Sunil Venkataram
Back to garden
•
🌰 seedling
Structural Change Theory Predicts AI Grows the Human-Intensive Economy
type/noteaieconomicslabor

Structural Change Theory Predicts AI Grows the Human-Intensive Economy

The historical pattern

Every major productivity shock has followed the same arc. Agriculture employed 40% of Americans in 1900. Mechanization made farmers vastly more productive. Food got cheaper. People did not spend the surplus on more food — they shifted spending to manufacturing, then to services. The automated sector shrank as a share of the economy despite producing more. The less productive sector grew.

Comin, Lashkari, and Mestieri formalized this in a 2021 Econometrica paper. Their core insight: demand is nonhomothetic. As people get richer, they do not buy proportionally more of everything. They shift spending toward sectors with higher income elasticity. Agriculture has low income elasticity (people can only eat so much). Services have high income elasticity (there is always a better restaurant, a more engaging experience, a more attentive doctor).

Their most striking empirical result: income effects account for over 75% of observed structural change patterns. Price effects — the standard story that automated sectors get cheap so people buy other things — explain only about a quarter. The dominant force is straightforward: richer people want fundamentally different things.

— Alex Imas, “What Will Be Scarce”, citing Comin, Lashkari, Mestieri (2021)

Application to AI

If advanced AI dramatically reduces the cost of producing a wide range of goods and services, this framework predicts a structural transformation along familiar lines:

  1. Automated sectors shrink as shares of GDP. Not because they produce less, but because their prices fall and income-driven demand shifts elsewhere.
  2. High-income-elasticity sectors grow. The question is which sectors have high income elasticity in a post-AI world.
  3. Baumol’s cost disease becomes a feature. The sectors that resist automation get relatively more expensive — and that relative expense is what keeps people employed there. The “stagnant” sector absorbs a growing share of spending and jobs because it cannot be automated.

This reframes the automation debate. The relevant question is not “which jobs can AI do?” but “what do people spend money on when commodities get cheap?” Historical data and current expenditure patterns both point toward the same answer: goods and services where human involvement is part of the value.

The expenditure evidence

BLS Consumer Expenditure Survey data shows the pattern already at work. Top-quintile households spend about 4.3x more than bottom-quintile households overall. But the ratios are far larger in categories with a strong human component — in-person dining, entertainment, education, personal services. Rich households do not just buy more stuff. They shift spending toward goods where experience, relationship, and social meaning matter more.

Hubmer (2023, Econometrica) confirms this with household-level data across the full universe of consumer spending: higher-income households spend relatively more on labor-intensive goods and services as a share of total consumption. Economic growth itself tilts demand toward sectors with higher labor content, even while other technological forces push the aggregate labor share downward.

— Hubmer, “The Race Between Preferences and Technology”

Why this differs from the displacement thesis

The cognitive displacement argument (see Jevons Paradox vs Cognitive Displacement - The Unresolved Tension) treats the economy as a fixed set of jobs: if machines can do them cheaper, the jobs disappear. Structural change theory treats the economy as a shifting composition of demand: as people get richer, they want different things, and those different things create new sectors and new employment.

Both can be partially right. Labor’s aggregate share may fall. But structural change predicts that the relational, human-intensive sector remains a substantial part of the economy — not because automation fails, but because automation’s success redirects demand toward what it cannot replace.

Key distinction

This is not a claim that labor’s aggregate share must rise. It may well shrink. The claim is about sectoral reallocation in rich economies: as AI makes commodity production cheap, spending and employment shift toward sectors where human involvement still carries value. Labor share can fall and the relational sector can still grow as a share of GDP.

Backlinks (5)
Connected Notes