What Rising Bond Yields Mean for the US Budget, the Economy and the Dollar

Research note. Computed from seven countries’ daily 10-year yields, the US Treasury curve and TIPS breakevens via FRED, the Treasury’s Fiscal Data (average interest rates, the Monthly Statement of the Public Debt and its CUSIP-level detail, the Monthly Treasury Statement), OMB’s fiscal-year tables, the Federal Reserve’s broad dollar index and our own price database, as of September 14, 2026. The prose is frozen at that date; the exhibits marked live update on their own schedules. The methods are in How we checked it below; our editorial and AI standards are in How we use AI.

• ·Since March 2 the 10-year is up +91 bp in the US and between +57 bp and +95 bp in the other six markets. The breakeven moved +11 bp all year; the real yield went from 1.93% to 2.60%.
• ·The 2-year moved from 17 bp below the daily effective fed funds rate on December 31 to 100 bp above it on September 11, with the rate itself at 3.63%: the market stopped pricing cuts and started pricing hikes.
• ·The nominal securities maturing in fiscal 2027 to 2029 carry coupons of 3.50%, 3.07% and 3.35%. Replaced at September 11 yields they add about $155 billion a year to interest on those securities once all three have rolled; TIPS ($725 billion of the maturities) are left out because their coupon is real.
• ·Holding debt at 98% of GDP needs the primary deficit to fall from 2.62% of GDP to 1.72% at today’s average rate, about $275 billion a year; a point higher on the rate doubles that.
• ·Since March 2 the S&P 500 is +10.9%, the equal-weight index +5.1%, utilities -11.7%; the 30-year mortgage rate, a nominal borrowing rate, went from 5.98% to 6.76% and housing starts are down 13.5% on the year.
• ·Weighted by the Fed’s 2026 index weights, the US yield advantage over the foreign yields we hold widened from 0.96 to 1.11 points; the broad dollar index still slipped, -1.3% for the year.

On September 11 the 10-year Treasury closed at 4.96%, up +91 bp from March 2, the last close before the diesel crack crossed $60 a barrel and the energy episode that produced a record retail diesel price began (that story is on the crack spread and diesel price pages). This piece asks what the yield does next to three things that depend on it: a government paying interest on $30 trillion, an economy priced off the risk-free rate, and a currency whose value is partly the rate it pays. I would use it the way it is built: as a schedule of what arrives when, so that the next data release can be judged against a number I committed to in advance.

Seven bond markets, one move

Since March 2 every major nominal 10-year has risen by a similar amount: the UK, Japan and Germany within nine basis points of the US, the euro area and Canada a little less, Australia the least. The split between real yields and expected inflation is measured here for the US only; the foreign exhibit is nominal. Six central banks (the ECB sets policy for Germany and the euro area alike) and seven fiscal positions produced one move, which argues for a shared factor before a local one. What that factor is comes next, as a hypothesis with evidence.

The composition of the US move is the first piece of evidence. Inflation expectations barely moved: the 10-year breakeven went from 2.25% at year-end to 2.36% on September 11, so the real yield rose from 1.93% to 2.60%. The second is the front end. The 2-year sat 17 basis points below the daily effective fed funds rate on December 31 (3.64% that day), a market pricing cuts; on September 11 it sat 100 basis points above it (3.63%), with CPI at 3.35%. The third is the curve: the 10-year minus the 2-year narrowed from 71 to 33 basis points, so the long end rose less than the short end. Taken together, the pattern is consistent with a repricing of the policy path. It does not exclude a term-premium rise: policy expectations and term premiums can rise together while the curve flattens, and the 10-year minus 2-year spread does not isolate the term premium.

Two cautions keep that a hypothesis. A breakeven is not a clean expectation; it carries an inflation-risk premium and a liquidity premium for TIPS, so “real yield” here is a proxy (the New York Fed and the Board’s own research on term-structure decomposition are linked in the sources). And the stronger-growth alternative has support: our Economy Health score rose from 57 at year-end to 71 on September 14, so some of the rise is a better economy demanding a higher real rate. The two stories part ways at the curve and at breakevens. Growth-led yields tend to come with a steeper curve and rising breakevens; this year brought a flatter curve and flat breakevens. That is why I weight the policy story, and why the first two markers at the end are about the front end and breakevens.

Can the government sustain its budget at these rates?

The Treasury borrows at the coupons it locked in when each security was sold. On August 31 those averaged 3.475% across $31.83 trillion of marketable debt, while the 10-year closed that day at 4.75% and on September 11 at 4.96%. The gap between the average coupon and the market is what refinancing closes, and the Treasury’s own CUSIP detail says when.

The whole-stock number comes first because it is the one people quote: $31.83 trillion times a 1.49-point gap is about $473 billion a year, and each trillion refinanced at the 10-year would be about $15 billion. Both are illustrations of the existing stock repricing at one yield, and they overstate the precision of anything that will happen: the actual cost depends on each maturing coupon, the tenor it is replaced with, and where yields are on the day. The ladder answers those questions for the next few years.

Read the shaded rows. Fiscal 2027 is the big one: $8.29T matures, $4.76T of it bills that already carry 3.82% on a bond-equivalent basis once every reopening issued by August 31 is weighted in, $223 billion of TIPS that are left out of the estimate, and the nominal securities together at an average 3.50%. Replaced at September 11 yields, that year adds about $64 billion a year; fiscal 2028 adds $56 billion and fiscal 2029 $36 billion, about $155 billion a year in all once the three have rolled, with $725 billion of TIPS maturities excluded. That is smaller than the whole-stock illustration and larger than nothing, and it is specific: the securities exist, their coupons are printed, and their maturity dates are fixed. The number moves with yields on the day each one rolls, which is the reason the tripwires below are about the Treasury’s realised average rate and not the market’s quote.

The bridge to the budget matters. That $155 billion is interest on marketable securities. Budget net interest ($970 billion in fiscal 2025, 18.5% of receipts) also includes interest on non-marketable debt such as savings bonds and state and local series, nets out the interest the Treasury pays its own trust funds, and is reduced by other interest income; it excludes the Federal Reserve’s remittances, which are counted as receipts. So the ladder’s figure is an input to net interest, and I do not treat the two as equivalent.

Whether that is sustainable is a question the projection cannot answer, because continued market access is something it assumes. What arithmetic can answer is the adjustment needed to stop the debt ratio rising. Fiscal 2025 ran a primary deficit, spending before interest minus receipts, of 2.62% of GDP, about $805 billion. With the average rate at 3.49% and nominal growth at its ten-year pace, debt held by the public stays at 98% of GDP only if that primary deficit shrinks to 1.72% of GDP: an adjustment of 0.90 points of GDP, about $275 billion a year, in some mix of spending and revenue, before any recession. At a point higher on the rate the adjustment is 1.83 points, about $563 billion. If the rate ever meets the growth rate, the whole primary deficit has to go.

The stress case is an accelerated one and is labelled that way: it compresses the two fiscal years of deterioration from 2007 to 2009 (receipts -18%, spending before interest +33.7%) into fiscal 2026, so its first year is worse than any single year in that episode. On today’s base it produces a deficit of 15.7% of GDP with interest at 24.6% of receipts, and by fiscal 2029, with the rate held, interest at 31.5% of receipts and debt at 137% of GDP. The fourth case is an explicitly generous sensitivity for the objection that no recession comes without rate relief: it cuts the average rate by the change in the fiscal-year average of the Treasury’s own marketable rate from fiscal 2007 to 2010, -0.66, -1.37, -0.39 points, to 1.07%, each year’s rate applied to that whole year’s interest. Interest then falls to 9.2% of receipts by fiscal 2029, and debt still reaches 129% of GDP, because the shock is to the primary balance and the rate can only soften the interest on top of it. Whether the relief would be available is the open question: the 2009 relief came with a starting debt of 35% of GDP and a Fed that could cut to zero, and this year’s bond market has been moving the other way.

What does it mean for the economy?

When yields rise because expected inflation rises, nominal earnings tend to rise with them. When they rise in real terms with expectations flat, the discount rate rises alone. That is this year’s mix, and it compresses valuation multiples, raises the hurdle for capital spending, and reprices anything owned for its yield. The evidence the site holds says the repricing has started in the rate-sensitive corners and has not reached activity or credit.

The table sorts into three groups. Housing, the most rate-sensitive activity, is where the move shows: starts are down 13.5% on the year and new home sales 6.3%, with the mortgage rate, a nominal borrowing rate, up from 5.98% to 6.76%. Credit has not tightened. Banks report 0% net tightening on business loans, the Baa spread is 1.64 points against the 1.68 known on March 2, and consumer credit is growing 2.6% a year. Activity is at trend. The Chicago Fed’s index sits at -0.05 on a three-month basis, and core capital goods orders are up 12.6% on the year in nominal terms. So the narrow conclusion the indicators support is that the economy has not slowed, and that the cost of money has risen for the parts of it that borrow long: housing first, then the utilities and real estate stocks that carry the most debt, -11.7% and -1.8% since March 2 against +10.9% for the index. Small companies are +9.1% over the same stretch. The “average company is being marked down” reading does not survive the data; the yield-owning sectors are the ones being marked down.

What does it mean for the dollar?

Higher US yields lift the dollar when they rise relative to the yields the dollar trades against. The Fed’s broad index weights 26 currencies; the five foreign 10-years the site carries (the euro area, Canada, Japan, the UK and Australia) are 45.6% of it, and Mexico, China and Korea, most of the rest, have no yield series here. On that partial, weighted measure the US yield advantage was 0.96 points on December 31, 0.99 on March 2 and 1.11 on September 14: a widening of 15 basis points, most of it from Japan and Canada rising less than the US. The broad dollar index is -1.3% for the year and -0.4% since March 2. So the rate gap widened a little and the dollar slipped a little, which is consistent with a small yield advantage being outweighed by something else, and I read the something else as the fiscal picture above, held loosely because half the index is unmeasured here.

The history test asks what the dollar did after prior stretches like this one: a real-yield rise of 75 basis points or more over six months with breakevens roughly flat. Since the breakeven record begins in 2003 there are three. The current move measures +68 bp over the last 126 sessions, just under the rule, so it is a comparison here; at a 65-point threshold the rule yields five episodes, the current incomplete one among them.

Three episodes are three stories, so they are read as stories. The dollar rose in all three over six months, and the one large gain, April 2022, came in a window when the US 10-year rose +1.27 points while Germany’s rose +1.38, the UK’s +1.88 and Japan’s +0.01. US yields did not rise alone that year; over calendar 2022 the German and UK 10-years rose more than the American one (+276 bp and +275 bp against +236 bp). The dollar’s 2022 gain lines up with the one market that did not move, Japan, and with the Fed hiking faster than the ECB and the Bank of England, which is a policy-path story more than a level story. In 2013 and 2023 Germany’s 10-year fell while the others were mixed (the UK’s rose 4 bp in the 2013 window, Japan’s 3 bp in 2023) and the US 10-year was about flat, and the dollar gained about what it gains in any six months. The equity column says the index did three different things; the study does not test sectors, so nothing here supports a claim about which parts lagged. The last column shows the US 10-year lower a year later in two of three; with three cases that is an observation and no rule. At the looser 65-point threshold the added completed episodes, June 2007 and February 2023, pull in a falling dollar and a mixed index, which is the reason the caption carries them.

My read on the dollar, held loosely: flat to weaker while the move stays global and the measured yield advantage stays near a point, with the fiscal arithmetic above the more likely driver than the rate gap. It becomes a stronger-dollar story if the foreign 10-years stop rising and the US does not, which is the third marker below.

What would change my mind

Five markers, each with a persistence rule so that a one-day print does not reverse the thesis. First, breakevens: a 10-year breakeven above 2.75% for ten consecutive sessions on Inflation Expectations would mean the move had become an inflation scare, and the economy section would need rewriting around nominal growth. Second, the front end: the 2-year falling back below the fed funds rate for ten sessions on Treasury Yields while the 10-year holds above 4.75% would mean the policy story had ended and the remaining rise was term premium and supply, a fiscal story. Third, the gap: the H.10-weighted US yield advantage above 1.50 points for ten sessions, shown with its session count on Global 10-Year Yields, would switch the textbook dollar case on. Fourth, the bill arriving: the Treasury’s average rate on marketable debt above 3.75% for two consecutive monthly releases on US National Debt, which tests the realised rate and not the market’s quote, since a narrowing gap could also come from falling yields. Fifth, the recession case starting: the Economy Health score below 50 for ten sessions with the 2-year still above 4.50%, which is the stress case without the rate relief the last two recessions had.

How we checked it

Every number here comes from official series we already publish, or from the Treasury’s own statements. In plain terms:

• ·Frozen and live. The prose quotes values pinned to the September 14, 2026 analysis snapshot (the Treasury curve, breakevens and the daily effective fed funds rate through the September 11 close; the Monthly Statement of the Public Debt of August 31). A script in the article’s folder reads the frozen block itself, re-derives every value from the committed data at those dates, and fails if either side changes. The reading box, the tables and the charts are labelled live and update on their own schedules.
• ·The seven yields are each country’s official daily 10-year benchmark (the euro area is the ECB’s AAA composite). “Since March 2” is the change from that day’s close, the last before the diesel crack crossed $60; the year-end column uses December 31.
• ·The real yield and the proxies. The real yield is the 10-year Treasury yield minus the 10-year breakeven (nominal minus inflation-protected), daily since January 2003; breakevens also embed an inflation-risk premium and a TIPS liquidity premium, so this is a proxy. The policy-expectations proxy is the 2-year yield minus the same day’s effective fed funds rate (the New York Fed’s daily EFFR, 3.64% on December 31, 2025 and 3.63% on September 11, 2026); the term-premium proxy is the 10-year minus the 2-year, which does not isolate the term premium. Both are crude next to the New York Fed’s ACM term-premium model, which is linked in the sources and which the site does not carry.
• ·The maturity ladder reads every marketable CUSIP on the Monthly Statement of the Public Debt (August 31, 2026), counts each once, and sums the outstanding amount by the fiscal year its maturity date falls in. The coupon of what matures is weighted by amount over the nominal securities: bills at the accepted-amount-weighted bond-equivalent investment rate across every auction of the CUSIP issued on or before the statement date, original and reopenings, from Fiscal Data’s auction records (the statement itself shows only the first auction’s discount rate, and 12 reopenings settled after August 31 are left out); notes and bonds at their fixed coupon. TIPS are counted in the amounts and excluded from the rate and from the increase, since their coupon is a real rate on an inflation-indexed principal. Floating-rate notes are in the amounts only. Each nominal security is placed in a replacement tenor by its original term (up to a year: 3-month; up to five and a half years: 2-year; up to twenty and a half: 10-year; longer: 30-year) and re-priced at the September 11 yield of that tenor; the “illustrative increase” is amount × (replacement yield − coupon), the change in annual interest on those securities. Both sides of the bill comparison are bond-equivalent yields, the basis Treasury itself recommends for comparing instruments. The whole-stock figures ($473 billion; $15 billion per trillion) apply the average-coupon-to-10-year gap to the entire marketable stock and are illustrations of the same arithmetic without the schedule. Budget net interest additionally covers non-marketable debt, nets intragovernmental interest and other interest income, and excludes Federal Reserve remittances, so the ladder’s figure is an input to it and never equal to it.
• ·The budget cases start from the federal-budget page’s arithmetic: receipts, non-interest outlays and nominal GDP grown at their fiscal 2015 to 2025 compound rates (4.89%, 5.70%, 5.33% a year), interest as the stated average rate times the prior year’s debt held by the public, the deficit added to that debt. The accelerated stress case applies the fiscal 2007 to 2009 changes in receipts (-18%) and in non-interest outlays (+33.7%; total outlays rose +28.9%, and matching the category is what moves the first-year deficit from 14.8% to 15.7% of GDP) to fiscal 2026, holds non-interest outlays flat for three years while receipts grow at the base rate (fiscal 2009 to 2012: outlays +0.3%, receipts +16.4%), then resumes. Nominal GDP is held flat in the shock year, as it was from 2007 to 2009. The rate-relief case is the same with the average rate cut by the change in the fiscal-year average of the Treasury’s average marketable rate from fiscal 2007 to 2008, 2008 to 2009 and 2009 to 2010 (-0.66, -1.37, -0.39 points), each year’s rate applied to the whole of that year’s interest, then held; it is an explicitly generous sensitivity. No policy response is assumed in any case; the Congressional Budget Office’s outlook is the reference forecast.
• ·The stabilizing primary balance is (r − g) ÷ (1 + g) × debt/GDP, with r the Treasury’s average rate on interest-bearing debt (3.49%, and a point higher) and g the ten-year nominal growth rate (5.33%). A negative result is the primary deficit that keeps the ratio flat; the gap is that number against fiscal 2025’s actual primary deficit, in points of GDP and in fiscal 2025 dollars.
• ·The episode study marks the first day on which the real yield sat at least 0.75 points above its level 126 sessions earlier with the breakeven within 0.25 points, one episode per 252 sessions, and grades the broad dollar index, SPY and the US 10-year over the next 126 and 252 sessions where complete; the same rule at 0.65 points is reported as a sensitivity. The foreign 10-year columns are each country’s change between the same two calendar dates as the US window (the last observation on or before the start date to the last on or before the US window’s end date). The baseline takes every 21st session of the record. The study tests the index and the dollar only.
• ·The dollar and the weighted gap. The dollar is the Federal Reserve’s Nominal Broad US Dollar Index (26 currencies, January 2006 = 100), distinct from the futures-market DXY. The weighted gap subtracts from the US 10-year the average of the euro-area, Canadian, Japanese, UK and Australian 10-years weighted by their 2026 shares of the broad index (21.0%, 12.8%, 5.2%, 5.2% and 1.4%); those five cover 45.6% of the index, and the site holds no yield series for Mexico, China or Korea and no bilateral exchange rates, so the measure is partial and the text says so. The weights are fixed at their 2026 values wherever they are applied to history; the chart is today’s weights over past yields, never the historical currency composition. The same weights module feeds the gap shown on the Global 10-Year Yields page.
• ·The economy exhibit lists, for each indicator, the last observation whose release date fell on or before March 2, 2026 (with the month it refers to) and the latest, with the release date, from the site’s own indicator pages: housing starts, building permits, new and existing home sales, core capital goods orders (nominal), the senior loan officer survey’s net share of banks tightening business-loan standards, the Baa corporate spread, the Chicago Fed activity index (three-month average; zero is trend growth), consumer credit, and the 30-year mortgage rate, which is a nominal borrowing rate. The equity rows are close-to-close ETF returns: SPY, RSP (equal weight), IWM (small companies), XLU, XLRE and IYT.