Japan vs. Korea: two developmental states
Two East Asian miracles, one model — or two? Set Japan and Korea side by side, and the comparison itself tells you which parts of the “developmental state” were the engine and which were the paint job.
The pilot agencies: MITI vs. the EPB
“The state led the market in critical ways. … The Japanese state’s priorities were developmental: the elite bureaucracy chose the industries to be developed, chose the best means of rapidly developing them, and supervised the competition in the designated strategic sectors.”
— Chalmers Johnson, MITI and the Japanese Miracle, 1982
Johnson invented the phrase “developmental state” to describe what he saw in Japan: not a planned economy, not a free market, but a third thing in which an elite bureaucracy steered private firms toward chosen industries. Every 2020s op-ed urging the West to “do industrial policy like East Asia” is borrowing Johnson’s thesis. So start where he started — and then add the country he never wrote about. Korea ran the same play, harder and faster, two decades later. Putting the two pilot agencies side by side is the only way to see what the developmental state actually is.
Both economies were doing the same thing in the abstract: catching up. A poor country converging toward the technological frontier can, in principle, grow faster than the frontier itself — it borrows proven technology rather than inventing it. The growth-theory baseline says convergence should happen on its own. The developmental-state claim is sharper: that the state could accelerate the convergence by steering capital into industries the market, left alone, would have reached more slowly or not at all. That is the claim the comparison has to test.
The apparatus has two parts. First, the late-development logic: the infant-industry argument that a new entrant facing established foreign competitors needs temporary shelter to survive its learning curve — the formal home of which is Ch 20 §20.8 (Contemporary Development), with the catch-up baseline in Ch 13 §13.6 (Convergence and Growth Accounting). Second, the state-capacity question: targeting only works if the bureaucracy doing it is coherent and relatively uncaptured rather than a vending machine for rents — the institutional-economics apparatus in Ch 18 §18.6 (Political Economy). Hold those two tools; the comparison uses them to ask not “did the state target?” but “what made targeting work in two places where it failed almost everywhere else?”
Two pilot agencies, each at its strongest
Japan — MITI as the coordinator
Take Johnson’s case at full strength. Japan’s Ministry of International Trade and Industry was a pilot agency, not a planning ministry. Through the 1950s and 1960s it controlled the levers that mattered most in a capital-starved economy: the allocation of scarce foreign exchange, the licensing of technology imports, and privileged access to cheap credit. It used those levers to steer private firms toward sectors it judged strategic — steel, shipbuilding, then automobiles and electronics — through what the Japanese call gyōsei shidō, administrative guidance. The crucial point, and the one that makes MITI sophisticated rather than dirigiste, is that it negotiated. The firms it was steering were already powerful incumbents with their own views; MITI cajoled, rationed, and brokered consensus rather than commanding. A reader sympathetic to Johnson should recognize this as the picture he drew: an elite bureaucracy and big business locked in a productive bargain, with the state setting direction and the market doing the work.
Korea — the EPB as the commander
Now Korea, at its own strongest — and not as Japan’s pale copy. When Park Chung-hee seized power in 1961 he built the Economic Planning Board as a super-ministry: budget authority, planning authority, and control over foreign capital concentrated in one place, its head a deputy prime minister. The EPB wrote five-year plans and drove the Heavy and Chemical Industrialization campaign of 1973–79 — the most aggressive targeting episode either country ever ran, conjuring POSCO’s steel, Hyundai’s shipyards, and a petrochemical complex more or less from nothing. The difference from MITI is not degree but kind: the Park state did not negotiate with strong incumbents because there were none. It had created the firms’ scale through credit allocation, and could therefore command where MITI could only coordinate. A reader sympathetic to Amsden and Wade should recognize this as the developmental state in its purest, most state-commanding form — the same apparatus as Japan’s, but with the power balance tilted decisively toward the state.
Here is the strongest case that the targeting was decisive, stated before we qualify it. POSCO became a world-class steelmaker after the World Bank refused to fund it on the grounds that Korea had no comparative advantage in steel. Japan’s automakers and electronics firms went from negligible to globally dominant under MITI’s steering. These were not obvious market outcomes; the agencies bet on industries the price signals of the day did not favor, and the bets paid off spectacularly. If you want to believe the state picked winners, this is the evidence — and it is real evidence, not a strawman.
The two ascents are visible side by side in the GDP-per-capita trajectories: Japan’s steep postwar climb from the 1950s, Korea’s even steeper compressed catch-up beginning in the 1960s, both pulling away from the developing-world pack. Open the GDP map on the East Asian trajectories to trace both lines against their peers; the year-event annotations mark the targeting episodes this stage describes. (The general question of why some countries make this climb and most do not is the work of Warum sind manche Länder reich und andere arm?; here we hold to the two cases that ran the developmental-state play.)
Targeting happened — but that is not yet the answer
So both states targeted sectors through a pilot agency, and the targeting unmistakably happened. That much is consensus. But notice what the side-by-side has already exposed: the two ran different power balances of the same apparatus — MITI negotiating with incumbents, the Park state creating and disciplining them. “The developmental state targeted industries” is true of both and tells you almost nothing about why it worked, because the sentence is equally true of Latin America, India, and much of post-colonial Africa, where targeting produced sheltered, inefficient firms that never grew up. The targeting is not the secret. Something else separated the two East Asian states from everyone else who tried the same thing — and finding that something is the whole game.
What Japan and Korea added was a test the protected firms had to pass. Subsidize a firm and shelter it from competition, and you have described both East Asia and the import-substitution failures — the difference is whether the firm faces a hard, external metric of whether the bet is paying off. Both East Asian states had that metric. It is the single most important thing in this entire story, and it is where the comparison turns next.
The test the firms had to pass
“In Korea, the government deliberately got the prices ‘wrong’ … But it disciplined subsidy recipients by imposing strict performance standards in exchange for the subsidies it distributed. The principle of reciprocity made all the difference.”
— Alice Amsden, Asia’s Next Giant: South Korea and Late Industrialization, 1989
Amsden’s book is the one that broke the comparison open. Her argument was not that Korea got the prices right — it deliberately distorted them — but that it disciplined the distortion. A subsidized firm had to deliver, and the metric of delivery was exports sold at world prices. This is the feature that the rest of the developing world, also subsidizing and protecting, did not have. If there is a transferable lesson in the whole developmental-state story, Amsden says, this is it.
The development-strategy choice of the postwar period came down to a fork. Import-substitution industrialization (ISI) built industries to serve the domestic market behind tariff walls; export-oriented industrialization (EOI) built them to sell abroad. The economics of the fork is in Ch 20 §20.8: orienting toward exports forces firm-level efficiency, because a firm that must sell into world markets at world prices cannot hide behind a protected captive market the way an ISI firm can. The open-economy mechanics — how export demand pulls an economy onto a different growth path — sit in Ch 17 §17.7 (Global Imbalances and Capital Flows).
Amsden’s “reciprocity” is not a named model in the textbook — it is the institutional move that turns export orientation from a market condition into a policy instrument. The state does not merely let firms export; it makes continued subsidy conditional on exporting. That converts the export market into a monitoring device: an uncorruptible, external test of whether each targeted bet is working, administered by foreign buyers the state cannot lean on. Hold that idea — subsidy made conditional on an external performance test — because it is what both East Asian cases share and what the failures lacked.
The same discipline, run two ways — and the foil that proves it
Korea — reciprocity as an explicit contract
Korea ran reciprocity at its most explicit, and this is Amsden’s central case. The Park state set export targets, monitored them firm by firm, and made continued access to subsidized credit, foreign exchange, and protection conditional on hitting them. A chaebol that met its targets was rewarded with more cheap credit to expand into the next sector; one that missed lost access. The discipline bit hard precisely because the export market is external and uncorruptible: a firm either sold abroad at world prices or it did not, and no amount of political connection could fake a foreign sales receipt. This is why Korea’s protected firms became globally competitive where the protected firms of Latin America did not — the export test culled the losers automatically, so the state did not have to be a perfect picker.
Japan — the same test, less contractual
Japan ran the same discipline in a less explicitly contractual register, and it would be a mistake to read its version as weaker. MITI tied access to scarce inputs to performance, and the sectors it targeted — autos, electronics, shipbuilding — were export sectors whose success was measured in world-market share, not domestic sales. Japan also ran the product-cycle move with discipline: it entered each sector at the point where it could compete on cost, proved itself in export markets, then climbed the value chain and vacated the low end to follower economies. The performance test was the same external one Korea used; Japan administered it through a negotiated relationship rather than an explicit target contract, but a firm that could not win abroad did not keep its privileged access. The discipline was real in both places. Its institutional form differed; its function did not.
Now the foil that makes reciprocity legible, drawn fairly. Latin American import substitution was not a stupid policy — it was a coherent response to a real structural worry that the periphery’s commodity-export position was a trap, and it built genuine industrial capacity for a time. What it lacked was the export test. Firms subsidized and protected to serve a captive domestic market had no external metric forcing efficiency; they sold behind tariff walls, and many never had a reason to become competitive. Same targeting, same subsidies, opposite outcome — and the variable that changed is the discipline. That contrast, not the targeting, is the comparison’s payload. The structuralist and dependency reasoning behind ISI, and why the mainstream verdict came down against it, is the spine of History of Economic Thought Ch.16 §16.3 (Prebisch-Singer, Dependency, and ISI); the heterodox-development lineage that Amsden and Wade extend — and its “getting prices wrong” quarrel with the neoclassical reading — runs through the same chapter.
The discipline is the transferable part
Export discipline is the load-bearing feature of the developmental-state template and the most-transferable lesson in it — this is the comparison’s central finding. It dissolves the picking-winners problem that Stage 1 left open: the state does not have to identify winners correctly, because the export test culls the losers for it. Both Japan and Korea had the mechanism; the import-substitution failures did not; the outcomes diverged accordingly. But notice the precondition already attached to even this most-portable feature. Reciprocity requires a state with the capacity to enforce the test — to actually cut off a politically connected firm that is failing — and a world market open enough to export into. A captured state cannot administer the discipline, and a closed global economy gives it nothing to test against. We will weigh both preconditions when we reach the transferability question. For now: the discipline travels, but it does not travel free.
The state targets sectors and disciplines them with the export test. But where did the money come from, and what kind of firm did the money build? Here, finally, the two cases diverge — and the divergence in what they built is what would later determine how each of them broke.
What the credit built: keiretsu vs. chaebol
In May 1961, within days of seizing power, Park Chung-hee had Korea’s leading businessmen arrested for “illicit wealth accumulation,” paraded through Seoul with placards reading “I am a corrupt swine.” Then he cut a deal: keep your fortunes, expand them — but build the industries the plan needs, on the state’s terms, or lose everything. The chaebol were not negotiated with. They were conscripted.
— The founding bargain of the Korean developmental state, 1961
That scene is the whole state-business power balance in one image. In Japan, MITI bargained with firms it had not created and could not simply command. In Korea, the state built the chaebol’s scale by deciding who got the cheap credit, and could discipline them from a position of dominance. Both states channeled the nation’s repressed savings into industry through directed credit — the same lever. But the lever built two very different kinds of firm, and the difference is where two developmental states stop looking like one model.
The financial machinery underneath both miracles was financial repression plus directed credit. Cap deposit rates below market levels and household savers earn little, but the captured savings can be channeled, at controlled rates, into whatever the state wants funded. The standard McKinnon-Shaw reading treats financial repression as a drag on development — it discourages saving and misallocates capital. East Asia inverted the conclusion: it used repression deliberately, as the pump that moved an unusually high household savings rate into targeted industrial investment. Whether the textbook owns a dedicated financial-repression section is a separate question — the corporate-governance and institutional-form apparatus this stage leans on is in Ch 18 §18.3 (Institutions and Development); the present-value-of-saving machinery that financial repression manipulates sits in Ch 24 §24.1 (Value Across Time).
The point to carry is that in both countries the credit-allocation system is the firm-structure mechanism. Control who gets the directed credit and you control which firms grow, how big, and in what shape. The bank-firm tie is a single lever seen from two ends — and Japan and Korea gripped it differently, which is why they ended up with structurally different firms.
Two solutions to one problem — neither the “right” one
Japan — the keiretsu and the patient main bank
The keiretsu was an elegant answer to the patient-capital problem. Each group clustered around a main bank that both lent to and held equity in its member firms, with the members cross-holding one another’s shares. Two things fell out of that structure. The cross-shareholding insulated managers from hostile takeover and short-term market pressure, giving them long horizons; and the main bank, holding both debt and equity, had the information and the incentive to monitor the firms closely. The state’s directed credit flowed through these banks, which did the granular monitoring the state could not do itself. For the heavy, long-gestation investments the development plan needed — steel, shipbuilding, where payback is measured in decades — this was close to ideal: patient capital with a competent monitor attached. Read on its own terms, the keiretsu is not a quaint relic; it is a coherent institutional solution to a hard problem.
Korea — the chaebol and the state credit spigot
The chaebol was an equally coherent answer to a different shape of problem — and emphatically not the keiretsu’s flawed cousin. Family-controlled, debt-financed, and diversified across wildly unrelated sectors (one chaebol might span shipbuilding, electronics, autos, construction, and insurance), it was funded not through equity-holding banks but through state-controlled banks acting as a credit spigot the government opened and closed. Family control gave fast, top-down decision-making; the state’s export discipline supplied the external accountability that a family firm’s ownership structure would otherwise lack. This structure was superbly suited to compressed development: a chaebol could storm into a new sector at scale, fast, because the state would fund the debt and the export test would referee the result. Where the keiretsu was built for patient depth, the chaebol was built for speed and reach — each fitted to the developmental tempo its country was running.
The state-business power balance is what the two firm-structures encode. The keiretsu were strong incumbents that Japan’s state had to coordinate with; the chaebol were state-built instruments the Korean state could command. Same lever — directed credit — pulled from opposite ends of the power relationship, producing a bank-centered patient structure in one country and a family-controlled high-debt structure in the other. The institutional-economics lineage that treats these as alternative coherent governance forms, rather than ranking them, runs through History of Economic Thought Ch.15 §15.4 (New Institutional Economics).
Mechanism is template; form is country-specific — and form encoded the fault line
Split the verdict cleanly. The directed-credit mechanism — financial repression channeling high savings into targeted investment — is part of the template; both countries used it, and it is consensus. The specific firm-structure form is country-specific: keiretsu and chaebol are two coherent solutions, not a better one and a worse one. And here is the finding that matters most: the divergence in form encoded a divergence in vulnerability. The keiretsu’s bank-centered patient capital was robust to short-term shocks but exposed to a slow asset-price deflation that would impair the banks at the system’s core. The chaebol’s high-debt, state-backed expansion was fast and scalable but exposed to a sudden stop in the external credit it ran on. The firms were built differently — and so, when the crises came, they broke differently. What they built determined how they broke.
Two developmental states, one core template, two firm-structures — and then both broke. Japan stagnated for two decades from the 1990s; Korea crashed hard in 1997 and recovered fast. Read as a natural experiment, how they broke and how they recovered is the test that finally tells us which parts of the template were robust and which were fragile.
The natural experiment, and the verdict
“Asian growth, like that of the Soviet Union in its high-growth era, seems to be driven by extraordinary growth in inputs like labor and capital rather than by gains in efficiency. … From the perspective of the year 2010, current projections of Asian supremacy extrapolated from recent trends may well look almost as silly as 1960s-vintage forecasts of Soviet industrial supremacy.”
— Paul Krugman, “The Myth of Asia’s Miracle,” Foreign Affairs, 1994
Krugman is the sharpest mainstream skeptic, and this is the right way to enter the verdict stage — because three years after he wrote it, the 1997 crisis hit, and he was widely read as prescient. His claim is serious: East Asian growth was perspiration, not inspiration — the mobilization of capital and labor, not rising efficiency — and input-driven growth eventually hits diminishing returns, just as the Soviet Union’s did. If he is right, there is no miracle to emulate and no template worth exporting. That claim has to be met at full strength before any verdict is earned.
This stage introduces no new apparatus; it aggregates the three axes already built and adds the machinery to read two crises. The 1997 Asian financial crisis was a sudden stop: the chaebol had borrowed short and in foreign currency to fund long, domestic investment, and when capital fled the region in 1997 the currency and maturity mismatches turned liquid firms insolvent overnight. Japan’s 1990s were the opposite shape — a slow domestic balance-sheet deflation as the late-1980s asset bubble unwound and impaired the keiretsu banks. The balance-sheet-recession apparatus that explains Japan’s lost decades in depth is the work of the sibling walkthrough What broke in 1990s Japan?; this stage reads Japan’s stagnation only as one side of the divergence.
Krugman’s claim has a formal form — Alwyn Young’s growth-accounting decomposition (1995), which attributed most East Asian output growth to measured input growth and comparatively little to total factor productivity. The growth-accounting method itself, and the residual it measures, is the apparatus in Ch 13 §13.6 (Convergence and Growth Accounting).
Growth accounting decomposes output growth into factor-input growth and a residual:
$$\frac{\dot Y}{Y} = \alpha\frac{\dot K}{K} + (1-\alpha)\frac{\dot L}{L} + \frac{\dot A}{A}$$where the residual $\dot A / A$ is total factor productivity. Young’s finding was that for the East Asian economies the first two terms — capital deepening and labor mobilization — explained most of the growth, leaving a modest TFP residual. Krugman’s “perspiration not inspiration” is exactly the claim that the residual was small.
You can grow by working smarter or by throwing in more workers and machines. Krugman’s claim is that East Asia mostly did the second — it mobilized an enormous amount of capital and labor and got the output you would expect from that, without much extra efficiency on top. The catch, which the verdict turns on, is whether mobilizing inputs at that scale is the boring part or whether it was the achievement.
Both broke — and broke differently
Japan — the lost decades
Japan’s asset bubble collapsed in 1990–91 and the economy slid into a deflationary stagnation that ran for two decades. Read at its strongest, this is the developmental-state-institutions-ossified story: the bank-centered patient-capital system that had funded the miracle became the zombie-bank system that prolonged the slump, rolling over loans to insolvent borrowers rather than recognizing losses; and the negotiated, consensus-bound state-business compact that built the boom now resisted the painful restructuring the bust demanded. The very features that made Japanese finance patient and stable in the ascent made it slow to clear bad debt in the descent. The deeper balance-sheet-recession mechanics belong to the sibling case; what matters here is the shape — a slow, internal, self-inflicted grind.
Korea — the 1997 crash and the fast recovery
Korea’s crisis was the opposite in tempo and reads two ways at once, both at strength. The chaebol-fragility reading: the high-debt, state-backed structure that powered compressed development was exactly the vulnerability — firms over-leveraged in short-term foreign borrowing, and the 1997 regional sudden stop left them insolvent, forcing an IMF program and a brutal recession. The developmental-state-robustness reading sits right beside it: the same state capacity that built the miracle executed the restructuring at speed. Korea returned to growth within roughly two years, restructured the chaebol under combined IMF and domestic pressure, and continued its climb to a high-income, technology-frontier economy. Korea did not get a lost decade. The structure was fragile to the shock that hit it; the state was robust enough to clear the wreckage fast.
Name the asymmetry honestly, because it limits what the experiment proves. Japan’s 1990s and Korea’s 1997 were not the same shock applied to two cases — they were different kinds of crisis: a slow domestic balance-sheet deflation versus a sharp external currency-and-maturity-mismatch sudden stop. So the natural experiment does not test robustness to an identical shock; it tests robustness to different shocks. That is weaker than a clean controlled comparison — but the informative finding survives it. The firm-structure divergence from Stage 3 predicted the kind of fragility each case displayed: the bank-centered keiretsu was felled by the slow bank-impairing deflation it was theoretically exposed to, and the high-debt chaebol was felled by the external sudden stop it was theoretically exposed to. The structures broke along exactly the fault lines their forms had drawn. The 1997 stress test, and the Asian-tiger export model it tested, is documented in History Ch.17 §17.5 (The 1997 Asian Financial Crisis); the Asian-tiger model that preceded China is §17.4.
Now Krugman and Young at full strength, because the skeptic deserves it. The growth-accounting critique is empirically serious: East Asian growth really was substantially input-driven — savings mobilized, labor pulled from farms into factories, capital deepened — and the measured TFP residual was more modest than “miracle” rhetoric implied. The 1997 crash looked like the predicted encounter with diminishing returns. A partisan of the growth-accounting view should recognize this as their argument, not a caricature of it. The divergence trajectories are visible in the lines: Japan flattening through the 1990s, Korea’s sharp 1997 dip and rapid rebound. Open the GDP map on the East Asian trajectories to see both the ascents and the post-crisis divergence in one view.
A real template, with preconditions that are hard to replicate
Aggregate the four axes. What is essential to the template: selective targeting, export discipline (the most-transferable feature), directed credit, close state-business coordination, and high state capacity underneath all of it. Both cases had all five. What is country-specific: the firm-structure form (keiretsu vs. chaebol), the state-business power balance (negotiate-with-incumbents vs. create-and-discipline), the timing (Japan’s earlier, base-supported ascent vs. Korea’s compressed catch-up two decades on), and the political context (postwar democracy under occupation vs. authoritarian developmentalism). Those varied, and the variation drove the divergent crisis vulnerabilities.
So is the template transferable? There is a real developmental-state template — but it ran on preconditions that are hard to replicate: a coherent, relatively uncaptured bureaucracy able to enforce the export test against politically connected firms; the Cold-War context that gave the exports an open US market, a security umbrella, and aid; a disciplined high-savings society to repress and channel; and a political window that let the state impose performance discipline on the firms it built. The template is real but it is not a recipe. A country with the apparatus and none of the preconditions does not get Korea — it gets failed industrial policy. The honest answer to “copy the developmental state” is: copy what, exactly, and into which preconditions?
The disagreement underneath is genuine, and worth naming at three layers rather than splitting the difference. At the frame layer, the 1990s “market-friendly” reading (the World Bank’s East Asian Miracle, 1993: fundamentals, human capital, and export orientation did the work; selective intervention was mostly neutral) and the “developmental-state” reading (Johnson, Amsden, Wade: the apparatus was load-bearing) were once genuinely distinct frames — but the modern consensus has substantially fused them, agreeing that export discipline and directed credit mattered and that they required state capacity and the geopolitical window. That frame gap is historically real but now narrowed; it is not two still-symmetric live options. At the method layer, Krugman and Young are right that the growth was input-driven — but mobilizing savings and labor into productive industrial investment at that scale and speed was itself the achievement. “Not a miracle” is a semantic deflation, not a refutation of the apparatus; the Soviet analogy fails precisely where it matters, because the Soviet input mobilization was not disciplined by an external export test and East Asia’s was. At the parameter layer, live empirical disputes remain — how large the targeting contribution was, how much of Korea’s 1997 was chaebol over-leverage versus external sudden-stop dynamics, how much of Japan’s stagnation was developmental-state ossification versus a balance-sheet recession any rich economy might have suffered. Those get settled with data, not rhetoric, and the modal estimates are adopted here without claiming the matter closed. The transferability lineage — Rodrik’s “one economics, many recipes,” the chastened-structuralist heir to this whole tradition — runs through History of Economic Thought Ch.16 §16.7 (the standing debate), with its current post-Washington-Consensus standing in Ch.17 §17.5 (the expanding frontier).
One model, or two?
Set side by side, the two miracles answer their own question. The four axes sort cleanly into shared template and country-specific contingency. Targeting happened in both, but targeting alone explained nothing — the same policy failed across Latin America, India, and Africa. Export discipline was the load-bearing transferable feature: the external test that culled the losers so the state did not have to pick winners perfectly. Directed credit was template; the firm-structure it built — Japan’s patient keiretsu, Korea’s high-debt chaebol — was country-specific, and that divergence in form pre-wrote the divergence in how each economy would later break. The post-crisis natural experiment, asymmetric as it was, confirmed the link: each structure failed along exactly the fault line its form had drawn, and Korea’s state capacity cleared its wreckage in two years where Japan’s ossified compact ground on for twenty.
So: one model, with two faces. There is a genuine developmental-state template — but it is not a recipe a finance ministry can download, because it ran on preconditions (state capacity, an open export market, a high-savings society, a political window for discipline) that most aspiring developers cannot assemble. That is why the question matters most where it is hardest to answer today. China’s state-capitalist model is the live contemporary heir to everything in this comparison, and whether it is a developmental state — and whether anyone can copy it — is the form the transferability stakes now take. The sibling comparative on planned-versus-reform paths, China vs. the USSR: two paths through planning, owns the China depth; here the comparison rests where it began — with two East Asian states that looked like one model from a distance, and turned out to share an engine and diverge on everything bolted to it.