Description
evaluate or estimate the nature, quality, ability, extent, or significance of; "I will have the family jewels appraised by a professional"; "access all the factors when taking a risk"
Case Studies for reassessing (home) regionalization
Abstract Alan Rugman and co-authors argue that globalisation, and with it global strategy, is a myth. This contention rests on a taxonomy of the world's largest firms based on their sales, showing an overwhelming share of home-regional firms. We question the rationales underpinning their classification scheme. When retesting the data using different schema we find that the original results are far from robust, with a significant share of firms attaining bi-regional or global status. Further longitudinal analysis shows that large firms increasingly are extending their sales beyond the home region. Our results defy regionalisation theory in its current form, and we call for refinements through further research. (2008) 39, 184-196. doi:10.1057/palgrave.jibs.8400345
Keywords: regionalisation; globalisation; semi-globalisation; global strategy; regional strategy; triad
INTRODUCTION
Alan Rugman and co-authors' recent empirical research has drawn attention to an important phenomenon: most MNEs' (multi- national enterprises) sales take place largely within their home region (Collinson & Rugman, 2008; Rugman, 2000, 2005; Rugman & Verbeke, 2004a, 2005). Their findings show that firms with a global sales presence (and thus global strategies) are more elusive than the business press and some textbooks would suggest. The authors have rightfully scolded researchers for their breathless commentary on ''globalisation'' and indiscriminate use of the prefix ''global'' (e.g., Rugman & Hodgetts, 2001). The finding of home-regionality has given rise to a new and ambitious research programme. A number of international busi- ness (IB) scholars are dedicating their theoretical efforts to explaining the empirical results of Rugman and co-authors (e.g., Management International Review, 2005/1, Special Issue). Given the contentious nature of the classification system used (see below), it is disconcerting that such theorising has proceeded without Rugman's empirical results having been tested for robustness. When researchers initially explore an issue, the adoption of ''first stab'' benchmarks to segment the phenomenon in question is inevitable and perfectly legitimate. But as we deepen our commit- ment to a particular research programme we must revisit those initial benchmarks to see how sensitive the findings are to the cut- offs imposed. To that end, in this paper we scrutinise the theoretical reasoning underpinning Rugman's classification system, and also explore practical aspects of its application. We then adjust the classification thresholds in line with our own arguments.
185
With relatively simple and justifiable adjustments in classification thresholds, considerable numbers of companies are shown to be active beyond the home region. Utilising longitudinal evidence we also highlight the growing share of bi-regional and, to a lesser extent, global firms between 1991 and 2001. Our findings call into question the pleas to IB researchers to abandon models and research ques- tions that include global strategies and to focus instead on intra-regional MNE activity.
, Global firms have less than 50% of sales in the home region and greater than 20% in each of the other two triad regions. This classification scheme is loosely based on Ohmae's (1985) ideas. Ohmae viewed a global presence (in terms of distribution) as imperative for firm success. His argument was founded on the need to recover the costs of innovation. As such, it was primarily an economies of scale and scope argument, although he also articulated the need for a global presence to avoid competitive blind spots. Only a balanced presence across the triad, then defined as the US, European Union and Japan, would render the MNE a triad power. Rugman adopts this need for balance in his definition of the global firm category. The imposi- tion of a minimum 20% host-region threshold is predicated on the roughly equal size of the three (broad) triad regions in terms of GDP:
The 20% figure is less than the one-third required for an equal triad distribution and so is biased downwards in favour of finding global MNEs. Conceptually it implies the successful deployment of customerend FSAs [firm-specific advantages] in three distinct markets. (Rugman, 2005: 64)1
THE EXISTING LITERATURE
Rugman's empirical research focuses on the world's largest firms as listed in Fortune magazine's Global 500 edition in 2002. He maintains that these firms are the major players in the IB world, collectively accounting for over 90% of the world's stock of foreign direct investment (FDI) and about half the world's trade (Rugman, 2005; but see Dunning, Fujita, & Yakova, 2007: footnote 4). Regional sales data for these companies is collected from annual reports, SEC filings and direct communications. Sales data is utilised because it ''constitutes a true performance measure at the output level'' (Rugman & Verbeke, 2004a: 7). Subsequent replication using company assets as the relevant criterion produced virtually identical results (Collinson & Rugman, 2008; Rugman & Brain, 2003). Utilising a three-region or ''triad'' perspective of global business activity, encompassing North America, Europe and Asia-Pacific, Rugman and co-authors seek to classify the world's largest firms as home-regional, bi-regional, host-regional or global. The classification employs the following (maximum and minimum) thresholds: , Home-regional firms have greater than 50% of sales in the home region. , Bi-regional firms have less than 50% of sales in the home region and greater than 20% in another region of the triad. , Host-regional firms, a special form of bi-regional firm (Rugman, 2005: 11), have greater than 50% of sales in a triad region other than the home region.
The 20% host-region threshold is used to define not only global firms but also the bi-regional firms, which are seen to deploy their FSAs successfully in the home region and one host region. The 50% home-region threshold is predicated on assumptions about the impact of the firm's geo- graphic sales distribution on the locus of both organisational decision-making and competitive advantage:
[W]e assume that a region representing more than 50% of total sales will systematically both shape and constrain most important decisions and actions taken by the MNE. It also implies a concentration of the MNE's customer-end related FSAs in that region. (Rugman, 2005: 64)
Rugman and co-authors then apply the 50% home-region threshold and 20% host-region thresh- old to classify the firms listed in Fortune's Global 500 list. They find an overwhelming preponderance of home-regional firms (see Table 1), a result that is
Table 1
Rugman's results (using 20% host-region threshold and 50% home-region threshold) Home-regional Of which solely domestic Bi-regional Host-regional Global No data available 120 Insufficient data available 15
320 (87.7%)
Source: Rugman (2005: 4).
54 (14.8%)
25 (6.8%)
11 (3.0%)
9 (2.5%)
186
largely invariant to the firms' industry or region of origin. Of the 365 firms for which sales data across the three triad regions was available, 2 a staggering 87.7% are homeregional, providing support for Rugman's claim that ''globalisation is a myth'' (2000: 163). We note in passing that, of the home- regional firms, 54 firms should more accurately be labelled as solely domestic. The remaining categories contain negligible numbers of bi-regional, hostregional and global firms. Since then, a substantial body of theoretical scho- larship has been built up in an attempt to explain this result of overwhelming home-regionality (e.g., Collinson & Rugman, 2008; Rugman, 2005; Rug- man & Verbeke, 2004a, b, 2005, 2007). In essence, Rugman and co-authors explain their finding on the basis of severe limits to the transferability and acceptability of firms' FSAs beyond the home region (Rugman & Verbeke, 2004a: 6). These limits to transferability and acceptability - which apply irrespective of whether FSAs are embodied in exports, transferred to licensees, or transferred to subsidiaries - are captured in the concept of (home) region boundedness of FSAs (Rugman & Verbeke, 2004a: 13). Any FSAs that may have been newly developed in host-region subsidiaries are equally limited in terms of transferability and acceptability, and typically cannot be deployed beyond the region in which they were created (Rugman &Verbeke, 2007: 202). Rugman and Verbeke (2007) elaborate their argu- ment by introducing the related concepts of liability of interregional expansion and liability of inter-regional foreignness. The lack of FSA deployment beyond the home region is explained by ''the additional costs of doing business abroad [which] are often much higher when venturing into other regions of the world than when expand- ing intra-regionally'' (Rugman & Verbeke, 2007: 201). The additional costs of doing business outside the home region stem from the need for expensive complementary FSAs that are necessary to render imperfectly non-location-bound FSAs applicable in host regions. Moreover, while certain (home coun- try) locationbound FSAs can be ''tuned up'' at little extra cost so as to be usable within the entire home region, this option is ruled out for inter-regional expansion. Overall, the cost penalty associated with expansion into host regions ''acts as an entry deterrent for many MNEs'' (Rugman & Verbeke, 2005: 14). In the final analysis, ''f firms have exhausted their growth in their home region of the triad and
still go into other regions, they then face a liability of [interregional] foreignness and other additional risks by this global expansion'' (Rugman, 2005: 1). As few firms are capable and willing to bear these costs and risks, they are destined to compete within the confines of their home region.
ASSESSING RUGMAN'S CLASSIFICATION SCHEME In this section we revisit the original empirical results that gave rise to (home-) regionalisation theory. At the outset, we contend that Rugman's classification system, utilising both a home- and a host-regional threshold, is overdetermined. Speci- fically, it incorporates an implicit hierarchy, with the 50% threshold for home-region sales over- whelming other considerations. If a firm exceeds this criterion then host-region sales become moot. Yet if we are to accept that 20% of sales in a host region is indicative of successful FSA deployment, why is such success negated by a large home-region presence? No strong theoretical grounds are offered for the precedence of the home-region threshold in the classification system, other than an assertion that homeregion sales in excess of 50% reflect and constrain decision priorities. We believe that the locus of decision-making is not a useful criterion in this instance. It will depend heavily on, inter alia, HQ's understanding of the host-regional context and the overall disposition of the firm (centralised vs decentralised), which in turn may be an outcome of the firm's administrative heritage and/or the power balance between HQ and the national/ regional operations (Bartlett & Ghoshal, 1989; Ghoshal & Bartlett, 1990). As such, inferences about decision-making rights based solely on a company's sales profile are troubling. More gener- ally, the introduction of decision-making is spur- ious in a classification system that is fundamentally about establishing firms' success at the output level and the deployment of FSAs in defined geographic regions (Rugman & Verbeke, 2004a: 7). The hierarchical nature of the classification system, with home-region sales taking precedent, means valuable information is ignored. For exam- ple, a company such as Volvo, which has more than 30% of its sales in North America, is classified as home-regional by virtue of having 51.6% of its sales in Europe. Crucially, 10 of the 25 biregionals recognised by Rugman (2005) had home-regional sales between 45 and 50%. A minor shift of their sales distribution in favour of the home region
187
(akin to Volvo) would pull these bi-regional competitors back into the home-regional category despite extensive sales (and putative successful FSA deployment) in a host region. Finnish telecom equipment manufacturer Nokia provides the most dramatic illustration of the problematic outcomes that result from the prece- dence of the home-region threshold. In 2001, Nokia reported 49% of total sales in the European home region, 26% in Asia and 25% in North America, making it one of the handful of global companies in Rugman's classification. By 2002, one year after Rugman's snapshot, Europe accounted for 54% of Nokia's net sales, with the Americas accounting for 22% and AsiaPacific for 24% (Nokia, 2003: 5). As a result, Nokia would have been categorised as a home-regional in 2002 on account of exceeding the 50% home-region threshold. We believe that any classification system that reassigns an erstwhile global company to the opposite end of the spectrum following a minor rebalancing of its sales portfolio (or a mere appreciation of the home-region cur- rency) is necessarily flawed. Further empirical issues spring from the 50% threshold. Home-region sales invariably include domestic (i.e., home country) sales. Yet it is unclear how we should interpret domestic activity in a taxonomy that is fundamentally about internatio- nalisation. Earlier, Rugman and Verbeke (1992) rightly argued that many FSAs may be location- bound (in the sense of country-bound), thus ham- pering the scope for international expansion. In the current classification scheme, success in transfer- ring, adapting and/or building FSAs in host regions is gauged by the firm's sales levels in those regions, but there is no equivalent measurement of the success in deploying FSAs within the home region exclusive of the home country. 3 As a result, a recognised home-regional firm may actually have little or no foreign sales in the home region, a fact also pointed out by Aharoni (2006) and Westney (2006). The evidence proffered by Rugman and co-authors on the 27 British firms among the Fortune Global 500 is a case in point (Rugman & Verbeke, 2007; Yip, Rugman, & Kudina, 2006). Average domestic sales for these companies were 52% while average homeregion sales stood at 64%. In other words, the home-region sales share net of home country sales was a mere 12%. The stylised average British firm would be classified as homeregional under the Rugman scheme, despite only one quarter of its foreign sales occurring in the home region and three quarters outside the home region.
For a broader sample, Osegowitsch and Sammartino (2007) report that the share of foreign sales beyond the home region is approximately two and a half times larger than foreign sales in the home region. In short, the use of a homeregional threshold without also accounting for domestic sales may obscure a firm's FSA reach in the home region beyond the home country. At the same time it may thwart an effective exploration of any inter-regio- nal (vs intraregional) liability of foreignness (LOF). We also have some empirical concerns with the 20% host-region threshold used for the bi-regional and global categories. The magnitude of the thresh- old may be difficult to justify since overall sym- metry across the regions in terms of GDP belies significant inter-industry differences. Respected market researcher Euromonitor reports highly skewed global demand in a variety of industries. For instance, the market size for dairy products - relevant for a Fortune Global 500 competitor such as Groupe Danone varies significantly, with Europe accounting for roughly twice the size of the North American market, and almost three times the Asia- Pacific market (Euromonitor, 2005). The same is true in pharmaceuticals, with 2001 sales estimated at US$172 billion for North America, US$125 billion for Europe and US$93 billion for Asia-Pacific (Euromonitor, 2000). 4 Demand for certain com- modities, on the other hand, is strongly skewed towards Asia. One of the most extreme examples may be the cement industry, with implications for a Global 500 company such as Lafarge. Rapidly growing emerging markets have the highest demand for cement. As a result, Asia accounts for more than two-thirds of global production volume, and China alone accounts for nearly 45% (Cembureau, 2007). World demand in dollar terms is bound to be less skewed, but no doubt still shows the (more devel- oped and slower growing) triad regions of Europe and North America lagging far behind Asia. While other industries show a more balanced distribution of demand across the triad, the above indicates that at least in some industries we should not expect a balanced sales distribution, even if a company's FSAs are of comparable strength in each region. The imposed minimum of 20% of sales in a particular host region may already exceed that region's proportional share of worldwide demand, and in some instances Rugman's benchmark may simply be too onerous. Moreover, the 20% host-region sales minimum, which is taken as evidence of successful FSA deployment in one or two host regions (to attain
188
bi-regional or global status), may be difficult to justify in view of the sampled companies' size. Takenaka was the smallest of the firms listed in the Fortune Global 500, with total sales in 2001 exceed- ing US$10 billion. Given the size of the firms considered, even the achievement of, say, 10% of total sales in a particular host region may be seen as evidence of successful FSAs deployment in that region. A business with sales in excess of US$1 billion cannot be established by catering to a few erratic customers, such as an expatriate community or similar niche, but requires significant firm-specific strengths to overcome any liability of foreignness. For instance, General Motors, a homeregional in Rugman's classification, may only have 15% of its sales in Europe and thus fail to satisfy the 20% hostregion threshold, but that represents 2001 sales in excess of US$25 billion. 5 A final empirical concern regarding the magni- tude of the 20% host-region threshold applies only to the global category. This category requires firms to have less than 50% of their sales in the home region and 20% or more in each of the other two triad regions. Given that the broad triad, on average, accounts for 90% of world demand in an industry, there is virtually no margin for error in a competitor that has close to 50% of sales in the home region. The corresponding arithmetic is o50 þ420 þ420 ¼ 90%. For instance, US-based 3M, a recognised bi-regional company, has 47% of sales in North America, 25% in Europe and 19% in Asia Pacific. As such, 3M's sales into the triad already exceed the stylised 90% of world demand claimed by the triad regions, and attaining global status according to Rugman's criteria may prove challenging. Since many bi-regionals as recognised by Rugman have home-region sales just below the 50% mark, achieving global status may be exceed- ingly difficult for them. In light of the concerns raised above, we argue that alternative classification regimes should be consid- ered to determine the status of firms. Most impor- tantly, the overdetermination of the current classification scheme must be remedied. For a simple alternative classification system, based on the geo- graphic deployment of FSAs, the use of host-regional sales thresholds would suffice. Firms that attain significant sales in one or two host regions are given bi-regional and global status, respectively. Firms that do not have significant host-regional sales are, by default, assigned to the home-regional category. 6 For such an alternative classification system, the definition of what constitutes ''significant''
host-region sales - so as to speak of successful FSA deployment in these regions - warrants further discussion. While 20% host-regional sales may seem a reasonable benchmark, there are arguments to support somewhat lower benchmarks, given both the heterogeneity in product market sizes and the huge size of the MNEs under examination (see earlier). There may also be merit in lower benchmark levels where the researcher is interested in the internationalisation process and the devel- opment of hostregional FSAs over time.
SENSITIVITY ANALYSIS OF THE DATA
In light of our theoretical and empirical concerns, and the alternative classification scheme sketched above, we engaged in sensitivity testing designed to ascertain the overall robustness of Rugman's results. Data concerning the 365 companies in the original sample were obtained directly from the Appendix in Rugman's (2005) book. Based chiefly on our arguments concerning the uneven division of demand across regions in some industries, we initially lowered the 20% host-region threshold to first 15% and then 10%, while retaining the 50% home-region threshold. As Table 2 shows, the bi-regional category weakens somewhat as 6-14 bi-regional firms shift to global, but, overall, the classification results are largely unchanged from Rugman's (Table 1). Given our views regarding the overdetermination of Rugman's classification system, we then abol- ished the 50% home-region threshold while retain- ing the original 20% host-region threshold. As a result, a company such as the above-mentioned Volvo, with 52% of its sales in Europe and 30% in North America, would now be considered bi- regional rather than home-regional. Similarly, Nokia would still be considered global in 2002 despite its European sales having nudged beyond the 50% threshold. Using this new system of classification, results vary markedly from Rugman's (Table 1). As can be seen in Table 3, 53 firms shift
Table 2 Results using 15 and 10% host-region thresholds and 50% home-region threshold Homeregional 15% threshold 10% threshold 320 (87.7%) 320 (87.7%) Bi-regional Host-regional Global
19 (5.2%) 11 (3.0%)
11 (3.0%) 11 (3.0%)
15 (4.1%) 24 (6.6%)
189
Table 3 Results using 20% host-region threshold, no homeregion threshold Home-regional 267 (73.2%) Bi-regional 87 (23.9%) Host-regional 2 (0.5%) Global 9 (2.5%)
produces contrasting findings, with global and bi- regional companies combined now accounting for a sizeable portion of the sample. We conclude that Rugman's classification is less robust than is desir- able. The numbers of ''newly found'' bi-regionals and, to a lesser extent, global companies clearly suggest that the home region is not the insurmoun- table barrier portrayed by Rugman and co-authors.
Table 4 Results using 15 and 10% host-region thresholds, no home-region threshold
LONGITUDINAL ANALYSIS
Homeregional 15% threshold 10% threshold 232 (63.6%) 200 (54.8%) Biregional 114 (31.2%) 122 (33.4%) Hostregional 1 (0.3%) 1 (0.3%) Global
19 (5.2%) 42 (11.5%)
from home-regional to bi-regional. A further eight firms shift from host-regional to bi-regional,7 while the global category remains unchanged. This casts some doubts on the robustness of Rugman's find- ings. In particular, the bi-regional category, which now accounts for a quarter of all firms, can no longer be dismissed as insubstantial. As a final step in our sensitivity analysis, we again dropped the 50% home-region threshold and utilised the less onerous 15 and 10% host-region thresholds. As is shown in Table 4, a strikingly different picture emerges once more. Global com- panies remain relatively few in number, although their share increases by 100-350% compared with Rugman's (Table 1). 8 More importantly, a very considerable proportion of the companies, around onethird in both instances, are now identified as bi-regionals. At this stage we remind the reader that the home-regional category in its present form actually contains 54 solely domestic firms. If these were set aside in a separate domestic category, global and bi-regional firms combined (133) would not be far behind the genuine home-regional MNEs (178) when adopting the 15% host-regional threshold. If the 10% threshold were to be adopted, their combined numbers (164) already exceed those of the genuine home-regional MNEs (146). These results confirm that it is the 50% home- region threshold, criticised earlier on theoretical and empirical grounds, that overwhelmingly drives Rugman's classification. Removing the overdeter- mination introduced by the home-region threshold
Rugman has bolstered his claims by arguing that home-regionals have dominated rankings of the world's largest firms for the past two decades. The situation is presented as unlikely to change in the foreseeable future. Efforts at intra-regional integra- tion and concomitant interregional protectionism (trade wars between the triad regions, barriers to regional entry, etc.) are seen to generate an enduring inter-regional liability of foreignness (Rugman & Verbeke, 2007). Rugman concludes that ''the end result is the persistence of MNEs that will continue to earn 80% or more of their income in their home triad region'' (2005: 63). To his chagrin,
some colleagues still seem to question these data. There must be a trend towards globalization over time they say. Well no - actually the aggregate [trade and FDI] data of chapter 11 strongly suggests the opposite; y Naturally these aggregate data trends are likely to be mirrored in the firm-level data. (Rugman, 2005: 239-240)
Scant longitudinal firm-level data is offered in support of these statements, however, and Rugman has been criticised for his lack of appreciation of the dynamics in firm internationalisation (Aharoni, 2006).9 Using company data available via Datastream we tested Rugman's claims for his original sample, which uses 2001 data. We sampled only those firms for which regional sales data was available for both 1991 and 2001. Not surprisingly, for a significant number of companies this was not the case, chiefly reflecting less stringent reporting requirements in the early 1990s, and the impact of mergers and acquisitions. In total we identified complete data sets for 198 firms, and the following results must be considered with the restricted sample in mind. 10 As shown in Table 5, we identify an unambiguous trend towards greater internationalisation beyond the home region between 1991 and 2001. Using Rugman's thresholds, the (small) numbers of bi- regional and global firms in 2001 represent an almost threefold increase on their numbers a decade
190
Table 5
Longitudinal classification (1991 and 2001) Year Homeregional 185 (93.4%) 166 (83.8%) 156 (78.8%) 133 (67.2%) 146 (73.8%) 115 (58.1%) 135 (68.2%) 92 (46.5%) Biregional 8 (4.0%) 20 (10.1%) 40 (20.2%) 58 (29.3%) 45 (22.7%) 68 (34.3%) 43 (21.7%) 74 (37.4%) Hostregional 3 (1.5%) 6 (3.0%) 0 (0.0%) 1 (0.5%) 0 (0.0%) 1 (0.5%) 0 (0.0%) 1 (0.5%) Global
Rugman's scheme (20% host-region threshold and 50% home-region threshold)
1991
2 (1.0%) 6 (3.0%) 1 (1.0%) 6 (3.0%) 7 (3.5%) 13 (6.6%) 20 (10.1%) 31 (15.7%)
2001
Our scheme I (20% host-region threshold, no home-region threshold)
1991 2001
Our scheme II (15% host-region threshold, no home-region threshold)
1991 2001
Our scheme III (10% host-region threshold, no home-region threshold)
1991 2001
Note: Of the 198 firms sampled, 53 (26.8%) were solely domestic in 1991. By 2001, only 22 firms (11.1%) were solely domestic.
earlier. Utilising our revised thresholds, the growth is not quite so stark, but still impressive off an already significant base. These results indicate that the situation is more fluid than realised by Rugman and co-authors. During the period of investigation, many former home-regional companies in our subsample were evidently able to successfully deploy their FSAs abroad.11
DISCUSSION
When researchers explore a new topic, the adoption of preliminary cut-offs to segment the phe- nomenon in question is often inevitable. It is good research practice, however, to revisit those initial thresholds at a later stage, to see how sensitive the findings are to the cut-offs originally imposed. To that end, in this paper we scrutinised the reasoning underpinning Rugman's chosen thresholds. We then manipulated the thresholds in line with our own arguments, and checked the robustness of the results obtained. Naturally, manipulating benchmarks in any (exhaustive and mutually exclusive) categorisation scheme will favour certain categories at the expense of others. We are conscious that in the context of this research endeavour there is no one host-region threshold that can claim to be the ''true'' threshold; an element of judgement pervades all the choices,
including ours.12 But we believe we have advanced empirical arguments in support of our judgements and, more generally, our call to relax Rugman's (20%) benchmark. With respect to the greater-than-50% home- region sales threshold we are far less open-minded. We believe it should be discarded for its incon- sistency with the avowed purpose of classifying firms based on the successful deployment of FSAs. Recall that the (20%) host-region threshold was adopted to indicate extensive customer-end FSAs in one or two host regions. The subsequent denial of these FSAs (and thus the denial of bi-regional or global status) on the grounds that home country sales greater than 50% of total sales imply ''con- centration of the MNE's customer-end related FSAs in that region'' (Rugman, 2005: 64) is untenable and suggests overdetermination. Our sensitivity analysis suggests that the 50% homeregion threshold is the primary driver of Rugman's findings. Once this threshold is aban- doned, and host country minimum thresholds relaxed to a less onerous 15% or 10% of sales, a significantly different picture emerges. Among the world's largest companies many are homeregional, but a substantial portion of them are bi-regional and a not insignificant number are global players.
Reassessing (home-)regionalisation
Thomas Osegowitsch and Andre Sammartino´
191
The lack of robustness in Rugman's empirical findings has serious theoretical implications. At present, Rugman argues that ''f firms have exhausted their growth in their home region of the triad and still go into other regions, they then face a LOF and other additional risks by this global expansion'' (2005: 1). Firms will stay at home because a home-regional presence
Potential reach of FSA Realised reach of FSA Global
Bi-regional Home-regional Sub-regional
Home Country
allows the MNE to gain all the potential economies of scale and scope and/or differentiation advantages within its home triad market. The additional scale, scope or differ- entiation advantages to be gained by going global are not sufficient to compensate for the enhanced risks. (Rugman & Brain, 2003: 7)
Firm A (1991)
Firm C (2001)
Firm B (1991) Firm B (2001)
Firm C (1991)
Firm A (2001)
These statements concerning the thrust of regionalisation theory in its current form are difficult to reconcile with the significant number of companies that have successfully pushed beyond the homeregion boundary. Results of our longitudinal analysis also cast doubt on Rugman's pronouncements concerning change over time as well as the lack of benefits from expanding outside the home region. Contrary to his claim that ''sales data for earlier periods [i.e., prior to 2001] are highly unlikely to provide much new information or cause us to reclassify more than a handful of the [y] firms'' (Rugman, 2005: 239), our analysis indicates strong growth in the bi- regional and global categories and a corresponding decline in the home-regional category for the period 19912001. We are left with the responsibility to explain the results our empirical investigation uncovered. Based on the observed (international) sales pattern at a given point in time, Rugman and co-authors draw conclusions regarding the deployability or reach of FSAs. 13 We agree with their proposition that sales are the decisive indicator of FSA reach: (international) sales reflect (international) customers' judgement about the attractiveness of the firm's goods or services and, ultimately, the strength and relevance of the underlying FSAs. If we accept this proposition we must also accept that (international) sales over time reveal the evolution of FSA reach. Based on our empirical evidence we must then conclude that FSA reach has beenexpanding from 1991 to 2001, allowing more firms to deploy their FSAs beyond the home region. To explain the observed trend, we deem it instructive, albeit simplistic, to conceptualise FSA reach as a series of concentric circles, as indicated in Figure 1. This reflects a view that, to paraphrase
Figure 1
Conceptualising FSA reach over time.
Westney (2006: 447), differences between homeregionalisation and bi-regionalisation and globali- sation may be merely quantitative rather than qualitative. At any one point in time, certain FSAs may be inherently national, as they are either irrelevant beyond the domestic context, or the cost of transfer/adaptation would outweigh the benefits obtainable. Other FSAs may be home-regional in reach, while still others may be bi-regional or even global. In addition to these established levels of analysis we would add the sub-regional level, as introduced by Enright (2005: 87). Sub-regions are groups of countries within a region that have much more in common with one another than with other countries in the region. These commonalities may reside along any of the four primary dimensions - cultural, institutional, geographic and/or economic (Ghemawat, 2001). This view of FSA reach as a series of concentric circles can be further refined to accommodate the results of our longitudinal analysis. One explanation why firms are increasingly spreading beyond the home region is predicated on demand-side arguments. Differences across countries and across regions - as conceived in terms of cultural, institutional and economic ''distance'' - may be diminishing over time. While it is clear that the world will remain in a state of ''semi-globalisation'' (Ghemawat, 2003) into the foreseeable future, it appears that at least along certain dimensions ''distance'' has shrunk, through factors such as new technologies, trade
192
and FDI agreements (bilateral and multilateral), and the emerging global mass media (Economist, 1997). In short, national as well as regional econo- mies have become more interdependent and, in the process, more compatible. Consequently, certain FSAs previously restricted to, say, the home country (Figure 1: Firm A in 1991), may be applicable to the entire home region 10 years later (Firm A in 2001). Other FSAs may have retained their level of reach as the dimension(s) relevant to the FSA failed to converge over time (Firm B in 1991 and 2001). A complementary explanation for the findings of our longitudinal analysis emphasises supply-side arguments. To that end we initially draw a distinc- tion between the potential and realised reach of FSAs and contend that FSA deployment is plagued by significant lag effects. For a variety of reasons it takes time to adjust the firm's geographic scope to the potential reach of its FSAs (especially since potential reach may be a moving target, as indicated above). These reasons include Penrosian constraints, 14 delays due to the challenges in assembling the required co-specialised assets (Teece, 1986) or, for companies in the earliest stages of internationalisation, ample opportunities for aggressive growth within the home region. 15 These lag effects may also help explain why, at present, home-regional sales account for a very large portion of total sales, and why Rugman (2005) and this investigation (still) find the home-regional category - however defined - the dominant cate- gory. To illustrate our point, Firm C in Figure 1 possessed FSAs with global reach in 1991, but had failed to realise their full potential. By 2001 the actual deployment of its FSAs was still below the potential, although by going bi-regional the firm had advanced beyond the situation a decade earlier. The longitudinal trends charted for the period 1991-2001 suggest that firms do eventually close the gap to the potential reach of their FSAs. Correspondingly, a significant number of erstwhile home-regional firms had already migrated into the bi-regional category and, to a lesser extent, the global category by 2001. If the trends presented in this paper persist, the home-regional category will be further weakened in years to come as more large companies assume bi-regional and possibly global status.
CONCLUSION
Rugman and co-authors' empirical research has drawn attention to a previously under-appreciated phenomenon: in IB, many companies operate
mostly within their home region. Their research was timely in that it helped pour cold water on the globalisation hyperbole. It has also given rise to the beginnings of a regionalisation theory intent on explaining why firms are compelled to restrict theiractivities to the home region. Our own results suggest that the case for homeregionalisation is overstated. In turn, regionalisation theory in its current form is straining against the evidence assembled in this paper. In the concluding paragraphs we offer a number of pointers to how this research stream may be advanced. Rugman and co-authors effectively dismiss bi-regional and global strategies, asserting that there are severe limits on the transferability and accept- ability of firms' existing FSAs beyond the home region. Moreover, ''the administrative heritage of most MNEs, undoubtedly conducive to home region market success, may well constitute an administra- tive rigidity when attempting to penetrate host regions'' (Rugman, 2005: 197). As a result, managers are advised to ''think regional [y] and forget global'' (Rugman & Hodgetts, 2001: 341). Yet all these claims rest squarely on Rugman's evidence that global and bi-regional firms are the exception in the world of big business. Our revision of the data - uncovering significant numbers of biregional firms [which ''may be regarded as partly global'' (Rugman, 2005: 4)] and global firms, and their share having grown markedly from a decade earlier - dilutes the case considerably. Regionalisation theory and its concepts, such as the interregional LOF, are fundamentally depen- dent on intraregional integration and enduring inter-regional differences. While intra-regional integration policies reduce the costs associated with deploying FSAs in the home region, in host regions the ''requirement for high, regionspecific 'linking' investments acts as an entry deterrent for many MNEs'' (Rugman & Verbeke, 2005: 14). While we confirm that sales of a large share of firms are concentrated mostly in the home region, we also identify a considerable and growing share of companies that manage to deploy their FSAs in host regions. This suggests that any interregional LOF is too small to generally deter firms from spreading beyond the home region. 16 In turn, this would suggest that inter-regional differences across the triad may be weak (and diminishing) and/or intra-regional integration less advanced than is commonly assumed.17 Taken to their logical endpoints, such conclu- sions ultimately lead to questions about the merit
193
of the (triad) region as a unit of analysis. The identi- fication of regional sales outcomes and region- based trends - as performed by Rugman and ourselves - may mask significant variation within regions. In fact, the extant literature on the inter- nationalisation of the firm (e.g., Benito & Gripsrud, 1992; Johanson & Wiedersheim-Paul, 1975) em- phasises the individual country, thus implying that firms are predominantly making country-by-coun- try choices in the deployment of their FSAs. Equally, notions of LOF (Zaheer, 1995) have over- whelmingly been theorised and explored at the country rather than the regional level. 18 Aggregat- ing country-based choices to the regional level, and inferring that firms are making regionby-region choices (and that, say, a LOF exists at a regional level), may hamper our understanding of actual firm strategies. In short, the usefulness of the (triad) regional level of analysis in this context is yet to be fully confirmed. One way to establish the merit of the (triad) regional level of analysis, and of the region-based concepts that make up regionalisation theory, would be the concurrent exploration of country- and region-level data. At present, no large-scale database offers the required detailed information on MNEs, but more manageable (industrybased) studies may allow such investigations. Ultimately, such research would also allow us to discern the level of differences within and across regions as perceived by MNE managers. Once the usefulness of the region as a level of analysis is established, attention can turn to refi- ning the corresponding theory. Initially, it would seem that the current, strong form of regionalisa- tion theory must be relaxed to accommodate a scenario where bi-regional and global firms are no longer exceptions. Further elaboration of the theory may be facilitated by a particular focus on biregional firms. Given their significant numbers and their inbetween status (in between home- regional and global), studying these firms would seem to hold great promise for a better under- standing of an LOF at the regional level. For instance, there are indications that any inter- regional LOF is far from uniform. Taking the 15% host-regional threshold (Table 4) we identify 114 bi-regional firms, 97 of which are North American or European in origin. Virtually all of these 97 firms are bi-regional by virtue of strong sales in each other's region rather than in the Asia-Pacific. Only four firms (Boeing, Ericsson, Manulife Financial and Motorola) have successfully deployed their
FSAs in the Asia-Pacific rather than in the other ''Western'' host region. This would seem to suggest that any interregional LOF that might exist is highly asymmetric. Regionalisation theory is also in need of more longitudinal studies, to establish a clearer picture of the dynamics in (intra- and inter-regional) expan- sion. At the most general level, future research should also strive for more fine-grained empirical models - by taking up some of the arguments raised earlier, using continuous instead of categorical measures, and adopting other innovative measure- ment approaches (e.g., Asmussen, 2006) - to get a better grip on regionalisation phenomena. Alan Rugman and co-authors have opened up an exciting new research stream. The data analysis presented herein casts doubts on regionalisation theory in its current state, and hints at some of the questions that remain to be addressed. Clearly, further empirical work and theorybuilding is warranted.
ACKNOWLEDGEMENTS
We gratefully acknowledge the assistance of Eric Quintane in compiling the data used in this paper. We also thank Editor-in-Chief Professor Arie Y Lewin and two anonymous JIBS reviewers for their efforts. NOTES Strictly speaking, if the broad triad accounts for roughly 90% of world output, a perfectly balanced global presence would require 30% in each triad market rather than 33.33%. 2 Accounting conventions give firms considerable leeway in how to report the geographic breakdown of their sales, and firms rarely follow the exact triad scheme. As a result, this entire research stream is affected by a significant level of ''noise'' in the data used. See Rugman (2005: Chapter 2) for a compre- hensive disclosure and a discussion of the inevitable limitations that arise from relying on sales data as reported in company accounts. 3 Setting a threshold for the share of home region ex domestic sales is an obvious solution, but complicated by the fact that the ''rest of home region'' varies with the company's country of origin. For instance, the size of the ''rest of home region'' differs markedly for a US firm compared with a Canadian firm. Approaches that account for the relative GDP of the home country (e.g., Asmussen, 2006) seem to hold great promise in this regard. 4 Not only sales but also profits may be highly skewed across triad regions. For instance, it is well
1
194
known that most big pharmaceutical companies, many of which feature in the Fortune Global 500, make about half of their profits in the US alone (Economist, 2007). 5 On the suggestion of a reviewer we set out to count the number of firms with host-region sales greater than US$1 billion. Sufficient data to perform this analysis was available for only 219 firms in Rugman's 365-strong sample. Of those 219 firms, 97 had sales greater than US$1 billion in both host regions, while 35 had attained such levels in one host region. The corresponding figures were 58 and 61 firms when the bar was raised to US$2 billion. 6 Clearly, such a simple scheme does not address our concerns regarding the inclusion of domestic sales among home-region sales. More complex classifica- tion systems may explicitly set thresholds for home- region sales ex domestic sales (weighted for the size of the home country within the home region). 7 Once the 50% threshold for home-region membership is removed, it becomes difficult to manage and rationalise the 50% benchmark for the special hostregional category. While it is a far from ideal categorisation, host-regional status now belongs only to those firms that achieve a market share greater than 20% in a host region and less than 20% in the home region. Most of the firms in the host-regional category tend to collapse into the bi-regional (and occasionally the global) category in subsequent tables. 8 Data insufficiencies bias this classification against finding global competitors, since Rugman's data contains numerous firms with no available information for a particular region (mostly Asia) and/or reported sales that make up significantly less than the 90% of total sales expected for the entire triad. For example, in 2001, pharmaceutical giant Glaxo reported European sales and US sales at 28.6 and 49.2% of total sales, respectively, rendering it a bi-regional in Rugman's classification scheme. Where the remaining 22.2% of sales took place was not disclosed by the company. It is likely that Glaxo's Asian sales during that year exceeded the 10% threshold, which would make the firm a global player according to one of our classifica- tion regimes. 9 Rugman (2005) does provide rudimentary tempor- al sensitivity analysis of his classification. For thatpurpose, he utilises updated 2002 figures for 60 firms in his original 2001 sample. Not surprisingly, a single year after the original analysis, ''only two of these sixty firms were recategorised'' (2005: 32). This hardly constitutes a strong test of robustness over time. 10 We are cautiously confident about the representativeness of our sample. w2-tests of Rugman's original sample and our own, based on firms' region of origin
and their distribution across the four categories in 2001 (using Rugman's classification scheme), were nonsignificant at the 1% level. 11 See Osegowitsch and Sammartino (2007) for a more detailed investigation of these trends, but see also Rugman and Oh (2007). 12 The definition of thresholds to distinguish global from bi-regional and home-regional firms is likely to remain a contentious issue unless the community of IB researchers can reach consensus as to what constitutes ''significant'' sales that warrant a particular categorisa- tion. We note in passing that there is also disagree- ment whether sales should be the relevant criterion (see Aharoni, 2006; Rugman & Verbeke, 2004a: 7). Clearly, such a situation is far from unique. After all, IB scholars are still debating exactly what constitutes a multinational firm (Aharoni, 1971; Ghoshal & Westney, 1993). 13 The notion of FSA country or region boundedness implies an exogenous phenomenon, determined entirely by environmental forces. We view the boundaries of effective FSA leverage as also governed by firmspecific aspects and as partly controlled by management. As a result we prefer the term ''FSA reach''. 14 The ''Penrose effect'' simply suggests that there are strict limits to a firm's growth rate due to dynamic adjustment costs that are incurred by firms trying to grow their productive resources. Penrose (1959) focused on one major source of dynamic adjustment costs, namely those attributable to the expansion of management resources. She insists that a firm's expansion requires the services of experienced internal managers. Hiring new managers is an inadequate solution, since only seasoned internal managers can undertake the coordination task inherent in firm expansion. As a result, the rate of growth is limited by the rate at which the firm can develop internal managers. Not only domestic expansion but also international expansion is subject to the Penrose effect (Tan & Mahoney, 2005). 15 We agree that firms typically begin their internationalisation in the home region. The home region is likely to contain some of the most similar (i.e., ''least distant'') countries. FSA deployment in these initial destinations is facilitated by institutional similarities, cultural affiliations, lower transportation costs, etc. Beyond the initial round of internationalisation, market choices are likely to be much less focused on the home region, as similar countries also exist in other regions. 16 Based on evidence presented elsewhere (Osegowitsch & Sammartino, 2007), we also dismiss the possibility that firms venture into host regions only once they have exhausted growth opportunities in the
195
home region, as a kind of ''last resort'' option to keep growing. Instead, it seems that many firms are simultaneously deploying their FSAs within and beyond the home region. 17 See also Aharoni (2006) and Westney (2006) on this point. 18 Hejazi (2007) is, to our knowledge, the only study trying to model a regional LOF in terms of MNE sales
(as well as asset, income and employment) distribu- tion. His methodology explicitly tries to disentangle country and regional effects. While the results must be treated as strictly preliminary, they do suggest that ''distance'' (as well as size differentials) between individual countries explains the sales distribution of US MNEs; regional (dummy) variables yield no additional explanatory power.
REFERENCES
Aharoni, Y. 1971. On the definition of a multinational corporation. Quarterly Review of Economics and Business, 11(3): 2737. Aharoni, Y. 2006. Book review: Alan M. Rugman, The regional multinationals. MNEs and global strategic management. International Business Review, 15(4): 439-446. Asmussen, C. G. 2006. Local, regional or global? Quantifying MNC geographic scope. SMG Working Paper 14/2006, Centre for Strategic Management and Globalization, Copenhagen Business School. Bartlett, C. A., & Ghoshal, S. 1989. Managing across borders: The transnational solution. Boston: Harvard Business School Press. Benito, G. R. J., & Gripsrud, G. 1992. The expansion of foreign direct investments: Discrete location choices or a cultural learning process? , 23(3): 461-476. Cembureau 2007. 2005 World cement production by region.http://www.cembureau.be/Documents/KeyFacts/STATISTICS/World Cement Production by Region.pdf. Accessed 24 March 2007. Collinson, S. C. & Rugman, A. M. 2008. The regional nature of Japanese multinational business. , 39(2): 215-230. Dunning, J. H., Fujita, M., & Yakova, N. 2007. Some macro-data on the regionalisation/globalisation debate: A comment on the Rugman/Verbeke analysis. , 38(1): 177-199. Economist 1997. One world?. 18 October: 79-81. Economist 2007. Billion dollar pills. 25 January: 61-64. Enright, M. 2005. Regional management centers in the AsiaPacific. Management International Review, 45(Special Issue 1): 5982. Euromonitor 2000. Pharmaceuticals: A world survey. World market overview. Global Report, 2 June, Global Market Information Database.https://www.euromonitor.com/. Accessed 3 January 2007. Euromonitor 2005. The world market for dairy products. Global Report, 7 June, Global Market Information Database. https:// www.euromonitor.com/. Accessed 3 January 2007. Ghemawat, P. 2001. Distance still matters. Harvard Business Review, 79(8): 137-147. Ghemawat, P. 2003. Semiglobalization and international business strategy. , 34(2): 138-152. Ghoshal, S., & Bartlett, C. A. 1990. The multinational corporation as an interorganizational network. Academy of Management Review, 15(4): 603-625. Ghoshal, S., & Westney, D. E. 1993. Introduction and overview. In S. Ghoshal and D. E. Westney (eds), Organization theory and the multinational corporation: 1-23. New York: St Martin's Press. Hejazi, W. 2007. Reconsidering the concentration of US MNE activity: Is it global, regional or national? Management International Review, 47(1): 5-27. Johanson, J., & Wiedersheim-Paul, F. 1975. The internationalization of the firm: Four Swedish cases. Journal of Management Studies, 12(3): 305-322. Nokia 2003. Nokia in 2002.http://www.nokia.com/A4126499. Accessed 28 March 2007. Ohmae, K. 1985. Triad power: The coming shape of global competition. New York: The Free Press. Osegowitsch, T., & Sammartino, A. 2007. Exploring trends in regionalisation. In A. M. Rugman (ed.), Research in global strategic management, vol. 13, Regional aspects of multinationality and performance. Amsterdam: Elsevier: 45-64. Penrose, E. T. 1959. The theory of the growth of the firm. Oxford: Oxford University Press. Rugman, A. M. 2000. The end of globalization. London: Random House. Rugman, A. M. 2005. The regional multinationals: MNEs and ''global'' strategic management. Cambridge: Cambridge University Press. Rugman, A. M., & Brain, C. 2003. Multinational enterprises are regional, not global. Multinational Business Review, 11(1): 3-12. Rugman, A. M., & Hodgetts, R. M. 2001. The end of global strategy. European Management Journal, 19(4): 333-343. Rugman, A. M., & Oh, C. H. 2007. Multinationality and regional performance, 2001-2005. In A. M. Rugman (ed.), Research in global strategic management, vol. 13, Regional aspects of multinationality and performance. Amsterdam: Elsevier: 31-43. Rugman, A. M., & Verbeke, A. 1992. A note on the transnational solution and the transaction cost theory of multinational strategic management. , 23(4): 761-771. Rugman, A. M., & Verbeke, A. 2004a. A perspective on regional and global strategies of multinational enterprises. Journal of International Business Studies, 35(1): 3-18. Rugman, A. M., & Verbeke, A. 2004b. Regional transnationals and triad strategy. Transnational Corporations, 13(3): 1-19. Rugman, A. M., & Verbeke, A. 2005. Towards a theory of regional multinationals: A transaction cost economics app- roach. Management International Review, 45(Special Issue 1): 5-17. Rugman, A. M., & Verbeke, A. 2007. Liabilities of regional foreignness and the use of firm-level versus country-level data: A response to Dunning et al. , 38(1): 200-205. Tan, D., & Mahoney, J. T. 2005. Examining the Penrose effect in an international business context: The dynamics of Japanese firm growth in US industries. Managerial and Decision Economics, 26(2): 113-127. Teece, D. J. 1986. Profiting from technological innovation. Research Policy, 15(6): 285-305. Westney, E. 2006. Book review: The regional multinationals: MNEs and ''global'' strategic management. Journal of International Business Studies, 37(3): 445-449.
196
Yip, G., Rugman, A. M., & Kudina, A. 2006. International success of British companies. Long Range Planning, 39(3): 241-264. Zaheer, S. 1995. Overcoming the liability of foreignness. Academy of Management Journal, 38(2): 341-363.
doc_743645027.docx
evaluate or estimate the nature, quality, ability, extent, or significance of; "I will have the family jewels appraised by a professional"; "access all the factors when taking a risk"
Case Studies for reassessing (home) regionalization
Abstract Alan Rugman and co-authors argue that globalisation, and with it global strategy, is a myth. This contention rests on a taxonomy of the world's largest firms based on their sales, showing an overwhelming share of home-regional firms. We question the rationales underpinning their classification scheme. When retesting the data using different schema we find that the original results are far from robust, with a significant share of firms attaining bi-regional or global status. Further longitudinal analysis shows that large firms increasingly are extending their sales beyond the home region. Our results defy regionalisation theory in its current form, and we call for refinements through further research. (2008) 39, 184-196. doi:10.1057/palgrave.jibs.8400345
Keywords: regionalisation; globalisation; semi-globalisation; global strategy; regional strategy; triad
INTRODUCTION
Alan Rugman and co-authors' recent empirical research has drawn attention to an important phenomenon: most MNEs' (multi- national enterprises) sales take place largely within their home region (Collinson & Rugman, 2008; Rugman, 2000, 2005; Rugman & Verbeke, 2004a, 2005). Their findings show that firms with a global sales presence (and thus global strategies) are more elusive than the business press and some textbooks would suggest. The authors have rightfully scolded researchers for their breathless commentary on ''globalisation'' and indiscriminate use of the prefix ''global'' (e.g., Rugman & Hodgetts, 2001). The finding of home-regionality has given rise to a new and ambitious research programme. A number of international busi- ness (IB) scholars are dedicating their theoretical efforts to explaining the empirical results of Rugman and co-authors (e.g., Management International Review, 2005/1, Special Issue). Given the contentious nature of the classification system used (see below), it is disconcerting that such theorising has proceeded without Rugman's empirical results having been tested for robustness. When researchers initially explore an issue, the adoption of ''first stab'' benchmarks to segment the phenomenon in question is inevitable and perfectly legitimate. But as we deepen our commit- ment to a particular research programme we must revisit those initial benchmarks to see how sensitive the findings are to the cut- offs imposed. To that end, in this paper we scrutinise the theoretical reasoning underpinning Rugman's classification system, and also explore practical aspects of its application. We then adjust the classification thresholds in line with our own arguments.
185
With relatively simple and justifiable adjustments in classification thresholds, considerable numbers of companies are shown to be active beyond the home region. Utilising longitudinal evidence we also highlight the growing share of bi-regional and, to a lesser extent, global firms between 1991 and 2001. Our findings call into question the pleas to IB researchers to abandon models and research ques- tions that include global strategies and to focus instead on intra-regional MNE activity.
, Global firms have less than 50% of sales in the home region and greater than 20% in each of the other two triad regions. This classification scheme is loosely based on Ohmae's (1985) ideas. Ohmae viewed a global presence (in terms of distribution) as imperative for firm success. His argument was founded on the need to recover the costs of innovation. As such, it was primarily an economies of scale and scope argument, although he also articulated the need for a global presence to avoid competitive blind spots. Only a balanced presence across the triad, then defined as the US, European Union and Japan, would render the MNE a triad power. Rugman adopts this need for balance in his definition of the global firm category. The imposi- tion of a minimum 20% host-region threshold is predicated on the roughly equal size of the three (broad) triad regions in terms of GDP:
The 20% figure is less than the one-third required for an equal triad distribution and so is biased downwards in favour of finding global MNEs. Conceptually it implies the successful deployment of customerend FSAs [firm-specific advantages] in three distinct markets. (Rugman, 2005: 64)1
THE EXISTING LITERATURE
Rugman's empirical research focuses on the world's largest firms as listed in Fortune magazine's Global 500 edition in 2002. He maintains that these firms are the major players in the IB world, collectively accounting for over 90% of the world's stock of foreign direct investment (FDI) and about half the world's trade (Rugman, 2005; but see Dunning, Fujita, & Yakova, 2007: footnote 4). Regional sales data for these companies is collected from annual reports, SEC filings and direct communications. Sales data is utilised because it ''constitutes a true performance measure at the output level'' (Rugman & Verbeke, 2004a: 7). Subsequent replication using company assets as the relevant criterion produced virtually identical results (Collinson & Rugman, 2008; Rugman & Brain, 2003). Utilising a three-region or ''triad'' perspective of global business activity, encompassing North America, Europe and Asia-Pacific, Rugman and co-authors seek to classify the world's largest firms as home-regional, bi-regional, host-regional or global. The classification employs the following (maximum and minimum) thresholds: , Home-regional firms have greater than 50% of sales in the home region. , Bi-regional firms have less than 50% of sales in the home region and greater than 20% in another region of the triad. , Host-regional firms, a special form of bi-regional firm (Rugman, 2005: 11), have greater than 50% of sales in a triad region other than the home region.
The 20% host-region threshold is used to define not only global firms but also the bi-regional firms, which are seen to deploy their FSAs successfully in the home region and one host region. The 50% home-region threshold is predicated on assumptions about the impact of the firm's geo- graphic sales distribution on the locus of both organisational decision-making and competitive advantage:
[W]e assume that a region representing more than 50% of total sales will systematically both shape and constrain most important decisions and actions taken by the MNE. It also implies a concentration of the MNE's customer-end related FSAs in that region. (Rugman, 2005: 64)
Rugman and co-authors then apply the 50% home-region threshold and 20% host-region thresh- old to classify the firms listed in Fortune's Global 500 list. They find an overwhelming preponderance of home-regional firms (see Table 1), a result that is
Table 1
Rugman's results (using 20% host-region threshold and 50% home-region threshold) Home-regional Of which solely domestic Bi-regional Host-regional Global No data available 120 Insufficient data available 15
320 (87.7%)
Source: Rugman (2005: 4).
54 (14.8%)
25 (6.8%)
11 (3.0%)
9 (2.5%)
186
largely invariant to the firms' industry or region of origin. Of the 365 firms for which sales data across the three triad regions was available, 2 a staggering 87.7% are homeregional, providing support for Rugman's claim that ''globalisation is a myth'' (2000: 163). We note in passing that, of the home- regional firms, 54 firms should more accurately be labelled as solely domestic. The remaining categories contain negligible numbers of bi-regional, hostregional and global firms. Since then, a substantial body of theoretical scho- larship has been built up in an attempt to explain this result of overwhelming home-regionality (e.g., Collinson & Rugman, 2008; Rugman, 2005; Rug- man & Verbeke, 2004a, b, 2005, 2007). In essence, Rugman and co-authors explain their finding on the basis of severe limits to the transferability and acceptability of firms' FSAs beyond the home region (Rugman & Verbeke, 2004a: 6). These limits to transferability and acceptability - which apply irrespective of whether FSAs are embodied in exports, transferred to licensees, or transferred to subsidiaries - are captured in the concept of (home) region boundedness of FSAs (Rugman & Verbeke, 2004a: 13). Any FSAs that may have been newly developed in host-region subsidiaries are equally limited in terms of transferability and acceptability, and typically cannot be deployed beyond the region in which they were created (Rugman &Verbeke, 2007: 202). Rugman and Verbeke (2007) elaborate their argu- ment by introducing the related concepts of liability of interregional expansion and liability of inter-regional foreignness. The lack of FSA deployment beyond the home region is explained by ''the additional costs of doing business abroad [which] are often much higher when venturing into other regions of the world than when expand- ing intra-regionally'' (Rugman & Verbeke, 2007: 201). The additional costs of doing business outside the home region stem from the need for expensive complementary FSAs that are necessary to render imperfectly non-location-bound FSAs applicable in host regions. Moreover, while certain (home coun- try) locationbound FSAs can be ''tuned up'' at little extra cost so as to be usable within the entire home region, this option is ruled out for inter-regional expansion. Overall, the cost penalty associated with expansion into host regions ''acts as an entry deterrent for many MNEs'' (Rugman & Verbeke, 2005: 14). In the final analysis, ''f firms have exhausted their growth in their home region of the triad and
still go into other regions, they then face a liability of [interregional] foreignness and other additional risks by this global expansion'' (Rugman, 2005: 1). As few firms are capable and willing to bear these costs and risks, they are destined to compete within the confines of their home region.
ASSESSING RUGMAN'S CLASSIFICATION SCHEME In this section we revisit the original empirical results that gave rise to (home-) regionalisation theory. At the outset, we contend that Rugman's classification system, utilising both a home- and a host-regional threshold, is overdetermined. Speci- fically, it incorporates an implicit hierarchy, with the 50% threshold for home-region sales over- whelming other considerations. If a firm exceeds this criterion then host-region sales become moot. Yet if we are to accept that 20% of sales in a host region is indicative of successful FSA deployment, why is such success negated by a large home-region presence? No strong theoretical grounds are offered for the precedence of the home-region threshold in the classification system, other than an assertion that homeregion sales in excess of 50% reflect and constrain decision priorities. We believe that the locus of decision-making is not a useful criterion in this instance. It will depend heavily on, inter alia, HQ's understanding of the host-regional context and the overall disposition of the firm (centralised vs decentralised), which in turn may be an outcome of the firm's administrative heritage and/or the power balance between HQ and the national/ regional operations (Bartlett & Ghoshal, 1989; Ghoshal & Bartlett, 1990). As such, inferences about decision-making rights based solely on a company's sales profile are troubling. More gener- ally, the introduction of decision-making is spur- ious in a classification system that is fundamentally about establishing firms' success at the output level and the deployment of FSAs in defined geographic regions (Rugman & Verbeke, 2004a: 7). The hierarchical nature of the classification system, with home-region sales taking precedent, means valuable information is ignored. For exam- ple, a company such as Volvo, which has more than 30% of its sales in North America, is classified as home-regional by virtue of having 51.6% of its sales in Europe. Crucially, 10 of the 25 biregionals recognised by Rugman (2005) had home-regional sales between 45 and 50%. A minor shift of their sales distribution in favour of the home region
187
(akin to Volvo) would pull these bi-regional competitors back into the home-regional category despite extensive sales (and putative successful FSA deployment) in a host region. Finnish telecom equipment manufacturer Nokia provides the most dramatic illustration of the problematic outcomes that result from the prece- dence of the home-region threshold. In 2001, Nokia reported 49% of total sales in the European home region, 26% in Asia and 25% in North America, making it one of the handful of global companies in Rugman's classification. By 2002, one year after Rugman's snapshot, Europe accounted for 54% of Nokia's net sales, with the Americas accounting for 22% and AsiaPacific for 24% (Nokia, 2003: 5). As a result, Nokia would have been categorised as a home-regional in 2002 on account of exceeding the 50% home-region threshold. We believe that any classification system that reassigns an erstwhile global company to the opposite end of the spectrum following a minor rebalancing of its sales portfolio (or a mere appreciation of the home-region cur- rency) is necessarily flawed. Further empirical issues spring from the 50% threshold. Home-region sales invariably include domestic (i.e., home country) sales. Yet it is unclear how we should interpret domestic activity in a taxonomy that is fundamentally about internatio- nalisation. Earlier, Rugman and Verbeke (1992) rightly argued that many FSAs may be location- bound (in the sense of country-bound), thus ham- pering the scope for international expansion. In the current classification scheme, success in transfer- ring, adapting and/or building FSAs in host regions is gauged by the firm's sales levels in those regions, but there is no equivalent measurement of the success in deploying FSAs within the home region exclusive of the home country. 3 As a result, a recognised home-regional firm may actually have little or no foreign sales in the home region, a fact also pointed out by Aharoni (2006) and Westney (2006). The evidence proffered by Rugman and co-authors on the 27 British firms among the Fortune Global 500 is a case in point (Rugman & Verbeke, 2007; Yip, Rugman, & Kudina, 2006). Average domestic sales for these companies were 52% while average homeregion sales stood at 64%. In other words, the home-region sales share net of home country sales was a mere 12%. The stylised average British firm would be classified as homeregional under the Rugman scheme, despite only one quarter of its foreign sales occurring in the home region and three quarters outside the home region.
For a broader sample, Osegowitsch and Sammartino (2007) report that the share of foreign sales beyond the home region is approximately two and a half times larger than foreign sales in the home region. In short, the use of a homeregional threshold without also accounting for domestic sales may obscure a firm's FSA reach in the home region beyond the home country. At the same time it may thwart an effective exploration of any inter-regio- nal (vs intraregional) liability of foreignness (LOF). We also have some empirical concerns with the 20% host-region threshold used for the bi-regional and global categories. The magnitude of the thresh- old may be difficult to justify since overall sym- metry across the regions in terms of GDP belies significant inter-industry differences. Respected market researcher Euromonitor reports highly skewed global demand in a variety of industries. For instance, the market size for dairy products - relevant for a Fortune Global 500 competitor such as Groupe Danone varies significantly, with Europe accounting for roughly twice the size of the North American market, and almost three times the Asia- Pacific market (Euromonitor, 2005). The same is true in pharmaceuticals, with 2001 sales estimated at US$172 billion for North America, US$125 billion for Europe and US$93 billion for Asia-Pacific (Euromonitor, 2000). 4 Demand for certain com- modities, on the other hand, is strongly skewed towards Asia. One of the most extreme examples may be the cement industry, with implications for a Global 500 company such as Lafarge. Rapidly growing emerging markets have the highest demand for cement. As a result, Asia accounts for more than two-thirds of global production volume, and China alone accounts for nearly 45% (Cembureau, 2007). World demand in dollar terms is bound to be less skewed, but no doubt still shows the (more devel- oped and slower growing) triad regions of Europe and North America lagging far behind Asia. While other industries show a more balanced distribution of demand across the triad, the above indicates that at least in some industries we should not expect a balanced sales distribution, even if a company's FSAs are of comparable strength in each region. The imposed minimum of 20% of sales in a particular host region may already exceed that region's proportional share of worldwide demand, and in some instances Rugman's benchmark may simply be too onerous. Moreover, the 20% host-region sales minimum, which is taken as evidence of successful FSA deployment in one or two host regions (to attain
188
bi-regional or global status), may be difficult to justify in view of the sampled companies' size. Takenaka was the smallest of the firms listed in the Fortune Global 500, with total sales in 2001 exceed- ing US$10 billion. Given the size of the firms considered, even the achievement of, say, 10% of total sales in a particular host region may be seen as evidence of successful FSAs deployment in that region. A business with sales in excess of US$1 billion cannot be established by catering to a few erratic customers, such as an expatriate community or similar niche, but requires significant firm-specific strengths to overcome any liability of foreignness. For instance, General Motors, a homeregional in Rugman's classification, may only have 15% of its sales in Europe and thus fail to satisfy the 20% hostregion threshold, but that represents 2001 sales in excess of US$25 billion. 5 A final empirical concern regarding the magni- tude of the 20% host-region threshold applies only to the global category. This category requires firms to have less than 50% of their sales in the home region and 20% or more in each of the other two triad regions. Given that the broad triad, on average, accounts for 90% of world demand in an industry, there is virtually no margin for error in a competitor that has close to 50% of sales in the home region. The corresponding arithmetic is o50 þ420 þ420 ¼ 90%. For instance, US-based 3M, a recognised bi-regional company, has 47% of sales in North America, 25% in Europe and 19% in Asia Pacific. As such, 3M's sales into the triad already exceed the stylised 90% of world demand claimed by the triad regions, and attaining global status according to Rugman's criteria may prove challenging. Since many bi-regionals as recognised by Rugman have home-region sales just below the 50% mark, achieving global status may be exceed- ingly difficult for them. In light of the concerns raised above, we argue that alternative classification regimes should be consid- ered to determine the status of firms. Most impor- tantly, the overdetermination of the current classification scheme must be remedied. For a simple alternative classification system, based on the geo- graphic deployment of FSAs, the use of host-regional sales thresholds would suffice. Firms that attain significant sales in one or two host regions are given bi-regional and global status, respectively. Firms that do not have significant host-regional sales are, by default, assigned to the home-regional category. 6 For such an alternative classification system, the definition of what constitutes ''significant''
host-region sales - so as to speak of successful FSA deployment in these regions - warrants further discussion. While 20% host-regional sales may seem a reasonable benchmark, there are arguments to support somewhat lower benchmarks, given both the heterogeneity in product market sizes and the huge size of the MNEs under examination (see earlier). There may also be merit in lower benchmark levels where the researcher is interested in the internationalisation process and the devel- opment of hostregional FSAs over time.
SENSITIVITY ANALYSIS OF THE DATA
In light of our theoretical and empirical concerns, and the alternative classification scheme sketched above, we engaged in sensitivity testing designed to ascertain the overall robustness of Rugman's results. Data concerning the 365 companies in the original sample were obtained directly from the Appendix in Rugman's (2005) book. Based chiefly on our arguments concerning the uneven division of demand across regions in some industries, we initially lowered the 20% host-region threshold to first 15% and then 10%, while retaining the 50% home-region threshold. As Table 2 shows, the bi-regional category weakens somewhat as 6-14 bi-regional firms shift to global, but, overall, the classification results are largely unchanged from Rugman's (Table 1). Given our views regarding the overdetermination of Rugman's classification system, we then abol- ished the 50% home-region threshold while retain- ing the original 20% host-region threshold. As a result, a company such as the above-mentioned Volvo, with 52% of its sales in Europe and 30% in North America, would now be considered bi- regional rather than home-regional. Similarly, Nokia would still be considered global in 2002 despite its European sales having nudged beyond the 50% threshold. Using this new system of classification, results vary markedly from Rugman's (Table 1). As can be seen in Table 3, 53 firms shift
Table 2 Results using 15 and 10% host-region thresholds and 50% home-region threshold Homeregional 15% threshold 10% threshold 320 (87.7%) 320 (87.7%) Bi-regional Host-regional Global
19 (5.2%) 11 (3.0%)
11 (3.0%) 11 (3.0%)
15 (4.1%) 24 (6.6%)
189
Table 3 Results using 20% host-region threshold, no homeregion threshold Home-regional 267 (73.2%) Bi-regional 87 (23.9%) Host-regional 2 (0.5%) Global 9 (2.5%)
produces contrasting findings, with global and bi- regional companies combined now accounting for a sizeable portion of the sample. We conclude that Rugman's classification is less robust than is desir- able. The numbers of ''newly found'' bi-regionals and, to a lesser extent, global companies clearly suggest that the home region is not the insurmoun- table barrier portrayed by Rugman and co-authors.
Table 4 Results using 15 and 10% host-region thresholds, no home-region threshold
LONGITUDINAL ANALYSIS
Homeregional 15% threshold 10% threshold 232 (63.6%) 200 (54.8%) Biregional 114 (31.2%) 122 (33.4%) Hostregional 1 (0.3%) 1 (0.3%) Global
19 (5.2%) 42 (11.5%)
from home-regional to bi-regional. A further eight firms shift from host-regional to bi-regional,7 while the global category remains unchanged. This casts some doubts on the robustness of Rugman's find- ings. In particular, the bi-regional category, which now accounts for a quarter of all firms, can no longer be dismissed as insubstantial. As a final step in our sensitivity analysis, we again dropped the 50% home-region threshold and utilised the less onerous 15 and 10% host-region thresholds. As is shown in Table 4, a strikingly different picture emerges once more. Global com- panies remain relatively few in number, although their share increases by 100-350% compared with Rugman's (Table 1). 8 More importantly, a very considerable proportion of the companies, around onethird in both instances, are now identified as bi-regionals. At this stage we remind the reader that the home-regional category in its present form actually contains 54 solely domestic firms. If these were set aside in a separate domestic category, global and bi-regional firms combined (133) would not be far behind the genuine home-regional MNEs (178) when adopting the 15% host-regional threshold. If the 10% threshold were to be adopted, their combined numbers (164) already exceed those of the genuine home-regional MNEs (146). These results confirm that it is the 50% home- region threshold, criticised earlier on theoretical and empirical grounds, that overwhelmingly drives Rugman's classification. Removing the overdeter- mination introduced by the home-region threshold
Rugman has bolstered his claims by arguing that home-regionals have dominated rankings of the world's largest firms for the past two decades. The situation is presented as unlikely to change in the foreseeable future. Efforts at intra-regional integra- tion and concomitant interregional protectionism (trade wars between the triad regions, barriers to regional entry, etc.) are seen to generate an enduring inter-regional liability of foreignness (Rugman & Verbeke, 2007). Rugman concludes that ''the end result is the persistence of MNEs that will continue to earn 80% or more of their income in their home triad region'' (2005: 63). To his chagrin,
some colleagues still seem to question these data. There must be a trend towards globalization over time they say. Well no - actually the aggregate [trade and FDI] data of chapter 11 strongly suggests the opposite; y Naturally these aggregate data trends are likely to be mirrored in the firm-level data. (Rugman, 2005: 239-240)
Scant longitudinal firm-level data is offered in support of these statements, however, and Rugman has been criticised for his lack of appreciation of the dynamics in firm internationalisation (Aharoni, 2006).9 Using company data available via Datastream we tested Rugman's claims for his original sample, which uses 2001 data. We sampled only those firms for which regional sales data was available for both 1991 and 2001. Not surprisingly, for a significant number of companies this was not the case, chiefly reflecting less stringent reporting requirements in the early 1990s, and the impact of mergers and acquisitions. In total we identified complete data sets for 198 firms, and the following results must be considered with the restricted sample in mind. 10 As shown in Table 5, we identify an unambiguous trend towards greater internationalisation beyond the home region between 1991 and 2001. Using Rugman's thresholds, the (small) numbers of bi- regional and global firms in 2001 represent an almost threefold increase on their numbers a decade
190
Table 5
Longitudinal classification (1991 and 2001) Year Homeregional 185 (93.4%) 166 (83.8%) 156 (78.8%) 133 (67.2%) 146 (73.8%) 115 (58.1%) 135 (68.2%) 92 (46.5%) Biregional 8 (4.0%) 20 (10.1%) 40 (20.2%) 58 (29.3%) 45 (22.7%) 68 (34.3%) 43 (21.7%) 74 (37.4%) Hostregional 3 (1.5%) 6 (3.0%) 0 (0.0%) 1 (0.5%) 0 (0.0%) 1 (0.5%) 0 (0.0%) 1 (0.5%) Global
Rugman's scheme (20% host-region threshold and 50% home-region threshold)
1991
2 (1.0%) 6 (3.0%) 1 (1.0%) 6 (3.0%) 7 (3.5%) 13 (6.6%) 20 (10.1%) 31 (15.7%)
2001
Our scheme I (20% host-region threshold, no home-region threshold)
1991 2001
Our scheme II (15% host-region threshold, no home-region threshold)
1991 2001
Our scheme III (10% host-region threshold, no home-region threshold)
1991 2001
Note: Of the 198 firms sampled, 53 (26.8%) were solely domestic in 1991. By 2001, only 22 firms (11.1%) were solely domestic.
earlier. Utilising our revised thresholds, the growth is not quite so stark, but still impressive off an already significant base. These results indicate that the situation is more fluid than realised by Rugman and co-authors. During the period of investigation, many former home-regional companies in our subsample were evidently able to successfully deploy their FSAs abroad.11
DISCUSSION
When researchers explore a new topic, the adoption of preliminary cut-offs to segment the phe- nomenon in question is often inevitable. It is good research practice, however, to revisit those initial thresholds at a later stage, to see how sensitive the findings are to the cut-offs originally imposed. To that end, in this paper we scrutinised the reasoning underpinning Rugman's chosen thresholds. We then manipulated the thresholds in line with our own arguments, and checked the robustness of the results obtained. Naturally, manipulating benchmarks in any (exhaustive and mutually exclusive) categorisation scheme will favour certain categories at the expense of others. We are conscious that in the context of this research endeavour there is no one host-region threshold that can claim to be the ''true'' threshold; an element of judgement pervades all the choices,
including ours.12 But we believe we have advanced empirical arguments in support of our judgements and, more generally, our call to relax Rugman's (20%) benchmark. With respect to the greater-than-50% home- region sales threshold we are far less open-minded. We believe it should be discarded for its incon- sistency with the avowed purpose of classifying firms based on the successful deployment of FSAs. Recall that the (20%) host-region threshold was adopted to indicate extensive customer-end FSAs in one or two host regions. The subsequent denial of these FSAs (and thus the denial of bi-regional or global status) on the grounds that home country sales greater than 50% of total sales imply ''con- centration of the MNE's customer-end related FSAs in that region'' (Rugman, 2005: 64) is untenable and suggests overdetermination. Our sensitivity analysis suggests that the 50% homeregion threshold is the primary driver of Rugman's findings. Once this threshold is aban- doned, and host country minimum thresholds relaxed to a less onerous 15% or 10% of sales, a significantly different picture emerges. Among the world's largest companies many are homeregional, but a substantial portion of them are bi-regional and a not insignificant number are global players.
Reassessing (home-)regionalisation
Thomas Osegowitsch and Andre Sammartino´
191
The lack of robustness in Rugman's empirical findings has serious theoretical implications. At present, Rugman argues that ''f firms have exhausted their growth in their home region of the triad and still go into other regions, they then face a LOF and other additional risks by this global expansion'' (2005: 1). Firms will stay at home because a home-regional presence
Potential reach of FSA Realised reach of FSA Global
Bi-regional Home-regional Sub-regional
Home Country
allows the MNE to gain all the potential economies of scale and scope and/or differentiation advantages within its home triad market. The additional scale, scope or differ- entiation advantages to be gained by going global are not sufficient to compensate for the enhanced risks. (Rugman & Brain, 2003: 7)
Firm A (1991)
Firm C (2001)
Firm B (1991) Firm B (2001)
Firm C (1991)
Firm A (2001)
These statements concerning the thrust of regionalisation theory in its current form are difficult to reconcile with the significant number of companies that have successfully pushed beyond the homeregion boundary. Results of our longitudinal analysis also cast doubt on Rugman's pronouncements concerning change over time as well as the lack of benefits from expanding outside the home region. Contrary to his claim that ''sales data for earlier periods [i.e., prior to 2001] are highly unlikely to provide much new information or cause us to reclassify more than a handful of the [y] firms'' (Rugman, 2005: 239), our analysis indicates strong growth in the bi- regional and global categories and a corresponding decline in the home-regional category for the period 19912001. We are left with the responsibility to explain the results our empirical investigation uncovered. Based on the observed (international) sales pattern at a given point in time, Rugman and co-authors draw conclusions regarding the deployability or reach of FSAs. 13 We agree with their proposition that sales are the decisive indicator of FSA reach: (international) sales reflect (international) customers' judgement about the attractiveness of the firm's goods or services and, ultimately, the strength and relevance of the underlying FSAs. If we accept this proposition we must also accept that (international) sales over time reveal the evolution of FSA reach. Based on our empirical evidence we must then conclude that FSA reach has beenexpanding from 1991 to 2001, allowing more firms to deploy their FSAs beyond the home region. To explain the observed trend, we deem it instructive, albeit simplistic, to conceptualise FSA reach as a series of concentric circles, as indicated in Figure 1. This reflects a view that, to paraphrase
Figure 1
Conceptualising FSA reach over time.
Westney (2006: 447), differences between homeregionalisation and bi-regionalisation and globali- sation may be merely quantitative rather than qualitative. At any one point in time, certain FSAs may be inherently national, as they are either irrelevant beyond the domestic context, or the cost of transfer/adaptation would outweigh the benefits obtainable. Other FSAs may be home-regional in reach, while still others may be bi-regional or even global. In addition to these established levels of analysis we would add the sub-regional level, as introduced by Enright (2005: 87). Sub-regions are groups of countries within a region that have much more in common with one another than with other countries in the region. These commonalities may reside along any of the four primary dimensions - cultural, institutional, geographic and/or economic (Ghemawat, 2001). This view of FSA reach as a series of concentric circles can be further refined to accommodate the results of our longitudinal analysis. One explanation why firms are increasingly spreading beyond the home region is predicated on demand-side arguments. Differences across countries and across regions - as conceived in terms of cultural, institutional and economic ''distance'' - may be diminishing over time. While it is clear that the world will remain in a state of ''semi-globalisation'' (Ghemawat, 2003) into the foreseeable future, it appears that at least along certain dimensions ''distance'' has shrunk, through factors such as new technologies, trade
192
and FDI agreements (bilateral and multilateral), and the emerging global mass media (Economist, 1997). In short, national as well as regional econo- mies have become more interdependent and, in the process, more compatible. Consequently, certain FSAs previously restricted to, say, the home country (Figure 1: Firm A in 1991), may be applicable to the entire home region 10 years later (Firm A in 2001). Other FSAs may have retained their level of reach as the dimension(s) relevant to the FSA failed to converge over time (Firm B in 1991 and 2001). A complementary explanation for the findings of our longitudinal analysis emphasises supply-side arguments. To that end we initially draw a distinc- tion between the potential and realised reach of FSAs and contend that FSA deployment is plagued by significant lag effects. For a variety of reasons it takes time to adjust the firm's geographic scope to the potential reach of its FSAs (especially since potential reach may be a moving target, as indicated above). These reasons include Penrosian constraints, 14 delays due to the challenges in assembling the required co-specialised assets (Teece, 1986) or, for companies in the earliest stages of internationalisation, ample opportunities for aggressive growth within the home region. 15 These lag effects may also help explain why, at present, home-regional sales account for a very large portion of total sales, and why Rugman (2005) and this investigation (still) find the home-regional category - however defined - the dominant cate- gory. To illustrate our point, Firm C in Figure 1 possessed FSAs with global reach in 1991, but had failed to realise their full potential. By 2001 the actual deployment of its FSAs was still below the potential, although by going bi-regional the firm had advanced beyond the situation a decade earlier. The longitudinal trends charted for the period 1991-2001 suggest that firms do eventually close the gap to the potential reach of their FSAs. Correspondingly, a significant number of erstwhile home-regional firms had already migrated into the bi-regional category and, to a lesser extent, the global category by 2001. If the trends presented in this paper persist, the home-regional category will be further weakened in years to come as more large companies assume bi-regional and possibly global status.
CONCLUSION
Rugman and co-authors' empirical research has drawn attention to a previously under-appreciated phenomenon: in IB, many companies operate
mostly within their home region. Their research was timely in that it helped pour cold water on the globalisation hyperbole. It has also given rise to the beginnings of a regionalisation theory intent on explaining why firms are compelled to restrict theiractivities to the home region. Our own results suggest that the case for homeregionalisation is overstated. In turn, regionalisation theory in its current form is straining against the evidence assembled in this paper. In the concluding paragraphs we offer a number of pointers to how this research stream may be advanced. Rugman and co-authors effectively dismiss bi-regional and global strategies, asserting that there are severe limits on the transferability and accept- ability of firms' existing FSAs beyond the home region. Moreover, ''the administrative heritage of most MNEs, undoubtedly conducive to home region market success, may well constitute an administra- tive rigidity when attempting to penetrate host regions'' (Rugman, 2005: 197). As a result, managers are advised to ''think regional [y] and forget global'' (Rugman & Hodgetts, 2001: 341). Yet all these claims rest squarely on Rugman's evidence that global and bi-regional firms are the exception in the world of big business. Our revision of the data - uncovering significant numbers of biregional firms [which ''may be regarded as partly global'' (Rugman, 2005: 4)] and global firms, and their share having grown markedly from a decade earlier - dilutes the case considerably. Regionalisation theory and its concepts, such as the interregional LOF, are fundamentally depen- dent on intraregional integration and enduring inter-regional differences. While intra-regional integration policies reduce the costs associated with deploying FSAs in the home region, in host regions the ''requirement for high, regionspecific 'linking' investments acts as an entry deterrent for many MNEs'' (Rugman & Verbeke, 2005: 14). While we confirm that sales of a large share of firms are concentrated mostly in the home region, we also identify a considerable and growing share of companies that manage to deploy their FSAs in host regions. This suggests that any interregional LOF is too small to generally deter firms from spreading beyond the home region. 16 In turn, this would suggest that inter-regional differences across the triad may be weak (and diminishing) and/or intra-regional integration less advanced than is commonly assumed.17 Taken to their logical endpoints, such conclu- sions ultimately lead to questions about the merit
193
of the (triad) region as a unit of analysis. The identi- fication of regional sales outcomes and region- based trends - as performed by Rugman and ourselves - may mask significant variation within regions. In fact, the extant literature on the inter- nationalisation of the firm (e.g., Benito & Gripsrud, 1992; Johanson & Wiedersheim-Paul, 1975) em- phasises the individual country, thus implying that firms are predominantly making country-by-coun- try choices in the deployment of their FSAs. Equally, notions of LOF (Zaheer, 1995) have over- whelmingly been theorised and explored at the country rather than the regional level. 18 Aggregat- ing country-based choices to the regional level, and inferring that firms are making regionby-region choices (and that, say, a LOF exists at a regional level), may hamper our understanding of actual firm strategies. In short, the usefulness of the (triad) regional level of analysis in this context is yet to be fully confirmed. One way to establish the merit of the (triad) regional level of analysis, and of the region-based concepts that make up regionalisation theory, would be the concurrent exploration of country- and region-level data. At present, no large-scale database offers the required detailed information on MNEs, but more manageable (industrybased) studies may allow such investigations. Ultimately, such research would also allow us to discern the level of differences within and across regions as perceived by MNE managers. Once the usefulness of the region as a level of analysis is established, attention can turn to refi- ning the corresponding theory. Initially, it would seem that the current, strong form of regionalisa- tion theory must be relaxed to accommodate a scenario where bi-regional and global firms are no longer exceptions. Further elaboration of the theory may be facilitated by a particular focus on biregional firms. Given their significant numbers and their inbetween status (in between home- regional and global), studying these firms would seem to hold great promise for a better under- standing of an LOF at the regional level. For instance, there are indications that any inter- regional LOF is far from uniform. Taking the 15% host-regional threshold (Table 4) we identify 114 bi-regional firms, 97 of which are North American or European in origin. Virtually all of these 97 firms are bi-regional by virtue of strong sales in each other's region rather than in the Asia-Pacific. Only four firms (Boeing, Ericsson, Manulife Financial and Motorola) have successfully deployed their
FSAs in the Asia-Pacific rather than in the other ''Western'' host region. This would seem to suggest that any interregional LOF that might exist is highly asymmetric. Regionalisation theory is also in need of more longitudinal studies, to establish a clearer picture of the dynamics in (intra- and inter-regional) expan- sion. At the most general level, future research should also strive for more fine-grained empirical models - by taking up some of the arguments raised earlier, using continuous instead of categorical measures, and adopting other innovative measure- ment approaches (e.g., Asmussen, 2006) - to get a better grip on regionalisation phenomena. Alan Rugman and co-authors have opened up an exciting new research stream. The data analysis presented herein casts doubts on regionalisation theory in its current state, and hints at some of the questions that remain to be addressed. Clearly, further empirical work and theorybuilding is warranted.
ACKNOWLEDGEMENTS
We gratefully acknowledge the assistance of Eric Quintane in compiling the data used in this paper. We also thank Editor-in-Chief Professor Arie Y Lewin and two anonymous JIBS reviewers for their efforts. NOTES Strictly speaking, if the broad triad accounts for roughly 90% of world output, a perfectly balanced global presence would require 30% in each triad market rather than 33.33%. 2 Accounting conventions give firms considerable leeway in how to report the geographic breakdown of their sales, and firms rarely follow the exact triad scheme. As a result, this entire research stream is affected by a significant level of ''noise'' in the data used. See Rugman (2005: Chapter 2) for a compre- hensive disclosure and a discussion of the inevitable limitations that arise from relying on sales data as reported in company accounts. 3 Setting a threshold for the share of home region ex domestic sales is an obvious solution, but complicated by the fact that the ''rest of home region'' varies with the company's country of origin. For instance, the size of the ''rest of home region'' differs markedly for a US firm compared with a Canadian firm. Approaches that account for the relative GDP of the home country (e.g., Asmussen, 2006) seem to hold great promise in this regard. 4 Not only sales but also profits may be highly skewed across triad regions. For instance, it is well
1
194
known that most big pharmaceutical companies, many of which feature in the Fortune Global 500, make about half of their profits in the US alone (Economist, 2007). 5 On the suggestion of a reviewer we set out to count the number of firms with host-region sales greater than US$1 billion. Sufficient data to perform this analysis was available for only 219 firms in Rugman's 365-strong sample. Of those 219 firms, 97 had sales greater than US$1 billion in both host regions, while 35 had attained such levels in one host region. The corresponding figures were 58 and 61 firms when the bar was raised to US$2 billion. 6 Clearly, such a simple scheme does not address our concerns regarding the inclusion of domestic sales among home-region sales. More complex classifica- tion systems may explicitly set thresholds for home- region sales ex domestic sales (weighted for the size of the home country within the home region). 7 Once the 50% threshold for home-region membership is removed, it becomes difficult to manage and rationalise the 50% benchmark for the special hostregional category. While it is a far from ideal categorisation, host-regional status now belongs only to those firms that achieve a market share greater than 20% in a host region and less than 20% in the home region. Most of the firms in the host-regional category tend to collapse into the bi-regional (and occasionally the global) category in subsequent tables. 8 Data insufficiencies bias this classification against finding global competitors, since Rugman's data contains numerous firms with no available information for a particular region (mostly Asia) and/or reported sales that make up significantly less than the 90% of total sales expected for the entire triad. For example, in 2001, pharmaceutical giant Glaxo reported European sales and US sales at 28.6 and 49.2% of total sales, respectively, rendering it a bi-regional in Rugman's classification scheme. Where the remaining 22.2% of sales took place was not disclosed by the company. It is likely that Glaxo's Asian sales during that year exceeded the 10% threshold, which would make the firm a global player according to one of our classifica- tion regimes. 9 Rugman (2005) does provide rudimentary tempor- al sensitivity analysis of his classification. For thatpurpose, he utilises updated 2002 figures for 60 firms in his original 2001 sample. Not surprisingly, a single year after the original analysis, ''only two of these sixty firms were recategorised'' (2005: 32). This hardly constitutes a strong test of robustness over time. 10 We are cautiously confident about the representativeness of our sample. w2-tests of Rugman's original sample and our own, based on firms' region of origin
and their distribution across the four categories in 2001 (using Rugman's classification scheme), were nonsignificant at the 1% level. 11 See Osegowitsch and Sammartino (2007) for a more detailed investigation of these trends, but see also Rugman and Oh (2007). 12 The definition of thresholds to distinguish global from bi-regional and home-regional firms is likely to remain a contentious issue unless the community of IB researchers can reach consensus as to what constitutes ''significant'' sales that warrant a particular categorisa- tion. We note in passing that there is also disagree- ment whether sales should be the relevant criterion (see Aharoni, 2006; Rugman & Verbeke, 2004a: 7). Clearly, such a situation is far from unique. After all, IB scholars are still debating exactly what constitutes a multinational firm (Aharoni, 1971; Ghoshal & Westney, 1993). 13 The notion of FSA country or region boundedness implies an exogenous phenomenon, determined entirely by environmental forces. We view the boundaries of effective FSA leverage as also governed by firmspecific aspects and as partly controlled by management. As a result we prefer the term ''FSA reach''. 14 The ''Penrose effect'' simply suggests that there are strict limits to a firm's growth rate due to dynamic adjustment costs that are incurred by firms trying to grow their productive resources. Penrose (1959) focused on one major source of dynamic adjustment costs, namely those attributable to the expansion of management resources. She insists that a firm's expansion requires the services of experienced internal managers. Hiring new managers is an inadequate solution, since only seasoned internal managers can undertake the coordination task inherent in firm expansion. As a result, the rate of growth is limited by the rate at which the firm can develop internal managers. Not only domestic expansion but also international expansion is subject to the Penrose effect (Tan & Mahoney, 2005). 15 We agree that firms typically begin their internationalisation in the home region. The home region is likely to contain some of the most similar (i.e., ''least distant'') countries. FSA deployment in these initial destinations is facilitated by institutional similarities, cultural affiliations, lower transportation costs, etc. Beyond the initial round of internationalisation, market choices are likely to be much less focused on the home region, as similar countries also exist in other regions. 16 Based on evidence presented elsewhere (Osegowitsch & Sammartino, 2007), we also dismiss the possibility that firms venture into host regions only once they have exhausted growth opportunities in the
195
home region, as a kind of ''last resort'' option to keep growing. Instead, it seems that many firms are simultaneously deploying their FSAs within and beyond the home region. 17 See also Aharoni (2006) and Westney (2006) on this point. 18 Hejazi (2007) is, to our knowledge, the only study trying to model a regional LOF in terms of MNE sales
(as well as asset, income and employment) distribu- tion. His methodology explicitly tries to disentangle country and regional effects. While the results must be treated as strictly preliminary, they do suggest that ''distance'' (as well as size differentials) between individual countries explains the sales distribution of US MNEs; regional (dummy) variables yield no additional explanatory power.
REFERENCES
Aharoni, Y. 1971. On the definition of a multinational corporation. Quarterly Review of Economics and Business, 11(3): 2737. Aharoni, Y. 2006. Book review: Alan M. Rugman, The regional multinationals. MNEs and global strategic management. International Business Review, 15(4): 439-446. Asmussen, C. G. 2006. Local, regional or global? Quantifying MNC geographic scope. SMG Working Paper 14/2006, Centre for Strategic Management and Globalization, Copenhagen Business School. Bartlett, C. A., & Ghoshal, S. 1989. Managing across borders: The transnational solution. Boston: Harvard Business School Press. Benito, G. R. J., & Gripsrud, G. 1992. The expansion of foreign direct investments: Discrete location choices or a cultural learning process? , 23(3): 461-476. Cembureau 2007. 2005 World cement production by region.http://www.cembureau.be/Documents/KeyFacts/STATISTICS/World Cement Production by Region.pdf. Accessed 24 March 2007. Collinson, S. C. & Rugman, A. M. 2008. The regional nature of Japanese multinational business. , 39(2): 215-230. Dunning, J. H., Fujita, M., & Yakova, N. 2007. Some macro-data on the regionalisation/globalisation debate: A comment on the Rugman/Verbeke analysis. , 38(1): 177-199. Economist 1997. One world?. 18 October: 79-81. Economist 2007. Billion dollar pills. 25 January: 61-64. Enright, M. 2005. Regional management centers in the AsiaPacific. Management International Review, 45(Special Issue 1): 5982. Euromonitor 2000. Pharmaceuticals: A world survey. World market overview. Global Report, 2 June, Global Market Information Database.https://www.euromonitor.com/. Accessed 3 January 2007. Euromonitor 2005. The world market for dairy products. Global Report, 7 June, Global Market Information Database. https:// www.euromonitor.com/. Accessed 3 January 2007. Ghemawat, P. 2001. Distance still matters. Harvard Business Review, 79(8): 137-147. Ghemawat, P. 2003. Semiglobalization and international business strategy. , 34(2): 138-152. Ghoshal, S., & Bartlett, C. A. 1990. The multinational corporation as an interorganizational network. Academy of Management Review, 15(4): 603-625. Ghoshal, S., & Westney, D. E. 1993. Introduction and overview. In S. Ghoshal and D. E. Westney (eds), Organization theory and the multinational corporation: 1-23. New York: St Martin's Press. Hejazi, W. 2007. Reconsidering the concentration of US MNE activity: Is it global, regional or national? Management International Review, 47(1): 5-27. Johanson, J., & Wiedersheim-Paul, F. 1975. The internationalization of the firm: Four Swedish cases. Journal of Management Studies, 12(3): 305-322. Nokia 2003. Nokia in 2002.http://www.nokia.com/A4126499. Accessed 28 March 2007. Ohmae, K. 1985. Triad power: The coming shape of global competition. New York: The Free Press. Osegowitsch, T., & Sammartino, A. 2007. Exploring trends in regionalisation. In A. M. Rugman (ed.), Research in global strategic management, vol. 13, Regional aspects of multinationality and performance. Amsterdam: Elsevier: 45-64. Penrose, E. T. 1959. The theory of the growth of the firm. Oxford: Oxford University Press. Rugman, A. M. 2000. The end of globalization. London: Random House. Rugman, A. M. 2005. The regional multinationals: MNEs and ''global'' strategic management. Cambridge: Cambridge University Press. Rugman, A. M., & Brain, C. 2003. Multinational enterprises are regional, not global. Multinational Business Review, 11(1): 3-12. Rugman, A. M., & Hodgetts, R. M. 2001. The end of global strategy. European Management Journal, 19(4): 333-343. Rugman, A. M., & Oh, C. H. 2007. Multinationality and regional performance, 2001-2005. In A. M. Rugman (ed.), Research in global strategic management, vol. 13, Regional aspects of multinationality and performance. Amsterdam: Elsevier: 31-43. Rugman, A. M., & Verbeke, A. 1992. A note on the transnational solution and the transaction cost theory of multinational strategic management. , 23(4): 761-771. Rugman, A. M., & Verbeke, A. 2004a. A perspective on regional and global strategies of multinational enterprises. Journal of International Business Studies, 35(1): 3-18. Rugman, A. M., & Verbeke, A. 2004b. Regional transnationals and triad strategy. Transnational Corporations, 13(3): 1-19. Rugman, A. M., & Verbeke, A. 2005. Towards a theory of regional multinationals: A transaction cost economics app- roach. Management International Review, 45(Special Issue 1): 5-17. Rugman, A. M., & Verbeke, A. 2007. Liabilities of regional foreignness and the use of firm-level versus country-level data: A response to Dunning et al. , 38(1): 200-205. Tan, D., & Mahoney, J. T. 2005. Examining the Penrose effect in an international business context: The dynamics of Japanese firm growth in US industries. Managerial and Decision Economics, 26(2): 113-127. Teece, D. J. 1986. Profiting from technological innovation. Research Policy, 15(6): 285-305. Westney, E. 2006. Book review: The regional multinationals: MNEs and ''global'' strategic management. Journal of International Business Studies, 37(3): 445-449.
196
Yip, G., Rugman, A. M., & Kudina, A. 2006. International success of British companies. Long Range Planning, 39(3): 241-264. Zaheer, S. 1995. Overcoming the liability of foreignness. Academy of Management Journal, 38(2): 341-363.
doc_743645027.docx