from banks and other financial intermediaries when intermediation is efficient (Hennart, 1994a).

The empirical evidence broadly confirms these predictions. Projects with good collateral are typically financed through loans (through bank financing or by selling bonds to the public) while those with poor collateral are financed through equity. Before WWI, US mining investments, which have poor collateral, were set up as free-standing firms, many of them registered in the UK and selling their shares on UK stock exchanges. US railroads, on the other hand, which could offer their real estate holdings as collateral, financed themselves by selling bonds to European investors (Wilkins, 1989). Today, the US has the largest equity market, and is home to the most savvy high-tech investors. As predicted, firms that carry out a foreign IPO (mostly in the US and the UK) are those based in countries with inefficient equity markets. Listing abroad allows them to raise more funds than listing at home (Caglio, Hanley, & Marietta- Westberg, 2016). The European, Israeli, and Chinese startups that list on US stock markets are overwhelming high-tech ones (Blass & Yafeh, 2001; Pagano et al., 2002). For many Chinese firms, accessing capital is the main reason why they enter the US, so they often have no other activity there. Lastly, as we will see below, high transaction costs in equity markets often push lenders and equity investors to entrust their funds to those they personally know; consequently, borrowers benefit from having close personal ties to lenders, and often locate next to them for that purpose, thus casting doubt on Brouthers et al.’s contention that capital access is always characterized by low embeddedness.

Innovation Outposts

Innovation outposts are the second new entry mode identified by Brouthers et al., who write that they are used to “cultivate relationships with foreign business networks to access their unique knowledge and resources” (p. 7). What do we know about them? Most are established in Silicon Valley and in Israel by Asian and European firms (Decreton, Monteiro, Frangos, & Friedman, 2021; Mind the Bridge & Nesta, 2018; Mind the Bridge, 2019a). A recent survey of innovation outposts in Silicon Valley (Mind the Bridge, 2019a) classifies them into four key types: Corporate Venture Capital Offices, hosting partners and associates of VC funds, the most common outpost with 76 firms using this type; R&D Centers with more than 50 employees  used by 70 firms; Corporate Innovation labs, which act as incubators and lean R&D centers – used by 66 firms; and Innovation Antennae, which consist of teams of up to ten employees engaging with local start-ups – the least common type with 34 firms, and the one Brouthers et al. seem to have in mind.6 The report documents the recent growth of these outposts, with 80% opened since 2010, and a shift towards leaner structures (Corporate Innovation Labs and Corporate Antennae), reflecting the trend towards open innovation.

For Brouthers et al., the novelty of innovation outposts lies in the fact that, in contrast to “traditional entry through internalization and control” (p. 19), they “do not require foreign equity investments, establishment of foreign subsidiaries or sales offices, or even formal contractual arrangements” (p.  6). In fact, most outposts in Silicon Valley and in Israel are set up as wholly-owned subsidiaries, and all involve internalization and control, since they are operated by employees (or partners in the case of some VC offices), and the management of international interdependencies by employees is what defines internalization. The fundamental question raised by innovation outposts is not there; instead, it is when and why firms will find it necessary to expand to a foreign location to access knowledge rather than source it from their home country HQ.

Establishing outposts abroad to be close to sources of knowledge is not a new phenomenon. Families in northern France running highly successful woolen mills at the turn of the twentieth century sent family members to live in Australia and Argentina to handle the firm’s raw wool supply. Setting up foreign R&D subsidiaries to tap local sources of knowledge is not new either. Ronstadt (1978), for example, discusses US MNEs establishing technology units in the mid-1950s and early 1960s to tap foreign technology. For example, Hewlett-Packard opened a research center in Bristol, UK, in 1984, to enlist European R&D organizations in the development of new digital technologies (Asakawa & Lehrer, 2003).

Foreign innovation outposts are only one way by which a firm can access external tacit knowledge. One interesting question is why this particular method is chosen over others. Knowledge can be acquired through a variety of mechanisms. It is often embedded in machinery and in parts and components, and so can be obtained by purchasing them (Mathews, 2002). It can also be accessed through consulting contracts. Specialists can be hired and brought to HQ (Zeng & Williamson, 2007). Joint ventures can be set up in the home country with foreign firms in which local firms can trade market access against the tacit knowledge of their foreign partners (Child & Rodrigues, 2005; Hennart, 2012). When should we expect firms to choose instead to set up an innovation outpost or a R&D facility abroad? I believe when two conditions are met: (1) it is difficult to persuade key foreign personnel who possess the wished-for tacit knowledge to move to the firm’s headquarters (Ronstadt, 1978); and (2) locating the R&D subsidiary abroad makes it possible to take advantage of knowledge spillovers which cause the productivity of the firm’s scientists to be higher than if they were transferred to the firm’s HQ (Hennart, 2012).

Virtual Presence

Brouthers et al. claim that virtual presence entry modes are a third category of new and non-traditional entry types. They write that such modes refer to “foreign entries in which no physical entry is undertaken for customer/user acquisition, although the firm appears to be in the country from the perspective of its buyers and users.” (p. 8). They go on to say that these modes are made possible by “recent advances in digital technologies that enable firms to enter    foreign countries by directly acquiring customers or users and often delivering products (3D printing) or services (downloads) while actually avoiding the need to establish formal foreign-based subsidiary units or export channels” (p. 8). They then conclude that

Virtual presence entry modes differ from traditional modes like exporting, licensing or franchising in two ways. First, virtual presence entry modes do not require the firm to set up operations of any kind (including cooperative modes like export agents, distributors, or export sales subsidiaries) in the foreign market, instead relying on digital technologies to reach customers/users. Second, through these digital channels, firms can leverage their advantages reaching large groups of potential customers/users and retain all the financial benefits of foreign entry (instead of sharing with others) because virtual presence entry modes provide firms with the ability to enter foreign markets and directly control the operation while minimizing investment in the foreign market (p. 8).

This, according to the authors, forces us to “re-conceptualize foreign entry as a process of customer/user acquisition instead of investments or asset acquisition” (p. 8).

Do changes brought about by digitization and internet-based distribution require us to search for new explanations, such as the authors’ EEE framework? It appears to me that it is clear that the goal of foreign market entrants has always been to sell to foreign customers, i.e., to acquire customers for their products.7 Selling to foreign customers through exports from a home base is usually the first-best option, but, in the case of mass-market goods, which typically have local substitutes, investing in a foreign manufacturing plant – a necessary evil since it involves significant costs and risks and requires some knowledge of the foreign market – is often necessary (Hennart, 2014; Hennart, Majocchi, & Hagen, 2021).

Is virtual presence, which involves selling from a home base directly to the foreign consumer, a new mode of foreign market entry? I do not believe it is. Sellers of specialized services often have their employees travel from the home office to visit their foreign customers, hence selling abroad without any  permanent physical presence there. Likewise, producers of global niche products and services who have a small number of internationally dispersed customers may serve them from home, because they prefer not to use intermediaries and because the small number of customers and their geographical dispersion does not justify setting up production or sales operations abroad; indeed, their foreign customers often contact them, obviating the need for foreign agents or distributors (e.g., Chandra, Styles, & Wilkinson, 2009; Hennart et al., 2021; Nordman & Mellen, 2008). Firms which have relatively unique products can also sell ex-factory, which means that it is their customers who handle logistics and custom clearance. Lastly, mail-order firms like La Redoute or Otto have been sending catalogues to foreign customers from their home base for years – for La Redoute, since 1929. In all these cases, firms sell abroad without foreign sales subsidiaries, agents, or distributors on the spot.

So what is different with virtual presence entry modes? Let us think of all the tasks a firm must accomplish to sell abroad. It must: (1) identify potential customers; (2) inform them of the existence and characteristics of its product or service; (3) persuade them to buy; (4) close the deal; (5) provide information, and sometimes training, on how to use the product; (6) process the purchase; (7) in some cases, arrange for credit and collection; and (8) deliver the product to the customer. A firm’s business model determines how complex, expensive, and time-consuming this process is (Hennart, 2014; Hennart et al., 2021).

Digitization and the internet have impacted this in two ways. In all business models, they have the potential to reduce communication costs,  i.e., to lower the costs of performing tasks (1) to (7). The content of some products and services (software, music, books, banking) can also be digitized, hence reducing the cost of delivering the product to the customer (task 8). Let us consider the impact on communication and transportation in turn. The reduction in communication costs brought about by the internet can potentially lower the cost and time needed to sell abroad (Yamin & Sinkovics, 2006). Beyond this, the virtual presence entry mode is indistinguishable from that of the exporters of specialized services or global niche products described above. Producers selling through virtual presence modes must still pay for distribution: they either have to design and run their own websites or subcontract this task to specialist firms, or they can place their products on the websites of retailers like Amazon and pay for access. If they are selling mass-market consumer products for which taste and use conditions vary across countries, they will have to adapt their communication and marketing mix to national differences. Singh, Furrer and Ostinelli (2004), for example, find that customers in Italy, India, the Netherlands, Spain, and Switzerland prefer the culturally adapted websites of B2C companies over standardized ones. Producers selling through virtual modes still have to invest in their own order fulfilment and billing, or subcontract it to other firms. Likewise, they must provide or subcontract logistical services, which sometimes require significant adaptation across countries (Peprah, Giachetti, Larsen, & Rajwani, 2021). Hence, as in the case of non-internet export channels, the virtual presence firms described by Brouthers et al. are likely to have to share some of the financial benefits of foreign entry with other firms, be it web designers, web platforms, fulfilment experts, or logistical service providers.

Digitization of content results in near-zero transportation costs, nonetheless, it requires that recipients have access to the requisite equipment and infrastructure. Besides this, virtual presence modes do not differ from the other forms of exporting we have reviewed above. As with non-digitized goods, the marketing mix of digitized products may have to be adapted on a country-by-country basis (Shaheer & Li, 2020). While sellers of digital products may not have to worry about local delivery if the local internet infrastructure is in place, there is still a need for billing and collection, which is sometimes subcontracted to specialists.8

Thus, while it is undeniable that digitization has reduced the cost of communication and, in the case of digitized products, delivery, it has not changed the nature of the process. Internet-enabled virtual presence firms have to deal with the same issues as those facing exporters with no physical foreign presence. Unless they sell global niche products, they will have to handle or avoid cultural and other types of country-specific differences (e.g., Holm, Decreton, Nell, & Klopf, 2017; Yamin & Sinkovics, 2006), and decide which tasks they should perform in-house and which to subcontract. The virtual presence firms identified by Brouthers et al. do not differ from other exporters with no foreign footprint (for example, mail-order firms) in their ability to “retain all the financial benefits of foreign entry (instead of sharing with others)”. All in all, the case of virtual presence exporters can be explained by the extant IB literature and  so does not call for new theories.

Managed Ecosystems

Managed ecosystems are Brouthers et al.’s fourth type of new foreign entry mode. Brouthers et al. claim that they are “a unique method of international entry facilitated by new digital technologies enabling firms to exploit existing resources in foreign markets through cooperation with local complementors and governments” (p. 13). As examples, they cite Just Eat, Airbnb, Uber, Groupon, TikTok, YouTube, and the Apple AppStore. Brouthers et al. argue that “reliance on complementors is different from traditional alliances and joint ventures as complementors are not hierarchically integrated partners or employees. Instead they are independent contributors satisfying country-specific user needs and enhancing platform value in local markets, even without a direct financial relationship with the platform firm” (p. 9). They add that “Managed ecosystems entry modes… require the firm to establish a presence in the foreign country to comply with local legislation, gain legitimacy, attract local complementors, and create value for customers/users” (p. 9), and this is why they classify them as high-embeddedness entry modes.

Is reliance on complementors unique to managed ecosystems? Does using foreign complementors necessitate having a physical presence in a foreign country? A good starting point to address the first question is to think of the basic function of platforms in terms of known categories. Some platforms, for example, Netflix, are the internet equivalent of brick and mortar retailers and distributors. They make or buy products and services which they sell to final consumers. Some platforms, on the other hand, could be called brokers or agents. They do not earn their keep by buying and reselling products, but are paid to facilitate transactions. For instance, eBay helps those who want to sell surplus items find buyers, Indeed.com helps people to find employers and eHarmony to find compatible life partners, Airbnb puts people with lodgings to rent in touch with those interested in short-term rentals, Uber helps to connect demanders and suppliers of taxi services, while Upwork and Toptal connect freelancers with firms eager to have specific short-term tasks carried out, for instance, programming or translation. These platforms are obviously an improvement over their non-digitalized counterparts, i.e., flea markets, job placement centers, marriage bureaus, tourist offices, radio taxi services, and bulletin boards. All two-sided platforms have a common interface with customers and complementors, but they differ in the control on behavior that they impose through that interface. In some cases, platforms exercise an amount of behavior control on complementors that is similar to that imposed by franchisors on franchisees, so they can be seen as franchisors (Hennart, 2019). Hence, while the internet and associated technologies have undoubtedly reduced communication and search costs, making electronic brokerage more efficient and increasing the geographic spread of participants, two-sided platforms and managed ecosystems are not fundamentally different from their pre-internet predecessors.

The use of complementors embedded in ecosystems also does not differentiate double-sided electronic platforms from other well-known entry modes. Brokers also rely on ecosystems. One can distinguish between three types of ecosystems: horizontal, upstream, and downstream. Horizontal ecosystems are found when the brokerage stage enjoys economies of scale in gathering and sorting information, but there are no economies of scale on either side of the platform, resulting in a few brokers dealing with a large number of buyers and sellers (eBay is an example). Upstream ecosystems arise when the minimum efficient scale at the upstream stage is smaller than at the downstream stage. Then, a large number of upstream firms feed content to a smaller number of distributors. This is the case of Apple’s App store and its Android counterparts where a large number of small entrepreneurial firms sell their apps to mobile telephone manufacturers who market them on their mobile phone platforms. The logic for this configuration is that apps are most efficiently developed by small firms, but distribution to final consumers is subject to large economies of scale. The pre-internet equivalent of this configuration is the biotech segment of the pharmaceutical industry, with the development of new biotech drugs best undertaken by small entrepreneurial firms, while a smaller number of large established pharmaceutical firms handle drug registration and marketing. In contrast, downstream ecosystems are observed when the upstream stage is subject to economies of scale, but the downstream stage is best undertaken by small independent complementors. In the Uber case, for instance, brand building and software development are subject to economies of scale, but the actual provision of taxi services is best accomplished by small firms or even individuals. The pre-internet equivalent of this entry mode is not alliances and joint ventures, as argued by Brouthers et al., but franchising networks. Such networks have been extensively discussed in the IB literature (e.g., Contractor & Kundu, 1998; Fladmoe-Linquist & Jacque, 1995; Hennart, 1982, 2000, 2019). McDonalds, for example, uses a large number of independent local complementors for the local production of its food, Coca-Cola has set up an ecosystem of licensed bottlers, and carmakers a network of franchised dealers.

The customers of McDonalds and Uber expect to be served by a large number of spatially dispersed local complementors. Using employees to perform these tasks would incur higher management costs than using franchisees, because franchisees keep the bulk of their earnings, and hence are more motivated to exert effort than employees who are paid a salary that is weakly related to their individual output. Consequently, in situations where effort is hard to monitor, it is more efficient to use franchisees than employees. On the other hand, if the bulk of customers are one-shot users, franchisees have incentives to free-ride on the brand by reducing quality (Brickley & Dark, 1987). Trademark owners therefore design production methods with built-in quality checks9 and have franchisees sign contracts under which they promise to adopt behaviors that safeguard quality. Hence, outlets will be run by employees of the trademark owner when it is difficult to design equipment and specify rules of behavior that ensure that quality standards are met, and when the cost of employee monitoring is low because the optimal number of outlets is small (Hennart, 1982, 1991). Franchising is optimal in the reverse case. This explains why the local production of fast food is usually franchised, while consulting and accounting firms usually man their offices with employees. Uber’s case is a perfect illustration of this. Uber would find it difficult to supervise drivers if they were company employees, so it is better off contracting them, but then it must curb their tendency to free-ride on quality. One difference between Uber and McDonalds is that technological advances make this task easier for the former than for the latter. While McDonalds must send inspectors to visit its franchisees, and, if contract breach is observed, persuade the courts to let it root out free-riders, Uber can expel drivers from the franchise with a single click. Uber drivers must also use the platform routing software, making it difficult for them to charge exaggerated amounts by taking circuitous routes. Payments are also made directly to Uber, thus curbing further opportunities for dishonesty. In-car cameras can monitor driver behavior. Hence, electronic platforms may achieve lower costs in controlling free-riding than traditional franchisors, and hence may make greater use of these hybrid forms of governance than traditional firms, everything else constant (Cuypers, Hennart, Silverman, & Ertug, 2021). Yet, as the extant franchising literature reminds us, free-riding is inherent in franchising, and Uber has not been able to eliminate all of it: customers have complained about dangerous driving, discourteous drivers, and even assaults (Rosenblatt, 2016; Tomassetti, 2016). In this case, as in that of the other entry modes analyzed by Brouthers et al., the difference is one of degree, not substance.

Brouthers et al. also argue that managed ecosystems require the firm to establish a presence in the foreign country, and hence are high-embeddedness entry modes. Here again, extant theory shows that this is not entirely correct. The early international strategy literature has argued that some global products can be sold worldwide as is, while others require tailoring to local needs, and thus a strategy of local adaptation (a multidomestic strategy) (Porter, 1986). Hennart (2014) and Hennart et al. (2021) contrast a mass-market business model, requiring country-by-country marketing mix adaptation and local production, with a global niche one, characterized by a relatively small number of customers with homogeneous tastes and price-inelastic demand, who can be served by export. While the former follow the Uppsala model of gradual firm internationalization based on experiential knowledge of target countries (Johansen & Vahlne, 1977), firms selling global niche products can scale-up production faster and enter many foreign countries simultaneously. This also applies to electronic platforms: the nature of the product/service sold determines how much local adaptation is needed, and hence their degree of country embeddedness. The success of both Uber and Toptal, a platform that matches expert programmers with firms, depends on the number and quality of complementors, drivers for Uber, and programmers for Toptal. However, while the specific location of complementors is crucial for Uber, since an abundant supply of Uber drivers in New York is of no use in Paris, it is not a concern for Toptal and its customers since they interact online. Consequently, we would expect Uber’s foreign expansion to follow an Uppsala pattern, i.e., a market-by-market strategy with the sequence influenced by psychic distance, while Toptal is more likely to enter many countries simultaneously.10 All in all, managed ecosystems are not so unique, and the extant literature which has analyzed their analogue counterparts has much to say about them.

Brouthers et al.’s Exploration–Exploitation–Embeddedness (EEE) Framework

Brouthers et al. use their four entry modes to build a prescriptive model of entry mode choice organized along two dimensions: (1) whether the entry is an exploration, which, along with Levinthal and March (1993:105) they define as “a pursuit of new knowledge” in contrast to “the use and development of things already known” (p. 12); and (2) the degree of needed embeddedness, which they define as the extent to which successful entry necessitates insertion in host--country networks. They see capital access as a high exploration mode requiring low embeddedness, innovation outposts as high exploration with high embeddedness, virtual presence as high exploitation with low embeddedness, and managed ecosystems as high exploitation modes requiring high embeddedness.

In the following paragraphs I argue that (1) the classification of entry modes as exploitation or exploration is not an improvement over those long since developed in the IB literature; (2) there are important differences in the required level of embeddedness within Brouthers et al.’s four categories of entry modes; and (3) Brouthers et al.’s EEE model lacks the predictive and prescriptive ability of the TCT of foreign entry modes because there is no direct relationship between the need for embeddedness and the use of particular foreign market entry modes.

Exploitation–Exploration

The IB literature has long acknowledged that FDI is driven by many motives. Dunning (1993: 56–57), for example, distinguishes between market-seeking, efficiency-seeking, resource-seeking, and strategic asset/capability-seeking investments. Hennart (1982, 1991) contrasts backward and forward vertical integration, with backward integration motivated by the need to access resources and capabilities, including financial capital, and forward integration by the desire to exploit knowledge and reputation. In contrast, March’s (1991) distinction between exploration and exploitation was intended to deal only with knowledge, and fits awkwardly with foreign entry modes. As Brouthers et al. note, March (1991: 71) defines exploration as “search, variation, risk-taking, experimentation, play, flexibility, discovery, and innovation”, not a particularly good description of the motivation for many types of foreign market entries. Chinese steel firms investing in iron ore mines in Australia, or listing their shares on the New York Stock Exchange, are not looking for experimentation, discovery, risk taking, and play. On the contrary, they want stable and secure access to iron ore and finance.

Embeddedness

Brouthers et al. define embeddedness as “the extent to which the firm is anchored in a particular space to be integrated into, or generate, local networks of economic and social relations… Embeddedness arises as firms rely on intense interactions and close relationships with external stakeholders (such as suppliers, customers, and complementors) in obtaining the use of critical resources” (p. 12). They see the need for embeddedness as exogenously determined, and write on page 13 that it refers to the degree to which “different activities of resource exploitation or exploration may be by their nature more or less locally embedded”.

Although Brouthers et al. argue that managed ecosystems must be highly embedded, and capital access lightly embedded, there are, in my view, significant differences in the degree of required embeddedness within these entry modes. Consider first managed ecosystems. Brouthers et al. hold that they call for high-embeddedness because they “rely on complementors in each country for value creation and often have to engage extensively with local stakeholders in overcoming legal and other government barriers” (p. 13). While it is true that some managed ecosystems (double-sided platforms) use foreign complementors, that does not imply that these complementors are always heavily embedded into their home countries. Uber operates in an industry that is locally regulated, and so must deal with each local context to successfully penetrate foreign markets. In contrast the main embeddedness of the top software writers who are complementors to Toptal is in their profession, not in their country of residence.11

Can one successfully access capital with low embeddedness? Brouthers et al. seem to think so. There are reasons to believe that this is not always the case. I argue above that the market for loanable funds is often subject to high transaction costs, and that this implies that one strategy for entrepreneurs with no track record heading firms that cannot offer good collateral is to try to forge personal relationships with lenders to reduce their risk perceptions. This seems important for equity financing. Bankman and Gilson (1999: 296) write that

It is commonplace that venture capital investors place enormous emphasis on the personal characteristics of the individual in whose innovation they invest. This is hardly surprising. Because of the uncertainty associated with the venture’s success, and because the venture’s value depends on investment in future growth options that, tautologically, are highly dependent on future managerial decisions, the marginal value of a “better” entrepreneur is considerable. Unlike established companies where most income derives from existing businesses whose value is less sensitive to future decisions, a new venture’s business is almost entirely dependent on the entrepreneur’s judgment. Thus the entrepreneur’s talents, independent of the character of the innovation, influence a much larger percentage of the venture’s value than in a more mature business.

As personal contacts are facilitated by geographic proximity, angel investors and venture capitalists “tend to invest in people and businesses that are geographically proximate to their homes” (Coyle & Polsky, 2013: 319). Sorensen and Stuart (2001) found this to be true for the domestic investments of US VCs. Consequently it makes sense for start-ups in need of finance to embed themselves into the places where venture capitalists live so as to develop closer relationships with them. Greenblatt and Keloharju (2001) have also shown that investors are more likely to buy the shares of companies that are located close to where they live. Discussing the reasons why 18% of Israeli start-ups have incorporated abroad, mostly in the US, Shulman (2020) points out that “the main reason why [Israeli start-ups] choose to incorporate outside Israel ...is that venture capital funds and other U.S. investors find it easier to invest in companies registered in-country”. To sum up, there is strong empirical support for the idea that those seeking equity and non-equity capital benefit from locating close to lenders, and hence that access to at least some forms of capital requires high embeddedness.

All in all, Brouthers et al.’s categorization of entry modes with regard to their required level of embeddedness is not persuasive. More fundamentally, their EEE model does not seem to be able to predict which actual foreign market entry mode will be chosen. Given an exogenously-determined need for embeddedness, what entry mode should we expect? Are we likely to see contractual or equity entry modes, and if the latter, wholly-owned subsidiaries or joint ventures? The authors leave this to future research, stating on page 12 that they “do not view the exploitation-exploration continuum or embeddedness as determinant of firms’ entry mode selection decisions; they are instead conceptual dimensions by which we make sense of and categorize non-traditional entry modes”. On page 17, they ask: “Once a firm decides on the right level of embeddedness in a foreign country, how is this task achieved in an efficient and effective manner?” This, of course, is what we expected their model to tell us. In contrast, I have shown that the extant TCT of entry modes has started to provide some answers to these questions.

One reason why Brouthers et al.’s EEE model is unable to predict entry modes is that, as predicted by the bundling model (Hennart, 2009; Verbeke, Coeurderoy, & Matt, 2018; Verbeke & Kano, 2015), the relationship between EEE and entry modes goes through the transactional properties of the resources being bundled. This suggests that Brouthers et al.’s exogenously determined level of embeddedness (how much embeddedness in host-country networks is required for successful entry) need not have any direct relationship to the entry mode chosen. It all depends on whether such embeddedness can be bought on efficient markets. If it can, entrants may not have to make physical investments to embed themselves into the target foreign country, but can contract instead with parties that do.

Take military procurement. Because arm purchases are in large part driven by political factors, it is extremely difficult for sellers of military equipment to sell to foreign countries that have their own weapons industry.12 Successful selling of weapons systems requires a high level of embeddedness into the target   country’s network of politicians and top military decision-makers. Kim (2019), however, shows that foreign arms makers can successfully overcome this liability of outsidership by relying on the social and political capital of professional lobbyists. He shows that, as predicted by the bundling model (Hennart, 2009), the impact of the need for embeddedness on entry modes will hinge on whether such embeddedness can be purchased on efficient markets. If this is the case, the firm will contract for it. If it cannot, it will have to choose another option. In the case of military hardware, it may decide to become an insider by joint venturing with a local firm or by starting to manufacture the weapons in the target market. Given an exogenous need for high embeddedness, the actual entry mode (i.e., whether the firm contracts with a local lobbyist, joint ventures with a local firm, or makes a direct investment in the target market) will thus depend on the transactional characteristics of this complementary resource.

Conclusions

Brouthers et al. claim that the four new entry modes that they highlight require new theories. This may well be, but what is needed first is a careful analysis of what is genuinely new about them, and what is similar to what we already know. This makes it possible to harness the insights of past work. I have followed this approach, and have explored the conceptual similarities between what Brouthers et al. label new and non-traditional entry modes and a number of entry modes which have already been discussed in the literature. This has allowed me to show that TCT has already started to identify their determinants. For instance, shifting registration to foreign capital-rich countries and listing on their stock markets as a way of accessing finance is a strategy that firms used extensively before WWI, and which has again become common. I discuss why this is the case. Some of the same reasons that persuade firms to locate close to sources of capital also push them to establish in innovation hubs such as Silicon Valley, a presence that can range from small Innovation Antennae to full-fledged R&D subsidiaries. I briefly sketch why this mode of knowledge acquisition is chosen over others. I also contend that virtual entry is the way many professional service firms and manufacturers following global niche strategies have been exporting their goods and services, and thus that it cannot be said that this entry mode is unique to internet-based firms. Also, managed ecosystems have similarities with innovation and franchising networks. I show how the extant literature can inform us on their challenges and opportunities. I conclude that the institutional and technological developments that have made the proposed new modes possible have led to changes in kind, not substance. This is why we can usefully apply some IB theories and concepts to start making sense of these advances in doing business. In contrast to that, I find Brouthers et al.’s exploitation–exploration–embeddedness framework unconvincing, because (1) the term ‘exploration’ is not appropriate to describe the motivation for most resource and strategic asset acquisition FDI; (2) most entry modes involve both exploration and exploitation; (3) there is considerable variation in embeddedness within what the authors call capital access and managed ecosystems, making their positioning within the authors’ 2 × 2 matrix arbitrary; and (4) the potential availability of intermediaries on efficient markets means that there is no one-to-one correspondence between the need for embeddedness in foreign country networks and a firm’s entry mode.

In their conclusion, Brouthers et al. write that the literature on their new entry modes “remains largely fragmented and disconnected from the broader entry mode literature. Because of this, current research is missing the vital opportunity to inform entry mode research and create opportunities for the cross-fertilization between non-traditional and traditional entry mode literature.” (p. 22). They argue that “further research is needed to….explain why firms choose these entry modes and what governance arrangements are required for these unique relationships with investors, network partners, complementors and customers/users” (p. 16). As I show, three of Brouthers et al. new entry modes, capital access, innovation outposts, and ecosystems, have already been integrated into a coherent theoretical framework, namely TCT (Hennart, 1991, 2010, 2012). This does not mean that these entry modes are not worth further study. On the contrary, Brouthers et al. make a valuable contribution in bringing them to our attention.

The four types of entry modes analyzed by Brouthers et al. are not the only ones that might repay further study. IB scholars have devoted a lot of attention to the entry strategies of large manufacturing firms selling mass-market goods to final consumers, strategies which have been well described by the traditional Uppsala model (Johanson & Vahlne, 1977). The existence of born global firms (e.g., Knight & Cavusgil, 2004) has shown, however, that this model is not a universal one, and that some firms are using other internationalization strategies. For example, those that follow global niche business strategies are likely to choose very different entry modes. For example, they tend to permanently serve their customers through exports (Hennart, 2014; Hennart et al., 2021; Zucchella, Hagen, Denicolai, & Masucci, 2016). More work is needed on these firms, as well as on firms that sell B2B and on professional service firms. Despite considerable historical evidence on the important role they play in international business (Jones, 1998), little has also been written on intermediaries, such as distributors, trading companies, investment advisors, and underwriters.13 Lastly, I believe that research on foreign market entry modes would benefit from taking a business model approach (Nielsen, Marinova, & Marinov, 2022). That approach is more comprehensive, because it looks at a firm’s complete value network (its customers, suppliers, partners, and distribution channels) (Teece, 2010), and also more granular, because it shows that subtle differences in a firm’s business model can have important consequences for its internationalization (Bohnsack, Ciulli, & Kolk, 2021; Giachetti, 2018; Hennart et al., 2021).

Notes

  1. 1

    Brouthers et al. write of the need “to reconceptualize foreign entry as a process of customer acquisition instead of investment or asset acquisition” (p. 8). Clearly, when firms invest abroad in manufacturing plants or in foreign sales subsidiaries, the ultimate goal is to facilitate sales to customers, often local but sometimes domestic as well.

  2. 2

    On p. 3, Brouthers et al. write that “TCT [transaction cost theory] continues to dominate our thinking to the extent that foreign subsidiaries are often viewed as the preferred organizational structure because of their capacity to protect from misappropriation firm advantages to be exploited (Hennart, 1991).” In fact, in an article that looks at the unsuccessful attempts by Algeria to use turnkey contracts as a substitute for foreign direct investment, I argue that this latter form is the best way to transfer tacit knowledge, but hasten to say that “transaction cost theory does not imply that traditional FDI is always superior to contracts” (Hennart, 1989: 215).

  3. 3

    Dell, for example, has used two separate sales channels in China: it uses employees to sell to large firms and educational institutions, and authorized distributors to sell to smaller and less experienced customers (Verbeke, 2009: 305)

  4. 4

    Some of today’s MNEs, for example, Rio Tinto and British Petroleum, started as free-standing companies (Corley, 1998).

  5. 5

    En.wikipedia.org, accessed on March 16, 2022. In 2020 Guyana Goldfields was acquired by Zijin Mining Group Co, Ltd.

  6. 6

    Mind the Bridge (2019b) reports on European innovation outposts in Israel. There R&D centers dominate, followed by Corporate Innovation labs, Corporate Venture Capital Offices, and Innovation Antennae.

  7. 7

    Sometimes, customers do not pay directly for the product, but provide sellers with personal data which they sell to third parties.

  8. 8

    Kaspersky, a Russian developer of anti-virus software, has contracted Digital River for the billing of some of its European customers.

  9. 9

    For example, fryers that stop working when the oil needs changing.

  10. 10

    Stallkamp and Schotter (2021) also argue that platforms like Toptal are likely to enjoy cross-country externalities, while those like Uber will enjoy within-country ones, and that this will result in different competitive patterns.

  11. 11

    Brouthers et al. concede that this might be the case when they write on p. 20 that “local embeddedness may vary based on the location-boundedness of complementors.” Yet, they place managed ecosystems in the high-embeddedness quadrant of Figure 2.

  12. 12

    In 2006 only 4% of US purchases of military equipment were awarded to foreign firms (Kim, 2019).

  13. 13

    However, see the literature on international trading companies (e.g., Ellis, 2001; Roehl, 1983).