Why coordination is a modest ambition, and why public investment compounds between sectors
Working in silos has been diagnosed many times, and always as something to cure. Evaluations find that many international development organizations, including United Nations agencies, work in silos. Organizational reviews find the same. Governments say it more politely, and the agencies write it into their own strategies. The finding is now old enough to be quoted without being tested.
I am not convinced the silo is the problem. Institutions are built around expertise, mandates and accountability, and for good reasons. Education needs people who know how a school system works: how teachers are hired, how a curriculum reaches a classroom, how a budget line becomes a textbook. Health needs people who know how a service reaches a patient. Some of the best results in development come from narrow work. Vaccination and cash transfers succeed because they are simple, fast and easy to hold someone responsible for. Whatever the cure is, it is not less specialization.
The trouble is in the space between the silos. A child does not live education, health, nutrition, protection and family income as five policy areas. A young person does not live skills, jobs, enterprise and migration as four strategies. Institutions stay separate because they must. Outcomes stay connected anyway. Between the silos, value is lost every year – and no report records it.
People who talk about joined-up work usually mean one of three things, and they mix them up. Coordination asks who is doing what, so that nobody gets in anybody’s way. Coherence asks whether policies pull in the same direction. Compounding asks whether one policy makes another worth more. The Organisation for Economic Co-operation and Development (OECD) has written about the first two for years, and governments have built machinery for both. The third is harder than it sounds. It asks what one investment returns if another it depends on is missing. That question changes what public money returns, and almost nobody is organized to ask it.
Beware: this is not a call to merge anything. An earlier generation of integrated planning merged budgets, staff and accountability into units nobody could hold responsible, and it failed for that reason. Delivery stays where it is. What has to be joined is how the problem is read and how the money is decided.
Investing where sectors meet pays, and the return is large enough to matter. In North Carolina, a programme for poor children from birth to age five combined health care, nutrition, support for parents, childcare and early learning, and its children were followed into adulthood. James Heckman and his colleagues put its return at about 13 percent a year. No single part produced that. It came from all the parts working on the same child, each making the others worth more.
Look where that return landed: better health, more years of school, less crime, mothers back at work. In most governments those gains would show up in several ministries. Had any one of them paid for the programme alone, it would have carried the whole cost and seen a fraction of the credit, decades later, in someone else’s statistics. No minister defends a budget line like that.
Failures land in the same place. A clinic treats a child for malnutrition and sends her home to a family with no income to feed her. A doctor identifies a child’s disability, and no school nearby can teach her. Every ministry has a mandate, the list of responsibilities it answers for in law and in the budget. These failures fall between two mandates, so they belong to neither and appear in neither ministry’s report. That suits everybody.
Call the place where two sectors meet an intersection. Nobody owns one. Budgets are voted by sector, and careers are built inside one. So the space stays empty, and not by accident. The silo owns the expenditure. The intersection decides much of its return. That leaves the space open to an outsider, usually an international development organization. It offers two things: a horizon longer than one budget and one government, and a view across mandates that domestic institutions have good reasons to keep separate. That is most of what such an organization brings to a country, and most of the reason anyone still listens to it.
The way in is to start somewhere other than the silos. Start with the person the policies are for, and move forward in time. What will she need to be able to do? What will decide whether she ever gets to use it? Which decisions taken this year make that more likely, or less?
Two established ideas ask exactly these questions. Human capital development asks what people will need to be capable of in the economy and the society that is coming. Intergenerational fairness asks what today’s decisions do to the chances of the people who come after.
The first reads across sectors. It counts what has to be built into a person before anyone earns anything back: health, schooling, skills. The second reads across time. It asks who pays for that, and whether the return reaches them. Both end at the same decision, and the decision is almost always a budget line.
Neither is waiting for an owner. Human capital is a World Bank platform: by April 2026, 97 economies had joined its Human Capital Project, and their finance ministers meet once a year to discuss it. Intergenerational fairness became European Union (EU) policy in March 2026, when the European Commission adopted its first strategy on it. The strategy promises an index and aims to carry the idea into the EU budget and the European Semester, the yearly review of national economic policies. A country negotiating to join the Union will meet the second idea whether it looks for it or not. Its finance ministry already sits inside the first.
The two fit because each is blind where the other sees. A human capital index can rise while the people it measures leave. A fairness account can look balanced while nothing is being built. Read together, they describe one decision from both ends.
Every family makes this decision. A household works out what to spend on a child, who pays, and when it comes back. A national budget is the same decision, made once a year and at scale. The family does the sum in one head. The state splits it across nine ministries, two international agendas and a dozen reports, and nobody adds it back up. Adding it back up takes two questions. Does this spending make other spending worth more? Does the return come back to those who paid for it? No government document answers either one today. Both answers have to be written in the terms a finance ministry reads: how much, paid by whom, spent on whom, and when and where the return arrives.
Call it the compounding test. The first question is the human capital question asked properly. The second is the fairness question asked plainly. Money that passes both does something neither index measures on its own, which is usually why it never gets voted.
The economics of the first question is settled. A child with good food and health, help before school, a school that builds on it, skills that stay useful and an economy that can use them has not received six policies. Six policies have accumulated in one person, and each raises the value of the next. Flavio Cunha and James Heckman called this dynamic complementarity: later investment pays more when there is more to build on. Skill begets skill. Fixing a problem at sixteen costs more and returns less than preventing it at three.
Here is what that looks like in two ministries’ budgets. A country trains young people for jobs its economy does not yet create. The training is good. The certificates are real. Their value depends on something the education ministry does not control, which is whether firms exist that want to hire. Where they do, every euro of training returns more than it cost, because the skill is used, taxed and passed on. Where they do not, the same euro buys an exit visa. The education ministry paid for the skill. The economy ministry owns the condition that decides what it is worth. Neither did anything wrong. The return on one ministry’s spending depends on the other’s, and that is what makes it an intersection. Between them most of the value is lost, and neither budget shows it.
That one case fails both questions. Training and jobs each make the other worth more, and only one was paid for. A country that pays to build a skill and then watches it work abroad has failed the fairness question too, whatever its training statistics say. The World Bank met the same problem inside its own index. A person can be healthy, well taught and unable to find anything worth doing, so the Bank built a version that counts only the human capital the labour market actually uses.
This is why the future belongs at the start of a strategy, not at the end. Most strategies begin with today’s system and its gaps. Begin instead with the country a child starting school this year will find at twenty-five. How many people, of what ages, doing what work? Foresight sketches that country. Human capital turns the sketch into a list of capabilities. Intergenerational fairness asks which decisions made now make those capabilities easier or harder to gain, and easier or harder to use at home.
National visions are usually written as growth, competitiveness, prosperity, cohesion and European integration. Those are results. They say nothing about what has to be built first. The useful question is what people must be able to do before that country is possible. Strategy then stops being a list of what each ministry will do. It becomes a judgement about which policies build those capabilities together, where one raises the return on another, and where one is quietly cancelling another. And all of it has to reach the budget, because governments deliver visions through money, not through strategies.
Finding an intersection needs no new model. Start with one capability, not one sector. Write down what must be true for it to form, then what must be true for it to be used. Mark what is missing today, and put a ministry’s name against each gap. The intersections are wherever two names sit on the same chain. One test separates a real intersection from two good things standing side by side. If one is funded and the other is not, does the first lose most of its value? If not, it was never an intersection.
The rest is a little work at the right moment. Once a year, read the country through the two questions, using the two indices and the government’s own data, and put the reading on one page rather than in a report. Do it before the budget ceilings are set, because afterwards it is commentary. Choose three or four intersections, not thirty, and put a cost on each, because a number can be argued for and an idea cannot. Take the page to the ministry that holds the money, not only to the one that holds the sector.
Every country already has the machinery. It has a budget structure that decides which costs get a line and an owner, and a cost with no owner is the first one cut. It has something that turns a commitment into a payment. And it has an outside review it cannot ignore, which sets the calendar. In Albania, at least until the end of 2027, these are the medium-term budget programmes; the Reform Agenda under the EU Growth Plan, which pays only for verified reforms and had paid Albania €212.8 million by May 2026; and the EU’s commitment to carry intergenerational fairness into its budget and economic reviews. Other countries have other names for the same three jobs. Nothing new has to be built. The machinery exists. What has to change is the habit – and habits cannot be ordered.
The usual objection is measurement. Compounding takes fifteen years to show, and budgets want proof in eighteen months. The answer is not a third index. The two that exist can be read against each other, for the same country, on the same page, against the same budget. Nor does anyone have to prove causality fifteen years in advance. What can be checked every year is whether the connection was planned and whether the next system was ready to build on it. That is short-term work with a long-term logic.
None of this can be sold by an organization that does not practise it. An agency that asks ministries to work across mandates has to show it can do so itself. UNICEF is the clearest test, because it stands at the start of the life all these chains run through. It works on nutrition, health, early childhood, education, protection and social protection. To a child they are not programmes. They are one life. Inside the organization they are separate programmes with separate results frameworks.
Take early childhood and the first years of school. A programme that builds children’s language before school creates what primary school is meant to build reading on. Whether that happens depends on the school knowing what the child brings, on teachers ready to use it, and on children who need extra help being found and followed. If none of that is planned, the language gain is real, the programme reports success, and the value fades in a classroom that never knew it was there.
Most country offices know this connection. Few write it down. It lives in the head of a good programme officer, and it leaves with her when she moves to her next posting. An organization’s knowledge should sit behind the programme, not in the head of whoever runs it.
That means writing down four things for each connection that matters. What the first stage builds. Where that matters next. What the next system must do to build on it. How anyone will know it was carried forward. The important links are well known: early language and later reading, nutrition and attention in class, early screening and timely support, cash transfers and children staying in school. None needs to be rediscovered in each country. Each needs a named institution responsible at the receiving end.
Reporting has to change too. Offices report what they supported and what was delivered: children reached, services provided, policies drafted. Those numbers stay. What is missing is what happened next. Did the next institution see what the child brought, and build on it? Nobody has to prove that one programme caused a result ten years later. The office runs the same yearly check as the budget reading and adds the one step only it can observe: whether the value was actually carried forward. That is where the value between the silos becomes visible, and almost no office writes it.
The instruments already exist. The four links belong in the country programme’s theory of change. The carried-forward check needs one indicator per critical connection, measured at the receiving institution and reviewed each year. The results framework only has to add that one question.
So the programme officer’s Monday question changes. She still asks what her programme contributes. She also asks what capability she is helping to build, what future it will have to work in, and what must connect to it before it becomes an opportunity. A specialist does not have to become a generalist to ask this. She only has to stop letting her specialty decide what she may think about.
The organizations face a change of their own. Knowledge is no longer scarce. Ministries, firms and researchers find evidence, data and foreign examples in an afternoon. What is scarce is knowing what matters, what connects to what, what is changing, and what to do about it. An organization built for yesterday’s problems can stay fully competent and still become less relevant. Competence is not a defence against change.
No strategy changes that. What changes it is what an organization pays for. People are rewarded for delivering their own part. Almost nobody is rewarded for making someone else’s work more valuable. That is exactly what this requires. Partnerships are judged by whether two organizations work on related subjects, when the real test is whether one makes the other more effective.
Working between sectors needs people who can read more than one, who know how a budget is really made, and who can turn evidence into something a minister will sign. That person works next to the specialist, not instead of her. The closest role today is social policy and public finance, which reads budgets but reports, like every section, on its own results. The role needs a mandate across sections, close to the deputy representative, and a result of its own: how much one programme’s value was carried into another. Until then it is nobody’s job description.
Timing matters as much as people. I believe that it is often better to be first than to be better. An organization that waits until a subject is an official priority walks into a crowded room. Intergenerational fairness is at that stage now, and in two years it will have an index and a queue of people explaining it. It is also better to define and own a category than to compete inside of someone else’s category. An agency that says it does education well is measured against everyone who does education. An agency that helps a country build its people’s capabilities, across a life and across generations, has no obvious competitor.
An idea also exists only where it is remembered. A framework a minister cannot repeat after the meeting does not exist outside the room. The two questions were built to pass that test. Does this spending make other spending worth more? Does the return come back to those who paid for it? Plain words, no diagram, short enough to survive a corridor.
In The Relevance Test I argued that importance is what an institution says about itself, relevance is what others decide about it, and relevance has become the scarce thing. This piece is about where it can be found. It will not be found in more sectors, more programmes or more coordination meetings. Everyone already offers those. It is found between the silos, where returns compound and almost nobody is looking.
An institution that can see that space, write down what must connect in it, put a cost on it and carry it to every table where the decision is shaped will not need to argue for its own importance. The numbers and those who decide will make that argument for it.
Money is spent inside the silos. What it is worth is decided between them. The silos will stay. There is little to do about that, and I believe they should. What must improve is the ability to see what happens between them, and to act while it still matters.
This argument began with the person the policies are for, and it ends there. A budget is signed in one afternoon, in one room, by people who each brought one column. The individual it lands on will never know which ministry paid for which part of that life. The individual will only know whether all that spending built abilities that became opportunities for a better life, at home.
Sources
OECD, Driving Policy Coherence for Sustainable Development, 2023; OECD and IIASA, Systemic Thinking for Policy Making, 2020.
World Bank, The Human Capital Project, fact sheet, April 2026; and The Human Capital Index 2020 Update, which introduced the utilization-adjusted index.
European Commission, Strategy on Intergenerational Fairness, COM(2026) 110 final, adopted 5 March 2026.
Flavio Cunha and James Heckman, The Technology of Skill Formation, American Economic Review 97(2), 2007.
Jorge Luis García, James Heckman, Duncan Ermini Leaf and María José Prados, Quantifying the Life-Cycle Benefits of an Influential Early-Childhood Program, Journal of Political Economy 128(7), 2020.
European Union, Regulation (EU) 2024/1449 establishing the Reform and Growth Facility for the Western Balkans; European Commission, Implementing Decision of 20 May 2026 approving the third release of funds to Albania, Montenegro and North Macedonia.
Dritan Shano, The Relevance Test, 31 July 2026.
