T10Y3M Meaning: Understanding the 10y / 3m Treasury Spread

The T10Y3M is the daily difference between the constant-maturity yield on the 10-year Treasury and the 3-month Treasury bill yield: a curve variable published by FRED continuously since 4 January 1982, and which the Federal Reserve Bank of New York turns, under a slightly different convention, into the single input of its probit recession-probability model.

Understanding what the T10Y3M means requires three elementary questions: what does each leg measure, how is the spread read day-to-day, and how does it differ from the cousin spreads circulating in financial press.

What each leg measures

The T10Y3M is a yield differential between two U.S. Treasury instruments of sharply contrasted maturities. The long leg is the 10-year Treasury constant-maturity yield, series DGS10 at the Federal Reserve Bank of St. Louis. The short leg of the FRED ticker is the 3-month constant-maturity yield, series DGS3MO; the series title spells it out, 10-Year Treasury Constant Maturity Minus 3-Month Treasury Constant Maturity. The difference DGS10 minus DGS3MO is published directly under the T10Y3M ticker, and the identity holds to the basis point across all 11,181 observations in the series.

The 10-year yield is not the coupon of a specific bond but a daily interpolation computed by the U.S. Treasury Department from active securities on the secondary market. This constant-maturity construction provides a continuous time series across all business dates, independent of issuance cycles. The 3-month yield follows the same principle, drawn from 13-week T-bills currently trading in the secondary market.

The gap between the two captures the average slope of the yield curve over the 3-month to 10-year interval. When the spread is positive (the 10-year yields more than the 3-month), the curve is said to be “positively sloped” or “normal.” When it is negative (the 3-month yields more than the 10-year), the curve is “inverted” over that interval. The absolute value of the spread reflects slope magnitude.

Reading the current value day-to-day

The T10Y3M is published daily at U.S. market close, generally with a two- to three-hour delay after 4:00 p.m. New York time. It is expressed as a percentage to two decimals, which in practice reads as basis points: a T10Y3M of +0.75 corresponds to a 75-basis-point gap between the 10-year and the 3-month in favor of the 10-year. A T10Y3M of -1.50 corresponds to a 150-basis-point inversion.

Across the published history, which begins in 1982, the average T10Y3M value is positive at +153 basis points, median +157. The extremes of the series itself are a peak of +518 basis points on 26 August 1982 and a trough of -189 basis points on 4 May 2023. Going further back means rebuilding the spread from the 10-year and the 3-month bill: that reconstruction puts the Volcker trough at -373 basis points on 10 December 1980, not in March. The eight sustained inversion episodes since 1968, whose full sequence the NY Fed T10Y3M recession signal documents, produced very uneven troughs, from -13 basis points in August 2025 to -189 basis points in May 2023, with a median across cycles of -52 basis points.

Crossing the zero boundary is the operational threshold tracked by the financial community as an inversion indicator. The crossing into negative territory is instantly publicized, but it is the duration and depth of the negative-territory sequence that carry the predictive information, not the crossing event itself. A short, shallow inversion may be a market event without macroeconomic weight: the three dips below zero in 2025, the deepest at -19 basis points, each lasted only thirteen to eighteen sessions. The 1998 false positive often cited here belongs to the T10Y2Y; the T10Y3M never went below zero that year, its 1998 low being +4 basis points on 22 September.

The DTB3 and DGS3MO conventions: why two series for the “3-month”

A technical nuance deserves noting for anyone downloading the raw series from FRED. Two series measure the “3-month Treasury yield,” with slightly different calculation conventions.

The DTB3 series (“3-Month Treasury Bill: Secondary Market Rate”) is the rate on a residual 13-week T-bill in the secondary market, quoted on a discount basis: the return is measured against the redemption value rather than the price paid, which puts it mechanically below a conventional yield. This is the historical Treasury Department convention since 1954, and it is the one underlying the NY Fed model, after conversion to a bond-equivalent basis.

The DGS3MO series (“3-Month Treasury Constant Maturity Rate”) is the constant-maturity yield, interpolated to exactly three months from active issuance-curve points. This is the parallel convention used by the Treasury since 1981 for international comparisons and for term-structure analyses.

DGS3MO sits systematically above DTB3, and the gap is a matter of convention rather than volatility: 12 basis points on average since 1981, 10 at the median, 17 since 2023, and as much as 142 at the peak of 1981. The division of labour runs the opposite way to the common account: FRED computes T10Y3M as DGS10 minus DGS3MO, while the Federal Reserve Bank of New York starts from the DTB3 discount rate, converted to a bond-equivalent basis, to feed its probit model. This convention divergence does not alter signal reading on macroeconomically relevant horizons (monthly, quarterly), but it explains small numerical differences between distinct academic publications.

For day-to-day macro observation, the FRED T10Y3M series suffices. For exact replication of the NY Fed probit model, the shortest route is its own monthly file, which publishes both legs, the spread and the probability: rebuilding DGS10 minus DGS3MO would only reproduce the T10Y3M that FRED already serves. The gap shows on the latest point, 0.87 at the NY Fed in August 2026 against 0.81 on T10Y3M.

T10Y3M, T10Y2Y, and other curve spreads: elementary distinctions

The T10Y3M coexists with half a dozen candidate curve spreads measuring different slopes or different combinations. Four deserve distinction to avoid conflating them in daily reading.

The T10Y2Y spread measures slope on the 2-year to 10-year interval (DGS10 minus DGS2). It is more frequently cited in general financial press because it inverts earlier than T10Y3M and is tracked tactically by the bond market. The T10Y3M vs T10Y2Y comparison details the eight historical inversions and explains why T10Y2Y’s earliness comes with lower predictive robustness.

The Fed Funds / 10-year spread measures the gap between the policy rate administered by the Federal Open Market Committee and the 10-year yield. Its reading is close to T10Y3M in magnitude (the Fed Funds rate and the 3-month being strongly correlated), but it is administered rather than market-derived, which introduces discontinuities at FOMC decisions. It is rarely used in academic research for this reason. Related reading: Comparing 2-year and 10-year across regimes.

Full-slope indicators (slope indicators) aggregate information from several curve points by principal-component decomposition. They are used in affine term-structure models such as Adrian-Crump-Moench, but they do not serve in daily macro reading: their interpretation requires technical understanding of the underlying models.

The near-term forward spread, proposed by Engstrom and Sharpe at the Federal Reserve Board in 2018, measures the gap between the eighteen-month forward implied yield and the current 3-month yield. It has predictive power comparable to T10Y3M over the 1973-2017 window, but it never left research status: the Board publishes no series for it, and the NY Fed page still serves only the 10-year minus 3-month term spread and its monthly probability. T10Y3M therefore retains its place as the single variable in that publication.

Why a single spread rather than the entire curve

A legitimate question arises: why reduce the information from the entire curve to a single spread, when the curve contains roughly twenty maturity points (one month, three months, six months, one year, two years, five years, ten years, thirty years, etc.)? Three arguments justify the parsimony of the T10Y3M choice.

First, statistical parsimony. Across a sample of eight documented recessions since 1968, a multivariate model with ten explanatory variables produces predictable overfitting, estimated coefficients capture sample noise rather than reproducible regularity. Estrella and Mishkin’s out-of-sample tests (NBER WP 5379, December 1995) showed precisely that models richer than T10Y3M lose performance on out-of-sample validation windows.

Second, interpretability. A single variable communicates easily, verifies easily, and debates easily. A linear combination of ten curve points requires communicating the weights, their temporal stability, and their theoretical justification. T10Y3M’s readability is a secondary but structuring argument for its institutional adoption.

Third, theoretical anchoring. The transmission channel documented in the literature (compression of bank net interest margins, contraction of credit supply, slowdown of activity) runs through the slope on the short-long interval, not through curvature or internal convexity. Reducing the curve to a binary spread isolates the macroeconomically relevant information, ignoring tactical information (short-term positioning, auction events, duration flows) that does not carry recession signal.

This parsimony explains why the NY Fed probit model remains univariate after three decades of methodological evolution. Attempts to enrich it by adding real indicators (industrial production, employment, leading indicators) have not significantly improved out-of-sample predictive power. T10Y3M has remained the canonical barometer precisely because it is sufficient, adding information does not advance marginal accuracy. The technical detail of this use is covered in the probit formalization of the T10Y3M signal, and the raw historical series is available in the T10Y3M dataset.

The typical inversion cycle in four phases

Across the eight inversions documented since 1968, a recurring temporal pattern emerges that helps situate the current T10Y3M value within a broader cycle, and that fits within the standard reading grid of central-bank policy and rate-cycle transmission on yield-curve evolution. This pattern breaks into four identifiable phases, whose combined reading with the instantaneous spread value enriches interpretation.

Phase 1, approaching the boundary. The T10Y3M, starting from a positive expansion-cycle value (typically 100 to 250 basis points), gradually declines over six to eighteen months under a Fed tightening cycle. The short leg rises with the Fed Funds rate; the long leg slowly embeds the expectation of return toward the long-run equilibrium rate. During this phase, the spread remains positive but compresses, an early warning that the bond market tracks attentively.

Phase 2, inversion and deepening. The T10Y3M crosses zero earlier than the standard account suggests: three weeks after the last Fed hike in 2006, two months in 2000, five months in 2019, and nine months before it in 2022, the 25 October 2022 inversion long preceding the final hike of 27 July 2023. It then keeps digging for three to eighteen months, reaching a trough whose depth varies by cycle, from -19 basis points in 2025 to -189 in May 2023, with a median across cycles of -52. During this phase the NY Fed probit model rises progressively, but its peak rarely clears 50 percent: 33.2 percent in June 1990, 46.3 percent in December 2001, 41.7 percent in March 2008, 37.9 percent in August 2020, and 70.9 percent in May 2024, the only reading above half.

Phase 3, un-inversion from below. The Fed begins cutting the Fed Funds rate, which drives the short leg of T10Y3M down faster than the long leg. The spread mechanically rises, crosses zero, then turns positive. This un-inversion from below, short-rate collapse faster than long-rate decline, is historically the profile observed ahead of the 2001, 2008, and 2020 recessions, and was reproduced in the 2024 un-inversion. It differs from an un-inversion from above (long-rate rise faster than short-rate decline), rarer in modern cycles but theoretically possible in a persistent inflationary shock. Further reading: T10Y3M: The August-to-December 2024 Un-Inversion, End of Record Inversion.

Phase 4, post-cycle normalization. After un-inversion, the T10Y3M continues climbing toward a positive expansion-cycle value, typically between 100 and 250 basis points, as the Fed reaches its new policy-rate plateau. This phase can last from one to five years depending on the depth of the subsequent recession and the speed of the Fed easing cycle. In the ongoing 2024-2026 episode, the climb is anything but linear: the spread dipped back below zero three times during 2025, in March, in June-July and in August, before settling at 80 basis points in May 2026 and 79 on 21 September 2026, the week after the target range was raised to 3.75-4.00 percent on 17 September. The NY Fed twelve-month recession probability stood at 28.8 percent in August 2026.

This four-phase segmentation is not a mechanical projection: the duration of each phase varies by cycle, and some episodes, the three 2025 dips in particular, depart from the standard pattern. But it provides a reading grid that contextualizes the instantaneous spread value within the broader curve movement.

Last updated — 6 October 2026