Research
The cooling divide: US homes with no air conditioning
Who the American households with no air conditioning are, and whether it is because they do not need it.
Direct answer
11.3% of US households (14.0 million homes) have no air conditioning of any kind, from our analysis of EIA’s RECS 2020 survey.
Climate explains only part of it: 1.86 million live where summers are hotter than Atlanta’s. The rate falls as income rises, from 18.7% under $20,000 to 7.1% at $150,000 or more. Renters go without at about twice the rate of owners.
How many US homes have no air conditioning?
14.0 million households. EIA’s headline is quoted constantly: 88.7% of American homes use air conditioning. It is a real figure and it is the right one. It is also the least interesting thing the survey says, because it flattens the 14.0 million households on the other side of it into a rounding error.
Those households are not distributed the way the summary implies. The share with no cooling ranges from 93.1% in Alaska to 3.5% in New Jersey, and the third-worst jurisdiction in the country is not a cold one.
43.0% of Hawaii’s households have no air conditioning, in a state with one of the three highest cooling loads in the country, at 4,154 cooling degree days a year.
HyreHVAC analysis: that pairing is the finding. A state can have one of the hottest cooling seasons in the country and the third-highest rate of homes without cooling at the same time, which means "they do not need it" cannot be the general explanation. Below, we test that explanation directly and it does not survive.
Who goes without air conditioning, in four numbers?
Is it because those homes are in cool climates?
Partly, but not mostly. The obvious reading is that homes without air conditioning are in places that do not get hot. The survey lets that be checked rather than assumed, because every record carries the cooling degree days for its location.
Climate does explain part of it. Households without cooling average 1,137 cooling degree days a year against 1,765 for households with it. That is a real difference, in the expected direction.
It does not explain most of it. Among households in truly hot places (more than 2,000 cooling degree days, hotter than Atlanta), 5.6% still have no cooling equipment.
That is 1,862,684 households, and it is the number this study exists to put on the record.
HyreHVAC analysis: climate sets the floor for how many homes can plausibly go without cooling; it does not account for who actually does.
Does income predict which homes have no air conditioning?
Yes, strongly. Cut the same households by income and the pattern is steady across all six bands: every step up the income ladder lowers the share of homes with no cooling. 18.7% of households earning under $20,000 have none, against 7.1% of households earning $150,000 or more.
The two ends are 2.6 times apart, far wider than the survey’s uncertainty. The standard errors are 0.83 and 0.53 percentage points respectively, so the confidence intervals do not come close to overlapping. This is a real gradient, not survey noise.
How does the share without cooling vary by group?
| Group | Households | No cooling | Standard error | Central AC |
|---|---|---|---|---|
| Under $20,000 | 18,734,082 | 18.7% | ± 0.83 | 47.4% |
| $20,000-$39,999 | 24,072,800 | 12.9% | ± 0.68 | 61.9% |
| $40,000-$59,999 | 19,607,254 | 11.6% | ± 0.71 | 67.7% |
| $60,000-$99,999 | 27,711,704 | 9.9% | ± 0.52 | 72.5% |
| $100,000-$149,999 | 16,541,269 | 7.3% | ± 0.60 | 76.9% |
| $150,000 or more | 16,861,915 | 7.1% | ± 0.53 | 78.9% |
| Owned | 82,920,129 | 8.8% | ± 0.27 | 74.7% |
| Rented | 40,608,896 | 16.5% | ± 0.60 | 52.2% |
| Mobile home | 6,832,499 | 11.2% | ± 1.06 | 58.0% |
| Single-family detached | 77,067,692 | 9.5% | ± 0.32 | 75.3% |
| Single-family attached | 7,451,177 | 11.8% | ± 0.87 | 69.4% |
| Apartment, 2-4 units | 9,341,795 | 19.0% | ± 1.28 | 41.0% |
| Apartment, 5 or more units | 22,835,862 | 14.3% | ± 0.74 | 53.3% |
| Built before 1950 | 20,256,553 | 16.7% | ± 0.81 | 43.3% |
| Built 1950-1969 | 25,235,020 | 11.9% | ± 0.61 | 61.7% |
| Built 1970-1989 | 34,637,134 | 12.1% | ± 0.57 | 67.5% |
| Built 1990-2009 | 33,316,668 | 8.3% | ± 0.47 | 80.7% |
| Built 2010-2020 | 10,083,651 | 6.9% | ± 0.69 | 84.8% |
Share of each group with no cooling equipment of any kind, 2020. Standard errors are computed from the 60 replicate weights supplied with the public use file.
Groups are shares within themselves, so each is comparable to the national rate of 11.3% rather than to the size of the other groups.
Are renters and older buildings more likely to go without?
Yes. Renters go without cooling at 16.5% against 8.8% for owners: close to twice the rate, and well outside the survey’s error bars. That distinction matters more than it looks, because a renter cannot install central equipment. The decision belongs to somebody who does not live in the building.
Building type follows the same logic. The worst-served housing in America is the small apartment building: 19.0% of households in buildings of two to four units have no cooling, against 9.5% of single-family detached homes.
Large apartment buildings sit between the two at 14.3%. HyreHVAC analysis: the most likely explanation is that small multifamily buildings are old, are rarely gut-renovated, and have no professional management to drive a capital upgrade.
The survey cannot separate those three from each other, and we will not pretend it can.
Age of the building runs the same way, from 16.7% of homes built before 1950 down to 6.9% of homes built since 2010. Cooling is largely something that arrived with the building.
Adding it to a home with no ducts is a different job from a replacement; the AC sizing estimator is the place to start.
What does going without cooling cost households?
More hardship, on the survey’s own questions. RECS asks whether a household went without food or medicine to pay an energy bill, and whether it kept its home at a temperature it felt was unhealthy. Both questions are asked of everyone, so they can be read against the cooling variables directly.
25.7% of households with no cooling reported forgoing food or medicine to pay an energy bill, against 19.2% of households with cooling. 11.8% kept their home at a temperature they considered unhealthy, against 9.8%.
HyreHVAC analysis: read this as association, not as cause. Households without cooling are poorer, and poverty drives both answers.
The finding worth taking away is narrower and still useful: the households most likely to lack cooling are the same households least able to absorb the cost of fixing it, which is exactly the group a market solution reaches last.
Income-qualified heat pump rebates were written for this group, though few states have opened them.
One cell we will not publish. RECS also asks whether anyone in the household needs cooling for a medical reason.
Among households with no cooling the estimate is 0.8%, but its relative standard error is 32%, above EIA’s own 30% suppression threshold.
The survey cannot support a claim about medically vulnerable households without cooling, so we are not making one, and we are saying so here rather than leaving the number out quietly.
How did we measure it?
EIA Residential Energy Consumption Survey 2020, public use microdata file v7. 18,496 responding households, weighted to represent 123.53 million occupied primary housing units. Downloaded September 4, 2026.
The RECS variable AIRCOND, which is asked of every household and covers cooling equipment of any kind: central systems, heat pumps used for cooling, ductless mini-splits, and window or portable units. A household with a working window unit is counted as having cooling. Evaporative coolers are included. Fans are not.
Computed from the 60 replicate weights shipped with the file, as the plain sum of squared replicate deviations. This was determined empirically rather than assumed: applying a Fay coefficient of 0.5, the natural reading and the convention in several other federal surveys, understates every standard error by a factor of about four.
The same machinery was run against EIA’s published "State Air Conditioning" table for all 51 jurisdictions, for both the share estimates and their relative standard errors. Maximum difference: 0.50 percentage points on the shares, which is the resolution at which EIA publishes them, and 0.18 percentage points on the RSEs. Every jurisdiction reproduces.
Cooling degree days are the CDD65 field on each record, which RECS derives from the household’s actual location, not from a state average. The "hot climate" cut is 2,000 CDD or more, chosen because it sits above every major northern metro and below the Gulf Coast, not because it produced a better number.
Any estimate with a relative standard error above 30% is reported as "too few respondents" rather than printed. That is EIA’s own rule, applied to our derived cells rather than only to theirs. One cell in this study hits it, and it is named in the text above rather than dropped.
Every figure in the prose, the tables and the charts on this page is read from one dataset built by script. A number in a sentence and the same number in the table below it cannot disagree.
What are the limits of this study?
RECS runs roughly every five years. These are the most recent household-level data that exist at this resolution, but they predate the 2021-2024 heat seasons and the Inflation Reduction Act. Where cooling has been installed since, this study will overstate the gap.
The survey records whether a household has cooling equipment, not whether it works or whether the household can afford to run it.
A home with a broken unit is counted as having cooling. The true share of households living without usable cooling is therefore higher than the figure on this page, by an amount RECS cannot measure.
State estimates rest on a few hundred respondents each: 439 in Washington, 282 in Hawaii. The standard errors in the table are the honest width of each estimate, and neighboring states in the ranking are frequently not distinguishable from one another. Read the map for its pattern, not for its rank order.
Income, tenure, building age and building type are heavily correlated with each other. This study reports each cut separately and does not attempt to isolate one from the others, because a cross-sectional survey cannot do that credibly.
93.1% of Alaskan households have no cooling, in a state averaging 2 cooling degree days a year. It is the one place where "they do not need it" is the whole explanation, and it is left in the map rather than trimmed, because removing inconvenient points is how maps start to lie.
Questions
How many US homes have no air conditioning?
Which state has the most homes without air conditioning?
Is it just that those homes are in cool places?
Do renters have air conditioning less often than owners?
What kind of housing is least likely to have cooling?
How accurate are these figures?
Written and audited by
HYRE HVAC Research Desk
Primary-source research, data analysis and fact checking
We are a research desk, not a sales floor. We read the federal microdata file, the statute or the manufacturer data sheet ourselves, and we publish the figure with the document it came from and the date we retrieved it.
Where a number cannot be traced to a primary source, we publish the shorter page and say what we could not verify.
The counts below are generated from the published pages themselves, last counted September 28, 2026, and they are what we have actually published rather than what we intend to.
- 13
- studies published
- 12
- federal sources read and cited
- 8
- studies published with their full dataset as CSV
- 51
- jurisdictions reproduced against EIA’s own tables
How this desk works
- Primary sources only. Federal data comes from the agency that collects it, in the file that agency publishes. We do not cite an article that cites a source; we download the source and compute the figure ourselves.
- We validate against the agency before we publish. First, we use each federal microdata file to reproduce the agency’s own published tables. Our cooling research reproduces EIA’s state estimates and standard errors for all 51 jurisdictions. That check caught a variance formula that was off by a factor of four.
- Every estimate carries its uncertainty. These are survey figures, not counts. Standard errors are computed from the replicate weights the federal file supplies and printed beside the estimate. An estimate too imprecise to publish is reported as such rather than printed.
- Nothing is typed by hand. Prose, tables and charts all read from one dataset built by script, so a number in a sentence and the same number in the table below it cannot disagree.
- We publish the data, not just the conclusion. 8 of our 13 studies offer the full computed table as a CSV download on the page, so you can check the analysis or disagree with it. A study without a row-level dataset gets no download link and claims none in its structured data.
- We correct in public. Where we have published a figure wrongly we fix the figure, rewrite any analysis that rested on it rather than patching the number underneath it, and leave a dated correction note on the page.
- We do not install or sell HVAC equipment, and we take no payment for placement, ranking or a favorable mention. Nobody buys a position on this site.
Data as of RECS 2020 public use microdata, v7. Authorship on this site is organizational: the analysis belongs to the desk rather than to a named individual, and we do not publish credentials we do not hold. Our editorial policy sets out how we source, date and correct what we publish.
Sources & retrieval dates
Cooling a home that has never had it
A house with no ductwork is a different job from a replacement. Size it before pricing it; the calculator shows the rule and its limits on the same page.