Do Cold Deaths Outweigh Heat Deaths?

Yes, for now. Cold kills about 9 to 17 times as many people as heat, and on the best global model the warming of 2000–2019 saved about 160,000 lives a year net. The saving is modelled, not counted, and goes mostly to cool, rich countries. With more warming the balance turns toward harm, first in hot, poor places.

By AI (Claude Opus 5.5, thinking level unavailable) Draft:  October 2026

"Although cold-related deaths currently exceed heat-related deaths, heat-related deaths are expected to exceed cold-related deaths in a high-warming scenario." — the Lancet Countdown, 20241

Summary

Yes, for now.  Cold kills far more people than heat, and on the best global model the warmer temperatures of recent decades have saved more people from cold than they have killed with heat.  That saving is a model result, it goes mostly to cooler and richer countries, and it shrinks as the world warms.  With more warming the balance turns toward harm, first in hot, poor places.

  1. Cold kills about 9 to 17 times as many people as heat.2 Antonio Gasparrini's study of 74 million deaths found that most of the cold toll comes on moderately cold days, not in cold snaps.3
  2. Recent warming has cut the net toll, on the best global model. Holding population and income fixed, Qi Zhao and colleagues found that the temperatures of 2016–19, compared with 2000–03, meant roughly 275,000 fewer cold deaths and 113,000 more heat deaths a year, a net saving of about 160,000.45 Bjørn Lomborg's headline figure of 166,000 is the same arithmetic, and on the numbers as published it is fair.6
  3. It is a modelled saving, not a count. Yuming Guo, one of Zhao's co-authors, says the paper does not show that the change was caused by climate change.7 The IPCC's health chapter finds only "variable evidence" that milder winters have reduced winter deaths in practice.8
  4. Heat deaths from warming are real. About 37 % of warm-season heat deaths in 732 locations are attributable to human-caused warming.9
  5. With more warming, the balance turns. Without adaptation, projections find a net rise in temperature deaths under high emissions everywhere except the coldest regions.1011 Allowing for adaptation and rising incomes, Tamma Carleton's team projects an average of 11 extra deaths per 100,000 people a year by the end of the century under a moderate scenario, with a range that still includes a net saving, and 73 per 100,000 under the highest.12
  6. The places where heat matters most are the least studied. The main studies have few or no data from Africa, the Middle East and much of South Asia.39

This article is one part of Does Warming Do More Good or Harm?, which weighs deaths against crops, sea level, extremes and ecosystems.  The economic estimates that put all of these on one scale are in What Do Economists Say Warming Will Cost?.  The warming itself, about 1.3 °C since the late 1800s, is measured in Is the Earth Warming?.

Part 1: A net saving so far

Fewer cold deaths are the benefit the "net good" side leans on most, and the data support it more than many readers would expect.

Cold kills far more people than heat.  Antonio Gasparrini and colleagues analysed 74 million deaths in 13 countries and found that "More temperature-attributable deaths were caused by cold (7·29%, 7·02–7·49) than by heat (0·42%, 0·39–0·44)."  Most of it came from moderately cold days, not cold snaps.3  A global model by Qi Zhao and colleagues, covering 2000 to 2019, put the toll at 5,083,173 deaths a year:  about 4.59 million from cold and 0.49 million from heat.4  For Europe's cities, a 2023 study corrected in 2024 finds 130,228 cold deaths a year against 13,589 heat deaths.13  The ratio in these studies runs from about 9 to 17 cold deaths for each heat death.2

Recent warming has cut the net toll, on the best global estimate.  Zhao's team held each grid cell's population, death rate and income at 2010 levels, so that the results "reflect the excess death burden purely caused by temperature change".4  Between 2000–03 and 2016–19, while their populated grid cells warmed 0.26 °C per decade, the cold share of deaths fell by 0.51 percentage points and the heat share rose by 0.21, "leading to a net reduction in the overall ratio".4  In deaths, that is roughly 275,000 fewer cold deaths and 113,000 more heat deaths a year, a net saving of about 160,000 a year, with an uncertainty range of about 70,000 to 240,000.5  The authors' own words are cautious:  "The results indicate that global warming might slightly reduce the net temperature-related deaths, although, in the long run, climate change is expected to increase mortality burden."4

Bjørn Lomborg made the same arithmetic into a headline:  "The total impact of more than 116,000 more heat deaths each year and almost 283,000 fewer cold deaths year is that by now, the temperature rise since 2000 means that for temperature-related mortality we are seeing 166,000 fewer deaths each year."6  His figures are close to those above.  The fact-checking site Climate Feedback published a reply from Yuming Guo, Zhao's second author.  Guo agreed that cold deaths fell, heat deaths rose and the net fell, but wrote:  "it is not correct to interpret that this net decrease was caused by climate change. We just estimated the trend of mortality burden related to non-optimal temperatures, but did not do further analysis to examine whether this change is due to climate change or other factors."7  The paper holds population and income fixed so that its trend reflects "the excess death burden purely caused by temperature change".4  Read together, the net fall is the modelled effect of the temperatures that actually occurred, but the paper does not separate human-caused warming from natural swings, and it does not show that the cold deaths its model removes were in fact avoided.  On the numbers as published, Lomborg's arithmetic is fair.  His critics are right that it is a short-term result from temperature alone, which the same paper expects to reverse.

Part 2: Three cautions

Three cautions keep this from being the last word.

  1. The heat deaths are real and attributable. Ana Vicedo-Cabrera and 64 co-authors compared the actual warm-season climate of 1991–2018 with a modelled climate without human influence. Across 732 locations, 37.0 % of warm-season heat deaths were due to human-caused warming, with the largest shares in Iran, Kuwait, the Philippines, Thailand and parts of Latin America.9 That study counts no cold deaths, so it does not net the two, but it shows the cost side is not hypothetical.
  2. Not every winter death is a temperature death. Kristie Ebi, a leading climate-health researcher, warns that "it is common, but incorrect, to assume a similar (but opposite) association occurs in the winter" as in summer.7 The IPCC's health chapter says that despite milder winters "there is variable evidence for a consequential reduction in winter mortality and susceptibility to cold over time", with some countries showing "stable or even increasing trends in cold-attributable mortality fractions".8 Zhao's net saving is a model result: it applies fixed temperature-death curves to the observed temperatures. It is not a count of deaths that did not happen.
  3. The poorest and hottest places are the least studied. Gasparrini's data had no locations in Africa or the Middle East,3 and Vicedo-Cabrera's lacked "large parts of Africa and South Asia".9 Those are where heat matters most.

Part 3: Turning with more warming

With more warming, the balance turns, first in hot places.  Gasparrini's 2017 projections, which assume no adaptation, found "a net increase in temperature-related excess mortality under high-emission scenarios, although with important geographical differences".10  Under the moderate RCP4.5 scenario, close to the IPCC's intermediate pathway of 2.1 to 3.5 °C by 2081–2100,14 the net change by the 2090s was close to zero or negative in North America, northern Europe, East Asia and Australia, and positive in southern Europe, South America and Southeast Asia.10  For Europe's cities, Pierre Masselot's team found in 2025 that "with no adaptation to heat, the increase in heat-related deaths consistently exceeds any decrease in cold-related deaths across all considered scenarios", though "only in Northern Europe the decrease in cold-related deaths slightly offsets the increase in heat-related deaths".11

Adaptation and wealth change the size of the answer.  Tamma Carleton and fifteen co-authors, including Michael Greenstone and Solomon Hsiang, built death-temperature curves from 40 countries that let richer and hotter places cope better.  They found that "today’s cold locations are projected to benefit, while today’s poor and hot locations have large projected damages".12  Their mean estimate of extra deaths at the end of the century "falls from 73 deaths per 100,000 under RCP8.5 to 11 deaths per 100,000 under the emissions stabilization scenario of RCP4.5", and under RCP4.5 the 10th to 90th percentile range runs from 36 fewer to 62 more.12  Ignoring adaptation and income growth, they say, would overstate the toll "by a factor of about 3".12

Objections and responses

  1. "Cold kills ten times as many people as heat, so warming saves lives." The ratio is right, and on Zhao's model the warming of 2000–2019 did save more lives from cold than it cost in heat.4 But the ratio describes today's climate, not the effect of the next degree. The extra deaths from each further degree of heat rise faster than the savings from each degree less of cold, and every projection reviewed here finds the net turning to harm under high warming, outside the coldest regions.1011 The Lancet Countdown, no friend of the benefit side, says the same thing from the other direction: cold deaths "currently exceed" heat deaths.1

Issues and unanswered questions

  1. How much winter mortality is really temperature? Zhao's net saving of about 160,000 lives a year assumes fixed temperature-death curves. The IPCC's health chapter reports only "variable evidence" that milder winters have reduced winter deaths in practice.8 If much winter excess comes from seasonal influenza and indoor crowding rather than cold itself, the cold benefit is smaller than the models say.
  2. The hottest, poorest places are the least measured. Africa, the Middle East and much of South Asia are missing or thin in every mortality study used here. These are the regions where heat matters most, so the global net is least certain exactly where the stakes are highest.
  3. Not fully read. The expression of concern attached to Masselot 2023 was not obtained.

Verdict

The verdict on temperature deaths:  on the best global model, the warming of the last few decades has saved more people from cold than it has killed with heat.  That saving is front-loaded and regional.  It goes mostly to cooler and richer countries.  Under the moderate scenario, the global net turns modestly harmful on average:  Carleton's mean of 11 extra deaths per 100,000 a year is roughly a million a year in a world of 9 billion, set against today's modelled saving of about 2 per 100,000, though their range still includes a net saving.15  Under high warming the heat side wins almost everywhere except the far north.

Whether this saving, set against crop losses, rising seas and the other effects, leaves the warming so far a net good or a net harm is weighed in Does Warming Do More Good or Harm?.

Sources

  1. 1.

    Romanello, Marina et al.  “The 2024 report of the Lancet Countdown on health and climate change:  facing record-breaking threats from delayed action.”  The Lancet 404(10465):1847–1896.  2024.  The yearly report of a collaboration of more than a hundred health and climate researchers, tracking health indicators of climate change.  Its heat-death figures are model estimates for people over 65, set against what ageing alone would have produced;  it does not net them against cold deaths.  Mirror:  PMC

    1. 1a.

      Section Key messages:  “Heat-related mortality of people older than 65 years increased by a record-breaking 167%, compared with the 1990s, 102 percentage points higher than the 65% that would have been expected without temperature rise”.  Section Indicator 1.3.1:  “the climatic suitability for the transmission of dengue by Aedes albopictus and Aedes aegypti increased by 46·3% and 10·7%, respectively, between 1951–60 and 2014–23”.  Section Indicator 1.1.5:  “Although cold-related deaths currently exceed heat-related deaths, heat-related deaths are expected to exceed cold-related deaths in a high-warming scenario.” ↩ ↩

  2. 2.

    Cold deaths divided by heat deaths:  Gasparrini 2015, 7.29 % ÷ 0.42 % = 17 3;  Zhao 2021, 4,594,098 ÷ 489,075 = 9.4 4;  Masselot 2023 as corrected, 130,228 ÷ 13,589 = 9.6 13. ↩ ↩

  3. 3.

    Gasparrini, Antonio et al.  “Mortality risk attributable to high and low ambient temperature:  a multicountry observational study.”  The Lancet 386(9991):369–375.  2015.  A time-series study of 74 million deaths in 384 places in 13 countries, 1985–2012, measuring how daily death rates rise on days colder or hotter than each place's optimum.  It established that cold, mostly moderate cold, accounts for far more temperature deaths than heat.  Mirror:  PMC

    1. 3a.

      Page 369:  “More temperature-attributable deaths were caused by cold (7·29%, 7·02–7·49) than by heat (0·42%, 0·39–0·44). Extreme cold and hot temperatures were responsible for 0·86% (0·84–0·87) of total mortality.”  The Lancet prints decimal points as raised dots. ↩ ↩ ↩

    2. 3b.

      Page 374:  “First, although this investigation includes populations with markedly different characteristics and living in a wide range of climates, the findings cannot be interpreted as globally representative. We did not include entire regions, such as Africa or the Middle East, and the assessment was mainly restricted to urban populations.” ↩ ↩

  4. 4.

    Zhao, Qi et al.  “Global, regional, and national burden of mortality associated with non-optimal ambient temperatures from 2000 to 2019:  a three-stage modelling study.”  The Lancet Planetary Health 5(7):e415–e425.  2021.  A global model built from daily deaths and temperatures in 750 places in 43 countries, extended to every populated grid cell on earth for 2000–2019.  It is the first global estimate of cold and heat deaths and of how they changed as the world warmed, holding population and income fixed.

    1. 4a.

      Page e420:  “Globally, 5 083 173 deaths (95% eCI 4 087 967–5 965 520) were associated with non-optimal temperatures per year, consisting of 4 594 098 cold-related deaths (3 337 222–5 640 617) and 489 075 heat-related deaths (304 216–732 518; table 1).” ↩ ↩ ↩

    2. 4b.

      Page e424:  “In this study, we explored the temporal change in temperature-related mortality burden by fixing socioeconomic factors (ie, population, mortality, and GDP per capita) at the 2010 level. This assumption allows the results to reflect the excess death burden purely caused by temperature change.” ↩ ↩

    3. 4c.

      Page e415:  “From 2000–03 to 2016–19, the global cold-related excess death ratio changed by –0·51 percentage points (95% eCI –0·61 to –0·42) and the global heat-related excess death ratio increased by 0·21 percentage points (0·13–0·31), leading to a net reduction in the overall ratio.”  Page e423:  “The global daily mean temperature increased by 0·26°C per decade during this time”. ↩ ↩ ↩ ↩

    4. 4d.

      Page e423:  “The results indicate that global warming might slightly reduce the net temperature-related deaths, although, in the long run, climate change is expected to increase mortality burden.” ↩

  5. 5.

    Zhao's 5,083,173 temperature deaths are 9.43 % of all deaths in the modelled cells 4, which implies about 53.9 million deaths a year.  A change of −0.51 percentage points in the cold share is about 275,000 fewer cold deaths;  +0.21 points in the heat share is about 113,000 more heat deaths;  the net −0.30 points (95 % range −0.13 to −0.44) is about 162,000 fewer (70,000 to 237,000) 4.  The figures compare the rates of 2000–03 and 2016–19 with population, mortality and income held at 2010 levels. ↩ ↩ ↩

  6. 6.

    Lomborg, Bjørn.  “The heresy of heat and cold deaths.”  lomborg.com.  n.d.  Lomborg's blog reply, written after September 2021, to a fact-check of his claim that warming has so far saved lives on balance, setting out his reading of Zhao and colleagues and of Gasparrini's projections.

    1. 6a.

      Section 166,000 avoided deaths:  “The total impact of more than 116,000 more heat deaths each year and almost 283,000 fewer cold deaths year is that by now, the temperature rise since 2000 means that for temperature-related mortality we are seeing 166,000 fewer deaths each year.” ↩ ↩

  7. 7.

    Climate Feedback (Science Feedback).  Global warming contributes to increased heat-related mortality, contrary to Bjorn Lomborg’s unsupported claims that climate change is saving hundreds of thousands of lives each year.  Sep 4, 2021.  A fact-check of Lomborg's claim by a site that recruits scientists to review climate statements, with comments written for it by Antonio Gasparrini, Yuming Guo and Kristie Ebi.  The comments are statements to the reviewers, not published papers.

    1. 7a.

      Section Yuming Guo:  “Cold-related mortality decreased and heat-related mortality increased from 2000 to 2019, causing a net decrease in total deaths. However, it is not correct to interpret that this net decrease was caused by climate change. We just estimated the trend of mortality burden related to non-optimal temperatures, but did not do further analysis to examine whether this change is due to climate change or other factors.”  Guo, of Monash University, wrote on behalf of the research network that produced the study. ↩ ↩

    2. 7b.

      Section Kristie Ebi:  “Because temperatures during summer have a strong, positive association with mortality, it is common, but incorrect, to assume a similar (but opposite) association occurs in the winter”. ↩

  8. 8.

    Cissé, Guéladio et al.  “Health, Wellbeing, and the Changing Structure of Communities.”  Climate Change 2022:  Impacts, Adaptation and Vulnerability.  Contribution of Working Group II to the Sixth Assessment Report of the IPCC, 1041–1170.  Cambridge University Press.  2022.  The IPCC's chapter on climate and health.  Unlike the summary for policymakers, it discusses cold-related deaths.

    1. 8a.

      Section 7.2.4.1, page 1074:  “Although there has been an observed increase in winter-season temperatures for a number of regions, to date there is variable evidence for a consequential reduction in winter mortality and susceptibility to cold over time due to milder winters; some countries demonstrate decreasing trends, while other countries show stable or even increasing trends in cold-attributable mortality fractions over time”.  Section 7.3.1.2, page 1092:  “strong regional contrasts in heat- and cold-related mortality trends are likely under a RCP8.5 scenario, with countries in the Global North experiencing minimal to moderate decreases in cold-related mortality while warm climate countries in the Global South are projected to experience increases in heat-attributable deaths by the end of the century”. ↩ ↩ ↩

  9. 9.

    Vicedo-Cabrera, Ana M. et al.  “The burden of heat-related mortality attributable to recent human-induced climate change.”  Nature Climate Change 11:492–500.  2021.  A study of warm-season deaths in 732 places in 43 countries, 1991–2018, comparing heat deaths under the actual climate with those under a modelled climate without human influence.  It measures the heat cost of warming so far and does not count cold deaths.

    1. 9a.

      Section Results:  “The overall estimate is 37.0% but this percentage varied widely across subregions and countries. The largest climate change-induced contributions (>50%) were in southern and western Asia (Iran and Kuwait), southeast Asia (Philippines and Thailand) and several countries in Central and South America”. ↩ ↩

    2. 9b.

      Section Discussion:  “we were not able to include locations in all world regions—for example, large parts of Africa and South Asia—due to a lack of the empirical data needed to estimate the exposure-response functions.” ↩ ↩

  10. 10.

    Gasparrini, Antonio et al.  “Projections of temperature-related excess mortality under climate change scenarios.”  The Lancet Planetary Health 1(9):e360–e367.  2017.  Death-temperature curves from 451 places in 23 countries applied to five climate models under four emissions scenarios, region by region, to 2099, assuming no adaptation and no population change.  It is the standard test of whether fewer cold deaths offset more heat deaths.

    1. 10a.

      Page e360:  “Results indicate, on average, a net increase in temperature-related excess mortality under high-emission scenarios, although with important geographical differences.”  Page e362:  “under the assumption of no adaptation or population changes”. ↩ ↩ ↩

    2. 10b.

      Section Appendix, Table S1:  Net change in temperature-related excess deaths, as a percentage of all deaths, 2090–99 against 2010–19, under RCP4.5:  North America −0.0;  Central America −0.5;  South America +0.7;  northern Europe −0.8;  central Europe +0.6;  southern Europe +1.6;  East Asia −0.6;  Southeast Asia +2.3;  Australia −1.0.  Under RCP8.5:  +1.4, +3.0, +4.6, −0.6, +3.5, +6.4, −0.1, +12.7 and −1.2. ↩

  11. 11.

    Masselot, Pierre et al.  “Estimating future heat-related and cold-related mortality under climate change, demographic and adaptation scenarios in 854 European cities.”  Nature Medicine 31:1294–1302.  2025.  Projections for 854 European cities to 2099 under three emissions pathways and 19 climate models, separating the effect of climate from that of an ageing population, with tests of how much heat adaptation would change the result.

    1. 11a.

      Page 1294:  “We showed that, with no adaptation to heat, the increase in heat-related deaths consistently exceeds any decrease in cold-related deaths across all considered scenarios in Europe.”  Page 1296:  “With no adaptation to heat, only in Northern Europe the decrease in cold-related deaths slightly offsets the increase in heat-related deaths”. ↩ ↩ ↩

  12. 12.

    Carleton, Tamma et al.  “Valuing the Global Mortality Consequences of Climate Change Accounting for Adaptation Costs and Benefits.”  NBER Working Paper 27599.  National Bureau of Economic Research.  2022.  Published in the Quarterly Journal of Economics 137(4):2037–2105 (2022);  page numbers here are those of the revised working paper.  Age-specific death-temperature curves from subnational data in 40 countries, allowed to flatten as places grow richer and as they adapt to their own climates, projected worldwide to 2100.  It counts both the deaths and the cost of the adaptation that prevents them.

    1. 12a.

      Section Abstract:  “Notably, today’s cold locations are projected to benefit, while today’s poor and hot locations have large projected damages.” ↩

    2. 12b.

      Page 22:  “The mean estimate of the mortality effects of climate change falls from 73 deaths per 100,000 under RCP8.5 to 11 deaths per 100,000 under the emissions stabilization scenario of RCP4.5. For RCP4.5, the median end-of-century estimate is 4, and the 10th to 90th percentile range is [-36, 62].” ↩ ↩ ↩

    3. 12c.

      Page 2:  “Perhaps most importantly, a failure to account for climate adaptation and the benefits of income growth would lead to overstating the mortality effects of climate change by a factor of about 3.” ↩

  13. 13.

    Masselot, Pierre et al.  “Excess mortality attributed to heat and cold:  a health impact assessment study in 854 cities in Europe.”  The Lancet Planetary Health 7(4):e271–e281.  2023.  A study of adults in 854 European urban areas, 2000–2019, estimating yearly deaths from cold and heat.  A coding error in the baseline death rates was corrected in July 2024, which cut both totals by about a third and left their ratio almost unchanged;  the numbers here are the corrected ones.

    1. 13a.

      Page e271:  “Across the 854 urban areas in Europe, we estimated an annual excess of 130 228 (empirical 95% CI 115 893–143 929) deaths attributed to cold and 13 589 (11 530–15 475)) attributed to heat.” ↩ ↩

  14. 14.

    IPCC.  “Summary for Policymakers.”  Climate Change 2021:  The Physical Science Basis.  Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, 3–32.  Cambridge University Press.  2021.  The summary of the IPCC's 2021 assessment of the physical science, line by line approved by member governments.  It is the mainstream statement of what has been observed and what is projected.

    1. 14a.

      Section B.1.1, page 14:  “Compared to 1850–1900, global surface temperature averaged over 2081–2100 is very likely to be higher by 1.0°C to 1.8°C under the very low GHG emissions scenario considered (SSP1-1.9), by 2.1°C to 3.5°C in the intermediate GHG emissions scenario (SSP2-4.5) and by 3.3°C to 5.7°C under the very high GHG emissions scenario (SSP5-8.5).” ↩

  15. 15.

    Scale only, not a figure Carleton's team prints as a count:  11 extra deaths per 100,000 people a year 12 in a population of 9 billion is 11 × 90,000 ≈ 990,000 a year;  their median of 4 per 100,000 is about 360,000.  Today's modelled net saving of about 162,000 a year 5 among 8 billion people is about 2 per 100,000.  Carleton's figure includes the benefit of adaptation and of rising incomes. ↩