Who Makes Markets?

In April 2023, the national security adviser of the United States walked onto a stage at the Brookings Institution and quietly retired an idea his own country had spent four decades exporting. Jake Sullivan did not frame it as a repudiation. He called it renewal. But the substance was unmistakable. The belief that markets, left largely alone, would allocate capital, talent and technology to their most productive uses had, in Washington's own telling, hollowed out the country's industrial base and left it dependent on rivals for the things that matter in a crisis.

The diagnosis itself was old news. Economists had circled these arguments for a decade. The shock was the venue. Here was a critique of what Sullivan called an oversimplified faith in markets, delivered from the institution that helped draft the original rulebook, by the government that had spent forty years enforcing it on everyone else. Larry Summers, no enemy of open markets, called it the most developed statement of the administration's economic thinking to date.

Within months the same move repeated on every continent. Brussels met the American subsidy wave with a Green Deal Industrial Plan. New Delhi widened production-linked payments across fourteen sectors. Seoul, Tokyo and Beijing pushed public money into semiconductors with a conviction that made the old question of whether states should pick winners feel already settled. The IMF's New Industrial Policy Observatory logged more than two and a half thousand deliberate interventions in that single year.

The tidy story here is a comeback: the state, long exiled by markets, striding back onto the field. That story runs the history backwards. The state never left. The four decades we call the free-market era ran on a particular and expensive kind of statecraft, independent central banks, competition courts, deposit insurance, the patent system, a global trade referee armed to punish. Building a market that looks like it runs itself is among the hardest things a government can do. In 2023 governments simply stopped pretending that was not what they were doing.

We talk about markets as if they were weather. Systems to shelter under, or to foolishly try to steer. It is a comforting picture and a false one. A market gets built, and it stays built only while public authority keeps supplying what no trader can manufacture alone: money that will still buy something tomorrow, a court that will enforce a bargain struck by a stranger, a standard that lets a buyer in Durban and a seller in Dar es Salaam agree on what a tonne of steel is. Take those away and the market shrinks toward whatever personal trust and the threat of force can still hold together.

The textbook casts government as a repairman. Markets do the real work; the state turns up afterward to price pollution, break up monopolies, fund the research no single firm will pay for. Useful, and true as far as it reaches. Buried inside it is a sequence, market first and state second, that no economic history supports. No market of any size has ever come before the authority that denominated its money and enforced its contracts. The repairman built the house.

Grant that the state furnishes the market and a fifty-year argument loses its point. Economic policy has been fought as a quarrel over size. Big against small, more against less. It is a clean axis, and it predicts almost nothing. It cannot tell you why the same subsidy grows a chip industry in one country and a cartel in the next.

The recent evidence, gathered by Reka Juhasz, Nathan Lane and Dani Rodrik in their reassessment of industrial policy, breaks the ideological script. The newest studies read more kindly on industrial policy than the last generation did, and not because governments got wiser. Economists got better at measuring what actually happened. What happened turned far less on how much a state spent than on how it was built to spend it.

Call it intervention and four different acts collapse into one flabby word. Each shapes the economy more directly than the last. Each demands more of the institution attempting it. They stack, and the order is not a suggestion.

The classic view stops at the first rung: set the rules, step back. Most of 2023 was governments lunging for the fourth, deciding on purpose which industries their countries would host. The two middle rungs do the quiet work that makes the top one survivable, and they are where a serious state gives itself away.

Everybody wants to be the architect. It comes with ribbon-cuttings, national champions, plans thick enough to photograph. It is also the rung that punishes a weak state hardest, because steering capital toward industries that do not yet exist is precisely where a captured or careless bureaucracy burns the most money. The glamour and the danger are the same fact seen from two sides.

Hand the same tool to two states and watch it split in half. A production subsidy tied to performance, run by people who can take it away, becomes a machine for learning. The identical subsidy, handed to an agency that cannot say no, becomes a pension for whoever lobbied hardest. Same instrument. Different state.

Capability is the word carrying the weight, and it has quietly hardened into a slogan, so it pays to be exact. It runs deeper than honesty or raw brainpower. Peter Evans got closest, decades ago, with a phrase that still earns its keep: embedded autonomy, a state close enough to industry to know what it needs and walled off enough to refuse it. Five capacities do most of the work.

The one everybody underrates is the third. Every industrial policy is good at starting things. The rare skill, the expensive one, is stopping them. South Korea's rise usually gets told as a run of clever bets. The part worth stealing is the discipline behind them. Support arrived with a test attached, and firms that missed their export targets lost their cheap credit, sometimes their owners. What made the test bite was that Koreans did not mark their own homework. The examiner was the world market. You can lobby a minister. You cannot lobby a German importer who simply buys someone else's steel. The state had borrowed a scoreboard it was unable to rig.

Pull on that thread and it runs through every case in this piece. States get good at the things they are scored on by someone they cannot overrule, and they rot at the things they grade themselves. It is the most reliable tell of capability I know, and it cuts hard against the reflex to fix a failing institution by reorganising it from the inside. The problem usually lies elsewhere, in the absence of a verdict the institution cannot argue with.

Seen this way, capability behaves like a muscle. It is specific to the task that built it, and it wastes when the task stops or the scoring stops. A state can be world-class in one room and helpless in the next, at the same moment, for the same underlying reason. Hold that thought. It is the whole of the South African story.

Put ambition on one axis and capability on the other and every state lands somewhere on a single plane. Ambition is how hard a government is trying to shape its economy. Capability is whether it can. The diagonal is where the two run even, a country attempting exactly what it can execute.

Distance from that line is the whole reading. Sit above it and capability is going spare; the state could afford to be bolder than it is. Sit below it and a capability gap has opened, the room between what a government promises and what it can deliver. Inside that room, subsidies leak, champions harden into monopolies, the master plan sets into a monument. Industrial policy leaves the gap untouched. It takes whatever the state already is, disciplined or captured, and multiplies it.

Singapore and Korea sit high and near the line, capable states carrying about as much as they can, which is a great deal. China sits well to the right of its own capability, betting that scale and speed will let it grow into ambitions it cannot yet execute. Washington and Brussels have lurched rightward since 2022, buying ambition faster than either has built the capacity to permit, staff and pour concrete. Their gaps are fresh, and, unusually, argued about in the open. Washington's own arithmetic makes the point. The Congressional Budget Office scored the Inflation Reduction Act's energy credits at about 391 billion dollars; within months Goldman Sachs put the likely bill near 1.2 trillion, three times higher, because nobody had written a ceiling into the credits. The most capable treasury on earth could not price its own flagship policy within a factor of three.

The dangerous ground is the lower right: high ambition, thin capability. It is also where the itch to copy runs strongest.

What makes the new industrial policy so contagious is that its instruments are portable and its foundations are not. A special economic zone can be announced in a budget speech. A national champion can be created by decree. A localisation requirement can be drafted in an afternoon. What cannot be legislated into being is the coordinating capacity that makes a zone attract real firms rather than tax arbitrage, or the exit discipline that stops a champion from becoming a protected monopoly feeding on the public purse.

Call it the cargo-cult failure. Across much of the developing world it is the common case: the visible architecture faithfully reproduced, the machinery underneath simply missing. The zone is built and stays empty. The subsidy flows and never learns to stop. The plan is launched with fanfare and quietly buried when the minister rotates out.

The obvious response, build the institutions first and do the industrial policy later, is half right and quietly ruinous. Capability is built by doing the work, under conditions where failure shows up fast and costs something real. It cannot be poured like a foundation slab and left to cure before the real work begins. Korea's bureaucracy did not qualify for industrial policy and then begin. It got good by running the policy against a scoreboard it could not silence, killing failures early and correcting in public. The path up the ladder is narrow, and it looks like this. Start one rung above where you honestly stand. Wire in a test you are not allowed to overrule. Keep failure cheap and visible, so the correction lands before the fortune is spent. Capability compounds only where the feedback stays honest.

No country shows the capability gap more painfully than the one this is written from. South Africa runs institutions much of the continent envies. Its Reserve Bank is independent and genuinely respected. Its courts proved, under real pressure, that they were still courts. Its Treasury and its statistician work to a standard wealthier countries would recognise. On paper, a state built to shape markets.

In 2023 that same state could not keep the lights on. South Africa lost a record 332 days to load-shedding, deliberate rolling blackouts to stop the grid from collapsing outright. Thirteen days in the entire year escaped a cut. The Reserve Bank, that admired institution, spent part of the year calculating what a sister institution's failure had cost: growth of about 0.3 percent, against the 2.3 percent it reckoned the economy could otherwise have reached. One world-class institution, reduced to measuring the wreckage left by a broken one.

Here the easy reading, that South Africa is simply a weak state, falls apart. The same country holds inflation inside a target band and collects its taxes through brutal years. Weakness is too blunt a word. Look closer and the pattern is stranger and more useful. Eskom was not always this. In the 1990s it was an engineering institution of real distinction, delivering some of the cheapest electricity on earth and named power company of the year by its own global industry. The capability was there, and then within a single decade it was gone. What changed was the scoreboard. The country's average competence held roughly where it had been. The Reserve Bank answers to a number it cannot fake and to markets it cannot command. Eskom, a monopoly answerable to one shareholding minister, could miss every target for years and still be marked as passing. Capability starved in the one place where nobody independent was keeping the score.

Look hard at the space between those two lines, because it says something sharper than measure what matters. Eskom's availability was measured the whole time. Published weekly, dissected by analysts, tabled in Parliament. The number was there all along. What was missing was any mechanism that made a powerful person feel it. Inflation bites a finance minister because the bond market reprices within the hour. A failing plant bit nobody who could fix it, not on any clock that touched a career. A metric with no consequence wired to it is decoration. A real scoreboard carries a cost. Measuring more was never the answer here. The answer was to attach a real cost to the measure, and hand the switch to someone the powerful could not call off.

Ambition was there in abundance. South Africa has energy master plans, localisation targets, a headline climate-finance deal worth billions of dollars. Stacked on top of a hollowed delivery institution, all that ambition produced neither the industry nor the credibility, only a widening gap between the press release and the wall socket. A green-industrial future cannot be architected on a grid that will not stay up. The fourth rung is out of reach while the second is on fire.

There is a darker turn in the same story, and it is the sharpest warning the continent has produced. Through the years now catalogued as state capture, South Africa did not lack the instruments of a developmental state. It had the state-owned enterprises, the procurement billions, the national champions, the whole visible apparatus of the top rung. Those instruments were turned, on purpose, into machinery for extraction. The tools that lifted Korea were used to strip a country. This is the capability argument at its bluntest. The instrument is neutral. Its sign, plus or minus, is set entirely by the hand that holds it.

Set beside that, the continent's quieter wins read differently. Botswana gets praised for prudence, which sells the achievement short. Its real feat was to refuse the top rung. It picked a narrow set of things and did them with unusual discipline, running diamond revenue through institutions walled off from the election cycle, banking the windfalls in a sovereign fund, holding corruption low enough to be regionally conspicuous. On the Capability Frontier it is one of the few states sitting comfortably above the line. Its constraint is ambition, and a dependence on diamonds that leaves it exposed. Botswana's next decade is the pleasant, hard question of how much bolder to risk becoming.

Rwanda is the mirror image, and the more contested case. A small state that has spent its scarce capacity on the machinery of delivery: a results-driven bureaucracy, a development board built for speed, and imihigo, public performance contracts that pin officials to targets ordinary citizens can watch them hit or miss. It is a home-grown attempt to manufacture the very thing Eskom lost, a scoreboard the powerful cannot quietly edit. Rwanda sits close to the line, ambition and capability climbing together. The debate around it is real and belongs in any honest account. Delivery has been bought alongside a narrow political space, and reasonable people doubt how long a model so dependent on one centralised will can hold. What is not in doubt is the sequence. Rwanda took the capability problem seriously first, and reached for ambition second. That order is the entire argument of this piece.

An idea earns its keep by surviving its own objections, and this one draws three that bite. The first is old and merciless. Name any scoreboard, make it the target, and people start improving the number rather than the thing. Goodhart's law, and it is why manufactured scoreboards rot. Officials hit local-content quotas by relabelling imports. Exam-driven schools turn out superb exam-takers. A metric the powerful cannot edit is rare precisely because power spends its days learning to edit metrics. Building the scoreboard is the easy day. Defending it, year after year, from the people it exists to judge, is the job.

The second objection cuts the other way. Some things worth doing cannot be scored on any honest short horizon. Foundational research, a semiconductor base, a generation of engineers: the payoff lands in decades, and a strict export test applied too early would have strangled them in the cradle. Korea's discipline worked because it was aimed at products that met a market within a few years. Point the same stopwatch at a twenty-year bet and you get timidity dressed up as rigour. The craft is fitting the clock to the wager, patience where the horizon is long, ruthlessness where it is short, and the wisdom not to confuse the two.

The third is the most uncomfortable, because it lands on the star witness. The tidy story of Korean export discipline is partly told backwards, buffed by the winners and quiet about the failures and the plain luck. Economists still argue over how much of the miracle the state truly caused. That does not dissolve the pattern, which turns up too widely to wave off, from Botswana's fiscal rules to the wreck of Eskom. It does demand some humility about any single case, this one included. An argument that cannot point to its own weak seams is selling something. This one has three, and they hold weight.

One question has to stay open, and pretending otherwise would be cheap. Much of 2023 was plainly reactive. A pandemic that exposed brittle supply lines, a war that turned gas into a weapon, a rivalry that made dependence feel like a loaded gun. Shocks fade, and the appetite for expensive intervention fades with them. A decade from now the subsidy race may well read as a costly detour that fractured trade and left a sprawl of half-built factories nobody needed. The sceptics hold real evidence, and they will press it hard.

The structural reading is harder to wave away. The forces pulling states back into the market, decarbonisation, the security premium on making critical things at home, the value of owning a frontier technology outright, show little sign of being temporary. If they hold, 2023 was the road itself bending, and what looked like a detour becomes the direction. Which reading wins comes down to something unglamorous. Whether the ambitious states build scoreboards honest enough to tell their real bets from their boondoggles, before the money runs out.

The temptation for African governments is to shop the menu, copy the loudest instruments, and call it a policy. The harder and more useful work points inward, and it cannot be outsourced. A handful of questions travel further than any master plan.

How to read this: the five points are the questions worth carrying, not a score. Select any point on the compass to read the question it stands for and the reasoning behind it in the panel beside it.

For African governments the lesson of 2023 has little to do with dosage. Stop litigating the size of the state. Start building its capability, and build it the only way it is ever built, by doing hard things under a scoreboard nobody can quietly rig. Markets are made. What decides the next decade is whether the states making them can be held to the result.