Every Spending Bill Should Show Up on Your Paycheck

When I first sketched how to pay for the transition, I picked a number.

“For ten years, tax every dollar of income above a universal exemption at a single 30% rate, whether it comes from wages, capital gains, dividends, or business profits. No deductions, no credits, no special rates.”

I still like the spirit of it. But a number was the weakest thing I could have picked.

Donella Meadows, in her 1999 essay Leverage Points: Places to Intervene in a System, ranked twelve ways to change a system.

Dead last, the weakest: “Constants, parameters, numbers (such as subsidies, taxes, standards).” Much higher up: who can see what’s happening, and the rules of the game.

A tax rate is a number. Congress can change it next year, and nobody will connect the change to anything.

So here’s the update. A flat-tax white paper, One Rate, One Formula, One Direction, makes a case I find convincing: don’t set the rate, set the rule that sets the rate. I’m adopting its formula, with changes, and this piece walks through both.

As a reminder, the whole series in one line: universal basic goods and services, powered by universal machine labor on demand, distributed by universal basic capital.

Taxes are how we pay for the transition to that. They should be built like the rest of it: simple, visible, and pointed in one direction.

Part 17: Keep Money, Add a Second Ledger kept money for whatever stays scarce, and added a ledger for physical things beside it. This piece is the money side: how the tax rate gets set while the floor is being built.

The first sketch, and why it changed

I published the first sketch alongside Part 7: We’re Less Than a Decade From the Next Economic System. It had four steps.

A decade of one rate, no loopholes. For ten years, one 30% rate on every dollar of income above a universal exemption. For the largest fortunes, a minimum tax on unrealized gains, so that “buy, borrow, die” (holding assets forever, borrowing against them instead of selling, and passing them on untaxed) stops working.
Reinvest the growth. As the deficit closes, put the growth dividend into community investment funds, loosely modeled on Norway’s sovereign wealth fund, that build community nodes and then gift them back.
Shift the base from labor to rents. Phase down payroll taxes, which effectively charge a fee for hiring a human. Replace them with taxes on things that stay scarce no matter how productive machines get: land values, natural resources, and permits for scarce compute, with land taxed by states and cities, where the Constitution allows it.
Let it shrink. As universal basic goods get produced locally and provided in kind, the money economy for essentials shrinks, and so does the need to tax it. What’s left is a small tax base on whatever stays scarce: land, luxury goods, and positional goods. At that point, the safety net isn’t a program funded by taxes. It’s the floor itself.

The case for Step 1 was the money on the table. The Joint Committee on Taxation projects that tax expenditures, the deductions, credits, and special rates in today’s code, will cost $2.3 trillion in fiscal 2026. That’s more than the entire deficit.

You can’t simply add those estimates up, because the provisions interact. But a clean base could plausibly close the deficit, and a decade of it could stop the debt from growing. Growth would pay the debt down from there: taxes stop the bleeding, and growth heals the wound.

Here’s what changed, and where each change lives below:

Step 1 became a formula. A rule sets the rate, capital pays more than wages, gains are taxed when sold or at death, and the exemption became a monthly cash floor.
Step 2 stays, and grows. Permit auctions and frontier profits go into the same community funds.
Step 3 stays, with a limit. The formula’s one rate replaces the payroll tax outright, but rents can’t carry the whole base.
Step 4 stays as the end state.

The formula

Every year, the rate is set by statute:

Rate = spending ÷ tax base + a debt surcharge.

That’s it. If Congress spends more, the rate rises the following year. If it spends less, everyone’s rate falls.

The surcharge is fixed in law and steps down as the debt shrinks relative to the economy:

8 points while debt held by the public is above 90% of GDP
6 points above 60%
4 points above 30%
2 points above zero
Then zero

Every dollar the surcharge raises is legally dedicated to paying down principal. Changing the formula, the schedule or the base takes a three-fifths vote. Spending bills still pass by simple majority, but their cost lands on the rate.

Put differently: every appropriation shows up on next year’s paycheck.

That’s the part Meadows would like. Under information flows, she wrote that “Missing feedback is one of the most common causes of system malfunction.”

Her example, as “the story goes”: a subdivision of identical houses, some with the electric meter in the basement and some with it in the front hall, where people could watch it spin. Same prices, no other change. The front-hall houses used 30% less electricity.

Today, federal spending has its meter in the basement. The formula moves the meter to the front hall. And it changes the rules of the game, which rank much higher on her list than any number.

Switzerland’s federal debt brake, approved by 85% of voters in December 2001 and first applied to the 2003 budget, ties the government’s spending ceiling to its receipts.

The Swiss set spending from revenue. This formula sets the rate from spending. Same loop, run in the other direction.

The honest math on the debt

The first sketch said a decade of one rate could stop the debt from growing, but couldn’t pay it off on its own. The white paper says its formula clears the public debt in 13 to 18 years.

Both can be true, but only on the white paper’s stated assumptions: a tax base of about $28 trillion growing 4% a year, spending reforms that take federal spending from about 22% of GDP to about 16.5%, and the surcharge doing the rest.

The starting point, as of October 1: debt held by the public is $32.4 trillion (total federal debt is $40.3 trillion, counting what the government owes its own trust funds).

Fiscal 2026 has closed. The Congressional Budget Office counts outlays of $7.4 trillion against receipts of $5.4 trillion, a $2.0 trillion deficit, with debt held by the public about the size of the whole economy.

So I ran the white paper’s numbers myself. Two things fell out.

First, the surplus is just the surcharge. If the rate is spending divided by the base, plus the surcharge, then whatever the formula collects beyond spending is the surcharge times the base.

Eight points of a $28 trillion base is about $2.2 trillion in the first year. Spending reforms don’t speed up the payoff. They lower the rate.

Second, the payoff slows as it works. Applied literally, the stepped schedule takes debt from about 100% of GDP to roughly a third of GDP in ten years. That’s the part the first sketch promised: taxes stop the bleeding.

But as the debt shrinks, so does the surcharge, and paying off the last dollars takes until about year 24, longer if the surcharge pauses.

To actually clear the debt in 13 to 18 years, you’d hold the surcharge at around 6 to 8 points until it’s gone. The white paper names this as a decision, “faster debt payoff vs. higher rates for longer,” and it’s the right way to frame it.

I don’t think the last dollar is the right goal, for two reasons.

First, eight points of surcharge pulls about 7% of GDP out of the economy in the first year. That’s a lot of spending to remove in one go, during a transition that could turn ugly fast.

Second, a country with zero debt has a problem almost nobody plans for. In January 2001, when surpluses looked big enough to retire the debt within a decade, Alan Greenspan warned the Senate that the government would soon run out of debt to buy back and start piling up private assets instead.

His advice: the government “should eschew private asset accumulation.” A federal government that owns stocks is a government with a reason to pick winners.

So here’s my pick: a gentler schedule with a finish line. The surcharge is 4 points while debt is above 60% of GDP, 2 points while it’s above 30%, and zero after that. Thirty percent is about where the debt stood in 2001, when it had fallen to 33.1% of GDP.

On the white paper’s own growth assumptions, that takes about 16 years, inside its 13-to-18-year range, just with a different finish line. The first-year surplus is about $1.1 trillion, close to the paper’s own estimate.

And after the surcharge ends, the formula still balances the budget every year, so the debt keeps shrinking relative to an economy that keeps growing.

Like the first sketch, every year count here would need to be modeled by the Congressional Budget Office before anyone should rely on it.

One more honest number: the rate itself. In year one, the formula comes out at about 30%. Here’s where that comes from:

About 26 points for today’s spending: $7.4 trillion divided by a tax base of about $28 trillion.
About 4 points of debt surcharge, while debt is above 60% of GDP.
Minus about half a point for customs duties, which stay outside the one rate.
Nothing extra yet for the cash floor I’ll get to below. In year one, it pays out what today’s programs already cost, and that’s already in spending.

That’s about 29.8%, blended. Split into the two rates below, labor pays about 27% and capital about 35%.

The floor then phases in. At its full size, about $1.4 trillion a year, it adds up to 5 points, which would put the rate near 34% to 35% if the debt were still above 60% of GDP. The real addition is smaller, because the floor replaces programs that today’s spending already pays for.

That one rate replaces both the income tax and the payroll tax. It’s also an average across every dollar, not a top bracket. For comparison, today’s top income-tax bracket is 37% in 2026, before payroll taxes.

Every point of rate on a $28 trillion base is worth about $280 billion, so how fast the spending reforms land matters a lot. The flat-tax white paper’s “near 30%” assumes they’re mostly in place on day one.

From there, the rate steps down as the economy grows, the essentials move in kind and the surcharge winds down. By the math in my whitepaper, it reaches 19% sometime between the mid-2030s and the mid-2040s.

Four changes to the formula

I’m adopting the formula with four changes, the same four in Section 16 of my whitepaper, The Information Economy.

1. One formula, two rates

The first sketch’s Step 3 said today’s system taxes human labor far more heavily than machines, and that has to flip. A single rate on everything doesn’t flip it. It just stops making it worse.

The economists Daron Acemoglu, Andrea Manera, and Pascual Restrepo found that the US tax system is “biased against labor and in favor of capital”.

Effective taxes on labor run 25.5% to 33.5%, while software and equipment were taxed at about 5% after the 2017 tax law. Their model’s optimal rates are 18.22% on labor and 26.65% on capital. Capital should pay more, by about eight points.

So the formula sets the labor rate, and capital income pays eight points more, written into the same law. That’s their gap, rounded.

Is that too high? There’s a ceiling to respect: above some rate, people simply stop selling assets, and the government collects less. The Penn Wharton Budget Model estimates the revenue-maximizing capital gains rate is 33% today, but 42% if gains are taxed at death instead of being wiped clean for heirs.

So the plan does both: gains are taxed when you sell or when you die, above an exemption. By my whitepaper’s math, the capital rate opens around 35% and would top out near 40% with the floor at full size, under that ceiling either way. It falls from there as the surcharge winds down.

The trade-off is real. Two rates invite people to relabel income.

Finland learned this after its 1993 dual income tax reform cut rates on capital income: economists Jukka Pirttilä and Håkan Selin found that some of the gain in taxable income was probably offset by self-employed people shifting income.

But notice which way the gaming runs. In Finland’s version, capital was taxed less, so income moved toward capital. Here, capital is taxed more, so income moves toward wages: reported on a paystub, withheld every two weeks, the most visible income there is.

An owner who pays herself a bigger salary instead of a dividend still costs some revenue. But it’s revenue you can see.

2. Hiring is never taxed more than automating

This one is a fix to the white paper itself.

Its business base says businesses deduct purchases from other businesses, “not wages or interest.” A robot bought from another company is a purchase from another business, so it’s deductible.

A worker isn’t. The business pays tax on the wage, and then the worker pays tax on the same wage again, even though the paper promises income is “taxed once.”

That’s a tax on hiring, and a subsidy to automating, built into the base. It’s the exact bias Acemoglu, Manera, and Restrepo warned about.

The fix is simple: businesses deduct wages as well as purchases. A worker and a robot are treated alike. Then the human-will rule doesn’t have to fight the tax code.

The original flat tax worked this way. Robert Hall and Alvin Rabushka’s business tax, in The Flat Tax, subtracts “wages, salaries, and pensions paid to workers”.

3. Downturns don’t raise rates

The raw formula has a nasty habit. In a recession, the tax base shrinks, so next year’s rate goes up, right when households and businesses can least afford it.

Economists call that procyclical. I’d call it kicking people when they’re down.

The fix comes straight from the Swiss. Their spending ceiling adjusts receipts by a cyclical factor, defined as “the ratio of trend gross domestic product (trend GDP) to current GDP.” In a boom, it forces a surplus. In a recession, it tolerates a deficit.

So the formula divides by the trend tax base, not last year’s. A bad year doesn’t move the rate. A bad decade still does, which is how it should be.

4. The surcharge pauses on data, not a vote

The surcharge pulls money out of the economy to pay down debt. That’s the point of it. It’s also the danger.

Even my gentler schedule withdraws about $1.1 trillion in its first year, around 3.6% of GDP. The white paper’s own schedule, run as written, withdraws about twice that.

In a normal economy, the debt dividend is worth it. In a demand collapse, it’s exactly backwards.

So the surcharge stops automatically when the dashboard in Plan for the Bad Transition shows a demand shortfall:

Long-term unemployment and median wages both getting worse
Labor’s share of income falling
The employment gap for young workers in AI-exposed jobs widening
Federal receipts coming in below forecast

No vote, no fight over whether it’s “really” a recession. The data trips it, and the data restarts it. A three-year phase-in, which the white paper already includes, helps too.

The cash floor is the progressivity

Here’s the biggest difference between the white paper and my first sketch.

The sketch had a universal exemption, because a flat rate with no exemption would hit the lowest earners hardest. The white paper drops the exemption entirely.

In its place, it consolidates roughly 80 overlapping benefit programs into one cash floor, a single monthly payment that phases out as income rises. In its words, “the floor is the progressivity.”

That’s not a new idea either. It’s a negative income tax. The concept goes back to Juliet Rhys-Williams in the 1940s, and Milton Friedman made it famous in Capitalism and Freedom (1962), where he argued it could replace the whole patchwork of support programs with one.

And it’s the same thing the research says to do when wages start falling.

Google DeepMind economists Julian Jacobs and Alex Imas recommend moving to a negative income tax when data shows “prolonged unemployment spells and falling median wages.” In Plan for the Bad Transition, that’s Switch B: a triggered version that switches on first, from the data, while the floor is being built. The universal floor below is where the cash path ends.

But the order matters. The white paper is blunt about its own plan: “A single rate is regressive relative to today.” So the floor has to be sized before the rate is set, not after.

Here’s my size: $4,800 a year per adult and $1,700 per child, paid monthly, to everyone, and indexed to prices.

Where those numbers come from: they’re 30% of the 2026 federal poverty guideline, which is $15,960 for one person, plus $5,680 for each additional person. At a 30% rate, a person earning up to the poverty line pays no net tax, and below it, they come out ahead.

That’s the first sketch’s universal exemption, paid out monthly. The exemption didn’t disappear. It became a check.

And there’s no separate phase-out. Everyone gets the same payment, and every dollar of income is taxed at the same rate, so the payment nets out gradually as you earn.

No cliffs where earning $1 more costs you $1.50 in lost aid. Friedman’s own 1968 example phased out at 50%; this one phases out at the tax rate itself.

The cost is real. The US has about 340 million people, 73 million of them children, so the floor costs roughly $1.4 trillion a year before counting the programs it replaces, about 5 points of rate.

That’s the price of not raising taxes on the bottom 40%. I’d pay it.

But not all at once. The floor phases in. In year one, it pays out only what today’s roughly 80 consolidated programs already cost, so the rate doesn’t jump.

While the floor is small, Switch B tops people up when wages fall. The floor grows with the tax base each year until it reaches its full size, about 5 points of rate.

One more nice property: the payment is fixed in dollars, and the rate falls over time. When the rate reaches 19%, $4,800 covers everyone’s first $25,000 of income. The floor gets more generous relative to the tax as the debt comes down.

Wait, a check? In Part 6: UBI Won’t Work. Universal Basic Goods Will., I argued against exactly that. I still think a check is the wrong destination. But it’s the right floor while the goods are still being built, and in a demand collapse, it’s the right stabilizer.

So the monthly cash floor is the end state of the cash path, not of the series. It shrinks as each essential moves in kind, leaving a little cash for extras and for whatever stays scarce.

The destination is the lifeline block from Keep Money, Add a Second Ledger: every household’s first block of the seven essentials (food, water, energy, housing, healthcare, transport and communication), plus personal AI and personal machine labor, provided in kind. Every need the lifeline block covers is a need the cash floor no longer has to.

Pay for outcomes, not activity

The white paper’s spending side aims at fewer dollars per outcome, not fewer services. Four reforms I’d keep:

Pay the same price for the same service. Medicare often pays more when a hospital outpatient department does what a doctor’s office does. CRFB estimates that site-neutral payment for all services could save almost $200 billion over a decade.
Pay Medicare Advantage what traditional Medicare would cost. Based on MedPAC’s data, CRFB estimates Medicare Advantage will be overpaid by $1.2 trillion from 2025 to 2034, mostly because plans code patients as sicker than they are and attract healthier ones.
Spend early. James Heckman’s research finds high-quality birth-to-five programs for disadvantaged kids can return 13% a year.
Housing First. As the VA scaled up housing without preconditions, homeless veterans counted on a single night fell 49% between 2009 and 2020.

Realized savings usually come in below the estimates. The white paper admits it: savings “are usually lower than scored.”

This July, CMS proposed site-neutral rates for outpatient imaging only, which it estimated would save $7 billion over a decade. Almost $200 billion on paper, $7 billion in practice. Plan on the smaller number.

What the spending cap costs

After the reforms, the white paper caps spending growth at inflation plus population. That’s what carries its long-run rate from 19% to 15%.

The white paper names its price: Social Security and Medicare benefits would track prices, not wages. Retirees keep their purchasing power, but they don’t automatically share in an AI boom.

I think that’s acceptable, but only for one reason. In this series, sharing in the boom doesn’t come through benefit formulas. It comes through universal basic capital: every resident’s share in the machines that produce their floor.

A retiree who owns a stake in their town’s power, food and compute shares in the boom directly. Without that, I wouldn’t adopt the cap.

Rents, frontier profits, and where the permit money goes

The first sketch’s third step was to shift the base from labor to rents. The research says rents alone can’t carry it.

Philip Trammell and Dwarkesh Patel point out that the share of global income paid for natural resources is “on the order of 5%”.

Land and resources alone can’t replace the income base. So the income base stays, with capital paying more, through the middle of the transition.

The genuinely new rents are compute and robots. AI 2040‘s Plan A imagines permits to build robots or compute “sold to the highest bidder,” with citizens owning shares in the proceeds. I’d take the auction, but the money doesn’t go into general revenue.

It goes into community investment funds that build nodes and compute cooperatives, then gift them back. If frontier profits pay down the formula’s rate, they just disappear into the budget. If they build nodes, they become universal basic capital.

What the white paper gets wrong

A few claims I’d drop or qualify:

“The same figure Hall and Rabushka reached.” The white paper notes that its long-run 19% matches Hall and Rabushka’s famous 19%. That’s a coincidence, not a confirmation. Their 19% matched federal revenue in 1993, on a different base: they set out to “tax consumption,” with a family allowance and business deductions for wages and investment. Different base, different budget, same number by accident.
Compliance cost “falls toward zero.” It would fall a lot. Not to zero. Two rates mean someone has to police the line between wages and capital income, and businesses still have to measure income.
Unrealized gains. My first sketch also proposed a minimum tax on unrealized gains for the largest fortunes. The Supreme Court left that question open in Moore v. United States (2024): “We do not decide that question today.” Taxing realized gains only keeps this plan clear of it. In Part 21: The World We Want Is Mostly Constitutional, I look at how to close “buy, borrow, die” without it.
Constitutionality. The income tax itself is on solid ground. The 16th Amendment lets Congress tax “incomes, from whatever source derived, without apportionment,” and nothing in it requires graduated rates. That’s a cleaner footing than the first sketch’s federal land tax, which I’ve since moved to states and cities.

Objections

Isn’t a flat tax a tax cut for the rich?

It can be, which is why this isn’t a flat tax in the usual sense. Capital income pays more than wages. Gains get taxed when they’re sold or passed on at death, at the capital rate, with no discount.

And the cash floor does the work that brackets used to do, out in the open, where it can be counted. If the floor is sized too small, then yes, this shifts the burden down. That’s why the floor gets sized first.

Won’t Congress just suspend the formula?

Maybe. I don’t think any law can fully bind a future Congress, and the Swiss put theirs in the constitution, which is a much higher bar here. The three-fifths lock helps.

But the real lock is visibility. Suspending the formula means voting to hide the price of what you’re spending. That’s a vote people would see on their next paycheck.

Why pay down debt during an AI transition?

Because interest buys nothing, and it’s already bigger than defense. Fiscal 2026 closed with net interest of about $1.1 trillion, more than the government spent on defense or Medicare. And if markets start to expect transformative AI, economists Trevor Chow, Basil Halperin and J. Zachary Mazlish show that long-term real interest rates should rise.

That makes carrying debt more expensive at exactly the wrong time. Paying it down while the transition is still gradual buys room for when it isn’t. And if it stops being gradual, the surcharge pauses on its own.

Isn’t the cash floor just UBI?

Close. Everyone gets the same payment, but every dollar you earn is taxed, so for most working adults it nets out, and they’re net taxpayers.

That’s a negative income tax: generous at the bottom, gone by the middle. And it’s a floor for the transition, not the destination. The destination is the things themselves.

One direction

The first sketch had a number. This plan has a direction.

Every spending bill shows up on your paycheck. Debt goes down, and the surcharge falls with it. Hiring is never taxed more than automating.

A recession never raises your rate. And as the lifeline block grows and the floor of goods gets built, the money economy for essentials shrinks, and so does the tax base needed to support it.

But all of this assumes the transition stays gradual. What if it doesn’t?

What if wages fall faster than the nodes come online, and the dashboard lights up in year two instead of year ten? That’s the next piece, Part 19: Plan for the Bad Transition.