Prices and the Two Economies#

← Part IX · Index

Thesis states it in one line: deflation in anything cognition-intensive; inflation in energy, land, healthcare, and skilled trades. This page works out what that actually does, because the consequences are larger than the observation.

The split, and why it is not ordinary relative-price change#

Relative prices always move. What is unusual here is that the split falls along a single clean axis - whether the good's cost is dominated by cognition or by something intelligence cannot manufacture - and that the axis cuts across nearly every consumption category simultaneously.

This is Game 3 showing up in the price index. Cognition approaches free; the inelastic complements do not; the ratio between them moves by orders of magnitude over a decade.

The measurement problem comes first#

Before the policy problem, there is a measurement problem, and it is severe enough to distort everything downstream.

Quality-adjusted deflation is largely invisible in official statistics. When a service becomes 10× better at the same price, or free at a lower quality that most users prefer, price indices capture almost none of it. This is the standard hedonic problem and AI makes it acute: a large share of the consumer surplus Game 3 predicts lands in unpriced or mispriced form - free tiers, bundled features, quality improvements, and time savings that never touch a transaction.

Three consequences:

1. Measured inflation overstates true inflation, because it captures the physical goods getting more expensive and misses the cognitive goods getting radically cheaper or better.
2. Measured real growth understates true growth, for the same reason. This partly explains why the J-curve may look flat even in a world where welfare is rising quickly.
3. Policy is set on the mismeasured series. Central banks respond to the index they have.

The transition may be considerably better for welfare than the statistics will ever show, and considerably worse for the politics, because people experience the inflating half directly and the deflating half as ambient. Nobody notices that analysis got free; everyone notices the electricity bill. → Energy

The policy problem: one instrument, two economies#

A central bank has one policy rate and an index averaging two opposite movements.

If cognitive deflation and physical inflation roughly cancel, headline inflation looks fine and the policy stance is simultaneously too tight for the deflating half and too loose for the inflating half. Neither sector gets appropriate policy; the average is meaningless.

Worse, the inflating half is exactly where policy is least effective. Raising rates does not build transmission lines, train electricians, or add housing supply - it mostly suppresses the investment that would eventually relieve those constraints. Monetary policy against supply-constrained inflation is close to counterproductive, and the 2020s already demonstrated this at smaller scale.

by the early 2030s, a live and serious debate about whether headline CPI is fit for purpose as a policy target, with proposals to target a services-and-shelter subindex or to exclude technology-driven deflation. ~50% that a major central bank formally revises its target definition or its stated tolerance band by 2035.

The distributional shape is what makes it political#

Low-income households spend disproportionately on the inflating basket - energy, housing, food, healthcare, transport. High-income households consume proportionally more of the deflating basket - software, media, education, professional services, discretionary information goods.

So the price split is regressive on its face, even though the aggregate welfare effect is positive and possibly large. The gains are real, widely distributed, and land in a form that is easy to under-notice; the losses are concentrated in the most salient monthly expenses.

This is the defining political-economy fact of the 2030s and it is more consequential than the employment story. Labor displacement is concentrated in specific cohorts and occupations; the price split touches every household every month.

The likely political responses follow directly, and are worth predicting because they will be described as reactions to AI when they are reactions to prices:

• Energy-price interventions - subsidies, caps, and large-load tariffs. → Energy
• Housing supply politics intensifying, since land is the purest inelastic complement there is
• Pressure on healthcare and education costs, being the two categories that already resisted every previous technology-driven deflation

How the split could fail to arrive#

The inflating half is inflating given current supply institutions, and that conditionality is the claim's main exposure. If permitting reform clears the energy build and housing liberalization spreads, the physical basket's inflation moderates and the split narrows to something an index can average without distortion - the benign failure. The malign failure is subtler: if cognitive deflation stalls because the complements inside services reassert themselves - the licensed body in the room, the liability holder, the distribution chokepoint - then measured services prices never fall much and the two economies quietly collapse back into one. The split is a prediction about institutions as much as technology, which is why it scores on the relative series B7 fixes, never on either half alone.

There is also a self-referential wrinkle worth naming: the inflating basket's service components are wage-driven, and rising trades and care wages are simultaneously this page's inflation problem and Assets' distributional compensation. The same number is the disease in one chapter and the cure in another - which is not a contradiction, but it does mean policy that suppresses the inflating half suppresses the one channel through which labor keeps a complement.

Indexation: where mismeasurement becomes money#

The measurement problem stops being academic at the point of indexation. Pensions, benefits, tax brackets, and wage agreements are indexed to headline CPI. For households that consume a meaningful share of the deflating basket, headline CPI overstates their experienced cost-of-living change, and indexed transfers quietly overcompensate - compounding the fiscal squeeze from the expenditure side. But the error reverses across the income distribution: for low-income households whose basket is dominated by the inflating half, the same index understates experienced inflation, and indexation undercompensates exactly the group the split hits hardest. One index, indexing two populations with opposite errors. Expect the fight over the measurement standard to arrive alongside the target-definition debate above, and for it to be conducted as a technical dispute while being a distributional one.

Baumol, inverted and weaponized#

The classic Baumol story: sectors with no productivity growth get relatively more expensive as wages rise economy-wide, which is why a haircut and a string quartet cost more each decade.

AI inverts the mechanism without changing the math. Cognitive sectors get an enormous productivity shock; everything else is a Baumol sector by comparison. And because the shock is so large, the relative-price divergence is not the gradual drift of the classic case but a step change compressed into a decade.

The classic case took a century and people barely noticed. This version does the same thing in ten years, which is precisely why it becomes political rather than merely interesting.

Headline CPI can hide both halves#

Uniform headline inflation or deflation is the wrong score. Software and info services can fall while shelter, energy, and care rise - headline looks "stable" while B7 is firing. Central banks that target headline or core will face a politics of relative prices that their mandate does not name. That is the register claim about monetary policy discomfort in the 2030s: not necessarily higher average inflation, but a split basket that makes every rate decision look like taking a side.

If Uncertainty 7 erodes the physical basket's scarcity, the split narrows late - a dated half-life, not a falsification of early-window B7.

Eight quarters is the bar for a reason. One energy spike or software discount season is noise. Require sustained relative divergence before rewriting politics or assets claims.

Observable: B7 - relative CPI/PCE component divergence between the cognitive and physical baskets. The central-bank prediction above is tracked in the register.

Related: Game 3 - Firms · Energy · Assets · 2028–2032 · Fiscal

View markdown source