What Is This?
In 1865, a British economist named William Stanley Jevons published The Coal Question, a book warning that Britain would run out of coal. His reasoning was counterintuitive enough that it became known as a paradox: improvements in steam engine efficiency, rather than reducing coal consumption, had caused it to increase dramatically.
Jevons had observed that as James Watt's improved steam engine made coal use 4-5x more efficient, the number of steam engines multiplied enormously. Cheaper operation meant wider deployment — in factories, mines, railways, shipping. Total coal consumption went up, not down.
His insight: when you make a resource cheaper to use by using it more efficiently, you lower the effective cost of getting value from it. Lower cost means people use more of it. More use means total consumption rises even as per-unit consumption falls. Efficiency improvements cause resource consumption to increase at the system level.
This is the Jevons Paradox, also called the rebound effect. It has been demonstrated repeatedly across 160 years of technological history and is currently reshaping debates about AI energy consumption, climate policy, and the limits of efficiency as a solution to resource constraints.^1
Why Does It Matter?
- Efficiency gains are the centrepiece of almost every plan to reduce resource consumption. More fuel-efficient cars, greener data centres, LED lighting, better insulation — all presented as solutions to energy use and emissions. Jevons says these solutions are incomplete at best and counterproductive at worst if they're not accompanied by absolute caps on consumption. Lower cost of operation → more operation → higher total resource use. The mechanism is not just empirically observed; it's economically inevitable in competitive markets.
- The AI version is playing out in real time. When DeepSeek released dramatically more efficient language models in early 2025, investors sold off NVIDIA and Anthropic on the assumption that AI compute demand would fall. Jevons predicts the opposite: cheaper AI inference means more AI will be run, more applications become economically viable, and total compute demand rises. NPR Planet Money ran a piece titled "Why the AI world is suddenly obsessed with Jevons paradox" — because the paradox perfectly predicts what happens when you make a technology radically cheaper.^2
- It applies to time and attention, not just energy. Productivity tools follow the same logic. Email made business communication faster — total communication volume increased. Smartphones eliminated downtime — leisure time became scarcer. Automation of routine tasks — the time "saved" gets channelled into more complex tasks rather than rest. This is why "I'll use this tool to free up time" almost never produces the promised time savings at the system level: freed capacity finds new uses.
- It's the most important framework for predicting second-order effects of any technological improvement. When a technology gets dramatically cheaper or more efficient, ask: what happens to total demand for what this technology provides? The answer is almost always "it increases substantially." This is the question that climate optimists, productivity gurus, and AI boosters consistently fail to ask.
- It undermines the "dematerialisation" thesis. Some economists argue that advanced economies are "dematerialising" — growing while using fewer physical resources. Jevons says this is often an accounting illusion: the resources are still being consumed, just elsewhere in the supply chain or in sectors that were previously uneconomical. Total global resource consumption has increased alongside efficiency improvements in every measured category.
Key People & Players
William Stanley Jevons (1835–1882) — British economist, logician, and one of the founders of modern economics. Published The Coal Question in 1865, which made him famous (John Stuart Mill praised it in Parliament). His core insight has proven more durable than almost any prediction made by Victorian economists.^3
Harry D. Saunders — Economist who formalised the macroeconomic case for the Jevons Paradox in the 1990s and coined the term "backfire" for the case where efficiency improvements increase total consumption beyond the original baseline. His work established the theoretical framework that makes Jevons's observation more than anecdote.^4
Joseph Khazzoom and Len Brookes — Economists who independently formulated what became known as the "Khazzoom-Brookes postulate" in the 1980s: that economy-wide energy efficiency improvements lead to increased energy consumption at the macroeconomic level. Their work revived Jevons for the energy policy debate.
Blake Alcott — Environmental economist who has written the most accessible and rigorous modern treatment of the Jevons Paradox, distinguishing between different types of rebound and when backfire (full Jevons) is most likely to occur.
Amory Lovins (Rocky Mountain Institute) — The most prominent critic of the Jevons Paradox's policy implications. Lovins argues that the rebound effect is real but manageable, and that efficiency improvements combined with correct price signals can still reduce total consumption. The debate between Lovins and Jevons advocates is the central dispute in energy policy economics.
The Current State
The rebound effect is empirically documented across every major domain where efficiency has improved. The dispute is about its magnitude: does efficiency improvement lead to partial rebound (consuming more, but not as much as the efficiency savings), full rebound (consuming exactly as much as before), or backfire (consuming more than before the improvement)?
The documented cases:
- Steam engines (1769–1900): Efficiency improved ~10x; coal consumption increased ~100x. Classic backfire.
- Vehicle fuel efficiency (1975–2025): Fuel efficiency of US cars roughly doubled; vehicle miles traveled tripled. Substantial rebound.
- Commercial aviation: Aircraft fuel efficiency per passenger-mile improved 70% since 1970; passengers carried increased ~10x. Clear backfire on total fuel.
- LED lighting: LEDs use 75-80% less electricity than incandescent bulbs; global light consumption has increased substantially as lighting became affordable in previously unlit areas and buildings dramatically increased their lit area. Full rebound in developing markets.
- Computing (Koomey's Law): Computations per watt have doubled every ~18 months since 1945; total energy consumed by computing has increased continuously. Backfire.
- AI inference (2024-2026): DeepSeek's R1 model and subsequent efficiency improvements reduced the cost of frontier AI inference by ~10-100x. Total AI API calls and AI-powered applications are projected to increase far more than 10-100x. Jevons in real time.^5
Where the debate stands:
The question of whether the full Jevons backfire (net increase above original consumption) occurs depends on price elasticity — how responsive demand is to lower prices. For energy used in production (industrial processes, computing), elasticity tends to be high and backfire is common. For end-consumer energy use (home heating, personal transportation), partial rebound is more common but still significant enough to undermine efficiency-only policies.
The policy implication is uncontested among economists who take Jevons seriously: efficiency improvements must be accompanied by absolute caps (carbon taxes, resource taxes, emissions trading) to prevent rebound from negating the efficiency gains. Efficiency alone is not a solution.
The broader application:
The paradox extends beyond energy to any domain where resources are scarce and efficiency can be improved. Time is the clearest current example. Every generation of productivity tools has promised to save time; every generation has found that time savings get reallocated into more tasks rather than leisure. The "four-hour workweek" is not a Jevons-violating invention — it works only for individuals who actively cap their workload, which is an artificial constraint operating against the Jevons mechanism.
Best Resources to Learn More
- Jevons Paradox — Wikipedia — The most comprehensive accessible overview, with historical examples and the full range of academic positions.^6
- NPR Planet Money: "Why the AI world is suddenly obsessed with Jevons Paradox" (Feb 2025) — The best short contemporary treatment, focused on AI.^7
- The Coal Question by William Stanley Jevons (1865) — Freely available online. Chapter VII is where the paradox is articulated. Remarkably readable for 19th-century economics.^8
- Small is Beautiful by E.F. Schumacher — The countervailing view: that efficiency improvements at the right scale with the right values can reduce consumption. The best statement of the Lovins position.^9
- MDPI Energies: Jevons Paradox and SDGs (2022) — The most current academic review, covering economy-wide evidence.^10