From William Stanley Jevons’s 1865 The Coal Question. Jevons observed that James Watt’s improved steam engine — which burned roughly a third of the coal of Newcomen’s earlier design — had not reduced British coal consumption. It had increased it.
“It is wholly a confusion of ideas to suppose that the economical use of fuel is equivalent to a diminished consumption. The very contrary is the truth.” — Jevons, The Coal Question, Chapter VII
The Mechanism
Greater efficiency lowers the unit cost of using a resource. Lower cost expands the set of applications where the resource is economically viable. Demand grows along three margins:
- Intensive — existing users use more. Cheaper miles per gallon, more miles driven.
- Extensive — marginal users adopt. Steam engines become affordable to smaller firms.
- Adjacent — new applications become viable. Industries that did not exist at the old price emerge at the new one.
If the combined response across those three margins exceeds the per-unit reduction, total consumption rises. That is the paradox.
When It Holds
The paradox is conditional, not axiomatic. It requires the price elasticity of demand for the resource’s services to be high enough that the rebound exceeds the efficiency gain.
It tends to hold when the technology is in early adoption with substantial unmet demand, the resource is cheap relative to the value it unlocks, and substitutes are weak. It tends to fail when the market is saturated, the resource is already a small share of the cost stack, or behavioural and regulatory constraints cap usage.
The Khazzoom-Brookes postulate is the formal macroeconomic version: at the economy-wide level, energy efficiency improvements tend to increase rather than decrease total energy consumption. The empirical strength of this claim is contested.
Rebound vs. Jevons Proper
These are different magnitudes and the distinction matters.
- Direct rebound: consumption rises but by less than the efficiency gain. Net consumption still falls. Most empirical estimates of energy efficiency programs sit here, with rebound effects of 10–30%.
- Jevons proper (backfire): consumption rises more than the efficiency gain. Net consumption increases. This is the strong claim and the rarer outcome.
Conflating partial rebound with full Jevons is sloppy. A 20% rebound still leaves 80% of intended savings on the table.
Common Misuse
The paradox is regularly weaponised to dismiss any efficiency intervention — fuel standards, insulation programs, AI compute optimisation. The dismissal usually skips the elasticity question entirely and assumes the rebound exceeds 100% by default. It rarely does.
The opposite error — assuming efficiency gains translate one-for-one into reduced consumption — is more common and equally wrong. Both errors come from refusing to model the demand response. The serious question is always: how elastic is demand for the service this resource provides, and what saturation point is the market near?
Related
- Goodhart’s Law — both describe an intervention that backfires because it changes the underlying incentive structure rather than just the measured outcome
- Chesterton’s Fence — the “Common Misuse” section’s warning against dismissing efficiency measures without modeling elasticity mirrors Chesterton’s Fence’s warning against removing a structure before understanding why it exists
- AI Will Not Replace Workers - Argument Map
Fact-Check Notes
No factual issues found. The Jevons quote and its attribution to Chapter VII (“Of the Economy of Fuel”) of The Coal Question (1865) were verified directly against the source text. The Watt/Newcomen “roughly a third” coal-consumption figure is within the range of commonly cited historical estimates (sources vary from ~25% to ~50% of prior consumption depending on which engines/years are compared), so it is not flagged as an error. The Khazzoom-Brookes postulate is real and the note that its empirical strength is contested is accurate and already appropriately hedged in the text. All three outgoing links resolve to existing files; two (Goodhart’s Law, Chesterton’s Fence) had non-obvious connections and now carry inline justifications above.