---
ticker: "USNTFS"
title: "U.S. Near-Term Forward Spread"
unit: "%p"
frequency: "Daily"
source: "Engstrom and Sharpe (2018); Gürkaynak, Sack, and Wright (2007)"
release: "Updated Daily"
category: "Yield Curve & Inflation Compensation"
country: "US"
language: "en"
canonical: "https://kred.dev/en/series/USNTFS"
license: "https://creativecommons.org/licenses/by-nc-nd/4.0/"
latest_value: 0.74
latest_date: "2026-10-16"
first_date: "1980-01-02"
observations_total: 11679
observations_shown: 120
---

# U.S. Near-Term Forward Spread

## Overview

The distance between the forward rate six quarters ahead and the three-month zero rate, both read from one fitted Treasury curve on the same day and expressed in percentage points.

## Key Figures

|  | Value | Date |
|---|---|---|
| Latest | 0.74 | 2026-10-16 |
| Change from previous | -0.07 | 2026-10-15 |
| Change over one year | +1.35 | 2025-10-16 |
| Highest on record | 4.28 | 1982-08-23 |
| Lowest on record | -7.47 | 1980-12-11 |
| Period covered | 1980-01-02 – 2026-10-16 |  |
| Observations | 11679 |  |

## Recent observations

| Date | Value | Change |
|---|---|---|
| 2026-04-27 | 0.05 | +0.03 |
| 2026-04-28 | 0.08 | +0.02 |
| 2026-04-29 | 0.11 | +0.04 |
| 2026-04-30 | 0.11 | -0.01 |
| 2026-05-01 | 0.11 | 0.00 |
| 2026-05-04 | 0.13 | +0.03 |
| 2026-05-05 | 0.15 | +0.02 |
| 2026-05-06 | 0.10 | -0.05 |
| 2026-05-07 | 0.14 | +0.03 |
| 2026-05-08 | 0.12 | -0.02 |
| 2026-05-11 | 0.15 | +0.03 |
| 2026-05-12 | 0.18 | +0.03 |
| 2026-05-13 | 0.19 | 0.00 |
| 2026-05-14 | 0.19 | +0.01 |
| 2026-05-15 | 0.29 | +0.10 |
| 2026-05-18 | 0.29 | 0.00 |
| 2026-05-19 | 0.34 | +0.06 |
| 2026-05-20 | 0.27 | -0.08 |
| 2026-05-21 | 0.26 | -0.01 |
| 2026-05-22 | 0.24 | -0.02 |
| 2026-05-26 | 0.20 | -0.04 |
| 2026-05-27 | 0.20 | -0.01 |
| 2026-05-28 | 0.19 | -0.01 |
| 2026-05-29 | 0.17 | -0.02 |
| 2026-06-01 | 0.19 | +0.03 |
| 2026-06-02 | 0.19 | -0.01 |
| 2026-06-03 | 0.21 | +0.03 |
| 2026-06-04 | 0.20 | -0.01 |
| 2026-06-05 | 0.23 | +0.03 |
| 2026-06-08 | 0.23 | 0.00 |
| 2026-06-09 | 0.22 | -0.01 |
| 2026-06-10 | 0.22 | +0.01 |
| 2026-06-11 | 0.16 | -0.06 |
| 2026-06-12 | 0.18 | +0.02 |
| 2026-06-15 | 0.17 | -0.01 |
| 2026-06-16 | 0.14 | -0.03 |
| 2026-06-17 | 0.12 | -0.02 |
| 2026-06-18 | 0.09 | -0.02 |
| 2026-06-22 | 0.10 | 0.00 |
| 2026-06-23 | 0.10 | +0.01 |
| 2026-06-24 | 0.03 | -0.07 |
| 2026-06-25 | 0.01 | -0.02 |
| 2026-06-26 | -0.01 | -0.02 |
| 2026-06-29 | -0.01 | -0.01 |
| 2026-06-30 | 0.02 | +0.03 |
| 2026-07-01 | 0.05 | +0.03 |
| 2026-07-02 | 0.06 | +0.01 |
| 2026-07-06 | 0.05 | -0.01 |
| 2026-07-07 | 0.08 | +0.03 |
| 2026-07-08 | 0.10 | +0.01 |
| 2026-07-09 | 0.09 | -0.01 |
| 2026-07-10 | 0.09 | +0.01 |
| 2026-07-13 | 0.11 | +0.02 |
| 2026-07-14 | 0.13 | +0.02 |
| 2026-07-15 | 0.12 | -0.01 |
| 2026-07-16 | 0.13 | 0.00 |
| 2026-07-17 | 0.11 | -0.01 |
| 2026-07-20 | 0.13 | +0.02 |
| 2026-07-21 | 0.14 | +0.01 |
| 2026-07-22 | 0.14 | 0.00 |
| 2026-07-23 | 0.15 | +0.01 |
| 2026-07-24 | 0.12 | -0.03 |
| 2026-07-27 | 0.10 | -0.02 |
| 2026-07-28 | 0.09 | -0.01 |
| 2026-07-29 | 0.14 | +0.05 |
| 2026-07-30 | 0.17 | +0.03 |
| 2026-07-31 | 0.20 | +0.03 |
| 2026-08-03 | 0.17 | -0.03 |
| 2026-08-04 | 0.13 | -0.03 |
| 2026-08-05 | 0.14 | +0.01 |
| 2026-08-06 | 0.18 | +0.03 |
| 2026-08-07 | 0.16 | -0.02 |
| 2026-08-10 | 0.20 | +0.04 |
| 2026-08-11 | 0.19 | -0.01 |
| 2026-08-12 | 0.19 | 0.00 |
| 2026-08-13 | 0.18 | -0.02 |
| 2026-08-14 | 0.22 | +0.05 |
| 2026-08-17 | 0.22 | 0.00 |
| 2026-08-18 | 0.23 | 0.00 |
| 2026-08-19 | 0.21 | -0.02 |
| 2026-08-20 | 0.25 | +0.04 |
| 2026-08-21 | 0.26 | +0.01 |
| 2026-08-24 | 0.23 | -0.02 |
| 2026-08-25 | 0.19 | -0.05 |
| 2026-08-26 | 0.21 | +0.02 |
| 2026-08-27 | 0.21 | +0.01 |
| 2026-08-28 | 0.21 | -0.01 |
| 2026-08-31 | 0.22 | +0.01 |
| 2026-09-01 | 0.25 | +0.03 |
| 2026-09-02 | 0.35 | +0.10 |
| 2026-09-03 | 0.65 | +0.30 |
| 2026-09-04 | 0.45 | -0.19 |
| 2026-09-08 | 0.74 | +0.29 |
| 2026-09-09 | 0.76 | +0.02 |
| 2026-09-10 | 0.88 | +0.12 |
| 2026-09-11 | 0.88 | 0.00 |
| 2026-09-14 | 0.89 | +0.02 |
| 2026-09-15 | 0.95 | +0.06 |
| 2026-09-16 | 0.94 | -0.01 |
| 2026-09-17 | 0.85 | -0.10 |
| 2026-09-18 | 0.96 | +0.12 |
| 2026-09-21 | 0.92 | -0.04 |
| 2026-09-22 | 0.88 | -0.04 |
| 2026-09-23 | 1.03 | +0.15 |
| 2026-09-24 | 1.04 | +0.01 |
| 2026-09-25 | 0.90 | -0.14 |
| 2026-09-28 | 0.96 | +0.06 |
| 2026-09-29 | 0.98 | +0.02 |
| 2026-09-30 | 1.02 | +0.04 |
| 2026-10-01 | 0.96 | -0.06 |
| 2026-10-02 | 1.06 | +0.10 |
| 2026-10-05 | 1.03 | -0.03 |
| 2026-10-06 | 0.98 | -0.05 |
| 2026-10-07 | 0.95 | -0.03 |
| 2026-10-08 | 0.89 | -0.06 |
| 2026-10-09 | 0.96 | +0.07 |
| 2026-10-13 | 0.89 | -0.07 |
| 2026-10-14 | 0.85 | -0.04 |
| 2026-10-15 | 0.81 | -0.05 |
| 2026-10-16 | 0.74 | -0.07 |

## Definition

USNTFS is the near-term forward spread, the distance in percentage points between the forward rate six quarters ahead and the three-month zero rate, with both legs evaluated on the same date from one fitted zero-coupon Treasury curve. KRED computes it from the published parameters of that curve by the construction its authors state (Engstrom and Sharpe 2018; Engstrom and Sharpe 2022), so what this page carries is an implementation of a stated formula on a stated input rather than a copy of anyone's numbers. The construction implemented here is the authors' current one, which takes both legs from this fitted curve, while their 2018 statement of the measure took them from a curve estimated by an arbitrage-free three-factor model instead (Kim and Wright 2005), so the inputs differ even where the formula does not. No machine-readable dataset from the authors is available for comparison, so the fidelity claim made here is to their stated formula and inputs and to nothing else.

The curve is a six-parameter extension of the parsimonious Nelson and Siegel (1987) form, fitted each day to Treasury coupon securities (Svensson 1994; Gürkaynak, Sack, and Wright 2007) and belonging to the family of methods that recover a zero-coupon curve from coupon prices (McCulloch 1971; Vasicek and Fong 1982; Bliss 1997). Two consequences follow for the near leg. It is read off the curve and is therefore not a traded bill, and the three-month point lies below the shortest maturity the curve publishes, which makes it an evaluation at the fitting floor of a curve estimated with bills left out, a design choice grounded in the idiosyncratic behaviour of bill yields (Duffee 1996). The far leg is an evaluation in the same sense, and the eighteen-month and twenty-one-month points it rests on are checked by bracketing, since the reconstruction test covers the twelve-month and twenty-four-month points that surround them with mean absolute deviations of 0.001797 and 0.001751 basis points.

The published span begins in 1980 rather than at the 1972 start of the original study, because the six-parameter regime of the input curve begins on 1980-01-02 and no earlier date can be evaluated on the same construction. An observed three-month cross-check exists only from 1981-09-30, so the twenty earliest month-ends carry no second measurement of any kind. Sixteen of those twenty sit below zero and they contain the sample's most extreme reading, −4.321575 pp at 1981-01-31, which nothing else corroborates, and for that reason the era statistics kept with this series describe 539 of the 559 month-ends rather than all of them.

Three statements bound what the series is. It is an observed forward spread that contains a term premium, since a forward rate is an expected path of short rates together with compensation for risk (Hicks 1939; Cochrane and Piazzesi 2005), so it is not risk-neutral and it is not a decomposition of any kind (Adrian, Crump, and Moench 2013). It is not the Korean near-term forward spread that KRED builds on risk-neutral yields with the premium removed, and the two measure different objects, so they are not comparable. And it is not offered as a statement about the business cycle, whatever the literature on the near-dated part of the curve reports (Engstrom and Sharpe 2019; Estrella and Mishkin 1998).

## Methodology

KRED evaluates both legs from the same date's curve parameters and takes their difference, applying no smoothing, no interpolation, and no rescaling. The far leg is the forward rate covering the quarter that begins six quarters ahead, obtained as seven times the twenty-one-month zero rate less six times the eighteen-month zero rate, and the spread is that forward rate less the three-month zero rate, in percentage points. The identity between this quarter form and the equivalent log-price form is asserted on every row at the frozen tolerance, so the construction carries no interpolation anywhere.

A daily row is consumed only when the six curve parameters and the published one-year and two-year zero columns are all present, and a row failing that screen is excluded rather than repaired. On the pinned vintage 527 of 12,148 daily rows since 1980 are screened out, 4.34 % of the panel, of which 526 are market holidays with every screened column absent and one, 2008-03-21, is a degenerate row carrying valid parameters while the published zero columns are withheld. That is coverage composition and not data loss, and the screen is never loosened to raise the count.

The near leg can be set against an observed three-month Treasury rate placed on the same continuous compounding, and that comparison is kept as an internal cross-check that is never published as a series of its own. Over 539 month-ends from 1981-09-30 the mean absolute difference is 14.41 bp, the maximum absolute difference is 116.26 bp, and the mean difference is +10.25 bp. The wedge is two-sided rather than merely positive on average, running from −69.9186 bp at 2023-06-30 to +116.2582 bp at 2008-10-31, and it is neither an effect of the compounding convention nor a quantity that shrinks as the sample advances.

Because the two measurements of the near leg differ by that wedge, they can fall on opposite sides of zero only where the published reading is smaller in absolute value than the wedge on that date, which confines any divergence to the region nearest zero. Across the 539 month-ends the two fall on opposite sides in 36, all 36 of them where the published reading is smaller than 0.50 pp in absolute value and none of the 356 outside that band, while within ±0.10 pp they diverge in 23 of 43. The episode structure differs accordingly, since what the published leg shows as one 27-month stretch from 2005-08 appears in the companion measurement as five fragments totalling 17 months, and over 2020-05 to 2021-01 the published leg is below zero at nine consecutive month-ends while the companion is above zero at eight of them.

The input curve is republished on a weekly cycle, so the trailing edge of the panel lags the calendar by up to about three business days. On the pinned vintage the published leg and the observed cross-check end on the same date, which means the cross-check is never fresher than the series it checks.

## Applications in Economics

A near-term forward spread compares where the market prices the short rate a year and a half out with where the short rate is now, so it reads the near-dated part of the curve rather than its whole slope. A forward rate is an expected path of short rates together with compensation for bearing risk (Hicks 1939), which is why the object is an observed price relation and not an expectation on its own.

The near-dated part of the curve is where policy expectations dominate, since the stance is set at the shortest maturity while transmission runs through longer rates (Bindseil 2004; Borio 1997). Longer spreads mix a larger and more variable premium into the same comparison, which is why the compensation component has been modelled separately in the term-structure literature (Cochrane and Piazzesi 2005; Adrian, Crump, and Moench 2013), and why Engstrom and Sharpe (2018) argue that the near-dated segment carries the information usually credited to the long slope (Engstrom and Sharpe 2019).

Three limits bound the reading. The measure contains a term premium and is therefore not a clean statement of expected policy, so a move in it can be a change in what the market expects or a change in what it charges. Readings close to zero are not determinate, because an observed short-rate cross-check parts company with the published near leg exactly in that region, so this series is read as a level rather than as a state. And in stretches where the policy rate sat at its effective floor the near leg is compressed against that floor, so the distance it can travel is bounded by the policy regime rather than by expectations alone (Wu and Xia 2016).

## Applications in Financial Markets

On a rates desk the measure is read as the market-implied distance between the short rate today and the short rate priced six quarters out, which is the horizon over which a policy path is usually traded. The position that expresses such a view is taken in the curve rather than in the spread itself, since the spread is not an instrument, and the exposure is measured by the duration that maps yield changes into price changes and managed by immunization defined on the whole term structure (Macaulay 1938; Fisher and Weil 1971).

Three measurement facts bear on that use. Both legs are read off a fitted curve rather than quoted, so the series inherits the fitting conventions of that curve and the differences among curve-estimation methods that the literature has compared (McCulloch 1971; Nelson and Siegel 1987; Svensson 1994; Bliss 1997; Gürkaynak, Sack, and Wright 2007). The near leg differs from an observed three-month rate by a wedge that is two-sided and, at the scale of this spread, large enough to matter, and the money-market literature explains why the shortest instruments carry idiosyncratic pricing of their own (Duffee 1996; Hamilton 1996; Furfine 1999; Bartolini, Bertola, and Prati 2002). And the panel's edge lags the calendar by up to about three business days, so the most recent value is not a same-day quotation.

Two uses are therefore ruled out. A threshold reading taken from this leg alone is not supported, because the second measurement of the near leg disagrees with it precisely where such a threshold would be crossed. And the series is not a substitute for a term-premium decomposition, since it contains the premium rather than isolating it (Adrian, Crump, and Moench 2013; Cochrane and Piazzesi 2005).

## Statistical Tests

USNTFS is a spread between two maturities of one fitted zero-coupon curve, the forward rate six quarters ahead less the three-month zero rate, so the statistical object is that constructed spread rather than a price observed in a market (Engstrom and Sharpe 2022). As a mean-reverting spread it is fitted with a constant and no trend, since a fitted trend on a spread of this kind is a specification error. On the 11,621 daily observations spanning 1980-01-02 to 2026-07-24 every procedure rejects its own null, so no clean order of integration is assigned. The augmented Dickey-Fuller test of Dickey and Fuller (1979), in the lag-augmented form of Said and Dickey (1984) with lag length set as in Ng and Perron (2001), rejects the unit root with a statistic of −5.28835 at p = 0.000, the nonparametric Phillips and Perron (1988) test also rejects with a statistic of −5.89707 at p = 0.000, and the Kwiatkowski et al. (1992) test rejects stationarity with a statistic of 2.51746 at p < 0.01. This is not a lone-dissent case, since the two same-null procedures agree with each other and the disagreement lies with the opposite-null test. The GLS-detrended escalation of Elliott, Rothenberg, and Stock (1996) was not computed for this member, and it confirms the augmented regression leg alone, which is the leg that already rejects, so it could not adjudicate a disagreement whose dissenting procedure is the stationarity test.

On the level the first-order autocorrelation is 0.994906 with an implied half-life of 135.714 observations, a descriptive persistence summary and not an integration claim, read with the small-sample downward bias of Andrews (1993) in mind. The Ljung and Box (1978) portmanteau on the level, refining the original Box and Pierce (1970) form, rejects the no-autocorrelation null at lags 10 and 20, with Q = 110216 and Q = 209721 and p = 0.000 in both cases. The multiple-break procedure of Bai and Perron (1998), computed by the dynamic-programming algorithm of Bai and Perron (2003), runs on the level because the level is the object this archetype fixes, and since the triplet certifies no clean stationary reading the dates that search returns are not citable numerals under the platform's break-date traceability rule and none is printed here (Perron 1989).

Three limits belong with the verdict. The tested object is a constructed spread rather than a traded price, since both legs are analytic evaluations of a fitted curve and the near leg sits below the shortest maturity that curve publishes, so measurement error in the near leg enters the object directly. Nothing here states which side of zero the spread lies on, and no duration or ordinal reading is drawn from it, because a second measurement of the near leg parts company with the published one in the region closest to zero, at a size recorded in the methodology. No claim about the business cycle is made either, since the discrimination results reported by the literature on the near-dated part of the curve are properties of those studies and not of this series (Engstrom and Sharpe 2018; Estrella and Mishkin 1998).

This spread is daily and has no defined seasonal period, so the seasonal-unit-root machinery of Hylleberg et al. (1990) and the Canova and Hansen (1995) seasonal-stationarity test are not applicable and are deliberately not run, the seasonal auxiliary regression being degenerate at a daily period (Beaulieu and Miron 1992; Ghysels and Osborn 2001).

## Frequently Asked Questions

### How is the near-term forward spread constructed?

The three-month zero-coupon rate is subtracted from the forward rate six quarters ahead, both legs evaluated on the same day from a single fitted zero-coupon Treasury curve. The original authors' construction is applied to the published curve parameters unchanged.

### How does the near-term forward spread differ from a long minus short slope?

It is built only from the near segment of the curve, so it does not span a decade of maturities and instead reflects the horizon over which policy expectations form. Both legs come from one fitted curve, so no maturity mismatch enters.

### Does KRED read a direction into the near-term forward spread sign?

It does not. KRED publishes the constructed distance and makes no claim about what a particular sign implies, leaving interpretation to the literature the construction follows.