Steering by an old map: why counting people is not a luxury

No sailor would cross an ocean with a chart drawn twenty years ago. Coastlines move, sandbanks shift and new harbours appear where there were none. Yet that is how many governments are steering today: by maps of their own people that are out of date, incomplete or missing altogether.

The census is that map. Every school built, every clinic staffed, every vaccine ordered and every parliamentary seat drawn depends on it.

And around the world, the map is fading.

A map with holes in it

Start with Africa. In a presentation to the UN Statistical Commission in March 2025, the UN Economic Commission for Africa reported that 41 of the continent’s 54 countries carried out a census in the 2020 round. It cited limited funding, rising costs, COVID-19 and insecurity as obstacles. Its estimate of the result is stark: one in three Africans remain uncounted in the 2020 round.

This is not only an African problem. A 2025 policy forum in Science by Jessica Espey, Andrew Tatem and Dana Thomson found, as the University of Southampton summarised it, that 24 of the 204 countries that ran a census between 2015 and 2024 had not published the results. Those 24 countries hold about a quarter of the world’s population.

Even wealthy, well-run statistical offices are missing people. The US Census Bureau’s own post-enumeration survey, published in 2022, found that the 2020 census undercounted the Hispanic or Latino population by 4.99%, against 1.54% in 2010.

South Africa’s 2022 census missed about 31% of people before adjustment, according to David Everatt, then outgoing chair of the country’s National Statistics Council. In 2024, two University of Cape Town demographers argued that the adjustment did not go far enough and called it the largest undercount the UN Population Division has recorded. Statistics South Africa rejected their concerns as unfounded. The disagreement itself tells you something: when a count goes badly, nobody can be sure how badly.

Twenty years without a count

Then there are the maps that were never redrawn at all.

Nigeria, Africa’s most populous country, last held a census in 2006. A census planned for 2023 was postponed and, as Vanguard reported in June 2026, still has no date. The commission’s former chairman said nearly ₦200 billion had already gone on preparations, including more than 760,000 digital devices, and warned that unused equipment risks becoming obsolete. President Bola Tinubu has said that “the population remains the greatest asset of the nation in the development process.”

I agree with him. But we do not leave our greatest assets uncounted for two decades.

To be fair, the reasons for delay are real. Vanguard lists funding, security, logistics and the political risk of a count close to the 2027 elections. I have worked in places where an enumerator cannot safely reach a village. Nobody should pretend a census is easy.

But difficulty is not a reason to steer blind. It is a reason to plan earlier and protect the work.

The charts we use between counts

A census is a snapshot every ten years. Between counts, countries rely on two other instruments, and both are weaker than most people realise.

The first is civil registration: the routine recording of births and deaths. According to a World Health Organization fact sheet published in 2024, nearly 40% of the world’s deaths go unregistered. In WHO’s African Region, only 44% of births and 10% of deaths are registered. When nine in ten deaths leave no official record, we are guessing at what kills people. As our editorial team noted in its piece on WHO’s latest global health estimates, the data is weakest exactly where the burden is shifting fastest.

The second is household surveys. The Demographic and Health Surveys, which covered 90 countries over four decades, lost their US government funding in February 2025. Philanthropy kept the core alive: the Population Reference Bureau reported in July 2026 that a $39 million Gates Foundation grant and $4 million from Bloomberg Philanthropies let ongoing surveys continue, but the programme is narrower than before and the Gates grant ends in 2028.

So the map is fading, and the sketches we draw between editions are smudged.

The case for skipping the count

Here is the strongest argument against my position, and it deserves a fair hearing.

Censuses are expensive, slow and politically risky. Meanwhile, satellite imagery, mobile phone records and statistical models can now estimate where people live with impressive detail. Why spend a fortune knocking on every door when a model can fill the gaps?

These tools are valuable. Everatt himself argues that big data could help future South African censuses. I want ministries using them.

But a model is only as good as the ground truth it is calibrated against. Strip away the count and you have a guess with good graphics. It is telling that one of the authors warning about this crisis, Andrew Tatem, leads WorldPop, a group that builds exactly these models. His verdict is plain: “accurate population data is not a luxury.”

Cutting the count to save money is penny wise and pound foolish. The saving is visible on one budget line. The cost is spread invisibly across every allocation that follows: the clinic built for a village that has emptied, the classroom missing in a town that has doubled, the vaccine order sized for children who were never counted.

Redrawing the map

What would taking this seriously look like? Four things.

1. Fund the census as infrastructure, not an event. A census needs a ten-year budget line protected from the annual scramble, the way a ministry protects a road programme. Equipment bought for a count that never happens is money lost twice.
2. Publish the results and the uncertainty. A count with an honest error margin is more useful than a precise-looking number nobody trusts. Release the post-enumeration findings alongside the headline figures.
3. Fix the continuous map. Civil registration is the cheapest way to keep the picture current between censuses. Link it to health facilities, so that every birth and death a clinic sees is recorded once and used many times, a principle I set out in Reusable public-health data starts before the analysis.
4. Train the people who read the map. Counts, surveys and models only help if analysts can combine them and leaders can question them. That is why professional education matters as much as the data itself, and why I keep returning to the questions leaders should ask of their evidence.

A map does not need to be perfect to be useful. It needs to be recent, honest about what it leaves out, and owned by the people who steer by it.

Every person left off the map is a person left out of the plan.

At Praxis, our research and data work helps institutions build the population and health evidence they can steer by. If your organisation is wrestling with that problem, let us talk.

About the author

Simon Aseno

President and CEO

Simon Aseno is President and CEO of Praxis Global Institute, where he leads work connecting research, technology, education and advisory to strengthen people, systems and institutions.