Two days ago, the New York City Health Department released its annual heat mortality report. The headline number: approximately 500 New Yorkers die from heat every year. Every year, as a baseline condition of summer in the city.
Most people carry a mental picture of heat death. It involves a construction site or a marathon, visible exertion under a punishing sun, a body overwhelmed by obviously extreme conditions. The picture is outdoors, public, acute. According to the data, it accounts for a small fraction of actual heat deaths.
The data describes something different: a person alone in a room without cooling, on a day that feels like ordinary summer, in a neighborhood that was made hotter than the one three miles away by decisions made decades before they were born. The death certificate, if one is issued, will probably say cardiac arrest or respiratory failure. Heat won't appear anywhere in the record.
Ordinary summer
The NYC Health Department counts heat deaths two ways. The narrow count tracks heat-stress deaths where a medical examiner identifies heat as the direct cause: roughly five per year. The broad count uses statistical modeling to estimate heat-exacerbated mortality, capturing people whose heart disease, kidney condition, or respiratory illness became fatal because of temperatures that wouldn't have killed them otherwise. That model begins accumulating excess deaths above 82°F, the city's median warm-season temperature. It's the broad count that produces the 500 figure.
And it's the broad count that reveals where those deaths actually cluster. Eighty percent fall on days between 82 and 94 degrees. Days that feel like summer. Days that don't make the news.
The National Weather Service issues heat advisories for New York City when the heat index reaches 95 to 100°F. That's when cooling centers open, public health messaging goes out, welfare checks begin. About 100 of the city's annual heat deaths occur on those advisory days. The other 400 happen when none of that infrastructure activates. Four hundred deaths a year falling into the space between what triggers an official response and what actually kills people.
Of New York City's roughly 500 annual heat deaths, 400 occur on days when no heat advisory is issued and no cooling centers open.
The advisory system does what it was designed to do: detect extreme heat events. Heat death in New York City is mostly cumulative, though, built up over ordinary warm days. And the window of cumulative exposure keeps widening. In the 1960s, heat wave season in major U.S. cities lasted about 24 days. By the 2020s, it had stretched to 73. Cities that averaged two heat waves per year now average more than six. The season during which 82°F days accumulate has tripled in length within a lifetime.
The room
Among New Yorkers who died directly from heat stress and whose records contained detailed information about location, 45% were at home. Among those who died at home and for whom air conditioning status was documented, none had a working AC in use. The sample is small: fifteen records with that level of detail. But the pattern holds across the broader data on heat-exacerbated deaths, which also occur disproportionately at home.
The NYC Comptroller's office found that among people who died at home during extreme heat, 42% had air conditioners that weren't working or weren't running, while 58% had no unit at all. More than 90% of New York City households have air conditioning, so the deaths concentrate in a specific slice of the population. Cost is the primary reason people don't run their units.
What makes an uncooled room dangerous is physics that most people don't think about. Without ventilation or mechanical cooling, indoor temperatures can exceed outdoor temperatures, particularly in older buildings with poor insulation and dark roofing materials that absorb and radiate heat. At night, when the body needs to recover from daytime thermal stress, an uncooled room holds the day's heat for hours after the air outside has cooled. The person inside never gets the physiological reset that survival requires.
The Health Department's report states it plainly: lack of access to home air conditioning is the most important risk factor for heat-stress death. Eighty-nine percent of those who died had at least one chronic health condition. Cardiovascular disease, substance use, mental health conditions, lung disease, diabetes. Millions of people manage these conditions successfully through an ordinary summer. The difference between managing them and dying from them, in the data, comes down to the room.
The map
Black New Yorkers die from heat stress at twice the rate of white New Yorkers: 2.3 deaths per million versus 1.0. The June 2025 heat wave, which killed 19 people from heat stress alone, disproportionately affected Black and Latino New Yorkers. Death rates were highest in Brooklyn and the Bronx, in neighborhoods with the most residents living below the federal poverty line.
This pattern scales nationally. In March, The Lancet Planetary Health published a Harvard study tracking 73.7 million Medicare beneficiaries across 27,926 ZIP codes over eighteen years. Each additional heat wave was associated with 8.83 excess deaths per 10,000 person-years. For Black individuals, the figure was 16.50. For white individuals, 5.85. Nearly three times the mortality burden per heat wave. The researchers noted this was the first study to establish a significant association between heat waves and mortality for Black individuals across the contiguous United States.
The study also found that neighborhoods with abundant green space showed a protective effect of 13.51 fewer deaths per 10,000 person-years per heat wave, though the confidence interval was wide (-24.92 to -1.79), meaning the protection is real but its precise magnitude uncertain. High-poverty neighborhoods showed significantly stronger associations between heat waves and death.
The researchers named the mechanism: redlining. The racially discriminatory housing policies of the 1930s through 1960s concentrated Black residents in neighborhoods that received less investment in tree canopy, parks, and infrastructure. Those neighborhoods are now measurably hotter. A peer-reviewed study of 108 U.S. urban areas found that formerly redlined neighborhoods average 4.5°F warmer than non-redlined neighborhoods in the same city. Ninety-four percent of the cities studied showed this pattern. A separate analysis by the National Community Reinvestment Coalition, a fair-lending advocacy organization, found that neighborhoods where lending was encouraged have roughly twice the tree cover of neighborhoods that were starved of investment, a finding consistent with the academic research.
The temperature in your neighborhood on a July afternoon is a consequence of where capital went and where it didn't, where trees were planted and where they weren't. Two days before Juneteenth, the data draws a line from the 1930s to this Thursday's forecast. The disparity in who dies from heat and the disparity in neighborhood temperature are the same disparity, measured at different points along a chain that determines, with striking precision, who is at risk right now.
Black households in cities like Chicago, Detroit, Minneapolis, and Pittsburgh are far less likely to have central air conditioning than white households. The conditions that make heat fatal, the cooling that prevents heat death, and the neighborhood temperatures that determine exposure all follow the same map.
What the count misses
New York City's methodology is unusually rigorous. Most cities don't do what NYC does. Which raises the question of what's happening in places that count differently, or don't count at all.
In Miami-Dade County, researcher Chris Uejio at Florida State University produced two heat death estimates for the same jurisdiction using different methods. The official count, based on death certificate coding: 34 per year. His excess mortality analysis, comparing actual deaths during hot periods to expected baselines: roughly 600, as reported by Scientific American.
That gap reflects how death certificates work. A medical examiner can classify a death as heat-related only if the deceased person's core body temperature is elevated, which rules out anyone whose body was found hours or days later. If someone with heart disease dies of cardiac arrest during a heat event, the certificate says cardiac arrest. Heat disappears from the record. The 1995 Chicago heat wave killed an estimated 700 people by excess mortality analysis, but only 465 appeared on death certificates as heat-related.
There is no nationwide standard for reporting heat deaths. Bexar County, Texas, which includes San Antonio, doesn't track heat deaths at all. The CDC's current national estimate is approximately 1,220 heat deaths per year. Excess mortality approaches produce numbers roughly double that. Some researchers argue the true figure could be far higher still, but the disagreement hinges on a specific methodological question: how you define the baseline of "expected" deaths during warm months and how confidently you can attribute excess deaths to heat rather than other seasonal factors. The wider you draw the attribution window, the higher the count. Where you draw it is a judgment call that different epidemiologists make differently, and the range of estimates reflects that judgment, not just different data.
What gets counted determines what gets planned for. If a city's official heat death toll is 34, the political case for expanding cooling infrastructure looks different than if the number is 600. The infrastructure of response is calibrated to the infrastructure of counting, and both are calibrated to a version of heat death that doesn't match the actual pattern.
The longer shadow
The Harvard study captured something most heat research misses by tracking mortality over an entire year following a heat wave. Heat damages kidneys, stresses cardiovascular systems, accelerates neurological decline. The study found significant associations between heat waves and deaths from heart attacks, Alzheimer's complications, pneumonia, and COPD well beyond the acute period. Some of those deaths show up weeks or months later, long after the temperature drops and the event is declared over.
This finding dismantles the emergency model of heat death more completely than anything else in the data. An emergency has a beginning and an end. A heat wave's mortality extends weeks, sometimes months, past the temperature drop. The damage accumulates in tissue, in organ function, in the slow deterioration of systems already under strain. By the time the death occurs, the heat wave that contributed may be weeks in the past. No surveillance system captures this. No death certificate connects a September cardiac arrest to an August heat wave. The Harvard study could detect the pattern only because it followed 73.7 million people for eighteen years and looked at annual mortality rather than daily spikes.
What the full picture shows
Heat kills mostly on days that feel like normal summer, mostly inside homes without adequate cooling, mostly among people who are older and who live in neighborhoods made hotter by decades of disinvestment. It kills through existing conditions that heat accelerates, which means the deaths often don't get recorded as heat deaths, which means the official count stays low, which means the response stays calibrated to a problem smaller than the actual one. And the deaths that happen weeks later, from accumulated damage, don't register in any count at all.
None of this is hidden. New York City counts it. Harvard researchers measured it across 28,000 ZIP codes. The temperature difference between redlined and non-redlined neighborhoods has been documented in 108 cities. The information exists. Most places simply lack the counting infrastructure that would make the full picture visible to the systems supposed to respond.
The data also reframes something closer to home than municipal counting systems. Deaths accumulate starting at 82°F, so the adequacy of your cooling is tested every warm day from June through September. Forty-five percent of direct heat deaths happen at home, and none of those people were using air conditioning. The room matters more than the forecast. If you know someone older who lives alone, someone managing a heart condition or diabetes, the data says the dangerous day is the ordinary Tuesday in July when nobody thinks to call.
The NYC Health Department report landed on a Sunday. By Thursday, temperatures in the city are forecast to reach the upper 80s. That's the range where the broad mortality model starts accumulating deaths. No advisory will be issued. No cooling center will open. In the Bronx or Brooklyn, in a building that holds the day's heat long after dark, someone will be deciding whether running the AC is worth what it costs. Someone else, who doesn't have that option, won't be deciding anything.
Both of those people are already in the data. Whether you know who they are, in your building, on your block, is something the numbers can't tell you.
- The cooling program gap: New York's federally funded HEAP Cooling Assistance benefit closed on June 5 this year, earlier than any previous year, leaving thousands of low-income residents unable to access free air conditioners before summer officially began.
- Heat waves arriving early: The March 2026 Southwest heat wave, which pushed Phoenix past 106°F, was deemed virtually impossible without human-caused climate change by World Weather Attribution, raising questions about whether heat risk calendars and cooling program timelines are built for a season that no longer exists.
- The national pessimism shift: A Pew Research Center survey from March 2026 found that 62% of Americans now say countries won't avoid the worst effects of climate change, a nine-point jump that includes a sharp increase among people who already accepted the science.
- Who bears the insurance cost: A U.S. Treasury analysis of 246 million homeowners insurance policies found that consumers in the highest-risk ZIP codes faced nonrenewal rates 80% higher than those in the lowest-risk areas, with 60% of Black homeowners facing extreme wind risk compared to 32% of white homeowners.