← How an Economy Grows and Why It Crashes
第 4 章 · 18 分钟
Specialisation and trade: comparative advantage and its edge
One identity, and where it is taken for an empirical finding.
概念地图
Gains from trade
Reallocating tasks by opportunity cost raises total output from the same inputs.
深色为本章概念，浅色为其他章
陪练模式
不想只是读？让它带你走一遍。
导师把这一章拆成小步，每一步先问你，再讲；你用自己的话答，它先认下对的部分再纠偏。用你自己的 AI，对话只存在这台设备上。
Locus01
延伸阅读
Ricardo · 1817 · Principles of Political Economy, chapter 7
Gave comparative advantage through the Portugal–England example: even when one side is more efficient at both goods, specialising by opportunity cost leaves both better off. He argued against grounding the gains from trade in absolute efficiency; the cost is that the model has two countries, two goods and one factor, and every conclusion lives inside that shell.
Foundational
Heckscher and Ohlin · 1920s–30s
Moved the source of comparative advantage from technology to factor endowments, and so derived conclusions about how trade changes the returns to factors at home. This put the distributional consequences of trade inside the model for the first time.
Turn
Prebisch and Singer · mid twentieth century
Argued that the terms of trade of primary exporters deteriorate over the long run, so static comparative advantage can lock a country into a low-value position. This is the longest-standing counterexample line against the Ricardian frame.
Counterexample
Schiff · 2010 · the chapters on trade between islands
Uses comparative advantage to explain specialisation between islands and treats the gains as self-evident. The book does not separate 'the total gain is positive' from 'everyone is better off'; the first crack lands here.
Subject of this course
机制02
Ranking by opportunity cost
One islander fishes fast and weaves slowly, another is slightly slower at both. Let each do the thing he gives up least by doing, and total output rises. What is doing the work is not who is stronger but what each gives up — which is what comparative advantage names. It requires no absolute superiority on either side, only that the two rank their opportunity costs differently.
机制Differences in opportunity cost, through a reallocation of tasks, raise total output from the same inputs.
可迁移性测试
Move it to a clinic: the attending punctures twice as fast as the resident, but his opportunity cost is a clinic slot, so the puncture goes to the resident and total patients seen rises. The isomorphism breaks in that medicine has a quality floor below which an efficiency gain is no gain.
机制03
Lock-in of the division
Once tasks are split by today's opportunity costs, whoever does a thing gets better at it, the costs diverge further, and the split hardens. The Prebisch line pointed out that this runs both ways: it can lock one side into a high-value position and the other into a low-value one. This is where the gains from trade meet the question of distribution — a larger total and everyone being better off are two different claims, and the book does not separate them.
机制An existing division of labour, through accumulated practice, widens the opportunity-cost gap further.
可迁移性测试
Move it to an engineering team: whoever first took over the deploy scripts writes them ever more fluently, until nobody else wants to or can, and the team's options narrow rather than widen. The isomorphism breaks in that a team can rotate people deliberately to break the loop, while countries have no such scheduler.
Derivation04
From differing opportunity costs to 'both sides gain'
What is being derived is the full form of Ricardo's conclusion, and what exactly it does and does not promise.
Two parties, two goods, labour the only input
Chosen, and the boundary of Ricardo's own model. With several factors the conclusion has to be rewritten.
Each party's opportunity costs are unchanged by specialising
Chosen: it assumes no scale effects and no learning. The second mechanism above violates exactly this.
The exchange ratio falls between the two opportunity costs
Chosen: it assumes the bargain does not land on an endpoint, where one side takes the whole gain.
推导 · 0 / 4
Worked derivation05
A larger total is not everyone better off
After island A specialises in nets and island B in fish, combined output rises by ten per cent. Does it follow that everyone on island B is better off?
Break the island into people, then look at whose opportunity cost changed.
裂缝06
Cracks in this framework
争议地形07
本书主张
Specialising by comparative advantage and exchanging freely leaves every participant with more; tariffs and quotas that obstruct it only make everyone poorer.
另一种看法
The structuralist and infant-industry line holds that static comparative advantage locks a country into low-value production, and that selective protection can buy a better long-run position.
分歧扎在
The disagreement is rooted in whether opportunity costs are exogenous. The book takes them as given by endowments and unchanged by specialisation; the other side takes them as a product of specialisation, so today's optimum need not be the long-run one.
什么证据能裁决
What would settle it is performance after protection ends: compare productivity growth in protected industries once protection is removed with comparable unprotected industries over the same period. Systematically higher growth afterwards would show opportunity costs can be moved by policy; equal or lower growth would show protection transferred rather than created.
The evidence points to 'most protection produced no such effect, with a few exceptions'. What is missing is an ex ante criterion separating the exceptions from the failures — nearly all existing work picks successful cases after the fact and then summarises their features, which is the degrees-of-freedom pump chapter 1 described.
Falsification08
Lock-in of the division as a preregistrable test
The mechanism under test: an existing division of labour, through accumulated practice, widens the opportunity-cost gap further.
Data source: product-level export data from UN COMTRADE, with a product complexity index
Product-level data show the composition of the basket changing, not just its total.
Sample: every economy with thirty unbroken years of product-level exports, resource exporters included
Excluding resource exporters would remove the strongest lock-in cases, erasing the very phenomenon under test.
Window: decade blocks, comparing the basket structure of consecutive blocks
Lock-in is a slow variable; short windows show only price swings.
Threshold: the more concentrated the initial basket in low-complexity products, the slower complexity rises over the next decade — a significantly negative coefficient
Registered on the direct meaning of the lock-in hypothesis.
Failure condition: the coefficient is insignificant, positive, or holds only once some period is excluded
The third counts as failure.
推导 · 0 / 3
接口09
Which slot it hangs on
挂在哪个槽位The conclusion you already hold is probably 'focus on what you are best at' — a judgement about allocating your own effort. This chapter goes at its criterion.
Does this chapter (a) replace 'do what you are best at', or (b) constrain it by swapping 'best at' for 'give up least by doing'?
慢变量Register two slow variables: the opportunity-cost ranking of the tasks on your desk, and how many new capabilities you have picked up in two years. Look at the second first — near zero means lock-in has already started.
小结10
本章小结
01Comparative advantage turns on differing opportunity-cost rankings, not on who is absolutely stronger.
02'Total output does not fall' and 'an interior price leaves neither side worse off' are both identities.
03Only 'real incomes rise after opening' carries empirical risk.
04A larger total does not entail everyone better off; compensation is a separate political decision.
05If opportunity costs move with specialisation, the static optimum can foreclose long-run options.
提取练习 · 合上书，先自己答一遍。
?Is 'combined output does not fall after specialising by opportunity cost' (a) an identity or (b) an empirical claim?
?Does the infant-industry argument attack (a) assumption one (two goods, one factor) or (b) assumption two (opportunity costs unchanged)?
Forced choice11
A · Identify the mechanism
Which of these is a mechanism?
二选一
Forced choice12
C · Locate the crack
In which situation does this framework give a confident and wrong answer?
二选一
读到这里，把它记为已读。