Macro ReviewHow to read
Outlook · Synthesis · Discount rate

Three pooled models centre the 10-year on 4.7% a year from now

Where could the 10-year be in a year, and which of its parts would do the moving?

Staff curve · 6 Nov 2026Parameters estimated on data to Oct 2026Updated 16 Nov 2026
10-year, twelve months on4.69%3.71%10th–90th pooled paths5.66%from 4.88% in Nov 2026 · pooled, four in five
, filtered0.69%0.29%across the three anchors0.97%Nov 2026 · survey anchor, bar spans three anchors
Expected short-rate average, 10 years4.19%Nov 2026 · filtered, ±0.15 at 90%
The 12-month band's record75%72 quarters from 2008 · Kupiec p 0.30holds its promise

In brief

The pooled range for the 10-year over the next year, the record that lets it stand here, and the parts that move it
  • In four of five pooled paths the 10-year zero sits between 3.7% and 5.7% in Nov 2027.→ Exhibit 1
  • Drawn at each quarter since 2008 on the data then published, the 12-month band held 75% of outcomes.→ Exhibit 3
  • Of the expected path's −0.18pp over the year, the supplies −0.08pp.→ Exhibit 2
  • The term premium reads 0.69% on the survey anchor; the other two anchors move it by up to 0.40pp.→ Exhibit 5
  • About 97% of the 10-year's 1.09pp gap over its long-run anchor looks persistent.→ Exhibit 6
What would change this view
What would change this viewTriggerTodayDecided by
The 10-year leaves the 12-month 80% band10-year zero below 3.71% or above 5.66% at end-Nov 20274.88% · 6 Nov 2026pendingBoard staff curve, end-Nov 2027
The 10-year leaves the 3-month 80% bandbelow 4.31% or above 5.35% at end-Feb 20274.88% · 6 Nov 2026pendingBoard staff curve, end-Feb 2027
The pooled band's record fails its coverage testsKupiec or independence p at or below 0.05 at a quarter-endKupiec p 0.30, independence p 0.57not crossedthe record, each quarter-end
The reasoning behind each point
  • Range
    In four of five pooled paths the 10-year zero sits between 3.7% and 5.7% in Nov 2027.

    From 4.88% in Nov 2026 (the Board staff curve's zero-coupon 10-year, latest day 6 Nov 2026), the pool of the model's own paths, the factor projection and the event calendar puts the median at 4.69% in twelve months and the 5th–95th at 3.43%–5.94%. In three months: 4.31%–5.35% (10th–90th), a band whose record does not yet pass its coverage tests, so it is drawn in the detail. The weights are the record's: statistical 0.17, factor 0.17, event calendar 0.67 (log score, shrunk). Model-conditional: what the three models imply if they are right, not a forecast.

  • Record
    Drawn at each quarter since 2008 on the data then published, the 12-month band held 75% of outcomes.

    72 quarterly origins from Jan 2008 to Oct 2025; the 80% band promises 80%. Kupiec's test of the miss rate: p 0.30; the independence test on the non-overlapping annual sequences (overlapping windows fail it by construction): p 0.57. Its continuous ranked probability score (lower is better) is 0.487 against the random walk's 0.531. Alone, the model's own paths held 57%, the factor projection 69% and the event calendar 79%. The 3-month band held 69% (Kupiec p 0.03). The data are masked to what was published at each origin; the parameters are this month's, fitted on the whole sample — a stated limit of the record.

  • Parts
    Of the expected path's −0.18pp over the year, the term premium supplies −0.08pp.

    The pooled mean path is additive: the 10-year's expected change of −0.18pp to Nov 2027 is the expected short-rate average's −0.10pp and the term premium's −0.08pp. Today the model reads the expected average at 4.19% (expected real rate 1.52%, expected inflation 2.66%) and the premium at 0.69%; r* 1.24% and π* 2.55% make the endpoint i* 3.78%.

  • Anchor
    The term premium reads 0.69% on the survey anchor; the other two anchors move it by up to 0.40pp.

    The identities fix the sum of the parts, not where the level splits: the three normalisations fit about equally well (log-likelihoods 463.6, 464.5, 466.3) and put the premium at 0.29% (real group), 0.97% (premium group), 0.69% (survey), against the filter's own 90% band of ±0.15pp. The survey anchor (the SPF's ten-year bill rate and CPI10 unbiased) is the headline; the range is printed as the anchors' answer, not as a probability.

  • Gap
    About 97% of the 10-year's 1.09pp gap over its long-run anchor looks persistent.

    A local level plus a transitory AR(1) on the 10-year less i* (survey anchor) since Jan 2001: persistent 1.05pp, transitory +0.04pp with a half-life of 2.3 months, so the expected gap is 1.07pp in three months and 1.05pp in twelve. With the level's variance at half and double its estimate the persistent share is 88% and 101%. Under the other anchors: real group gap 0.69pp, 96% persistent; premium group gap 1.37pp, 97% persistent.

  • 10-year
    The 10-year, at 5.12%, sits 106bp above the level a hold at today's rate would justify.

    It is 100bp above the SEP path's fair value. A path's fair value is the 10-year that policy path justifies at today's term premium: 4.065% on a hold, 4.115% with one further 25bp, 4.165% on the grid's +50bp path — every one below today's par (16 Nov 2026), so the market prices more tightening than any policy path in the grid; the market's own path, the ACM-fitted 10-year, is 5.111%. On this page's rates beta — a regression estimate, not the valuation identity that Scenarios and Playbook & targets use — completing the hold path would give the S&P 500 back +11.27% (+10.73% on the SEP path).

  • So what
    On this page's model, completing the SEP path would give the S&P 500 +10.7% and homebuilders +24.5%.

    The beta is −1.08% per +10bp on the 10-year for the S&P 500. Homebuilders are the most rate-sensitive of the 24 (−2.38% per +10bp on the 10-year) and consumer staples the least (−0.54%); read the direction and the ranking, since a weekly beta applied to a 100bp move is linear where the world is not. The 90% Newey–West bands are −2.96 to −1.80 and −0.89 to −0.19. The S&P 500 actually returned −1.62% since the September decision, of which the rates factor taken 1.17% (beta band −1.75 to −0.60%), leaving −0.45pp it does not explain. The SEP path's +10.73% for the S&P 500 is this page's model — model-conditional, what the model implies under stated inputs, not a forecast: a total-effect beta (−1.08% per +10bp from 156 weekly returns since 24 Nov 2023, regressed on oil, the ten-year, HYG − LQD and the dollar with no market factor, so the index's own co-movement with yields in that window is inside it) times the −100bp still to come, with the other three factors held and no horizon. The Scenarios page prices the same paths two other ways — the valuation identity, with the premium and growth held, and the factor engine's draws conditioned on the path — and sets the three models side by side; they are not expected to match, and that page says why. The day-0 pass-through of a policy surprise to the 10-year is 0.42 (4.2bp per +10bp of surprise, 1988–2023), which is why a path for the policy rate does not translate one-for-one into the factor this page uses.

  • Quarter
    Over the next quarter the factor model's S&P 500 median runs from −1.1% to −0.9% across the paths, all inside the range with no scenario.

    The no-scenario 10th–90th band over 1 quarter is −11.6% to +15.8%. At a year the paths' medians lie within a fraction of a point of one another, because the model's stock–bond correlation reverts towards zero; a quarter is where they part.

  • Paths
    The rate paths put the 10-year between 4.07% and 4.17%; none of them carries odds.

    The Fed paths — hold path −106bp, SEP path −100bp, above the dots −96bp — each set the 10-year at the fair value What is priced gives a policy path: the distance from today's yield, so a path asks what happens if the 10-year travels that distance, not when or why. The path above the dots takes the policy-rules grid's one path above the SEP's, +50bp then hold; its fair 10-year still sits below today's yield, so its move (−96bp) is the smallest fall of the three — the market prices more tightening than even that path. The 4 Hormuz paths are listed without odds and are not projected: they carry no oil level yet; the formula cannot price them either, because no elasticity of index earnings or of the premium to oil exists among the site's estimates. No path carries a probability: no source the site reads states one. Setting the 10-year conditions the draws rather than replacing them, so the other factors keep their tails.

  • Formula
    At the hold path's 10-year the model's value is still −29.7% from the price; the formula's full pass-through, +29.6%, is arithmetic, not a target.

    Valuation-model arithmetic: the Valuation page's two-stage model — price over trailing earnings = k [(1 + g_L) + H (g_S − g_L)] ÷ (r − g_L) — at its priced inputs (k 0.84, E0 ÷ P 3.985%, g_S 15.1% fading to g_L 4.0% over 6 years, r = 5.12% + 3.502pp), which return today's price. A path moves rf by its full distance and holds the premium and growth — full pass-through, the extreme case — so its move from today is (r0 − g_L) ÷ (r1 − g_L) − 1: +29.6% on the hold path, +27.8% on the SEP path, +26.0% above the dots. The response is convex — −22% per point of the discount rate at today's rate — because r − g_L is 4.62pp today. Oil, credit and the dollar are not modelled on the formula: the site estimates no elasticity of index earnings or of the premium to them.

  • Market
    The market prices the 10-year at 5.12%, 106bp above the fair value of a hold.

    From What is priced (16 Nov 2026): the hold path's fair value is 4.065%, the SEP path's 4.115% and the market's own, the ACM-fitted 10-year, 5.111%; the 1y1y forward sits +115bp over the policy rate; the SEP medians are 4.1 for 2026 and 4.1 for 2027, 3.2 in the longer run (16 Sep 2026).

  • Unconditional
    The factor model's 1-year central band for the S&P 500 runs from −19.3% to +35.6%.

    Over 1 year the S&P 500's 10–90 band is −19.3% to +35.6% with the median at +4.7% — close to the cash leg because the simulation carries no drift, not because the engine expects it.

  • Projection coverage
    The published band held 82.4% of outcomes over the 9 quarter-ends kept out of its calibration, from 30 Jun 2024.

    Calibrated: the 10–90 band held 77.1% of 214 asset-quarters since 31 Mar 2022 (74.0–79.9%) against a nominal 80%. Calibrated, so the factor projections head this page beside the engine cones: the width scale estimated on the quarter-ends to 31 Mar 2024 (k = 1.07) held 82.4% of the held-out quarter-ends from 30 Jun 2024 (76.4–88.2), which contains 80%, and the published bands carry it. The 10–90 band held 79.4% (76.8–82.0) with fixed betas and a GARCH-t residual, 78.0% (75.0–80.9) with the betas drawn and 77.1% (74.0–79.9) with the residual bootstrapped; widened by k = 1.07 it would hold 81.3% (78.5–84.0) — applied. Drawing the betas moves the pooled coverage −1.4pp; resampling the residual instead of simulating it moves it −0.9pp. Before any scale the band held 76.4% of the first 9 quarter-ends (to 31 Mar 2024) and 77.8% of the 9 since. A scale estimated on the first half moves the second to 82.4%, which holds, so it is applied. The backtest published before 24 September left the residual out of the band it scored and mapped the four factor-proxy ETFs through betas, so it tested a narrower object than the page drew. The verdict sits on a knife-edge: the uncalibrated band's interval misses 80% by 0.1pp, and under 1 of the 4 alternative seed triples the spec lists it contains 80% instead — so the page's placement owes as much to the simulation's seed as to the bands.

Next: Sector & style outlook — each basket a year ahead

Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculations · data to 6 Nov 2026cross.discount_rate-read

Where the 10-year could be in twelve months, on three models pooled

The 10-year 's pooled range month by month for a year, from today's filtered state, beside its own history

Exhibit 1 · The 10-year Treasury yield's history and its modelled range a year ahead

Zero-coupon 10-year yield, %, month-end, Board staff curve; then the range of three combined models' paths, Nov 2026 to Nov 2027
Range: simulated shocks
About
About this exhibit
Line chart of the 10-year zero-coupon yield at month-end (Board staff curve), followed by the combined three-model range of paths over the next twelve months
The 10-year and its pooled range for the next twelve months
Percent, month-end, Board staff zero-coupon curve; then the pool of three models' paths from Nov 2026.
The pool mixes whole paths from the measurement model's own simulation, the factor projection and the event calendar, weighted by their record (S 0.17, F 0.17, E 0.67). Model-conditional: what the three models imply from today's filtered state if they are right, not a forecast. The band's record is the next section.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-2

Exhibit 2 · Expected change in the 10-year, split into short-rate and term-premium parts

Change in the combined models' mean from Nov 2026, pp, quarterly to Nov 2027; the 10-year's change equals the expected short-rate average's plus the premium's
Range: simulated shocks
By10-yearExpected short-rate averageTerm premium
Feb 2027−0.05pp−0.03pp−0.01pp
May 2027−0.09pp−0.06pp−0.03pp
Aug 2027−0.14pp−0.08pp−0.06pp
Nov 2027−0.18pp−0.10pp−0.08pp

From 4.88% today: expected short-rate average 4.19%, term premium 0.69% (filtered, survey anchor).

About this exhibit
Table of the change in the combined models' mean from the origin month, in pp, at each quarter of the next year, for the 10-year, the expected short-rate average and the term premium
What moves the expected path
Change in the pooled mean from Nov 2026, pp. The mean is additive: the 10-year's change is the sum of its parts'.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-3

Drawn at each quarter since 2008, the pooled band held about as often as it promised

Coverage — how often the outcome fell inside the band drawn at the time on the data published by then — for each family and the pool, with two tests of the misses

Exhibit 3 · 10-year yield outcomes against the band drawn a year earlier

%, quarterly origins Jan 2008 – Oct 2025; black: outcome inside the 10th–90th band drawn 12 months earlier, red: outside; held 75% of 72, Kupiec p 0.30
Range: out-of-sample recordAbout
About this exhibit
Chart of the 10-year outcome at each quarterly origin since 2008, against the 10th–90th band drawn twelve months earlier; dots are black inside the band and red outside
The 12-month band at each quarter, against the outcome
Percent; dot black inside the 10th–90th drawn 12 months earlier, red outside. Held 75% of 72; Kupiec p 0.30, independence p 0.57.
Each band is the published pool as the rule would have weighted it then, drawn from the model filtered on the data published by the origin's month-end. The parameters are this month's, fitted on the whole sample: the data are real-time, the parameters are not. 12-month windows overlap, so the independence test runs on the non-overlapping annual sequences.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-4

Exhibit 4 · 10-year band coverage and forecast scores for each model and their combination

Quarterly origins since Jan 2008, 12 and 3 months ahead: share of outcomes inside the 10th–90th band, %, Kupiec and independence p, CRPS and log score
Range: out-of-sample record
Family12 months3 months
Held %Kupiec pIndep. pCRPSLog scoreHeld %Kupiec pIndep. pCRPSLog score
Pool, the record's weights (published)750.300.570.487-1.38690.030.020.279-0.74
Pool, equal weights760.450.520.479-1.38690.030.030.281-0.77
Statistical (the model's own paths)57<0.010.410.503-1.7259<0.010.790.290-1.10
Factor (direct projection)690.030.580.488-1.4767<0.010.450.280-0.77
Event calendar790.860.520.490-1.39730.160.040.284-0.79
Random walk with the 10-year's own past changes64<0.010.640.531-1.41730.160.400.290-0.80
Pool of statistical and factor63<0.010.530.484-1.4867<0.010.450.282-0.85
Factor as a Bayesian VAR (scenarios only)26<0.010.420.828-4.0555<0.010.270.370-1.58
… with a sum-of-coefficients prior35<0.010.670.654-2.4557<0.010.690.329-1.19
Event calendar, no volatility scaling780.640.210.480-1.27790.77<0.010.282-0.75
About this exhibit
Table for quarterly origins since 2008, at twelve and three months: band coverage, Kupiec and independence p-values, CRPS and log score; rows for the published combination, equal weights, each of the three models, a random walk and variants; shaded when coverage is far from nominal
Each family's record, and the pool's
The 10-year at quarterly origins since Jan 2008: coverage of the 80% band (shaded when more than 6pp off), Kupiec p, independence p, the continuous ranked probability score and the log score.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-5

The parts of the 10-year, and how much of today's gap looks persistent

The measurement model's filtered expected short-rate average, , and π*, with the range the three anchors give, and the gap over the long-run anchor

Exhibit 5 · Estimated parts of the 10-year yield, and the term premium by anchor

%, monthly since Jan 2001, one-sided estimates with 5th–95th bands, survey anchor: expected short-rate average, term premium, r* and π*; dashed: the premium under other anchors
Range: other models
About
About this exhibit
Monthly line chart of the model's one-sided estimates of the 10-year's parts on the survey anchor (expected short-rate average, term premium, r* and π*) with their bands; dashed lines show the term premium under the other two anchors
The 10-year's parts, filtered, with the other anchors' premium
Percent, monthly, one-sided estimates (survey anchor) with their 90% bands; dashed: the premium under the other two anchors.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-6

Exhibit 6 · Persistent share of the 10-year's gap over its long-run level, by anchor

In pp, the 10-year less its long-run level (r* + π*), Nov 2026: a lasting and a passing part with its half-life, by anchor
Range: assumed valuesAbout
AnchorGapPersistentHalf-life, monthsIn 3mIn 12mShare at ½ · 2×
Survey anchor (headline)1.091.05 97%2.31.071.0588% · 101%
Real-rate sources' average0.690.67 96%2.10.680.6783% · 102%
Affine premia's average1.371.32 97%2.51.341.3290% · 100%
About this exhibit
Table with one row per anchor: the 10-year less its long-run level (r* + π*), its persistent part and share, the transitory part's half-life, the expected gap in three and twelve months, and the persistent share at half and double the level's variance
The gap over i*, persistent and transitory
The 10-year less i* = r* + π*, pp: a local level plus a transitory AR(1); the persistent share at half and double the level's variance.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-7

The fair 10-year of each stated policy path, at four term premia

Each path's ten-year average funds rate plus a — today's, the , the 2023 peak and today's plus 50bp — and the paths quarter by quarter: arithmetic on today's curve and a stated path, not a forecast

Exhibit 7 · Fair 10-year yield for each policy path at each term premium

%: each path's ten-year average plus a term premium; paths start at IORB 3.90% and take the longer-run 3.2% from Q1 2029; arithmetic, not a forecast
About
Policy path10-year average of the pathToday's ACM 10-year term premium
+0.73pp
2015–19 ACM average
−0.29pp
2023 ACM peak
+0.46pp
Today +50bp
+1.23pp
Hold at IORBno steps3.34%4.07%3.05%3.80%4.57%
One further +25bp, then holdlevel after steps 4.15%3.39%4.12%3.10%3.85%4.62%
+50bp, then holdlevel after steps 4.40%3.44%4.17%3.15%3.90%4.67%
Cuts of 25bp a quarter from 2027 Q1 to 3.00level after steps 3.00%3.21%3.94%2.93%3.67%4.44%
Cuts of 25bp a quarter from 2027 Q3 to 3.00level after steps 3.00%3.26%3.98%2.97%3.72%4.48%

The bold cell (hold, today's ACM) gives 4.07% against a par 10-year of 5.12%: a gap of +106bp, of which +105bp is the ACM expected-rate component (4.39%) above the flat path's average (3.34%) and +1bp the par-versus-fitted-zero difference. The gap is not zero by construction.

About this exhibit
Grid: the fair 10-year for the stated policy paths (hold, +25bp, +50bp and two cut paths) at four term premia (the latest ACM, the 2015–19 average, the 2023 peak, the latest +50bp), with each path's ten-year average.
Arithmetic on today's curve and a stated policy path — not a forecast
Arithmetic on today's curve and a stated policy path — not a forecast. 10-year = the path's ten-year average + a term premium (Q1 2027 to Q4 2036). Paths start at IORB 3.90% and are flat at the SEP longer-run median 3.2% from Q1 2029; today's par 10-year 5.12%
Source: Federal Reserve Bank of New York (EFFR record, HLW and LW r*, ACM); Board of Governors (projections, IORB); Macro Review calculations · data to 16 Nov 2026rates.rules-7

Exhibit 8 · The policy paths for the funds rate, quarter by quarter

% at each quarter to Q1 2029 and one quarter beyond; every path is at the longer-run median from there to Q4 2036
QuarterHold at IORBOne further +25bp, then hold+50bp, then holdCuts of 25bp a quarter from 2027 Q1 to 3.00Cuts of 25bp a quarter from 2027 Q3 to 3.00
Q1 20273.90%4.15%4.40%3.90%3.90%
Q2 20273.90%4.15%4.40%3.65%3.90%
Q3 20273.90%4.15%4.40%3.40%3.90%
Q4 20273.90%4.15%4.40%3.15%3.65%
Q1 20283.90%4.15%4.40%3.00%3.40%
Q2 20283.90%4.15%4.40%3.00%3.15%
Q3 20283.90%4.15%4.40%3.00%3.00%
Q4 20283.90%4.15%4.40%3.00%3.00%
Q1 2029 longer run from here3.20%3.20%3.20%3.20%3.20%
Q2 20293.20%3.20%3.20%3.20%3.20%
About this exhibit
Table of the stated policy paths' funds rate at each quarter from the next quarter to the quarter after the paths reach the longer-run median.
The paths, quarter by quarter
% at each quarter to Q1 2029 and one quarter beyond; every path is at the longer-run median from there to Q4 2036
Source: Board of Governors (projections, IORB); Macro Review calculations · data to 16 Nov 2026rates.rules-8

The named policy paths, and what completing each would be worth

What each path assumes and the fair 10-year it gives, how each is put together quarter by quarter, and the rates contribution still to come for every equity asset if the 10-year completes the move — , on a total-effect beta

Exhibit 9 · Each policy path's fair 10-year, what it assumes and the distance left

%: fair 10-year = the path's ten-year average short rate + the latest ACM term premium (0.725pp); 'still to move' against the 5.12% par 10-year, 16 Nov 2026
About
PathFair 10-yearStill to moveWhat it assumesAfter the stepsThe numbers behind it
Hold at IORBHold at IORB4.065%−105.5bpThe Committee holds at IORB 3.90% (the target-range midpoint 3.875%) through 2028 Q4, then the Committee's longer-run 3.2% from 2029 Q1. The policy-rules page's grid averages that path's expected short rate over ten years (3.340%) and adds the ACM term premium of the term-premium page's latest reading (0.725pp, 13 Nov 2026): a fair 10-year of 4.065%.3.90% through 2028 Q4, then the Committee's longer-run 3.2% from 2029 Q1
  • IORB · 17 Sept 20263.90%
  • Target-range midpoint · 17 Sept 20263.875%
  • Fair 10-year, hold path · 13 Nov 20264.065%
One further +25bp (the SEP path)One further +25bp, then hold4.115%−100.5bpOne further +25bp and then a hold at 4.15% through 2028 Q4, then the Committee's longer-run 3.2% from 2029 Q1 — the Committee's own median for 2026 (4.1%) is 22bp above the range midpoint, 1 further move. The same grid cell at today's term premium: 4.115%.4.15% through 2028 Q4, then the Committee's longer-run 3.2% from 2029 Q1
  • SEP median funds rate 2026 · 16 Sept 20264.1%
  • SEP 2027 · 16 Sept 20264.1%
  • SEP longer run · 16 Sept 20263.2%
  • Fair 10-year, one further +25bp · 13 Nov 20264.115%
What the market pricesACM-fitted 10-year5.111%−0.9bpWhat the market itself prices: the 1y1y forward 5.05% sits +115bp above IORB 3.90% (2y1y 4.96%, +106bp), and the ACM decomposition's fitted 10-year of 13 Nov 2026 is 5.111%, 0.9bp below the par 10-year; its expected-rate part averages 4.386% over ten years against the hold path's 3.340%.the forwards as the curve prices them; the ACM fit carries its own expected-rate path to ten years
  • 1y1y forward · 16 Nov 20265.05%
  • 2y1y forward · 16 Nov 20264.96%
  • 1y1y − IORB · 16 Nov 2026+115bp
  • ACM-fitted 10-year · 16 Nov 20265.111%
+50bp, then hold (above the SEP path)+50bp, then hold4.165%−95.5bpThe one path above the SEP's in the policy-rules grid: a step of +50bp to 4.40% in the first quarter, held through 2028 Q4, then the Committee's longer-run 3.2% from 2029 Q1 — 25bp above the SEP path and 50bp above the hold. At today's term premium its fair 10-year is 4.165%, still 96bp below today's yield: the market prices more tightening than even this path. A path +50bp above the SEP's own medians is not a cell the grid publishes.4.40% through 2028 Q4, then the Committee's longer-run 3.2% from 2029 Q1
  • Funds rate after the step · 16 Sept 20264.40%
  • SEP 2027 · 16 Sept 20264.1%
  • Fair 10-year, +50bp then hold · 13 Nov 20264.165%
About this exhibit
A table of the policy paths (hold, SEP, the market's own, two steps then hold): fair 10-year, distance still to move, what the path assumes, the rate after the steps, and the dated figures behind it.
The paths and what defines them
Each path is a fair value for the 10-year, not a forecast: the policy-rules page prices the path's average expected short rate over ten years and adds the ACM term premium of the term-premium page's latest reading (0.725pp, 13 Nov 2026). Today's par 10-year is 5.12% (16 Nov 2026). Every policy path holds its stepped level through the years in the "After the steps" column and takes the Committee's longer-run rate after them — the assumption a reader most often misses.
Source: U.S. Treasury; Federal Reserve Board (FOMC projections); New York Fed (ACM term premium); Macro Review calculations (fair values as on Policy rules)cross.priced-11

Exhibit 10 · Each policy path against the Fed's projections and market forwards

The policy-rules page's stated paths, %, quarterly: each holds its level through the stepped years, then takes the longer-run 3.2% from 2029
About

The grid takes the longer-run 3.2% from 2029 Q1; the September SEP also carries a 2029 median of 3.6%, which the grid's paths do not use — the dot for 2029 sits above the steps for that reason.

About this exhibit
A chart of the stated policy paths, % quarterly, with the SEP's year-end medians (dots), the market's 1y1y and 2y1y forwards (diamonds) and the ACM ten-year expected-rate average (dotted line).
Each policy path, quarter by quarter
The policy-rules page's stated paths, %, quarterly: each holds its level through the stepped years, then takes the longer-run 3.2% from 2029.
Dots: the SEP's year-end medians (16 Sept 2026, filled; June projection, hollow). Diamonds: the market's 1y1y (5.05%) and 2y1y (4.96%) forwards, each at the middle of its year. Dotted line: the market's expected-rate average over ten years from the ACM decomposition (4.386%, 13 Nov 2026). Each path's fair 10-year is the average of its quarterly rates plus today's term premium; the paths table prints them.
Source: U.S. Treasury; Federal Reserve Board; Macro Review calculations · data to 16 Nov 2026cross.priced-13

Exhibit 11 · Each equity asset's return if the 10-year reaches its SEP-path level

%: rates beta per +10bp, 156 weeks to 13 Nov 2026; rates return since 16 Sep 2026 and still to come at the path's 4.115%
About
Assetβ per +10bp90% bandRealisedsince the start dateStill to comeon this pathCompletionFlagActualResidualStabilityfive-factor beta
Indexes
S&P 500SPY−1.08-1.60 to -0.55−1.18%-1.75 to -0.60+10.73%5.49 to 15.98−0.12moved away−1.63%−0.45%not published
S&P 500 equal weightRSP−1.27-1.77 to -0.78−1.43%-1.98 to -0.87+13.03%7.96 to 18.09−0.12moved away+3.52%+4.95%holds (100% sign)
Nasdaq-100QQQ−1.17-1.83 to -0.50−1.25%-1.95 to -0.54+11.38%4.90 to 17.85−0.12moved away−6.00%−4.76%holds (100% sign)
Russell 2000IWM−1.97-2.54 to -1.40−2.12%-2.73 to -1.51+19.40%13.81 to 24.98−0.12moved away−3.86%−1.74%holds in sign (100% sign)
Developed ex-US (EAFE)EFA−0.87-1.44 to -0.30−0.94%-1.56 to -0.33+8.63%2.99 to 14.27−0.12moved away−2.27%−1.32%not published
Emerging marketsEEM−0.93-1.51 to -0.35−0.97%-1.57 to -0.37+8.84%3.34 to 14.33−0.12moved away−10.39%−9.43%not published
Sectors and industries
TechnologyXLK−1.32-2.12 to -0.52−1.38%-2.21 to -0.55+12.59%4.99 to 20.19−0.12moved away−10.32%−8.94%holds (100% sign)
Communication servicesXLC−0.72-1.37 to -0.07−0.77%-1.47 to -0.07+7.04%0.68 to 13.41−0.12moved away−5.16%−4.39%does not hold (89% sign)
Consumer discretionaryXLY−1.40-1.97 to -0.82−1.60%-2.25 to -0.94+14.58%8.59 to 20.56−0.12moved away+7.47%+9.07%does not hold (90% sign)
Consumer staplesXLP−0.54-0.89 to -0.19−0.63%-1.03 to -0.22+5.72%2.01 to 9.42−0.12moved away+11.86%+12.48%holds (100% sign)
Health careXLV−0.61-1.15 to -0.06−0.71%-1.35 to -0.08+6.49%0.68 to 12.29−0.12moved away+12.42%+13.13%holds (100% sign)
FinancialsXLF−1.28-1.91 to -0.65−1.44%-2.14 to -0.73+13.12%6.64 to 19.59−0.12moved away+4.37%+5.81%does not hold (58% sign)
IndustrialsXLI−1.29-1.81 to -0.76−1.40%-1.98 to -0.83+12.82%7.57 to 18.06−0.12moved away−1.54%−0.14%holds (100% sign)
EnergyXLE−0.68-1.51 to 0.16−0.74%-1.66 to 0.18+6.79%-1.61 to 15.20−0.12moved away+0.11%+0.85%holds (98% sign)
MaterialsXLB−1.25-1.82 to -0.67−1.39%-2.03 to -0.75+12.71%6.87 to 18.55−0.12moved away+2.67%+4.07%holds (100% sign)
UtilitiesXLU−1.05-1.59 to -0.50−1.19%-1.81 to -0.58+10.91%5.25 to 16.56−0.12moved away+7.56%+8.76%holds (100% sign)
Real estateXLRE−1.85-2.30 to -1.39−2.07%-2.58 to -1.56+18.93%14.26 to 23.60−0.12moved away+4.23%+6.31%holds in sign (100% sign)
SemiconductorsSMH−1.70-2.77 to -0.64−1.73%-2.81 to -0.65+15.78%5.90 to 25.65−0.12moved away−15.34%−13.61%holds (100% sign)
Regional banksKRE−2.22-3.14 to -1.29−2.43%-3.45 to -1.42+22.24%12.94 to 31.53−0.12moved away−0.19%+2.24%does not hold (68% sign)
HomebuildersXHB−2.38-2.96 to -1.80−2.68%-3.33 to -2.03+24.50%18.53 to 30.47−0.12moved away+4.93%+7.61%holds (100% sign)
Oil & gas E&PXOP−0.65-1.67 to 0.38−0.69%-1.78 to 0.40+6.29%-3.65 to 16.24−0.12moved away−5.79%−5.10%does not hold (68% sign)
TransportsIYT−1.47-2.14 to -0.80−1.66%-2.41 to -0.91+15.12%8.27 to 21.97−0.12moved away+4.77%+6.42%holds (100% sign)
REITsVNQ−1.88-2.34 to -1.41−2.10%-2.62 to -1.58+19.21%14.47 to 23.96−0.12moved away+3.84%+5.94%holds in sign (100% sign)
RetailXRT−1.43-2.07 to -0.79−1.64%-2.37 to -0.91+14.99%8.31 to 21.68−0.12moved away+8.52%+10.16%holds (100% sign)

Completion is realised ÷ (realised + still to come); β and the log-to-simple scale cancel, so the ratio and its flag are the same for every row of a column — what the beta decides is the size of the two amounts, and the 90% band under each is the width the standard error gives them. Greyed betas are those whose 90% band includes zero. The stability column is the Factors page's verdict on its five-factor rates beta (exposure beyond the market), the nearest published judgement on whether an asset's rate sensitivity holds; it is not a verdict on the four-factor beta used here.

About this exhibit
A table of the equity assets: rates beta per +10bp with its band, the rates contribution realised since the start date and still to come on the selected path, completion, flag, actual return, residual and the Factors page's stability verdict.
Priced against each path: all 24 assets
the 16 September decision (+25bp) (16 Sept 2026) against the path “One further +25bp (the SEP path)”. β is the four-factor rates beta over 156 weeks to 13 Nov 2026 with its 90% Newey–West band — a total effect: it includes what reaches the asset through the stock market's own response to yields; realised = β × +11bp ÷ 10 × the window's log-to-simple scale; still to come = the same with the −101bp still between today's 10-year and 4.115%. The SEP path's +10.73% for the S&P 500 is this page's model — model-conditional, what the model implies under stated inputs, not a forecast: a total-effect beta (−1.08% per +10bp from 156 weekly returns since 24 Nov 2023, regressed on oil, the ten-year, HYG − LQD and the dollar with no market factor, so the index's own co-movement with yields in that window is inside it) times the −100bp still to come, with the other three factors held and no horizon. The Scenarios page prices the same paths two other ways — the valuation identity, with the premium and growth held, and the factor engine's draws conditioned on the path — and sets the three models side by side; they are not expected to match, and that page says why.
Source: Betas computed here on Yahoo Finance adjusted closes (licensed), the Treasury par curve and the Board's H.10 (public); the paths' fair values are copied from the policy-rules page and its ACM decomposition.cross.priced-10

What the market prices, reading by reading

The 10-year, the fair value of each policy path, the forwards and the Committee's projections the paths are measured against

Exhibit 12 · Market rates and Fed projections the paths are measured against

Readings copied from What is priced, %, each with its date (13 Nov 2026 to 16 Nov 2026; SEP medians 16 Sep 2026)
ReadingValueDate
10-year par yield5.12%16 Nov 2026
Fair value of a hold4.065% · +106bp below today13 Nov 2026
Fair value of the SEP path4.115% · +101bp below today13 Nov 2026
Fair value of +50bp, then hold4.165% · +96bp below today13 Nov 2026
The market's own path (ACM-fitted 10-year)5.111%16 Nov 2026
1y1y forward − IORB+115bp · 5.05% vs 3.90%16 Nov 2026
SEP medians: 2026 · 2027 · longer run4.1 · 4.1 · 3.216 Sept 2026
Completion of the SEP path since the September decision-0.12 · moved away (+11bp of −90bp)16 Sept 2026
S&P 500 still to come on the SEP path, that page's beta+10.73% (hold +11.27%) · β −1.08 per +10bp, 156 weeks, no horizon16 Nov 2026
About this exhibit
Table of the market readings with dates: the 10-year par yield, the fair value of each policy path, the model-fitted 10-year, 1y1y less IORB, the SEP medians, the SEP path's completion and the S&P 500 still to come
What the market prices, beside the set
Copied from What is priced by its published keys and re-compared at every update: the observed path the branches are measured against. The three Fed branches pin the ten-year at the distance to each fair value below; the term-premium page gives the adverse one.
Source: U.S. Treasury; Federal Reserve Board (FOMC projections); New York Fed; Macro Review calculations (as on What is priced)cross.projections-10

Each family's range, the weights, the sources' biases and the shorter band

The three families' ranges at 3 and 12 months, the weights the record would give, each source's estimated bias, and the 3-month band while its record fails

Exhibit 13 · Projected ranges for the 10-year yield and its parts, by model

%: 10th · median · 90th percentile at Feb 2027 and Nov 2027 from Nov 2026; finer parts from the statistical and factor models only
Range: simulated shocks
FamilyPartFeb 2027Nov 2027
Published pool10-year4.31% · 4.82% · 5.35%3.71% · 4.69% · 5.66%
Expected short-rate average3.82% · 4.16% · 4.48%3.39% · 4.10% · 4.75%
Term premium0.28% · 0.67% · 1.08%-0.08% · 0.62% · 1.28%
Statistical (the model's own paths)10-year4.49% · 4.83% · 5.19%4.05% · 4.69% · 5.31%
Expected short-rate average3.89% · 4.15% · 4.41%3.63% · 4.07% · 4.52%
Term premium0.39% · 0.68% · 0.96%0.16% · 0.62% · 1.07%
Factor (direct projection)10-year4.31% · 4.85% · 5.41%4.01% · 4.90% · 5.66%
Expected short-rate average3.89% · 4.20% · 4.49%3.54% · 4.22% · 4.77%
Term premium0.40% · 0.65% · 0.96%0.27% · 0.68% · 0.99%
Event calendar10-year4.27% · 4.82% · 5.40%3.63% · 4.64% · 5.70%
Expected short-rate average3.79% · 4.16% · 4.50%3.29% · 4.08% · 4.80%
Term premium0.22% · 0.67% · 1.15%-0.20% · 0.58% · 1.42%
Pool, equal weights10-year4.36% · 4.83% · 5.32%3.87% · 4.75% · 5.55%
Expected short-rate average3.85% · 4.16% · 4.47%3.49% · 4.11% · 4.70%
Term premium0.35% · 0.66% · 1.01%0.09% · 0.64% · 1.13%
Statistical and factor poolExpected real short-rate average1.28% · 1.52% · 1.73%1.05% · 1.48% · 1.89%
Expected inflation, 10 years2.54% · 2.65% · 2.77%2.43% · 2.64% · 2.85%
r*1.02% · 1.26% · 1.44%0.89% · 1.27% · 1.59%
π*2.46% · 2.55% · 2.65%2.41% · 2.57% · 2.74%
TIPS liquidity premium0.30% · 0.56% · 0.83%0.06% · 0.54% · 0.93%
10-year TIPS yield2.17% · 2.59% · 2.99%1.65% · 2.47% · 3.15%
About this exhibit
Table of the 10th, median and 90th percentiles at three and twelve months for the 10-year, the expected short-rate average and the term premium, under the published combination, each model and equal weights; finer parts from the statistical and factor models
Each family's range, and the parts' ranges
10th · median · 90th, percent, at 3 and 12 months from Nov 2026; the parts from the pool of the statistical and factor families (the event family has no finer split). Model splits: no outcome scores them.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-8

Exhibit 14 · Weight given to each model in the combined 10-year range

Weights (ratios) from the models' scores on 72 forecasts 12 months ahead, to Oct 2025, pulled halfway toward equal, beside equal weights; not odds
FamilyRecordEqual
Statistical (the model's own paths)0.170.33
Factor (direct projection)0.170.33
Event calendar0.670.33
About this exhibit
Table of each model's weight in the combination, the record's weight beside equal weights
The families' weights
Today's weights on 72 scored 12-month origins: the log score's, shrunk halfway to equal; model weights, not odds.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-9

Exhibit 15 · Estimated bias of each published measure the model reads, with its error

Percentage points, ± standard error: the constant the model removes from each source before combining (survey anchor); anchor measures at zero; data to 6 Nov 2026
MeasureBias treated asBias± s.e.
1y real rate (GSW 1y − Cleveland 1y)free−0.660.41
Cleveland 1y expected inflationfree−0.190.24
HLW r*free+0.850.20
LW r*free+0.990.20
Lubik–Matthes r*free+0.620.19
SEP longer-run funds rate − 2free+0.500.21
DKW expected real short rate, 10yfree−0.370.07
SPF BILL10 − CPI10anchor (set to zero)±0.000.00
DKW expected inflation, 10yfree+0.150.02
Cleveland 10y expected inflationfree−0.250.02
SPF CPI10anchor (set to zero)±0.000.00
DKW TIPS liquidity premium (anchor)anchor (set to zero)±0.000.00
Kim–Wright term premiumfree+0.130.09
ACM term premiumfree+0.310.09
SF Fed (CR) term premium, observedfree+0.380.09
About this exhibit
Table of the estimated bias (pp, ± standard error) of each published measure the model reads, from real rates and expected inflation to r* models and term premia; anchor measures set to zero
What each source reads against the model
The estimated bias of each published measure, pp, with its standard error: a constant the model takes off that source before it pools it (survey anchor).
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-10

Exhibit 16 · 10-year yield outcomes against the band drawn a quarter earlier

%, quarterly origins since Jan 2008; black: outcome inside the 10th–90th band drawn 3 months earlier, red: outside; held 69% of 75, Kupiec p 0.03
Range: out-of-sample recordAbout
About this exhibit
Chart of the 10-year outcome at each quarterly origin since 2008, against the 10th–90th band drawn three months earlier; dots are black inside and red outside
The 3-month band at each quarter, against the outcome
Percent; dot black inside the 10th–90th drawn 3 months earlier, red outside. Held 69% of 75; Kupiec p 0.03, independence p 0.02.
Each band is the published pool as the rule would have weighted it then, drawn from the model filtered on the data published by the origin's month-end. The parameters are this month's, fitted on the whole sample: the data are real-time, the parameters are not. 12-month windows overlap, so the independence test runs on the non-overlapping annual sequences.
Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-11

Appendix

How the model is specified, how the families are drawn and pooled, what the record scores, and every source behind the page291 checks, all pass
About this page

One measurement model reads seventeen published measures of the 10-year's parts — the staff curves, four estimates of the neutral real rate, surveyed and market inflation expectations, three term-premium models — and filters them into one estimate of each part with its error. Three models draw the next twelve months from that state: the model's own dynamics, a projection on seven fundamentals and market factors, and the calendar of Fed decisions and Treasury refundings. Their paths are pooled by their record, and the record — every model re-drawn at each quarter since 2008 on the data then published — decides which bands stand on the page. Everything forward here is model-conditional: what the models imply if they are right, not a forecast.

The measurement model: seven parts, seventeen measures, two exact identities

yt(10)=rt∗+κggt+πt∗+κqqt+rtpt+irpt,ytR(10)=rt∗+κggt+rtpt+lpty^{(10)}_t = r^*_t + \kappa_g g_t + \pi^*_t + \kappa_q q_t + rtp_t + irp_t, \qquad y^{R(10)}_t = r^*_t + \kappa_g g_t + rtp_t + lp_t
rt∗=rt−1∗+ηtr,πt∗=πt−1∗+ηtπ,xt=μx+ϕx(xt−1−μx)+ηtx    (x=g,q,rtp,irp,lp)r^*_t = r^*_{t-1} + \eta^{r}_t,\quad \pi^*_t = \pi^*_{t-1} + \eta^{\pi}_t,\quad x_t = \mu_x + \phi_x (x_{t-1} - \mu_x) + \eta^x_t \;\; (x = g, q, rtp, irp, lp)
mj,t=Zjst+bj+εj,t,κ(ϕ)=1−ϕ120120 (1−ϕ)m_{j,t} = Z_j s_t + b_j + \varepsilon_{j,t}, \qquad \kappa(\phi) = \frac{1-\phi^{120}}{120\,(1-\phi)}
SymbolMeaningFromAssumed or estimated
y(10)y^{(10)}, yR(10)y^{R(10)}the nominal and TIPS 10-year zero-coupon yieldsBoard staff curves (feds200628, feds200805)–
mj,tm_{j,t}, bjb_ja published measure and its estimated biasthe seventeen sources; biases estimated–
κg\kappa_g, κq\kappa_qthe ten-year average of an AR(1) gap's pathfrom the estimated φ–

Maximum likelihood with an exact diffuse start; filtered (one-sided) states are printed, never smoothed ones, except where a chart says so.

Three families, one pool, one record

p(yT+h)=∑fwf pf(yT+h),wf∝exp⁡(∑olog⁡pf(yo)) shrunk halfway to 1/3p(y_{T+h}) = \textstyle\sum_f w_f\, p_f(y_{T+h}), \qquad w_f \propto \exp\Big(\sum_{o} \log p_f(y_o)\Big)\ \text{shrunk halfway to } 1/3
CRPS=E∣X−y∣−12E∣X−X′∣,LRuc∼χ12, LRind∼χ12\text{CRPS} = \mathbb E|X - y| - \tfrac12 \mathbb E|X - X'|, \qquad LR_{uc} \sim \chi^2_1,\ LR_{ind} \sim \chi^2_1
SymbolMeaningFromAssumed or estimated
pfp_fa family's distribution of whole pathssimulated–
wfw_fthe family's weightthe record's log scores, with outcomes seen by thenestimated

Model-conditional, not a forecast; the record re-draws each family on the data published at each origin.

Exhibit 17 · Specification

How the numbers are made, step by step

  1. The measures. Source: the Board, five regional Feds and the SPF, each at its publication date. Transform: seventeen monthly measures with their publication dates. Check: the two new sources' own checks (2 checks, all passing).
  2. The measurement model. Source: step 1. Transform: three normalisations, estimated monthly, filtered to today. Check: the identity rows hold; an independent filter reproduces the likelihood (4 checks, all passing).
  3. The families and the pool. Source: step 2 with the factors and the event calendar. Transform: the model's paths, the factor projection, the event paths; the pool of whole paths and the draw files. Check: every path meets the identities; the pools use only days seen; quantiles reproduce from the files (5 checks, all passing).
  4. The record and the placement. Source: steps 1–3 at each quarter since 2008. Transform: coverage, Kupiec, independence, CRPS, log score; the placement. Check: the summary recomputes; the placement follows the tests; no late data (3 checks, all passing).
  5. The gap. Source: step 2. Transform: a local level plus AR(1) on the 10-year less i* under each anchor. Check: the closure formula (2 checks, all passing).
  • Specification
    Model-conditional; seventeen measures, seven parts.

    The fans on this page are model-conditional distributions, not forecasts: they are what three fitted models imply from today's filtered state if the models are right. The measurement model (an unobserved-components model in state-space form, core.ucm_rates) reads seventeen published measures of the 10-year's parts — the Board staff nominal and TIPS curves (exact identity rows), the 1-year real rate, Cleveland's 1- and 10-year expected inflation, four r* models (HLW, LW, Lubik–Matthes, the SEP's longer-run rate less 2), DKW's expected real rate, expected inflation and liquidity premium, the SPF's ten-year bill rate and CPI, and three affine term premia (Kim–Wright, ACM, the SF Fed's Christensen–Rudebusch) — and filters them into r*, a real-rate gap, π*, an inflation gap, the real term premium, the inflation risk premium and the TIPS liquidity premium, with a shared error for the affine models and a bias for each source.

  • Identity
    Every path adds up.

    y = E[ī] + TP, with E[ī] = r* + κ_g·g + π* + κ_q·q and κ(φ) = (1 − φ^120)/(120(1 − φ)) the ten-year average of an AR(1) gap: today κ_g = 0.299 and κ_q = 0.199. The TIPS yield is r* + κ_g·g + rtp + lp. Every path of every family meets these to machine precision.

  • Anchors
    Three normalisations, the survey anchor the headline.

    The identities fix the sum, not the split of the levels: three normalisations are fitted (the real-rate level as the average of its six sources; the premium level as the average of the three affine models; the survey anchor) and printed as the range, the survey anchor the headline. Log-likelihoods real-group 463.6, premium-group 464.5, survey 466.3.

  • Regularisation
    A noise floor and trend caps.

    Every non-exact measure's noise has a floor of 5bp; r*'s monthly volatility is capped at 0.04pp and π*'s at two thirds of it. Half-lives (months): g 25, q 16, rtp 19, irp 4, lp 11. Parameters estimated on 1999-01 to 2026-10 and filtered to Nov 2026.

  • Families
    The model's paths, the factor projection, the event calendar.

    Statistical: the model's own simulation from the filtered state and its covariance. Factor: for each horizon and part, the model's forecast error regressed by ridge (penalty equal to the sample size, standardised factors, no intercept) on seven factors as a reader had them — the policy gap, core PCE inflation's gap to π*, unemployment's six-month change, oil's twelve-month change, the stock–bond correlation, the 10-year's realised volatility and the Bund spread — with a path adding one historical month's joint block of errors. Event: the model's mean path plus empirical shocks on the calendar a reader could see — scheduled FOMC decisions and refunding auction days from their own pools, ordinary days in monthly blocks, centred and scaled by trailing volatility (filtered historical simulation); ACM's risk-neutral change is permanent, its premium change decays at the model's persistence.

  • Pool
    Whole paths, the record's weights.

    The pool mixes whole paths, so every pooled path keeps its identities. Weights: equal until 36 scored origins, then the log score's, shrunk halfway to equal; today S 0.17, F 0.17, E 0.67. The draw files hold 6,000 pooled paths (seed 202611) with their sha256 in this page's data.

  • Record
    Quarterly since 2008, on the data then published.

    At each quarter from Jan 2008, every family is re-drawn from the model filtered on the data published by that month's end and scored at 3 and 12 months against the staff curve's outcome: coverage of the 80% band, Kupiec's test on every origin, Christoffersen's independence test on the non-overlapping subsequences (the median p), the CRPS and the Gaussian log score. A band leaves its fold when both p-values exceed 0.05. Limits: the parameters are this month's (fitted on the whole sample) — the data are real-time, the parameters are not; the macro drivers are at their latest release; 12-month windows overlap.

  • Cadence
    Estimated monthly, filtered daily.

    The parameters are re-estimated once a month; each day's update filters to the latest published day with them.

Source: Board of Governors of the Federal Reserve System; Federal Reserve Banks of New York, San Francisco, Richmond, Cleveland and Philadelphia; BEA; BLS; EIA; U.S. Treasury; Deutsche Bundesbank; Yahoo Finance; Macro Review calculationscross.discount_rate-1

Before you read

Macro Review is a personal research site, still being built.

Nothing on it is investment advice, a recommendation or a solicitation.

Check every figure for yourself. Each names its publisher — confirm it there before you rely on it.

Figures can be revised, delayed or wrong.

Cones, fans, scenarios, fair values and “what followed” tables are conditional on the model each page names: they describe, they do not predict.

No warranty is given; use is at your own risk.