Macro ReviewHow to read
Outlook · Transmission · Transmission

Mortgage rates carry 96% of a 10-year move within a year

How much does each shock move what the market expects?

Estimated on data to Oct 2026 (mortgages to Dec 2025)Inputs read to 17 Nov 2026Updated 17 Nov 2026
10-year +100bp → mortgage rate, a year on+96bp+76bp at 3 months · data to Dec 2025
Pump price +10% → headline CPI+0.45%+0.45% in 1991–2008 · data to Oct 2026
Funds rate +100bp → 2-year, same month+58bp +48bp · data to Oct 2026
S&P 500 per +1pp on the 10-year−21.6%valuation-model arithmetic, premium and consensus growth held · 16 Nov 2026
Credit +1%, market taken out → S&P 500−0.07%raw +2.71 is mostly the equity in high yield · to 13 Nov 2026

In brief

What each channel does, in a line each; what would change the table beneath
  • Mortgage rates have carried 96% of a 10-year move within a year, and 76% within a quarter.→ Exhibit 1
  • A 10% pump-price rise has lifted headline CPI 0.45% within the month, on 1991–2026 data.→ Exhibit 3
  • In the month of a 100bp move in the funds rate, the 2-year has moved 58bp.→ Exhibit 3
  • Each +1pp on the 10-year takes 21.6% off the S&P 500, on the valuation model with premium and growth held.→ Exhibit 5
  • Net of the market, a 1% move in high-yield credit has come with a −0.07% S&P 500 move the same day over the last year.→ Exhibit 5
What would change this view
What would change this viewTriggerTodayDecided by
Pump prices reach CPI gasoline as estimatedNov 2026 CPI gasoline, unadjusted, m/m outside the model's 90% band once all surveys are inpump −5.6% on 2 of 5 surveys → −6.1 to −4.0%pendingCPI for Nov 2026, 10 Dec 2026
Pump prices reach headline CPI as estimatedNov 2026 headline CPI, unadjusted, m/m outside the model's 90% band−0.30 to 0.20% on the same surveyspendingCPI for Nov 2026, 10 Dec 2026
Mortgage rates follow the 10-year as estimatedthe at a year leaves 79–113bp per 100bp on the database's next months96bp (to Dec 2025)pendingFHFA's next quarterly mortgage database release (no release calendar dates it)
The reasoning behind each point
  • Mortgages
    Mortgage rates have carried 96% of a 10-year move within a year, and 76% within a quarter.

    New-purchase rates from the National Mortgage Database. The long-run level regression gives 1.01 per point of the 10-year (1998–2025); the survey of 1973–2019 gives 68% at a year. The database's rate is the rate at origination, which trails the rate locked, so the first month carries little.

  • Pump prices
    A 10% pump-price rise has lifted headline CPI 0.45% within the month, on 1991–2026 data.

    Through CPI energy (+5.0%) and the gasoline index (+9.3%); per 10% sustained, +0.59% by the third month. The earlier half of the sample gives +0.45, the later +0.43: the pump weighs more in the index when it is dearer.

  • Policy rate
    In the month of a 100bp move in the funds rate, the 2-year has moved 58bp.

    The 1y1y forward has moved 48bp. Delivered moves, most of them expected: the curve moves before the rate does, so this is how much of a move is still to come in its own month, not the effect of a surprise (the hike page estimates that). 1990–2008 61bp, 2009–2026 45bp.

  • Valuation
    Each +1pp on the 10-year takes 21.6% off the S&P 500, on the valuation model with premium and growth held.

    From Scenarios: the slope at today's inputs, the premium held at 3.50pp — the premium priced at consensus near-term growth, 15.1% (FactSet), not the Earnings model's 5.8% a year. The model is convex, in % from today: Hold path −106bp → +29.6%; SEP path −100bp → +27.8%; Above the dots −96bp → +26.0%. Exact given its inputs; what can be wrong is an input.

  • Credit
    Net of the market, a 1% move in high-yield credit has come with a −0.07% S&P 500 move the same day over the last year.

    From Factor analysis (13 Nov 2025 to 13 Nov 2026, 252 sessions). Net of the 10-year only the beta is +2.71: most of that is the equity in high yield. The excess bond premium's route to the equity premium is not modelled, and neither is the dollar's to import prices.

  • Shocks
    A +10bp FOMC-day surprise in the 2-year has moved homebuilders −0.81% the same day since 2012.

    The response to an identified shock: the daily return of homebuilders regressed on the 2-year yield's change on the 122 FOMC statement days since 2012 (t −3.4, Newey–West, 4 lags; 90% band −1.21 to −0.42 per +10bp; Benjamini–Hochberg q 0.003 across 30 responses). The response holds in sign over the rolling three-year windows (sign share 100%). Out of sample, the regression fitted before the last 24 decisions explained 16% of the variance of those days' returns and called the sign of 75% of them. Other responses clearing the same bar: REITs −0.75%, real estate −0.64%, utilities −0.63%. A daily window is a coarse instrument — the day's move in the 2-year is the announcement's and whatever else the day brought — and the other kinds of shock (growth news, inflation news, oil supply, term premium) have too few identified days to be estimated yet: their counts are printed.

Next: Range and scenarios — the range around each asset

Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-read

Mortgage rates carry 96% of a 10-year move within a year; the pump reaches CPI at once

How much of each move has reached its target, month by month, and one row per channel: the shock, the response, how fast, the band, the sample, the record and the page it lives on

Exhibit 1 · Share of each shock passed through by month, for each channel

Share passed through as a ratio, 0 to 12 months after a sustained move, 90% band, data to Oct 2026; dashed at one: full pass-through
About

Horizontal axis: months after the shock. The dashed line is full pass-through.

About this exhibit
Line chart of the share of each sustained move passed through (one = all of it), month by month for a year after the shock, with 90% bands, for three channels: pump price → CPI gasoline, 10-year → mortgage rate and funds rate → 2-year
How much of each move has reached its target, month by month
The share of each sustained move passed through, 0 to 12 months after it (1 = all of it), with 90% bands.
The target's response over the shock's own cumulative move at the same horizon (). At three months: CPI gasoline 0.98, the mortgage rate 0.76, the 2-year 0.50; at twelve: 0.99, 0.96, 0.58.
Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-2

Exhibit 2 · Estimated response of each target to each shock, one row per channel

Each channel's shock, response (%, bp), horizon, 90% band, sample and record on other samples, and its page; data to Oct 2026
ChannelShockResponseHorizon90% bandSampleRecordLives on
Pump price → CPI energy → headline CPIestimated+10% on the monthly average pump priceheadline CPI +0.45% (energy +5.0%, gasoline +9.3%)the same month; +0.59% per 10% sustained at 3 months0.42 to 0.48%Apr 1991 – Oct 2026419 observations1991–2008 +0.45%; 2009–2026 +0.43% (impact, per 10%)CPI
The 10-year → the new-purchase mortgage rateestimated+100bp on the 10-year (monthly average), sustained+76bp at 3 months, +96bp at 1276% passed through in a quarter, 96% in a year; level beta 1.0179 to 113bpAug 1998 – Dec 2025329 observations1998–2011 110bp; 2012–2025 92bp; survey of 1973–2019 68bp (at 12 months)Financing & affordability
The policy rate → the 2-year and the 1y1y forwardestimated+100bp on the federal funds effective rate (monthly average)2-year +58bp, 1y1y +48bp in the monththe same month; the 2-year carries 58% of the funds rate's own path at 12 months43 to 73bpAug 1990 – Oct 2026435 observations1990–2008 2-year 61bp, 1y1y 56bp; 2009–2026 2-year 45bp, 1y1y 18bpPolicy rates
The 10-year → the S&P 500, through the valuation identityfrom another page+100bp on the 10-year, so on the required return; premium and growth heldS&P 500 −21.6% per point at today's rate (the identity's slope)at once: an identity has no lag–1inputs of 16 Nov 2026exact given its inputs; convex, so a fall moves it more than a rise1Valuation
Credit → the S&P 500from another page+1% on the credit factor (high-yield return net of the 10-year), dailyS&P 500 +2.71% the same daythe same session2.24 to 3.18%13 Nov 2025 – 13 Nov 2026252 observationswith the market also taken out of the factor, −0.07 (t −1.7): most of the beta is the equity in high yield; the excess bond premium or the HQM spread → the equity premium is not modelledFactor analysis
Crude oil → the pump price, core PCE prices, the unemployment rate, real GDPestimated+10% on WTI (monthly average), sustainedthe pump price +6.5% at 12 months; core PCE prices +0.15% at 12 months; the unemployment rate −0.33pp at 12 months; real GDP +0.23% at 4 quarters12 months for the monthly targets; 4 quarters for real GDP0.05 to 0.25%Apr 1991 – Sep 2026419 observations1991–2008 core PCE +0.03%; 2009–2026 core PCE +0.22%; 1991–2008 real GDP −0.20%; 2010–2026 real GDP +0.21%Thematic impact
The policy rate → core PCE prices, the unemployment rate, real GDPestimated+100bp on the federal funds effective rate (monthly average), sustainedcore PCE prices +0.16% at 12 months; the unemployment rate −0.58pp at 12 months; real GDP +0.66% at 4 quarters12 months for the monthly targets; 4 quarters for real GDP0.07 to 0.25%Apr 1991 – Sep 2026419 observations1991–2008 core PCE +0.12%; 2009–2026 core PCE +0.13%; 1991–2008 real GDP +0.21%; 2010–2026 real GDP +0.53%Thematic impact
The dollar → import pricesnot modelled—not modelled———the BLS import and export price indexes are not among the files the site reads; nothing is estimated in their placePPI
  1. The 10-year → the S&P 500, through the valuation identity: none: an identity is exact given its inputs; Hold path −106bp → +29.6%; SEP path −100bp → +27.8%; Above the dots −96bp → +26.0%. Exact by construction: an identity of price, earnings, growth and the discount rate, with no out-of-sample record to hold. What can be wrong is an input, and the sensitivity to it is convex — −22% per point of the discount rate at today's rate, more for a fall than for a rise.
About this exhibit
Table with one row per channel: shock, response, horizon, 90% band, sample, record on sub-samples and home page; rows for pump price → CPI, 10-year → mortgage rate, policy rate → 2-year and 1y1y, 10-year and credit → S&P 500, oil and policy rate → macro, and the dollar (not modelled)
The transmission table: one row per channel
Each channel's shock, response, horizon, 90% band, sample and record on other samples, and the page it lives on.
Estimated rows describe what followed moves like the shock in their samples (); the S&P 500 rows are their pages' own numbers. Each band is printed in its row's own unit.
Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-1

A 10% pump-price rise lifts headline CPI 0.45% within the month

The estimated responses at every horizon from the month of the shock to a year later; every horizon's share as one picture; the two rows taken from other pages

Exhibit 3 · Cumulative response to each shock, month by month for a year

Change since the month before the shock, 0 to 12 months after, with its 90% band; CPI in %, rates in bp, per unit of shock
About
A 10% rise in the monthly pump price
cumulative change in each index, per cent · months after the shock
A 100bp rise in the 10-year (monthly average)
cumulative change, basis points · months after the shock
A 100bp rise in the funds rate (monthly average)
cumulative change, basis points · months after the shock
About this exhibit
Small line charts of the cumulative change for a year after a shock, with 90% bands: CPI indexes per pump-price rise, the mortgage rate per 10-year rise, and the 2-year and 1y1y per funds-rate rise; each shock's own series drawn too
What followed each shock, month by month
Cumulative change in each series 0 to 12 months after the shock, per unit of shock, with 90% bands.
Local projections: each line is the change from the month before the shock to h months after. The shock's own series is drawn too — a move that lasts carries more of itself — which is why the table prints shares.
Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-3

Exhibit 4 · Share passed through by channel, each month for a year

Share passed through as a ratio, 0 to 12 months after the shock, data to Oct 2026; blue positive, orange negative, darker when more reliable
Months afterCPI gasoline ÷ pumpCPI energy ÷ pumpHeadline CPI ÷ pumpMortgage ÷ 10-year2-year ÷ funds rate1y1y ÷ funds rate
00.930.500.040.150.580.48
10.980.540.050.520.580.43
20.990.560.060.740.550.35
30.980.580.060.760.500.29
40.980.600.060.780.420.22
50.980.610.060.850.360.13
60.990.640.070.910.340.11
70.990.660.070.970.310.07
80.990.660.070.920.330.09
90.980.650.080.930.420.22
100.980.640.080.950.500.32
110.980.630.080.950.550.38
120.990.650.080.960.580.43
About this exhibit
Heat map of the share passed through at each month for a year after the shock for six channel pairs, coloured by sign and reliability: CPI gasoline, CPI energy and headline CPI per pump price; mortgage per 10-year; 2-year and 1y1y per funds rate
Every horizon's share passed through, as one picture
The share at each month after the shock, by channel; blue positive, orange negative, darker as its t grows, hatched below 1.645.
A hatched cell's 90% band includes zero. Each share is the target's response over the shock's own cumulative move at the same horizon, both fitted on the same rows; its band, t and the rows fitted are in the hover and in the Table view.
Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-8

Exhibit 5 · S&P 500 sensitivity to the 10-year yield and to high-yield credit

Valuation model, % from 16 Nov 2026, for each Scenarios 10-year path; S&P 500 per +1% credit factor, 13 Nov 2025 – 13 Nov 2026
Scenarios branch10-year pinS&P 500, % from today
Today, no pin±0bp±0.0%
Hold path−106bp+29.6%
SEP path−101bp+27.8%
Above the dots−96bp+26.0%

−21.6% per point of the discount rate at today's rate, the table's row: the model is convex, so a fall moves it more than a rise.

Credit factor, +1%S&P 500t
Net of the 10-year+2.71%+9.6
Net of the market and the 10-year (the tile)−0.07%−1.7

The credit factor is HYG log return; its comparison series is HYG − LQD. 252 sessions, Newey–West with 4 lags.

About this exhibit
Two small tables: the S&P 500's % change from today on the valuation model for each Scenarios path's 10-year, with its slope per point; and the S&P 500's same-day beta to the credit factor, with and without the market
The S&P 500 rows, as Scenarios and Factor analysis print them
% from today on the valuation model for each 10-year pin (inputs of 16 Nov 2026); the S&P 500's credit beta, with and without the market.
The premium (priced at consensus near-term growth, 15.1%, not the Earnings model's 5.8%) and the long-run rate are held in the valuation model's answer, which is printed as % from today, not as an index level. The credit beta runs 13 Nov 2025 to 13 Nov 2026.
Source: the Scenarios and Factor analysis pages; Macro Review calculations · data to 16 Nov 2026cross.transmission-4

How each asset has moved on FOMC days

Each asset's same-day return per +10bp FOMC-day move in the 2-year since 2012, with its 90% band — a response to an identified shock, estimated on daily returns

Exhibit 6 · Each asset's same-day response to rate moves on FOMC days

Same-day return, %, per +10bp FOMC-day move in the 2-year on 122 days, 25 Jan 2012 – 16 Sep 2026; bars: 90% band; faded where not significant

Response per +10bp FOMC-day move in the 2-year: the same-day move, in %. A daily window is a coarse instrument — the day's move in the shock variable is the announcement's and whatever else the day brought. Rolling 3y: the page's stability rule over three-year windows of shock days. Out-of-sample: the regression fitted before the last 24 shocks of this kind, scored on them.

About this exhibit
Dot-and-bar chart of each asset's same-day return per FOMC-day move in the 2-year, over the FOMC statement days since 2012, with 90% bands; responses that do not clear the multiple-testing bar are faded; a toggle switches to the alternative instrument
Homebuilders moved −0.81% per +10bp FOMC-day move in the 2-year
FOMC day: change in the 2-year yield (122 days, 25 Jan 2012 – 16 Sept 2026): each asset's daily return regressed on the identified shocks, the other kinds' days held in; dot = response per +10bp FOMC-day move in the 2-year (faded where it does not clear Benjamini–Hochberg), bar = ±1.645 × Newey–West SE (4 lags on 122 shock days); 30 responses tested. Identified days: monetary policy 244 · growth news 10 · inflation news 4 · oil supply 105 · term premium 36.response to an identified shockNot estimated — growth news: 5 identified days, below the 48 the record needs; inflation news: 4 identified days, below the 48 the record needs; oil supply: 14 identified days, below the 48 the record needs; term premium: 36 identified days, below the 48 the record needs.The inflation news is the raw change in the nowcast; the bias-corrected line is not yet published.
Source: Shock table computed here from the Board's FOMC calendar, Treasury par yields, the nowcast pages, the Hormuz log, EIA weekly stocks and the auctions and decomposition pages; responses on Yahoo Finance prices (licensed).cross.factors-12

Do the estimates hold across samples?

Each estimated channel on two halves of its sample, and on every ten-year window

Exhibit 7 · Each channel's estimate on each half of its sample, and in full

Each estimated channel refitted on halves that share no month, % or bp, with its 90% band; last column: the full sample, data to Oct 2026
EstimateFirst halfSecond halfWhole sample
Pump price → headline CPI, same month (per 10%)0.45pp0.41 to 0.491991–20080.43pp0.39 to 0.472009–20260.45pp0.42 to 0.48
Pump price → CPI gasoline, same month (share)0.890.87 to 0.911991–20080.950.94 to 0.962009–20260.930.91 to 0.94
10-year → mortgage rate at 3 months (share)0.690.56 to 0.831998–20110.840.68 to 1.002012–20250.760.66 to 0.86
10-year → mortgage rate at 12 months (share)1.100.70 to 1.491998–20110.920.68 to 1.152012–20250.960.79 to 1.13
Funds rate → 2-year, same month (share)0.610.43 to 0.791990–20080.450.02 to 0.892009–20260.580.43 to 0.73
Funds rate → 1y1y, same month (share)0.560.33 to 0.791990–20080.18−0.33 to 0.692009–20260.480.31 to 0.65

The Monthly Interest Rate Survey, a second mortgage survey on Aug 1973 – Apr 2019, gives a twelve-month share of 0.68 (±0.12); the level regression of the mortgage rate on the 10-year gives 1.01 (±0.05).

About this exhibit
Table of the estimates refitted on two non-overlapping halves of each sample and on the whole, with 90% bands: pump price → headline CPI and CPI gasoline; 10-year → mortgage at three and twelve months; funds rate → 2-year and 1y1y
The same estimates on each half of the sample
Each estimated channel refitted on two halves that share no month, 90% bands beneath; the last column is the whole sample.
Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-5

Exhibit 8 · Each channel's pass-through estimate over rolling ten-year windows

Each estimate, as a ratio or in %, refitted on the rolling ten-year window ending at each month to Oct 2026, with its 90% band
Pump price +10% → headline CPI, same month
window ending in the month shown
10-year → mortgage rate, share passed through at 3 months
window ending in the month shown
Funds rate → 2-year, share in the month of the move
window ending in the month shown
About this exhibit
Small line charts of an estimate refitted on the ten years ending at each month, with 90% bands: pump price → headline CPI; 10-year → mortgage share at three months; funds rate → 2-year share in the month
Each estimate on every ten-year window
The same estimate refitted on the 120 months ending at each month, with its 90% band; blank until the first full window.
A line that wanders outside its own earlier band is an estimate that has changed. The window counts the months the regression may read, lags included.
Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-6

The 's −100bp on the 10-year would move mortgage rates −96bp within a year

: each Scenarios branch's ten-year pin taken through the mortgage channel, beside the valuation model's answer for the S&P 500 as Scenarios prints it

Exhibit 9 · Mortgage rate and S&P 500 change under each Scenarios rate path

Model-conditional, bp: each Scenarios path's 10-year move from 5.12% (16 Nov 2026) times the 12-month mortgage share, 90% band; S&P 500 % from latest
About
Pin · SEP path−101bp on the 10-yearChannel · mortgage share at 12 months0.96 (0.79 to 1.13)Effect · the mortgage rate a year on−97bp (−114 to −79bp)· the S&P 500 +27.8% from today on the valuation model, the premium and consensus growth held
Scenarios branch10-year
pin → level
S&P 500
valuation model
Mortgage rate, a year on
90% band
Hold pathThe ten-year at the hold path's fair value−106bp4.07%+29.6%from today−101bp−120 to −83bp
SEP pathThe ten-year at the SEP path's fair value−101bp4.12%+27.8%from today−97bp−114 to −79bp
Above the dotsThe ten-year at the fair value of a hike path above the dots (+50bp, then hold)−96bp4.17%+26.0%from today−92bp−108 to −75bp
Term premium upThe ten-year up by the term premium's last three monthsNot projected: no decision by the author
Hollow dealNot projected: no ten-year pin: oil is not modelled through the identity — no elasticity of index earnings or of the premium to it exists among the site's estimates, and none is invented here
AttritionNot projected: no ten-year pin: oil is not modelled through the identity — no elasticity of index earnings or of the premium to it exists among the site's estimates, and none is invented here
EscalationNot projected: no ten-year pin: oil is not modelled through the identity — no elasticity of index earnings or of the premium to it exists among the site's estimates, and none is invented here
Genuine dealNot projected: no ten-year pin: oil is not modelled through the identity — no elasticity of index earnings or of the premium to it exists among the site's estimates, and none is invented here
About this exhibit
Chain strip and table: for each Scenarios path, the 10-year pin and level, the S&P 500 % from today on the valuation model, and the mortgage-rate change a year on with its 90% band; the oil paths are not projected
What the Scenarios paths would move through these channels
Model-conditional: each branch's 10-year pin times the mortgage share at 12 months, with its 90% band; the S&P 500 as Scenarios prints it.
The pins come from What is priced, through Scenarios; the S&P 500 column is the valuation model's answer with the premium and growth held. Sensitivities with their models named, not positions or targets. The 10-year moves from 5.12% on 16 Nov 2026; the theme pages take these rows for the branches they name.
Source: the Scenarios page (the pins and the S&P 500 column); FHFA; Federal Reserve H.15; Macro Review calculations · data to 16 Nov 2026cross.transmission-7

What the estimates rest on

Each input's file, date and sample; the months left empty

Exhibit 10 · Data inputs behind each estimate with their dates, samples and gaps

Each input's publisher file, day read, first and last observation, and every month left empty, each named; data to Oct 2026
InputPublisher's fileRead fromRuns
Weekly U.S. Regular All Formulations Retail Gasoline Prices (Dollars per Gallon)EMM_EPMR_PTE_NUS_DPGw.xls17 Nov 202620 Aug 1990 – 9 Nov 2026
CPI-U U.S. city average, not seasonally adjusted: all items, energy, gasoline (all types)cu/16 Nov 2026Jan 1985 – Oct 2026
NMDB average contract rate, new home-purchase mortgages, all marketsnmdb_new_monthly_national.csv17 Nov 2026Jan 1998 – Dec 2025
MIRS contract rate, all homesmirs_t17_all_homes.xls17 Nov 2026Jan 1973 – Apr 2019
H.15 monthly averages: 10-year Treasury constant maturity, federal funds effective rateH15.zip17 Nov 2026Apr 1953 – Oct 2026
Treasury daily par yield curve, nominalpar_nominal_*.csv17 Nov 20262 Jan 1990 – 16 Nov 2026
BLS household survey: the unemployment rate (LNS14000000), seasonally adjustedln/ln.data.1.AllData.sel.tsv.gz16 Nov 2026Jan 1990 – Oct 2026
BEA NIPA table T20804: the PCE price index excluding food and energy (DPCCRG), monthlyNipaDataM.txt17 Nov 2026Jan 1959 – Sep 2026
BEA NIPA table T10106 line 1: real GDP, chained dollars (A191RX), quarterlyNipaDataQ.txt17 Nov 20261947-Q1 – 2026-Q3
Cushing, OK WTI Spot Price FOB (Dollars per Barrel) (monthly; the level is not displayed — only log changes and responses are)RWTCm.xls17 Nov 2026Jan 1986 – Oct 2026
Months left empty, and why
  • Oct 2025
    BLS CPI

    no index published for the month; left empty — every regression row that reads it is excluded, never filled

  • Jan 1990
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • Feb 1990
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • Mar 1990
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • Apr 1990
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • May 1990
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • Jun 1990
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • Jul 1990
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • Aug 1990
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • Oct 2025
    macro channel inputs

    a monthly value is missing; left empty — every regression row that reads it is excluded

  • 17 Nov 2026
    BLS import and export price indexes

    not read: the dollar → import prices channel is not modelled and says so

About this exhibit
Table of each input: publisher file, date read, first and last observation, plus every month left empty and why
What the estimates are made from
Each input's publisher file and the day it was read, its first and last observation, and every month left empty.
Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-9

Appendix

: what each estimate is, how it is fitted, what it cannot say, and every check behind it113 checks, all pass
About this page

Every scenario runs from a factor to an expectation: a pump price to the CPI, a 10-year to the mortgage rate, the policy rate to the front of the curve, a discount rate to the index. This page puts the channels in one table — each with its shock, its response, how fast it arrives, its band, the sample it was estimated on and how it held on other samples — and says where a channel is not modelled. The estimates describe what followed moves like the shock in the past; they are not forecasts.

Local projections and the pass-through share

yt+h−yt−1=αh+βh st+∑j=1Lγh,j′ Δxt−j+ut+h,h=0,…,12y_{t+h} - y_{t-1} = \alpha_h + \beta_h\, s_t + \sum_{j=1}^{L} \gamma_{h,j}^{\prime}\, \Delta x_{t-j} + u_{t+h}, \qquad h = 0, \dots, 12
θh=βh target / βh shock,se⁡(θh) by the delta method\theta_h = \beta_h^{\,\text{target}} \,/\, \beta_h^{\,\text{shock}}, \qquad \operatorname{se}(\theta_h)\ \text{by the delta method}
band: θh±1.645 se⁡(θh)(90%)\text{band: } \theta_h \pm 1.645\, \operatorname{se}(\theta_h) \quad (90\%)
SymbolMeaningFromAssumed or estimated
sts_tthe shock: the month's change in the pump price (log), the 10-year or the funds rateEIA; Federal Reserve H.15–
yt+h−yt−1y_{t+h} - y_{t-1}the target's change from the month before the shock to h months afterBLS CPI; FHFA; Treasury–
Δxt−j\Delta x_{t-j}the monthly change of every series in the channel, 6 lags (controls)the same files–
βh\beta_hthe response at horizon h per unit of shockleast squares, Newey–West errors with h lagsestimated
θh\theta_hthe pass-through share: the target's response over the shock's own cumulative response at the same hthe two regressions on common rowsestimated

Model-conditional and descriptive: what followed past moves like the shock in each sample, not an identified causal effect and not a forecast. The two equations of a share are fitted on the same rows and their Newey–West cross-covariance enters the delta-method error. The S&P 500 and credit rows are copied from Scenarios and Factor analysis, with their own models.

Exhibit 11 · Definitions

  • Model
    Model-conditional, and descriptive.

    Model-conditional: every response on this page is a local projection's description of what followed moves like the stated shock in its sample, or the valuation identity's answer with its inputs held — not a forecast and not an identified causal effect. Each estimated row regresses the change in its target from the month before the shock to h months after on the shock, with six lags of every series' monthly change as controls, h = 0 to 12; Newey–West standard errors with h lags, bands at ±1.645 standard errors (90%).

  • Share
    A response over the shock's own path.

    A shock that lasts moves its own series again the next month: a 1% rise in the monthly pump price has been followed by more, and a 100bp move in the funds rate by further moves. The pass-through share at horizon h is the target's response over the shock's own cumulative response at the same h, both fitted on the same rows, with the delta-method error from the two equations' Newey–West cross-covariance — what a sustained move of the stated size carries through.

  • Pump price
    EIA's survey against BLS's indexes.

    EIA's Monday survey averaged over each calendar month (the count of surveys is kept; a month with none would stay empty), from Sep 1990, the month after the last month EIA's early, irregular survey skipped. BLS's unadjusted indexes, because the pump price is unadjusted and the two share its seasonal; with twelve lags, which span the seasonal, the impact is +0.44% per 10% against +0.45% with six. BLS published no CPI for October 2025: that month is left empty, never filled, and every regression row that would read it is left out; the count of rows is printed.

  • Mortgages
    The rate at origination, against the 10-year.

    The National Mortgage Database's average contract rate on new home-purchase loans (all markets), Jan 1998 to Dec 2025, against H.15's monthly 10-year. It is the rate at origination, a month or two after the lock, which is why so little arrives in the first month. The level regression (1.01 ± 0.03) is a long-run check, not an estimate of speed; the Monthly Interest Rate Survey's 1973–2019 contract rate is a second survey on a second sample.

  • Policy rate
    Delivered moves, not surprises.

    The funds rate is H.15's monthly average effective rate. The 2-year is Treasury's par yield and the 1y1y forward is bootstrapped from every day's par curve exactly as the Treasury curve page does it (checked day by day against that page), then averaged by month. Delivered moves are mostly expected, so the curve has already moved; the identified response to a surprise is the hike page's.

  • Identity
    From Scenarios, with its inputs.

    Copied from Scenarios: the two-stage identity's slope per point of the discount rate at today's inputs, and its exact answer for every ten-year pin Scenarios prices, with the premium (3.502pp, priced at consensus near-term growth of 15.1%; the Earnings model's own near-term growth, 5.77% a year, is not this input), the long-run rate (4.0%) and the fade held, inputs of 16 Nov 2026. With those held the price is proportional to 1 ÷ (r − g_L), so the slope is −1 ÷ (r − g_L) and a rise moves the index less than a fall of the same size.

  • Credit
    From Factor analysis, with its form.

    Copied from Factor analysis: the S&P 500's beta on the credit factor (HYG log return, net of the 10-year) over one year of sessions, Newey–West with 4.0 lags; the equity factor set nets the market out too, and the tile reads that beta. How the credit factor is constructed is set out on that page. No estimate on the site maps the excess bond premium or the HQM spread to the equity premium.

  • Dollar
    Not modelled.

    Not modelled: the BLS import and export price indexes are not among the files the site reads, and the table does not substitute another series for them.

  • Macro rows
    Oil and the funds rate to core PCE, unemployment and real GDP.

    The two macro rows: WTI's monthly average (EIA; only its log changes and responses are printed) and H.15's funds rate, each as the shock in a local projection of core PCE prices (BEA's monthly index excluding food and energy, 100 × log), the unemployment rate (BLS) and, for the oil row, the pump price, with 6 lags of every change as controls; real GDP (BEA, chained dollars, 100 × log) on quarterly averages with 4 lags. The printed elasticity is the share at 12 months (4 quarters for GDP) per sustained shock of the stated size, with the delta-method band (kind: estimation error). These are descriptions of what followed delivered moves, anticipated or not; the identified response to a policy surprise is the hike page's, and Outlook › Thematic impact multiplies each theme branch's pin by these shares.

  • Paths
    A pin through a share.

    Model-conditional: each Scenarios branch's ten-year pin times the 10-year → mortgage share at 12 months (0.961 ± 0.104), with the band from the share's; the S&P 500 column is the identity's answer as Scenarios prints it. A branch Scenarios does not project is not projected here.

  • Same numbers
    No random draws.

    Least squares on fixed samples: no random draws, so two runs on the same files return the same numbers.

  • Licence
    Whose numbers these are.

    Every input is a federal statistical series; the rows from other pages carry their pages' own attributions.

Source: EIA; BLS; FHFA; Federal Reserve H.15; U.S. Treasury; Macro Review calculations · data to Oct 2026cross.transmission-10

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.