Climate Damages Will Concentrate Among Populations Least Able to Adapt

? amir-jina frames climate-impact research around a practical problem: physical changes in temperature, rainfall, storms, and sea level do not by themselves tell policymakers what climate change will do to people. The relevant effects are on mortality, work, crops, energy use, incomes, and the ability to live safely—and on how those effects should shape mitigation and adaptation decisions.

That distinction matters because climate policy weighs the costs of acting against the damages avoided by acting. Jina argues that policy debates have often put more structure on the cost side: the expense of energy policy, fuel-efficiency standards, renewables, and other efforts to limit warming. For a long time, the benefit side was largely guesswork. Impact research attempts to estimate the damages that lower warming might avert.

The central finding is not that each additional degree produces a uniform increment of harm. The relationship is usually non-linear. A degree of warming can reduce a cold-related burden in one place while pushing another further into dangerous heat. Where a population begins, how much its climate changes, and what protection it can afford all shape the result.

Jina illustrates the point with projected U.S. summer temperatures under a high-emissions scenario. Illinois, whose historical summer temperatures sit around Mexico’s average, would by late century become hotter in summer than every U.S. state is today—roughly comparable to current Indian summers. Texas, Louisiana, and Oklahoma would reach average summer conditions comparable to present-day Saudi Arabia. The point is not merely that places become hotter, but that the consequences of a shift depend on the climate from which they begin.

The presentation briefly uses ecological suitability as a bridge to the human case. Species have ranges of temperature and precipitation in which they can flourish, with conditions worsening outside them. Under a high-emissions projection, the core climate-suitability zone for Canadian aspen shifts north and becomes disjoint from its present range. Trees and corals cannot relocate themselves at the pace that such an envelope moves.

Humans have more ways to alter their own suitability envelope. Clothing, heating, buildings, and air conditioning make many otherwise difficult environments livable. But those protections are neither complete nor free, and their availability is unequal.

Heat makes that constraint particularly clear. The relevant measure is not merely air temperature but wet-bulb temperature: a thermometer wrapped in water captures whether evaporation can cool a sweating body. As conditions approach body temperature, evaporation no longer provides adequate cooling. Jina says that at roughly 33°C wet-bulb temperature, the body begins to heat internally and cannot cool itself without an external mechanism. Air conditioning is the prevailing technological answer, but it raises energy demand and remains unaffordable for many people.

Estimating climate impacts requires more than observing that hotter and colder places have different outcomes. A simple comparison across locations confounds temperature with everything else that differs between them: income, education, public policy, geography, history, and other persistent local conditions.

Jina’s mortality example makes the problem concrete. If researchers pooled observations from two locations and drew a line through their average temperatures and mortality outcomes, they could infer that warmer temperatures lower mortality. Yet that result might simply reflect that the warmer location is richer, healthier, or better served. The cross-place correlation can point in the wrong direction.

The alternative is to use a location as its own control. Rather than compare Chicago with another city, researchers compare Chicago on otherwise similar dates across time: the same place, season, and day of the week, while accounting for measurable trends and differences. Within-location weather variation can then help identify how temperature affects mortality, labor supply, crop yields, electricity demand, crime, or another outcome.

That approach has helped drive the expansion of empirical climate-impact research over the past decade. It can also reveal adaptation: if the relationship between heat and harm becomes flatter over time, people, institutions, or infrastructure may have become better able to manage high temperatures. And it can identify thresholds that are invisible in mild places but emerge where heat becomes more extreme.

The choice of outcome matters. Counting only death certificates that identify heat as a cause misses deaths in which heat aggravates another condition. Someone with cardiovascular problems may die during extreme heat without being recorded as a heat death. Jina therefore favors all-cause mortality: examine every death and test whether mortality rises when temperatures vary, after controlling for other factors.

Even that involves judgment. Mortality may spike during a heat wave and then fall below its expected trend in the following days or weeks because some frail people die earlier than they otherwise would have. Jina calls this forward displacement; researchers also use the term “harvesting,” which he dislikes. In the work he describes, heat effects are often assessed over roughly two to three weeks, and sometimes as long as two months. Deaths within that window may be treated as deaths that would have occurred anyway. This is not only a statistical choice, he says, but also a moral choice about what “would have happened anyway” means.

Within-location weather variation is useful because it can be treated as close to random. But it does not settle every question about a persistently warmer world. Climate change is not merely a string of hotter versions of today’s days; it can induce slower changes in behavior, technology, institutions, and settlement. El Niño and major disasters offer complementary evidence on broader environmental disruption.

Empirical response curves show why average warming cannot be converted directly into average damages.

For maize, the relationship has a sharp threshold. At lower temperatures, warming can improve yields. But once temperatures exceed roughly 29°C, or about 85°F, yields fall steeply. Jina describes a single day around 35°C, or 95°F, as associated with about a 5% decline in yield. The significance of that slope depends on exposure: a place that rarely reaches the damaging range faces a different risk from one that enters it regularly.

Mortality follows a U-shape. Both cold and hot days increase deaths, with a middle range at which temperature-related mortality is minimized. High-risk labor has a threshold-like response as well: workers put in fewer hours or become less productive during very hot days. Jina stresses that this is not simply leisure replacing work. Lost work is not fully made up through another activity. Electricity demand is also U-shaped, rising as people heat themselves in cold conditions and cool themselves in hot ones.

Other outcomes offer fewer compensating gains. Violent crime rises approximately linearly with maximum daily temperature in the relationship Jina presents, while property crime rises and then flattens. Where a response is linear or where a high-heat threshold is one-sided, warming can produce losses nearly everywhere rather than a balance of regional gains and losses.

Projecting those relationships to late-century conditions under RCP8.5 produces a distinctive U.S. geography while holding adaptation fixed. In colder northern areas such as Minnesota and North Dakota, warming can reduce exposure to dangerous cold without yet moving populations far up the hot side of the mortality curve. Across much of the South, average conditions already lie on the harmful side of that curve; additional warming raises mortality. Energy expenditures follow a related pattern, declining where reduced heating needs dominate and rising where cooling needs grow.

When the presentation aggregates direct effects across mortality, crop yields, labor, energy, and crime using monetary values including prevailing wages, crop values, and the Environmental Protection Agency’s value of a statistical life, damages concentrate in the South. Mortality is especially consequential in that aggregation because it is a large non-market harm.

The accompanying county map makes the distributional implication visible: projected direct damages, expressed as a share of county GDP under RCP8.5 in 2100, are concentrated across lower-income Southern counties rather than spread evenly across the country. The following income-ranked comparison is an estimate under that scenario and fixed-adaptation assumption.

Jina describes a gap of somewhere between five-to-one and 20-to-one across counties. The precise estimates carry uncertainty, but the pattern follows from the combination of initial climate, the shape of temperature-response curves, and the value of damage relative to local income.

Adaptation means flattening the temperature-response curve. In the idealized case of complete adaptation, a severe hot or cold day would no longer change mortality attributable to temperature. People would still die on those days, but temperature would not raise or lower the death rate.

That is not what researchers observe. Jina says complete adaptation is seen essentially nowhere, including in the United States. Protection is imperfect because it carries real costs. A fully climate-controlled environment—his deliberately extreme example is an Abu Dhabi-style indoor ski slope—could insulate people from ambient temperature, but only through major expenditure and energy use.

Global mortality evidence suggests adaptation occurs along more than one dimension. People accustomed to high temperatures can change behavior: avoid strenuous activity in the hottest part of the day, shift schedules, stay hydrated, and organize daily life around heat. For people over 65, the effect of a 35°C day relative to a 20°C day falls as average local climate becomes hotter. Jina treats that as evidence of acclimatization in a broad sense, encompassing behavior and institutions as well as biology.

But income constrains how far adaptation can go. Low-cost measures can help, yet air conditioning and other capital-intensive protections remain out of reach for many populations that need them most. Oslo, in Jina’s comparison, loses cold days and gains days in a comparatively temperate range, a shift that looks beneficial on its mortality curve. Accra, Ghana loses cooler days and gains extreme-hot days, shifting further onto the steeply harmful part of its curve.

Electricity consumption makes the constraint visible. In wealthy settings, energy demand rises for cooling in heat and heating in cold. In the lower global income deciles, however, the relationship between temperature and aggregate energy use is close to absent. That does not mean heat is costless. Jina’s interpretation is that households cannot purchase enough cooling for the damage to appear as higher electricity consumption.

Jina places Mexico around the eighth global income decile in the energy analysis, while the United States sits in the richest decile. By roughly the seventh decile and below, he says, aggregate energy data no longer show appreciable temperature-driven increases in energy use. This global energy analysis differs from the fixed-adaptation U.S. projections: it estimates adaptation from historical relationships between income and cooling-related energy use.

Neither approach resolves the future. Countries may grow faster or slower than projected; cooling technology may change; energy systems may become cleaner or more efficient; institutions may adapt differently; and migration may alter where people face exposure. More efficient air conditioners, powered by renewable energy and able to provide more cooling with less energy, could change projected risks. Impact research can help identify where such investments would have the greatest protective value.

Jina lists climate-linked effects across mortality, illness, vector-borne disease, labor productivity, absenteeism, agricultural yields, livestock, energy demand, coastal risk, wildfires, migration, infrastructure, crime, conflict, and even the language people use online. Climate conditions run through economic and social life through many channels, not a single damage function.

For global mortality, the work he describes uses subnational, age-specific records covering 55% of the world’s population and roughly 400 million deaths. The familiar U-shape appears, particularly for people over 65: both extreme cold and extreme heat raise mortality. The shape and intensity of the response vary by age, climate, and capacity to adapt.

Already-hot places can face greater harm even when they warm more slowly than high-latitude regions. The determining issue is proximity to damaging thresholds. Tropical areas may begin closer to the steep heat-damage range, while lower income limits access to protective infrastructure. The presentation’s global mortality-risk map and regional comparison place much larger mortality burdens, as a share of GDP, in already hot regions including Pakistan, Bangladesh, sub-Saharan Africa, and India than in the United States or European Union.

A whole-economy analysis reaches a related conclusion. The relationship between annual average temperature and GDP per capita is non-linear. Jina is explicit that GDP is not a comprehensive measure of climate damage: it omits non-market harms, including mortality. But it is a measure of productivity. Countries such as Germany, the United Kingdom, and France may see little effect from modest warming, or some benefit if they sit on the upward side of the estimated curve. Indonesia, India, and Nigeria lie on the steeply declining side, where further warming is associated with sharper GDP losses.

Historical estimates presented in the talk associate observed climate change from 1961–2010 and 1991–2010 with lower GDP per capita in many tropical countries and gains in some cooler ones. Jina’s work in progress makes the distributional comparison directly. At 3°C of warming, he estimates that roughly 80% to 90% of cumulative damages fall on the poorest half of the world’s population, which accounts for roughly 18% of historical and projected emissions through 2100.

The presentation’s cumulative-damages chart ranks the global population from poorest to richest and shows the damage curves rising rapidly before the midpoint of the distribution, while cumulative emissions rise much more gradually. The scale of damages changes across warming scenarios, Jina says, but the concentration of harm among lower-income populations remains visible even at 2°C.

Weather-based causal designs compare conditions within a place over time. Jina argues that researchers also need evidence from environmental disruptions that are broader, longer-lived, and geographically extensive.

El Niño is a tropical Pacific cycle lasting roughly four to seven years and affecting about 30% of the global climate. Its relationship to climate change is ambiguous, but it shifts temperatures and precipitation across large areas. That makes it useful, in Jina’s account, for studying consequences beyond the contrast between a hot day and a cooler day in the same city.

He describes data from 80 surveys across roughly 30 countries, beginning in the 1980s, that find lower weight-for-age among children under five during El Niño conditions than during La Niña conditions. For the 2015 El Niño, Jina says the estimate implied roughly 7 million additional children entering acute starvation levels—more than in the first year of COVID, despite that pandemic’s supply-chain disruptions.

He also cites research on conflict onset. In countries whose average climate shifts with El Niño, annual conflict risk rises from a baseline near 3% to roughly 6% during a large El Niño. In countries not exposed to the related shift in average climate, that risk remains relatively flat. These are not direct forecasts of climate change; they are evidence that large environmental shifts can alter food security and conflict risk through channels that isolated weather comparisons may not fully capture.

Disasters pose a related question: do hurricanes and typhoons simply interrupt economic activity, or do they alter an economy’s growth path? Jina says theory has supported several possibilities, including creative destruction, rebuilding into more productive capital, recovery to the prior trend, and lasting damage. Better measurement of storm exposure is necessary to distinguish among them.

The hurricane-growth trajectory he presents is a gradual divergence from a pre-storm growth trend rather than an immediate economic collapse. A single 90th-percentile cyclone event is associated with a predicted 2.9% loss over 20 years. Repeated storms layer those effects on top of one another, creating a persistent drag in disaster-exposed economies. Jina says hurricane or typhoon effects can lower growth by a couple of percent for up to 20 years after a strike, making recurring exposure especially consequential in places such as the Philippines.

The United States is not exempt. Jina says U.S. data also show persistent effects and describes the country as relatively less adapted in some respects, including through its flood-insurance arrangements. He points to Tatyana Deryugina’s work on East Coast hurricanes: government transfers rise after a storm and continue increasing for up to a decade. In Jina’s account, transfer effects ten years later are ten times greater than damage estimates from the year of the hurricane.

Employment losses and unemployment insurance are among the mechanisms he identifies. More broadly, social safety nets already provide a form of climate adaptation even though they were not designed or budgeted as such. As climate conditions shift, those systems can come under increasing strain.

Jina does not treat high-emissions projections as precise forecasts. Asked about the use of RCP8.5, he says he used it primarily to make geographic and distributional patterns visible, not to offer definitive damage totals. Lower-warming scenarios reduce aggregate damages, while the allocation of harm toward lower-income populations remains robust.

For the sectors included in the estimates he discusses, Jina gives central figures of roughly $3 trillion in annual damages at 2°C of warming and $5 trillion at 3°C, while stressing substantial uncertainty. In his formulation, research knows more about who is likely to suffer from climate change than about the exact total cost.

The projections combine local estimates of how temperature, sea-level rise, extreme precipitation, and other conditions affect observable outcomes; measures of factors that mediate those effects, such as income and average climate; and projections of future climate and incomes. Some estimates hold adaptation fixed to expose the distribution implied by present relationships. Others incorporate historically observed adaptation, such as the relationship between income and cooling-related energy use. Neither approach can fully resolve what people, institutions, technologies, or migration patterns will do over the next century.

Migration is a major unknown. The baseline does not assume that hundreds of millions of people will readily relocate from highly exposed places to safer, cooler ones. Jina is skeptical of arguments that treat such movement as an easy solution: a world able to move populations on that scale, he says, would look unlike the one that exists today.

The value of the estimates is not that projected damage is inevitable. It is that they locate exposure and identify what would have to change to reduce it: affordable cooling, efficient clean energy, disaster financing, and social protection capable of absorbing the risks where the estimated burden is already concentrating.