Residential Energy Use

All residential buildings in the U.S. together consume 9,481 trillion Btu of energy per year, spending $232.75 billion.

The vast majority of energy use goes to space heating, water heating, air conditioning, and refrigeration. Air conditioning and refrigeration are nearly 100% electric around the country, while heating space and water is done almost entirely with natural gas and propane. A significant portion of energy usage is categorized as “Other” - a long tail of mostly electrified conveniences.

The next graph is a detailed look at uses of electricity. Although electricity makes up a small share of space heating and water heating, these are still the most significant uses of electricity.

Most electricity consumption comes from activities that follow daily patterns. While the refrigerator runs consistently, washing machines may run any time, and microwaves use too little power to matter, three highly correlated activities - air conditioning, space heating, and water heating - together consume 56.6% of electricity.

Water heating (15.5% of electricity usage) is primarily used for showers and faucets, which are primarily used in the morning and at night. And heating and cooling demand (41.1% of electricity usage) is correlated across households as the temperature changes throughout the day. These factors, plus the general increase in activity during the day versus at night (e.g. for lighting and other things), are what gives the grid its load shape.

Heating is nearly universal (ranging from 87% of square feet in the South to 93% of square feet in the Midwest), but the presence of air conditioning varies more by region.These regional differences are driven by the climates found in the regions.

The graph below breaks down the square footage of the four Census regions by climate. The size of the block represents the number of square feet of a given climate (X-axis) in the given region (Y-axis). Darker colored squares have a greater proportion of air conditioned square feet.

The relatively low adoption of air conditioning in the Northeast is because that part of the country is very cold. And the relatively low adoption of air conditioning in the West is because much of that part of the country (or rather, much of the square footage in that region) enjoys the mild weather of the Pacific coast.

Although there’s nearly as much cooled square feet of housing as heated square feet, residential buildings use much more energy for heating than for cooling. Partly it’s because cooling is a more efficient process for thermodynamic reasons. It takes less energy to remove heat than to create heat.

It’s also because heating is typically used more frequently than cooling. Winter lasts longer than summer in most places. And while hot days are never more than 30 degrees hotter than a comfortable temperature, on cold days temperatures can drop much more than 30 degrees below the setpoint.

The next graph again breaks down the country’s square footage¹ by Census region and climate, but now the shading represents the energy intensity of space heating (energy used for heating per square foot of heated space). Space heating is most intense in the cold regions because it’s used more there. (The same is true for air conditioning in hot regions).

As I’ve said, the heating of space and water is mostly powered by natural gas, and electricity mostly powers everything else. But heating is also powered by propane and fuel oil / kerosene, and this fuel mix differs regionally. The Midwest and West both have ~80% natural gas, while the South has the most electricity and the Northeast has by far the most fuel oil in their heating fuel mix.

These differences actually have more to do than the year in which a house was built than its region. Older homes are more likely to use propane or fuel oil for space heating because the natural gas system was only completed in the 1960s. Newer homes are more likely to use electricity for space heating because electric heat pumps were first commercialized in the 1990s.

Because the South has experienced more economic growth than the Northeast since 1960, homes in the South are newer than in the Northeast. 32.6% of housing units in the Northeast were built before 1950 (compared with 7.9% in the South).

Let’s now narrow our focus to spending (in dollars) on electricity. We’ve already seen the long tail of electricity-power activities and the amount of energy that each consumes. Let’s see the cost of that total consumption in dollars.

Residential spending on electricity totalled $161.5B. The average household spent $1,380/yr on electricity in 2020 (which is actually quite a bit lower than other estimates I’ve seen - but I’ll use it for this article). Comparing that figure to the 2020 BLS consumer expenditures survey, that’s about as much as the average household spent on apparel, education, and gasoline but quite a bit less than was spent on entertainment and food.

This graph shows the average annual cost of an end use to a household that actually has that end use. For example, the average household does not spend $302/yr on pool pumps, but the average household that has a pool spends $302/yr on the electricity for their pool pump.

What stands out to me here is how cheap the energy costs are for incredibly useful things. The marginal cost of using your microwave, dishwasher, and washing machine is measured in cents, not dollars.

And then some things are surprisingly expensive. Everyone knows that air conditioning and heating consume a lot of power - we have the trope of the father admonishing the children to conserve air conditioning - but I think a lot fewer people think about water heating, even though in this study it’s even more expensive. Maybe the father’s efforts are better spent encouraging shorter showers. Even still, the marginal cost of a warm shower is measured in cents, not dollars.

Pool pumps seem expensive. (I’m not sure how pool heating is accounted for, since there’s a “hot tub heaters” category but not one for pools. Maybe pool heating is included in overall water heating?) But water is heavy and it takes a lot of energy to move heavy things. Plus pool pumps run continuously.

EV charging costs $280/yr on average. That’s enough to make a noticeable difference on the monthly bill, but it’s tiny compared to what people typically spend on gasoline.

Looking at adoption of various household energy uses, I’m shocked by how low the adoption of dishwashers and clothing washers/driers is. It’s commonly said that when countries first begin to develop, the first major energy-related purchase that households tend to make is a refrigerator, typically followed by air conditioning, televisions, and washing machines. Those first three items have nearly universal adoption even among Americans in deep poverty, but washing machines are owned by only 56.4% of households earning less than $5k/yr.

I don’t think it’s the cost of the appliances that’s limiting their adoption. Refrigerators have a similar upfront cost, and higher operating costs, compared with washer/drier sets. Although the refrigerator has significantly more utility, I still find it surprising that adoption of these other appliances is so much lower.

I think this is partly explained by the fact that those appliances require plumbing hookups, while refrigerators can be plugged into a regular socket. The lowest income people are much more likely to live in older buildings and apartments, which are less likely to have hookups to allow for dishwashers.

Electricity can be a particularly stressful budget item for many people. Although utility bills are low compared to most other spending categories, it’s such an essential service that there aren’t many options for cutting back without discomfort. When seeking to lower electricity spending, most people first seek to use less heating or air conditioning.

The EIA survey indicates that 9.9% of households have left their home at an “unhealthy temperature” to reduce their bill. This varies by income from 24.8% of households earning <$5k/yr to 4.0% of households earning >$150k/yr. 10.0% of households (<$5k=26.6%, >$150k=1.5%) received a notice advising them that their service will be disconnected if they don’t pay by a certain date. But actual disconnections are rare. Only 0.2% of households experienced a disconnection for nonpayment. (Presumably this figure has the same income distributional characteristics as the others, but the number of households experiencing disconnection was too low for the survey to break it down by income.)

Now let’s look at the adoption of various energy technologies that are changing how residential buildings interact with the grid. For each technology I’ll look at what building structure or household demographic statistics most strongly predicts adoption.

The first thing I notice is that respondents weren’t even asked about battery storage. When the next EIA survey comes out, I bet that they’ll ask about residential battery storage, and I bet that solar adoption will be much higher.

The technology with the most adoption is smart meters (or AMI - Advanced Metering Infrastructure). Maybe this is unsurprising, since it’s the only technology that consumers generally don’t choose for themselves; it’s installed for them by the utility and the cost is socialized.

A building’s AMI adoption is most strongly predicted by its structure type. Just 14.8% of large apartment buildings have AMI, while 32.8% of detached single family homes do. This is probably because the metering infrastructure for a detached single family home is much simpler than for a large apartment complex.

Adoption of back-up generation is surprisingly high at 14.3% overall. This is typically a portable gasoline-fueled generator with enough power for essential loads, costing around $1,000. Adoption is most strongly predicted by the structure type, since it’s hard to imagine how an apartment dweller could opt into backup generation.

Back-up generation is apparently popular in places with frequent natural disasters, like the hurricane-prone Gulf Coast. But the data shows higher adoption in New England (22.3%) than the East South Central and South Atlantic (which include the most hurricane-prone areas in the U.S.). Maybe the cold temperatures in New England make back-up power more safety-critical?

The level of adoption of back-up power is also interesting because it’s a strong indication of how much people value having reliable power. Everyone knows that power outages happen. The average electricity customer experienced 366 minutes of power outages in 2023 (and much less if you exclude extreme events like hurricanes). And most people know that a backup generator can be purchased for around $1,000. So when someone buys a generator, they’re saying that avoiding 366 minutes of outage is worth $1,000 to them. And when someone doesn’t, they’re saying avoiding outages is worth less to them.

Residential electricity customers vary widely in how much they’d pay to avoid outages. Typically this is correlated with income (and we see this in the EIA data here), but there are other factors like health concerns. Elderly people may be less able to bear the discomfort of outages, and people who use medical equipment may require electricity for their health.

But when the grid goes down, it goes down for everyone. You can’t pay more or less to get more or less reliable electricity service (except by buying on-site back-up power). Investments in reliability are socialized over the whole system, regardless of how an individual customer values reliability. Utilities and grid operators commission studies to estimate the weighted average value of reliability to all customers, but it’s very challenging and those studies have a suspicious tendency of landing on round numbers like $5,000/MWh and $10,000/MWh.

Moving on to smart thermostats, we see that income most strongly predicts adoption (<$5k=3.1%, >$150k=25.7%). That makes sense because you can install a smart thermostat in any structure, even as a renter. Its only value is as a minor convenience. And there aren’t many income-based subsidies for smart thermostats.

Solar adoption varied most widely by climate. Mixed-dry climates have a 7.8% adoption rate while adoption is lowest in mixed-humid areas (1.7%). I actually think that Census region is the driving factor due to state incentives, but climate region happens to have a bit more signal because although it’s closely associated with Census region it also picks up things like income effects.

EVs are a similar story. EV adoption varies most widely with income, followed by climate region.