Hidden Rules of Development

What Botswana’s Success Story Never Had to Explain

Stephen Brien — Tue, 01 Sep 2026 07:24:09 GMT

The short version

Botswana grew faster and longer than almost any economy on earth, saved its windfalls, and built one of Africa’s most competent, least corrupt states.
Capital and skilled labour compounded at nearly ten per cent a year. How much better Botswana got at using them turned negative, and has been fluctuating ever since.
In 2024, growth turned negative, and the party that had ruled uninterrupted for fifty-eight years lost power that same year.
Was the money behind Botswana’s priority ever at risk, judged by an outsider, or actually stoppable if the answer came back no?
Was the declared goal, repeated for decades, ever what the state was actually organised around? Neither question has a settled answer yet.

This post is the first of a series on Botswana. It follows an examination of eight economies split by whether their firms were ever put to a market test with the power to fail them, four that renewed, four that stayed inert, and a ninth, Japan, where that same test read a coming stagnation two decades before the numbers did. Botswana breaks the pattern from the other side. By every mark that separated the renewers from the rest, it should belong with the four that renewed. It never did, and nothing was ever built with the power to make a wrong answer cost it anything.

In 1973, in the eastern half of Botswana, a state-backed venture began extracting copper and nickel from the ground at Selebi-Phikwe. A town was planned around it from the start, meant to outgrow the mine that founded it and become self-sustaining in its own right. The town grew into the country’s third-largest urban centre. But it never became self-sustaining, and nobody ever planned for what actually happened instead.

The mine lost money almost immediately. By 1982, nine years in, the losses had reached a quarter of a billion US dollars. Three times over in those nine years, the government’s response was never to ask whether the venture should continue.

A single loss can be a shock. Three rescues in nine years can’t be. The government wasn’t deciding anything new by the second or third time. It was repeating a decision already made, without ever asking again whether it was still right.

Each time, the government found a way to keep the mine going without ever asking whether it should. It deferred the royalties it was owed (payments for the minerals it took out), rebated the company’s power and water charges, and turned most of the senior debt into notes, a form of IOU carrying no maturity date at all. A fourth rescue, in 2014, did the same. A further 3.3 billion pula owed to the government was mostly converted into equity, ownership shares in the company, and the rest paid in cash. The government then declared it, on paper, debt-free.

The mine closed in October 2016. Cabinet decided it should close. In the end, the court that formally shut it down only delivered what the government had already decided for itself.

More than six thousand people lost their jobs at the mine alone. Thirty-three thousand people had built a town around a purpose that no longer existed. The government’s own explanation was that reviving the venture would now cost close to five billion pula. A commission of inquiry convened in 2025 is still examining whether the closure was justified. None of that cost was carried by anyone who had a hand in the decision.

Forty-three years in operation. At least four rescues. One closure.

That is what this government’s money looked like when nothing was ever asked of it in return.

An asset, a place, a payroll, kept going for four decades past the point anyone could say why.

This isn’t the Botswana most people picture. The image is diamonds, run cleanly by Debswana for fifty years, not a copper-nickel mine that never turned a profit. Selebi-Phikwe makes visible a pattern in Botswana’s history that usually runs through money spent too diffusely to see, a tax allowance here, an easy loan there, an allocation renewed quietly year after year.

What follows is the rest of the record: the credentials Botswana earned by every conventional measure, the number that says growth never actually made the country better at using what it had, three more measures that converge on the same answer, and the question this post keeps coming back to about whether any of it was ever made to cost anyone anything.

I: It earned every credential a success story needs

By every measure economists use to name a success story, Botswana was one.

By any conventional measure, it succeeded

By every conventional measure of development success, Botswana had already earned its credentials. Whatever else is true of it, failure isn’t the word anyone uses.

At independence in 1966, it was among the poorest countries on earth. What followed was one of the fastest sustained growth records anywhere. Among the World Bank Growth Commission’s list of successfully growing economies, it is the only resource-dependent economy whose growth period persisted.

None of the diamond windfall was squandered the way windfalls usually are. Botswana ran budget surpluses through most of the 1980s and 1990s. It built a currency, a central bank, and a planning system that other finance ministries came to study. Its flagship diamond venture has been run cleanly for fifty years, with no domestic partner holding a stake and no sign of insiders diverting it to their own benefit. One independent assessment, in 2010, called Botswana the least corrupt country in Africa.

What secured property rights explain, and what they don’t

Few small economies have been studied this closely. Daron Acemoglu, Simon Johnson, and James Robinson (AJR) made this exact case. Their detailed argument is that Botswana’s institutions, which secure property rights and constrain its political elite, explain why good policy was chosen and sustained. In their account, those institutions let ordinary people, not just a narrow elite, get ahead. Their own account never hid the rest of the picture. Inequality stayed extreme for decades. The ruling party faced no credible opposition for fifty-eight years. Its own political style, they said themselves, was soft authoritarian.

Subscribe now

AJR’s own account settles the credentials question convincingly. It was never trying to settle a narrower one. Did that concentration of power ever bend what the money was actually for, not just whether it ever cost anyone anything? None of that concentration of power changed their verdict, because their own standard was never fairness. It was whether a state preys on its own economy or protects it.

Botswana passed. It had the capital, and the deliberately built skills to put it to use, the discipline demonstrated for decades, and, by any conventional measure, the state. The elite behind it was disciplined.

What the elite’s discipline was never required to prove is the harder question.

II: The number that says it never transformed

Four different measures point the same way, and the finding survives its two most obvious challenges.

Productivity turned negative in 1990, and never recovered

Botswana had every conventional ingredient for transformation, and still never transformed. Its growth increasingly came from adding more capital and labour, rather than from getting better at using what it already had.

Total factor productivity growth was positive from 1975 to 1990, even as capital compounded at 9.1 per cent a year and the educated workforce at 9.6. Then it turned negative, and it has been fluctuating rather than recovering ever since. World Bank growth accounting for the years since 2002 finds it turning negative again through the 2000s, recovering to as much as a third of growth in the early 2010s, then fading to a tenth of it by the late 2010s.^[1] For thirty-five years now, the story has been the same. Capital and labour kept compounding. The underlying ability to use them better did not.

Everything went in. This country did not get better at making things.

This isn’t a failure to grow. It’s a government that added more without ever getting better at using it.

That same conclusion shows up in what the economy actually makes, not only in how efficiently it makes it. Manufacturing’s share of national output rose through the middle decades and settled back to roughly three per cent. That’s almost exactly where it stood at independence, even as almost everything else about the country changed.

That is the puzzle. Slow growth isn’t the problem. Botswana grew about as fast as any economy in the world. Underinvestment isn’t the problem either. It invested at a rate most finance ministers would envy.

What does it actually take to turn investment into an economy that keeps finding new things to be good at? And why did a government this capable, this disciplined, and this well-resourced never manage it?

Complexity and innovation rankings say the same thing

Botswana’s ranking on the Harvard Growth Lab’s index of economic complexity, a measure of the diversity of an economy’s exports, fell from seventy-fourth in 2012 to ninety-third in 2024. Further evidence of an economy not transforming. A different measure, the same answer as total factor productivity’s zero.

The blunter number tells the same story. Diamonds still account for eighty-five per cent of everything Botswana sells abroad, and half of everything the government collects in revenue.

Invention gives a starker number still. The Global Innovation Index ranks 132 economies each year on how much they invest in innovation and what they get out of it. Botswana’s input rank, 74th, lands almost exactly where the complexity index placed it, on a completely different measure. On the output side, it falls to 94th. Its own report card states the finding in plain language. It produces less than its investment would predict. It falls below what a country at its income level should manage.

A shortfall of return, not of ambition.

Debswana, responsible for most of what Botswana sells abroad, has never had to clear either the complexity ranking or the innovation ranking. It has been a joint venture with the same foreign partner since the start, never required to win a market on its own. Fifty years on, no other company based in Botswana has come anywhere near competing among the world’s most advanced firms.

Four measures now, each built for a different purpose, by people who weren’t thinking about Botswana at all: total factor productivity, these two indices, and Debswana’s own market test, sitting inside the country’s own largest company.

Neither Botswana’s size nor a passing resource shift explains the finding away

Four separate observations point the same way, though not through any method validated across many countries. Some of them are structurally biased toward a small, resource-heavy economy scoring modestly, regardless of how well it’s governed. Total factor productivity isn’t: it has no relationship to a country’s size, so a small economy can’t explain why it reads zero.

Seen against other small, resource-dependent economies, the complexity finding holds up. Botswana’s rank, 93rd of 145 in 2024, sits just ahead of Namibia, Mongolia, and Gabon, all comparably sized and resource-dependent, but nowhere near Slovenia, Estonia, or Lithuania, small economies that actually diversified and sit inside the top thirty. Diversifying away from a resource base may simply be harder for a small economy to do, whatever its government looks like.

Subscribe now

On innovation, size alone doesn’t sort these economies into one group. Mongolia’s output rank has stayed ahead of Botswana’s, while investing less to get there. Something other than size has held Botswana back.

TFP’s own number could still be an artefact of a different kind, a one-off productivity gain from shifting into a newly discovered resource, fading once the shift completes. That doesn’t explain a rate that stayed negative for over a decade.

However it’s measured, and however closely the objections are examined, the lack of transformation does not go away.

President Masisi said as much himself. In a May 2021 address after a Cabinet and BDP retreat, he tied Botswana’s high post-independence growth to public-service quality, and its far lower growth since the 1990s to its decline, citing a “litany of unfinished or poorly executed projects.” He said Botswana could not “remain locked in the thinking of the unproductive eras and expect to escape the middle income trap.” Botswana’s own head of state had named the stall.

III: None of the ordinary explanations survive Botswana’s own record

Not one of the ordinary explanations for a lack of transformation, a missing plan, a missing capacity, the resource curse, a lack of nerve, survives against Botswana’s own record.

The usual explanations, and why Botswana’s own literature already knows them

Development economics has a standard toolkit for why a growing economy fails to transform. A government without a plan. A state too weak to carry one out. A windfall that becomes a curse, financing everything except the will to change. An elite too cautious to spend its own political capital. Botswana’s own literature has already tested versions of every one of them.

Samatar credits the same discipline AJR does, a cattle-owning elite that chose developmental leadership over its own short-term gain. Hillbom’s own account goes further, denying Botswana developmental-state status altogether, on the grounds that it never built the diversified private sector or industrial strategy that model requires. Her own label for it, a development-oriented gate-keeping state, describes an economy the state sector dominates and a resource-dependent elite controls, not one built to transform itself.

Before this account can say what actually explains the absence, it has to clear the same four explanations any development economist would raise first.

It had a plan, and the capacity to act on one

If any of the rival explanations that follow holds, the four measures of absence above are explained away rather than diagnosed.

Botswana never lacked a plan. Every development plan from the 1970s onward named diversification as an objective, several of them as the objective. Nor was the state too small to act on one. It owned the country’s largest manufacturing enterprise, a development bank, a development corporation, and a housing corporation. It has planned in five-year cycles since before independence.

Three separate programmes were built specifically to diversify the economy away from diamonds, the National Development Bank from 1965, the Botswana Development Corporation from 1970, and the Financial Assistance Policy from 1982 to 2001. None of them did.

And it wasn’t for lack of an easy environment to start one. In 2010, Botswana ranked among the most economically free countries in Africa, ahead of most of its East Asian comparators. That freedom never produced the diversification it was supposed to invite.

Whether it could build something was never the missing thing. What it was ever asked to build is a different question. And this isn’t, in any simple sense, the resource curse, a windfall spent in ways that let corruption or debt take hold instead of forcing hard choices.

The windfall wasn’t squandered, and this isn’t a lack of nerve

Botswana kept its exchange rate honest for decades, ran surpluses instead of spending its windfalls as they arrived, and came through crashes in commodity prices that flattened its neighbours.

But discipline only rules out part of the story. Capacity is the part discipline alone doesn’t touch. What eased was the pressure to ever use it. Even the literature most sympathetic to state-led development agrees that capacity was never enough by itself.^[2] Nor is this a government without nerve. In 1967 it vested its own president’s tribe’s mineral rights in the new state, a transfer of value away from where it could easily have stayed, not a favour kept for himself.

This state was not sitting still. Across sixty years it kept trying something different, and kept adapting what it tried. What it never did was let anyone with the standing to say no check whether it had worked.

That accounts for four rivals, and not one of them is still standing.

Each of those deserves a full hearing, and each gets one later in this series. So does the question of whether the HIV epidemic distorts the productivity numbers themselves, not only the choices behind them.

Once they are properly cleared away, what survives isn’t another circumstance. Nor is it the question a government this honest and this capable usually gets asked, whether it did enough.

It’s a narrower one. That question can’t be answered by a plan or a speech.

A declared priority costs nothing to state and everything to enforce.

The only place left to look is what standard the money's recipients actually had to meet, a standard that shifted five times across sixty years, not the fixed market test Korea, Taiwan, and Singapore's own firms faced throughout.

Subscribe now

IV: No wrong answer was ever allowed to cost anyone anything

For most of its history since independence, the government could change what it was doing without ever facing an independent verdict that forced either its choices or the government itself to change. That long insulation is now being tested.

Selebi-Phikwe was judged only by the people who paid for it

The government financed the mine and decided what counted as reason enough to keep it going. It kept the mine going for four decades without ever facing a ruling from outside itself that could have triggered an end to the support. That’s what it looks like when the government never had to face an independent verdict on its own decision. It is an umpire who also owns one of the teams, ruling that his own team played fair.

That pattern isn’t a rule yet. It’s closer to a shape worth carrying loosely into what follows, one that only needs to sharpen where a period’s own evidence asks for it.

Was the money ever put to a real reckoning for what it produced? Did anyone with the standing to judge it ever get the chance? At Selebi-Phikwe, the honest answer to both was no.

A declared goal, and five postures that moved past it

There’s a second question behind whether Selebi-Phikwe’s money was ever reckoned with, less about any single decision than about the whole sixty years. Every plan since the 1970s has declared economic transformation the goal. Did that goal ever actually organise what the government did with its money, or only what it said, a gap between the two?

National Development Plan III put diversification and jobs at the centre of the government’s own strategy, and the state built real machinery around that goal, feasibility studies, agencies to promote investment and exports. Over the decades, what the money was actually organised around moved through five different postures one after another: a grant, then a resource rent (income from owning the resource, not producing it), then employment, then facilitation, then an unwinding.

By 2010, two in five formal jobs (registered, salaried jobs, as opposed to informal work) in the country were in government itself. That was not what transformation looks like by anyone’s definition, but perhaps closer to what a durable political coalition looks like when the state is what holds it together.^[3] The same elite that ran the economy also ran the government, without ever facing an organised rival’s electoral verdict. Even so, this single number can’t settle, on its own, whether a payroll that size was one of the things the state’s five shifting postures were actually for, or only their side effect.

Nothing ever forced that gap into the open

Nothing in the record examined so far shows a moment where an adverse verdict on one of those five postures triggered the government to change course, a programme shut down on someone else’s ruling, a strategy publicly dropped, someone held to account. That’s an absence in the record examined, not proof that no such moment, or response, ever occurred. Selebi-Phikwe doesn’t count as one of these moments either: it closed on the government’s own verdict, not anyone else’s.

Whatever this pattern is called, it isn’t cultural persistence, an old habit nobody actively chose. It’s what a declared goal looks like when nothing forces a government to test it against what it’s actually doing.

The same absence held in politics. Botswana held real, competitive elections, and the ruling party could, in principle, have lost any one of them. For fifty-eight years, it didn’t. For nearly six decades, the government never had to face an electoral verdict that could have triggered a change of government, or made a wrong decision cost it anything.

The insulation is now being tested

None of this is settled history. The two conditions that let the government avoid a real reckoning for sixty years, cheap growth and an unbeaten position at the polls, have just weakened at the same time.

National income grew 11.9 per cent in 2021, then 5.5, then 3.2, before contracting by 3.0 per cent in 2024. Mineral revenue fell from thirteen per cent of national income to roughly seven between 2022 and 2024, over the same stretch. The pressure that revenue had been holding off is arriving late, not never.

That same year also ended the ruling party’s run. For the first time since independence, an electoral verdict actually went against it.

Both cushions gave out in the same year.

Whether the two arrived together by coincidence or because of the same thing underneath both is not yet answerable. But together they close off the easiest answer of all, that this insulation is safely in the past.

The UDC coalition government that inherited this has already made its own declarations. It launched an Economic Transformation Program in mid-2025, naming services, manufacturing, tourism, and renewable energy as the sectors to build. A five-year, hundreds-of-billion-pula spending plan backs it. Most of that reads like the same pattern: a named goal, a plan, more machinery.

One piece is genuinely different, if it happens rather than stays declared. The government has pledged to break up the state power utility into separate generation, transmission, and distribution businesses, open to private investment. It is the first proposal in this record to expose a piece of the state’s own economy to competitors it doesn’t control. Whether that actually happens, or becomes one more exception managed on the state’s own terms, isn’t answerable yet either.

That doesn’t prove anything about what comes next. It’s the first real test of whether the government can keep operating this way without either cushion. The posts that follow put some version of that question to each period of Botswana’s history in turn.

Subscribe now

What follows

In the coming posts, each period in Botswana’s history will receive its own analysis, and the series will close by asking what it all adds up to.

They cover founding decisions, before there was a country or any rent to make it easy. What a government does with money once its own reason for raising it is satisfied. Fifteen years spending more on industry than ever before or since, paying for jobs not results. A government that wrote its own passivity into doctrine. And what happens to a state’s discipline once its cushion runs out.

Any one of these posts alone reads like competent government meeting a hard constraint, not a government hiding its goal. The sequence, across sixty years, look like something else, a government that never stopped declaring transformation as its goal, and never let its real priority be tested against that declaration, whatever it actually was in a given decade.

The series will conclude by taking rival accounts seriously, including the simplest, that prices and incentives explain everything, and testing what survives. And a closing post asking what sixty years add up to, and what it still costs the country.

The next post in the series goes back to 1963, three years before independence, to find the pattern already forming.

The production of this post has been AI-assisted. For more details, see the About page.

Subscribe now

Japan's Lost Decade Was Twenty Years In The Making

Stephen Brien — Fri, 14 Aug 2026 07:30:41 GMT

The short version

Japan ran one of the most transformational governing orientations on record. Then, in a budget signed off in January 1973, it shifted.
Almost nothing else changed. The ruling party stayed in office; the famous economic bureaucracy stayed intact and kept intervening exactly as before.
The power and the machinery held. Only what it was made to protect changed.
The reckoning, a productive system that kept performing but stopped renewing itself, did not surface until the 1990s, twenty years on.
Fifty years later, Japan still has not put a new productive direction back at the top of what it protects.

This post continues the series analysing the varied fortunes of the World Bank’s own Growth Commission countries. So far the focus has been on four that ran a transformative governing orientation and kept climbing in the decades since the Commission’s own assessment, and four that never exhibited that orientation and where inertia took over. This post examines Japan, another of the Commission’s success stories,^[1] and looks at how its governing orientation shifted long before its growth trajectory stalled.

On 15 January 1973, officials in Tokyo signed off on a budget already three weeks behind the government’s own schedule. The general election that December had just returned Tanaka Kakuei to office; his second cabinet was not yet a month old, and a fight inside the government over the year’s pension increase, health-insurance financing and teacher pay ran past the normal December deadline and into a final round of negotiation in mid-January before anyone would sign.

The economy around them was still growing at close to ten per cent a year.^[2] The figure they set for social security was up almost twenty-nine per cent on the year before, the largest jump in a decade, and it carried the total past two trillion yen for the first time. A standard reference account of the ministry’s postwar role describes exactly the power that the budget put to use: the Ministry of Finance, it records, “regulated the allocation of resources in the economy, affected the distribution of income and wealth among the citizenry... and promoted economic growth and welfare.”^[3] The January 1973 budget is where that power tipped toward the distribution half of that sentence.

This was nine months before the oil embargo that is often remembered as the end of the Japanese miracle.^[4] The embargo was a second test of the state’s own commitment, and Japan met it by absorbing the shock into a budget already committed to widening its focus, not by reconsolidating around production. It was years before growth visibly fell. And it is the day a state’s priorities can be seen shifting, independent of how the 1990s story would later turn out.

A government with a growth record like post-war Japan’s rarely gets questioned again. However, growth shows an economy kept moving, not what a government chose to protect once real trade-offs began to bite.

This account traces three things in sequence: the governing orientation Japan actually ran under before 1973, disciplined enough to help earn Japan its place on that list; the single budget when that orientation noticeably changed; and the two decades it took to reach anything a growth chart would show.

Subscribe now

I: What Japan built

What characterised Japan’s early orientation was a deliberate, sustained choice to put transformation ahead of every rival claim on it, held to even when relenting would have been easier.

Japan enacted that transformational priority for two decades, from the early 1950s to January 1973, in three ways: governing terms set in advance, conflicts resolved the same way regardless of who lost, and external commitments that made a later retreat costly, not just unlikely.

The early priority, sector by sector

That governing orientation actually starts showing up in the 1950s. Steel and shipbuilding carried it first, built on rationed dollars and licensed technology, sent to win the same world orders as anyone else. In December 1960, the cabinet staked itself on the Income Doubling Plan, a promise to double national output within ten years.^[5]

A promise costs nothing. What made Japan’s direction developmental was where the scarce yen was directed when two claimants pulled. It went to the one that could win abroad, against a verdict no ministry could soften, while the one that could not was allowed to shrink. A governing circuit, trade-ministry bureaus and development-bank credit, carried that discipline into the economy, sector after sector.

The priority moved by sector as the decades passed, and the test got harder. Through the 1960s it fell on the carmakers, where the state’s own protective instinct was tested directly. MITI’s own bid to merge them into two or three sheltered national champions failed in the Diet ^[6], so Toyota and Nissan went on competing against each other at home, unsheltered, and later won share in America’s market on the strength that competition had built.

Had the shelter gone through, Japan’s carmakers might owe their strength to protection rather than to a market test they could not dodge.^[7]

By the 1970s, the test got harder still: the market test fell on a product with nowhere to hide at all, the semiconductor, sold from the start into a single world market.

While the orientation was live, a new sector took this unrigged test about once a decade: steel and shipbuilding in the 1950s, cars in the 1960s, semiconductors in the 1970s. The same rule was applied each time, regardless of which sector it fell on.

The state did not block competition once a sector reached the market. That doesn’t prove the earlier rationing, the dollars and technology that got steel and shipbuilding there in the first place, was itself efficient. A more open capital market might have reached the same result with less waste; nothing here rules that out.

The rule that cut the other way

The same rule cut the other way too: not just tolerating new rivals, but letting the state’s own past investment fail. That is where applying it cost something real. A sector that had once been a state priority was not spared the same verdict once it stopped winning; the transformational orientation held, meaning the state even abandoned what it had itself built up.

Coal shows this most clearly. A 1955 rationalisation act started sorting mines into those still worth state investment and those marked for closure. The miners were not a constituency the state could safely ignore: their union’s political tenacity was real, and avoiding the social unrest a wholesale abandonment might have triggered was itself part of why the state built such an extensive support system for displaced workers. What it did not do, even so, is protect the mines themselves once they failed the test. Textiles lost on the same terms.

It costs a state nothing to step back and let a new sector prove itself. It costs something real to walk away from a sector the state already backed, and coal and textiles are where Japan paid that price.

A founding conflict fought by someone else

Japan didn’t have to break its own elite to prioritise transformation and expose it to a market it could not rig. The reason matters, because who did the breaking decides whether it reflects Japan’s own choices at all. Korea broke its tycoons; Taiwan stripped its industrialists of their rents. Japan’s powerful pre-war interests had already been broken by a foreign hand. The American occupation dissolved the zaibatsu holding companies from 1946 and redistributed roughly a third of the farmland before the developmental state got going.^[8] This is a founding conflict: an inherited privilege curtailed before new terms can even operate, performed by someone else.

Inheriting the breaking proves nothing on its own about the government that inherits it. What demonstrates the orientation is what came after, when reversing course was still a live option.^[9]

A priority Tokyo chose to keep enforcing

The occupation had dissolved Japan’s zaibatsu in 1946. It left the banks standing, though, and around them the broken zaibatsu regrew through the 1950s into a new form, the keiretsu. The breaking was the occupation’s doing. The regrowth, and the terms it happened under, were Tokyo’s own choice.

Once the occupation ended and the keiretsu had regrown, Tokyo could have treated them the way the old zaibatsu had been treated before the war: backed no matter what they did. However, it did not. A keiretsu firm’s position stayed conditional on continuing to perform, not on who it already was.^[10] Firms could import only what their exports paid for; when a firm faltered badly enough for its bank to step in, the price was never only money.^[11]

Coal already illustrated this rule once, as a sector the state simply let fail through ordinary policy. This time it returned as a fight. The Mitsui Miike coal strike of 1960, the largest labour dispute of the postwar era, was a lockout Mitsui Mining imposed on miners resisting that same rationalisation. More than 30,000 miners held out for 312 days, backed by one of Japan’s most powerful unions.^[12]^[13] Management and the state prevailed anyway.

This was genuinely contested: the state had to win a real fight to make the same rule hold. The pattern is a government enforcing its own terms, in a transformational conflict of its own making, years after anyone could have made it do so. The same choice, in its harshest form.

Bank rescues carried that same discipline into finance: a firm’s claim on capital held only as long as it could still perform, not because of what it already was. Yamaichi, one of Japan’s four largest securities houses, came close to failing in 1964. Its own banks set the price first. By May 1965, they had forced through internal reconstruction, branch and staff cuts, a tighter grip on costs, not a blank cheque. The state’s own backstop followed only once those terms already held.^[14]

A decade later, cars came up again: Sumitomo Bank moved into the failing carmaker Mazda not with a takeover but with its own executives installed inside the company, an arrangement the sitting president could not simply outlast.

Yamaichi and Mazda were exactly the kind of insiders a government usually protects on reputation alone: large, established, close to the financial and industrial elite. Tokyo let its own banks impose real losses on them anyway, on the same terms steel, cars, and semiconductors had already been tested on: what mattered was productive use of resources, not who held them.

The harder test

A bank that expects the state to cover its losses has no reason to impose real terms on an incumbent it rescues: that is the gentler test. Tokyo did not let banks assume that. The Bank of Japan’s own loan to Yamaichi’s banks came only after they had forced the cuts through themselves, not before, and MOF and BOJ were reluctant to make that loan look automatic.^[14-1] That reluctance was the state’s hand in what looked like a private decision. It set the expectation Sumitomo Bank was operating under a decade later, when it rescued the failing carmaker Mazda by installing its own executives inside the company.

Prioritising transformation over vested interests looks like this: not a slogan, but a government willing to let its own establishment absorb real costs rather than exempt it from the same test.

That is a narrower claim than saying the system was open to any newcomer, but it is stronger evidence: powerful insiders can fight back, so a government willing to cost them anyway is much harder to fake than one that merely tolerates new entrants.

The harder test held both times: real costs at Yamaichi in 1964, a real loss of control at Mazda in 1974. It did not hold forever. Yamaichi itself is where it failed: the firm collapsed outright in November 1997, brought down by losses concealed through off-book transfers to related companies, more than three decades after its own 1964 rescue, one of the largest corporate failures in Japan’s postwar history.

A commitment with the exit removed

Japan had a third way of keeping transformation the state’s own priority: it bound itself to obligations a future government could not quietly waive. In 1964, Japan took on the rich-country club’s own trade obligations, current-account convertibility and OECD membership, over the objections of the industries the move exposed to open competition.^[15] Korea did not make the same choice for another thirty-two years. A protected economy that opens on a treaty’s own schedule has given up the easy option of changing its mind.

Subscribe now

This is the floor the rest of the story needs: Japan had a transformative governing orientation worth the name, evidenced on the record before the 1973 turn.

That separability, orientation tracked through dated acts rather than assumed from who held power or what tools the state had, is what makes a change in it detectable at all. What comes next puts that same transformative orientation through a different kind of test: not whether it holds, but whether it moved.

II: How Japan turned

The turn in Japan’s governing orientation looks like a single moment. It was not: January 1973 only ratified what three years of separate decisions had already built.

Three years welded into one budget

In late 1970, the Diet removed from the country’s basic pollution law the clause requiring environmental protection to stay “in harmony with the sound development of the economy”. That shifted the balance of considerations away from prioritising growth. That same year, the government began paying rice farmers to grow less. The same logic reached retail in 1973: small retailers got the one thing Japan’s governing terms had always refused a firm, protection from whatever would displace it. From 1974, a large store needed their approval before it could open, not just their customers’ business once it had.^[16]

Welfare built up its own record over the same years. The 1972 budget brought free medical care for the elderly. Then, on 15 January 1973, a budget: social security up 28.8 per cent, the largest jump in a decade, environment up 61.7 per cent, requested public investment up 38 per cent.^[17]

The press called it the First Year of Welfare; a cabinet decision the following month gave the shift its own name and its own governing document, the Economic and Social Basic Plan, subtitled “for a dynamic welfare society.” The reforms were built to last, not simply to win an election: pensions were indexed to prices that same year, an escalator that ratchets, and the share of national income going to social security climbed from 6.5 per cent in 1973 to 13.7 per cent by 1982.

That indexing is itself a governing term, the same kind of standing rule that once locked transformation in place, now locking in something else. Some of that social security increase was arithmetic, not discretionary: Japan’s fertility rate had already fallen below replacement, and the country had already crossed the standard threshold into an ageing society, so a rising elderly share was setting a floor under pension spending on its own.^[18]

What had begun as a single budget was now a standing claim, indexed to keep growing on its own: a second priority added to production, not yet a replacement for it, widening what Japan’s governing terms protected.

The man presiding over it fit the shift: Kakuei Tanaka, prime minister from July 1972, was a machine politician, a man who had risen through trading public works and favours for votes.^[19] A leader with no developmental formation, pouring real ambition into redistribution, was what the widening of purpose looked like from the inside.

A ratification, not a start

Two ordinary explanations could account for that widening without needing any shift in the state’s own priorities. The demographic floor was one: a rising elderly share setting pension spending on a path no government could easily have avoided. A leader facing that same ageing electorate has an ordinary incentive to spend on transfers, whatever the state’s own priorities happen to be. Either alone, or both together, could explain the direction this one budget took.

Neither explains what came after: fifty years without a single government even attempting to re-point the state toward a new productive frontier.

January 1973 is not the moment the turn was decided. It is the moment three years of separate decisions got welded into a single new floor under what the state protected.

A turning point is often the moment an accumulation gets ratified rather than the moment it began, and treating the ratifying act as the whole cause risks making a change look more sudden than it actually was.^[20]

What stayed exactly where it was

Japan’s orientation shifted from transformation towards maintenance and distribution. Almost nothing else changed. The party stayed in office. The rural base that was its electoral spine stayed exactly where it was, though what that base wanted from the state had itself begun to shift, from investment in yields to support for prices. The machinery, the trade ministry that ran industrial policy, held too, and kept winning for another decade.

This wasn’t the kind of overhaul economists usually mean by a paradigm shift: the machinery itself never changed, only what it was told to protect, first alongside the old ambition, and then, increasingly, instead of it.^[21]

Japan’s priorities reranked. Its machinery never did.

None of this required Tokyo’s elites to want it. An unused capacity to resist counts for as much as a declaration that costs nothing, and nothing in Japan’s own governing terms required its leadership to hold the line against this pressure. It did not.

Korea faced the same 1970s shock and answered differently: its cabinet doubled down on heavy industry within days of the oil embargo instead of pivoting to compensation, discipline that kept renewing itself for decades, though not permanently, since the same political capacity to discipline can fade once business and civil society grow more assertive of their own legal rights.^[22] The same shock did not force the same response. Korea kept production first; Japan let something else take its place.

Rice was where that same absence of resistance ran longest, and cost the most.

A closer look at rice

Japanese farming had been an engine through the developmental years, and it earned that role the same way every other sector did: land reform built a mass of smallholders, and the state backed them because their surplus helped finance industry, a claim rice made good on by producing. The land didn’t change. The farmers didn’t change. What changed was the claim farming was allowed to make: from a claimant tested by yield to a claimant protected from the test itself. In 1970, three years before the oil shock, the government organised a nationwide cartel paying farmers to leave paddy fields idle, holding rice prices up through managed scarcity rather than yield.

Introduced as temporary in 1970, the cartel was still in full force a decade later, no longer the exception but the sector’s new governing terms. By the early 1980s the state was paying over 360 billion yen a year to keep rice off the market.

The domestic spending figure is not the only place the shift shows. A world food-price spike in the early 1970s should have narrowed farm protection everywhere: when world prices rise, the gap between them and a country’s own protected price narrows on its own. That is exactly what happened almost everywhere else: France, Germany, Britain, even Denmark all saw their own measured rate fall hard. Japan’s own rate barely moved through 1973-75. It did not narrow. It kept climbing instead, more than doubling by 1987. The same yen the state once spent on transformation, it now spent on maintenance.

What kept it going

Fifty years is a long time for an ordinary trigger to keep working. What actually sustained the shift in priorities?

Welfare’s own turn had three plausible sparks: a faction’s own rise inside the ruling party, dressed up in policy language; an electoral system built to pay politicians for spending on narrow, organised constituencies, doing exactly what it was built to do; or a ruling party buying off the specific threat it had just faced at the polls, the way it had done before.^[23] Any of the three could have lit it, and none of them needs to be wrong.

None of them could keep it burning past its own limit: a faction only as long as its own fortunes hold, an electoral system only as long as it stays in place, a threat only as long as it takes to pass. All three limits were reached. The widening outlived the faction supposedly driving it. It outlived the electoral system itself, abolished in 1994. It outlived the 1971 scare that first triggered it, by decades, still resolving the same way long after all three were gone.

Rice tells a narrower story. Rural votes shifted, so the shelter grew, and nothing about the state’s own priorities needed to move. That’s probably true, for rice. It doesn’t reach steel, or the banks, protected on exactly the same terms.

The faction, the electoral system, the 1971 scare, rural votes: each one can take credit for a spark somewhere. None can take credit for the fifty years after. What lasts that long is not a spark. It is Japan’s own governing orientation, shifted once and never shifted back, needing no fresh faction, no fresh incentive, and no fresh threat to keep winning in whatever sector it touched next.

What is not yet shown is what it cost, or when that cost arrived.

III: What Japan stopped checking

Japan’s economic engine kept winning for another decade, but not because its governing orientation was still testing it: that had already thinned. What carried the economy forward was the momentum built before it did, already being dragged down even while it still looked like winning.

A stock, not a flow

By the early 1980s, nine years after a budget had let something else outrank production, Japan’s chipmakers led the world’s most important memory-chip market. A lead like that could mean one of two things: a state still testing its own firms hard enough to keep them out front, or a sector coasting on a capability built years before, no longer being tested at all. From the outside, the two look the same. Inherited capability keeps scoring well long after the checking that built it has stopped.

A complexity ranking, a patent count, a market share: each one reads the stock, what a country has already built, not whether anyone is still testing it. Contestability reads the flow instead: whether a new entrant can still dislodge the firms the numbers celebrate.

That is the direct evidence, and the only kind built to catch a state that has stopped testing itself.

The semiconductor drive launched in 1976 produced real results through the late 1970s: by 1982, Japanese firms held about seventy per cent of the American market in the leading generation of memory chip. The drive itself ran only four years, and its own account credits what came after to differences in corporate and financial structure, not to continued state effort.^[24]

Japan’s broader numbers matched the same shape: high in the complexity of what it made, among the leaders in patents per head, its frontier goods carried by home-grown firms no ministry shielded. None of that tells you whether Japan was still testing its own firms, or coasting on capability that testing had already built.

The very industry the 1976 drive built eventually passed to a Korean rival. By 2012, Samsung led the world memory-chip market outright, and Japan’s makers, merged into a single firm, went bankrupt the same year. An outsider took the frontier corner away, dislodging the exact firms the numbers had been celebrating. Complexity rankings and patent counts never have to sit that test.

A firm’s claim to its own ownership could be tested too, not just its product, and capital markets were where that happened. The state treated the two very differently. Japan’s own 1964 OECD accession forced its capital markets open, exposing incumbent management to a threat product competition never carried: a hostile foreign bidder. The Ministry of Finance worked alongside business to build a defence against exactly that. Cross-shareholding climbed from 18.4 to 27.5 per cent of listed shares between 1965 and 1973, and financial institutions from 23.3 to 33.8 per cent.^[25]

The defence worked fast: the market for corporate control closed within a decade of 1964. Product markets stayed genuinely open for decades longer, all the way to the Samsung handoff in 2012.

Starting without stopping

The state kept starting new things: a solar programme in 1974, that same chip drive in 1976, a new energy agency in 1980, a basic-research programme still running from 1981, a computing project in 1982, a chipmaker in 2022, the government has now backed with more than fourteen billion dollars. On the surface, that looks like the same orientation that built steel and cars in the first place, still picking industries, still putting money behind them.

But starting something new was never the hard test. When money isn’t scarce, funding a promising bet costs little in the sense that matters: no rival claim has to lose for it to go ahead. Prioritising transformation was never just about what got funded. It was about whether the bet had to answer for itself once it stopped delivering, whether the state was willing to stop paying for a failure.

None of these later programmes ever faced that test: every one of them ran through the firms already there, and none was ever cut loose or made to answer for itself the way an exporter was. Starting kept happening. Stopping did not. It is the clearest case in the whole series of a state willing to start but never to stop.

A boundary, not a choice

The clearest sign that the orientation had already shifted sits inside the machinery itself, not just in what the state kept building. The same machinery had already picked up a second job to match the state’s new priorities: by 1978 the Depressed Industries Law re-tasked MITI’s own tools, the ones that had once enforced the market test on growth sectors, to manage the decline of the ones that were failing it. A designated industry now drew up an equipment-disposal plan for MITI’s approval, backed by policy-based finance, tax relief, and exemption from the Antimonopoly Law itself. Same ministry, same statute, a different job standing beside the old one: the tools built to check a claimant were now excusing one.

The job grew, not shrank. A 1983 extension widened the exemption from fourteen designated industries to twenty-six, each one a claimant excused from competing at all. It is the same cartel logic that rice already showed, protection that grows once it starts, now running through the very machinery built to enforce the opposite. The pattern outlasted the law’s own list, too: cartels in cement, aluminium, steel, and petrochemicals kept sustaining high domestic prices and blocking new entrants well beyond the 1970s.

The government’s own development-finance program shows the same claim losing, but on a slower timeline than the budget of 1973. The channel that once decided which claimant got the state’s scarce credit, the trade-ministry bureaus and development-bank credit of the governing circuit itself, had already been draining away from productive industry before the January budget. The share earmarked for strengthening key industries ran above 20 per cent in the mid-1950s and was already near 3 per cent by the years around 1973, never to recover. The share for housing and living standards passed 40 per cent over the same stretch, on its way past half by the 1990s.^[26] The circuit did not close in a single year. It had been drifting shut since the 1950s; the January budget only confirmed a direction the credit data had already long set.

One industry shows a different kind of limit. Software rewarded fast entry, workers who moved between firms, and ideas that moved between projects; the institutions built around large manufacturers and coordinated catch-up fit that world poorly, and Japan never mastered it.^[27]

It is tempting to call that fate: institutions built for one world, unable to survive the next. But Japan’s institutions had already proven they could build from nothing. They built steel. They built cars. They built semiconductors. The same machinery now failing at software is the machinery that pulled off all three. What they could not do was rebuild themselves a second time, into something suited to a world that rewarded the opposite of what they were built for. That is a narrower limit than fate. It is a boundary on what the machinery itself could do, not a change in what the state chose to protect.

Software shows what the machinery could no longer do. The same years show what the state had already stopped asking it to do. By the 1990s, the market test had not applied to a state-backed investment in over a decade, capital could not discipline a keiretsu firm’s management, and the credit circuit had stopped rewarding performance.

Zombie lending, a bank letting a weak firm keep borrowing rather than forcing it to fail, was not a new kind of failure by the time it appeared. It was the only kind of failure left to a machinery that had already lost every other way of testing a firm. None of it was a capability failure like software’s. It was the orientation itself, no longer asking anything to answer for itself.

Subscribe now

Twenty years before the reckoning

The state’s priorities shifted in January 1973, with growth near ten per cent. All of it, the credit that stopped pulling toward growth sectors, the Depressed Industries Law, the software Japan never mastered, took a generation to surface as the stagnation of the 1990s. That a different line in a different budget might look like the turning point just as easily is a fair challenge. Only the 1990s could confirm that this line, and not another, was the one that mattered.

No single decision was ever going to say so outright, and no memo would either. A minister does not announce a lower rank for transformation. What shows instead is a lean: thousands of individually defensible choices, a pension law here, a price support there, a budget line nudged half a point, each one too small on its own to call a turn. Add them up, and the direction is there anyway, the way you only see what has been heating a pot all along once you stop watching the water minute to minute.

This is not drift, an institution quietly accomplishing less over time without anyone redirecting it. It is a compounding of two dated, separable effects: a reprioritisation, shown plainly, in a budget, on a date, while power and machinery held still, and a failure to rebuild, shown too, in that stalled credit and the tools handed a second job managing decline.

One piece of the second effect is worth noting. In 1979, the bureaus themselves proposed a re-pointing, MITI floating the idea that Japan become a “knowledge-intensive industrial nation,” a frontier agenda in computers and advanced manufacturing. No prime minister of the years that followed ever took it up as his own the way a cabinet had signed Income Doubling in 1960. The chance was there. It went untaken.

The 1978 law sat only five years downstream of the 1973 budget. The 1979 proposal that went untaken sat six. Nothing on the scale of the VLSI drive itself ran again for fifteen years, not until 1995. An administration whose own instruments had already been redirected toward managing what was shrinking had correspondingly less machinery left pointed at what might have grown next.

The lever that gave Tokyo its grip on steel, cars and chips in the first place was always a specific one: rationed dollars, licensed technology, credit a firm could not get anywhere else. Even the 1976 chip drive, the same case already read here as a stock rather than a flow, shows that lever weakening. By then, Tokyo was a funding partner in a joint programme, not a gatekeeper firms could not get around. What followed is credited to the firms’ own corporate and financial structure, not to continued MITI machinery.

A government that wanted to act on that 1979 proposal would have had to build a new lever for a new circle of firms still hungry enough for what only the state could ration. It could not simply redirect the old one. A targeted lever like this one always carries its own use-by date: it disciplines only for as long as its targets still need what it rations, and Japan’s own champions were, by design, the firms most likely to outgrow that need first.

That reorientation reached the governing terms and the governing circuit themselves, not only the machinery already named. It did not cause the 1990s slowdown; the growth-accounting record already explains the deceleration on its own terms. What it shaped was the form the slowdown took once it arrived: firms kept alive rather than let go, and the restructuring an unrigged test would once have forced never came, because the machinery built to run that test had already been repointed to protect what was shrinking rather than discipline it.

Whether Japan ever undid what it did in 1973, or whether the shifting priority simply stayed, is the question fifty years of evidence now have to answer.

IV: Why Japan never turned back

Once Japan re-ranked its own priorities, undoing the choice got harder with each year, not easier. Fifty years on, no government has yet re-oriented the state toward a new productive frontier.

Past the point of undoing

The reckoning that arrived in the 1990s did not produce a course correction. Japan’s governing orientation, once it widened in 1973, never narrowed back.

Part of the reason is political, and it sits in the years right after 1973. A coalition built around a settled priority tends to grow strongest exactly when nothing disturbs it, and for twenty years, nothing did.

Part of the reason is more mechanical: the same lever already traced here, the one that worked only as long as Tokyo’s champions still needed it, had already been spent by 1979.

Political turnover was the first real chance to bring back the discipline the ten-per-cent years once enforced: a new government declaring a new productive standard and holding itself to it. The Liberal Democratic Party lost power outright for the first time in 1993-94. None came. Germany, facing the same pressures from the same decade, did exactly that where Japan didn’t: after a comparable seven-year stall, 1998 to 2005, Chancellor Schroder announced the Agenda 2010 reforms in 2003 and re-pointed the state back toward its productive base, the reforms themselves rolling out over the following two years.

Germany got a reset. Japan’s own chance had already come and gone without one.

More chances followed, and the same test went unanswered each time. The LDP lost power a second time, from 2009 to 2012, with no new standard declared either. Then Abenomics, from 2013: cheap money and a long fight against deflation moved the demand side, but no cabinet ever declared a new productive direction the state could be held to. Growth steadied without the one thing that would have restored the discipline. Fifty years on, no reversal of the widening has ever been dated at all.

None of that made the shift in orientation permanent on its own. A theory of entrenched coalitions predicted the same outcome for Germany, yet Germany broke out. One more possibility is that the persistence traces back to Tanaka’s administration in the 1970s specifically, not to a governing orientation with a life of its own beyond it. If that were the whole explanation, it should have faded once his own party lost power. It did not: the LDP fell twice, and the shift in governing orientation survived both falls.

A gesture, not yet a test

The government now in office is gesturing toward a reversal. Sanae Takaichi’s talk of economic security and her state-backed bet on the chipmaker Rapidus mark the most explicit reach for a declared productive frontier since the 1979 proposal that went untaken. What would make it more than a gesture is an act to match it: a verdict Rapidus is actually held to, a test it can fail, the same kind of act January 1973 was, only pointed the other way. No such dated mechanism has been found yet. That is a gap in the evidence, not a confirmed absence.

Even a clean verdict would only put Rapidus where Tanaka’s budget stood in January 1973, not where the orientation stands now. Fifty years without a government undoing it, through two changes of the party in charge, is what confirmed the shift. Nothing today can confirm the shift this early. Whether this one is different is the only open question left in a fifty-year record.

What holds together

Two explanations have been offered for why the governing orientation never narrowed back: a political one, that the coalition around it entrenched because nothing challenged it for twenty years, and a mechanical one, that the state had already spent the one tool it would have needed to redirect the economy. Neither is enough by itself. The political one alone doesn’t fully work: the same coalition-entrenchment logic predicted Germany too, and Germany broke out. Add the mechanical fact, though, and the persistence stops being mysterious: even a government that wanted to break out would have lacked the specific tool Germany didn’t need to lack.

The explanation goes beyond capacity, though: Japan was not just unable to reverse course. No government even wanted to. Not until the one now in office.

Widen the lens, and four things now hold together: the orientation was real, it turned on a specific act, its cost stayed obscured for twenty years, and fifty years on, nobody has yet undone it.

V: What Japan proves, and doesn’t

Japan’s priorities changed years before the cost was visible, and no one was watching for it. Whether other countries follow the same pattern is still an open question. But the standard instruments would have missed it either way: they stop watching exactly when a country starts looking like it has arrived. Watching what a government protects, not what it grows, works, and Japan is the proof.

What was revealed

Japan’s own trade-off was revealed in a single budget, on 15 January 1973: three years of separate decisions welded into one document that, for the first time since the postwar reconstruction began, let welfare and protection stand alongside production at the top of what the state defended. The party in office held. The machinery that ran industrial policy held. Only the priority moved, and it moved on a date anyone could have read at the time.

What can be said of Japan’s own budget was there to be read, in public, on a single date, whether or not anyone actually read it that way at the time. Twenty years passed before anyone did. That gap is not evidence that the document was unreadable. It is evidence of how easily a legible document goes unread when nothing forces the reading.

The 1973 budget isn’t the whole case. By then, Tokyo had already shown, more than once, that its own machinery would hold an insider to a real standard: a 1955 act that let the coal industry fail rather than protect it, the Mitsui Miike lockout that broke a 312-day strike by one of the country’s most powerful unions,^[12-1]^[13-1] the 1964 rescue of Yamaichi that arrived only after its own banks forced real cuts through first.^[14-2]

The same machinery that had already imposed on its own establishment real losses, twice over, in the two decades before 1973, is the machinery that was still standing, unchanged, on the other side of the budget. What moved was the priority a tested machinery was made to serve, not the machinery’s willingness to enforce a standard.

What was overlooked

For twenty years, the same economic system kept winning on the strength it had already built. It was never once asked to answer in the way it had answered for the old priority. A growth rate does not differentiate between a system still being tested and one coasting on what it built before the testing stopped. What does is what a state newly chooses to protect, visible only once you stop reading any single year and add the choices up. It shows a change in priorities before its consequence, not after, something no single year’s growth figure ever could.

Arrival is not a finish line a country crosses. It is the moment the test that built its success runs out.

A state that finds a new test to answer to keeps being tested. A state that doesn’t starts quietly living off what the old test already built, with nothing in a growth chart to show the difference.

What the 1990s exposed was not a mystery once the 1973 budget was understood for what it was: a productive system that had quietly stopped rebuilding for what came next, years after it had stopped being asked to.

It was not a production failure. Measure output per person of working age rather than per head, and the aftermath wasn’t the disaster the aggregate figures suggest: it rose more than twenty per cent between 2000 and 2013, ahead of the United States over the same years. Japan kept producing, per worker, about as well as anyone.

What it lost is narrower and harder to see in a growth chart: it lost its recurring prioritisation of transformation. Japan never again pointed its production at a new frontier the way it did in 1960. The same failure of contestability shows up again in the circle of firms the state itself protected: the country that once forced its own chipmakers to answer to a live market lost DRAM leadership to Samsung, and watched the firm its chipmakers had merged into go bankrupt in 2012.

Nor did the shift ever revert, for the reasons already traced: political entrenchment, and a lever the state had spent by 1979 and never rebuilt. Whether the gesture now underway breaks that pattern is still an open question.

What slowdown pace theories missed

Japan’s own case explains something the standard story about its stagnation has never quite captured, not just that growth slowed, but why the slowdown took the shape it did. Catch-up was ending. The oil shock was real. No government could have kept growth at ten per cent forever, and nothing here says otherwise. What those causes don’t explain is the shape of what followed: a country that could have gone looking for its next productive frontier let something else stand at the top instead, and never fully re-pointed.

That gap is exactly what the standard instruments were never built to see. The World Bank’s own Growth Commission put Japan on its list of thirteen success stories in 2008, thirty-five years after the 1973 budget, a perfectly fair verdict on whether Japan had arrived, but arrival was never the question. Complexity rankings and patents per head are the same: real measures of what an economy can do, not of which direction its government is pointing it. None of them were ever going to catch a change in priority, because none of them were looking for one.

The scholarship built to explain the stagnation has the same blind spot, one level up. A productivity-collapse account, a balance-sheet-recession account, and a monetary-mismanagement account each answer a real question, how fast, or how well, did growth come back, and none had reason to ask what the state chose to redirect production toward once it did. A theory of pace and a theory of priority are different theories, built to catch different things.

One strand of that scholarship gets closer than the rest. Zombie lending, banks keeping insolvent firms alive through subsidised credit rather than letting them fail, turns out to be describing the same failure from the inside: firms that exited Japan’s domestic market were, on average, more productive than the ones that survived, the identical failure of contestability inside Japan’s own domestic economy that DRAM’s handoff to Samsung already showed abroad. Two separate bodies of evidence, one testing exporters against the world, the other testing ordinary firms against each other at home, arrive at the same place.

What one country can prove

What Japan’s own record does prove is a revealed priority: a change visible before its consequence arrived, dated to a single budget, while growth was still running near ten per cent. No account of the stagnation gets there first; none of the slowdown theories were ever asking that question.

What it does not prove is that the shift caused what followed in the 1990s. Catch-up ending and the oil shock explain why growth slowed; the shift in priority explains only the shape of what happened next. Nor does it prove that Japan is the rule rather than the exception: whether the same pattern holds elsewhere is still an open question.

What it does prove is that the method works. Reading any government’s priorities the same way means watching what it protects once protecting everything is no longer possible, not waiting for the growth numbers to confirm it decades later. Japan is the proof.

The production of this post has been AI-assisted. For more details, see the About page.

How a Government’s Real Priorities Get Revealed

Stephen Brien — Tue, 04 Aug 2026 07:00:08 GMT

The short version

Malaysia and China’s policies look alike but reveal opposite priorities. Malaysia shelters its champions, tests its zone; China shelters a few firms, tests the rest.
A government’s real priority is revealed at the exact moment it can’t protect everything. Whether that priority lasts for years is separate.
Three questions sharpen in turn: what a firm must deliver, whether growth can spread, and which protected interests it would threaten.
Real cost is the clearest sign: China let its own financial firm fail, Taiwan stripped import licences, Singapore cut workers’ savings.
But cost alone isn’t proof: Indonesia ousted its oil chief over hidden debts worth a third of the economy, yet established no test on firms’ productive activity.
The tell for real priority: does a government hold firms to a standard it can’t rig, and let winners grow or wall them off?

The previous post showed how eight economies were sorted by their governing orientation: whether a firm answered to a standard its government could not lower. This post asks what a single revealing moment can show about which priority a government actually protects. A later post asks whether that moment recurs into a governing orientation durable enough to survive successive leaders and crises, and how early such a reading can be trusted.

Malaysia and China both built economic zones.

From outside, the policy looked much the same, with fenced areas, foreign investment, exports and firms exposed to international customers.

Malaysia placed export firms inside its zones. Those firms faced a genuine test. They had to win foreign customers, and the government could not simply declare them successful.^[1]

But the test stopped at the boundary. Beyond it stood protected domestic champions (firms the state had chosen to shelter), politically allocated ownership and markets the zone firms could not readily take from them. The export economy could grow without displacing the claims governing much of the economy outside it, such as an ownership right, a licence or a protected position a favoured actor could lose.

China also kept a protected core. State firms in oil, telecommunications, power and other commanding sectors remained sheltered.

But much of the new activity developed beyond such firms’ automatic claim. New firms could enter, expand and compete in a margin, a space at the edge of the economy, that the old core did not wholly control. State firms entering that space could lose. Hsieh and Song estimate that more than 83 per cent of the state-owned industrial firms operating in 1998 had closed or been privatised by 2007.^[2] That scale of exit carried a real cost in jobs lost, not only a firm-level statistic.

Both governments sheltered champions. Both exposed other firms to demanding markets.

Malaysia enclosed the test. China enclosed the champions. The same instrument, and the opposite thing inside the wall.

Investors, donors and policymakers must judge a government’s direction long before any growth figures exist to help them. Two kinds of evidence let them judge it. The ordinary terms a government sets for its firms are one, and what it does the one time protecting a favourite would cost something real is the other. What follows builds the first kind of evidence, then tests it against the second, and asks at the end what the two together are actually worth.

I: What identical policies can hide

Two governments can use the same tools for opposite purposes. What separates them only becomes visible once each can no longer protect everything.

What the zones don’t settle

Malaysia enclosed the test, and China enclosed the champions, but that contrast alone is not a complete description of either economy. Malaysia’s zones generated genuine production, employment and learning, while China’s exposed margin remained extensively shaped by the state.

The narrower contrast concerns the relationship between protection and the emerging path, whether the firms subject to it can reach, expand along and ultimately reorder the terrain through which the economy is changing.

That zones comparison tells us something important, but not yet enough. Malaysia kept the demanding test apart from the protected domestic order. China allowed much of the new economy to develop beyond the old core’s automatic claim.

But arrangements can be bent, waived or reversed once powerful interests are affected. But arrangements can be bent, waived or reversed once powerful interests are affected. What those choices reveal is the priority a government preserves when it can no longer protect every purpose and claim at once.

What a revealed priority means

Development plays out across decades, crises and successive leaders, not one leader’s tenure or one test. But it starts somewhere, with a single government, at a single moment, when its terms are tested, and something has to give.

Call what that moment shows a government’s revealed priority, the purpose it protects, once, when it can no longer protect everything. It is not an assertion about the whole state. A government may protect transformation in one domain while sheltering privilege in another. The inference stays limited to the specific path, domain and period the evidence actually covers, never generalised beyond them.^[3]

Declared ambition, administrative capability and a revealed priority answer different questions. Ambition identifies where a government says it wants to go. Capability identifies what the state can carry into action, however skilfully. A revealed priority identifies which purpose actually prevailed, this time, when the government could not preserve everything it valued.

China and Malaysia could both administer zones and attract export investment. Capability alone did not tell us what each arrangement was politically organised to protect. All eight governments in the previous post’s comparison declared development, and substantial administrative capability appeared on both sides of the eventual divide between the economies that went on to sustain transformation and those that did not.^[4] Capable machinery can direct credit towards exporters, but it can also enforce ownership targets, protect favoured firms or extract resources for political maintenance.

Ambition identifies the destination. Capability supplies the machinery. A revealed priority identifies which purpose actually governed, this time.

This post is concerned with a transformative revealed priority, one in which productive transformation takes priority over a politically valuable claim capable of governing the consequential economic path. Reading it takes two kinds of evidence, before any growth figures exist. The ordinary governing terms firms face support an inference about intention; a conflict in which those terms turn politically costly reveals what actually gets prioritised.

Whether a revealed priority recurs into a governing orientation durable enough to survive successive leaders and crises is a separate, longer question. A later post takes it up. This one stays with the here and now.

Governing terms point to intention. Conflict reveals a priority. Whether it recurs into an orientation is for later.

II: What the terms make conditional

Long before the terms are tested, a government has decided who may enter, what support depends upon, what counts as success and who may alter the answer.

What a firm actually delivers

Plans identify the direction a government declares. Governing terms show how it has organised economic activity around that direction. They establish who may enter, what support is available, what firms must do to qualify, what counts as success or failure, and who is entitled to different treatment.

Those terms also include the fiscal arrangements surrounding a government’s plans, what the government taxes or exempts, which revenues it protects or earmarks, where it directs procurement (what the government buys, and from whom) and capital spending, and what purposes public investment is built to serve.

Tax concessions, infrastructure, guarantees and budget allocations may make a declared productive path materially possible. They may instead preserve incumbents, distribute rents (payments that flow from a protected position rather than from producing anything) or finance political maintenance (paying off the coalition that keeps the government in power).

None of these instruments discloses its purpose by itself. A development bank may finance firms expected to build productive capability, or preserve borrowers whose political position makes repayment optional. Public investment may open a path to new producers, or strengthen the actors who already control access to that path. The intention becomes clearer from who receives the advantage, what activity it enables and what its beneficiaries must deliver in return.

Two comparisons make the point concrete. One isolates what a supported firm must deliver in return; the other isolates whose access the government chooses to make dependable in the first place.

Korea and Malaysia both reached for the same tools, among them credit, protection, tax privileges and administrative preference. What a firm had to do to keep them differed completely. In Korea, access to state-directed credit was tied to export performance. Foreign customers supplied an answer the government could not wholly control. Either the firm could sell what it produced or it could not. Korea’s governing terms indicated that support was intended to build productive capability tested beyond the domestic political system.

Malaysia organised its advantages around two purposes. Export manufacturers received tax holidays (a set period with no tax owed), duty-free inputs and industrial-estate infrastructure in return for establishing internationally competitive production. In the domestic economy, licences, directed credit and procurement were attached principally to Bumiputera (ethnic-Malay and indigenous) ownership and participation. The beneficiary qualified because of who owned the firm and the distributive purpose it served. Malaysia’s fiscal architecture therefore supported export production while securing an ethnically defined share of ownership and economic control.^[5]

The contrast was between what a firm had to deliver and whose economic position the government intended to advance.

Subscribe now

Whose access counts as dependable

External markets are one way of supplying a meaningful standard, but not the only one. Competitive procurement, published licensing criteria, creditor rules or independent technical assessment may perform a similar role. What matters is that the condition is specified before the result is known and cannot easily be rewritten afterwards for one particular firm.

Singapore offered foreign manufacturers land, infrastructure, tax concessions, training and administrative support without requiring the resulting activity to be owned or controlled by an established domestic business group. The qualification was the productive activity the firm proposed to establish. Ownership by a politically favoured constituency was not itself one of the intended results (constituency here means any support base a government needs, not a body of voters).

Indonesia also used licences, credit, contracts and protection to promote investment and industrial activity. But politically connected conglomerates enjoyed forms of access that ordinary entrants could not assume. Their place within the governing coalition helped make licences, finance and protected opportunities dependable.

Singapore’s terms indicated that the government intended to secure productive activity regardless of who owned it. Indonesia’s terms indicated that development would proceed partly through a politically connected domestic business structure whose claims were themselves made valuable.

The two comparisons expose different dimensions of intention. Korea and Malaysia differed in what a supported firm had to deliver. Singapore and Indonesia differed in whose access the government chose to make dependable.

What actually decides who prospers

Formal rules do not necessarily settle either question. Rules may appear open to all while the most valuable guarantees, land, finance or procurement remain reserved for recognised insiders. An entrant may technically qualify for support but still have to negotiate every permission, while an established firm can plan on the assumption that access comes with its position.

Weak formal institutions don’t mean uncertain treatment for everyone. A government can make dependable commitments to a favoured few while leaving everyone else exposed to its discretion.^[6] A governance or institutional-quality index can’t tell those apart, a blindness built into the measure itself. It scores how rule-governed a system looks, not who the rules are written to protect.

This same blindness affects real capital. A score built for this purpose cannot tell an investor or donor which government is open to genuine discipline and which merely looks orderly. Both can score identically well.

The relevant distinction is therefore not simply between rules and deals. It is whether the terms provide an identifiable basis for differential treatment, or make the political identity of the firm itself the basis on which access depends.

Consistency does not require uniformity. A government may deliberately offer different terms to exporters, infant industries (new sectors not yet able to compete internationally), domestic producers or firms entering strategically important sectors. Those differences may be central to its economic strategy. The question is whether they correspond to a stated purpose and an intelligible condition, or place particular firms outside the standard altogether.

That leads to the central question:

What must a firm do to prosper under this system?

Must it improve, export, innovate or win customers? Or must it control a licence, maintain a political connection or hold a protected position?

The distinction is not between supported and unsupported firms. A government seeking transformation may direct credit, protect infant industries, subsidise learning and rescue firms from temporary difficulty. What matters is what the firm must deliver in return, and whether continued support remains conditional on that delivery.

The firms supported, the form the support takes and the conditions attached to that support therefore support a provisional inference about intended priorities. They indicate what the government means to reward, whose access it intends to make dependable, and which claims it has placed beyond an independent standard.

These signals do not yet show what happens when adhering to those terms threatens an interest the government also values.

III: Where the terms stop applying

As the zones comparison already suggested, a test can be real and still be contained. Where a government lets it reach is itself part of what the terms reveal.

How far each test was allowed to go

Governing terms also reveal intention through the reach permitted to new productive activity.

Transformation occurs when the ability to invent and make hard things becomes widely embedded in an economy, reshaping the work people do, the capabilities suppliers acquire and the range of goods and services the country can produce. Growth may instead remain concentrated within a productive enclave, never asked by the wider economy to compete beyond it.

The economic distinction between an enclave and a more deeply connected growth sector is familiar. The additional question here is political. Are the enclave’s firms and competitive terms allowed to reach beyond it, drawing in workers and capital and altering domestic markets and arrangements as they grow, or does the protected order retain control over the finance, markets and access through which wider expansion would have to pass?

The answer points towards what the government intends to prioritise. A government willing to let new productive activity reshape firms, workers, suppliers and patterns of ownership is accepting that transformation may unsettle politically valuable claims. A government that confines the same activity within carefully maintained boundaries indicates that those claims remain part of what it intends to protect.

Malaysia illustrates the second configuration. Its zone, already tested against foreign customers and already walled off from the protected domestic order, generated exports, employment and learning without displacing national champions, altering the wider ownership settlement or carrying its competitive terms into the domestic economy.

The productive test was real. The transformation it might generate was contained.

China protected a commanding state-owned core, but allowed much of its emerging economy to develop beyond that core’s automatic claim. Coastal zones, township enterprises (collectively owned local firms operating outside the state plan) and later private firms could enter new activities, attract labour and capital, form supplier relationships and expand into markets not reserved for the old plan. Protected enterprises remained important, but they were not granted an entitlement to control or seize every new source of growth.

China’s own reach went further still. Much of the old core stayed enclosed, but new activity could draw growing parts of the economy and society into a different productive system.

Taiwan offers another version of this broader reach. Its emerging export economy was not confined to a sealed enclave or reserved for established licence-holders. The removal of import-licence rents reduced incumbents’ control over entry, while new domestic manufacturers and suppliers could participate in expanding export activity.

The terms allowed productive learning to spread through firms and supply networks, changing not only what particular exporters sold but what a widening part of the economy was capable of making.

Why the boundary counts as a choice

Three different boundaries, drawn in three different places. Were any of them actually a choice? The boundary is not simply where reform has not yet reached. Malaysia’s ethnic-equity exemption for its export zones is a named, structured carve-out, not an absence of policy. China’s accession to the WTO (World Trade Organization) extended new terms to the exposed margin while leaving the core’s protections outside that same negotiation.

A named exception had to be deliberately drafted by someone, so it counts as real evidence of intent. A general silence counts as no evidence either way. The topic may simply not have been reached yet. Both look like active design, not passive lag.

Malaysia’s own natural-resource wealth may have made that wall cheaper to hold. Palm oil, tin, and later oil and gas gave its protected domestic economy an alternative source of return, one that did not depend on ever competing past the wall.^[7] A business class with a comfortable rent has less pressure to fight its way past that wall than one with no other income at all. This does not make the wall less deliberate. It may explain why holding it cost the government less than it would have cost elsewhere.

The growth path need not be fully visible when its terms are first established. A government may decide who may enter, which claims have priority and where institutional boundaries lie before knowing which sectors or firms will become decisive. Those choices are informative because they show whether emerging activity has room to spread, connect and displace, or whether its effects have been bounded in advance.

A design fixes its own reach in advance. It does not fix, in advance, which side of that reach the terrain ends up rewarding. That second fact needs the growth itself to already be visible.

The inference that transformation is prioritised strengthens as investment, entry and productive capability accumulate. By the time China entered the WTO in 2001, the exposed part of its economy was already large enough to show that it was not merely a peripheral enclave. The accession protocol extended and clarified the terms connecting that activity to the international economy.^[8]

The question is not whether incumbents remain protected somewhere. It is whether their claims control the route through which productive capability might become widely embedded.

The same test now reads as a provisional inference rather than a question. Reshaping the wider economy supports transformation’s priority; containing it supports the opposite reading.

That is evidence of intended priority. The harder question is what happens when the expansion permitted by those terms begins to impose real costs on interests the government values.

What the threat actually targets

Once a growth path can spread, the next question is what its expansion would displace.

That expansion may weaken established claims over finance, licences, protected markets, ownership, guarantees or control of entry. The threatened object is not necessarily the incumbent firm itself. More often, it is the bargain (the implicit deal) securing its position, the expectation that competition will remain limited, credit will remain available, ownership will remain untouched, or that failure will not bring the usual consequences. That is a position the surrounding terrain has never required the incumbent to defend.

The character of that bargain provides further evidence of governmental intention. An incumbent may receive dependable treatment on terms unavailable to comparable entrants. Where the governing terms leave that exclusivity open to erosion by new firms and productive activity, they suggest that transformation is intended to take priority over the incumbent claim. Where the terms preserve the position regardless of what the emerging path produces, the incumbent bargain remains among the purposes the government intends to protect.

The relevant protection may not reside in a single rule. A tariff may fall while procurement continues to favour the same producer. A loan may be called in (the lender demanding immediate repayment) while refinancing arrives through another public institution. A licence may lapse while a related company receives its replacement. Ownership may change formally while effective control, protected access and the expectation of rescue remain intact.

The inquiry must therefore follow the substantive privilege across instruments.

What actually allows the incumbent to exclude entrants, command resources or retain its position after a poor result? Protected finance, control of a licence, preferential procurement, restricted ownership or an understood right to restructuring? And do the ordinary arrangements leave that claim exposed, or preserve several routes by which it can be restored?

A government can alter a peripheral concession while leaving the protected bargain untouched. Removing one subsidy may mean little if the firm retains exclusive access to finance and customers. Opening the market on which its position depends, ending its control over entry or making continued support conditional would place the underlying claim at risk.

The arrangements described earlier denied some inherited interests an automatic entitlement over new activity. Singapore’s established local business class did not receive a reserved right to own or control the foreign-led manufacturing sector. China’s planning institutions did not retain authority over every firm and market emerging beyond the state-owned core.

This evidence is not simply that new activity existed outside the old structure. It is that particular incumbent claims over ownership, allocation and entry were not built into the terms of the new activity.

Subscribe now

Brazil, and the capture problem

Brazil provides a contrasting configuration. Its industrial policies could impose conditions concerning investment, domestic production and exports. Yet established manufacturers occupied central positions within networks of suppliers, dealers, workers and public agencies. Their privilege did not depend on one tariff or loan that could be cleanly withdrawn. It rested on a wider bargain in which maintaining domestic production, employment and the industrial coalition around the firm formed part of what state support was expected to secure.

Brazil’s automobile sector made this visible. A peripheral producer could face the stated condition directly, while an established manufacturer’s obligations could be adapted around its wider Brazilian operations and its importance to the domestic industrial structure.

The governing terms pointed towards an intention to expand industrial production, but also to preserve the position of the firms through which that production had been organised. Their established place within the system was no longer merely an instrument for achieving the policy purpose. It had become part of the purpose itself.^[9]

This alone does not distinguish a government’s own protective priority from ordinary regulatory capture (where the industry being regulated ends up controlling the regulator). An organised incumbent, left to its own devices, would seek exactly this kind of accommodation regardless of what the government intended. Brazil’s auto sector, read on its own, cannot rule that out.

The same favour fits both stories equally well. But capture means the industry controls the regulator, so a captured government could never let a live test threaten the firm it answers to; one confirmed instance of support staying genuinely conditional is proof the government, not the firm, still held the pen.

For Brazil’s auto sector, the revealing question is narrower still, whether the established manufacturers’ claim on state support ever stayed conditional on productive delivery, or whether preserving their position had become an objective in its own right.

Identifying the claim at risk sharpens the same inference. Either way, this remains an inference, not yet a settled fact.

Three questions in escalating order, what a firm must deliver, whether that activity can reorder the economy, and which incumbent claims would have to give way. Each sharpens the last.

Prices and incentives explain what a government does under a given set of terms. They do not explain which terms it chooses to set, or uphold, when it has genuine room to do otherwise. The harder evidence comes when the government can no longer preserve both the threatened bargain and its own terms, and must decide whether that bargain will actually be protected.

IV: What the government protects under pressure

A government's priorities become clearest when a favoured claim seeks exemption, whether before its terms can operate or once they impose a real loss."

When governing terms become costly

Conflict begins when a government can no longer preserve both a politically valuable claim and the terms through which it intends the productive path to proceed.

It can arise at two moments. Sometimes an inherited privilege already controls the finance, licences, markets or authority the new path requires, and the government must curtail that claim before different terms can operate. Sometimes the terms are already in place, produce an adverse verdict and threaten a valued actor with a material consequence.

The first reveals priority through the terms the government is willing to establish. The second reveals it through the terms the government is willing to enforce.

Korea’s corporate reorganisations in 1986 and Indonesia’s Timor car project illustrate the second moment (Timor was a national-car scheme built around the president’s son). Korea placed fifty-six troubled firms into state-supervised reorganisation. Their factories and productive capabilities could be preserved, but their owners and managers could no longer assume that support would leave their positions untouched.^[10] When the requirements attached to national-car status threatened Timor, by contrast, the surrounding terms were adapted so that the favoured project kept its privileges.^[11]

The material consequence may concern equity (ownership shares), ownership, management, finance, licensing, protected access or an assumed guarantee. The organisation itself need not disappear. A government may preserve factories, jobs and productive capability while diluting owners (cutting their ownership share), replacing managers or withdrawing protection. A rescue may preserve productive value while removing privilege, or preserve both together.

As before, what matters is which purpose the intervention protects.

The same asymmetry recurs elsewhere. Similar accommodations protected Malaysia’s national carmaker Proton, Brazil’s informatics firms and Thailand’s connected banks when existing terms threatened politically valuable claims.^[12]

The government’s response provides stronger evidence than the terms alone, since it shows which priority prevails once both cannot be preserved.

Both forms require further checks. Did the threatened claim matter to the government? Did the government possess a meaningful choice? And did establishing or enforcing the terms serve productive transformation rather than another purpose?

Did the claim matter to the government?

A conflict reveals prioritisation only when allowing the consequence to stand puts something the government values at risk. The severity of the loss to the target is not enough. What matters is the political cost to the government.

Brazil halted Simca’s production for six months after the carmaker failed to meet Brazil’s domestic-content conditions. The sanction was severe for the firm, but Simca was a recent entrant without a deeply rooted network of suppliers, workers or political patrons whose cooperation the government depended upon.^[13] Enforcing the standard required little sacrifice from the government itself.

Taiwan had rationed foreign currency, limiting who could import, and the resulting scarcity gave those licences real value. Dismantling that system was more important to Taiwan’s subsequent development, but presents a related caution.

The reforms destroyed valuable rents and overcame a prolonged deadlock within the administration. Yet the industrialists benefiting from the old arrangements were not central to the exiled KMT’s (Kuomintang’s) political base, while American pressure helped break the deadlock.^[14] The reform may have been highly consequential for transformation without imposing an especially costly loss on a constituency the government depended upon.

Singapore’s 1986 reduction in employer contributions to the Central Provident Fund reached a different kind of claim. It reduced the retirement provision accumulated for workers and imposed a cost on organised labour, whose cooperation was embedded in the PAP’s (People’s Action Party’s) governing settlement.^[15] The decision therefore put at risk a relationship the government had strong reasons to preserve.

These examples differ not chiefly in the severity of the consequence, but in what the government itself put at risk. Brazil could sanction a relatively dispensable entrant without testing whether transformation outranked something it seriously valued. Taiwan removed important licence rents, but principally from interests outside the KMT’s governing coalition. Both decisions may have advanced development while revealing comparatively little about prioritisation.

Singapore’s CPF reduction was more revealing because the loss reached into the PAP’s own governing settlement (its broader political arrangement) with organised labour. It risked political support and the cooperation of a constituency on which its governing model (its overall approach) depended.

The value of the claim must therefore be established independently of its eventual fate. Did it carry coalition support, employment, financial resources, control over production, patronage (jobs and favours handed out to keep supporters loyal) or the cooperation of actors capable of obstruction? The more the government stood to lose by denying protection, the more clearly its response reveals which purpose it placed first.

Subscribe now

Did the government own the choice?

Political cost reveals prioritisation only where the government also possessed a feasible alternative. Could it realistically have rescued the institution, guaranteed its debts, waived the standard, changed the ownership arrangement or placed the loss somewhere else?

During the Asian financial crisis of 1997, Thailand suspended fifty-eight finance companies and ultimately ordered fifty-six of them closed. Thailand’s closures imposed substantial losses on owners and patrons connected to the country’s political and financial order. But they occurred amid a collapsing currency, systemic financial distress and an IMF (International Monetary Fund)-supported programme that reinforced the restructuring.^[16] Thai authorities retained some room to shape the process, yet the range of viable alternatives had narrowed sharply.

The closures may have helped create a more disciplined financial system. But because Thailand had little effective freedom to preserve the institutions on their previous terms, the decision itself provides weak evidence that the government was prioritising transformation over connected claims. It shows what happened under constraint more clearly than what Thai leaders would freely have chosen.

The government’s freedom to choose matters because it determines what the decision can tell us. Where the government retains meaningful control over the result, its choice may reveal which purpose it protected. Where crisis, creditors or external conditions effectively determine the outcome, the decision reveals the force of the constraint more clearly than the government’s own priority.

The bankruptcy of the Guangdong International Trust and Investment Corporation provides a freer decision. GITIC was one of the provincial “window companies” through which Guangdong, China’s richest province, raised money from foreign banks. Because it was an arm of the provincial state, lenders treated its debts as carrying an implicit guarantee from Beijing.

When GITIC became insolvent in 1998, the Guangdong authorities proposed a rescue. Beijing rejected it. China possessed the legal authority to intervene, and its foreign-exchange reserves dwarfed the liabilities at issue many times over.

The government could have paid.

Instead, Beijing placed GITIC into bankruptcy in 1999 and allowed foreign creditors to bear losses.^[17] The guarantee investors had counted on did not survive the only test that mattered. It was not acting under an external programme, facing an empty treasury or accepting an outcome for which no plausible alternative existed. That refusal therefore provides unusually clear evidence of a priority expressed through governmental choice.

That does not yet tell us which purpose Beijing placed first. It tells us that the consequence was sufficiently its own decision for that purpose to be investigated.

A government can also author a decision before the conflict arrives. It may adopt a law, treaty or independent institution precisely to make politically convenient exceptions harder to grant later. The free choice that matters is the one that created the constraint, not the one taken at the moment the constraint bites. Although its discretion is restricted at the final moment, the consequence may still reveal prioritisation if the government freely created the constraint and continued to uphold it once doing so became costly.

A sceptic could press further: if a law authored the constraint, something authored the choice to pass that law, and so on without end: an infinite regress. Every causal explanation faces the same regress; bounding it at the observable record is how any such inquiry is ordinarily conducted, not a special exemption invented for this one. That regress has to stop somewhere: this account asks only whether the constraint relevant to this test was authored within the observable record, not to trace free choice back indefinitely.

This contrast is with terms principally defined and enforced by others. Compliance under an externally controlled constraint carries less information about national prioritisation because the government may have no effective opportunity to choose otherwise. Authorship of the terms that determined where the loss fell matters more than how much discretion remained at the final moment.

V: What the conflict was actually for

Cost and agency only show that a conflict was real. Whether it actually served transformation is what decides whether the inference holds, and that verdict reaches beyond this one conflict.

Was transformation the priority?

Political cost and governmental authorship show that a decision expressed a real priority. They do not yet tell us what that priority was. Whether the government was establishing new terms or enforcing existing ones, the final test is whether its choice enabled or preserved productive transformation.

The strongest evidence lies in the fit between the government’s stated productive objective, the privilege standing in its way and the change eventually made. The inference strengthens where curtailing that privilege allowed the transformative terms to operate, and where punishment, consolidation or fiscal necessity explain the decision less convincingly.

Taiwan’s reforms of 1958–60 show how that connection can be traced. The licence rents (the scarce-licence income already discussed) were not transferred to a more loyal group. They were curtailed so that the export strategy could operate.^[18] The fit between the obstacle removed and the path enabled makes the episode evidence of transformative purpose.

The inference remains bounded, as before, to this path, within this domain, across this period, and protected interests survived elsewhere besides. Taiwan therefore gives strong evidence about the purpose of the reform, but weaker evidence of political sacrifice within the governing coalition.

American pressure explains why reform happened when it did. It does not explain the specific design Taiwan’s own technocrats chose (the appointed experts running the reform, not elected politicians), or the multi-year implementation they carried out once the decision was made. The government owned the how, even if it did not fully own the whether.

Singapore’s 1986 CPF cut, already established as a cost imposed on a valued constituency, brings the checks together more fully.^[15-1] The claim curtailed was also the direct source of the competitiveness loss the reform addressed. Singapore accepted that political cost precisely to preserve the terms under which its productive economy could compete.

Indonesia’s treatment of Pertamina shows why the connection cannot be assumed. When the state oil company’s debt crisis erupted in 1975, Pertamina’s president, Ibnu Sutowo, controlled resources and patronage important to the military and Suharto’s regime. Removing him and curtailing Pertamina’s commercial empire confronted a valuable insider.^[19] Concealed debt equal to roughly a third of Indonesia’s economy is itself a finding: no productive test, no creditor discipline, was operating on Pertamina at all before the crisis forced the reckoning.

But national solvency and political recentralisation explain the intervention more directly than transformation. The decision did not clearly expose producers to a new productive standard or open a path previously blocked by Pertamina’s privilege. A powerful actor lost his position, but the productive terms of the economy were not evidently changed.

The contrast clarifies the distinction. Removing a powerful actor is not enough. The evidence points towards transformation when the claim curtailed was obstructing an already identified productive path, and its removal allowed that path’s terms to operate. A ruler who removes a rival while preserving equivalent shelter for loyal beneficiaries has changed who holds privilege, not the basis on which it is held.

What the three checks add up to

Malaysia and China opened this post looking alike. What told them apart was where each let its market test reach: Malaysia enclosed the test, China enclosed the champions. These checks, run on Taiwan, Singapore and Pertamina, show why that reading can be trusted.

Taiwan, Singapore and Pertamina show the same three checks converging in different ways: political cost, a government that owned the choice, and a change that served the declared productive path rather than some other purpose. Where they converge, prioritisation shows up in the terms a government is willing to establish just as clearly as in the consequence it is willing to impose.

Taiwan and Singapore make the same point from opposite directions. A case can be strong on one check and weak on another, as Taiwan is. Taiwan gives the sharpest evidence here for what a reform was for, and one of the weakest for what it cost the governing coalition. Or it can be strong on every check at once, as Singapore is. The checks sort independently; they do not average into one verdict per case.

The same caution applies to any diagnostic built from more than one imperfect signal, not only this one.

But the three checks only establish that a conflict was genuine, and that it served the productive path rather than some other purpose. They say nothing about how far the answer reaches.

A real test and a wide test are not the same thing: the first tells you the terms are genuine, the second tells you whether they are allowed to matter.

For anyone assessing a government, look first at whether its ordinary terms already set a demanding, uncontrollable standard, something a firm could fail. Then wait. A government that has not yet faced a costly test of that standard has not yet revealed a priority. The absence of a conflict is not evidence of restraint, only evidence that the test has not arrived.

That market test is not the whole answer. Once it has been passed, ask a second question. Can the firms that passed it expand into the wider economy, or does something else keep them walled off? A real test that never leaves its enclave has told you less than it first appears to: it shows a configuration that survived, not one the wider economy ever pressed to spread.

What a government does the first time protecting a favourite would cost it something real tells you more than any plan, capability score, or declared ambition examined so far.

Subscribe now

Notes

Malaysia’s electronics exporters were exempted from the ethnic-equity rule that governed much of the protected domestic economy. The zone developed real export capability, but the exemption also bounded its political reach: exposure operated where it did not displace the champions outside the zone. See the World Bank’s account of Malaysia’s electronics cluster and the preceding post’s Malaysia case.↩︎
Hsieh and Song, “Grasp the Large, Let Go of the Small” (2015), report the 83 per cent figure. They caution against treating state-firm closure as the direct engine of growth: private-firm expansion accounted for much more. The statistic is used here to show that entry and displacement operated on the exposed margin, not that closing state firms caused transformation.↩︎
Developmental-state scholarship has shown how governments may support firms while making that support conditional on productive performance. Soft-budget-constraint theory identifies the opposite condition, in which organisations expect political connection to protect them from the consequences of failure. Political-settlements accounts, and Dercon’s account of governments gambling on growth despite real political risk, explain why some governing coalitions sustain growth-promoting arrangements while others preserve rents, ownership and political position. This post’s contribution is diagnostic rather than explanatory: it identifies how, after the fact, a government’s own gamble can be read from its ordinary terms and a single conflict, not why some governments make that gamble and others do not.↩︎
The claim is deliberately narrower than saying capability does not matter. The preceding post showed that capability does not sort these eight: Malaysia scores close to China on broad contemporary state-effectiveness measures, while Korea and Taiwan’s reputations for even-handed administration strengthened mainly after their take-off; Singapore is the clear case of high early capability. Different measures rank bureaucratic construction and impartial application differently. The point here is only that administrative machinery does not identify the purpose it will carry.↩︎
In Malaysia, qualifying export manufacturers received Pioneer Status corporate-tax holidays, duty-free imports of inputs and equipment, exemptions from the New Economic Policy’s Bumiputera equity requirements and industrial-estate infrastructure at Bayan Lepas. Indonesia supplies an important caution. Oil revenues financed rice, roads and schools, providing substantial evidence of developmental ambition, while connected firms remained protected from productive discipline. Fiscal arrangements can therefore strengthen an inference about intended prioritisation, but cannot by themselves establish which purpose will prevail once that prioritisation becomes politically costly.↩︎
Pritchett, Sen and Werker, eds., Deals and Development (2018), distinguish open deals from closed ones, and ordered enforcement from disordered enforcement. The case this argument leans on is ordered but closed: a government enforces its side of the bargain reliably, but only for a closed group, while everyone else faces discretion. Reliability is not synonymous with openness. The question here is whose deal is dependable, what conduct it rewards, and whether that dependability can survive failure.↩︎
Malaysia’s commodity exports, palm oil, tin, and later oil and gas through Petronas, are an established feature of its economy. The point is offered as a live alternative worth weighing, not a settled cause: a resource-based cushion may have lowered the domestic business class’s incentive to contest exclusion from the export zones, alongside, not instead of, the government’s own choice to keep the boundary in place.↩︎
WTO accession did not place Chinese policy beyond all later revision. Its narrower significance is that the exposed sector was visible by 2001 and the accession protocol and working-party commitments increased the diplomatic, legal and commercial cost of selective discrimination or restored protection.↩︎
Helen Shapiro’s studies of Brazil’s automobile industry document both sides of this pattern. The earlier GEIA regime enforced domestic-content conditions against relatively weak entrants, while later BEFIEX export bargains with the established carmakers were repeatedly renegotiated rather than withdrawn. The claim that the established manufacturers had become part of the protected domestic bargain is an inference from this differential treatment and from their position within networks of suppliers, dealers, workers and public agencies, not a contemporaneous statement of governmental motive.↩︎
Amsden reports that fifty-six firms were under government-directed reorganisation in 1986 rather than being left to ordinary liquidation. The episode supports the narrower distinction: public support could preserve factories and capabilities without guaranteeing that existing ownership and management would remain untouched. It does not show that all fifty-six firms were sanctioned specifically for missing export targets. Kang’s competing account argues that mutual dependence between the state and the chaebol constrained how much genuine discipline could be imposed on either side, regardless of what enforcement looked like on the surface.↩︎
Indonesia’s 1996 national-car programme granted exceptional tariff and tax privileges to Timor Putra Nasional, controlled by President Suharto’s son. When domestic production and local-content requirements could not initially be met, a special accommodation allowed Korean-built Kia cars to be imported and treated as qualifying national cars, while state banks supplied roughly US$690 million in finance. The accommodation, rather than the project’s later termination under WTO and IMF pressure, is the evidence used here of terms being adapted around a politically favoured project.↩︎
The four compressed cases are developed in the preceding post. Malaysia’s Proton retained protection without a terminating market test; Indonesia’s Timor car received large state-bank lending to the president’s son; Brazil’s informatics reserve imposed substantial costs while delaying access to technology; and Thailand repeatedly protected connected financial interests. The cases differ institutionally. What they share is that a standard was altered or suspended when it threatened a politically valuable claim.↩︎
The Brazilian authorities withheld foreign exchange from Simca after it failed to meet domestic-content requirements, stopping production for about six months. The sanction was real, although it did not expel the firm from the programme. The description of enforcement as politically inexpensive is a comparative inference from Simca’s relatively weak position within the Brazilian industrial structure, rather than a motive explicitly stated in the historical record.↩︎
Before reform, Taiwan rationed foreign exchange through import licences that generated premiums reportedly reaching 350 per cent on some goods. A prolonged deadlock between Finance Minister P. Y. Hsu and K. Y. Yin broke after Washington warned in February 1958 that aid would be reduced without reform. Between 1958 and 1960, the government dismantled the rationing and certificate system and unified the exchange rate. Import controls proper remained substantially in place for longer. The affected industrialists were not central to the émigré KMT’s political base, which limits the episode’s value as evidence of sacrifice within the governing coalition.↩︎
Singapore’s new growth relied heavily on multinational firms rather than an established local business constituency. The later cost stated here is narrower and directly observable: the official CPF contribution-rate series records the sharp 1986 reduction in employer contributions after the 1979 high-wage turn. This does not prove that every labour claim was subordinated to transformation; it identifies one consequential reversal borne by a core constituency.↩︎↩︎
The comparison concerns authorship, not the merits of closure. Thailand began suspending finance companies before the IMF programme, so its closures were not simply foreign orders. The later restoration examples come from the wider stalled record: Indonesia’s Bank Andromeda assets rapidly reappeared under another licence, the IPTN aircraft programme ended under external pressure, and Malaysia’s Renong was rescued on terms preserving the insider behind it.↩︎
GITIC was one of Guangdong’s provincial “window” companies, borrowing abroad on the widely held assumption that Beijing stood behind it. The central government rejected a provincial rescue in October 1998 and pushed the trust into bankruptcy in early 1999, leaving its debts to foreign creditors substantially unpaid. The important qualification is that the refusal can also be read as fiscal federalism: the centre declining to absorb a province’s liabilities. The wider restructuring of China’s state sector makes that reading incomplete, but GITIC alone cannot establish the direction of the whole economy.↩︎
The deadlock and reform sequence are detailed in the licence-rent footnote above; Hsu resigned and Yin implemented the reform in stages. As with that finding, this is stronger evidence of a specific privilege removed for a declared path than of a regime risking a coalition pillar. Import controls also remained after the rationing system was dismantled.↩︎
Pertamina’s concealed debts reached roughly US$10.5 billion, close to one-third of Indonesia’s economy. The government assumed the liabilities, dismantled much of the company’s commercial empire and removed its president, General Ibnu Sutowo, despite his position within the military and patronage system. The episode demonstrates the state’s ability to confront a powerful insider, but its immediate purposes were national solvency and the recentralisation of financial authority, not the imposition of a productive-performance test.↩︎

The Cost of Prioritising Transformation: What Sorted Eight Growing Economies

Stephen Brien — Fri, 31 Jul 2026 09:25:52 GMT

The short version

Eight economies grew fast and alike for a generation, then split: four reached the technology frontier; four never did.
All eight had firms selling into world markets and facing standards their governments did not control.
Governments did control which domestic firms had to meet those standards, and which were protected from them.
Behind that choice sat a governing orientation: when transformation conflicted with another political priority, which one came first?
That answer showed in whether an incumbent retained unconditional shelter, whatever it delivered, with power to shape or obstruct new growth.
Four countries denied that privilege: two withdrew it outright; two built beyond its reach. Four others preserved it and remained inertial.

The first post in this series set out the puzzle; the second ran the usual explanations against these eight: investment, industrial policy, openness. None explained the split. One thing did: market exposure. This post identifies the governing orientation behind that exposure and shows, case by case, how it separates the eight. The next asks whether it could have been read at the time, and how to distinguish the real signal from a counterfeit.

In May 1961, a week after seizing power in Seoul, South Korea’s new military government arrested the fifty-one richest men in the country. It held them for weeks and released them only when each had signed the same undertaking, to hand his fortune to the state’s industrial plan.

Those undertakings were largely never enforced. What the men got instead was an assignment: an industry each, and credit on terms the state set, through banks it had just nationalised. Those arrests are the famous part of this story and the least important part of it. What matters is the arrangement that replaced them.

Ten years later and a few thousand miles west, a government facing a comparable crisis chose the other road. After the 1969 race riots, Malaysia chose the guarantee: an open-ended stake in whatever it built, with no closing date, rather than making its business class earn its place.

The problem was not redistribution itself. It was that the new stake became a durable entitlement rather than a privilege conditional on productive performance.

Both governments said they wanted development. Only one refused its own insiders the privilege of not paying for it.

An unconditional privilege is a standing guarantee that a favoured firm will be sheltered whatever it delivers^[1].

Market exposure is a fact about firms, not about countries, and Malaysia had both kinds at once: firms tested by the world market and firms sheltered from it. Its Penang electronics zone held firms to a genuine export test, one a firm could fail, and they climbed from bare assembly toward design and higher-value work^[2].

But the zone sat outside the country’s own customs border under the Free Trade Zone Act of 1971, and what it made could not be sold at home except as an import, capped at a fifth of a firm’s output and paying full duty. So the discipline stopped at the zone’s boundary, and the domestic champions outside the zone never had to meet the same test.

Korea and Malaysia are one pair. This wider set is eight countries, matched on where they started, how fast they grew, and how long the boom lasted, then split four and four^[3]. That split is not driven by income; all eight closed some of the gap. It is a matter of capability: whether the ability to invent and make hard things became widely embedded in an economy, to reshape its workers, its suppliers, and what it produces. Or it remained within a pocket, walled off from the protected core.

I: What a government is choosing

No government can be taken at its word about what it is aiming at. What shows is what it does, and what it refuses to keep in reserve.

No government decides whether the world market is demanding. What it decides is narrower, and entirely political: which of its own firms must answer to it, and which are protected. Behind that choice sits a governing orientation. It is what a government reaches for and prioritises when the pressure to choose can no longer be deferred, and transformation conflicts with something else it wants. The comfort of its coalition. The survival of a favoured firm. A budget it would rather not break. A community it has promised a fixed share.

That was the choice Malaysia’s 1969 settlement made explicit, and Korea’s monthly export-promotion meetings made just as explicit in the other direction.

A governing orientation is transformative when remaking the economy takes precedence over competing priorities, including the interests a government would ordinarily protect. In examining governing orientation, this post focuses on one particular conflict: the survival of privileged firms.

Crossing those firms is among the most visible and politically costly choices a government can make. Whether the priority arises from genuine political commitment or from the constraints facing the government remains disputed^[4]. This post asks only how that priority can be observed in practice.

This is where a simple theory of state capture should be most persuasive.

Every coalition that loses its protection in the cases that follow was organised and influential enough to defend it. What separates the eight is that some governments still withdrew that protection, while others did not.

A governing orientation is a disposition, not a recorded decision, so it cannot be observed directly. Declarations do not reveal it. What does is the choice a government makes when transformation conflicts with a favoured interest. The clearest test of that orientation is whether a protected firm is allowed to fail on the same terms as everyone else.

Declarations are cheap

Transformation as a priority is exactly the kind of thing every government will claim, and the claim on its own carries no information. All eight countries announced that they wanted to transform, and most built an apparatus to prove it. Korea’s junta announced its own five-year plan before the machinery to enforce it existed. Taiwan’s government declared the same export-first ambition. China called its opening “reform,” a word every government reaches for.

By itself, a declaration states an ambition. It records what a government claims to pursue, allowing its later choices to be judged against that claim. Without it, an action may be visible while its purpose remains unclear, and only a choice made when priorities conflict reveals which one the government places first.

Singapore’s Lee Kuan Yew named the goal at a May Day rally in 1960, before the institutions needed to pursue it even existed.

Seven years later, in 1967, he described the choice his government had made between competing aims. It had decided “the priorities between national purposes and social purposes” and had “put national purposes first.”

This government had promised free schooling and free medical care, yet fell short on both. It lost two by-elections and governed with a majority of one. Its later actions therefore clarified the earlier declaration. When national transformation conflicted with immediate social commitments and political security, the government showed which it placed first.

Brazil, Malaysia, Indonesia and Thailand declared too, but declaring didn’t sort anything at the time. Brazil promised fifty years of progress in five and built a capital city to be proud of; Malaysia declared development and ethnic redistribution side by side, and in practice redistribution governed. Indonesia hit its numbers while the patronage ran offstage, and Thailand declared as often as the others without ever making the claim expensive.

All eight declared a goal. Only four proved they meant it, once that goal grew costly.

Subscribe now

What shows the orientation

A governing orientation is invisible until something forces a choice; then two things can be observed. First is the act, what a government did when transformation and its own political interest pulled apart. An act signals the orientation when it advances the declared aim at a political cost to the government, which is why moves against a government’s own insiders reveal it most clearly. Refusing to rescue a failing champion, forcing a bloated group to sell what it had accumulated, withdrawing a privilege a firm had built its business around.

A declared aim and a powerful insider’s position can sit together for years. Then a moment comes when they cannot, and what a government protects at that moment is what it was actually pursuing.

Such acts are difficult to interpret because reform, a power grab and a private grudge can look identical from the outside. They are also rarer than the export-discipline literature implies. Of the four economies that transformed, only Taiwan provides a clearly dated move against firms holding the privilege that is not entangled with a rival explanation, and even that imposed little political cost.

The second source of evidence is the arrangement itself, the rules and institutions that determine what a firm must do to retain its credit, licence or other privilege. It is also whether those requirements can be waived for a favourite.

This can reveal a government’s orientation without any dramatic confrontation.

What matters most is what the government put in place before pressure arose. Did it preserve the option to rescue or exempt a favoured firm later, or did it close off that option in advance? A government committed to transformation may therefore design its institutions so that the conflict never arises.

Closing off that option is itself revealing. Every government has some power to protect a favoured firm. An arrangement that limits or removes that discretion commits the government in advance to letting the rule apply.

Korea’s export-credit system, examined below, is what that commitment looked like in practice: credit tied to shipment figures, not political standing.

Two features show what the arrangement prioritises: how widely the rule applies, and whether the government can waive it. Neither is enough alone. A rule that covers every firm but bends for favourites is largely symbolic. A rule that cannot be waived but excludes the protected champions leaves the privilege untouched.

The orientation’s sharpest test

A failing favoured firm creates the clearest conflict between transformation and the protection of vested interests. Malaysia’s Renong is the clearest failure of this test, rescued repeatedly rather than let go, its owners’ position preserved throughout. This test is not simply whether the company survives. Governments may have good reasons to preserve productive assets, employment or financial stability, and the four that transformed sometimes did so. By contrast, the sharper question is whether the insiders who held the privilege survive on the same terms.

A rescue can preserve the enterprise while its owners lose equity, its managers lose control, or its political sponsors lose the protection they had secured.

This effect is also prospective. Firms that know failure may cost their owners control or future support have reason to improve before the test arrives. A subsidy, tariff or cheap loan is conditional only if poor performance can eventually cost the protected interest something real.

The test’s relevant unit is the protected interest holding the privilege, not always identical to the company carrying its name.

No documented Korean rescue from this period restores an unconditional shelter outright. Malaysia’s later rescue of Renong changed nothing. The arrangement returned to exactly the shape it had always had.

A government may rescue a productive enterprise without insuring its owners. When favoured insiders retain control and protection despite failure, support has become a guarantee rather than a discipline.

The same test applies before failure arrives, not only after it.

Withdrawing that privilege before success has arrived makes the choice harder to dismiss as reform theatre. A government that does this accepts immediate political risk without yet having results to vindicate its decision^[5].

The treatment of firms that failed and the arrangements governing those that did not divide the eight countries four and four. That division is evidence of differing governing orientations.

Where no favoured firm failed, the inquiry moves back to the arrangement itself. That question becomes whether any protected interest retained a claim strong enough to control or obstruct the path of new growth. Declarations settle neither test. What happens when performance and protection collide reveals priority, or rules written in advance to determine the answer do.

II: The path no privilege governed

Among the four that transformed, one governing orientation took four different forms. Each denied an incumbent’s privilege the power to govern its growth: some by making a favourite face a test it could fail, others by building the growth where no favourite’s claim reached at all.

Successful cases are easiest to misunderstand at exactly this point. A government can break its incumbents spectacularly, nationalising, expropriating or jailing them, and still leave the privilege itself in place. It may simply hand the economy to a sheltered sector that never has to earn its keep^[6]. In that case, ownership changes, but the barrier to transformation remains.

Breaking an incumbent changes who owns the firm; withdrawing its privilege changes whether it must still earn its keep.

A government can equally deny a privilege without ever confronting anyone. Building the new growth where those who would have claimed privilege have no standing to claim it counts as the same refusal, even though nothing is taken and no dramatic act enters the record. What is given up is the ready coalition, not a fight that can be dated. What gets built instead reveals the same priority as a confrontation: protecting an established interest was the easier course. The government did not take it.

This is why the fate of the privilege matters more than the fate of the incumbent. The successful countries did not all attack protected firms, but each prevented an unconditional privilege from governing the path of new growth. A privilege governs a path when its holder can capture what the path produces or block what would displace it, as Korea’s own credit rules show below.

Four governments made that choice in four different ways.

Korea: privilege made conditional

Korea’s governing orientation was transformative. For fourteen years, the state built a system in which no privilege was meant to capture what its path produced or block what would displace it. That is not the 1961 arrests’ doing. Those arrests were coercive, and the pledges signed on release were mostly never collected^[7]; a consolidating junta could stage that scene alone.

What consolidation cannot explain is the fourteen years after. That was a state that made every firm’s privilege conditional^[8] on what it shipped abroad, and held that condition against real, repeated pressure to abandon it.

The arrangement that made this possible had a fixed address. From February 1965 until his death in October 1979, Park Chung-hee presided monthly over an export promotion committee at the Blue House, where ministries and business leaders met to clear what stood in exporters’ way, officials reporting back on whether the country’s targets were being met^[9]. Credit ran through state-nationalised banks, on terms set against those same figures, not political standing. This is Amsden’s account of Korean discipline, and it holds^[10], a standing, personally run mechanism, not a single gesture.

Domestic pressure to shelter connected capital never went away. That same industrial drive sheltered domestic sectors behind tariffs the export lines did not carry, and a chaebol’s, a family-controlled industrial conglomerate’s, credit ran to the group as a whole, not to each plant, so a sheltered domestic arm never had to answer for what it shipped. That tariff wall stopped at the group’s export line; the wall never crossed it.

Nor were the state’s own friends spared. They were mostly bailed out at least once each, and in 1986 fifty-six firms sat under state reorganisation rather than being allowed to fail^[10-1]. What the rescues did not do was restore the privilege outright; support stayed tied to the shipment figures, not handed back the way it would be to a protected incumbent.

The strongest objection to this account is Kukje; when Korea broke up the group in 1985, the standard account, which Amsden shares, read it as a badly run conglomerate punished for failing^[10-2]. Korea’s own Constitutional Court found otherwise in 1993, ruling the dissolution arbitrary, outside the law, and the president’s personal doing. Its chairman had given less to a presidential foundation than his rivals; his companies went to men who gave more.

That is real, documented abuse of the same arrangement this argument rests on, and it does not undo fourteen years of monthly export promotion meetings. One arbitrary dissolution is a genuine cost the mechanism could be turned to, not evidence that the mechanism was never real in the first place.

Nor did 1987 kill the orientation behind the arrangement; it changed what the orientation had to answer to. Winning office now required the conglomerates’ own money, a lever privilege had never held before, precisely because the same discipline that had tested them had also grown them large enough to change what they could demand of it^[11].

For fourteen years, the system Park built made every firm’s credit answer to what it shipped, in design if not in every documented instance^[10-3]. That, not the 1961 arrests, is what marks the orientation as transformative, and the harder politics that followed changed its terms without ending it.

Taiwan: the licence premium stripped

Taiwan’s transformative orientation shows not in attacking privilege everywhere, but in removing it where it obstructed the new growth. It built new institutions to advance the same priority, and left protected interests alone where they could not govern that path.

Where Taiwan did remove privilege, it acted only after a years-long political deadlock finally broke. P.Y. Hsu, the finance minister who controlled foreign exchange policy, spent years defending a rationing system that handed import-licence holders a premium of up to 350 per cent on some goods; K.Y. Yin, who held no formal authority over the policy at all, argued to abolish it.

Washington’s February 1958 confirmation that aid would be cut without reform is what actually broke the deadlock, not a change of mind in Taipei. Told the president had sided with Yin, Hsu resigned his post rather than administer a policy he had spent years resisting; Yin took over the very agency built to run the old system, and set about dismantling it.

Between April 1958 and July 1960 the five-rate structure Hsu had defended^[12] collapsed by stages into one, at NT$40 to the dollar, and the rationing that had handed a favoured few that 350 per cent premium was gone.

That victory was narrower than it looks. Those import-licence holders it cost were never the ruling party’s own people to begin with. The KMT (Kuomintang), an émigré party barely a decade removed from the mainland, had no local business base of any kind to protect, on either side of this fight^[13]. What it did not touch was the tariff wall built up separately; import controls stayed close to where they started for another decade, sidestepping the harder confrontation a full opening would have required.

Where Taiwan built anew, it did so separately and fifteen years later, using a different institution to advance the same priority. The Industrial Technology Research Institute, founded in 1973, transferred its process technology and engineering staff to found UMC in 1980, and repeated the transfer, production line included, in 1987 to help found TSMC, a joint venture with a state development fund as majority owner and the Dutch firm Philips supplying capital and licensed technology^[14]. Hsinchu Science Park, opened in 1980 on the Stanford model, gave both a home.

Clearing a rent is one kind of evidence. Building an institution whose spin-offs had to compete in a global market, fifteen years on, is another. Taiwan gave both.

What Taiwan left untouched was its own party-controlled enterprise network, because it no longer governed the path of new growth. It sat beside the new export path rather than across it. That let it escape the same standard until the late 1980s, sheltered three decades longer than the licences broken in 1958. That is where the path was tested and where it was not, drawn exactly around the new growth, and nowhere else.

Transformation ran exactly as far as Taiwan let it, and no further.

Singapore: built past the claim

Singapore’s transformative orientation shows in two linked choices. First, it built the new economy without granting local business a protected claim on it. Later, once low wages threatened to hold that economy back, it deliberately sacrificed the advantage they provided to force further upgrading.

This first choice was embodied in the Economic Development Board, founded in 1961. It sent its own officers abroad to court multinationals directly, including Texas Instruments, Fairchild, Seiko and National Semiconductor, and built Jurong’s industrial estate speculatively, with roads, power and factory shells in place before a single firm had committed to filling them. Local businessmen who might have expected a protected stake were never given one. They were the natural coalition through which to build, and Singapore built past them.

A privilege handed to a national champion buys loyalty.

That same support handed to a foreign firm buys none, and Singapore extended it anyway. In doing so, the government accepted political risks of its own, cutting workers’ retirement savings^[15], breaking election promises, nearly losing office and dismantling the labour wing the party itself had been built on^[16]. Malaysia took the politically easier course and built its domestic business class into a constituency instead.

The second choice came once the low-cost model had served its purpose. The National Wages Council, tripartite and founded in 1972, first coordinated wage restraint through the 1970s. It then administered the Corrective Wage Policy of 1979 to 1981, recommending increases of roughly 20 per cent a year to price low-value production out and push the economy towards higher-value work^[17]. A government chiefly concerned with preserving employment and the growth model would have kept the low-wage advantage. Singapore chose instead to put that model at risk to move beyond it.

Singapore’s case does not rest on a dated confrontation with a firm. No incumbent stood directly on the path of the new growth. The evidence lies instead in the political and institutional choices through which the government built that path, and in the interests it declined to protect while doing so.

Even so, the state did shelter part of its own economy.

Its telecommunications operator and port authority held long statutory exclusivities that continued deep into the boom^[18]. But those protections did not govern the manufacturing growth being built beside them, and those industrial estates, including Jurong, remained answerable to the same commercial demands as the multinational firms using them.

The absence of a local champion does not alone prove deliberate refusal. Singapore may initially have lacked domestic firms capable of leading exports. What makes the case more revealing is that the government chose policies that did not create a protected local claim, then abandoned the low-wage model once preserving it would have obstructed further upgrading. The first choice kept incumbents from governing the growth path. The second showed that transformation mattered more than the coalition, or the low-wage advantage on which the path had first been built.

Subscribe now

China: built outside, then bound

From late 1978, China let a new economy grow up outside the old plan, even though its expansion weakened the planners and institutions the party had long relied on. Once that exposed sector had become the new growth, Beijing accepted international commitments that made selective protection and reversal more difficult. Two linked choices, in that order, are the shape of China’s transformative orientation.

From late 1978, farmers in villages like Xiaogang began keeping, and increasingly selling, what they grew above a state quota, the household responsibility system that would spread nationwide within five years^[19].

In August 1980, Beijing designated Shenzhen and two neighbouring Guangdong sites Special Economic Zones, a fourth followed that October in Fujian^[20]. Both moves, the household responsibility system and the SEZs, were domestic, years before any foreign government asked for either.

Deng Xiaoping never broke the old central planners; the reformers let a new economy grow up beside them instead, in the coastal zones, the township enterprises, and the private firms outside the old plan, the first half of the pattern. It grew where the old guard had no claim. Then came the second; Beijing bound that new sector to a rule its own incumbents could not lobby away^[21], its 2001 entry into the World Trade Organisation. The veto of the old guard was not defeated. It was made moot.

That pattern reached further than any single rule. Inside the party, local officials’ promotion prospects came to depend on their region’s growth, an effect strongest at the provincial level and weaker nearer the top, where factional ties mattered as much or more^[22]. Even the people running the system had a stake in the reforms’ own success, not just the firms growing under those reforms. Career incentives explain why officials cooperated with the plan; they do not explain why the plan’s own designers gave up the power to reverse it.

No firm here shows China in confrontation. It sheltered its commanding-heights firms and never withdrew that shelter.

That is not a break in the pattern. The interest transformation had to outrank here was never a firm. It was the plan itself, and the party whose legitimacy had been built on it.

Those walls around that plan were real. Private firms were kept out of oil, telecoms, power and the rest of the commanding heights, yet the growth did not come from behind those walls. It came from the sectors outside, where state firms held nothing like the core’s shelter. More than four in five state industrial firms of 1998 were shut or sold by 2007, and the private firms that grew over those same years account for the bulk of what followed^[23].

A government only performing reform could have kept an exemption in reserve, a carve-out for a connected firm. Beijing gave up that room instead. By 2001 the exposed sector was the new growth, and WTO accession narrowed Beijing’s power to shelter it just as it mattered most.

China’s premier, Zhu Rongji, went to Washington in April 1999 with concessions he had already joked would see him “kicked out of office again” if he described them at home.

Clinton declined the deal, the administration posted a summary of the concessions online, and officials in Beijing read what their premier had offered. Zhu was called a traitor and offered to resign more than once. He did not withdraw the concessions^[24].

Two qualifications. That shelter has since grown back, the state sector reconsolidated and the exposed private sector squeezed, so exposure to the world’s verdict was a phase, not a fixed trait, and does not undo what happened while it held^[25]. Property and finance stayed shielded throughout and sit outside this account. The claim tracks what an economy makes and sells, not what it lends or builds for itself.

How much of this account is available in the rules themselves, before the outcome arrives to confirm it, is the next post’s opening question.

One orientation, four methods

All four shared a transformative governing orientation, expressed as a refusal. Neither capture nor block appears among the four^[26].

Four avoided it two ways.

Korea and Taiwan withdrew a privilege already held. Singapore and China built the new growth where existing privilege had no standing to reach it at all, a bypass. Nothing is taken, and no act reaches the record, but the refusal is the same.

A bypass is harder to date than a broken promise, not harder to establish. Political science treats a government’s sustained refusal to give an available claimant what it was positioned to demand as a real, attributable choice^[27], and both cases had one, Singapore’s businessmen, China’s planners. Route depended on room, not just will. Korea and Taiwan’s incumbents already held the space; Singapore and China had room to build a parallel economy from nothing.

China alone also kept a shelter, its commanding-heights core, real but never in the new growth’s way. No competitive reach into the exposed sector^[28] meant nothing could capture or block it. The usual proxy for reform seriousness, selling off the state sector, would have scored this wrong at the commanding heights, even as China sold off most of its industrial state sector everywhere else. That same question asked of Malaysia’s zone reaches the opposite verdict here. China kept its commanding heights and transformed; Malaysia built a real zone and didn’t.

Those methods, and the kind of privilege denied, varied further still: Taiwan abolished a rent the state had created, Korea made one contingent on performance^[29], Singapore and China denied privilege any purchase on the new growth at all.

One outcome sets Singapore apart. Its frontier activity was real, but much of it was owned elsewhere; local companies did about a fifth of business research and a tenth of private patents^[30]^[31]. Korea, Taiwan and China came to own what they built; Singapore didn’t. That orientation was identical, only the outcome diverged.

Whether this account could have been reached at the time, and whether it can be told apart from a strongman’s purge, a lender’s orders, or a shortage no government could have refused, both resurface at this post’s close.

What convinces here is cost. In every case, something real was given up.

III: The path a privilege still governed

Each of the other four countries had its own governing orientation, and in each, something else outranked transformation when the two conflicted. What varied was the competing priority; what did not was the result. The shelter held, and the firms it protected never faced what the rest of the economy did.

Preserving the privilege: the inertial four

Each of the inertial four had a governing orientation aimed elsewhere. Each let an incumbent’s privilege govern the path that carried its growth, and each retained meaningful policy instruments through which that privilege could have been narrowed.

What decided whether those instruments were used was never the government’s capacity to act, but whether the target sat inside its own coalition or outside it. Several imposed high costs while doing it, and none of those costs fell on the firms that held the privilege.

Malaysia. Malaysia‘s governing orientation was ethnic-redistributive, not developmental. After the 1969 race riots, the New Economic Policy made a rising Malay share of the economy the price of ethnic peace, enforced through equity quotas a firm could not simply outgrow.

Transformation asks which firm can compete. That Malaysian settlement asked which race owned it, and when the two conflicted, the settlement won. Breaking the bargain risked the violence it was created to prevent; forgoing transformation risked only a slower growth rate, and no government spends its own survival to buy a growth number.

Malaysia’s prime minister said as much himself in 1986, at a recession’s worst. He would modify the equity rules to meet the employment ones, so investors knew when the government would forgo equity for jobs. That same choice held for decades after, through three Renong rescues and a 1998 finance minister sacked the moment his reform to restructure the crony conglomerates was closest to succeeding.

Electronics was the one sector exempted from the ethnic-equity rule, a real concession, not an oversight. The government gave up ground on its own core commitment to capture the investment. That is why its exporters could grow to two-fifths of all exports behind a wall no export test was ever built to cross. In practice, the Penang zone confirms the settlement rather than contradicting it; exposure was allowed only where it could never touch a firm the settlement protected.

The best-known account of this pattern says Malaysia simply lacked the test^[32]. It had one. Losers were culled at the factory door, on foreigners, inside the zone. A zone that genuinely tests firms and a settlement that genuinely protects them are not supposed to coexist this comfortably. That they do is the case’s real difficulty, not a loose end to explain away. It fails the orientation test, whether what the world rewards ever reached the firms whose shelter governed the path. It never did.

Indonesia. Indonesia’s governing orientation was patronage-preserving: President Suharto’s closest business partner was a Fujian-born trader named Liem Sioe Liong, whose Salim Group grew, on the back of that friendship, into the largest of the connected conglomerates. Discipline a stranger’s firm and the state loses nothing. Discipline Liem’s, or one like it, and the ruler risks the coalition that keeps him in power. That asymmetry, not any lack of state capacity, decided which firms the arrangement tested and which it never touched.

This same asymmetry explains the one clear exception. In 1975, when General Ibnu Sutowo’s Pertamina had run up debt worth roughly a third of the economy, the technocrats broke up his empire and removed him regardless of his rank. He fell for threatening the state’s solvency, on which the whole coalition depended, not for failing to compete, which threatened only himself.

The president’s son’s $690m “national car” never faced that test, because a failing car project threatened nothing but its own balance sheet. It imported Korean cars wholesale, made almost nothing itself, and held privileges no competing assembler could get. If capability had set the line, it should have failed as visibly as Sutowo’s empire did. It never was, because unlike Sutowo, it never put the coalition itself at risk.

Thailand. When General Sarit Thanarat seized power in 1957, he fused the military-bureaucratic state directly to ethnic-Chinese banking dynasties. Thailand‘s governing orientation was coalition-preserving from that fusion on, its ruling apparatus never separate from its biggest commercial families. This was not an ally kept close, as in Malaysia and Indonesia, but a fusion neither could unwind.

The going rate for that fusion was frank. By his death in 1963 his own estate ran to some 2.8 billion baht, fifteen companies, and a share of the beer monopoly. That was not a lapse the system meant to prevent. It was what the coalition cost, and what it bought.

That is why the 1997 cull reached fifty-eight finance companies within weeks, and did not reach the banks until a separate reckoning the following year. The finance companies sat outside the coalition Sarit had built; the banks sat inside it.

Its carmakers show the same principle in a different form. The one standing rule counted how much of a car was assembled locally, a measure that never asked whether the firm behind it could compete.

Yet the state was not short of capacity. In 1984 the prime minister held a currency devaluation against a general who denounced him on television, spending real political capital on the exchange rate. Its investment board approved certificates for over a decade without once withdrawing one for poor performance. Capacity for macro discipline and firm discipline both existed. Only the second was never used on the firms that mattered, the ones whose failure would have meant testing the coalition’s own partners.

Brazil. In 1964, a military coup stopped a populist government from redistributing to workers, backed by the landowners and industrialists who wanted exactly that stopped. From that moment on, Brazil’s governing orientation was elite-preserving. What followed had one constituency to answer to, and it was not labour.

An unelected regime with no ballot box could not afford to lose that constituency’s backing, while workers had already been stripped of any power to reward or punish the regime in return. Testing the coalition was existential. Testing workers cost nothing at all.

That is why the same regime could jail union leaders and cut real wages for years, as sustained policy, at no political cost for over a decade. By 1974 the minimum wage bought about half what it had bought in 1960, and workers had nothing left to withdraw.

Its landlords and its computer industry, shielded by law from foreign competition^[33], were a different matter. They were the only support the regime had, and neither was ever asked to earn what workers were made to pay for. The computer industry’s shelter was written into law; the landlords’ rested on a law never enforced against them. Both held through the same years the wage floor was falling.

That same shelter reached even a foreign firm, because by the time anyone considered enforcing it, Fiat sat atop a pyramid of Brazilian suppliers, dealers, and workers who depended on the plant staying open. Its export deal required 155,000 engines a year; when it missed, the target was rewritten rather than the deal pulled. In Korea, the deal was subject to the test; in Brazil, the test was subject to the deal.

A price sustained that long, paid entirely by the powerless and never asked of the propertied, is a ranking this regime held, not a limit on what it could do.

None of this was a shortage of capacity. Malaysia administered a national resource-transfer program for two decades; Indonesia dismantled a state within a state when Sutowo’s debts threatened it; Thailand’s own investment board held enforcement power it simply never used against a firm that mattered. Malaysia’s two-fifths of exports running through a genuinely competitive zone, and Thailand’s fifty-eight firms liquidated in weeks, both look like evidence of transformation until you ask which firms they left untouched.

The chances they let pass

Preserving the privilege at the decisive moment might mean the orientation was absent, or only that acting was too costly then. A government’s orientation can be tested more than once. Any moment that lowers the political cost of withdrawing a privilege is a second chance to see whether the orientation was ever there. What the inertial four did later, when the cost of acting fell, resolves this distinction. If the orientation had been dormant, the moments that made withdrawal cheap were when it should have surfaced.

Three kinds of pressure lower that cost, each harder to resist than the last. A corruption probe makes moving against a protected insider look like duty rather than betrayal. An internal power struggle can break a coalition’s old bargains outright. And a financial crisis lets a government drop a failing champion and blame the shock. Each of the inertial four met at least one of these pressures, and let it pass every time.

Indonesia’s headline corruption prosecutions proved narrower than they looked^[34]. Separately, the financial crisis produced its own version of the same pattern. One of its biggest conglomerates ceded its bank and a hundred companies, kept the one arm that mattered, and used it to buy back most of the rest^[35].

Brazil’s corruption drive was real but bounded, its biggest conviction later annulled and its meatpacking principals back on the board within six years^[36]. The reform that ended automatic cheap state credit withdrew something real, though the guarantees that mattered survived it, carried forward by a different government than the one that built them. Where prosecution came, it fell on individuals, was time-limited, and left the shelter intact.

Only Malaysia had a real internal power struggle, fought within the ruling coalition and lost there: an Attorney-General removed, a Deputy PM sacked, and the investigating task force disbanded^[37]. Had the reformers won it inside UMNO (the ruling party since independence), it would have counted. Instead the 1MDB reckoning came only in 2018, after a change of government, never from the coalition that had run the country; the earlier state financed its own carmaker’s rescue rather than expose it to failure^[38].

A financial crisis is the cheapest of the three, because it asks the least of a government. When a weak champion is already failing, the government needs only decline to catch it. Thailand’s 1997 cull came at that exact moment, and it fell on the finance companies, not the banks that anchored the system. Malaysia in its 1997-98 crash and Brazil in its 1982 debt crisis each reached the same point, where letting go took no more than standing still. And each instead reached in to hold the firm up^[39].

When one did let go, the hand was not its own. Indonesia closed Bank Andromeda in 1997 under IMF pressure, then took over an existing small bank and rehoused the same assets and staff inside it within weeks, a switch the resolution programme’s own credibility never survived. Its prestige aircraft venture was switched off only when the Fund finally forced the issue. Pressed to let its debt-laden champion Renong face the market, Malaysia rescued it instead, on the terms already seen.

One qualification matters here. Thailand’s and Malaysia’s financial-crisis windows were the same 1997 shock, not two independent tests, so what follows rests on three mechanisms across the four countries, not four separate trials of each.

Across three different kinds of moment, each of the four inertial governments had at least one chance to drop a privilege at no cost to itself. Every time, the privilege held. That convergence across three unrelated mechanisms is what rules out the alternative explanation. Not that the orientation was present but too expensive to spend; it was never there to spend. That same test that made a declaration cheap makes this cheap too, measured by what it cost the coalition, not that it happened.

Subscribe now

The later record

Symmetrically, the same test runs in reverse, more briefly, for the four transformers. That discipline they set was not a single year’s stance: Korea’s export review ran fourteen years, Taiwan carried the same discipline through chiefs who overlapped in office from the 1950s into the 1980s^[40], Singapore went on raising its own costs, and China bound itself two decades after the opening began.

Two of the four democratised across this same period. Korea’s arrangement did not fully survive its own 1987 transition; Taiwan’s did, carrying its economic discipline through its own political liberalisation in the years that followed^[41].

On both sides of the split, then, the later record repeats the decisive moment: each government’s orientation held. The four that guaranteed privilege kept guaranteeing when stopping was cheap, and the four that denied it kept denying even when quitting would have cost nothing.

IV: What this establishes, and what it does not yet

Across all eight countries, the same distinction holds. Either privileged interests were prevented from controlling the path of new growth, or they retained the power to shape it. What remains is to test whether this explanation survives the main rival accounts and to define where its own limits lie.

What this establishes

The problem is a familiar one^[42]. Organised interests can seize the gains from growth and turn them towards preserving their own position. This post has provided a case-by-case account of which of the eight governments prevented that seizure, rather than merely promising to do so.

Within this matched set, similar starting conditions produced two outcomes: four economies became transformative, while four remained inertial. In every case, governing orientation came down to the same test: when transformation conflicted with another priority, did transformation come first?

A transformative orientation denied an incumbent’s privilege the power to govern the path carrying its growth, however well connected the owners it disappointed. That denial reads the same whether won by breaking a privilege outright or building past it before any claim was staked; both cost a government something only one that meant it would pay. An orientation pointed elsewhere accommodated that privilege instead, and let it govern there. A privilege governs a path when the firms holding it can capture what it produces or block what would displace them.

What it does not yet

The split itself describes an observable institutional pattern, not yet the governing priority behind it. That pattern consists of the rules and practices governing privileged firms: which interests receive protection, what they must do to retain it, whether exceptions can be made for favourites, and whether those interests can control or obstruct the path of new growth. Governing orientation is the priority inferred from that pattern. The distinction is easy to collapse, but the evidence establishes the arrangement more directly than the motive or disposition behind it.

The test also has limits of application. It cannot classify a government where no meaningful conflict has arisen. A windfall may absorb the pressure and postpone the choice^[43]. An economy that never attempted to build a technology frontier of its own may never have put a champion’s privilege to this test^[44]. A young or unusually fortunate government may simply not yet have faced a moment in which protection and transformation came into conflict. In each case, the absence of a revealing choice is not evidence either way.

There are limits to what can be inferred from the eight cases themselves. The cases were selected before their outcomes diverged^[3-1], rather than assembled around the eventual finding. But the criterion was refined against them whenever a case exposed a weakness in an earlier formulation. It is therefore possible that the final wording fits these cases more closely than it would fit others^[45]. Nor can a matched set of eight exclude every nearby explanation, including one based on shared regional or historical conditions rather than governing orientation^[46]. A stronger test would apply the criterion to cases selected before their outcomes were known.

The evidence also supports a narrower conclusion about state capacity. General administrative weakness does not appear to explain the split, although the comparison does not establish that all eight possessed the same sector-specific technical capabilities. The missing element was therefore not simply an administrative tool that an outside donor could provide. The deeper constraint may have been the government’s priorities, the political room available to act on them, or some interaction between the two.

Finally, the evidence does not all carry the same weight. A politically difficult act is more revealing than an easy gesture, while an institutional arrangement maintained through repeated tests is stronger evidence than either taken alone. The comparison identifies this hierarchy but does not turn it into a precise rule.

What stands, then, is a four-and-four division in the treatment of privilege. What remains less certain is how confidently that pattern reveals governing orientation, how early it could have been recognised, and how reliably it can be distinguished from coercion, external pressure or structural necessity.

This post began with a puzzle about why eight countries that grew alike eventually diverged. It ends with a question that can be applied beyond the sample: does any favoured interest still possess a privilege strong enough to control or obstruct the path of new growth?

The same question applies across both routes identified here. A government answers it directly when it withdraws the shelter of a favoured firm and allows failure to carry consequences for the insiders behind it. It also answers it by building the new economy where protected interests cannot claim control, then maintaining that boundary when it has the opportunity to relax it.

The wider implication is that governing priorities need not remain wholly hidden. A firm’s productive capability is usually treated as measurable, while a government’s orientation is treated as a matter of rhetoric or conjecture. But orientation becomes more observable when repeated choices and institutional rules reveal which interests are protected when protection conflicts with transformation. That evidence is less precise than a productivity statistic, but far more revealing than a declaration alone.

Notes

North, Wallis and Weingast’s natural-state/open-access framework (Violence and Social Orders, 2009) reads this same territory at the scale of a whole economy. Their test requires impersonal access to political and economic organisations alike. The test this post applies is narrower and sits inside their natural state, not beyond it. China and Singapore pass this narrower test while remaining, on the wider one, natural states by the same criteria scholars working in their tradition already apply to each.↩︎
The zone’s climb from assembly toward design is documented in the trade literature on Penang specifically: firms there were confined to simple assembly through the 1970s and 1980s, and by the 2010s some had moved into component design and testing, Penang hosting global design centres for firms including AMD and Altera (Athukorala, 2014). The aggregate skepticism about Southeast Asian technology dependence is not, on its own, a finding about this zone; Studwell’s own account of Malaysia centres on Proton and Perwaja and never engages Penang’s electronics sector directly.↩︎
The eight were fixed before any of them had split, picked in the earlier posts as economies matched on their growth record and their declared ambition, then watched as they diverged. So the cleavage reported here is a finding about a set chosen in advance, not a set assembled around the finding. How common the division is across developing economies at large is a separate question this matched design cannot answer.↩︎↩︎
This is Khan’s own challenge to any account that reads withholding a privilege as a chosen priority: whether a coalition can withhold a privilege, on his account, is set by the distribution of holding power a settlement already encodes, not by what any government would prefer. The next post takes that challenge on directly rather than assuming it away here.↩︎
What such a cost signals, and what firms then believed, is the third post’s subject. The question here is narrower, which move exactly marks a transforming government; the negative cases here pin it down, denying an unconditional privilege where the rents and the power sit.↩︎
Beyond the eight, several governments broke a dominant incumbent class far more violently than any renewer and still did not transform, because they handed the economy to a sheltered state or patronage sector that never met the world’s test, so the way out stayed open even though the old owners were gone. Nasser’s Egypt, Velasco’s Peru and Nyerere’s Tanzania are the clean peacetime cases: each broke a dominant incumbent class, at real cost to it, and each stalled. Waldner’s account anticipates exactly this, that intense elite conflict pushes rulers toward broad side-payment coalitions and the exemption-preserving state, so dramatic confrontation tends to produce the open exit, not close it.↩︎
Several of these were coercive acts, and some the work of governments that curtailed political liberty as well as economic privilege: arrest without trial, expropriation, the suppression of unions and opposition. To describe them is not to endorse them, and the same powers have elsewhere produced only predation. The claim here is causal, not moral. Bearing a cost against its own most powerful firms shows a government meant to change its economy, not that its means were lawful, gentle, or right.↩︎↩︎
Kang reads Korea’s discipline not as an autonomous state rewarding performance but as a mutual-hostages bargain between a strong state and a few concentrated groups, in which the chaebol kept large rents and, on his reading, used personal ties to avoid sanction rather than earn conditional support. That is a real dispute, not a reconciled one; what survives it is taken up at ^[7-1].↩︎
Irwin, D. A., “From Hermit Kingdom to Miracle on the Han: Policy Decisions that Transformed South Korea into an Export Powerhouse,” NBER Working Paper 29299 (2021), p.31, describes the committee’s composition and monthly business: government ministries, the central bank, and the Korea Trade Promotion Agency sat with industrial firms, chambers of commerce, and banks and financial institutions, and firms used the meeting to raise export obstacles while officials reported on whether the country’s targets were being met. The paper does not give the committee a Korean name, and does not describe individual firms’ shipments being reviewed line by line; both claims appeared in an earlier draft of this section and have been removed as unsupported by the source cited for them.↩︎
The review ran monthly from February 1965 until Park’s death in October 1979, 177 meetings in all, chaired at the Blue House. On whom the discipline fell, Amsden (1989, p.16) is direct: the sternest discipline reached large firms “no matter how politically well connected.” The bailout and reorganisation figures given in the text above are hers. A further limit belongs with them: no named Park-era firm stripped of credit for missing a target has been located in the sources held for this post: the mechanism is documented, a clean single instance is not. On the Kukje case taken up in the text above, Amsden groups it with Korea Shipbuilding as “badly managed chaebol… that the government recently punished with dismemberment”; the 1993 ruling is reported by UPI, 29 July 1993, and by Jane Han, “Ex-Kukje Chairman Dies,” The Korea Times, 29 March 2009. She does name other firms that lost state credit, Shinjin and Asia Motors among them, but as mismanagement or market failure rather than a missed export target. The strong version of the performance claim also has a serious dissenter. Kang (2002) argues that Korean subsidies were not exchanged for performance standards at all, reading the era’s overcapacity as evidence that allocation stayed political, and names Amsden’s performance-standards passage as his target. The reading here does not need the strong version. It needs only that credit ran through the state and was set against exports. But Kang’s challenge cuts deeper than that alone answers. If credit tracked exports in name only, and firms bought their way out of sanctions through personal ties, the arrangement was conditional on paper and unconditional in practice, which this post’s own test would then read as a standing shelter, not its withdrawal. The 1986 reorganisations are compatible with either account. ↩︎↩︎↩︎↩︎
The 1987 turn is not only a political shock landing on the arrangement from outside it. The same discipline that scored credit against export performance is what grew the chaebol into entities large enough that, once winning office required outside money, only they could supply it.↩︎
Pre-reform Taiwan rationed foreign exchange through import licences that handed recipients premiums as high as 350 per cent on some goods (Irwin 2021, p.8, citing Lin 1973). P.Y. Hsu, the finance minister who controlled the system, defended it for years against K.Y. Yin, who held no formal authority over the policy; the deadlock broke only in February 1958, when Washington confirmed aid would be cut without reform. Hsu resigned rather than administer the change; Yin took over the agency built to run the old system and dismantled it in stages: a dual rate announced 12 April 1958, a devaluation to NT$36 that November, certificate and official rates merged August 1959, rationing itself ended that September, full unification at NT$40 in July 1960. Import controls proper, as distinct from the rationing that generated the licence premiums, were not substantially relaxed for years more (Haggard and Peng 1994, p.76, cited in Irwin 2021, p.26). Irwin, D. A., “Economic ideas and Taiwan’s shift to export promotion in the 1950s,” The World Economy 46(4) (2023): 969-990, also circulated as NBER Working Paper 29298 (2021), pp.8, 19-26; Wade, R., Governing the Market: Economic Theory and the Role of Government in East Asian Industrialization (Princeton University Press, 1990), Appendix B and pp. 118-119, 392-393.↩︎
Taiwan’s better-known reform is the land redistribution of 1949 to 1953, often cited as its transformative credential. It fell on a landlord class the KMT had no ties to either, which makes it a weaker sign of intent on the same grounds as the licence-stripping, not an independent piece of evidence, so it is not relied on here.↩︎
ITRI, a government applied-research institute, was established 5 July 1973. It transferred its pilot production line, process technology and engineering staff to found UMC on 22 May 1980, and repeated the transfer in 1987 to help found TSMC, structured as a joint venture with a Taiwan government development fund as majority shareholder, Philips as a substantial minority partner supplying capital and licensed manufacturing technology, and Taiwanese private industrial capital making up the remainder; exact ownership percentages are reported inconsistently across secondary sources and are not given here pending verification against TSMC’s own primary filings. Hsinchu Science Park, opened 15 December 1980 and explicitly modelled on Stanford Industrial Park, sited both firms. Amsden, A. & Chu, W.-w., Beyond Late Development: Taiwan’s Upgrading Policies (MIT Press, 2003); Computer History Museum, oral history interview with Chintay Shih, on the ITRI-to-TSMC technology and staff transfer, https://archive.computerhistory.org/resources/access/text/2012/02/102745999-05-01-acc.pdf; founding dates from ITRI’s and UMC’s own published institutional histories.↩︎
Lim Chong Yah on the National Wages Council, Malaya Law Review 23(1) (1981); CPF employer-contribution rates, 1955-2026 (dataset), Central Provident Fund Board, via data.gov.sg: https://data.gov.sg/datasets/d_98ffa142ae0dec40391f78f81d26aca9/view. The employer rate held at 25 per cent through July 1985 before the cut to 10 per cent took effect in April 1986.↩︎
van Elkan, R., “Singapore’s Development Strategy,” in Bercuson (ed.), Singapore: A Case Study in Rapid Development (IMF, 1995); Employment Act 1968 and Industrial Relations (Amendment) Act 1968 (Singapore Statutes Online).↩︎
The cost this post reads is what a government was willing to make its own insiders absorb. It is not a ledger of what labour absorbed to get there. Singapore cut retirement savings and dismantled its own labour wing; China’s transition idled a large share of state-sector workers; Korea’s discipline ran under restricted independent unions throughout. Those costs are real, and they are not weighed here against the ones that are.↩︎
Singapore’s state-linked utilities held monopoly power of their own for decades, legislated or natural. The Port of Singapore Authority and the Telecommunication Authority of Singapore both sat among the statutory boards whose earnings the government relied on for public-sector savings, a position documented at least into the mid-1980s. Huff, W. G., The Economic Growth of Singapore: Trade and Development in the Twentieth Century (Cambridge University Press, 1994), p.334; Lim, Linda, Pang Eng Fong and Ronald Findlay, “Singapore,” in Findlay and Wellisz (eds.), The Political Economy of Poverty, Equity, and Growth (World Bank/Oxford University Press, 1993), pp.108–109.↩︎
Lin, Justin Y., “Rural Reforms and Agricultural Growth in China,” American Economic Review 82, no. 1 (March 1992): 34–51. Lin’s account confirms decollectivization ran from village-level experiments in 1978 to full official sanction only by late 1981, reaching 98 per cent of production teams by the end of 1983: a rollout, not a single legal event.↩︎
The Standing Committee of the National People’s Congress approved Shenzhen, Zhuhai and Shantou as Special Economic Zones on 26 August 1980; Xiamen followed that October.↩︎
What the reading here adds is the firm-level thing you can observe.↩︎
Li, H. and Zhou, L.-A., “Political Turnover and Economic Performance: The Incentive Role of Personnel Control in China,” Journal of Public Economics 89 (2005): 1743–1762, finds a real promotion-growth link at the provincial level. Shih, V., Adolph, C. and Liu, M., “Getting Ahead in the Communist Party,” American Political Science Review 106, no. 1 (2012): 166–187, finds no significant effect on advancement into the Central Committee’s top bodies, where factional ties dominate instead.↩︎
Hsieh, C.-T. and Song, Z., “Grasp the Large, Let Go the Small: The Transformation of the State Sector in China,” Brookings Papers on Economic Activity (Spring 2015): 295–366, at p. 297. More than 83 per cent of the state-owned firms in the industrial sector in 1998 were shut down or privatised by 2007, and the growth of private firms accounts for 70 to 80 per cent of aggregate growth after 1998. The authors are explicit that the exit of state firms contributed little directly, about 3 per cent of industrial growth over the period, and they caution against a simple markets-over-Mao reading of China’s growth. The claim here rests on their private-sector finding, not on the exit statistic. The authors themselves call that private-sector share an unexplained residual, not a settled causal account of what drove it, and this claim asks no more of it than they do.↩︎
Joseph Fewsmith, “China and the WTO: The Politics Behind the Agreement,” NBR Analysis 10, no. 5 (National Bureau of Asian Research, November 1999). Fewsmith records Clinton declining the deal on 7 April, the attacks on Jiang Zemin for “selling out the country,” and Zhu having “offered up his resignation repeatedly in the months since.” The remark about being kicked out of office is Zhu’s own, to the Wall Street Journal (Gerald F. Seib, “Kosovo, China May Spur Some Isolated Thoughts,” 14 April 1999, p.A28), quoted directly in Devereaux, C., Lawrence, R.Z. & Watkins, M.D., Case Studies in US Trade Negotiation, Vol. 1: Making the Rules (Institute for International Economics, 2006), ch.6, p.273 n.75. ↩︎
Since 2020, the same reconsolidation has reached firms well outside the state sector too. Ant Group’s initial public offering, on track to be the largest in history, was suspended two days before its scheduled listing in November 2020; a state campaign under the banner of “common prosperity” followed within the year; and the property developer Evergrande defaulted in December 2021, its restructuring later collapsing into compulsory liquidation.↩︎
Acemoglu and Robinson’s political replacement effect (2006) names a nearby mechanism: incumbents block innovations that would erode their own political position, even when adopting them would raise aggregate output. What this post’s test adds is not the mechanism but the measurement: it asks the question at the level of a standing arrangement, readable off what a government has built or refused to build, rather than at the level of any single innovation. Across all eight cases here, what the record shows is block and exemption, not capture; whether a privilege can also govern a path by capturing what it produces, rather than only holding it off, is a form this test has not yet been asked to find.↩︎
Holland, A. C., “Forbearance,” American Political Science Review 110, no. 2 (2016): 232–246, and Forbearance as Redistribution: The Politics of Informal Welfare in Latin America (Cambridge University Press, 2017), treats a government’s deliberate non-enforcement of a claim it could have honoured as a real, attributable political choice, not an absence of temptation. The literature does not settle this cleanly: Hanson, J. K. and Sigman, R., “Leviathan’s Latent Dimensions,” Journal of Politics 83, no. 4 (2021): 1495–1510, warn that state non-action can equally reflect incapacity as choice. Neither study addresses industrial protection specifically; both are imported here from adjacent literatures, not native to this argument’s own.↩︎
SingTel and PSA, and China’s commanding-heights core, differ from Malaysia’s champions in one respect. Nothing about a telecoms monopoly or a state oil major stood in competitive reach of export manufacturing, so there was no rival path for either to block or capture even in principle. Malaysia’s champions stood in exactly the competitive reach SingTel and PSA did not: the zone next door.↩︎
Khan’s account of rents for learning runs closest of all: a rent made conditional on a monitored performance the state can withdraw support for, close kin to the conditional-versus-unconditional privilege drawn here. Khan reads that conditionality as an output of what a political settlement can enforce; this post treats withholding the privilege as the move a coalition makes, or fails to make, inside the room the settlement leaves it.↩︎
Singapore’s Agency for Science, Technology and Research reports locally-owned companies performing 19.5% of business R&D and receiving 10.6% of private-sector patent applications in 2023. Both figures are upper bounds, since the definition counts a firm up to 70% foreign-owned as local. Its three Fortune Global 500 entries are Trafigura, Wilmar and Olam, a commodity trader, a palm-oil group and a food company.↩︎
Hong Kong’s exclusion, taken up at ^[36-1], rests on total research spending, not on who owns it. Singapore’s inclusion here rests on ownership share, not on the total. A reader checking the two boundary cases against each other deserves both measured the same way: on the same total-spending basis (World Bank gross domestic expenditure on R&D, per cent of GDP), Singapore ran at roughly 1.7 to 2.2 per cent across 1998 to 2022, two to four times Hong Kong’s, which stayed under one per cent across nearly the same span.↩︎
The nearest widely-read version of this comparison is Studwell’s How Asia Works (2013), which popularised for a general readership an export-discipline mechanism whose scholarly roots run through Alice Amsden (1989) and Robert Wade (1990); it contrasts the Northeast Asian successes with the inertial economies of Southeast Asia and centres export discipline: holding subsidised firms to an export test and culling the losers. The two scholars’ own accounts are broader than the popularised version. Amsden’s own term is reciprocity, a state-firm exchange of support for performance, not one confined to exports; Wade’s governed market names credit allocation, trade protection and technology policy as one coordinated package. Taiwan’s upstream state sector does not supply a clean parallel. Wade’s own account groups China Steel and China Petroleum among the public enterprises that “tend to make big profits” (1990, p.180), and records complaints that China Steel’s domestic pricing ran up to 20 per cent above the landed cost of Japanese steel, a margin waived only for steel bound for export, not shown below cost (p.131). Wade raises, but explicitly does not resolve, whether any Taiwanese public enterprise subsidised exporters this way (p.143). The clearer analogue to below-cost export pricing in his book is Korean: Hyundai’s export cars, priced below production cost with domestic sales covering the gap (p.310). Studwell codes Malaysia as simply lacking the test.↩︎
Luzio, E. & Greenstein, S., “Measuring the Performance of a Protected Infant Industry: The Case of Brazilian Microcomputers,” Review of Economics and Statistics 77(4) (1995): 622–633. https://doi.org/10.2307/2109811 ↩︎
Winters, J. A., Oligarchy (Cambridge University Press, 2011), p. 184 n.54, Cambridge Core; Robison, R. & Hadiz, V., Reorganising Power in Indonesia (RoutledgeCurzon, 2004), pp. 208–209, Taylor & Francis.↩︎
Robison, R. & Hadiz, V., Reorganising Power in Indonesia (RoutledgeCurzon, 2004), pp. 190–191, 193 (Taylor & Francis); “How Salim Group reemerged from near bankruptcy to vast conglomerate,” The Jakarta Post, 9 June 2014.↩︎
World Bank, “Understanding the Effects of the Taxa de Longo Prazo (TLP) Reform in Brazil,” Staff Note 1, July 2017. The stated rationale is fiscal and monetary throughout; the note makes no reference to corruption or Lava Jato. On the principals: STF annuls all Lava Jato acts against Marcelo Odebrecht, 21 May 2024, upheld 6 September 2024 (STF); JBS shareholders return the Batista brothers to the board, 26 April 2024, six years after they left it (Agriculture Dive). On the operation’s durability, Prado, M. M. & de Assis Machado, M. R., “Using Criminal Law to Fight Corruption,” American Journal of Comparative Law 69(4) (2021), 834–879 (DOI).↩︎↩︎
Al Jazeera, “Malaysia PM sacks deputy amid corruption scandal,” 28 July 2015 (deputy prime minister sacked; attorney-general removed); The Edge Malaysia, “1MDB special task force disbanded, says MACC,” 6 August 2015. The reasons given for the attorney-general’s removal are contested on the public record and no motive is asserted here.↩︎
The Edge Malaysia, “Proton gets RM1.5b loan from govt, with conditions for restructuring attached,” 8 April 2016; Zhejiang Geely Holding Group, “China’s Zhejiang Geely Holding Group Signs Final Contract with Malaysia’s DRB-HICOM,” 22 June 2017; Bhattacharya, P. & Hutchinson, F. E., “Shifting to a Higher Gear: The Saga of Malaysia’s National Carmaker Proton,” Trends in Southeast Asia 2024/9, ISEAS–Yusof Ishak Institute, April 2024.↩︎
Both were solvency crises a loan could fix, not forcing ones; a cash shortfall does not by itself prevent withdrawing the privilege, as governments beyond this roster have shown under one. What the two cases show is that neither government made the move, not that the shortfall forbade it.↩︎
Taiwan’s continuity ran not through one agency but through a succession of economic chiefs who overlapped in office from the 1950s into the 1980s, carrying the same discipline across every reorganisation. The chip-maker Taiwan spun out of a state research institute in the late 1980s was exposed to a genuinely competitive global market for its foundry customers, unlike a firm sheltered behind a domestic review, though the state kept a large ownership stake throughout rather than exiting: a champion tested by the market, not one released from the state. Amsden and Chu’s dedicated study of Taiwan’s upgrading policies (2003) documents sustained state involvement of this kind as the pattern, not a clean handoff to private hands.↩︎
Taiwan lifted martial law in 1987, the same year Korea’s own transition began, and completed its own democratisation over the following decade.↩︎
Mancur Olson, The Rise and Decline of Nations: Economic Growth, Stagflation, and Social Rigidities (Yale University Press, 1982). Olson argued that distributional coalitions, organised interest groups that accumulate in societies left undisturbed by war or upheaval, turn toward redistributing income to their own members rather than growing the economy as a whole; this post’s own test asks, case by case, which of these eight governments actually broke that pattern rather than merely declaring an intention to.↩︎
A windfall, oil rents, a bailout a creditor will fund, absorbs the pressure that would otherwise force a government to choose between rescuing a failing champion and letting the change bind. This is not a description of Indonesia or Malaysia, both of which met at least one such moment in this post, a corruption probe or a financial crisis, and each answered it the way the arrangement predicts. A windfall can postpone a reckoning; it did not excuse either government from facing the ones that actually arrived.↩︎
The mirror boundary sits with economies that got rich hosting a frontier rather than building one, Hong Kong the clearest. Its domestic economy was not free of protected champions: the two power companies held a guaranteed return on their fixed assets and twice blocked government attempts to cut it, and note issue was a privilege conferred on chosen banks. What Hong Kong did not do was build a technology frontier of its own, spending under one per cent of GDP on research and development in every year from 1998 to 2021. The condition traced here is about building a frontier, not hosting one.↩︎
“Wording” undersells what actually moved during these repairs. Three levers did real work: which unit holds the privilege, a firm or the group around it; whether a case is read by a single act or by the whole arrangement; and where a path’s boundary is drawn. Naming the wording alone understates how much room the criterion had to bend toward the data it was being tested against.↩︎
Within this set the discriminator also tracks a regional line: the four transformers share a Confucian-heritage administrative tradition that the inertial four do not. A reading that rests on inherited culture rather than a political choice would predict the same split, and this matched eight cannot rule it out from the inside. That test needs a wider set than eight, and is owed the same treatment as the open questions above. A nearby account, Cold War patronage rather than inherited culture, does not survive this same set: Thailand was itself a frontline Cold War ally, a SEATO member hosting the air bases that flew most of the American bombing campaign in Indochina and receiving over $2 billion in American military assistance between 1950 and 1987, and remained inertial; Singapore and China transformed without any comparable patronage at all.↩︎

What I’ve been reading

Stephen Brien — Sat, 25 Jul 2026 11:29:43 GMT

This week’s two posts asked the same question from opposite ends: what makes a change to a country’s rules actually last? Why Political Reform Is Difficult to Sustain worked it out in general; Whatever Trump Builds, the Next President Inherits ran the same test over a live example. The focus was on how a reform is made to stick, and why so many changes that look like clean wins quietly come apart.

WHAT HOLDS A RULE IN PLACE

One idea runs under both posts, borrowed from economics: a country’s rules hold not because they are written down but because everyone expects everyone else to keep to them. Change a rule, and nothing moves until the new expectation becomes common knowledge. Not just believed, but believed to be believed.

How Common Knowledge Builds and Weakens Societies Steven Pinker, interviewed by Yascha Mounk

The jump from “everyone knows” to “everyone knows that everyone knows” is what lets a crowd act as one. It is why the child who says the emperor has no clothes changes everything, while a thousand private doubts change nothing. Pinker draws the darker version too: an autocrat who controls who may gather and who may speak is not really censoring opinions; he is stopping them from becoming common knowledge.

The power of leadership in changing social norms in heterogeneous societies Fabio Galeotti, Jona Krutaj and Marie Claire Villeval

When private preferences shift only slowly, a minority cannot dislodge an entrenched norm on its own: everyone sits still, waiting for everyone else. Add a leader who publicly commits to the better rule, and expectations snap onto it. It is the crystallising move. Not new information, not a change of heart, just a single visible act that lets each person see the others have seen it too.

Institutions, Technology and Prosperity Daron Acemoglu

In his Nobel lecture, Acemoglu ties why some countries prosper to who actually holds power, and to the gap between power on paper and power in fact. That gap is the wall: the limit nobody can measure until someone leans on it.

PAYING FOR A REFORM, AND MAKING IT HOLD

First, a reform has to be paid for, and the payment that proves a government means it is the cost that lands on its own side, not only on its enemies. Then it has to be made to hold, because entrenchment is cheaper than transformation and a win that is not sealed comes loose. These pieces show both problems, mostly in a development setting, where the law on paper is always the easier part.

Rethinking State Capacity: A Practical Framework for Better Policy Implementation World Bank Governance blog

A practitioner’s framework for the distance between a rule that has been passed and a rule that actually runs. The part a donor can fund is the statute; the part no outside money can buy is what the people inside the system expect one another to do.

State Capacity and American Power Carnegie Endowment

State capacity, usually a thing we measure in poorer countries, turned on the United States. It is examined as a slow decay rather than a sudden break.

The rise and reversal of bureaucratic capacity: lessons from Chinese history Pavithra Suryanarayan

The reversal case, and the most bracing. State capacity is not a ladder you climb and then stay on; it can be taken apart on purpose by elites who find it threatens them. Suryanarayan uses Chinese history to show a capable bureaucracy being hollowed out to protect the very people it might otherwise have taxed or exposed. Capacity is a choice in both directions, which is why a win left unsealed does not simply sit there, safe.

The Death of Direct File Don Moynihan

A clean, recent case of a good reform coming apart for want of entrenchment. The IRS’s free tax-filing tool worked and was popular, and it was killed anyway. It is not really a story about the merits. It is a reform that was never sealed, and so cost nothing to reverse. Even a demonstrably good policy does not hold for free.

Merit vs. Tenure: Reforming Federal Firing Statecraft

The same fight at the level where it bites: job protection versus firing at will. Whether you change the people, wear down the working norms, or rewrite the law itself decides how hard the change is to undo — argued here by the people who have to live with the answer.

CAPTURING THE SCOREKEEPER

The cheapest way to shift a settlement is not to win the fight but to capture whoever keeps the score. Once a change counts only when a referee certifies it, the referee is the soft target. These pieces run from the legal theory of that capture to the live American case: the official who publishes the jobs numbers.

The Law of Constitutional Capture Laurence Claus

The legal spine of the idea. Claus describes how a government built for the common good can be rebuilt from the inside to serve its leadership by shifting the shared sense of where the real rule lies, until a shadow system of loyal officials becomes the thing everyone reads by.

The Strongman Presidency William Howell and Terry Moe, interviewed by Santi Ruiz

Howell and Moe on how a federal workforce that “were mostly mailmen” became the terrain of presidential control. This is the referee question in practice, and why wearing down the norms buys the most lasting effect for the least resistance, because no ordinary-politics defence and no clear statutory text ever catches it.

When trust in official statistics declines Nicholas Bloom, Erica Groshen, Duncan Hobbs and Michael R. Strain

After the August 2025 firing of the head of the Bureau of Labor Statistics, the authors estimate economic-policy uncertainty jumped more than 50% in the following week, with a hit to output that may have run to around $20 billion. The numbers measured by an outsider can be trusted, while the ones a graded official reports cannot be.

Whatever Trump Builds, the Next President Inherits

Stephen Brien — Thu, 23 Jul 2026 06:31:31 GMT

The short version:

Judge a presidency by two things, not the noise: what a leader will pay to get his way, and what he can make outlast him.
This presidency splits into two opposite projects: tariffs, deportations, and the efficiency drive are serious but point outward, and little is built to last.
The real reform points the other way, inward: on who counts as one of us, and on the referees who keep the government’s score.
What he has genuinely sealed strengthens the presidency itself, for whoever holds it next of either party, not his own grip on it.
His boldest reaches, over who is a citizen and who may vote, keep stalling, and not against the opposition you would expect.
One move above all would turn a stronger presidency into a captured one: making defeat impossible, not merely unlikely. Whether he has made it is what the argument turns on.

Across this series, we have taken apart what it actually takes to change a country’s rules: pushing on a wall to find whether it holds, exploiting the window when it opens, imposing the cost on the blockers and insiders who anchor the old arrangement, and sealing the change so it survives its author. President Trump’s second term is that whole sequence run in the open, in real time. This post asks what any of it will amount to.

It was over before lunch. On the morning of 20 January 2021, a new president sat down at the Resolute desk and, with a stack of executive orders, undid four years of Trump’s work in a single afternoon. The travel ban, revoked. The border emergency, ended. The pipeline, cancelled. What had taken a whole presidency to build came apart at the speed of a signature.

And yet not all of it. This same handover that erased almost everything left two things standing, untouched and beyond the new president’s reach: a tax cut written into law in 2017 ^[1] and more than 200 judges appointed for life. The pen that undid the rest could not touch them.

That is the puzzle Trump’s second term puts back in front of us. Much of what a leader does washes away the moment he loses power. A little of it outlasts him for a generation. In practice, the difference is rarely how loud the fight was, or how hard it was fought. It is something quieter and easy to miss, and once you learn to see it, it changes what to make of everything else he and other leaders do.

I. Watch What He Pays, and What Outlasts Him.

Motive and volume tell you nothing about what lasts. Two questions do the work: what he will pay to get his way, and what he can seal.

What everyone already sees

Two things about this presidency are not in dispute, and neither needs arguing.

He treats limits as things to test, not obey. He goes at the courts, at Congress, at the unwritten rules of the office, testing which are load-bearing and which are only reputations no one has checked. And he manufactures his own openings. He does not wait for a crisis to justify a move; he declares the emergency and makes the occasion that licenses the move.

Everyone sees this, and everyone reaches for the same two ways to read it. The first is motive. Is he a fraud, or a dictator? You cannot know in real time, and neither can the people insisting they do; his true intentions are not available for inspection, and the daily argument that claims to read them is guessing.

The second is volume. How extreme, how loud, how many norms broken in a week? Volume is the easiest thing in politics to manufacture, yet it tells you almost nothing about what will remain.

All of it is vivid, and all of it is real. But none of it tells you what he is seeking to build, or whether any of it will outlast him.

Cost, Then Seal

A different pair tells you what he will pay and what will last, and neither is motive or volume.

The first is cost. How far will the president go to get his way, and what will it cost him to remove those standing in his way? The firings, the lawsuits, the political capital spent, the fight joined. A leader who will not bear a real price to clear his blockers does not mean it. Cost is the sign that he is serious.^[2]

The second is the seal. Once a change is made, what would it cost the next administration to undo it? A statute, a lifetime appointment, a court precedent is expensive to reverse. An executive order costs an afternoon. This seal is the sign that a change will last.

A leader can pay dearly for a change and still leave it in a form his successor can undo at a stroke.

That is exactly what we saw on that January morning. Four years of executive action gone by the afternoon, while the 2017 tax cut and the lifetime judges stood untouched. Here, the difference between what vanished and what survived was not how hard it had been fought for. It was whether it had been sealed.

The Seal Was Never in the Ink

So far, the seal looks like a matter of form. On paper, a statute holds, an executive order does not. History says it is not that simple, and the complication matters.

This seal was never really in the ink. Nothing in the American system is harder to undo than a constitutional amendment. Prohibition was one, and it lasted thirteen years. What makes a change hold is that enough people come to expect it will hold, and act as if it will, until that expectation becomes the stable state nothing pushes against; the statute or the ruling is only where it gets written down.^[3]

Real transformation is not the speech or the signing ceremony. The machinery underneath matters: who answers to what, who bears a real risk and who is shielded from one. It is whether that machinery now runs on different rules. Leave it untouched, and nothing has changed, however loud the announcement.

That is why the seal is the sign to watch. What it would cost the next government to undo a change, not what it cost him to make it, is the outward mark that the expectation has set hard. What that cost measures is the real resistance a reversal would meet, not the form the change took. Even the grandest form can be cheap to undo once no one defends it. The point returns.

The cost and seal assess a reform’s effectiveness, rather than its merits. They weigh the sealed over the eroding. That is what lets a policy’s enemies and its friends read the same result off it.

A worse change, properly sealed, outlasts a better one left in pencil.

The same test can be applied to the previous president. Biden also had both the instincts this presidency is identified with.

He tested limits: his student-loan forgiveness reached for a spending power Congress had never granted, to see whether the wall was load-bearing. It was. Struck down by the courts, blocked the next year again, gone without an election ever being needed.^[4] Pencil.

And he manufactured his own openings. Short of the sixty votes the Senate’s rules demand, he drove his Inflation Reduction Act through on budget reconciliation, which needs only fifty, and that one stuck.

Written into law, it could not be swept away with a pen; the administration that followed had to pass its own act of Congress even to cut it back, and even then the law survived. Ink. Same president, same instincts, one effort gone and one still standing, and the difference was not which was bigger or bolder or better loved, or which side wanted it to survive. It was the seal.

A Democrat’s reach holds up no differently than a Republican’s. Neither party gets a discount. A test that only ever grades the side you’re worried about was never neutral to begin with. Both cases, though, are already settled; the same test, turned on a presidency still underway, can only place a bet.

II. The Tariffs Were Serious. They Just Weren’t Aimed at Home.

Almost none of the economic agenda, the tariffs and deportations and cuts, is built to last: it is serious, but aimed outward, at outsiders, not at home.

The problem is always outside

Trump’s economic agenda, the one everyone argues about, is by this test the least likely part of this presidency to last.

The basis of this agenda does not indicate that there is something wrong with the arrangement at home. It says the problems come from outside, from foreign competitors, from trading partners who got the better of us, from migrants. So the cure is always aimed outward, at them.

Immigration feels like the exception, but as an economic matter Trump’s agenda treats it as a foreign problem, cheap outside labour undercutting American wages, and the enforcement lands on the workers. Raids, deportations, the pressure falls on the migrants themselves. Employers who depend on that labour and profit from it are not the focus of it.^[5]

That matters, because those employers are the domestic insiders who anchor the arrangement, and most of them are his own coalition, in agriculture, construction, hospitality. An administration genuinely bent on transforming the labour settlement would have to make them pay. This one mostly spares them and hits the outsider, proof the pressure is not landing where the arrangement would actually have to change. That is the shape of a fight that faces outwards.

Entitlement cuts split along the same seam, and it is worth separating them. Trimming Medicaid or food stamps for citizens is ordinary socio-economic policy, an argument about how large the welfare state should be.^[6] But cutting benefits specifically for immigrants and non-citizens is not an economic move at all. It is a statement about who is entitled to the common pot, which is to say about who counts as one of us. The first belongs here. The second belongs to a different reform.

Subscribe now

Real Cost, Not Durable

None of this is to say the grievances driving this agenda are invented or the effort fake. The lost factories, the stagnant wages, the disorder at the border were real, and Trump has spent real capital on all three.

These tariffs are the clearest case. He imposed them, they fell on American consumers, and when China retaliated against the farm states he paid the farmers billions to hold the policy rather than fold.

He had made exactly this choice once before, in his first term, and paid a steeper price for it; doing it again, six years and a lost election later, is about as clear a sign as politics offers that the price is one he means to pay.

By the first question, cost, that is a serious move. But it is aimed at a foreign trade arrangement, not at a deal among Americans, and the one inward cost he moved to neutralise, the retaliation against his own farm states, he bought off almost as fast as it landed. Still, the cost to consumers he simply let stand.

DOGE tells the same story from the far side of the ledger. Of the savings it announced, only a few cents on each promised dollar has been independently confirmed,^[7] a great deal of noise and almost none of it written into anything a successor could not undo by choosing differently.

Run the two together, and it is not only the direction that repeats. It is the beneficiary. Together, the employers spared by the immigration raids and the farmers paid off after the tariffs are the same group: his own coalition, sorted by organisation, not by nationality, which is also why the consumers footing the tariff bill, just as much at home, got no such rescue.

The cost points outward. The protection points inward.

So one half of the presidency is serious, outward-facing, and unsealed. Whatever else it is, it is unlikely to outlast him in that form, though a tariff regime markets come to expect as permanent would be a different kind of seal than this test was built to read, and one it has not checked for.

But it leaves the question that immigration raised and did not answer. Trump’s labour and fiscal fights all face outward; if there is a real reform in this presidency, it faces the other way, inward, at home.

III. The One Reform His Own Senate Hasn’t Given Him

This is the deepest of his inward reaches: who counts as a citizen and who may vote, and it has stalled, held up not by his enemies but by his own side.

The deepest reform, and why it’s stuck

Benefit entitlements were only the opening move on who counts as one of us. Trump’s deeper reach goes further, into citizenship and the vote itself.

This is the most profound internal question a country can have, and it is the one that earlier immigration fights avoided. Here, Trump faces inwards and reaches for durable form.

Two moves show it. First, he has tried to redefine who is a citizen, issuing an executive order that narrows birthright citizenship.^[8] And, going further, the SAVE Act, an attempt to change who may vote by statute, reaching for a binding law an order could never be.

To deliver a new settlement here, whom on his own side would he have to make pay? On the labour half, the answer was clear: the employers. And he spared them.

On the franchise, the insider he would have to make pay sits closer to home, in his own party’s Senate ranks; the SAVE Act passed the House and dies in the Senate, and not only on Democratic votes.

Twice his allies tried the same route around sixty that carried Biden’s Inflation Reduction Act, attaching it to a reconciliation bill needing only a simple majority. Both times it fell short, on the same four Republican defections: Collins, Murkowski, McConnell, Tillis. He has pushed for it twice, and both times the same four votes held the line. Even his own party has a line it will not cross for him.

Even at the height of his power, his own party’s Senate has not given him the votes, and that, not Democratic opposition alone, is why the deepest reform he says he wants remains unsealed.

He reaches, and he is stopped short of the decisive price, partly by his own side.

Contested both ways, one outcome

The SAVE Act is the hardest test of the method so far, because the two sides do not just disagree about Trump. They disagree about the facts underneath him, and neither side is necessarily arguing in bad faith.

Non-citizen voting is already a federal crime, so one side reads the SAVE Act as basic integrity, closing a real gap between a sworn attestation and documentary proof, and the other reads it as a paperwork hurdle aimed at a vanishingly rare problem, whose real effect would be to keep eligible citizens from registering.

The evidence genuinely splits, and neither side’s numbers settle it cleanly. State audits turn up non-citizen votes in the low dozens across tens of millions of ballots.

Though an audit will find only the cases it is built to look for. Some conservative surveys estimate far higher registration, a different count of a different thing than an audited vote, and those figures rest on self-reporting that is hard to verify independently.

None of that needs resolving here, because it does not change what happened next. Trump’s birthright order was struck down by the Supreme Court on the merits: not on a technicality, but because the claim itself failed. The SAVE Act sits stalled in the Senate. Believe either side of the argument, and the act still never became law; the order still never survived review. He reached for ink on the deepest question of all, twice, and was blocked both times, whoever you think was right to try.

He paid a real price for it, against his own side and not only his opponents. Unlike the tariffs, where he spared his own coalition, this is the one reach where the cost landed on an ally by name, not a rival. It still bought him nothing that lasts here. Cost and seal are different questions, and this is where they come apart hardest so far.

Subscribe now

IV. Where the Reform Is Real, He Has Barely Made Any of It Permanent.

This is aimed at the referees who keep the government’s score. Three ways to move them, and only one, writing the rules down, truly binds a successor.

The Own-Side Test

Twice now, the test has been whether a cost lands on his own side or only on outsiders. Trump’s economic agenda failed that test. Real effort, but his own coalition spared; the cost has been left with consumers. This reach into who belongs passed it, barely. His own Senate, not just his opponents, is what finally stopped him.

Between a clear failure and a narrow pass, the pattern is worth naming properly, because there is a third place to test it, what he has done to the referees who keep the government’s score.

Genuine reform of a settlement can take two shapes. One involves fights, bearing a cost against the insiders who anchor it, in the open, where they cannot mistake what is happening. The other goes around them, building the new arrangement beside the old one until the insiders’ veto has nothing left to veto.

Which shape it takes says less about how serious a government is than about how entrenched the insiders are and how much of the old order has to give way. A government too weak to win the fight is not thereby excused, only pointed toward the other route. What both share is the end state: either way, what the insiders held stops being theirs alone to keep.

There is a catch, and it is worth naming up front. In a divided federal country, the people who hold the power to enforce the rules are often your political opponents. So a cost imposed on them, by either route, looks significant but proves little. It is what any president will spend to defeat a rival, not evidence that a move survives.

The real test is whether the cost reaches his own side: an ally, a donor, a favoured industry losing something as well. A successor can simply reverse a cost borne by opponents. Cost borne by his own side is different: a successor inherits it, and cannot undo it even if they wanted to. That is exactly the distinction that separated failure from a narrow pass.

A move that only ever costs the other team is not a reform of the arrangement. It is a change of whose arrangement it is.

Even cost paid on both sides can still wash away at the next election, which is what the seal is for. Is he reaching to write the new terms down so they would bind his own successors, or only to install friendlier people?

Trump’s sharpest targets are a particular kind of office, and it helps to see the kind before the people occupying them. A referee is any office whose job is to keep the score impartially for everyone^[9]: the inspectors general, the prosecutors, the officials who publish the government’s own numbers, the officers of the court. The presidency has always held referees it does not fully control, and that check is part of the settlement.

Three ways to move a referee

There are three different things a president can do to a referee, arranged here from least to most binding, and they are not equally hard to undo.

At the first level, he can change the people by firing office-holders and appointing loyalists. That is ordinary politics turned up loud; Andrew Jackson built a whole spoils system on it in 1829^[10], and every president since has appointed his own. It reverses at a stroke, because the next president appoints his own in turn.

At the second level, he can change the norms, which is what the firing achieves beyond the personnel.

Remove enough independent officials, and you teach everyone who remains, and everyone who comes after, that independence is what gets you removed.

The office’s working sense of what it is for, and whom it answers to, adjusts to what now keeps an official in the job, without a word of the law changing. That reverses too, but not at a stroke, because a norm is an expectation, and expectations are rebuilt slowly.

At the third level, he can change the written terms of reference, amending the statute itself so the office answers differently by law.

The Inspector General Act of 1978 wrote the watchdogs into being, in the wave of reforms that followed Watergate; the Pendleton Act of 1883, passed after a disappointed office-seeker shot President Garfield, wrote the spoils system Jackson had built back out of existence, putting federal jobs beyond the president’s personal gift. To change what these offices are, in a way that binds the next president, you would have to change those laws.

Change the people, and everyone recognises ordinary politics. Change the statute, and courts and Congress settle it on the merits. Change the norms, and neither applies: no ordinary-politics defence, no statute’s clear text to argue over, only a changed expectation nobody can point to and say, there, that crossed a line. That is where the fiercest disagreement actually lives. That is exactly why the norms level is where a president buys the most lasting effect for the least resistance: no ordinary-politics defence catches it, and no statute’s text does either.

So the same drift reads as capture to one side, referees being quietly disabled, and as correction to the other, referees that had already tilted being brought back to neutral. Which view is right turns less on what happened, which both sides can usually agree on, than on what one already believed about the institutions being moved.

Mostly Talk, Almost No Statute

Trump is overwhelmingly at the first two levels, changing the people and wearing down the norms, and almost nowhere at the third. He has fired inspectors general, roughly seventeen in a single January night. He fired the commissioner who publishes the jobs numbers after a weak report. He pushed prosecutors out and sanctioned law firms for representing clients he opposed, losing to them four times out of four in court.^[11] He cut the Justice Department’s voting-rights office from around thirty lawyers to a handful.

The shape of that record is telling: his losses cluster where the referee sits outside his control, in the courts. The one clear defeat came from the one referee structurally outside the executive branch. And not one of these costs lands on an ally, a donor, or a favoured industry, only on career officials and outside opponents.

Real capital, real litigation. But almost none of it written down. Fire the inspectors general, and the next president appoints new ones; amend the Inspector General Act, and he cannot.^[12] He has fired. He has not amended.

With one exception, and it is the important one. He won a Supreme Court ruling, Trump v. Slaughter, that a president may remove the heads of the independent agencies at will; the Federal Reserve pointedly carved out in a companion case, Trump v. Cook, the rest now fireable.

That is a genuine change at the third level, written by the Court and binding on successors. This does not rewrite what each office is for, but it strips the tenure that gave those offices the ability to act fully independently.

A ruling, though, is the softest form of that seal: the same Court reached it by overturning a ninety-year-old precedent, and a later one could revisit it as readily. It binds while the expectation behind it holds, which for now it does.

Even the deepest seal is not much sturdier than this once the expectation behind it goes. Prohibition at least took a second constitutional amendment to undo. Reconstruction didn’t even need that: after the Civil War, the Fourteenth and Fifteenth Amendments wrote Black citizenship and the Black vote into the Constitution itself, about as sealed a change as the American system can produce.^[13]

Within a generation it was gone. Enforcement was left to federal troops and officials whose independence was never secured, and once that protection was gone, Jim Crow rolled the settlement back for the better part of a century. Nobody repealed anything; the words just stopped being enforced, because nothing remained to hold them to their shape. A ruling this recent has barely finished setting the expectation it depends on.

FDR met the same kind of limit in 1937: his attempt to rewrite the Supreme Court’s terms by packing it was refused by his own party too.^[14] It is the same wall his own Senate had just shown Trump, on a different branch entirely. Trump did not try to pack the Court.

He was confirmed in the power by it, which is less reassuring than it sounds to one legal tradition, though a rival one reads it as simply correct.^[15] It is also the one durable change in this Part that cost him nothing to win, exactly what a rightly decided question should cost. The power he won passes, in full, to whoever holds the office next.

V. What he leaves behind

Measure what has actually hardened, and it enlarges the office for any successor, not the man. Then weigh the one move, capturing the count, that would change that, and the charge that he is reaching for it.

What actually stuck, and for whom

By his second term, Trump had learned to write in ink. His first had been governed by the pen and washed away entirely; what matters now is who this new ink belongs to.

The permanent tax law. The lifetime judges. The ruling that allows a president to fire agency heads. The immunity the Court granted the office in Trump v. United States. Every one of them outlasts him. Some, the removal power and the immunity, are office-powers any successor of either party can pick up; others, the lifetime judges, are a conservative bench seated for a generation that an opponent inherits as a constraint, not a tool. None of them does anything for this man’s grip on the office.

Whatever he has built into permanence, a successor from the other party sits down at the same desk and finds the same expanded powers waiting.

His most profound reaches, meanwhile, have sealed nothing at all. Birthright citizenship, struck. The SAVE Act, stalled. The reform he pursued hardest, on the question of who belongs, is the one with the least to show for it.

Do not mistake any of this for reassurance. A power that every future president inherits is not a small thing. It is a larger one, because it compounds. Each expansion becomes the floor the next occupant inherits, not a choice available to make again from scratch.

Power of this kind can be clawed back, but the asymmetry is steep. An expansion is cheap and can be seized alone; a reversal is dear and slow, needing a change of coalition or of court, and often a crisis to force either, the way the post-Watergate reforms needed a Watergate.

The ruling that allows this president to fire the watchdogs will allow his successors to fire theirs. We have watched at least one clear case of this in living memory. Surveillance and war powers built after 2001 under George W. Bush, over the Democratic Party’s loud objection, passed intact to Barack Obama, who had campaigned against them, and he kept them and used them.

An office gives back only what it is forced to.

This power belongs to the office, not to Trump personally, and that is not the same as nothing to worry about. It means the office, enlarged permanently, and the licence widened with it,^[16] for everyone who comes after: the forward worry, not the absence of one.^[17]

The One Thing He Hasn’t Touched

One thing remains to account for: the count, the process that certifies who actually won an election. What matters here is only what he has actually done to it.

Defences around it are thinning, starting with the same voting-rights office already cut to a handful of lawyers. The Federal Election Commission has had no quorum since the spring of 2025, longer already than the three lapses that preceded it since 2008, and cannot act. Orders demanding proof of citizenship were issued and struck down, and a separate effort to access state voter rolls has gone zero for nine in the courts.

Congressional maps were redrawn hard on both sides; a Republican map in Texas was allowed to stand for the coming midterms, and a Democratic map in California was upheld against challenge.^[18] All real, all on the record.

But the count itself, the law that governs how the result is certified and finalised, he has not touched. A law can stand untouched while the people who work it change, which is why what follows watches the officials as closely as the statute.

That count has been the American crisis point before, in 1800 and again in 1876, and each time the answer was to write the rules down harder, once the danger had already announced itself as a crisis.^[19]

After the 2020 election, the switch, the point where a certified result could be flipped, was reached for directly. He pursued slates of alternate electors, pressured his Vice President to reject the count, and a crowd of his supporters broke into the Capitol, halting the certification for hours and forcing Congress to evacuate.

This time, the system answered the same way it had in the 1880s, with a law, the bipartisan Electoral Count Reform Act of 2022, which hardened exactly the step that had been tested, making the Vice President’s role plainly a matter of form and raising the thresholds for objection.^[20]

Each time the switch was reached for, the response was to seal it tighter. That 2022 seal is the thing now untouched. The switch was reached once in the last term; that reach produced the harder law now in place, and there has been no fresh attempt against it since.

The concern, taken seriously

Where this can end is already visible, because others have got there before, and not all in the same place. A leader can tilt the field, the way Viktor Orbán rewired Hungary’s rules, maps, courts and media over sixteen years until his elections were no longer fully fair, without ever touching the count.

And leaving the count alone was what undid him, because the ballots were still tallied honestly, defeat stayed possible, and in 2026, on record turnout, voters handed Péter Magyar’s Tisza party a two-thirds majority and ended the Orbán era.

Or a leader can take the count itself, the way Venezuela’s chavistas packed the electoral council so the result was theirs to declare, and no vote has removed them since. Of the two, only taking the count makes defeat impossible rather than merely unlikely, so long as elections are still held at all.

But the American count is harder to take than either example makes it look, because there is no single council to pack or office to seize, only fifty separately-run state tallies, each certified by its own officials, courts and rival-party watchers, none of whom a president appoints or can fire.

That decentralisation is a real defence against capture, but not the deepest one. It locates that danger precisely: not in the apparatus at large but in the few central officials, in the few central states, where the pressure in 2020 actually fell.

That is the concern that matters, and it deserves its strongest form, not its easiest. The easy form is that Trump has already taken the count. He has not, as we have seen, and that version collapses on contact. The serious form is that the ground for it is being laid, and it rests on three things: what he did, what he is doing now, and what to make of both.

The first is a matter of record. In 2020, he tried to reverse an election he had lost, through alternate electors and pressure on his Vice President. And he maintains to this day that it was stolen, a claim rejected in every one of the sixty-plus cases that tested it.

The second is happening now, though it is easy to overstate. That a president replaces officials with his own is ordinary; everyone does it, and it reverses at a stroke. What is not ordinary is narrower and harder to undo.

The office has gained powers and immunities it lacked, in rulings that allow a president to fire more officials beneath him and shield his own official acts from prosecution. And the people and norms that stopped him in 2021 no longer look as immovable as they did. The rulings are on the record. What they mean is not.

The third is what to make of the first two, and it is where the two sides part. One side reads them as preparation, a demonstrated intent from 2020 now riding a rising capacity. The other reads the same record as an assertive president gathering power the way presidents do, with 2020 behind him and no sequel in view. The concern is not that a seizure has occurred, but that the grounds for one are being laid.

And you cannot answer that concern with the reassuring facts, because it was never about them. That the certification statute is untouched settles little. Capture of this kind never needed the statute, only the people who work it and the expectation of which way they will jump.

The line he has not crossed

So set the trajectory aside and name the one thing no assumption can argue away; the final say over who won is still beyond any single side’s control. That floor holds. For now, the worst and most provable move, the durable capture of the count itself, has not been made.

Hold the two apart, because collapsing them is how both sides go wrong. Changing who may vote, which the SAVE Act would do, is a contested reform of the franchise, the players. Capturing who counts the votes is a different order of thing, the scorekeeper, and it is untouched.

The one ruling that blurs the line, the removal power, cuts both ways. It enlarges the office for anyone. And it wears at the defences around the count, not by reaching the state officials who do the certifying, who remain outside it, but by hollowing the federal backstops that stand behind them. That is why it is the hardest case of all.

Because that floor is the part that is truly falsifiable, it is worth naming what would breach it. Watch, in the year ahead, for any of these, each a move on the declaring of the winner rather than the counting of the ballots:

The 2022 certification law amended to reopen the routes it closed, or repealed;
The officials who certify a swing state’s result replaced with people pledged to a particular outcome;
A state certifying its result, or casting its electoral votes, for the candidate its own voters rejected;
A refusal to certify a lost result that the courts do not swiftly reverse;
The machinery of the count pulled out of the states and into hands that a president can direct.

Any one of these would turn a reform of the office into a reach for the man. None has happened. That is not a promise that none will; it is a line drawn beforehand, so that if it is crossed, you will know which one it was.^[21]

VI. Reformed the Office. Not Himself.

The two agendas resolve into a serious reformer of the office, not of his own place in it; why the same acts look like capture to one side and correction to the other; and the single line left to watch.

The Seizure Is the Loud Fear. The Ratchet Is the Quiet One.

Two projects, not one. An economic agenda that faces outward, written almost entirely in pencil. And a political agenda, aimed at who counts as one of us and at the referees, that has reached for statute and already enlarges the office for whoever holds it next.

Both views are looking at the wrong half. His critics see recklessness in the tariffs; his champions see resolve in the deportations. Both are watching the agenda written in pencil. The more serious reform, the one that actually holds, is the one almost no one is arguing about. He is a serious reformer of the office, not of his own place in it.

And the consequence that matters is not the one the argument is loudest about. It is not that this man seizes the presidency; it is that the presidency he hands on is a bigger and less bounded office than the one he was handed, for whoever comes next, a ratchet that turns more easily than it turns back.

The seizure is the loud fear. The ratchet is the quiet one, and it is the one worth betting on.

Why the two sides can’t agree

In practice, the two sides of the political aisle rarely dispute what he has done; the firings, the orders, and the rulings are on the record, and anyone can count them. They dispute what those acts mean, and the usual way of putting it, neutral referees against an establishment that had already chosen a side, is too crude. Officials who keep the machinery of government are, whatever their politics, small-c conservatives by disposition. Their instinct is to hold the existing order in place and slow whoever would remake it.

And that order is not neutral. It was laid down over a century by a bipartisan establishment, as much old New-England Republican as New Deal Democrat, and it leans by its nature against an insurgent working to overturn it. What he is moving on really is the institutional order, the law included, not simply a nest of partisans. So the officials really are ranged against him. Part of what they are defending is the law itself, which is why the resistance runs deeper than interest.

The nearest precedent is the Saturday Night Massacre. Nixon ordered his own Attorney General, Elliot Richardson, to fire the prosecutor investigating him; Richardson refused and resigned rather than do it, and his deputy, William Ruckelshaus, refused and resigned right after him. Neither was an opposition partisan. Both were the president’s own men, unwilling to let his order stand in for the law.

Three different kinds of insider drew the same line in the same argument: an elected senator, a career official, a president’s own appointee. Resistance to a serious reform does not live in one branch or one job title; it lives wherever someone is personally answerable for what the rule says.

That is not a fact about this president alone. Any serious insurgent, of any stripe, would draw the same resistance from the same order, for the same reason, and the resistance alone would tell you nothing about whether the insurgent was right, or whether the resisters were as disinterested as they looked.

Still, the officials really are ranged against Trump, and they really are defending the law, not just their own interest; both are true at once, and neither settles the fight: whether you read what he is doing as capture or as correction turns on which side of that older quarrel you already stood.

The objection can run deeper than that, though, and it deserves its own hearing: that an inspector general or a voting-rights office is itself a species of unaccountable power, a fair complaint, not a fringe one, and a president may have real grounds to move on one. That complaint, pursued by changing the statute rather than the people who staff it, would not be the target of anything said here.

But the establishment has a blind spot of its own. The same neutral, professional administration can be every bit as unaccountable as the watchdog offices it defends. At other points in its own history, it has been the instrument of a substantive political and moral program of its own choosing, not one imposed on it from outside.^[23]

Where the disagreement stops

Disagreement about institutional neutrality runs deep, but it does not run everywhere. The count is a referee of a different kind, one that admits no such quarrel: it either reports who won or does not, and its neutrality is not a matter of which side you stood on. That is where the danger finally lies, not in the fate of the other referees.

Your starting assumptions about those institutions do much of the work in reaching it, though not all, the acts are real, and there are interpretations they will not bear. That is not a reason to relax. It is a reason to look for the one act whose meaning no assumption can change.

That concerning act is the durable capture of the count. And he has not made it.

The line drawn here is narrow and exact, not a verdict on the whole man or the whole term, only on whether the count itself has been taken. This says where the worst has not happened. It does not say the worry was misplaced, and it does not close the file.

Treating the clearance itself as suspect, as though the whole exercise exists to find fault with this president specifically, is a mistake: the same test, applied earlier to Biden, could have gone either way. It is not built to convict. On this one point, it clears him. Passing it here is not a grudging result. It is the result the rule was built to give to anyone who earns it.

The count is objective, but that does not make it invulnerable. A true result cannot be faked, but it can be produced under conditions designed to yield a predetermined outcome.

Sudden Change Has a Checklist. Slow Doesn’t.

The office, the referees, the count: three different fronts, and on each one the same question. Does it still belong to whoever holds the job next, or only to him?

The defences around the count need not be broken to give way. They can be worn down instead, left to rust one defensible step at a time, until nothing dramatic has happened and everything has changed.

That is the more likely of the two, and the one that cost and seal can most easily miss; the way Orbán and, for a time, the chavistas both worked, tilting referee after referee, long before anyone reached for the count (in Orbán’s case, before an honest count finally ended him). It is the blind spot the two questions named from the outset, the eroding they weigh less readily than the sealed.

Both matter, not the noise: the sudden move that would cross the line, and the slow wearing that would dissolve it. The first comes with a checklist, the tripwires named above, any one of which would tell you it had happened, at a stroke, turning a change in what the presidency can do into a change in who is allowed to hold it.

The second offers no such list, and no single moment you could show a doubter and say, there, that was it. The subtlest form leaves even the defences standing.^[24]

That missing line is not a reason to worry less. This slow path is how democracies are increasingly lost now^[25], not in a single night but in a hundred defensible steps, none of them the crossing; and the vigilance it asks is the harder kind, catching the drift while it can still be named and turned back.

Whoever comes next

Measured against that question one more time: does it belong to whoever comes next, or only to him? The office, the referees, and the count each answer differently. The office is enlarged, and it already belongs to whoever comes next, not to him, though not every part of it in quite the same way. The referees split by belief, not fact, and will keep splitting long after this presidency ends. The count alone has a clean answer, for now: still watched, not yet taken.

The issues of cost and seal are not unique to the agenda of any one president. When the next one comes, of either party, vowing to remake the country or accused of seizing it, they will not have changed. What will they pay, and what can they seal? The real reform, if there is one, is always somewhere quieter than the argument. Here is where it landed this time.

Only one move would turn the office he has enlarged into his to keep rather than theirs to inherit: capturing the count. This is the one thing he has not done. Whatever he builds, the next president inherits, unless that line gives way. The two questions, and the signs that would say the line had given way, are not this piece’s any longer. They belong to whoever is still watching, whoever holds the office next.

Notes

The 2017 Tax Cuts and Jobs Act was deficit-financed (the Congressional Budget Office scored it at $1 to 1.5 trillion over ten years), with the cost diffused across future taxpayers, not assigned to any named program. Its durability comes from being written as permanent law, not from how it was paid for.↩︎
The borne-cost test applies costly-signalling theory. A signal separates the serious from the bluffing only when it is dear enough that a bluffer would not send it (Michael Spence on market signalling; James Fearon, “Signaling Foreign Policy Interests,” 1997). What is new here is carrying that logic into domestic rule-change, where the cost falls on the leader’s own coalition rather than a foreign rival, and running it across two terms split by a lost election, so the comparison separates resolve from circumstance.↩︎
That a settlement holds only because enough people expect everyone else to keep to it, a self-fulfilling convention rather than a command, is an old idea about law, in David Lewis (Convention, 1969), H.L.A. Hart (the rule of recognition), and Oliver Wendell Holmes (law as “systematized prediction”), and reached independently in game theory by Ken Binmore. Constitutional theorist Laurence Claus restates it directly (”a legal system is a signaling system”) and names the failure mode, constitutional capture, in which a leader spins webs of dependency until his own will becomes the new focal point for shared expectation, read in place of the law (The Law of Constitutional Capture, 2025). The seal, here, is that expectation set hard.↩︎
Biden v. Nebraska (2023) struck down the first debt-cancellation plan as exceeding the authority Congress had granted; the narrower SAVE repayment route that followed was blocked by a federal appeals court in 2024.↩︎
Second-term enforcement has concentrated on the arrest, detention, and deportation of migrant workers, workplace raids included; prosecutions or penalties against the employers who hire them have not been a comparable focus. This is a claim about where the weight falls, not that no employer has been touched.↩︎
Trump has repeatedly promised not to cut Social Security or Medicare; the second term’s tax-and-spending law instead reduced Medicaid and food-stamp (SNAP) eligibility and funding.↩︎
Independent audits have confirmed only around $2 billion of the Department of Government Efficiency’s claimed savings, a small fraction of a headline figure that ran through the year into the tens, and by some accounts the hundreds, of billions. “A few cents on each promised dollar” is that gap.↩︎
Birthright citizenship, that anyone born on American soil is a citizen, is that same Fourteenth Amendment guarantee, extended to the children of immigrants by the Supreme Court in United States v. Wong Kim Ark (1898). Trump’s order reaches to narrow the very seal the Reconstruction passage below describes.↩︎
The distinction between rivals and blockers is Milan Svolik’s (The Politics of Authoritarian Rule, 2012); the referee, the office whose job is neutral certification rather than rivalry or policy, is the category added here. On why the same acts read so differently across the aisle, Graham and Svolik (2020) find that voters punish undemocratic conduct only weakly, and less as polarisation rises.↩︎
Trump hung a portrait of Jackson in the Oval Office and has invoked him as a forerunner. Parallels (the Bank War, the defiance of Worcester v. Georgia, the spoils system later undone by the Pendleton Act) are standard ground.↩︎
This piece scores what Trump has done as president; what has been done to him is not its subject. His own prosecutions, the New York conviction and the federal indictments, were brought, and in the New York case concluded, while he held no federal office, in his capacity as a candidate and private citizen between his two terms. That is a different question from the one asked, and is not scored here. Character, as a candidate or as president, is not what this piece measures. His approach to the office is.↩︎
The Inspector General Act, amended in 2022, requires the president to give Congress thirty days’ notice and reasons before removing an inspector general. Several of the January 2025 dismissals were challenged as failing that test; whether they did is a live legal question left open here. Not every referee has even that much cover. U.S. Attorneys, for instance, carry no comparable statutory protection and serve at the president’s will, so for some offices there was no statute to amend in the first place.↩︎
A shorthand for what the two amendments practically secured. Formally, the Fourteenth Amendment’s citizenship clause is universal in its wording, “all persons born or naturalized in the United States”, but its purpose was to make citizens of the formerly enslaved. In turn, the Fifteenth barred the denial of the vote on grounds of race, enfranchising Black men; Black women waited on the Nineteenth Amendment, in 1920. That Fifteenth’s promise was itself hollowed by Jim Crow and made real again only by the Voting Rights Act of 1965 and the federal enforcement behind it, which is why the second seal held where the first had not. That expectation, this time, was defended.↩︎
Roosevelt’s 1937 bill proposed adding up to six justices to the Supreme Court, nominally to ease the workload of those over seventy, in practice to secure a majority for New Deal laws the Court kept striking down. His own party’s Senate Judiciary Committee reported it adversely, and it never reached a floor vote.↩︎
A Court whose majority his own first term appointed is not necessarily a disinterested arbiter, and being handed power by a friendly court that has come to read your will into the law is one of the ordinary ways capture arrives. What keeps this short of that is process, not sympathy. That ruling came through adversarial litigation a loser can still contest, and a later Court can revisit, not a change in how the Court decides. The distinction is real, and it thins the day the Court stops being an arbiter and becomes an instrument. The ruling also has a substantive case behind it, not only a friendly bench: on the unitary-executive view, running from Federalist 70 to Justice Scalia’s lone Morrison v. Olson dissent (1988) and treated by that tradition as since vindicated, it is the removal-shielded independent agency that was the innovation, not the president who can fire its head. On that view the ruling is correction, on the merits, not capture.↩︎
What a president may do and be tolerated in doing is the range sometimes called the Overton window, and it widens by attrition. A line crossed often enough at no cost becomes the new baseline, no longer a line at all. Most norm-breaking is noise that fades with the news cycle; what lasts is the crossing that goes unpunished until reversing it would itself look like the radical move. That hardened tolerance is a seal like any other, measured by what it would cost to forbid again what has become ordinary, and a successor inherits it whether or not a single law has changed.↩︎
There is a deeper worry the office/man distinction does not reach, and it is Laurence Claus’s. A move can leave a leader’s personal grip untouched and still change the constitutional settlement rather than operate within it. If expanding presidential immunity and the removal power shifts where officials look for the operative rule, from the law toward the executive’s will, the settlement has moved toward capture no matter who inherits the larger office. “It passes to either party” answers the fear that this man is entrenching himself; it does not answer the fear that the office now sits closer to a leader’s will and further from the law. This post treats that structural question as the count question, the single line whose crossing would make the shift decisive. Claus would locate it earlier, in the character of the rulings themselves.↩︎
Redrawing districts for partisan advantage (gerrymandering, after an 1812 Massachusetts map) is a two-century habit of both parties. In the current cycle a Republican legislature’s map was upheld for the coming midterms, and a Democratic-controlled state answered with a ballot measure authorising its own.↩︎
In 1800, a tie in the electoral college threw the choice to the House and nearly cracked the young republic; the answer was the Twelfth Amendment. In 1876, a disputed election was settled only by a backroom bargain, after which Congress passed the first Electoral Count Act to keep it from happening again.↩︎
Originally, the 1887 Electoral Count Act had let a single member of the House and a single member of the Senate force a floor vote on any state’s electoral count; the 2022 reform raised that to one-fifth of the members of each chamber, so an objection now needs broad backing even to be heard.↩︎
The forward sort is a standing bet, and it can be lost two ways. If a properly sealed change (a statute, a lifetime appointment) were undone inside a single term without an election turning the government over, the “ink survives” half would fail. If a pencil change (a purge, an executive order) outlived a successor who took office determined to reverse it, the “pencil decays” half would fail.↩︎
COINTELPRO, the FBI’s covert counterintelligence program, ran from 1956 to 1971; Hoover’s own directive described its purpose as working to “expose, disrupt, misdirect, discredit, or otherwise neutralize” the groups it targeted, most of which, the Senate’s Church Committee found in 1976, had no tie to a foreign power at all, non-violent civil rights organisations included. Attorney General Robert Kennedy authorized wiretaps on Martin Luther King Jr. in 1963, and the following year the Bureau anonymously mailed King a letter urging him toward suicide.↩︎
Capture, in the end, is a change in what people expect. The count can hold on paper, every official still in post, and still be lost the day the ones who matter come to expect that the leader’s will is what finally counts, and jump that way unbidden. Nothing is seized; the expectation moves, and a checklist built for acts will never catch it.↩︎
The slow, individually defensible, on-its-face-legal accumulation described here is how the democratic-backsliding literature frames erosion, chiefly Steven Levitsky and Daniel Ziblatt (How Democracies Die, 2018; Tyranny of the Minority, 2023) and Nancy Bermeo’s “executive aggrandizement” (2016). This piece departs from that work in that it draws the decisive line: not at the gradual tilt, which it grants is the likelier danger, but at the durable capture of the count, the one move that ends the contest. What the two questions add is an ex-ante durability test, run on a live administration move by move, that the literature theorises but does not apply.↩︎

Why Political Reform Is Difficult to Sustain

Stephen Brien — Mon, 20 Jul 2026 07:31:25 GMT

The short version:

A government can pass every law on the books and still be bluffing, and break the law in secret and still be changing the country for good.
What holds a country’s rules in place is not law but what its powerful groups expect one another to do. Of four moves against that belief, only one can reset it.
That is why transformation is rare. A defeat re-locks for nothing, dropping the country back onto what everyone already expected; a win must pay to break a standoff the rest are waiting out.
The cheapest way to move a settlement is not to win the fight. It is to capture whoever keeps the score.
Two marks tell you whether a reform is real: did any cost fall on the government’s own side, and was the win made common knowledge? Fail either, and it is just a performance.

This is the last of five posts on how a country’s rules actually change. The first four each took one of a reform’s four moves on its own; this one reads all four together. It stands on its own, so the earlier posts are useful, not required.

In 2020, the Indian government passed three laws to remake its farm economy, with every formal mark of power behind them: the votes, the majority, the whole machinery of the state. A year later, it repealed them, and the repeal held, with nothing required to keep it in place.

In 1978, eighteen farmers in a Chinese village called Xiaogang broke their commune back into private plots in secret, an illegal act that could have had them shot. Within a few years the whole country was farming their way, and it never went back. The legal reform vanished; the illegal one became the rule.

The law is not the enemy either. In 1688, England forced new terms on the crown and wrote them into statute the year after, an act every bit as formal as India’s. That settlement held for three centuries. Writing it into statute was not incidental to its holding. The vote itself was the public act that led the crown’s creditors to believe it and price their money on its holding. Government borrowing rose more than tenfold in the nine years after.

Three reforms, three surprises, and not only three: the same reversal runs through the developing world, where a reform passes with every formal mark of success, and a year later nothing beneath it has moved, while somewhere else the same law on paper takes and holds. You cannot tell the real from the performed from the statute, and the outcome that would settle it arrives too late to act on.

This post asks what decided each one, finds the answer in what a government cannot fake, and turns it into a way to catch a fake reform the morning it lands, long before the results arrive.

I. What a country’s rules really are

What a reform truly acts on, the four moves it can make, and what tells a real reform from a mere performance.

A country’s deepest rules are not its laws. They are the settlements those laws rest on, standing and largely unwritten deals among its powerful groups, held in place less by law or force than by what each group believes the others will do.

The four moments of a reform

First, the wall gets tested. Most limits a government treats as fixed are real, but some are only reputations no one has tested, and no one knows which until someone leans on one.

Then, the window opens, or is forced. The moment to act is rarely handed down by a crisis; more often it is made, and a made opening is a wager, not a method.

Then, the cost gets paid, on the ledger. Talk is free, so what proves intent is a borne cost, and the cost that proves a government means to change the rules is the one that falls on its own coalition, not only on its enemies.

Finally, the win gets sealed, or it does not. A win changes nothing until it is made common knowledge through a scorekeeper the losers cannot lean on, so that the new line is public (the way England’s 1688 vote was) and not just a claim the next government can quietly erase.

Sealing is the move competent governments most often skip, because the thing it must overcome is the one opponent you cannot point to: the spread-out expectation of everyone not in the room.

The seal also has a design that decides whether it holds. One that must be re-performed by hand decays as soon as the performing stops; one lodged with a keeper who profits from the new line holds itself, kept up by nothing but the keeper’s own interest.

The reading that gets all three cases right

Each of the three reforms played out differently from how they looked. India had every mark of power and went nowhere. Xiaogang had none and became the rule. 1688 looked as formal as India, yet held for three centuries. What each looked like was the law. What decided it was the settlement underneath.

What misled the usual readings was the formal marks themselves, the votes and the majority and the machinery, and not some missing detail beside them.

A government really moving one of these deals and one only performing it look the same in the speeches and the results. They come apart on what neither can fake.

Three of the four moves, testing a limit, seizing a moment, paying a cost, can only shake a settlement loose. Only the fourth, sealing a win, can settle it onto something new.

One caution travels with this from the start: judging whether a rule will hold is not judging whether it should.^[1]

The belief layer

Every settlement runs on belief. Each player goes along because of what it expects the others will do, and those expectations, being about things that rarely happen, are rarely tested. This sits on top of real power, not instead of it. Who actually holds the guns and the money is real. But layered on the power is a structure of belief, whether India’s farmers and traders actually expect the new order to hold.

It is the belief that reform usually forgets. India changed the rules on paper and left the belief where it was. China changed the belief and left the paper to follow. This is the layer that a generation of legal-transplant and good-governance programmes funds hardest and moves least: the paper they can write, the belief they cannot legislate.

The rules on paper are the easy part; belief is what determines whether the reform survives contact with the people it governs.

Belief’s independence from power is only partial. The Xiaogang farmers’ move also shifted what the officials above them were willing to enforce. Belief rides on power; it is not free of it.

That split between the belief layer and the power beneath it is an idea long worked out in economics: that a country’s institutions are not rules on paper backed by force, but arrangements no one breaks first, because each player keeps to them only as long as it expects everyone else to.^[2] That idea tells you what a settlement is; the cut that follows tells you which of the four moves can move it.

II. The one move that can change them

Three of the four moves can only shake a settlement loose. What the fourth does that they cannot, and why that makes real transformation so rare.

Why only the seal re-coordinates

Each of the first three moves reaches something real beneath the settlement. Doubting a wall probes the limit it names, making a moment takes the opening it offers, and paying a cost meets the price on the ledger. The wall test, in particular, can go further and move the hard power a limit rests on. But changing what a settlement rests on is not the same as moving it.

A settlement coordinated on belief can absorb a changed limit, or a shift in the balance of power it need not yet reckon with, and still not move. The reason is a standoff.

In a boardroom where every director privately thinks the chief executive has to go, no one says it first. Nothing moves, though they all already believe it, until someone names it aloud and the room turns against him in an instant, because each suddenly knows the others were thinking it too. But turning against the old chief is only half of it. The room still has to settle, out loud and together, on who runs things now, and that second act does not follow from the first on its own.

Any settlement holds the same way. Even when every player privately expects the old order to give, no one acts on that expectation until each can see that the others see it too.

Romania in December 1989 was that boardroom at national scale. Ceaușescu had ruled for decades on the shared belief that everyone would keep obeying, whatever they privately felt, a wall that looked unassailable until it was tested.

At a televised rally meant to shore him up, the crowd began to jeer, live, and each person, seeing the others jeer and the cameras catch it, knew at once that the others knew too. The old expectation broke in minutes, and within a week his rule was gone.

Private knowledge, however widespread, leaves the standoff standing. The settlement does not shift onto the new line until each can see the others see it.

What only the seal does is reach the coordinating layer, the shared expectations players read each other by. That is the belief layer named above. It makes the change common knowledge, not merely known but known to be known, so that the settlement can re-coordinate on the new line.

A settlement can be shaken loose from beneath, its prop collapsed or its coercive floor broken, and still not settle onto anything new; destabilising the old line and re-coordinating onto a new one are two different acts, and only the seal does the second.^[3] This is why the vote in 1688 mattered more than the law it passed.

North and Weingast read 1688 as the moment the crown’s hands were tied. Parliament won the purse, and a king barred from taxing at will was barred from defaulting too. That constraint was real. What is added to that account is why creditors trusted it to hold. The vote answered that. It let each creditor watch the others begin to believe, and a promise no one trusts the rest to keep binds nothing.

Subscribe now

This is the one claim the series of posts could not make until the four moves were laid side by side. But two things follow from it that the equilibrium view it borrows does not already hold.

First, transformation is not merely rare but asymmetric, cheap to reverse and costly to win. Second, because only the seal leaves a datable mark, a reform can be read from outside before any result is in.

Another way a settlement changes

Everything above concerns deliberate moves, the ones a reformer can choose and an observer can catch in real time. A settlement can also change without any of them. It happens when the things it rests on drift far enough that the old expectation quietly stops holding, and belief re-coordinates after the fact rather than on purpose. This slow route is real, and the four moves, being all deliberate, do not name it.^[4]

Japan’s post-war developmental drive faded this way. Its ministries, big firms, and banks pulled together while everyone expected the growth machine to keep delivering. As catch-up growth ran out, that expectation drained away, and by the early 1970s the government’s push to build had given way to managing what it had.

The deal underneath held; what turned was the orientation on top of it. The machine ran on for years under its own momentum, so the turn stayed invisible, plain only in the stagnation of the 1990s. No reformer sealed the shift; no public act marked its end.

So a settlement changes in one of two ways. A reformer seals a win on a single day, or the ground drifts too slowly for anyone to choose the moment or catch it as it passes. Japan drifted, which is why its turn left no date to point to. Narrow the claim to the deliberate route, and only the seal leaves a dated mark. That mark is the whole reason a reform can be read from the outside at all.

Why transformation is rare

A defeat seals itself, and a win does not. When India repealed its farm laws, the reversal locked in instantly and for nothing, because it confirmed the expectation everyone was already coordinated on and dropped the country back into the groove it was in. That is why reforms so often go to die exactly at the seal.

A reform is the opposite. It has to pay the one-time cost of moving the whole coordinated set to a new point. The premium is paid once, at the crossing, and it is paid against every group that arranged its life around the old line and will defend that arrangement.

The reason a reform must pay that weight is the same standoff as before. No one shifts to the new line until enough others already have, so a reform must pay to break the deadlock.

Paul Volcker’s Federal Reserve did exactly this to inflation. By 1979 prices rose fast because everyone expected them to, and everyone raised their own prices and wages first so as not to be left behind; the same standoff read one more way.

Volcker broke it by forcing interest rates high enough to cause a deep recession, paying in lost output and jobs to move the whole economy onto a new expectation, that prices would hold. The cost was borne once, at the crossing, and once the new expectation set, it held for a generation.

That the path a settlement can take depends on where it starts, and on what each player expects the others to expect, is the general shape of these problems, worked out formally by Acemoglu and Jackson.

Their result is the deeper claim, and it holds without anything in this post. What it lacks is a direction. In their model, a settlement slides back as easily as it climbs; adding that direction is what turns the rarity from a matter of chance into a matter of cost.

Hold the seal and the new point costs no more to keep than the old one did; all that then remains is that the sealing not decay, so an unsealed win can still come loose.

Entrenchment is cheaper than transformation.

That price gap, not any shortage of courage, is why real change is rare.

The rarity is conditional. Most reforms never reach the seal at all; they are crushed at the wall, at the first move. So the seal is not the bottleneck for every attempt, only for the few that clear the first three moves, where it is the step most often skipped or botched.

And the seal is not a one-way gate. An unsealed win has not been made to last, which is not the same as saying it will not last. Deng’s reforms stalled after the 1989 crackdown emboldened their opponents, and in 1992 he toured the southern provinces to re-seal them in public, winning back a win that had come loose.

Controlled evidence backs the mechanism. The largest coordinated experiment on voter information found that private facts barely moved voters while public ones moved them much more, and random corruption audits in Brazil bit hardest where a local radio station carried the verdict to the whole town.

The evidence is suggestive, not settled, and gathered at the scale of voters rather than standing deals among the powerful, but it points the right way.

Subscribe now

III. Where the fight actually is

One implication follows straight from the seal. Because a change only counts once a scorekeeper certifies it, the cheapest way to move a settlement is to capture the scorekeeper, not to win the fight.

The scorekeeper is itself a settlement

The seal needs a scorekeeper, someone whose count of who-believes-what is itself believed. The seal’s whole job is to make a change common knowledge, and common knowledge can form only around a single count everyone reads together and trusts. The office that publishes an act of parliament is one such keeper, the record a country reads to know a law is truly in force. The body that certifies the winner of an election is another.

A trusted scorekeeper is not a fixed feature of a country but rather something the players have agreed to believe. It is a smaller settlement resting underneath the big one, held up by belief in the same way and movable by the same four moves.

So capturing the scorekeeper is not a fifth kind of move outside the four. It is the seal again, reached one level down. Move the small settlement that keeps the count, and the big one has to follow, because the count is what everyone goes by. The cut, the line drawn earlier between what a move touches and what only the seal reaches, does not break here; it nests.

The boardroom has its own level above it. The directors who could remove the chief are themselves a small settlement, with their own seats and their own rules for filling them. Capture that board, and you have moved the body that moves the chief, without ever facing him across the table.

The cheapest attack

The natural expectation, with four moves laid out, is that the cheapest attack must be some new and heavier one, a costlier commitment than any of them. It is the opposite.

If a result counts only after an independent scorekeeper certifies it, then the cheapest attack on any settlement is not to win the underlying fight at all. It is to capture the scorekeeper, to move a sub-settlement instead of the whole one.

The underlying fight is a standoff no one wants to move on first. The scorekeeper has no such standoff protecting it, because it is smaller, less watched, and the count it keeps is the one everyone reads to learn what everyone else believes. Capturing it need not win any single mind; change that count, and each mind shifts, seeing the others shift with it.

Such cheapness is conditional, not fixed. Capture is cheap exactly where the scorekeeper is weakly held, and dear where its independence is itself common knowledge.

The thing worth guarding in any settlement, then, is the independence of whoever keeps that score.^[5] This is the settlement’s own strength turned against it.

A deal that holds only because everyone expects everyone else to hold cannot be beaten head-on, but change the one count they all read, and they all have to follow, because no one can afford to be last to move.

Rwanda, in miniature

Rwanda shows the cheap version of this in miniature. Its districts are scored in a single annual ceremony, and two kinds of numbers emerge from it. One is farm figures, inflated, maize by fivefold, because the very officials being graded report them up.

The other is health figures that are real, because they are measured by outside surveys the government cannot quietly revise. Same government, same ceremony, same year. Only the keeper of the score is different, and the truth changes with it.

That is a captured reporting number, not yet a captured national settlement, but the mechanism is the same. Capture the keeper, and you manufacture common knowledge directly, without ever winning the fight underneath. The rule generalises to any figure a state reports. The numbers an outsider measures can be trusted; the ones the graded officials report up cannot.

The stakes beyond the reporting desk

The highest-stakes version of this is not Rwanda’s reporting desk; it is a constitutional order. In April 2022, Tunisia’s Kais Saied seized control of the country’s independent electoral commission by decree, replacing most of its members; three months later, the commission he had just appointed organised and certified the referendum that cemented his one-man rule. Russia’s Constitutional Court did the same for Vladimir Putin in 2020, resetting his term count to zero two days after he asked.

Neither man won a fight against a rival for power. Each captured the body whose word would make the result real, then let that body do the rest. Capture the scorekeeper and the fight need never happen, which is why it is the move a ruler reaches for to keep power, not to govern well.

IV. Reading a reform

Two questions you can put to any reform in real time, whether its costs fell on the government’s own side and whether its win was made common knowledge. The rest is what the reading still cannot tell you, and the one fact it all comes down to.

How to read a reform

A real reform leaves marks a government cannot fake. Two of them you can read the moment they happen; a third you cannot read in advance at all. Take the two readable ones first, and neither is what the government announces.

A government remaking a settlement has to spend against the interests the old deal protected, which means some of its costs fall inward, on its own coalition. That first mark, the ledger, is the one hardest to fake and the cleaner of the two.

A government whose costs fall only outward, on rivals and outsiders and never on its own coalition, has told you it is bluffing before any programme arrives to be judged.

That is a clean disqualifier, and you can catch that mark the morning the costs land. Egypt spent twenty years under Nasser and Sadat imposing high costs on landlords and foreign capital while sparing the officers and the subsidised cities on which its rule actually rested. The one time a crisis forced the cost inward, Sadat’s 1977 subsidy cut, the coalition threw it off within two days.

The second mark, the seal, is what a government does with a win once it holds it. A real remaking is made common knowledge through a scorekeeper the losers cannot reach, an office, a record, a body whose count they cannot quietly revise. What a government cannot fake is not the announcement but the sealing: whether the new line was placed where its rivals can no longer quietly reopen it. A win only declared, a law left in a drawer, is one the next pressure will open.^[6]

The third thing cannot be read in advance at all. That is the wall, whether the government’s push will meet a limit that holds.

Most walls have a solid floor that everyone correctly gauges. A ruling party will not vote itself out of existence, a landed elite will not surrender its estates for the asking. But above that floor is a wide band of untested strength whose height nobody knows until it is tested.^[7]

New Zealand’s protected economy looked beyond reform in 1984. Then the finance minister Roger Douglas swept the protection away in a few years. Because a wall’s strength is unknown until the test, the reformer about to triumph and the one about to be destroyed look identical going in. We remember the ones who guessed right and forget the rest, which is why the wall is offered as illustration, not as a list of cases the reading called.

Analysing the opening case

Take the reform this post opened with and read it by the two marks. India’s government had the votes, made its move, and won the vote, so every loud signal said the change was real.

Interpreted by the marks, it was a fake, and you could have said so the morning the costs landed. The cost fell outward, on the farmers, not inward on the government’s own coalition; and nothing was sealed, no scorekeeper, no common knowledge, only a law in a drawer that the next pressure could open. Whether the farmers’ resistance was also a real wall test the government lost is a separate question that this account leaves open.

Subscribe now

The direction of the cost is why the government could fold. Spending against its own coalition is a bet a government cannot cheaply unwind. Having hurt its own side to make the change, it would have every reason to hold rather than take that loss for nothing. Spending only on outsiders is a bet the Indian government never placed. So when the farmers held out, nothing of the government’s own was staked on the reform. A change no insider had to absorb was one it had no reason to defend.

India’s repeal a year later reads the opposite way. It sealed itself, because it confirmed what the country already believed. The legal reform was the weaker one all along, and the interpretation that gets both cases right asks not what a government announced, but what it could not fake.

That is what reading a reform this way buys: it overrules the signals everyone else trusts. The votes, the announcement, the early results are exactly what a settlement held by belief gives off to look solid while nothing beneath it has moved. The interpretation sets all three aside.

So it is a far better lie detector than a crystal ball. Fail either mark and the reform is a performance; pass both, and it may be real. You get a clean no early; you never get a promised yes, and on the wall, whether a limit will hold, you get nothing at all, because that is the one thing no one can tell in advance.

Where the reading stops

The reading holds only where a settlement runs on coordinated belief, which is most of the time. Two kinds of case fall outside it, and each is a standing way to misread a reform.

The first rests on a hard material prop, like oil or mineral income the powerful control directly. There the prop, not the seal, decides how far anything can move. The belief is still there; it is simply pinned to the resource.

Everyone assumes two things that a seal cannot move: that the resource will last, and that one group keeps privileged access to it. Both rest on a fact anyone can already see: who holds the taps. A seal can make a new political line common knowledge, but it cannot make someone else the owner of the oil. So the coordination holds until material control of the resource shifts, and a reform reaches only that far.

When Saudi Arabia locked scores of its own princes and tycoons in the Ritz-Carlton in 2017, the costs fell squarely inward, the surest mark of a real move, and yet little was remade, because a settlement floating on oil never has to accept a verdict it cannot rig. Just as a drifting settlement changes with no seal, an oil-propped one holds against every seal.

The second is direction: only moves aimed at the domestic deal count. A government also acts abroad, on trade partners and foreign states, and a move aimed abroad can look like bold reform while leaving the home settlement untouched, its costs paid by foreigners or consumers no coalition member answers to. Reading a foreign-facing move as a settlement move is the commonest way to misread a reform: it spends outward by design, so the ledger’s inward-cost mark never even applies.

Can the reading be wrong?

Those two limits are about where the reading applies. A sharper objection cuts at the theory itself: that anything resting on “belief” bends to fit any outcome after the fact. It is a fair charge, and it has an answer. The ledger is scored the day the costs fall and the seal the day the win is won, both before any result, so a wrong call shows up as a wrong call and cannot be talked away.

Only the wall is unreadable in advance, which names a rule worth stating. A settlement of belief is falsifiable downward, never upward.

Among the deliberate moves, a sealed win should hold longer than an unsealed one, since an unsealed win has never made the change common knowledge and can still come loose. That difference in durability is itself a claim the world can contradict, one the framework was not built to fit.

Two honest caveats remain. The seal claim is a mechanism, not a scored record: every case here fits it, and the case that would break it, a public seal that still came apart or a lasting change with no seal to name, is not among them. And the split between what a settlement rests on and the belief that reads it is a working one, not two real tiers; a fuller account would treat them as one surface bending over years.

What you are left holding

Those four moves, the wall, the window, the ledger, and the seal, work the same for a predator as for a reformer. A ruler grabbing power for himself seals his win by the same marks a genuine reform leaves, on every move but one. Those marks tell you a rule will hold, not who it will hold for, whether the new order is open to all comers or closed around the winners.^[8] They cannot tell reform under the rules from revolution against them.

One thing is left holding the weight, and everything above follows from it.

A country’s real rules live in what its powerful believe about one another.

So a reform that crosses a desk is read by two marks, not by the votes, the announcement, or the early results, which are exactly what a results framework counts. Read that way, it shows itself as a performance long before the result that would otherwise have arrived too late to matter.

Which is why so much reform in the developing world fails while looking, on paper, like success: it never cost the government's own coalition anything, and was never sealed beyond the losers' reach, so it stayed a law in a drawer. The statute is the part an outsider can fund; what a country's powerful expect of one another is the part no outsider can move.

That holds the same way in a one-party state and a democracy, in a Chinese village and an English revolution. What you are reading, in the end, is never the law. It is the belief beneath it, and whether anyone has managed to move it.

Subscribe now

What I’ve been reading

Stephen Brien — Sat, 18 Jul 2026 08:44:09 GMT

THE PERFORMANCE TEST

The shared question across these three pieces: what distinguishes the four East Asian economies that kept climbing toward the technology frontier from the four that grew fast for a generation and then settled into a lower rate of growth? Liao opens the historical case. Cherif and Hasanov supply independent empirical corroboration from IMF research. Noah Smith engages with the mechanism directly and adds a twist to it.

Industrialization Before Liberal Democracy, Part I: Asia — leonliao.substack.com — 12 July 2026

The historical pattern is more decisive than the political science literature usually allows. In South Korea, Taiwan, Singapore, Japan, and Hong Kong, rapid industrialisation preceded liberal democracy by decades. Military rule, one-party governance, or colonial administration came first; democratic transitions arrived only after industrial societies, export-competitive manufacturing sectors, and large middle classes had already formed. Liao’s claim is structural rather than normative: early industrialisation requires the kind of concentrated political authority and long-run policy consistency that competitive democratic politics does not easily generate.

The piece acts as provocation. It establishes the sequence without fully resolving whether causation ran from authoritarian capacity to industrial output or from industrial success to democratic demand. What it does establish cleanly is that the sequencing argument cannot be dismissed as an artefact of particular cases. Five economies, different colonial histories, different political arrangements, the same order every time.

Modern industrial policy: The Asian miracles’ blueprint for developing economies Reda Cherif and Fuad Hasanov — VoxDev / IMF — 19 March 2026

Two IMF economists, working from a large cross-country dataset, arrive at the same architecture. Their three principles for the Asian miracles — promoting sophisticated export-oriented industries, fostering innovation and competition, ensuring accountability for public support received — collapse into two distinct categories. The first two are inputs: which sectors a country builds and which policies a government runs. The third is the test: whether a firm that fails to win export markets actually loses its backing.

These are senior IMF economists saying, on a major IMF platform, that what separated East Asian transformation from ordinary growth was disciplined conditionality — the willingness to actually withdraw support from firms that did not perform. Governments that ran industrial policy without this mechanism produced growth; those that ran it with the mechanism produced transformation. That the same conclusion is reached by independent routes — historical case analysis on one side, IMF cross-country work on the other — is precisely the corroboration that single-method studies cannot provide.

Updated thoughts on industrial policy — Noahpinion — 20 April 2026

Smith endorses the export-discipline playbook but adds a twist. When governments channel support through cheap bank loans rather than direct grants, the mechanism breaks down: failing firms do not simply exit, because their banks have too much to lose by letting them go. The lenders’ incentive becomes to keep those companies alive rather than allow a restructuring that would expose the banks’ own losses. This is Japan’s 1990s, not Korea’s 1980s. The mechanism meant to generate competitive selection instead generates zombie lending.

The objection hits China’s current version of the strategy harder than it hits the Korean original, where government support came mainly through direct fiscal transfers — so a failing firm actually lost its backing. But it raises a general point: the performance test works only when underperformance has real consequences, and directing support through banks can nullify those consequences even when firms are nominally exposed to market competition.

Smith does not push this to a conclusion. The piece ends with the observation that distinguishing between industrial policy types — FDI promotion, export subsidies, protectionism — matters more than the term “industrial policy” alone can carry. That decomposition is overdue.

KOREA AND CHINA

Korea and China transformed; the mechanisms were not the same. Lane provides the causal anchor for Korea. Amodio et al. decompose the mechanism into its components. Magoon, writing from a different starting point entirely, independently identifies the institutional solution China found to the accountability problem that the Korean literature takes for granted.

Manufacturing Revolutions: Industrial Policy and Industrialization in South Korea Nathan Lane — Quarterly Journal of Economics, 140(3): 1683–1741 — 2025

The most careful recent attempt to separate the effect of Korea’s industrial policy from everything else happening at the same time — Cold War market access, Japanese capital, cheap labour — using detailed historical records on which industries received support and how individual firms responded. The key result is that the Heavy and Chemical Industry Drive generated lasting changes in what Korean firms could produce. Companies that received support under the Drive were making more sophisticated goods, at higher capability levels, a decade later. This is what distinguishes industrial policy that builds lasting capability from industrial policy that merely subsidises what firms are already doing.

The policy worked, but it also channelled resources toward large, politically connected conglomerates in ways that could easily have become permanent. What prevented that from happening was Korea’s willingness to let connected firms fail. The 1997 crisis is the clearest demonstration, though the pattern had been set earlier. Lane’s paper provides the evidence for the first claim. The 1997 crisis literature is where the second is tested. Both are needed to read the episode accurately.

How export promotion and technology transfer powered South Korea’s industrialisation Francesco Amodio, Jaedo Choi, Markus Poschke, Michael Sposi — VoxDev — 19 February 2026

The companion piece to Lane, focused on how the transformation actually worked inside Korean firms. The central finding is that export promotion and technology transfer had to operate together: neither alone produced the gains; the combination allowed firms to acquire new production methods at scale and, over the period, to shift from simple consumer goods — clothing, textiles — to complex manufactures like machinery and electronics. The export requirement kept firms answerable to foreign buyers who could walk away, while the technology transfer component gave them something new to learn. Remove either element and the dynamic collapses.

The caveat the authors state carefully is that success depended on implementation capacity and on selection criteria — specifically, export performance — that constrained how much officials could direct support toward favoured firms rather than capable ones. Both conditions are hard to replicate, and neither is automatic. The piece ends with the honest observation that the countries most likely to benefit from a Korea-style strategy are those that already have a state capable of credibly enforcing those conditions. That narrows the set of plausible candidates considerably.

The Puzzle of China’s Economic Success: How a Centralised State Created Competition Among Local Officials — frompovertytoprogress.substack.com — 14 July 2026

China is the anomaly in the transformation literature. Its institutional profile — no pretence of administrative neutrality, party control of enterprise, no genuine discipline from foreign buyers at the outset — fits none of the standard accounts of how transformation happens. China transformed anyway. Magoon’s explanation is the regional tournament. Local officials competed against each other on measurable growth metrics, with promotion for success and removal for failure, creating results-based competition within a politically unified structure. Successful local experiments were systematically identified and replicated nationally; Beijing set the targets and kept score.

The institutional logic aligns with the Amodio, Huang, Lei, and Poschke analysis of China’s rise, which identifies the same regional tournament mechanism. What the two pieces together establish is that China did have a performance test; it ran inside the party hierarchy. Whether that internal competition is replicable elsewhere is a question neither piece fully resolves, and there are strong reasons to think it is not.

How China became the world’s factory Amodio, Huang, Lei, and Poschke — VoxDev — 18 February 2026

Where Magoon explains the regional tournament mechanism, this paper documents the external levers that made China’s transformation work: tariff sequencing, WTO accession, special economic zones, and processing trade — the arrangement by which Chinese firms imported components, assembled them, and exported the finished goods. Each mechanism is tracked with data.

The sequencing matters as much as the policies themselves. Tariff reductions were broadly simultaneous rather than sector-specific — falling sharply across the economy in the 1990s as China pursued WTO accession, which it achieved in 2001. The combined reduction in import and export tariffs intensified domestic competition and drove productivity gains. Special economic zones created bounded spaces where global competitive conditions applied, raising investment, wages, and firm entry. Processing trade — importing components duty-free, assembling domestically, re-exporting — provided a lower-cost entry route into global supply chains, though the paper notes that processing firms typically remained less productive and captured less value than ordinary exporters.

Taken together, these mechanisms show how a state without the clean administrative apparatus of 1970s Korea nonetheless engineered genuine competitive exposure — through external commitments and bounded zones rather than through direct performance conditions. The Magoon piece gives you the tournament logic; this one gives you the machinery.

THE MIDDLE-INCOME TRAP

Three cases and a methodological challenge to the concept that contains them. Masina’s “dependent development” account of Thailand and the World Bank’s “competitive pressure deficit” account of Malaysia are in partial tension with each other, and neither fits Indonesia, where the problem was sound macroeconomic management that never translated into industrial capability. Kumagai provides the regional frame and the contemporary constraint. Katz arrives from outside the case literature to ask whether the trap is a structural phenomenon or a measurement artefact — a prior question the cases cannot settle on their own.

Thailand and the Myth of the Middle-Income Trap — pietromasina.substack.com — 13 February 2026

Masina’s reframing is the most structurally challenging piece on the list. Thailand’s stagnation does not result from reaching a natural development ceiling; it results from a model built around foreign investment, in which multinationals retained core technologies and high-value functions while Thailand supplied assembly labour and infrastructure. The country, in Masina’s phrase, “has never been allowed to climb very far at all.” Vietnam’s competitive threat now reveals what was never built: if Thailand had developed genuine industrial sophistication, cheaper Vietnamese labour could not simply replace it. Its substitutability is the evidence.

The prescription of deliberate state-led policy to build home-grown productive capacity is formally identical to what Cherif and Hasanov identify as the missing element in strategies that fail to transform. The political economy is different. Masina’s argument implies the obstacle is partly of external origin: not only domestic policy failure but also the terms on which Thailand was integrated into international production networks. Whether Thai firms ever faced the kind of demanding foreign buyers who would have forced them to upgrade, or whether the buyers retained the technology and left only the assembly work, is a question the piece raises without fully resolving. It is the right question.

The Final Stretch: Lessons from Malaysia’s Road to High-Income Status Mariana Vijil, Mauricio Pinzón-Latorre, Yew Chong Soh, Siyun He — World Bank Development Talk — 16 April 2026

The World Bank’s diagnosis of Malaysia’s stall lands on insufficient competitive pressure. Malaysia has genuine strengths: management practices, access to finance, and a diversified industrial base with deep integration into global production networks. But firms are underinvesting in new technology and innovation. The authors trace this to sheltered markets: companies insulated from serious competition have little reason to keep improving. The prescription is to expand market access and reduce trade barriers so that innovation becomes a survival requirement rather than a policy aspiration.

The gap between prescription and diagnosis is worth dwelling on. Opening markets to more competition is a structural condition; it does not by itself create the pressure to keep improving that the report’s own analysis says is missing. A firm that can enter a market but face little real threat of losing it faces very different incentives from one that must keep winning customers who can easily go elsewhere. The World Bank is recommending the ingredient while the diagnosis points to the absence of the test. Whether Malaysia’s protected industries can be made genuinely exposed to that kind of competition — given the affirmative action framework that has shaped Malaysian economic policy since the early 1970s — the paper does not say.

ASEAN Faces a Tighter Middle-Income Trap Satoru Kumagai — East Asia Forum — 10 September 2025

Kumagai provides the broader regional framing and the contemporary constraint. ASEAN’s middle-income economies face weak institutions, insufficient domestic innovation, and excessive reliance on foreign firms — and, unlike their Northeast Asian predecessors, they have not produced homegrown companies with the global scale to pull domestic suppliers into more sophisticated production. The window for the classic escape route is also narrowing: geopolitical uncertainty and U.S. tariff pressure are reshaping the international supply chains through which the standard development playbook assumed access could be obtained.

The prescriptions — deeper regional integration, reform of institutions that currently distribute patronage rather than reward performance — describe the destination rather than the path. The piece is most useful for establishing the constraint that today’s would-be transformers face that 1970s Korea did not: the external environment in which export discipline worked then no longer exists in the same form, and the geopolitical logic is moving in the opposite direction.

China’s Slowdown and the Myth of the Middle-Income Trap — richardkatz.substack.com — 5 May 2026

The most direct challenge to the concept itself. Katz’s central point is methodological: the World Bank’s definition of middle-income status measures countries’ income in dollars at market exchange rates, which overstates how far behind middle-income countries actually are in terms of what their citizens can afford. Adjust for what money buys in each country, the standard approach for making living standards genuinely comparable across borders, and the picture shifts considerably. On that basis, Katz finds that 55 of the 57 countries in his sample showed positive per capita income growth over three decades, with only two recording absolute decline. China’s recent slowdown, on this reading, reflects specific policy choices under Xi rather than a structural ceiling the country was always going to hit.

The objection does not dissolve the transformation question. Even if most countries eventually converge toward higher living standards, the rate of convergence varies widely. And the gap between those that reached the technology frontier and those that remained well short of it is real in Katz’s own data. It is a genuine caution against treating stagnation as permanent rather than prolonged, and against presenting the gap between transformers and stallers as starker than the evidence allows. If a substantial share of apparently trapped economies are simply decelerating from exceptional earlier growth rather than hitting a structural ceiling, that is a different phenomenon from the one the case literature describes. And the distinction matters for what policy can plausibly achieve.

An Open Market, a Sound Currency, a Capable State — and Still No Transformation

Stephen Brien — Thu, 16 Jul 2026 07:30:27 GMT

The short version

The textbook says fix three things first: open the market, competitive currency, capable state. This post tests that advice on economies that had them and stalled anyway.
Each was fully in place in an economy that stalled: Thailand’s market, Indonesia’s currency, Malaysia’s state.
None separated them from those that transformed. They’re ingredients a country can hold; what sorts them is a test its firms can’t escape.
This doesn’t prove market exposure caused the divide, only that its rivals didn’t. Where the test came from is the next post.
The checklist can buy growth, not transformation. The tell: can this firm actually lose, to a rival, buyer, or price the government can’t control? If not, you’re buying growth.

This is the second post in a series on why sustained transformation has been so rare. The last post found one thing that tracked the split between four economies that kept climbing and four that stalled. It was market exposure: whether a firm answered to a home protector or a buyer beyond its government’s reach. This post asks the harder question first: could something more ordinary, and far easier to copy, have done the same work?

For thirty years, Indonesia did almost everything the textbooks ask of a poor country. Its economic team, US-trained technocrats the press called the Berkeley Mafia, broke a hyperinflation that had run past six hundred per cent, held the budget close to balance, and kept the currency honest. In 1978, with oil revenue pouring in and every incentive to spend, they deliberately weakened their own exchange rate to keep the country’s farms and factories competitive abroad.

It was the kind of discipline almost no government manages. By every visible measure, Indonesia was one of the best-run economies of its kind. And it built almost nothing the world wanted to buy.

Indonesia is not a freak case; it is the pattern this post is about. Malaysia built one of the most capable states in the developing world and stalled at middle income. Thailand had a large, open market to sell into for years and stayed put. The answers a textbook gives all turn up present and switched on in an economy that grew fast for a generation, then settled and went no further.

Whatever separated the four that climbed from the inertial four, the textbook never names it.

The wider gap between those two groups is what the series is about. The four that kept climbing reached the technology frontier; the four that stalled built real industries and then stopped, and sit a generation later still short of it, better off than before but stuck below the neighbours they once grew alongside. A policy can buy either, and most that buy growth never reach the second.

Four rivals, one fair test

That single line the last post found, whether a firm answered to a protector at home or a buyer beyond the government’s reach, splits the four transformers, the renewers of the last post, from the inertial four almost perfectly. But a clean split is not a proven cause. The two groups differ in many other ways that could explain how far they diversified and how much they innovated.

Four rival explanations are the most direct to test. Three are standard advice given to any developing country: win access to export markets, keep the currency competitive, build a capable state. These three are not strawmen; each is the centre of a major reform agenda: open trade, a stable economy, and the state-capacity movement, and each is mainstream advice for good reason. The fourth candidate explanation is that exposure to global financial shocks can stymie growth.

One larger rival the first post already removed, so this one leaves it aside: the older claim that an activist state picking industries did the work. Industrial policy on its own cannot be the divider: it ran as high in the economies that stalled as in those that renewed. What can is a single attribute of it, its disciplined core, support tied to a standard the government could not relax, which is the external market exposure test this series tracks.^[1]

Teasing them apart is not trivial. These eight countries are all past growth successes, so the ordinary ingredients of growth are common to both sides, and the trick is to tell a real difference from a coincidence.^[2]

The bar is the same for each, and fixed in advance. Market exposure drew a clean line between the groups; the question is whether any rival explanation draws one as clean. A rival fails the moment one economy is found that had it, fully switched on, and stalled anyway. That does not make it useless, only not the thing that sorted the eight.

The test runs only one way. That multiple winners all shared a feature proves nothing, because winners are alike in dozens of ways at once, and any of those shared traits could be the real cause.

Three rival explanations now face that bar. The fourth, financial shocks, is held for the end, where it helps the case rather than rivals it.

One distinction runs under everything that follows. Each of the three is something a country holds: a market, a rate, a bureaucracy. Market exposure is not something a country has, but a test its firms must keep passing.

A market to sell into

“Get your firms into the world market” is the first instruction in almost every reform program, and for good reason: a poor country’s home market is small, the world’s is not. But simple market access is not what separated these eight, and seeing why protects a government from buying the wrong kind of it.

Access means two different things. It can mean a market to sell into, one that buys what a firm already makes. Or a market that makes a firm climb, one that keeps buying only if the firm keeps mastering harder work. Transformation is that climb, so it is the second meaning the claim needs.

All four transformers built themselves on exports, so the suspicion is fair: perhaps they faced markets that pushed their firms upward, and the inertial four did not. What each economy’s firms actually faced shows why that suspicion does not hold.

The plain meaning comes first: a market to sell into. The economy that had the least of it was China, which entered its industries decades late, into a field its neighbours had crowded, and had to win its buyers rather than inherit them. It transformed anyway; its 2001 entry to the World Trade Organisation only confirmed a competitiveness it had already built over two decades.^[3] A ready market to sell into is not required, or the transformer with the least of it would have failed.

Thailand shows the reverse. Through the Vietnam War years, American military spending ran near four per cent of its GNP, up to two-fifths of export earnings. It was an enormous market that bought whatever Thailand offered without ever asking it to make anything harder. Thailand stayed inertial. A market that only buys is not enough.

That leaves the harder question: what a market that makes a firm climb actually does. From outside, a market pushing firms upward and firms climbing on their own look identical. So to call the market the cause, you have to catch it doing the pushing, not accommodating a climb that would have happened anyway.

Korea, Taiwan, and Singapore climbed, from cheap goods to world-class ones. But they climbed with heavy help: Cold War markets held open as alliance policy, capital and technology from Japan and the United States. The obvious winners fit the idea that demanding buyers pushed them up, but cannot confirm it. When several advantages always travel together, none can be singled out as the cause; with that much help in play, the market’s push cannot be distinguished from the rest.

China came closest to isolating it, entering late and without a Cold-War ally’s inherited market access, yet it still went from cheap assembly to electronics. But China was hardly unaided. Its export climb leaned heavily on foreign firms plugging into East Asian production networks, and it ran an aggressive industrial policy up the same ladder, subsidies and directed credit. So even the strongest case cannot confirm it: its own confound was foreign parents and state push, moving with the market and never behind it alone.

Subscribe now

Malaysia is the case that settles it, the one place market access was forced and isolated. Its Penang electronics zone required firms to export the bulk of their output, a condition the trade data enforced, not at an official’s discretion. The firms it bound were mostly foreign, and they did climb: the multinational plants in Penang moved over the years from bare assembly toward design, testing, and higher-value work.

But they climbed as arms of foreign parent companies. Malaysia’s own suppliers remained lower down the chain, and the country saw no frontier firm of its own emerge. Forced access lifted the branch plants that faced it; the capability went home with the parent company. A market that a country hosts is not the same as its own firms being made to climb.

Market access does not track the divide: China had the least and transformed, Thailand had the most and stalled. As something a country can hold, a market to sell into cannot be what divided the eight. Present, yet inert.

What survives is not access but the pressure it is mistaken for: firms that must keep making harder things or lose their buyers, and so can never rest. That is a pressure firms are put under, not a market a country holds. Next, the price of the money itself.

Getting the price right

“Get the exchange rate right” is advice every finance ministry has heard, and it is well supported by evidence. Dani Rodrik finds undervalued currencies raise growth across the developing world, and nothing here disputes it. A cheap currency makes exporting profitable across the board, pulling capital toward the tradable industries where growth is meant to come from. Yet it is still not the divider, and here the reason is almost arithmetic.

A cheap currency faces the same bar. Did the transformers keep theirs cheap and the inertial four not, or did both? How cheap each of the eight kept its money falls cleanly on one side.

The hardest case for the idea is Indonesia, where money was not just cheap but very well managed.^[4] Its technocrats took inflation from six hundred per cent to single figures, and in 1978, mid-oil boom, devalued anyway to stop oil earnings from pricing exports out.^[5] Thirty years of that discipline built no lasting tradable industry, no industry that could survive selling to foreign buyers. If cheap, well-run money were the key, Indonesia would have transformed. It did not.

Indonesia is not a fluke. Malaysia held the ringgit fixed at 3.80 to the US dollar for seven years after 1998, a rate judged undervalued when the peg was set; Thailand devalued in 1984 to protect exporters. Neither transformed. The winners were cheaper for longer, true; but a genuinely cheap currency still sat on the stalled side for years and pulled nothing across.

That same fact turns up on the winning side. Korea and Taiwan were both named currency manipulators by the United States in the late 1980s, for holding their currencies down against pressure to rise. China ran the longest version: its 1980s devaluation opened the renminbi, which stayed cheap for a generation. China’s undervaluation was smaller than once claimed, but three of the four transformers kept the price genuinely low as well.

A cause that divided the eight would sit on one side, not both. This one sits on both. By that bar, the exchange rate is out.

But out as the divider is not out altogether. Rodrik’s finding stands. Cheap money does raise growth. So if the currency did not divide them, what did it do?

Brazil, the fourth staller, answers that, and it could answer this way only because its home market was large. It ran the cheap-currency strategy more deliberately than anyone else, behind the highest walls: multiple exchange rates, tariffs, and export subsidies, maintained for half a century.

Its firms sold to a captive Brazilian public who, with imports blocked, could take their custom nowhere else, and there were enough of them to build a real industrial base without ever facing the world. Keep the price cheap behind walls like that, and industry grows and grows but cannot outlast the day they fall.

The firms exported too, cheaply, on that same currency and those subsidies. But they never needed to win the foreign buyers who could always walk away; a home market that large kept them alive regardless. So nothing forced them to improve. When the walls came down after 1990, and Brazilians could finally buy imports, their own customers could at last refuse them, the cheap currency that had flattered their books was gone, and the firms, never made competitive, collapsed.

Brazil industrialised; it did not transform.

The currency makes it plainest. A level is not a selector. A cheap currency is one price for every firm. It is low for the exporter that just beat a demanding buyer, and just as low for the protected firm that never faced one. It cannot ask which of them earned its edge.

So how much can the exchange rate explain? Where a firm must win foreign buyers free to walk away, a cheap currency raises the reward, and the firms already winning pull further ahead. Where a firm lives off a captive home market, it only makes their goods cheaper and improves no one. It can widen a gap real competition is already opening; it cannot open one. Singapore, fitting neither side, is left where it lies.

Getting the price right was never a substitute for the test; at most it sharpened one already running. Which leaves the obvious question: if a price cannot sort firms, what kind of state can?

A capable state

“Build state capacity” is the reflex answer when a government, or the donor advising it, asks what to fix first, and it is not hollow. A capable state can watch firms, hold them to targets, and let even a favourite fail. It is also the most expensive answer to get wrong, because building a bureaucracy takes a decade. And it is not the thing that sorted them.

If it were, two things would have to hold. The winners would need more capacity than the four that stalled, and they would need it before they pulled ahead, not after. On amount first, the winners did not have more.

If capacity were the answer, the states with the most of it would be the ones that transformed. They are not. On the World Bank’s Government Effectiveness measure, Malaysia scores near the eightieth percentile, higher than four-fifths of the world’s governments, above every economy that stalled and level with China, a country that renewed. Malaysia stalled all the same. So the single best summary measure of capacity already fails to sort the eight: the highest-scoring of the stalled economies outranks a winner.^[6]

Maybe one measure is too blunt, so split capacity into its two real parts. The first is how well the state is built: whether it hires on merit, pays properly, offers a career. The second is how evenly it applies its rules to everyone alike.

Neither part sorts the eight. On how it is built, Singapore, Korea, and Taiwan score high early, but so does Malaysia, well above Thailand or Brazil. On even-handedness the order actually inverts: Brazil is more even-handed than any stalled economy, close to Korea as recently as 2024, while China, a renewer, is nearly the least even-handed of all.^[7] So amount, in every form, is not what separated them: a stalled economy sits among the winners, and on even-handedness beats them.

The timing settles it. Korea and Taiwan are remembered for model bureaucracies, but the memory runs backwards. Through their 1960s and 1970s takeoff, both scored at or near zero on even-handed administration, below Malaysia; the clean, professional states they are famous for arrived only after both democratised in 1987, long after they had pulled ahead. They earned the reputation only after they had already won.^[8]

Singapore is the one exception: a strong state, and early. China is the harder case: a genuinely capable state that has never been an even-handed one, scoring low on that measure then and now. That leaves only one winner in four whose even-handed administration came before the divide opened; for the others it came afterwards, or never. A cause cannot come after its effect.

Capacity, then, sorts the eight neither by amount nor by timing. Which raises the real question. If Malaysia had a capable state, what was it doing? Its uneven rules were not a fault; they were policy. Malaysia’s New Economic Policy of 1971 committed the state, openly and for the long term, to favouring one community, a generation before growth stalled.

Petronas shows the state at full power: superbly run, staffed by genuine technocrats, it once withheld royalties from a state that had voted for the opposition. It could compete abroad and hold the ruling coalition together at once, and when those aims collided the coalition won. South Africa makes the same point from another angle: a genuinely capable core, aimed at sharing income at home rather than at any test set abroad.^[9]

Subscribe now

The capacity was real; it was aimed where the coalition required, not where a test abroad would have pointed it. And the aim came first. The favouritism was law in 1971, and the state grew less even-handed from the early 1980s, before growth stalled, so it was cause, not symptom.

So state capacity did not decide the divide, and could not have: it sat on both sides of the split, missing where the theory needs it and present where it does not. But “not the cause” is too blunt. It is not what starts transformation, since the winners pulled ahead before their administrations were clean, and it is not sufficient, since Malaysia’s competence bought it nothing.

Nor is it idle. The buildable half, hiring on merit, paying for skill, the winners had at takeoff, as did Malaysia: shared background, not an edge. The half that moves is even-handedness, which, in Korea and Taiwan, strengthened as they closed in on the frontier. So a competent administration is something frontier growth develops and leans on, not a lever a country starts with. It is neither the trigger nor the guarantee.^[10] Even that the eight cannot prove: China sits at the frontier with one of the least even-handed states.

That is the real lesson for anyone who reaches for capacity first. A bureaucracy is never neutral. Unlike a market that can sit unused or a currency left mispriced, it is always doing something, and gets better at whatever it is aimed at. What a state cannot buy is the choice of where to aim.

A capable state aimed at the wrong target stalls no less for being capable.

Three ingredients and a test

Three eliminations are not yet an explanation. Unless the three unsuccessful candidates share one reason, this is just a list of dead ends. They do not: the three line up too neatly to be chance. What they share is that a country can simply hold each one, and none of them asks a firm for anything in return.

None of that is a knock on the three. Each is genuinely good for growth, and these economies grew fast for a generation on them. But growth per se was never the puzzle; all three factors contribute to it. The puzzle is what separated the four that kept transforming from the four that grew and then went inert, still good, just not great. And on that, the three come up empty: each was present on both sides of the divide.

Market exposure is different in kind. A firm does not own it. A firm faces it; in a buyer, it must win, or a rival its government cannot call off. And it is this that makes the other three factors bite. A cheap currency helps only when firms must sell abroad to survive; a capable state only when it holds firms to a standard instead of sheltering them; market access only when a firm has to win the sale, not just reach the market.

This test is not just competition by another name. Ordinary competition a government can still blunt, sheltering a firm or tilting the field; here the verdict is beyond reach. A buyer who stops calling cannot be lobbied or appealed to by the firm or by its government.

An ingredient is a stock. It sits at whatever level it sits at and does not accumulate. A test is a standard applied again every year. The firms it weeds out stay weeded out; the survivors keep facing it and keep improving.^[11] So a standard, not a stock, is the better candidate for the lasting distance between growth and transformation, a standing pressure a laggard cannot close by acquiring more of any one ingredient.

Of the four candidates, only market exposure can compound, and that is what a laggard cannot buy its way past. Each round starts from the last round’s survivors, so the distance can widen where a fixed advantage would remain. The one long-run series, patents per head, shows Korea, Taiwan and Singapore well clear by 1990, with China the late exception, so the widening is better read as a gap that opens at the sort and holds, not a rate the eight cases prove.^[12]

So the three ingredients get a country its growth; they cannot, by themselves, carry it the rest of the way.^[13]

Four candidates collapse into one shape: three ingredients that lift everyone, and the test that decides whether the lift becomes transformation.

The whole argument fits in one view:

In practice, the test is not exotic. It is a firm that keeps its state backing only while it keeps winning buyers abroad, and loses it when it stops; a national champion genuinely allowed to fail; a protected market thrown open so a domestic favourite must beat an import it cannot block. The three ingredients can all be in place, and none of that be true.

That distinction also cuts at the get-the-basics-right consensus: the three work alongside the test; they do not create it.

And because a test pays off only as round after round of firms are filtered out, a reform judged too early will look like a failure long before it has had time to work.

None of this is a new prescription. It is the property the familiar ones rely on: promote exports, honour your deals, discipline firms by their sales abroad,^[14] each working only so far as the standard falls to someone the government cannot later lean on.

So the thing that actually separated the eight is starting to show itself. It is the discipline the ingredients answer to, not one more ingredient. Where it comes from, and what to call it, is the next post’s subject.

The crisis that revealed the test

There is one explanation for the divide that has nothing to do with any test: bad luck.

That rival is different in kind: not something an economy carried in, but something that hit it, a financial crisis. Any official who has lived through a crash will feel the intuition: maybe the inertial four were simply unlucky, struck harder by the era’s international financial crises than the four that kept climbing, and knocked off course by them. A severe crisis can bankrupt firms and cost a country a decade, so if those blows fell unevenly, the divergence is misfortune, not any test.

For that to be true, the crises had to fall harder on the inertial four than the renewers, and each stall had to begin when its blow landed, not before. So two questions: did the crises sort the eight, and did each stall start when its own crisis struck?

Severity comes first. On the standard measure of how hard a crisis hits, how far a country’s currency fell and how much reserve it burned defending it, the 1997 crash put Thailand deepest in the red, with Malaysia, Korea, and Indonesia close behind. Korea, a renewer, was hit as hard as anyone, and took the IMF’s largest-ever rescue. The two renewers that escaped, Singapore and China, did so by policy, not luck. Taiwan had taken its own hit earlier, in 1974. Severity does not sort the eight.

Then timing. The case the luck story rests on is Brazil, a miracle knocked off course by the 1982 debt crisis and then a currency collapse in 1999. But the dates do not fit.

Brazil was already failing before 1982. Its productivity growth had stalled by the late 1970s, the boom propped up by cheap foreign borrowing. The blow was not even exceptional, since Korea carried more debt than Brazil into the same years and kept growing through it. And the stall outlived the crisis: the debt was cleared in 1994, yet the plateau ran on regardless. A crisis this ill-timed cannot have created the divergence, only hurried it.

So the pattern the luck story needs is not there. Exposure does not sort the eight: a renewer took the hardest blow, and the renewers that escaped did so by design, not fortune. Nor do the stalls begin when the blows land: Brazil’s ran before its crisis and outlasted it. By its own test, crisis vulnerability is out as the divider.

Yet ruling it out as the divider does not rule it out as evidence. In 1997 the crisis did something: it sorted. If it did not create the divide, what was that sorting?

One crisis hit Korea and Malaysia together, and within months they ran opposite experiments. Korea took the IMF’s rescue with restructuring attached and forced its largest firms through it. Malaysia refused it, raised capital controls, barriers stopping money leaving the country, and sheltered its connected firms most of all, a defensible recovery on some readings, a sheltering all the same.^[15] That shock was the same; the difference was all in the response.

That choice was shaped by debt and politics, granted.^[16] But the direction was set long before: Korea had tied its firms’ support to winning foreign buyers for years, Malaysia had sheltered its own regardless.

The test being tracked is whether a firm must keep winning buyers who are free to walk away. It runs every day: the firms that face it and pass grow stronger, while the ones kept sheltered from it slowly hollow out.

A crisis does not start that reckoning; it forces it into the open, all at once. Korea, which had been running the test, used the crash to enforce it; Malaysia, which had been dodging it, dodged again. Each did under pressure what it had done all along.

1997 did not administer the test. It collected on a verdict the test had already reached.

This is not fresh evidence: Korea and Malaysia are two of the eight the pattern was built from. Whether they sorted the same way in 2008 and 2020 is a check for later work.^[17]

So the fourth rival ends differently from the first three. Those were real ingredients that sat inert on both sides. Crisis vulnerability is not even that: the exposure the luck story needs was never there; at most a crisis hurried a divergence already moving. Yet it is the least dismissible of the four.

And the fork is no historical curiosity. Every government meeting a shock today faces the same choice. The shock may be a debt crisis, a bank run, a currency collapse; the choice does not change: use the emergency to make the reckoning land, or buy the connected more time.

Subscribe now

What is settled, and what is not

The result is narrower than it looks. Four rivals were tested against eight economies that grew alike and then split. Three of them, a market, a competitive currency, a capable state, are all genuinely good for growth, but each turned up on both sides of the divide, so none can be what separated the transformers from the rest.

Only market exposure drew a clean line, and it did so because it is not an ingredient at all. It is a test: whether a firm must answer to a buyer, a price, a rival its government cannot lean on. The ingredients matter, but they are not the secret sauce; the test is.

One question tells you whether a country has it, and an official can ask it of any firm their own policies favour: can this firm actually lose, to a rival the government cannot overrule, a buyer it cannot lean on, a price it cannot fix? If the honest answer is no, the policy is buying growth, not transformation.

Two limits sit on the claim itself. A financial crisis can set a country back a decade, but it only reveals what a country’s firms already are; it does not make them. And market exposure stays a candidate, not a verdict: it names its own undoing, since an economy that put its firms to a real, unrelenting test and still built nothing tradable would sink it. None of these eight is that case.

It leaves other questions untouched. It says nothing about who paid to build the capability, the suppressed wages, the broken unions, the land taken; that cost is real, and a different question from this one. Nor can it say why some economies’ firms had to face the test at all, when a large domestic market or oil income could spare them; that is a question for a later post.

What none of this changes is a practical asymmetry. A government can copy the ingredients this year: peg the currency, professionalise a ministry, sign a trade deal. It cannot copy the standard the same way, because the standard works only when the government cannot overrule it: a test the government controls is one it can always soften.

Building industry is not the same as building the thing that keeps it improving.

A country can pour capital into factories and, a generation on, find itself optimising yesterday’s growth engine while the frontier moves on without it. For any economy industrialising today, the question is not whether to build the factories. It is whether anything will force them to keep getting better once the first wave of growth fades.

Which leaves the question this post has not answered. Two of these economies put their firms to the test, and two did not. Why the test took hold where it did is a question about origins, not outcomes. It is where the next post begins.

Eight Miracles Grew Alike. Why Did Only Four Transform?

Stephen Brien — Mon, 13 Jul 2026 06:20:18 GMT

The short version

The World Bank’s 2008 Commission on Growth and Development named thirteen genuine postwar miracles; this post lines up eight of them and asks what split them.
Not accumulation, and not industrial policy: investment, foreign capital, and directed credit all ran as high in the four that stalled as in the four that renewed.
What divided them was one thing, usually attributable to a specific act: a change in what each economy’s firms answered to.
The renewers bound their firms to a standard set outside the state’s reach, a foreign buyer’s verdict no minister could lower, and one no failing firm could plead its way out of.
The usual accounts already say a good state disciplines its champions and makes them export, but most states will quietly ease a struggling champion’s terms.
The rare thing, and the hardest to copy, is a test the government itself cannot move.

This is part of a series on why sustained economic transformation is so rare. Earlier posts met two of these economies one at a time: Indonesia, which managed its money better than almost any comparable country for thirty years and still never built the firms that could have made it rich; and Brazil, a certified miracle that sits today almost exactly where it started. Here the series steps back from the single cases to the pattern across them. How the difference arose and why some governments could sustain it while others could not are questions that later posts take up.

In 1983, Malaysia set out to build a national car. The Proton, championed by Prime Minister Mahathir Mohamad, produced its first model two years later, and the state ensured it would sell: duties on competing imports climbed over the years to 300 per cent. Behind that wall, Proton took most of the Malaysian market.

Korea had made the same bet a little earlier, but in the opposite direction. Hyundai, which built its first car in the 1970s, was pushed outward. It shipped the Pony to Canada, then the Excel to the United States in 1986, into markets where nobody had to buy a Korean car and almost nobody had heard of one. It sold poorly at first and had to improve to survive.

Forty years on, the two cars carry the difference. Hyundai is one of the largest carmakers on Earth, selling in nearly every market. Proton never made it abroad; in 2017, unable to compete, Malaysia sold nearly half of it to a Chinese company to keep it alive. Same industry, same era, same national ambition. One firm was made to face the world. The other was protected from it.

Same league, different frontiers

These two cars are one instance of a divide that runs through all eight of these economies. Two groups of four began at similar incomes and grew at the same fast pace for the same length of time, then split: four climbing on, four stalling. All eight grew, and all eight outran their peers, so growth is not what divides them.

Four did something rarer: South Korea, Taiwan, Singapore, and China kept renewing as they approached the technology frontier, generating new world-beating industries as old ones matured, and went on climbing after the fast-growth years ended. The other four grew hard and then ran out of road, short of it. That divide, renewal rather than growth, is the one worth explaining, and the one the usual explanations miss.

Korea stood at eleven per cent of American income per head in 1960 and reached fifty-one by 2022; Singapore went from twenty-three per cent to one hundred and fifteen; Taiwan from nine to fifty-seven. All three now sit at or near the frontier and have not stopped.

China is the hardest of the four to place. It rose from seven per cent of American income per head in 1960 to thirty by 2022, still closing the gap: it began transforming later than the others, on thinner early evidence, and only lately pulled clear. It stays in view as the case that most tests the pattern, not as a loose end.

Four others grew quickly within the same generation, and against their own past, every one of them succeeded. Malaysia rose from eleven per cent of American income to thirty-two per cent; Thailand from five per cent to eighteen per cent; Indonesia from four per cent to eight per cent. Then inertia set in, short of the frontier.

Brazil grew hard into the 1970s and then stalled short: a certified miracle that, sixty years on, sits a single point from where it started, twenty-two per cent of American income in 1960 and twenty-three today.

The difference is not that the four renewers simply climbed higher on the same ladder; they kept building new rungs, industries that had not existed before, while the others reached the top of the ladder they were on and stopped.

None of the four went inertial because they hit some low ceiling. Malaysia gave out at the very income Korea passed through on its way to fifty-one; the two ran the same stretch of road, and only one kept going.

How the comparison works

Comparing these eight is closer to a matched comparison than to lining up eight stories.^[1] No one designed the split; history held most things equal and changed only the outcome, so the eliminations that follow are evidence, not opinion. This is the logic of a controlled experiment: hold everything else constant, change one thing, and infer the cause from the difference.

Its power is in what it rules out. If a factor ran just as high in the inertial four as in the four that kept climbing, it cannot be what separated them, whatever the textbooks credit it with. Investment and foreign money: if the inertial four had as much as the renewers, none of them is the answer. That lets us cross candidates off the list without arguing each one on its merits.

It narrows the field; it does not deliver a verdict. Eight countries are not a laboratory, and several real differences run all the way through, land equality among them.^[2] So the plan is three steps: clear away the answers that ran equally in both groups, find the one thing that did separate them, and trace it to something concrete in the record, an act you can point to that came before the split.

Get all three, and we have a strong candidate, but not necessarily a proven cause. That will be a job the next post takes up.

It was not how much they put in

The standard account credits foreign investment with more than money. Multinationals are supposed to bring technology, management practice, and a ready-made route into global markets, and that, the story goes, is what made the difference. As a story about how much foreign money came in, it does not hold up under comparison. Korea and Taiwan attracted almost no foreign direct investment during their transformation decades, at below one per cent of GDP for most of the period. Malaysia attracted four to five per cent, Thailand two to four.

Thailand shows what that access bought. By the 2000s, it had become one of the world’s largest vehicle-assembly hubs, almost entirely on foreign nameplates, Toyota, Honda, Isuzu, without producing a domestically owned automaker to rival Hyundai. Two of the inertial four were more open to foreign capital than two of the four that renewed.

High investment is the classic engine of catch-up growth: more capital per worker should mean more output, and for decades this was credited with East Asia’s rise. Investment rates do not draw the line, though. Thailand reached forty-two per cent of GDP in fixed investment (factories, machinery, buildings) before 1997, more than Korea ever managed, much of it in real estate and speculative finance rather than productive capacity. Malaysia hit forty-four, and both saved about as much as Korea.

Urbanisation is supposed to move workers out of low-productivity farming and into cities, where productivity gains are concentrated. Here the pattern runs backwards. China, which renewed, was among the least urbanised of the eight; Brazil, which went inertial, among the most, exactly the reverse of what the story predicts, so it cannot be urbanisation either.

On the standard scorecard, the renewers matched or trailed the inertial four on every accumulation mark, and transformed anyway. What none of them captures is what an economy was accumulating for.

This sharpens an older argument: Krugman and Young once warned that the East Asian miracles ran on perspiration rather than inspiration and might one day run dry. These eight push the point further. Within this set, the pattern flips: the economies that accumulated hardest, Thailand and Malaysia, are the ones that stalled, not the ones that transformed. Accumulation does not separate the miracles that transformed from those that ran out of road.^[3]

It was not the state, or simply the market

Development economists usually reach for a different kind of explanation next: industrial policy, in which a country’s government actively directs where its resources go. Tested the same way, it fails most clearly of all to separate the two groups.

Singapore built almost no national champions of the kind Malaysia or Brazil ran behind tariffs; it kept its economy open to the world’s firms and let them carry capabilities in from outside, even as the state itself stayed a major owner of the firms it did hold, from the port to the airline to the banks. Korea did the reverse: state-directed credit, protected home markets, state-backed conglomerates, monthly meetings at which firm-level export numbers were read down the page. Both renewed.

Malaysia ran the Proton car project under tariffs that reached 300 per cent, a heavy-industry programme, and a national-preference system. Brazil built Embraer, ran the BEFIEX export scheme, and sheltered its industries for decades. Both went inertial.

If active industrial policy produced renewal, Malaysia and Brazil, which ran large industrial-policy programmes, would have renewed. If only the absence of industrial policy could produce renewal, Korea, which ran a heavily directive industrial policy of its own, should have stayed inertial instead. Neither happened. Measures such as tariffs, state ownership and directed credit either run at least as high in the inertial group as in the renewers, or cut clean across the divide.

Tariff levels alone are a crude measure. They say nothing about what was attached to them, and that condition, what a firm had to do to keep the protection, is a difference worth holding onto. That is true of industrial policy itself: the bare presence or absence of directed credit and targets does not separate these eight either, but a conditional version, one whose test a government cannot later soften, might, and that is the variable the next sections isolate.

This cuts across a thirty-year debate. The developmental-state scholars, Amsden, Wade and Studwell among them, credit activist states that picked industries and directed credit; a market-friendly tradition credits openness and free prices. Korea and Singapore sit on opposite sides of that choice, and both reached the frontier all the same. The axis the debate runs on is not the axis that divides these eight.^[4]

Political openness fares no better. Two of the renewers, Singapore and China, kept their politics closed and effectively one-party throughout. Two of the inertial four, Brazil and Indonesia, opened theirs, democratising in 1985 and 1998, and went inertial all the same. Autocracies sit among the winners and democracies among the losers; whichever did the work, it was not the openness of the politics.^[5]

Between how much an economy took in, how its government ran things, and how open its politics was, none of these holds. Something else was doing the work.^[6]

What does separate them?

Every measure examined so far asks how much arrived: how much capital, how much schooling, how much support. Closing the income gap on rich countries, convergence growth, did happen in both groups. But closing the gap is not the same as reaching the technology frontier, and none of those measures asks the question that matters here: what kind of capability an economy came to possess.

By capability, this post means one thing throughout. It is an economy’s own power to make hard things and invent new ones, held inside its firms rather than borrowed from foreign parents.

What follows tests that capability in three places, moving from the easiest to observe to what underlies it: how diversified an economy’s output became, whether it grew firms of its own in the hardest technologies, and how much invention, relative to its size, sits behind that.

That is the half of the same debate the accumulation numbers could never reach: the capability that got built, what Nelson and Pack set against the accountants.

A diversified economy signals real transformation rather than mere growth: it means firms have assembled enough diverse capabilities to produce a wider range of sophisticated products, not simply more of what they already made.

It also marks the shift from catching up to leading. Catching up can happen inside a narrow set of activities, pouring more capital and labour into the same goods; staying at the frontier means constantly finding new, more challenging things to make, which shows up as an economy’s export basket growing steadily more diverse and complex.

Hyundai’s own arc traces the same shift at company scale, from an adapted, licensed compact car in the 1970s to a firm now building some of the most complex vehicles in the industry. Korea’s national export basket moved the same way: simple, labour-intensive goods in the 1960s giving way to ships, machinery, and electronics, each harder to make than the last.

Economic complexity measures whether the things an economy exports require capabilities few other countries hold, the ECI in the Harvard rankings. On the Harvard Growth Lab’s rankings, Korea, Taiwan and Singapore have all ranked in the world’s top ten since 2012, and China has climbed into it, from seventeenth to tenth. Thailand, Malaysia, Indonesia, and Brazil have not: every reading from 2012 to 2024 places each of the four somewhere between the low twenties and the high sixties, never closer than 12 places to the nearest of the four renewers.

Two of the four in the inertial group, Thailand and Malaysia, place in the high twenties and low thirties rather than near the bottom; Indonesia and Brazil, despite qualifying as bona fide growth miracles, sit closer to the bottom of the complexity rankings than to the top.

Complexity measures what an economy makes, not who owns the capacity to make it; on that first measure the line is real but still gradational.

A difference in kind

The four economies that renewed look to have crossed into something the other four never reached: invention-bearing capability, and for three of them, ownership of it too. Call it a capability gap.

Homegrown firms that compete globally in the hardest technologies are a sharper test than export data alone, because a country can diversify its exports simply by hosting multinationals that bring their own capabilities. Owning the firm means the capability lives within the economy, not just in a factory that a foreign parent could relocate.

That distinction is what separates catching up from renewing. Catch-up growth can run on imported capabilities: license the technology and host the assembly line. Exports diversify without the country ever learning to invent on its own. Sustaining growth beyond the frontier requires firms that can continue generating new capabilities themselves, which is exactly what a homegrown presence in the hardest industries demonstrates.

Among the Fortune Global 500, the inertial four countries essentially have no homegrown firms in tradable technology. Their largest global firms are state-owned oil and gas companies: Pertamina, Petronas, PTT, Petrobras. The four that renewed field Samsung and SK Hynix in chips, TSMC in fabrication, Hyundai in vehicles, Huawei and BYD in electronics and electric cars.

The inertial four field state oil companies. The four that climbed field Samsung, TSMC, Hyundai, and Huawei.

Singapore is the exception among the four, and an instructive one. Its entries on that list are a commodity trader, a palm-oil group and a food company, not a manufacturer or a technology firm. Singapore reached the frontier by hosting it rather than owning it.1

What kind of deal a firm grew inside shows up in the firm itself. The inertial four’s leading firms are what closed, protected deals tend to produce: firms that made their money from resource exports or a protected home market, large, connected, and profitable, but never meant to outcompete anyone.^[7]

Renewers grew the opposite: firms that lived or died on winning in open export markets. The gap lies in the kind of firm a deal produces, not only in the firms themselves.

Patents test the same thing more precisely than trade data can: not just whether an economy makes sophisticated things, but whether it invents them, which is a genuine, protectable claim to have created something new rather than assembled, licensed, or copied it.

Subscribe now

Catching up mostly runs on borrowed technology: adapting, adopting, and reverse-engineering what already exists elsewhere, which takes little of a country’s own invention. Pushing past the frontier leaves nothing left to borrow; an economy has to generate the next idea itself. Patents are measured per million residents, not in raw totals, so a big country earns no credit for being big.

In 1990, Taiwan already filed patents at 36 per million people and Korea at 5, roughly 20 to over 100 times the best inertial rate, Brazil’s 0.27. By 2000, Taiwan had reached 260 per million, Korea 74, Singapore 60. China, the fourth renewer, was filing at 0.13 per million that year, behind Malaysia and behind Brazil, indistinguishable from the group it would later leave. Three renewers had separated. China had not yet.

By 2020, China’s rate had reached 18 per million: an order of magnitude below Korea, Taiwan, and Singapore, who were now filing in the hundreds, but also an order of magnitude above Malaysia, the best of the inertial four, still under 3. China had crossed a line the other three renewers had crossed decades earlier. The split that had been three against four for most of the period being studied became four against four only in recent memory.

This was never a difference in how much a country has. It is a difference in how many of a country’s own people are doing the inventing, and for three of the four renewers that difference has held since before this comparison begins.

Economic diversification showed the split by degree; firm ownership and invention read as a difference in kind, not just degree. On invention the four separate from the inertial four cleanly. On ownership three of them do, and Singapore does not. That is the one place these measures disagree.

This is the frontier renewal from the opening, now measured. It says where the four arrived; it does not say what got them there. Whatever explains the split will have to be a plausible cause of these three outcomes: diversification, homegrown firms, and invention.

The one act the four had in common

Trace that gap back, in the cases where the record is clearest, and one thing co-varies with it. It is not a policy. It is a change, sometimes a single dated act, sometimes a slower shift, in what a country’s leading firms had to do to survive. They had to prove themselves against real competition, not rely on a protector who could be persuaded, delayed, or outlasted, provided the change comes before the split it explains.

Taiwan in the late 1950s had no oil, no minerals, and no room to keep weak firms going. Like most developing economies, it protected its domestic market with import barriers, and on top of that ran an overvalued currency that starved exporters of what they earned abroad. Between 1958 and 1960 the reformers around K. Y. Yin fixed the second problem, not the first. They devalued and unified the exchange rate, gave exporters foreign currency at world prices, and ended the quarterly rationing of foreign exchange.

The effect was to change what was profitable. A firm that had lived behind the old rate found the shelter gone; a firm that could sell abroad could now get what it needed to do so. The economy’s reward turned on one thing. Make something foreigners will choose to buy.

Korea reached the same place by opposite means. Park Chung-hee chaired monthly export-promotion meetings from 1965 that tracked targets firm by firm. The credit a firm could draw on moved with its exports, and banks lent to exporters on cheaper terms than anyone else got. Over the following decade, a firm could lose its General Trading Company status, the designation that carried preferential export financing, for missing its export targets.

Support was traded for performance rather than given for free, the mechanism Alice Amsden named reciprocal control, and Korea is the case she built it from.

The discipline grew messier as the stakes rose: the 1970s heavy-industry push leaned more on after-the-fact mergers than up-front conditionality, and not every rescue went as cleanly as the export system implies.

Singapore did it a third way, and its case carries a caveat the other three do not: a trading city-state this small never had the peasant agriculture or protected market the others removed, so what counts here is one narrow property, not its size. No firm-level conditions at all, and almost no protection to remove in the first place. It ran an economy close to a free port, near-zero tariffs on manufactured goods, no meaningful import-substitution phase, and its leading firms faced imports and foreign rivals from the start.

Where the state did hold a stake, in the port operator, the national airline, the domestic bank, it held it through a commercial vehicle, Temasek, that expected a market return and offered no shelter from competition. Cost was the wider lever. It worked in two stages. From 1968, two laws stripped unions of independent bargaining power and put wage-setting in the state’s hands, holding labour costs below Korea’s, Taiwan’s, and Hong Kong’s while the country built an export base.

A later, three-year wage-correction policy, 1979 to 1981, pushed wage costs up sharply by design, pricing the cheapest, most labour-intensive work out of Singapore, though workers’ real pay rose only modestly, about three per cent a year.

Singapore is also the case that pulls two ideas apart. It imposed almost no export requirement, yet its firms were still judged by buyers no minister could lean on. The operative thing was never being made to export; it was being answerable to a verdict the state could not write.

China did it later and more slowly, and unevenly.^[8] Deng’s reforms from 1978 left the old state sector largely protected, with soft loans, guaranteed inputs, and little exposure to failure. A narrower channel opened elsewhere. Township-and-village enterprises operated outside the plan altogether. They answered to buyers rather than planners, while the coastal special economic zones were established from 1980, drawing firms into foreign joint ventures and intensifying competition.

Chinese tariffs stayed high nationwide through this period; the exposure ran through those two channels specifically, the special zones and the enterprises outside the plan, not yet through the border as a whole.

Taiwan’s mechanism was close to pure market. Strip away the currency distortion and let world prices sort firms directly. Korea’s was closer to pure state: a government that tracked and rewarded firms itself, though strictly by their performance abroad. Singapore and China sat in between, a wage floor forced by the state, a plan unwound in stages by the state, each pushing firms toward a test that neither state nor firm fully controlled.

What stayed constant was not the specific instrument but the kind of deal each state struck with its firms. Structurally, the deal had two features: it was ordered, honoured once made rather than quietly renegotiated, and open, available to any firm that could meet it rather than to a connected few.^[9]

The standard that a firm had to meet came from outside the state’s reach, a foreign buyer’s choice, a world price, a competing firm’s better product, and that is the one thing a government cannot fake.^[10] Binding your own firms to that standard is a choice a state makes; the standard itself is one it cannot write, and it is what the four had in common.^[11]

The difference is not how hard a government leans on its firms but whether a failing one can plead its way out. Korea leaned as hard as any state ever has, directing credit and reading export numbers firm by firm, and still it renewed, because the firm that missed had no one at home to petition. The verdict was a foreign buyer’s, and the buyer never took the call.

An official can be talked into shifting a goalpost. A buyer in another country cannot, because the plea never reaches him.

That criterion is drawn from these four cases, not tested against them yet; a fifth renewer whose government kept the scoring in its own hands, free to ease it for a struggling firm, would break it.

There was a second way the four differed, beneath the tools: how open the deal was. Singapore was open to all comers; Korea reserved access to its champions for a favoured few, then held that few to the same external test.

The tool is the thing most often copied: a country reaching for Korea’s directed credit or Singapore’s open door is reaching for opposite instruments that sat on the same side of the line.^[12]

Subscribe now

This is the ordinary engine of catch-up: exposure to a frontier above you drags you upward, faster than staying sheltered from it ever could, though the specific mechanism here, a competitive test a firm cannot escape, is a narrower descendant of Gerschenkron’s backwardness-as-advantage, not his own claim.

This is also what separates market exposure from much export promotion. A government can run the whole apparatus, export targets, directed credit, monthly performance reviews, and still judge for itself whether the targets were met and forgive the firm that missed them; that is only a deal with itself.

Look for that same exposure among the inertial four, and in every case it is missing, for a different, specific reason each time.

A protector who could be reached

The arbiter finding only holds if its absence is just as visible on the other side of the line, if the four that went inertial can be shown, case by case, to have had no such standard, and for a different, specific reason each time.

In each of the four, the protector was real, and could always be reached. Malaysia’s own HICOM heavy-industry programme, launched in 1980, gave Proton state credit and a protected market, but attached no test it was ever made to pass. Proton kept its tariff walls and its licences regardless, and by the mid-1990s had built no meaningful export presence at all.

Thailand’s Board of Investment took an even lighter-touch approach: a certificate granting tax holidays and import-duty waivers, approved once and never revisited. Over more than a decade and hundreds of certificates issued, not one has been documented as withdrawn for poor performance. That same pattern shows up at the top: Charoen Pokphand, Thailand’s largest conglomerate and family-run since 1921, spent the following decades growing into size and reach across its business, but never into a firm whose own engineering others paid to license.

Brazil ran a version of Korea’s own test, on paper. Fiat’s entry into the protected domestic market carried a real federal export quota, four hundred million dollars over ten years. But its 1973 deal let it fill that quota by pooling exports from its Brazilian tractor and truck operations, so the new car plant never had to meet the target on its own. The number was real; the discipline on the carmaker was not.

Indonesia’s clearest case was never built to face anyone. Suharto’s son was granted exclusive status as the nation’s official “pioneer” maker, with tax and import exemptions. The firm built nothing. It imported finished cars from Korea and sold them, rebadged, as Indonesia’s own national car.

The constant here too was the arbiter’s absence. In every case, success or failure was judged by someone the firm’s own government could still reach and, when it mattered, still capture: an official, a licence board, a price council, a family connection. Call it exposure to a negotiable protector, a test that could always be renewed, rewritten, or waived once it became inconvenient.^[13]

Among these four, the top firms became fixtures, and stayed so for decades. Thailand’s largest private conglomerate has grown for a century largely unchallenged; Indonesia’s tycoons outlasted the ruler who made them; Malaysia’s were minted by privatisation and kept by connection; Brazil’s were financed into national champions by a state development bank. None of them ever had to answer to a buyer who could walk away.^[14]

A firm that owes its position to a negotiable protector has every reason to keep that protector negotiable, and the bigger the firm grows, the more it must spend to keep it that way.

None of this was a failure to climb. Each of the four climbed hard to a comfortable, defensible position, and then settled into holding it rather than pushing past it. So the inertia was never a ceiling; the opening already showed Malaysia stalling where Korea climbed on. It was a peak nothing was pressing them to leave.

Winners who could still lose

On the other side of the line, among the four economies that kept renewing, firm survival meant the opposite. Samsung staked its future on selling memory chips into a world market that owed it nothing, a verdict its own government could not fake and could not simply override, and lived only because foreign buyers kept choosing it. That is the reverse of a protector who can be reached.

A firm like that could not be kept alive from inside; it had to be chosen from outside, again and again, by people free to choose someone else.

This is a different kind of openness from the trade openness already ruled out, not whether foreign firms can get in, but whether a country’s own champions can still be allowed to fail.

The same thing shows at the level of a whole economy, in whether its biggest winners can still lose. Among the renewers, they could. Korea broke up Daewoo, its second-largest conglomerate, in the shakeout after 1997, and let Kia fail.^[15] Taiwan built no equivalent chaebol at all: an economy of small firms, none of them insulated by scale. China’s champions are made and unmade in a domestic competition fierce enough to kill most of the entrants.

China is the exception still in motion: it opened its deal wide, and under a tightening state is now testing whether it will keep that deal open once its own winners would rather it did not.^[16]

This is also more plausibly what produced the differences already observed: a diversified export basket, homegrown firms in the hardest industries, patents relative to population. A firm judged by an independent arbiter has a reason to upgrade, invent, and diversify that a firm judged by a negotiable protector never acquires. The test does not just correlate with those outcomes; it is a plausible source of them.

It may work as much by attrition and compounding as by persuasion.^[17] An external test removes firms that cannot meet it and forces the rest to keep adapting, so capability is less something taught and more something a firm develops under pressure. And what stands compounds: a firm that wins abroad earns the revenue, the exposure, and the standing to win again. A sheltered firm compounds nothing. A small early difference, weeded and compounded for a generation, becomes the distance between a frontier and a plateau.

An unenforced target culls no one, and HICOM had targets too. What the four renewers had, and the four inertial economies lacked, was a bite: a real, live consequence that actually lands when a firm fails, not a box ticked once at approval and never checked again.^[18] Malaysia’s targets went unenforced. Thailand’s certificates asked for nothing to begin with. Korea built the bite in from the start, and Samsung’s survival depended on it.

There is a deeper reason a country can copy the policy and never build the capability. Capability is not installed; it grows. A government can set what its firms must survive, and it cannot hand them what surviving builds. The bite sets the conditions under which capability can accumulate; accumulation is done by firms over years, or not at all. Which is why a country can import Korea’s targets, as Malaysia did, and none of Korea’s capability with them.

All of that explains catch-up; whether the same test keeps biting at the frontier, once the winners it produced would rather it stopped, is a harder question the series takes up later.

Subscribe now

The cleanest split, not yet a cause

This is the opening post of a series asking what separated the renewers from the inertial four. Tested against investment, foreign capital, savings, urbanisation, and political openness, one thing survives where a thirty-year argument between developmental-state and market-friendly economists has not. It is not whether a state directs or a market decides. It is whether a struggling firm can win relief by pleading at home, or must clear a bar set abroad that no minister can lower for it.

Korea authored almost everything about its firms’ lives except the verdict abroad. Call it market exposure, the arbiter, the test, the bite, all the same thing seen at different points.^[19]

This difference does not show up in a cross-country dataset; it comes from each country’s case record. And something that lines up with the outcome need not be what produced it. The cleanest split is a candidate, not yet the cause.

Two questions remain, each for a later post. Does market exposure survive against its own direct rivals: state capacity, the 1997 shock, market access, and currency management? And how did some governments come to put their own firms to a test they could not later soften, while others sheltered theirs for a generation? The first is the next post’s job; the second is where the series turns after.

The external standard comes down to one question that the existing accounts skip: not whether a deal is honoured, but who the firm has to satisfy. Malaysia honoured its deal with Proton for thirty years and never made it compete. What separated the renewers was not whether the deal held, but who set the bar it had to clear: an outside buyer with no reason to be kind, not an official who could be persuaded to move the goalposts.

This is a small turn on a familiar idea. The discipline literature already locates the test in the market’s verdict; the turn is to ask not whether a state will enforce a standard but whether a failing firm can still plead its way out of it. Every government here could have rewritten the rules; what separated the eight was whether the score was kept where no single firm could reach it.

So the next time a government claims credit for a rules-based, consistently enforced industrial deal, the question is not whether the rules were upheld. It is who wrote them, and whether that author could still change the terms once the firm began to struggle.

A deal a state can still rewrite for itself is not a test. It is a promise.

Notes

A word on what this comparison can and cannot do. Setting the two quartets side by side is powerful for crossing candidates off: anything that ran just as high among the inertial four cannot be what divided them. It is far weaker at settling what did the dividing. Several things still separate the two groups throughout, land equality most sharply, and this design cannot adjudicate between them. So the external test is not the last survivor of a clean elimination; it is the candidate that survives elimination and that a case-by-case reading of the record continues to point to. That is a strong lead. It is not a proof, and the next post has to adjudicate; this one cannot.↩︎
Real economies differ in a hundred ways at once, and these two quartets were not identical in every respect but one. Growth sits in a world of high causal density, where many things move together and share drivers: an economy sitting on oil or minerals can bankroll roads, schools, and factories without any firm ever facing a foreign buyer, so several candidates co-move. Land equality is the sharpest of the differences that survive to the end; a government that starts with land already equal can discipline its elites without any foreign test at all. Land in fact enters twice, and the two must be kept apart: as a starting endowment (how equal holdings already were, a rival origin the series tests later) and as land reform, a deliberate policy discipline in Studwell’s account, Korea’s, Taiwan’s and China’s post-war redistribution seeding the smallholder surplus that manufacturing grew on. Note too that the inertial four’s largest firms are the state oil and gas companies named earlier: the same resource rival seen from the firm’s side, a government sitting on rents can bankroll its economy without any firm ever having to win a foreign buyer. This is why the question has been argued for thirty years without agreement, and why there is some value in an explanation simple enough to say why so many candidate causes move together in the first place.↩︎
Krugman and Young argued that East Asian growth was mostly driven by factor accumulation rather than productivity, and warned that it might exhaust itself as the Soviet Union’s had. They were comparing East Asia to the rest of the world; the inversion that matters here shows up only when the same comparison is run within the miracle set, as the body does.↩︎
Both traditions treat the choice between an active state and an open market as the explanation for who rose and who did not. The most forceful recent statement of that market-friendly tradition is Arvind Panagariya's Free Trade and Prosperity (2019), which credits openness rather than industrial policy for the East Asian miracle and turns the revisionists' own Korea evidence against them: growth fell during the interventionist heavy-industry drive and recovered after it was abandoned, and Korea was already highly open before the drive began. His case against the picking-winners reading reaches the same place this section does from the opposite direction — the state-versus-market axis does not explain who rose. The claim here is narrower: not that either side is right, but that the axis they share does not sort these eight, since a directed economy (Korea) and an open one (Singapore) both transformed.↩︎
This lands inside a real, unsettled fight in political economy. Some scholars find no link between democracy and growth; others find a positive one. Nothing here resolves that debate; it just shows regime type doesn’t sort these eight. The thing that does sort them is still a capture question, only relocated: the institutionalist tradition asks whether whoever holds political power can capture the rules that bind them; this post asks whether whoever holds power can capture the standard a firm must clear. Same structural concern, different arena. Whether that economic non-capture is sustainable under concentrated power over generations, the Singapore and China question, is left open here.↩︎
This mechanism isn’t new on its own. It is close to Amsden’s reciprocal control and to the deals-and-development literature’s idea of an honoured deal. Amsden’s Korea/Latin America contrast already turns on whether a state is autonomous enough from its own champions to make them actually lose when they fail, a capture point close to the one drawn here. The step beyond her is small but real: even a state autonomous enough to enforce its own metric can quietly soften it, because it still owns it. What these eight cases add is the finding that this axis, not the state-versus-market choice that’s occupied both traditions for thirty years, is the one that actually separates a good grower from a renewer. The mechanism has since received independent peer-reviewed confirmation. Amodio, Choi, Poschke and Sposi (2026), reviewing Korea’s Heavy and Chemical Industry Drive through firm-level data, identify export performance as the criterion that makes public firm support “less likely to be distorted by domestic factors such as lobbying or market power” — a formulation that converges on this post’s external arbiter from a different empirical route.↩︎
The four-firm shorthand behind this contrast, magicians, rentiers, powerbrokers and workhorses, is Eric Werker’s four-markets typology (HBS Technical Note 713-026, 2012), taken up in Pritchett, Sen and Werker’s Deals and Development. It sorts firms by where profit comes from, competition or rents, and which market they serve, export or domestic: magicians win in export markets on productivity, rentiers export resource rents, powerbrokers milk protected home markets, workhorses are ordinary domestic competitors. The renewers’ champions are magicians; the inertial four’s flagships are rentiers, the state oil companies, and powerbrokers, the protected national champions like Proton.↩︎
China is the shakiest of the four renewer cases on this measure. Its exposure came later, narrower, and less evenly than Taiwan’s, Korea’s, or Singapore’s, and its patents stayed indistinguishable from the inertial four until roughly the turn of the century. That it belongs in this group is largely inferred from where it ended up, not from an independent reading of how exposed its firms actually were in 1980. This is the case the next posts treat as still being tested, not yet settled. The processing trade data clarify the mechanism behind the patent lag. At its peak in the 1990s and 2000s, processing trade — importing inputs duty-free, assembling for re-export — accounted for more than half of China’s exports. But processing firms were typically less productive and captured less value than ordinary exporters; the capability accumulated through the transition out of processing trade, not through processing itself. China’s patents tracked this: indistinguishable from the inertial four while the economy ran on assembly, beginning to separate only as ordinary exporters displaced processing firms. The exposure ran through the special zones and the enterprises outside the plan; it ran on assembly first and on capability only later.↩︎
The deals-and-development literature’s own “ordered” category means only that a deal is honoured, not reneged on. This post’s cut is narrower: it’s about who sets the standard a firm must meet, not just whether the deal is kept.↩︎
An external verdict is not physically impossible to override. A government can subsidise or bail out a failing exporter, as Korea sometimes did in the 1970s. The difference is that the verdict has already been returned by someone the state does not control, so overriding it means visibly paying to keep alive a firm the market has already rejected, and that bill grows with the firm. A negotiable protector never issues a verdict at all; there is nothing to overturn, only a target to quietly move. Why some governments were willing to pay the first bill, and others reached for the second, is the question the later posts take up.↩︎
This is closest to what Joe Studwell calls export discipline, which already locates the test in the world market’s verdict rather than in the state’s intentions. The one thing these cases add is Singapore: discipline is upheld there with almost no state-imposed export requirements to enforce, which pulls the two ideas apart. What matters is not that a firm was made to export but that it was judged by someone its government could not reach, exporting being the commonest way to arrange that, not the thing itself.↩︎
This is Andrews, Pritchett and Woolcock’s form-versus-function point about copying institutions, applied to industrial policy. A government can import the instruments, the targets, the directed credit, the review meetings, and none of the arbiter that gave them their bite, because the bite lived in who returned the verdict, not in the apparatus. The instrument is the form; the external test is the function.↩︎
A reader who works with Mushtaq Khan’s political-settlements framework will recognise these inertial cases as familiar ground: a government fails to enforce a target not for want of the right kind of standard but because the firms sit inside the ruling coalition’s own distribution of power, and it has neither the holding power nor the incentive to make them lose. That is a serious rival to the reading offered here, and it is not settled in this post. Whether “who sets the standard” and “who holds the power to enforce it” are two names for one thing, or two things that happened to move together across these eight cases, is exactly the question the later post on why some governments could impose the test has to answer. The Korean record holds both things at once. Kim, Lee and Shin (2021), exploiting plant-level data from the HCI Drive, find lasting gains in targeted region-sectors alongside disproportionate resource flows to large plants owned by business groups — the firms politically positioned to capture the policy rather than simply earn it. ↩︎
One honest caveat: the clearest renewer failures, Daewoo and Kia, cluster right around the 1997 crisis, so contestability and crisis-timing are tangled together here. The next post argues the same shock produced opposite responses in the two groups, which is a start on untangling it, not a full answer yet.↩︎
Daewoo’s collapse and Kia’s receivership both sit inside the 1997 Asian financial crisis, which cost plenty of Korean workers their jobs and savings well beyond these two firms. The point isn’t that failure arrived cheaply or was well managed, only that these economies kept a standing threat of failure hanging over their biggest firms. The crisis itself is a rival explanation the next post takes up directly. A deliberate limit alongside this: the comparison asks only what separated the growth outcomes and brackets who bore the costs and how the gains were shared. Singapore held wages down by stripping unions of independent bargaining. That the external test sorted the outcomes says nothing about whether it sorted them justly; the distributional ledger is real, it is simply not what this comparison measures.↩︎
The honest test of this criterion is not a hypothetical fifth country. It is China, here, now. China renewed, but its government retains more power to rewrite the terms its own champions face than Korea’s or Taiwan’s did once their firms were selling into world markets, and on patents it sat indistinguishable from the inertial group until recent memory. Whether Beijing keeps binding its winners to a standard it could lift is exactly the question the criterion turns on. That the clearest live test of the idea sits inside the sample, still unresolved, is a feature of an honest reading, not a hole in it. But it does mean China is evidence still being taken, not evidence already in.↩︎
There’s a live debate in trade economics about whether exporting makes firms better, or whether already-better firms are the ones that choose to export. The same question applies here at country scale, and eight cases can’t settle it: renewer states might just have exposed firms that were already positioned to survive exposure. What the eight cases do show is that exposure and capability outcomes showed up together in a pattern the usual rivals can’t explain. Proving exposure did the work, rather than just rode along beside whatever did, needs a finer instrument than a comparison this size can supply. Recent firm-level evidence offers a partial resolution. Amodio, Choi, Poschke and Sposi (2026) find the two processes are complementary rather than sequential: productivity gains from new technologies raise a firm’s probability of winning foreign buyers, and participation in export markets strengthens the incentive to pursue further technological improvements. The reinforcement runs in both directions. That finding does not settle the selection question at country scale, but it closes the gap between “firms get better because they export” and “better firms happen to export” — both mechanisms were active, and each fed the other.↩︎
Dani Rodrik has long argued that industrial policy works not when a state picks winners but when it is willing to let losers go: support has to carry a real penalty for failure, not become an unconditional entitlement. That is the same discipline described here as the bite. What these eight cases add is a claim about where the penalty has to come from. Rodrik’s disciplining state sets and enforces the metric itself; the finding here is that a metric a government both sets and judges stays soft, because the government can always move it, so the cases that renewed are the ones where the standard was set outside the state’s reach, by a buyer or a world price it could not rewrite. A fair objection runs alongside: isn’t an externally-judged test just well-designed industrial policy? Largely yes, and that is the point. Korea’s directed credit sorted firms because it was tied to export orders no ministry controlled; Malaysia’s, attached to targets the ministry could waive, sorted no one. Same tool, opposite result.↩︎
Chalmers Johnson’s plan-rational/market-rational distinction and Peter Evans’s embedded autonomy are earlier statements of the general intuition that what a firm must answer to, more than the stock of policies around it, shapes whether it improves. Evans’s disciplining relationship sits inside the state-firm nexus, a negotiated bond the post’s external arbiter deliberately sits outside; the nearer predecessors of that specific cut are Amsden, Studwell and Rodrik. What is specific here is a testable version: exposure to a standard set outside the state’s reach, checked against a named set of cases rather than asserted in general (both authors are in Further Reading below).↩︎

The strongest example is ST Engineering: majority Temasek-owned, genuinely inventive, S$420m of R&D in FY2024 at 3.7% of revenue, with its own products in satellite ground systems and video analytics.

What I’ve been reading

Stephen Brien — Sat, 11 Jul 2026 08:22:40 GMT

STATE CAPACITY

Three pieces sharing a common question: what actually produces state capacity, as distinct from the institutional forms that are supposed to carry it? Henn and Robinson show that the prior condition for Africa’s state-building deficit was deliberate preference, not inherited incapacity — which changes what solving it would require. Opalo works forward from that diagnosis toward a practical prescription. Liao provides the sharpest available image of what organisational connective tissue looks like when it actually works.

Africa as a success story: Political organisation in pre-colonial Africa Soeren Henn & James Robinson — VoxDev / NBER Working Paper 34546 — 24 February 2026

Africa in 1884 had approximately 45,000 independent political communities against roughly 50 in Europe, not a gap in capability but a deliberate choice. Central Africans organised political life around “wealth in people” rather than territory, keeping scale small to protect community autonomy, and on their own terms this worked. The problem is that small-scale organisation optimised for internal autonomy was systematically exposed to external conquest. It was larger polities, China, Japan, Thailand, that resisted colonialism. The implication for state-building is that the prior condition blocking consolidation was rational preference, not incapacity, which is a harder thing to change.

The mechanics of building strong states in 21st century Africa — An Africanist Perspective — 3 July 2026

Opalo’s organising claim is that Africa is under-governed rather than over-governed, and that most state-building prescriptions are calibrated for the opposite problem. His prescription: form must follow realistic functions. Build on what the state can actually deliver, starting with subnational self-government and local service delivery, rather than copying institutional structures from states with different capability profiles. Brazil’s 5,570 functioning municipalities against Nigeria’s largely hollow local government areas is the comparison. Whether this resolves or simply relocates the preference for community-scale autonomy that Henn and Robinson identify depends on whether those subnational units eventually build fiscal linkages upward.

The Organizational Party: The Real Source of China’s State Capacity — leonliao.substack.com — 4 July 2026

Liao’s argument is that China’s capacity derives not from bureaucratic structures or market design but from the Communist Party’s organisational depth — its ability to link people, resources, discipline, information, promotion, and tasks into a single system reaching from the centre to the village. “China’s market economy has never been only a market economy. It is a market economy deeply shaped by political organisation.” The reform era created “organised market competition” in which local governments competed for growth within party-set frameworks of cadre evaluation — liberalisation structured by political organisation, not instead of it. The piece identifies what Africa’s preference for small-scale autonomy has systematically failed to build: the connective tissue that enables implementation at scale.

DEVOLUTION

Two pieces on the same reform mechanism producing opposite outcomes. The Brazil paper finds that municipal splits in the 1990s delivered real development gains to peripheral communities with no visible cost to the rest. Pakistan’s 18th Amendment, by contrast, devolved power from the centre but stopped at the provincial level, creating a new tier of patronage rather than reaching the citizen. The variable between them is not whether power moved downward but how far, and whether the institutions it landed in had genuine authority over the services people depend on and genuine accountability to the people receiving them.

Decentralising development: Economic effects of government splits Ricardo Dahis & Danielle Szerman — VoxDev — June 2026

Between 1988 and 1997, Brazil went from 4,124 to 5,507 municipalities; 39% of eligible districts applied to split. Peripheral areas that broke away saw nighttime luminosity increase by roughly 40% over fifteen years; the parent headquarters gained modestly; the remaining districts showed essentially no change. The mechanism was not money alone; federal transfers rose roughly 15%, but it was decision-making autonomy over services previously controlled at a distance. The gains came from matching the scale of governance to the scope of the problem, eliminating capture, and creating accountability to citizens who had previously been peripheral.

Pakistan’s Half-Devolution — macropakistani.substack.com — 4 July 2026

Pakistan’s 18th Amendment transferred power from Islamabad to the provinces, but local government’s share of total public spending fell from 10% in 2005 to 5% by 2024. Provinces absorbed the new resources — 82% of incremental transfers went to current expenditure — and became intermediate patronage centres; Punjab has not held local elections since 2015. Where Brazil’s splits created genuine decision-making authority at community level, Pakistan’s devolution created a new tier of resource capture between the centre and the citizen. The constitutional form changed; the function did not.

SOUTH AFRICA

Five pieces that converge on a single structural diagnosis: the GNU has the form of shared, depoliticised governance, but the informal architecture hasn’t changed. The ANC study group piece shows that party control of the public service persists beneath the coalition’s surface. The Khaas piece on the Madlanga Commission shows what capable institutions can do, and how exceptional their doing it remains. Havemann shows the strategic and fiscal vacuum at the centre. Axelson et al. confirm that the fiscal experiments have been backfiring. And the comparator countries piece maps the distance between where South Africa is and what the institutional architecture for escape from the middle-income trap actually looks like.

Party before country: How secretive ANC study groups undermine GNU’s work Daily Maverick — dailymaverick.co.za — April 2026

Before departments appear in front of parliamentary portfolio committees, senior officials brief ANC MPs separately in party study groups; by the time the formal committee sits, the deliberations have taken place, the reports have been approved, and the outcome, as Deputy Minister Holomisa put it, is “largely a foregone conclusion.” This predates the GNU and spans portfolios — Basic Education, Defence, and Human Settlements are all cited. The GNU’s formal architecture presupposes a depoliticised public service neutral between coalition partners; the informal architecture has not changed. The GNU has the form of a shared executive; the function remains party-controlled.

Twelve months that changed the Republic — khaas.substack.com — 5 July 2026

One year of the Madlanga Commission and the contrast with Parliament’s parallel ad hoc committee is the real subject. The Commission was established to investigate criminality and institutional capture in the criminal justice system. It operated with rigour and genuine consequence: careers ended, powerful figures exposed. The parliamentary committee produced procedural disorder, political grandstanding, and shielded connected witnesses. Both are formally accountable to the same constitutional order; one worked, one did not. South Africa contains genuine islands of institutional capability — the Commission, SARS at its best — but no governing orientation that could make those capabilities systemic.

How can South Africa decisively turn the corner? — havemania.substack.com — March 2026

Three growth conferences in March 2026 — ERSA, News24 On the Record, SA-TIED — produced the same finding from different audiences: the GNU has no coherent economic strategy. Raising the top income tax rate from 41% to 45% lost the fiscus R6.5 billion; the affected group’s share of personal income tax fell from 11.5% to 10.6%. Attempting to regulate 13,000 spaza shops yielded 200 registrations. These are not random errors but errors in the same direction: interventions that look like governing but subtract from the conditions for growth. Nobody in the coalition owns an agenda willing to defend specific trade-offs.

Laffer curve confirmed for South Africa’s top earners @JustinSandefur flagging Axelson et al., forthcoming AEJ — 8 July 2026

The Axelson et al. paper confirms what Havemann’s R6.5 billion referenced: the 2017 increase from 41% to 45% placed South Africa on the wrong side of the Laffer curve. The elasticity of taxable income among top earners is estimated well above 1 (against an international consensus of 0.4), meaning the income effect outweighed the rate increase and revenue fell. Sandefur’s nuance is worth noting: avoidance and income restructuring are the more likely mechanisms, not labour supply reduction, which points to base broadening rather than rate cuts as the relevant lever. The directional finding stands regardless, and for a state with the GNU’s fiscal constraints, it is not merely an academic result.

The Comparator Countries — Economic Growth and Development Series 1.3 (GovCompass) — nexuspolitics.substack.com — 9 July 2026

Six countries at roughly South Africa’s current stage of development — Chile, Vietnam, Malaysia, South Korea, Botswana, and Poland — and what they did to escape the middle-income trap. South Korea’s R&D spending exceeds 4.5% of GDP, compared with South Africa’s 0.8%. Malaysia’s Economic Planning Unit produced 11 five-year plans with genuine implementation authority; South Africa’s National Development Plan assumed capacity the state lacked. Botswana required mineral revenues to fund investment rather than consumption; South Africa has no sovereign wealth fund. The comparator countries did not succeed because they found better diagnostic frameworks. They built their versions of the organisational connective tissue that Liao identifies in China. And that is what South Africa’s NDP assumed it could bypass.

Why a Win Is Harder to Seal Than a Defeat

Stephen Brien — Thu, 09 Jul 2026 07:30:31 GMT

When a government wins a political fight, the country can go on behaving as though the wall is still standing. A private win is only a rumour.
When a government loses a fight, the wall holds for free: everyone was already walking around it.
When securing change, nothing is free: the wall has moved, but everyone else is still walking around where it used to be.
For a political win to stick, three things have to happen: a real test, common knowledge of the result, and an independent scorekeeper to confirm it.
Most transparency reforms only give governments the first of these. That is why so much disclosure changes so little.

This is the fourth post in a short series on how a country’s working rules change — not the statute book, but the coordinated expectation about who holds what. The first post established what that expectation is: a wall that holds because everyone assumes everyone else is still walking around it. The second question asked what happens when someone tests it — Edward Heath confronted the miners in 1974 and found out the hard way. The third was about being serious enough to hold the test — Thatcher spent years stockpiling coal before she faced the NUM in 1984. This post is about what comes after a win, and the step almost every capable government skips.

In Rwanda, every district mayor signs a public contract once a year. Two very different kinds of numbers come out of it. The farm figures were fiction. The health figures told the opposite story.

The contract lists what the district will achieve over the coming twelve months: clinics opened, fields planted, roads graded. At year’s end, the results are scored, and the districts are ranked before an audience. The system is called imihigo, after an old word for a vow made aloud, and it is the same ceremony for every target, run by the same officials in the same districts.

When independent researchers checked the harvest claims against a proper field survey, the reported food output for the main crops ran 192 per cent above what the survey found, and for wheat nearly nine times above.^[1] Rwanda’s fall in child deaths over the same years was real, and it is among the best documented anywhere in Africa.

Same government, same ceremony, same officials. What separated the honest column from the invented one was not effort, and it was not sincerity. It was who did the counting. The health gains were measured by international household surveys that the government did not run and could not quietly revise. The harvests were reported up through the same chain of officials whose own performance was being judged on them.

It is also where the usual remedy for crooked government fails. That remedy is more transparency, more information in citizens’ hands.

But the farm figures did not fail for want of information; they failed at the source. The fault was never how much people knew. It was who they had to believe. A finding that runs up through the people being graded cannot carry the whole country’s beliefs, not because it is necessarily wrong, but because everyone knows it could have been adjusted on the way.

The deal nobody writes down

The explorer, the political actor willing to test whether the wall is real, found the moment and paid the price. The wall gave.

A settlement, the real rules, as opposed to the written ones, does not run on the true balance of power.^[2] It runs on what each player believes the others will do. Those beliefs are almost never tested; they can stand for a generation with no one finding out whether they are real. A standing deal is held up by nothing more solid than coordinated expectation, which is why the hard part of changing one is never the changing. It is the belief.

Legislation is itself a form of common knowledge: a statute, once passed, is published, on the record, citable by anyone. But a country’s deepest rules are not the ones in the statute book. They are the standing deal among the people who actually hold power: who runs the courts, who keeps the books, who certifies who won, and what it costs anyone who tries to take a role that isn’t theirs. A law can restate that deal, or it can misstate it — and only the second case needs anything sealed at all.

For years, British governments kept the coal miners at bay with a quiet back-room arrangement — neither testing the line nor displaying it, each side behaving as though the deal held because everyone expected the others to.

The explorer’s test changes the real balance. It does not move the beliefs of everyone who wasn’t in the room when the line gave way.

That victory, by itself, changes nothing.

A private win is only a rumour

The win is real. The balance of power has genuinely shifted, and the people who lost know they lost.

But almost nobody else does, and that is what matters, because a settlement never runs on the true balance of power. It runs on what each player believes the others will do, and on what each believes the others believe. A private win moves the real balance and leaves those beliefs untouched.

The next morning, every other player still expects the old line to hold, still expects the others to expect it, and so still behaves as though nothing has moved. The old line is gone, and the country keeps walking around where it used to be. Each player keeps to the old path not because they still believe the old line holds, but because they believe everyone else does. And acting alone on a doubt, before you can see that others have moved, is a gamble you lose by yourself.

That kind of expectation is self-confirming. It holds not because everyone still believes the old line is real, but because each person’s next move depends on what they think everyone else expects.

That stickiness reaches even genuinely good work. Indonesia’s economic technocrats, the group nicknamed the Berkeley Mafia, won real fights for sound macroeconomic management across three decades under Suharto, from inside the machinery of the state. Ireland’s state investment agency (IDA Ireland) built and then defended one of the most effective investment-promotion operations anywhere. Both were genuine victories, won against real resistance and, by the previous post’s measure, real enough to prove direction. Neither reset the wider settlement around it.

What those wins lacked was not merit but visibility. They may never have carried stakes large enough to shift how the whole country expected things to work, and we cannot rerun them in public to know. What we can say is that they achieved what they achieved without ever making it known, and a win nobody outside the room can see is, to the rest of the country, a rumour at best.

Winning the fight was never the hard part. Making the win known is.

Subscribe now

Telling people is not enough

The largest experiment ever run on this question says it is not that simple, and the way it fails is the clue to how it works.

A consortium of researchers ran six coordinated trials across five countries (Benin, Brazil, Burkina Faso, Mexico and Uganda) — each designed and publicly logged before any data were collected, a method that rules out selecting results after the fact. They handed voters solid, benchmarked information about how their politicians had actually performed. When the results were combined across all six trials, the effect on how people voted was almost exactly zero. Honest information, delivered straight to the people who could act on it, moved nothing on average.^[3]

The same programme then tried a variation. In three of the studies, the information was delivered not quietly to each voter but in a public setting, where each person could see that their neighbours were receiving it too. There, the effect was large.^[4]

The information was identical in both versions. The public setting changed not just what each voter knew but what each believed the others now knew too. The experiment cannot fully separate that from the simpler effect of being seen to receive it. A private fact, however true, leaves you unsure whether anyone else will act on it, and so you sit still. A public one removes that obstacle.

There is a precise name for this: common knowledge, the state in which everyone knows something, and everyone knows that everyone knows it, and so on up. The extra layer matters because each player’s decision to act depends not just on what they know but on what they expect others to do. And that expectation only changes when they can see that the others have received the same information. That gap between information and something close to common knowledge is the line this whole argument runs along.^[5]

The win has to be sealed in public.

The staged event

The first way to push a result from private fact to common knowledge is the staged event.^[6] The clearest modern case changed no law at all.

By the start of the 1990s, China’s reform drive had lost momentum, and the men who wanted it stopped were gaining ground. In early 1992, Deng Xiaoping, by then retired from formal office, travelled south to Shenzhen and the other coastal zones and said, in public and over and over, that economic reform would go on, and that anyone who stood against it would have to step aside. For several weeks, the central newspapers said nothing, which meant each official could not be certain that the others had heard it, even if they had, because hearing it in the southern press was not yet something everyone could see everyone else had done.

The contest was not about policy; the policy was a decade old. It was about whether his words would be allowed to become common knowledge again. Once the southern press carried them and the Party Congress later that year wrote his line back into doctrine, the matter was settled, not because a rule had changed but because every official now knew that every other official had heard it.

The tour was theatre in the exact sense that counts: a way of making one fact known to everyone at once, and of remaking that common knowledge when it had begun to slip.

The staged event need not even be a confrontation. At the 1995 rugby World Cup final in Johannesburg, Nelson Mandela walked onto the pitch wearing the green Springbok jersey, the shirt that white South Africa had worn through the apartheid years and black South Africa had loathed. The gesture changed no statute and settled no policy. What it moved was what each side believed the other would now accept. A nationwide broadcast tells you what happened; a stadium lets each person see that the others have seen it too, which is what changes what you expect they will now accept.

A reconciliation that until then had been negotiated in rooms became, in a single afternoon, something the whole country knew it shared.^[7]

Not all staged events work this way. In 2016, the Indian government cancelled 86 per cent of all the cash in circulation in a single prime-ministerial broadcast (a larger stage than anything in this section), and the settlement did not move.

The standing scoreboard

Call it a scoreboard, a public, hard-to-fake measure that everyone can watch as it moves.

The bond market functions as a single market, with no institution to run it. A government’s borrowing cost — the interest rate it must pay to raise money — moves in public every day, in response to what each lender believes the others believe about whether the rules will hold, and no government can quietly revise that rate on its own debt.

When Brazil’s federal government began auditing municipalities at random and releasing the results before elections, corrupt mayors lost re-election far more often, and the effect was sharpest in towns that had a local radio station to carry the findings. The audit was the test. That radio was what turned a submitted report into something the whole town knew it shared. The two are different things, and both are required.^[8]

The World Bank’s Doing Business rankings were supposed to be exactly this: a hard-to-fake measure run by an institution with no stake in any country’s result. They were also edited.

Subscribe now

When the whole deal is rewritten

In 1688, the English Parliament, backed by a Dutch invasion, deposed James II and replaced the old settlement outright. The new rules were written down in public the next year, in the Bill of Rights: the Crown could no longer suspend laws at will, keep a standing army in peacetime, or raise taxes without Parliament’s consent. This was not one official being restrained. It was the whole standing deal between Crown and Parliament torn up and rewritten, in statute, where anyone could read it.

This was reform of the constitutional order, not reform within it: the deposition broke the rules that existed in order to install rules built to bind harder afterward. And it held, because everyone could see that it held, in two ways at once. One influential argument reads the same public statute as the standing scoreboard described in the previous section: the Crown’s borrowing costs fell as lenders came to believe a king checked by Parliament would not seize their money or tear up his debts.

How much of England’s later credit boom that settlement explains is itself disputed; the public rewriting of the deal carries the point on its own.^[9]The other mechanism needed no market at all. Parliament sitting on a fixed schedule and courts ruling in public are built to keep re-certifying a settlement on a standing basis, a recorded vote or a published judgment doing the work without anyone needing to stage anything or hold a stake in the outcome.

Deng’s tour, Mandela’s gesture, and 1688 did not hold the same way, and the difference is the whole lesson of how to make a win last. Deng’s tour had to be remade the moment the common knowledge began to slip; he, and then his successors, had to go out and re-perform it by hand.

1688 needed no one to re-perform anything: the creditors confirmed it for free every morning in the price of the King’s debt, because their own money now rode on the new limits holding, and Parliament and the courts confirmed it every time they sat, because that was simply what they were for. A staged event has to be re-performed; a scorekeeper that runs on its own interest or its own design re-performs itself. That built-in permanence is also what makes a constitutional order, of everything in this post, the single most valuable machinery to capture.

So the way to make a win permanent isn’t to seal it more tightly. It is to seal it to a keeper who profits from the new line, so that holding the rule up becomes someone’s daily self-interest rather than the reformer’s standing chore. If the only person who can re-perform the seal is the reformer, they will be re-performing it forever, as Deng was.

Even so, by hand or by interest, a win has to be held up somehow, where a defeat needs nothing. That is the puzzle. Why should a victory that has already been won need any holding up at all, when a defeat needs none?

What would refute this

The grand cases here, Deng and Mandela and 1688, are illustrations, not experiments. We cannot rerun them with the publicity removed to see what would have happened, and a vivid story can always be told two ways. The one piece of genuinely controlled evidence in this post, the public arm of the voter-information trials, rests on a large effect from a single study, making it the more fragile half of an otherwise clean finding. So treat what follows not as a proven law but as the simplest account that fits what these cases share, an account that still has to survive its own thin spots.

The controlled evidence, the voter-information trials, reaches only part of the claim that a settlement moves when a real result is made common knowledge. The trials vary whether a result is made public, and the Rwandan record varies who keeps the score, but whether a test is real is determined by illustration, not by an experiment.

Below that sits a scope question. The controlled result concerns how people voted in five countries, while the claim here concerns how a country’s powerful behave; the elite cases extend the mechanism into that arena rather than proving it there.

But a prior condition applies: whether a result can be made common knowledge at all depends on whether the country has the wiring to carry it. The Brazil effect was sharpest where a local radio station could carry the audit findings, and the public-delivery result needed a room where neighbours could see one another receive the news. What that wiring does not decide is which way the result falls. That is still the reformer’s work, not the radio’s.

What would sink the account? One that cannot fail is not worth much. This one fails if a real result is made genuinely common knowledge, proved, public, and certified by a scorekeeper the players cannot lean on, and the settlement still does not move.

Three things decide whether the seal — the act of making a win common knowledge — works, and all three are observable before the settlement moves: was the test real, was the scorekeeper independent of the scored, and which way did the result fall. None of these is hidden. The first condition is the ground the first post in this series covered: whether the wall was real in the first place, or only a guess no one had tested. A case that clears all three and still does not move is the verdict that would refute the account; the cases that follow have not yet had to face one.

A stage with nothing on it

A stage works only when there is a real result beneath it to make known.

In November 2016, the Indian government, without warning, cancelled banknotes worth about 86 per cent of all the cash in circulation, in a single televised address. The drama was total, and the audience was an entire nation.

But the move tested no organised holder of power. It was aimed at an abstraction, “black money,” and almost all of the cancelled notes came back through the banks.^[10] The spectacle was real, and the result was nothing, so there was nothing to seal, and the settlement did not so much as flinch.

The test is the first half of the act, the ground the first post in this series covered. Publicity is the second. Without a real test underneath, it is a press conference about nothing.

The catch: who keeps the score

A scoreboard is worth only as much as the independence of whoever keeps it.

That is precisely what the two imihigo columns showed. The farm numbers were reported up by the people being graded on them, so they could not be trusted, and they were not. The health gains were confirmed by household surveys run from outside that chain, so they could be, and they were. A scoreboard kept by the players is meant to produce common knowledge of nothing except the score those players wanted.

That same failure happens at the grandest scale. For years, the World Bank ranked countries by the ease of doing business, and governments revised their laws to climb the rankings. Then an investigation found that senior Bank figures had altered the underlying data, improving China’s position in one edition under pressure tied to a fundraising drive. In Rwanda, the score was bent by the people being graded; here, it was bent from inside the body that kept it. Either way, the most authoritative scoreboard in the field turned out to be editable by an interested party, and the Bank shut it down.

The Doing Business collapse shows what an independent scorekeeper actually is.

A trusted scorekeeper is not a piece of administrative plumbing you can simply install.

Its authority is itself something the players have agreed to believe — a smaller settlement held in place by the same logic as the big one — and, like any such arrangement, it can come apart. The Doing Business collapse was exactly that: a shared belief that a scorekeeper was coming undone in public.

The more authoritative a scoreboard becomes, the stronger the pull on anyone positioned to reshape it, so the credibility that makes it useful is precisely what makes it worth capturing.

The trust in the scorekeeper works the same way. People believed the Bank’s ranking only as long as everyone could see everyone else believing it; once everyone could see that belief failing, it was worth nothing. A scorekeeper’s independence has to be common knowledge, just as a result does. Whether a society can sustain a scorekeeper that the powerful cannot lean on depends on the same power balance as the big settlement it sits beneath. That is what makes sustaining an independent scorekeeper hard; it is not what makes it wrong.

Subscribe now

When the seal is stamped the other way

The seal records whatever the test found, whichever way it fell.

In 2020, the Indian government passed three laws to deregulate agricultural markets, taking on the entrenched network of state-regulated trading yards and the farmers tied to them. The protests were vast and public, and ran for more than a year.

The government then repealed the laws.^[11] The reversal was as visible as the laws had been, and that visibility did its own work. That reversal made the opposite result common knowledge. Every future Indian government now knows, and knows that the others know, that this particular line held. The settlement moved from untested to settled, from reputation to record, only the record ran the other way. The test was run in the open, and the settlement won, and the public defeat sealed the line as firmly as a victory would have, in the settlement’s favour rather than the reformer’s.

Publicity is a tool, not a force. Publicity works in the direction the crowd already leans.

Why is a win so much harder to seal than a defeat

A settlement’s rule moves — the powerful actually change how they behave, not merely what they privately know — only when the result of a real test is converted into common knowledge by a scorekeeper the players trust. The test must be real, not a spectacle over nothing; common knowledge, not private information; and the scorekeeper independent of the people being scored.

Coordinated expectations are self-confirming. A belief a whole society has coordinated on for years is a default everyone snaps back to, and a single contrary event does not dislodge it, because each person needs a reason to think everyone else has moved too.

Sealing a defeat is easy, and it lasts, because a defeat confirms the expectation that the players were already coordinated on. No new coordination is required, only the continuation of what each player was already doing. Nobody has to be persuaded that the wall is real; they were already behaving as if it were. That is why the Indian government’s reversal was sealed instantly and needs no maintenance.

This is the normal case: the players’ standing expectation already favoured the old line.

The reforming government may itself have expected to win, but what the public defeat confirms is the expectation everyone else was already coordinated on.

Sealing a win is the opposite. A victory has to hold a brand-new expectation up against the full weight of the old one, which every player has spent years arranging their behaviour around. The old expectation does not vanish when the wall falls; it has gravity, and it pulls. That is why Deng’s southern message had to be remade the moment it slipped, and why a reforming win needs a staged event or a standing scoreboard to hold it, where a defeat needs neither.

The asymmetry is not obvious, which is exactly why the step gets skipped.

Entrenchment is structurally cheaper than transformation. The machinery a defeat gets for free, a win has to build and then keep running.

Formal modelling shows how one prominent figure’s visible move can reset what a community expects of itself,^[12] and how new expectations can erode back toward the old norm unless something sustains them.

What this post adds is the asymmetry itself and the design rule that follows: a win held by a keeper who profits from the new line holds, because the terrain now rewards defending it. One that only the reformer can re-perform by hand fades the moment he stops.

The need for a scorekeeper that the players cannot lean on is older, and so is the idea that a self-interested party can hold a commitment in place. What is new is putting them together. The staged seal and the standing one are the same move made two ways: one dear and fragile, one cheap and lasting. The formal model, in any case, treats the scorekeeper as a feature you install rather than, as the Doing Business collapse showed, a smaller settlement that can come apart.

This single asymmetry is what was behind three things that looked separate. It is why a staged event and a scoreboard are needed at all, and why both have to repeat. It is also why the seal stamps a defeat so much more readily than a victory, and why the step gets skipped.

The quiet competent state’s mistake

Someone has to doubt that the wall is real. Someone has to pay to test it. And the result has to be made common knowledge, certified by a scorekeeper that the rest of the players actually believe.

The third step is the one a capable, low-key government skips most reliably, and the asymmetry explains why this is structural rather than mere oversight. An agency that wins inside its own walls is fully rewarded there. The discipline works, the technocrats are respected, and the job by its own lights is done.

Carrying that win out into the open, holding it up against the whole country’s standing expectation and paying to make it common knowledge, is the second and far costlier climb, and the success of the first is exactly what makes the second look unnecessary.

The adversary with no address

But why should it look unnecessary when its whole point is that the work is not yet done? The answer is the property that sets the seal apart from every move before it. Every other fight in this story is against something you can see. The interest that holds the old line, the rival who would block you, the union that can shut down the country: these have faces and addresses, and you know when you have beaten them.

The seal is the one move whose adversary cannot be located. What it has to overcome is not a person or a faction but a state of belief, the standing expectation spread across everyone who was not in the room, held by people who may not know they hold it, and held only because each expects all the others to. There is no door to kick down, no signature to extract. The thing the seal must defeat is everywhere and nowhere at once.

So a capable government skips the seal and feels it prudent to do so. It has just beaten a real opponent it could see, and the books are closed. The seal asks it to go on spending, to stage the event or build the scoreboard, against an enemy it cannot point to, for a victory it cannot photograph. To a results-minded official, that is money spent on nothing, and the better the agency, the faster the cut, because a disciplined state is the one with the least patience for outlays that show no return.

The invisibility is not a detail. It is the same property that makes a settlement a settlement — a belief no one can see is the hardest to move, so the seal is most costly to make for the same reason the old expectation was so strong to begin with.

So the quiet state settles at a stable optimum at the wrong scale. A win that the country never sees leaves the old expectation standing at full strength, because nothing was ever put up against it.

The practitioner’s lesson here runs counter to standard advice. The popular prescription for cleaner government is to disclose more, publish more, and put information in citizens’ hands. The mainstream of that research on transparency targets the volume of disclosure; it has paid less attention to who certifies it.

The voter-information trials show that information delivered privately moves little, and the imihigo and Doing Business cases show that a scoreboard kept by the scored is worse than none. What actually moves a settlement is narrower and harder than most transparency reforms allow: a real test, common knowledge, and an independent scorekeeper.

That independence is a political achievement, not a feature of any regime type. The cases in this post run from a one-party state to a young democracy because the mechanism does not turn on who governs but on whether the scorekeeper can be leant on, and, as the radio cases showed, on whether the country has the wiring to carry the result at all.

So there is a single test for any transparency reform, and it is not whether more was disclosed. It is whether the result was made common knowledge, certified by a scorekeeper, the players had no way to lean on. Most disclosure fails it, which is why so much of it changes so little.

Announcing a result is not the same as making it common knowledge. An announcement still leaves each person unsure whether anyone else will act on it, which is the gap the public-delivery trials measured. Better communications plans do not close it.

How a rule changes

For a generation, British governments handled the coal miners with a quiet back-room accommodation. The first post in this series used Edward Heath as its example of the explorer who finds the wall real: he tested the miners in 1974 and lost his government on the answer. Margaret Thatcher came back to the same wall in 1984. The previous post traced how she had prepared the ground years in advance (coal stockpiled, mobile police ready) so that when the confrontation came, she could hold it open rather than take a private settlement. She won.^[13] What mattered was not only that she won but that she refused the back room and forced the verdict into the open, so that every other union and every future government revised what it believed was possible.

A boundary moved in private is a rumour; a boundary moved in public is a rule.

But the two are not mirror images. A wall that holds seals itself the moment it is seen to hold, because everyone was already braced for it; a wall that falls has to be shown to have fallen, again and again, against the pull of everyone’s old expectation.

Entrenchment is cheaper than transformation. So if you mean to move a settlement and not just win inside it, spend on the part that is never the fight: making the result common knowledge, and keeping it that way, through a scorekeeper the losers cannot touch.

Subscribe now

South Africa Is Too Strong to Change

Stephen Brien — Mon, 06 Jul 2026 07:30:33 GMT

The short version

South Africa’s debt has more than tripled since 2008, with no default, no bailout, no run on the currency.
That looks like strength, and it is. The question is what the strength is for.
A capable state built to share out what it has funds that settlement from its own savers, in its own currency.
No outside creditor ever arrives to call time. A weaker state would have been forced to change long ago; this one can keep paying.
So the strength does not loosen the trap. It trades the shock that would force the question for a larger, internal reckoning, spread thin enough for the settlement to absorb.
Which is why the most capable state in the region is the one going backwards.

This is the second of two posts on South Africa, and part of a longer series on why economic transformation is so rare. The first post showed a state that is capable yet pointed in the wrong direction, built to share out what already exists rather than to build more of it. This one is about what that capability does over time. Readers who have followed the series will recognise the shape of the answer from Indonesia, which managed its money better than almost any country of its kind for thirty years and still ran itself onto the rocks in 1997. South Africa is the same machine, and it is still running.

In October 2025, South Africa came off the global money-laundering grey list. A month later, one of the big ratings agencies raised the country’s credit score for the first time in over sixteen years. The news was reported the way good news is reported: a corner turned, the cleanup paying off, the country let back into the world’s good books. Borrowing, everyone agreed, would now get a little cheaper.

Here is what the celebration sat on. South Africa’s public debt has more than tripled as a share of its economy since 2008, growing far faster than the country’s income. In most of the developing world, a debt climbing like that ends in a familiar way: a default, an IMF programme, a run on the currency, a queue outside a bank. South Africa has had none of it. The state kept borrowing, the markets kept lending, and the settlement the borrowing pays for went on exactly as before.

A debt that size should have forced a reckoning years ago. None came.

And that is the strange thing the upgrade papers over. The most capable state in the region has been sliding quietly backwards for a decade, borrowing more all the while, and the crisis that everywhere else does the forcing has simply never arrived.

The absence of a reckoning is not a footnote to the decline. It is the engine of it.

Part I — The danger that looks like safety

Everything that should reassure is part of the machinery of the decline: the upgrade, the cleared auction, the debt the markets are glad to hold. The first question is why nothing has forced the country to stop.

Capable, and going backwards

The standard advice for a country that cannot grow is to build state capacity first, and growth will follow. For a state built to share out what it has rather than build more, that advice is worse than incomplete. Capacity does not break the trap; it deepens it.

South Africa is where the process is visible because it did the first part. It built a Reserve Bank as credible as any in the developing world, a Treasury that ran the public finances by the book, courts independent enough to jail a former president for contempt, a revenue service once studied as a model abroad.

This is not a country that ignored the advice; it is the closest thing the advice has to a success story. And across the same decade it built and polished all of that, income per person began to fall, and unemployment sat above thirty per cent.

That is the surface, the cleanup, the upgrade, an economy still going nowhere. Capability and backsliding are connected, but not in the way the advice assumes.

The first post in this pair named the one thing such a state lacks, a test its own firms can fail. Call it the firm-test, whether set by open competition or by a government with the will to impose it. It closed on the financing structure that lets the state absorb its shocks.

No wall, no turn

The reason no reckoning ever came lies in how systems change. A country holds its course until it hits a wall it cannot get over, a crisis it cannot absorb, a bill it cannot put off. Capability is, among other things, the power to absorb walls, and the forcing event delivers the reckoning that argument and elections rarely manage, because the settlement’s beneficiaries are also its voters. The ballot box tends to ratify the arrangement, not end it.

In South Africa the arithmetic is blunt. Around twenty-eight million people, approaching half the population, draw a social grant, a monthly welfare payment from the state, and the public payroll and its unions sit at the core of the governing party’s base. No majority is going to vote that away. That is why it takes a wall.

Not every crisis forces, though. A settlement built around distribution throws off one kind of emergency: a solvency squeeze, a commodity shock, a run of bad budgets. Call these the payable crises. They hurt, but money ends them, and when they pass, the people the settlement protects are still in place.

South Africa has met a string of them this past decade, from the 2015 currency scare that followed a finance minister’s overnight sacking to the fiscal shock of the pandemic, and has come through each with the arrangement intact. The forcing crisis is the other kind, the one money cannot end, the one that puts the model itself on trial. (An earlier post drew the distinction.)

A distributive state is very good at surviving the first kind. So the second kind never arrives, because the forcing crisis is the payable one left to grow.

Capability is the strength to survive the payable crisis, and that is the edge the standard prescription misses. The more of that strength a state has, the fewer forcing moments it meets, and the longer it stays exactly where it is. Competence does not steer a system toward reform; it quietly absorbs the emergencies that would otherwise force the question.

The advice assumes a stronger state does its job better, whatever the job is. But it does only whatever it already does better, and if that is funding a distribution that does not grow, more strength buys more years of exactly that. Capability is neutral about direction; the settlement is not. It is the settlement’s instrument, pointed at its own survival.

Funded from within

South Africa’s strength in absorbing crises comes down to one specific structure that its capacity builds and keeps funded, a financial channel. A credible central bank, a deep and regulated savings system, market access held even through the capture years, all of it feeds that single channel, and the channel is the whole of its resilience. That is the sense in which strength is the trap.

The government borrows in its own currency, the rand, from a deep pool of domestic savings kept at home by rules designed to keep it safe. More than ninety per cent of the debt is owed in rand, and that savings pool is roughly two and a half times the size of the economy, far more than the state has ever needed to borrow.

That independence has already been tested. When overseas investors lost their nerve and sold through the capture years, domestic institutions absorbed the stock without the auctions faltering, because a captive base asks nothing about what the borrowing pays for.

For the money that keeps it running, South Africa answers to no creditor outside its own borders.

What a short fuse looks like

That self-funding structure is what a forcing crisis would have to break, and it is exactly what most distributive states lack. The instructive cases are the ones that hit the wall through the channel that South Africa does not run on:

Argentina in 2001 had promised to swap one peso for one dollar on demand, a currency peg, and borrowed in dollars to match; when the dollars ran out it could no longer keep the promise, and the country defaulted on roughly $100 billion in debt, the largest sovereign default to that point.
In 1994, Mexico funded itself with short-term debt whose value was tied to the dollar, so when confidence faltered, the peso collapsed, the debt fell due faster than it could be refinanced by borrowing afresh, and only a United States-led rescue stopped the bleeding.
In 2001, Turkey had a central bank that took its orders from the government and a banking system thick with bad political loans; a public quarrel between the president and the prime minister was enough to set off a run.

In each case, the wall came through hard-currency debt that the state could not print its way out of. A state can always print more of its own currency to cover a debt written in that currency, so that debt can never be called in by force; a debt in money it cannot print has to be found, and can run out.

South Africa has neither the dollar debt nor the captured central bank. It carries none of the original sin that broke the other three, the curse of owing in money you cannot print.^[1] It prints the currency in which its debt is written, and it owes that debt, in the main, to its own people.

The channel determines the length of the fuse. Borrow abroad in money you cannot print, and the fuse is short; borrow at home in your own currency, and it is the longest fuse there is.

What sets the longest fuse available is not how capable a state is, but what currency it owes in. Capacity is not irrelevant here. It determines how quickly a short fuse burns, which is why Turkey’s captured central bank cracked under a quarrel. What no amount of it can do is turn a short fuse into a long one.

The rescue South Africa never has to call

In a distributive state, the rescuer that eventually arrives is the IMF, which refinances the emergency and, in doing so, locks the arrangement in place.^[2] South Africa has internalised the rescue. When the IMF arrives, it arrives with conditions; the lender is also the discipliner. South Africa’s own captive bond market does the lending, and none of the disciplining.

Markets still discipline South Africa, and the rating agencies have cut it to junk, judging it risky. But a downgrade only makes borrowing dearer, and dearer borrowing is something a government can carry for years. What they cannot do is force the reckoning that comes when lenders stop rolling over the debt, because its captive base means access is never quite lost. South Africa is the same mechanism running at full strength. It is a distributive settlement that can fund itself in its own currency for as long as its own savers keep their money at home. The one thing that has forced distributive states everywhere to a reckoning, the day their outside funding runs out, is the one thing South Africa has never had to face.

Part II — Why waiting makes it worse

A capable state can keep paying to put the reckoning off. Each year it does, the bill grows larger, and the economy it protects drifts further from anything that lasts.

What the good news buys

The deferral is working; the reckoning is held off another year. The grey-list exit and the upgrade that opened this post are genuine achievements, but right now they make it cheaper to keep funding the settlement exactly as it stands.

For a state aimed at sharing out what exists, an upgrade or a grey-list exit is not a sign the trap is loosening. It is a sign the deferral just got cheaper.

The political opening of 2024 tells the same story from the other side. For the first time since 1994, the ANC lost its majority and had to govern in coalition. If any event was going to reopen the menu, this was it.

Two years on, the coalition has left the economic core in the same hands, defended the empowerment rules rather than loosening them, and let the public-sector wage bill, already about a third of all government spending, keep climbing. To seat every partner, the cabinet swelled past seventy ministers and deputies, among the largest in the world. The menu reopened, and the government ordered the same meal as before.

The wall is being built higher

The deferral is not neutral. Every crisis a capable state absorbs, and every yearly commitment it hardens, raises the size of the shock that the next correction would have to clear. Every year of deferral lays another course on it: a higher wage bill, a longer grant roll, and a debt whose interest bill is now the fastest-growing line in the budget, swallowing around a fifth of every rand the state collects, more than the state spends on either the health of its people or the social grants that anchor the settlement. The country is building the wall it will one day run into, and building it higher the longer it waits.

A government disciplines its own base only when inaction has become more expensive than action, and only while the crisis is still small enough to leave a government standing to act on it. Correction rather than collapse, big enough to force the hard choice, small enough to survive making it. Call that the correction window.^[3]

A capable state spends its competence absorbing precisely the correction-sized shocks that would have fallen inside that window. Absorbing a shock does not clear the underlying imbalance; it postpones the bill and lets it grow, so each shock the state survives is replaced by a larger one. So the shocks that finally get through are the ones that have grown past it.

In 2008 the debt was a quarter of national income, and a correction then would have hurt but held; today it is three times that, and the window has narrowed with every budget since. Indonesia ran exactly this for thirty years, a competent state that absorbed every survivable shock until the only one left was past the size at which any government could act.

For a state aimed at distribution, absorbing every crisis does not merely postpone the reckoning. It enlarges it.

Rented, not owned

The wall is not the only thing the deferral builds. It also buys capability the country will never own, and one industry makes it plain. South Africa’s car-assembly subsidy cannot be withdrawn or reconditioned, because its beneficiaries, the assemblers and the organised workers who staff them, sit inside the governing coalition, so the test that would make the sector earn its support is never added.

And the support bought world-class assembly, BMW, Mercedes, Volkswagen and Toyota all building for export from South African plants, but it asked nothing back: the carmakers run the return-on-capital maths and build where it pays, so the subsidy buys their presence, not their commitment.

By every measure normally used to score industrial policy, this is a success: output, exports, quality, jobs. The failure is the one those measures cannot see, whether anything was built that lasts. Capability built with no demand to sustain it is rented, not owned, and it leaves when the maths turns. The maths is already turning, as the world moves to electric vehicles and the plants that pay to build here today need not tomorrow. Three decades of subsidy, and nothing owned for a reckoning to fall back on.

The maths is not the state’s alone to set. A mineral economy keeps the rand strong, because commodity exports pull the currency up. A strong rand makes what South Africa produces dearer for foreign buyers, so the export industry struggles to compete, while safe domestic returns remain high. That inheritance makes the firm-test harder to impose. It does not impose it on the state’s behalf. The orientation is defended, not merely inherited, and the longer the settlement holds, the deeper the defence digs in.

Part III — Why the comforts don’t hold

Three reassuring stories suggest themselves: that the trap is not really so binding, that the strength already built is a loaded spring a crisis will release, and that a capable state is at least better placed than the ones that have already hit the wall. This part takes each apart.

Brazil had the channel too

The reassuring reading is that the trap is not so binding, that this channel is no iron law. The whole argument now rests on one claim, that it is the channel, not the capability, that sets the length of the fuse. That claim stands only if it survives the one country that had the channel and cracked anyway. A country built like South Africa on the money side, yet with a different ending, isolates the channel as the thing on trial. If it were the whole story, that country could not have cracked.

Brazil is that country, and it was no Argentina. It did not borrow in a currency it could not print. In 2015 and 2016, it had exactly what protects South Africa today: an inflation-targeting central bank that markets believed in, a deep local-currency bond market, its debt owed largely at home, and a debt load to match. It had the long fuse, not the short one. And it cracked anyway: a deep recession, a downgrade to junk, a president impeached. The channel did not save it.

It did not save Brazil because the channel is not a permanent feature. Credibility is earned, and it can be spent. Brazil’s distributive politics burned through its credibility, and once the markets stopped believing, the captive financing stopped being captive. So the deferral South Africa enjoys is not a law of nature. It is a contingent condition, holding only as long as the credibility behind it holds.

The earlier post on Brazil told this from the inside. Brazil shows that the channel can be overridden by the politics running through it; South Africa shows what the channel buys for as long as the politics leaves it intact.

South Africa is already some way down that road. All three of the big rating agencies have cut it to junk, for the same reason they cut Brazil. A settlement that keeps growing while the economy does not.

And it has brushed against the cliff itself. The currency scare of 2015, when a president sacked his finance minister overnight, was a few days in which the markets briefly bet the discipline was ending, selling the currency and demanding more to lend, in the very years Brazil was cracking for good. South Africa stepped back, and the minister was reversed within the week. Brazil did not step back. That is the whole of the difference between them, and nothing guarantees South Africa steps back the next time.

The kind of wall Brazil hit matters, though. What it hit was not a default and not an IMF programme, but a recession and a political rupture, painful but survived. Higher capacity bought Brazil a softer wall, not no wall at all. Capacity does not abolish the forcing crisis. It defers it, and it cushions it, and in a distributive state, both of those buy the orientation more time. And that is the unsettling part: the softer wall is South Africa’s best case, not its worst. When the channel finally gives, what Brazil got is the hopeful end of the range.

The loaded spring recoils

There is a hopeful reading of all this, and it usually arrives here. The capability South Africa has built, the argument would run, is stored work waiting on an aim. Point it at production, and the long first chapter of a transition, the part that takes a generation, is already written. A turn would not have to start from nothing.

That much is true, but it assumes the spring releases the way the hopeful reading needs it to. The record says otherwise. When a forcing crisis finally reaches a distributive state, it usually produces not a developmental turn but austerity, default, or populist entrenchment.

Here is the range a forcing crisis hands a distributive state, and the range South Africa is facing:

In 1997, Indonesia hit its wall with regime collapse and years of lost growth, not reform.
Greece after 2010, which had given up its own currency and thus had no soft adjustment to fall back on, endured a decade of austerity and emerged on the far side with the same economy, only smaller.

Korea in 1997 is the one clean case of a crisis that opened onto transformation, and even Korea turned on an export orientation it already held before the crisis hit; the crash restructured the conglomerates, but it did not invent the developmental aim that did the restructuring.

The pattern is not encouraging. The wall, when it reaches a distributive state, breaks the model far more often than it remakes it.^[4]

A crisis can restructure an orientation a country already holds; it cannot supply one it lacks.

The reason is not bad luck. The deferral does a second kind of damage, on top of enlarging the reckoning. The same years that entrench the orientation also strip out what a turn would need. A state that spends a generation rewarding the firms best at working the government, rather than the ones best at making things, has nothing competitive left to put in their place when the old order finally gives. A forced reckoning cannot call a new economy into being. It can only work with what the old one left behind, and a distributive state leaves behind firms built to dodge the test, not to pass it.

That is South Africa’s position exactly. Korea, when its crisis came, had an export orientation a generation in the making, something for the crash to restructure. South Africa has spent its own generation building the other thing, rented assembly, and an economy organised to share out what the mines earn rather than to make what the world will choose to buy. When its wall comes, the crisis will find nothing of the kind to work with.

The bigger the deferral, the bigger the break

The last reassuring reading is that a state this capable is at least better placed than the ones that have already crashed. Indonesia is the closest thing South Africa has to a mirror: the ratchet is the same, and only the fuse differs. It was the one competent, well-managed state that ran this same ratchet all the way to the wall.

Indonesia’s deferral ran on an oil-funded buffer, and when its wall arrived it came through the dollar. Its big business groups had borrowed abroad in dollars, so the day the rupiah, Indonesia’s currency, fell, their dollar debts suddenly dwarfed what they could pay, and they were bust overnight. In 1997 the catastrophe reformed nothing. It cleared the board.

That is the part that should concern South Africa. Indonesia was no failed state; like South Africa, it ran an economy of firms that worked the government rather than the world. When its deferral broke, there was nothing of the kind a transformation needs to take the board’s place, so the crash simply cleared it. South Africa has built the same economy.

South Africa’s deferral runs on its own currency, the one channel Indonesia lacked. That is why it has not had its 1997. South Africa is Indonesia’s ratchet still mid-turn: its strength defers even better than Indonesia’s oil did, so the base it would one day have to break keeps growing.

There is an unwelcome corollary. A reckoning deferred this long does not have to arrive the way Indonesia’s did, in a single crash; an own-currency system more often fails slowly, through a weaker rand and prices that climb faster than wages. South Africa is already living the slow version of it. The rand has lost more than half its value against the dollar since 2011, and what wages can actually buy has barely moved in a decade: the reckoning arriving in instalments. But slow or sudden, the base it would have to unwind grows larger every year. The better the deferral, the bigger the adjustment it stores.

Part IV — What it costs, and who pays

The bill for never being made to choose does not disappear. It is paid, a little at a time, by the very people whose savings funded the delay.

The danger in the safest asset

For a state aimed at distribution, “build capacity first” is not neutral, and it is not merely insufficient. It can be counterproductive, counter to the field’s first instinct. The capability thus built is the ability to defer the one force that would have changed the state’s direction, thereby strengthening the very thing that holds the orientation in place.

Economics files monetary sovereignty, the power to borrow in a currency you print, under insurance, and the economist is right. A state that owes in its own currency cannot be forced into default. The safety is real. What makes it dangerous here is what the safety buys.

For a country bent on building, removing the threat of default is a foundation. For one bent on sharing out what exists, it removes the one reckoning that would have put the model on trial and turns it into a slow internal adjustment that a distributive coalition can absorb for years without ever reordering itself.

And that is the whole of the harm. The forcing crisis was never a reliable cure. Its one real value was that it ended the deferral and made the model defend itself at all.

A slow erosion makes no such demand. The cost it carries is paid outward, a little at a time, by the very savers who fund the deferral in the first place. Their money, held at home by the rules that keep it safe, buys the bonds that postpone the reckoning, and then pays for the postponement through a currency that weakens under them year by year.

A missed payment lands as a single event, one creditor and one moment, a refusal made all at once. A weakening currency takes a little from each saver on no particular day, with no moment to refuse and no one to refuse to. The structure that funds the deferral is the structure that scatters its cost, which is why the fuse burns so slow.

In South Africa the loop has a name. The government’s largest single creditor is the Public Investment Corporation, Africa’s largest asset manager, which invests the public sector's retirement savings and lends much of that money straight back to the state. The deferral is funded, in good part, out of the savings of the people the state employs, and paid for, a rand at a time, by everyone whose money the same weak currency is quietly eroding.

So monetary sovereignty spares a distributive state not a likely turn, but the only moment it is ever made to ask the question. It leaves the orientation not merely unturned but unexamined, indefinitely. What reads as the safest asset a developing country can hold is, for a distributive one, the longest fuse, and the longest fuse is the most dangerous to sit on.

Why “stabilise first” builds the trap

It also inverts the usual order of operations. The standard counsel is to stabilise first, build capacity next, and wait for reform to follow. In a distributive state, that sequence builds the deferral engine before anything has fixed what the engine is pointed at. Capability is worth building, and any transformation needs it; the error is in building it first while the orientation stays distributive, and trusting that reform will follow.

South Africa has run that sequence in full. It stabilised its public finances in the 1990s, built credible institutions through the 2000s, and even wrote the reform down, the National Development Plan, unveiled with fanfare in 2012 and broadly endorsed. More than a decade later it remains largely on the shelf, waiting for the one thing that would make a distributive state act on it, a reckoning it cannot defer.

This updates a real and serious consensus. The state-capacity literature reads the usual failure as a state that has the form of capability without the function, the look of a developmental agency without the delivery.^[5] South Africa is the other failure, the one the consensus is not built to see: genuine function, pointed at the wrong end. Building more of the same capacity does not edge it toward transformation. It digs the channel deeper. It is also why the shift the field keeps waiting for, from an economy of deals to one of rules, never comes here. The crisis that would force it is the thing the capable state is built to postpone.^[6]

The boundary should be kept honest. Capacity-deferral guarantees the non-turn, not that a turn, once forced, would succeed. The absence of a forcing crisis explains why nothing has changed, not what change would look like if it came.

None of this is a case against the distribution itself. The grants hold millions clear of destitution, and that is no small thing. It is a case about the one test, market or state, that the settlement has never been willing to add alongside them.

What strength was for

South Africa built a real state and aimed it at sharing out what exists. The strength it built to do that, funding the settlement in its own currency, clearing its own bond auctions, answering to no outside creditor, is the same strength that keeps the one corrective crisis permanently out of reach. Each year it works, the reckoning moves further off, and the settlement it protects drifts further from anything that grows.

This returns the argument to 2024, the coalition still in place. The coalition reopened the menu, and the question the first post left open, whether South Africa will turn, this one can answer in part. It will not turn from the capability it already has, nor from the absence of crisis, which is precisely the problem. It will turn, if it turns, only when something arrives that the capable state cannot absorb.

This is one of the quieter reasons transformation is so rare: the states most able to build its preconditions are often the least likely to be forced to use them.

South Africa’s debt tripled, and no creditor ever came to collect. That was not the danger passing. It was the danger doing its quiet work: every absorbed year hardened the orientation, raised the wall another course, and hollowed out the economy that would one day have to survive it. The state has spent thirty years making the reckoning come later and larger, onto an economy with less left to survive it. That is what its strength has been for.

In this series

South Africa: Capable Enough to Fail — the statics half: capacity and orientation are separable, and South Africa’s is aimed at distribution.
Why Africa Keeps Getting the Wrong Kind of Crisis — the payable-versus-forcing distinction, extended here as the internalised rescue.
Indonesia: The Ratchet That Built the Crash — the same ratchet run all the way to the wall, on a different fuse.
Brazil — the confound: credibility is earned and can be spent, so the channel can fail.

Notes

“Original sin” is Barry Eichengreen and Ricardo Hausmann’s term for a country’s inability to borrow abroad in its own currency, which forces it into foreign-currency or short-term debt and a currency mismatch: when the local currency falls, the debt burden jumps and can become unpayable. Argentina, Mexico, and Turkey all carried it; South Africa, borrowing at home in rand, largely does not. The thesis is well established; what is new here is the inversion: escaping original sin removes the very crisis that would otherwise force a distributive state to change. (Barry Eichengreen & Ricardo Hausmann, “Exchange Rates and Financial Fragility,” 1999.)↩︎
An IMF programme is more than a loan. It arrives with conditions, fiscal targets, sometimes the restructuring of state firms or the opening of markets, and it works in part as a borrowed commitment: the discipline a government accepts from outside when it cannot impose enough on itself. Paul Collier called such external bodies agencies of restraint, institutions a government leans on to bind itself when its own promises will not hold. Those conditions are built primarily to restore solvency and market access, and even where they have pushed liberalisation they have not supplied the firm-level discipline that forces firms to compete, so a programme can stabilise a distributive economy without redirecting it. A state that finances itself never has to accept that bargain. South Africa’s own savers have (so far) always lent, so the outside discipline, and the leverage that travels with it, has never arrived. (”Agency of restraint” is Paul Collier’s term, 1991.)↩︎
The intuition is Hyman Minsky’s. A system that absorbs every shock grows steadily more fragile, because each absorbed shock removes a correction and lets the underlying imbalance build, until the one that finally arrives is too large to contain. Minsky drew it for financial markets, where a long calm breeds the leverage that produces the crash. The correction window applies the same logic to a distributive state’s fiscal and political commitments, which a capable government keeps absorbing until the adjustment they require is one no government can survive. (After Hyman Minsky’s financial-instability hypothesis: stability is destabilising.)↩︎
Economists have long debated whether crises trigger reform. The optimistic “crisis hypothesis” says they do; the record is more mixed, and the cases of real transformation tend to be ones where the developmental orientation predated the crisis. The rule here lands on the sceptical side: a crisis can restructure an orientation, not originate one. (On the contested “crisis hypothesis”: Dani Rodrik, Understanding Economic Policy Reform, 1996; Allan Drazen & Vittorio Grilli, 1993.)↩︎
The state-capacity literature’s leading account of failure is Matt Andrews, Lant Pritchett and Michael Woolcock’s Building State Capability (Oxford University Press, 2017): the state that takes on the form of a capable institution without its function, what they call isomorphic mimicry. South Africa is the case that the account is not built to see, a state whose function is real and pointed at the wrong end.↩︎
Deals and Development (Lant Pritchett, Kunal Sen and Eric Werker) reads economies like South Africa’s as run on particular deals rather than impersonal rules, and asks what moves a country from the one to the other; the first post set South Africa’s industrial bargains in those terms. The framework answers the question of what drives the shift in terms of changing elite incentives. But it does not name a single forcing mechanism. This post supplies one: the shift is forced by a crisis the old arrangement cannot absorb, and a state with South Africa’s fiscal strength keeps absorbing it, so the force never arrives. Capability can be a brake on the move from deals to rules, not just a precondition for it. (Deals and Development, Lant Pritchett, Kunal Sen & Eric Werker, Oxford University Press, 2017; the “single forcing mechanism” reading is my characterisation of a gap in the framework, not a quotation from it.)↩︎

South Africa: A World-Class State Solving the Wrong Problem

Stephen Brien — Thu, 02 Jul 2026 07:30:13 GMT

The short version

South Africa cleaned up: tthe lights came back on, the grey-list was lifted, the credit rating moved.
The economy did not grow, and because the decline had started before Zuma and continued after, the problem was not capability.
It was what the capability was pointed at.
South Africa built the institutions its governing settlement required; it never built the ones that hold firms to account for producing more.
The question it raises is structural: not whether the government wanted to develop, but whether what the settlement distributed made a firm-test architecturally available at all.

This is part of a series on why economic transformation is so rare. Readers who have followed it have already met the variations: Malaysia, which made the transition to a productive economy on thin institutions and just enough developmental purpose to make it stick; Indonesia, which managed its money better than almost any comparable country for thirty years and still never built the firms that could have made it rich. South Africa is the mirror: stronger institutions than either, and none of the transformation.

Part I — The puzzle

The blast furnaces at Vanderbijlpark have been going cold.

For most of a century, this stretch of veld south of Johannesburg made the steel that built South Africa. The works traces back to ISCOR, the state’s own iron and steel corporation, set up in the 1920s as the proudest object of the country’s industrial ambition. The mines, the railways, the towers of Johannesburg: South African steel, much of it from here.

In December 2025, ArcelorMittal, which took over ISCOR’s steelmaking after the company was privatised in 2004, said it was winding down its long-steel business. Around 3,500 jobs go with it

In the surrounding municipality of Emfuleni, more than half the workforce is already without a job, up from about a third in 2011. A boatbuilder who once employed 25 people is down to 15 and says she is living hand to mouth.

This is what the failure of South African development looks like from the ground.

Now drive an hour north, to Pretoria, and watch the same country do something completely different.

Same state, two faces

On the evening of 9 December 2015, Jacob Zuma fired his finance minister. Nhlanhla Nene was gone by nightfall, replaced by a backbench MP almost nobody in the markets could name.

The reply came not from parliament but from the price of money. Over two days, the rand fell from about 14.6 to the dollar to around 15.4, and the interest the government had to pay to borrow long-term jumped by around one-and-three-quarter percentage points. On a total debt pile the size of South Africa’s, that translates automatically into billions more in annual interest payments, without any legislation or vote required. By the Sunday, Zuma had folded. The backbencher lasted a weekend, and Pravin Gordhan was back at the Treasury.

Four days. A president with a parliamentary majority and his own party behind him, forced to climb down by his central bank, his finance ministry, and a bond market that would not pretend the appointment was normal.

Same state. Two faces. One forced a president to back down in four days. The other could not keep a single steel mill alive.

Capable at the top, lethal at the bottom

Most accounts of South Africa see only one half of that contrast.

The South African Reserve Bank is independent by constitutional design, and it has gone to court to keep it that way. When a sitting president spent public money upgrading his private home, the Constitutional Court ruled he had broken his oath, and later jailed him for contempt when he defied a commission of inquiry.

The Treasury was targeted by the capture project (the organised effort, centred on the Gupta family, to redirect state institutions toward private extraction) and survived. The Zondo Commission spent four years documenting how the Guptas hollowed out one state company after another. It concluded that the failure to take the Treasury was almost a miracle.[h]

Under Pravin Gordhan in the 2000s, the South African Revenue Service became the kind of institution other governments sent delegations to study. An IMF benchmarking rated it world-class.^[1] Revenue tripled. A South African paid their tax online, on time, to an agency that worked.

Then it was deliberately broken. In 2014, Zuma appointed Tom Moyane as head of SARS. The commission that later examined what happened did not reach for diplomatic language: the day Moyane arrived was a calamity for the institution. The unit that handled large taxpayers was gutted. Around two hundred senior managers were pushed out; the executive committee was stripped bare. Skilled people left in their hundreds, and the institutional memory departed with them. The Revenue began to miss its targets.

And then it was rebuilt. After a judicial commission found Moyane had deliberately disabled the institution, a new commissioner was appointed in 2019. He restored the specialist unit that had handled large taxpayers, rebuilt the executive structure, and drew back enough of the expertise that had walked out to make the collection machine run again. By 2024/25, SARS was collecting record gross revenue of 2.3 trillion rand. A tax service built to world-class standards, broken on purpose, and rebuilt, all inside two decades.

A weak state could not have done any of this. By the standards of the developing world, these are exceptionally strong institutions. South Africa’s budget documents are rated among the most transparent in the world, year after year, ahead of most wealthy democracies, a measure of how well the distributional machinery runs, not whether it points toward development.

The strength, though, is not uniform. Where South Africa is capable, at the Reserve Bank, the Treasury, and the courts, it held the line against capture. Where it is not, there was no line to hold.

In Hammanskraal, a township inside the boundaries of the national capital, water that the municipality could not keep clean killed at least 31 people in a cholera outbreak in 2023.[ac]

Capable at the top, lethal at the bottom.

Such unevenness is not random. It maps onto which functions serve the political settlement (the underlying bargain that determines who the state is actually built to serve) and which do not. The machinery of rule-keeping, at the apex, is where the political coalition built around sharing out what the state controls most needs capability. The machinery of service delivery, at the municipal level, is where it has no comparable stake.

Two kinds of capability are in play here, and South Africa has built only one: the stabilising and rule-keeping kind, a credible treasury, a tax service that collects, courts that enforce. The second is the productive kind, the ability to expose firms to a test they can actually fail: whether they can produce something the world will choose to buy, enforced by a judge (a market or a government with real authority) whose verdict cannot be appealed away. South Africa has not built that. The choice of which to build was never neutral: a state that means to keep sharing out what it has needs a treasury that can keep borrowing, and has no use for a test its protected firms could fail. That enforcement muscle already exists. What the capable state has lacked is not force but direction.

The recovery that wasn’t

Before the orientation argument can stand, it has to survive the simplest counter: that the stagnation was Zuma’s, not the settlement’s. The dates rule that out.

After Zuma fell in 2018, the recovery arrived on schedule.

Load-shedding came to an end; by 2025, the lights had stayed on for the better part of a year. South Africa came off the global money-laundering grey-list (which raises the cost of cross-border transactions) in October 2025. The following month, it earned its first credit upgrade from a major ratings agency in over 16 years.

On every measure of a state pulling itself back from the edge, the country passed.

Yet the economy did not grow.

Income per person is still below where it stood in 2007. Growth has stayed under one per cent per year for a decade, slower than the population. Unemployment sits near a third of the workforce. On the broad measure, which counts the people who have given up looking, it runs above forty per cent.

A child who started school the year Zuma took office has now spent her entire education inside an economy that has produced no net gain in income per head.

This is the fact that the corruption story cannot absorb. That story said that if you clean it up, growth would return. South Africa cleaned it up. Growth did not return.

The dates give it away. The decline did not begin with Zuma. Fixed investment was already sliding by 2008, the year before he took office, and the public-sector wage bill had been reset upward by the salary deals of 2007 to 2010. And the drag deepened after the looters were gone: the post-2018 wage settlements, the grant rolls still climbing. The drag brackets the capture decade on both sides. Capture made things worse. It did not create them.

The post-2018 period brought its own headwinds: COVID, a generation-high rate environment, an Eskom infrastructure stock depleted over a decade. Those are real.^[2]

A settled system has a resting state, a position it slides back to once whatever pushed it away lets go.^[3] It returns there because the people it serves have every interest in returning it there. The looting pushed South Africa toward something worse. Once the looters were out, it settled back.

The country did not fail to recover. It recovered to a place that does not grow.

Subscribe now

What the inheritance cannot explain

The second counter is structural: that an economy shaped by resource extraction cannot sprout factories regardless of what the state does. The same president, the same resource base, and one institution captured and the other spared. Someone made that choice. The inheritance sets the difficulty; it did not make the choice.

The resource-inheritance explanation says the economy was built to dig things up and ship them out, and an economy shaped that way does not sprout factories on command.

The inheritance is real. But it cannot tell you why SARS was captured and the Reserve Bank was not, under the same president, in the same years.

Nor can it explain the recovery without growth. If the inheritance alone bound the economy, repairing the state after 2018 should have made no difference either way. Instead, the repair worked, the institutions came back, and the economy returned to exactly the low-growth place it had occupied before. Some of the economists who built the structural case now argue the old picture has frayed: that South Africa’s commodity dependence today reflects a fragmented, capacity-starved state more than one dominant complex.^[4]

Other economies transformed from worse starting points. The inheritance story struggles with cases like these. Korea in 1960 was a war-ruined exporter of wigs and plywood; Mauritius had no industrial base, no Cold War patron, an economy organised entirely around sugar, and is now the wealthiest country in sub-Saharan Africa by income per head.^[5] The structural account also fails a closer test. The same inherited structure. The same president. The same resource base. One institution captured; the other left alone. Someone made that choice. The inheritance sets the difficulty. It did not make the choice.

So corruption was a costly detour. The inheritance is real ground. Not one of these explanations is the thing itself. What is left, underneath all of them, is what the state is for.

Part II — What the state is for

The blast furnaces and the bond market are the same puzzle, seen from opposite ends. One direction: a president falls in four days. The other: thirty-five hundred jobs end without a fight. The question Part I left open is not about capability. It is about what the capable state is actually for.

The answer turns on a single concept: governing orientation. Not what a state says it wants (every state says it wants growth) but the prior belief about what an economy is for that sits upstream of (that is, structurally prior to) every budget and shapes which policies survive contact with hard choices. South Africa’s governing orientation is distributional, built to share out what exists, not to build more of it. Not as a finding reached by reading between the lines, but as a commitment that runs through three decades of choice. That is not a failure to try harder. It is what the settlement was built to produce. Three decades of budget choices, firm-level conditions, and costly commitments all confirm it: the capable state has been built to protect distribution, not to force productive transformation.

Orientation is not the same as political will, and it is not the same as declared intent. A government can have vast political will and still have a purely distributional orientation: it expends that will defending what it shares, not redirecting it. The distinction matters. Will-based accounts predict that finding a determined leader solves the problem; the orientation reading predicts that the direction of the state precedes and shapes any policy choices. A government whose underlying orientation is distributional can declare developmental intent and still mean it. The governing orientation encapsulates not sincerity but what happens when that intent meets the first policy that costs the settlement something.^[6]

Governments changed; the orientation did not

South Africa’s distributional orientation did not arrive with the first post-apartheid budget; it did not persist because it was culturally inherited. It was renewed, decade by decade, by the structure of who the settlement rewarded: the same priorities, the same people, the same direction.

The movement that took power had been forged in a long fight against a racial corporatism that enriched a few and excluded the many. It arrived regarding business as something to be transformed and corrected rather than trusted to build. That distrust was earned. But it was the structure of the settlement, not the founding distrust, that reproduced the orientation across governments with very different styles. Subsequent governments did not re-choose it. They inherited a settlement that, through every budget they wrote, kept rewarding the same priorities and the same people, reproducing the orientation rather than choosing a new one.

The cost was structural, not moral. An orientation built on a founding distrust of enterprise sits upstream of every budget, pointing both spending and conditions in the same way: toward sharing out what exists and toward holding the levers through which the sharing is done. What a society chooses to honour, the grant or the maker, shapes what its capable state builds. The settlement does not merely hold firms to distributional conditions; it signals to every firm and every entrepreneur what is honourable, and in South Africa what the settlement honours is holding the terms, not improving performance. In South Africa, that settled long before the first post-apartheid budget was written.^[7]

What governments kept doing alongside it was declare a different orientation. Mbeki’s government made “the developmental state” (a model of government actively steering industrial transformation) official language.^[8]

The National Development Plan of 2012 asked for 5.4 per cent growth; almost none of it was implemented, and Trevor Manuel, who chaired the commission that wrote it, later described a national habit of producing brilliant documents that sit on shelves. The intent kept being declared. The commitment kept not being made.

What the intent story misses is what sits above it. South Africa has never lacked developmental declarations. It has lacked an orientation that could sustain them when they cost the settlement something.

Governments changed; the orientation did not.

None of this is hidden: it is the visible shape of a politics organised around sharing out what exists rather than building more of it. The budget tells you which one the state actually chose.

What the budget rewards

A government’s orientation is not what it says it wants. It is what its budget actually rewards when the claims compete.

South Africa’s budget gives a consistent answer. The public-sector wage bill has climbed from about 5.6 per cent of GDP at the end of apartheid to over ten per cent. The wider social wage (government grants and free services reaching more than half the population) takes above 60 per cent of non-interest spending. Set against that, fixed investment (spending on factories, infrastructure, and equipment) fell from just over twenty per cent of the economy in 2008, before Zuma took office, to a low of around thirteen per cent in 2021, barely recovering to just over fourteen per cent by 2024, against a national development plan that called for thirty. In 2021, the government withdrew 20.5 billion rand from its own infrastructure fund to settle a wage bonus. That is the orientation made visible in a single transaction: when a public-sector pay deal and a capital budget went head-to-head, the pay deal won.^[9]

That orientation has not merely held; it has deepened: the wage bill nearly doubled as a share of GDP over the democratic period, and with each budget cycle, the coalition around the existing distribution grows more embedded and the prospect of reorientation more remote. Building more state capacity of the same kind extends the problem rather than approaching it.

The same logic shows up in who gets the jobs. For two decades, the governing party deployed its own people across the state and the companies it owns, treating senior posts as positions to be handed out rather than filled on merit. That is not, in itself, theft. It is distribution by another name. The state itself becomes something to be shared among the coalition.

Call it a distributional settlement (a political bargain built around sharing what already exists), run by a capable state. Systems become what they reward: not because a culture of distribution has taken hold, but because the coalition built around sharing out what the state controls has every reason to keep the state pointed in the same direction.^[10] And a state excellent at that task becomes the instrument through which the settlement sustains itself.

Subscribe now

Discipline ran toward the books, not the firms

The clearest test of orientation is what a government is willing to absorb to enforce it. Rhetoric is cheap; a budget can be spun; but the costs a government pays against its own base are the thing hardest to fake.

South Africa’s record is unambiguous. GEAR (the Growth, Employment and Redistribution strategy, adopted in 1996) is the case that settles it.^[11] To cut the deficit and earn an investment-grade rating, the ANC imposed austerity on its own alliance, faced down the union federation and the Communist Party, and drove the policy through over their objection.^[12] Mbeki invited his critics to call him a Thatcherite. That is a genuinely costly commitment, paid against the government’s own base.

The question is not whether the commitment was real. It is what the commitment was for. The pain was borne for solvency and creditworthiness, to keep the budget sound enough to continue funding the settlement. A state that means to sustain its distributional programme has every reason to stay solvent, and GEAR was that. It hit its fiscal targets and missed its stated goals of six per cent growth and four hundred thousand jobs a year; manufacturing shrank right through it. The discipline ran toward the books, not toward the firms.

The same government that would discipline its own union allies to protect the budget would not discipline a single connected firm to make it compete. That asymmetry is the orientation made visible. The state will spend political capital to distribute grants, the wage bill, and empowerment. It will spend political capital to stay solvent: GEAR, inflation targeting, and the 2025 exit from the grey list.

What it has almost never spent political capital on is forcing its own side to get better at making things. The will to pay costs was there. What it never paid for was productive transformation.

Inputs, not outputs

The budget and the costs the state is willing to bear are two kinds of evidence about governing orientation. The third kind is how the state conditions what firms actually do, the governing circuit, the set of conditions it runs on firms, and the direction those conditions point. All three converge on the same answer.

South Africa does reach into the productive economy directly. Black economic empowerment shifts ownership and management toward black South Africans and certifies firms on a scorecard. Local-content rules require that certain goods be made at home. Centralised wage bargaining (agreements between unions and major employers, extended by law to every firm in the sector, including those that were never at the table) sets the floor beneath the hiring decision.

But look at what every one of these conditions turns on: who owns the firm, who runs it, who works for it, and what those workers are paid. None asks what the firm makes, how efficiently it makes it, or whether it can sell abroad at a competitive price. Every element (the equity mandate, the local-content rule, the wage floor, the procurement contract) is distributing something to someone. Whether the firm gets better at making things is the one question the scorecard never asks.

Those conditions govern what goes into the machine. They do not ask what comes out.

The distributional-regime account explains the insider-outsider structure; what it leaves open is why the settlement has reproduced itself across governments with very different styles, and why a state capable of rebuilding a world-class tax service has never applied the same capability to production.

That is not an oversight. The conditions run on inputs because the inputs are what is being distributed: protected wages to the workers who already have jobs, the insider/outsider structure that Seekings and Nattrass traced from apartheid into democratic South Africa.^[10-1] Ownership to the connected. Procurement contracts to the coalition. South Africa’s wage-setting practices are ranked among the most rigid globally, reinforced by ownership mandates, local-content rules, and a sprawl of state-owned monopolies; much of it helps explain why a third of the workforce cannot find work. The third who cannot find work are not the settlement’s failure. They are its boundary: the people whose inclusion would require dismantling the protections through which the distribution runs. Every element of that list is distributing something to someone. The unfreedom is not an accident sitting beside the distribution. It is how the sharing-out is done.

A state that means to share out what exists governs by holding the levers, because the levers are what it shares.

Proceeds, not terms

Malaysia also ran a politics of ethnic redistribution. Its New Economic Policy reserved ownership stakes, jobs and contracts for the Malay majority, the Bumiputera (literally ‘sons of the soil’, the indigenous-majority designation under Malaysian law). On its face, more aggressively distributional than anything South Africa attempted.

But Malaysia distributed the proceeds of a working export engine. In Penang’s Free Trade Zones in the 1970s, foreign electronics firms were exempted from the Bumiputera ownership rules. The price of the exemption was performance: to keep it, a firm had to export at least eighty per cent of what it made.^[13] Outside the zone, distribute. Inside it, a condition that the world market enforced. The world market did the testing; but sustaining the zone’s exemption from the redistribution rules required the state to hold the two-track distinction against political pressure from constituencies left outside. Exactly the kind of cost a distributional settlement will not bear.

The freedom to run that condition came from the shape of Malaysia’s political deal. The NEP was built around distributing proceeds, ownership stakes, business opportunities, a growing Bumiputera share of accumulated wealth. A settlement about who receives the proceeds can accommodate a condition on how they are generated. South Africa’s settlement is built differently: the controls are not about distributing proceeds but about who can own businesses, who gets the jobs, and what wages are paid. Changing those terms is not reforming the distribution. It is ending it.

What determines whether a firm-test is available is not how distributional a settlement is (Malaysia’s was very distributional) but what it distributes: proceeds, which an output-test can coexist with, or the terms of production themselves, which an output-test would dismantle.

A test on what comes out is compatible with a settlement built around sharing out what comes in. It is not compatible with a settlement built around sharing out who goes in.

The distinction (whether a settlement distributes proceeds or the terms of production) is what determines whether a firm-test can be threaded through the politics at all.

This mechanism runs as follows. A settlement that distributes ownership stakes can ring-fence a productive zone because firms inside, if they prosper, enlarge the pool from which stakes are drawn; the distribution is supplemented, not threatened. A settlement that distributes the terms of production cannot ring-fence for the same reason: a firm inside the perimeter operating at market wages and subject to performance-based exit is, by definition, a firm whose workers and owners are operating outside the settlement’s primary protection. The zone does not supplement the distribution; it suspends it. Once what goes into the machine is the thing being distributed, no one can ask whether the machine should run differently. The question threatens the distribution.

Subscribe now

There is no South African Penang

The question Penang raises for South Africa is obvious. If Malaysia could ring-fence an enclave where the output-condition replaced the redistribution rule, why not South Africa? The zones exist. The legislation exists. What has never existed is a suspension of the redistribution conditions inside the perimeter.

South Africa has built special economic zones since the 1990s. The Special Economic Zones Act of 2014 offers firms inside the perimeters a lower tax rate and customs exemptions. What it does not offer, and what no amendment has ever added, is an exemption from the Black Economic Empowerment Act or from sector-wide wage agreements set by bargaining councils (statutory bodies where major employers and unions negotiate, with results legally binding on every firm in the sector, even those that never participated in the negotiation).

In April 2025, the Centre for Development and Enterprise proposed trialling a zone exempt from strict labour laws as a way to test a mass-employment model. The proposal’s own description is telling: workers hired on commercial terms, wages agreed between firm and employee, firms that cannot perform let go without a bargaining-council fight. The proposal is, among other things, a proof of absence: it names as a novel experiment what would be unremarkable if any enclave had ever run on those terms. That this required a formal proposal in 2025 is the observable event the orientation account predicts.[ae] The special-arrangements clause in the Act, which in principle allows zones to operate under different laws, has never been triggered to suspend a single redistributive condition.

There is no South African Penang. Not because no one thought of it. The clause that could enable it exists in the law. The enclave has never been available: not because the enabling clause is absent (it is in the Act) but because what the settlement distributes makes suspending redistribution conditions inside the perimeter structurally unavailable under a settlement built on sharing out the terms of production. That structural constraint is conditional, not permanent: it holds for as long as the settlement distributes terms rather than proceeds. The 2024 configuration makes that prior question newly contestable.

The condition it never ran

The Penang enclave worked because it contained one condition the governing circuit had never run: a test that firms could fail. Without it, state support in South Africa has produced a consistent pattern. In no case was there a test the firm could fail before the state committed ongoing support.

The projects South Africa did invest in show the pattern at the engineering level. Engineering capability of genuine ambition was present, a state-backed electric vehicle (the Joule) and a nuclear programme that drew international attention (the Pebble Bed Modular Reactor), but neither was required to demonstrate, before the money went in, that it could recoup its costs. Both collapsed when that question eventually had to be answered.

The car industry shows it at scale. Decades of support for foreign assemblers (Volkswagen, Toyota, BMW, Mercedes-Benz, Ford) produced world-class vehicle assembly and made the sector South Africa’s largest manufacturing export. The assembly plants did not fall short of world-class. They reached it, which is exactly what an incentive structure that asks about quality and volume, and never about whether the design is yours, produces. The intellectual property and brands stayed offshore; the presence remained contingent on the incentive lasting rather than on the industry becoming competitive. In no case was there a condition the industry could fail. It is a bargain the state reliably honours, and one closed to any newcomer who might do it better.^[14]

The result cannot be attributed to South Africa’s distributional intent. Indonesia’s government set out to build a national industry with an explicitly transformative purpose, and it ended up with tenancy, not transformation. Its dominant auto group, Astra International, assembles and distributes foreign marques without owning the design; the one bid for a national car, the Timor programme under Tommy Suharto, survived on political connection rather than performance (the firm-test was waived from the start) and it died in the 1997 crisis. The firmest transformative intent in the region; the same rented result. A decade earlier, in a different sector: when Pertamina’s debts threatened the national budget in 1975, Indonesia’s government culled its largest state oil company to protect fiscal discipline (exactly as South Africa’s discipline has run) and never once used the moment to require that a firm improve its productive performance. The same absence, a different domain.

The two tracks differed not in what they enforced (both asked whether the firm could sell at world prices) but in the political unit of enforcement. Malaysia exempted an enclave; Korea made the world-market test the price of support economy-wide. Both ran the same condition. South Africa ran neither.

The condition can be run by either route. China and Korea ran it through state direction, ownership requirements, performance targets, support withdrawn from firms that failed the test, on political terms that the South African settlement had foreclosed. Singapore ran the same condition through open markets and world-price exposure. The route differed. So did the political conditions under which each route was available; the claim here is only about the presence of the condition, not the methods by which it was sustained. What distinguished China, Korea and Singapore from South Africa and Indonesia was not the money spent or the declared intent. It was whether the support carried a condition the industry could fail.

The governing circuit explains why. A performance condition would ask whether the firms that hold the settlement’s benefits deserve to hold them. The settlement cannot ask that question.

The firm-test is a necessary condition, not a sufficient one. Thandika Mkandawire’s survey of African developmental states locates the scarcity not in the instrument but in the political conditions that make it enforceable.^[15] For South Africa, the prior question is not how the condition would be administered. It is what makes it unavailable under this settlement.

Setting the firm-test absence alongside the budget and the costs the state has absorbed: the three faces of the orientation read together.

Across all three of its faces (what the budget rewards, what conditions it runs on firms, and what it has been willing to enforce) South Africa’s governing orientation is distributional. The budget rewards wages and grants over investment. The firm-level conditions ask only about inputs, not performance. And the political pain it has absorbed has always protected solvency or distribution, never productive transformation.^[16]

Part III — Implications and constraints

Part II ended with what three decades of budgets, conditions, and costly political choices all confirm: the orientation is distributional, and it runs deep enough to reproduce itself. That is the diagnostic half. The harder question is the other half.

Why has nothing enforced a correction in thirty years? The answer rests on a single asymmetry. South Africa’s fiscal discipline faces outside enforcement it cannot override. The equivalent enforcement for productive discipline (holding firms that carry the economy to account for whether they improve or fail) has never arrived.

The enforcement that never arrived

The Nene firing puts the proof on the table. Zuma fired his finance minister. The bond market reacted. The rand fell sharply, and the cost of government borrowing jumped. That reaction (not a political campaign, not legislation, not a change of government) forced the reversal. Within four days, Zuma climbed down. No market participant ordered the reversal. The reaction made the original decision too expensive to hold.

That pattern has a prehistory. In 1985, Chase Manhattan refused to roll over (that is, renew) its loans to the apartheid state, which suddenly owed what it had borrowed all at once. That crisis helped drive the apartheid government toward the negotiating table: a weaker state hitting a wall that a stronger one would have ridden out.

The bond market imposed fiscal discipline without being asked. No comparable outside enforcement has ever arrived for productive performance. The reason traces to the founding.

The transition that followed apartheid was a negotiated bargain. Universal suffrage was the prize; constitutional protection of much of the existing economic order was the price. Votes went to everyone. Property, the commanding firms and banks, the Reserve Bank’s independence, and the civil service’s sunset clauses (guarantees that kept apartheid-era civil servants in their jobs through the transition) were locked in.

The new majority inherited political power and a deliberately narrowed set of economic levers. That narrowing was real; within it, the orientation was still a choice, and a different choice existed, named, at the founding.

In 1993, the ANC’s own economists argued for a more investment-led, state-directed path: one that would have kept the Reserve Bank under the elected government and built productive capacity directly. The proposal was rejected the same day it was formally presented. Within weeks, the ANC accepted a standard IMF Letter of Intent (a formal policy commitment to orthodox economic management in exchange for IMF support) without changes.^[17] The alternative, faster liberalisation, capital account opening, and letting the world market discipline domestic firms directly, did not advance either.

Two organised factions held power, and neither included the firm-test. The unions were aligned with wages and redistribution: COSATU’s explicit commitment in 1991 was to a high-wage system and, in the long run, to socialism, structurally incompatible with a programme built on export-performance conditionality (the requirement that firms receiving state support meet export targets, the kind of condition Malaysia ran in Penang).^[18] The technocrats who controlled economic policy preferred strict budget discipline.

Both options (developmental-industrial and market-liberal) were claimed by neither faction. No organised force had both the interest and the power to carry either against the coalition’s existing beneficiaries.

Khan’s reading of developmental state failures converges on this: a government does not build the instrument that would break it.^[19] Each budget cycle that rewards the settlement’s insiders for being insiders, not for what they produce, is a cycle that deepens that alignment. The firms that hold the settlement’s benefits are the same firms whose position in the coalition makes a test of those benefits politically unavailable. The orientation does not merely persist. The budget recruits for it.

Thirty years of recruiting for the wrong orientation also shape the wrong firms. The survivors in South Africa’s deals environment are those best at managing official relationships, not those best at making things. The relationship-managing firms that the environment selects for are not merely the beneficiaries of the distributional settlement: they are the settlement’s enforcement constituency. Their political stake in the existing conditions is the downstream product of the upstream selection. Remove the conditions, and you expose firms that adapted to avoid competition, not firms ready to face it.

The settlement blocked the firm-test at the founding, and the budget has reproduced the coalition that prevents it ever since.

Subscribe now

The discipline that worked

The ANC’s economic technocrats (the ones who had designed GEAR and declined both the developmental-industrial option and the union-aligned path) did impose discipline. When they put GEAR together, the Growth, Employment and Redistribution programme, they did not consult the alliance. They imposed it. And it hit its deficit target.

What made that possible was a condition the government could not avoid: the bond market’s signal is public, readable by every participant, and inescapable for every official; it tests fiscal credibility continuously, regardless of political will. No official can pretend otherwise. GEAR worked by aligning with that enforcement rather than fighting it. It cut the deficit because the bond market was already demanding a cut, so enforcement and policy were pulling in the same direction. That said, the mechanism constrains rather than prevents gradual deterioration (South Africa’s debt-to-GDP ratio has risen through the same period), because the signal operates on fiscal credibility, not on the stock of debt.

Fiscal credibility is automatically enforced. Productive performance is not. No daily price tells a government its firms are falling behind; no automatic mechanism demands a response. The signals for it that do exist are neither public nor inescapable. A government can ignore a productivity report in a way it cannot ignore a bond yield. In thirty years of South African history, there has been no analogous productive forcing moment: no export-performance crisis, no productivity collapse that reversed automatically.

The government has remained its own would-be enforcer against its own beneficiaries, and has predictably never pulled the trigger. The hard part was never the test. It was being the one to enforce it.

Both routes, the same obstacle

The bond market enforced fiscal discipline without requiring the government’s cooperation. Could the world market have done the same for productive performance in South Africa? In Malaysia, the condition attached to state support was a number anyone could read. Export eighty per cent of what you make. The test was whether the world market would buy it. The enforcement was automatic: if the world market would not buy what you made, no official could revise that verdict.^[20]

South Africa’s protected firms were never in that position. Selective tariffs and non-tariff barriers kept import competition out of the sectors where the settlement’s insiders operated. Removing the domestic conditions would not have changed that: the equity mandates and local-content rules were one layer of protection; the tariff walls were another. The world market cannot automatically test a firm that does not compete in it.

The firm-test does not have to come from the state. Korea and Malaysia attached conditions to support; Singapore simply opened the market. Either route exposes firms to a verdict the world market delivers. In South Africa, neither route was ever on the agenda. The tariffs are not an administrative remnant from a previous era. They are benefits for the firms and workforces behind them, and removing them is the same political cost as imposing conditions directly. Removing barriers is a one-time political act: pay the cost once, and the world market delivers a continuous non-negotiable verdict no official can amend. Imposing conditions is every decision to hold the line against a connected firm that claims an exception. The friction differs in implementation. Reaching the moment of removal, however, faces the same settlement veto as imposing conditions directly: both require taking away rents that insider firms and their workforces currently hold. The settlement veto operates above both.

There is one corner of the economy where market opening did happen, and it is the exception that confirms the rule. With the blackouts at their worst, the government raised the cap on how much power a private firm could generate, then removed it altogether, and private and rooftop generation rushed into the gap. The lights came back on as much because the state let go of something it had only been throttling as because it repaired Eskom. That is the orientation account’s own prediction in miniature: the state frees what it was merely strangling and keeps its hold on what it distributes through. The gain was real, the one clearly growth-positive move in a decade. But it did not scale into a recovery, because the sectors the settlement reaches into pull any such gain back toward the floor. The crack lets in light. It does not open the door.

Neither route found a carrier

State-imposed performance conditions or market exposure: neither has found a coalition willing to carry it through. The workers locked out of formal employment, and new-entrant firms that would gain from barrier removal, are a real constituency. They face not only tariff barriers but labour market rules (bargaining council wage extension, dismissal constraints) that protect existing employed workers at their direct expense. They have not found a political carrier any more than the developmental-industrial option found one in 1993. The controls that protect the settlement’s insiders (the wage floor, the equity mandate, the bargaining-council extension) are the same controls that lock outsiders out; the arrangement has no exterior, and the constituency it excludes is the constituency that would need to break it.

The performance test on firms was not absent because the skill was unavailable, the intent was insincere, or the mechanism was unknown. Each was tried, or named, or both. No outside enforcement has ever arrived for productive performance. That makes developmental reorientation a political problem, not a technical one, not an institutional one.

The capable state that cannot enforce the firm-test is not merely failing. It is succeeding at something else. South Africa is not a capable fiscal state that also happens to have a weak productive record. It is a settlement that rewards solvency and distribution: an institutional architecture that delivered exactly those two things, enforced in one domain by a judge the settlement cannot override, and unenforced in the other because no such judge exists.

A state can be highly capable and not developmental, because the two properties are independent.^[21] South Africa’s numbers carry the proof: institutions at the apex that, where they have not been hollowed, remain formidable, alongside twenty years without real income growth. Malaysia’s institutions were thinner when it built its export engine. The differentiating variable was not institutional depth; it was what the governing circuit asked of firms. South Africa’s institutions are the muscles a developmental state is built from. What they lack is a governing circuit that asks something of them.

The failure is structural and political: neither route has found a carrier, no outside enforcement has arrived for productive performance, and the institutional capability that exists has been pointed at something other than development.

That political structure has had three decades to compound. Three decades of productive conditions declared and not enforced create a credibility deficit: even if the next government announces an export-performance condition, the first test is whether it holds the line when the first connected firm complains. The track record is what the market reads, not the announcement. The first government to actually hold the line does not merely face a political cost. It faces the cost of convincing firms (domestic and foreign) that this time the enforcement is real and will not be reversed.

So the first real sign of a turn will not be a new plan. It will be a connected firm going bankrupt (or losing its state support for missing a target) and no rescue arriving. Everything else, an upgraded agency, a better governance ranking, a new industrial strategy, is consistent with the orientation staying exactly where it is. The absence of the rescue is the observable event. A government that ‘holds the line’ by intention is not.

The one-party grip that closed off the developmental option for thirty years broke in 2024. When the ANC lost its majority and entered coalition, the factional arithmetic changed in ways that matter: the menu reopened not just on paper as in 1993, but in the arithmetic of who can now demand what.^[22] The 2024 coalition is the first to include a veto actor (the DA) whose institutional rewards do not depend on sustaining the existing distributional architecture. Whether this veto changes what the settlement distributes, rather than merely who runs it, is the empirical question 2024 opens.

Whether 2024 is a juncture

That question is the post that follows this one; what this post has established is the frame for asking it.

Whether 2024 is such a forcing juncture is the question. The conditions for developmental reorientation are not built incrementally; they emerge at specific historical moments, a debt crisis, a change in coalition, a forcing event that no government can absorb. Opening a moment is not the same as using it: what determines whether the moment produces reorientation is what forces the question inside it, not merely that it has arrived.

South Africa has not faced a productive-performance crisis that moved the way the bond market moved. The settled system’s resting state has a monetary prop: because the country finances its deficit by borrowing in its own currency from domestic institutions required to hold its bonds, no external creditor can recreate what Chase Manhattan imposed in 1985. The capable state can keep absorbing the fiscal shocks that would otherwise force a reckoning. The monetary prop does more than buy time. A state that can absorb fiscal shocks indefinitely never faces the kind of productive-performance crisis that forces reorientation.

The instruments built to rank those institutions do not merely fail to see the direction. They reward it. South Africa scores at or near the top of the global Open Budget Index across every survey round since 2010 (including the entire state-capture decade) straight through a period in which real income growth collapsed and unemployment expanded to nearly half the workforce. A state expert at distributing inputs at scale, at running the audit, the tender, the wage settlement, scores well on exactly those measures.^[16-1] What they cannot distinguish is whether that expertise points toward development or away from it. The accountability tools designed to catch that failure cannot see it. They cannot enforce what they cannot measure.

South Africa is insulated from both directions: no external creditor can force a productive reckoning, and no ranking instrument can see that one is needed.

The bond market delivered its verdict in four days. Thirty years of productive underperformance have moved nothing. The asymmetry that opened this piece is the asymmetry that closes it, still intact, still waiting for enforcement that has not arrived.

What South Africa shows is that capability is not the variable. The question any government watching this needs to answer first is what the settlement actually distributes. A settlement distributing proceeds can add an output test alongside the redistribution; one distributing the terms of production (who owns, who works, what wages are paid) cannot. The test would end the distribution. That prior question comes before will, before institutional design, before any enforcement mechanism. Those questions are only reachable once the geometry of the settlement is developmental. Getting that wrong is how a capable state spends thirty years solving the second problem while the first one holds.

Subscribe now

What I’ve been reading

Stephen Brien — Tue, 30 Jun 2026 16:51:12 GMT

Pieces I’ve been reflecting on this month make analogous arguments from different angles: the standard toolkit of development economics looks shakier than the field is willing to admit. What’s left when you strip out the tools points somewhere uncomfortable — and, from a pro-growth perspective, clarifying.

Has the manufacturing route closed?

’s How I Became a Manufacturing Skeptic is the most significant self-revision in mainstream development in years. Rodrik built more of the intellectual case for manufacturing-led growth as the replicable East Asian lesson than almost anyone else, and now argues it isn’t replicable. Even the most successful current industrialisers (Bangladesh, Vietnam, Cambodia) are absorbing far less labour into manufacturing than South Korea or Taiwan did at equivalent income levels. The escalator, he argues, has broken. He also appeared on the Ideas in Development podcast to discuss what services-led transformation would actually require — worth listening to alongside the piece.

’s Why the World Bank changed its mind on industrial policy registers a different part of the same problem. The Bank’s conversion to an industrial policy reflects geopolitical repositioning: Western states competing with China in semiconductors and EVs, not new evidence of development. The industrial policy being exported is designed for middle-income strategic competition, not for structural transformation from a low base. Importing the vocabulary while removing the conditions doesn’t transfer the lesson; it transfers the label. has drawn out what this means for Africa specifically here — the risk that low-income economies adopt a policy framework calibrated for a stage of development they haven’t reached. (I’ve made a version of this argument here.)

’s Petronas: The Exception That Could Not Be Repeated makes this concrete. Malaysia built a genuine developmental institution, but the conditions were constitutive: a specific political coalition, commodity rents at the right moment, bureaucratic insulation that proved unrepeatable. The lesson is about which conditions can be reproduced, not which interventions can. Copy the instrument; miss the point. I’ve written about the governing coalition behind Petronas here. The argument is that Malaysia’s orientation toward distributing resource rents into productive investment, rather than capturing them, is what gave Petronas its exceptional character and why that character didn’t transfer.

Growth, evidence, and what the field won’t say out loud

’s Economic growth is enough makes an uncomfortable empirical case. Constructing a non-monetary composite of basic human wellbeing across 169 countries, he finds it correlates with GDP per capita at approximately 0.9. If that holds, the case for prioritising targeted programmes over market-enabling institutional reform rests on a far weaker foundation than the mainstream has assumed; resource allocation within the field may have been systematically wrong for decades. A companion visualisation piece makes the long-run growth facts hard to argue with.

’s International development: From moralizing to calculations to laissez-faire to rhetoric adds the structural explanation for why a finding like Pritchett’s hasn’t changed the field’s behaviour. Reviewing the history of development policy at the UN, he shows that each phase (basic needs, structural adjustment, human development, the current normative regime) was driven by the interests of powerful states rather than by evidence that prior frameworks had failed. The current “rhetoric” phase combines high normative ambition with almost no operational content and no accountability for delivery.

Together they make a challenging point for the field: development is calibrated around what is legible and fundable for donors, not around the market-enabling reforms and growth dynamics that actually produce what it claims to want.

The constraint nobody in development wants to name

The piece on the ambition of African elites is direct in a way that most development writing avoids. African industrial transformation will not come from external capital or institutional blueprints. It requires domestic elites with the strategic orientation and risk appetite the task demands, and the current generation largely lacks these qualities. The argument is uncomfortable precisely because it doesn’t translate into a programme (but that does not diminish its relevance).

’s Making lasting peace in the Sahel makes the same point in a different register. The Sahel now accounts for half of global terror deaths. Military and technological responses cannot substitute for the elite bargains that give armed factions a stake in stability; without those bargains, security operations produce temporary displacement rather than durable peace. The limiting factor is not operational. It is political: who has an interest in what gets built, and whether that interest runs toward creating conditions for productive private activity rather than capturing what already exists.

Both pieces point to the same constraint: not technical, not financial, not even institutional in the conventional sense, but about the orientation of the people at the top of the domestic political economy. That is more challenging terrain than development writing usually occupies.

The implicit argument running across all three sections is that development organisations are several steps removed from their own stated objectives. If the manufacturing route is narrowing, if targeted programmes are weaker than the growth-wellbeing correlation implies, and if the real constraint is whether domestic elites will create the conditions for markets and productive investment to take hold, then most of what the field does addresses secondary variables at best. The field knows this, as people in all fields know things they cannot act on institutionally. The interesting question is not whether the diagnosis is correct but who is positioned to make it matter.

A Government Whose Costs All Fall Outward Is Bluffing

Stephen Brien — Mon, 29 Jun 2026 07:30:33 GMT

Governments announce reform constantly, and are believed by no one.
Reagan fired eleven thousand air traffic controllers in a day, and was believed at once. A borne cost cannot be faked.
What proves intent, and what tells reform from mere resolve, is who is made to bear the cost.
Spending against enemies is free; spending against your own side — the people the old settlement was built to protect, and hardest of all the people who keep you there — shows which way you mean to go.
The ledger says more than the speeches, and what the ledger rules out is plain before any reform arrives.

This is the third of the “How a Rule Changes” posts. The first asks whether a wall a reformer treats as fixed is real; the second asks where the moment to lean on it comes from. This one reads the ledger. Every government announces reform; almost none pay the cost that would make anyone believe it. What proves intent — and what tells reform from mere resolve — is who is made to bear the price.

In the summer of 1981, the United States found out that its new president meant what he had said, and it found out in a single afternoon.

Ronald Reagan had told the air traffic controllers, a union that had endorsed him only the year before, that if they walked out, he would dismiss them. They walked out. He dismissed more than eleven thousand of them at a stroke, and barred them from federal employment for life. After that, when the administration said it would not be held to ransom, the people who might have tested the claim did not test it. They believed it, and they behaved as though it were true.

Plenty of governments say as much, and no one believes them. They declare that the old settlement is finished, that the rules have changed, that this time it is serious, and the firms, officials, and union bosses who would have to act on the declaration nod along and carry on exactly as before. Words are the same. The result is not.

The difference between Reagan and those governments that no one believes is that he paid the price. Talk is cheap, and action is not. The people who would have to act on a new commitment have watched a hundred announcements come to nothing, and have learned to treat each fresh one as worth close to nothing.

What moves them is not the announcement but the cost that comes after it: the price the leader himself paid, and the people he made bear it.

Reagan fired the controllers, and they did not come back. That was not a regrettable side effect of his commitment. That was the commitment. A cost already borne is the one thing a government cannot fake, because a government that did not mean the thing would never have paid for it.

The firing proves he was serious. It does not tell you what he was serious about. That might have been no more than a president breaking a union that had defied him. The aim is the harder thing to read, and it turns on a different question: not whether he paid, but who he made pay.

The cost has to fall on the powerful

Cost tells you whether a government means reform, and which way it is trying to go.

Not every cost carries the message. A cost aimed at an enemy carries none. Crushing an opponent is what governments are expected to do, and it tells the people watching only that the regime still dislikes the people it already disliked, which they knew anyway.

The cost that carries the message is one that falls on a power the old order was built to protect — and hardest of all, on the government’s own side. A leader who means only to keep power gains nothing from breaking the interests that the current arrangement was built to reward.

That difference is where the cost lands. And the cost does more than prove the leader means the change. It shows the kind of change he means: the settlement, the arrangement of who gets protected and who profits, is going to shift, and his own side will have to live under the new one too. A cost can land on either of two kinds of targets. One is power entrenched but external to the leader’s coalition. That is the union that brought down the previous government, or the foreign investors that a new ruler inherits.

The other, and the stronger, is the cost a leader makes his own side carry, because only a leader who means the change would spend against the people who keep him in power.

Reagan’s case is the sharpest version of the rule, because the union was his own backers, not merely his opponents. The union that had endorsed him a year earlier ran a system the country could not do without, which is what gave their strike its leverage. Breaking them, at the moment their strike could ground the country, imposed a cost on the people whom the old settlement was built to keep on side.

Peter the Great paid the same kind of price in a harsher coin. When he came home to Moscow in 1698, set on hauling Russia into the European age, the thing in his way was his own elite.

Those boyars, whose standing rested entirely on the world he meant to abolish, stood against him; so did the streltsy, the hereditary musketeers of his household guard, who had just risen against him.

He shaved the boyars’ beards with his own hands at the reception that welcomed him home, and later taxed any man who kept his. He had the mutinous streltsy executed during the winter and left the bodies in plain sight across Moscow.^[1]

Peter’s cruelty was real, and it is not the thing being praised here. What counts is where the cost fell. It fell on the men a ruler is supposed to keep on side, which is what made it impossible to mistake.

Who the cost is aimed at

You prove you mean a thing by paying a price a faker would not pay.

Economists have a name for it: costly signalling. Michael Spence found the pattern in the job market, where a degree signals less what it taught than the cost a weaker candidate could not have borne; Dani Rodrik brought it to development, where a reformer proves credibility by going further than a faker would, by accepting conditions too costly to bluff.

Rodrik’s signal is aimed at the outside audience: the creditors and investors whose confidence the reformer needs to win. To them, a costly move reads as resolve, proof he will hold his commitments.

The inward cost points the other way, at the leader’s own side, the people who keep him in power. They are the audience he cannot escape. An outside interest can be bought off or faced down; his own side cannot. So a cost turned on them is one he can neither dodge nor fake.

To them, it reads as more than resolve: a direction that will cost them. The outside audience reads resolve; the inside audience reads direction. Only the second can tell you which way the government means to go.

That same inner circle has long been studied in the work on coups and how rulers fall. There, elite insiders who can retaliate read a costly move as a signal about the leadership, and that gives them reason to act. But that work asks one question of the move: will the ruler keep his power? This post asks the other one, the one a bluffer would never want raised: where does he mean to take the country?

Among inward costs, size matters too. The higher and more visible the cost, the louder it speaks: a year-long strike says more than a single sacking.

And it matters how freely the cost was chosen, a question that sharpens as the cases pile up. A leader forced into a move tells you less than one who paid when he could have done nothing.

What a government is really trying to do

Reagan’s controllers and Peter’s boyars sit centuries apart, yet the cost says the same of both: a government that makes its own side bleed means what it says. But meaning what it says is not the same as reforming. A government that breaks a union holding the economy to ransom shows it will not be ruled by an organised interest. It does not yet show whether it intends to build a different kind of economy.

What tells reform from resolve is which interest the cost falls on. The reforming cases are the ones where it falls on the very interests a country would have to break to build something new.

The clearest case runs through China, and it predates the reforms China is remembered for. In February 1980, Deng Xiaoping and the reformers stripped Mao’s old guard, the security chief Wang Dongxing among them, from the inner ranks of the Politburo. There was not yet a single market reform to point to, no special economic zones, no farmland handed back to families.

These were not external enemies. They were senior men of the reformers’ own party, the faction whose whole standing rested on the policy the reform meant to overturn.

The move is not China’s alone, and the interest it strikes can be larger than that of a faction of the party. When Park Chung-hee seized power in South Korea in 1961, he had the country’s leading businessmen arrested, the men whose fortunes made them the one private power a directed economy could not leave standing, and held over them the threat of losing everything they owned.

Then he let them go, on one condition. They were to build the industries the state told them to build. That cost fell on exactly the people the old order had rewarded most, and it bought their place in a new one.

A century earlier, the reformers of Meiji Japan had struck a larger interest still, and one that was their own. They were samurai, and to build a modern state, they abolished the samurai, stripping their own class of its stipends and swords, and crushing the rebellion that the abolition provoked.

Reform turns inward

A reformer has to spend against his own side in the end. The interests that block a new economy, the blockers, are his own: they sit among the very people the old settlement was built to reward. A government that never turns a cost inward, whose costs fall only on enemies and outsiders, has left the blockers exactly where it found them. It is reshaping nothing.

Reform here means a direction of travel. A reforming government is remaking the settlement to build a different kind of economy, not merely re-dividing the same output among different people.

Whether it reaches that economy by state direction or by the market is not the test. Every case so far has been a government taking the economy firmly in hand, but a reformer can break with the old settlement by the opposite route. Estonia did, throwing its borders open to trade after 1991. The blow fell inward, on the state industries it had inherited from the Soviet order, the very interests a market economy had to sweep aside. Exposed to competition, they folded, and what survived Estonia rebuilt on a single flat tax. Its free-market road and Deng’s state direction run to the same place, on the same side of this line.

Most leaders are unwilling to spend against their own side, which is one of the main reasons transformation so rarely comes.

Ruling reform out

Those who never pay it give themselves away. The ledger does one thing better than anything else: it catches a false reformer. When a government’s costs all fall outward, never once on its own side, you do not need to wait for the programme to fail. Reform is already ruled out.

For twenty years, Egypt under Nasser and Sadat imposed one high cost after another on the landlords, on foreign capital, on the dispossessed old regime. The costs consistently bypassed the coalition’s core beneficiaries: the officers and the subsidised cities. When Sadat’s 1977 bread-subsidy cut finally turned the cost inward, the coalition refused within forty-eight hours, and the government folded.

Egypt was forced inward and refused. Venezuela was never forced at all, and reshaped just as little.

Hugo Chávez built his Bolivarian Revolution on costs aimed outward: he expropriated foreign firms and the old business elite, and faced down the opposition. The one interest never made to pay was his own, the military and the loyalists who lived off the country’s oil rents. Because the cost never turned inward, nothing was rebuilt. When the oil price fell, the economy collapsed, but the order only changed hands: the rents passed to whoever held the guns.

Which way is it trying to go

Ruling out the all-outward fakes was the easy half. The difficult half is interpreting the cost that does turn inward

Such a cost can only tell you so much. It proves a leader is serious; it cannot always tell you what he is serious about. The danger is that every inward cost could be mistaken for consolidation, a ruler clearing rivals under the cover of reform. The test that tells them apart is whether the target posed a survival threat.

A leader out to remake the order goes after blockers, the insiders whose position rests on the old way, whether or not they threaten him personally. A leader out only to keep power goes after rivals, the men who could unseat him, and leaves alone any insider who is no threat to him at all.

The power-keeping leader is the common one. Milan Svolik shows that when authoritarian rulers fall, it is almost always insiders who remove them, which is why managing the inner circle, not managing the public, is the game most rulers are playing.^[2]Mushtaq Khan finds that most of the bargains holding a ruling elite together are built to keep power, not to build an economy. That is exactly why the cost can read both ways: the men a ruler most wants gone are often the same men whose standing a real reform would end.

So when the target is a man who is both a blocker and a rival at once, the two kinds of leaders act alike, because each has his own reason to strike. Here, the cost alone cannot determine which leader you are watching.

Even so, three signs help you lean one way before the results are in:

The first is timing: did the cost fall early, before any reform had paid out a reward?
The second is the target’s position: was it built to guard the old order, or only to threaten the leader?
The third is the pattern: a single purge followed by the leader’s allies growing rich looks like a man securing himself, while a steady run of blows against the interests that block the economy looks like reform.

Deng’s 1980 purge is the case where the signs resolve. Wang Dongxing was both a keeper of the old order and a rival worth fearing, a man either kind of leader had reason to remove. But the cost fell before any reform reward existed, against men whose whole position depended on a reversal of course. That early, against insiders with that much to lose, it points toward reform.

In Mohammed bin Salman’s case, they cannot. When he locked Saudi Arabia’s princes and financiers in a Riyadh hotel in 2017 and stripped them of a fortune, the cost was real: no bluffer would have paid it. But the targets pointed both ways at once. Some had grown rich under the old order, the kind a reformer must break; others were simply his rivals; and the National Guard he seized from a cousin was a seat of power, not the old money a reform would have to go after.

You can read that much from the cost alone, without waiting to find out what Vision 2030 delivers. The plan will be judged when it arrives, based on what it does. The cost has already said all it can. It proves he meant it. It cannot, on its own, show which way he is going.

Making a cost stick

Everything so far has rested on one thing: the cost being one the government cannot take back. A cost the government could quietly reverse would prove nothing, whether you are trying to rule out a fake or spot a real reformer.

There are two ways to make a cost stick. One is to sink a cost: to break, in plain view, an interest you were never meant to touch, at a price that can never be recovered. The other is to tie your hands: to bind yourself to an outside rule, and let the market or a treaty hold you to what your own word could not.

Margaret Thatcher sank one against the miners. Their union was powerful enough to black out the country; a decade earlier, it had brought down Edward Heath’s government. Thatcher did not leave the rematch to chance. Acting on a detailed contingency plan she had commissioned years earlier, she had coal stockpiled at the power stations, non-union haulage lined up, and mobile police ready. When the confrontation came in 1984, the state could withstand a long strike rather than be starved into a quick settlement.

She then fought the strike out over a long and bitter year. That year was for something more than winning. It was to show she would not fold.

Israel tied its hands instead, and broke one of the worst hyperinflations in history. The cure was well understood; what every earlier plan had lacked was a way to make it stick. So in a single all-night cabinet sitting in 1985, with prices losing their value by the week, Israel froze prices, wages and the exchange rate together, then legislated away its own power to print money to cover the deficit. That closed the one escape route the others had left open.

Promises can be broken by morning. A law that a government writes against itself cannot.

Which way it points

Both moves leave a commitment the leader cannot quietly take back, which is why they are so easily confused. But they differ in who the commitment is aimed at, and that difference is what matters.

Tying your hands points outward. The commitment is a promise to the markets or to a treaty partner, an audience whose confidence the government is trying to win. Schelling charted this move half a century ago.^[3]

Sinking a cost can do the rarer thing. It can fall inward, on the government’s own side, on the people who could punish it for the blow, the way Reagan’s firing fell on the union that had backed him. Only that inward blow shows which way a government means to go.

Zhu Rongji’s case shows the line between them. By the late 1990s, China's more addressable reforms were long done, and the difficult ones were blocked by the ministries and state firms that stood to lose from them. Zhu Rongji tied China to the World Trade Organisation on terms no successor could quietly undo. The promise pointed outward, to the markets and the WTO. The cost fell inward, on the protected state firms.

But Zhu did not choose those firms; the terms did. That is the limit of tying your hands: it can send a cost inward, but it cannot choose where that cost lands. Only sinking a cost lets a leader pick the target and land the blow himself.

What the cost always settles

Tying your hands is the one move that gives away that choice. Everywhere else, even when a crisis forces a leader’s hand, the target stays his. A crisis can dictate the scale of the adjustment, but it never determines who bears it, nor does the lender who arrives with it. The IMF can set the size of the cut; it cannot set the address.

The same crisis and the same creditor reached Seoul and Jakarta in 1997. Korea forced the conglomerates that ran its economy to fail or merge; Suharto shielded his own family’s firms and let the cost fall on the fuel subsidies the poor depended on, until the streets brought him down.^[4]

Why some commitments last

An inward cost proves a leader is serious. It does not prove his reform will outlast him. That is a separate question, and a more interesting one. Some commitments hold once their author is gone; others fade with him, and the difference lies in what the cost reached. The clearest case is one in which the cost was unmistakably real, yet the reform faded anyway.

Bolivia is that case. The man who took back the presidency in 1985 to break the hyperinflation, Víctor Paz Estenssoro, was the same man who, a generation earlier, had built the order that bred it. Because he had built that order himself, his blow could reach only as far as what he had made: he could strike the union he had empowered, but not the rentier state underneath it. He had founded the 1952 revolution that nationalised the mines and made the miners' union the strongest force in the country.

He shut most of the state mines, put more than twenty thousand of his own miners out of work, and faced the union down under a state of siege.^[5]

Paz did sink a cost, and the miners were publicly and irreversibly broken. But the blow fell on the union, not on the order that had made the union strong. That order was a state whose business was to live off the country’s resource wealth and parcel it out to whoever held power, and he left it in place. Cutting the weed is not the same as sinking a cost into the root. As an earlier post in the series showed, the inward cost bought stabilisation and little else; a generation later, when gas rents replaced Bolivia’s tin revenues, the same order rebuilt itself around new claimants.

Deng’s commitment compounded where Paz’s faded, and not because Deng was more serious than Paz. Where Paz cut the weed, Deng struck the root. Deng's 1980 purge eliminated the faction that would have returned the country to the Maoist line. The blows that came after did the same. Farm decollectivisation, the special economic zones, Zhu Rongji's WTO terms: every one fell on a group that had profited from the old order, an order that the previous cost had already weakened.

A commitment lasts where the cost reaches the state’s structure, the mechanism that rewards the old coalition, and fades where it strikes the coalition members but leaves that structure intact. Even then, breaking the structure only clears the way; building a new economy on the ground it held is a further test, and no act of will guarantees it.

The claim is a conjecture, not a law, and two cases cannot make it one. Its strength is that it can be proved wrong: it fails the first time a cost reaches the root, and the old order rebuilds itself anyway.

Comparing the two cases cannot separate the depth of the cost from the system that absorbed it. A single-party state can keep the interests it breaks from regrouping; a democracy with organised opposition cannot. So part of why Deng’s reform held and Paz’s faded may lie as much in the political system as in the cost, which is the sharper question to take into the next case.

An inward cost proves a government means its reform. It does not prove the reform will outlast the people who profit from the old way.

What the ledger shows

Reagan was believed in a single afternoon. He paid inward, where no bluffer would. Where a government means to take a country is not in what it announces, but in who it makes pay. The inward cost, borne against the blockers, the powerful that the old order protects and, hardest of all, the government's own side, is the one that shows direction. Bearing it binds the leader to his course in a way no speech ever could.

The thing worth reading, then, is the cost a government is willing to turn on its own supporters. It is the one signal a bluffer could never fake, and it shows before a single programme arrives.

The ledger is surest when it catches a fake. When a government's costs all fall outward, on enemies and outsiders and never on its own side, it is reshaping nothing, and you can see that long before any programme arrives to be judged. Egypt and Venezuela are both proof, one folding the instant it was forced inward, the other never turning inward at all.

What the ledger confirms, it confirms only tentatively. An inward cost tilts the odds toward reform without settling them, and even one truly borne can leave the old order standing if it struck the coalition but not the structure that fed it. Where a state lives off oil or minerals, as Bolivia lived off tin, even that tilt blurs: the resource income flows, however the economy is run. So an inward cost only reshuffles which insiders collect it.

The credibility scorecards, the international indices and assessments that grade the seriousness of reforms miss all of this. They read the cheap signal, the words a bluffing government would gladly have said, and skip the one entry a bluffer could never have made.

And even a commitment proven beyond doubt settles only half the question. A proof by one leader is not yet a rule the country can count on, the kind that binds whoever comes next. How a single leader’s proof hardens into that rule, the seal that still holds once the man who paid for it is gone, is the harder problem, and the one the next post takes up.

For now, the cost is what to judge by. Every government announces that it means to change things; almost none will pay the price that makes the claim believable. The speeches are the part designed to fool you. The ledger is not.

Indonesia: The Ratchet That Built the Crash

Stephen Brien — Thu, 25 Jun 2026 07:30:45 GMT

Indonesia was brilliantly managed. It was never managed toward what transformation requires.
Every crisis the state absorbed was one it could no longer force it to change.
The cost of a manageable correction, one bad enough to matter, small enough to survive, rose with every competent rescue.
By 1997, the window had closed. The crash that arrived was beyond the size that a government could act through.
The regime fell. The governing orientation did not.

The previous post on Indonesia showed that its governing orientation, what the state was built to reward and what it refused to rescue, remained the same under distinctive development approaches. Picking up on ’s Indonesian trilogy, this post addresses what that governing orientation left undeveloped and why it survived the 1997 crash that destroyed the regime that built it.

In 1975, Indonesia’s governing circuit reached down and, within months, culled its largest state company to protect the budget. That was decisive and correct. Each crisis that followed was met the same way: the books steadied, the budget held. Absorbed jolts did not vanish. They migrated into the banks and waited.

In 1997, the same state, facing a regional crisis, could not bring itself to discipline the conglomerates whose unbacked-dollar borrowing was destroying the currency. Indonesia’s economy shrank by thirteen per cent in a year. Suharto was gone within months. The same kind of state, the same concentration of power, and a different outcome.

A peculiar kind of failure

Indonesia was not a country that lacked the skill to run the firm-level test it never ran. By every visible measure, it was brilliantly run; it managed money better than almost any developing country of its kind.

Competence is not like reserves, more or less of one stock. A competence aimed at one target gets better at hitting it, and is rarely turned on another without external pressure.

That competence was the trap. A state good enough to absorb every crisis it faces removes the pressure that might have forced it to develop differently. Nothing smaller than the catastrophe it has been deferring will correct it.

That cost moved in only one direction, because every competent rescue also enlarged the base any future correction would have to break. The window for a manageable correction, a crisis severe enough to force the hard choice, yet survivable enough to leave a government standing to make it, did not slam shut. That window narrowed, year by year, until it was gone.

The rescue and what it protected

The 1975 Pertamina rescue was decisive and right. Indonesia’s largest state company had borrowed more than it could repay; the government moved fast, broke it up, and protected the budget. The first post covers the story in full.

That was also the first turn of a lock. The rescue protected the budget, the currency, and the country’s ability to pay its bills, but did nothing to address whether any Indonesian firm was good at making things. The economists were good at keeping the books straight, and because that worked, a different question was never forced to the surface: why are our own firms not getting better?

That correction could not be forced. If the books are sound and the country is paying its way, there is no visible failure pointing at the firms. The only thing that makes a government ask the hard question about its own companies is a crisis so bad it cannot be handled any other way, and Indonesia kept finding another way. When oil revenues fell in 1983, it devalued and cut spending; when prices collapsed further in the mid-1980s, it let the currency fall again. Each time, the books steadied, and no connected firm was required to do anything differently. Each competent response raised the cost of the next one.

Indonesia had plenty of crises. None of them forced it to discipline its own firms, because the firms were the coalition.

A government does not build the instrument that would break it.^[1]

That settlement was always the constraint. The ratchet is what turned a static arrangement into a compounding trap.

Every year, the arrangement held; the base grew as the Suharto family moved into business and the favours multiplied. The larger the base grew, the more a government would have to act against its own backers to discipline any part of it, and the higher the political cost of doing so.

A government can be pushed past that cost by a crisis severe enough that inaction is plainly more costly than acting. But the size of crisis required to force that choice grew with the base, and the state’s competence kept actual shocks below it.

The better a state is at managing the crises it faces, and the more each rescue expands the protected base, the more likely the reckoning arrives past the point where anything can be done. The contrast with Korea makes the mechanism visible.

Korea in the same decades had no oil revenues and ran persistent deficits financed by foreign borrowing. Exports were not a performance target; they were the mechanism by which the Korean model paid for itself. The fiscal buffer that made Indonesia’s macro-competence possible was the same buffer that made the hard choice permanently avoidable.

Brazil, in the same years, borrowed recklessly and blew up in 1982; ordinary Indonesians were spared that. Fiscal management and economic transformation are different exams — and Indonesia passed the first, permanently, at the cost of the second.

The firms otherwise positioned to demand rules-based treatment were folded into the protected base rather than left outside it. That channel through which the arrangement might have been corrected from within was closed. That closure belongs to the arrangement, not to whoever is running it. The logic of protect the base, absorb the costs does not change when the political superstructure does.

Subscribe now

The capability that was never demanded

A trap like this exacts a price that, while it is being paid, is invisible. Not money, which Indonesia had, and not even growth, which it got. The price is foregone export competence: the capability that cannot be built while success looks like this.

The previous post traced the manufacturing boom and its limits. Discipline that arrived, from foreign buyers who would take their orders elsewhere, landed in the open sector and never reached the connected core. It was growth the country hosted, not capability it built.

The clearest case is the one industry where Indonesia did build real depth. Astra, whose rise and post-1997 reinvention Pietro Masina has chronicled, grew from a 1957 trading house into the partner that Toyota, Daihatsu and Honda built their Indonesian manufacturing around. Astra trained real engineers, built genuine suppliers, mastered modern manufacturing. Yet in four decades it never designed a car of its own.

Astra could have made that climb. Korea’s Hyundai designed its own cars from a smaller domestic base. Policy choices, the profitability of the partnership model, the sheer cost of frontier innovation — all played some role in why the climb was not made. The partnership model also reflected Indonesia’s entry position in the regional production hierarchy: an assembler in a supply chain where design authority stayed in Japan.

Entering a global value chain at the assembly end is a starting position, not a ceiling — Korea’s manufacturers moved from licensed assembly to proprietary design. What the arrangement did not supply was the one demand that might have forced the attempt regardless: no one set the harder standard.

That partnership was profitable and safe, and nothing in Indonesia’s arrangement rewarded the risk of trying for more. By 1997, the crisis posed a different question: could Astra survive? Astra could, through restructuring, and by passing to foreign ownership: Jardine Matheson, a Hong Kong trading house whose history traces to British imperial commerce in Asia, became its largest shareholder.

But survival is a solvency question. No one asked the design question during the recovery, and no one has asked it since.

The distinction is between what Indonesia built — manufacturing to international standards, training engineers, developing suppliers — and what it did not: the capacity to design, to own technology, to govern an industry’s direction. Indonesia crossed the first threshold; the arrangement never demanded the second.^[2]

What a firm reaches for is set by what it is compelled to attempt, not by what it is capable of building. The compulsion arrives from buyers who will take their orders elsewhere, from a state that will withdraw its support, from competitors who will displace it.

None of those demands reached Astra on the question that mattered: whether it would ever design what it assembled, not because market competition was absent, but because the arrangement’s protection meant Astra’s prosperity did not depend on responding to it. That is not a separate failing on top of the coalition and the oil. That is the same pattern, visible at the level of a single firm; and the constraint was in the arrangement, not in Astra.

The deepest cost of building one kind of competence is that it quietly removes the pressure to build the next, in Indonesia’s case, the competence at making things the world will pay for.

The reckoning the ratchet built

In 1997 the lock finally jammed. Indonesia’s government had let the banking system swell through the boom years, much of it lending to the conglomerates that owned the banks, on the same logic that let everything else off the hook, the rot that the regional contagion would expose rather than create.

Indonesia’s currency collapsed. Indonesia’s economy shrank by around thirteen per cent in a single year, and dozens of banks failed. And Suharto, after thirty-one years in power, was gone within months.^[3]

Indonesia’s banking system failed more completely than any of its neighbours’, and the ratchet, the mechanism by which every absorbed crisis raised the minimum shock required to force the next correction, explains the specific reason why. The connected conglomerates had borrowed abroad as freely as they borrowed at home, in dollars, against no requirement that what they built could ever repay it. So when the rupiah fell, the firms that had never had to be good were insolvent all at once.

That is not the whole story of 1997, which also runs through political panic over the succession, a botched international rescue, the exchange-rate regime, and sheer contagion. But the firm-level rot was the less examined part. The ratchet explains why no smaller, survivable shock had forced the choice first, not why the crisis, when it came, was as large as it was.

A government disciplines its own base only when a crisis makes inaction more costly than acting — and only when the crisis is still survivable enough to leave a government standing to act on it. Correction rather than collapse: big enough to force the hard choice, small enough to make it through. The ratchet ensured that was the one size of crisis Indonesia never had to face.

The one shock it could not absorb was past that line by the time it arrived, not because 1997 was unusually large, but because the country’s skill had consumed every correction-sized crisis before it. In 1997, catastrophe reformed nothing. It cleared the board.

Subscribe now

The lock outlived the locksmith

The story should end in 1998, with the regime that built the trap swept away. It does not.

Old machinery came back in new clothes. There is a vast new sovereign wealth fund, Danantara, to pick winners and channel capital, its architecture as familiar as its ambition is loud. There is a drive to process the country’s own nickel ore rather than export it raw, announced as the start of an industrial transformation.

And the politics that ties them: Joko Widodo, who rose from outside the elite and governed by drawing it in rather than making any of it fail. He handed power to Prabowo Subianto, the general he had once run against.

Those the old arrangement had made powerful carried their assets, networks, and positions into the new one. Prabowo had not arrived from outside the settlement; he was Suharto’s former son-in-law and a Special Forces commander through the regime’s final years. He spent a quarter-century reconstituting his political position across the transition. Major conglomerates lost assets during the post-1997 restructuring and rebuilt most of them.

The political superstructure was remade, and something real changed with it: the current coalition builds infrastructure and processes minerals, roads, refineries, downstreaming. That is developmental ambition of a kind the Suharto era did not systematically pursue.^[4]

Masina argues that many elite actors now have a genuine material interest in this upgrading; their returns depend on infrastructure and industrial investment, not merely on rent extraction. That shift is real.

Nor is this a question of political form. The countries that made the transformation leap ran under equally authoritarian regimes, as Korea’s case showed, and what set them apart was the firm-level test. Democracy does not supply the test either: any government needs political backing, and in Indonesia that backing rested on the same firms a productivity test would put at risk.

Nickel processing moves the country from digging ore to refining it, what Masina calls green re-commodification: a genuine structural crossroads between technological command and a reorganised commodity role. That history suggests the commodity path. Indonesia remains positioned at the commodity tier of a global value chain whose higher rungs — battery chemistry, cell manufacturing, vehicle design — are controlled elsewhere. One more route to growing rich without learning to climb.

Astra’s response to the electric vehicle transition tells the same story at the firm level: distributing Chinese-made EVs rather than developing proprietary technology; the assembler posture migrated intact to the new industrial order.

Through the 2000s, a long commodity boom carried the recovery back to respectable growth; the economy could rebound without ever being held to that standard.

The distinction is between upgrading that builds scale and upgrading that builds capability: the current coalition is doing the first; nothing in the arrangement has yet made the second unavoidable.

A governing orientation — what a state is built to reward, including at the level of the firm — does not live in the regime. It lives in the settlement. Changing just the regime does not reset it. That enduring settlement did the work. That regime was the surface.

What competence cannot buy

The competence trap is invisible at the level of macro indicators. A country managing its budget, absorbing its crises, and growing looks like a success — and on those measures, it is.

The signal is elsewhere: in whether firms in the protected core are being required to get better, and in whether the arrangement is building the pressure that would force the harder capability or steadily eliminating it. Fiscal prudence and structural transformation are different exams; passing the first tells you almost nothing about whether a country is passing the second.

The standard response to economies in difficulty, stabilise the macro, build institutional capacity, improve governance, addresses the supply side of state capability. What Indonesia illustrates is a demand-side failure: not a state lacking competence, but one applying it in a direction that permanently removed the pressure on its firms. A macro rescue that delivers stability without changing what firms are required to attempt can strengthen a competence trap rather than break it: the fiscal buffer buys more years of deferred reckoning, the correction threshold climbs. Stabilisation can be the ratchet’s best friend.

The warning is not about weak states. It is about strong states aimed at the wrong target — and about how invisible that distinction is in the data that gets watched. A state can be skilful, prudent, praised, and trapped by all three at once. A state can be competent enough to fail.

Subscribe now

Indonesia Spent Thirty Years Arguing About the Wrong Thing

Stephen Brien — Mon, 22 Jun 2026 07:30:50 GMT

Indonesia halted 600 per cent inflation, rode two oil booms without ruin, fed itself, and pulled most of its people out of poverty.
It wanted more than growth. It set out to become a different kind of economy, and paid real prices to try.
It never arrived. No Indonesian-owned firm became a global technology leader the way Samsung or TSMC did.
Its reformers and planners argued over which road would get it there.
This is about the one thing that sat unchanged under both roads, and why the argument decided nothing.

Indonesia is one of the thirteen economies the World Bank singled out for growing fast enough and for long enough to make its shortlist of sustained success. For most of that run, until 1997, it outgrew its peers. Yet it never joined the smaller group that makes the final leap to becoming rich. has provided a more contemporary analysis of the current challenges the country’s economy faces. This post looks back at what Indonesia’s state was built to reward, and what its ruling coalition was actually optimising for during its earlier growth period.

In the mid-1960s, with prices rising at around 600 per cent a year and the economy in ruins, Indonesia’s new ruler did something most strongmen never do. Suharto handed the wreckage to a circle of academic economists, most of them trained at Berkeley.

People called them the Berkeley Mafia, and within a few years inflation was in single figures and the foreign lenders had been talked back to the table.1

Then came the oil booms of the 1970s, the windfall that wrecked Nigeria and Venezuela.2 It did not wreck Indonesia. In 1978, with no crisis forcing its hand and the oil still flowing, the government deliberately devalued its currency to keep its farms and factories competitive, the prudent move almost no government makes in the middle of a boom. The money went into rice, roads and schools. Indonesia fed itself by 1984. Poverty fell from around 60 per cent of the population in 1970 to about 11 per cent by 1996.3

By the standards of its time, Indonesia did almost everything a poor country is told to do, and did it well. And it still never made the leap to rich. It rose to middle income and stopped there. No Indonesian-owned firm ever reached the technology frontier the way a Samsung or a TSMC did.4 That is the puzzle. Not a country that failed for want of competence, but one that had it in full and stalled anyway.

The country that got it right

Getting it right meant more than running a tidy economy. Plenty of countries manage that and never become rich, and nobody is surprised. What makes Indonesia a genuine puzzle is that it wanted the leap, and paid for it.

Suharto called himself the Father of Development and meant it as a programme, with five-year plans that set targets for industry, not just output. Its planners argued that cheap-labour growth was a trap that would move on the day wages rose somewhere cheaper, so the country had to master real industry or stay poor for good. That was its declared developmental intent, and Indonesia had it in full.

It backed that intent with real political prices. The austerity that broke inflation and the deliberate devaluation in a year of plenty were costly commitments, paid before any reward was in sight, the kind nobody pays unless they mean it. Both were paid to steady the nation’s books, not to make a single firm do well.

And the oil was doing something quieter at the same time. It did not just pay for the rice, the roads and the balanced books. It was also buying the loyalty of the people who kept the regime in power, and buying it with no strings attached.

So the explanation does not appear to be that something was missing. The intent was real, the prices were paid, and the oil meant the money was there too. Indonesia had everything a country is told it needs. If the answer is not in what it lacked, it has to be in what it did with all of it: what its state was built to reward, and what it was content to leave alone. It disciplined its budget without mercy, and its firms not at all.

Indonesia against itself

Indonesia did not pick one development road and hold to it. Over thirty years, as oil prices rose and fell, it ran both rival models the field still argues about. When the oil money was thin, and Jakarta needed its creditors back, the economists won. They took the free-market road: free the banks, deregulate trade, let cheap labour attract foreign factories, take a modest place in the world market, and grow rich slowly.

When the oil flowed, and the state could afford its ambitions, the engineers won.

Their man was B. J. Habibie, a German-trained aeronautical engineer and a Suharto favourite. Their road was a matter of national pride: skip the cheap stuff, build the hard, high-technology industries, make your own planes and steel.5 When the two camps clashed in the open, Suharto sided with the engineers, backing Habibie against the economists’ loud, public objection that plane-building made no economic sense.

The usual way to explain a country that climbs and then stops short is to set it beside another that did not, and name what the second had that the first lacked. The natural comparison is Korea, which started just as poor and made the leap Indonesia never did. But two countries differ in a thousand ways, and any of them could be the reason one pulled ahead.

Indonesia narrows the guesswork. It ran both roads on itself, the policy changing while the institutions, the ruling coalition and the resource base all held still underneath, which is about as close as development comes to a natural experiment in whether the road is what decides.

The whole debate assumes the two roads lead to different places. In Indonesia, they led to the same outcome. Under both, the economy climbed to the same middle-income plateau and then stopped. If a thing can change while the result stays the same, it was not the thing deciding the result.

The road changed. The plateau did not. So the road was not the cause.

It is not a clean laboratory. Neither road was ever run as an economy-wide programme; the reforms reached the open sector while the favoured firms sat in their protected niches, often at the same time. So it does not isolate a cause the way a real experiment would.

But what it does show is narrower, and it holds whichever camp held the pen. One set of firms stayed exactly where it was, untouched, and the plateau never moved with the road.

One distinction does the work here. There are two very different disciplines in this story. One is discipline over the books, holding down inflation, debt and the budget, which Indonesia ran superbly. The other is discipline over firms, making the companies meant to carry the economy improve or fail. That second kind is not something only a strong state can impose; an open market delivers it too. So the question is not simply state versus market. It is whether anything, by any route, ever forced a leading Indonesian firm to get better or get out.

The same test, withheld two ways

What separates the economies that arrive from those that stall is not ambition, nor money for champions. It is a test that a leading firm can actually fail, and be made to pay for failing.

Two things can impose it. A market imposes it from the outside by allowing a firm that cannot compete to lose customers and go under. A state imposes it from the inside, through the machinery that carries a decision from the top down to the single firm, its governing circuit. It backs a firm, then pulls the backing the moment the firm underperforms. Either way, something has to be willing to let the weak firm lose.

Indonesia’s two roads look like opposites. The market picks the winners, or the state does. But beneath the policy, the governing circuit never changed. Whichever road the country was on, the same machinery sorted its firms the same way, and it kept the connected core, the big conglomerates tied to the people in power, beyond the reach of any condition at all.

The free-market road withheld the test one way. The reformers freed trade and finance, and the market’s discipline did reach the open sector, the part of the economy exposed to foreign competition. The firms there were made to compete, and many did well. But that sector was mostly foreign-owned, so it improved firms that were not Indonesian. It was real, but the company it sharpened belonged to someone abroad.

The connected core sat in protected niches the circuit never opened to that competition. A market can only discipline the firms it can reach, and these the state kept out of reach.

The build-champions road withheld the test the other way. Here the circuit was generous, backing its chosen firms lavishly. But it attached no condition any of them could fail. It gave, and never once threatened to take back.

Whichever camp held the pen, the connected core never faced a condition it could fail.

It shows up inside a single firm. The Salim group was the largest of the connected conglomerates.6 When the economists freed finance, its bank gorged on cheap credit. When the engineers built champions, its other arms drew protected rents, the easy profits a sheltered position throws off. Two different roads, two different sources of advantage, and under neither did any part of it ever face a test it could fail.

Salim was not a weak firm being carried. It was an exceptional one, superbly good at the only thing the system actually rewarded: staying close to power and turning that closeness into cheap money and protected markets. A firm like that did not need fixing. It was already excellent, at the wrong thing.

This is what the missing test looks like when it is present. South Korea, in the same years, made even its largest firms export or die: meet a hard external standard or lose the support.7

Korea’s version was neither clean nor permanent. The same directed credit left its big firms dangerously over-borrowed by 1997, and the export test was one part of a longer recipe that ran through land reform and a tightly held financial system. The point that matters here is narrower and survives all of that: a leading Korean firm faced an external standard it could actually fail to meet, and the leading Indonesian ones, on either road, never did.

The machine worked

The objection to meet is that it could not. Disciplining firms is hard, harder than managing a budget. Perhaps the governing circuit, for all its reach, was never strong enough to make a favoured firm sink or swim, and “never built it” is a polite way of saying “was never able to.”

The answer is a single date. In 1975, the company that ran Indonesia’s oil nearly took the country down with it.

Pertamina was run by a general, Ibnu Sutowo, who, through the early 1970s, had turned it into an empire of hotels, steel mills and a rice estate, borrowing quietly to do so.

In February 1975 it could not repay a short-term loan, and the hidden total surfaced: around US$10.5 billion, close to a third of the entire economy.8 The economists running the budget did not flinch. They took the debt onto the state’s books, broke up the empire, and removed Sutowo.

That is not a weak state. That is a government reaching down and culling its single largest champion, fast, to protect the nation’s finances. That is the governing circuit at work, the machine that runs from the top of the state down to a single firm. It plainly existed, and it could reach the very top.

The question is not whether Indonesia could run a test. It did run one, and ran it well. The question is what the condition was for. And here is what the Pertamina rescue quietly reveals. The circuit selected on solvency, never on productivity. Sutowo was removed for threatening the nation’s finances, not for failing to build anything good.

It was aimed elsewhere

Aimed at productivity, the same circuit would have produced the opposite of what Indonesia did. IPTN, the state aircraft company Habibie built, is the inverse case.

Over its life, the state poured billions into it, with no clean year-by-year total to pin down. It never had to win over a real customer; its planes went at a loss to state airlines that had little choice. When money ran short for its flagship jet in 1994, the government quietly moved US$185 million out of a reforestation fund to keep it going.9 The engineering was rarely the problem; turning it into an aircraft a buyer would actually choose was the condition it had to meet.

A firm that threatened the budget got the Pertamina treatment. A firm that lost money for decades building planes nobody bought got another grant. The list runs on. The same decade handed a presidential son a clove monopoly, gave a Suharto friend the entire plywood export trade, and built a national car that turned out to be a rebadged Korean Kia. Each was sheltered. None was ever asked to earn the shelter.

The pattern has a precise, testable form. Across the New Order, Suharto’s three decades in power, I can find no connected-core firm that was ever cut off from state support for failing to compete.10 The open sector is another matter: there, foreign-owned exporters were disciplined all the time, by their buyers.

To break the claim, you would need a single, connected champion whose protection (its tariff wall, captive procurement, directed credit, or monopoly licence) was stripped away for missing its targets while it was still solvent. A firm cut loose because it was going broke, or because its patron fell from favour, does not count; that is solvency or palace politics, not whether the firm was any good.

Why the circuit was aimed at solvency

If the state could run the governing circuit and aim it at solvency, the question becomes: why? The answer is not a mistake or an oversight. It is interest.

Suharto’s rule rested on keeping a particular set of people loyal: the generals, the Chinese-Indonesian business families who financed the regime, and in time his own children. They were the coalition, the base whose backing kept him in power, and the bargain that held them together was the country’s political settlement: the left destroyed, the unions broken, the press tamed, the army inside the government, and the technocrats kept in a protected circle of their own.11

The favoured firms were not bystanders to that coalition. They were it. So a condition that could make a connected firm fail was not a neutral tool the state had neglected to build. It was a weapon, and it could point only at the regime's own base.

That is why the regime enforced one discipline ferociously and refused to enforce the other. The macro kind defended that base. The whole arrangement ran on the state’s solvency: the subsidies, monopolies and cheap loans the coalition lived on all flowed from one pool, the oil money and the public budget, and a bankrupt state has nothing to pay out. So when Pertamina’s borrowing threatened to drain that pool dry, removing Sutowo did not attack the coalition. It defended it.

A productivity condition cut the other way. It would have done nothing to protect the pool and would have been taken directly from a single favoured firm. Worse, it would have changed what a place in the coalition was worth. That place was a guaranteed return that never depended on being any good; a test a firm can actually fail makes it conditional: perform or lose the support, which is exactly what Korea did when it told even its biggest firms to export or die.

Held to that standard, any Indonesian insider could be cut. The bargain that bound the coalition promised security; a productivity test offered risk, for an economy-wide gain the coalition did not capture. A government does not build the one instrument that would break it. That is Mushtaq Khan’s point about political settlements, seen on the factory floor: a ruling coalition can sustain only the disciplines its own survival permits.

So “could not” is the wrong answer.

The circuit could reach the top, as Pertamina showed, and the nerve to use it was there. What stopped a productivity test was not a missing skill but a standing interest. Turning that instrument on a connected firm would have starved the coalition that was the regime. Whether Indonesia could also have mastered the harder, different craft of judging firms on what they made is a question it never had reason to ask.

The rules relaxed for the connected

The windfall the technocrats managed so well was also what paid for the coalition’s loyalty, and did so without conditions. The regime could keep its base content with shelter and rents straight out of the oil account, so it never had to make those firms productive in order to afford them. That is a quieter curse than the one that wrecked Nigeria. The oil did not bloat the currency and choke off industry here; it bought the regime the freedom never to discipline the industry it had.

The kind of regime Suharto ran is not the difference; it’s worth comparing with two states of the same kind. South Korea and Taiwan were no gentler: all three were hard authoritarian states that jailed opponents and tolerated little dissent.12 The difference between them and Indonesia was never how hard the state could be. It was on whom it was willing to be hard. Korea was willing to be hard on its biggest firms.

Indonesia was willing to be hard on anything except them.

The question usually asked of a whole country can be put to a single firm. Not whether a nation’s institutions are open or closed, but whether one connected firm inside it could ever be allowed to fail for being no good at what it did. In Indonesia, over thirty years and on both development paths, the answer was no. Not because the state lacked the nerve to enforce it, but because the firm that would fail such a test was, in the end, the regime itself.

What a state is built to reward

Indonesia had every piece a developmental state is supposed to have: the declared intent to build industry, the costly commitments paid to back it, and a governing circuit that plainly worked, the machinery that culled Pertamina in 1975. Nothing was missing.

The trouble was that all of it pointed one way. The intent was loudest in the boom years, exactly when champions were handed support with no condition attached; the commitments were paid to steady the nation’s books; and the circuit, when it reached down, reached for solvency. Every piece was aimed at the nation’s finances.

All three stopped at the same line, and for the same reason. A firm-level condition would have fallen on the connected core, which was the coalition. So the intent could be declared but never enacted, the commitments paid but never to make a single firm improve, the circuit built but never aimed at whether a firm was any good, because all three, turned that way, would have broken the base the regime stood on.

And the arrangement renewed itself. Sheltering the core was exactly what kept the core loyal, so every year of protection made the next year’s protection more necessary, not less. The thing that made it impossible was the same thing it would have destroyed.

And no force inside the system pushed the other way, either. In the countries that broke open, it was a homegrown export sector that did the breaking: firms that had passed the world’s test and grown strong on it eventually had the weight, and the reason, to force the closed arrangement open and demand the same rules for everyone.

Indonesia got the boom but not the bloc. After the mid-1980s reforms, a real export industry emerged: garments, footwear, and electronics. By the early 1990s, manufactured goods had risen from around 2 per cent of the country’s exports to about half. But it ran on foreign factories chasing cheap labour, not on firms Indonesia owned, and its momentum faded before any home-grown champion grew strong enough to demand anything. The force that elsewhere cracks a closed system from the inside never formed.13

Three things, then, and one direction. The governing circuit, the machine that runs from the palace to the factory floor, has a reach and a direction. Its reach was never in doubt; Pertamina settled that. Its direction had a name. The country’s governing orientation is not what a government wants but what it actually rewards and refuses to rescue, read off the cases rather than the speeches.

A country gets the economy it honours, and Indonesia honoured the men close to the palace and the symbols of a modern nation, never the unglamorous business of being better this year than last. The closer a firm sat to power, the safer it was from ever having to improve.

The settlement on the factory floor

The theory of political settlements already explains a ruling coalition can sustain only the disciplines its survival permits. What the case adds is two things that theory describes but does not supply. It gives a way to see that constraint, and a way to test it.

The first is a way to see it. Governing orientation is the firm-level signature of the settlement, read off what the state rewards and refuses to rescue rather than off the distribution of power above it. The constraint stops being something you infer from politics and becomes something you can read on the factory floor.

The second is a way to test it. With the settlement, the coalition and the resource base held fixed and only the policy road varying, Indonesia’s own swing back and forth shows the orientation sitting underneath the road, not set by it. Between them, the two moves make the settlement’s grip observable and testable rather than merely asserted.

A more capable state, the objection runs, could have done both, sheltered its coalition and still disciplined its firms. It could not. A state strong enough to cull a failing firm is strong enough to shelter a favoured one, and Suharto’s used that strength to shelter. The discipline that actually transforms an economy is the kind no government can quietly switch off, the kind an open market imposes and no palace phone call can lift.

The call did not need to lift it. Its work came earlier, in keeping the connected core out of the open market at all, so the market’s discipline never reached the firms that mattered. To shelter a firm was to take it out of the one discipline that would have transformed it, so no state, however capable, could do both at once.

In Indonesia, almost nothing was ever beyond the reach of that phone call.

The argument that missed the point

The famous thirty-year argument between Indonesia’s free-marketeers and its industrial planners was an argument about the wrong thing. Both sides were fighting over which road to take. Neither was fighting over the only question that separates the countries that arrive: whether any firm on either road could ever be made to fail. The road was never the variable. What the state was built to reward was.

And the lesson is not that Indonesia lacked discipline. The intent to build industry turned into one condition, the nation’s ability to pay its bills, never a condition that a company be any good at what it made. That translation, what a state’s stated purpose becomes in the things it actually rewards, is its governing orientation. Discipline is not a quantity a state has more or less of; it is an aim, and Indonesia’s was fixed on solvency, never on productivity.

This is also where the story’s one ugly debt has to be named. The freedom that let Suharto’s economists act so decisively rested on the mass killings of 1965 and 1966 and the silencing of everyone who might have objected. It is a set-aside cost, but it should not be buried.

The diagnosis leaves a harder question behind it. A state this competent could have aimed the circuit at productivity, and chose not to, in the only sense that matters, because the firms it would have broken were the base the regime stood on. To turn it on them, Suharto would have had to remake that base, the settlement his own power rested on.

That is exactly what the leaders who transformed their countries did. Park Chung-hee reset the terms by which Korea’s big firms operated when he made them export or die. It is the choice a transformation demands, and the one Suharto would not make.

Why a country that managed everything else so well never made it, and what the gap finally cost when the reckoning came, is the rest of the story.

Reform Windows Are Made, Not Found

Stephen Brien — Fri, 19 Jun 2026 07:31:29 GMT

The short version:

Reformers are taught to wait for a crisis to open a window.
Most crises open nothing, and plenty of reform happens with no crisis at all.
The window is not the event. It is the moment someone treats as an opening.
That moment can be generated: stake a government on a referendum, invite a creditor to impose the cuts it cannot, or move before a collapsed state sets.
But manufactured windows are bounded and often fail. Making one is a wager, not a method.

The second of the “How a Rule Changes” posts. The companion post asked whether a wall a reformer treats as fixed is real; this one asks where the moment to lean on it comes from, and what a self-made window still cannot give

In 1967, Botswana was among the poorest countries on earth. There was no crisis, no emergency forcing anyone’s hand. The country was simply poor.

That is the year its government quietly settled who would own the diamonds.

Geologists had found a diamond pipe at Orapa, in the heartland of the Bamangwato, the tribe of the president, Seretse Khama. Under the old order, mineral rights belonged to the tribes. Across 1967, the tribes signed those rights over to the national state, one by one.

Khama’s own tribe signed first, in April; the rest followed in July, on identical terms, the rights passing to the state “without compensation.”1

No door swung open here. There was nothing to climb through. Khama moved the boundary while the prize was still a rumour, and by the time the diamonds were real the question of who owned them had long been closed, the fight that might have split the country quietly avoided. He made the opening himself.

The opening we are taught to wait for

We do not usually picture reform that way. We picture a window of opportunity, opening from the outside. An old order fails. A crisis suspends the rules. A reformer climbs through before the gap shuts.

There is good reason for the picture. The economist Leszek Balcerowicz called these moments extraordinary politics, the brief spell after a collapse when shared crisis and public trust let a prepared reformer push through what normal politics would block.2 Ireland in 1958 is the type. Forty-five thousand people were leaving every year, the population had fallen to 2.8 million, and the sense had set in that the whole model had failed. A country failing on its own terms is what cracks open a settlement, the standing bargain over who gets what.3

Sometimes the crisis is sharper. By mid-1991 India had reserves down to a few weeks of imports and was flying gold to London and Zurich to stay solvent; out of that emergency came the reforms that remade the economy.4

So the story holds, as far as it goes. Crises do open windows.

But the crisis is not the window

It does not go far, and it fails in two places at once.

The first is familiar. “Never waste a crisis,” the political adage runs; the trouble is that most crises are wasted anyway. The loan arrives, the emergency passes, and the settlement that produced it is left standing. Nigeria turned a decade of oil windfalls into a debt crisis and diversified into nothing; Venezuela spent a bigger one a generation later and collapsed when the price fell. The rescue refinances the very arrangement that needed rescuing, which is most of the story of crises that lead nowhere.5

The second is less familiar and more damaging. Reform happens with no crisis at all. Australia floated its dollar in 1983 and cut its tariff wall through the decade, none of it forced by an emergency. Colombia, the one major Latin American economy to sidestep the debt crisis of the 1980s, threw open its trade and financial markets under César Gaviria in 1990, with no collapse to compel it.

And economists who have gone back over the record find the supposed link between crisis and reform surprisingly weak. When reform fails to follow a crisis, the theory simply says the crisis was not bad enough, which makes it almost impossible to disprove.6 The hypothesis is more widely held than the evidence for it warrants.

Put the two together and the picture inverts. If most crises open nothing, and reform can come without one, the crisis was never the window. Ireland used 1958; a dozen countries had a year just as bad and used none of it. Khama had no crisis and opened a window anyway.

Peter the Great did not wait for one either. He made his window by coming home. A year abroad had shown him what a modern state could be, and he returned to Moscow in 1698 treating Russia’s backwardness, the permanent condition everyone else endured, as an opening rather than a fact of life. He began remaking the country by decree at once.

But decree reached only so far. What carried the rest of Russia with him was the war that followed. When the Swedes destroyed his army at Narva in 1700, backwardness stopped being a tolerable habit and became a death warrant, and reform turned from the tsar’s whim into the country’s survival.

The window is not the event but what someone makes of it. No crisis or collapse is a window until a reformer treats it as one and moves through. That is why a window can be made where nothing has opened at all. That it needs someone to act is the easy half; the harder, rarer half is making the opening last.

Subscribe now

Making the moment

Once you see that, you see that the moment can be made.

Albert Hirschman named the craft half a century ago. He called it reform-mongering: the unglamorous art of working a resistant system, timing a push, building odd-bedfellow alliances, turning disorder into pressure.7 Hirschman’s reformer mostly took the openings that came. The harder move is to manufacture one.

The plainest way is to go and get a mandate. In early 1992, F. W. de Klerk was negotiating the end of apartheid and losing the support of his own side; his party had just lost a by-election to the right, and his white electorate was turning against the talks.

So he called a whites-only referendum on a single question: should the reform process continue? He staked his government on the answer, and he could have lost.

He won, 68.7 per cent on an 85 per cent turnout, and used the scale of it as licence to do what his side had feared. He manufactured the conditions for a mandate. (Though the larger opening behind him was lit elsewhere. The Berlin Wall had just fallen, and Soviet backing for his opponents had collapsed, and he did not make that shock.)

There is a bolder version, and from outside it looks like no move at all. A government can import the pressure. When a government cannot push a painful reform through at home, it can invite someone outside to force it, a lender whose conditions it privately wants, so the loss arrives wearing a label that reads “we had no choice.”

In December 1990, the new president of Uruguay, Luis Alberto Lacalle, wanted to cut spending and privatise one of the world’s oldest welfare states, but his own coalition would not let him. So he signed a programme with the International Monetary Fund.8

He did not need the money. The balance of payments was in surplus, reserves covered nearly eight months of imports, and the government drew less than a tenth of the loan. It did not want the loan. It wanted the conditions, so that a vote against the programme became a vote against the Fund.

But made windows fail

So Lacalle made his window. Then he lost it. Within two years, a referendum repealed the privatisation law by nearly three-quarters of the vote. He had manufactured the moment, and the public slammed it shut.

Greece shows the same pattern, more starkly. By 2015 the country was surviving on bailout loans from the other eurozone governments and the European Central Bank, and the price of each loan was deeper austerity. Alexis Tsipras did not want to sign. So in July he put the creditors’ terms to a snap referendum and urged Greeks to reject them, betting that a clear No would strengthen his hand when he went back to negotiate. Greeks gave him his No, 61 per cent of it.

It bought him nothing. The creditors held a harder lever than any mandate. They had already let the banks close and frozen the emergency funds that kept the cash machines working, and within a week Tsipras signed a bailout harsher than the one the vote had rejected.9

These are the made windows we can see. The ones we cannot see are the larger number of leaders who manufactured a moment and were simply crushed or reversed, and left no trace. For every Khama there are many of them. That is the discipline the whole idea needs. Manufacturing a window is a wager, not a method, and few of the windows so made ever last. You can open one and still lose it.

The widest window

The widest opening of all comes when a state collapses, and nothing has set in its place. Collapse shows two things at once. The move inside the window is what matters, and what you fail to lock becomes the next wall.

When an old order falls, its assets do not vanish. Someone is holding them: the directors in their factories, the managers in the ministries, the people who ran the state’s property the day before the state stopped existing. The first question the new state faces is how much of the country its writ can override, and how much it must leave in the hands that already hold it.

Estonia and its neighbours faced that question in the same few years and answered it differently. In 1991, the new government in Estonia restored property to the families who had owned it before 1940.

Mart Laar’s coalition then sold the large enterprises to outside investors rather than their directors, and let the loss-making Soviet plants close. Each move was the state insisting that it, not the man in possession, decided who owned the country. The dispossessed fought back, through the Russian-speaking towns of the north-east and through Moscow. The state held.10

The new states in Russia and Ukraine responded differently. They confirmed the men in possession. Soviet assets passed to insiders, in Russia through the rigged loans-for-shares auctions of 1995, and the possessors became the country’s real power. When the Ukrainian government tried in 2005 to claw back a single privatised steel mill and resell it, it recovered nearly six times the insiders’ price. But by then the possessors were strong enough to bury the wider effort, and the prime minister who led it was gone within months.

The window was the same. What each state did inside it was not. And the part Russia did not do is the part that lasts. The oligarchy it let form in the open moment hardened into a wall of its own, and the strongman who eventually broke it did so to take its place, not to free what it had captured. A window not used closes into a wall.

What the moment cannot give you

A window of opportunity is not a date on the calendar. It is a moment that someone decides to treat as one. It can be handed to you by a crisis, though most who receive it waste it. It can be inherited at a founding, if you build before the opposition forms. Or it can be made by escalating a mandate, summoning an outside enforcer, or catching a state before it sets.

Where the opening comes from matters far less than whether anyone moves through it. You cannot conjure one from nothing, and the making fails as often as it works.

And even the supply is not the reformer’s to write. India’s reformers in 1991 did not summon their crisis; it was handed to them by a war in the Gulf, which spiked the price of the oil they had to import and cut off the remittances their workers sent home.

What the crisis could not hand them was the plan, which they had prepared in advance. A blueprint for liberalisation, the M document drafted by Montek Singh Ahluwalia, had been circulating in the government since 1990, so when the window opened they could move at once.

Whether a creditor’s programme is there to be courted, whether a collapse arrives to be used, lies beyond any single country’s control. It turns on the price of commodities, the price of money, the end of someone else’s empire. But not every opening waits on those forces. Khama waited for none of them; he made his opening with no crisis to hand him one. The reformer governs what is made of the moment, and sometimes, as in 1967, whether there is a moment at all.

Subscribe now

What the reformer can still do

What separates a reformer who moves a settlement from one who only describes it is a short and unforgiving discipline. Most of it runs counter to the reflex to wait.

They do not wait for a crisis. Most crises change nothing, and the one severe enough to force change is usually too destructive to use well; reform reaches countries in calm water more often than the crisis story admits.

They read the moment, not the calendar. The cheapest move is to see an opening where everyone else sees a settled fact, which is the timing twin of the question the companion post asked about walls, not whether the limit is real but whether the moment is already open.

They strike while it is cheap. The cost of a fight climbs with the stakes, so the move comes against a protected interest before the prize has a price and its defenders have organised. That, not luck, was Khama’s edge in 1967.

And they move fast, because the opening shuts. Balcerowicz gave the post-collapse window a year, maybe two, before ordinary politics returns. Whatever is done has to be done while it is open.

When they manufacture a window, they treat it as a wager and nothing more. The escalated mandate and the summoned enforcer are real moves, but they failed Tsipras, and they failed Lacalle, and no reform should rest on the manufactured window holding.

The last move is the hardest and lies beyond this post because it is barely about the window at all. It is to make the reform hold before the window closes. The failures have something in common. None of them failed for want of an opening. Lacalle found one, and his law was repealed; Tsipras found one and lost it in a week; Russia had the widest opening of the century and let it set into an oligarchy. The window was never the hard part. Holding what you win in it is.

What one leader forces through in a single open moment is not yet what a country believes, and whether it can be made to hold is the next question, and the harder one. The companion to this post began with walls that look fixed and are not. A wall, it turns out, is often only a window that someone failed to keep open.

The Wall Nobody Has Tested

Stephen Brien — Wed, 17 Jun 2026 16:37:41 GMT

The précis:

Everyone agrees on the reform that would lift growth, and everyone agrees the politics forbids it.
But “politically impossible” usually means a wall no one has tested, only a guess about who would resist.
The explorers we remember are mostly the ones who guessed right; the many who were wrong vanished without a name.

This post belongs to a broader inquiry into why economic transformation is so rare: why a few governments break the bargains that constrain their economies, while most never try. Those bargains, the political settlements that decide who wins and who loses, are often treated as fixed. It asks what it takes to move them, beginning with the first question: which of those limits are real, and which only appear to be.

In November 1978, in a village called Xiaogang in eastern China, eighteen farmers signed a contract that could have had them shot.

It did something the state had banned for twenty years. It broke the village’s collective farm back into family plots, each household keeping whatever it grew above the government quota. This was private farming in a country that had abolished private farming. The farmers understood the danger, and they wrote it into the contract: if any of them were jailed or killed for what they were about to do, the others would raise his children until they came of age.

The fear in that clause is the place to start. It is not fear of a bad harvest. These were good farmers; they knew the family plots would out-produce the commune, because they had done so before the communes came. What they could not predict was what the Party would do once it found out.

The harvest came in larger than any in years, and the blow the farmers had braced for never landed. The local officials who were meant to punish them looked the other way. The provincial leadership quietly let the experiment stand. By 1982 the central government had endorsed household farming for the whole country, and within a year nearly every farming family in China was doing what those eighteen had been ready to die for.

The wall they expected to hit was not there.

The farmers made the first move, but it survived only because everyone above them who could have crushed it chose not to, each of them no longer sure the old rule could be enforced. Presidents and cabinets make the same gamble, with far more riding on it. Every so often a reform that everyone calls politically impossible simply happens. The puzzle is why.

Subscribe now

Why the map says impossible

The usual explanation for why a move like that should fail is simple enough. A country’s politics rests on a settlement: a standing bargain over which groups are protected and which are not. A reform that threatens the protected is the one everyone calls politically impossible, and the impossibility is treated as part of the terrain, a mountain range the road has to bend around.

Even so, deferring to those limits is usually sound. The economist Mushtaq Khan put the careful version of the advice plainly: build your reforms to fit the power you actually have, not the power a textbook prefers. Work with the grain. Most of the time he is right; the powerful really do strangle what is built against them.

The farmers in Xiaogang went straight at the grain. And the grain gave way.

Up close, a settlement looks like part of the country itself, and its limits look like permanent facts. Whether they truly are is the question.

The wall nobody has tested

Some are, some aren't. And the ones that matter are the easiest to miss.

Most limits are real, and everyone can see it. No government needs to test whether a ruling party will vote itself out of existence, or whether a landed elite will hand over its estates without a fight. The answer is plain in advance, so no one wastes themselves finding out.

Other limits only look that solid. No one knows how hard the resistance would really be, because no one has tried.

And the not-knowing runs both ways. The groups contending for power, as Mushtaq Khan puts it, are “unsure of their real holding power and that is why contests and conflicts happen.”1 In December 1989 Ceaușescu called a rally to show he still commanded Romania; the crowd jeered, and the authority he had wielded for a generation was gone within the week. A wall can look solid to the very people standing behind it.

Until someone leans on it, such a limit is only a guess about what would happen if they did, and the guess hardens the longer it goes untested. A rule nobody has broken in living memory has a reputation, not a record.

So not everything called politically impossible really is. Some of it is a guess mistaken for a measurement. Telling the difference asks for judgement before nerve, and the only way to be sure is to test it.

The leader who goes to find out

The same not-knowing that lets a wall stand is what lets it move.

Consider the eighteen farmers. The balance of power in that village can tell us their reform was possible. It cannot tell us why they were the ones to test it, or why then.2

There are two familiar ways a government comes to move against the powerful. Sometimes it misjudges the balance and is punished for the mistake; sometimes the balance has quietly shifted first, and the move only registers a change already underway. Call them the misjudged push and the late arrival. Xiaogang is neither. The farmers did not misread the danger; they wrote the risk of execution into their contract. And the commune was intact, the balance unshifted until they moved.

So there is a third way. A government can move against the powerful for a single reason, to find out whether the limit is real. Governments have done this again and again, though the move itself is rarely named for what it is.

This is a different kind of actor. Most leaders treat a hard limit as a fact and accept it; this one treats it as an open question and pushes to find out.3 Call it the explorer.

The figure is older than any reform ministry. When Caesar brought his legion across the Rubicon, the law prohibiting it was plain, and the penalty was death; what no one knew was whether the Republic could still enforce that law against a general willing to find out. It could not. Pompey and the Senate fled the city, and the wall that had held Rome’s commanders in check for centuries turned out to have been an assumption no one had tested.

Most explorers fail

The explorers we can name are mostly the ones who were right. De Klerk tested whether his own side would hold the line for apartheid; Gandhi gambled that the Raj could not enforce its laws against a whole nation’s defiance.

The ones who leaned on a wall that held, and were broken on it, mostly vanish from the record. The famous few we remember are the ones who lost loudly: Mossadegh, nationalising Iran’s oil against the British and the oil majors in 1951, gone in a coup by 1953; Allende, restructuring Chile against the propertied and the army, gone the same way in 1973. They stand in for a far larger number who tried smaller things and disappeared without a name.

Nor is the failure confined to coups and the developing world. Edward Heath fought the February 1974 election on a single question, whether Britain was governed by its elected government or by the coal unions; the unions’ power held, and he lost the vote and his office with it. In a settled democracy, the price of a failed test is a career rather than a life, but the wall is found just as real.

So any list of explorers flatters the type, and most who lean on a real limit are crushed by it.

It would be tidy to cast these explorers as cool experimenters, calmly probing the edge of the possible. Mostly they are not. What sets the explorer apart is a stubborn refusal to take a limit on trust, and a willingness to test it when everyone else accepts it.

Subscribe now

The signs of an explorer

The explorer finds where a settlement’s edges really are. It takes a limit that was just a guess and turns it into a measurement.

You can tell an explorer’s move while it is being made, before anyone knows whether the wall will hold. Two things show it.

The price is named in advance: the Xiaogang farmers wrote theirs into the clause about who would raise their children, and Heath staked his government on a single question. And the danger falls on the mover’s own side. That is what separates an explorer from an opportunist: the opportunist picks off a weakness everyone can already see, while the explorer moves where his own side believes the wall is real, as de Klerk did against the apartheid his white base wanted to keep.

There is a complication when the wall sits inside a government’s own coalition, where the cost of breaching it can be spread. With inflation running near 500 per cent in 1985, the limit everyone in Israel assumed untouchable was the Histadrut, the trade-union federation that was Labour’s own base. A national-unity government under Labour’s Shimon Peres tested it anyway, agreeing an emergency package that froze and cut real wages, and the wall gave.

But it gave partly because the cost was spread: the plan’s lead author was the coalition’s Likud finance minister, not Labour. A protected interest your own supporters defend is easier to break when a coalition partner shares the blame.4

Going in, the losers looked no different from the winners. Mossadegh and Allende staked everything on walls they could not know would hold, and the walls held all the same. What shows is the move, not how it ends.

What does not show is the purpose. What grows in the space the move opens is decided elsewhere, by the aim behind it. The same move can serve different ends. When Putin broke the oligarchs who had made and funded him, he tested a settlement as surely as Deng Xiaoping did in opening China. Both leaned on a limit no one was sure would hold, and both found it gave. One freed an economy; the other tightened his hold on a state.

The move is seldom just the protagonist’s. Great powers had a hand in more than one of these moves; the two who fell loudly fell to foreign hands as much as to domestic ones. The tailwind is real. But a push from outside still has to be taken up by someone inside who treats the limit as open and pays to find out where it lies; the patron supplies the wind, not the decision to sail into it.

And it belongs to no particular kind of person. Eighteen farmers can make it, and so can a president. But the result reaches far beyond whoever makes it: when a settlement’s lines move, everyone inside them lives with the outcome, and most of those people had no part in the contest at all.

The board is never set

A settlement can be read like a chess board. The pieces are the country’s organised interests, and the way they stand is the distribution of power a reformer inherits. A strong player glances at the board and says who is winning; the structural account does the same with a settlement, and calls the reform lost before it is tried.

But a forecast is not a result. A position that looks lost can hide a move no one has tried.

In 1984, the structural account would have called New Zealand’s economy beyond reform. It was among the most protected in the rich world, its farm subsidies and tariff walls treated as permanent features of the country.

Then a Labour finance minister, Roger Douglas, swept them away in a few years, and the interests that were supposed to stop him never did.5

This is the part the structural account leaves blank. It carefully explains what a given distribution of power makes possible. On who actually acts within that space it goes quiet, falling back on a vague phrase: political will.

The explorer is what that phrase points to: not a force from outside the struggle for power, but that struggle at the moment its outcome is still open. Which of a settlement’s possible outcomes arrives is not settled by the balance of power alone; it turns on whether someone moves to find out.

And a test does more than measure a limit. A test that holds moves it. The structural account reads a settlement; the move changes it. So among the outcomes the balance leaves open, agency is not a leftover the theory can set aside. It is the part that chooses.

None of this promises the move will work. Most do not; the explorers who fail far outnumber the ones who succeed. An open outcome does not imply a winnable one. What is always open is the chance to find out, to learn which limits are real and which are only assumed, at a price the mover accepts.

So the board is never simply set. A settlement that looks finished is usually one no one has tested lately, and the scholarship that maps the balance of power has underrated what a move against it can do.

The wider conditions do much of the loosening: a fiscal crisis strands an old model, an empire retreats, a technology rots the ground beneath a settled interest. But conditions only move the odds; they never reveal which walls would actually give. The wall looks fixed; whether it really is, only the test can say.

Playing the position

So far, the explorer can look like someone who walks up to a wall and shoves, and the ones we have watched mostly got crushed doing exactly that. But the head-on charge is the crude form of the move, not its definition. The skilled explorer does not charge; he tests obliquely, in moves too small to provoke the answer he fears. It is the same move, played well, and it is how most walls that fall are brought down. There are many ways to do it, and two are especially clear.

The first is to build a position, the way a strong chess player does. He rarely wins by snatching pieces; he makes his position unassailable move by move, until the loser sees the game is lost and concedes. A reformer does the same, improving his standing quietly so that by the time any real exchange comes it is already half-decided.

Rome learned this twice over. Caesar crossed his Rubicon and won the gamble, only to be stabbed for it within five years. His heir, Augustus, moved in small steps instead, each one too modest to provoke a challenge. He beat his rivals one at a time, handed the Republic back to the Senate while quietly keeping the armies, and took up its powers one by one, until the Republic was gone and no single act had marked its end.

The other is to change the board. When the main one is locked, build another beside it. Deng Xiaoping did not open China by confronting its central planners. He built Shenzhen, a zone where different rules ran, small enough to be tolerated and real enough that the people who prospered inside it became a lobby for more.

Bismarck used the same move against the socialists his ban had failed to crush. He built the world’s first modern welfare state beside them, insuring German workers against sickness and old age, to draw their loyalty toward the state that paid them rather than the movement that promised them. An interest you cannot beat head-on can sometimes be outflanked by a new one raised on open ground.

Where the walls come from

Every wall was once a question someone tested. A political settlement is not a permanent feature of a country; it was made. Each limit a reformer inherits as fixed is the place where an earlier contest was settled and left to harden. The board the structural account reads as given was built, move by move, by people who tested limits before them. Structure is not the opposite of agency; it is prior agency that has set.

Korea and the Philippines show it. The two began the post-war years alike, both poor, agrarian, and run by landowners, and then took different roads. The most commonly named difference is land. At the end of the 1940s, Korea broke the power of its landed elite through sweeping reforms, a generation before its industrial rise; the Philippines never did, and its great families stayed intact. That this set the two on different paths is well understood.6

What is easy to miss is that Korea’s reform was not handed down by history. It was an explorer’s move: a government testing whether it could break owners who could have stopped it. The Philippines never made that move, and reform after reform died in a congress controlled by its families.

The open ground Korea’s later reformers inherited, and the wall the Philippines kept striking, were both, in part, the residue of a single choice made decades before: to push against a settlement that protected landlords, or to leave it be.

So the wall one government treats as permanent is often just the last explorer’s victory, set in concrete, and it sets harder the longer no one leans on it, the way a reputation calcifies into a fact when it goes untested. What was made by moves can be remade by them. A wall is only a window that has closed. That is the deepest sense in which the board is never set.

What it costs to mean it

The move is a wager, and running it is never free. It is also what makes a larger wager possible. Development, as Stefan Dercon describes it, takes an elite staking its country’s future on growth,7 and that bet needs a settlement that someone has already moved.

Protected interests do not give way because they have been doubted. They give way, if they give way at all, because a price has been paid that they cannot match or undo.

The farmers of Xiaogang wrote their willingness to pay into the contract, in the clause about who would raise their children. Most explorers never have to honour a clause like that. Some do.

A price like that, named in advance and falling on the mover’s own side, is not one a leader who only wanted to keep power would pay. It is the surest sign the move is a test and not a calculation. What that price buys, and whether the gain holds once it is paid, is the harder question, and the one this series takes up later.

So the reformer’s first correction is to the map. Most of a settlement’s limits are real, and whoever forgets it is soon punished. But some are only assumptions, holding because the cost of finding out has kept anyone from trying. The discipline is to read “politically impossible” as a claim about what no one has yet tested, not a fact about the country.

That asks a particular cast of mind of anyone who means to move a settlement, not just describe one. The question that decides things was never whether a reform would work, but whether those who could stop it really could: a matter of power, not policy. A wall with a real record of stopping people is one thing; a wall with only a reputation is another, and only a test tells them apart. A long record is a guide, not a proof.

The walls worth the price are the ones a reformer’s own side believes are real, where the fear runs closest to home. None of it carries flattering odds. It is a way of acting, not a way of winning.

Testing the wall is only part of why transformation is rare. The move still needs a moment to make, which the companion post, The Window You Open Yourself, takes up, and a team to hold the new order once the wall is down, the ground of Why Africa Keeps Getting the Wrong Kind of Crisis. But the wall has to be found first.

Accept the whole settlement, and you keep the economy it was built to protect. But you have to accept only the real walls.

Subscribe now

Hidden Rules of Development

What Botswana’s Success Story Never Had to Explain

I: It earned every credential a success story needs

By any conventional measure, it succeeded

What secured property rights explain, and what they don’t

II: The number that says it never transformed

Productivity turned negative in 1990, and never recovered

Complexity and innovation rankings say the same thing

Neither Botswana’s size nor a passing resource shift explains the finding away

III: None of the ordinary explanations survive Botswana’s own record

The usual explanations, and why Botswana’s own literature already knows them

It had a plan, and the capacity to act on one

The windfall wasn’t squandered, and this isn’t a lack of nerve

IV: No wrong answer was ever allowed to cost anyone anything

Selebi-Phikwe was judged only by the people who paid for it

A declared goal, and five postures that moved past it

Nothing ever forced that gap into the open

The insulation is now being tested

What follows

Further reading

Japan's Lost Decade Was Twenty Years In The Making

I: What Japan built

The early priority, sector by sector

The rule that cut the other way

A founding conflict fought by someone else

A priority Tokyo chose to keep enforcing

The harder test

A commitment with the exit removed

II: How Japan turned

Three years welded into one budget

A ratification, not a start

What stayed exactly where it was

A closer look at rice

What kept it going

III: What Japan stopped checking

A stock, not a flow

Starting without stopping

A boundary, not a choice

Twenty years before the reckoning

IV: Why Japan never turned back

Past the point of undoing

A gesture, not yet a test

What holds together

V: What Japan proves, and doesn’t

What was revealed

What was overlooked

What slowdown pace theories missed

What one country can prove

Further reading

How a Government’s Real Priorities Get Revealed

I: What identical policies can hide

What the zones don’t settle

What a revealed priority means

II: What the terms make conditional

What a firm actually delivers

Whose access counts as dependable

What actually decides who prospers

III: Where the terms stop applying

How far each test was allowed to go

Why the boundary counts as a choice

What the threat actually targets

Brazil, and the capture problem

IV: What the government protects under pressure

When governing terms become costly

Did the claim matter to the government?

Did the government own the choice?

V: What the conflict was actually for

Was transformation the priority?

What the three checks add up to

Further reading

Notes

The Cost of Prioritising Transformation: What Sorted Eight Growing Economies

I: What a government is choosing

Declarations are cheap

What shows the orientation

The orientation’s sharpest test

II: The path no privilege governed

Korea: privilege made conditional

Taiwan: the licence premium stripped

Singapore: built past the claim