On paper, Arcelia looks like a poor-but-average Mexican town. It is located in Guerrero, Mexico’s second poorest state.
Official data gives it a score of 0.714 – firmly in the “high development” band on the UN’s Human Development Index (HDI).
Just then a satellite spots Arcelia. Using artificial intelligence to analyze what it sees, it gives a low score of 0.617.
According to the UN’s own classification, it is no longer high but medium development – a different development level and a different policy-reality for 33,000 people.
Arcelia is not a special case. More than half (58%) of the global population is at the wrong development level, because the average of official data is too broad to see them. This is the central finding A study published in the journal Nature Communications By researchers from Stanford University in the US and the United Nations Development Programme.
“Nearly half of the world’s poorest countries have not conducted a census in the last 10 years,” said study co-author Hannah Druckenmiller. He highlighted the need for updated and accurate information to ensure that public policy matches people’s day-to-day needs.
An accurate HDI score matters for aid delivery
human Development Index This is not just a ranking. “This could determine the allocation of global resources,” say the study authors. The size of the areas that are prioritized for assistance.
Getting it wrong at the local level means resources can miss those who need them most. The problem is that the HDI only provides a score for entire countries. It was not originally conceived to differentiate between provinces or municipalities within a country.
But in a simulated aid program for Mexico, which targeted the poorest 10% of the country’s population, researchers found that adding municipality-level data increased their understanding of people’s development status – levels of poverty and wealth, education and health – by more than 11 percentage points.
HDI fixed one blind spot and created another
For decades, measuring development meant measuring gross domestic product (GDP) – the total economic output of a country. The problem is that GDP may increase, which will benefit only some people, while others will remain uneducated, sick, or poor.
In 1990, the United Nations introduced the Human Development Index to fix this.
“[HDI] Looks at average gross national income per capita, average years of schooling or expected years of schooling, and average life expectancy for each country, and adds them together to create an indicator of well-being that runs from 0 to 1,” said Sabina Alkire, director of the Poverty and Human Development Initiative at the University of Oxford. Alkire was not part of the new study.
The HDI is based on data from the UN’s own agencies, the World Bank, and national household and census surveys. It has become the world’s most widely used alternative to GDP.
But researchers in the 2026 study realized that the HDI is still a less than accurate measurement – the national averages presented in the HDI reveal little about what is happening inside a country at the local level.
For example, take a look at the map below.
This map presents one color per country. More developed nations are blue. Large parts of Africa and South Asia are orange and red. It’s a useful view of global human evolution – but a very basic one.
Take Mexico: Nationally, 130 million people are represented by a single color – blue. One development score for all.
But the level of development may vary from province to province.
In 2019, a team led by Jeroen Smits and Iñaki Parmanayer took the HDI to the province-level, detailing 1,739 provinces in 159 countries. they called it Subnational Human Development Index.
This changed Mexico’s approach: it now had 32 points instead of just one.
The north and center of Mexico still appear blue. Meanwhile, the south – especially Guerrero, Oaxaca, Chiapas – had turned light blue.
It was more detailed than the original HDI, but each block still showed only a single average. And we now know that even within a province, development scores can vary.
There are 81 municipalities in Guerrero alone – all given the same number in the SHDI. The 2026 study set out to go one step further and uncover the truth about life in each of those municipalities using satellite data.
What can satellites see and what can’t they?
The Stanford team fed satellite images with known province-level HDI scores into a machine learning model and let the algorithm find patterns.
Statistical correlations emerged about road density, building patterns and night-time lighting – man-made signatures of income and education. Capturing less visible health consequences from space proved difficult.
The map below shows what the model predicted – an HDI score for each of Mexico’s 2,500 municipalities, on the same scale as the previous two maps. Only the difference from Map 2 matters.
What appeared to be a uniform light blue in Guerrero broke down into multiple shades of blue and fragments of local realities.
Take Arcelia again. With data at the province level, Arcelia had a score of 0.714 – “high development.” However, with satellite data, it gets a score of 0.617 – “medium development.” That’s a big difference to potentially 33,000 people.
One step forward – but not the whole picture
As director of the Oxford Poverty and Human Development Initiative, Alkire has spent two decades developing poverty measurement tools used by governments around the world.
Alkire called the new Stanford study a step further: “We are in a time of innovation as a community working on measurement,” Alkire told DW. “These types of studies are fantastic because they are leading the way on a wide scale.”
But both Alkire and the study’s authors note that satellites also see only part of the story. They do not provide good data on health development.
“A malnourished child is not visible by night light,” Alkire said. The authors themselves stated that their estimates explain only 29% of the within-province HDI variation in Mexico.
Therefore, it is unlikely that satellites alone will provide a complete picture of human evolution. “Based on satellite alone, I don’t think so,” Alkire said.
But satellite data has been shown to be a valuable addition, especially where surveys are too expensive or too slow – satellites complement ground-level data but cannot replace it.
Edited by: Zulfikar Abbani