A negative income tax for land

An introduction to the mechanism at the heart of Groundshare: Milton Friedman's negative-income-tax design, applied not to what people earn but to how much of the nation's ground they hold — and why that structure does what neither the traditional Georgist campaign nor a tax-funded basic income has managed to do. Canonical figures: NUMBERS.md; genealogy: analysis/prior-art.md; political record: analysis/political-viability.md.

1. Friedman's machine, briefly

In 1962 Milton Friedman proposed replacing the welfare state's maze with one instrument: pick an income threshold; above it you pay tax, below it the tax runs backwards and pays you. One schedule, no programs, no caseworkers, no cliff between "recipient" and "taxpayer" — everyone stands on the same line, and where you stand determines whether money flows in or out.

The negative income tax never passed under its own name, but its skeleton is load-bearing across the modern transfer state: the EITC in the United States, and in Canada the GST credit and the Canada Child Benefit — cheques computed against a threshold, delivered through tax machinery, no stigma attached.

It also has a known flaw, and the flaw is the interesting part. An NIT on income must tread carefully around incentives: every dollar of support withdrawn as you earn acts like a tax on working, so the schedule needs a carefully chosen slope, and arguments about that slope never end.

2. Now point it at land instead

Take Friedman's machine and change one input. Not what you earn — what you hold of the nation's ground.

• The threshold is the equal share: Canada's assessed land value divided by its population — about $164,000 per resident on the net basis (≈$655,000 for a family of four).
• The schedule is one straight line through that point, at a modest annual rate on the difference.
• Hold less land value than your household's combined shares, and the line runs negative: you're paid. A household holding no land receives the full flow — about $9,000 per person per year.
• Hold more, and you pay — on the excess only, at the same rate. The median household never crosses the line: an average-size household starts paying only above ≈$456,000 of land held, roughly an $870,000 property.

Formally: each household's settlement is T = r × (L_held − n·share) — the same expression as Friedman's NIT with land value in place of income. The sum of every T in the country is zero by construction. The government keeps nothing; the payments from over-holders are the payments to under-holders.

And here the known flaw vanishes. Land is the one base an NIT can't distort, because land cannot respond: it can't work fewer hours, move offshore, hide in a numbered company's basement, or quit in discouragement. The incentive-slope debate that has dogged the income NIT for sixty years has no purchase here — which is why the land version needs no kink, no phase-out, no taper design at all. One rate, one line. Friedman himself called the land value tax "the least bad tax"; a land NIT is his own transfer design mounted on his own favourite base.

There is one upgrade beyond the mechanical. Friedman's income threshold is a policy choice — a poverty line someone picked, defensible and forever debatable. The land threshold is a measurement: total ground divided by everyone with an equal claim to it. Nobody sets it; it's computed. The moral argument happens once — does the ground belong to all of us equally? — and after that the schedule contains no further knobs for a legislature to fight over except the rate.

3. Why not the traditional Georgist route?

The classic Georgist program — tax land values, use the revenue to replace taxes on work and buildings — is better economics than almost anything else a century of economists has agreed on. It also has one of the worst political records of any policy ever seriously proposed. Pittsburgh's graded tax died in 2001; Altoona's full LVT lasted five years; the UK enacted land taxes twice (1910, 1931) and repealed them both without meaningful collection; Vancouver itself was the world's celebrated single-tax city until the 1913 bust, then spent decades retreating. (The full record: analysis/political-viability.md.)

The pattern isn't bad luck. The traditional design presents every landholder with a bill and nobody with a cheque. Its losers are concentrated, visible, and organized; its winners get lower other taxes — diffuse, invisible, and grateful to no one. It creates no constituency, and its revenue goes into a treasury, where every spending fight becomes the tax's fight too.

The NIT structure inverts every one of those liabilities:

• The majority sees only a cheque. Roughly two-thirds of households are under their share — the average renter household receives ≈$21,400/yr, and even the average owner-occupier household receives ≈$3,700/yr. A household below its share never sees a charge, a bill, or a form — there is no gross tax anywhere in its life.
• There is no revenue. ΣT = 0. Nothing flows to government, so there is no pot to accuse anyone of misspending, no program to defund, no annual budget fight to lose. The classic attack on the single tax — "and then the state controls all that money" — has no target.
• The constituency is built in. Alaska's Permanent Fund Dividend has survived four decades untouchable for exactly this reason: every resident gets a payment with their name on it, every year. The traditional LVT has no Alaska; the land NIT is structured as one.

What's traded away is real and should be said plainly: the Georgist program raises revenue to untax work, and Groundshare deliberately doesn't — it settles title instead. Hybrids exist (a rate above the settlement rate, with the excess replacing income tax), but the lead design buys moral clarity and a constituency at the price of raising nothing.

4. Why not tax-and-dividend — the UBI route?

The modern proposal closest to this one is the "universal land dividend": levy a land value tax on everyone, pay everyone an equal dividend from the proceeds. Arithmetically, it is identical to the land NIT — the same line, drawn as two flows instead of one. Saez and Zucman have shown formally that a proportional levy plus equal demogrant and a threshold negative tax are the same object.

Canada just ran the definitive experiment on why the drawing matters. The carbon fee-and-dividend was a textbook two-flow design, and on the Parliamentary Budget Officer's direct-cost numbers most households received more in rebate than they paid. It died anyway — killed in three words — because the two flows are experienced completely differently: the fee was felt at every fill-up; the rebate was an unlabelled quarterly deposit nobody connected to it. The gross flow was visible and the net was a modelled abstraction, and the visible number votes.

The netted design closes that gap structurally, not rhetorically:

• The net number is the only number. No gross levy exists for anyone under their share — there is nothing to photograph at the pump. For payers, one bill on the excess; for everyone else, one cheque. The felt number and the true number are the same number.
• The cheque is property income, not a rebate. A UBI is support; a dividend from a tax is your own money partially returned. The settlement payment is rent — income from a share of the country you already hold. That is the Alaska frame, the only frame with four decades of political survival, and it changes the clawback politics too: settled in design, receipts here are non-taxable and excluded from income tests (analysis/dividend-treatment.md), the way the Canada Child Benefit already works.
• No trillion-dollar churn. The standard objection to UBI — "it costs $400 billion a year" — is a gross-flow illusion, but under tax-and-dividend the illusion is true on paper: the treasury really does collect and disburse the full pool. Netting deletes the churn. The whole scheme's cash movement is the sum of net positions, a fraction of the gross, and no government account ever holds the pool.
• And unlike a general-revenue UBI, nothing here touches work. A UBI funded from income or consumption taxes gives with one hand and takes from paycheques with the other. The land NIT's entire base is value nobody produced. What you earn, build, and save never enters the formula.

5. The same line, three drawings

6. What this page is not claiming

Not that the arithmetic is new — it's Friedman's, and the equivalence to tax-plus-dividend is a theorem, not a secret. Not that the idea of sharing land value is new — that claim is 250 years old and the genealogy is published (analysis/prior-art.md). And not that presentation abolishes opposition: a median Vancouver detached owner pays ≈$49,000 a year under this schedule and will fight it under any name, the settlement inherits assessment risk in full, and a hostile campaign will call it a home tax regardless. The claim is narrower and, we think, sturdier: of all the ways to draw the same line, this is the only one whose visible objects — a stated share you own, one net payment — match the design's actual moral content, and the only one structurally shaped like the single scheme of its kind that has ever survived contact with voters.

Sources

1. Milton Friedman, Capitalism and Freedom (1962), ch. 12 (the negative income tax); his "least bad tax" remark on land value taxation.
2. Saez & Zucman on the equivalence of linear taxes-plus-demogrant and threshold negative taxes (see analysis/prior-art.md, §4).
3. Parliamentary Budget Officer distributional analyses of the federal fuel charge; Department of Finance, "Removing the consumer carbon price, effective April 1, 2025" — full citations in analysis/political-viability.md.
4. Alaska Permanent Fund Dividend history — analysis/political-viability.md.
5. All dollar figures: NUMBERS.md (net basis, wave 17 onward); incidence from analysis/household-incidence.md; land-tax campaign record from analysis/political-viability.md.

Groundshare — a proposal in open development. Every number traces to a cited public source with its retrieval date; corrections are published, not erased. Rebuilt 2026-08-02 from the repo's research files.