The Geography of Global Vulnerability

INFORM builds risk from three dimensions, each scored from 0 to 10. Hazard and exposure combines natural hazards (earthquakes, floods, tsunamis, cyclones, drought and epidemics), weighted by how many people live in their path, with human hazard, which captures current conflict intensity and the projected probability of violent conflict. Vulnerability combines socio-economic deprivation, inequality and aid dependency with vulnerable groups: the uprooted, the undernourished, children under five and people recently hit by disasters. Lack of coping capacity combines institutions — governance and disaster risk reduction — with infrastructure: communications, roads, power, water and access to health care.

The three are combined by a geometric mean, so a country cannot offset a very high score on one dimension with a low score on another, and risk is high only where all three are at least moderate. Each layer below uses INFORM's own five-class thresholds. The overall risk map shows the familiar belt: the Sahel, the Lake Chad basin, the Central African Republic and the Democratic Republic of the Congo, South Sudan, Sudan, Ethiopia and Somalia, then Yemen, Syria and Afghanistan, with Haiti, Myanmar, Pakistan and Papua New Guinea outside it.

Switching layers shows how the belt is assembled. The natural-hazard map is centred on Asia and the Pacific rim, from Japan and the Philippines through Indonesia, Bangladesh and India to Iran, Türkiye and the Andes. The conflict map runs across the Sahel and the Middle East and into Ukraine, Myanmar and parts of Latin America. The vulnerability and coping maps are mostly a map of income, darkest across sub-Saharan Africa. The compound layer marks countries in the top third of all three dimensions at once.

Counted by population, the five INFORM classes divide the world unevenly. On the mid-2026 index, 16 countries with 837 million people are classed very high risk and a further 28 with 1.2 billion are classed high, together 2.03 billion people or 24.7% of the 8.2 billion covered. The medium class holds the largest share, 3.3 billion, because it includes India; low and very low hold the remaining 2.8 billion. The count of countries is spread more evenly than the count of people, which is why maps of countries tend to overstate how much of the world is at high risk and understate how many people are.

Compound risk needs a stricter test than the overall class, because a geometric mean can reach the high class through two bad dimensions and one middling one. Here a country counts as compound if it is in the top third of all 191 countries on all three dimensions at once: hazard and exposure of 5.0 or more, vulnerability of 4.6 or more and lack of coping capacity of 5.1 or more. Thirty-three countries meet the test, home to 1.53 billion people or 18.6% of the world. Another 30 countries, with 0.9 billion people, are in the top third on two dimensions.

The compound countries form a coherent geography. Twenty of the 33 are in sub-Saharan Africa, five in the Middle East and North Africa, three each in East Asia and the Pacific and in Latin America and the Caribbean, and two in South Asia: Afghanistan and Pakistan. Pakistan and Nigeria alone account for nearly half a billion of the 1.53 billion. Most countries meet the test on all three dimensions together or on none. Hazard, vulnerability and weak coping capacity cluster, and that clustering is what makes the belt so persistent.

If hazard decided risk, countries with the same hazard score would have similar risk. They do not. Seven countries share a hazard-and-exposure score of 8.3 out of 10: Bangladesh, the Democratic Republic of the Congo, India, Myanmar, the Philippines, Somalia and Sudan. Their overall risk runs from 5.2 in India and 5.3 in the Philippines to 7.8 in Sudan, 8.0 in the Democratic Republic of the Congo and 8.2 in Somalia. The difference is entirely in what the hazard meets: India's lack of coping capacity scores 3.9 and the Philippines' 3.8, while the Democratic Republic of the Congo scores 8.2 and Somalia's vulnerability 9.4.

The wider pattern holds across the index. Thirty-seven countries have a hazard score of 7 or more, and their overall risk ranges from 3.3 in Israel to 8.3 in South Sudan. Ten of them, with 2.5 billion people — Russia, India, the Philippines, Mexico, Indonesia, Israel, Colombia, Türkiye, Egypt and Thailand — stay in the medium class or below despite very high hazard, because their coping capacity scores 4.3 or lower. These are the exposed-but-capable countries. They are not safe — a medium-class country can still suffer a catastrophe — but their hazards rarely become national crises that need outside help.

A variance decomposition makes the point formally. Because risk is the geometric mean of its three dimensions, the variance of its logarithm across countries splits exactly into contributions from each. Hazard and exposure accounts for 38% of the cross-country variation in risk, vulnerability for 32% and lack of coping capacity for 29%. Hazard is the largest single contributor, but less than two-fifths of the total. Most of what separates countries is how their societies, economies and institutions handle the hazards they face, and much of that can be changed.

INFORM's hazard dimension has two parts, and they behave very differently. Natural hazard has a rank correlation of 0.47 with overall risk across countries; human hazard, which measures current conflict intensity and the projected probability of violent conflict, has a correlation of 0.82. Every one of the 16 very-high-risk countries scores 8.4 or more out of 10 on human hazard, and eight of them reach 9.7 or above. Their natural-hazard scores range from 2.6 in Burkina Faso to 6.6 in Somalia, which places 11 of the 16 in INFORM's very low to medium natural-hazard classes.

The scenarios in the chart below recompute every country's risk with INFORM's own aggregation formula, changing one input at a time. If every country's human hazard were capped at the global median of 2.1, no country would remain in the very-high class and the population in high- or very-high-risk countries would fall from 2.03 to 1.00 billion. If natural hazard were set to zero everywhere — a physical impossibility, used here only as a bound — 12 countries would remain very high risk and 1.65 billion people would remain in high-risk countries. Of the 44 high- and very-high-risk countries, 40 also score high for human hazard.

None of this means natural hazards are unimportant. They kill and impoverish people every year, and for small island states and deltas they are existential. The point is how they interact with conflict. Conflict adds its own hazard, raises vulnerability by uprooting people and destroying livelihoods, and dismantles the institutions and infrastructure that would otherwise cope. A drought in a country at peace is a hard year; a drought in a country at war can become a famine. The International Committee of the Red Cross has observed that of the 25 countries most vulnerable to climate change on the ND-GAIN index, 14 are mired in conflict.

INFORM publishes a trend edition that recomputes every year from 2017 to 2026 with a consistent methodology, which allows the components to be followed over time. The clearest change is in the infrastructure part of coping capacity — communications, power, water, sanitation and access to health care. It improved, by 0.2 points or more, in 154 of 191 countries and worsened in two; the cross-country average fell from 3.81 to 3.31. Mobile phones, internet access, electricity and basic water services spread across the developing world through the decade, and each makes warnings travel faster and makes recovery easier.

The institutional part of coping capacity did not move. It improved in 53 countries and worsened in 54, and its average was 5.23 in 2017 and 5.24 in 2026. Governance and disaster risk reduction are harder to build than phone networks and easier to lose. Over the same years, human hazard rose in 104 countries and fell in 30; its average rose from 3.08 to 3.68, and the number of countries in INFORM's high or very-high human-hazard classes rose from 39 to 46. Vulnerability, which includes displacement, food insecurity and recent shocks, worsened in 95 countries. Natural hazard, which in INFORM is mostly structural exposure, barely changed.

The net result is a slow divergence. Average risk edged up from 3.90 to 4.03, but the number of countries classed as very high risk rose from 10 to 16, with the increases concentrated among countries where conflict spread: Burkina Faso's risk rose from 5.3 to 7.2 as its human hazard went from 3.8 to 9.4, and Haiti's from 6.0 to 7.5 as gang violence drove its human hazard from 2.3 to 7.4. At the other end, India's risk fell from 6.1 to 5.3 as vulnerability and coping both improved, and China's from 4.3 to 2.9.

Coping capacity follows income closely. Across 189 countries, the logarithm of GDP per capita explains 76% of the variation in INFORM's lack of coping capacity: each doubling of income goes with a score about 0.76 points lower. That relationship is a reminder that development is the most powerful resilience programme there is. But the 24% that income does not explain is large, and it is the more useful part for policy, because it shows what countries at the same income have achieved and others have not.

Countries below the line cope better than their income predicts. Burundi, Rwanda, Malawi, India and Bangladesh, along with Uzbekistan, Tajikistan, Kyrgyzstan, Georgia, Armenia and Jordan, score between 1.2 and 2.0 points better than the line — often through public health systems, community-based disaster management or deliberate investment in basic services. Countries above the line cope worse than their income predicts. Many are resource-rich, such as Equatorial Guinea, Gabon, Angola, Libya and Venezuela, or fragile, such as Chad, Haiti, the Central African Republic and Papua New Guinea. Their income has not turned into institutions and services.

The benchmark used here is deliberately modest. For each country, every component of vulnerability and coping capacity is moved to the 25th percentile of its own World Bank income group — the level that the best quarter of its income peers already achieve — if it is not there already. Hazards are unchanged. Recomputed with INFORM's formula, coping capacity alone at the peer benchmark would cut the population in high-risk countries from 2.03 to 1.67 billion; vulnerability alone to 1.45 billion; both together to 0.94 billion, with no country left in the very-high class. Institutions do more of the coping work than infrastructure: on their own, they lower it to 1.83 billion; infrastructure alone barely moves it, because infrastructure has already improved.

Sorting the 44 high- and very-high-risk countries by what drives their risk gives three groups. The largest, 33 countries with 1.59 billion people, are conflict-locked: high human hazard combined with high lack of coping capacity. They include all 16 very-high-risk countries and run from the Sahel and the Horn to Yemen, Syria, Afghanistan, Myanmar, Pakistan, Haiti and Venezuela. Here the binding constraint is conflict and the institutions it destroys. Coping investments still save lives, but on their own they cannot move these countries out of high risk.

The second group, seven countries with 377 million people, are conflict-exposed but coping: Bangladesh, Ecuador, Guatemala, Iraq, Kenya, Rwanda and Ukraine score high on human hazard but have coping capacity below the high class, often through deliberate investment. Their priority is to protect what works — warning systems, health systems, social protection — from the political and security shocks that threaten it, and to reduce vulnerability, which is where the peer benchmark moves them most. The third group, four countries with 68 million people, are fragile without major conflict: Madagascar, Zambia, the Republic of the Congo and Libya, where weak coping and high vulnerability meet moderate hazards.

A fourth group sits outside the high classes but matters as a model: the exposed but capable. Ten countries with 2.5 billion people — India, Indonesia, the Philippines, Mexico, Colombia, Türkiye, Egypt, Thailand, Russia and Israel — have hazard scores of 7 or more but coping capacity of 4.3 or better, which keeps them in the medium class or below. Their experience with early warning, social protection and disaster risk reduction is the most transferable evidence available to the countries above them, and much of it was built at middle or lower-middle income.

Snapshot scores are from the INFORM Risk Index Mid-2026 (version 0.7.3, released 1 September 2026) by the European Commission's Joint Research Centre, for 191 countries, with INFORM's own class thresholds, population table, World Bank regions and income groups. Trends are from the INFORM 2026 trend edition (2017–2026), which applies a consistent methodology across years; its 2026 values precede the Mid-2026 update and are not mixed with it. Changes of less than 0.2 points are treated as unchanged. Population shares use INFORM's 2026 populations throughout.

INFORM combines two components a and b as 10 − (√(a′·b′) − 1) × 10/9, with a′ = 0.9 × (10 − a) + 1, and overall risk as the geometric mean of hazard and exposure, vulnerability and lack of coping capacity. The author checked that this reproduces every published score to within rounding (maximum difference 0.05) and uses it for all scenarios and the calculator. Top-third thresholds are the 67th percentiles across countries. The variance decomposition splits the variance of log risk into each log dimension's covariance with log risk, divided by three. The income fit is ordinary least squares of lack of coping capacity on log GDP per capita, for the 189 countries with GDP data.

Scenarios change one set of inputs and recompute the index; they are not causal estimates or forecasts. The peer benchmark sets each vulnerability and coping component to the lower of its current value and the 25th percentile of its income group. The map uses Natural Earth 1:50m boundaries via world-atlas in an Equal Earth projection; small states are drawn as circles, territories INFORM does not score are grey, and boundaries do not imply endorsement. Related reports in this series: polycrisis resilience, the polycrisis century and resilience and cascading risks.