Addressing the Pollution–Poverty–Climate Feedback Loop

The poverty-to-pollution link runs through fuel and infrastructure choices that poor households and poor governments make under constraint. Wood, dung and crop waste are free or cheap where gas and electricity are unaffordable or unavailable. Old coal plants, diesel generators, two-stroke engines and open burning of rubbish are what cities can afford when capital is scarce. The pollution-to-poverty link runs back through bodies and budgets. Outdoor PM2.5 caused an estimated 3.8 million deaths in 2017, most of them in Asia and Africa, and much more illness, lost work and lower productivity.

The climate links are where the loop closes. Most of the combustion that produces PM2.5 also produces CO₂, and much of it produces black carbon, which warms the air and darkens snow and ice. Sulphate particles do the opposite: by reflecting sunlight and brightening clouds, they mask roughly half a degree of warming, according to the IPCC's latest assessment. Climate change then feeds pollution back. Heat accelerates ozone formation, drought and heat bring larger fires and more dust, and stagnant air traps smog.

The last two links, from climate to poverty and from poverty to climate vulnerability, are the subject of reports 2026-56 and 2026-58 in this series. Climate shocks leave lasting income losses in poorer economies, and human development is among the strongest protections against them. This report focuses on the shared root: the combustion that connects the pollution and climate nodes and is shaped by poverty. That is where the cheapest integrated interventions lie.

Multiplying each country's source shares by its attributable deaths gives a global picture for 2017. Coal combustion, in power plants, industry and homes, accounted for about 610,000 of the 3.8 million deaths, oil and gas for 530,000, and biofuel burning for 690,000. Together, fuel combustion accounted for 1.8 million deaths, or 48%. Desert and anthropogenic fugitive dust accounted for 880,000, farming (mainly ammonia from fertiliser and manure, which forms secondary particles) for 370,000, and open fires and agricultural burning for 120,000. The rest came from waste, solvents, shipping and other sources.

The mix varies with income. In low-income countries, dust accounts for 45% of death-weighted exposure and open fires for 10%, because Sahelian dust and savanna burning dominate and fuel use is small. Lower-middle-income countries, including India, Pakistan, Bangladesh, Nigeria and Viet Nam, are the most combustion-heavy: biofuel accounts for 25% of exposure and coal and oil and gas 14% each. In upper-middle-income countries coal leads (19%), driven by China and South Africa. In high-income countries oil and gas (20%) and farming (16%) lead, as coal has largely been cleaned up.

The combustion share is the overlap between air-quality policy and climate policy. Replacing a coal plant with solar, a diesel bus with an electric one or a wood fire with LPG or an electric cooker cuts PM2.5 and CO₂ together, and usually black carbon too. Dust and ammonia are different. They need land, water and farming measures, and they offer no direct climate co-benefit beyond methane, which is a large part of farming's climate impact. Section 3 shows which roots dominate where.

Classifying each country by its largest human source of PM2.5 gives five archetypes. The biomass loop covers 45 countries with 3.3 billion people, including India, Indonesia, Pakistan, Brazil, Nigeria, Bangladesh, Ethiopia, the Philippines and Viet Nam. Here residential and other biofuel burning is the largest human source, and clean cooking and heating are the first lever. The coal loop covers 18 countries with 1.7 billion people and 1.6 million attributable deaths: China, Türkiye, South Africa, the Republic of Korea, Ukraine and much of the Balkans. Replacing coal in power, industry and homes is the decisive lever, and the climate co-benefit is largest.

The oil-and-gas loop covers 57 countries, including the United States, Russia, Mexico, Egypt, Iran and most of Europe, where transport, power and industrial fuels lead. The fire loop covers 31 countries, mostly in Central and Southern Africa, among them the DR Congo, Tanzania, Angola, Mozambique, Madagascar and Zambia. Savanna, forest and crop burning lead there, and median extreme poverty is 43%, the highest of any archetype. The farming loop covers six countries, including Japan, Germany, Poland and the Netherlands, where ammonia from agriculture is the largest human source.

Desert dust overlays all of these. In 38 countries, mostly in North Africa, the Sahel and the Middle East, half or more of PM2.5 is dust, and human sources form a smaller, though still deadly, remainder. There, the integrated agenda pairs combustion measures with land restoration, dust-control and heat-health measures. The archetypes are a starting point, not a verdict. India's PM2.5, for example, is 29% biofuel, 17% coal, 14% oil and gas and 15% dust, so every lever matters, but the order of priority differs from China's.

Dividing each country's fossil-fuel air-pollution deaths by its fossil CO₂ emissions gives a simple measure of the health dividend of climate action: how many lives a year each million tonnes of fossil CO₂ carries. In 2017 the figure was 105 deaths per megatonne in lower-middle-income countries, 94 in low-income countries, 44 in upper-middle-income countries and 9 in high-income countries. It is highest in Nepal (about 450), Bangladesh (250), Myanmar (220), India (118) and Pakistan (109), and lowest in Australia, Canada, Norway and the Gulf states, at 1 to 2.

The gap reflects three things: how dirty the combustion is (old coal plants, diesel and two-stroke engines, few emission controls), how many people breathe the smoke (density and proximity) and how vulnerable they are (age, health and background exposure). It means that the low- and lower-middle-income countries, which emitted 11% of fossil CO₂ in 2017, bore 34% of the deaths from fossil-fuel air pollution. High-income countries emitted 44% of the CO₂ and bore 11% of the deaths.

For integrated policy, the implication is direct. Clean-energy investment in densely populated, polluted, poorer economies buys health benefits that dwarf those available elsewhere, on top of the same climate benefit per tonne. Climate finance allocated by tonnes alone misses this. Weighting it by co-benefits, as the deaths-per-tonne measure allows, would steer more of it to South Asia and the fastest-growing cities of Africa.

Across developing countries, PM2.5 exposure rises with the share of people living below $8.30 a day (r = 0.40). The worst combinations are in South Asia, where annual exposure was 60–80 µg/m³ in India, Nepal, Bangladesh and Pakistan in 2017 and most people live below $8.30, and in West Africa, where dust adds to biomass smoke. Upper-middle-income countries are less polluted on average (35 µg/m³, population-weighted), and high-income countries far less (14 µg/m³). The WHO guideline is 5 µg/m³.

Poverty works on both sides of the link. It determines which fuels households and cities use. In 2017, most people in 56 of these countries relied mainly on polluting cooking fuels, and biofuel burning in homes drives a large share of outdoor as well as indoor pollution in South Asia and Africa. It also limits defences: air filters, cleaner homes, healthcare and the ability to move away from polluted areas are unaffordable for most. Pollution in turn lowers earnings through illness, lost workdays and reduced cognitive and physical productivity, which entrenches poverty.

The exposure gap is closing slowly. Satellite-based estimates show developing-country exposure falling by a median of about 3 µg/m³ between 2017 and 2023, much of it in China and East Asia. South Asia, the Sahel and West Africa remain far above every WHO interim target. Integrated pathways that reach poor households first, through subsidised clean cooking, clean public transport, waste collection and enforcement against open burning, protect those most exposed while cutting emissions.

Not every pollution cut helps the climate in the short run. Sulphate and other reflective aerosols, mostly from coal and oil, cool the planet by scattering sunlight and brightening clouds. The IPCC's Sixth Assessment estimates they mask about half a degree of warming. As China cut sulphur emissions by more than 70% between 2006 and 2017, and shipping fuel was cleaned in 2020, some of that masking was lost. For policy makers this creates a trap: pursuing clean air alone, which is overwhelmingly worth doing for health, adds near-term warming.

The IPCC's scenario data quantify the trap and the way out. In a high-emission pathway (SSP3-7.0), short-lived climate forcers add about 0.18°C of warming between 2020 and 2050. If air pollution is cut strongly without touching methane, the figure rises to about 0.26°C, because the cooling aerosols disappear. If methane is cut alongside, short-lived forcers add only about 0.02°C. Methane, mainly from fossil-fuel extraction, landfills and agriculture, is a potent warmer and an ozone precursor, so cutting it improves air and climate at once.

The UNEP and Climate and Clean Air Coalition Global Methane Assessment found that cutting human methane by 45% this decade would avoid nearly 0.3°C of warming by 2045. Each year it would also avoid 260,000 premature deaths, 775,000 asthma-related hospital visits and 25 million tonnes of crop losses. Black carbon from diesel, kerosene lamps, brick kilns and open burning is another double win. Integrated pathways must therefore bundle three things: cutting combustion at the root, which removes both CO₂ and particles; cutting methane; and cutting black carbon. The aerosol trap then becomes a manageable side effect rather than a reason to delay.

An integrated pathway starts from a country's dominant source and chooses the levers that cut pollution, emissions and poverty together. In the biomass loop, the lead lever is clean cooking and heating with LPG, electricity and biogas, paired with crop-residue and waste management. It cuts household and outdoor PM2.5, saves women hours a day, reduces pressure on forests and, where the wood is unsustainably harvested, cuts emissions. In the coal loop, the lead lever is replacing coal in power, industry and homes with renewables, storage and efficiency, plus pollution controls on plants that remain during the transition. That makes it the largest climate co-benefit and, in China, the largest health gain.

In the oil-and-gas loop, the levers are clean transport (electric buses, two- and three-wheelers, rail and walkable cities), fuel and vehicle standards, and cleaner power. Methane from oil and gas operations should go first. In the fire loop, where poverty is deepest, the levers are fire management, alternatives to slash-and-burn and residue burning, and secure land rights. In the farming loop, they are better fertiliser and manure management, which cuts ammonia and nitrous oxide and, through manure and rice, methane.

Three levers apply everywhere. Cutting methane through leak repair and flaring bans, landfill gas capture and better rice and livestock practices avoids the aerosol trap. Cutting black carbon from diesel, brick kilns and kerosene adds health and climate gains. Monitoring air quality makes progress visible, and many African cities still have no air-quality monitors at all. Financed together and targeted where the deaths per tonne are highest, these levers turn three separate problems into one investment case.

The archetype layer shows the five loops: biomass across South and South-East Asia, Brazil and much of East and West Africa; coal in China, South Africa, Türkiye and the Balkans; oil and gas across the Americas, Europe, Russia and the Middle East; fire across Central and Southern Africa and Australia; and farming in a handful of European and East Asian economies. Countries where dust supplies half or more of PM2.5 are outlined in the fossil, biofuel and dust layers.

The fossil layer shows where fossil fuels supply the largest share of PM2.5: Southern Africa (South Africa, Eswatini, Botswana, Lesotho), Türkiye, the Balkans and Eastern Europe, Korea, China and the United States. The biofuel layer highlights Central America, South and South-East Asia, the Great Lakes of East Africa and, through wood heating, the Nordic and Baltic countries. The dust layer covers the Sahara, Sahel, Arabia and Central Asia. The death-rate layer shows outdoor PM2.5 deaths per 100,000 people, highest in the Balkans and Eastern Europe, China, Egypt, the Caucasus, Central Asia and India, where pollution meets older populations and high rates of heart disease. The deaths-per-tonne layer peaks across South Asia, and the poverty layer shows where the loop meets deep deprivation.

Hover over or tap a country for its card: its 2017 source mix as a stacked bar, PM2.5 in 2017 and 2023, attributable deaths with their uncertainty range, fossil CO₂ per person, deaths per million tonnes of CO₂, poverty and clean cooking. Selecting a country also loads it into the pathway builder in section 9. The sortable table below lists every country.

Choose a country and set how far each source is cut: coal combustion, oil and gas combustion, biofuel burning, open fires and farming emissions. The builder starts from the country's 2017 source shares. It removes the chosen fraction of each source's contribution to population-weighted PM2.5 and scales attributable deaths in proportion. It also estimates the fossil CO₂ avoided, using the country's 2017 emissions from coal, oil and gas, and the number of people moved to clean cooking where most cook with polluting fuels.

The preset buttons load each archetype's integrated pathway: the lead lever cut by 80% and the others by 30–50%. Comparing countries shows the logic of section 4. The same pathway saves many more lives per tonne of CO₂ in Bangladesh or India than in Canada or Germany, and in countries where dust dominates, combustion cuts leave much of the PM2.5 in place.

The calculation is first-order. Real exposure–response curves are flatter at very high concentrations, so proportional scaling overstates deaths avoided in the most polluted places until levels fall further. Cutting a source also changes secondary chemistry non-linearly, and the 2017 baseline has since changed. Treat the results as the shape of the answer, not a forecast. The builder covers outdoor PM2.5 only; indoor smoke from cooking adds a large, separate health gain.

Integration starts with planning. National climate plans, air-quality plans and poverty strategies are usually written by different ministries with different targets. Countries can instead set joint targets: PM2.5 alongside CO₂ and methane in nationally determined contributions, and clean cooking and clean transport counted in both health and climate budgets. Source apportionment like the data here tells each country which sources to target first. Its archetype is a sound starting point, refined with local monitoring.

Measurement is the second requirement. Most low-income countries, and many African cities, have few or no ground-level air-quality monitors. They rely on satellites and models that cannot easily attribute pollution to local sources. Low-cost sensor networks, reference monitors and regular emissions inventories make pollution visible and enforceable. They also allow health benefits to be counted in project appraisal, where they are routinely left out, so the deaths per tonne of section 4 can guide spending.

Finance is the third. Integrated projects such as clean-cooking programmes, coal-to-renewables transitions, electric public transport and methane capture often have the highest combined returns but fall between funding windows. Climate funds count tonnes, health budgets count patients and development lenders count kilometres. Blended programmes that pay for verified outcomes on all three, carbon finance with integrity safeguards, and reform of fossil-fuel subsidies that keep polluting fuels cheap can align money with the loop. The poorest households, who breathe the dirtiest air, should be reached first through subsidies for clean fuels and protection from the cost of transition.

Coverage: 156 countries with at least one million people. Source shares and deaths come from McDuffie et al. (2021), Supplementary Data Files 1 and 2: 2017 population-weighted PM2.5, and fractional contributions by sector and fuel from GEOS-Chem sensitivity simulations with CEDS emissions. Fuel groups are coal, oil and gas, and biofuel. Dust combines AFCID and windblown dust; fires combine agricultural waste burning and other open fires. Farming is the agriculture sector, mainly ammonia. Deaths use the GBD 2019 exposure–response functions, with 95% intervals.

Source deaths are country shares × country deaths, summed (fossil 1.14 m; the paper's non-linear global estimate is 1.05 m). Archetype is the largest of biofuel, coal, oil and gas, fires and farming; "dusty" means dust supplies 50% or more of PM2.5. Deaths per megatonne is fossil-fuel PM2.5 deaths ÷ fossil CO₂ (coal + oil + gas, Global Carbon Budget via Our World in Data, 2017). Poverty: World Bank PIP, $3.00 and $8.30 a day (2021 PPP), 2017. PM2.5 2023: WDI. Short-lived forcers: IPCC AR6 WG1 Chapter 7 data repository.

Caveats: the source shares are modelled, for one year (2017), and do not cover household (indoor) air pollution or ozone. Multiplying shares by deaths ignores non-linearity. Deaths per tonne mixes local pollution intensity with population density. Archetypes ignore secondary sources. The pathway builder is first-order, as explained in section 9. The cross-country correlations are not causal.