Mapping the Interdependencies of Twenty-First-Century Risk

The network places the 18 pressures on a ring, grouped by domain, and draws a line between any two whose rank correlation across countries passes the chosen threshold. Line weight shows strength; dashed lines show negative links, where one pressure tends to be high where the other is low. At the default threshold of 0.4, the raw network has 56 links, 49 of them crossing domains. The most connected pressures are the child share of the population, population growth, weak governance, the electricity gap and urban growth: each is linked to many others.

Switching to income-controlled links redraws the web using partial correlations that hold the log of GDP per capita constant. The question becomes: among countries of similar income, do these pressures still travel together? At a threshold of 0.25 the controlled network has 38 links, 30 of them crossing domains, and the hubs change. Population growth, resource dependence, conflict hazard, urban growth and air pollution move to the centre; the digital gap, food insecurity and governance, so central in the raw web, lose most of their links.

Hover over or select a pressure to see its links and their strength. The heatmap below the network shows every pair at once. Because the analysis is cross-sectional, a link says nothing about direction: conflict may drive displacement, displacement may fuel conflict, or a third factor may drive both. The network is a map of co-occurrence for the evidence in later sections to explain.

Ranking each pressure by its correlation with low income separates the web into two kinds of risk. The digital gap (0.88), the child share of the population (0.87), the electricity gap (0.80), weak governance (0.79), food insecurity (0.71), population growth (0.68) and urban growth (0.59) are tightly bound to poverty. They co-occur because poverty generates all of them: low income means fewer people online and on the grid, higher fertility, weaker states and less secure diets. Much of the density of the raw web — and much of the rhetoric of interlocking crises — reflects this shared root.

Four pressures barely track income. Uprooted people (−0.03), weak growth (0.12), natural hazard exposure (0.19) and youth unemployment (−0.19) are found at every income level. Natural hazards strike rich and poor alike; displacement follows conflict and disaster rather than poverty as such; youth unemployment is highest in southern Africa and parts of North Africa and the Middle East, many of them middle-income economies. External debt service runs the opposite way to poverty (−0.54): it is heaviest in middle-income economies that borrow on commercial terms, not in the poorest countries, whose debt is often concessional.

This matters for policy because poverty-bound pressures share a remedy. Raising incomes and building basic services — power, connectivity, schooling, health — weakens many links at once, and the evidence from this series is that it can be done: the resilience-stack report shows electricity, water and connectivity spreading fast in the past two decades. The pressures that do not track income need their own instruments: disaster risk reduction, refugee protection, labour-market policy and debt management. No growth strategy will remove them.

The links that survive income control describe the interdependence that development alone will not dissolve. The strongest cross-domain link is between conflict hazard and uprooted people (partial correlation 0.65): at any income level, conflict and displacement come together. Conflict is also linked to weak governance (0.45), resource dependence (0.43), natural hazard exposure (0.36) and drought (0.35). The environmental links fit UNHCR's finding that three-quarters of forcibly displaced people live in countries highly vulnerable to climate hazards.

A second hub is resource dependence. Countries that rely heavily on natural-resource rents have faster population growth (0.50), faster urban growth (0.42), more conflict (0.43) and weaker governance (0.36) than countries of similar income. This is the cross-country signature of the resource curse: rents that substitute for taxation weaken the bargain between state and citizens and give armed groups something to fight over. Population growth itself is linked to urban growth (0.76), the electricity gap (0.43) and weak growth (0.35): fast-growing populations stretch infrastructure faster than it can be built.

One surviving link needs a caution. Air pollution is linked to conflict (0.64) and resource dependence (0.47) after income control, but the satellite measure of fine particles includes desert dust, so it partly maps the Sahel and the Middle East, where conflict and oil are also concentrated. The link is geographic as much as causal. Inflation keeps modest links with weak governance (0.25) and conflict (0.20), and weak growth with food insecurity (0.26). External debt service, by contrast, is almost unconnected once income is controlled: it is a channel of global finance rather than of local co-occurrence.

A cluster is a set of three pressures, from at least two domains, in which every pair is linked with income held constant (partial correlation of 0.25 or more). Clusters are where reinforcement is most plausible: if A goes with B, B with C and C with A, a shock to any one can propagate around the triangle. Thirty such clusters exist among the 18 pressures. They are not causal loops, which would need evidence of direction, but they mark where to look for them.

The strongest clusters fall into three families. The first joins resource dependence with rapid population growth and either a young population or fast urban growth (average link 0.56). The second joins population growth, urban growth and the electricity gap (0.54): fast-growing cities outrunning infrastructure. The third joins conflict with uprooted people and weak governance (0.45), and with air pollution and resource dependence. Heat appears in a cluster with population growth and a young population (0.47), placing the hottest countries among the fastest-growing.

The families overlap through shared nodes, which is what makes them a web rather than a list. Population growth sits in the first two families; resource dependence links the first and third; conflict anchors the third. A country with all three — oil or mineral dependence, fast population growth and active conflict — sits where the three families meet, and section 5 shows that such countries are among the most unstable in the world.

Interdependence shows up within countries as stacking. Each country receives a score for each of the four driver domains — environment, economy, demography and technology — equal to the average of its percentile ranks on that domain's indicators, and a domain counts as stressed if the score is in the top third. Ninety-nine countries, with 3.3 billion people, are stressed in none; 44, with 3.2 billion, including India, Indonesia, Brazil and Bangladesh, in one; 23 in two; 20 in three; and 5 — Sudan, South Sudan, Yemen, Syria and Mozambique — in all four.

Instability rises step by step with stacking. Among countries stressed in no driver domain, 3% are in the top third for instability; with one domain, 30%; with two, 74%; with three, 85%; with all four, every one. The median instability score rises from 0.31 to 0.91, and across all countries the average driver score and the instability score correlate at 0.86. Part of this is shared construction — poverty raises both — but the gradient is too steep and regular to be only that, and it is consistent with the reinforcing clusters of section 4.

The map shows the geography. The countries stressed in three or four domains run from the Sahel through Nigeria and Central and East Africa to the Horn, Yemen, Syria, Iraq, Afghanistan and Pakistan, with Mozambique and Angola to the south — the same belt that appears in this series' maps of vulnerability and future exposure. Switch layers to see each domain alone: the environmental map is broader, taking in South Asia and the Middle East, while the technological map is mostly sub-Saharan: 42 of the 59 countries in its top third.

A cross-country network captures pressures that vary between countries. It misses pressures that act on the whole system at once. The World Economic Forum's Global Risks Report 2026 ranked geoeconomic confrontation — sanctions, tariffs, export controls and the strategic use of economic tools — as the top risk for 2026, followed by state-based armed conflict, extreme weather, societal polarisation and misinformation. Half of its respondents expected a turbulent or stormy two years, and 68% a multipolar or fragmented order over the next decade. These are links between states, not within them, and they shape every national web at once.

Technology is the clearest case. The technological pressures measured here are access gaps, which are mostly poverty. But the WEF survey's fastest-rising long-term concern is the adverse outcomes of artificial intelligence, which climbs from 30th in the two-year outlook to 5th over ten years. Misinformation and disinformation rank among the top short-term risks. These pressures are concentrated in connected, high-income societies, where they interact with polarisation and trust in institutions. The digital divide and the risks of digital saturation are two different technological pressures, and only the first appears in this network.

Finance is the third. Debt service barely links to other pressures across countries once income is controlled, but globally it ties them together. UNCTAD reports that developing countries paid $921 billion in net interest on public debt in 2024, and that 61 spent at least a tenth of government revenue on interest. Higher global interest rates, a strong dollar or a commodity shock move every country's debt burden at once, then constrain spending on adaptation, health and education. The global layer is where shocks synchronise; the national web is where they land.

A network view changes how priorities are set. In a list of risks, each gets its own programme. In a web, the most effective interventions are those at nodes with many strong links, because weakening a hub weakens every path through it. The analysis identifies four hubs. Peace and institutions anchor the conflict family: conflict hazard and weak governance link to displacement, resource dependence, drought and natural hazards even among countries of similar income. Conflict prevention, mediation and building state capacity are therefore the interventions most likely to break several reinforcing clusters at once.

Economic diversification addresses the resource-dependence hub, which links to population growth, urban growth, conflict and governance beyond what income explains. Moving revenue from rents to taxes, investing rents transparently and building other sectors loosen the resource curse at its root. The demographic transition — driven by girls' education, health care that lowers child mortality, access to family planning and urban jobs — reduces the population-growth hub, which links to urban growth, infrastructure gaps, weak growth and heat exposure. It works slowly, which is why it must start early.

The fourth hub is the poverty core itself. The digital and electricity gaps, weak governance, food insecurity and young populations are bound together by low income, so broad-based growth and basic infrastructure weaken all of them together. Pressures outside the core need their own tools: disaster risk reduction for natural hazards, protection and integration for the displaced, debt restructuring and cheaper finance for middle-income economies, and governance of artificial intelligence and information integrity at the global level. The web is dense, but it is not uniform. Knowing where its hubs are is the first step to cutting it.

Indicators are oriented so that higher means more pressure. From INFORM Risk Mid-2026: natural hazard, drought, uprooted people, human (conflict) hazard, governance and food security scores. From ERA5 via Our World in Data: 1991–2020 mean temperature. From SatPM via Our World in Data: population-weighted PM2.5, which includes natural dust. From the World Bank WDI: median inflation 2019–2024; average external debt service as a share of GNI 2019–2024 (118 developing economies only); latest youth unemployment; average natural-resource rents 2019–2023; negative average per-capita growth 2015–2024; average urban growth 2019–2024; the shares offline and without electricity. From UN WPP 2024: population growth 2025–2050 and the under-15 share in 2025.

Links are Spearman rank correlations across countries with pairwise-complete data. Income-controlled links are partial Spearman correlations holding the log of GDP per capita (PPP, latest year) constant, for the 184 countries with GDP data. Domain scores average a country's percentile ranks within each domain; a domain is stressed at a score of 0.67 or more. Clusters are triples from at least two domains with all three partial correlations at 0.25 or more. The node layout is a ring grouped by domain.

Correlations describe co-occurrence, not causation or direction. Several INFORM indicators are composites that partly reflect poverty, which inflates the raw web — one reason for presenting income-controlled links. The PM2.5 measure carries desert geography. External debt service covers only developing countries. Global pressures that do not vary across countries are discussed from the literature, not measured. The map uses Natural Earth 1:50m boundaries via world-atlas; boundaries do not imply endorsement. Related reports: the polycrisis century, the pollution–climate nexus and the geography of vulnerability.