Macro at a glance
Trajectory Enquiries
Six questions people ask about the US federal debt, answered in order, one number each. Every figure comes from the Treasury, the CBO or the Fed and carries its own date; the core figures update live from Treasury on each visit. The last question has no single number, so it gets a map instead.
Line 1How big is it?
Stock
How big is it?
StockTotal public debt outstanding: the debt held by the public plus what the government owes its own trust funds. It crossed $40 trillion on 19 August 2026. The part that trades in the market, marketable debt, is about $31.8 trillion.
Which debt? Three measures circulate and they differ by trillions. Total public debt ($40.1T) is the headline. Debt held by the public is the one economists compare with GDP. Marketable debt is what reprices at auction. This page names the measure every time.
What would change this answer: debt held by the public falling as a share of GDP for two fiscal years in a row. The dollar total almost never falls, so the size of the economy is the fair test: between 1946 and 1974 the ratio fell from 106% to 23% while the dollar debt kept rising. Check it against OMB Historical Table 7.1 or CBO each February. The strongest opposing case: size alone does not decide solvency. Japan carries more than twice the US ratio and still borrows cheaply, because who holds the debt, and in what currency, matters as much as how big it is.
Receipts: Treasury FiscalData Debt to the Penny, MSPD Table 1 · CBO, The Budget and Economic Outlook 2026–2036
Line 2How fast is it growing?
Pace
How fast is it growing?
PaceAt the last twelve months' pace, the debt adds another trillion dollars about every four and a half months. Going from $5 trillion to $6 trillion took about six years (1996–2002).
Days taken to add each trillion, $2T to $40T
Each bar is the gap between one trillion-dollar crossing and the next. The bars are jagged, not smooth: debt-ceiling suspensions produce catch-up bursts ($25T arrived 28 days after $24T). The trend through that noise is the point. Hatched bars ($2T to $5T) are estimates: Treasury's daily record starts in April 1993, so the $1T to $4T crossing dates are drawn on a straight line between fiscal-year-end totals. Treat them as accurate to a few months: the same method puts $5T in November 1995, but the daily record shows February 1996, held back by the 1995–96 debt-limit standoff. Before any of this, the first trillion took about 192 years, from 1790 to around October 1981.
What would change this answer: the trailing twelve-month pace lengthening past 250 days per trillion and staying there for two quarters, with no debt-ceiling suspension inside the window. That would mean the deficit itself is shrinking, not just being delayed. The strongest opposing case: a trillion is not a fixed yardstick. It was about a third of the economy in 1981 and is about a thirtieth now, so crossings would come faster even with a stable deficit. Line 1's ratio to GDP is the fairer measure of whether the pace is a problem.
Receipts: Treasury FiscalData Debt to the Penny, 1 Apr 1993 – 2 Sep 2026 · Historical Debt Outstanding (fiscal-year-end totals, 1790–1992, for the estimated bars)
Line 3How soon does it reprice?
Structure33% within a year
How soon does it reprice?
Structure33% within a yearAbout a third of marketable debt matures within twelve months, so about a third of the book picks up a new interest rate inside a year. Bills, which roll every few weeks, make up most of it. In September 2014 the share was 24%.
Why this line comes before price: structure decides how fast a change in market rates reaches the government's bill. A short, bill-heavy book passes a rate rise through in months, not decades. Over the past three years the book has been getting shorter, not longer: Treasury's average maturity has slipped from about 75 months at its 2023 peak to 70.0 months in July 2026.
What would change this answer: the bill share back at or below 20% of marketable debt for two quarters (MSPD, monthly), and average maturity lengthening in two successive Treasury refunding presentations. That would mean Treasury is locking in term again, and a rate rise would take longer to reach the bill. The strongest opposing case: a short book is not only a risk. Bills have often been the cheaper way to borrow, and if rates fall, a short book passes the saving through just as quickly.
Receipts: MSPD Table 1 · Treasury presentation to TBAC, Q3 2026 (pp.23–24) · GAO-26-107529 (Figs 4–6)
Line 4What does new debt cost?
Price
What does new debt cost?
PriceThis is the gap between what the market charges for new ten-year money (5.26%) and the average rate the government pays across all its interest-bearing debt (3.49%). While that gap stays positive, every refinancing raises the average.
Average rate paid on the whole debt, 2001–2026, against today's 10-year
September readings each year, plus the January 2022 low and the latest month. The dashed line is today's 10-year yield for comparison only; it is not a history. The yields above read live from Treasury's daily par yield curve on each visit.
What would change this answer: the 10-year falling below the average rate paid. Then refinancing would start lowering the bill, as it did through most of 2001–2021. Or, more quietly: the average rate paid stops rising for several months while the 10-year stays above 4.5%. That would mean the book is absorbing higher rates better than the repricing story assumes, and the cost is proving chronic rather than acute.
Receipts: Treasury daily par yield curve (live, home.treasury.gov) · FiscalData Average Interest Rates on Treasury Securities
Line 5What is it crowding out?
Squeeze54.7% of the deficit
What is it crowding out?
Squeeze54.7% of the deficitThis is the share of FY2025's federal deficit that was net interest. In FY2009 the share was 13%. More than half of this year's new borrowing is now borrowing to pay interest on old borrowing.
Net interest as a share of all federal outlays, FY1940–FY2025
The share view allows for growth and inflation: today's 13.84% is high, but still below the FY1996 peak of 15.45%. Switch to dollars to see the same series before that adjustment.
One source at a time: Treasury's gross interest, CBO's net interest and this page's implied run-rate (debt × average rate ≈ $1.40T) are three different concepts. They are never added together or swapped mid-sentence.
What would change this answer: net interest's share of all federal outlays falling for two fiscal years in a row, as it did after the FY1996 peak, or CBO's next baseline moving the FY2036 "cents left" figure back above zero. Either would mean revenue or growth is outrunning the interest bill. Check the first against OMB each year and the second at each CBO baseline (usually January–February). The strongest opposing case is already on this line: at 13.8% of outlays, interest is still below the FY1996 peak of 15.5%, and after that peak the share fell for eight straight years, to 7.0% by FY2004.
Receipts: OMB Historical Tables 1.1, 3.1 and function 900 (net interest FY1940–FY2025; chart folded in from Red Ink Rising, 28 Sep 2026) · CBO Feb 2026 baseline, Table 1-1 · CBO Historical Budget Data (Feb 2026 vintage), FY1962–FY2025, and CBO Feb 2026 baseline, FY2026–FY2036 (share-of-revenue view, folded in from The Wiggle Room, 30 Sep 2026) · deficit comparison folded in from The Snowball (OMB Table 1.1, FY2006–FY2025), 30 Sep 2026
Line 6How does this end?
The loop, and the ways outFive exits, one checklist
How does this end?
The loop, and the ways outFive exits, one checklistThe five answers above form a loop. Interest widens the deficit, the deficit adds to the debt, the debt is refinanced at the new price, and a higher price means more interest. Governments have left loops like this in five ways. Below is each one, when it last happened, and the sign that would show it starting now.
How the US got from 106% to 23% of GDP, 1946–1974
Percentage points of the 83-point fall. Acalin & Ball, IMF Working Paper 24/5. A third of the fall came from paying lenders less than the market rate, and nobody announced it.
Grow out of it
Not visible1946–74: growth did 32 of the 83 points. It needs nominal growth above the average rate paid for years.
The tellDebt/GDP falling while deficits continue.
Run surpluses
Not visible1998–2001 was the last run of federal surpluses. CBO's baseline has none through 2036.
The tellA primary surplus in the Monthly Treasury Statement over a full fiscal year.
Financial repression
Watching1942–51: the Fed pegged bills at ⅜% and the long bond at 2½%, and held about three quarters of bills by late 1945.
The tellA yield ceiling, a negative real policy rate, compulsory holders, or the Fed owning a large share of bills. See the checklist below.
Inflate it away
Watching1970s: real returns on Treasuries turned negative for years. Headline PCE is 3.7% (Jul 2026).
The tell10-year yields staying below inflation. Today the 10-year sits about 1.4 points above PCE.
Change the terms
No sign1933: Congress voided the gold clause in Treasury bonds, and the Supreme Court let it stand (Perry v. United States, 1935). In 1979 a back-office failure delayed some bill payments for a few days.
The tellAny change to coupon, maturity or payment terms on existing debt, or a forced exchange.
Repression checklist: what separates managing a market from squeezing lenders
One "yes" in four. The lever being pulled is shortening the debt, not capping yields or rewriting terms. If a second box turns, the repression line above changes from Watching to Present.
The strongest opposing case: Howard Marks and Stanley Druckenmiller both describe the cost as chronic, not acute. It is "an invoice", not a crisis. Marks rates a failed auction as improbable and dollar reserve status as secure for now. Paul Singer said in October 2013 that the conditions for a loss of confidence were "here and now". Nearly thirteen years later that had not happened. The structure described here is measurable; its timing is not.
Receipts: IMF WP/24/5 (Acalin & Ball) · NY Fed Staff Report 913 (Garbade) · Perry v. United States (1935) · BEA Personal Income and Outlays, Jul 2026 · Treasury buyback results · Oaktree memo, 22 Sep 2026
Trajectory Enquiries
Seven questions about the price of the US stock market, asked the way Warren Buffett once suggested: set the value of Corporate America against the size of the economy it lives in. We ask what the ratio measures, what it reads now, what history calls that, what it could mean, and whether to believe it. Every figure carries its source and date.
Line 1What does the Buffett Indicator measure?
The rulerShares ÷ output
What does the Buffett Indicator measure?
The rulerShares ÷ outputover the economy
Take what investors say US companies are worth, all their shares at market prices, and divide by what the whole US economy produces in a year. At 100% the two are the same size. Above 100%, the market prices Corporate America at more than a year of the nation's output.
The ruler is crude on purpose. It needs no view on earnings, interest rates or which companies matter. That is its strength as a long-run gauge and its weakness as anything more precise, which Line 7 takes up.
The sum, Q2 2026
Lines 2 and 3 take the top and the bottom of this sum in turn.
“Probably the best single measure of where valuations stand at any given moment.”
Warren Buffett, Fortune, December 2001
What would change this answer: a better ruler. Nothing changes what the ratio measures; what can change is whether the relationship it assumes, between company value and national output, still holds. Line 7 sets out that case.
Receipts: Federal Reserve, Financial Accounts of the United States (Z.1), release of 11 Sep 2026 · BEA, National Income and Product Accounts · both via FRED
Line 2What is Corporate America worth?
The top of the sum
What is Corporate America worth?
The top of the sumAt the end of June 2026 the shares of US nonfinancial companies were worth $83.1 trillion. That is $13.0 trillion more than three months earlier, a rise of 18.6% in a single quarter, and $17.7 trillion more than a year earlier.
Since 1952 this value has multiplied 526 times, growing about 8.8% a year. Since 2009 it has grown about 12.4% a year.
What would change this answer: the next quarter's print. The Fed revises these figures, sometimes by hundreds of billions (Q1 2026 moved up $0.5 trillion this release), so treat the latest quarter as provisional until the next release confirms it. Q3 2026 data is due in mid-December.
Receipts: Federal Reserve Z.1, Nonfinancial Corporate Business; Corporate Equities; Liability, Level (FRED NCBEILQ027S), quarterly, end of period, not seasonally adjusted
Line 3How big is the US economy?
The bottom of the sum
How big is the US economy?
The bottom of the sumIn the second quarter of 2026 the US economy was producing goods and services at a rate of $32.6 trillion a year, in current dollars. Over the past 20 years that figure has grown 4.4% a year; over the past ten, 5.7%.
Since 1952 the economy has grown 90 times over. Company values grew 526 times. The chart shows the two racing, and where they parted company.
Company values and the economy, trillions of dollars, 1952–2026 (log scale)
Second-quarter readings each year. On a log scale, equal slopes mean equal growth rates: the lines run side by side for four decades, then shares pull away. Where the green line sits above the grey, the ratio is over 100%.
What would change this answer: revisions and a recession. BEA revises GDP for years after first publication, and a fall in nominal output would lift the ratio even with share prices unchanged. The two are measured differently: company value is a stock on one day, GDP a yearly flow.
Receipts: BEA, Gross Domestic Product, current dollars, seasonally adjusted annual rate (FRED GDP), Q2 2026 third estimate · Federal Reserve Z.1 (FRED NCBEILQ027S)
Line 4What is it now?
Today's reading
What is it now?
Today's readingCorporate America is priced at 2.55 times a year of the US economy. That is the highest of 298 quarterly readings since 1952, 26 points above the previous record of 228.8% set at the end of 2025, and more than half as high again as the 2000 dot-com peak of 162.6%.
The lowest reading was 32.2%, in the second quarter of 1982.
The Buffett Indicator, quarterly, 1952–2026
Grey below 125.6%, the line that marks the top fifth of all readings. Green at or above it. Hover or tap a quarter for its reading.
What would change this answer: a reading back below 228.8% would end the record, and one below 162.6% would put the ratio back inside anything seen before 2020. At today's GDP, that second line means company values of about $53 trillion, some 36% below June's.
Receipts: MoneySpider calculation, FRED NCBEILQ027S ÷ FRED GDP × 100, 1952 Q1 – 2026 Q2. The numerator is the Fed's series, not the Wilshire 5000, which is a licensed index.
Line 5How would we describe that?
The verdictSignificantly overvalued
How would we describe that?
The verdictSignificantly overvaluedWe sort all 298 readings since 1952 into five equal groups. The top group starts at 125.6%. Today's 255% is about twice that line, and the ratio has sat in the top group every quarter since the end of 2013.
These labels are ours. They describe where a reading sits in this series' own history; they are not Buffett's thresholds or anyone else's.
What would change this answer: one quarter below 125.6%, which at today's GDP means company values under about $41 trillion, roughly half of June's. The bands themselves move slowly as each new quarter joins the history; this quarter lifted the top line from 124.0% to 125.6%.
Receipts: MoneySpider's own quintiles of the full 1952–2026 series (cuts at 47.8 / 65.0 / 81.3 / 125.6%), recomputed each quarter
Line 6What could that MEAN?
Three averages, one sum
What could that MEAN?
Three averages, one sumIf you MEAN the average since 2009, …
GDP today: $32.6T. In ten years at this rate: $50.1T.
Company values in 2036 compared with today's $83.1T, if the ratio returns to each average
The sum on each row: that average ratio × GDP in ten years' time, compared with company values at the end of June 2026. Before dividends. The bar scale runs from −60% to +60%.
What would change this answer: faster growth or a ratio that stays high. If the next ten years' growth matched the last ten's (5.7%), the since-2009 case turns slightly positive. And no rule says the ratio must return to any average: it has stayed in the top band for twelve years.
Why it is arithmetic, not a forecast: every row assumes a destination and a speed. We do not say the ratio will go to any of these averages, or when. The point is the size of the gap between today's reading and the ratio's own history, expressed in dollars.
Receipts: averages of the MoneySpider Buffett Indicator series (Lines 4 and 5) over three windows · growth presets from BEA GDP, Q2 to Q2 · same method as the CAPE calculator (Mathematical Scenarios) on our Micro page, with a different ruler
Line 7Should we believe the ruler?
Both casesAs a gauge, yes. As a clock, no
Should we believe the ruler?
Both casesAs a gauge, yes. As a clock, noThe ratio says US shares are expensive against the economy, by more than ever before. It cannot say when, or whether, that gap closes. There is a serious case that the ruler itself has stretched, and a serious case that it has not stretched this far.
Why the ruler may overstate it
- Global earnings, domestic output. The top of the sum values US companies' worldwide profits. The bottom counts only output produced inside the US. The more US companies earn abroad, the higher the ratio can sit.
- A different kind of company. Software and other asset-light firms need less capital per dollar of sales than the manufacturers of 1960, so each dollar of output can support more market value. Inferred
- Higher profit margins. Profits have taken a larger share of the economy than they did for much of the series' history. Inferred
- It is a poor clock. Research since Campbell and Shiller (1988) finds valuation ratios say something about the next ten years and very little about the next one. This ratio entered its top band at the end of 2013 and company values have since risen more than threefold.
Why it may not stretch this far
- No precedent. 255% is 26 points above any earlier quarter and 92 points above the 2000 peak. The structural arguments on the left were all available in 2000 too.
- A race that cannot run forever. Company values have grown 12.4% a year since 2009 against the economy's 4.9%. Over long spans, profits cannot outgrow the economy that pays them without taking an ever larger share of it.
- The rates behind the rise have gone. Much of the post-2009 climb came when the 10-year Treasury yield was low; it was under 1% in 2020 and was 5.29% on 30 Sep 2026 (Treasury par yield). Confirmed
- A second ruler points the same way. Our Micro page measures the S&P 500 against its own ten-year earnings (CAPE, on Cape Fear), which is also near the top of its history.
What would change our reading: two things, stated in advance. The ruler has stretched for good if the ratio holds above 162.6%, the 2000 peak, through a US recession. It is reverting if it falls back below 125.6% for two consecutive quarters, with or without a crash in prices, since fast GDP growth could do it slowly.
Receipts: J. Y. Campbell and R. J. Shiller, “The Dividend-Price Ratio and Expectations of Future Dividends and Discount Factors”, Review of Financial Studies (1988) · US Treasury daily par yield curve (10-year) · Cape Fear, moneyspider.net/micro
Trajectory Enquiries
Six questions about what has happened to the US dollar since August 1971, when it stopped being convertible into gold. We ask the official price index, then gold, then the checkout and the pay packet, and set out why their answers differ. Every figure carries its source and date. The gold price updates live on each visit.
Line 1What does the CPI say?
The official index
What does the CPI say?
The official indexA dollar in August 1971 bought what about 12 cents buys today. Consumer prices are 8.2 times higher, a rise of about 3.9% a year for 55 years. Not every decade looked the same: prices rose 8.4% a year through the 1970s and under 2% a year in the 2010s.
What a 1971 dollar buys, in cents, 1971–2025
Annual averages of the CPI-U, not seasonally adjusted, with 1971 set to 100 cents. The fastest fall came in the 1970s. The 2021–2025 surge shows as a second, shorter drop.
What would change this answer: nothing on this line changes the past. The rate going forward is what moves. Three years in a row at the Fed's 2% target would pull the 55-year average down only slightly.
Receipts: BLS Consumer Price Index for All Urban Consumers, all items, not seasonally adjusted (FRED CPIAUCNS)
Line 2What does Troy say?
Gold, per troy ounce
What does Troy say?
Gold, per troy ounceIn August 1971 the official price of gold was $35 an ounce. Today an ounce costs about $4,050. Read literally, a 1971 dollar now buys under 1% of the gold it used to. That is gold's testimony. Line 3 asks whether to believe it.
Gold, US dollars per troy ounce, 1971–2025, plus today's spot
Annual averages from the US Geological Survey; the last point is today's spot. On a log scale, equal distances mean equal percentage moves. The dashed line is the $35 official price.
“You have to choose between trusting to the natural stability of gold and the natural stability of the honesty and intelligence of the members of the government. And, with due respect for these gentlemen, I advise you, as long as the capitalist system lasts, to vote for gold.”
— George Bernard Shaw, 1928
What would change this answer: a sustained fall in gold while the CPI keeps rising, as happened for 21 years after 1980. Gold's verdict on the dollar can reverse, and has.
Receipts: US Geological Survey, Data Series 140 (gold, 1971–2022) and Mineral Commodity Summaries 2026 (2023–2025) · live spot: gold-api.com, indicative, not an official fix
Line 3Why don't they agree?
Gold against the indexGold prices fear and rates too
Why don't they agree?
Gold against the indexGold prices fear and rates tooFrom 1971 to 2025, on annual averages, gold rose about 80-fold and consumer prices about eightfold. If gold only tracked inflation, the two lines below would move together. They don't. Between 1980 and 2001 gold fell 56% while prices more than doubled.
Gold and consumer prices, both set to 100 in 1971, log scale
Both series are annual averages, based on 1971's market gold average of $41.4, not the $35 official price. The distance between the lines is the part of gold's move that inflation does not explain.
What would change this answer: gold and the CPI moving in step for a decade. That would suggest the fear and rates premium had gone. Nothing in the record since 1971 shows that.
Receipts: USGS Data Series 140 and MCS 2026 · BLS CPI-U · World Gold Council, China official gold purchases, Aug 2026 · Federal Reserve FEDS Note, 3 Sep 2026
Line 4What did it cost at the till?
Groceries and energy6 of 8 rose less than the CPI
What did it cost at the till?
Groceries and energy6 of 8 rose less than the CPIWe took eight everyday prices that the Bureau of Labor Statistics has tracked since about 1980. Six rose less than the CPI over the same years, so after inflation they got cheaper. Ground beef rose much more than the CPI. Piped gas rose slightly more.
Change in price after inflation, from each item's start year to August 2026
Each bar compares the item's price rise with the CPI's rise over the same years. Above zero, the item beat inflation. Below zero, it got cheaper in real terms.
Then and now
| Item | Start | Then | Aug 2026 | Item × | CPI × | 12-month avg |
|---|
BLS US city average prices. "Now" is the August 2026 print, the latest available. The last column averages September 2025 to August 2026 for comparison (BLS published no October 2025 food, electricity or piped-gas prices). Monthly food and energy prices are noisy: eggs averaged $4.25 across calendar 2025 during the bird-flu shortage.
Petrol, right now: petrol has risen sharply since March 2026, from $2.96 a gallon in January to $4.20 in August (BLS). The Energy Information Administration's weekly survey showed $4.47 in the last week of September. Even at $4.47, petrol is up about 3.6 times since 1980 against 4.1 times for the CPI, so it is still cheaper after inflation, by about 12%.
What would change this answer: a different end date moves individual items. On every end date we tested (calendar 2025, the last 12 months, August 2026 alone), at least five of the eight items rose less than the CPI.
Receipts: BLS Average Price Data, US city average (FRED APU0000702111, 708111, 709112, 703112, 717311, APU000074714, 72610, 72620) · BLS CPI-U · EIA weekly retail gasoline, regular (FRED GASREGW), 28 Sep 2026
Line 5What did it cost in hours of work?
Prices against pay7 of 8 take less work
What did it cost in hours of work?
Prices against pay7 of 8 take less workMeasured in minutes at the average hourly wage, seven of the eight items take less work to buy than they did at their start year. A pound of coffee took 27 minutes in 1980 and takes 17 today. Ground beef is the exception. Gold runs the other way: an ounce cost about 11 hours of work in 1971 and about 105 hours in 2025.
Minutes of work to buy each item, start year and August 2026
Grey dot: start year. Green dot: August 2026. Electricity is shown per 100 kWh and piped gas per 10 therms so they fit the same scale.
The caveat: the wage series covers production and nonsupervisory workers, about four in five private employees. It is an average, not a median, and it excludes benefits. Real hourly pay rose less than 9% from 1971 to 2025, and about 17% from 1980, so the shorter times come partly from modestly higher real pay and partly from these prices falling after inflation.
Receipts: BLS Average Hourly Earnings of Production and Nonsupervisory Employees, total private (FRED AHETPI, from 1964) · BLS Average Price Data · USGS gold
Line 6Does the CPI tell the whole story?
Both casesIt depends whose basket
Does the CPI tell the whole story?
Both casesIt depends whose basketThe CPI is an average over a basket no single household buys. Lines 4 and 5 show that groceries and household energy mostly rose less than the index. So the index's faster rise has to come mostly from elsewhere, largely services: above all shelter, its largest single component, along with medical care and education. A household that rents, or has just bought a home, can feel a different inflation from the one published.
Why people think it understates
- Housing. From January 1983 the CPI measures homeowners' costs as an equivalent rent, not house prices or mortgage payments.
- Quality adjustment. When a product improves, part of its price rise is counted as more quality, not inflation.
- Substitution. The chained CPI (from 2002) assumes people switch to cheaper goods when prices rise.
- No one's basket. The weights are national averages; renters, drivers and older households each face different ones.
The official finding the other way
- The Boskin Commission (1996) estimated the CPI overstated inflation by about 1.1 points a year (range 0.8–1.6). Several of the method changes on the left were made in response.
- The till test. On this page, six of eight directly priced items rose less than the index. That fits an index that does not hide grocery or energy inflation.
- What we do not use: alternative "true inflation" series whose methods are not published. They cannot be checked, so they are not cited.
Coming in a later edition: housing, healthcare and education, the categories where individual experience and the index diverge most. Each needs its own like-for-like price series before it goes on this page.
Receipts: BLS, "Changing the way we measure homeowners' costs" (1983) · Advisory Commission to Study the Consumer Price Index (Boskin Commission), final report, Dec 1996 · BLS C-CPI-U documentation