The road from Kampala to Fort Portal has been under construction for as long as anyone in our delegation can remember. White dust coats the trees on either side of it, as if the rains had been replaced for a season by snow. Our driver picks his way around the potholes; the Toyota raises a long pale wake behind it. The road is, in theory, the approach to Uganda's self-declared tourism capital. In practice, it is the road tourists complain about and Ugandans no longer bother to. Like many infrastructure projects in this country and others, it is delayed, and its delay has become a fact of life rather than a story. The reward for the bumpy hours is the air. The Rwenzori region opens onto trimmed tea plantations and crater lakes; the mountains, when they are not under cloud, look like the kind of geography invented by a Renaissance painter who had never seen Africa and was making it up. Most of the vehicles on this road are heading further south to Bwindi, where their occupants will stand at conversational distance from a gorilla. We turn off earlier. We are not here for the gorillas.

In Fort Portal, our delegation — a small number of representatives from European embassies — is welcomed by Florence. She is the lawyer who, for more than two decades, has led a regional anti-corruption coalition operating across this part of western Uganda. The meeting room is small and decorated, almost crowded, with awards. The rest of the office breathes humility. Water bottles are distributed. We sit. She gets straight to business.

The number first. Her organisation, she tells us in the level voice of someone who has been asked to repeat it more than once, has recovered around six hundred million shillings of taxes that had been embezzled or stolen by local government. Roughly one hundred and fifty thousand dollars. Then she turns the page of a notebook and starts going through the rest: the contracts that were never executed, the materials that never arrived, the work that was billed but not done. She reads the list the way an accountant reads a ledger. She is not theatrical about any of it.

The method is monitors. The Coalition trains local people — at last count, more than a thousand of them, across roughly ten districts in south-western Uganda — to look at a public construction project, look at the document that says what is supposed to be built and how, and report any gap between the two. The document is called a Bill of Quantities; the Coalition has access to them through a working partnership with the Public Procurement and Disposal of Public Assets Authority. The reporting tool is unglamorous: a small mobile application called the Contract Monitoring System, which our governance programme had helped the procurement authority to put together. Florence is the first person to point out, with no false modesty, that the tool is not revolutionary. What is revolutionary, she says, is that the monitors actually use it.

She takes us outside to show us. A few minutes' drive from her office is a recently completed health facility. It is the kind of building that, on paper, was going to deliver a step change in care for a particular sub-county: maternity, surgery, basic equipment, the works. When her team did their site visit at the time of handover, the theatre was empty. The contractor had built the room and signed the form. The equipment for which the room had been built was simply not there. What was supposed to be a place where caesareans could be performed was, on the day it was opened, a four-walled gap in the supply chain. Her team refused to let the file close. The equipment, eventually, arrived.

She summarises the lesson by way of a sentence I have not stopped thinking about. "If you are not present," she says, "contracts are not implemented as they should." She does not deliver it as a slogan. She delivers it as a finding.

This work, it should be said, is not a hobby. Her monitors — and she does not emphasise this, because she does not want us to sentimentalise her people — sometimes receive death threats. They show up on construction sites, in red dust at the end of a working day, holding the document that the contractor was supposed to deliver against, and they ask, in front of witnesses, whether what was specified is what is there. This is, to put it dryly, a profession with friction. Suppliers who are over-quoting connive with officials who are signing; officials who are signing connive with structures that are protecting; the chain is not always easily broken from the bottom. And yet the monitors, who are unpaid volunteers and who have nothing to gain from this beyond the public good of the thing, keep doing it. Florence calls them, with affection, an army.

The arithmetic, even on the back of an envelope, is the kind that ought to wake up a treasury. Six hundred million shillings, recovered by one organisation, in roughly ten districts, over a period of years. Uganda has nearly a hundred and fifty districts. If even a fifth of them had a coalition like this one, the recovered figure would not be in the hundreds of millions but in the tens of billions, every year, with no aid budget required to produce it. And the recovery figure, in any case, undersells the scale of what is happening. Most of the public money saved by this kind of work never appears on a recovery list at all; it is the shape of the contracts that the contractor decided to deliver properly because somebody, this time, was watching.

This is, in my experience, the part of the development story that gets the least airtime in donor capitals. There is a great deal of talk in Berlin and Brussels about strengthening anti-corruption institutions — the office of the Auditor General, the Inspectorate of Government, the Public Procurement Authority. Those are necessary. They are not sufficient. The line of defence that actually moves money is the layer underneath them — the small NGO with a battered Toyota and an army of volunteers who happen to know which contractor is whose nephew. It is the cheapest layer of governance you can buy. Every shilling spent on a coalition like Florence's returns several thousand to the public purse, and the figure is conservative.

It is also, of course, the layer that faces the most friction. Small civil society organisations in Uganda, as in many countries, operate in an administrative environment that places significant compliance demands on them — operating permits, recommendation letters, memoranda of understanding, district by district. These requirements are not costless. Florence, who is a lawyer, can navigate them; many of her smaller peer organisations cannot. The structural irony is not subtle: the organisations whose work recovers public money are often spending their limited resources on the administrative burden of being allowed to exist. This is a pattern worth naming, not because it is unique to Uganda, but because the consequences are particularly visible in a context where the civil society layer is already thin.

I left Fort Portal that evening with the small, embarrassing realisation that I had spent the morning being more useful than I had been in months. Not because of anything I had done — I had been a guest in a chair with a water bottle, and I had asked the kind of questions guests in chairs with water bottles are expected to ask — but because the people I had been brought to meet were doing, with no fuss and very little money, the kind of work that moves a country.

Donor budgets, in my experience, prefer to pay for grand things: ministries, summits, frameworks, capacity-building plans whose acronyms multiply across the years. The bet I came to believe in, by the time my own posting in Uganda ended, is the unfashionable opposite. Find the lawyer in the small office decorated with awards. Find the army of unpaid monitors with mud on their shoes. Pay for the petrol of the next Toyota. Underwrite the legal fees the next time a contractor sues for defamation. Where you can, stay out of the way. The work has its own velocity. What it needs from the outside world is the room, and the cover, to keep going.

If you are not present, Florence said, contracts are not implemented as they should. I think about that line a great deal now. It applies to construction. It applies, I have come to believe, to most things.

The guards in our condominium in Nakasero spent most of their nights on TikTok. They were polite, alert when they needed to be, and, in the long stretches between, illuminated by the small rectangle in their hand. The videos played at full volume because nobody had taught them, or anyone else around them, that headphones were a thing you bought. Around two in the morning the soundtrack of the building was a mosaic of looped Ugandan dance, Nigerian comedy, and whichever American conspiracy had made it that week to the ear of a man earning eight dollars a shift. I would walk past them on the way back from a late dinner and feel, every time, the same uneasy mixture of affection and dread. They were not lazy. They were captured.

Uganda is the second-youngest country in the world. Median age, around fifteen. Half the country has not yet sat for a final school exam, and the other half is barely older than that. A forty-something European like me — in the demographic vocabulary that operates here — could already plausibly be a grandfather. The smartphone, more or less absent from this country a decade ago, is now in the pocket of fifteen million Ugandans, around two-thirds of all adults; the figure has doubled since 2020, pushed by a flood of cheap Chinese handsets that made connectivity finally affordable for people who had never before had a written address. This is the most digitally connected generation Uganda has ever produced. It is also, in my observation, one of the most exposed.

The exposure is what is rarely named in the cheerful presentations about Africa's digital leap. Owning a smartphone is not the same as being able to read what it is doing to you. The condominium guards on TikTok did not know that the application logs the device's location every few minutes; they did not know that scrolling for three hours teaches the algorithm exactly which kind of misinformation will keep them awake for a fourth; they did not know what the words ad identifier and data broker referred to, and could not be reasonably expected to. In Europe, GDPR is at least a four-letter argument the user has the right to make. In Uganda, that argument has no statutory ground to stand on. The country's data is being harvested at the speed at which its phones are being charged, and a sizeable share of it leaves the country before the user has finished the video.

The line is old; in Uganda it has begun to feel literal. The same data flows that allow platforms to sell better-targeted advertising to a Kampala middle class also create, as a by-product, extraordinarily detailed pictures of individual behaviour — where people go, who they call, what they read, what they share. SIM-card registration, mandatory and tied to national ID, makes that picture easy to assemble and easy to cross-reference. This is not unique to Uganda: it is the architecture of the modern digital economy, operating in a context where the legal frameworks that might constrain it are still being built. The journalist, the community organiser, the twenty-year-old in Kawempe who reposts the wrong cartoon at three in the afternoon — all are generating data whose downstream uses they cannot fully anticipate or control.

The commercial and the political dimensions of this are not, across much of the developing world, two separate problems. They share the same pipeline. Companies built in California and Beijing do the heavy collection; the resulting data exhaust has value far beyond advertising. What this means for civic life — for the space available to journalists, activists, opposition voices, ordinary citizens with inconvenient opinions — depends enormously on the strength of the legal frameworks that govern it. Where those frameworks are weak or absent, the asymmetry between those who collect data and those who generate it becomes very wide very fast.

Around all of this sits the next thing. Artificial intelligence, in the form most Ugandans will encounter it, is already arriving. It is arriving as an interface in English, trained predominantly on English, with the cultural gravity of the places where its training data was scraped. During a workshop with civil-society partners I once asked ChatGPT, in front of the room, to produce images of a doctor, a nurse, and a ship captain. The doctor was a white man. The nurse was a white woman. The captain was another white man. I had opened the application from a hotel in Kampala. The model had not noticed.

This is not a quirk that will be fixed by a sterner prompt. It is the predictable consequence of training large language models on a corpus that vastly over-represents the global North. Take Luganda, the most widely spoken language in Uganda, with roughly twelve million speakers — about as many as Modern Greek. Modern Greek has around two hundred and sixty-one thousand Wikipedia entries. Luganda has about four thousand four hundred. Even Swahili, spoken in some form by close to one hundred and fifty million people across East Africa, has only around one hundred thousand articles to its name. The disparity is not an oversight. It is the shape of the internet. When a model trains on the internet, it trains on this shape, and it returns predictions that preserve it.

African knowledge — medicinal, agricultural, legal, narrative — enters the room as a footnote, if it enters at all. Out of Africa's roughly two thousand languages, a great many are now under quiet pressure not just from urbanisation and migration but from the linguistic gravity of the model that increasingly answers everyday questions. A child in Kotido who asks an AI tutor for help with her homework will, in the absence of deliberate intervention, get a tutor that does not speak her mother's language and does not know her grandmother's stories. The model is not malicious. It is simply a mirror of who has spent the last twenty years writing the web.

What follows from all of this is not a counsel of despair. It is, if anything, the opposite. Uganda's young, smartphone-fluent generation is in a better position to shape what AI does in this country than any cohort that came before. They are early. They are numerous. They are, when you put a microphone in front of them, exceedingly clever about the texture of the digital life they actually lead. What they are missing — what the country is missing — is a layer of organised civil society capable of translating that fluency into policy: a data-protection law with teeth, an enforcement body that functions independently, a curriculum that treats digital literacy not as a vocational add-on but as a basic civic skill, on the order of knowing how to read a contract.

The governance programme I worked on tried, in its modest way, to fund this layer: the small NGOs that read draft bills line by line, the lawyers who could explain to a parliamentary committee what biometric ID does and does not protect, the trainers who would sit with a group of teenagers in Mbale or Gulu and walk through what their phone was actually doing while they slept. None of this work made the front page of any newspaper. It was unglamorous, often boring, and almost always underfunded next to the next round of equipment for the ministry. It was also, looking back, the work that mattered most — because every year that passed without it was a year in which the surveillance infrastructure got cheaper and the protection infrastructure got more expensive.

The phrase that kept circling in drafts of this essay — if you are not at the table, you might be on the menu — is the kind of slogan that ages well only when it is taken literally. There is, in fact, a table: it is the room where the rules for AI, for data, for platform liability, for biometric ID are being written. Uganda is mostly not in that room. The European Union is. The United States is. China is. The platforms themselves, increasingly, are. A country with the median age of fifteen and fifteen million smartphone users is, in that room, an item of business rather than a participant.

What I think about, when I think about the guards on TikTok, is not their credulity. It is the unfairness of asking them to be the last line of defence. A teenager in Kawempe who has not been taught what an advertising identifier is should not be the person on whom the country's digital sovereignty depends. That is the work of a state, of a parliament, of an organised civil society, of a generation of Ugandan technologists who exist and whom the country has not yet decided to take seriously.

The smartphone in the guard's hand is not the problem. The problem is that almost everything around the smartphone — the law, the literacy, the language of the model, the silence of the regulator — has been built somewhere else, by people whose appetite his country has not yet learned to refuse. There is still time. The generation that will set the terms of Uganda's digital life is in secondary school right now. What they are taught, and what they are protected from while they learn it, is the cheapest and most consequential investment the country could make. It is also the one most likely to be deferred, because the returns do not show up in this electoral cycle. The bet a serious country would make is the one that is, for the moment, hardest to make: that the table is more useful than the menu, and that the difference between the two is decided long before anyone sits down to eat.

On the slopes of Mount Moroto, in Karamoja, a language called Soo is spoken by about thirty people. They are the last. When they die, the language dies with them — and so does a way of naming the wind, the shape of a herd at dusk, the specific gradient of a particular kind of grief. Linguists have given Soo a code. UNESCO has placed it on a list. None of that is the same as having a child who learns it.

I lived in Uganda for almost three years, and the slow disappearance of cultural worlds is not the kind of fact that makes the front page of the New Vision. It is a quiet attrition, distributed across more than seventy languages, fifty-six ethnic groups, dozens of musical traditions, and a national museum that is itself slowly turning into an exhibit of how a country forgets itself. The building in central Kampala is a dilapidated pavilion of badly preserved bits of Ugandan history — bows, drums, a few colonial photographs — arranged with an apologetic logic that fails both as scholarship and as spectacle. A country with the cultural depth of Uganda deserves a museum that knocks you flat. Instead, it has one that whispers an apology and asks you to keep moving.

This is the part of "development" that everyone agrees is important and almost no one funds. In our governance programme, the creative sector was not an afterthought; it was a pillar. The reason was not sentimental. The International Finance Corporation puts the global creative economy at around two trillion dollars and fifty million jobs. About half of those jobs go to women. The sector employs more young people than any other sector worldwide. In Uganda, where the median age is sixteen, the arithmetic is not subtle.

Already, without much policy help, Uganda's creative industries generate close to one and a half billion dollars a year. Most of it is music, and most of the music is carried by a handful of names — Eddy Kenzo, Jose Chameleone, Sheebah Karungi — whose biographies are themselves arguments for cultural investment. Sheebah grew up in the Kawempe ghetto, raised by a single mother who disapproved of her daughter dancing in clubs. She started at fifteen. Her first hit, "Ice Cream," landed in 2011, and since then she has been the country's leading female artist and one of its most uncompromising voices on a woman's right to decide what happens to her body. There is no governance reform paper, no mid-term review, no Theory of Change diagram that produces what Sheebah has produced. A talented teenager from a slum, a microphone, and the freedom — fragile, contested — to use it.

Look, then, at how a country that produces a Sheebah budgets for the conditions that produce her. In the 2024/25 financial year, the Ugandan state allocated about nine billion shillings to the creative industries — roughly two and a half million dollars. The same year's budget allocates about a thousand times more to governance and security. Tourism, which depends almost entirely on culture and natural beauty without paying directly into either, gets four hundred and thirty billion shillings, a figure the government is proud of. The implicit theory is that culture is the substrate other sectors mine for revenue. There is a pattern here that repeats across many countries: cultural production is supported when it generates export revenue or national pride, and less so when it generates questions.

This is where the conversation usually gets dragged back into a debate about budgets and priorities, and where I think it has been wrong for a long time. The bottleneck on Uganda's creative sector is not money. The bottleneck is the political weather that artists are able to work in. History suggests that when that weather turns, the first squeeze is rarely on the economy and almost always on the culture: the comedians, the bloggers, the radio DJs, the choreographers. This is not unique to any one country — it is a pattern visible from Budapest to Kampala to Caracas. What makes it particularly acute in parts of the Sahel is the near-total absence of the infrastructure that cultural production requires. In Chad, when we worked to open N'Djamena's first permanent art gallery through AFROTOPIA, we found that painters lacked basic materials — frames, brushes, paint. Musicians had no access to recording studios or even instruments to rehearse with. Photographers and filmmakers had no editing software. Unlike Senegal or Burkina Faso, which have built genuine cultural ecosystems over decades, Chad had almost nothing — not because the talent was absent, but because the conditions for it to survive had never been created.

Felwine Sarr, the Senegalese philosopher who wrote Afrotopia, makes a point that has stayed with me through every restructuring of every Ugandan governance programme I worked on: the future of a society cannot be imagined only by people working in ministries or generals in the army. It has to be imagined collectively — by thinkers, writers, painters, architects, musicians, activists, civil servants, troublemakers. A country whose imagination is concentrated in two or three offices on Parliament Avenue is a country preparing itself, without knowing it, for stagnation. A country whose imagination is distributed — in the Kampala cartoonist, the Kabale singer, the Mbale poet, the Karamoja festival dancer — is a country that has a chance of writing its own next chapter rather than having one written for it.

This, I think, is what we as development partners often missed: we treated culture as the soft annexe of the real work. We funded electoral observation missions, anti-corruption training, judicial reform, capacity-building for ministries — important work, work I would do again. But culture understood intuitively what our logframes did not always capture: that the deepest shifts in a society are rarely produced by a bureaucrat reading a manual on procurement. They are produced by a thirty-year-old woman holding a microphone in a stadium, telling other women they could leave. By a cartoonist with a pen. By a museum that tells the truth about who a country has been and who it could be. That is where lasting change lives. We were sometimes the last to notice.

Which brings me back to Soo, and the thirty people on Mount Moroto. When a language disappears, what disappears with it is not a vocabulary but a way of seeing. The Tepeth speakers of Soo carry inside their grammar a particular relationship to a particular mountain, a way of differentiating types of cloud, a memory of certain animals, a manner of addressing the dead. None of this is preserved by writing it down in a notebook. It is preserved by being spoken to a child who answers back. There is no donor programme that will save this. There is only a country that decides, or fails to decide, that its own diversity is worth protecting — and that means, in practice, protecting the conditions in which thirty people on a mountain can keep speaking their language without being told, in seventy different ways, that doing so is backward, irrelevant, or politically inconvenient.

The cheapest budget line in the Ugandan state is the one that decides whether, in fifty years, anyone will still know how to name the wind on Mount Moroto. It is also, by a clear margin, the most important.

Years ago, stuck at a traffic light in Kampala, I greeted a street child in Karamojong. I was using one of the three or four sentences I had picked up on trips to the north-east, and I delivered it badly. The boy answered in kind and walked away. The driver looked at me with a strange expression. "So you speak that foreign language?" he said. I told him it wasn't foreign. It was a Ugandan language. Karamoja, I said, is part of Uganda. He thought about this and said nothing.

This is the opening note of any honest essay about Karamoja: it is a Ugandan region that the rest of Uganda has never entirely accepted as its own. People in the central region carry the kind of prejudice you only carry against neighbours you don't visit. The Karamojong are perceived as violent, dangerous, primitive. Most Ugandans encounter them as the shoeless street children at car windows, brought down from the dry country by traffickers and parked at intersections. None of which has made the country any more curious about who lives there.

The drive from Kampala to the regional capital is the simplest education. The banana groves and sugar cane around Jinja thin out into the rocky landscape around Soroti, and then into something stranger: a wide, dry-eyed, semi-arid country that gets perhaps seven hundred millimetres of rain a year, less than two-thirds of what Kampala gets. The Karamojong have lived here a very long time. The grammar of survival is pastoralism. The accounting unit is the cow. In the centre of every manyatta — the cluster of huts that constitutes a homestead — the cattle stand as a quiet collective bank account. They are wealth, identity, and sacrament at once. A young man without enough cows cannot marry. A young man without enough cows is, in a sense the rest of Uganda no longer fully understands, not yet a person.

A colleague once told me about his mother. She had travelled from her home in a small Karamoja town to Kampala for medical treatment, and the doctors told her, kindly, that she should stay a few more nights before going home. The next morning she was gone. She had slipped out of the hospital because she could not bear to be away from her cattle any longer. I told this story to friends in Europe, and most of them found it touching — the elderly mother and her stubbornness — and missed what it was actually about, which was not stubbornness. It was a relationship between a person and a herd that I, the foreigner, do not have an analogue for. The cow is what builds the house and what sleeps inside it. The cow is what marks every important event in a life and what is rarely killed for food, because killing the cow is the last thing you do.

It is also, in this same economy, what determines whether a woman can be married, and to whom, and at what age. A bride price in Karamoja can run to a hundred cows, sometimes more — among the highest on the continent — and the math of one hundred cows is not romantic. It pushes marriages later for poor men, earlier for desperate families, and concentrates power in the hands of older men who happen to have herds.

Closer to the centre of every Karamoja village is a tree, and under the tree, the Akiriket. This is the council of elder men, the body that decides social norms, settles land disputes, and hears the cases no formal court will reach. The men sit together on the traditional stools. Women are not in the circle at all — not at its edge, not within earshot. What looks ceremonial is in fact administrative; the Akiriket is one of the most consequential institutions in northern Uganda, and women have, for as long as anyone alive can remember, been excluded from it.

This is the institution that an organisation called NARWOA — the Nakere Rural Women Activists, based in Kotido — set out to put a chair into. The project was modest in scale and fundamental in ambition: build groups of older women whose collective authority would make it harder for the elders to decide questions about women's lives without consulting any. It is the kind of intervention that reads as boring on a donor report and is anything but boring in practice. It also complicates what an outsider thinks is supposed to happen. One of the women's-rights organisations on the eastern edge of the region was led by a man, who routinely spoke for the women he was meant to represent. My team argued about this. Did a women's-rights organisation need to be led by a woman? I have heard arguments on both sides and I am not sure either is fully right. What I am sure of is that male allies will be necessary, that the conversation will be slow, and that anyone who pretends to know the answer at the planning stage has probably not yet been in the room.

A familiar objection used to surface, in those rooms, every time we spoke about gender programmes in Karamoja: what about the boys. In Karamoja, where the literacy rate for men sits at around thirty-seven per cent, the literacy rate for women is sixteen point four per cent: about one woman in six can read. To say in good faith that we should worry about the boy child first is to claim, in effect, that a sixteen-per-cent literacy rate for women is an acceptable starting point. It is not, on its face, an equity argument.

The deeper difficulty is the one Silvia Tamale names so well: if African women are to successfully challenge their subordination, they need to develop home-grown conceptualisations that capture their specific political realities and traditional world-views. A white European man designing a women's-rights programme in Karamoja is, by definition, not the right person for this. He can make himself useful, sometimes, by sitting in the right rooms, by funding the right partner, and by understanding the limits of his own usefulness. He cannot do the conceptual work.

What I keep coming back to, when I think about Karamoja, is the geometry of the meeting. The men on the traditional stools. The tree above them. And no woman in the circle at all — the absence so complete that no one present needs to think about it. I think about the moment when a stool is added to the circle, and a woman sits on it, and the meeting stops being something decided about her without her in the room. That is not a project. That is a different country.

Karamoja deserves the chance to be that different country, on its own terms. And Uganda, in turn, deserves to discover that one of its most distinctive regions is not a foreign language but a vocabulary the rest of the country has never bothered to learn.

Every cultural space makes a decision about who belongs in it. Usually the decision is not announced. It is made through architecture — the absence of a ramp, the height of a stage — through pricing, through the language of a programme note, through the implicit understanding of who the evening is for. The people excluded by these decisions rarely appear in anyone's account of the evening, because their absence is the precondition for the event taking place as planned. They are not there to object.

In 2022, working through a project called AFROTOPIA at the cultural centre ACAMOD in N'Djamena, we tried, on two occasions that year, to make different decisions. The two events were superficially unlike each other. One involved migrants in a transit centre, invited to paint and dance and play music over four weeks in the early summer. The other was a celebration of Chad's national holiday on the 11th of August, designed explicitly around the presence and participation of artists and thinkers with disabilities. What they shared was a single underlying conviction: that a cultural space which excludes the people it has simply never bothered to include is not a neutral space. It is a space that has made a choice, and called it normal.

AFROTOPIA had taken its name from the philosopher Felwine Sarr, whose argument about African futures was, at its core, an argument about imagination: stop measuring these societies against a modernity designed elsewhere, and start asking what they might become on their own terms. For us, in the specific context of N'Djamena — a city without a permanent art gallery until we opened one, a capital where the national museum was gently collapsing into itself — the project was more immediate than philosophical. In the first months after reopening the centre, we had organised eleven concerts, open-mic evenings, a vernissage, exhibitions, and training workshops. The artists who performed brought with them the instruments of Chadian musical tradition — the sanza, the balafon, the kundi, the kakaki — filling a space that had been silent for years. It was about opening a space and then being deliberate, each time, about who entered it.

The migrants came in the mornings, twice a week through June, when the Sahelian heat was still manageable. They were staying at the International Organization for Migration's transit centre across the city — most of them benefiting from voluntary return assistance, which means they had already made the decision to go home and were waiting for the logistics to align. That particular limbo has its own texture. The future is settled in outline; the past is not something you dwell on; and the present is a room in a compound with time that needs to be survived as much as spent. Into this situation we brought brushes and instruments and fabric and four artists who asked, more or less, what people wanted to make.

The development sector has a complicated relationship with this kind of proposition. Psychosocial support, art therapy, cultural programming — these categories exist in logframes, but they occupy a defensive position. You justify them differently than you justify a water pump. The evidence base is presented as thinner, the outcomes as harder to quantify. And yet the evidence is not thin at all: the relationship between creative expression and mental wellbeing in displacement contexts is well-established. What the workshops provided — structure, a skill being acquired, the presence of attentive artists, the small rhythm of arriving somewhere with a purpose — addressed something real in people whose days had otherwise lost their shape.

What I keep coming back to, thinking about those mornings, is a distinction that feels important. The water pump and the vaccination campaign restore what was taken. Art does something different: it asserts what remains. To choose a colour, to decide how a line falls, to learn three chords — these are acts of agency in a situation designed, structurally, to suspend it. The transit centre decided when people ate, where they slept, what their next step would be. The patio at ACAMOD was one of the few places where the decision belonged entirely to the person holding the brush.

On the 17th of June the patio filled. Paintings were hung on the walls — landscapes of countries left behind, faces of people missed, images of peace, chosen by the participants themselves in conversation with the artists over the preceding weeks. A group played music. There was dancing. None of it was polished. All of it was specific, which is the only quality that distinguishes genuine expression from decoration. The event had cost less than a thousand euros in total. I have sat in donor planning meetings where that sum was spent on a single line of catering. The disproportion is not an accusation; it is an observation about what we have decided dignity costs.

Two months later, on the 11th of August, Chad celebrated its national holiday. The 11th of August is Independence Day — the day a country performs its idea of itself, assembling the symbols and sounds it has decided represent the nation. AFROTOPIA held an event that evening that was, in its own quiet way, a counter-proposal to that performance. We called it the Inclusive National Holiday. The question it posed was a simple one: who is usually absent from the stage on the day a country celebrates itself?

The answer, in Chad as in most places, includes persons with disabilities. Not because they are unwelcome in any explicit sense, but because the default settings of public cultural life — the inaccessible venues, the absence of sign language interpretation, the stages designed for audiences assumed to be able-bodied — function as a quiet, structural exclusion. Most cultural events in N'Djamena, as the concept note we had written that year observed, lacked even basic provisions for people with limited mobility. The exclusion was not intended. It was simply never thought about.

The evening began with slam poetry on the theme of inclusion, performed by Troubadour and Prince Bakounini. A roundtable followed, with the writer Sosthène Mbernodji — president of the Association of Chadian Writers and Authors — alongside a media promoter and a humanitarian inclusion specialist, debating what concrete changes would make Chadian cultural life genuinely accessible. Then came dance, and a concert featuring Yegsonne and Françoise Modjimadji — both artists with disabilities, both, that evening, on the stage.

That last detail matters more than it might seem. Persons with disabilities are often spoken about in cultural contexts — as subjects of compassion, objects of policy, beneficiaries of inclusion initiatives. Less often are they the artists, the performers, the ones the audience has come to hear. The Inclusive National Holiday was not primarily an advocacy event, though it produced recommendations. It was a concert. The argument it made was made through the music, by the simple fact of who was standing at the microphone.

There is something the two events share, beyond their common location and their common year. Both were built on the same recognition: that visibility is not a bonus, not a nice-to-have added once the serious work is done. It is itself a form of serious work. The migrants who painted on the walls of the patio were not, in those mornings, cases awaiting resolution. The artists who performed on the 11th of August were not ambassadors for a category. They were people making things, in front of other people, in a space that had decided they were worth the invitation.

Cultural spaces, at their best, are where a society rehearses its better possibilities. They are where the people normally relegated to the margins get to occupy the centre, if only for an evening — and where that occupation, repeated enough times, starts to feel like something other than an exception. AFROTOPIA was a small project in a city that most of the world could not find on a map. But the question it was asking is not a small one: who does the stage belong to, and who has been waiting, just outside, for someone to notice they were there?

Governance is one of those words that has been stretched so far it has nearly lost its shape. Donors use it to justify conditionalities. Governments invoke it to signal reform intent. Academics write entire careers around its definition. And yet, stripped of jargon, it is a simple idea: governance is about whether state power is used for the benefit of citizens at large, or for the interests of the connected few.

That simplicity matters. Because when governance fails, it is not an abstraction that suffers — it is people. Children sitting in classrooms without textbooks. Patients in health centres without medicines. Bridges that collapse because procurement contracts were awarded on the basis of connections rather than competence. Roads left unrepaired not because funds were unavailable, but because those funds never reached their destination.

Good governance has many dimensions — transparency, accountability, participation, the rule of law, efficiency, equity. It is sometimes taught as a checklist, and that is where things go wrong. Governance is not a list of institutional boxes to tick. It is a set of relationships: between the state and its citizens, between public officials and the law, between those with power and those affected by it.

At its core, good governance requires three things. First, solid mechanisms for policy development — processes that are inclusive, evidence-based, and responsive to the needs of all citizens, including those who are easiest to overlook. Second, effective implementation — the administrative capacity to translate policy into services, and the management culture to do so honestly. Third, and perhaps most importantly, credible accountability — institutions that are genuinely independent, that can investigate, report, and where necessary, sanction. Supreme Audit Institutions, human rights commissions, independent judiciaries: these are not bureaucratic niceties. They are the mechanisms through which citizens verify that power is being used on their behalf.

The independence of these institutions is not a procedural detail. It is the difference between a state that can be held to account and one that cannot. When an Auditor General can be removed by the executive for inconvenient findings, or when a human rights body is starved of funding, the entire architecture of accountability is hollowed out. What remains is the form of governance without its substance.

When we think about democracy, we think of elections. But elections take place every five years. In between, there are thousands of days during which government decisions are made, budgets are spent, contracts are awarded, and services either reach people or they do not. Democracy is not only the moment of the vote — it is everything that happens in between.

This is where citizen engagement becomes critical, and where its traditional forms fall short. Participation in budget hearings requires literacy, time, and proximity to district offices. Filing a complaint with an oversight body requires knowledge of how to do so and confidence that something will happen. For most citizens — particularly those in rural areas, or those from marginalised groups — these pathways are effectively closed.

Digital tools have the potential to change this. Not as a silver bullet, but as a genuine leveller — lowering the cost of participation, making technical processes accessible, and creating channels through which ordinary citizens can contribute to accountability in real time. The CitizensEye app developed for Ghana's Audit Service, and the Citizen Feedback Platform now used by Uganda's Office of the Auditor General, are examples of what this looks like in practice: citizens reporting geolocated service failures directly to auditors, generating evidence that shapes what gets audited and where public attention is directed.

Digitalisation offers something that previous governance reforms could not: the possibility of evidence-based decision-making at scale, in real time, based on the actual behaviour of systems and citizens rather than the assumptions of planners sitting in capital cities.

Consider a practical example. Traffic management in Kampala. An AI system analysing vehicle flow and road conditions across the city could tell the Kampala Capital City Authority not just where potholes exist, but which ones are driven over most frequently — the ones whose repair would reduce the most damage to the most vehicles, for the most people, every single day. Every shilling of public maintenance spending invested where it has the greatest impact on the greatest number of people. A utilitarian calculus in the best sense — maximising social welfare from constrained public resources, in the spirit of what Locke articulated about the purpose of political authority: to serve the governed.

The same logic applies to hospital allocation, to school infrastructure, to water access. Data — properly collected, properly disaggregated — allows planners to understand not just where need exists in aggregate, but where investment would have the greatest measurable effect on human outcomes. But disaggregation matters enormously here. Aggregate data conceals inequality. An analysis of crime across a city might suggest that street lighting is needed in certain commercial districts. But if the objective is to protect women and children from gender-based violence, the relevant data is different. The tool is only as equitable as the question it is asked to answer.

None of this is inevitable. Digitalisation does not automatically produce better governance — it amplifies the governance that exists. In the hands of accountable, citizen-oriented institutions, digital tools and AI can be transformative instruments of equity. In the hands of unaccountable ones, they can become instruments of surveillance, exclusion, and control.

We are living through a moment of acute pressure on the institutions and norms that underpin international cooperation. Belligerent nationalism is on the rise. Multilateral frameworks are under strain. The temptation in many capitals is to turn inward — to treat governance as a domestic matter, immune from external scrutiny or solidarity.

This is precisely the wrong response. The challenges that matter most — climate change, forced displacement, transnational corruption, the governance of AI — cannot be addressed by any single state acting alone. They require the kind of sustained cooperation that only robust international institutions can provide. And they require domestic governance frameworks strong enough to resist capture by narrow interests.

We have learned, through hard experience across many countries and many decades, that open societies do not sustain themselves automatically. They require active maintenance: independent oversight institutions, a free press, civil society organisations with the space to operate, and the political will to protect all of these when they come under pressure.

Lake Chad has lost roughly 90% of its surface area since the 1960s. The causes are multiple — climate change, population growth, irrigation — but one factor that received far less attention was governance. No institution, in any of the four countries that share the lake, was effectively monitoring how water was being used, by whom, or whether the rules that existed on paper were being applied in practice. The oversight gap was as significant as the rainfall deficit.

The question we worked on from 2013 onwards, through GIZ's Good Financial Governance programme in Africa, was whether Supreme Audit Institutions — the independent bodies that audit government spending and performance — could be part of the answer. SAIs are typically associated with financial probity: tracking whether money was spent as authorised. But performance auditing, the practice of evaluating whether government programmes actually achieved their objectives, opens the door to much broader questions. Did water management policies work? Were regulations enforced? What happened to the communities they were meant to protect?

Working with AFROSAI, the African umbrella body for SAIs, we supported the audit institutions of Nigeria, Niger, Chad and Cameroon to conduct a joint environmental performance audit of the Lake Chad Basin. It was completed in 2015 and presented to President Idriss Déby of Chad — the first time on the African continent that a joint environmental audit had been submitted to a head of state.

What did it change? The honest answer is: it is hard to know. The audit generated findings and recommendations; whether those recommendations translated into policy change remains difficult to trace. What it did demonstrate was that cooperation between national SAIs on a shared natural resource was feasible — that institutions which normally operate within strictly national mandates could align their methodologies, agree on findings, and speak with a common voice. That procedural achievement matters more than it might seem, in a region where cross-border cooperation on shared resources has often stalled for political reasons.

What remains hard: environmental performance auditing requires technical expertise that most African SAIs do not have in-house. Geology, hydrology, ecological monitoring — these are not auditing skills. The capacity gap is real, and filling it takes years, not programme cycles.

Audit reports, in most countries, are not read by citizens. They are technical documents, written in bureaucratic language, published months or years after the events they describe, and distributed to committees that may or may not act on them. The feedback loop between what citizens experience — the pothole, the absent nurse, the school with no books — and what auditors investigate is almost entirely broken.

The idea behind CitizensEye, developed in 2018 with Ghana's Audit Service, was simple: use a mobile app to let citizens report service delivery failures directly to the Auditor General, geolocated and categorised, in real time. The logic was borrowed from consumer platforms — the same instinct that makes someone leave a review on TripAdvisor — applied to public services. If citizens could flag issues as they encountered them, auditors would have a continuously updated picture of where the system was failing, grounded in lived experience rather than administrative self-reporting.

The Ghana launch attracted significant media attention. The harder question — whether the Audit Service had the internal capacity and political will to systematically act on citizen reports — was something we grappled with from the beginning, and did not fully resolve. A tool is only as useful as the institution it feeds into.

In Uganda, the model was adapted through the Citizen Feedback Platform, developed under the Strengthening Governance and Civil Society Programme, with civil society organisations — particularly the Civil Society Budget Advocacy Group — serving as the link between communities and the Office of the Auditor General. By 2024, 495 citizens across 20 local governments had submitted 726 reports. In at least one documented case, a primary school with 346 pupils flagged through the platform received construction materials within weeks.

What remains hard: digital tools work best for people who are already relatively connected — literate, smartphone-owning, comfortable with apps. The communities with the most acute service delivery failures are often the least reachable through digital channels. Bridging that gap requires civil society partners willing to do sustained, offline community mobilisation. The technology is the easy part.

Most governance programmes, if you look closely, reach a relatively narrow slice of the population: educated, urban, male, middle-aged. The people who attend stakeholder consultations, who engage with civil society organisations, who have the time and the confidence to show up — these are rarely the people most dependent on public services working well. The paradox of participatory governance is that the people whose participation matters most are usually the hardest to reach.

The Civil Society in Uganda Support Programme (CUSP), co-funded by the EU and Germany, tried to take this seriously. Working across all regions of Uganda, including the West Nile refugee-hosting areas, it supported youth parliaments, vocational programmes for young people with disabilities, sport-based civic education for girls in marginalised communities, and AI innovation workshops for civil society organisations learning to use new tools. The premise was that participation is not a single thing — it looks different for a 19-year-old woman in Karamoja than for an NGO officer in Kampala, and programmes need to be designed accordingly.

By 2024, seven sub-national youth parliaments had brought together 310 young people whose resolutions fed into national policy debates. The HEVOT vocational centre in West Nile opened to serve young people with disabilities who had previously had no pathway to skills training. An EU Youth Sounding Board gave twenty young Ugandans — including doctors, teachers, and female entrepreneurs — a formal advisory role with the EU Delegation.

What remains hard: inclusion is expensive and slow. It requires working through intermediaries — civil society organisations, community groups, local government structures — that have their own capacities, limitations, and interests. The further you go from the capital, the thinner the civil society ecosystem becomes. And the political environment matters enormously: in contexts where civic space is constrained, the organisations that can reach the most marginalised communities are often under the most pressure.

Tassilo von Droste zu Hülshoff grew up between Germany and France, educated at a bilingual school in Saint-Cloud before moving through Nancy, Sciences Po Paris, and eventually London, where he completed a master's degree in Environmental Law and Sustainable Development at SOAS. It was an education that kept circling the same question from different angles: what does it mean for a state to be accountable to the people it governs?

That question became a career. Over eighteen years working primarily with GIZ — Germany's international development agency — he has lived and worked in Mauritania, Brussels, South Africa, Chad and Uganda, always somewhere at the intersection of public institutions and the citizens they are supposed to serve. His early work was in public financial management: the unglamorous but foundational task of helping governments build the systems that track whether public money reaches its destination. From there, he moved into the world of supreme audit institutions — the independent bodies that hold governments to account for how they spend — advising AFROSAI, the African umbrella organisation for national auditors, and leading GIZ's Good Financial Governance programme across the continent from Pretoria.

A recurring frustration with that work was that audit reports, however rigorous, too often disappeared into bureaucratic silence. The findings were sound; the accountability loop was broken. That frustration eventually produced CitizensEye — a mobile application developed in 2018 with Ghana's Audit Service that allowed citizens to report service delivery failures directly to the Auditor General, geolocated and in real time. It was the first tool of its kind for a supreme audit institution in Africa, and the model that later became Uganda's Citizen Feedback Platform, now used across twenty local governments.

His most recent role, as Head of Programme for the Strengthening Governance and Civil Society Programme in Uganda, brought together the threads of the preceding years into a single mandate: a €44.7 million initiative working across all regions of the country on civil society capacity, oversight, and inclusive governance. The work ranged from youth parliaments in Karamoja to vocational training centres for persons with disabilities in West Nile to AI innovation workshops for civil society organisations in Kampala.

Alongside his work in development cooperation, he has written — on governance, on culture, on the relationship between institutions and the societies they shape. He co-authored La Révolution Wikipédia, an early study of the encyclopedia's transformation of knowledge and authority, and later wrote Die Entwicklungspolitik der Europäischen Union, an introduction to EU development policy. He has also worked as a freelance journalist, contributing to travel guides and publications in French, German and English.

He is based in Düsseldorf.

Since I started working in development cooperation, my obsession has been a simple question: what does good governance actually do for people? Not in the abstract — in the clinic that gets built, the contract that gets implemented, the budget that reaches the school instead of disappearing on the way. The question sounds obvious. The answer, as anyone who has spent time in this field knows, is rarely straightforward.

There is no shortage of tools for measuring governance. The Corruption Perceptions Index. The World Bank Governance Indicators. Freedom House. The Rule of Law Index. Each of them captures something real. But taken individually, each tells only part of the story — and the part they tell is often the part that confirms what the commissioner of the index already believed.

What has changed, and what prompted this essay, is that it is now possible to bring these datasets together and interrogate them in ways that weren’t practical before. To ask not just what one index shows, but what happens when you hold all of them against each other, and then hold the whole picture against what actually happens to people’s lives. Correlating governance scores with child mortality, life expectancy, electricity access, and income inequality across a hundred and fifty countries is not a definitive answer to anything. But it is a way of forming new hypotheses — and, more usefully, of questioning the assumptions you walked in with.

The surprises were immediate. How large was my surprise to find that Belarus — a country with one of the lowest political freedom scores in the world — has a Gini coefficient of 24.4, making it one of the most equal countries on earth, more equal than France, Germany, or the United Kingdom. Or that nine out of twelve countries in what I will call the authoritarian development cluster sit below their income-group peers on inequality. Or that Freedom House scores, the most direct measure of democratic rights, have almost zero correlation with income distribution across countries.

There are more surprises in the data. I want to share them — and to be honest about what they do and do not mean.

There is a fact embedded in the 2022 governance data that does not resolve comfortably into either side of one of development’s oldest debates. China, with a press freedom score of 16 out of 100 and a Freedom House rating of 9, has an HDI of 79 — higher than Brazil, Colombia, or South Africa. Vietnam, rated 19 on political rights, has a life expectancy of 73 and a UHC service coverage score matching Romania. Russia, despite a WGI governance score of 19, has an HDI of 82, placing it above much of Latin America. These countries do not merely survive their governance deficits; they appear, in aggregate human development terms, to flourish despite them.

This is not a marginal anomaly to be explained away. When you run a regression of composite governance scores against HDI across 77 countries, you get an r of 0.85 — a strong positive correlation that seems to confirm the conventional wisdom. But when you disaggregate governance into its two main components — state capacity on one side, accountability and rights on the other — and test each separately, a sharper and more troubling picture emerges.

The Capability Advantage

The governance indices in common use actually measure two quite different things. Some — the World Bank Governance Indicators, the Chandler Good Government Index — assess what a state can do: its effectiveness at delivering services, its regulatory coherence, the quality of its institutions as instruments of administration. Others — Freedom House, the Reporters Without Borders Press Freedom Index, the Transparency International Corruption Perceptions Index — assess how a state is constrained and observed: whether citizens can hold it to account, whether the press can investigate it, whether corruption can be named and punished.

Building a composite “capability score” (averaging WGI and Chandler) and a composite “accountability score” (averaging Freedom House, press freedom, and CPI) across the 56 countries for which both are available, the correlation with HDI breaks significantly in the capability direction: capability → HDI yields r = 0.916, accountability → HDI yields r = 0.831 — a gap of nearly 0.09 Pearson points that is substantively meaningful. For life expectancy the gap is larger still: capability r = 0.882, accountability r = 0.766.

This is not surprising once you think about the mechanisms. What actually produces low child mortality, high life expectancy, or broad UHC coverage? Schools built and staffed. Vaccination programmes designed and delivered. Roads connecting rural populations to clinics. Water infrastructure maintained. All of these require state capacity — the ability to plan, fund, contract and execute. A free press does not, in any direct sense, build a hospital.

The authoritarian development model, when it works, works precisely here. China, Vietnam, Singapore, Thailand, Morocco and Kazakhstan all sit at the top of the capability-versus-accountability gap table: their state effectiveness scores substantially exceed their political freedom scores, and the gap ranges from +7 points (Morocco) to +33 (Singapore). These are states that have invested in the machinery of service delivery while suppressing the mechanisms of political accountability. In the cross-section, the investment shows.

The Regional Pattern

The capability-accountability gap is not randomly distributed. Breaking it down by region reveals a structural geography of governance that maps closely onto the developmental state literature.

East Asia and Pacific shows the largest positive gap: +5.7 points on average, meaning the region’s states are more effective than they are accountable. MENA follows at +4.1. Western Europe sits near parity at −1.4 — the only region where capability and accountability are roughly matched, which is the distinguishing feature of consolidated liberal democracies. North America is similar.

Then the pattern inverts. Sub-Saharan Africa shows −9.2: African states on average have more political pluralism than administrative effectiveness. Latin America is the most extreme case at −14.4, with Brazil (−21), Bolivia (−20), Peru (−18), and South Africa (−15) all having significantly stronger accountability than capacity. These are countries with competitive elections, independent judiciaries, free presses — and chronically weak bureaucratic effectiveness.

This is a recognisable finding in the literature. Francis Fukuyama has argued that the sequence matters as much as the content: states that built administrative capacity before democratising tend to have stronger institutions than states that democratised before building capacity — much of sub-Saharan Africa and Latin America, where political competition arrived before the administrative infrastructure to support it. The data appears to bear this out, at least in cross-section.

The Inequality Puzzle

If the authoritarian development model delivers better-than-expected human development outcomes, one natural suspicion is that it does so unequally — purchasing aggregate HDI gains while concentrating income at the top. The Gini coefficient data complicates that story considerably.

Across 101 countries with both governance and Gini data, the correlation between governance and inequality is negative but weak: composite governance score → Gini yields r = −0.229. Better governance is associated with lower inequality, but the relationship is modest. More striking is what breaks down when you separate the governance dimensions. The Chandler Good Government Index — the most capability-oriented of the eight indices — shows the strongest association with lower inequality (r = −0.417). Freedom House, the most direct measure of democratic rights, shows essentially no relationship with inequality at all (r = −0.093). The BTI governance index, which explicitly tracks democratic transformation, yields r = −0.003 — effectively zero.

The implication is uncomfortable for democratic theory: in the cross-sectional data, political rights do not reduce income inequality. Administrative competence does — through the mechanisms of functional taxation, effective social transfer programmes, and public service delivery that reaches the poor. Democracy, on its own, is not a redistribution engine.

Against this backdrop, the authoritarian cluster produces a finding that many readers will find counterintuitive: most of these countries are more equal than their income-group peers, not less. Nine out of twelve countries in the cluster fall below their peer-group average Gini coefficient. Belarus (24.4) is one of the most equal countries in the dataset, more so than every Western European state except Slovakia and Slovenia. Azerbaijan (26.6) and Kazakhstan (29.2) are more equal than Germany (32.4) or the United Kingdom (32.4). China (35.7), Russia (35.1), Vietnam (36.1), and Iran (35.9) all sit comfortably below their upper-middle-income peer average of 39.1.

The mechanism, however, matters enormously. The low Gini scores in the former Soviet states reflect wage compression and state management of labour markets rather than democratic redistribution. These are economies in which wages are administratively constrained, informal sector activity is limited, and the range of income outcomes is narrow for structural rather than redistributive reasons. The same Gini coefficient that in Denmark reflects generous social insurance and progressive taxation in Belarus reflects the absence of the income variance that market economies generate. Equality measured this way and equality in lived experience are not the same thing.

The outlier cases are particularly revealing. South Africa (Gini 63.0) is the most unequal country in the dataset by a wide margin, and is also one of the clearest cases of accountability without capacity: strong constitutional protections, competitive elections, an independent judiciary and press — yet administrative effectiveness that has consistently failed to translate political commitments to redistribution into actual redistribution. For these countries, the post-apartheid inheritance of concentrated land ownership and racialised labour markets represents a form of structural inequality that conventional governance reforms are poorly equipped to address, regardless of their accountability credentials.

What Accountability Doesn’t Explain — in the Short Run

The short-run picture is uncomfortable for advocates of accountability-first development strategies. Between 2015 and 2022, the correlation between changes in Freedom House scores and changes in HDI was only r = 0.30. Countries that improved political rights gained an average of 3.3 HDI points over the period; countries that backslid gained 2.4. The gap exists but is modest — and substantially smaller than the equivalent relationship between capability improvements and HDI gains (r = 0.37).

More troubling still: every major democracy that experienced significant political backsliding in this period continued to improve or hold stable on HDI. India (Freedom House −11 points) gained 1 HDI point. Brazil (−8) was unchanged. Poland (−7) gained 2. Hungary (−6) gained 2. The United States (−7) gained 1. Democratic erosion, at least over a seven-year window, leaves no visible trace in the human development data.

This is the empirical core of the authoritarian competence argument, and it deserves to be taken seriously rather than explained away. If accountability were the primary driver of development outcomes, we would expect its deterioration to register — and it largely does not, at this timescale.

What Accountability Does — Over Time

The picture changes when you shift the question from level and trend to variance and fragility.

Among the 20 countries with the highest accountability scores (above 70 out of 100), the average HDI is 92.8, with a standard deviation of just 3.2. Among the 16 countries with the lowest accountability scores (below 40), the average HDI is 58.0 — and the standard deviation is 13.4, more than four times larger. This group includes China (79) and Vietnam (73) at the top, alongside Somalia (30), Chad (40), and Niger (39) at the bottom. The spread is enormous.

Accountability, in other words, appears to function less as an accelerator of development and more as a floor under it. High-accountability states do not necessarily develop faster than high-capacity authoritarian states in any given period. But they almost never fall to the bottom. Low-accountability states produce a wide and unpredictable distribution of outcomes, from Singapore and China to South Sudan and North Korea. This is consistent with recent empirical work by Imam and Temple (IMF Working Paper, 2025), who find — using a panel of 108 developing countries from 1970 to the present — that in democracies, growth spells are longer-lived, output collapses less likely to deepen, and periods of stagnation less likely to presage further decline.

This logic extends to inequality. Democratic redistribution systems — social insurance, progressive taxation, collective bargaining — are institutionally self-reinforcing: they create constituencies that defend them, and they survive leadership change. Compressed wages under administrative control do not. Russia’s Gini rose from roughly 27 in 1990 to over 40 during the transition decade — one of the largest and fastest inequality increases in recorded history, directly produced by the removal of the administrative compression that had held it down.

The Fragility the Data Cannot See

There is a deeper problem with reading the authoritarian overperformers as a governance success story: the data is a cross-section, and cross-sections cannot observe collapse.

Venezuela had strong state capacity in the oil era — and an HDI that rose steadily through the 1970s and 1980s. The collapse of the 1990s and 2000s was not gradual; it was sudden, and it was directly enabled by the absence of accountability mechanisms that could have forced course correction before the institutions were hollowed out. China’s zero-COVID reversal in late 2022 — a policy implemented with ruthless effectiveness for three years and abandoned overnight — is a smaller but structurally identical case: a state capable of extraordinary administrative performance in one direction, with no mechanism for self-correction when the direction was wrong.

The capability trap, as Andrews, Pritchett and Woolcock describe it, operates precisely here. States can build impressive service delivery systems on foundations of political control, and those systems will show up well in cross-sectional data. But the same concentration of power that enables rapid implementation also eliminates the feedback loops that sustain performance over time. The low Gini scores in Belarus, Azerbaijan and Kazakhstan are, in a sense, a governance artefact: they reflect the ongoing capacity of those states to compress economic outcomes. That capacity is not institutionally self-sustaining. When it shifts, the inequality will surface.

Two Authoritarian Models: What Fiscal Transparency Reveals

The Open Budget Survey, published by the International Budget Partnership, offers a way to test a more precise version of the accountability hypothesis. Rather than measuring political rights, it measures something narrower and more technical: whether governments publish their budget documents online, in a timely manner, and with sufficient detail to support public scrutiny. It covers 125 countries including China, Russia, Vietnam, Kazakhstan, and most of the authoritarian cluster. It is, in effect, a measure of fiscal transparency — the dimension of accountability most directly relevant to whether public money reaches public services.

The hypothesis is compelling: authoritarian states may be able to suppress political rights while still maintaining the fiscal discipline that service delivery requires. Without transparent budgets and functional audit systems, even a capable state cannot reliably direct spending toward intended outcomes. Fiscal transparency, on this account, is the minimum condition for the authoritarian development model to work.

The data offers partial but not clean support. Within the 33 semi-authoritarian states in the sample (defined as Freedom House score below 60 and press freedom below 50), the correlation between OBI budget transparency and HDI is r = 0.567 — a meaningful positive relationship. Countries with higher budget transparency within this group average an HDI of 68.6; those with low transparency average 51.7. Russia (OBI 66, HDI 82), Turkey (OBI 64, HDI 85), Kazakhstan (OBI 63, HDI 80) and Thailand (OBI 60, HDI 80) all sit in the high-transparency, high-development quadrant. The hypothesis appears to hold for them.

Then China disrupts the pattern. With an OBI of 20 — near the bottom of the cluster — and an oversight score of just 30 out of 100, it achieves an HDI of 79. Iran (OBI 0) achieves 78. Belarus (OBI 0) achieves 80. Algeria (OBI 15), Tunisia (OBI 16): both achieve HDI above 73. These are states with almost no fiscal transparency by international standards — and reasonable to strong human development outcomes.

This points to two structurally distinct authoritarian development models, not one. The first — call it the transparency-managed model — is exemplified by Russia, Turkey, Kazakhstan and Vietnam: states that suppress political competition but publish budget documents and maintain formal audit institutions. Their fiscal accountability is partial but real. The second — the opacity-managed model — is exemplified by China, Belarus, Iran and Algeria: states that achieve service delivery through direct party or state control of resource allocation, without the transparency layer. The budget process is opaque, but the political apparatus substitutes for the market and civil society feedback mechanisms that transparency normally provides. Spending reaches hospitals and schools not because the budget is publicly scrutinised, but because the party has both the incentive and the administrative capacity to direct it there.

The mediation analysis supports this framing. When OBI scores are controlled for in the governance → HDI regression, the partial correlation drops from r = 0.781 to r = 0.618 — a fall of 0.16, indicating that fiscal transparency explains part, but only part, of why governance predicts outcomes. The residual relationship — governance predicting HDI even after transparency is controlled for — reflects the opacity-managed model: the states where good outcomes arrive without the transparency that normally enables them.

The durability question follows directly. The transparency-managed authoritarians have built fiscal systems with some embedded correction mechanisms: audit reports exist, budget execution data is published, legislative oversight bodies are functional even if toothless. These structures create at least a weak feedback loop between spending and outcomes. The opacity-managed states have no equivalent. Their development gains depend entirely on the continued alignment between the political apparatus’s incentives and the population’s welfare — an alignment that is historically contingent and not institutionally guaranteed. The data cannot observe this risk in a cross-section. But the pattern of gains it would predict — strong performance during periods of political stability and resource abundance, rapid deterioration during shocks — is precisely what Venezuela illustrated between 1980 and 2010.

A Provisional Synthesis

The data does not support the conclusion that accountability is irrelevant to development. It supports a more precise and more uncomfortable claim: that state capacity is the more proximate determinant of short-run human development outcomes, while accountability determines the stability and predictability of those outcomes over time.

Capability delivers. Accountability sustains.

The inequality findings add a third dimension. Administrative capacity — the same dimension that predicts HDI gains in the short run — also appears to be the mechanism through which states achieve more equal income distribution, not democratic accountability. But the equality produced by administrative compression is qualitatively different from the equality produced by democratic redistribution: the first is fragile and dependent on continued political conditions; the second is institutionally self-reinforcing.

Singapore, which has substituted administrative management for democratic redistribution over the long run, shows a Gini of 47.3 — the highest of any high-income state in the dataset. South Korea and Taiwan, often cited alongside Singapore as evidence that autocratic development works, are today full democracies; their development trajectories cannot be cleanly separated from that transition.

For development practitioners, the sequencing question — build capacity first, then accountability, or pursue both simultaneously — is not fully resolved by the cross-sectional data. What the data does suggest is that accountability substantially narrows the range of outcomes that states can fall into. And that the kind of equality produced by accountable states is more durable than the kind produced by administrative compression, however impressive the latter looks at any given moment.

A press freedom score of 16 correlates with an HDI of 79 and a Gini of 36 today. An Open Budget Index of 20 — near the bottom of the global ranking — coexists with functioning schools and hospitals. Whether those numbers still look like a success story in 2035 is a question the cross-section cannot answer — but the theory of accountability, and the history of states that tried to sustain development without it, suggests they should give us pause.

Data sources: This essay draws on the Governance Compass, which aggregates eight major governance indices across 160–180 countries from 2015 to 2025, and a companion correlations tool comparing governance scores against World Bank WDI and UNDP development outcome series. Gini data is from the World Bank Poverty and Inequality Platform (most recent year available per country). All analysis reflects data as of 2022–2025.

A note on process: This analysis was developed with the assistance of AI tools for data processing and writing. The data compilation, the questions driving the analysis, and the judgements about what the findings mean are the author’s own; the computation and much of the drafting were AI-assisted.