When efficiency makes us consume more

Savings per unit, falling cost, new uses and total demand. Theory, formulas, worked examples and applications.

Saving per unit does not guarantee saving in total

The mechanism: cheaper, then used more

A more efficient car lowers the cost of every kilometer traveled, and that lower cost can, in a perfectly rational way, encourage driving more kilometers than would be driven with a less efficient car. A cheap, low-consumption LED lighting system can encourage multiplying the number of light points installed, instead of keeping the same total amount of light with less energy. A data center that becomes more efficient can offer cheaper digital services, which in turn stimulates an increase in total demand for those same services. In all three cases the same underlying economic mechanism is at work: part of the technical saving achieved by the efficiency improvement gets "recovered" as higher consumption, instead of translating entirely into less total energy consumed. This phenomenon is known as the rebound effect, first observed in 1865 by economist William Stanley Jevons, who noted that improvements in the efficiency of coal-fired steam engines had paradoxically led to greater total coal consumption, not less.

This does not mean efficiency is pointless

It is easy to misread the rebound effect as an argument against technical efficiency itself, but that reading is wrong: without the efficiency improvements achieved, considerably more energy would be needed to obtain exactly the same final services. What the rebound effect teaches is not that efficiency is pointless, but that improving performance per unit is not the same as reducing total consumption, and that both magnitudes need to be measured and watched separately, instead of automatically assuming that a technical improvement in performance simply translates into a proportional saving in the whole system's overall consumption.

Not every rebound carries the same moral weight

If a home reduces its heating consumption thanks to better insulation, and uses part of that economic saving to be able to maintain an indoor temperature that was previously unhealthy because it was too expensive, that specific rebound is providing genuine, legitimate wellbeing to whoever experiences it — it is not waste, it is a previously unmet basic need that can now be covered. That is not at all the same as using that same economic saving to keep an empty building's façade lit all night with no functional purpose whatsoever. Both are technically "rebound effect" in the strict economic sense of the term, but one meets a real need and the other meets none at all: the size of the rebound matters, but so does what specific service that rebound is funding.

The rebound can also be material, not just energy-related

The same mechanism extends beyond the energy consumed while using a device: cheaper, more efficient electronic devices can speed up the rate at which equipment that still works perfectly is replaced, and make it easier for the total number of devices a person or household owns and keeps running simultaneously to grow. The sense that "the cloud" is a practically unlimited resource arises precisely because each individual digital operation costs very little in terms of energy and money, but the whole system's total consumption — adding up every data center and every telecommunications network in the world — keeps growing steadily, precisely because that low individual cost encourages multiplying the total number of operations performed.

Efficiency accompanied by sufficiency

The reasonable response to the rebound effect is not to give up on improving technical efficiency, but to always pair it with a question about the sufficiency of the service actually needed: clearly define what level of service is reasonable, improve the available technology to provide that service at the lowest possible cost, and actively monitor the whole system's total consumption, not just isolated per-unit performance. Reasonable usage limits, pricing that honestly reflects the service's real cost, longer equipment service life, and urban planning that reduces the need for unnecessary trips can all help preserve the technical saving achieved without denying the legitimate wellbeing that efficiency itself makes possible.

Efficiency with sufficiency

Improving something's efficiency does not guarantee reducing its total consumption: measure and watch both magnitudes separately.
Ask what specific service an economic saving is funding before judging the rebound: covering a previously unmet need does not carry the same weight as purely ornamental consumption.
The rebound can also be material, not just energy-related: a cheaper device can speed up the replacement of equipment that still works.
Pair the technical improvement with a question of sufficiency: clearly define what level of service is reasonable, not just how to provide it more cheaply.
Reasonable usage limits, pricing that reflects the real cost, and longer equipment service life help preserve the saving achieved.

Don't give up on efficiency out of fear of the rebound

The rebound effect is not an argument against improving technical efficiency: without those improvements, obtaining the same services would cost considerably more energy. The problem is not efficiency itself, but assuming it is enough on its own to reduce total consumption.

The numbers, with their source

The rebound can cancel out anywhere from a tenth to more than half of the expected saving

Studies on the rebound effect's magnitude produce highly variable results depending on the sector and method used. Computable general equilibrium (CGE) models find an average rebound of around 58% — meaning more than half of the energy saving expected from an efficiency improvement fails to materialise. Studies using other methods to estimate economy-wide rebound find an even higher average, around 71%. By contrast, so-called "direct rebound effects" — simply using a specific device more because it has become cheaper to run — usually come in at 10% or less in most studies. In the residential sector, some studies find relatively high rebounds, between 56% and 80%.

Sources: Carbon Brief — Why 'rebound effects' may cut energy savings in half; ACEEE — The Rebound Effect: Large or Small?.

Lighting: 3,000 times cheaper, 40,000 times more consumed

The most cited historical example of the rebound effect is lighting. In the United Kingdom, the real price of artificial light — adjusted for inflation — fell about 3,000-fold between 1800 and 2000. Over that same period, artificial-light consumption per person multiplied by 13,000, and the country's total consumption by 40,000, far beyond what any forecast based purely on technical efficiency gains would have predicted. Between 1900 and 2000 alone, the average Briton went from consuming 75 times more artificial light than their 1900 ancestor.

Source: historical analysis of the Jevons effect in lighting, collected in Medium — When LEDs Made the Night Brighter: A Jevons Paradox Story, and in historical lighting-price sources cited by Our World in Data.

What this table is missing

No rebound percentage is valid outside its specific context: the sector, the country, the starting income level and the type of energy service — heating, transport, lighting, digital services — substantially change the real magnitude of the rebound observed in each case.

Sources

The IEA analyses the role of energy efficiency at system scale in Energy efficiency.
Carbon Brief summarizes research on the rebound effect's magnitude in Why 'rebound effects' may cut energy savings in half.
The American Council for an Energy-Efficient Economy (ACEEE) reviews the empirical magnitude of the rebound effect across different sectors.
Historical sources document the rebound effect's origin in William Stanley Jevons's 1865 observation on coal consumption.