Sunil Venkataram
Back to garden
•
🌰 seedling
Demand Elasticity Determines Whether Automation Creates or Destroys Jobs
type/noteailaboreconomics

Demand Elasticity Determines Whether Automation Creates or Destroys Jobs

The logic

A firm has 10 workers, each producing one unit per day. AI makes each worker 10x more productive. Costs fall, prices fall. What happens next depends entirely on consumers.

Elastic demand (elasticity > 1): The price drop causes a more-than-proportional increase in quantity demanded. Output expands so much that the firm needs more workers to meet it, even though each worker is 10x more productive. Net effect: more hiring.

Inelastic demand (elasticity < 1): The price drop causes a less-than-proportional increase in quantity demanded. The firm produces roughly the same output with fewer workers. Net effect: displacement.

The calculator example makes it concrete. If each worker can now make 10 calculators instead of one, but consumers still want roughly the same number of calculators, the firm fires 9 of the 10 workers. But if cheaper calculators unlock massive new demand (schools, developing countries, niche applications), the firm hires more workers to meet it.

Connection to Jevons paradox

Jevons paradox (efficiency increases total consumption of a resource rather than decreasing it) is the elastic-demand case. When steam engines made coal more efficient, coal consumption skyrocketed because new applications became viable. The same logic applies to labor: if AI makes a worker far more productive and demand for that product is elastic, the sector ends up with more workers.

The existing note on Jevons Paradox vs Cognitive Displacement - The Unresolved Tension frames this as an unresolved empirical question. The O-ring automation framework from Gans and Goldfarb adds precision: it identifies demand elasticity as the specific variable that determines which side wins in each sector. The resolution will be sector-specific, not economy-wide.

Which sectors are elastic vs. inelastic

This is the practical question. Some rough predictions from the Imas article:

Likely elastic (automation creates jobs):

  • Software development. Cheaper code production unlocks applications that were previously too expensive to build. The TAM of software is far from saturated.
  • Legal services at the consumer tier. Most people who need legal help cannot afford it. If AI drops the cost of basic legal work by 10x, millions of new clients appear.
  • Consulting. Better output at lower prices can expand the client base beyond Fortune 500 companies into mid-market and small business.

Likely inelastic (automation destroys jobs):

  • Trucking. Demand for freight is tied to economic output, not to the cost of trucking itself. Cheaper trucking does not cause people to ship 10x more goods.
  • Warehousing. Similar story. Throughput demand is downstream of retail volume, not warehouse labor costs.
  • Routine financial analysis. The number of quarterly earnings reports to analyze does not grow because analysis gets cheaper.

The hard cases are sectors where elasticity is genuinely uncertain: healthcare, education, creative work. These could go either way depending on regulatory, cultural, and pricing dynamics that are hard to predict in advance.

Why this framework is useful

Most AI-and-labor analysis treats “automation” as a single phenomenon with a single direction. The demand elasticity lens says the outcome is sector-specific and predictable in advance if you can estimate the elasticity. Two sectors with identical levels of AI capability can have opposite employment outcomes.

This also explains why historical analogies are unreliable without specifying the demand structure. ATMs and bank tellers (elastic: more branches opened), manufacturing robots and factory workers (inelastic in mature markets: fewer workers), spreadsheets and accountants (elastic: more businesses could afford accounting). The technology is not what matters. The demand curve is.

Backlinks (4)
Connected Notes