The Jevons paradox is the observation that when technology makes the use of a resource more efficient, total consumption of that resource may rise instead of fall. Greater efficiency lowers the cost of each unit of useful work. If demand responds strongly enough to the lower cost, through heavier use and through new uses that were not economic before, the increase in activity outweighs the saving per unit.
The idea is named after the English economist William Stanley Jevons, who set it out in 1865. It is a central point in debates about energy efficiency and climate policy, and it has been widely cited in discussion of computing and artificial intelligence, where the cost of a given task has fallen steeply. The paradox describes a possible outcome, not a law: whether it occurs depends on how sensitive demand is to cost.
Origin
In The Coal Question, Jevons examined Britain's dependence on coal. Many of his contemporaries believed that more economical steam engines would stretch the country's reserves. Jevons argued the reverse. James Watt's engine used far less coal per unit of work than its predecessors, and that was exactly why steam power spread from pumping mines to factories, railways and ships. Coal consumption had multiplied as engines improved. He concluded that it was a confusion of ideas to suppose that economical use of fuel meant reduced consumption.
Mechanism and conditions
The effect works through price. An efficiency gain reduces the effective cost of the service the resource provides, such as motion, light or computation. Consumption of the service then rises. Economists call this response the rebound effect and measure it as the share of the expected saving that is lost to extra use. A rebound of more than 100 percent, where resource use ends up higher than before, is called backfire, and this is the Jevons paradox in the strict sense.
Backfire requires elastic demand for the service: a given percentage fall in cost must bring a larger percentage rise in use. Where there is inelastic demand, as with home heating in a rich country, efficiency does cut consumption, though by less than engineering estimates suggest. Rebound also has indirect forms. Money saved on one thing is spent on others that use the resource, and cheaper inputs raise output across the economy.
Worked example
The change in resource use can be written as:
- and are resource use before and after the improvement.
- and are the quantities of the service consumed before and after.
- and are efficiency before and after, in units of service per unit of resource.
Suppose efficiency doubles, so , and the cost of the service halves. If consumption of the service rises by 50 percent, resource use is of its former level, a fall of 25 percent where engineers expected 50 percent. The rebound is half the expected saving. If consumption of the service instead rises to 2.5 times its former level, resource use is , a rise of 25 percent. That is backfire.
Evidence
Historical cases of backfire are easiest to find where a technology is young and its cost falls by large multiples. Lighting is the standard example: each move from candles to gas, to electric filament and to more efficient lamps cut the cost of light, and societies responded by using vastly more of it. Computing shows the same pattern. For mature uses in developed economies, studies of direct rebound in car travel and household heating usually find that a minority of the saving is lost, well short of backfire. Estimates of economy-wide rebound are larger and more uncertain. The view that efficiency raises total energy demand at the level of the whole economy is associated with the economists Daniel Khazzoom and Leonard Brookes.
Relevance to policy and investment
For climate policy, the paradox implies that efficiency standards alone may not reduce greenhouse gas emissions as much as planned. Many economists argue that efficiency measures work best when combined with a price on the resource, such as a carbon tax, which keeps the cost of use from falling and charges for the externality.
In markets, the argument resurfaced in early 2025, when the release of a low-cost Chinese AI model raised doubts about demand for chips and data centres. Technology executives and analysts invoked Jevons to argue that cheaper computation would widen the use of AI and so increase total spending by each hyperscaler on computing capacity and electricity. The same reasoning appears in thematic investing in fields where costs follow a steep learning curve or fall with economies of scale. Critics caution that the paradox is not automatic. It holds only if demand is sufficiently elastic, and suppliers can still suffer if prices fall faster than volumes rise, since lower marginal cost tends to be competed away.