Protected Into Invisibility: Data Poverty and Fragility

*We promised to leave no one behind. But you cannot reach a person your systems cannot see, and decades of missing data, some of it the unintended cost of our own caution, have quietly turned a promise of inclusion into a machinery of exclusion. This is part one: the trap, how it compounds, and the uncomfortable role our own protective instincts play in it.*

In Afghanistan, my team built a drought severity map to support the response to a deepening agricultural crisis. It did its job well. District by district, it showed where conditions had tipped toward famine, where harvests had failed, where the need for agricultural, nutrition, food, WASH, and health support was most acute. We could point to the worst-hit areas with real confidence. What we could not do was name a single person inside them. Behind the shaded polygons were families living through famine-like conditions, the people who were supposed to receive that agricultural, nutritional, food, water, and healthcare assistance, and there was no individual beneficiary record of who they were, where exactly they lived, or what they had already lost. The map could tell us, with authority, where the catastrophe was. It could not tell us who, by name, was caught in it. And in a system that allocates assistance against evidence, a person no record can resolve sits dangerously close to a person who does not count.

That gap has stayed with me because it exposes a paradox at the centre of modern humanitarian and development work. The global community has made "leave no one behind" the moral spine of the 2030 Agenda. And yet leaving no one behind begins with a precondition we are still failing to meet. Everyone has to be counted. The people most at risk of being left behind are, with bitter regularity, exactly the people our data systems never saw in the first place. We built a promise of inclusion on top of an evidence base structured around exclusion, and then expressed surprise when the same people kept getting missed.

This is not a marginal problem affecting a few edge cases. It is foundational, and the numbers are staggering once you go looking for them.

Around 800 million people still lack any official proof of legal identity, down from over a billion a decade ago but still a population larger than most continents. More than 110 low and middle-income countries lack functional civil registration and vital statistics systems, the unglamorous infrastructure that records births and deaths. The poorest fifth of the global population accounts for more than half of all unregistered births. A child never registered at birth begins life statistically invisible, and that invisibility compounds across a lifetime: no documented identity, no claim on services, no entry in the datasets that decide where schools, clinics, and emergency relief are sent.

The result is that national averages, the figures that dominate policy, routinely mask the people furthest behind. Aggregate progress on poverty or school enrolment can conceal stagnation or decline for the hardest-to-reach groups, because those groups are precisely the ones underrepresented in the data that produces the average. Migrants are largely absent from official global statistics. Stateless populations are, almost by definition, erased from the records of the states that deny them. The people we most need to see are the people our instruments are worst at seeing.

What makes this so hard to escape is that it is not a single gap. It is a feedback loop that deepens with every turn, and I have watched it turn.

It starts with absence. A community is not registered, not surveyed, not connected to the systems that generate data, whether because it is remote, poor, displaced, marginalised, or some compounding combination of all four. Because it is absent from the data, it is absent from planning. Services are not sited there, programmes are not designed for it, funding formulas pass it over, because the evidence that would justify investment does not exist. Because investment passes it over, the community grows more marginalised, more remote from the formal systems, and therefore even harder to count next time. Each cycle, the gap widens. The under-counted become the under-served become the more-invisible. Decades of this produce communities that have been failed so consistently they have effectively dropped out of the official picture of their own countries.

Fragility accelerates every stage of this loop. In conflict-affected and fragile contexts, the registration systems are broken or were never built, collection is dangerous and partial, populations move and scatter, and whole groups may be deliberately excluded from official statistics by authorities with an interest in their invisibility. I have written before about how the hardest places to operate end up with the thinnest, least trusted data, and this is the development-scale version of that same injustice. The places with the deepest, most entrenched need are the places where the evidence of that need is weakest, and the global architecture, which allocates attention and resources against evidence, reads weak evidence as weak need. It is the same dynamic I traced in disaster finance in Invisible Disasters, Invisible Funding: missing data is treated as missing problem. In fragile contexts, it is usually the opposite.

Here is the uncomfortable turn, and the reason I wanted to write this rather than another tidy lament about capacity gaps. Some of this invisibility is the unintended consequence of our own caution.

The humanitarian and development sectors have spent the last decade, rightly, building a serious ethic of data responsibility: do no harm, minimise what you collect, protect what you hold, never let a dataset become a weapon against the people it describes. I believe in all of it. I have made the protective call myself, choosing to collect less in dangerous contexts precisely because data you do not hold cannot be stolen or coerced from you. That instinct has prevented real harm, and I would defend it again tomorrow.

But every protective choice has a shadow, and we have been slow to look at it honestly. Data minimised is also data missing. Granularity withheld to protect a vulnerable group is also granularity unavailable to advocate for that group, to fund services for it, to prove it was left behind. There is a growing body of work on what scholars have called the surveillance gap, the harms of extreme privacy and data marginalisation, and it names something practitioners feel but rarely say. That being unseen is not safety. For the already-marginalised, absence from the data is frequently just another form of exclusion, this one with our fingerprints on it. It is the same pattern I have described in voice infrastructure inequality, where the populations our technologies serve worst are the ones already furthest from the table. We worried, correctly, about the harm of being counted in the wrong way. We attended far less to the harm of not being counted at all.

This is sharpened by the fact that much of our data-protection apparatus was designed in and for the global North, around individual privacy in high-capacity, rights-protecting states. Applied bluntly to populations whose problem is not over-surveillance but non-existence in the record, those frameworks can push the most vulnerable further into the dark. A privacy regime that makes it harder to count an undocumented migrant, a displaced minority, or an informal settlement does not always protect those people. Sometimes it simply guarantees they will remain unfunded and unreached, protected into invisibility. The road to that outcome is genuinely paved with good intentions, which is exactly what makes it so hard to challenge.

So we are left with a hard picture. The people we have pledged hardest to reach are the people our systems are structured to miss; fragility deepens the gap at every turn; and some of the missing data is the unintended price of protective choices we made for genuinely good reasons and never went back to weigh. Naming that is not an argument for abandoning data responsibility. It is an argument for finishing it, for completing a half-built ethic that learned how to protect people from the harm of being counted badly but never learned to protect them from the harm of not being counted at all.

That is the problem. I do not think the answer is to swing back toward collecting everything about everyone, and I am wary of anyone who frames this as a simple choice between visibility and protection. There is a better path, one that treats representation as a right while keeping people safe, and it has a name. I propose data equity as the way to address these unintended consequences of data protection, and that is the subject of part two. What data equity actually means, why it is not the opposite of data protection, and what pushing for it looks like in practice.

What gets counted gets funded. What gets missed stays vulnerable. In part one I have tried to show how the missing got missed. In part two, I want to argue we can do something about it.