What Is Demand Elasticity?
Demand elasticity tells you how strongly customers respond when something changes. Usually that change is price. If the price falls and people buy much more, demand is elastic. If buying barely changes, demand is inelastic.
The idea matters well beyond an economics class. It is one reason a claim that “AI made this work cheaper” does not tell you what happens next. Lower cost per unit, higher total demand, more output, and more jobs are four different claims. They need four different kinds of evidence.
The basic idea
Price elasticity of demand is the percentage change in quantity demanded divided by the percentage change in price. Economists usually compare the absolute size of those changes because price and quantity demanded normally move in opposite directions.
Demand is elastic when quantity changes proportionally more than price. It is inelastic when quantity changes less. If both change by the same percentage, demand is unit elastic.
Those labels only make sense for a defined product or service, market, customer group, comparison, and period. Demand for tax preparation this year is not the same market as demand for all professional services over the next decade.
A simple example
Suppose a service cuts its price by 10 percent and the number of subscriptions sold rises by 15 percent. Using those percentage changes, the absolute elasticity is 1.5. Demand is elastic because the quantity response is larger than the price change.
That result describes this service, for these customers, during the period measured. It does not establish the elasticity of the whole industry. Change the market, timeframe, available substitutes, quality, or customer income, and the result may change.
Why demand elasticity matters for AI
AI can reduce the labour, compute, or time required to produce one unit of work. But customers may never see a lower price. A business might keep the saving as margin, improve quality, shorten waiting time, or increase capacity instead.
Even if the customer-facing price falls, demand still has to respond. Economist Erik Brynjolfsson describes the mechanism this way: if AI lowers labour and price per unit but demand expands only slightly, labour demand may fall. If lower prices unlock much more demand, output and possibly employment may rise.
That is economic reasoning, not a general jobs forecast. More output does not prove more employment, better wages, or a fair distribution of the gains. Likewise, AI exposure tells us which tasks a system may affect; it does not tell us how employers or customers will respond.
There is also a possible rebound effect: making something more efficient can increase total use. Shane Greenstein’s analysis of AI and Jevons’ paradox and work by Luccioni, Strubell, and Crawford on AI rebound effects explain why that possibility deserves attention. But rebound is conditional. Saturation, substitutes, supply limits, and other bottlenecks can stop demand from growing enough. Our explainer on Jevons’ paradox goes deeper into that distinction.
Most important: the reviewed research does not establish one economy-wide demand-elasticity estimate for AI-assisted output.
Test an AI productivity claim
When someone says AI made work cheaper or more productive, ask:
- What specific output became cheaper, faster, better, or more available?
- Did customers actually see that change?
- What happened to quantity demanded, in which market, and over what period?
- Did total output rise more than labour per unit fell?
- What happened to total hours, headcount, wages, quality, and review work?
A task map can help define the work and output before you measure the result. If the claim cannot answer these questions, it is incomplete.
Bottom line
Demand elasticity measures response. It helps explain why the same productivity improvement can lead to different outcomes in different markets. It cannot predict jobs by itself.
The practical rule is simple: define the output and market, measure what changed for customers, track demand and output, and then measure labour outcomes separately.