Which market next: CAGE ranks them, Uppsala sizes the step

Ghemawat (2001) tells the story of Tricon Restaurants International, the international arm of the company that runs Pizza Hut, Taco Bell and KFC, choosing where abroad to concentrate its capital. By the size of fast-food consumption, Mexico (US$700 million) ranked 16th of 20 markets. Adjusted for distance, it rose to second (Ghemawat, 2001, p. 147). That is the first half of choosing the next market: which country deserves priority (Tricon already had restaurants in 27 countries (Ghemawat, 2001, p. 147)). The second half is how much to commit.

This post uses the CAGE framework for the first half and the Uppsala model for the second, then tests both against three public cases: Huawei in Egypt, Huawei in Thailand and Tesla in Shanghai. The route in is covered in how to choose a foreign-market entry mode, fast against step by step in the born-global post.

Is choosing the next country the same question as setting the step size?

No: two questions, two tools. The source itself says so: Johanson and Vahlne (1977) write that they concentrate on extending operations in individual markets (Johanson and Vahlne, 1977, p. 23). The order of countries enters only as an observation: the sequence in which firms set up in countries "seems to be related" to psychic distance, the sum of factors preventing the flow of information (Johanson and Vahlne, 1977, p. 24).

How does CAGE rank candidate countries?

Look at each candidate through Ghemawat's four groups, then adjust the size of the opportunity for distance instead of looking at size alone. The usual tool is country portfolio analysis, which focuses on GDP and consumer wealth and so ignores the costs and risks of a new market (Ghemawat, 2001, p. 138).

The size of the effects comes from Frankel and Rose estimates he cites: countries 5,000 miles apart trade only 20 per cent of what they would if 1,000 miles apart; a shared language multiplies trade by three (Ghemawat, 2001, p. 138; Ghemawat, 2001, p. 142). Those are figures from 2000, from an unpublished working paper. Industries differ: food touches consumers' identity and so is sensitive to culture, electricity not at all (Ghemawat, 2001, p. 140; Ghemawat, 2001, p. 142). Here is how he adjusted for Tricon:

One limit: distance also depends on the company (Tricon already owned more than four-fifths of its Mexican outlets and held 38 per cent of the local market (Ghemawat, 2001, p. 147)), and this is a restaurant chain in 2001, not a small workshop.

How does Uppsala size the next step?

Risk is existing commitment multiplied by uncertainty, and step size is capped by the risk you can tolerate. Johanson and Vahlne (1977) split commitment into the amount of resources put in and how hard they are to move to other uses; a marketing organisation built around the firm's products, with its customer relations, is a deeper commitment than cash (Johanson and Vahlne, 1977, p. 27). Scale-increasing steps are taken while existing risk is below the tolerable level; large steps appear in firms with large resources or little uncertainty (Johanson and Vahlne, 1977, p. 30).

Experience reduces uncertainty. Objective knowledge can be taught; experiential knowledge can only be learned by doing, and abroad a firm must build it up in the country itself (Johanson and Vahlne, 1977, p. 28). The short rule: small steps, unless the firm has very large resources, conditions are stable and homogeneous, or it has much experience of similar markets (Johanson and Vahlne, 1977, pp. 30-31). The paper gives no numerical threshold: how small is small is for you to set.

Dunning (1988) notes that his 1976 paper listed a type of ownership advantage that comes from being a multinational in the first place (Dunning, 1988, p. 2). In my linking of OLI to Uppsala, experience in the first market is an asset carried to the next. From that I assemble a rough grid: distance sets the order of trial, experience sets the step size.

Did the three public cases climb the ladder or skip rungs?

All three have earlier rungs in the sources themselves, but step size can be compared only for Thailand, and only in part.

Huawei in Cairo. The release calls it the eighth OpenLab for enterprise customers, located in Egypt "to serve and influence all of Northern Africa". Ni Zheng gave three reasons: the Cairo ICT market is relatively mature and influences surrounding countries, Egypt recognises the significance of digital transformation, and more than half of the top 15 African universities in the UK QS ranking are in Egypt (Huawei, 2017). These are criteria for choosing a base, not distance. The site ran for a year as an R&D unit before opening to customers; I read that as an uncertainty-reducing step before the public one, and the release does not explain it.

Huawei in Thailand. The third data centre, planned for 2021, costs 700 million baht, equal to the first two combined: the Eastern Economic Corridor in 2018 and Bangkok in 2019 (Xinhua, 2020). The total invested in the country doubles in one step. Uppsala does not define a "small step", so I do not say this breaks the rule. Abel Deng cited Thailand's 4.0 policy and the February 5G licence auctions (Xinhua, 2020). Johanson and Vahlne say uncertainty can fall when politics stabilises (Johanson and Vahlne, 1977, p. 30); joining the two is my inference.

Tesla in Shanghai. The report lists what Tesla already had in China before the groundbreaking: cars on sale there (the report gives the Model 3's price in China), more than 3,350 charging stations of two kinds (over 1,500 fast chargers and 1,850 destination chargers) built over six years, and the first R&D centre outside the United States, opened in Beijing in October 2018. The plant's capacity of 500,000 cars is about twice the 245,240 Tesla delivered worldwide in 2018 (Xinhua, 2019). The report gives counts of charging stations, not money, so I cannot compare the plant with the earlier steps. It ties the plant to a new policy allowing wholly owned subsidiaries, of which Tesla was the first to take advantage (Xinhua, 2019). Dunning (1988) notes that some government intervention can encourage or discourage inward direct investment (Dunning, 1988, p. 4). This gate seems to have opened because a rule changed, not because Tesla learned more; the source does not say whether Tesla waited for that policy.

What the test shows: the order of rungs appears in all three sources. Uppsala's thresholds cannot be checked, since no source speaks of earlier resources or experience. CAGE ranking cannot be checked either.

How would an invented rice-noodle workshop rank three markets?

Rank by distance first, then adjust the step size for the experience already in hand. The workshop, the owner's relative and every rating below are invented by me; I call it Workshop A so as not to match a real brand.

Workshop A in Can Tho has 90 workers, makes rice noodles, sells in Vietnam and has exported a few trial lots. Three candidates: Cambodia, the Philippines and Australia. I rate near, middle or far on Ghemawat's four columns and do not add scores, because he gives no scale.

Add one invented fact: the owner's niece managed an Asian grocery in Sydney for six years. Market-specific knowledge comes mainly from the market itself (Johanson and Vahlne, 1977, p. 28); the Pharmacia case suggests someone who has been in the country can shorten the ladder (Johanson and Vahlne, 1977, p. 24).

The invented result: Cambodia goes first, near and cheap to learn from, through a distributor with one trial container. Australia is not dropped, but its first step is a small lot through the niece's shop, asking for feedback on packaging and taste. Before each step the workshop writes down the maximum loss it can bear in each country. This suits you if you have two to four candidates and no firm order from any. It does not suit you if a customer has already come to you: then the question is fast against step by step, as in the born-global post. If the big step is buying a whole chain, read Walmart in Germany.

Going deeper: what did I read, and what could I not check?

I read Johanson and Vahlne (1977) in full through an automated text conversion of the PDF; the risk formula lost its operator symbol, so I use only what follows from the formulas right after it. Their data are studies of Swedish firms published between 1966 and 1975, by the reference list, and they themselves say predictive value is limited (Johanson and Vahlne, 1977, p. 23).

I read Ghemawat (2001) from a 10-page scan, using automated character recognition and checking the main figures against the page images; the scan lacks two interleaved pages, so page numbers come from the page footers. I read Dunning (1988) from page 1 to 5 in an automated text conversion and use only two passages.

For the three cases I read Huawei's release, the Xinhua report on China Daily and the Xinhua report on the Pudong site in full. All are one-sided sources, one piece per case; I did not read financial reports or independent press, and I do not know the outcomes. The library's CAGE framework summary has only the four groups, so CAGE details come from Ghemawat's own article. Joining CAGE with Uppsala, the step-size grid and Workshop A are my inference, not checked against data on Vietnamese firms.