In 1955, 970,000 black children were enrolled in South African schools. By 1967 the figure had reached 2.24 million. By 1973 it was 3.31 million – a 131 per cent jump in twelve years. And it happened even though school was not compulsory for a single one of those children, and even though almost everyone, parents included, knew the schools were failing them.

The schools were chronically underfunded and overcrowded. Most teachers had no tertiary training. Most pupils never made it from primary school to high school, let alone to matric (the final school-leaving examination). And still the numbers enrolling grew every year. The historian Oliver Aiken calls this the paradox of Bantu Education: a system designed to give black children less, swamped by black families demanding more.

Why would a parent enrol a child in a school they knew was failing? Part of the answer is simple: any education was worth more than none. A literate child could find work a non-literate one could not. Families were making an investment – in the ordinary economic sense of the word – and they understood what they were doing.

We remember the uprising of 16 June 1976 as a revolt against Afrikaans, and as the moment Black Consciousness arrived in the streets. Both are true. But they explain why the students marched, not why they marched that year. For that, it helps to look at the economy.

The timing of Soweto appears to have been shaped, to a surprising degree, by the South African business cycle: by a boom that allowed white voters to ignore black schooling, a downturn that prompted business to help pay for it, and a recession that withdrew the funding at a particularly bad moment.

Start with the boom. Through the 1960s the economy grew quickly – six to seven per cent a year for much of the decade. Employers’ demand for skilled work was met comfortably by white workers, whose higher salaries they could afford to pay. What they wanted from black workers was narrower: literate, low-skilled, low-wage labour. Bantu Education was built to supply close to that, and little more. As Hermann Giliomee notes, white South Africans of the time largely did not support large-scale spending on black schooling. There was little economic reason, from their point of view, to train skills that employers did not yet intend to buy.

So the system was kept small by design. Spending on Bantu Education was capped at R13 million; anything beyond that had to come from taxes levied directly on Africans themselves. Black parents, unlike white parents, paid for school fees and textbooks out of their own pockets. By 1971–2, the state spent roughly eighteen rand educating a white child for every one rand it spent on a black child. In the nominally independent Homelands – the rural territories to which the apartheid state assigned black South Africans – schooling was funded worse still.

Into that gap stepped the parents. In 1967, Tswelelang Higher Primary in Meadowlands set out to build extra classrooms by levying a R2 tax on local residents and running a bottle-cap collection drive. The Department refused to match a cent of what they raised. When the Urban Bantu Council asked the liberal-leaning Johannesburg City Council to help lift the monthly schooling levy from 18 to 38 cents, the government dragged its feet, conceded two cents, and left the council R55,000 in debt by 1970.

By 1971 Soweto had twenty secondary schools – roughly one for every 80,000 urban African families. To put that in perspective: a township of more than a million people had the secondary-school capacity of a small country town. When residents pressed again for the higher levy, the deputy minister, Piet Koornhof, told them any extra money would be spent not in Soweto but in the Homelands, where black secondary schooling was supposed to happen anyway.

Then the cycle turned. The downturn of 1969 achieved something a decade of liberal argument had not: it changed the minds of business leaders. For years, liberals – here, those who wanted apartheid’s economic restrictions loosened – had warned that Bantu Education would eventually leave the economy short of skilled black workers. In a slump, with white labour looking expensive and scarce, that warning began to look practical rather than theoretical.

Companies that had ignored black schooling now moved to fund it. In August 1971 Reckitt and Colman pledged R100,000 for bursaries; Barlow Rand set up a foundation with some R700,000; in February 1972 Harry Oppenheimer’s Anglo American gave R160,000 towards a junior secondary school in Soweto; The Star launched its TEACH fund. Because these gifts flowed through the City Council and private hands, they bypassed the R13 million ceiling.

Government shifted too. A pragmatic, verligte (reform-minded) wing of the National Party had decided, as Jonathan Hyslop puts it, to accept ‘the continued existence of the urban black working class’ – and to spend accordingly. Between 1969 and 1972 it dropped job-reservation rules, eased restrictions on black urban employment, and, most importantly, in 1972 shifted the financing of black education off African taxation and onto the state’s general revenue. The 38-cent levy was finally granted. Spending on black education rose from R55 million in 1970–1 to R97 million by 1973–4.

The effect was substantial. In four years the number of secondary schools in Soweto doubled. Secondary enrolment climbed from 12,656 pupils in 1972 to 37,656 in 1976 – an increase of more than 170 per cent.

This sets up the puzzle. Why did Soweto erupt precisely when demand for schooling was at its peak, and when, for the first time, both business and the state were moving to meet it?

Several factors coincided – and most of them were economic.

First, the timing was unfortunate. The global recession of 1973–5 reached South Africa, and the government reversed course and cut spending. As one historian puts it, the building of Bantu Education’s infrastructure ‘slowed down just at the moment that the greatest demands were being made on the system.’ The chart above shows the squeeze: growth fell from 6.7 per cent in 1974 to barely positive in 1975 and 1976, and into contraction by 1977.

Second, the same demand for skilled black labour that had opened company chequebooks now drew the best teachers away from the classroom. A factory or clerical job paid better than teaching, and teaching – long one of the few respectable careers open to educated Africans – became less attractive just as the schools needed staff most.

Third, the Department chose this moment to shorten black schooling from thirteen years to twelve, combining two cohorts of pupils into one in 1976. The result was a ‘bulge’ of younger children pressed into schools that had specialised in senior grades and were not equipped to receive them. Philip Bonner’s assessment is blunt: ‘Learning was virtually impossible and the pass rate plunged to lower levels than ever before.’

Into that overcrowded, underfunded and overstretched system, the government introduced a deeply unpopular measure. The 50–50 rule – that half of all lessons be taught in Afrikaans – had for years been applied loosely; by 1964, 95 per cent of secondary classes were taught in English. But hardliners wanted it enforced. In 1976 Vorster appointed the conservative Andries Treurnicht as deputy minister, and on 11 June Treurnicht ordered five striking Soweto schools to obey the rule. Soweto had planned, as in earlier years, to resist quietly. This time the Department forced the issue. The sense of ownership that residents had built over their schools, Aiken writes, ‘collapsed in an instant.’

None of this reduces Soweto to economics. The students who marched were refusing a language that stood for their subjection, and they carried a new confidence – Black Consciousness – that told them they need not accept it. Bantu Education was, at its core, a racist project designed to keep black South Africans politically and economically confined. The business cycle does not explain that, and it is not meant to.

What economics helps explain is the timing. It suggests why the system was so close to breaking in June 1976, rather than five years earlier or five years later: a school system swollen by demand, briefly funded, then squeezed again in a recession – all within half a decade. The Afrikaans decree was the immediate trigger. The economic conditions help explain why the system it met was already under such strain.

In a post on this blog in October last year, I shared a simple count: of the 42 students from Africa enrolled in the top 50 economics PhD programmes in the United States, two were South African. Brazil alone had 47 – more than the whole of Africa combined. In total, 137 came from Latin America, more than three times the African number. Not one South African was studying at an Ivy League department.

This lack of representation matters because the leading economics departments offer training that is hard to find elsewhere: advanced methods, exacting standards, and daily contact with the researchers who set the international research agenda. As I wrote then, ‘being trained by the best is much more likely to also make you the best’. And South Africans bring something their classmates cannot – a close knowledge of this country’s history, institutions and communities. We want more of them asking the questions that matter here, with the tools to answer them well. Many will return; those who stay abroad tend to keep working on South African questions, and to open doors for the students who follow.

That post did something I did not expect. A South African donor read it and asked a simple question: how can we help? I gave them an ambitious answer. Find students with the potential to earn a PhD at a top-50 department, bring them into honours and, for some, a master’s at Stellenbosch, and then train, mentor and fund them until they can apply abroad with confidence. They liked what they heard and backed our ambition.

We did not invent this approach. Latin America has shown, over many years, that it can work – in Brazil, Chile and Colombia. The striking thing is that the change came from individual universities, not from governments: a department decides to compete internationally, and slowly builds the record to do so. In 2024 alone, one institution – the São Paulo School of Economics – sent eleven students abroad to begin PhDs. We are forming partnerships with several of these universities to learn from what they have done, and from the mistakes they made along the way.

Today we launch the South African Economics Pipeline project. What the programme offers is concrete. From next year, each bursar receives R200,000 a year, enough to study full-time. They complete honours and, where the potential is clear, an 18-month master’s. Each is paired with two mentors – a Stellenbosch academic and an established international scholar in their field. Every November they attend a Summer School of PhD-level short courses taught by visiting scholars, and they take part in the new PSG–SU Economics PhD Workshop, which brings strong young researchers together at Stellenbosch. (More on this PhD Workshop in another announcement over the next few weeks.)

Our long-term aim is to build technical capacity within African universities and the South African government, and it already has support within Treasury. ‘The depth of South Africa’s technical policy capacity is a matter of national interest, and I welcome a programme designed to add to it’, says Dr Duncan Pieterse, Director-General of the National Treasury. ‘I am glad to be named as endorsing this aim.’

Applications are open from today, and closes 31 August. The first cohort begins in January 2027. We are looking for students that studied at South African universities with a strong record in economics, mathematics and statistics, the curiosity to work on hard questions, and the patience to spend several years building real expertise. If that sounds like you – or a friend – you can read the details and apply at econpipeline.org.

We are also looking for donors. Our main donor, The Millennium Trust, funds five bursaries, alongside everything that surrounds them: the Summer School, the mentor pairings, additional teaching, the support through the application process. That infrastructure is now built, and it can carry far more than five students. What limits us is bursaries. Adding one costs $15,000 per student per year. If you would like to put another South African on the path to a top PhD programme, you can find more info here or simply reach out to me.

The pipeline has been thin for a long time. It does not have to stay that way.

On the afternoon of 11 June 2010, in a packed Soccer City, Siphiwe Tshabalala collected the ball on the left, took one touch, and smashed it into the top corner. It was a brilliant goal, nominated for a Puskas, but it was a magic moment because of what it signaled. Sure, Bafana Bafana would go on to draw that match against Mexico, then crash out in the group stage. But nobody in that stadium was thinking about the standings. They were thinking that South Africa, and Africa, was on the world’s stage.

Like most South Africans, I remember that moment – and what turned out to be one of the best World Cups – very well. And in the epilogue to my book I used that month to make an argument about economic growth. Want to win the economic World Cup, I wrote, then stop hiring expensive coaches and top-down strategies. Give every child a soccer ball. Empower the next generation, and let them design their own future.

I still believe the spirit of that line. But sixteen years of football – and sixteen years of economics – have taught me it was incomplete. A football, it turns out, is not enough.

Let me show you why, using the World Cup itself as a teacher.

Start with a simple question. Each World Cup is awarded to a host nation. Where does that host sit on the global income ladder when the tournament arrives?

I ranked every country by GDP per capita – the Maddison Project’s long-run income data – in each World Cup year, and measured the host against the world’s frontier economy, the United States.

My hunch was that hosts have crept closer to the frontier over time. The data says something else. For sixty years, hosts hovered between a third and four-fifths of US income: Switzerland and Sweden near the top, Brazil and Mexico far below. The 1990s and 2000s had World Cups in countries close to the frontier, followed by South Africa, Brazil and Russia less than half of US GDP. Then, a reversal, with Qatar in 2022 at 255 per cent of US income, the richest host ever, by a distance.

In one decade FIFA swung from the poorest host to the richest. South Africa’s World Cup was sold as development – a coming-out party for a young democracy and a whole continent. Qatar’s was a petro-spectacle, air-conditioned stadiums paid for with gas. So much for the idea that hosts edge ever closer to the frontier. If anything, the choice of who hosts has come seems unmoored from how rich a country actually is.

Perhaps, instead, there is a correlation between income and victory? Let’s look at the final itself. In every World Cup final since 1930, I compared the two finalists’ GDP per capita.

In 13 of 22 finals, the poorer nation lifted the trophy. Brazil, never close to the income frontier, won five times, usually against richer European opponents. On any given afternoon, ninety minutes is a great leveller.

And yet zoom out and the picture inverts completely. Only eight nations have ever won the World Cup. Plot the concentration of titles across the world’s countries, the way economists plot the concentration of income, and the contrast is brutal.

Income across nations has a Gini coefficient – the standard measure of inequality, where zero is perfect equality and one is one country owning everything – of about 0.52. The distribution of World Cup glory has a Gini of 0.97. The people’s game, supposedly the most democratic sport on earth, hoards its ultimate prize more tightly than the world economy hoards money. Football is radically equal for ninety minutes and radically unequal across a century.

If money does not buy trophies, what does? As I wrote before, football is unusually generous with evidence, because it records everything.

For the 2022 World Cup I counted, for each squad, how many players earned their living at clubs in Europe’s ‘Big Five’ leagues – England, Spain, Germany, Italy and France – and compared that to how the team performed.

The relationship is clear: roughly every three or four players plugged into those elite leagues is worth an extra goal of margin over the tournament. Teams do well when their best talent reaches the steepest learning environments and tests itself against the best, week in and week out.

It is the same lesson I drew from the Africa Cup of Nations, and it is an old lesson in economics too: openness beats protection, and skills are built where competition is fiercest. Note the cautionary dot, though. Germany arrived in 2022 with more Big-Five players than anyone – and went home in the group stage. Access to the best matters; on its own, it is not enough. You still need the system around it to function.

Now combine the lessons. Football has been getting more equal where it matters for spectacle – the gap between the strong and the weak on the pitch has narrowed. The economy has been doing the opposite.

Start with China and India. As the two of them industrialised, well over a third of humanity saw their incomes race towards rich-world levels – the largest catch-up in history. Count every person equally, weighting each country by its population, and the gap between the world’s economies has been shrinking since around 1990. By that measure, the world is converging.

Measure the very same thing with a simple standard deviation, though – treating tiny Togo and giant India as equal points – and you get the opposite: divergence, dominated by the many small, poor economies that really have fallen further behind. And here football picks the second measure. A World Cup does not count people; on the pitch nobody weights by population, and tiny Togo has exactly as many shots at the trophy as giant India. By the tournament’s own arithmetic – one country, one team – the economic gap between the contenders really has widened. That divergence line is the world as football itself counts it.

And yet, on the pitch, the gap has done the opposite. Since 1950 the average winning margin in World Cup matches has fallen by more than a third, from about 2.3 goals to 1.4; minnows now routinely hold giants.

Why can football close gaps that economies cannot? Because footballers can move easily and the rules are the same everywhere. A century ago, barely one World Cup player in twenty turned out for a club abroad; by 2026 it is seven in ten.

A talented teenager from Addis to Zagreb can be scouted, signed and schooled in Madrid, and bring that learning home. Knowledge, coaching and competition flow across borders far more freely in football than the institutions, capital and know-how that actually make countries rich. The pitch is flat. The world is not.

Which brings us back to 2010, and to my own analogy.

When I wrote ‘give every child a soccer ball’, I meant education. Empower people with skills and they will build the economy themselves. It was the optimism of the moment, and it was half right. But the defining finding of development economics over the past fifteen years is uncomfortable: schooling is not the same as learning.

Plot every country’s expected years of schooling against its learning-adjusted years – the World Bank’s attempt to count how many of those years actually produce reading and reasoning – and almost everyone sits below the line where school equals learning. South Africa sits a long way below it. Our children spend about 10.2 years in school but acquire the equivalent of 5.6 years of genuine learning. Four and a half years simply evaporate. We handed out the soccer balls, but they never had fields to play on, or coaches showing up for training sessions.

And the consequences show up exactly where you would expect.

Here is South Africa’s income per person since Tshabalala scored. The gold line is what happened: essentially nothing. In 2022 the average South African was no richer than in 2010. The dashed line is where the National Development Plan – our own grand strategy, drafted in that hopeful era – hoped we would be. The gap between them is a lost generation, roughly 60 per cent of income that never materialised. Tshabalala’s goal was a flash of individual genius with no system behind it. So, it turned out, was the promise of 2010.

The lesson is that a child with a ball and nothing else does not become a World Cup player. He needs a pitch that is not a rubbish-strewn field, a coach who has himself been coached, a league that lets him play competitive matches, a referee who applies the rules fairly, and the freedom to move to wherever the game is hardest. The ball is the cheap part. The ecosystem is the hard part – and it is the part that wins tournaments and grows economies.

That is what I would change about my old analogy. I would aim even higher for ordinary people, and be honest about what that takes. Growth is not handed down by a clever coach with a five-year plan, and it does not bubble up automatically from a warehouse of soccer balls either. It comes from building the whole apparatus around the talent: real learning rather than mere schooling, institutions that enforce the rules, markets competitive enough to sharpen skill, and openness that lets people test themselves against the best.

When the tournament kicks off again in 2026 – across the United States, Canada and Mexico, three hosts spanning almost the full income ladder – watch which underdogs hold the giants. The beautiful game will, once more, flatten the field for ninety minutes at a time. The harder, slower, less glamorous work is building an economy that can do the same for a lifetime.

South Africa is back, too, playing the opener against Mexico, in a repeat of 2010. But look at where Bafana start.

Of all 48 squads heading to North America, theirs is among the least exported – barely one player in eight at a Big-Five club, while Spain and England field almost nobody who is not. If the lesson of this piece holds, that is exactly the gap to watch: on the pitch in 2026, and in the economy long after.

South Africa spends just under a trillion rand a year on public procurement – about 15% of GDP, on the best recent system-wide estimate of R931.5 billion in 2021/22. On 16 April 2026, Finance Minister Enoch Godongwana published 100 pages of draft General Regulations under the Public Procurement Act 28 of 2024. The rules scale by contract value: set-asides on contracts up to R20 million, ownership-linked pre-qualification from R20 million to R100 million, and subcontracting conditions above R100 million. On top of this sit minimum allocation targets across 21 designated-group permutations in Annexure 2, and a centralised supplier database, with compliance ultimately the Auditor-General’s to verify. With the Act and the companion Tribunal Regulations, this is the most far-reaching procurement-law reform in a generation.

What does the recent international evidence suggest the consequences will be?

What the regulations actually do

The new rules tighten as the contract gets larger. For contracts below R20 million, the contract is set aside for firms 100% owned by members of one of the designated categories – black people, black women, women, people with disabilities, military veterans, small enterprises, co-operatives, and so on. For contracts between R20 million and R100 million, bidders must pre-qualify by demonstrating either that at least 40% of their past procurement went to 51%-black-owned and -managed enterprises, or that they will subcontract 30% of the contract to 100%-owned enterprises in one of the identified categories. For contracts above R100 million, an institution may, after a feasibility assessment, make subcontracting a condition of the bid; where it does, every bidder must subcontract at least 25% of the contract value to 100%-owned enterprises in the identified category.

Annexure 2 sets a further layer of obligations: minimum set-aside targets across 21 designated-group permutations, expressed as a share of the institution’s annual procurement budget. The headline targets are 30% to black-owned firms, 15% to black-women-owned, 18% to women-owned, 4% to firms owned by people with disabilities, 2% to firms owned by military veterans, 30% to small enterprises, and on through twenty-one rows. An institution does not apply all twenty-one: it selects up to five categories and must meet the Annexure 2 targets for those it picks. The targets overlap – a black woman can count towards several rows – and summed they far exceed any budget, so the real burden is not arithmetical impossibility but administrative complexity, audit risk, and the pull towards chasing measurable categories over delivery.

This is a structural change, not just an adjustment. Under the previous regime, the Preferential Procurement Policy Framework Act, price counted for 80 of every 100 points (or 90 on larger contracts), with the balance awarded for B-BBEE status, so every compliant firm could still bid and could still win on price. The new rules replace points-based preference with hard exclusion: on a set-aside contract, only bidders meeting the ownership criteria are eligible to bid at all.

None of this is to dismiss the draft wholesale. Parts of it are genuine improvements that should survive any redraft: it requires institutions to publish their procurement plans and awarded contracts, it consolidates a famously fragmented system into a single, clearer framework, and its transformation objective is constitutionally legitimate. The question is not whether to pursue transformation through procurement, but whether outright ownership-based exclusion is the instrument that will achieve it. The international evidence is the test.

A debate already in motion

South African procurement-law scholarship has argued for several years that the system is over-regulating its way into worse outcomes. Geo Quinot – my colleague at Stellenbosch and director of the African Procurement Law Unit – has made this case across a sequence of papers. In a 2020 piece for the African Public Procurement Law Journal, he described ‘an extreme level of fragmentation’, with dozens of subordinate instruments layered on top of the primary statutes. By 2023, in Ruch Prawniczy, he was arguing that the granular, ‘band-aid’ approach ‘threatens to undermine the very constitutional foundations of the procurement system’. The Auditor-General’s own data are consistent: only 36% of national and provincial entities were fully compliant with procurement legislation in 2019/20.

In 2025, with Jonathan Klaaren, Ryan Brunette and Ron Watermeyer, Quinot extended the argument in APPLJ. The current paradigm, they wrote, ‘is more appropriate for off-the-shelf products and well-defined services than for public infrastructure and other complex purchases’. They proposed graduating rule-strictness to institutional capacity – relaxed rules for entities able to be trusted with discretion, tighter rules for those that cannot.

In his 2024 UNCITRAL chapter, Quinot put the diagnosis plainly: ‘Where capacity across the procurement system is low and where there are serious integrity concerns, it is common to limit discretion in procurement decisions … by way of strict rules.’ Strict rules and bureaucratic capacity, on this reading, are substitutes: a country leans on the one when the other is weak.

The conceptual argument is already well developed in the South African legal literature. What this post adds is the empirical scaffolding from the recent international economics literature, which the local debate has not yet engaged with. Two papers, both in the American Economic Review, are particularly relevant.

Lesson one

The first paper, by Bosio, Djankov, Glaeser & Shleifer (AER 2022), assembles a dataset of procurement laws, practice, and outcomes across 187 countries. The authors distinguish ‘laws’ (what is on the books) from ‘practice’ (what bureaucrats actually do), and ask which predicts outcomes – integrity of the process and quality of public works.

Three findings stand out. Practice predicts outcomes. Laws predict practice. But laws, on their own, do not predict outcomes. Put differently: once you know what bureaucrats actually do, the rule book adds no further information about whether the system delivers. Stricter regulation on paper has no measurable association with better procurement performance.

The puzzle resolves once one looks at state capacity. Stricter laws improve outcomes in low-capacity countries, where unconstrained discretion would more often be misused. They worsen outcomes in high-capacity countries, because the rules end up blocking welfare-improving discretion – including the discretion to exclude a low-quality bidder who happens to meet every formal criterion. In Bosio et al.’s framing, regulation and public sector capacity are substitutes – the same conceptual point Quinot makes in his 2024 chapter, but supported by cross-country regression evidence rather than legal reasoning alone.

Two further patterns are worth noting. In low-capacity countries, the law is stricter than practice: rules aspire, bureaucrats fall short. In high-capacity countries, practice is stricter than law: capable officials voluntarily exceed the rule book. The well-functioning systems are not the most heavily regulated ones. They are the ones with capable people who do not need the regulation to do the right thing.

So where does South Africa sit on this picture? Recomputing the Laws index from Bosio’s openICPSR replication file gives a baseline of 2.1 out of 4 – already slightly above the global linear fit at the country’s level of income, meaning the country is already a bit more heavily regulated than the typical country at its income level. Estimating where the new regulations would put it is necessarily approximate: Bosio’s original survey instrument was designed around a hypothetical US$2.5 million road-resurfacing case rather than South Africa’s set-aside architecture. We walked through all 24 Law-side questions, mapped each to a specific gazetted regulation, and applied three pre-registered scenarios: conservative (flip only on explicit textual mandate), central (flip on a reasonable legal reading), and aggressive (flip on intent). The full decision matrix is in the replication folder. The central estimate is 3.07 out of 4, with a conservative-to-aggressive range of 2.39 to 3.83. Under any of these scenarios, South Africa moves from “slightly above” the global trend line to the upper third of the distribution, alongside countries with considerably weaker public sector capacity. Nothing in the new regulations builds the underlying capacity that, on Bosio’s evidence, is the only factor shown to matter for outcomes.

Lesson two

The second paper, by Carril, Gonzalez-Lira & Walker (AER 2026), asks a different question: should the same procurement rules apply to a stationery order as to a wastewater treatment plant?

US Defense procurement runs on a regulatory threshold that helps answer it. Below US$25,000, contracts can be negotiated without public notice. Above US$25,000, contracts must be publicly advertised. Carril and co-authors compare otherwise-similar contracts that fall just above and just below this threshold, isolating the causal effect of forced public advertising. They find that above the threshold, the extra publicity brings in about 60% more bids and prices fall by about 6%. But cost overruns rise by 7%, and delays rise by 8%.

Almost all of the deterioration is concentrated in complex contracts – services, customised work, anything the buyer cannot fully specify in advance. The within-firm decomposition isolates the mechanism: when a given contractor wins under more publicity, it does not deliver any worse than the same contractor under less. The damage comes from which firms end up winning. Forced advertising pulls in marginal entrants who underbid the established suppliers, win, and then fail to deliver. Economists call this adverse selection at the entry stage: forcing competition changes the type of firm that competes, and on complex contracts it is the lower-quality entrants who win.

A complexity-tailored rule – heavy publicity for commodity contracts, lighter publicity for complex ones – would save the DOD about US$104 million a year, about 2% of relevant spend.

This is the same argument Klaaren, Brunette, Quinot and Watermeyer make in their 2025 APPLJ paper: rules should differentiate by procurement type. What Carril adds is the causal magnitude. Uniform rules across heterogeneous contracts have a measurable cost, concentrated where contracts are most complex – exactly where governments most need procurement to work.

South Africa’s new regulations apply uniformly. The R20 million threshold is the same for a stationery contract as for a wastewater treatment plant. The R100 million subcontracting condition, where an institution imposes it, can apply to a building lease and to a rail-signalling system alike. Most of the country’s high-profile procurement failures of the past decade – Eskom’s builds, the Transnet rail expansion, failing municipal water plants – sit in the complex category Carril’s paper identifies as most vulnerable to blanket rules.

Three predictions

What do these two papers suggest about how the new regulations are likely to perform?

First, costs will rise on average, though the evidence cuts both ways and design is decisive. Set-asides are not simply anti-competitive: the most directly relevant study, Cappelletti & Giuffrida’s (2024) decade of US federal procurement, finds that the entry of targeted firms more than offsets the exclusion of others, so the number of bids actually rises. But the same study finds that execution suffers. Small-business set-asides raise cost overruns by 16.2%; the more restrictive disadvantaged-business set-asides – those reserved for narrowly defined ownership groups – raise overruns by 30.9%, with delays lengthening by roughly 5% and 9% respectively. The mechanisms are adverse selection and weaker performance incentives. On preferences specifically, the picture is genuinely mixed: they can lower costs (Krasnokutskaya & Seim 2011; and, as reported in Cappelletti & Giuffrida 2024, Nakabayashi’s study of Japanese set-asides) when they draw genuinely new competitive entrants into a thin market, but they raise costs (Marion 2007 finds a 5% bid preference raised California highway costs by 3.8%) when the favoured pool is shallower than the market it displaces. South Africa’s new design narrows eligibility towards firms 100%-owned by members of whichever designated category the institution selects – a thinner pool than the 51%-black-owned market most firms have grown into under B-BBEE – and the rule sits on top of a budget split many ways. On the design choices made, and with many reserved categories of the restrictive 100%-ownership type, the new regulations fall on the cost-raising side of the evidence.

Second, execution on complex contracts is likely to worsen. This is Carril’s adverse-selection result applied to South African infrastructure procurement. When eligibility for a high-stakes tender is determined by ownership rather than ability to deliver, the firms that win have been selected on the wrong basis, and the cost of that selection error shows up at the execution stage rather than at award.

Third, the pre-qualification tier invites collusion, and the compliance burden will slow decisions. The R20 million to R100 million pre-qualification stage turns procurement into a two-stage process, and any pre-selection stage is a gift to cartels: a partial cartel can coordinate its first-stage bids to crowd competitive rivals out of the second stage and push prices up (Seibel & Skoda 2026). When Slovakia abolished pre-selection, procurement savings rose. South Africa’s construction sector, where collusion is documented in Competition Commission investigations and where the ‘construction mafia’ is a live enough concern that the draft addresses it directly, is fertile ground for exactly this. At the same time, the compliance burden climbs. This is Quinot’s point, made sharper by the new annexure: the categories an institution selects must each be tracked, met and audited, each carrying personal liability for accounting officers. South African courts have accepted that holding officials personally liable for procurement decisions has a ‘chilling effect’ on their willingness to take decisions. The new regulations multiply the compliance tests a single decision must pass to stand up to audit, so decisions are likely to slow, particularly where the compliance position is unclear.

Conclusion

Bosio et al.’s reading of their own results is that the regulation of procurement helps, but mostly in lower-capacity countries where unconstrained discretion is more often misused. South Africa has its problems, but it is not at the bottom of the global capacity distribution. It is in the middle, having lost ground over the past decade. The policy direction the evidence supports is not more law, but more capacity. A procurement system designed to substitute for a weak state tends, over time, to keep the state weak.

There is a parallel risk for the firms the policy means to help. Set-asides can keep firms alive without making them competitive: winners survive on reserved demand rather than rising productivity, showing no measurable productivity premium (Cappelletti, Giuffrida & Rovigatti 2024). A regime meant to build black-owned firms can instead entrench their dependence on the next set-aside. The right test of preferential procurement is whether it builds competitive firms, not whether it reserves contracts.

The conceptual case has been made in the South African legal literature for several years. What the international economics evidence adds is the magnitude: cross-country regressions on 187 countries showing the substitutes relationship, and a causal estimate from US Defense procurement that quantifies what uniform rules cost on heterogeneous contracts. The public-comment period on the draft regulations is open.

On the available evidence, the draft regulations as written are likely to worsen South African procurement outcomes, and they should be substantially redrafted before being gazetted in final form.

South Africa’s supermarkets are well stocked. By the crude measure of calories, the country produces and imports more than enough to feed every child in it. And yet more than one in five young South African children - by some counts closer to one in four - is stunted: too short for their age, the mark of poor nutrition or repeated illness in the first years of life.

This is not a story about famine. South Africa grows and imports plenty. Yet stunting here is severe and stubborn. That is the puzzle. If the food is there, why are so many children still too short for their age, and what can actually be done about it?

Read more

We came round a bend on a dirt road in the centre of the Kruger National Park two weeks ago. A pack of wild dogs had just pulled down an impala, a few meters from the car, and two of them were tearing the body apart while the animal, by all appearances, was not yet dead. It was an extraordinary sighting. It was also genuinely gruesome.

Wild dogs hunt this way out of necessity. They lack the heavy jaws and long, dagger-like canines of a lion or a leopard, which can kill large prey with a single suffocating bite to the throat. A wild dog cannot. So the pack pulls its prey down and simply begins to eat, shearing into the soft parts – the belly, the rump, the head – often before the animal has died. Efficient. Awful to watch.

It was not, on paper, an easy eight days. Many roads were closed after the floods a few months earlier. The grass was long, which makes spotting animals harder, and two small children make for few leisurely drives. On the first day, a boomslang, one of the most venomous snakes in the country, dropped out of a tree onto the hide we were about to enter. And yet we saw lion, leopard, elephant, buffalo, and those wild dogs. Most people alive today will never see any of it.

Kruger turned 100 yesterday. The instinct on a centenary is to do one of two things: to celebrate, or to mourn the poaching and the crumbling infrastructure. I want to do something else, because I think the most interesting thing about Kruger is an economic point that most people get backwards. We usually assume that wildlife and economic growth are enemies – that more people and more money must mean fewer animals. For a park like Kruger, the relationship runs the other way. Kruger is not only a contributor to South Africa’s economy. It is a beneficiary of it. Animals, it turns out, do well where people do well.

To put it into perspective, consider a strange but revealing exercise from the team at Our World in Data: weigh every mammal on the planet and ask how the total divides up. Wild mammals account for just 5 per cent of it. Humans are 36 per cent, and our livestock and pets – mostly cattle – make up the remaining 59 per cent. Our own bodies outweigh every wild mammal on Earth by more than seven to one. All the world’s dogs, taken together, weigh roughly as much as all the wild land mammals combined. Birds tell the same story: poultry outweigh wild birds by more than two to one. The biomass of wild mammals has fallen by something like 85 per cent over the past 100,000 years, most steeply since we learned to farm.

This collapse has a name. The late economic historian Eric Jones, in a book published shortly before his death, called it the ‘drawdown’ – the long extermination of wild animals that ran alongside, and helped to power, the rise of the modern economy. His claim is deliberately uncomfortable: that ‘without animal bloodshed and captivity, modern economies, societies, settlement and agriculture might not have emerged as they did’. The wealth that now lets us mourn the wild was built, in part, by destroying it.

In short, wild animals at scale have become a rare commodity. The places where they still exist in large numbers are not accidents of wilderness left over from an older world. They are deliberate, managed, expensive choices. Kruger, at nearly 2 million hectares – a long, thin park running down most of South Africa’s border with Mozambique – is one of the largest such choices on the planet. That is the first thing to keep in mind.

How that exception came to exist is a more tangled story than the gift-shop version suggests.

When the park was proclaimed in 1926, the New York Times ran the charming headline, ‘Biggest Zoo is a Vast Jungle’. Kruger was assembled from two older game reserves, the Sabi (proclaimed in 1898) and the Singwitsi (1902), and those reserves had been created, somewhat awkwardly, to protect game stocks for commercial trophy hunting. Conservation, in other words, began as an investment strategy. Farmers and the companies that had given up their land expected to recoup their losses through hunting once the herds recovered. The land itself was handed over precisely because it was thought agriculturally useless.

This does not mean the land was empty. As the historian Jacob Dlamini has shown, the lowveld had long been home to an African economy built on hunting and the ivory trade. Within a year of his appointment as head ranger, James Stevenson-Hamilton had driven roughly 2,000 people off the land around the Sabi, cutting them from the means that sustained them. They gave him the name ‘Skukuza’ – the destroyer.1 In a cruel twist, that name now belongs to the park’s largest rest camp and its administrative heart.

For a while the project looked likely to fail. Neighbouring farmers, whose livestock made easy prey once hunting was banned inside the reserve, dismissed it as a ‘government lion-breeding concern’ and lobbied to have the animals exterminated. And then, in the space of a few decades, the mood flipped. The ‘Great White Hunter’, in William Beinart’s phrase, became the ‘Great White Safari Guide’. The newspapers that had wanted the lions shot now competed to be the scheme’s loudest champion.

What changed was partly sentiment and partly money. The campaigners had a powerful example to point to: Yellowstone, the world’s first national park, which had demonstrated something governments needed to hear. Viewing wildlife, it turned out, was a legitimate and financially viable use of land. You could make more from tourists with cameras than from farmers with rifles. The choice of name sealed the deal. Calling it the Kruger Park – after the old Boer republic’s president, Paul Kruger – was an act of reconciliation between English- and Afrikaans-speaking whites, part of the same nation-building decade that gave South Africa a new flag. It was also, as the historian Jane Carruthers notes, a small fabrication: Kruger was not a particularly keen conservationist, and proclaimed the Sabi reserve only under public pressure. But the myth did its work.

With the park proclaimed, all that remained was to fill it. In 1928 the South African Publicity Association launched an advertising crusade aimed squarely at the United States. ‘A new bid for American tourists and their dollars is being made,’ reported the New York Times, ‘and South Africa is the bidder.’ For every 10,000 American visitors, the planners reckoned, something like a million pounds would follow. The bet paid off, and what drove the growth that followed was, tellingly, a set of luxuries: the private motor car, the camera, the cinema reel. Tourists were even reassured that they need not panic if a lion stared at their vehicle, since the animal had ‘probably never seen a car before’.

Look at the visitor numbers and you are really looking at South Africa’s economic history in miniature. In 1927 the park received 27 visitors. The next year, 650. By 1938, more than 38,000. The park welcomed its millionth visitor in the 2002/03 season and climbed to a peak of about 1.9 million in 2017/18, roughly doubling every decade for most of a century. The pandemic knocked it down to 1.3 million in 2021/22, and it has since recovered to about 1.87 million. A 2018 projection had visitor numbers doubling again to 3.65 million by 2029. That now looks fanciful, and the reason is simple: visitor growth stalled when the economy stalled.

This is the heart of it. A safari is a luxury good, the kind of thing people buy more of as they grow richer and forgo when times are hard. That holds for foreign tourists choosing between South Africa and Kenya, and it holds for South African families deciding whether they can afford the week away at all. The prices have always tracked a market – a rest-camp hut once went for a few rand a night – and the gates open widest when wallets are full.

But the dependence runs deeper than ticket sales. Protecting wild animals is enormously expensive. It takes rangers, fences, veterinarians, anti-poaching patrols, and roads and bridges – including the ones the recent floods washed away. That money comes from tourism revenue and from a state with the fiscal capacity to fund what tourism does not. Kruger, the profitable giant, helps cross-subsidise the smaller parks in the national network. When the economy is strong, visitors come, budgets hold, and the pressure to poach a rhino for quick cash eases. When the economy is weak, every one of those supports buckles at once. Put differently: the wild dogs we watched on that road depend, in the final analysis, on whether the South African economy works.

Which is why the cracks matter, and why I do not want to paper over them. Some of the units we stayed in had clearly seen better years, while the rondavels at Skukuza are being rebuilt in luxury. It is, in a small way, the whole South African story – the widening gap between what still works and what is falling apart. And in the same week we were there, something happened that had never happened in the park’s hundred years: two South African visitors were stabbed to death at Crooks Corner, in the far north where South Africa, Mozambique and Zimbabwe meet. Their car is still missing, and the likeliest explanation is a crossing from across the border. Kruger was supposed to be the safest place in the country, somewhere you could sleep without fear. This does not mean the park is failing. It means the park is not sealed off from the country, or the region, around it. Insecurity is what arrives when economies fail on both sides of a border.

Jones had a name for the second act, too. Where economies grow rich enough, he argued, the killing eases and some species stage what he called the ‘Great Rebound’ – a recovery bought rather than stumbled upon. That is what Kruger is: not a fragment of old wilderness that survived us, but a rebound we chose to pay for. Jones left the harder question open. The West’s pillaging has subsided; will the rest of the world repeat it as it develops, or learn to afford the rebound instead? Kruger is South Africa’s attempt at the second answer – and it holds only for as long as the economy does.

The lesson? If we want lions and leopards and wild dogs to still be here at the bicentenary, in 2126, it is not enough to guard the fence. We have to grow the economy that pays for the fence, staffs the gate, fixes the bridge, and fills the camps with people who can afford to come. Conservation and prosperity are not rivals fighting over the same hectares. They rise and fall together. A hundred years ago a newspaper called this place a vast jungle and marvelled that anyone would protect it. The task of the next hundred years is to keep it that way – which, in the end, means keeping South Africa a place where people, and therefore animals, can do well.

She is a founder in KuGompo City, South Africa (formerly East London). She has a small team, a product that works, and a handful of paying customers who keep coming back. What she does not have is money to grow. The bank offers her a loan, secured against assets she does not have. A government programme offers a modest grant and a place on a training workshop. Neither fits.

She knows exactly what she wants: capital that shares the risk if the bet goes wrong, and brings help if it goes right. Economists call it equity. She just needs a partner.

A remarkable new study shows she is in the majority. A team of economists surveyed 4,444 startup founders across 51 African countries, then ran an experiment that asked them to rate realistic-looking investment offers, one attribute at a time. Founders across the continent prefer equity to debt, and they prefer it by a wide margin. How wide? The experiment lets us put a price on it: swapping a debt contract for an equity one is worth about as much to a founder as cutting the loan’s interest rate by eleven percentage points.

That finding looks like a story about finance. But it is really a story about people. Equity is how you finance a bet on talent and ideas; debt is how you finance assets. A startup economy that runs on equity therefore only works if a country keeps producing people worth betting on – and the managers and engineers they will need to grow. South Africa does not produce nearly enough of them, and it produces them unequally. Here’s why it matters.

Read more

I don’t remember the exact date I became an Arsenal supporter. I think it must have been in 2010, around the time of the FIFA World Cup in South Africa. All my friends who followed English football are Chelsea or, worse, Man City supporters. I guess it was a combination of Arsene Wenger, who had studied economics, and his philosophy of investing in young, promising players and playing an attractive brand of attacking football. And the creative midfielder Cesc Fabregas (whom I had seen playing live in South Africa, a dull win in Johannesburg). And the Dutch striker Robin van Persie (also live, a much more lively quarter-final against Brazil). I was an avid rugby supporter (less now, with kids, though I’ll watch every Bok game) and especially cricket supporter – see this post.

But football drew me in, and I suspect a large part of that is the community. I’ve been reading Andrew Mangan’s Arsenal blog almost daily since 2010. A quick sum: Andrew has published at least 10,000 daily posts since I first read his blog, and I’m confident I must have read at minimum two-thirds of those. That excludes all the news articles and player ratings after every game, and the two weekly podcasts. In short, Arsenal, through the eyes of Andrew (and his team), has been a big part of my life for the last fifteen years.

There have been some ups. Arsenal have won four FA Cups (2014, 2015, 2017 and 2020). The most memorable was the 3-2 win against Hull City in 2014 – 0-2 down after eight minutes, level by half-time, then Aaron Ramsey – one of my favourite all-time Arsenal players – winning it in extra time. (Other favourites: Diaby, Ozil, Odegaard. I have a thing for a passing midfielder.)

But there have been many, many downs. It’s been the Premier League that’s been the biggest unclaimed prize. We finished second in 2016, a year we really should have won it, but until 2023, we never really looked like winning it. And then something changed.

The change had happened earlier, when a young Mikel Arteta became the new head coach in December 2019. He soon won the FA Cup, but his project was a process rather than a quick piece of silverware: bold decisions, big-name players sold, young stars promoted through the system, and – just like Wenger before him – sustained investment in youth.

It took a while, but success came. First qualification for Europe’s second competition (the Europa League), then, finally, the Champions League. In the Premier League, Arsenal began to move up the table, finishing second three years in a row. But for many, success only means one thing: winning silverware.

And, so, on Wednesday last week, as Man City scraped to a one-all draw with Bournemouth, Arsenal became champions of England. I was crouched on the cold cement floor of a rondavel in the Kruger Park, watching the last ten minutes on my phone, trying not to wake the kids. And then the utter relief at the final whistle. The deathly silence of the Kruger rondavel – interspersed, perhaps, by a distant howl of a hyena – juxtaposed against the wild celebrations on the streets of London and across the world. Perfection.

Why Arsenal won this year will be a topic of analysis for weeks on blogs and podcasts. I present one piece of evidence, comparing the late-Wenger era to the Emery and Arteta eras. Arsenal did not score more goals this past season (compared to the Wenger and Emery eras), but we did concede far fewer goals on a consistent basis. Winning is as much about defense as attack.

We’ve also been doing better against our main rivals over the last few seasons: for the last three seasons, we’ve not lost twice against the same rival.

And compared to last season, even though we lost more games, we managed to turn draws into wins. Arsenal had eight one-nil wins this season; five of the eight came on the road.

But though the numbers will tell their own story, this season will be remembered for one thing: anxiety. Sixteen years (for me) of waiting. Three years of being runners-up, of bottling (I’m a Protea fan, I understand, see the above-mentioned post), of being good-but-never-great. By May, we were not so much hopeful as battle-fatigued – shellshocked from leading the table into spring twice and being overhauled, punch-drunk from a third long (Northern Hemisphere) autumn of holding our nerve, hollowed out by the slow attrition of caring this much for this long. Arsenal – and here I include everyone: the manager, players, fans – simply could not afford another second-place, another close-but-no-cigar. We simply had to win, especially being ahead in the title race since game week seven.

And then the release that came on Wednesday. Like all fans, I enjoyed all the celebration videos. But the most special moment – the one that had me in tears – was this short post following the news:

Before my trip to Kruger, I sent Andrew a mail to ask if I could analyse Arseblog’s text (despite the anxiety, I never stopped believing we’d win this year). Andrew responded with a kind, supportive message. So I asked Claude Code to help me create a few graphs. Enjoy.

But first: How incredible is it to post about a single football team daily for more than two decades? Arseblog and the team running it is entirely supported by reader contributions. They have a Patreon account. I joined them recently – long overdue – and, if you’re an Arsenal supporter, I hope you do too.

Andrew’s daily blog posts can be analysed in multiple ways. I’ve only done it since July 2010, around the time I started reading, because I did not want the Wenger era to dominate the analysis.1 The first thing I did was analyse the length of the posts. The length has shortened somewhat in recent years to around 1,000 words, though with the notable exception of the very recent run-in. (The day after Arsenal became champions, Andrew wrote a 2,000+ word post.)

I then had Claude do a sentiment analysis of Andrew’s posts and compared it to a rolling average of the points per game. Below it, I then plot the difference. Above zero (in teal shading) means that the blog mood is higher than results would predict; Andrew was more bullish than the team deserved. Below zero means the opposite.

I find the results fascinating. In the Wenger years, Andrew was more pessimistic than Wenger’s results perhaps deserved. This turned around significantly after Emery’s appointment; the blog was consistently more optimistic than the results deserved. And so, too, for the start of the Arteta era, though his sentiment became more ‘balanced’ later.

The difference hides the variation in Arsenal’s performance, though. The large, positive difference at the start of the Arteta era was not an outlandish belief in Arteta’s ability; it was simply the result of Arsenal’s poor performance. The following graph illustrates this best. I plot an ‘event study’ type graph of sentiment before and after Emery and Arteta’s appointments. Andrew was clearly eager to see Wenger’s successor succeed; by the time Arteta was appointed, the future looked bleak, and the new manager received little ‘additional’ support from fans.

Why would I read about a faraway football club on an almost daily basis? I find Andrew’s pieces thoughtful and funny, a rare combination. He calls a spade a spade without the need to create artificial clickbait rage; the benefit, perhaps, of becoming a blogger before the era of social media. So how could I visualise this combination of wit, wisdom and, well, the willingness to call a spade a spade? Let’s count the ‘swear words’ per 1,000 words over the period.2

There is a clear downward trend from the start of the analysis to around 2014. Andrew did note in his email to me that:

Perhaps one thing to note/consider is that there’s likely to be a change in language/vocab over such a long time period by virtue of the author growing older/maturing along the way. But I'll leave that for you to think about!

The evidence seems to suggest he matured around 2014. He was then roughly the same age as I am now, which is good news, I guess.3

There is something faintly absurd about a South African economic historian crouched on the cement floor of a Kruger rondavel in the middle of the night weeping at the result of a London football match. There is also, I have decided, nothing more human than it. Arsenal finally gave me a trophy. Andrew, unknowingly, gave me sixteen years of company while I waited. I owe both.

A photograph crossed my Facebook feed last week. Three girls in front of a tent flap. The eldest stands in a green-print dress; the youngest sits on a wooden chair in the middle; the third, in navy, leans on the chair-back. All three have very short, almost shaven hair. The original caption notes the probable reason: poor camp hygiene, treated by shears. The photograph is from the Frankfort concentration camp, taken between 1900 and 1902. It has been colourised by Tinus le Roux, then sharpened with AI, and published in the second edition of his book, The Boer War in Colour.

I looked at the for a long time, because they looked somehow familiar. And then I read the caption and turned cold. The three girls have the same surname as me: Fourie.

The surprising thing was how surprised I was by my own reaction. I have spent the last decade counting people who died young. Most of human history, when written as a graph, is the death of children. Before vaccination and knowledge about hygiene and sanitation, roughly half of all children did not see their fifth birthday. This was true everywhere across the world before 1800, including the death notices of the early twentieth-century Cape Colony I’ve analysed and written about. A few years ago, I even co-authored a paper on mortality in these particular camps.1

And yet a single photograph stopped me in my tracks.

This essay is about why. It is also about what that asymmetry – the gap between what we know and what we feel – means for those of us whose craft is to count…

Read more

Meet Carel. In February 1834, on a farm outside Worcester, a 55-year-old enslaved coppersmith slipped his owner’s grip and disappeared. He had done it before. Sometime, somewhere along the way, he had picked up an Irish-sounding alias: ‘Mr Paddy’. Under that Irish pseudonym he hired himself out to one employer after another – a forge here, a building site there – living and working as if the law had never named him another man’s property. His owner, Schalk Willem van der Merwe, eventually placed an advert in the Government Gazette of 7 March 1834. It listed Carel’s height, his bald head, the burn marks on his arm, the fluency of his Dutch. Van der Merwe never recovered him. The next entry in his slave register simply ends.

Why did Carel run when he did? You can answer that with cruelty. You can answer it with longing. Both answers are true, and both are insufficient.

In a paper just published in the Asia-Pacific Economic History Review, Karl Bergemann, Gabriel Brown and I offer a third answer that sits alongside those other two. Carel ran when he did because the labour market had shifted under his feet – and he was the kind of man best placed to take advantage of the shift. Here’s why it matters: even under the most brutal of institutions, people still respond to economic incentives. Histories that ignore those incentives end up underselling the agency of the very people they want to honour. Economics, used carefully, gives us a lens onto choices that the archive otherwise leaves dark.

Start with a simple question. Why does anyone bother to coerce labour at all? It is, after all, expensive. You have to pay for overseers, walls, advertisements, recapture. The economist Daron Acemoglu and Alexander Wolitzky asked exactly this in a 2011 paper, and their answer is short: coercion pays only when workers have nowhere better to go. The moment a worker’s outside option – the wage they could earn somewhere else, the place they could blend in, the legal crack they could slip through – improves, the maths of coercion changes. The cost of holding people in place rises. People start to leave. Owners start to look for cheaper ways to get the work done.

That gave us a sharp, testable prediction for the Cape in the 1830s. Britain’s Slavery Abolition Act came into force at the Cape in December 1834, but it did not deliver freedom on that day. Instead, it began a four-year apprenticeship: the legal status changed, but the work continued. Wage labour, however, started to expand. Free Black and Khoe workers were already moving across the colony as masons, harvesters, wagon drivers, smiths. Recaptured Africans were finishing their indentures. Punishment books and guardian offices were chipping away at owners’ arbitrary power. None of this added up to liberation. It did, however, add up to a real improvement in the outside option for someone bold enough to try the wage economy without permission.

If the model is right, the most skilled enslaved people should have been the ones who ran first after December 1834. They had the most to gain by selling their hands as free workers. The least skilled, by contrast, should not have run any more than before – and might even have run a little less, since full freedom was now only four years away.

Testing this needed two pieces of data that nobody had ever combined before.

The first was a hand-built dataset of 689 runaway adverts placed between 1830 and 1838 in the colony’s two leading newspapers, De Zuid Afrikaan and the Cape of Good Hope Government Gazette. These adverts are not perfect. Owners only paid to advertise the runaways they thought worth recovering. But for the people they did describe, they described in detail: name, age, skills, clothing, the names of farmers who had hired them, even the pseudonyms they used.

The second was the Cape’s slave emancipation dataset, compiled from valuation rolls drawn up in 1834 to calculate the compensation paid to owners. Every enslaved person on it has an official price. That price was set by assessors using auction records: skilled workers were valued more, unskilled less, the sick and elderly less again. Imperfect, yes. But for our purposes, valuation is a usable proxy for the wage someone could plausibly earn outside slavery. We matched 344 of the runaway adverts to individuals in the emancipation dataset.

That gave us the building blocks of a difference-in-differences design – a workhorse tool of empirical economics. The idea is straightforward. Take two groups: one expected to respond to the change (high-valuation enslaved people) and one not (low-valuation). Compare how their escape rates moved before and after December 1834. The ‘high minus low’ comparison sweeps out anything common to both groups – weather, harvest cycles, the colony’s wider unrest. What remains is the differential response. That residual is what the model says should jump.

The picture is the argument. Before 1834, the two lines move roughly together. Both rise as the 1830s wear on. After December 1834, they separate. Escapes among low-valuation enslaved people drop sharply: they no longer have much to gain from a dangerous flight when legal freedom is only a few years away. Escapes among high-valuation enslaved people, however, stay high for two more years – before falling again as 1838, the date of full emancipation, comes within sight.

Translated into numbers: across the full 1830–1838 window, the diff-in-differences estimate is a 38 to 61 per cent rise in escapes for the higher-valued group, depending on the specification. But this average understates the moment itself. When we zoom in tightly on the months either side of December 1834 – using a separate technique called regression discontinuity in time, designed precisely to catch sharp breaks – the high-valuation jump is over 100 per cent above the pre-emancipation mean. The effect then fades, gradually, as full freedom approaches and the gain from running early shrinks back to nothing.

In short: when freedom-on-paper arrived in December 1834, the people best placed to earn a wage outside slavery moved first. The model called the shot.

It is entirely reasonable to push back here. Does this mean cruelty did not matter? Of course it does not. The sheer number of low-valuation runaways – women, children, the elderly, the unskilled – is testament to the desperation that drove flight throughout the colony’s history. Our estimates capture only the marginal response to a change in incentives. They do not, and cannot, explain every flight. The data have their limits too: adverts only record runaways their owners thought worth advertising; valuations capture skill imperfectly; and we cannot cleanly separate the channels through which outside options improved – legal reform, expanded wage demand, easier passing as free.

What the result does say is that even under chains, the calculus of the labour market reached the people inside it. Carel’s flight was not random. It was the choice of a man with rare skills, walking towards the place those skills could feed him.

Economics is often accused of speaking only about prices and markets, and of having little to say about the parts of life that actually weigh on us. Slavery is the hardest case. And yet here it is the economic lens – outside options, incentives, marginal response – that lets us read a fragment of the lives of people the colonial archive usually buries beneath the words of their owners. Those people noticed when their world changed. They acted on the change.

That is the discipline at its best: a way of taking seriously the choices of people in markets, including the people the law refused to count as people. We have many of these stories still to tell. Karl is currently writing a book that will tell more of them.

A fortnight ago, in The oldest tool in the book, I wrote about the knappers of Jojosi: the people who, for more than 100,000 years, walked back to the same hornfels outcrop above a KwaZulu-Natal river to chip blades from the same dark, fine-grained rock. I explained that they are a reminder that what makes us human is not our strength or speed, but a skill that can be taught, traded, and passed on. Skills produce the surplus that economists since Adam Smith have called the wealth of nations. Multiply that across millions of people, and you have the long walk out of subsistence.

The walk was, for most of human history, painfully slow. As Joel Mokyr – the 2025 Nobel laureate in economics – has been reminding us for forty years, growth before 1750 was ‘slow, intermittent, and reversible’. Then came the Industrial Revolution, the marriage of artisanal know-how and codified science. ‘Technology pulls itself up by the bootstraps,’ Mokyr writes, ‘by giving scientific researchers vastly more powerful tools to work with.’ We are, today, at another bootstrap moment. Artificial intelligence has not erased the value of skills: it has raised the premium on knowing what to do with the machine in front of you. The marginal product of a clever question, asked of a clever model, is enormous. The marginal product of someone who can only consume what the model spits out is, well, marginal. Skills, again, separate the best from the rest.

Machines augment rather than displace the skills of an economic historian. What are these skills? Anne McCants put it well in a 2020 essay in the Journal of Interdisciplinary History. The economic historian, she wrote, lives in the ‘structural hole’ between two disciplines. From economics we borrow theory, the discipline of asking what would have happened otherwise, and the courage to put a number on a question. From history we borrow context, narrative, archival patience, and a humility about generalisation. The result, McCants argues, is a kind of social science that ‘disrupts inevitabilities’, ‘widens our horizons of empathy’, and shields us from a misguided nostalgia about the past.

Put differently: we are the people in the social sciences who are obliged to look at both the spreadsheet and the parish register, and to be suspicious of anyone who looks only at one.

That is a hard discipline to practise. Real archives are messy. Real datasets have holes. Real arguments must respect chronology and causation. And the bar for publishing in our field keeps rising: cleaner data, richer context, sharper identification, more robustness checks. All of that is expensive in time, the only resource a working academic truly cannot replenish.

Tools that ease the load

Which brings us to the point of this post. Over the past three months I have been building a set of tools – ‘skills’, in the language of Claude Code – that aim to give the economic historian back some of that time, without sacrificing rigour. Each was born of a small frustration in my own workflow. Each is named after someone whose own work cleared the path. You can find them all listed on my website and available for download on Github. Let me describe what each does, and why it matters.

/tyler is named after Tyler Cowen, the George Mason economist behind Marginal Revolution and the most voracious public reader of academic work I know of. Cowen’s habit is to read fast, write briefly, and synthesise widely. The skill borrows the spirit. /tyler turns a folder of academic PDFs into a token-efficient markdown wiki: a single 25-page paper costs an AI roughly 12,000 tokens, and a 100-paper literature review will not fit in a typical session. Tyler shrinks the lot to a 40,000-token index – titles, authors, abstracts, keywords, JEL codes – with the full text on demand. For almost every paper, or every set of papers, that lands on my desk now, I first ask Claude to /tyler them. I can then ask one question of a hundred papers at once: which of these find a positive coefficient on schooling? Which use a difference-in-differences? Which cite Mokyr? It is, in short, a research assistant who has read everything by lunchtime.

/janluiten is named for Jan Luiten van Zanden, Professor of Global Economic History at Utrecht and my own PhD advisor. Jan Luiten taught me to ask, before any new project: what is the big story, why is it relevant to the field, and why should anyone want to know? The skill is built on those three questions. It is the supervisor everyone needs, and the mentor that even senior researchers can benefit from. Have a new paper idea? First ask /janluiten. It will tell you, gently but firmly, when the idea is not yet worth a year of your life – and, just as usefully, when it is.

/diebolt is named for Claude Diebolt, the CNRS research professor and founding editor of Cliometrica, who has done more than almost anyone to keep cliometric standards alive in continental Europe. (Claude was also the first journal editor to give Dieter von Fintel and me a chance, with our 2010 Cliometrica paper on inequality at the Cape – the kind of editorial generosity that opens careers.) The skill mimics what Claude does as an editor every week: a managing-editor agent reads the paper, identifies the methods actually used – difference-in-differences, an instrumental variable, a structural model, a long-run wage series, an archival reconstruction – and assigns a panel of referee ‘gurus’ chosen to match those methods. Each guru reads the paper in isolation, the way the Economic History Review or the Journal of Economic History would send it out, and writes a full report. The editor synthesises. If you have Codex installed, a second model family (GPT-5.4) runs an independent audit of the methodology reports and even cross-verifies the R scripts against the statistics in the paper – two-model verification where it matters most. You decide which revisions to act on. The point is to find the weak spots in your argument before a real reviewer does, when there is still time to do something about them.

/kris is named for Kris Inwood, the Guelph economic historian whose long collaboration with South African scholarship – and whose extraordinary 2020 donation of his personal library to LEAP – made many of the citations this skill is designed to verify reachable in the first place. /kris hunts hallucinated references. Two model families work in parallel: one cascades through CrossRef, OpenAlex, Semantic Scholar, and arXiv; the other resolves DOIs, monitors retractions, validates ORCIDs, and cross-checks the Internet Archive. Disagreements trigger an adversarial round. If a citation in your bibliography does not exist – or has been retracted, or has the wrong year, or attributes a paper to the wrong scholar – /kris will catch it before your reviewer does.

/EHRstyle encodes the Economic History Review’s submission rules into a LaTeX template. The good news is that EHR now accepts LaTeX manuscripts; the less good news is that the journal’s house style is exacting – UK spelling, anonymity protocols, the journal’s peculiar footnote citation form, the pre-1992 ‘second series’ convention. /EHRstyle handles the lot. (You will still want to look carefully at the shortened titles in the footnotes – the skill is good but not infallible there.)

/tanniedi is named for Di Kilpert, the language editor who has copy-edited many of LEAP’s papers over the years and, in the process, taught a generation of LEAP students what an em-dash is for and what a comma is not. The skill packs a LaTeX project into a clean Word document for editing – the format Di prefers – and then unpacks the tracked changes back into the original `.tex`. No more retyping equations. No more lost cross-references. Just text out, edits in.

Skills for a research lab

Claude Code skills are also wonderful for standardising team work. I direct LEAP, the Laboratory for the Economics of Africa’s Past, and over the past month I have built a small family of LEAP skills that let us apply our brand consistently across everything we publish – papers, slides, letters, replication packages, public-facing data. If you run a team, academic or corporate, and you have not yet codified your house style as a set of skills, this should be high on your priority list. The cost of doing it once is low. The cost of not doing it – every postdoc reinventing your colour palette, every RA exporting a slightly different version of the logo, every reference letter laid out by hand – is hours that nobody gets back.

A few examples. /LEAPstyle is the master file: colour palette, ggplot2 theme, beamer template, LaTeX preamble, writing voice. Anyone in the lab can ask Claude to apply it and get output that looks unmistakably LEAP. /LEAPstructure scaffolds a new research project from scratch – the canonical `submission/`, `release/`, and `archive/` layout, with a manuscript skeleton, a `pipeline.R` stub, a CC BY 4.0 licence, and a project-level CLAUDE.md, all wired together so that a co-author or a future replicator can find their way around without asking. /LEAPdata turns the messy reality of a research folder – a tangle of xlsx files, training labels, PII, intermediate joins – into a clean, public, CSV-only GitHub data release with a README, a CODEBOOK, machine-readable variable definitions, and an R loader. /LEAPletter generates reference letters and formal correspondence on the official LEAP–Stellenbosch University letterhead, with the four-colour accent strip and the right signature block in the right place; what used to be a fortnightly act of fiddling with margins is now a single prompt.

I also use this approach for my blog. /OLWsocial is the one worth describing here, because it is the kind of skill any team with a public-facing channel could build for themselves. Given a finished blog post, /OLWsocial reads the text, identifies the core argument and the most provocative claims, and assembles a complete social media package: platform-specific posts for LinkedIn, X, Substack Notes, and Instagram, each with a different hook and the right length and tone for the platform; quote cards rendered as images, with the right font and colour palette; and any graphs from the post adapted for social – resized, retitled, with the chart junk stripped out so that a small phone screen can still read them. The output is a single folder that the social media manager can work from directly. What used to be half a day of cropping, retyping, and second-guessing is now a five-minute review.

Kabbo and the long walk

Most of these skills are aimed at economic historians. One is not – or not exclusively. Kabbo – available at kabbo.app – is a publication-pipeline manager for scholars. List, search, create, update, analyse. Spot stalled drafts. Set reminders. Export BibTeX. See, at a glance, what your team is up to this quarter.

For years I kept track of my own work in a Microsoft Publisher document: little coloured boxes for ideas, drafts, papers under submission, revise-and-resubmits, accepted, published. Every time something moved – a referee report arrived, an editor said yes, a paper went out the door – I had to drag the tiny box one column to the right. By hand. Over Christmas, watching how quickly the new AI tools could do this kind of thing, I realised the manual dragging was finished. I opened a Lovable account and began to play. Kabbo.app is one consequence of those fun hours. I have since moved the codebase from Lovable to Claude Code, and I expect to make a lot of additions over the coming months. I know of at least one academic department who have started to use it, which is fantastic. Closer integration with Claude Code and Codex is on the way, as are tools for team leaders. I am very eager for more feedback on how to make it better. (It is free, by the way.) Sign up on this blog and stay updated.

I named it Kabbo for a reason. //Kabbo – the name means ‘dream’ in |Xam – was a Bushman storyteller from the northern Cape who, in the 1870s, was force-marched in chains to Cape Town’s Breakwater Prison for stock theft. There the German linguist Wilhelm Bleek and his sister-in-law Lucy Lloyd took him into their Mowbray home and, over many months, recorded his stories. The notebooks they filled together preserved a |Xam world that would otherwise have vanished entirely. Kabbo spoke often of the journey he longed to make: the long walk back to his land, to his family, to the work of telling stories. He died before he could complete it. But the pipeline of his stories survived – and it survives still.

Like Kabbo, every scholar walks a long road carrying a bag of half-told stories, hoping to get them home before the light goes. The right tools, used well, help more of them home.

South Africa’s first attempt at a national AI policy was withdrawn sixteen days after it was gazetted. The reason given was that several references in the bibliography turned out not to exist. The deeper problem was that the withdrawn draft read as if artificial intelligence were a technology the South African government could control. The evidence runs the other way. South Africans are already using these tools, learning from them, and reshaping their working days around them. The next draft has to start there.

So what should it say? Begin with what the data show.

The Bureau for Economic Research’s 2026Q1 outlook survey asked South African managers and professionals, and a separate panel of private individuals, two simple questions. How much, if anything, has AI added to your productivity over the past three years? And how much do you expect it to add over the next three?

Three things in this picture matter. The bloc of managers reporting no material impact collapses from 18 per cent to 3 per cent: in three years, AI moves from optional to unavoidable in professional work. The largest positive band, five per cent or more added, more than doubles for managers, from 35 to 82 per cent, and roughly doubles for private individuals, from 27 to 57 per cent. There is also a thin negative tail that grows in both groups, particularly among private individuals, who report that AI has reduced their productivity and expect it to keep doing so. South Africans are adopting this technology unevenly, and the unevenness is unlikely to be undone by legislation.

The Visa and Discovery Bank SpendTrend26 report shows AI subscriptions among the fastest-growing spending categories in the country, with virtual-card payments to international AI providers rising sharply through 2025. The Anthropic Economic Index, which records anonymised global usage of Claude, provides a closer look at how it is being used.

South Africans lean more on AI for direct task delegation (“just do this for me”) and less on learning-style use (“help me understand”). Coursework is over-represented relative to the global pattern, as is work-task use. This is a country using AI to compress time on tasks people already know how to do, slightly more than to acquire new knowledge through it. We did not need an Integrated AI-Powered Monitoring Centre to know this, an agency the withdrawn draft proposed. We know this because Anthropic already publishes its data and an existing institution, the BER, ran a survey.

The point is: these initial, small surveys show South Africans have already adopted AI. The relevant question now is whether the conditions exist for consumers to extract real value from what they are already paying for, in foreign currency, every month.

In a new working paper, economists Brynjolfsson, Collis, Eggers, Kazinnik and Nguyen at Stanford’s Digital Economy Lab estimate the consumer surplus generated by tools like ChatGPT, Gemini, Claude and Copilot in the United States. Their method is direct: representative samples of US adults are asked how much compensation they would require to give up access to these tools for a month. The mean answer rose from US$98 in 2025 to US$124 in early 2026. Aggregated across an adult user base that grew from 98 million to 115 million, total consumer surplus rose from about US$116 billion to about US$172 billion in less than a year (Brynjolfsson, Collis, Eggers, Kazinnik and Nguyen 2026, Stanford Digital Economy Lab).

Two features of that estimate matter. The first is its size: the surplus accruing to users is more than ten times the providers’ US revenues. The second is who captures it. Brynjolfsson and co-authors find that usage frequency, workplace use, and paid-subscription status are the strongest predictors of valuation.

A note of caution belongs here. Babina (2026, NBER Working Paper 35123) reviews the firm-level evidence on AI and emphasises that different datasets capture different things (invention versus use, internal capability versus outsourcing, realised activity versus investor expectations), and that conclusions are sensitive to which measure is used. We are early in measuring AI’s economic effects on firms. But the consumer-surplus evidence is hard to read except one way: this technology’s welfare benefits live with the people who use it. A policy that taxes, throttles, or audits its way into that surplus takes gains that are already accruing to ordinary users.

The picture for South Africa is therefore one of distribution. We are on the right side of the productivity opportunity but on the wrong side of the diffusion gradient.

If the empirical position is that people are using this technology, gaining from it, and capturing most of the surplus, what should a policy do? The withdrawn draft proposed, almost from scratch, a National AI Commission, an AI Ethics Board, an AI Regulatory Authority, an AI Ombudsperson, an AI Insurance Superfund modelled on the Road Accident Fund, a National AI Safety Institute, an Integrated AI-Powered Monitoring Centre, and a National AI Regulatory Forum to coordinate at least seven existing regulators that already struggle to coordinate among themselves. Each body, on paper, is defensible. Together, they sit a long way from where the academic literature on AI policy has settled.

Four findings from that literature explain the distance.

The first is about knowledge. Agrawal, Gans and Goldfarb (2019, Innovation Policy and the Economy) wrote what is now the canonical economist’s statement on what governments should do about AI. They treat modern AI as a sharp fall in the cost of prediction and argue that there is no single “AI law” to be written. The relevant policy levers (privacy, trade, liability, labour, competition) already sit inside existing departments and regulators. AI policy is therefore a portfolio of adjustments to existing policy domains, not a new architecture. A document that commits in advance to certifying high-stakes applications, mandating impact assessments, requiring explainability and licensing high-risk deployment is pretending to a model of the technology that no one possesses, and that the people writing the policy possess least. The pretence has costs. It pushes innovation offshore, raises fixed compliance costs disproportionately on small firms, and freezes a regulatory image of the world that will be obsolete by the time the regulator is staffed.

Gans (2026, Journal of Law and Economics) extends the same logic to legal design. He compares ex ante licensing regimes for AI training data with ex post, fair-use-style regimes that compensate harmed parties after the fact. Because transaction costs are high and harms are hard to identify in advance, ex post mechanisms generally outperform ex ante licensing on welfare. The implication for the SA draft is that where information about harm is missing, design rules so the state acts on observed harm rather than on imagined risk. Pre-clearance architectures buy comfort by transferring the cost to the wrong people.

The second finding is about institutions. Metcalf (2025, AI & Society) argues that AI safety regulation is structurally vulnerable to capture by powerful incumbent firms. The mechanism is familiar from a long literature: information asymmetry, concentrated influence, and an agency dependent on the regulated for technical expertise. What is striking is that the conclusion comes from a scholar broadly sympathetic to safety regulation. In his account, capture is the default outcome unless the rules of appointment, conflict, sunset, and review are designed to resist it. Better intentions do not fix it. The withdrawn draft was almost entirely silent on those rules: how members are appointed and removed, what conflicts must be declared, how budgets are capped, what sunset provisions apply, what appeals are available, how performance is measured against standards set in advance. Where those rules are missing, distributional questions get decided inside administrative processes by whoever happens to be in the room.

The third finding is about heterogeneity and sequencing. Agrawal, McHale and Oettl (2026, NBER Working Paper 34953) study AI in science and document a “jagged frontier”: AI’s marginal returns and failure modes differ sharply across stages of work and across domains. The implication for policy is that comprehensive, equally weighted plans have lower expected welfare than narrower interventions targeted at the bottlenecks that actually bind. The withdrawn draft was unusually comprehensive: six pillars, fifteen building blocks, a catalogue of interventions under each, all weighted roughly equally. What it lacked was sequencing: a small set of tractable problems whose partial solution would generate visible pressure on adjacent constraints. Comprehensive plans look impressive on paper. But they do not learn.

The fourth finding is the most uncomfortable for the proposed architecture, because it is about the regulator. Jin, Sokol and Wagman (2026, NBER Working Paper 35010) build a formal model of AI-augmented enforcement and find a sharp threshold result: incremental investment in AI monitoring does not, on its own, improve deterrence. Below a recognition threshold, more monitoring produces more detected violations without changing firm behaviour, because firms can adapt around imperfect detection faster than the regulator can respond. Effective AI enforcement requires sufficient investment, adequate human review capacity, and careful management of false positives, capacities the SA draft assumed rather than demonstrated. The draft modelled its AI Insurance Superfund on the Road Accident Fund, whose accumulated deficit, claim backlog, and governance record are matters of public record. And the document itself was withdrawn, after sixteen days, because nobody had checked the references. They are evidence about the regulatory capacity the document was asking citizens to assume. Jin, Sokol and Wagman’s result formalises what the withdrawal showed: under-resourced regulators do not fail safely. They add cost without changing behaviour.

What does the literature leave for policy to do? Quite a lot, but not what the draft proposed. The aggregate productivity gains from past general-purpose technologies came overwhelmingly from reallocation: resources moving from less productive to more productive firms, workers shifting between occupations, capital being redeployed faster than incumbents would prefer. Policy that wants to capture the AI dividend should protect that mechanism. The frictions that bind it are familiar and unglamorous: electricity reliability, spectrum allocation, the cost and speed of cross-border data transfer, the tax treatment of intangible investment, exchange-control rules on software IP, the depth of local venture capital, and the mobility of skilled labour. None of these requires a new regulator. All are addressable by departments that already exist.

What about concentration?

A serious objection to anything resembling a hands-off posture is that AI markets show strong tendencies toward concentration, and that doing little risks letting a small number of foreign firms set the terms on which South Africans access the technology. The objection is real, and the empirical economics is on the side of the people raising it.

Korinek and Vipra (2025, Economic Policy) show that compute, data and talent generate scale and scope economies large enough to produce natural-monopoly dynamics, tipping, and vertical integration in foundation-model markets. Athey and Scott Morton (2025, NBER Working Paper 34444) formalise the welfare cost: workers displaced by AI can be hurt twice, first by labour-market substitution, and then again through monopolistic pricing of the AI services that have replaced their tasks. Hadfield and Koh (2025, NBER chapter c15305) add that an economy of increasingly autonomous AI agents will need legal infrastructure for identity, registration and accountability that does not yet exist.

The notable feature of these papers is what they recommend. None calls for a new AI regulator. Korinek and Vipra and Athey and Scott Morton both argue for explicit competition oversight applied through institutions a country already has: merger review, scrutiny of exclusionary contracting, attention to upstream concentration in cloud and model layers. Hadfield and Koh argue for extensions to property, contract and agency law administered by ordinary courts and registries. South Africa’s Competition Commission, CIPC, and judiciary already have the relevant mandates. What they need is the empirical capacity to apply them to AI markets and the political backing to take on incumbents in cloud, foundation-model and infrastructure provision. That is a capacity-building problem for existing regulators. It is not the case for nine new ones.

There is a real risk in writing a policy that says only “step back”. AI, like every general-purpose technology before it, will produce concentration, dependence, and identifiable harms. Some will need a public response. The state should do the small number of things only it can do: basic research funding, the electricity grid, transparency on automated decisions affecting individuals, liability defaults for identifiable harms, and competition oversight applied through the institutions we already have. It should stop pretending it can do the rest. Begin with the two or three rules that matter most. Implement them with care. Defer the larger architecture until the smaller one has been shown to work.

AI is the defining general-purpose technology of this era, and it is producing real, measured, internationally documented gains in the working lives of the people who use it well. South Africans are already among those people, with no help from government and increasingly large gains in prospect. The question is whether national policy will help them go further or get in their way.

That is the test the next draft has to pass.

The full report, “Putting AI to work for South Africans”, is available at from the BER Website.

References

Agrawal, A., J. Gans, and A. Goldfarb. 2019. “Economic Policy for Artificial Intelligence.” Innovation Policy and the Economy 19: 139–159.

Agrawal, A.K., J. McHale, and A. Oettl. 2026. “AI in Science.” NBER Working Paper 34953.

Athey, S., and F. Scott Morton. 2025. “Artificial Intelligence, Competition, and Welfare.” NBER Working Paper 34444.

Babina, T. 2026. “Understanding Firms’ AI Efforts and Their Economic Impact.” NBER Working Paper 35123.

Brynjolfsson, E., A. Collis, F. Eggers, S. Kazinnik, and D. Nguyen. 2026. “What is Generative AI Worth?” Stanford Digital Economy Lab working paper, April 2026.

Gans, J.S. 2026. “Copyright Policy Options for Generative Artificial Intelligence.” *Journal of Law and Economics* 69(1): 1–19.

Hadfield, G.K., and A. Koh. 2025. “An Economy of AI Agents.” Chapter in The Economics of Transformative AI. Chicago: University of Chicago Press / NBER. https://www.nber.org/books-and-chapters/economics-transformative-ai/economy-ai-agents.

Jin, G.Z., D.D. Sokol, and L. Wagman. 2026. “Adaptive Enforcement with AI-Augmented Monitoring.” NBER Working Paper 35010.

Korinek, A., and J. Vipra. 2025. “Concentrating Intelligence: Scaling and Market Structure in Artificial Intelligence.” Economic Policy 40(121): 225–256.

Metcalf, T. 2025. “AI Safety and Regulatory Capture.” AI & Society, published online 3 August 2025.

On 8 May 1996, Nelson Mandela rose in the Constitutional Assembly to celebrate a document that had taken two years of negotiation, compromise, and no small amount of faith. ‘The Constitution whose adoption we celebrate,’ he said, ‘constitutes an unequivocal statement that we refuse to accept that our Africanness shall be defined by our race, colour, gender or historical origins.’ The chairman of that Assembly was a young Cyril Ramaphosa – later honoured with the Order of the Baobab in Silver for his work on the text.

Nearly thirty years later, the constitution is still standing. That fact alone deserves attention. In the mid-1990s, scholars worried openly about how long it might last. A new democracy, deep inequality, a dominant party, a history of political violence – the ingredients were not obviously those of constitutional longevity. Yet here we are. The Bill of Rights remains intact. The Constitutional Court has overruled sitting presidents. The document has survived changes in government, an attempted insurrection, and the slow erosion of state capacity that has come to define the post-Mandela era.

Here is a question the anniversary invites: did the particular design of the constitution – proportional representation, a parliamentary executive, a unitary state – shape South Africa’s economic trajectory? Would the country have grown faster, or spent less, had the Constitutional Assembly made different choices?

Read more

A standard economic intuition is that inefficient institutions are competed away over time. Slavery, by that logic, should have collapsed long before any abolitionist movement arose. A slave will value their own freedom above any market price. Why, then, was the contract that would have priced this so rarely written?

Warren Whatley is Professor Emeritus of Economics at the University of Michigan and a leading scholar of the economic history of Africa and African Americans. His new book, Slavery, Freedom and Development: How Africa Became the Mirror Image of Europe (Cambridge University Press), examines why the two sides of the Atlantic moved in opposite directions over six centuries.

The conversation opens in Baghdad in 1974, where Warren spent a year as an undergraduate exchange student. Working with contacts to Stokely Carmichael and Bob Moses, he helped form a Pan-African Student Union at the University of Baghdad. The episode that stayed with him was a delegation of Berber students arguing that they were African and had never been Arabised. Their case forced him to revise the Philadelphia-shaped notion of African identity he had arrived with. The book is dedicated to comrades from that year, with whom he has since lost contact.

The central argument is the following. Between roughly the thirteenth and twentieth centuries, Western Europe moves out of feudalism toward modernity, while Africa over the same period moves in the opposite direction. The two trajectories are not merely different. They are systematically inverse. Freedom develops in Western Europe as a property of place, captured in the legal idea of “free air”, while the Atlantic slave trade reshapes African political and social life around the logic of capture.

Whatley proposes a mechanism that sits between the supply-side and demand-side accounts of the trade. Call it the gun-slave cycle. European buyers exchanged firearms alongside other goods for captives. Once one society armed itself, its neighbour faced a choice: arm and raid, or be raided. The result is a prisoner’s-dilemma arms race that spreads region by region and changes the norms of conflict.

This sets up the book’s hardest question. Roman law contained the peculium: a mechanism by which an enslaved person could accumulate property and purchase their freedom. The historical record shows that wherever manumission was allowed, slaves paid above the market price for their own bodies. Yet such contracts were rarely offered, and in the United States South they were sometimes illegal. Had that path been open, Whatley argues, the subsequent history of race in the United States would look quite different. Why the market did not produce abolition on its own is, in his view, an open question for economists.

Two threads carry the argument toward the present. The first is patriarchy: Western slavery fixed slave status to the mother, which allowed free men to father both free and enslaved children and made the system reproducible without internal contradiction. The second is customary law, which in much of Africa absorbed kinship-embedded slavery and carried its norms forward into the colonial and postcolonial period. Whatley sees this as the most plausible site for expanding freedom today. On reparations, he is cautious. The arms-race framing complicates any single line of attribution within Africa, and he points instead to the local: customary law, gender equality, and political participation as the practical terrain on which freedom is won.

The song for this episode is What Home Could Be, a Suno-generated piece built from Warren’s own research. Have a listen!

This is the last podcast of season 2. A special thank you, first, to Vasti Calitz and Andri Burnett of VoiceNote productions for their behind-the-scenes work to turn our casual conversations into professional productions. And then to the brilliant Jonathan Schoots, who, brilliant as he is, is just not quite smart enough to know how to accurately price his time. Until then, I – and the many listeners to the podcast series – are the lucky beneficiaries.

Over the years, I have heard a great many apologies from students who missed a test or a deadline. Some are reasonable. Some are creative. I once had a student who, even after I had granted an extension, returned with a doctor’s certificate. Her finger was injured. She could not type. She showed me the plaster. I was close to saying I would happily accept an essay produced by the other nine fingers, even if there was the occasional letter missing, but feared she might escalate the matter.

But the most common excuse, by some distance, is this one: ‘Sir, it was a human error.’

I have always found that fascinating. On the surface, it is a kind of shrug. There was no good reason. Move along. But it carries something useful: there is a person in the room. There is a doctor, a finger, a certificate and a story that ties it all together. You can question them. You can correct them. That is what makes it human.

On Sunday, 26 April 2026, South Africa’s Minister of Communications and Digital Technologies, Solly Malatsi, withdrew the Draft National Artificial Intelligence Policy. Hardly anyone had read it. The Minister was withdrawing it because the bibliography was full of, well, ghosts. Several of the references at the back did not exist. They had been hallucinated by an LLM, copied into a Cabinet-approved policy, and waved through every layer of review until the public got hold of them.

A second fact, less remarked on, matters more. The body of the policy carried no in-text citations. The references at the back were never linked to claims at the front. Even the real ones had not been used. The bibliography was decoration before it was hallucinated.

Together, these two facts mean something simple. The document was not built on research at all.

The most common public reaction was to demand that someone be fired. The Minister has already suspended two of them.

My contrarian view is that this is the wrong response. The reason is that these hallucinated references are the only reason we now know that the policy was not built on research. A more careful drafter would have produced a polished bibliography of real papers, none of them actually used to write the text, and the Cabinet would have approved an AI policy that had been generated rather than researched. We would have read it, found nothing obviously wrong, and let it pass.

Firing the people who let the hallucinations through punishes the one feature of the document that gave its emptiness away. The lesson the next drafter will learn is cosmetic: make the bibliography look real next time.

The hallucinated references were, in this respect, a gift. They told us, before any expert could, that the underlying analysis was not real.

The philosopher of science Douglas Allchin reminds us, in his excellent recent book Toward a Philosophy of Error in Science, of something the school textbooks tend to obscure. Science is a long sequence of errors, each one slowly displaced by a less wrong successor. The Earth sat at the centre of the universe for fourteen centuries on the authority of Ptolemy, until Copernicus put the Sun there instead. Doctors bled their patients on the authority of Galen for nearly two thousand years, until germ theory replaced the four humours. Geologists read the rocks off the Book of Genesis until nineteenth-century fieldwork showed they could not have been laid down by a flood. Stomach ulcers were caused by stress, every doctor knew, until two Australians proved they were caused by a bacterium. Errors are how the system makes progress. What the system requires is that errors be locatable – visible enough that someone can come along and correct them.

The chain of correction is held together by citations. A bibliography is the visible portion of that chain – an audit trail of error. It is how a careful reader locates the basis of a claim and, once located, questions it.

A bibliography unconnected to the text cannot do this work. There is no path back, because there was no path forward in the first place. What you have is a list of names at the end of a document, and the presence of the list is meant to substitute for the work the list refers to.

Faking such a list used to be more expensive than producing it honestly. It took longer to invent plausible references than to cite real ones. AI inverts the asymmetry. This is George Akerlof’s market for lemons, applied to scholarship: once readers know some references may be fake, the signal value of all of them drops. Faking is now cheaper than honesty. The only thing that distinguishes the fake bibliography from a real one, at a glance, is whether the names happen to exist. In the draft AI policy, several of them did not. That is how we caught it.

The fix is not to be more careful with AI. The fix is to do the research.

That means reading the studies that bear on the question. It means asking what the evidence actually says about, say, the labour-market consequences of automation in middle-income countries, and citing the answer. AI can help with parts of this. It can summarise long papers. It can draft. It can even verify references; there are dozens of online tools for the verification step, and I will soon report on skills I’ve written to help with this. But AI cannot decide which papers bear on your claim and read them for you. That part is still ours.

What such a policy might actually look like, built on the research we already have, is the subject of next week’s post.

In the meantime, I will be grading student essays, waiting for the inevitable email from those who missed the deadline.

‘Sir, it was a human error.’

It always is.

The average South African adult works 15.7 hours a week.

Not the average of those employed. Not the average of those in the labour force. That is the number you get when you take every adult in the country, employed or not, young or old, and divide the hours they spend in paid work by the number of them. Across the 160 countries in a new database by Amory Gethin and Emmanuel Saez, the people of only nine countries work less. Italy, at 15.3, is one of them. France, at 16.1, sits just above us.

Why that coincidence – a South African working week nearly identical to Italy’s – is the question I investigate in this post. Because 95 years ago, someone predicted that exactly this will happen.

In 1930, in the depths of the Great Depression, John Maynard Keynes published Economic Possibilities for our Grandchildren. He argued that, once you looked past the cyclical gloom, the real trajectory of the world economy was astonishing. ‘The standard of life in progressive countries one hundred years hence,’ he wrote, ‘will be between four and eight times as high as it is.’ Compound interest and compounding knowledge would, within a century, reduce the working week to fifteen hours or so. ‘The economic problem,’ he predicted, ‘may be solved, or be at least within sight of solution, within a hundred years.’

His deadline was roughly 2030. We are four years out. And in South Africa, arithmetically, he was right. But for the wrong reasons…

Read more

Imagine you could rewind about twelve thousand years. In the hills of the Fertile Crescent, something extraordinary was beginning – and, over the millennia that followed, in the Yangtze basin, in southern Mexico, and in the highlands of New Guinea too. People were gathering wild grasses and planting the seeds in soil they had cleared. They were penning sheep and goats instead of only hunting them. They were scratching the earth with digging sticks and waiting, on purpose, for something to grow.

This is the Neolithic Revolution. After the Industrial Revolution, it is probably the most consequential event in human history. It produced the first food surpluses – and, over the millennia that followed, the cities, writing, specialised labour, kings, priests, soldiers, merchants and bureaucrats that a surplus makes possible. It reshaped almost every aspect of what it means to be human.

So how does the new draft South African History curriculum teach this pivotal moment? Under a Grade 5 topic titled ‘Food’, built around the question: ‘How do we know what people ate in the past?’

Not ‘Why did people start farming?’ Not ‘What changed when they did?’ What did they eat.

For the past few weeks, the public debate about the amended History Curriculum and Assessment Policy Statements has focused on one thing: its Africanness. Is it too Afrocentric? Does it sideline the French Revolution, the civil rights movement, Jan van Riebeeck, the Holocaust? Is it patriotic history dressed up in new clothes? These are serious questions, and serious historians have asked them.

But there is one major question that the curriculum has failed to address that I think is the more important one.

For most of human history, almost everyone was poor. Not poor as we understand the word today. Poor as in one or two dollars a day, in today’s money, for nearly all of humanity. Life expectancy of about thirty years. Child mortality of around forty per cent. Famine always one failed harvest away.

Then, roughly 250 years ago, something changed. In pockets of north-western Europe, and eventually across much of the world, living standards began to rise. First slowly. Then astonishingly fast. The average human today is eighteen times richer than in 1800. Global child mortality has fallen from 43 per cent to under four. Extreme poverty, which afflicted nearly half the world as recently as 1980, has fallen below ten.

This is the story of how humanity escaped poverty. It is the central story of the past two centuries. And it is still being written, not least in South Africa, where we remain caught between the world we came from and the one we are trying to build.

Here is the striking thing. In the new draft curriculum, this story is almost entirely absent.

Economic history – the study of how societies become prosperous – has been sidelined. The word ‘productivity’ does not appear in the Intermediate Phase curriculum. The idea of institutions as the rules that determine who benefits from trade and innovation is nowhere. Comparative advantage, structural transformation, human capital: absent. The Industrial Revolution is taught as a story of child labour and European imperialism, not as the moment when, for the first time in history, humans figured out how to generate widespread prosperity. The Mineral Revolution is taught almost entirely through the lens of exploitation. South Africa’s economic story, in Grade 12, ends in the 1930s. Nothing on what happened after 1994. Nothing on why growth stalled after 2008. Nothing on what young South Africans might do about it.

Put differently: a learner can complete twelve years of History without once being asked why some societies are rich and others are poor. A learner can finish Grade 12 without hearing an optimistic, evidence-based account of how humans have, repeatedly, figured out how to live better.

This begs the question: what is the purpose of historical learning? In the post-apartheid era, much has been said about the importance of History for nation-building: preparing young learners to be well-informed, responsible citizens. In 2018, a Ministerial Task Team recommended, rather ambitiously, that History should be made compulsory up to matric. That all sounds rather admirable, except that it’s unrealistic if no one wants to take the subject to begin with.

In the United States the number of students graduating with humanities degrees has fallen for eight consecutive years. History as a major has collapsed. Though the subject has always been relatively small, at its peak History majors accounted for 5.7% of US bachelor degrees in the late 1960s, in 2019 that number was less than 1.2%. Why? Setting aside questions of employability and earnings for History graduates, the bigger problem seems to be that History has a reputation problem. At best, the subject is perceived as dry fact, at worst, it gets dismissed as irrelevant and impractical.

So what’s the problem? It seems it is a question of relevance. The subject is important, but in an abstract sense. People want to know more about the past because they believe historical knowledge is intrinsically important. But a subject which is abstractly important struggles to attract builders – the people who want to make things, start things, fix things. Those students go elsewhere. They head for engineering, computer science, business where the link between action and outcome are more readily apparent.

A history that fails to give learners an account of how change actually occurs – how poor societies have, in specific and identifiable ways, become richer, freer, healthier – fails to empower them. Empowerment comes from whether a subject explains how change happens. Economic history excels at identifying what incentives people faced, how institutions shaped behaviour, how small change compounds over time. It turns facts into mechanisms and supplies actionable insights.

Without that, how do we teach a young South African that poverty is not destiny?

None of this is to say that slavery, colonialism and apartheid should be taught less. They should be taught well – and economic history can help teach them better.

Why did the transatlantic slave trade happen? Not because Europeans were uniquely cruel. But because an institution took hold that made human beings into property – a form of capital that could be bought, mortgaged and put to work on sugar plantations at a scale no free labour market could supply. And because that demand, meeting European firearms, transformed African political economies too: kingdoms that could capture and sell people acquired guns, wealth and the coercive power to capture more. The trade wrote itself into institutions on both sides of the Atlantic. And why were its consequences so lasting? The economist Nathan Nunn has shown that African countries from which more enslaved people were taken are significantly poorer today – not because of some inherent disadvantage, but because the slave trade destroyed social trust. Communities that had to fear enslavement by their own neighbours grew mistrustful of strangers, and that mistrust still shows up, measurably, in surveys today.

Economics does not diminish the horror of slavery. It deepens it, by showing that the damage extended far beyond the millions of lives directly destroyed.

In the report I submitted to the Minister last week, I make a simple argument. Nothing I am suggesting requires a new curriculum. Nothing demands extra teaching time. It is a shift in framing: bring the economic dimension back into topics already taught, so that learners leave school with a fuller picture of the human past.

Take Grade 6, where learners study the trans-Saharan trade. The topic already describes gold, salt and cowrie shells moving across the desert. Add one question: what made Mali rich? A big part of the answer is exchange – Mali controlled the gold fields of Bambuk and Bure to the south, the Saharan salt mines lay a thousand miles to the north, and caravans crossing between them enriched elites on both ends. Mali taxed, conquered and enslaved too. But a great deal of its fabled wealth came from trade rather than plunder – and that is a powerful corrective to the assumption that African wealth, wherever it appears, must have been taken from someone else.

Take Grade 8, where learners study the Industrial Revolution. The framing is ‘what were the inventions, what were the social consequences?’ Add one more question: why did it happen in Britain, and not in China or India, both of which had been much richer a few centuries earlier? The answer – which involves institutions, property rights, scientific culture and incentives for innovation – is one of the most important questions in all of history. It is also directly relevant to a country still negotiating its own industrial transformation.

Take Grade 12, where we currently stop the South African economic story in the 1930s. Bring it forward. Teach the Mandela-Mbeki years, when growth returned and poverty fell. Teach the post-2008 stagnation. Teach state capture and the dismal economic performance, exposing the contrast with other developing countries that have outpaced us. Learners deserve to understand the country they are inheriting.

I do not make this case only as a critic. The ten South African chapters of my book Our Long Walk to Economic Freedom cover precisely this territory, from pre-colonial economies to the present. From next year, the full book will be freely available – I would love to see it in the hands of every History teacher who wants it. Alongside the book, a free-to-read graphic novel based on ten of its chapters will be published in January 2027, aimed at a high school audience. Both are offered as resources for any teacher, learner or curriculum developer who wants to bring this story into the classroom.

This is a debate worth having. The Minister and her officials have opened the curriculum for public comment precisely because they know that an early draft is not the final text. There is time to add the economic dimension without disturbing the structure they have built.

Luckily, we still have the opportunity to change this. A review commissioned by the Department of Basic Education in 2024 found that enrollment into high school-level History has actually increased in recent years. Tellingly, the report suggests this may be because History is perceived as easier than subjects like Physics. All the same, we risk losing this golden opportunity. As I wrote in my report, History education has always been important but it is especially important today. The generation now entering South Africa’s schools will inherit an economy that has not grown in per capita terms for over a decade, where unemployment stands above 30 percent and where inequality remains among the highest in the world. These are economic outcomes, and they have historical causes. A History curriculum that equips learners to understand those causes, and to imagine how they might be overcome, is a necessity.

If we learn anything from history, it is that the future is likely to be better. But it requires us to wish it so.

An edited version of this post first appeared on News24 – my final column for the platform, as my contract came to an end in April. From here on, the twice-monthly columns will continue exclusively on Our Long Walk. Writing like this is now supported directly by readers, and if you’ve valued the column over the years, a paid subscription is the most direct way to keep it going. I would be grateful for the support.

Yesterday I gave a talk to twenty-five of Stellenbosch University’s Heads of Department about artificial intelligence. I wanted to write it up, partly because the argument is easier to follow as an essay than as twenty slides, and partly because I suspect others might find it interesting.

One thing to mention first: When I started preparing the talk a week ago, two of the tools I ended up using did not yet exist. Opus 4.7, the model I use most, was released only last Thursday. Claude Design, which I used to build the slides, appeared the next day. If I had given the same talk a month ago, several of my examples would not have been available. A week from now, parts of this essay will already read as slightly dated. When I wrote a post here listing the AI tools I use, only six months ago, almost nothing on that list would survive today. That is the tempo we are working at, and it is the first thing a Head of Department needs to understand.

It helps, before going further, to place this in the frame economic historians use. General-purpose technologies are an old story: the plough (see an earlier piece here), steam, electricity, the computer. Each reallocated labour rather than eliminating it, but each also rewarded some people, places, and organisations sooner than others. The interesting question has never been ‘will it replace us?’ It has always been ‘where does labour go, and who gets there first?’ AI fits that pattern, with one new wrinkle. It automates cognition itself, and its speed of diffusion is unusual even by the standards of earlier GPTs. That combination is what makes the present moment genuinely different…

Read more

A good friend recently began to work for a large firm in South Africa. He tells me that he’s been surprised at the productivity of the team; having just returned from a venture firm in Europe, where he thought they were working hard to keep the dream of future gazillions alive, he expected a slightly more relaxed lifestyle here. But it is just the opposite. Their main markets are outside of South Africa, so they have to compete not only against the best in South Africa, but the best globally. Success, here, doesn’t just mean higher (future) returns for the owners. It means higher salaries, more jobs – for both skilled and semi-skilled workers – more taxes for government. In short: it means economic growth.

His story got me thinking. South Africa has a Department of Small Business Development. It has programmes, grants, and incubators dedicated to helping entrepreneurs start small. It does not have a Department of Large Firm Competitiveness. Perhaps it should.

That is because the uncomfortable truth is that small firms – however admirable their hustle – are not the ones that will move the needle. Large firms will.

That claim sounds unfashionable, even harsh. So let me show you the evidence.

A striking new paper by Zhang Chen, forthcoming in Econometrica – one of the most prestigious journals in economics – documents a simple but powerful fact: as countries grow richer, the right tail of the firm size distribution thickens. In plain language: economic development means more large firms, and larger large firms. This is not a comparison of apples and oranges – rich countries versus poor countries at a single point in time. It holds within countries over time. As an economy grows, its biggest firms grow even bigger relative to the rest. Chen shows this for a panel of 33 OECD countries from 2008 to 2017 and for the United States over four decades. The relationship is robust across sectors and across different ways of measuring firm size.

The pattern is remarkably persistent. Spencer Kwon, Yueran Ma, and Kaspar Zimmermann collected data on the size distribution of US businesses stretching back a full century. Their finding, published in the American Economic Review in 2024, shows that in the early 1930s, the top 1 per cent of US corporations held about 70 per cent of total corporate assets. By 2018, that figure had risen to 97 per cent. The top 1 per cent’s share of receipts – a proxy for sales – climbed from around 60 per cent to over 80 per cent. This is not a recent Silicon Valley story. It is a hundred-year secular trend. Each generation, it turns out, believes itself to be witnessing something new. Each generation is wrong. Bigness has been rising for as long as we have measured it.

But why? Is it just that rich countries happen to have big firms, or is there a deeper mechanism at work?

Chen’s theoretical contribution offers a compelling answer. He builds a model of what he calls ‘idea search’: firms improve their productivity by learning from other, more productive firms. Think of it as economic osmosis. A mid-sized manufacturer does not figure out a better logistics system from scratch. It learns by hiring someone who worked at a bigger firm, by observing what a competitor does, or by absorbing knowledge through a supply-chain relationship. When firms search for better ideas – new technologies, new processes, new ways of organising production – they are more likely to learn from the most productive firms in the economy. And those are, almost by definition, the large ones.

This creates what economists call a ‘diffusion externality’. When a large firm innovates or adopts a better process, the benefits do not stop at its factory gates. Its suppliers learn. Its competitors are forced to adapt. Its former employees carry knowledge into new ventures. The entire economy benefits. Put differently, when a large firm in Johannesburg figures out how to do something better, that knowledge ripples through its suppliers, its competitors, and eventually the entire economy.

The problem is that large firms do not capture the full social value of this diffusion. They invest in idea search for their own benefit, not for the benefit of the firms that learn from them. The result: they under-invest relative to what would be socially optimal. Chen shows that policies favouring large firms can correct this market failure and improve welfare. This is not a defence of cronyism or monopoly. It is a precise economic argument about externalities – the same logic that justifies subsidising education or vaccinations.

What does this mean for South Africa?

Start with exports. Marianne Matthee and her co-authors showed, in a 2017 study published in the South African Journal of Economics, that only large, productive firms manage to export manufactured goods from South Africa. Small firms lack the scale, the knowledge, and the networks to compete in international markets. If we want South Africa to earn foreign exchange, we need firms big enough to break into those markets.

And here is the rub. South Africa already has firms like that – just not nearly enough of them. SARS tax data show that barely a thousand firms employ 1 000 or more workers. They make up roughly half a per cent of all formal firms, yet they account for fully half of all formal employment.

The trouble is what those giants do. South Africa’s large firms tend to come in two flavours: resource extractors that ship commodities, and domestically oriented champions in retail, banking, and telecoms that sell mostly to South Africans. The category we are missing – the large, export-oriented, internationally competitive manufacturer of the kind my friend now works for – is precisely the one Chen’s model tells us matters most for growth.

Then consider market structure. Friedrich Kreuser, Michael Kilumelume, and Rulof Burger, in a 2024 UNU-WIDER working paper, studied markups and mergers in the South African economy. They found that large mergers affect market dynamics through supply-chain integration – a channel that is often overlooked in the competition debate. The policy question is not whether large firms should exist. It is whether they face enough international competition to keep them sharp. A large firm shielded by tariffs and red tape is a drag on the economy. A large firm forced to compete globally is an engine of growth.

To put it into perspective: a single globally competitive firm like Shoprite employs over 150,000 people across Africa. It trains workers, builds cold-chain infrastructure, and anchors local economies in towns where it is often the largest private employer. Discovery, Naspers, Bidvest – these firms sit at the centre of vast supply chains, generating demand for thousands of smaller suppliers. They pay the bulk of corporate tax. They invest in the technologies that other firms eventually adopt. A thousand survivalist spaza shops, however important for the livelihoods of their owners, do not create the same kind of systemic growth. They are a symptom of an economy with too few good jobs, not a cure.

Of course, this does not mean small firms are worthless. They provide livelihoods. They are seedbeds for entrepreneurship. A handful will hopefully grow into the large firms of the future – and a healthy economy needs that pipeline. But the policy emphasis is backwards. Small firms benefit most when large firms thrive – as suppliers in their value chains, as employers of skilled workers who later start their own businesses, as anchors of local economies that generate demand for everything from catering to courier services. Chen’s model makes this precise: the diffusion externality flows from large firms to the rest of the economy, not the other way around.

The problem is not that South Africa has too few small firms. It is that too few of its large firms compete globally. And when those firms are hemmed in by red tape, unreliable electricity, crumbling rail networks, and policy uncertainty, the damage is not limited to their balance sheets. It cascades through the entire economy.

Which brings us back to my friend. His large firm’s intensity is not an accident. It is the price of global competition. And it is precisely that competition – gruelling, relentless, sometimes exhausting – that makes the firm a growth engine for the rest of the economy. Every rand it earns abroad, every idea it absorbs and diffuses, every worker it trains lifts the broader economy in ways that no government programme for small business can replicate.

The lesson? If South Africa wants jobs, stop romanticising smallness. Make sure the big firms can compete with the best in the world. Their success will not just enrich their shareholders. It will ripple through the entire economy – into the salaries of their workers, the orders of their suppliers, and the tax receipts that fund schools and clinics. Bigness, it turns out, is goodness.

Imagine a small group of people walking across a grassland in what is now KwaZulu-Natal. The year is 220,000 BCE. They are not hunting. They are not gathering food. They are heading, with intent, to a particular outcrop of dark grey rock above the Jojosi River. They have walked this route before. So did their parents, and their parents’ parents, probably for 5000 generations or more. The stone they want is hornfels – a hard, fine-grained rock baked by ancient magma into something that flakes beautifully. They will arrive, knock large blocks into long blades, and carry the blades away to use somewhere else. And then, over the next hundred thousand years, others will keep coming back to do the same thing.

That picture comes from a paper just published in Nature Communications by Manuel Will and an interdisciplinary team working at the Jojosi Dongas, an eroded landscape in the grasslands east of the Drakensberg. The team excavated several thin layers of stone tools across three sites, fitted hundreds of flakes back into the cores they came from, and dated the sediments above and below. The result is the first clear, dated record of people returning to a single rock source for the same purpose – and only that purpose – for more than 100,000 years.

Until very recently, the standard story of how Stone Age people got their tool stone went something like this. They were on the move – hunting, gathering, following water – and when they happened across a useful rock, they picked it up. Archaeologists call this ‘embedded procurement’: stone-collecting was a by-product of doing something else. Making a special trip to a particular place to get a particular rock was thought to be rare, and late, perhaps the sort of thing Neolithic farmers did when they sank flint mines into the chalk of southern England. Pleistocene hunter-gatherers, the textbook said, did not really do this.

Jojosi shows that they did. And much earlier than anyone thought.

Read more