If We Ran Congress for a Day: A Plan to Get the Debt Under Control, Step by Step

Updated October 14, 2026. We corrected several figures, rebuilt the year-by-year arithmetic, and strengthened the plan. Our first version overstated what several items save inside the ten-year window and how fast the savings grow: it had them reaching 3 percent of GDP a year by 2037, when its own policies level off closer to 2. To make up part of the difference we added two items, a half-point payroll tax increase and a cap on the tax exclusion for employer health plans, and we turned our automatic triggers into a real debt brake; with them, savings reach about 2.4 percent of GDP by 2037. We also moved the Medicare age change onto the same under-55 cohorts as Social Security, which keeps our promise to near-retirees but saves almost nothing inside the window. Budget figures reflect CBO's October 8 estimate for the full fiscal year; market figures are as of early October, after the Fed's September 16 rate increase.

The federal government ran a deficit of about $2.0 trillion in fiscal 2026, which ended September 30. That is roughly 6.2 percent of GDP. Debt held by the public sits at about 101 percent of GDP. Interest on that debt now costs more than $1.1 trillion a year, more than we spend on defense. Nobody handed us the gavel. But the arithmetic does not care who holds it, and this piece is about what the arithmetic demands, why, and who pays.

None of the numbers above is a crisis. Countries can carry a great deal of debt for a long time. The problem is the direction. The Congressional Budget Office's baseline has debt reaching 120 percent of GDP by 2036, 144 percent by 2046, and 175 percent by 2056, with interest costs rising from 3.3 percent of GDP to nearly 7. Those projections were finished before February's tariff ruling, which makes them slightly worse. That path does not end in a specific disaster on a specific date. It ends with fewer options, higher rates, and a government that spends its first trillion dollars every year paying for decisions it made in the past.

So this is our attempt to do the thing pundits rarely do: write down a complete plan, with numbers, and explain every step as we go. We will define the terms, set a target, show where the money is, lay out the plan, run the math year by year, and then tell you where it breaks. Along the way we flag which items are load-bearing walls and which are bargaining chips, because that distinction is where fiscal plans live or die.

One warning up front. The ten-year figures below are rough. Most are drawn from CBO's published menu of deficit-reduction options and from the work of nonpartisan scorekeepers, adjusted for what has changed in 2026. Read them as accurate to within a third, not to the dollar. The debt paths are our own arithmetic, and we show how to reproduce them. What matters is the shape of the plan and the order of magnitude of each piece.

Key terms and one equation

Deficit versus debt. The deficit is a flow: how much more the government spends than it collects in one year. The debt is a stock: the sum of every past deficit, less every past surplus. This year's deficit becomes next year's debt.

Gross debt versus debt held by the public. You will see headlines about a national debt of $40 trillion; it crossed that line in August. That is gross debt, and roughly $8 trillion of it is money the government owes itself, mostly to the Social Security and Medicare trust funds. The number economists watch is debt held by the public, about $32.4 trillion, because that is the amount that has to be sold to actual investors. When we say debt-to-GDP in this piece, we mean this one, and it is right around 100 percent.

The primary deficit. This is the deficit excluding interest payments. It matters because interest is the consequence of past decisions and the primary deficit is the current one. If you want to know whether today's policy is sustainable, you look at the primary balance. Ours is a little under 3 percent of GDP in deficit, which is to say that before paying a dollar of interest, we are already short by about $900 billion.

r and g. The letter r is the average interest rate the government pays on its debt. The letter g is the nominal growth rate of the economy, real growth plus inflation. These two numbers, and the gap between them, determine whether a given debt load shrinks or grows relative to the economy that carries it.

Everything in fiscal policy reduces to one line:

Plug in today's numbers. CBO puts the average rate on the debt near 3.4 percent and nominal growth near 4.3, so r − g is about −0.9. With debt at 100 percent of GDP, that term is worth about −0.9 points a year: growth is quietly eroding the debt for us. But the primary deficit is about +3 points. Net, debt-to-GDP rises about 2 points a year, and ten years of that takes you from 100 percent to roughly 120, which is CBO's number.

That cushion is not permanent, even in CBO's own projections. CBO has the average rate on the debt rising to 4.2 percent by 2056 as low-coupon debt rolls over at today's higher rates, while real growth slows to 1.5 percent. Put those together and r catches up with g around the middle of the 2030s and runs above it after that. The friendly term in the equation becomes an unfriendly one within a decade, which is why the baseline accelerates in the 2040s.

Read the equation once more, because it contains the entire strategy. There are only three ways to bend the line: shrink the primary deficit, raise g, or lower r. Congress controls the first directly, influences the second weakly, and controls the third barely at all. A plan that depends on r or g doing the work is a hope, not a plan.

Where we stand, October 2026

Two things in that table changed in 2026, and both matter.

First, the deficit was supposed to be $1.9 trillion. It came in at $2.0 trillion, about $140 billion more, largely because in February the Supreme Court struck down the tariffs that had been imposed under emergency powers. About $130 billion was refunded during the year, and the replacement tariffs raise less: customs collections ended the fiscal year below the year before. CBO's August update, which reflects tariffs in place as of July 31, puts the cost over the next decade at about $0.9 trillion, roughly a third of the tariff revenue the baseline had counted on. Keep this episode in mind. It is the reason for one of our rules later.

Second, rates went up, not down. In September the 10-year yield passed 5 percent for the first time since 2007, and the 30-year climbed above 5.5. CBO's baseline assumed the Fed would be cutting this year. Instead, the war with Iran pushed oil above $100 and inflation to 4.2 percent in May, and on September 16 the Fed raised rates for the first time since 2023. Interest on the debt rose 11 percent in fiscal 2026 alone. Every Treasury the government rolls over at 4.5 to 5.6 percent pushes r, the average rate on the whole stock of debt, closer to g. That −0.9 cushion in the equation is shrinking faster than CBO assumed, and it is the single biggest change to the fiscal picture this year.

Why this isn't Greece: the dollar

Readers who remember Greece in 2010 or Argentina in any decade tend to ask the obvious question at this point: at 100 percent of GDP and rising, why has nothing snapped? The answer is the currency the debt is written in, and it cuts both ways.

The United States borrows in dollars, and the United States is the only entity on earth that can create dollars. That single fact removes the failure mode that destroyed Greece and Argentina. Greece borrowed in euros, a currency it could not print, so when the market stopped lending, the country was out of options: it could not pay, and it could not manufacture the money to pay. Argentina borrowed in dollars because nobody would lend it pesos at a tolerable rate, and every time the peso fell, its dollar debts grew in local terms until they were unpayable. A country that borrows in a currency it does not control can be forced into default by a change in market mood. A country that borrows in its own currency cannot be forced into default by its lenders: there is always a buyer of last resort, the central bank, and there are always enough dollars to make the coupon payments.

There is one exception, and it is self-inflicted. The statutory debt limit caps federal borrowing no matter what Congress has already voted to spend, and hitting it without raising it is the one way the United States could miss a payment. Last year's tax law raised the limit to $41.1 trillion, and it is likely to bind again around the spring of 2027. The 2011 standoff over it cost the country its AAA rating at S&P.

That is the reassuring half. The other half is that the risk does not disappear; it changes shape. For a foreign-currency borrower, the risk is default. For an own-currency borrower, the risk is inflation and the currency. The government can always pay, but if it pays by leaning on the central bank, the dollars it pays with are worth less, and the people who lent them, along with everyone holding cash and fixed claims, absorb the loss. This is not a hypothetical. It is how heavily indebted own-currency countries have most often reduced their real debt burdens, including the United States after 1945, when a decade of inflation running above bond yields quietly erased a large share of the war debt in real terms. Britain did the same for thirty years after the war. Bondholders were paid every penny they were promised, and the pennies bought much less. The main exception shows what the alternative costs: Britain after Waterloo paid down its war debt with a century of budget surpluses while holding to the gold standard, which is to say it deliberately gave up the printing press.

The dollar has a second advantage that Greece never had: the world wants it. Roughly $9 trillion of Treasuries, close to 30 percent of the debt held by the public, sits in foreign hands. Central banks hold it as reserves, exporters park their surpluses in it, and anyone anywhere who needs a safe asset denominated in the currency that global trade and finance run on buys it. That standing demand lets the United States borrow for less than its fiscal numbers alone would command. Economists call the discount a convenience yield, and on $32 trillion, every quarter point of it is $80 billion a year. The broader advantage is called the exorbitant privilege. It is real, but it is a discount, not a guarantee of cheap money: this autumn the United States has been paying more to borrow for ten years than France or Italy. It is also not a law of nature. It is a habit that other holders of capital could break if the trajectory ever looked unmanageable, and a 30-year yield near 5.6 percent is starting to test it.

Two other own-currency borrowers show the range of outcomes. Japan carries gross debt above 200 percent of GDP without a crisis, because nearly all of it is owed to Japanese savers and to the Bank of Japan, at rates the central bank held near zero for most of three decades. It has paid for that with a generation of stagnation and a currency that has lost much of its value. Britain, in the autumn of 2022, announced unfunded tax cuts and watched its bond market revolt within days, yields spiking until the prime minister fell. Britain did not default and was never going to. It simply discovered that even a country that prints its own money is charged a price when lenders lose confidence, and the price is set in r.

So the dollar means there is no cliff, no single day on which the United States cannot pay, unless Congress chooses one through the debt limit. What there is instead is a slow ratchet, running through the interest rate and the price level, and a choice about which of the two absorbs the adjustment. That is the choice the next section takes up directly. It is also why, when we come to the investor section at the end, whether the Fed will tolerate inflation turns out to matter more than any single line in the budget.

What actually stops us: the auction and the printing press

A reader put the sharpest version of the question to us this way. If the government can always create the dollars, what actually stops it from borrowing far more than it does? And if lenders ever balk, why not simply print the money and stop worrying about inflation? The honest answer is that nothing mechanical stops it. There is no number at which the machine seizes. What there is instead are two prices, both set by other people's expectations, and both of which feed back into the deficit itself.

The first price is the yield at auction. The Treasury sells new debt several times a week: bills that mature in days or months, notes and bonds that run two to thirty years. Most of what it sells simply replaces debt that is maturing; the genuinely new borrowing, about $2 trillion a year, is the deficit. The buyers are:

• primary dealers, a couple of dozen large banks and broker-dealers who are expected, as a condition of their status, to bid for a share of every auction;
• foreign central banks and sovereign funds;
• pension funds and insurers;
• money-market funds;
• and, increasingly, hedge funds.

Because the dealers must bid, a Treasury auction cannot fail the way a corporate bond deal fails, with the issuer walking away unfunded. What can happen, and what people mean when they call an auction weak, is that the debt clears at a lower price and a higher yield than the market expected an hour before the sale. A run of weak auctions at the long end, with foreign buyers stepping back and the dealers absorbing more than they want, is the market repricing the government's credit in the open.

Why does that matter more at 100 percent of GDP than it did at 35? Because of the loop. At today's debt, every one-point rise in the average rate the government pays adds about $320 billion a year to interest, roughly 1 percent of GDP, and the interest is borrowed too. Higher yields mean more issuance, more issuance means weaker auctions, and weaker auctions mean higher yields. In the equation, that is r rising toward and past g, and the scenarios further down show what it does to the path. This loop is how a country that cannot be forced into default still ends up in a crisis: the crisis is not a missed payment, it is the price. Britain in 2022 walked into it in a week. Canada, another own-currency borrower, walked into it over 1994 and early 1995: the Wall Street Journal called it an honorary member of the Third World, and Moody's cut its rating. Neither country defaulted. Britain's prime minister fell; Canada's government cut spending hard and had a surplus within three years.

The second price is the price level. "Printing money" has a precise modern form: the Federal Reserve buys Treasuries in the market and pays for them with newly created bank reserves. It did this by the trillions after 2008, and the inflation that critics predicted never arrived, because the economy had idle workers and idle factories and the new money mostly sat in bank reserves. The same was true in the first months of 2020. That is the true half of the "just print it" argument, and it is worth conceding plainly: in a deep slump, the printing press is not inflationary, and refusing to use it is a mistake.

The other half arrived in 2021 and 2022. Roughly $5 trillion of pandemic spending had been enacted, the last $1.9 trillion of it in March 2021. Much of that money reached an economy that was by then recovering fast, while the Fed was still buying $120 billion of Treasuries and mortgage securities a month. Inflation reached 9 percent, its highest in forty years. That is the mild version. The severe version is the one every country that has funded itself at its own central bank eventually reached: Weimar Germany, Zimbabwe, Venezuela, Argentina repeatedly. Once an economy is producing everything it can, new dollars do not create new goods; they compete for the same goods, and the price of everything, including what the government itself buys, rises to absorb them. A government that finances itself by printing is running the same deficit as before, in a currency it is simultaneously devaluing, which is why the deficits of hyperinflating countries get larger in real terms, not smaller.

So why not accept, say, 5 or 6 percent inflation for a decade and let it erode the debt, as the country did after 1945? Four reasons, in ascending order of importance.

First, inflation is a tax, and a badly designed one: it falls on cash, on wages that lag prices, on fixed pensions, and on anyone who saved in dollars, and nobody voted for it.

Second, it only works by surprise. Once lenders expect 5 percent inflation, they demand 5 percentage points more yield to compensate, and the real cost of borrowing does not fall at all. The post-1945 erosion worked because the Fed held yields down by agreement with the Treasury until 1951 and because nobody expected the inflation, and that trick is not available twice.

Third, once a country has surprised its lenders, it pays for the memory: they add a premium for the risk that it happens again. That is why Argentina borrows at double-digit real rates. The United States has imposed inflation losses on its lenders only as a side effect of other mistakes, in the 1970s and again in 2021–22, never as a deliberate way to pay its bills, and it borrows at a real rate of a little over 2. The exorbitant privilege described above is, precisely, the market's confidence that the dollar will not be debased on purpose, and printing to fund the deficit spends that confidence to pay this year's bills.

Fourth, and this is the institutional point, the Federal Reserve is independent for exactly this reason: so that no Congress can order the printing press turned on. Economists call the failure of that firewall fiscal dominance, the state in which the central bank sets rates to keep the Treasury solvent rather than to keep prices stable. Kevin Warsh, who became chair in May, has rejected suggestions that he will be more accommodating to the White House than his predecessor, and September's rate increase is consistent with that. His talk of a new Fed–Treasury accord is the thing to watch, because closer coordination is how fiscal dominance usually begins.

Which brings us to the answer. What stops the United States from borrowing without limit is not a wall; it is a slope. Two prices, the yield the market charges and the value of the dollar, rise as confidence falls, and both feed back into the deficit, so the slope steepens as you go. It is gentle for years, which is why nothing has snapped, and then, as Britain discovered, it is vertical for a week. Social Security is the version of this everyone can see, because the law forces it to pay out only what comes in and puts a date on the shortfall. The federal government as a whole has no such law. Its Social Security moment arrives not on a date but at a price, and the plan in the rest of this piece is, in the end, a plan for keeping that price low.

Why have debt at all?

A reasonable person, having seen the numbers above, asks why a government should carry debt in the first place. Why not pay it off, all $32 trillion, and never borrow again? The question deserves a real answer rather than a shrug, because the answer explains what "under control" should mean.

Start with what a Treasury security is to the rest of the financial system. It is the world's benchmark safe asset:

• Money-market funds hold it so that your cash is available tomorrow.
• Banks hold it as the liquid reserve regulators require.
• Pension funds and insurers hold it to match promises they have made decades out.
• The repo market, which moves trillions of dollars a day and keeps every other market functioning, runs largely on Treasury collateral.
• The Federal Reserve conducts monetary policy by buying and selling it.
• Every corporate bond, mortgage, and municipal loan in the country is priced as a spread over it.

A government with no debt has no Treasury market, and a modern financial system without one would have to invent something to replace it, on far worse terms.

This is not a thought experiment. In early 2001, with the budget headed for a fourth straight surplus, CBO projected surpluses large enough to retire all the debt that could be redeemed within about a decade. The Federal Reserve studied what it would use for open-market operations, and what banks and pension funds would hold, in a world without Treasuries. Then a recession, two rounds of tax cuts, two wars, a new drug benefit, and a financial crisis arrived, and the question answered itself. But it was a genuine worry among serious people, and it should tell you that zero is not the goal.

The second reason is that debt is a shock absorber. Wars, recessions, and pandemics arrive without warning and cost a great deal, all at once. A government that could not borrow would have to raise taxes or cut spending in the middle of the emergency, deepening it. A government that can borrow spreads the cost over the years and the generations that benefit from the emergency having been handled. This is called tax smoothing, and it is the same logic by which a family takes a mortgage on a house rather than saving for thirty years and buying it in cash. Borrowing to buy a house you will live in for decades is sensible. Borrowing to pay this month's groceries is not. The federal analog of a mortgage is a war or a bridge; the analog of groceries is a permanent structural gap between what the government promises and what it collects. The problem is not that we have a mortgage. It is that we are putting groceries on the card.

The third reason is arithmetic. Paying off $32 trillion in fifteen years would take total surpluses averaging about 5 percent of GDP. That means swinging the primary balance by nearly 10 points of GDP from where it stands today. It would take a decade and a half of tax increases and spending cuts far larger than anything in the plan below, sustained through at least one recession, with the deflationary effect of pulling that much money out of the economy every year. The United States has reached zero exactly once: in January 1835, when Andrew Jackson finished paying off the roughly $58 million he had inherited, in an era when the federal government was a rounding error in the economy. Britain spent the century after Waterloo working its Napoleonic debt down, and never to zero. No large modern economy has done it, and none has tried.

So the right question is never "debt or no debt." It is the question a lender asks a business: is the debt growing faster than the borrower's capacity to carry it, and what is it being used for? A growing company with modest permanent leverage is normal and healthy. A company borrowing every year to cover operating losses, at rising rates, is not, whatever its history. That is the standard we are going to hold the federal government to, and it is why the targets below are expressed as ratios and trajectories, not a number to be paid down to zero.

What "under control" means

There are two different goals people mean when they say "fix the debt," and they cost very different amounts.

Stabilization means debt-to-GDP stops rising. From the equation, that requires a primary deficit no larger than (g − r) × debt-to-GDP. With g − r at 0.9 and debt at 100 percent, a primary deficit of about 0.9 percent of GDP holds the ratio flat. That is roughly a $700 billion-a-year improvement on where we are today. But the cushion is shrinking. Once r catches up with g in the mid-2030s, stabilizing takes a balanced primary budget, and after that a surplus.

Reduction means the ratio falls, which requires a primary balance better than the stabilizing level. Since nobody controls r − g, the only way to be sure of that is a primary surplus.

Our goal is to stabilize first and reduce after, with three targets:

1. Cut the primary deficit by at least 2 percent of GDP by 2031, enough to stop the debt ratio rising within five years of enactment.
2. Reach a primary surplus by the mid-2030s, growing toward 1 percent of GDP as the slower benefit changes mature.
3. Have debt-to-GDP falling every year from 2034 on, with the debt brake in Step 7 to push it there if interest rates do not cooperate, up to a limit we explain below.

Our central projection uses CBO's interest rates. In it, debt stops rising at about 107 percent of GDP in 2033, slips below 100 in the mid-2040s, and reaches about 86 percent by 2056. The full range, including what happens if rates stay where they are this autumn, is in What it does to the debt.

Why 3 percent? Because in the long run, debt-to-GDP settles at the deficit divided by nominal growth. A 3 percent deficit with nominal growth near 3.7 percent, CBO's long-run figure, converges toward about 80 percent of GDP. That is roughly where the country sat just before the pandemic, and a level it carried without strain. Three percent is the number in the bipartisan resolutions now pending in both chambers and the number the Treasury Secretary has publicly committed to. In Warren Buffett's half-serious proposal, it is the deficit above which every sitting member of Congress should be ineligible for re-election. We did not invent it.

Here we have to be straight about timing. The resolutions want 3 percent by the end of fiscal 2030. A plan that phases in gently cannot get there that fast, and ours does not. Under CBO's interest rates, our central projection has the deficit falling from about 6.2 percent of GDP in fiscal 2026 to under 4 percent by 2036, and to 3 percent by the mid-2040s. Reaching 3 percent by 2035 would take roughly another 1 percent of GDP a year, about $3 trillion more over the decade, on top of a package that is already larger than anything Congress has passed. We think a slower path that can pass beats a faster one that cannot.

Why not balance the budget? Because it is unnecessary and, done quickly, recessionary. Once the primary balance is comfortably in surplus, the debt ratio falls even while the overall budget runs a modest deficit, the way a growing company can carry modest permanent debt. The record shows it: debt-to-GDP fell from 106 percent to 23 percent between 1946 and 1974 while the government ran an overall deficit in most of those years. It ran a primary surplus in most of them, which is the part that matters. Fast growth helped, as did the inflation and capped interest rates described earlier.

Two comforting stories, and where the money is

Before the plan, we need to retire the two most popular fantasies, one from each side of the aisle.

"Just tax the rich." The top 1 percent of households earn roughly a fifth of all income, call it $3.5 trillion. Raising their effective tax rate by ten full points, across every dollar and with no behavioral response, produces about $350 billion a year. The deficit is $2.0 trillion. Even confiscatory rates on the top do not close it, and real-world avoidance shrinks the take well before you get there. Taxing the top more is part of any honest plan. It is not most of one.

"Just cut the waste." All non-defense discretionary spending, which is every federal agency, program, and grant that is not the military or an entitlement, totals about $950 billion. Eliminating it entirely, every national park, air-traffic controller, and FBI agent, would close less than half the gap. The federal government is not primarily a collection of agencies. It is, in the line usually credited to former Treasury under secretary Peter Fisher, an insurance company with an army.

Here is what the government actually spends its money on:

Any plan that leaves the three big programs untouched is arithmetic theater. Likewise, revenue is about 17 percent of GDP against spending near 24, and no advanced economy closes a gap of nearly seven points with income taxes alone. Every other advanced economy also taxes consumption broadly, through a VAT or a goods-and-services tax, and we are the outlier. So the plan has to touch the big three programs and the broad tax base. The only real question is how.

The plan: design rules

The fiscal deals that worked, from the 1983 Social Security rescue to the 1990 budget agreement and the 1997 balanced-budget deal, shared a few features, and we adopted them as constraints.

No group bears more than about 40 percent of the adjustment, because otherwise the coalition that has to pass it does not exist.

Nobody within ten years of full retirement age, 67 today, sees a cut in scheduled benefits or a later eligibility age. The moral idea is simple: you can only change promises to people who have time to adjust. The 1983 rescue is the model. It raised the retirement age in a law passed in 1983 that did not begin to bite until 2000.

Two changes do reach today's retirees, and we flag them wherever they come up:

• The switch to chained CPI for the annual cost-of-living adjustment.
• Higher Medicare premiums for those with high incomes. The 1983 law set the precedent for this one too: it began taxing benefits for higher-income retirees the following year.

The VAT will also raise prices once. Because Social Security is indexed to prices, checks rise with that one-time step, though retirees spending out of savings will pay the VAT like everyone else.

Everyone gives something and gets something. The sweeteners in Step 6 are not pork but the price of passage.

Eligibility ages and benefit formulas phase in over fifteen to twenty years; tax changes do not. Slow benefit phase-ins protect people's retirement plans and spread the political cost across many Congresses. Taxes start within one to two years, because the arithmetic needs the money early, and because tax rules, unlike benefit formulas, can fairly change with a year's notice.

Only revenue that requires an act of Congress counts, for reasons the tariff episode makes plain in Step 7.

Score it conservatively. We assume enactment in 2028, with most tax changes taking effect in January 2029 and the VAT in January 2030. We also take an allowance off the tax items for the ways they interact. If enactment slips a year, every date below slips with it.

What follows is organized the way a bill would be: the two big programs first, then revenue from the top, then revenue from the base, then discretionary spending, then the offsets, then the rules that hold it together.

Step 1: Social Security

Social Security is pay-as-you-go. Today's workers pay a 12.4 percent payroll tax, split between employer and employee, on wages up to a cap of $184,500, and that money pays today's retirees. The surpluses that built up from the mid-1980s to 2020 were parked in trust funds. Costs now exceed income and the funds are being drawn down.

The retirement fund is projected to run dry in the fourth quarter of 2032. At that point the law allows only what is coming in to be paid out. That means an automatic, across-the-board cut of about 22 percent to every retirement and survivor check, with no vote. The combined retirement and disability funds would follow in 2034.

The trustees' latest estimate of the 75-year shortfall is 4.42 percent of taxable payroll, up sharply from 3.82 percent a year ago. The biggest reason is that they cut their long-run fertility assumption from 1.9 to 1.75 children per woman. Lower immigration and last year's tax law, which reduced income taxes on benefits, account for most of the rest.

1a · Raise the full retirement age from 67 to 70, two months per birth year, starting with people born in 1972, roughly everyone under 55 today. The age reaches 70 for people born in 1989. Each one-year increase is roughly a 6 to 7 percent cut in monthly benefits at whatever age a person claims, so this is a benefit reduction disguised as a calendar change.

The case for it is longevity: life expectancy at 65 is five to six years longer than when monthly benefits began in 1940. The objection is that it is regressive, because longevity gains have gone mostly to higher earners and a 68-year-old roofer cannot work three more years. Our answer is to pair it with a hardship pathway through disability insurance for physically demanding occupations, and with the progressive formula change in 1d.

Ten-year savings are tiny, around $10 billion, because the first affected group does not reach 62 until 2034. Our first version counted $100 to $350 billion here, which assumed a start date our own hold-harmless rule forbids. The long-run effect is large: on the actuaries' arithmetic, roughly a quarter of the 75-year gap.

1b · Eliminate the payroll-tax cap, with no additional benefit credit above the current cap. Earnings above $184,500 currently pay no Social Security tax at all. Applying the 12.4 percent rate to every dollar of wages, starting in January 2029, is the largest revenue item available inside the program.

It goes further than restoring the old balance. In 1983 about 90 percent of all wages fell under the cap; today about 82 percent do, because earnings at the top have grown fastest. This would tax all of them.

The objection is a fair one: it turns Social Security from insurance, where contributions relate to benefits, into a plain tax on high earners. That is true, and we would rather say so plainly than let the fund run dry. It also stacks. Combined with the top bracket in 3g, a dollar of wages above $600,000 would face a federal rate of roughly 55 percent before state taxes, against about 36 percent on large capital gains. That gap is an invitation to relabel pay as gains, which is why 3e matters and why we take our largest allowance for avoidance on these items in Adding it up.

Ten-year revenue is about $1.5 trillion. That is deliberately cautious. Scaling CBO's 2024 estimate for a narrower version, which taxes only earnings above $250,000, would put full repeal closer to $2 trillion. On the actuaries' arithmetic, 1b alone closes more than half of the 75-year gap. Load-bearing wall.

1c · Switch cost-of-living adjustments to chained CPI, for Social Security and the other federal benefits indexed to prices. Benefits rise each year with a consumer price index. The chained version accounts for people substituting cheaper goods when prices rise and runs about a quarter point lower per year. Most economists consider it the more accurate measure, the tax code already uses it, and a quarter point that compounds for decades is real money.

The objection is that a quarter point a year is a 5 percent cut after twenty years. It lands hardest on the very old, whose spending is heavy on medical care that chained CPI may not capture well. Our answer is a small "longevity bump" at age 85 that offsets the compounding for the oldest beneficiaries.

The switch starts with the adjustment paid in January 2029. Cost-of-living adjustments still pass through the VAT's one-time price step from Step 4 in full. Ten-year savings are $200 to $250 billion, most of it in Social Security.

This is a load-bearing wall for a reason that has nothing to do with the money. It is one of only two items that reach current retirees, and it is what makes the rest of the package look fair to everyone else.

1d · Make the benefit formula more progressive, for people born in 1972 or later. Benefits are computed by applying replacement rates to slices of a worker's average lifetime earnings: 90 percent of the first slice, 32 percent of the next, 15 percent of the top. Reduce that top factor from 15 to 10, phased in over ten birth years. It trims benefits only for the highest earners, the same people who gained the most from the longevity math behind 1a. Ten-year savings are close to zero, because the first affected group reaches 62 in 2034, and the long-run effect is modest.

1e · Raise the payroll tax rate by half a point, a quarter point each on workers and employers, starting in 2030. This is the piece that finishes the job. Without it, the changes above leave Social Security close to solvent but not quite there; with it, they close the 75-year gap with a little room. For a worker earning $60,000, it costs $150 a year. Ten-year revenue is $500 to $600 billion.

Inside the ten-year window, Step 1 comes to about $2.3 trillion, nearly all of it from the payroll tax changes in 1b and 1e. With those in place, the retirement fund no longer runs dry in 2032. The benefit changes do their work after 2035, which is what the hold-harmless rule requires, and by the 2050s they are among the largest items in the plan.

Step 2: Medicare

Medicare Part A, which pays hospitals, is funded mainly by a 2.9 percent payroll tax. Its trust fund is projected to run short in the second quarter of 2033, when incoming revenue would cover about 89 percent of costs. Parts B and D, which pay doctors and for drugs, are about 75 percent funded from general revenue, with premiums covering the rest. That general-revenue share is where Medicare quietly drains the Treasury. Medicare is also the program whose future cost is least knowable: medical inflation of 5 to 6 percent a year breaks any plan, while a technology-driven slowdown would rescue one.

2a · Raise the eligibility age from 65 to 68, two months per birth year, for the same people as 1a: those born in 1972 or later. The first affected group turns 65 in 2037, and the age reaches 68 for people born in 1989. The Medicare and Social Security ages move in lockstep, so the two-year gap between them stays what it is today. People in the gap get a subsidized bridge, described in Step 6, so the change saves money rather than simply shifting costs onto 66-year-olds.

Ten-year savings are close to zero. Once fully phased in, it saves something like 0.1 to 0.2 percent of GDP a year, net of the bridge; CBO's estimates for a smaller increase, to 67, found only modest net savings, because so many people in the gap move to other subsidized coverage. It is also deeply unpopular, and because it saves almost nothing inside the window, it will be the first bargaining chip a negotiator reaches for. Trade it only for something that saves as much in the 2050s.

2b · Expand income-related premiums for Parts B and D. The standard Part B premium covers about a quarter of the program's cost. Beneficiaries with incomes a little above $100,000 for a single person already pay surcharges that raise their share to between 35 and 85 percent, but fewer than one in ten pay them.

The change has two parts. First, raise each surcharge tier by 10 points of cost, so the top tier pays 95 percent. Second, freeze the income thresholds for ten years, so that by the mid-2030s roughly one beneficiary in six pays a surcharge.

This is the second of the two items that reach current retirees, and it reaches only those with high incomes; Medicaid continues to pay premiums for low-income enrollees. Ten-year savings are $100 to $200 billion, phased in from 2029. Our first version counted $300 billion. Raising that much would also mean higher premiums for middle-income retirees, which this plan does not do.

2c · Expand drug-price negotiation. The 2022 law lets Medicare negotiate prices on a capped number of drugs, rising to 20 a year. A drug becomes eligible once it has been on the market for seven years, or eleven for biologics. Keep those waiting periods, which is where most of the innovation incentive lives, and remove the cap. Any drug past its waiting period with more than $200 million a year in Medicare spending becomes eligible. That ends more of the arrangement in which the United States pays two to three times what other rich countries pay for identical molecules. Ten-year savings are $200 to $400 billion.

2d · Site-neutral payment. Medicare pays more for the same procedure in a hospital outpatient department than in a doctor's office. Paying the same rate for the same service ends the incentive that has driven hospital systems to buy physician practices purely to reclassify their billing. It is a payment cut for hospital outpatient departments, and we say so; it phases in over four years. The Medicare payment advisory commission has recommended this for years. Ten-year savings are $200 to $250 billion, and the hospital lobby will fight it hardest.

2e · Fix Medicare Advantage coding. Private Medicare Advantage plans are paid more for sicker enrollees, as measured by diagnosis codes the plans themselves submit, and they have become very good at finding diagnoses. Medicare already trims plan payments by 5.9 percent to offset this. The advisory commission estimates the real coding difference is closer to 10 percent.

The fix closes that gap and adopts the commission's recommendations: base payments on two years of diagnoses, and ignore diagnoses that turn up only in the plans' own home visits. That saves $350 to $500 billion over ten years.

Plans will respond partly by trimming the extras they advertise, such as dental and gym benefits, and they will call it a cut to seniors. But the overpayment is what pays for those extras. Because it also inflates the standard Part B premium, the fix lowers premiums for everyone else. The large insurers will fight it as hard as the hospitals fight 2d.

Step 2 comes to about $1.1 trillion over ten years, most of it from what Medicare pays hospitals, insurers, and drug makers rather than from what beneficiaries pay.

Step 3: Capital and the top

This is the part of the plan people expect. It is necessary and it is not sufficient. Keep the subtotal, about $1.8 trillion over ten years, in view against a deficit of about $2 trillion a year. Everything here takes effect in January 2029.

3a · Two new long-term capital gains brackets: 28 percent above $1 million of gains, 32 percent above $5 million, with the existing 3.8 percent investment-income surtax on top. Long-term gains are taxed today at 0, 15, or 20 percent depending on income. The tax is due only when you sell, which is why raising the rate too far causes people to hold forever and revenue to fall. Estimates of the revenue-maximizing rate cluster in the high twenties to low thirties, and that is where we stop. Ending the step-up in 3b pushes that ceiling higher, because the option of holding until death is what makes gains taxes so easy to avoid. Ten-year revenue is $100 to $200 billion, modest and mostly symbolic.

3b · End the step-up in basis at death, with a $5 million exemption per person and installment terms for farms and family businesses. Today, when you die, the cost basis of your assets resets to market value, and a lifetime of gains is never taxed. Combined with borrowing against appreciated assets while alive, this is the "buy, borrow, die" strategy. Of every item in the tax code, this is the hardest to defend on any principle and the cleanest to fix. Gains would be taxed at death as if sold, and the tax would be deductible against the estate tax so the same dollar is not taxed twice. Ten-year revenue is $200 to $300 billion. Load-bearing wall.

3c · Estate tax: exemption back to $5 million per person, indexed, rate to 45 percent, and close the trust-based workarounds. Wealth-transfer taxes have among the lowest economic costs of any tax; nobody works less because of a tax paid after dying. The current exemption of $15 million per person means the tax barely exists. The bigger leak is the planning industry: grantor trusts, dynasty trusts, and valuation discounts, which a reform would need to address directly. Ten-year revenue is about $300 billion.

3d · Corporate rate from 21 to 25 percent. Estimates of how much of the corporate tax ultimately falls on workers through lower wages, rather than on shareholders, range from about a fifth to more than half, so this is less of a pure tax on capital than it looks. Above the high twenties, profit-shifting and inversions begin again; 25 is comfortably below that and well below the pre-2017 rate of 35. Full expensing for research and equipment stays in place to protect investment. Ten-year revenue is $450 to $500 billion.

3e · Close carried interest, tighten the pass-through deduction, and cap like-kind exchanges. Carried interest lets private-equity and hedge-fund managers treat their compensation as capital gains. The pass-through deduction gives a 20 percent deduction to business income earned through partnerships and S corporations, which mostly benefits high earners. Like-kind exchanges, limited to real estate since 2018, let investors defer gains indefinitely by swapping properties; we would cap the deferral at $500,000 a year. With top wages taxed far more heavily than gains under this plan, these are the doors through which pay gets relabeled, and closing them protects the revenue in 1b and 3g. Ten-year revenue is $150 to $250 billion combined.

3f · Raise the minimum tax on U.S. companies' foreign profits to 15 percent, from about 12.6 percent today. Multinationals pay a reduced rate on foreign earnings. Raising it to the 15 percent floor that most large economies have now adopted removes much of the remaining incentive to book profits abroad. Ten-year revenue is about $200 billion.

3g · Top income-tax bracket back to 39.6 percent above $600,000 for married couples. This restores the pre-2017 top rate at roughly the pre-2017 threshold, adjusted for inflation. Together with 1b it produces the stacked rate on top wages described above, which is one more reason the avoidance allowance on these items is our largest. Ten-year revenue is about $250 billion.

Step 4: The broad base

This is the hard part, and the part that separates a real plan from a campaign speech.

4a · A 5 percent federal value-added tax, with a rebate for the bottom 40 percent of households, starting in January 2030. A VAT is a sales tax collected in stages. Each business pays tax on its sales and receives credit for the tax paid on its purchases, so the tax lands on final consumption and is much harder to evade than a retail sales tax. Every other developed country has one, and a U.S. rate of 5 percent would still be a quarter of the European average.

It is the one lever that raises money at the scale of the problem, and it does not penalize saving or investment the way income taxes do. In the OECD's ranking of taxes by their drag on growth, only property taxes do less harm. The objection is that it is regressive, since poorer households spend a larger share of their income. The answer is the rebate described in Step 6, which makes the net effect progressive through the middle of the distribution.

Three more things to know:

• It needs time to build. The collection machinery takes about eighteen months, which is why it starts in January 2030.
• It raises prices once. It produces a one-time step up in the price level of a few percent, which the Fed would need to look through rather than fight. With inflation above target this year, that is not a small ask, and it is another reason not to start sooner.
• Retirees are compensated. Social Security and other benefits indexed to prices rise with that step. Scorekeepers handle this through a standard offset for the income and payroll taxes the VAT displaces, which comes to about the same thing, and our figures include it.

Ten-year revenue is $2.5 to $3 trillion before the rebate and about $2 trillion after it. The load-bearing wall of the entire plan.

4b · A financial-transactions tax: 0.1 percent on stock trades and 0.02 percent on the notional value of derivatives, starting in 2029. Ten cents per $100 traded is negligible for a long-term investor and existential for a high-frequency strategy that turns over its book daily. Supporters say it curbs churn; critics say it thins liquidity and pushes trading offshore, and several countries that tried one repealed it.

It is the most economically debatable item in this plan, kept because the money is large: $700 to $800 billion over ten years, in line with CBO's earlier estimates for a tax at this rate. CBO's newest version taxes trades at a tenth of this rate and raises about $300 billion; revenue rises far less than the rate because trading shrinks as the tax rises, which is also why we take a quarter off our figure in Adding it up. It is also a bargaining chip. We would trade it for two more points of VAT, with the rebate scaled up, which raises about the same.

4c · A carbon tax at $40 per ton with a border adjustment, on a trigger. A carbon tax charges fossil fuels in proportion to their carbon content. A border adjustment taxes imports the same way, so that domestic producers are not disadvantaged. Economists across the political spectrum favor it as the cheapest way to price emissions. But legislating a new fuel tax during an energy shock, with gasoline above $4 this spring, is a non-starter. So it is enacted with the package and starts in January 2030, or earlier if oil has averaged below $70 for a year. Ten-year revenue is $500 billion to $1 trillion, depending on when it starts and how fast emissions fall.

4d · Cap the tax exclusion for employer health insurance. Premiums for job-based coverage are exempt from income and payroll tax without limit. It is the largest tax break in the code. It is worth the most to people with high incomes and expensive plans, and by hiding the cost of those plans from workers, it pushes health spending up.

We would cap the exclusion at the 75th percentile of premiums in 2029, indexed to general inflation afterward. It binds first on the most expensive plans, and gradually on more as health costs outrun prices. Congress tried a version of this before, as the "Cadillac tax," and repealed it in 2019 before it took effect. This one works through ordinary income and payroll taxes, which makes it more transparent and harder to dodge. Ten-year revenue is $300 to $400 billion, rising over time, and the payroll-tax share flows to Social Security and Medicare.

That figure is deliberately cautious. CBO puts a similar cap, which bites sooner because it is pegged to 2026 premiums, at about $520 billion over 2028–2034 alone. Ours starts from a higher threshold, and the Cadillac tax's fate is a reason not to count every dollar.

Step 4 comes to about $4.6 trillion before the rebate in Step 6.

Step 5: Discretionary restraint

Discretionary spending is the roughly one-quarter of the budget that Congress appropriates each year: defense, and the agencies. CBO's baseline already assumes those appropriations grow only with inflation, so real savings have to come from growing slower than that. The rules:

• Caps for ten years, starting in fiscal 2029. Non-defense appropriations grow half a point slower than inflation; defense grows a full point slower.
• An audit with teeth. The Pentagon must pass an audit, with efficiency targets attached, as the means of finding the defense savings.
• War costs, handled honestly. Costs of active operations, including those against Iran, are funded outside the caps, as in Iraq and Afghanistan. But CBO must certify them as operational, which closes the old loophole of moving base spending into war accounts. And they count against the debt brake in Step 7, so a war that adds to the debt raises the VAT rather than disappearing from the books.

The 2011 caps, though loosened several times, restrained spending for most of a decade, and this is a milder version. Ten-year savings are about $800 billion, about two-thirds of it from defense. We put it last because it is the smallest lever, not because it is unimportant: it is what convinces people the entitlement changes are not the only sacrifice being asked.

Step 6: Sweeteners, and how the bridge works

You cannot take more than $10 trillion out of an economy and give nothing back, and no deal that passed ever tried. The 1983 Social Security rescue paired benefit taxation and payroll-tax increases with a retirement-age increase delayed so far into the future that nobody voting on it would feel it. The 1986 tax reform paired the closing of loopholes with lower rates. The 1990 agreement paired tax increases with spending rules both parties could point to. Sweeteners are how a package with something for everyone to hate becomes a package with something for everyone to defend. Here are ours, about $1.65 trillion over ten years, and exactly how each one is supposed to work.

The VAT rebate, about $750 billion over ten years. Every adult in a household in the bottom 40 percent of the income distribution receives $800 a year, and every child $400, in today's dollars. The IRS pays it quarterly as a refundable credit, the way Canada has paid its goods-and-services tax credit since that tax began in 1991.

The amounts are not arbitrary: at a 5 percent rate, $800 is the VAT paid on $16,000 of taxable spending per adult, roughly what a household in the bottom two quintiles spends on goods the tax reaches. The credit phases out between the 40th and 50th percentiles so there is no cliff, and it rises in step if the debt brake ever raises the VAT. Eighty-five million adults and twenty-five million children at those amounts comes to about $80 billion a year today.

The effect is that the bottom 40 percent pay no net VAT at all, the middle pays some, and the top pays it in full. That is what turns a regressive tax into a mildly progressive one.

The insurance bridge for people in the Medicare gap. The Medicare age change in 2a starts with people turning 65 in 2037, so the bridge costs almost nothing before then. Once the age change is fully phased in, it costs roughly $20 to $50 billion a year in today's dollars. It still deserves the most explanation, because it is the difference between a Medicare age increase that saves money and one that merely shifts costs onto 66-year-olds. Here is the mechanism.

Expanded earned-income and child tax credits, about $400 billion. The earned-income credit is a refundable payment to low-wage workers that grows with earnings and then phases out. The child credit is a per-child reduction in tax that is only partly refundable to families whose incomes are too low to owe much. We would raise the credit for childless workers, who currently receive very little, and make the child credit fully refundable. Together that puts about $40 billion a year back into the households most exposed to the new taxes. It is aimed at working families specifically because they carry the VAT, the extra half point of payroll tax, and the health-exclusion cap, and get little from Steps 2 and 3.

Permanent premium caps on the health exchanges, about $300 billion. Exchange subsidies have alternated between generous and stingy depending on which Congress last touched them. We would make the generous version permanent, so that nobody's net premium exceeds a fixed share of income and the cliff at four times the poverty line is gone for good. That is what makes the bridge in 2a credible; a 66-year-old is not going to accept a promise of exchange coverage whose subsidy expires with the next election.

Investment provisions for business, about $200 billion. Full expensing, the ability to deduct the cost of research and equipment in the year it is bought rather than over many years, is the single most pro-growth feature of the current tax code. It stays in place at the higher corporate rate. On top of it, an investment credit for domestic manufacturing gives the companies paying 25 percent instead of 21 something concrete to build with. The point is to make the corporate rate increase a tax on profits, not on investment.

Step 7: The rules

This step is short and, we think, the most important thing in the piece, because the history of fiscal repair is a history of good plans being undone.

Rule 1 · Only statutory revenue counts. In February, CBO had about $3 trillion of ten-year deficit reduction on its books from tariffs. One Supreme Court ruling, and the weaker replacement tariffs that followed, erased about a third of it within months; CBO's August update put the cost at roughly $0.9 trillion. The replacement tariffs rest on other grants of presidential authority and could change again. Revenue that exists at the discretion of one branch of government, or one court, is not revenue you can build a plan on. Nothing in this package rests on it, and if the remaining tariffs go too, the brake in Rule 2 is what responds.

Rule 2 · A debt brake. The equation in section one has a term Congress cannot control, r − g, so we pre-legislate the response:

• The test. From fiscal 2034, debt held by the public must fall as a share of GDP every year.
• The response. If it does not fall, the VAT rises by one point the following January, with the rebate rising in step. It keeps rising a point a year until debt is falling again, up to a ceiling of 10 percent.
• Reversal. Steps come back off, one point at a time, only once debt is below 90 percent of GDP and falling.
• Recessions. The brake skips any year the economy is in recession, and the year after, because you do not tighten into a downturn.
• No escape hatch. Nothing requires a vote in either direction, and waiving the brake takes 60 votes in the Senate.

Switzerland has run its federal budget under a constitutional debt brake since 2003, and Sweden under spending ceilings and a surplus target since the late 1990s. Ours is narrower, because it works through one tax that already exists. In our central projection it never fires. Under CBO's interest rates without credit for lower debt, it fires twice. If rates stay where they are this autumn, it reaches its ceiling by the end of the 2030s and holds debt near 113 percent of GDP instead of letting it climb. What it does to the debt shows all three.

Rule 3 · A commission with a fast-track vote. Bills to create a bipartisan fiscal commission have been introduced in both chambers in this Congress. The version that works is the one modeled on military base closures: the commission writes a single package, Congress gets one up-or-down vote, and no amendments are allowed. That is how you pass things nobody wants to own individually. Pair it with two-year budgeting, so that the annual shutdown theater stops consuming the calendar.

Rule 4 · The ten-year hold-harmless. Nobody within ten years of full retirement age sees a cut in scheduled benefits or a later eligibility age. Two changes reach today's retirees: chained CPI, and higher Medicare premiums for those with high incomes. Both are named in the plan, and nothing else may be added to that list. This is non-negotiable, and it is what lets everything else through.

Adding it up

The savings arrive quickly, because most of the money comes from taxes that start in 2029 and 2030. They reach about 0.7 percent of GDP in 2029, 2 percent by 2031, and roughly 2.4 percent, about $1.2 trillion, by 2037. After that they keep growing slowly as the retirement-age, formula, and Medicare changes mature, toward 3.4 percent of GDP by the mid-2050s. The chart below runs the primary balance forward, which is the number the plan actually controls.

Interest is the other half of the deficit. With less debt to service, and with the lower rates CBO's own models associate with less debt, interest costs in 2036 run near 3.9 percent of GDP instead of CBO's 4.6. The total deficit falls from about 6.2 percent of GDP in fiscal 2026 to about 3.8 percent in 2036, and to 3 percent in the mid-2040s. The 3 percent target is a destination, not a starting point, which is why we said so up front.

What it does to the debt

Here we have to be honest about what a plan can and cannot promise. We ran the package through the equation year by year, from 2026 to 2056, under four assumptions about r − g. The policy is identical in every line; the only thing that changes is the interest rate the market charges.

• Today's cushion persists. r − g stays at −0.9. The optimistic case.
• CBO's rates, with CBO's credit for lower debt. r − g rises from −0.9 to zero by 2035 and to about +0.6 by the mid-2050s, as in CBO's projections. CBO's models also say that each point of debt-to-GDP adds about 2 basis points to interest rates. We give the plan three-quarters of that, a year late, to allow for the time it takes old debt to roll over. This is our central case.
• CBO's rates, no credit. The same path, with no allowance for the lower debt.
• Rates stay high. All interest rates run about 0.6 points above CBO's path and work their way through the debt by the early 2030s. That is a moderate reading of this autumn: the 10-year yield is nearly a full point above CBO's forecast for the decade, short rates less. CBO's own October analysis finds that with no plan, rates half a point above its forecast would put debt at 124 percent of GDP in 2036, and rates 1.5 points above would put it at 133.

We run the no-plan baseline at CBO's rates and at the high rates. The debt brake operates on every plan line, as it would in law. We also show the high-rate case without it, so you can see what the brake is worth.

Read down the rows. In the central case, the plan stops the debt ratio from rising at about 107 percent of GDP in 2033. It brings the ratio below 100 in the mid-2040s and to the mid-eighties by 2056, against 180 percent with no plan. The brake never fires.

The other two CBO-or-better cases bracket it. If today's cushion lasts, the same plan takes debt back to the low seventies by 2056, roughly where it stood a decade ago. Give the plan no credit for lower debt and the brake fires twice, the VAT settles at 7 percent, and debt declines more slowly, to the high nineties.

If rates stay where they are this autumn, the plan on its own cannot hold the line: debt creeps up to about 123 percent by 2056. The brake takes the VAT to its 10 percent ceiling by the end of the 2030s and holds debt near 113 percent. That sounds like failure until you run the no-plan baseline at the same rates: it passes 200 percent of GDP in the mid-2050s. In that world Congress would have to go further than the brake can on its own, but it would be starting from 113, not from 200.

That is the honest answer to "does this fix it." The plan controls the primary balance, and the primary balance is necessary, but the market decides how much credit you get for it. Which is exactly why Rule 2 exists.

One more subtlety, and it is the most important one for investors. The relationship runs both ways. A credible plan lowers r, because the term premium investors demand on long Treasuries is partly compensation for fiscal risk. A 30-year yield near 5.6 percent is, in part, the market's vote on the current trajectory. Our central case gives the plan only the credit CBO's models imply; it gives none for a smaller term premium. Enact something like this and the bad rows become less likely on their own. Do nothing and they become more likely. Fiscal credibility is a self-fulfilling asset, and its absence is a self-fulfilling liability.

What could go wrong

The Medicare cost curve. This is the single biggest unknown. If medical inflation returns to 5 or 6 percent a year, the plan needs a second round of health reform. If new drugs and automation bend costs down even a point a year, the plan over-delivers. The weight-loss drugs are the obvious hope. So far, though, CBO has found that covering them for obesity would raise federal spending rather than lower it, because the savings from healthier patients arrive slowly and fall short of the cost of the drugs. We would call it a coin flip.

Interest rates. The high-rate row above is not a stress test invented for the occasion; if anything, it is a little milder than where long rates sit this autumn. If they stay there, the brake reaches its ceiling and Congress has to do more.

Growth. CBO assumes real growth of about 1.8 percent a year after 2026, slowing to 1.5 percent by the 2050s, largely because the workforce is aging and immigration has fallen. Every half point of extra trend growth improves the thirty-year path by around 20 to 25 points of GDP. Immigration policy, permitting, and whether AI productivity gains show up in the aggregate data are the levers. We assumed nothing from any of them, which is the correct amount to assume when building a plan and probably the wrong amount when forecasting.

War and energy. The war with Iran has already pushed oil, inflation, and interest rates up this year. A longer war adds to the debt directly, and an energy shock that keeps inflation high keeps rates high. Because war costs count against the debt brake, they cannot hide, but they still have to be paid for.

Recessions. You do not tighten into a downturn. The plan starts in 2028 and phases in, the automatic stabilizers stay in place, and the brake skips recession years. A recession in the phase-in years delays the path by two or three years; it does not invalidate it.

Political durability. In 2001 the government had a surplus and a projected path to zero debt. Within three years, tax cuts, two wars, and a new drug benefit had reversed it. Surpluses get spent. That is not cynicism; it is the empirical record, and it is why Rules 1 through 4 are in the plan rather than left to good intentions.

Why it hasn't happened

No piece of this plan could pass the current Congress on its own. Each of the biggest pieces has entrenched opponents:

• The VAT. Much of the right sees it as a money machine for bigger government, and much of the left sees it as a tax on the poor.
• The carbon tax. Carbon pricing has been politically radioactive since the climate bill died in 2010.
• The entitlement changes. They are toxic to whoever moves first.

The bond market, while grumbling, has not forced the issue. Auctions clear, the dollar remains the reserve currency, and a 30-year yield near 5.6 percent is uncomfortable but not yet a crisis.

Every major fiscal deal in modern history came with a deadline attached. The 1983 rescue happened months before the Social Security checks would have bounced; the 1990 and 2011 agreements came under debt-ceiling and market pressure. The next deadlines are already on the calendar.

What would make it happen sooner is the thing nobody wants: a string of weak Treasury auctions, a sharp move in long rates, or a ratings action that forces the question. This autumn has offered a preview of the second. What would make it happen later is a growth surprise that makes the problem look smaller for a few more years. Our guess is that the forcing event arrives before the trust-fund date, not after. We also expect the package that passes to look a great deal like the one above, assembled in a hurry by people who will insist they never wanted any of it.

What this means if you own assets

We will keep this general, because it is education and not advice. But readers of this newsletter are not here for civics.

The long end of the Treasury curve is where the fiscal story gets priced. Long-term yields have risen much more than the Fed's rate this year. Estimates of the term premium, the extra yield investors demand for holding long bonds instead of rolling short ones, have climbed to levels not seen in more than a decade. Part of that premium is fiscal risk. A credible plan compresses it; continued drift widens it. Long-duration assets of every kind, bonds and the equities valued off them, are exposed to which way that goes.

The inflation path is the other channel, for the reasons in the section on the dollar and the one on the printing press. Countries that borrow in their own currency have often resolved this position through a decade of inflation running above the rate they pay on their bonds, quietly transferring wealth from savers to the Treasury.

The Fed under its new chair has so far acted as if it will not play that role, raising rates in September with the deficit above 6 percent of GDP. His talk of a new Fed–Treasury accord is the thing to watch. If that resolve holds, the adjustment comes through real interest rates rather than through inflation. That is a different world for asset prices than the one most of us grew up in: a better one for anyone holding cash, and a worse one for anyone holding duration.

The tax items are worth reading twice. Last year's law raised the estate exemption to $15 million and kept the step-up in basis, and provisions like these tend to flip when control of Congress flips. Anyone doing multi-decade planning should treat the current rules on step-up, estate exemptions, and capital-gains rates as among the most likely to change, in either direction.

The best outcome for asset owners is a boring one. A phased, credible plan passed before a crisis produces lower long rates, a smaller risk premium, and a Fed that can actually cut into a slowdown. The worst outcome is a plan written in a week, under duress, after a market event. History says we usually get the second one. The arithmetic says we do not have to.

Sources and a note on the numbers

Budget figures:

• Current-year deficit, revenue, and interest figures are from CBO's Monthly Budget Review for September 2026 and its February 2026 Budget and Economic Outlook: 2026 to 2036.
• Long-term figures are from CBO's long-term budget outlook data for 2026 to 2056.
• The effect of the tariff ruling is from CBO's updated projections of tariffs as of July 31, 2026, published August 20.
• Debt-limit timing is from the National Taxpayers Union and the Bipartisan Policy Center.

Social Security and Medicare:

• Trust-fund dates and the 4.42-percent-of-payroll shortfall are from the 2026 Social Security and Medicare Trustees Reports.
• Estimates for individual Social Security changes draw on the actuaries' provision estimates.
• The Medicare Advantage figures draw on MedPAC's March 2026 report and CRFB's overpayment estimates.

Markets and rates:

• Yields are Treasury constant-maturity data via FRED.
• The Fed's rate is from its September 16, 2026 policy statement.
• The high-rate scenario follows CRFB's estimate of what persistently higher rates would cost. CBO's October 8 analysis of two higher-rate scenarios is the official benchmark.

Legislation and overviews:

• The bipartisan 3-percent resolutions are S.Res.654 and H.Res.981.
• The commission bills are summarized by the Committee for a Responsible Federal Budget.
• The Brookings update on the federal budget outlook (March 2026) is the best single overview of the long-run arithmetic.

The ten-year figures for individual policies are estimates assembled from several sources and scaled to the 2028–2037 window: CBO's Options for Reducing the Deficit, the actuaries' estimates, MedPAC, and analyses by nonpartisan budget groups. They are static, rounded, and should be read as accurate to roughly a third. The debt trajectories in What it does to the debt are our own calculations; they are illustrative, not forecasts, and anyone with a spreadsheet can reproduce them from the inputs below.

Disclosure: our model portfolios hold Treasury securities and other assets whose prices are sensitive to the interest-rate and fiscal developments discussed here; subscribers receive every position and every change as it happens. Nothing here is individualized investment advice, and nothing here is a prediction of what Congress will do. Our full Disclosure Policy applies.

Tools we use — see our full toolkit (some links earn us a commission).