Macro ReviewHow to read
Thematic · The hike's transmission

The Committee's own path puts the 10-year at 4.12%; the market is higher

What does a quarter-point hike do when bonds move with stocks and oil against them, and how fast?

Said 20 Sept 2026Data to 16 Sept 2026 (markets) · projections on data to Dec 2023 · updated 17 Nov 2026
The 10-year if the market priced the Committee's path4.12%5.12% today · −100bp to go · · trigger 4.37% · not crossed
The mortgage rate on the same path6.17%7.17% today · spread held · model-conditional
The S&P 500's forward P/E on the same path22.8×18.7× today · premium and growth held · model-conditional
Unemployment first responds4 monthsindustrial production at month 1 · core prices not within three years · 1990–2023, supply-driven months
2-year since 11 Sep+11bp11 Sep 2026 → 16 Sep 2026 · +7bp on the decision day

Activity moves first; core prices not within three years

What does a quarter-point hike do when bonds move with stocks and oil against them, and how fast?
  • In the bonds-with-stocks, oil-against state a quarter-point hike moves jobs and output first, and core prices not within three years: holding (said 20 Sep 2026).→ Exhibit
  • In supply-driven months unemployment rose within a year of a tightening surprise (+1.27pp at twelve months, weak instrument): read the sign, not the size.→ Exhibit
  • The data do not rank the buckets as the classification does: at two years the ratio is 0.58, the other 0.74.→ Exhibit
  • The 10-year sits 100bp above the 4.12% the Committee's own median path implies at today's .→ Exhibit 1
  • If the 10-year reached the Committee's path, the fair-value model would put the S&P 500 +27.8% if the premium and growth were unchanged.→ Exhibit
What would change this view
What would change this viewTriggerTodayDecided by
The market stops pricing beyond the the 10-year below the SEP-path grid cell plus 25bp, 4.37% (re-logged 17 Nov 2026; 4.37% as first logged 20 Sep 2026)5.12% (16 Nov 2026); on 11 Sep 4.96% (11 Sep 2026)not crossedeach Treasury close
The rate-sensitive prices respond faster than estimatedtheir contribution to core PCE falls by more than 0.34pp+0.10pp since Jul 2026 (1.15 → 1.25, Sep 2026)not crossedPersonal Income and Outlays, 25 Nov 2026
Unemployment overshoots the estimate's bandup more than 2.20pp from its level at the decision±0.00pp since Aug 2026 (4.10 → 4.10, Oct 2026)not crossedEmployment Situation, 4 Dec 2026
The reasoning behind each point
  • Verdict
    In the bonds-with-stocks, oil-against state a quarter-point hike moves jobs and output first, and core prices not within three years: holding (said 20 Sep 2026).

    As written on 20 Sep 2026: “In an episode like this one a 25bp move does not show up as lower core inflation inside three years — the part of core PCE a policy rate is supposed to reach contributes +0.34pp more at twelve months on these estimates, not less, and the core price index's own response never leaves its 68% band — while the activity side moves first (unemployment +1.27pp and industrial production −7.5% at twelve months). What the move reaches quickly is the long end, and the market has already priced it: the 10-year at 5.01% sits 89bp above the 4.12% the rules grid puts on a path with one further move at today's term premium.” The verdict is “holding” because no trigger of the falsifier table has crossed; the scorecard sets what the log said against what has happened since.

  • Responses
    In supply-driven months unemployment rose within a year of a tightening surprise (+1.27pp at twelve months, weak instrument): read the sign, not the size.

    Normalised to a surprise that raises the 1-year by 25bp: unemployment +1.27pp at twelve months and +2.33pp at twenty-four, industrial production −7.5% at twelve. The 68% band first holds away from zero for three months at month 4 for unemployment and month 1 for production; the core price index's own response is never distinguishable from zero over three years (−0.25 log points at twelve months, SE 0.41). These are descriptions of what followed surprises like this one between 1990 and 2023, not forecasts.

  • Buckets
    The data do not rank the buckets as the classification does: at two years the rate-sensitive ratio is 0.58, the other 0.74.

    The rate-sensitive bucket's ratio is 0.58 ± 0.37 and that of the bucket a policy rate is not supposed to slow 0.74 ± 0.58. Both are measured against the core price index's response on the six-equation common sample (315 months, +1.33 per unit of surprise), which differs in sign from the core-only response at the same horizon (333 months, −0.71) — the first is the ratio's denominator, the second is the response the timing table prints. In the linear state the denominator is −0.15 with |t| 0.14: too close to zero for a ratio to mean anything, so the linear ratios are not printed.

  • Priced
    The 10-year sits 100bp above the 4.12% the Committee's own median path implies at today's term premium.

    Holding gives 4.07% and the forward curve as it stands — the 1y1y forward 5.05% against IORB 3.90%, 115bp — gives the ACM fitted 5.11%. The grid prices stated paths; the market's own reading is the fitted yield, and the two are not the same object.

  • Map
    If the 10-year reached the Committee's path, the fair-value model would put the S&P 500 +27.8% if the premium and growth were unchanged.

    The identity is Scenarios' (the page's ending copies it with the hold and market paths). This page's own map, in the fold, uses a regression estimate instead of the identity — the factors page's rates beta (+11.4%) — and a one-stage multiple (22.8× against 18.7× today). The same move takes the purchase mortgage rate to 6.17% with the NMDB spread held and the equal-weight conditions index to −0.28σ from −0.13σ. Every cell is model-conditional: one identity, everything else unchanged, and the macro rows scale a +25bp response linearly to the path's impulse.

  • Instrument
    The instrument is weak: first-stage F 5.37 on the orthogonalised surprise against the convention of 10, with data ending December 2023.

    The slope is +0.716pp (t 2.32); the raw surprise gives F 27.6. Every normalised response is divided by that state's impact coefficient of the 1-year, which is 0.565pp with |t| 1.31 in the supply state — below 3 in every state — so the level of each response is uncertain by much more than its band shows. The raw per-unit coefficient is printed beside the normalised one throughout (unemployment at twelve months: +2.88pp per unit of surprise, +1.27pp after normalisation).

Next: Transmission — the standing table of channels

Source: Bauer–Swanson surprises; FRED; Treasury; the pages cited · data to the dates shownthemes.hike_transmission-read

What the view expected against what has happened, test by test

The policy paths and the view's tests: what was written on 20 Sep 2026, what the data show, and each named verdict
WhatSaidHappenedVerdict
PathsThe three paths the map pricesSaid · 20 Sept 20263 paths named, no odds: Hold — one and done · SEP — one further +25bp · Market — the forward curve as it standsHappenednot yet decided — the four quarters after the 16 September decisionVerdictopen
TestThe market stops pricing beyond the SEP pathSaid · 5 Oct 2026the 10-year below the SEP-path grid cell plus 25bp, 4.37% (re-logged 17 Nov 2026; 4.37% as first logged 20 Sep 2026)Happened · 16 Nov 20265.12% (16 Nov 2026); on 11 Sep 4.96% (11 Sep 2026)Verdictnot crossed
TestThe rate-sensitive prices respond faster than estimatedSaid · 20 Sept 2026their contribution to core PCE falls by more than 0.34ppHappened · Sep 2026+0.10pp since Jul 2026 (1.15 → 1.25, Sep 2026)Verdictnot crossed
TestUnemployment overshoots the estimate's bandSaid · 20 Sept 2026up more than 2.20pp from its level at the decisionHappened · Oct 2026±0.00pp since Aug 2026 (4.10 → 4.10, Oct 2026)Verdictnot crossed

The view is holding: the word follows a rule set out in the Appendix, from the log and the tests above, never by hand.

About this exhibit
A scorecard showing, for the policy paths and the view's tests, what the dated log said, what has happened, and the verdict
What was said on 20 Sept 2026, and what has happened since
The paths the map prices and the three tests of the view: what the dated log said, what the data show, and the verdict
Source: the theme's dated log (said 20 Sept 2026); Treasury; BLS; BEA · data to 16 Sept 2026themes.hike_transmission-20

When each variable first responds to a rate surprise, and its peak

Months after a tightening surprise until each response first holds away from zero, and its peak; supply-driven months, Jan 1990 to Dec 2023
061218243036months after the surprise2-year yieldmoves on the day of the decisionIndustrial productionfirst at month 1, peak 34Unemploymentfirst at month 4, peak 3510-year yieldfirst at month 7, peak 19Core pricesno reading within 36 months
A dot marks the month a response first holds away from zero; a bar runs to its peak. The 2-year moves on the day, activity within months, the 10-year later; core prices never leave their band within three years — which is the view's point.
About this exhibit
A timeline diagram, months after a tightening surprise in supply-driven months, marking for the 2-year yield, industrial production, unemployment, the 10-year yield and core prices the month each response first stays clear of zero and the month of its peak
Activity moves within months; core prices show nothing within three years
When each response to a tightening surprise first holds away from zero (68% band, three months running), supply-driven months, 1990–2023
Source: Bauer–Swanson monetary-policy surprises; FRED; Macro Review calculations · data to Dec 2023themes.hike_transmission-21

The view, and the claim as written

The author's claim on 20 Sept 2026, kept as written; a figure revised since is marked where it stands

In the bonds-with-stocks, oil-against state a quarter-point hike moves jobs and output first, and core prices not within three years — holding

As written on 20 Sept 2026

In an episode like this one a 25bp move does not show up as lower core inflation inside three years — the part of core PCE a policy rate is supposed to reach contributes +0.34pp more at twelve months on these estimates, not less, and the core price index's own response never leaves its 68% band — while the activity side moves first (unemployment +1.27pp and industrial production −7.5% at twelve months). What the move reaches quickly is the long end, and the market has already priced it: the 10-year at 5.01% sits 89bp above the 4.12% the rules grid puts on a path with one further move at today's term premium.

State-dependent local projections on 386 months between January 1990 and December 2023, the pandemic stretch excluded, with the Bauer–Swanson orthogonalised surprise as the instrument. Three things qualify every number: the first stage is weak (F 5.37 against the convention of 10; the raw surprise gives 27.6); the coefficient each response is divided by to read '+25bp on the 1-year' is 0.56pp with |t| 1.31 in the supply state and below 3 in every state, so levels are far less determined than ratios; and the instrument ends December 2023, so nothing after that identifies anything. The reach ratios put the not-reachable bucket (0.74 ± 0.58) above the reachable one (0.58 ± 0.37) — the data do not rank the buckets the way the classification does, and the two are not distinguishable from each other.

StatusactiveSaid20 Sept 2026

Argued in the : 5 dated documents, listed with their sections under Sources in the Appendix.

Source: the September 2026 report and its research notes (5 documents, listed in the Appendix)themes.hike_transmission-1

The market prices more than the Committee's own path

If the market priced the Committee's path, where would the 10-year, the mortgage rate and the multiple be?

Exhibit 1 · Key market and macro readings, latest and under each policy path

Model-conditional: each cell is the reading at that path's 10-year; impulses in bp after 16 Sep 2026: hold ±0, SEP +25, market +115
About
MetricToday
±0bp
Hold
±0bp
SEP
+25bp
Market
+115bp
Identity and source
Rates
10-year Treasury par yieldas of 16 Nov 20265.12%4.07%4.12%5.11%model-conditional the rules grid cell for the path at today's ACM term premium; the market column is the ACM fitted 10-yearTreasury curve (16 Nov 2026); Policy rules grid cells
30-year purchase mortgage rateas of Oct 20267.17%6.12%6.17%7.16%model-conditional the 10-year's move with the NMDB spread held at its twelve-month mean (204.2bp)Financing & affordability (Oct 2026)
Valuation
S&P 500 forward price/earningsas of 16 Nov 202618.66×23.01×22.80×18.70×model-conditional PE × [1 − Δ10y ÷ (10-year + ERP − g)] with the premium (3.41pp) and growth (4.0%) heldValuation (16 Nov 2026)
Conditions
Conditions index, equal weightas of Sep 2026−0.13σ−0.29σ−0.28σ−0.13σmodel-conditional Σ w_i Δx_i ÷ σ_i over the 10-year and the real rate, the breakevens held so both move with the 10-year (w 0.143, σ 1.804 and 1.998)Financial conditions (Sep 2026)

Every cell is model-conditional: it is what the metric would read if the 10-year moved to the path's cell and the identity beside the row held with everything else unchanged. The macro rows scale this page's +25bp responses linearly to the path's impulse, which the estimates do not license beyond the range of surprises they were fitted on — the market column's 115bp is 4.6 times the normalisation.

About this exhibit
A table of market and macro readings, including the 10-year yield, the mortgage rate, the S&P 500 forward P/E and the conditions index, today and under the policy paths (hold, SEP, market), every cell model-conditional
The Committee's own path puts the 10-year at 4.12%, not 5.12%
Model-conditional throughout: each cell is what the metric would read if the 10-year moved to that path's cell and the identity beside the row held with everything else unchanged. It is not a forecast and no probability is attached to any column. The policy impulse over the next four quarters is Today ±0bp · Hold ±0bp · SEP +25bp · Market +115bp.

The formulas behind the policy-path table

How each path becomes table cells: the 2-year's pass-through, the S&P 500's forward price-to-earnings ratio and the conditions index, with the source pages' latest inputs
About
  • 2-year
    Day-0 pass-through.

    Each 30-minute change around an FOMC announcement regressed on the surprise, 1990–2023 (310 announcements, 35 of them unscheduled), with a constant and White standard errors. The 2-year loads 0.723 (t 19.1, R² 0.79) on the surprise, the 10-year 0.420 and the 30-year 0.253.

  • Multiple
    The forward price/earnings identity.

    PE × [1 − Δ10y ÷ (10-year + premium − growth)] with the premium 3.41pp and growth 4.0% held: the denominator is 4.527 at a 10-year of 5.12% and a multiple of 18.66× on 16 Nov 2026. Holding the premium and the growth rate is the assumption doing the work here, and the payout-basis growth convention of the valuation page applies.

  • Conditions
    The equal-weight conditions index.

    Σ wᵢ Δxᵢ ÷ σᵢ over the 10-year and the real rate, the breakevens held so both move with the 10-year: weight 0.143 each, σ 1.804 on the 10-year and 1.998 on the real rate, so the index moves 0.151σ per percentage point of 10-year. Today it reads −0.13σ (Sep 2026).

About this exhibit
Text cards giving the formulas behind the policy-path table: the 2-year's day-0 pass-through from a rate surprise, the forward P/E identity, and the equal-weight conditions index, each with its inputs
The identities behind the map
Each cell of the map is one of these three arithmetic steps applied to the path's 10-year; the inputs are the source pages' own figures, printed so the cells can be redone by hand.
Source: Reference copies as listed on the map; the day-0 pass-through is estimated here on the surprise workbook's own FOMC sheet.themes.hike_transmission-14

What followed a surprise of this size, horizon by horizon

How did unemployment, production, prices and the 10-year respond in episodes like this one, and when?

Responses of yields and jobs, output and prices to past rate surprises

Response, pp or %, to a surprise lifting the 1-year by 25bp, months to 36; red: supply-driven months; dashed: all months; Jan 1990 to Dec 2023
About

weak normalisation Every response is divided by that state's impact coefficient of the 1-year on the surprise, 0.565pp with |t| 1.31 — below 3 in every state, so the level of each response is far less determined than its band shows and the same shading applies to state S, state D and the linear projection alike. The raw coefficient per unit of surprise is printed beside the normalised one under every chart: the divisor, not the data, is what makes the normalised numbers large.

1-year Treasury par yield
pp · horizon in months
Peak +1.47 at month 10 (t +1.37) · 12 months +1.26 · per unit of surprise +2.84
68% band away from zero for three months from month 8 · 353 months fitted
10-year Treasury par yield
pp · horizon in months
Peak +1.58 at month 19 (t +2.22) · 12 months +0.83 · per unit of surprise +1.88
68% band away from zero for three months from month 7 · 336 months fitted
Unemployment rate
pp · horizon in months
Peak +3.09 at month 35 (t +2.04) · 12 months +1.28 · per unit of surprise +2.88
68% band away from zero for three months from month 4 · 333 months fitted
Industrial production, total index
log points (≈ %) · horizon in months
Peak −10.62 at month 34 (t −1.96) · 12 months −7.52 · per unit of surprise −17.00
68% band away from zero for three months from month 1 · 333 months fitted
Core PCE price index
log points (≈ %) · horizon in months
Peak −0.63 at month 28 (t −0.69) · 12 months −0.25 · per unit of surprise −0.56
68% band never stands away from zero for three months in a row · 333 months fitted
Rate-sensitive, with a lag — cumulative contribution
pp · horizon in months
Peak +0.58 at month 36 (t +1.67) · 12 months +0.34 · per unit of surprise +0.78
68% band away from zero for three months from month 0 · 315 months fitted
Not rate-sensitive — cumulative contribution
pp · horizon in months
Peak +0.71 at month 32 (t +2.41) · 12 months +0.26 · per unit of surprise +0.59
68% band away from zero for three months from month 2 · 315 months fitted
About this exhibit
Small charts of how the 1-year and 10-year yields, unemployment, industrial production, core PCE prices and two price buckets' contributions responded to a surprise that lifts the 1-year by the normalisation: supply-driven months (with bands) against all months (dashed); a toggle switches the state
Responses to a surprise that raises the 1-year by 25bp
Local projections on 386 months, Jan 1990 to Dec 2023 with the pandemic stretch excluded; horizons 0 to 36 months. Shaded bands are 0.994 and 1.645 standard errors (Supply state (F → 1)); the dashed grey line is the linear projection. These are descriptions of what followed surprises like this one between 1990 and 2023, not forecasts.
Source: Estimated on this site from Treasury par yields, BLS unemployment, the Board's G.17 industrial production, BEA's core PCE price index and the rate-sensitivity classification, with the Bauer–Swanson surprise from the Federal Reserve Bank of San Francisco.themes.hike_transmission-10

Each response by variable and state: peak, timing and size

Responses, pp or %, to a 25bp surprise to the 1-year: peak and its month, first month clear of zero, later horizons, since Jan 1990
ResponseStatePeakAttFirst visible at12 months24 months36 monthsMonths
1-year Treasury par yieldppSupply state (F → 1)+1.47raw +3.3210+1.378 m+1.26± 1.21+0.53± 1.89−1.32± 1.68353
1-year Treasury par yieldppDemand state (F → 0)+7.06raw +7.6436+2.0832 m−1.67± 2.45+0.63± 3.32+7.06± 3.40333
1-year Treasury par yieldppLinear (no state)+1.23raw +1.9836+1.1734 m+0.34± 0.85+0.63± 1.15+1.23± 1.06333
10-year Treasury par yieldppSupply state (F → 1)+1.58raw +3.5719+2.227 m+0.83± 0.83+0.93± 0.81−0.47± 0.59336
10-year Treasury par yieldppDemand state (F → 0)−3.47raw −3.759−2.227 m−2.99± 1.90−0.73± 1.90+2.27± 1.58355
10-year Treasury par yieldppLinear (no state)+1.15raw +1.8526+2.1117 m−0.16± 0.66+0.60± 0.63+0.40± 0.55333
Unemployment rateppSupply state (F → 1)+3.09raw +6.9835+2.044 m+1.28± 0.93+2.33± 1.43+3.00± 1.53333
Unemployment rateppDemand state (F → 0)+1.61raw +1.7418+0.68never+0.18± 1.72+0.92± 2.82−1.04± 2.76337
Unemployment rateppLinear (no state)+1.86raw +2.9825+1.453 m+1.00± 0.88+1.86± 1.33+1.54± 1.24333
Industrial production, total indexlog points (≈ %)Supply state (F → 1)−10.62raw −24.0034−1.961 m−7.52± 3.63−8.84± 4.79−10.36± 5.43333
Industrial production, total indexlog points (≈ %)Demand state (F → 0)−12.34raw −13.3515−1.529 m−7.51± 5.84−6.87± 9.22+2.07± 7.35343
Industrial production, total indexlog points (≈ %)Linear (no state)−13.24raw −21.1915−2.790 m−10.21± 3.65−11.42± 4.99−11.25± 5.03343

The response at a horizon is the projection's own coefficient at that horizon, not a sum over horizons; the standard errors are Newey–West with h lags.

About this exhibit
A table of every response variable in three states (supply-driven, demand-driven, all months): peak and its month, t, the first month clearly away from zero, and the response at later horizons
Peak, first visible, and the response at 12, 24 and 36 months
39 rows: 13 response variables in three states, every value normalised to +25bp on the 1-year at impact in that state. "First visible at" is the first horizon whose 68% band stays away from zero for three consecutive months.
Source: Estimated on this site; the sources of each response variable are listed under Sources.themes.hike_transmission-11

Three tests of the view, none crossed

What would change the view?

What would change the view, test by test

Thresholds from the September report; status from the latest data. 0 crossed · 0 closing · 3 not crossed · 0 pending · 0 not tracked here.
About
StatusTestLatest readingThresholdIf crossedThreshold source
not crossedThe 10-year falls below the SEP-path grid cell plus 25bp5.12% (16 Nov 2026); on 11 Sep 4.96% (11 Sep 2026)Treasury curve< 4.37%The market has stopped pricing beyond the path the Committee's own median describes; the gap this page reads as 'pricing still to complete' has closed, and the map's SEP column stops being the interesting one.Logged at 4.373% on 20 Sep 2026; re-logged at 4.365% on 17 Nov 2026 after the rules grid moved — the rules grid's cell for the SEP path (one further +25bp) at today's ACM 10-year term premium, plus 25bp (4.115% on the grid of 16 Nov 2026, plus 25bp). The test uses the latest row; all 9 stay in the record.The author's decision of 5 Oct 2026 · re-log the threshold after the rules grid moved · 5 Oct 2026
not crossedThe reachable bucket's contribution to twelve-month core PCE falls by more than this page's twelve-month response+0.10pp since Jul 2026 (1.15 → 1.25, Sep 2026)Sources & rate sensitivity≤ −0.34 ppThe transmission has arrived faster and in the opposite direction to the one estimated here: the part of core inflation a policy rate is supposed to reach is falling by more in a few months than these projections put on twelve, and the page's central estimate is wrong in sign as well as in speed.The threshold is minus the twelve-month response of the reachable bucket's cumulative contribution in the supply state, normalised to +25bp on the 1-year, as estimated by the edition of 20 Sep 2026 (+0.3439pp, 68% band +0.20 to +0.48). It is an estimate, not a published figure: this update check compares it with the current estimate and fails when they part.this site's calculation (edition of 20 Sep 2026) · local projections, state S, b_reachable at h = 12 · 20 Sept 2026
not crossedUnemployment rises from its level at the decision by more than the upper edge of this page's twelve-month band±0.00pp since Aug 2026 (4.10 → 4.10, Oct 2026)Unemployment≥ 2.20 ppThe labour market has moved by more than the widest reading this page's twelve-month response supports, so the activity leg of the transmission is stronger than estimated and the whole normalisation is too small.The threshold is the upper edge of the 68% band on the twelve-month unemployment response in the supply state, normalised to +25bp (+1.27pp ± 0.93). The base is the latest month published at the decision (August 2026); the September print moves the base when it arrives. the update's check compares the threshold with the current estimate and fails when they part.this site's calculation (edition of 20 Sep 2026) · local projections, state S, u at h = 12, 68% band · 20 Sept 2026
Source: the September 2026 report (thresholds); the pages named in each row (readings)themes.hike_transmission-2

The three paths the map prices, without odds

Which policy paths does the map price?

The policy paths the table prices, and what each assumes

Each path over the year after the 16 Sep 2026 decision, named: hold, SEP and the market's forward curve, with no odds
About
  1. Hold — one and doneNo further move: the rules grid's hold path, which starts at IORB and is flat at the SEP longer-run median from 2029 Q1. Policy impulse 0bp on the 1-year.
  2. SEP — one further +25bpThe rules grid's hike25 path, the September SEP's own median of one further move this year. Policy impulse +25bp.
  3. Market — the forward curve as it standsThe 1y1y forward less IORB as Treasury curve prints them (+115bp on 16 September 2026); the fair 10-year in this column is the ACM fitted yield rather than a grid cell.

The map is arithmetic on published objects, not a probability-weighted view: each column is a stated policy impulse and the site's own metrics under it. No odds are attached to any column here, and none is quoted from anywhere else.

About this exhibit
Text cards defining the policy paths the table prices for the year after the September decision: hold (no further move), SEP (one more hike) and market (the forward curve)
The three paths the map prices
The four quarters after the 16 September decision
Source: the site's specification for this page · 19 Sept 2026themes.hike_transmission-scenario-pricing_paths

What each path would move, on the Scenarios set

: the three paths this page's map prices, as the Scenarios set and What is priced carry them

Each policy path's 10-year yield, mortgage rate and S&P 500

Model-conditional: each path's 10-year, against 5.12% on 16 Nov 2026; mortgage-rate change at a year, with band; S&P 500, premium and growth held
Hold — one and donethe Hold path on the Scenarios set
The 10-year this path justifies4.07%−106bp from 5.12% on 16 Nov 2026
Mortgage rate a year on−101bp90% band −120 to −83bp

S&P 500 +29.6% (9,610) if the premium and growth were unchanged — the fair-value model's arithmetic on that 10-year, not a forecast.

SEP — one further +25bpthe SEP path on the Scenarios set
The 10-year this path justifies4.12%−101bp from 5.12% on 16 Nov 2026
Mortgage rate a year on−97bp90% band −114 to −79bp

S&P 500 +27.8% (9,477) if the premium and growth were unchanged — the fair-value model's arithmetic on that 10-year, not a forecast.

Market — the forward curve as it standsthe What the market prices on the Scenarios set
The 10-year this path justifies5.11%−1bp from 5.12% on 16 Nov 2026

a path What is priced prices, not a Scenarios branch: the S&P 500 and the mortgage rate are not projected on it

The same policy paths as on The decision vs the source of inflation and Housing supply 2027–28, under this page's names for them.

The channel these paths run through · The policy rate → the 2-year and the 1y1y forward: +100bp on the federal funds effective rate (monthly average) → 2-year +58bp, 1y1y +48bp in the month, the same month; the 2-year carries 58% of the funds rate's own path at 12 months; 90% band 43 to 73bp. Record: 1990–2008 2-year 61bp, 1y1y 56bp; 2009–2026 2-year 45bp, 1y1y 18bp. The transmission table carries it with its sample and its model.

What each path would do to core PCE, unemployment and real GDP is read on Thematic impact, under Outlook.

About this exhibit
Cards, one per policy path (hold, SEP, market), giving the 10-year each path justifies against the anchor, the mortgage-rate change a year on with its band, and the S&P 500 under the fair-value model
What would each path move?
Model-conditional: the ten-year column is the fair value the rules grid gives each policy path at today's term premium (What is priced); the S&P 500 column is the valuation identity's answer to that ten-year with the premium and near-term growth held (Scenarios); the mortgage column is the ten-year's move through the transmission table's pass-through at a year, with its band (Transmission). They are sensitivities with their models named, not positions or targets.
Source: the Scenarios and Transmission pages; Macro Review calculations · data to 16 Nov 2026 and Oct 2026themes.hike_transmission-3

Around the decision, the buckets and the dated record

The curve and the cross-section around 16 September, how each bucket of core inflation responded, the dated events, and the report's exhibits

2-year and 10-year yields and the S&P 500 since the September report

Daily change since the 11 Sep 2026 close: yields in basis points, S&P 500 in log points; dotted line: the 16 Sep 2026 decision

Indexed to the close of the data lock: yields in basis points, the index in log points.

About this exhibit
Daily lines of the change in the 2-year and 10-year Treasury yields (bp) and the S&P 500 (log points), all zero at the close of the report's data date, with the decision marked
The 2-year, the 10-year and the index, indexed to 11 Sep
Yields in basis points and the index in log points, both zero at the close of 11 Sept 2026; the dotted line is the decision on 16 Sept 2026.
Source: Treasury curve; the index is calculated by Macro Review and its level is not displayedthemes.hike_transmission-16

How market readings changed before, on and after the September decision

Change 11 Sep 2026 to the eve, on 16 Sep 2026 and since; yields and spreads in bp, funds and the index in log points
MetricLock → decision eveDecision dayDecision → latestOn 11 SepLatest

Each column is read between the last session on or before its dates in the source page's own series, so a row whose page has not published since the decision shows no session and no value. Yields, spreads and the cross-sectional readings also carry their level at each date; fund and index rows are log changes only, and no price or index level is shown. The Board's broad dollar and the weekly security-market-line readings are published with a lag, so their latest session is the September report date (11 Sep) itself.

About this exhibit
A table of market readings (yields, spreads, funds, the index, oil), each with its change from the report's data date to the decision eve, on the decision day and from the decision to the latest session, plus levels
Since the September report (11 Sep)
Read between the last session on or before each date in the source page's own series. 0 of the 28 rows have no session after the decision — their source pages publish only through 16 Sept 2026 — and those cells say so instead of carrying the decision day's number forward.
Source: Treasury curve, Treasury curve, Performance and Risk sentiment, with the WTI spot price and the index calculated by Macro Review; fund and index rows are log changes only and no level is shown.themes.hike_transmission-15

Each price bucket's response at two years, against its assumed rate sensitivity

Each bucket's contribution to core PCE, pp, 24 months after a 25bp surprise, and its ratio to core's; supply-driven months since Jan 1990
About
BucketWhat the classification assumesResponse at 24 monthsRatio to core68% band
Rate-sensitive, with a laga funds-rate change acts on this price through demand, with a lag — the part the instrument is aimed at+0.776pp± 0.580+0.583SE 0.371+0.21 to +0.95
Partly rate-sensitivepart of the price responds to the policy rate, part does not−0.359pp± 0.392−0.270SE 0.469−0.74 to +0.20
Priced off asset valuesreached only through asset prices, not through demand for the good−0.050pp± 0.395−0.038SE 0.318−0.35 to +0.28
Raised by a hike (imputed bank services)imputed bank services: a higher policy rate raises the measured price−0.052pp± 0.390−0.039SE 0.314−0.35 to +0.27
Not rate-sensitiveno channel through which a policy rate acts on this price+0.982pp± 0.553+0.738SE 0.577+0.17 to +1.31
The two denominators. The ratio's denominator is the core price index's response on the six-equation common sample: +1.330 log points on 315 months, |t| 1.00. The core-only equation, which is the one the timing table prints, is fitted on 333 months and gives −0.711 (SE 1.599) at the same horizon — a different sample and a different sign. Both are printed because neither stands in for the other.

A ratio means nothing when its denominator is indistinguishable from zero, so a state's ratios are shown only when the denominator's |t| reaches 0.5. The linear (no state) denominator is −0.151 with |t| 0.14: too close to zero for a ratio to mean anything, so its ratios are described here rather than printed.

About this exhibit
A table of the five price buckets: what the classification assumes, each bucket's estimated response at two years, its ratio to the core index's response, and the band, by state
Assumed against estimated: the classification's label beside the estimated response ratio
Each bucket's response at 24 months divided by the core price index's response at the same horizon and in the same state, so the normalisation cancels. Both equations are fitted on the 315 months where all six are observed.
Source: Buckets from the site's classification of BEA's PCE components by channel (the Sources & rate sensitivity page); the responses are estimated here.themes.hike_transmission-12

The dated events behind the view

Events with their source; scheduled dates are marked, and events read from a data series say so
  • 29 Dec 2023
    The Bauer–Swanson surprise series ends: the workbook's 'update 2023' sheets carry no announcement after December 2023, so every projection on this page stops there.

    Federal Reserve Bank of San Francisco, monetary-policy-surprises-data.xlsx · Monthly (update 2023) and FOMC (update 2023) sheets · 19 Sept 2026

  • 5 Mar 2026
    The site's correlation quadrant turns to the supply reading (stocks with bonds, against oil) and stays there; the transition variable used here reads F = 0.93 in September 2026.

    Correlations & regime: pair by pair · quadrant strip · 17 Sept 2026

  • 11 Sept 2026
    The September report’s data lock: the 10-year at 4.96%, the 2-year at 4.63%, futures about 86% for a September move.

    Treasury curve · series n2y, n10y · 17 Sept 2026

  • 16 Sept 2026
    The FOMC raises the target range; the 2-year closes +7bp and the 10-year +1bp on the day, the 2s10s slope 6bp flatter.

    Treasury curve · series n2y, n10y, s2s10s · 17 Sept 2026

  • 17 Sept 2026
    Interest on reserve balances moves to 3.90%: the 1y1y forward is 115bp above it, which is the map's market column.

    Policy rules · decomp.forwards · 17 Sept 2026

  • 20 Sept 2026
    This page is built: state-dependent local projections on 386 months, first-stage F 5.37, effective supply-state sample 193.8 months.

    this site's calculation · edition of 20 Sep 2026 · 20 Sept 2026

  • 30 Sept 2026
    BEA annual update and the August PCE print: the reach buckets and their contributions move with it, and this page updates after the Sources & rate sensitivity page.

    research note: Inflation sources outlook (13 Sep 2026) · §1 · 13 Sept 2026

  • 28 Oct 2026
    FOMC (27–28 Oct): the first test of whether the SEP column or the hold column is the one the map should be read down.

    the September 2026 report · §4.3 what to watch · 13 Sept 2026

Source: the September 2026 report and the pages cited per rowthemes.hike_transmission-4

The September report's exhibits and the live page for each

Each named exhibit as the September report printed it, fixed at its 11 Sep 2026 data date, with the page that now carries it live
No.StemExhibitLive page
E-H1_responsesResponses to a surprise that raises the 1-year by 25bp, supply state against the linear projectionThe hike's transmission
E-H2_timingPeak horizon, first visible at, and the response at 12, 24 and 36 months, by variable and stateThe hike's transmission
E-H3_reach_ratiosAssumed against estimated: the classification's label beside the estimated reach ratioThe hike's transmission
E-H4_pricing_mapThe drift to full pricing: what the site's own metrics would read under each pathThe hike's transmission
E-H5_event_studySince the lock: the curve, the funds and the cross-section around the 16 September decisionThe hike's transmission
136E-W1_source_instrument_matrixSource → instrument: what a 25bp move reaches, by source of the current inflationSources & rate sensitivity
63E11_expectations_tableWhat is priced, called and bet for 16 SeptemberPolicy rates
About this exhibit
A table of the September report's exhibits behind this theme: title and the live page that carries each
The September report's exhibits, and where each lives now
Exhibit numbers as the report printed them on 15 September; each stays fixed at the 11 September data date and points to the page that carries it live
Source: the September 2026 report, exhibit listthemes.hike_transmission-5

Method

Specification, , the sample, the instrument and its limits, the sensitivities, and the sources112 checks, all pass
About this page

The September decision raised the target range in an episode the site's own correlations read as supply-driven. This page asks what a move of that size did to the macro factors in episodes like it, how long it took, and what the site's other pages would read if the market finished pricing the path — the estimates are descriptions of 1990–2023 data, the map is arithmetic on published figures, and neither is a forecast.

The state-dependent local projection

yt+h−yt−1=F(zt−1) [αhS+βhS Δi^t+γhS′Xt−1]+(1−F(zt−1)) [αhD+βhD Δi^t+γhD′Xt−1]+εt+hy_{t+h} - y_{t-1} = F(z_{t-1})\,\big[\alpha^{S}_h + \beta^{S}_h\, \widehat{\Delta i}_t + \gamma^{S\prime}_h X_{t-1}\big] + \big(1 - F(z_{t-1})\big)\,\big[\alpha^{D}_h + \beta^{D}_h\, \widehat{\Delta i}_t + \gamma^{D\prime}_h X_{t-1}\big] + \varepsilon_{t+h}
F(z)=11+e−γz,Δit=π MPSt⊥+ut  (first stage)F(z) = \frac{1}{1 + e^{-\gamma z}}, \qquad \Delta i_t = \pi\, \text{MPS}^{\perp}_t + u_t \;\text{(first stage)}
SymbolMeaningFromAssumed or estimated
yt+h−yt−1y_{t+h} - y_{t-1}the change in the outcome h months after the surprise (unemployment, production, prices, the 10-year)––
Δi^t\widehat{\Delta i}_tthe 1-year yield's change, instrumented by the orthogonalised monetary-policy surprise (Bauer–Swanson)the surprise series, to December 2023–
F(z)F(z)the weight on the supply state: z is the stock–bond minus stock–oil correlation, standardised––
βhS\beta^{S}_hthe response in the supply state, divided by the impact coefficient to read '+25bp on the 1-year'––
Xt−1X_{t-1}controls: lags of the outcomes, the instrument and the state––

The diagnostics behind the estimates

The impact coefficient, the effective supply-state sample, the bucket ratio and the peak, as estimated
Impact on the 1-year, per unit of surprise+0.565ppsupply state · t 1.31 · first stage F 5.37 (convention 10)
Unemployment at its peak, supply state+3.09ppmonth 35 · t 2.04 · +1.27pp at twelve months · normalised to +25bp on the 1-year
Response ratio, rate-sensitive prices0.5824 months · SE 0.37 · not rate-sensitive 0.74 · common sample 315 months
Effective supply-state months193.8Σ F(z) over the 386 months estimated · against 438 parameters a regime-specific VAR would need
Source: Macro Review calculationsthemes.hike_transmission-22

The instrument

About this exhibit
The monthly surprise and its orthogonalised version, in percentage points. Both end Dec 2023, so every projection on this page does. The first stage of the change in the 1-year on the orthogonalised surprise is F 5.37 on 385 months.
Source: Monetary policy surprises: Federal Reserve Bank of San Francisco (Bauer and Swanson 2023). The state is calculated on this site from daily stock, bond and oil moves. (data to Sep 2026)themes.hike_transmission-6

The transition variable

About
About this exhibit
ρ(stocks, bonds) − ρ(stocks, oil) over 63 sessions at each month end, standardised over 1990–2023, to Sep 2026. Above zero is the supply reading: stocks moving with bonds and against oil.
Source: Monetary policy surprises: Federal Reserve Bank of San Francisco (Bauer and Swanson 2023). The state is calculated on this site from daily stock, bond and oil moves. (data to Sep 2026)themes.hike_transmission-7

The state weight

About this exhibit
F(z) = 1 ÷ (1 + exp(−γz)) with γ = 1.5, so F → 1 is the supply state; the threshold variant is the same variable read at zero. Σ F(z) over the months estimated is 193.8.
Source: Monetary policy surprises: Federal Reserve Bank of San Francisco (Bauer and Swanson 2023). The state is calculated on this site from daily stock, bond and oil moves. (data to Sep 2026)themes.hike_transmission-8

The instrument is weak: first-stage F 5.37 on the orthogonalised surprise against the convention of 10, with data ending December 2023.

About
First stage: Δ 1-year on the surpriseSlopeSEtFMonths
MPS_ORTH, excluded samplethe page's own first stage+0.7160.309+2.325.37385
MPS, excluded sample+1.2420.236+5.2527.60385
MPS_ORTH, inclusive sample+0.9710.358+2.717.37407
MPS, inclusive sample+1.4250.252+5.6531.89407
MPS_ORTH, to 2020-02+0.5430.316+1.722.97361
MPS, to 2020-02+1.0870.243+4.4720.01361

The strong-instrument convention is F ≥ 10. The page's own first stage is 5.37; the read says so, and every response carries the weak-normalisation note.

SensitivityMonthsFirst stage FImpact on the 1-year, state SUnemployment
peak · 12m
Industrial production
peak · 12m
Core PCE
peak · 12m
Rate-sensitive
peak · 12m
Not rate-sensitive
peak · 12m
10-year
peak · 12m
The raw surprise (MPS) in place of the orthogonalised one38627.60+0.828t +2.17+2.32+1.17−6.54−5.68−0.82−0.53+0.40+0.06+0.23+0.11+1.20+0.70
The inclusive sample (the pandemic stretch kept)4087.37+0.192t +0.39+11.26+6.35−29.62−23.36+1.25−0.97+1.83+0.21+2.40+0.74+3.83+1.52
daily-window proxy, not the 30-minute construction31 proxy months from Jan 2024 to Jul 2026, 22 statement days4176.33+0.576t +1.52+2.83+1.15−9.77−6.46−0.63−0.23+0.66+0.27+0.72+0.23+1.64+0.57

Sensitivity B — the 2024–26 extension, Method only. Daily close-to-close changes in the 1-year par yield on the 22 FOMC statement dates the Board's calendar page lists from Jan 2024 to 16 Sep 2026, summed by month and appended to the monthly instrument; the sample runs to Jul 2026, the last month every series is published (417 months estimated). This is a daily-window proxy, not the 30-minute construction: a daily window contains everything else that happened that day, and the proxy is made from the policy rate the first stage regresses on, so its first stage (6.3) is arithmetic, not identification. It is shown to say what three more years of data do to the shape, not to re-identify anything.

About this exhibit
The first stage and the three sensitivities
The instrument's strength, and what the projections look like when the shock, the sample or the end date changes. Sensitivity B — the 2024–26 daily-window extension — is shown here only: it is a proxy for the 30-minute construction, not the construction itself.
Source: Estimated on this site; the surprise workbook is the Federal Reserve Bank of San Francisco's.themes.hike_transmission-18

Sensitivity A — the linear proxy-SVAR

1-year Treasury par yield
pp · horizon in months
Peak +0.43 at month 9 · 12 months +0.40 · 24 months +0.13
10-year Treasury par yield
pp · horizon in months
Peak +0.23 at month 18 · 12 months +0.16 · 24 months +0.12
Industrial production, total index
log points (≈ %) · horizon in months
Peak −1.07 at month 9 · 12 months −0.78 · 24 months +0.14
Unemployment rate
pp · horizon in months
Peak +0.30 at month 36 · 12 months +0.25 · 24 months +0.20
Core PCE price index
log points (≈ %) · horizon in months
Peak −0.41 at month 36 · 12 months −0.22 · 24 months −0.28
HQM 10-year corporate spread
bp · horizon in months
Peak +35.33 at month 3 · 12 months +5.11 · 24 months +1.86

Identification check: with the instrument replaced by the 1-year's own reduced-form residual the impulse vector equals the Cholesky column with the 1-year ordered first — the largest difference is exactly zero. The projections above, not this VAR, are the page's estimate; this is the linear reference the specification asks for.

About this exhibit
A VAR(12) on 1-year Treasury par yield, 10-year Treasury par yield, Industrial production, total index, Unemployment rate, Core PCE price index, HQM 10-year corporate spread over the same sample (362 rows), the policy shock identified by regressing the 1-year's reduced-form residual on the surprise (F 5.97, t 2.44) and reading the impulse vector off the covariance of the other residuals with the fitted values. Responses are normalised to +25bp on the 1-year at impact; bands are 500 moving-block bootstrap replications with a block of 12.
Source: Estimated on this site from the same series as the projections.themes.hike_transmission-17

Specification, sample and conventions

  • Theme
    What the page reads, and what it does not claim.

    Everything in this section is a model-conditional description of what followed surprises like this one between 1990 and 2023, not a forecast and not a target. The page asks what a 25bp move does to the macro factors in an episode like this one, how soon, and what the site's own metrics would read as the market completes the pricing. The responses are estimated here from 1990–2023 data; the map and the event study are arithmetic on other pages' published published figures, copied through the manifest and compared with the source on every update.

  • Specification
    The projections, their controls and their standard errors.

    State-dependent local projections (Jordà 2005; Auerbach and Gorodnichenko 2012; Ramey and Zubairy 2018), monthly, horizons 0–36: y(t+h) − y(t−1) = F(z(t−1))·[α_S + β_S·s(t)] + (1 − F(z(t−1)))·[α_D + β_D·s(t)] + Σ(j=1..6) Γ_j·X(t−j) + e(t+h), with X = [d_i1, d_ip, d_u, d_p_core, s]. The controls enter in both states rather than interacted: the state splits the shock, not the dynamics. The design has 34 regressors against 32 in the linear projection, which is estimated alongside as the reference. Standard errors are Newey–West with h lags (Bartlett, no small-sample correction); bands are 0.994 and 1.645 standard errors. Responses: 1-year Treasury par yield, 10-year Treasury par yield, Unemployment rate, Industrial production, total index, WTI spot price, Core PCE price index and the rest of the thirteen.

  • Units
    Why every response is divided by an impact coefficient.

    Every response is rescaled so the shock is the one that moves the 1-year by 25bp on impact in that state: β̃_h = β_h ÷ β(i1,0) × 0.25. That divisor is 0.565pp with |t| 1.31 in the supply state, 0.271pp with |t| 0.66 in the demand state and 0.400pp with |t| 1.86 in the linear projection — below 3 everywhere, so every state carries the weak-normalisation note and the raw per-unit-of-surprise coefficient is stored and printed beside the normalised one. Dividing by a small, imprecise impact is what makes the normalised numbers large; their ratios are far better determined than their levels.

  • First stage
    How strong the instrument is.

    First stage: the change in the monthly average 1-year par yield on the month's surprise, with a constant and a White covariance, on the projections' own 385 months — slope +0.716pp per unit, t 2.32, F 5.37. That is below the strong-instrument convention of 10 and the read says so. The raw surprise (MPS) gives 27.6 on the same months and is carried as the sensitivity; the inclusive sample gives 7.4 and the pre-2020 sample 3.0.

  • Sample
    1990 to 2023, with the pandemic stretch excluded.

    Jan 1990 to Dec 2023 — the instrument ends December 2023, so every projection does. The pandemic stretch Mar 2020 to Dec 2021 is excluded: an observation enters a regression only if every month it reads (the control lags, the base of the change, the shock month and the far end) lies inside the sample, which leaves 386 months and 373 usable rows at horizon 0 falling to 333 at 36. The rate-sensitivity buckets' chain starts Jan 1992 (its first year is burn-in), so their equations are shorter. The inclusive sample is carried as a sensitivity.

  • State
    The transition variable, and what it is checked against.

    z = ρ(stocks, bonds) − ρ(stocks, oil) over 63 sessions at each month end, from S&P 500 log returns, minus the daily change in the 10-year par yield as the bond-price proxy, and EIA daily WTI log changes (8456 common sessions from 3 Jan 1990); standardised over 1990-01–2023-12 (mean −0.215, standard deviation 0.577). F(z) = 1 ÷ (1 + exp(−γz)) with γ = 1.5, the Auerbach–Gorodnichenko value on a standardised variable, so F → 1 is the supply state. It is checked against the site's own correlation quadrant: 100.0% of the 84 supply-quadrant weeks since Jan 2006 fall in months with F ≥ ½ and 93.6% of the 746 demand-quadrant weeks in months with F < ½, against a floor of 70% each way. The hard variant 1[z > 0] classifies the same months.

  • No regime VAR
    Why the regime-specific estimate is a projection, not a VAR.

    Why no regime-specific VAR: with γ = 1.5 the effective supply-state sample is Σ F(z) = 193.8 months of the 386 estimated, against 438 parameters per regime for a VAR(12) on 6 variables. The state-dependent projections are the regime-specific estimate this data size supports.

  • Proxy-SVAR
    Sensitivity A, and the identification check it passes.

    Sensitivity A — the linear proxy-SVAR (Mertens and Ravn 2013; Gertler and Karadi 2015): a VAR(12) on i1, y10, ip, u, p_core, hqm over the same sample (362 rows), the policy shock identified by regressing the 1-year's reduced-form residual on the surprise (F 5.97, t 2.44) and reading the relative impulse vector off the covariance of the other residuals with the fitted values; responses normalised to +25bp on the 1-year at impact, bands from a moving-block bootstrap (block 12, 500 replications). With the instrument replaced by the 1-year's own residual the impulse vector is the Cholesky column with the 1-year ordered first, to 1.1e-14 — the identity that says the identification is doing what it claims.

  • Extension
    Sensitivity B: a daily-window proxy for 2024–26, Method only.

    Sensitivity B — the 2024–26 extension, Method only. Daily close-to-close changes in the 1-year par yield on the 22 FOMC statement dates the Board's calendar page lists from Jan 2024 to 16 Sep 2026, summed by month and appended to the monthly instrument; the sample runs to Jul 2026, the last month every series is published (417 months estimated). This is a daily-window proxy, not the 30-minute construction: a daily window contains everything else that happened that day, and the proxy is made from the policy rate the first stage regresses on, so its first stage (6.3) is arithmetic, not identification. It is shown to say what three more years of data do to the shape, not to re-identify anything.

  • Ratios
    What a response ratio divides by, and when it is not printed.

    The ratio is the bucket's response divided by the core price index's at the same horizon and in the same state, so the normalisation cancels and the standard error comes from the two equations' joint Newey–West covariance. Both equations are fitted on the months where all six are observed, which is not the sample the core-only response is fitted on — both are printed. A ratio means nothing when its denominator is indistinguishable from zero, so a state's ratios are shown only when the denominator's |t| reaches 0.5: the linear state's does not and its ratios are described rather than printed.

  • Instrument
    The surprise series, its citation and its end date.

    Bauer, Michael D., and Eric T. Swanson (2023), "A Reassessment of Monetary Policy Surprises and High-Frequency Identification", NBER Macroeconomics Annual 37; data from the Federal Reserve Bank of San Francisco. The series ends December 2023. The workbook's monthly sheet runs 1988-02 to 2023-12 (431 months) and its FOMC sheet carries 361 announcements from 4 Feb 1988 to 13 Dec 2023, 68 of them unscheduled. The surprises are interest-rate changes over 30-minute windows around FOMC announcements: Eurodollar futures to 2022, SOFR futures from January 2023. MPS is the first principal component of the ED1–ED4 changes scaled so the effect on ED4 is one; MPS_ORTH is MPS orthogonalised on the six macro-financial variables of the paper's Table 1.

  • Rounding
    Why the instrument's identity check is not run at machine precision.

    The workbook's monthly sheet is published rounded, so the check that each month's MPS equals the sum of that month's announcement surprises is run at a tolerance of 2e-5 rather than at machine precision; the worst difference across the 431 months is 1.37e-5. Months without an announcement carry an exact zero and are checked at zero tolerance.

  • Pass-through
    The day-0 regression the 2-year row uses.

    Day-0 pass-through: each 30-minute change around an FOMC announcement regressed on the surprise, 1990–2023 (310 announcements, 35 of them unscheduled), with a constant and White standard errors. The S&P 500 column is 100 × the log change.

  • Map
    What a model-conditional cell means.

    Every cell is model-conditional: it is what the metric would read if the 10-year moved to the path's cell and the identity beside the row held with everything else unchanged. The macro rows scale this page's +25bp responses linearly to the path's impulse, which the estimates do not license beyond the range of surprises they were fitted on — the market column's 115bp is 4.6 times the normalisation.

  • Event study
    How a column with no session is read.

    Each column is read between the last session on or before its dates in the source page's own series, so a row whose page has not published since the decision shows no session and no value. Yields, spreads and the cross-sectional readings also carry their level at each date; fund and index rows are log changes only, and no price or index level is shown. The Board's broad dollar and the weekly security-market-line readings are published with a lag, so their latest session is the September report date (11 Sep) itself.

  • References
    Taken from the published pages, never recomputed.

    Every live number on this page is a copy of a published US Macro page, taken with that page's date, its as-of date and the value's place in it. Nothing is recomputed. On every update each copied value is compared with the source page, matched by date label rather than position; a mismatch, or a source page whose own checks fail, stops publication.

  • Status
    How a trigger's status is set.

    crossed — the threshold condition holds on the latest observation (for a run, on each of the last n); closing — not crossed, the series has printed since the September report's data date (11 September 2026), and the gap to the threshold is half or less of the gap on that date; not crossed — otherwise; pending — the observable is dated in the future or no page of this site carries it yet; not tracked here — an event, or subscription data this site does not display. Colours flag attention, not good or bad news for the book.

  • Odds
    None are attached to any path.

    Odds come only from the September report's documents and are kept in a dated file, one row per revision with the reason and the source document; this site never revises them. Each path states its basis: market-implied (a quoted market price), model (a documented model of this site's own) or the author's judgement (a dated assessment anchored to the quoted evidence). Only market-implied and model odds are drawn as odds; the author's judgement is shown in the folded record of each set, labelled with the date it was made.

  • Checks
    What must pass before the page is published.

    111 checks run on every update, all of which must pass before the published figures is written: the rate-sensitivity buckets' Fisher identity; the instrument against the workbook's own FOMC sheet; the page's coverage; the state's standardisation, its threshold variant and its agreement with the site's quadrant; the first stage's sign and sample; the response ratios' common sample; the SVAR's Cholesky identity and its normalisation; the day-0 pass-through; the map's identities against the Rules & paths, Factor analysis, Valuation, Financial conditions and Financing & affordability pages; the event study's slope and breakeven identities; the logged trigger thresholds against the estimates they were written from; and every copied observation against its source published figures.

  • Licences
    What is never displayed.

    The S&P 500 index, the sector and bond funds and the WTI spot price are vendor data, shown with their attribution. This page works in log changes throughout because that is what the projections estimate. The Treasury curves, the BLS unemployment rate, the Board's G.17 and H.15, BEA's PCE price index, the Treasury HQM curve and the SF Fed surprise workbook are published. The rate-sensitivity classification and everything estimated here are computed on this site.

Source: The site's own method for this page; publisher attributions are kept verbatim under Sources.themes.hike_transmission-19

Why a regime-specific vector autoregression is not estimated

Why no regime-specific VAR: with γ = 1.5 the effective supply-state sample is Σ F(z) = 193.8 months of the 386 estimated, against 438 parameters per regime for a VAR(12) on 6 variables. The state-dependent projections are the regime-specific estimate this data size supports.

About this exhibit
With γ = 1.5 the effective supply-state sample is Σ F(z) = 193.8 of the 386 months estimated, against 438 parameters per regime for the VAR the sensitivity fits.
Source: Macro Review calculations on this page's state seriesthemes.hike_transmission-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.