How the 2-Year Treasury Leads the Fed (1976–2026): 50 Years of Cycle Reversals

Why the 2-Year Leads: A Mechanical Explanation

The 2-Year Treasury yield is, by construction, a forward-looking instrument. Its current level reflects the market’s expectation of the average policy rate over the next 24 months, plus a term premium. The Fed Funds Effective Rate, by contrast, is a backward-looking summary of past FOMC decisions — the rate the Fed has already set. Read alongside: the relationship between Treasury issuance and gold demand.

This is the simplest version of the mechanical explanation: a forward-looking instrument tends to move before a backward-looking one whenever the policy path is expected to change. If a critic argues the lead is “trivial,” that is precisely the point — the lead is structural, not predictive in the sense of revealing private information. The 2-Year reflects publicly available information about the likely future path of policy, processed by market participants who collectively price the curve.

But “trivial” understates what the historical record allows us to test. Three observations make the relationship analytically interesting:

1. The lead is variable, not constant. If the relationship were purely mechanical, one would expect the lead to track a stable horizon. Instead, the lead between a DGS2 major pivot and the Fed pivot ranges from 1 to 7 months across the 9 strictly paired cases. The longer leads captured by Chart 2 — out to 34 months in some cases — come from a different measure: the distance from the DGS2’s local peak or trough (within a 36-month window) to the Fed pivot, rather than from the DGS2’s own confirmed cycle reversal. Variability under both measures indicates that markets sometimes anticipate policy changes well in advance, and sometimes move only just before.
2. The Fed sometimes acts against what the 2-Year had signalled. The December 1983 Fed hike pivot occurred one month after a DGS2 easing pivot — the market disagreed, and the Fed reversed course twelve months later in December 1984. This is documented in the table below.
3. The signal asymmetry matters. Deep inversion of the spread (DGS2 well below Fed Funds) has historically been a more consistent leading indicator of Fed cuts than positive spreads have been of Fed hikes. This is documented in the section on the spread as a real-time signal.

The Full Record: 15 Fed Pivots Since 1976

The table below lists every Fed cycle reversal identified by the pivot algorithm, with three measures of DGS2 behavior in the run-up: the 6-month DGS2 change preceding the Fed pivot date, the date and value of the nearest DGS2 local extremum within a 36-month window (peak before an easing pivot, trough before a hiking pivot), and the strict-paired DGS2 major pivot date when one exists in the 24-month window prior.

* 1993-04 trough notation: the local minimum is 3.84% but DGS2 had been declining from a 1992 peak; the trough marks the start of the trend reversal toward the 1994 hike pivot. “Anticipated” = DGS2 6m prior change exceeds 25 bp in the Fed’s eventual direction. “Strict pair” = a same-direction DGS2 major pivot identified by the same algorithm applied to DGS2, within the 24-month window prior to the Fed pivot date.

Two Frameworks: Retrospective Lead Time vs Real-Time Signal

The lead times in Chart 2 are measured retrospectively. For each Fed pivot that has already happened, we look backward to find the DGS2 extremum or major pivot. This is useful for characterising the historical relationship but not for real-time use — at any given month, an observer does not know whether the current DGS2 move is a “peak” or “trough” until well after the fact.

A real-time framework is provided by the spread itself: DGS2 minus Fed Funds, observable each month. The spread does not require knowing whether a peak has formed; its level alone places the current observation in one of five historical regimes, each with its own distribution of subsequent Fed actions. This is the framework explored in the next section.

The two frameworks are complementary, not competing. Retrospective lead-time analysis answers “when DGS2 moved before each Fed pivot, by how much?”. Real-time spread analysis answers “given today’s spread, what has historically happened to the Fed Funds rate over the following 12 months?”. Both are needed to characterise the relationship fully.

The Spread as Real-Time Signal: Five Regimes

The DGS2-minus-Fed Funds spread is observable at any moment and admits a direct interpretation: it is the market’s expected average Fed Funds rate over the next 24 months minus what the Fed has actually set, plus a small term premium. When the spread is sharply negative, market participants are collectively pricing a future Fed Funds rate below the current setting — that is, expecting cuts. For the complete series, see the 10-year term-premium data.

The historical record provides a rough base rate for each spread regime. The table below summarises forward 12-month Fed Funds behavior conditional on the spread regime at the observation date, computed over the 585 monthly observations (1976-06 to 2025-02) for which a 12-month forward window was available in the May 2026 edition. The base rates are kept as published; the September 2026 refresh adds observations, not a recomputation.

Asymmetry note. The signal is sharper on the negative side. Deep inversion (32 months, around 5% of sample) preceded a Fed cut within 12 months in 84% of cases. Positive spreads above +100 bp (95 months, 16% of sample) preceded a Fed hike in only 48% of cases, with 39% preceding a cut instead. Positive spreads frequently arise during late-cycle easing phases, when the market expects continued cuts despite already low Fed Funds — a configuration that has not historically preceded hikes.

The Pearson correlation between the spread and the subsequent 12-month change in Fed Funds is +0.34. This is a modest correlation, consistent with the spread being one informative signal among many rather than a deterministic predictor.

When the 2-Year Was Wrong

Any signal that worked in 13 of 15 cases did not anticipate the Fed in the other 2. Examining these two cases, together with the noisier paired-pivot record, identifies the boundary conditions of the relationship.

January 1990: the Fed cut while DGS2 was rising

In the 6 months leading to the Fed’s first easing pivot of 1990 (January 1990), the DGS2 had risen 27 basis points — moving against the Fed’s eventual direction. The longer-horizon picture is more nuanced: DGS2 had peaked at 9.68% in March 1989 and was 159 bp below that peak by January 1990, so the 12-month view shows DGS2 declining. The 6-month window simply caught a counter-trend bounce. The strict-paired DGS2 major pivot in August 1989 also identifies the cycle reversal 5 months before the Fed acted. So this case is best characterised as a true anticipation by DGS2, with a short-term move against the eventual direction in the final 6 months. A companion piece: the chronology of cuts into a bull market.

June 2021: the Fed led the 2-Year

The unique case where the Fed pivoted before DGS2 is the June 2021 shift from the pandemic floor. As reported in the Federal Reserve’s record of FOMC rate decisions. DGS2 had moved only +6 bp in the 6 months prior — barely off the all-time low of 0.12% reached in February 2021. The DGS2 major pivot, identified by the algorithm at July 2021, came one month after the Fed’s pivot. This is consistent with the unusual policy environment: the Fed was emerging from an explicit yield-curve-control-adjacent regime where short-rate forward guidance had suppressed front-end volatility. The 2-Year was anchored by guidance, not by market expectations — so it could not move first.

What these cases share

The exceptions cluster around regimes where front-end rates do not freely reflect market expectations — explicit forward guidance (2021) and short-term counter-trend bounces inside longer trends (1990). When the 2-Year is anchored by policy commitment rather than expectation, the lead-lag relationship breaks down. This is itself a useful diagnostic: when forward guidance is binding, the spread loses information value.

September 2026 in Real Time

The May 2026 edition of this study closed with the spread at −17 bp, in the NEUTRAL regime, and with both series drifting lower. What followed is the first out-of-sample test of the lead-lag record since publication.

The 2-year yield bottomed at 3.38% on 27 February 2026 and closed at 4.67% on 15 September, the day before the FOMC raised its target range to 3.75–4.00%, the first increase since July 2023 (Federal Reserve press release of 16 September 2026). Over the same period the effective federal funds rate did not move: 3.64% in January and February, 3.63% from May to the eve of the decision. On the monthly means used throughout this study, DGS2 rose from 3.47% in February to 4.22% in August, +75 bp over six months, three times the 25 bp threshold that defines an anticipated reversal in the table above. On 16 September itself the 2-year closed at 4.74%, its 2026 high, above the top of the new target range.

The spread told the same story in the five-regime framework. It left the NEUTRAL band in May (+37 bp), reached +59 bp in July and August, and stood near +88 bp on the partial September mean. The POSITIVE regime is, on the historical base rates, the least decisive of the five on the hiking side: a hike within twelve months followed 51% of its 249 months, a cut 18%. The regime alone was close to a coin flip. The direction was carried by the level of the 2-year, which had climbed above the funds rate by more than the FOMC’s usual 25 bp increment four months before the Committee moved.

Two cautions apply. First, the daily figure of +129 bp and the monthly figure of +75 bp measure different things, close-to-close extremes versus monthly means, and do not reconcile to each other. Second, the September 2026 hike is not yet a dated reversal under the pre-registered pivot rule: the rule requires a 100 bp follow-through within 24 months, which only the coming FOMC decisions can supply. Until then the 2026 episode is a real-time observation set beside the fifteen dated reversals, not a sixteenth row in the table. The six months added to the dataset in September 2026 carry no new pivot flag for that reason.

Current State and Levels to Watch

As of August 2026 (monthly means): DGS2 stands at 4.22%, the Fed Funds Effective Rate at 3.63%, and the spread at +59 basis points. This places the latest full-month observation in the POSITIVE regime (+25 to +100 bp), which historically preceded a Fed hike greater than 25 bp within 12 months in 51% of cases, a cut in 18%, with a median forward change of +26 bp. On 16 September 2026 the FOMC raised its target range to 3.75–4.00%; the 2-year closed at 4.74% that day and the partial September mean puts the spread near +88 bp, close to the +100 bp boundary of the VERY POSITIVE regime.

Methodology

Data

• DGS2. US Treasury Constant Maturity 2-Year, daily series from FRED, aggregated to monthly mean. Start: 1976-06.
• FEDFUNDS. Federal Funds Effective Rate, monthly series from FRED. Coverage to 2026-08 at the refresh of 18 September 2026 (2026-02 in the May 2026 edition).
• NBER recession dates. Standard NBER business cycle reference dates.
• Sample window. 1976-06 to 2026-08, the overlapping window of both series, n = 603 monthly observations. The pivot flags, the table of reversals and the regime base rates are those of the May 2026 edition; the six months added in September 2026 are observations, not new pivots.

Pivot algorithm (pre-registered)

A “major pivot” in either series is defined as the first month in which all three conditions hold:

1. The 12-month change of the 3-month rolling average of the series reverses sign relative to the prior month.
2. The series moves at least 100 basis points in the new direction within the following 24 months (confirms the reversal is not a short-lived wobble).
3. At least 12 months have elapsed since the previous major pivot (prevents double-counting of two-sided wobbles inside a single cycle).

Applied to monthly Fed Funds, this identifies the 15 reversals listed. Applied to monthly DGS2 mean, it identifies 16 reversals. Sensitivity at the 75 bp and 125 bp thresholds yields 16 and 13 Fed reversals respectively, with the same core ordering. The 13/15 anticipation count is stable across these threshold variants.

“Anticipated” definition

A Fed pivot is classified as anticipated if the DGS2 monthly mean changed by more than 25 bp in the Fed’s eventual direction over the 6 months preceding the pivot date. At the ±10 bp sensitivity threshold the count remains 13/15 (the borderline 2021-06 case at +6 bp stays below the threshold); at ±50 bp it falls to 12/15 (the 1994-02 case at +47 bp falls out).

Spread regime thresholds

The five regimes are defined by spread thresholds at ±25 bp and ±100 bp, applied to the monthly DGS2 minus Fed Funds difference. These thresholds were chosen before computing forward-action statistics and correspond to commonly cited demarcations in the literature (e.g., the ±25 bp band is a typical FOMC inter-meeting move). Sensitivity at ±50 bp and ±150 bp shifts the regime counts but preserves the qualitative ranking of conditional cut/hike frequencies.

Forward Fed action measurement

For each observation date, the 12-month forward Fed Funds change is the difference between Fed Funds 12 months ahead and the current month’s Fed Funds. “Cut within 12 months” means this difference is less than −25 bp; “hike within 12 months” means greater than +25 bp. The 25 bp threshold is a standard FOMC move increment.

Limitations

• The window of 1976-06 to 2026-08 covers six recessions and multiple monetary regimes. Stationarity of the lead-lag relationship across regimes is not formally tested.
• The 36-month window for the local-extremum lead-time analysis caused censoring in one case (1980-11). A longer window would identify a longer lead in that case.
• Monthly aggregation of DGS2 smooths through intra-month volatility relevant to high-frequency analysis but does not affect cycle-level conclusions.
• The Fed Funds Effective Rate is itself a market-determined outcome of the Fed’s target range. In the post-2008 administered-rate regime, it tracks the IORB / ON-RRP corridor very closely and behaves as a policy rate; in earlier periods it could diverge slightly from the target.
• The relationship is partly endogenous: the Fed observes the 2-Year when setting communication. This does not invalidate the historical record but does limit how much one can attribute causation to either side.

Reproducibility

The full dataset, including the derived columns used in every figure in this study, is published as an open dataset under CC BY 4.0. See the dataset link below.