Home › Guides › How to Read the Yield Curve
A picture of how much interest the U.S. government pays to borrow money for different lengths of time.
The U.S. government borrows constantly. When it needs cash, it sells a piece of paper called a Treasury bond. The buyer hands over money today, and the government promises to pay it back later - in one month, six months, a year, ten years, or even thirty years. The interest rate it pays depends on how long it's borrowing for.
A picture of how much interest the U.S. government pays to borrow money for different lengths of time.
The U.S. government borrows constantly. When it needs cash, it sells a piece of paper called a Treasury bond. The buyer hands over money today, and the government promises to pay it back later - in one month, six months, a year, ten years, or even thirty years. The interest rate it pays depends on how long it's borrowing for.
If you plot all of those interest rates on a chart - with the length of the loan along the bottom and the interest rate going up the side - you get the yield curve. On a normal day, the line slopes upward. Investors want extra reward for tying up their money longer. Lending the government money for a month feels safe; lending it for thirty years feels riskier, because anything could happen between now and 2056, so investors demand a higher rate.
You hear individual points on this curve quoted in the news constantly - "the 10-year is at 4%," "the 2-year jumped 5 basis points." The shape of the whole line matters because it tells you what bond investors collectively expect. A steep upward curve usually means people think the economy will keep growing. A flat or downward-sloping curve usually means they're nervous.
The yield curve also touches your life directly. Your mortgage rate moves with the 10-year point. Your savings account follows the very short end. Car loans, business loans, even what your retirement portfolio is worth - almost everything traces back, in some way, to where this curve sits and which way it's been moving lately.
Yields plotted across every Treasury maturity (1M to 30Y); the shape encodes growth and rate expectations.
Each point on the curve is a yield published by the U.S. Treasury - the rate of return an investor would earn buying a bond of that maturity at today's price and holding it to maturity. Treasury publishes these as constant-maturity yields for thirteen standard tenors: 1-, 2-, 3-, 4-, and 6-month, plus 1-, 2-, 3-, 5-, 7-, 10-, 20-, and 30-year (this site charts eleven of them; the 2-month was added in 2018 and the 4-month in 2022). They're recomputed every trading day from real secondary-market quotes on actual Treasury bonds, then smoothed so even tenors that didn't trade an exact-maturity bond that day still get a quote.
The shape comes from two sources. First, expectations about where the Fed will set short-term interest rates over the life of the bond - a 10-year yield is roughly the average expected short-term rate over the next ten years plus a premium. Second, the term premium - the extra compensation investors demand for the uncertainty of locking in money for longer periods. When either piece moves, the curve moves.
An upward-sloping (or "normal") curve is the historical default. It implies the market expects rates and inflation to drift higher over time, or at least not collapse. A flat curve - short and long yields close together - signals uncertainty about where the economy is going. An inverted curve - short yields above long yields - is unusual and important. It typically means the market expects the Fed to cut rates aggressively, and the Fed usually only cuts aggressively when growth is slowing or a recession is approaching.
Who's actually trading these bonds? Primary dealers (about two dozen large banks the New York Fed designates as official Treasury market-makers), foreign central banks, U.S. and foreign pension funds, mutual funds, insurance companies, hedge funds, and a growing share of retail buyers through brokerages and money-market funds. Together they trade well over $700 billion of Treasury bonds on a typical day - it is the deepest, most liquid bond market in the world.
The curve interacts with almost every other rate in the financial system. The 30-year fixed mortgage rate runs roughly 150–200 basis points above the 10-year Treasury yield (the spread is mostly compensation for prepayment risk and the credit risk of the homeowner). Corporate bond yields are quoted as a spread to the same-maturity Treasury. Swap rates, futures, options, and even foreign-currency carry trades all tie back to where the curve sits today.
A concrete example: on a typical day in mid-2024 you might have seen 3-month at 5.30%, 2-year at 4.70%, 10-year at 4.30%, 30-year at 4.45%. The 10-year minus 2-year would be roughly negative 40 basis points there - an inversion. The 30-year sitting just above the 10-year is called positive curvature at the long end, and it is the more common long-end shape historically.
CMT yields (DGS1MO..DGS30) with shape driven by expected short-rate path + term premium.
The long-end yield decomposes into two parts. The expectations component is the average expected short rate over the bond's life - essentially what an investor would earn rolling overnight Treasuries from now until maturity. The term premium is everything else: compensation for duration risk (long bonds lose more when rates move), inflation uncertainty, and supply-demand imbalances. Both components are unobservable directly; they have to be estimated with affine term-structure models. The two standard published decompositions are Adrian-Crump-Moench (ACM, NY Fed, updated daily) and Kim-Wright (Federal Reserve Board, updated monthly).
Why this matters: a flattening curve isn't a single story. It could be falling rate expectations (the Fed is expected to cut), falling term premium (investors want less compensation for duration), or both. Interpretation requires the decomposition. During QE episodes, the term premium gets compressed - sometimes negative - by Fed balance-sheet purchases that suck duration out of the private market. Reading a flat curve in 2014–2016 as a recession signal was a common mistake; the flatness was largely term-premium compression.
The classic recession signal is yield-curve inversion. The cleanest empirical track record belongs to the 10Y-3M spread (Estrella & Mishkin, 1996; subsequent Fed staff papers). The 10Y-2Y is what bond traders watch in real time because the 2-year prices in roughly two years of Fed expectations, which traders find more actionable. The 10Y-3M has inverted before every U.S. recession since the late 1960s; the 10Y-2Y (data begin in 1976) has inverted before every recession since, with lead times of 6–24 months. The 2019 inversion preceded COVID; the 2022–24 inversion preceded an extended period without an NBER-confirmed recession - the longest such gap on record.
Treasury Constant Maturity methodology: the U.S. Treasury computes these yields from close-of-day bid-side quotes on the on-the-run securities at each benchmark tenor - composites of dealer quotes collected by the Federal Reserve Bank of New York - fitted with Treasury's monotone convex method (which replaced the earlier quasi-cubic Hermite spline in December 2021). The fitted par curve delivers a single daily quote at each standard tenor even when no exact-maturity bond traded. On-the-run (OTR) bonds - the most recently issued at each major tenor - trade with a small specialness premium because they are heavily used in repo markets and as benchmarks. Off-the-run bonds trade at slightly higher yields. Because its inputs are on-the-run quotes, the CMT yield tracks the slightly richer OTR level, which matters when comparing CMT spreads to spreads built from off-the-run or seasoned issues.
Practitioner trading: rates traders express views on the curve as butterflies (long the wings, short the belly, or vice versa), curve steepeners and flatteners (long one maturity, short another in beta-weighted notional), or outright duration trades. DV01 - dollar value of a 1bp move - scales roughly with modified duration times price. A 10-year note has about 9 years of duration; a 30-year bond has about 20. DV01-weighting is essential when constructing curve trades because a parallel shift will produce different P&L on different maturities.
Hedging context: corporates issuing 10-year debt typically hedge their funding-rate exposure with 10Y Treasury futures or swap-vs-Treasury basis trades. Mortgage originators hedge pipeline duration with TBA (To-Be-Announced) MBS contracts and 5Y/10Y Treasury futures. Pensions running LDI (Liability-Driven Investment) strategies match long-dated liability discount rates with 30Y Treasury, long-dated TIPS, and pay-fixed swaps. These three flows - corporate, mortgage, pension - explain most of the marginal demand at different points of the curve.
Regulatory context: the Supplementary Leverage Ratio (SLR) makes Treasury holdings expensive in dealer capital, which has been a structural drag on dealer market-making capacity since 2014. Treasury market liquidity stress events - March 2020 dash-for-cash, October 2014 flash rally, August 2023 long-end backup - often trace partially to SLR-binding dealers unwilling to absorb supply. The Fed's Standing Repo Facility (SRF), launched in 2021, provides a ceiling backstop at the top of the target range.
Common practitioner mistakes: (1) reading curve shape in isolation without checking whether term premium has moved; (2) treating an inversion as a recession timing signal rather than a probability shift; (3) ignoring the on-the-run vs off-the-run distinction when measuring tight spreads; (4) using CMT 30Y minus CMT 2Y as a 30-year-minus-2-year spread without realizing it folds together two different points on the term-premium curve; (5) treating each daily move as informative when typical day-to-day noise is several basis points.
Monotone convex par curve (replaced the quasi-cubic Hermite spline in Dec 2021); FRED DGS series; decomposition methodology matters.
Historical episodes worth knowing: the 1979–82 Volcker era saw the curve invert deeply (sometimes by more than 300 bps) as the Fed crushed double-digit inflation; the 2000 inversion preceded the dot-com recession; 2006–07 inversion preceded the Global Financial Crisis; the 2019 inversion preceded COVID; the 2022–24 inversion was the longest of the modern era and the deepest since the early 1980s. Each cycle has its own term-premium backdrop, so cross-cycle inversion-depth comparisons without that context are misleading.
The CMT methodology is documented in the U.S. Treasury's "Treasury Yield Curve Methodology" technical paper, originally codified in the 1980s and refined repeatedly since. Since December 6, 2021 the daily par yield curve has been built with the monotone convex method (in the Hagan-West family): instantaneous forward rates are bootstrapped at the input maturities, then joined with monotone convex interpolation, which keeps the forward curve well-behaved and reprices the inputs exactly. Bills enter via their bond-equivalent yield. The earlier quasi-cubic Hermite spline, which fit the inputs with continuous first and second derivatives, was retired at that date; Treasury's parallel testing showed the switch moved published rates by well under a basis point. (The bid-yield inputs with a 30-day maturity filter sometimes described in this context belong to the Federal Reserve Board's separate Gurkaynak-Sack-Wright research curve, which fits a Svensson model to off-the-run issues.) The real (TIPS) par yield curve is constructed analogously and published at 5-, 7-, 10-, 20-, and 30-year tenors; TIPS themselves are auctioned at 5, 10, and 30 years (the 30-year returned in 2010, and the 20-year TIPS was discontinued after 2009).
Term-premium estimation is a Kalman-filter exercise on an affine term-structure model. ACM uses five pricing factors and is updated daily by the NY Fed at https://www.newyorkfed.org/research/data_indicators/term-premia-tabs. Kim-Wright (2005) uses three factors and is updated monthly. The two diverge most during balance-sheet-expansion episodes: the ACM 10Y term premium fell steadily through the QE era, turned negative for the first time around 2014–2015, and stayed negative or near zero for much of 2015–2020, consistent with Fed purchases compressing term premium below its historical neutral level. Kim-Wright shows the same broad decline at somewhat different levels. Neither is right - they are different model fits with different identification assumptions.
Auction supply dynamics matter at the long end. The quarterly Treasury Refunding Announcement (QRA) sets coupon issuance sizes for the next quarter, released at 8:30 AM ET on the Wednesday of refunding week, typically early Feb, May, Aug, and Nov (sometimes the last Wednesday of the prior month). The August 2023 QRA upsized 10s and 30s by more than the market expected, and the 10Y yield backed up to about 5.0% over the following weeks. The TBAC (Treasury Borrowing Advisory Committee) deliberations preceding each QRA are public and worth reading for institutional context.
On-the-run versus off-the-run specialness: OTR bonds trade with a repo specialness - they earn a premium in the repo market because dealers and hedge funds need them to deliver against short positions. Specialness compresses OTR yields by 1–10 bps versus comparable off-OTR bonds. When a new bond becomes the on-the-run at quarterly refunding, the previously-OTR bond loses its specialness and re-rates wider. Curve traders monitor this rotation closely.
QT (quantitative tightening) and SOMA portfolio dynamics: the Fed's System Open Market Account holds roughly $4–6T of Treasuries depending on the cycle (the Treasury-only peak was about $5.8T in mid-2022). SOMA reinvestment caps - the Fed's published monthly runoff schedule - interact with bills-vs-coupon issuance to alter how much duration supply hits the private market. The Fed's decisions to slow QT in mid-2024 and then end runoff effective December 2025 meaningfully changed the private-market duration-absorption requirement at the long end.
Cross-asset implications: the curve shape feeds into the discount rate for equity valuation - long-end yields anchor terminal-value discounting for long-duration growth stocks, which is partly why 2022's long-end back-up compressed Nasdaq multiples so heavily. FX carry trades depend on cross-currency rate differentials at specific maturities. Mortgage convexity hedging - MBS investors selling duration as rates fall and buying it as rates rise - creates non-linear feedback into the 5Y–10Y belly that can exaggerate curve moves during fast rate selloffs.
Recent literature worth knowing: Engstrom & Sharpe (2018, FEDS Notes) argue the near-term forward spread (18m-3m forward) is a better recession predictor than the slope spread; the NTFS turned negative in late 2022 alongside 10Y-3M and 10Y-2Y. Bauer & Mertens (2018, FRBSF Economic Letter) re-examine recession-predictive power across various spreads. Cieslak (2018) and Cieslak-Pflueger (2023) have papers on term-premium dynamics in the post-2008 regime.
Research caveats: the NBER Business Cycle Dating Committee announces recessions with an average 12-month lag, so "inversion was wrong" claims made during the inversion window are premature. The 10Y-3M signal has Type II error risk in zero-bound environments because the 3M cannot meaningfully turn negative. Inversion is a probability shift, not a deterministic indicator - base-rate probability of recession within 24 months following a deep inversion has historically run roughly 70–80%.
Operational note for this site: the CMT yields published here come from FRED's H.15 series (DGS1MO through DGS30), pulled daily. The 1-month maturity began publication 2001-07-31; the 30-year was suspended from 2002-02-15 to 2006-02-09 (Treasury wasn't issuing 30-year bonds during that gap); the 20-year was suspended from January 1987 through September 1993. Chart axes that span those windows will show gaps in some series - that is real data, not a bug.