“Flexible” or “fickle”

Picture a weekly meeting. The manager from the Chinese headquarters says: “Last month’s numbers are in. Customer acquisition cost for product A is too high, so we’ll move the budget to B and change some channels too.” To the headquarters team this is the most ordinary of sentences: read the data, change the approach, iterate fast. Many Chinese internet companies learned this as a basic skill in the competition of their home market. The local partner across the table may be thinking something else: three months ago, when we signed the memorandum, you said A was the core product. I hired two people for it and negotiated with three distributors, and now one sentence moves everything?

Scenes like this are common in cross-border work. Behind them are two different understandings of good business conduct, which this book calls two business “operating systems”.

The rule-based commitment game. Here the core value of a business relationship is predictability. If I agreed, I do it; if circumstances change, we first apply the contract, and where the contract is silent we sit down and renegotiate. A commitment is valuable precisely because it limits one’s own future choices: Schelling pointed out in The Strategy of Conflict that a party able to bind itself credibly often obtains better cooperation (Schelling 1960). On this understanding, frequent changes of plan use up the value of commitment.

Outcome-based adaptive search. Here the core value is the result. A plan is only the current best guess about the best path; when new information arrives the guess should be revised, and holding on to a plan the data have already refuted is the irresponsible choice. On this understanding, frequent adjustment is diligence and sticking to the plan is rigidity.

Both understandings have merit and both have costs. The trouble comes when each side reads the other’s behaviour through its own system: what one side sees as flexible pragmatism, the other may decode as erratic and unpredictable; what one side sees as keeping faith, the other may see as rigid and blind to data.

High and low context

The anthropologist Edward Hall proposed an influential pair of concepts: high-context and low-context communication (Hall 1976). In high-context communication much of the meaning lies not in the words themselves but in relationships, setting and shared background knowledge, and the speaker assumes the listener “gets it”. In low-context communication meaning is carried mainly by explicit language, and important things are written down and spelled out. Hall placed China, Japan and many Asian societies near the high-context end, and Germany, Switzerland, the United States and the Nordic countries near the low-context end.

The pair helps with the weekly meeting above. When the headquarters team changes the plan, it often assumes the partner also understands the background that “the market has changed”, so it sees no need to explain; a partner who relies more on explicit agreement sees an unexplained change as a change without a reason.

Hofstede’s dimensions

More often used in quantitative work are Hofstede’s cultural dimensions. They began with a questionnaire survey of IBM employees and were later extended to six dimensions: power distance, individualism, masculinity, uncertainty avoidance, long-term orientation and indulgence (Hofstede 2001; Hofstede et al. 2010). The most relevant to this book are:

• Uncertainty avoidance: how far members of a society feel uneasy with ambiguity and the unknown and want rules and plans to reduce them.
• Long-term orientation: how far a society values persistence, thrift and preparing for the future, rather than tradition and present obligations.

These scores help us ask questions but cannot answer them directly, and they come with important cautions.

Trust in joint ventures

International joint ventures are a natural laboratory for cross-cultural cooperation. Two parent companies contribute resources and run a business together; they must cooperate while each pursues its own interests. The literature repeatedly emphasises that a joint venture’s stability depends heavily on mutual trust, and that trust co-evolves with learning and with the forms of control used as the venture proceeds (Inkpen and Beamish 1997; Inkpen and Currall 2004).

One change, several decodings

To make “decoding” precise, consider a simple Bayesian calculation. Suppose that, to the local partner, the foreign team may be one of three types: committed, rarely changing its plan; adaptive, changing its plan when the market state shifts and not otherwise; and opportunist, changing its plan whenever it privately pays, regardless of the market state. The partner starts out thinking the team is most likely one of the first two, with only a 10 % chance that it is an opportunist.

The key point is that the adaptive and the opportunist types can change their plans with almost the same frequency; they differ only in whether the changes are correlated with the market state. If the partner cannot see the market state, he cannot tell the two types apart.

# How a local partner decodes a plan change. Three types of foreign team, as the partner sees them:
#   committed   - changes plan in 5 % of months, whatever happens;
#   adaptive    - changes plan when the market state shifts (95 %), almost never otherwise (2 %);
#   opportunist - changes plan in 35 % of months, whenever it privately pays, ignoring the state.
# The state shifts in 30 % of months, so the adaptive team also changes plan in about 30 % of months.
# The partner starts with prior 0.45 / 0.45 / 0.10 and watches 12 months.
PRIOR = {"committed": 0.45, "adaptive": 0.45, "opportunist": 0.10}

def p_change(team, shift):
    if team == "committed":
        return 0.05
    if team == "adaptive":
        return 0.95 if shift else 0.02
    return 0.35

def posterior(months, sees_state):
    """months: list of (shift, change). If the state is not verifiable, the partner averages over it."""
    post = dict(PRIOR)
    for shift, change in months:
        for t in post:
            if sees_state:
                p = p_change(t, shift)
            else:
                p = 0.3 * p_change(t, True) + 0.7 * p_change(t, False)
            post[t] *= p if change else 1 - p
    z = sum(post.values())
    return {t: v / z for t, v in post.items()}

year = [(True, True)] * 4 + [(False, False)] * 8          # 4 shifts, 4 changes, all in shift months
odd = [(True, True)] * 3 + [(True, False)] + [(False, True)] + [(False, False)] * 7
cases = (("A  4 changes, reason not verifiable", year, False),
         ("B  4 changes, each with the agreed state indicator", year, True),
         ("C  as B, but one change came without the indicator", odd, True))
for label, months, sees in cases:
    post = posterior(months, sees)
    print(label)
    print("   " + "   ".join(f"{t} {PRIOR[t]:.2f} -> {post[t]:.2f}" for t in PRIOR))
# A  4 changes, reason not verifiable
#    committed 0.45 -> 0.01   adaptive 0.45 -> 0.81   opportunist 0.10 -> 0.18
# B  4 changes, each with the agreed state indicator
#    committed 0.45 -> 0.00   adaptive 0.45 -> 1.00   opportunist 0.10 -> 0.00
# C  as B, but one change came without the indicator
#    committed 0.45 -> 0.00   adaptive 0.45 -> 0.87   opportunist 0.10 -> 0.12

In case A the 4 changes almost rule out the committed type (0.45 falls to 0.01), but they raise the probability of the opportunist type from 0.10 to 0.18, nearly doubling it. The reason is not that the partner is suspicious; it is that, to an observer who cannot see reasons, the adaptive and the opportunist behave alike. In case B the same 4 changes become strong evidence that the team really is adapting to the market as agreed. Case C shows that this judgement tolerates error: one deviation brings the opportunist probability back only to 0.12 and does not make trust collapse.

This small model captures one of the book’s basic points: how the other side reads your change depends not only on what you changed, but on whether they can see the reason for it, and on whether that reason was agreed before the change. One reason high-context communication causes trouble in cross-border ventures may be exactly this: the reasons stay inside the team.

The limits of this chapter

Reducing disagreements in cross-border ventures to “Chinese flexibility versus Western contract-keeping” is an oversimplification this book tries to avoid. The countries of Southeast Asia differ greatly among themselves; many Southeast Asian partners are quite close to China on Hofstede’s dimensions, as Chapter 5 shows with data; and many local partners in Southeast Asia are themselves ethnic Chinese entrepreneurs, whose disagreements with Chinese teams are often not about culture but about information, incentives and institutions. The model in this chapter also describes only one mechanism: how missing verifiability makes adaptation look like betrayal. The next chapter turns to a more basic question: what is flexibility actually worth?