Macro ReviewHow to read
Macro · Policy & rates · Inside the 10-year

Most of the 10-year's 71bp rise in a year is real, not inflation

What is the 10-year Treasury yield made of — expected policy, expected inflation, or compensation for risk?

Data to 6 Nov 2026 · updated 16 Nov 2026ACM to 13 Nov 2026
Real parts of the year's move+58bpto 6 Nov 2026 · of the zero-coupon 10-year's +71bp · inflation +13bp
10-year zero-coupon yield4.88%6 Nov 2026 · 1y +71bp · par 5.12% on 16 Nov, 1y +98bp
, four models' median0.89%0.64%four models' term premium1.75%6 Nov 2026 · band 0.64–1.75% · survey-anchored 1.75%, ACM 0.64%
, 10-year average3.12%3.12%four models' expected path4.22%6 Nov 2026 · survey-anchored; other models 4.01–4.22%
Expected inflation, 10 years ()2.30%6 Nov 2026 · 2.82%

In brief

What moved the 10-year and what it is made of, in a few lines; what would change the read beneath
  • Since the 28 Oct 2026 decision the 10-year fell 10bp; the models' median fell 6bp.→ Exhibit 2
  • Of the 10-year's +71bp in the year to 6 Nov 2026, +58bp came from the real parts, not .→ Exhibit 1
  • Four models put the 10-year term premium between 0.64% and 1.75% on 6 Nov 2026.→ Exhibit 6
  • The survey-anchored model puts the at 3.12% over ten years, against 4.22% on the ACM model.→ Exhibit 7
  • All four models put the term premium on 6 Nov 2026 above 62% of days since 2004.→ Exhibit 4
What would change this view
What would change this viewTriggerTodayDecided by
Inflation compensation overtakes the real parts of the year's moveinflation compensation (expected inflation plus its risk premium) explains more of the year's move, in bp, than the real partsreal +58bp, inflation +13bp (6 Nov 2026)not crossedeach weekly update of the Board staff curves
The forecasters' path catches up with the models'forecasters' expected ten-year average bill rate reaches 4.01%, the lowest model path3.00% (2026 Q1 survey)not crossedthe 2027 Q1 Survey of Professional Forecasters (about 15 Feb 2027)
The year's rise in the premium is undonethe term premium (survey-anchored) falls below 1.16%, its level a year earlier1.75% (6 Nov 2026)not crossedeach weekly update of the Board staff curves
The reasoning behind each point
  • Since the decision
    Since the 28 Oct 2026 decision the zero-coupon 10-year fell 10bp; the models' median term premium fell 6bp.

    From the close of 27 Oct 2026, the eve of the decision, to 6 Nov 2026: real parts +7bp, inflation compensation −17bp (the breakeven −9bp). Change in the term premium: Survey-anchored −6bp, Kim–Wright −6bp, ACM +3bp (DKW: the DKW file runs to 22 Oct 2026, before the decision; it adds each month about two weeks after the month ends); the expected path took −12bp to −3bp. Each model splits its own fitted 10-year, whose move over these days differs from the Board's curve by at most 1.3bp.

  • The year
    Of the 10-year's +71bp in the year to 6 Nov 2026, +58bp came from the real parts, not inflation compensation.

    The zero-coupon yield from the Board staff curve, to 6 Nov 2026; the par yield most pages print was 5.12% on 16 Nov 2026, +98bp in a year. From 6 Nov 2025: expected real rate +12bp, real term premium +46bp, expected inflation −1bp, inflation risk premium +14bp. The Board's DKW model, to 22 Oct 2026, agrees: real +67bp, inflation +26bp.

  • The band
    Four models put the 10-year term premium between 0.64% and 1.75% on 6 Nov 2026.

    Survey-anchored 1.75%, DKW 0.90% (22 Oct 2026), Kim–Wright 0.87%, ACM 0.64% (ACM 0.72% on 13 Nov 2026). The premium is what is left after the expected path, so the band is the disagreement about that path.

  • The path
    The survey-anchored model puts the expected policy path at 3.12% over ten years, against 4.22% on the ACM model.

    On 6 Nov 2026: DKW 4.07% (22 Oct 2026), Kim–Wright 4.01%. The survey path is the forecasters' own ten-year real bill rate (15 Feb 2026: 3.00% nominal) moved by κ = 0.246 of the change in the one-year real rate since; their BOND10 (4.00%) implies a premium of 1.00pp.

  • History
    All four models put the term premium on 6 Nov 2026 above 62% of days since 2004.

    Kim–Wright's 0.87% (88th percentile), DKW's 0.90% (85th), ACM's 0.64% (62nd), the survey-anchored 1.75% (97th; anchored on the latest first-quarter survey and the assumed κ). Over three months the 10-year moved +11bp: real parts +11bp, inflation compensation ±0bp.

  • Record
    Since 2020, 3 of 8 premium readings beat a no-change forecast of the 10-year; before 2020, only one did.

    Out of sample, 12 months ahead, expanding window from Jan 2004. Best since 2020: DKW term premium (R²_OS 0.14, Clark–West t 2.2). One tightening cycle and a few dozen independent years' worth of months; the page prints no forward read.

Next: Policy rules — where the textbook rules put the rate

Board of Governors, Federal Reserve Bank of New York, Philadelphia Fed and Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-read

Since the 28 Oct decision the median premium fell 6bp; over the year the move was mostly real

The move since the last FOMC decision split into the expected path and the on four models; the 10-year split each month into four parts, and the year's move split the same way

Exhibit 2 · The 10-year's change since the latest Fed decision, split by model

Each model's 10-year change since the decision, bp, 27 Oct 2026 (its eve) to 6 Nov 2026; bars: path and premium; diamond: the total
About
About this exhibit
Stacked bars per model (survey-anchored, DKW, Kim–Wright, ACM): change in the model's own 10-year split into expected path and term premium, bp, from the decision's eve to the latest day; diamond = the model's 10-year change.
The move since the 28 Oct 2026 decision, model by model
bp · from the close of 27 Oct 2026, the eve of the decision, to 6 Nov 2026 · bars add to the diamond
The Board's zero-coupon 10-year moved −10bp: real parts +7bp, inflation compensation −17bp. DKW: the DKW file runs to 22 Oct 2026, before the decision; it adds each month about two weeks after the month ends.
Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-11

Exhibit 3 · Each model's 10-year change since the decision, split into path and premium

bp · since the decision, 27 Oct 2026 (its eve) to 6 Nov 2026; DKW over its own days · each model splits its own fitted 10-year
ModelIts 10-year, bpExpected path, bpTerm premium, bpBoard's curve, bp
Survey-anchored−10−3−6−10
DKW (Board)the DKW file runs to 22 Oct 2026, before the decision; it adds each month about two weeks after the month ends––––
Kim–Wright (Board)−11−5−6−10
ACM (New York Fed)−9−12+3−10

Median change in the term premium across the 3 models −6bp (range −6 to +3bp); in the expected path −5bp. The breakeven moved −9bp.

About this exhibit
Table per model (survey-anchored, DKW, Kim–Wright, ACM): change in its own 10-year, expected path, term premium and the Board's curve, bp, from the decision's eve to the latest day (the numbers behind the chart beside it).
Since the 28 Oct 2026 decision: path and premium on each model
bp · 27 Oct 2026 to 6 Nov 2026 (DKW over its own days) · each model splits its own fitted 10-year
Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-12

Exhibit 4 · The 10-year zero-coupon yield split into expectations and premia, month-end since 2004

%, month-end, 2004 to Nov 2026 (default: last ten years); expected real short rate and expected inflation from the forecasters' survey, the premia from yields.
About
About this exhibit
Month-end stacked bars of the 10-year zero-coupon yield in four parts (expected real short rate and expected inflation from the forecasters' survey, real term premium and inflation risk premium from yields), the yield as a line, since 2004 (default: the last ten years).
The 10-year in four parts, month-end
% · month-end, 2004 to date · bars add to the line (the Board's 10-year zero-coupon yield)
The expected real rate and expected inflation come from the forecasters' survey; the two premia are what the yields leave. Four models on 6 Nov 2026: Survey-anchored 1.75%, DKW 0.90% (22 Oct 2026), Kim–Wright 0.87%, ACM 0.64%; ACM 0.72% on 13 Nov 2026.
Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-1

Exhibit 1 · The 10-year's change split into its parts, by horizon and estimate

Change in the Board's 10-year zero-coupon yield, bp: one month, three months and a year to 6 Nov 2026; the site's split beside DKW's.
About this exhibit
Stacked bars: change in the Board's 10-year zero-coupon yield split into expected real short rate, expected inflation, real term premium and inflation risk premium, over one month, three months and a year to the latest day; the site's split beside the Board's DKW model; diamond = total change.
The move split four ways, this page and DKW
bp · changes to 6 Nov 2026 (DKW to its own last date) · bars add to the diamond
Source: Board of Governors (staff curves, DKW, Kim–Wright); Philadelphia Fed (SPF); Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-2

The premium is 0.64–1.75%, depending on which expected path you believe

Five estimates of the 10-year , and the expected path of short rates each one implies, with the forecasters' own

Exhibit 5 · Term premium and expected policy path across the models, with their range

% · on 6 Nov 2026 (DKW on its own last date) · the band is the lowest to the highest of the models
About
Ten-year averageSurvey-anchoredDKW (Board)Kim–Wright (Board)ACM (New York Fed)Band
1.75%0.90%0.87%0.64%0.64–1.75%
3.12%4.07%4.01%4.22%3.12–4.22%

At κ 0.044 rather than 0.246 — the survey anchor's own parameter range — the survey-anchored real term premium is 1.35–1.45% and the expected real rate 0.72–0.82%. ACM 0.72% on 13 Nov 2026.

The one-year rate one year ahead on the expected path: ACM 4.61% (13 Nov 2026), Kim–Wright 4.11% (6 Nov 2026); median 4.23% (6 Nov 2026). The Treasury 1y1y on the Board's curve, premium included, 4.56% (6 Nov 2026).

About this exhibit
Table: 10-year term premium and expected policy path (ten-year averages) on four models (survey-anchored, DKW, Kim–Wright, ACM) on the latest common day, with the lowest-to-highest range.
The band: premium and path across four models
% · on 6 Nov 2026 (DKW on its own last date) · the band is the lowest to the highest of the four
Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-10

Exhibit 6 · The 10-year term premium on each model and the forecasters' own, monthly

%, month-end, 2004 to Nov 2026; term premium = the 10-year less the expected path of short rates; dashed: forecasters' ten-year yield forecast less their bill-rate forecast
About
About this exhibit
Month-end 10-year term premium on four models (survey-anchored, DKW, Kim–Wright, ACM) plus the forecasters' own premium (dashed), since 2004, recessions shaded.
Five estimates of the 10-year term premium
% · month-end, 2004 to date · = the 10-year less the expected path of short rates
Four models on 6 Nov 2026: Survey-anchored 1.75%, DKW 0.90% (22 Oct 2026), Kim–Wright 0.87%, ACM 0.64%; ACM 0.72% on 13 Nov 2026. The forecasters' own premium is their ten-year yield forecast less their ten-year bill-rate forecast, asked once a year. At κ 0.044 rather than 0.246 the survey-anchored real term premium is 1.35–1.45%.
Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; NBER (recessions); Macro Review calculations · data to 6 Nov 2026rates.decomposition-3

Exhibit 7 · The ten-year average short rate expected by the models and forecasters

%, month-end, 2004 to Nov 2026; average short rate expected over the next ten years; dashed: the forecasters' ten-year bill-rate forecast, which steps each first-quarter survey
About
About this exhibit
Month-end average short rate expected over ten years on four models (survey-anchored, DKW, Kim–Wright, ACM) plus the forecasters' own ten-year bill-rate forecast (dashed, steps each first-quarter survey), since 2004.
The expected policy path over ten years, four ways
% · month-end · the average short rate expected over ten years; the forecasters' own step changes each first-quarter survey
On 6 Nov 2026: Survey-anchored 3.12%, DKW 4.07% (22 Oct 2026), Kim–Wright 4.01%, ACM 4.22%.
Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; NBER (recessions); Macro Review calculations · data to 6 Nov 2026rates.decomposition-4

The is the extra yield for holding a 10-year bond rather than rolling short bills. Nobody observes it: it is the 10-year yield less the of short rates, and each model expects a different path — surveys a lower one, the market-fitted models (, Kim–Wright, ACM) a higher one. So the band is wide because the expected path is uncertain, not because the yield is. On this page the premium is split in two: the and the .

Only the stock–bond return correlation explains the real premium in every sample

Monthly regressions of the and the inflation risk premium on market conditions, with their stability across samples

Exhibit 8 · Market conditions against the real term and inflation risk premia, since 2007

Month-end regressions, Jan 2007 to Nov 2026: t-statistic and the premium's change with a typical move, bp; whether the sign holds in every sample
Real term premium: R² 0.60, 239 months; 2026-11 +64bp from the fit (typical miss 42bp)
Real term premium ont-statisticTypical move, bpSame sign in every sample
Realised 10-year yield volatility (bp a year, 63 days)−0.8−7yes
Stock–bond return correlation (63 days)+9.1+44yes, significant
Coupon share of marketable debt (%)−0.4−3no — do not use
10-year Bund yield (%)+1.7+29yes
10-year JGB yield (%)−0.4−5yes
Inflation risk premium: R² 0.14, 239 months; 2026-11 +29bp from the fit (typical miss 28bp)
Inflation risk premium ont-statisticTypical move, bpSame sign in every sample
Cleveland Fed 1-year − 10-year expected inflation (pp)−1.7−8no — do not use
Stock–bond return correlation (63 days)+1.9+8yes, significant
VIX−1.7−6no — do not use
Realised 10-year yield volatility (bp a year, 63 days)+1.3+6no — do not use

MOVE is replaced by realised 10-year yield volatility from the Board staff curve (the ICE BofA MOVE index is not among the site's sources); the S&P 500 by SPY (SPDR S&P 500 ETF Trust) for the equity leg of the stock–bond correlation.

About this exhibit
Two tables of month-end regressions since 2007: the real term premium on five market conditions and the inflation risk premium on four; t-statistic, change with a typical (one standard deviation) move, bp, and whether the sign holds in every sample.
What has gone with the two premia since 2007
Month-end regressions, 2007-01 to 2026-11, Newey–West errors (12 lags) · effect of a typical (one standard deviation) move, bp
Source: Board of Governors (staff curves, DKW, Kim–Wright); Cleveland Fed; Treasury (MSPD); Deutsche Bundesbank; Ministry of Finance Japan; Cboe (VIX); Yahoo Finance (SPY); Macro Review calculations · data to 6 Nov 2026rates.decomposition-5

Exhibit 9 · Each market condition's link to the premia, by sample period

t-statistics, a ratio, in each sample from 2007 to 2026; a driver whose sign flips describes one period, not a relation.
Real term premium on2007–2015108 months2016–2026131 months2020–202683 months
Realised 10-year yield volatility (bp a year, 63 days)−2.3−0.3−1.9
Stock–bond return correlation (63 days)+3.5+3.5+3.8
Coupon share of marketable debt (%)−1.6−1.8+0.7
10-year Bund yield (%)+1.9+6.6+7.7
10-year JGB yield (%)−1.9−3.5−3.4
Inflation risk premium on2007–2015108 months2016–2026131 months2020–202683 months
Cleveland Fed 1-year − 10-year expected inflation (pp)+1.2−2.5−3.0
Stock–bond return correlation (63 days)+2.4+4.6+2.1
VIX+0.6−4.2−4.1
Realised 10-year yield volatility (bp a year, 63 days)+0.9−0.3−0.3
About this exhibit
Table: t-statistics for each market condition on the real term premium and the inflation risk premium in each sample period (2007–2015, 2016 on, 2020 on).
The same regressions in three samples
t-statistics by sample, Newey–West (12 lags) · a driver whose sign flips is a description of one period, not a relation
Source: Board of Governors (staff curves, DKW, Kim–Wright); Cleveland Fed; Treasury (MSPD); Deutsche Bundesbank; Ministry of Finance Japan; Cboe (VIX); Yahoo Finance (SPY); Macro Review calculations · data to 6 Nov 2026rates.decomposition-6

How far this split agrees with the published models

Each component against the Board's model and the premium and expected path against and ACM, levels and changes

Exhibit 10 · The site's split against the published models: correlations and average gaps

Monthly since 2004: mean difference (site less model), bp, correlation of levels and of monthly changes, and values on 6 Nov 2026
This page againstCorrelation, levelsCorrelation, changesMean difference, bpToday, this pageToday, the model
E[r̄] vs DKW0.670.08+280.82%1.25%
E[π̄] vs DKW0.48-0.05−102.30%2.82%
RTP vs DKW0.490.38−291.35%0.79%
IRP vs DKW0.720.25+90.41%0.11%
E[π̄] vs Cleveland Fed0.500.10+352.30%2.74%
IRP vs Cleveland Fed0.190.04−270.41%0.47%
Term premium vs Kim–Wright0.710.82−121.75%0.87%
Term premium vs ACM0.390.72−221.75%0.64%
Term premium vs DKW0.660.48−201.75%0.90%
Expected path vs Kim–Wright0.720.06+113.12%4.01%
Expected path vs ACM0.460.13+233.12%4.22%
Expected path vs DKW0.740.03+193.12%4.07%
About this exhibit
Table: each part of the site's split (expected real rate, expected inflation, real term premium, inflation risk premium, term premium, expected path) against DKW, the Cleveland Fed, Kim–Wright and ACM: correlation of levels and of monthly changes, mean difference, bp, and the latest values.
This split against the published models, in levels and monthly changes
Monthly since 2004 · correlation of levels and of monthly changes, mean difference (this page − the model), bp · today, 6 Nov 2026
Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-7

As a guide to the next year, 1 of 8 readings worked before 2020 and 3 after

The record of each premium reading as a predictor of the 12-month change in the 10-year, against a no-change forecast

Exhibit 11 · How premium measures forecast the 10-year's change a year ahead

Out-of-sample R², a ratio: each measure's forecast of the twelve-month change in the 10-year, refitted from Jan 2004, scored from Jan 2013 against no change.
About
ReadingR²_OS 2013–26R²_OS 2013–19R²_OS 2020–26Clark–West t, 2020–26Right sign, 2020–26, %Top − bottom quartile, bp
DKW term premium−0.07−0.35+0.14+2.273−117
Kim–Wright term premium−0.07−0.28+0.09+1.766−109
ACM term premium−0.16−0.50+0.10+2.361−86
Cycle gap: 10-year − (r* + expected inflation)−0.21−0.06−0.32+1.561−97
Survey-anchored term premium (RTP + IRP)−0.20−0.01−0.34−0.834−48
Gap to the drivers' fitted value (real-time)−0.39+0.12−0.69−1.435−48
Policy gap: 1-year real rate − r*−0.33−0.10−0.50−3.120−27
Curve slope: 10-year − 1-year zero−0.17−0.23−0.13−0.841−34

In sample (Hodrick errors) every slope is negative — a high premium has come before a smaller rise — but out of sample only the last cycle rewards it. This page makes no forward read.

About this exhibit
Table: out-of-sample record of the premium and gap measures as forecasts of the twelve-month change in the 10-year: R²_OS over the full, pre-2020 and post-2020 samples, Clark–West t, share with the right sign, and top-less-bottom-quartile spread.
Premium readings as a guide to the next year's change
Out of sample: 12-month change in the 10-year, regression refitted monthly from 2004-01, scored from 2013-01 against no change · R²_OS above 0 beats no change
Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-8

Appendix

The identities, the survey anchor and κ, the as-published calendar, the drivers and the record — and every check74 checks, all pass
About this page

A 10-year yield is what investors expect short rates and inflation to average over ten years, plus two premia for bearing risk: the real term premium, for tying money up at a fixed real rate, and the inflation risk premium, for the chance that inflation surprises. None of the four is observed. This page takes expected inflation and the expected real rate from the forecasters' survey, reads the two premia off the nominal and inflation-protected yields, and sets the result beside three Federal Reserve models. The premium is what is left after the expected path, so the models disagree about the premium exactly as much as they disagree about the path — that is why the band is wide.

The four parts of the 10-year

yt(10)=Et[rˉ]+Et[πˉ]+RTPt+IRPty^{(10)}_t = \mathbb{E}_t[\bar r] + \mathbb{E}_t[\bar\pi] + \mathrm{RTP}_t + \mathrm{IRP}_t
yt∗(10)=Et[rˉ]+RTPt+LPty^{*(10)}_t = \mathbb{E}_t[\bar r] + \mathrm{RTP}_t + \mathrm{LP}_t
bt=yt(10)−yt∗(10)=Et[πˉ]+IRPt−LPtb_t = y^{(10)}_t - y^{*(10)}_t = \mathbb{E}_t[\bar\pi] + \mathrm{IRP}_t - \mathrm{LP}_t
SymbolMeaningFromAssumed or estimated
y(10)y^{(10)}10-year nominal zero-coupon yield, continuously compoundedBoard staff curve (SVENY10)–
y∗(10)y^{*(10)}10-year TIPS zero-coupon yieldBoard staff TIPS curve (TIPSY10)–
E[πˉ]\mathbb{E}[\bar\pi]average CPI inflation expected over ten yearsPhiladelphia Fed SPF, CPI10, as published–
LP\mathrm{LP}TIPS liquidity premiumBoard DKW model, as published–
RTP, IRP\mathrm{RTP},\ \mathrm{IRP}real term premium and inflation risk premium — the identities' residuals––

The expected real rate, anchored on the survey

Et[rˉ]=(BILL10−CPI10)v+κ (rt(1)−rv(1))\mathbb{E}_t[\bar r] = (\mathrm{BILL10} - \mathrm{CPI10})_v + \kappa\,(r^{(1)}_t - r^{(1)}_v)
rt(1)=yt(1)−πte,(1),κ=1−ρ4040 (1−ρ)r^{(1)}_t = y^{(1)}_t - \pi^{e,(1)}_t,\qquad \kappa = \frac{1-\rho^{40}}{40\,(1-\rho)}
SymbolMeaningFromAssumed or estimated
vvthe latest first-quarter survey known on day t (published about 45 days after its quarter begins)––
BILL10\mathrm{BILL10}forecasters' ten-year average 3-month bill ratePhiladelphia Fed SPF–
y(1)y^{(1)}1-year zero-coupon yieldBoard staff curve (SVENY01)–
πe,(1)\pi^{e,(1)}1-year expected inflationCleveland Fed model, as published–
ρ\rhoquarterly persistence of a gap between the 1-year real rate and its ten-year expectation: 0.9assumed (the spec); not estimated hereassumed
κ\kappashare of a move in the 1-year real rate carried into the ten-year average: 0.246, computed from the assumed ρ over 40 quartersassumed (follows from ρ)assumed
κlow\kappa_{\text{low}}the low end, 0.044: DKW's own pass-through, the slope of monthly changes in DKW's expected real rate on the 1-year real rate (Feb 2004–Oct 2026)estimated here on the Board's DKW seriesestimated

κ = 0.246 rests on the assumed ρ = 0.9; the low end, 0.044, is the one estimate of it on the page, and the premium moves with κ as the κ range shows (the survey-anchored premium at the low κ is a series of its own).

What goes with the premia (in-sample, descriptive)

RTPm=α+β′xm+εm,se^(β) Newey–West, 12 lags\mathrm{RTP}_m = \alpha + \beta' x_m + \varepsilon_m,\qquad \widehat{\mathrm{se}}(\beta)\ \text{Newey–West, 12 lags}
SymbolMeaningFromAssumed or estimated
xmx_mmonth-end realised yield volatility, stock–bond correlation, coupon share, Bund and JGB yields (RTP); the Cleveland near-term gap, stock–bond correlation, VIX and yield volatility (IRP)––

Since the last FOMC decision, model by model

Δym(10)=ΔEm[i]+ΔTPmfrom the decision’s eve to the latest day\Delta y^{(10)}_{m} = \Delta \mathbb{E}^{m}[i] + \Delta \mathrm{TP}^{m}\quad\text{from the decision's eve to the latest day}
f1y1ym=2 Em ⁣[y(2)]−Em ⁣[y(1)]f^{m}_{1y1y} = 2\,\mathbb{E}^{m}\!\left[y^{(2)}\right] - \mathbb{E}^{m}\!\left[y^{(1)}\right]
SymbolMeaningFromAssumed or estimated
mma model: survey-anchored, DKW, Kim–Wright or ACM; each splits its own fitted 10-year––
Em[i]\mathbb{E}^{m}[i]the model's expected path: the average short rate it expects over ten years––
Em[y(n)]\mathbb{E}^{m}[y^{(n)}]the model's n-year expected-path yield, continuously compoundedACM: the risk-neutral yield (ACMRNY); Kim–Wright: the fitted yield less its term premium–
f1y1ymf^{m}_{1y1y}the average short rate the model expects in the second year: the Treasury 1y1y forward with the model's own premium stripped––

The record: premium readings and the next year's change

yt+12−yt=a+b xt+uty_{t+12} - y_t = a + b\,x_t + u_t
ROS2=1−∑t(Δyt,12−f^t)2∑tΔyt,122R^2_{OS} = 1 - \frac{\sum_t (\Delta y_{t,12} - \hat f_t)^2}{\sum_t \Delta y_{t,12}^2}
SymbolMeaningFromAssumed or estimated
f^t\hat f_tforecast from the regression refitted each month on the pairs already realised (expanding window)––
ROS2R^2_{OS}above zero: the reading beat a forecast of no change––

Exhibit 12 · Inside the 10-year: definitions

  • Identities
    Four parts, three yields.

    Nominal 10-year = expected real short rate + expected inflation + real term premium + inflation risk premium; TIPS 10-year = expected real short rate + real term premium + TIPS liquidity premium; breakeven = nominal − TIPS = expected inflation + inflation risk premium − liquidity premium. All three yields are the Board staff zero-coupon curves (continuously compounded); the breakeven is exact on the two curves and is checked against the Board's own published breakeven.

  • Anchor
    The survey, moved by κ.

    Expected inflation is the Survey of Professional Forecasters' ten-year CPI forecast as published. The expected real rate starts from the same survey's ten-year bill-rate expectation less its CPI forecast (first-quarter surveys) and moves day by day with κ times the change in the one-year real rate since that survey was published. κ = 0.246 is the ten-year average weight of a transitory real-rate gap with quarterly persistence 0.9; 0.044, the pass-through the DKW model itself shows, is the low end (6 Nov 2026: expected real rate 0.72–0.82%, real term premium 1.35–1.45%). The New York Fed's r* sits well above the survey's real bill rate on average, so it is kept for the cycle gap only.

  • Calendar
    As published, never earlier.

    As published: the SPF from 45 days after its quarter begins, the Cleveland Fed's model month from day 14, DKW 15 days after each date, HLW r* 60 days after its quarter ends, the MSPD 7 days after month-end. The decomposition's day is the last day both Board staff curves are published (6 Nov 2026); ACM runs to 13 Nov 2026.

  • Benchmarks
    Four published models.

    DKW (Board, daily since 1983; TIPS parts since 1999) publishes all five parts and passes its own two identities; Kim–Wright (Board) and ACM (New York Fed) publish the premium and the expected path of the nominal yield only; the Cleveland Fed publishes expected inflation, the inflation risk premium and a real risk premium monthly, and its ten-year 'real rate' is a real yield, so its parts do not add to an observed yield — a benchmark, not an identity.

  • Drivers
    Descriptions, not causes.

    Month-end regressions since 2007 with Newey–West errors (12 lags). MOVE is replaced by the realised volatility of the 10-year zero-coupon yield (63 days, annualised, from the Board staff curve) and the S&P 500 by SPY in the stock–bond correlation (63 days of SPY log returns against the negative daily change in the 10-year yield). Descriptions of the sample, not causes; a driver whose sign flips between samples is marked.

  • Record
    Out of sample, no forward read.

    Each reading predicts the 12-month change in the 10-year zero-coupon yield from a regression refitted every month on the pairs already realised (from 2004-01, at least 60 months), scored from 2013-01 against a forecast of no change. Clark–West t with 12 lags; the hit rate is the share of months with the right sign. The in-sample slope uses Hodrick (1992) errors. The HLW r* in the cycle and policy gaps is the current one-sided file, re-estimated with each release — a mild look-ahead. The page makes no forward read from any of it.

  • Since the decision
    From the decision's eve, model by model.

    The change since the last FOMC decision runs from the close of the decision's eve — the last day both Board staff curves are published before the decision day, since the statement comes at 14:00 New York, inside that day's close — to the last day both are published; the decision's date is the one the Policy rates page records. On each model the change in its own 10-year is the change in its expected path plus the change in its term premium: the survey-anchored split on the Board's zero-coupon yield; Kim–Wright and ACM on their fitted yields over the same days; DKW, a monthly file, from its last day before the decision to its latest day, and no row until that day is on or after the decision. A model's total differs from the Board's curve by its fitting error; the curve's own change is printed beside each row.

  • One year ahead
    The 1y1y without its premium.

    The one-year rate one year ahead on each model's expected path is twice its two-year expected-path yield less its one-year yield, continuously compounded: the average short rate the model expects in the second year, which is the Treasury 1y1y forward with the model's own term premium stripped. ACM publishes the expected-path yields at one and two years (risk-neutral yields) and Kim–Wright the fitted yields and premia from which they follow; DKW publishes the 5- and 10-year maturities only and the survey-anchored path is a ten-year average, so neither carries it. The median is of the 2 models on the days all of them publish. It is the expectations ledger's market-implied policy path (the author's decision of 3 Oct 2026 (D-2026-10-03-C)).

  • Caveats
    Estimates, not prices.

    Every premium here is a model estimate, not a price. The survey anchor moves in steps (one survey a quarter, a bill-rate expectation once a year), so too little of the yield's day-to-day variation is assigned to expectations and the survey-anchored premium inherits the yield's persistence. SPF expectations have sat below realised bill rates since 2022; if the survey is slow rather than right, this premium is overstated and the affine models are closer.

Source: Board of Governors (staff curves, DKW, Kim–Wright); Federal Reserve Bank of New York (ACM); Philadelphia Fed (SPF); Cleveland Fed; Macro Review calculations · data to 6 Nov 2026rates.decomposition-9

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.