# Fiscal - a tax base built on labor income

← [Part IX](README.md) · [Index](../README.md)

---

The slowest-moving item in this document and possibly the most consequential, because it constrains the state's ability to respond to everything else in it. → [State capacity](../03-domains/contested/state-capacity.md)

**Advanced-economy tax systems collect predominantly from labor income.** Payroll taxes, income taxes, and the social insurance contributions attached to them are the backbone. That base was chosen because labor income was large, stable, visible, and hard to move.

Every one of those properties is weakening at once.

## The arithmetic of the squeeze

Three pressures, all arriving in the same window and all pointing the same way:

**1. The base narrows where it is easiest to collect.** If cognitive labor income falls as a share of national income while capital income rises, the tax base shifts toward the category that is mobile, deferrable, realizable at the holder's discretion, and internationally arbitraged. **Capital income is harder to tax not because of rates but because of timing and location**, and no amount of rate-setting fixes that.

**2. Demand for transfers rises exactly when the base narrows.** Displacement, retraining, and the [apprenticeship-gap](../06-uncertainties/apprenticeship-gap.md) response all cost money, and they come due during the same period. Countercyclical demand against a structurally weakening base is the classic fiscal trap.

**3. Demographics are already doing this independently.** Ageing populations shrink the working-age tax base and expand pension and healthcare obligations regardless of AI. **AI does not create this problem; it arrives on top of it, and the two compound.** Most fiscal projections for the 2030s already show strain before any AI assumption is added.

## How the squeeze claim could be wrong

The squeeze is a story about shares, and governments budget in levels. If AI raises total income enough, labor's *share* of the tax base can fall while labor-tax *receipts* still grow in absolute terms - in which case treasuries feel nothing on the horizon that matters politically, and the adaptation this page predicts gets deferred a decade. That is the main way to be wrong here, and it is the benign way: the problem would still be real, just slower than the transfer demand it races against.

The second failure mode is compositional. The squeeze assumes displaced cognitive labor income is not replaced by other taxable labor income. If [Game 4](../02-games/4-labor.md)'s reallocation toward care and trades runs faster than its displacement - and [demography](demography.md) gives reasons it might - the labor base rotates rather than shrinks, and payroll receipts hold up even as their occupational composition transforms underneath. **Watch receipts by income type, not employment counts:** [B10](../07-indicators/diffusion/labor.md) is defined on the revenue side for exactly this reason.

## The intersection with rates

[Rates](rates.md) argued that a productivity step-up raises the real neutral rate. Apply that to a large stock of public debt and the interest burden rises with it.

This produces a genuinely awkward configuration:

> **The scenario where AI works best for growth is also the scenario where debt service costs most.** Higher productivity means higher rates means higher interest expense on existing debt, and the growth dividend has to exceed that before it is fiscally net positive.

It usually does - growth raises the denominator faster than rates raise the numerator, for plausible parameters. But **the timing is wrong**: rates reprice immediately and the growth dividend arrives over a decade. The interval in between is fiscally uncomfortable, and it coincides with peak demand for transition support.

## The level-of-government asymmetry

The squeeze is not uniform across the state. Central governments collect from labor and capital income and carry the transfer obligations - they get the worst of both sides. But **local governments in most federal systems collect from property, and property sits on the winning side of [Assets](assets.md):** land appreciation fattens the property-tax base exactly as the income-tax base thins. The plausible result is a decade in which municipalities hosting datacenters and appreciating land are flush while national treasuries strain - an inversion of the usual fiscal-federalism pattern, and one that shapes where infrastructure gets welcomed. The datacenter-siting politics in [Energy](../01-substrate/energy.md) are partly this: the level of government granting the permit is the level capturing the tax, while the level bearing the transfer burden has no seat at the siting table.

## What actually gets proposed, and what actually happens

| Proposal | Assessment |
|---|---|
| **Robot / automation taxes** | Popular, incoherent. Taxing capital equipment discourages exactly the investment that raises productivity, and "robot" is undefinable in tax law. **Very unlikely to survive drafting**, likely to be proposed repeatedly. |
| **Higher capital income and corporate taxation** | The obvious move; constrained by mobility and by tax competition. Requires multilateral coordination, which is why the OECD minimum-tax effort is the relevant precedent - slow, partial, and real. |
| **Land value taxation** | The textbook answer to the exact configuration in [Assets](assets.md): it taxes an inelastic complement, cannot be avoided by relocation, and does not distort supply. Also the textbook answer for 150 years without adoption. Political economy, not economics, is the obstacle. |
| **Consumption taxation** | The path of least resistance and the most likely default. Broad, hard to avoid, administratively simple - and **regressive**, which compounds the distributional problem in [Prices](prices.md) rather than offsetting it. |
| **Compute or energy levies** | Interesting because the base is physical, immobile, and metered - genuinely hard to avoid. Directly conflicts with industrial-policy goals, which is why it has not happened. Watch for it to appear as a **ratepayer-protection measure** rather than as a tax. → [Energy](../01-substrate/energy.md) |

> **Prediction:** the actual response through 2035 is **incremental consumption-tax expansion plus partial capital-income tightening**, not any structural redesign. Base rate on tax systems being fundamentally redesigned outside a war or a fiscal crisis is very low, and nothing here forces one within the window.

## UBI, honestly

Universal basic income dominates public discussion of this topic and deserves a direct assessment rather than an omission.

**The case for:** it is administratively simple, it directly addresses income loss without requiring the state to predict which jobs disappear, and it is robust to being wrong about the details of the transition.

**The case against, on this document's analysis:** it is expensive precisely when the base is weakest, and - more fundamentally - **it addresses the wrong problem.** [Game 4](../02-games/4-labor.md)'s conclusion is that the sharpest harm is the destruction of the novice-to-expert pathway, not mass unemployment. A transfer does not rebuild an apprenticeship system. It substitutes for income while leaving the human-capital formation problem entirely intact, which means the *next* cohort faces it worse.

> **Assessment:** meaningful UBI in a major economy by 2035 is **~15%** likely. Far more likely, and more useful: expanded conditional transfers, wage insurance, and subsidized training - none of which are new instruments, all of which are politically achievable, and all of which target the actual mechanism better.

**The higher-leverage fiscal intervention is on the training side, not the transfer side** - subsidizing the junior-hiring externality directly, since [Game 4](../02-games/4-labor.md) identifies it as a commons tragedy where firms pay full cost and capture a fraction of the benefit. That is a textbook case for a Pigouvian subsidy, it is cheap relative to UBI, and almost nobody is proposing it.

### State capacity is the binding constraint on every fiscal fix

Instruments above assume a state that can design, pass, and administer them. [State capacity](../03-domains/contested/state-capacity.md) is uneven and, in several high-income jurisdictions, thinning on technical hiring just as the measurement and enforcement load rises. A correct fiscal diagnosis with a state that cannot run the program is not an operational plan. Prefer interventions that use existing administrative rails (payroll tax credits for junior headcount, expanded apprenticeships through existing guilds, land-value components inside existing property tax) over greenfield agencies. Complexity is not free when [B10](../07-indicators/diffusion/labor.md) is already warning that the base is under pressure.

**Demography drives half of this page even if AI TFP misses.** Aging raises transfers and care spend on a path already locked in. AI multiplies the squeeze; it did not invent it. → [Demography](demography.md)

---

**Related:** [Assets](assets.md) · [Prices](prices.md) · [Rates](rates.md) · [Game 4 - Labor](../02-games/4-labor.md) · [State capacity](../03-domains/contested/state-capacity.md) · [Uncertainty 3](../06-uncertainties/apprenticeship-gap.md) · [B10 - tax-base indicator](../07-indicators/diffusion/labor.md)

**Observable:** [B10](../07-indicators/diffusion/labor.md) - labor-related tax share of total revenue vs transfer and debt-service outlays.

**Next:** [Index](../README.md) · [Protocol](../RESEARCH.md)