Evidence of a Decline in Electricity Use by U.S. Households
It has been slowing down for decades, but is electricity use by American households now going down?
Americans tend to use more and more of everything. As incomes have risen, we buy more food, live in larger homes, travel more, spend more on health care, and, yes, use more energy. Between 1950 and 2010, U.S. residential electricity consumption per capita increased 10-fold, an annual increase of 4% per year.
But that electricity trend has changed recently. American households use less electricity than they did five years ago. The figure below plots U.S. residential electricity consumption per capita 1990-2015. Consumption dipped significantly in 2012 and has remained flat, even as the economy has improved considerably.
Source: Constructed by Lucas Davis at UC Berkeley using residential electricity consumption from EIA, and population statistics from the U.S. Census Bureau.
The decrease has been experienced broadly, in virtually all U.S. states. The figure below shows that between 2010 and 2015, per capita residential electricity consumption declined in 48 out of 50 states. Only Rhode Island, Maine, and the District of Columbia experienced increases.
Source: Constructed by Lucas Davis at UC Berkeley using residential electricity consumption from EIA, and population statistics from the U.S. Census Bureau. Electricity use per capita is measured in megawatt hours.
This pattern stands in sharp contrast to previous decades. During the 1990s and 2000s, for example, residential electricity consumption per capita increased by 12% and 11%, respectively, with increases in almost all states. Previous decades experienced much larger increases.
So what is different? Energy-efficient lighting. Over 450 million LEDs have been installed to date in the United States, up from less than half a million in 2009, and nearly 70% of Americans have purchased at least one LED bulb. Compact fluorescent lightbulbs (CFLs) are even more common, with 70%+ of households owning some CFLs. All told, energy-efficient lighting now accounts for 80% of all U.S. lighting sales.
It is no surprise that LEDs have become so popular. LED prices have fallen 94% since 2008, and a 60-watt equivalent LED lightbulb can now be purchased for about $2. LEDs use 85% less electricity than incandescent bulbs, are much more durable, and work in a wide-range of indoor and outdoor settings.
Source: Energy.Gov, “Revolution…Now”, September 2016.
Is this really big enough to matter? Yes! Suppose that between LEDs and CFLs there are now one billion energy-efficient lightbulbs installed in U.S. homes. If operated 3 hours per day, this implies savings of 50 million megawatt hours per year, or 0.16 megawatt hours per capita, about the size of the decrease above. Thus, a simple back-of-the-envelope bottom-up calculation yields a similar decrease to the decline visible in aggregate data.
No other household technology is as disruptive as lighting. Incandescent bulbs don’t last long, so the installed stock turns over quickly. Air conditioners, refrigerators, dishwashers, and other appliances, in contrast, all have 10+ year lifetimes. Thus, although these other technologies have also become more energy-efficient, this can’t explain the aggregate decrease. The turnover is too slow, and the gains in energy-efficiency for these other appliances have been too gradual for these changes to explain the aggregate pattern.
Traditional economic factors like income and prices also can’t explain the decrease in electricity use. Household incomes have increased during this period, so if anything, income effects would have led electricity use to go up. Moreover, between 2010 and 2015, the average U.S. residential electricity price was virtually unchanged in real terms, so the pattern does not seem to be the result of prices.
Another potential explanation is weather. The summer of 2010 was unusually hot, so this partly explains why electricity consumption was so high in that year. But the broader pattern in the figure above is clear even if one ignores 2010 completely. Moreover, I’ve looked at these data more closely and there is a negative trend in all four seasons of the year: Summer, Fall, Winter, and Spring.
This is not the first time in history that lighting has experienced a significant increase in energy-efficiency. In one of my all-time favorite papers, economist Bill Nordhaus examines the history of light from open fires, to candles, to petroleum lamps, to electric lighting. Early incandescent lightbulbs circa 1900 were terribly inefficient compared to modern incandescent bulbs, but marked a 10-fold increase in lumens per watt compared to petroleum lamps. However, as lighting has become cheaper, humans have increased their consumption massively, consuming thousands of times more lumens than they did in the past.
Economists refer to this price effect as the “rebound effect”. As lighting becomes more energy-efficient, this reduces the “price” of lighting, leading to increased consumption. An important unanswered question about LEDs is to what extent will these energy efficiency gains be offset by increased usage? Will households install more lighting now that the price per lumen has decreased? Will households leave their lights on more hours a day? Outdoor lighting, in particular, would seem particularly ripe for price-induced increases in consumption. These behavioral changes may take many years to manifest, as homeowners retrofit their outdoor areas to include additional lighting.
It is not clear yet whether U.S. household electricity use has indeed peaked or this is just a temporary reprieve. Probably the biggest unknown in the near future is electric vehicles. Currently only a small fraction of vehicles are EVs, but widespread adoption would significantly increase electricity demand. It is worth highlighting, though, that this would be substitution away from a different energy source (petroleum), so the implications are very different from most other energy services.
Over a longer time horizon there will also be entirely new energy-using services that become available, including services that are not yet even imagined. The 10-fold increase in electricity consumption since 1950 reflects, to a large degree, that U.S. households now use electricity for many more things than they did in the past. The recent decrease is historic and significant, but over the long-run it would be a mistake to bet against our ability to consume more energy.
Keep up with Energy Institute blogs, research, and events on Twitter @energyathaas.
Suggested citation: Davis, Lucas. “Evidence of a Decline in Electricity Use by U.S. Households” Energy Institute Blog, UC Berkeley, May 8, 2017,
For more see Davis, Lucas W. “Evidence of a Decline in Electricity Use by U.S. Households,” Economics Bulletin, 2017, 37(2), 1098-1105.
Lucas Davis View All
Lucas Davis is the Jeffrey A. Jacobs Distinguished Professor in Business and Technology at the Haas School of Business at the University of California, Berkeley. He is a Faculty Affiliate at the Energy Institute at Haas, a coeditor at the American Economic Journal: Economic Policy, and a Research Associate at the National Bureau of Economic Research. He received a BA from Amherst College and a PhD in Economics from the University of Wisconsin. His research focuses on energy and environmental markets, and in particular, on electricity and natural gas regulation, pricing in competitive and non-competitive markets, and the economic and business impacts of environmental policy.
While the results of this article are true, the measure of electric use per capita seems misleading.
A better index of electric usage would be total usage divided by housing units less vacancy if you can’t get actual customer/meter data from utilities.
We have seen our residential usage follow the 2000-2015 pattern in the first graph over the last twenty years, as have other utilities in CA. The number of customers (meters) continues to increase but the average usage per meter decreases. The size of the households varies from the single occupant to multiple families per house and doesn’t necessarily correlate with electric usage.
The adoption of LEDs has led to a large part of the usage decline. Included in this transition is the switch from CRTs (monitors & TVs) to LED monitors and LED TVs. A 55″ LED TV is rated at about 135 kWh per year, which is about 25 kWh more than a 60 watt incandescent burning 5 hours per day.
The number of EVs will continue to increase and increase electric usage. Part of this increase will be selling to a retail customer at night rather than selling energy at a loss. Depending on how fast storage will be adopted, there might even be a 13:00-17:00 duck curve charging periods for EVs.
A very large factor that is missed is the impact of the government. Yes, I know that economists think that the government can’t do anything and it is the magical market that causes all change. But, higher codes and standards lower electricity mandates for energy usage by new devices has a very large impact on energy use. When your 5/10/15 year old washer/refrigerator/dishwasher/TV/cable box dies, the new one that you buy will use less energy than your old item even if the least expensive item is purchased. Many of the more high priced items (eg Bosch) have even lower energy usage.
The post notes that the occupants per unit has been pretty steady between 2.53 and 2.56, so the per capita metric is equivalent to yours.
Declining energy consumption is a phenomenon throughout the OECD and is consistent with increasing deployment and use of a range of energy efficient devices domestically, commercially and industrially. Documented falls in electricity consumption have also being occurring across Europe and the UK.
As with other OECD markets, in Australia, electricity consumption has also been falling since 2008, and while some changes are undoubtedly a consequence of structural adjustments in the economy through changes in manufacturing, the largest change can be attributed to efficiency. Within the residential market, the largest gains in Australian homes (where cooling, not heating, is the primary issue) is greatly improved efficiency of refrigerators and reverse-cycle air conditioners. While white goods do have an average life of 10 years, the greatest recent energy efficiency gains arose from energy efficiency standards rolled out in Europe and the US in the early 2000s, the impacts of which emerged from factories by the mid 2000s. Given the average of 10 years, around half of white goods are likely to have been replaced up to 2016.
And of course the advent of compact fluorescents and now LED globes has also had an impact (Australia’s phase out of incandescent globes started in 2009.)
The downward trend in Australia has been further enhanced by the massive addition of solar on homes in Australia since 2010 reducing on-grid demand through self-production – more than 1 in 5 homes in Australia now has solar PV, and still growing.
Interesting article, although it is curious that it omits any references at all to the efficiency of the building fabric. Better insulation of walls, attics, floors, plus effective weatherisation of doors plus low emissivity glazing : all these make homes not just more comfortable, but less gas guzzling.
However, the point of this blog is that building stock turnover is so slow that these changes cannot explain the magnitude of the usage reductions.
Overall, energy efficiency building refurbishment each year saves more energy that new construction , particularly if the vacated older home continues to be lived in.
That’s true, but refurbishment is progressing at a slow rate. This elements needs much more emphasis and support by energy efficiency programs. There have been several posts on this blog about this issue.
Thoughtful article and comments above. For your consideration.
U.S. residential power rates have been rising faster than CPI since 2005. Only other time this was true in history was late 70’s early 80’s during similar subsidized RE push. If you believe shadowstats(dot)com more than BLS (I do), then CPI has been consistently under-reported since mid 1980s, real wages (for those employed) have been flat, true unemployment (measured the same way it was for great depression) is today still north of 20%, and GDP has been shrinking since 2001 due to de-industrialization. In this macroeconomic environment, one would expect to see a decline in electricity usage regardless of efficiency, and the largest declines in the most rural and impoverished areas. Whereas, if the decline was due only to increasing efficiency, the decline should be in the wealthier areas where there is faster appliance turnover and more disposable income to spend on expensive energy efficiency upgrades like rooftop solar, geothermal heat systems, CFL and LED lighting upgrades, smart thermostats, home networking, etc. When the data clearly show southern states leading the decline in electricity consumption, it is much more likely that we are seeing the effects of rising energy poverty rather than a watershed reversal of Jevons’ Paradox. Rising energy poverty, particularly in electricity, has also hit the UK, Spain, Italy, Greece, Denmark, and Germany. Would be silly to think USA is the exception. This is damage largely self-inflicted by aggressive, expensive, yet ineffectual western government policies enacted in the name of affecting the Earth’s climate.
Here’s a detailed critique of Shadowstats as a reliable source for these metrics: http://www.economonitor.com/dolanecon/2015/03/31/deconstructing-shadowstats-why-is-it-so-loved-by-its-followers-but-scorned-by-economists/
That is a very thoughtful critique by Ed Dolan, and I agree with him that Williams overstates inflation considerably. But, it is important to note that he ultimately argues to correct Williams value, not to CPI-U, but to 2.45% above CPI-U. Even a value 2% above CPI-U reverses GDP growth to GDP decline from 2001 to 2015. Dolan does not here attack Williams’ unemployment statistic.
I would criticize Dolan’s attempt to reconstruct GDP using raw counts of car purchases. This industry has been heavily off-shored during this time, preferences have shifted from domestic to imports, purchases have lost ground to leases, and private and public debt has exploded. Any calculation or historical reconstruction of GDP also needs to back out the increasing fraction of overhead costs (not productivity) that the financial and government sectors represent.
Do you have any critique on the substance of my comment in regards to explaining why the consumption reductions are weighted toward poorer states. Is this not the exact opposite of what should be expected if it was due to technology turnover to more efficient devices? I’m not arguing that efficiency gains are not happening, but that rising energy poverty is the more dominant force. We have objective metrics of price and wages to make the energy poverty case.
I found other critiques of ShadowStats GDP estimates as well, but I’m not going to engage in the validity of a website that clearly has substantial credibility problems. My point is if it is so bad on just one point, why should we trust it on others. You need to bring more supporting evidence than simple assertions to back up this set of alternative facts. (And while I use that pejorative term, you may in fact have supporting evidence, but you haven’t presented it.)
As for the CPI measurement debate, there are many views on what’s a valid approach due to the problem of defining the “market basket.”
On energy efficiency gains, we expect that technology rollout lags in poorer areas, but not that they don’t happen at all. It just means that we can expect to see the efficiency gains in those states later than in the wealthier states. That California is an early adopter and that its gains are smaller in the more recent years is consistent with that.
Had to dig through two paragraphs of scorn to find your answer. So you believe the reason for the disparity between poor and rich states is due to early adoption v. slower adoption of EE tech. That would be plausible if efficiency was a step function rather than a continuous one where everyone is moving on their portion of the continuum simultaneously. However, CA and NY are still avant-garde in energy policy, pricing, residential solar adoption, and other things that should be accelerating their efficiency gains compared to the American Southeast. Interesting that EIA, like me, is predicting rebound in national residential electricity consumption. Their prediction for 2018 is an increase larger than the decline in 2016 and 2017 ( p. 18 of just published STEO https://www.eia.gov/outlooks/steo/pdf/steo_full.pdf ). Doesn’t fit your theory or preconceptions. Will you also dismiss this?
Ike, here’s another critique of the GDP measure in ShadowStats. There are many more on the web: http://globaleconomicanalysis.blogspot.com/2012/05/gdp-real-gdp-and-shadowstats-theater-of.html
I don’t understand your point about EE being a step function. As the easiest, cheapest measures are adopted, as they have been in California, the states that haven’t yet adopted those measures can make bigger gains relative to California in future years. California has been in the cutting edge for 40 years, which gives plenty of room for lagging states to be coming up from behind.
As for EIA forecast, that agency ALWAYS forecasts load growth. Here’s the 2013 Annual Energy Outlook that forecasts continued growth. https://www.eia.gov/outlooks/aeo/pdf/0383(2013).pdf And this from the Feb 2015 STEO: “Electricity Generation. EIA forecasts that U.S. electricity generation will grow by an average of 1.0% 2015 and 0.9% 2016.”
(And I find it ironic that you criticize government sources for empirical data, but then turn to government data for a forecast…)
Data in EIA reports are probably trustworthy. Their predictions are rubbish. At the time the FPC became the FERC, some of us including myself were offered the opportunity to join the EIA.
I decided not to do so when one of the representatives of the management warned against short term forecasts, because they could too easily be shown to be wrong!
U.S. residential power rates have been rising faster than CPI since 2005. Only other time this was true in history was late 70’s early 80’s during similar subsidized RE push. If you believe shadowstats.com more than BLS (I do), then CPI has been consistently under-reported since mid 1980s, real wages (for those employed) have been flat, true unemployment (measured the same way it was for great depression) is today still north of 20%, and GDP has been shrinking since 2001 due to de-industrialization. In this macroeconomic environment, one would expect to see a decline in electricity usage regardless of efficiency, and the largest declines in the most rural and impoverished areas. Whereas, if the decline was due only to increasing efficiency, the decline should be in the wealthier areas where there is faster appliance turnover and more disposable income to spend on expensive energy efficiency upgrades like rooftop solar, geothermal heat systems, CFL and LED lighting upgrades, smart thermostats, home networking, etc. When the data clearly show southern states leading the decline, it is clear which case is more likely true.
I thought I left this post already. Well to redo it….
EV adoptions are on a rise. To consider their impact:
Average EV runs 3.5 miles / kWh.
Assume 1,000 miles / month for BEVs and 500 miles electric only / month for PHEVs with a 50/50 mix.
Assume 80% of charge at home
Load: ((1000 + 500) / 2) / 3.5 * 80% = 171 kWh per month per average vehicle.
Ignores line and conversion losses.
Not sure how big of an impact this has, but we have been moving away from CRT/plasma TV’s to LED as well.
I had thought that average household size [millenials moving back with parents] might have an effect; but find that the average has stayed at 2.53-2.56 between 2006/2016.
Seems like greater household/ rooftop solar would be a significant factor.
Maybe a look at household / per capita energy consumption would be a place to look at to include impact of switching from electric to gas cooking. Eating out may reduce home energy consumption – i see that grocery store sales [nationwide] dropped for the first time in many years [perhaps first time ever]
I agree average household size could be a factor. Looks like average HH size declined 2% from 2010 to 2015 (https://www.statista.com/statistics/183648/average-size-of-households-in-the-us/)–which is roughly the magnitude reduction 2010-2015 from hypothetical LED energy savings cited in this article.
Wouldn’t a decrease in average household size lead to an increase in per capita usage? If the numerator stays the same but the denominator decreases, use per capita should increase not decrease. Also, if millennials are moving out and occupying previously unoccupied premises, wouldn’t the additions of lights, refrigeration, etc. add to the numerator?
I guessed at what the yearly values are in your first graph and reproduced the approximation with the Y-axis (electricity use per capita measured in MWhs) scaled at 0 rather than starting at slightly less than 3.6 MWhs per capita. The well-pronounced curve in your graph looks a whole lot less impressive when the scaling changes. The downward trend looks to have started in 2007 with an upward spike in 2010 before resuming the downward path again in 2011 with a large decline in 2012.
You indicate there is a 10-fold increase in the 61 years from 1950 to 2010 which is a 4% annual increase. Assuming that the 2010 value is approximately 4.7 MWh per capita would mean the 1950 figure is about .434 MWh per capita. Since the 1990 figure looks to be about 3.75 MWh that would mean an annual growth of nearly 5.4% between 1950 and 1990. The growth between 1990 and 2010 is only about 1.1% annually. From 2011 to 2015, there was some decline in 2011 that looks similar to year over year fluctuations that happen at other times in the period covered (for example, 1996 to 1997 or 2005 to 2006, 2007 to 2008 or 2008 to 2009). The big decline occurred in 2012 and usage has trended up in 2013 and 2014 before experiencing a decline in 2015. All the values between 2012 and 2015 look similar to year-to-year variation that is evident at other periods in the graph.
It doesn’t appear to me that the combination of CFL and LED bulbs can explain this. In fact, 2009 to 2010 appears to be the largest single-year positive change in MWhs per capita in the 21 years covered in the graph. It seems like a strong assumption that the average use of ALL lights is 3 hours per day. As more lights become CFL or LED that would seem to mean that the saturation is starting to go more from high use lights like those in the living room towards lights in kitchens and hallways that are going to get way less than 3 hours of daily use.