Two Trillion Dollars of Exceptions

Section 1The question before us

Two men earn a million dollars in the same year. One of them earns it by working and pays roughly four hundred thousand dollars. The other earns it because assets he already owned became more valuable, borrows a million against them, spends it, and pays nothing at all. Neither has broken a law. Neither has done anything clever. They are simply standing in different places in a code that treats them differently on purpose.

This paper is not about whether taxes are too high. That is a separate argument and a legitimate one, and the numbers in Section 2 will not comfort either side of it. This paper is about whether the code applies the same rule to everyone standing in the same position, and it plainly does not.

A market economy requires that outcomes be determined by what you produce. Ours increasingly rewards where you stand in the code.

What is at stake is larger than the revenue. A tax system is one of the few places where every citizen deals with the government directly, and it is therefore one of the places where a country either builds confidence in its institutions or spends it down. A person who believes the code is a maze that rewards whoever can afford the best guide does not simply resent the wealthy. He concludes that the rules are for people like him and the exceptions are for everybody else, and that conclusion spreads well beyond taxation.

Americans do not need a system that takes more. They need one they can believe everyone is standing in. Knowing that your neighbor, your employer, and the largest fortune in the state are all subject to the same rule is worth something that does not appear on any revenue estimate, and losing it costs something that does not either.

There is a version of the argument for a flatter, simpler tax that comes from the political left and a version that comes from the right, and they meet in the same place: a code with $2.3 trillion of exceptions is not a tax system that has been designed. It is a sediment, laid down one favor at a time, and every layer of it was somebody's good idea.

1.2   Our own share of the failure

The people who complain most about the tax code are frequently its beneficiaries, and that includes the author of this paper and a good many of its readers.

The mortgage interest deduction is a loophole. So is the exclusion for employer-provided health insurance, the largest single break for most working families. So is the preferential rate on capital gains and dividends, the deduction for charitable giving, the exclusion for retirement contributions, and the twenty percent deduction for pass-through business income that this author has claimed. Each of them was enacted for a reason somebody could defend. All of them together are the thing we say we oppose.

A movement that argues against government picking winners cannot exempt the provisions that pick us. If the objection to the code is that it is riddled with special treatment, the objection has to survive the discovery that some of the special treatment is ours. Anything less is not a principle. It is a negotiating position.

Section 2What the numbers actually show

Before proposing anything, it is worth establishing what has and has not happened to American taxation, because most of what is said about it in public is wrong in one direction or the other.

2.1   The great increase happened once, in thirty years

From 1820 to 1862 federal revenues averaged about 1.8 percent of GDP and never exceeded 3 percent. Across 1863 to 1913 they averaged roughly 3 percent, peaking near 6 at the end of the Civil War and falling below 2 by 1913. Then the income tax arrives, and by 1945 federal revenue is 19 percent of GDP.2

Everything Americans argue about was settled between the Sixteenth Amendment and the end of the Second World War. The century before it and the eighty years after it are both remarkably flat.

2.2   Revenue has not moved since 1950. Spending has.

Federal revenue has averaged about 17 percent of GDP over the past fifty years, and fiscal 2024 came in at 17.1. That stability held through a top marginal rate of 91 percent under Eisenhower, 70 under Carter, and 28 after 1986. Whatever Congress does to the rate schedule, roughly the same share of the economy arrives at the Treasury.2

Federal spending over the same fifty years averaged 21.2 percent and reached 23.4 percent in fiscal 2024. The deficit ran 5.9 percent of GDP in fiscal 2025 against a fifty-year average of 3.8.2

Taxes have been flat for seventy-five years. Spending has not. Any paper claiming otherwise in either direction is arguing with the arithmetic.

And one figure that will discomfort readers of this site: counting federal, state, and local government together, total United States taxation was 25.6 percent of GDP in 2024 — down from a peak of 28.3 percent in 2000, and 31st out of 38 OECD countries against an average of 34.1.4 By the standards of the developed world, the United States is a low-tax country. We include that because it is true, not because it helps.

Section 3Two trillion dollars of exceptions

Here is where the code actually does its work.

The Joint Committee on Taxation projects that tax expenditures — the exclusions, deductions, credits, preferential rates, and deferrals written into the code — will total $2.3 trillion in fiscal 2026, up from $2.2 trillion in 2025, and $11.7 trillion across the five years from 2025 through 2029. The ten largest alone exceed $1.4 trillion, about two-thirds of the total.1

Set that beside what the government actually collects. In 2025 federal individual and corporate income taxes brought in $3.1 trillion while the code gave away $2.2 trillion. The giveaway exceeds what the United States spends on Social Security, on Medicare, or on national defense. It exceeds the entire fiscal 2025 deficit of $1.8 trillion.1

Individual provisions accounted for about $2.0 trillion of the 2025 total and corporate provisions for $264 billion — which is worth stating plainly, because the popular picture of the tax code as a corporate escape hatch is not what the numbers show. The largest single item is the exclusion for pension contributions and earnings, at $383 billion. The others at the top are the preferential rates on dividends and long-term capital gains, and the exclusion for employer-sponsored health insurance.1

Note also that the list keeps growing. The 2025 tax act added exemptions for tip income and overtime income, an additional standard deduction for seniors, and immediate expensing for factories, most of them scheduled to expire after 2028.1 Each has a constituency. Each was defended as relief. None was subjected to the scrutiny a $50 billion appropriation would receive, because spending through the code does not look like spending.

3.1   What the money is actually for

An aggregate that large invites the assumption that it consists of obscure carve-outs. It does not, and a reader who intends to argue the number in public should know what is in it before somebody else looks it up.

The ten largest individual provisions, in round figures: the exclusion for pension and IRA contributions and earnings, $383 billion. Reduced rates on dividends and long-term capital gains, $304 billion. The exclusion for employer-paid health insurance, $226 billion. The premium tax credit, $135 billion. The child and dependent credit, roughly $130 billion. The exclusion of capital gains at death, about $76 billion. The earned income credit, about $67 billion. The exclusion of gain on the sale of a principal residence, about $63 billion. The pass-through business deduction, about $60 billion. The mortgage interest deduction, about $30 billion. Below those sit the research credit — the largest corporate item — accelerated depreciation, charitable giving, the state and local deduction, municipal bond interest, and several hundred smaller provisions.1

Four purposes account for most of the money: retirement saving, health coverage, investment income taxed at preferential rates, and support for families and low-wage work.

Almost none of this is gamesmanship. It is ordinary provisions claimed by tens of millions of ordinary households.

That is the honest finding and it cuts against the easy version of this argument. The narrow, reverse-engineered provisions that Section 8 addresses are real and corrosive, and they are a small share of the dollars. In dollar terms, closing the exceptions is mostly a change to the taxation of the middle class, and any paper implying the money is hiding somewhere else is misleading its readers.

Two further cautions on the figure itself. The refundable portions of the earned income and child credits are spending delivered through the code, and Treasury classifies them as outlays while the Joint Committee combines them — so part of the total is spending that only looks like a tax break because of where it was written. And the Joint Committee states expressly that listing a provision implies no judgment about whether it is good policy. The number measures deviation from a baseline, not waste.1

This is the same argument this organization made about contracts and subsidies in Issue Paper No. 7, moved into a place where it is harder to see. A check to a named company is a favor everyone can recognize. A provision drafted so that three firms qualify is the same favor, delivered where no appropriations committee votes on it and no reporter covers it.

Section 4Who actually pays

A claim heard constantly on our side of the argument is that the bottom half of earners pay no taxes. A claim heard constantly on the other side is that the wealthy pay nothing. Both are wrong, and the accurate account is more useful than either.

4.1   The visible system is progressive

Federal income tax is steeply graduated, and analyses that assign the corporate tax burden to shareholders and workers still find the overall federal system progressive. In 2022 the top one percent of returns paid 40.4 percent of federal individual income tax while reporting 22.4 percent of adjusted gross income — a share of the tax roughly double their share of the income. Anyone arguing that high earners bear no federal burden is arguing against the data.8

Three qualifications belong with that figure, because it is quoted more often than it is explained.

It describes one tax. Individual income tax is roughly half of federal revenue; counting payroll, corporate, excise, and estate taxes, the top one percent's share of all federal taxes falls to something nearer 25 percent, and counting state and local taxes it falls further.8

It is a share of tax, not a rate, and it means nothing without the income share beside it. The evidence of progressivity is the ratio of 40.4 to 22.4, not the 40.4 alone.

And it moves a great deal — 45.8 percent in 2021, 40.4 in 2022, 38.4 in 2023 — because it tracks capital gains realizations, and 2021 was the largest realization year in four decades. Anyone still citing the 2021 figure has chosen a year.8

4.2   And the measure cannot see what Section 6 is about

This qualification is different in kind from the other three, and it is the reason we set the statistic out at length rather than simply conceding it.

These distributions rank taxpayers by adjusted gross income. AGI is a narrow concept, as the organizations that publish the figures say plainly: it excludes government transfers, the value of employer-provided health insurance, unreported income, municipal bond interest, and imputed rent.8 It also excludes, entirely, unrealized appreciation and the proceeds of borrowing.

So consider the arrangement described in Section 6. A man owns appreciated stock, sells none of it, and lives on fifty million dollars borrowed against it. His adjusted gross income is close to nothing. He does not appear in the top one percent of this distribution. Depending on his dividends, he may appear in the bottom half of it.

The statistic offered to prove the code is progressive is computed on a base that structurally omits the income we say is escaping.

That is not an accident of measurement. AGI measures taxable income, and the entire object of buy, borrow, die is to hold economic gain that never becomes taxable income. The metric and the arrangement are defined against one another, which means the metric is silent about precisely the population Reform II addresses.

We want to be careful about what this does and does not establish. It does not show that the code is regressive, or that high earners pay little — among people whose income appears on a return, the schedule is steeply progressive and this paper says so twice. It shows that the most common objection to Reform II is an answer to a question we did not ask.

4.3   The invisible system is not

Payroll taxes are the largest federal tax most working people pay, and they apply from the first dollar. State and local taxes are regressive in nearly every state. Property taxes ride into rent. Excise and sales taxes fall on consumption, which is nearly all of a modest income and a small fraction of a large one. And taxes levied on businesses are embedded in the price of everything a household buys, whoever writes the check.

Who remits a tax is not who bears it. That is not ideology, it is price theory, and almost all public argument about taxes ignores it.

4.4   And the highest marginal rates in America are paid by the working poor

This is the finding that should end the argument about who is carrying what.

A household on assistance that loses a dollar of benefits for every additional dollar earned faces an effective marginal rate of 100 percent. Where several programs phase out across the same income band, the effective rate can exceed 100 percent — the household is poorer for having earned more. No billionaire in American history has faced a marginal rate approaching that.

The system taxes the transition out of poverty harder than it taxes anything else, and then we are surprised that the transition does not happen.

This paper does not resolve that, because the phase-out schedules sit in benefit programs rather than in the tax code and belong to a separate paper. But no honest description of who pays what in America can omit it, and anyone who believes in work incentives should find it the single most objectionable feature of the whole arrangement.

Section 5Reform I: Graduated, prospective, and honest at the bottom

Rates should be graduated, applying only to the dollar earned above each threshold, and changes should apply prospectively. Effective marginal rates — counting benefit phase-outs alongside tax — should never exceed the top statutory rate for any household at any income.

The first two provisions describe the system we already have, and we state them because they are worth defending rather than because they are novel. A graduated schedule applied at the margin does not punish the next dollar of a person's earnings at the rate applied to his whole income, and much popular resentment of progressive taxation rests on believing that it does. Retroactive changes to settled transactions are a separate wrong, and they undermine the planning on which every long-lived business depends.

The third provision is the real one, and it follows directly from Section 4.4. If a top statutory rate of thirty-seven percent is the most the country is willing to impose on its highest earners, then it is the most that should be imposed on a household earning thirty thousand dollars — counting everything, tax and phase-out together. That is a principle a person of any politics can state in one sentence and no one can defend violating.

5.1   A negative rate at the bottom is a rate, not a loophole

A graduated schedule need not begin at zero. It can run negative at the bottom, cross zero somewhere in the middle, and rise from there — paying out at low incomes, collecting at higher ones, on one continuous curve. That is not a subsidy bolted onto a tax system. It is a tax system with a wider range.

The idea belongs to Milton Friedman, who proposed a negative income tax in Capitalism and Freedom in 1962 as a replacement for the patchwork of categorical programs — a single schedule, administered by one agency, with no caseworker deciding who deserves what. The earned income credit enacted in 1975 is its partial descendant.7

We raise this because it settles a classification question that matters. When this paper calls for closing preferences, it is not calling the earned income credit a loophole. The test that separates the two is simple and applies neutrally: does the provision turn on how much you earned, or on what you did with the money?

A provision keyed to income is part of the rate schedule. A provision keyed to a choice — buying rather than renting, insuring through an employer rather than independently, saving rather than spending — is a preference, and it is a preference whether it appears as a deduction, an exclusion, or a credit.

By that test the earned income credit is a rate. The mortgage deduction, the employer health exclusion, and the retirement exclusion are preferences. The child credit sits between them: adjusting liability for household size is a defensible way to measure ability to pay, which is a rate-schedule function, but a fixed credit per child phasing out at four hundred thousand dollars of income is doing more than that. We think it is partly both and we would rather say so than classify it conveniently.

One consequence follows immediately, and it is uncomfortable for the current schedule rather than for the concept. If the payments at the bottom are part of the rate curve, then their phase-outs are part of the rate curve too — and the curve as presently drawn contains a stretch where the marginal rate exceeds 100 percent. That is not an argument against a negative rate. It is an argument that the schedule has been drawn badly, which is precisely what the third provision of this reform corrects.

Section 6Reform II: Borrowing against appreciated assets is a realization

Where a taxpayer above a substantial threshold pledges appreciated assets as security for borrowing, the proceeds should be treated as a realization of gain to the extent of the appreciation, with basis adjusted so the same gain is never taxed twice.

The strategy this addresses has a name — buy, borrow, die — and it is not exotic. Buy an asset. Watch it appreciate. Never sell it, because selling triggers tax. Borrow against it instead, at rates available to people with large portfolios, and live on the loan, which is not income because a loan must be repaid. Die holding it, at which point the heirs take a stepped-up basis under section 1014 and the appreciation of an entire lifetime is never taxed as income to anyone.5

No step in that sequence is aggressive. It requires no shelter, no offshore entity, and no lawyer willing to take a position. It is simply what the code says to do.

6.1   Congress has already decided this three times

This is the part of the argument we did not expect to find, and it is the reason the reform is narrower than it sounds.

Installment obligations. Under section 453A(d), if a taxpayer pledges an installment obligation as security for debt, the net proceeds of that borrowing are treated as a payment received on the obligation. Borrowing against the asset is realization. That has been the law since 1987.3

Constructive sales. Under section 1259, enacted in 1997, a taxpayer holding an appreciated financial position who enters into a short sale, an offsetting notional principal contract, or a forward to deliver substantially identical property is treated as having sold it, and must recognize gain at fair market value. Congress decided that a transaction economically equivalent to a sale would be taxed as one.3

Retirement plan loans. Under section 72(p), an amount received as a loan from a qualified employer plan is treated as a distribution — taxable income — above statutory limits.3

A factory worker who borrows sixty thousand dollars against his retirement plan has taxable income. A man who borrows six hundred million against appreciated stock does not.

That is the whole of the case. We are not proposing a new theory of income. We are proposing that a rule Congress has already applied to installment sellers, to sophisticated hedgers, and to ordinary workers with 401(k) accounts be applied to the transaction where it matters most.

6.2   Why we tax the transaction and not the balance sheet

Other approaches to the same problem have been proposed, and we should say plainly that we reject them — not as a matter of tactics, but because we think they are wrong.

A Senate proposal in 2021 would have marked tradable assets to market annually for taxpayers above $100 million in income or $1 billion in assets, roughly 700 people, with a deferral charge on non-tradable assets at sale.6 An administration proposal the following year would have imposed a 20 percent minimum tax on total income including unrealized gains for households above $100 million in net worth.6 Both would tax appreciation that nobody has done anything with.

Tax the transaction, not the balance sheet. A price that moved is not an event. A decision to spend the money is.

Three reasons, in order of how much weight we put on them.

A transaction is something a person chose to do. The value of an asset moves because strangers traded something similar on a Tuesday. The owner did nothing, decided nothing, and received nothing. Taxing him for it makes the government a silent partner in every price movement in the country and makes his liability a function of other people's behavior. Borrowing against the asset is different in kind: he went to a lender, signed an agreement, and took cash. That is an act, and acts are what a transaction tax is for.

A transaction tax can always be paid. Mark-to-market can present a farmer, a founder, or a family business with a bill and no money to pay it, forcing the sale of the very asset being taxed — which is how a tax intended to reach the very rich ends up breaking up the merely illiquid. Our proposal cannot do that, because the transaction being taxed is the one that produced the cash. The money to pay the tax arrived in the same wire.

And it needs no valuation machinery. Marking assets to market requires the government to determine what a private company, a partnership interest, or a piece of farmland is worth, every year, against a taxpayer with every incentive and every resource to argue. Our proposal asks no such question, because a lender has already answered it. Somebody with his own money at risk decided what the collateral was worth and how much to advance against it. That number is more reliable than anything an appraisal fight would produce, and it costs the government nothing to obtain.

The constitutional question is real — whether Congress may tax appreciation that has not been realized remains unsettled — and our proposal avoids it. But that is a convenience rather than the reason. We would take the same position if the question were settled tomorrow in the other direction.

We would also close the second half of the sequence: appreciated assets should not receive a stepped-up basis at death. The Joint Committee has estimated the cost of that single provision at roughly $60 billion a year.1 A reform that taxes the borrowing but leaves the step-up in place merely moves the timing.

Section 7Reform III: Close the preferences, including ours

Tax expenditures should be eliminated, and the great majority of the revenue used to lower rates across the schedule rather than to fund new spending. Every remaining provision should carry a sunset and require affirmative reenactment.

We are aware of how that sentence reads when the list is specific, so here is the list. The exclusion for employer-provided health insurance. The preferential rate on capital gains and dividends. The exclusion for retirement plan contributions and earnings, the single largest item at $383 billion. The mortgage interest deduction. The charitable contribution deduction. The twenty percent pass-through deduction. The state and local tax deduction. The exemptions for tip and overtime income enacted in 2025.1

Those are the loopholes. There is no separate category of disreputable preferences claimed by other people. Two-thirds of the money is in ten provisions, and the ones at the top are claimed by tens of millions of ordinary households, including most readers of this paper.

We exclude the earned income credit from that list, and the child credit in part, for the reason given in Section 5.1: a provision keyed to how much a household earned is part of the rate schedule rather than a preference, whatever the accounting convention calls it.

A reform that closed only the preferences we do not use would raise almost nothing and would deserve to fail. We would rather lose readers by saying so than publish a paper that pretends the money is somewhere else.

7.1   The objection that does not depend on arithmetic

Every argument so far has been about money. This one is not, and it is the stronger case.

A preference exists to drive behavior, and not everyone can perform the behavior. The mortgage deduction rewards buying over renting. The employer health exclusion rewards getting coverage through a job rather than purchasing it yourself. The retirement exclusion rewards having enough margin at the end of the month to save some of it. In each case a person who makes a different choice — or who cannot afford the choice at all — pays more so that his neighbor may pay less.

This is Congress deciding what kind of life it prefers you lead, and pricing the alternatives accordingly.

That is not a loophole in the pejorative sense. Nobody is cheating. It is a deliberate use of the tax code to make some lives cheaper than others, enacted by people who thought the lives they were subsidizing were better ones. Sometimes they were right.

A person who believes citizens should order their own lives ought to object to the arrangement whether it raises revenue or loses it, and whether or not he personally qualifies. The renter subsidizing the homeowner, the self-employed contractor subsidizing the salaried employee's health plan, the worker with nothing left to save subsidizing the worker who has — none of them chose to fund somebody else's preferred arrangement, and none of them was asked.

This is why we would close the preferences even if the revenue were returned to the penny, and even if the code got no simpler. The objection is not that they cost money. It is that they are a judgment about how to live, made by a legislature, and charged to the people who chose otherwise.

7.2   Why a sunset is the load-bearing provision

The United States has done this before and it did not hold. The 1986 reform broadened the base, eliminated a long list of preferences, and cut the top rate to 28 percent — and within a decade the preferences had returned, because each one returns separately and nobody organizes against the reappearance of a provision worth two billion dollars.

That is the actual failure mode, and rate reduction alone does not prevent it. Every preference should therefore expire on a stated date and require an affirmative vote to continue, so that the default is a clean code and the burden of persuasion falls on the beneficiary rather than on the reformer.

Section 8Reform IV: No provision may identify its beneficiary

No tax provision should be drafted so that its beneficiaries can be identified in advance. Where a provision's eligibility criteria are narrow enough that the qualifying taxpayers are effectively a named list, it should be treated as what it is — an appropriation — and enacted, scored, and voted on as one.

This is the same rule Issue Paper No. 7 proposed for direct subsidies, applied to the channel that is harder to see. A provision available to any firm meeting general criteria is a policy. A provision whose thresholds, effective dates, and definitions were reverse-engineered so that three companies qualify is a favor, and the fact that it arrives as a subtraction rather than a check does not change what it is.

Two mechanical requirements would do most of the work. Every new preference should be accompanied by a published estimate of the number of taxpayers expected to qualify — a provision expected to benefit fewer than a stated number should face a presumption against enactment. And every preference above a dollar threshold should be scored, reported, and reviewed on the same schedule as direct spending, because a hundred billion dollars forgone and a hundred billion dollars appropriated are the same hundred billion dollars.

Section 9Reform V: A code that stays fixed

Creating or expanding a tax preference should require a supermajority of both houses. Eliminating one should require a simple majority. The definition of income should be governed by a general principle and an anti-avoidance rule rather than an enumerated list, so that it can reach arrangements nobody has yet invented. And the whole reform should phase in over years rather than arrive at once.

The first four reforms clean the code. This one is about whether it stays clean, and without it the answer is no.

9.1   The tax on the tax

Americans spend roughly 7 billion hours a year complying with the federal tax code. Valued at ordinary wage rates that is something near $390 billion of lost time, and taxpayers pay another $148 billion out of pocket for software and professional preparation. The total runs above $500 billion, about 1.8 percent of GDP.9

Complying with the corporate income tax costs the country more than the corporate income tax collects.

That comparison is not rhetorical. The compliance burden is larger than the 1.7 percent of GDP the corporate income tax raises, and roughly twenty-three times the entire budget of the Internal Revenue Service.9 Nine in ten Americans use a paid preparer or commercial software, because the alternative is not realistically available to them.

None of that money buys anything. It does not build a road, treat a patient, educate a child, or fund the government. It is spent entirely on determining what is owed under rules Congress wrote, and it is a tax on the tax — paid by the people least able to hire their way out of it, and paid in the only currency nobody gets back.

Senator McCain used to warn against creating industries that depend on a problem persisting. The tax preparation and tax planning professions are the clearest example in American life. Their existence is not anyone's fault and their practitioners are not villains — most of them are helping neighbors navigate something the neighbors did not write. But a half-trillion-dollar industry now exists because the code is difficult, and no constituency of comparable size exists for making it simple.

The number that explains how it got this way: since 2000, Congress has made 9,630 changes to the tax code. That is roughly one per day, for a quarter of a century.9

9.2   Hard to complicate, easy to clean

The usual proposal is a supermajority requirement for tax increases, and it would defeat this reform rather than protect it.

A supermajority to change taxes applies to eliminating a preference exactly as much as to raising a rate — Washington State's version says so explicitly, covering reductions in tax breaks. Adopt that rule and $2.3 trillion of exceptions become permanent, protected by any thirty-four senators who benefit from them. It is a ratchet pointed at the reformer.

We propose the inverse, and the asymmetry is the whole point. Creating or expanding a preference requires two-thirds of both houses. Removing one requires a simple majority. The code becomes difficult to complicate and easy to clean, which is the direction of travel we want and the opposite of what supermajority proposals normally produce.

Sequence matters absolutely. Locking a dirty code is worse than not locking it, so the cleaning in Reforms III and IV must come first, in the same legislation.

9.3   Freeze the principle, not the definition

Here is the objection that would sink a durability rule drafted carelessly, and it is a serious one.

Nobody writing the Sixteenth Amendment imagined a person living for decades on money borrowed against stock he never sold. Nobody today can name the arrangement that will be obvious in 2050. Freeze a definition of income and you have guaranteed that it will be gamed, because the people who will game it are better paid and more motivated than the people who drafted it.

So the answer cannot be a better list. It has to be a standard that reaches arrangements nobody anticipated.

Separate the two things. Freeze what should be stable — the rate schedule, the prohibition on named beneficiaries, the bar on retroactivity, the cap on effective marginal rates. Those are principles and they do not need to evolve. Leave flexible what must evolve — what constitutes a realization, governed by a stated principle rather than an enumerated list. Something on the order of: any transaction by which a taxpayer converts appreciation into present economic benefit is a realization.

And add a genuine general anti-avoidance rule, amendable by simple majority. Canada, Australia, the United Kingdom, India, and the European Union all have one. The United States has only a narrow codified economic substance doctrine, which is why each new arrangement requires its own act of Congress.10 A country that must legislate against every new structure individually will always be a decade behind the people designing them.

9.4   Phase it in, and return the payroll money too

Two enactment provisions, both of which decide whether the reform survives contact with actual households.

Phase in over five to ten years, with existing mortgages grandfathered. The 1986 Act phased several provisions for the same reason, and the sectors most affected here — housing, employer benefits, charitable giving — need time to reprice rather than a cliff.

Return the payroll tax revenue as payroll rate reduction. This is technical and it is the difference between a reform households accept and one they revolt against. The exclusion for employer-paid health insurance shelters payroll tax as well as income tax, and the Joint Committee's published estimate of that preference expressly excludes the payroll effect.1 The provision is therefore worth considerably more than the $226 billion on the list, and repealing it raises payroll revenue that an income tax rate cut cannot give back. A middle-income family with employer coverage comes out behind unless part of the offset arrives as a lower payroll rate. Draft it that way or the reform fails on its first pay period.

Section 11What we are not claiming

We are not claiming this raises or lowers total taxation. The proposals here are about the structure of the code, not its level, and a reader who wants a smaller government should want that argument won on the spending side where Section 2 shows the movement actually occurred.

We are not claiming the wealthy pay nothing. The federal system is progressive and the people who say otherwise are wrong. What we claim is narrower and more troubling: that two people with the same economic gain can owe entirely different amounts depending on whether the gain arrived as wages or as appreciation they borrowed against.

We are not claiming that anyone using these provisions has done anything wrong. Buy, borrow, die requires no aggressive position. A person who declines to use it while his peers do is not virtuous; he is worse off, and his family bears the cost. The behavior is rational. The rules that make it rational are the problem.

We should also disclose a position we hold and are not arguing for here. This author would prefer no income tax at all. Taxing what a person earns rather than what he consumes penalizes the thing an economy most needs people to do, and requires an apparatus of reporting and enforcement that a consumption-based system would not. We have looked for a viable path to replacing it and we have not found one. Every serious proposal we have examined either raises far less than the government currently spends, falls hardest on people with the least, or would in practice be enacted alongside the income tax rather than instead of it — which is the worst outcome available.

So we propose to repair the system we have rather than to describe a better one we cannot get to. A reader who thinks that insufficiently ambitious may be right. We would rather publish a reform that could happen than a preference that cannot.

And we are not claiming this would be popular. It would cost most readers of this paper money in the short run, and Section 7 lists precisely how.

Section 12The argument you can carry

The paper compressed to what a person can remember and repeat.

And the one that is ours rather than theirs. The mortgage deduction is a loophole. So is the health exclusion, the retirement exclusion, the charitable deduction, and the pass-through deduction this author claims. There is no separate category of disreputable preferences belonging to other people. If we will not give up ours, we do not have a principle. We have a negotiating position.

Section 13Conclusion

Most argument about American taxation is about the rate, and the rate is the part that has changed least and mattered least. Seventy-five years, four top rates between 91 and 28 percent, and the same seventeen percent of GDP arriving every time.

What has grown is the structure around the rate — a code that now forgoes more revenue in exceptions than the federal government borrows in a year, laid down one defensible provision at a time by people who were not doing anything wrong. Each layer had a sponsor, a rationale, and a constituency. Together they are a system in which what you owe depends less on what you earned than on what form it arrived in and how well advised you were.

The repairs proposed here are not radical and not new. Tax the dollar above the threshold. Treat a transaction that converts appreciation into spendable cash the way the code already treats three narrower versions of the same thing. Stop writing provisions that identify their beneficiaries. Close the exceptions — all of them, ours included — returning the money as lower rates rather than as new spending. And then make the thing hard to complicate again, because the last time this country cleaned its tax code the exceptions were back within a decade.

That last repair is the one we would defend hardest. A code amended nine thousand times in twenty-five years is not being governed. It is being negotiated, continuously, by whoever is in the room — and the half-trillion dollars Americans spend each year working out what they owe is the invoice for that negotiation, sent to people who were never party to it.

A free economy depends on the proposition that what a person keeps is determined by what he produced. Every exception in the code moves that determination somewhere else: to a lawyer, to a lobbyist, to an accident of which asset class the money happened to arrive in. Two trillion dollars is a great deal of determination to have moved.