--- title: "What's Inside 90 Percent" canonical: "https://anchor.is/docket/issue/52/lab/article/40733/whats-inside-90-percent" publication: "The Docket" publisher: "The Docket" language: "en-US" guide: "https://anchor.is/docket/llms.txt" format: "Disaggregating Impact" author: "Seth D. Carrington" author_url: "https://anchor.is/docket/team/30" issue: 52 section: "lab" section_name: "The Lab" published: 2026-10-02 updated: 2026-10-03 description: "The survey says 90 percent of Americans are affected by heat. Hospital records, school closings, and utility data show how many different things that single number is covering up." word_count: 2292 source_links: 14 --- # What's Inside 90 Percent *By [Seth D. Carrington]()* *October 2, 2026* [Read on the website]() AP-NORC [surveyed 1,165 adults](https://apnorc.org/projects/most-americans-have-been-affected-by-extreme-heat-this-year/) in late July 2026 and found that 90 percent said extreme heat had affected at least one area of their daily lives. Two years ago the figure was 83 percent. Since then it's been getting quoted the way round, alarming numbers get quoted, dropped into policy briefs and news ledes as though it settles something. It settles very little. It gathers experiences of entirely different kinds and consequence into one pile and asks the word "affected" to do about nine jobs at once, most of them badly. Eight areas of life were on the questionnaire. Here's how they came out: | Domain | Major impact | Minor impact | Combined | |---|---:|---:|---:| | Electricity bills | 46% | 35% | 81% | | Outdoor activities | 30% | 38% | 68% | | Exercise routine | 21% | 30% | 51% | | Sleep | 17% | 33% | 50% | | Travel or vacation plans | 15% | 27% | 42% | | Pets | 15% | 27% | 42% | | Job or commute | 12% | 21% | 33% | | Weddings, reunions, similar events | 12% | 19% | 31% | Respondents could answer major impact, minor impact, no impact, or not applicable. Reschedule a jog and you're in the 90 percent. Land in an emergency department with heatstroke and you're probably also in the 90 percent, though the survey [doesn't ask about symptoms or medical care](https://apnorc.org/wp-content/uploads/2026/08/July-2026-Topline-Final.pdf), so you'd have to decide which of the eight boxes your heatstroke belonged in. Two things about the table before we move on. The denominators include everybody. Twenty-two percent marked "job or commute" not applicable: retirees, people home with kids, the unemployed. So the 33 percent reporting a job effect is figured against a population that includes people without jobs. And AP-NORC collected income, age, race, and housing tenure, then published none of it crossed against the heat questions. Whether renters had a different summer from owners, or households at $25,000 from households at $150,000, is something this survey could answer. It hasn't. What follows puts the survey's self-reports next to institutional records that come at heat from the other direction: clinical cases, dollars, lost school days, workplace injuries. Different agencies built them for different purposes, covering different populations over different stretches of time. They don't line up with the survey or with each other. The mismatches are the useful part. ## The bill and what it doesn't say Eighty-one percent said heat affected their electricity bills, the survey's biggest number and its emptiest. "Impact on electricity bills" contains no dollars, no income, and no way to tell noticing a bigger figure on the statement from deciding which other bill goes unpaid this month. The questionnaire never asks whether you have air conditioning, whether it works, or whether you can afford to run it. A household that paid an extra $18 without thinking about it and a household that sat at 85 degrees because 75 would have broken the budget both show up as "major impact" if that's the box they checked. EIA's [monthly data](https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_5_03) put the national average residential price at 18.34 cents per kilowatt-hour in June 2026, up about 12 percent from two years earlier. The agency's [September forecast](https://www.eia.gov/outlooks/steo/tables/pdf/sf02.pdf) projects the average summer residential bill at $196 a month against $178 in 2024. That $18 is a national average. It runs from $173 a month in the West North Central states to $226 in the West South Central, and it tells you nothing about how the increase landed inside any given city, or what share of a household's income is going to cooling. A $196 bill on $120,000 a year is a line item. The same bill on $24,000 is a decision about what else doesn't get paid. I wrote earlier this year about Houston's cooling ordinance, which [requires landlords to provide the equipment](https://anchor.is/docket/issue/48/lab/article/37324) and says nothing about who pays for the electricity to run it. The landlord buys the box, the tenant feeds it. Same gap here, one level up. The survey and the price data agree that bills went up. Neither can tell you who was inconvenienced and who was in trouble. What the increase costs measured against what the household can absorb is the number that would matter to anyone writing a program, and nothing in wide use measures it. ## Who ends up in the emergency department The survey's nearest approach to health is sleep and exercise. Routine, not medicine. It never asks whether you had heat exhaustion, heat cramps, heatstroke, or saw anybody about it. CDC watches the far end of the range, the people sick enough to generate an emergency department record. Its [September report](https://www.cdc.gov/mmwr/volumes/75/wr/mm7536a2.htm) counted 145,321 heat-related ED visits in 2026 through September 10. During the late-June and early-July heat dome, when roughly 221 million people across 42 states got at least one day of major or extreme heat risk, 21,775 heat-related visits turned up across seven of the federal health regions. The demographic finding is specific. Men aged 26 to 54 had the highest heat-related visit rates in six of those seven regions, and several regions hit their highest daily rates since 2018 around the July 4 weekend. The survey, for its part, found adults in the West reporting the steepest increases in major effects on exercise, outdoor activities, and job or commute. Whether those are the same people is unknowable. The survey doesn't ask about medical events and the surveillance doesn't ask about disrupted lives. One asks people how the summer went. The other counts the ones who ended up on a gurney. In between is everybody who felt bad and handled it at home, or never connected the dizziness to the weather, or couldn't afford to find out. The survey might catch some of them under "sleep" or "exercise." It might not. It wasn't built to look. Even the hospital picture is contested on its own ground. CDC's count captures visits coded as direct heat illness: heatstroke, exhaustion, cramps. But heat also drives cardiovascular, kidney, and respiratory trouble that gets coded as cardiovascular, kidney, and respiratory trouble. Some researchers argue the direct-heat numbers badly understate the clinical burden. Others argue that once you start attributing kidney failure to the weather, attribution stops meaning much. Either way, the boundary around "heat-related visit" is drawn where diagnostic convention put it, which is not necessarily where the body put it. ## The school days that don't exist in the data There's no question on the survey about school closings, early dismissals, child care, or lost classroom time. A mother who left work to collect a kid from a building nobody had cooled would have to file that under "job or commute," assuming it occurred to her. In Philadelphia, the School District [sent 52 schools home early](https://www.inquirer.com/education/philadelphia-school-district-early-dismissal-heat-20260831.html) in the opening days of the school year because those buildings lacked adequate cooling and the forecast cleared the district's 85-degree threshold. The [same 52](https://www.philasd.org/blog/2026/05/19/virtualmay20/) dismissed early again on September 3. Early-childhood programs shortened their days, after-school activities were canceled, roughly 28,000 students affected. In a [previous piece](https://anchor.is/docket/issue/50/lab/article/38889) for this publication I ran the 2026 list against earlier years. Fifty-one of the 52 were on the heat-dismissal list in 2023 and again in 2024. The buildings are the same buildings. Nobody cooled them. That's 104 building-days of shortened instruction in the first week of school, sitting in district records and local press coverage, and the national survey can't detect a minute of it. There's no category for it. The survey has no place for a parent at one of those 52 schools, yet a retiree in Scottsdale nudging her thermostat feeds the same 90 percent through the electricity-bill question alone. No federal source tracks heat-related school dismissals. It happens district by district, gets written down in building notices, and nobody adds it up. ## Work, as a survey checkbox Thirty-three percent of all adults reported some effect on their job or commute. That item makes no distinction between a longer walk across a parking lot, a schedule change handed down by a supervisor, a shift missed for illness, wages lost, or an injury. It folds indoor and outdoor, remote and on-site, salaried and hourly into one percentage calculated against a population that includes people who don't work. The occupational records are measuring something much harder to reach. BLS [counted 48 fatal workplace heat injuries](https://www.bls.gov/iif/fatal-injuries-tables/fatal-occupational-injuries-table-a-9-2024.htm) in 2024, 40 of them outdoors and 6 indoors, 24 in construction. For nonfatal cases the most recent detailed data combine 2023 and 2024: about [7,100 cases](https://injuryfacts.nsc.org/work/safety-topics/exposure-to-environmental-heat/) involving days away from work, job restriction, or transfer. OSHA's [2026 National Emphasis Program](https://www.osha.gov/sites/default/files/enforcement/directives/CPL_03-00-024_0.pdf) says in its own text that its statistics "likely understate the magnitude of heat-related injury, illness, and death." The agency figures its proposed federal heat standard would cover roughly 36 million workers, which is not a count of people harmed but a count of people the government thinks are exposed enough to need a rule. Between the tens of millions the survey has reporting some job effect and the 7,100 recorded nonfatal cases over two years lies an enormous population nobody has numbers for. Men and women who slowed down, stretched their breaks, got dizzy on a ladder, went home wrung out and useless. The survey counts them if they thought to mention it. The injury records count them only if the event was bad enough to be recognized, reported, and classified. Most heat exposure at work happens between those two thresholds. Economists have tried to model the resulting productivity losses across the whole workforce and come back with figures that don't resemble each other, depending on what they assume about acclimatization, work intensity, and what counts as lost output. The spread is informative in itself: the space between the two instruments is real, and nobody has a defensible method for measuring what's in it. ## Side by side None of these records was built to talk to the others. The survey takes a reading of perception across a season. Emergency department data track acute cases almost as they happen; the electricity figures are prices and monthly averages. School records are particular buildings on particular days, and the occupational numbers arrive a year or two late, counting only what got classified. Line them up anyway and the same hole shows up in each. The survey can tell you how many people noticed heat in their routines and nothing about how badly any of them were hurt. The institutional records can tell you who needed an ambulance, which schools shut early, and who died on a job site, while missing the much larger number who never crossed an administrative threshold. Neither goes anywhere near what a given disruption cost a particular household in money, time, health, or opportunity, set against what that household could stand to lose. The survey recorded income, age, race, housing tenure, and employment status, then published none of it crossed against a single heat item. The numbers that would show whether "affected by heat" means something different depending on who you are sit in a dataset nobody has opened in public. ## What we still can't answer What share of heat's damage to people's health never reaches an emergency department? CDC counts visits coded as direct heat illness. The survey doesn't ask about symptoms. Everyone between those two instruments, managing it at home or not recognizing it for what it is, stays invisible, and as the attribution argument among researchers shows, even the boundary of what counts as a heat-related visit is unsettled. How many districts run heat-dismissal protocols like Philadelphia's, and how many instructional days go out of the national total every September? Nobody is keeping that score. When heat at work produces not a recordable injury but shortened hours, lost output, or a man who is simply used up by August, where does that register? Not in the BLS counts, and not reliably in a survey item that lumps it with a hot commute. The 36 million workers OSHA would bring under a standard are living somewhere between "affected" and "recorded case," and no measurement system we have describes that place. The 90 percent will keep getting cited through next summer and the one after, as evidence that heat has become a universal American experience. In the narrowest sense it has. On every question that determines what actually happens to anybody — who was harmed, how badly, at what cost, with what recourse — the figure has nothing to say, and was never designed to. That part isn't AP-NORC's fault. The harder thing to explain is why, having collected the information that would answer some of it, nobody has put it on the table. ## Things to follow up on... - **CDC tracker maintenance window:** CDC's [Heat and Health Tracker](https://ephtracking.cdc.gov/Applications/heatTracker/) was down for planned maintenance at the research cutoff, which means weekly regional ED visit rates that would add geographic detail to the MMWR's published findings were temporarily unavailable. - **Summer bills still incomplete:** EIA's actual residential electricity data run only through July 2026, with [August figures scheduled for October 23](https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_5_03), so the full-summer cost picture currently depends on a forecast rather than recorded prices. - **Occupational data lag widens:** BLS's newest detailed nonfatal heat-illness estimates [combine 2023 and 2024](https://injuryfacts.nsc.org/work/safety-topics/exposure-to-environmental-heat/), and no complete 2025 or 2026 federal citation count has been published, meaning the occupational record falls further behind the survey with each passing summer. - **Philadelphia's affected enrollment unknown:** The district's latest published school-level enrollment file covers [October 2025](https://www.philasd.org/research/), and the next update isn't expected until November 2026, so the exact number of students dismissed early in September remains an estimate rather than a count. ## More - [The Lab — Issue 52]() - [Issue 52]() - [Publication guide]()