Code Behind Growth

In 2019, the Masakhe Ladies' Stokvel in Cape Town, forty women, agreed to put R2,500 into a common pot every month. Each month one took the whole pot home, around R100,000, on the strict condition that it go to building materials and labour. They rotated until everyone had received. Within three years the group had built thirty-six houses. No bank issued a loan. No regulator approved a product. No economist recorded the transaction. And it worked better than most formal credit would have.

That arrangement has a name, the stokvel, and it is not a relic. More than eleven million South Africans belong to one, moving around fifty billion rand a year through roughly eight hundred thousand groups. So here is a question development policy keeps getting wrong: if a system this large, this durable and this trusted is not a real institution, what is?

Spend any time near these systems and the official vocabulary starts to grate. We call them informal, as if they were a rough draft of the real economy, a phase to be grown out of. After enough fieldwork you begin to suspect the labelling is backwards. So it helps to borrow a frame from a different trade entirely.

Ask anyone who builds software and they will tell you a machine runs in layers. At the bottom sits a kernel almost no user ever sees, the part that manages memory and brokers trust between everything above it. On top of the kernel runs an operating system. On top of that run the applications you actually open. Pull out the kernel and the brightest application on the screen goes dark.

An economy is built the same way. The layer we measure, the banks, the listed firms, the figures that move markets, is the application layer. It is the part with a screen. Beneath it runs an operating system that policy rarely inspects: the stokvel, the trader who advances stock on a handshake, the savings circle that clears every Friday. And beneath that runs a kernel made of the one asset no balance sheet can hold. Trust. Read the stack from the top and informality looks like noise. Read it from the bottom and it is the thing holding the rest up.

The orthodox view treats informality as a transitional defect to be regularised away. Register the traders, title the land, bank the unbanked, and the informal will dissolve into the formal as the economy matures. The difficulty is that informality has refused to shrink on cue. By the International Labour Organization's reckoning, the large majority of Sub-Saharan employment is informal, and it has stayed that way through decades of growth, through structural adjustment, through the mobile-phone revolution. A defect does not survive that much pressure. A load-bearing wall does. And on the measures that actually matter to the people inside it, the informal system tends to win.

There is a deeper reason the system stays invisible, and it has little to do with oversight. The figures cannot see it because they were never built to. National accounts, the machinery that produces GDP, were assembled in the 1930s and 1940s to do one urgent job: count market production, first to size the Depression, then to mobilise economies for war. The man who built the first United States accounts, Simon Kuznets, warned the Senate in 1934 that the welfare of a nation could scarcely be inferred from such a measure. The warning was filed and forgotten. Anything outside the market boundary, subsistence, care, the pot that clears on a handshake, was left out by definition, not by mistake. So when the data calls the stokvel invisible, the data is reporting the edge of its own lens, and mistaking that edge for a hole in the economy. We have been reading a continent through an instrument designed to miss the part that holds it up.

The word informal does the damage. It implies an absence: no rules, no enforcement, no structure. Sit in on a stokvel meeting and the opposite is true. There is a constitution, a treasurer, a payout order, and an enforcement mechanism most magistrates would envy, the cost of facing your neighbours. The same architecture recurs across the continent under different names: the esusu of Nigeria, the susu collectors of Ghana, the chama of Kenya. Different language, identical machine.

Here is the part the orthodox account never quite explains. Why is formal credit so punishing for exactly the people who most need it?

The textbook answer is risk. The honest answer, after enough time near a loan book, is distance. A bank charges a thin-file borrower a steep rate because it cannot see them. No salary slip, no history, no collateral it can repossess cheaply. So it prices the darkness. It builds branches to find you, scoring models to guess at you, collections teams to chase you, and a margin on top for the chance it guessed wrong. Strip all of that away and most of the rate was never the cost of money. It was the cost of not knowing who you are.

A stokvel pays none of it, because the lender and the borrower share a taxi rank. Proximity quietly does the work an entire risk department is built to approximate. The rate falls to zero not because the group is generous, but because at close enough range there is almost nothing left to price.

Look again at that rate sliding to zero, because the zero is a trick of the eye. The rand figure is real enough. But nobody walks into a stokvel pledging nothing. They pledge something a bank could not book as an asset and could never repossess: their standing in a group they cannot leave. Miss a payment at a bank and you lose a score, and at worst a thing you owned. Both can be rebuilt. Miss your turn in a stokvel and you lose your neighbours, your name at every door on the street, the next ten years of being someone people trust. No loan book on earth holds collateral that bites that hard.

So the two systems are not pricing different risks. They are pricing the same risk in two currencies. The bank denominates distance in money and secures it against property it can seize. The stokvel denominates distance in belonging and secures it against something it can never seize and you can never escape. Measured in the only currency that matters to the person borrowing, the interest-free loan is the most expensive credit in the country. Which is precisely why it is the one almost no one defaults on.

The most influential case for formalisation came from Hernando de Soto: the poor sit on trillions in dead capital, assets they hold but cannot borrow against for want of formal title. Give them title, and the dead capital wakes up as collateral, credit and growth. It is an elegant theory, and the evidence for it has been quietly deflating. Large titling programmes have repeatedly failed to produce the promised lending, because the binding constraint was almost never the paper.

What formalisation usually does instead is swap a working informal institution for a formal one with the right form and the wrong function: a titling office that cannot process claims, a registry no one trusts, a bank that still will not lend against the freshly titled plot. The form of formality arrives, the function does not, and the system it displaced is left weaker for the disruption. The reform takes aim at the coordination people genuinely rely on and hands them a costume in return.

The question was never how to formalise the informal economy. It is how to build with it. Diagelo reads informal systems the way a good engineer reads existing infrastructure: as something to connect to, reinforce and extend, not bulldoze and pour again from scratch.

There is hard proof this works, and it scales. When one large South African bank let stokvels open and run accounts digitally without dissolving the groups, members earned about 256 million rand in interest in 2025 and avoided some 211 million in withdrawal fees, money returned not by changing how people save but by changing the plumbing around them.

Kenya ran the bigger experiment. M-Pesa did not march people into branches; it digitised the money-by-trust behaviour they already practised, riding a dense network of corner-shop agents rather than replacing it. Traced over a decade by the economists Tavneet Suri and William Jack, mobile money lifted an estimated 194,000 households, about two percent of the country, out of extreme poverty, with the largest gains going to women who moved out of subsistence farming and into business. Function was added at the scale of a nation. The institution underneath was left intact.

Return to those forty women and their thirty-six houses. No statistic recorded the pot, no capital flowed into it, no formal institution touched it, and it built more real wealth than the formal credit those women could never have reached. Multiply that by eleven million members, then by a continent, and you are no longer looking at the margins of the economy. You are looking at the operating system, the layer everything else runs on top of, hidden in plain sight because it does not arrive in the legal form we were trained to recognise.

So before the next reform sets out to formalise, bank or title, it is worth asking the engineer's first question: what already works here, and why? Read the kernel before you rewrite it. It has been running the real coordination all along, while the layer with the screen took the credit. The instrument was built for another economy. What it cannot count is not what is missing here. It is what holds everything up.