Macro ReviewHow to read
Cross Asset · Fair value · Valuation

At the the is 42% below the price; no target

A two-stage payout model on this site's 5.8% near-term earnings growth (consensus 15.1%) and a 5.2-point premium values the index at 4,338: a model reading, not a forecast. What is the S&P 500 worth, and what premium is the market paying for equity risk?

Treasury yields · 16 Nov 2026Index earnings · Sep 2026; consensus · 11 Sep 2026S&P 500 · 16 Nov 2026Updated 17 Nov 2026
Gap to the model's value−41.5%the model's value at the yield-gap premium 4,338 · · 16 Nov 2026price above the model's value
()+3.50ppvs 5.17pp yield-gap premium · 16 Nov 2026below the model's premium
8.62%as priced: 10-year 5.12% + 3.50pp · 16 Nov 2026
S&P 5007,41718.7× forward earnings · 16 Nov 2026
Cyclically adjusted P/E39.498th 1881 · Nov 2026
10-year Treasury yield5.12%real 2.92% · 16 Nov 2026

In brief

The , the and what would change this view
  • Today's price implies an equity risk premium of 3.50pp over the 10-year, against Damodaran's 4.20% on 1 Oct 2026.→ Exhibit 2
  • Slower near-term growth takes 1,512 points off the price and a higher premium 1,566.→ Exhibit 1
  • Across the premium bases the model's value runs from 3,448 to 5,904; the price is above every one.→ Exhibit 9
  • One input at a time, near-term growth moves the model's value most: 3,072 to 5,777.→ Exhibit 7
  • Year to date to 16 Nov 2026, earnings added 18.8 to the S&P 500 and higher yields took 27.5.→ Exhibit 12
What would change this view
What would change this viewTriggerTodayDecided by
The premium the price offers climbs halfway to the implied ERP > 4.34pp3.50pp (16 Nov 2026)not crossedeach close
…and not because the 10-year fell10-year Treasury yield < 4.87% (−25bp)5.12% (16 Nov 2026)not crossedeach Treasury close
Stocks and bonds stop moving together (the comparison months change)63-day stock–bond correlation < 0+0.10 (16 Nov 2026)not crossedeach close
The price falls to the model's valueS&P 500 ≤ 4,3387,417 (16 Nov 2026)not crossedeach close
The reasoning behind each point
  • Implied ERP
    Today's price implies an equity risk premium of 3.50pp over the 10-year, against Damodaran's 4.20% on 1 Oct 2026.

    The implied equity risk premium (ERP) is the return over the 10-year Treasury (5.12% on 16 Nov 2026) that makes the two-stage model return today's price, with near-term EPS growth at the consensus 15.1% fading to terminal growth of 4.0% over 6 years; the cost of equity is the two added, 8.62%. The model's value is read at the yield-gap premium, 5.17pp: today's implied premium plus 1.67pp, the distance from today's yield gap (the earnings yield less the 10-year, −1.14pp) to its median in the 64 past months since 2005 when stocks and bonds moved together (+0.53pp) — an assumption that the gap reverts, not a median of past premia, which this page's inputs cannot give (consensus growth has no monthly history). On the valuation history's rules the median implied premium of those months is 5.04pp (3.20–5.58pp between the quartiles). Month by month since 1970 the median is 5.55pp; on the valuation history's own rules today reads 2.41pp, the 4th percentile of its record. Aswath Damodaran's implied premium (NYU Stern) was 4.20% on 1 Oct 2026 against a 5.29% T-bond, on dividends plus buybacks and a five-year consensus; his series is drawn beside this page's as a cross-check. Bridged one of his conventions at a time: terminal growth = the 10-year +1.04pp; his cash yield −0.94pp; his growth input −0.04pp; his date −0.12pp; his five-year-then-stable model vs this fade +0.25pp; his 3.70% (trailing cash yield) → 4.20% headline +0.51pp — from 3.50pp to his 4.20%.

  • The model's value
    Slower near-term growth takes 1,512 points off the price and a higher premium 1,566.

    The S&P 500 closed at 7,417 on 16 Nov 2026. Step by step in index points: near-term growth −1,512 (consensus 15.1% against the site's 5.77% a year, from a fan with no edge over a fixed mean), the 10-year ±0 (held where it is), terminal growth ±0 (the same on both sides) and the premium −1,566 (3.50pp to 5.17pp). The model's value is conditional on these inputs — the model's value at the yield-gap premium 4,338 (−42%) · record unproven · no target: whether the gap has predicted returns is tested on the valuation history page.

  • Premium basis
    Across the premium bases the model's value runs from 3,448 to 5,904; the price is above every one.

    today's implied premium, at the site's growth 3.50pp → 5,904 (−20%); yield-gap premium 5.17pp → 4,338 (−42%); full-history median, as a distance 6.64pp → 3,516 (−53%); yield-gap premium, months when stocks and bonds moved apart 6.79pp → 3,448 (−54%). Each bar spans near-term growth from its 10th to its 90th percentile (−4.9% to 17.9% a year); drawn together with the premium the model's value runs from 2,912 to 6,386 (10th–90th percentile of 20,000 draws, not a probability). Every basis is read at the site's near-term growth (5.77%, from a fan with no edge over a fixed mean); the full-history median enters as its distance from today on the history's own rules, so every row is on this page's rules. The model's value is a function of the premium basis.

  • Sensitivity
    One input at a time, near-term growth moves the model's value most: 3,072 to 5,777.

    One input moved to each end of its range, every other input at its headline value and the premium held: near-term growth 3,072–5,777; the payout 3,338–4,879; the premium 3,976–5,216; terminal growth 4,164–4,670; the 10-year 4,173–4,518; the transition period 4,268–4,479; the term-premium model 4,320–4,338. The premium × terminal-growth table frames the base case; the terminal-growth steps hold the cost of equity. At the payout Damodaran measures on this page's earnings (0.67) the model's value is 3,998 with the premium held, and 4,243 (−42.8%) re-solved on the headline's rules, where the implied premium falls to 3.14pp.

  • What moved
    Year to date to 16 Nov 2026, earnings added 18.8 log points to the S&P 500 and higher yields took 27.5.

    With distributions reinvested the same window returns +9.4%, the figure the Performance page prints (to 16 Nov 2026). The price moved +8.2 log points (+8.5%) to 16 Nov 2026: earnings +18.8, the 10-year's +94bp −27.5, the premium +16.9 (the implied ERP −0.61pp). Near-term growth is not measured: no consensus capture at both ends (captures on 31 Dec 2025, 31 Mar 2026, 30 Jun 2026, 1 Sep 2026, 11 Sep 2026), so its change sits inside the premium. By window: Year to date: price +8.2, earnings +18.8, rates −27.5, premium +16.9; Since the Strait closed: price +7.6, earnings +13.2, rates −31.3, premium +25.7; Since the 29 Jul FOMC: price +1.4, earnings −1.7, rates −12.4, premium +15.5; Since 11 Sep: price −3.2, earnings −1.5, rates −4.8, premium +3.0; Since the 16 Sep FOMC: price −1.9, earnings −1.5, rates −3.3, premium +2.9.

Next: Statements value — the price on the index's statements

Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Yahoo Finance; Macro Review calculations · data to 16 Nov 2026cross.valuation-read

The market pays 3.50pp for equity risk; the uses 5.17pp

The implied premium since 1970 beside Damodaran's; his premium bridged to this page's one convention at a time; then the two-stage model and every input, with its source, date, kind and range

Exhibit 2 · Implied equity risk premium since 1970, beside Damodaran's

Premium over the long rate that returns each month's price through the two-stage model, pp, monthly since 1970; dot, latest; dashed, Damodaran's; shaded, 2008–09 write-downs.
About this exhibit
Monthly line since 1970: the implied equity risk premium on the valuation history's rules, with the latest reading on this page's inputs, Damodaran's monthly premium (dashed), the full-history median and the 2008–09 write-downs shaded.
The market pays 3.50pp for equity risk, against Damodaran's 4.20%
The each month since 1970, on the valuation history's rules (2.41pp in Nov 2026); today on this page's inputs 3.50pp (16 Nov 2026). Dotted: Aswath Damodaran's monthly implied premium (trailing 12-month cash yield), 3.70% on 1 Oct 2026; his headline figure 4.20%.
Shaded "write-downs": Jun 2007 – Dec 2010. From Jun 2007 the 2008–09 write-downs cut trailing reported earnings by 92% to their trough, so the premium read from them dives; from Jan 2010 the history's model cannot price the month at all (no line). Damodaran's cash-yield series runs through them. Two bases for one idea. This page solves the two-stage model at today's price with consensus near-term growth fading over six years to the Fed's terminal growth, a payout of 0.84 of earnings (measured: dividends plus buybacks over as-reported earnings) and the 10-year Treasury; the history recomputes the same model each month from what was known then (the surveys' terminal growth, the earnings model's near-term growth, Shiller's long rate). Damodaran uses dividends plus buybacks and a five-year consensus against the 10-year T-bond; the levels differ by construction, the movements are the cross-check.
Source: Robert J. Shiller data; Federal Reserve Bank of Philadelphia; U.S. Treasury; Aswath Damodaran, NYU Stern; Macro Review calculations · data to 16 Nov 2026cross.valuation-28

Exhibit 3 · From the implied premium to Damodaran's, one convention at a time

Implied equity risk premium, pp, 16 Nov 2026; each bar is the change when one more input takes Damodaran's value; the bars add up.
About this exhibit
Waterfall in pp from this page's implied premium to Damodaran's: each bar the change when one more input takes his convention.
From this page's 3.50pp to Damodaran's 4.20%, one convention at a time
The premium today's price implies, re-solved as each input moves to convention: terminal growth at the 10-year, his dividends-plus-buybacks yield, his five-year growth, his date (1 Oct 2026), then his own model. Percentage points.
Each bar is the change in the implied premium when one more input takes his value, the others as the bar before; the bars add to the whole move. The first four are solved on this page's two-stage model; the last two are his models on his inputs, which reproduce his published figures. His cash return is 0.67 of this page's trailing earnings (0.73 of his own 2025 earnings; the inputs table reconciles the two), against the 0.84 this page measures. Model-conditional: a reconciliation of two readings of one price, not a forecast.
Source: Aswath Damodaran, NYU Stern; U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Federal Reserve; Macro Review calculations · data to 16 Nov 2026cross.valuation-34

Exhibit 4 · The two-stage model and each input, with its source and range

The formula, then each input at 16 Nov 2026: value (%), kind, source, date and range; below, the model's value at five payout ratios.
P∗=k E0 [(1+gL)+H (gS−gL)]r−gL,H=T2,r=rf+ERPP^{*} = \frac{k\,E_0\,\left[(1+g_L) + H\,(g_S - g_L)\right]}{r - g_L}, \qquad H = \tfrac{T}{2}, \qquad r = r_f + \mathrm{ERP}
InputValueKindSourceDateRange used
Trailing earnings yield (E₀ ÷ P)3.99%observedE₀: Shiller's trailing 12-month reported EPS (the Earnings page's) ÷ the S&P 500's close16 Nov 2026–
Payout (dividends + buybacks ÷ earnings)0.84measuredS&P Dow Jones Indices, the four quarters to 2025 Q3: dividends $665bn + buybacks $1,020bn over as-reported earnings $1,995bn (E₀'s basis), the latest quarter the publisher's preliminary estimate; on operating earnings 0.78; the former convention 0.85 is a sensitivity row; measured by Damodaran: 0.73 in 2025 on his earnings, 0.67 over the last 12 months on this page's E₀ — Damodaran's 0.73 is his 2025 cash return (dividends + buybacks, $198.98) over his 2025 earnings ($271.52); his 0.67 is his trailing 12-month cash return to 1 Oct 2026 ($198.17) over this page's E₀ ($295.52, Shiller's reported EPS): the cash is −0.4%, the earnings +8.8%30 Sept 20250.65 to 0.95 assumed values
Near-term EPS growth, consensus15.1%consensusFactSet Earnings Insight, calendar-2027 EPS growth (a one-year rate)11 Sept 2026–
Near-term EPS growth, the site's5.77%site estimatethe Earnings page's fan, bridged to EPS, a year over two years — from a fan with no edge over a fixed mean16 Nov 2026-4.87% to 17.85% spread of inputs
Transition period to terminal growth6 yearsconventiona convention of this site, not estimated–4 to 10 assumed values
Terminal growth4.0%conventionthe Fed's longer-run projections, real GDP 2.0 + PCE inflation 2.016 Sept 20263.7% to 4.5% assumed values
10-year Treasury yield5.12%observedU.S. Treasury par yield curve16 Nov 20264.87% to 5.37% named paths, no odds
Equity risk premium: the yield-gap premium5.17ppsite estimatetoday's implied premium plus the distance from today's yield gap to its median in the 64 past months when stocks and bonds moved together — an assumption that the gap reverts, not a median premium16 Nov 20264.11pp to 5.74pp spread of inputs
Cost of equity, at the model's premium (10-year + premium)10.29%derivedthe two rows above, added16 Nov 2026–

The headline at each payout: measured, the former convention, and the second source

the headline uses the measured payout (the measured payout: S&P Dow Jones Indices' dividends plus buybacks over as-reported earnings, the four quarters to 2025 Q3 ($1,685bn ÷ $1,995bn = 0.84, the latest quarter the publisher's preliminary estimate); the former 0.85 convention is a sensitivity row (the author's decision of 3 Oct 2026 (D-2026-10-03-12)); on operating earnings the same cash is 0.78. S&P Global's newsroom lists no S&P 500 buyback release after this one (18 Dec 2025), so the payout is that release's, 4 quarters behind the latest completed quarter (the author's decision of 5 Oct 2026 (D-2026-10-05-03))); the former convention of 0.85 is a sensitivity row. Damodaran's 0.73 is his 2025 cash return (dividends + buybacks, $198.98) over his 2025 earnings ($271.52); his 0.67 is his trailing 12-month cash return to 1 Oct 2026 ($198.17) over this page's E₀ ($295.52, Shiller's reported EPS): the cash is −0.4%, the earnings +8.8%.

PayoutkPremium held: the model's valueImplied premium at consensusYield-gap premiumRe-solved: the model's value
Measured: S&P DJI, four quarters to 2025 Q3 (as-reported earnings)the headline0.844,338 (−41.5%)3.50pp5.17pp4,338 (−41.5%)
The former convention, 0.850.854,365 (−41.1%)3.53pp5.20pp4,345 (−41.4%)
S&P DJI, the same quarters on operating earnings0.783,998 (−46.1%)3.14pp4.81pp4,243 (−42.8%)
Damodaran, 2025 (his earnings)0.733,763 (−49.3%)2.89pp4.56pp4,169 (−43.8%)
Damodaran, trailing 12 months (this page's E₀)0.673,444 (−53.6%)2.55pp4.22pp4,059 (−45.3%)

Premium held: the yield-gap premium stays at 5.17pp while the payout moves. Re-solved: the implied premium is solved again at each payout and the yield-gap premium is that plus the same 1.67pp distance, as in the headline. Model-conditional. On the former convention of 0.85 the same inputs give the model's value 4,345 (−41%), an implied premium of 3.53pp and a yield-gap premium of 5.20pp; at the measured 0.84 they give 4,338 (−42%), 3.50pp and 5.17pp. The payout moved on 18 Dec 2025's release being read as measured, by the author's decision of 3 Oct 2026, not on new data.

Fuller and Hsia's H-model: growth starts at g_S and falls in a straight line to g_L over T years, then holds. With g_S = g_L it is the one-stage payout model. Model-conditional: the value is what these inputs imply, not a forecast or a target. The full symbol table is the Appendix's Model tab.

About this exhibit
The two-stage model's formula, then a table of every input with value, kind, source, date and range; below, the headline at each payout ratio.
The two-stage model and its inputs
The model, then every input it uses: its value, what kind of input it is, where it comes from, its date, and the range the sensitivity section moves it across.
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Federal Reserve; Earnings page; Macro Review calculations · data to 16 Nov 2026cross.valuation-30

Slower near-term growth takes 1,512 points off the price and a higher premium 1,566

From today's price to the in index points, one input at a time, and the 's alternatives

Exhibit 1 · S&P 500 price to the model's value, one input at a time

Index points, 16 Nov 2026: the S&P 500, the change as near-term growth then the premium move to the model's values, and the model's value.
About this exhibit
Waterfall in index points: the S&P 500, the step from moving near-term growth to the site's view, the step from moving the premium to the model's, and the model's value.
Where the gap comes from
From the S&P 500 at 7,417 (16 Nov 2026) to the model's value, 4,338: each step moves one input from the price's value to the model's, the others held. Model-conditional.
The order is near-term growth, the 10-year, terminal growth, then the premium; in logs the steps add to the whole move, and in index points they add to the difference between the price and the model's value. Not drawn, because they move nothing: rates (the 10-year) (the 10-year held at 5.12% (±25bp are scenarios, not an expectation)); terminal growth (terminal growth 4.0% in the price and in the model alike). The near-term growth step: near-term EPS growth from the consensus 15.1% (FactSet, operating EPS) to the site's 5.77% a year (reported EPS, from a fan with no edge over a fixed mean): the step carries the accounting basis and the growth view together — splitting them needs a reported-basis consensus, which no source the site reads publishes.
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Yahoo Finance; Macro Review calculations · data to 16 Nov 2026cross.valuation-8

Exhibit 5 · The model's value at each transition period to terminal growth

Index points and gap to the price, 16 Nov 2026: near-term growth fading over 4, 6 (used) and 10 years, the premium re-read at each.
Range: assumed values
6 years (used)4,338−41.5% · premium 3.50pp
4 years4,538−38.8% · premium 3.13pp
10 years4,003−46.0% · premium 4.25pp
About this exhibit
Table: the model's value and gap to the price with near-term growth fading to terminal growth over the transition the page uses and its shorter and longer alternatives, the implied premium re-read at each.
The transition period
The model's value with near-term growth fading to terminal growth over other lengths; the premium re-read from the price at each. The tornado and the calculator hold the premium instead: 4 to 10 years give 4,268 to 4,479.
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Yahoo Finance; Macro Review calculations · data to 16 Nov 2026cross.valuation-9

Near-term growth moves the most, from 3,072 to 5,777

The value by premium and with the base case framed; one input at a time, widest swing first; terminal growth moved with the held

Exhibit 6 · The model's value by equity risk premium and terminal growth

Index points, two-stage model at near-term growth 5.77%, 10-year 5.12%, 6-year transition; framed, base case; outlined, the value nearest the latest price, 7,417.
Equity risk premium ↓Terminal growth →
3.00%3.50%4.00%4.50%5.00%5.12%= the 10-year
3.50pp (the price's premium at consensus growth)4,9435,3775,9046,5607,3977,633
3.67pp4,7995,2065,6976,3037,0697,283
4.17pp4,4184,7565,1595,6456,2456,410
4.67pp4,0924,3784,7135,1115,5935,724
5.17pp3,8124,0564,3384,6705,0645,170
5.67pp3,5673,7784,0194,2994,6274,714
6.17pp3,3523,5353,7433,9824,2594,332
6.67pp3,1613,3223,5033,7093,9454,007

Read across a row: terminal growth moves the value as much as the premium does. The last column sets terminal growth equal to the 10-year, as Damodaran does.

About this exhibit
Grid in index points: the model's value at each equity risk premium (rows) and terminal growth rate (columns); base case framed, the cell nearest the latest price outlined.
The model's value by premium and terminal growth
Index points on the two-stage model, near-term growth 5.77%, the 10-year 5.12%, a 6-year transition. Framed: the base case; outlined: the cell nearest today's price, 7,417. Model-conditional.
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Federal Reserve; Macro Review calculations · data to 16 Nov 2026cross.valuation-35

Exhibit 7 · The model's value with each input at either end of its range

Index points, 16 Nov 2026; one input at a time, others at headline values, premium held; orange, its low end; blue, its high end.
About this exhibit
Tornado: the model's value with each input moved to the low (orange) and high (blue) end of its range, the others at headline values, widest swing first.
What moves the model's value most
One input at a time moved to each end of its range, every other input at its headline value and the premium held; widest swing first. Model-conditional.
The diamonds are the other reading, the page's own steps, which hold fixed the gap between the site's premium and the premium the price implies and re-read the latter at each payout, fade and long-run rate: the long-run rate then enters the priced and the expected state alike and cancels by construction (across the SEP participants' range it moves the gap −41.4% to −41.7% there, against −43.9% to −37.0% with the premium held), the payout ratio's swing shrinks from 20.8 to 6.1pp, and the direction reverses for the long-run rate and the fade; for near-term growth, the 10-year, the term-premium model and the premium the two readings are one number.
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Yahoo Finance; Macro Review calculations · data to 16 Nov 2026cross.valuation-25

Exhibit 8 · The model's value at each terminal growth rate, cost of equity held

Index points and gap to the price, 16 Nov 2026: terminal growth 3.0% to 5.0% around the Fed's 4.0%, cost of equity held at 10.29%.
Range: assumed values
3.0% (−1.0)3,812−48.6% · −7.1 points of gap
3.5% (−0.5)4,056−45.3% · −3.8 points of gap
4.0% (used)4,338−41.5%
4.5% (+0.5)4,670−37.0% · +4.5 points of gap
5.0% (+1.0)5,064−31.7% · +9.8 points of gap
About this exhibit
Table: the model's value and gap at terminal growth rates around the Fed's longer-run rate (the one used marked), cost of equity held.
Terminal growth moved
The model's value with moved around the Fed's longer-run rate, the cost of equity held at 10.29% (the 10-year plus the yield-gap premium, the model's; as priced it is the lower figure in Key data).
Source: Federal Reserve; U.S. Treasury; Robert J. Shiller data via multpl.com; Macro Review calculations · data to 16 Nov 2026cross.valuation-32

The runs from 3,448 to 5,904 across premium bases

The model's value by premium basis, each bar near-term growth's 10th to 90th percentile, the price as the line; then the calculator on your own inputs

Exhibit 9 · The model's value on each premium basis, against the price

Index points, 16 Nov 2026, at the site's near-term growth; dot, centre; bar, growth's 10th–90th percentile; last row, all inputs drawn; line, the S&P 500.
About this exhibit
Range chart in index points: the model's value on each premium basis, each bar near-term growth's 10th–90th percentile, plus all inputs drawn together; vertical line, the S&P 500.
The model's value by premium basis
The model's value in index points on each premium basis, 16 Nov 2026, every row on this page's rules and at the site's near-term growth (from a fan with no edge over a fixed mean); each bar spans that growth's 10th to 90th percentile, the dot and the figure are the centre. Model-conditional.
The same two-stage model and inputs on every row; only the equity risk premium changes. The row a reader should take as the page's reading is the yield-gap premium (today's implied premium plus the yield gap's distance to its median in months when stocks and bonds moved together); the others show how much the answer depends on that choice. Damodaran's premium is not a row: on his own conventions his model returns the index by construction, and the bridge above reconciles his premium to this page's. The last row draws near-term growth and the premium together (the 10th–90th percentile of 20,000 draws, the 25th–75th thick) — the spread of the inputs, not a probability.
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Federal Reserve; Macro Review calculations · data to 16 Nov 2026cross.valuation-29

Exhibit 10 · Calculator: the model's value on inputs you choose

The same two-stage model on inputs you choose, index points; it opens on the page's headline inputs of 16 Nov 2026.

Your inputs — set to the page's headline inputs, so the value below is the headline

Premium basis
Transition period
The model's value on your inputs4,338−41.5% against the S&P 500 at 7,417

the cost of equity sits 6.29pp above terminal growth; a 0.1pp change in either moves the value by about 2%.

Premium the price implies at your near-term growth (5.77%)2.56pp
… at consensus near-term growth (15.1%)3.50pp
Terminal growth the price implies at your premium6.80%
Cost of equity on your inputs10.29%
2,4427,862

Dots: the page's premium bases (Exhibit above); line: the price; diamond: your inputs. Opened on the page's inputs, so the value equals the page's headline; move any control to try your own.

About this exhibit
Calculator: the two-stage model's value, gap, implied premium, implied terminal growth and cost of equity on inputs the reader sets.
Try your own inputs
The same two-stage model on inputs you choose; it opens on the page's headline inputs, and a link to the page carries your inputs.
Source: the page's inputs (U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Federal Reserve); your changes · data to 16 Nov 2026cross.valuation-33

At the site's own growth the price implies a 2.56pp premium

The premium the price implies at the site's growth and at consensus, the it implies on each premium basis, and the

Exhibit 11 · What the latest price implies: premium, terminal growth and cost of equity

16 Nov 2026: implied premium at the site's and consensus near-term growth, pp; terminal growth implied on each basis at consensus growth, %; cost of equity, %.
Implied ERP at the site's near-term growth (5.77%)2.56pp
Implied ERP at consensus near-term growth (15.1%)3.50pp
, as priced8.62%10-year 5.12% + 3.50pp
, at the model's premium10.29%the 10-year + the yield-gap premium
Implied: 3.50pp4.00%
Yield-gap: 5.17pp5.79%
Full-history: 6.64pp7.37%
Yield-gap, apart months: 6.79pp7.53%

The terminal growth the price implies on each premium basis, against the 4.0% the model assumes, with consensus near-term growth. The price fixes one number, the premium less growth, and these are its two readings.

About this exhibit
Two tables: the premium the price implies at the site's and consensus near-term growth, the cost of equity as priced and at the model's premium; then the terminal growth the price implies on each premium basis.
What the price implies
Today's price read back through the same model: the premium it implies at each near-term growth rate, and the terminal growth it implies on each premium basis. A reading of the price, not a forecast.
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Yahoo Finance; Macro Review calculations · data to 16 Nov 2026cross.valuation-20

Year to date, earnings added 18.8 and yields took 27.5

The move over each window in log points: earnings, near-term growth where it is measured, the 10-year, and the premium

Exhibit 12 · S&P 500 price change split into earnings, yields and the premium

Log points (≈ %): earnings, near-term growth, the 10-year and the premium; tabs choose the window, year to date by default, 31 Dec 2025 – 16 Nov 2026
About this exhibit
A waterfall per window (year to date by default; the tabs choose another): the change in the S&P 500's log price split into earnings, near-term growth, the 10-year and the premium.
What moved the price
Year to date: log points × 100, 31 Dec 2025 – 16 Nov 2026; near-term growth not measured.
Δlog price = Δlog earnings + Δlog multiple, and the multiple's move splits into the 10-year's observed change and the residual premium; near-term growth is measured only where a consensus capture exists at both ends of the window, and is otherwise inside the premium.
Source: Yahoo Finance; U.S. Treasury; Robert J. Shiller data via multpl.com; Macro Review calculations · data to 16 Nov 2026cross.valuation-11

The S&P 500 trades at 39.4× earnings (98th percentile)

Multiples and the earnings-yield spreads against their own history — yield spreads, not the implied premium

Exhibit 13 · Valuation measures against their own history

Latest reading (× or %), 20-year mean and standard deviation of monthly readings, and percentile since each series began; to 16 Nov 2026.
About
MeasureLatest20-year meanσPercentileSince
Forward price/earningsFactSet's 19.1× of 11 Sep, moved with the S&P 500's price since (0.977×) · FactSet Earnings Insight, 11 Sep 2026, moved with the index · 16 Nov 2026no monthly history behind it: the mean, the standard deviation and the percentile are not computed, and the quoted averages stand in their placequoted: 5-year average 19.8× (FactSet Earnings Insight, 11 Sept 2026) · 10-year average 19.0× (FactSet Earnings Insight, 11 Sept 2026)18.66×not computed–––
Trailing price/earnings100 ÷ the trailing 12-month earnings yield, monthly · Shiller's trailing reported EPS over the S&P 500's close; history: Shiller data via multpl.com · 16 Nov 2026today's reading is the Earnings page's E₀ over today's close; the statistics are the monthly seriesrange over the whole series 13.50× to 123.46×25.10×25.55×16.2974th2005 (261)
Cyclically adjusted price/earningsreal price ÷ the ten-year average of real earnings, monthly · Robert J. Shiller, Irrational Exuberance data · Nov 2026Shiller's own column; his September price is the 1 September closerange over the whole series 4.78× to 44.20×39.37×27.88×6.1898th1881 (1,751)
Trailing earnings yieldtrailing 12-month GAAP earnings ÷ price, monthly · Shiller's trailing reported EPS over the S&P 500's close; history: Shiller data via multpl.com · 16 Nov 2026range over the whole series 0.81% to 7.41%3.99%4.52%1.2226th2005 (261)
Dividend yieldtrailing 12-month dividends ÷ price, monthly · Robert J. Shiller, Irrational Exuberance data · Sep 2026Shiller's dividend column runs one quarter behind his price column; the ETF-derived basis (SPY's distributions) is the monthly series belowrange over the whole series 1.06% to 13.84%1.07%1.85%0.410th1871 (1,869)
Cash yield (dividends + buybacks)trailing 12-month dividends + buybacks ÷ the index, monthly · Aswath Damodaran, NYU Stern (monthly implied premium workbook) · Oct 2026his series starts in Jan 2009, so the mean and standard deviation run over the months there arerange over the whole series 2.57% to 7.15%2.59%4.35%0.911st2009 (213)
Earnings yield less the 10-yeartrailing earnings yield − the 10-year par yield, monthly averages · Robert J. Shiller (trailing earnings) and U.S. Treasury par yields · 16 Nov 2026range over the whole series -2.91pp to 5.44pp-1.14pp1.62pp1.683rd2005 (261)
Earnings yield less the real yield (10-year TIPS)trailing earnings yield − the 10-year real par yield, monthly averages · Robert J. Shiller (trailing earnings) and U.S. Treasury real par yields · 16 Nov 2026range over the whole series -1.05pp to 7.33pp1.07pp3.69pp1.725th2005 (261)

The trailing multiple and the earnings yield are one number read two ways, so their percentiles are mirror images by construction. The forward multiple is FactSet's of 11 September, moved with the index since, and has no monthly history here: its mean, standard deviation and percentile are not computed, and the quoted five- and ten-year averages stand in their place. Shiller's dividend column runs a quarter behind his price column, so that row is dated its own month.

About this exhibit
Table of valuation measures (forward and trailing P/E, CAPE, earnings, dividend and cash yields, two yield spreads): latest, the long mean and standard deviation, and the percentile since each began.
Valuation measures against their own history
Today's reading, the mean and standard deviation of the last 20 years of monthly observations, and the percentile of today's reading since each series began (Damodaran's cash yield since his series began, Jan 2009).
Source: Robert J. Shiller, Irrational Exuberance data; multpl.com; FactSet; Aswath Damodaran, NYU Stern; U.S. Treasury; Macro Review calculations · data to 16 Nov 2026cross.valuation-19

Exhibit 14 · Trailing yield spreads over the 10-year since 2005, by basis

Percentage points, monthly, Jan 2005 – Sep 2026: trailing earnings yield less the 10-year Treasury and TIPS, and SPY distributions plus 4.0% less the 10-year.
About
About this exhibit
Three lines, monthly since Jan 2005: trailing earnings yield less the 10-year Treasury, less the 10-year TIPS, and SPY distributions plus the long-run rate less the 10-year.
The premium on the trailing bases
Percentage points, monthly Jan 2005 – Sep 2026: earnings and distributions already reported, less the 10-year. The forward bases are on the card beside this one — quarter-end readings over a far narrower range, which this axis flattened onto the floor.
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Yahoo Finance; Macro Review calculations · data to 16 Nov 2026cross.valuation-1

Exhibit 15 · Forward yield spreads over the 10-year at each quarter-end capture

pp, captures of FactSet's forward P/E, 31 Dec 2025 to 11 Sep 2026: payout basis at 4.0% growth; forward earnings yield less the 10-year.
About
About this exhibit
Two series at the author's quarter-end captures of FactSet's forward P/E: the payout basis and the forward earnings yield less the 10-year.
The premium on the forward bases
Percentage points at each of the author's quarter-end captures of FactSet's forward P/E, 31 Dec 2025 – 11 Sept 2026. Split from the trailing bases beside it: a handful of readings inside a fraction of their range, which is not a span one axis can carry with them. The payout basis is at g = 4.0%, the one long-run rate; the issue printed it at 3.0% (the reconciliation fold).
Source: U.S. Treasury; Robert J. Shiller data via multpl.com; FactSet; Yahoo Finance; Macro Review calculations · data to 16 Nov 2026cross.valuation-2

Where do other assets sit?

Real yields, credit spreads, gold and the dollar against their own histories

Exhibit 16 · Real yield and credit spread, gold and the dollar against their history

Latest reading (%, bp or $), its date, and its percentile since the start shown; daily or monthly as each row states.
ReadingLatestDatePercentileSinceWhere
10-year real yield (TIPS par)U.S. Treasury, daily par real yield curvepercentile of daily readings2.92%16 Nov 2026100th2003as on Treasury curve
Corporate spread, 10-year HQM less 10-year TreasuryU.S. Treasury HQM corporate bond yield curve and Treasury yields (the credit page's own series)percentile of monthly readings; high = credit cheap78bpSep 202620th1990as on Corporate spreads
Gold (GLD), priceYahoo Finance GLD close (level cross-checked with the Performance page's)percentile of daily closes since the fund's first$352.9016 Nov 202695th2004this page
Gold against the 10-year real yield: 156-week betaYahoo Finance GLD and Treasury real par yields, weekly (Friday) changesGLD's weekly log return per percentage point of weekly change in the real yield; correlation −0.16; over all 1147 weeks since 2004 the beta is −6.2 (correlation −0.28)−4.3% per pp13 Nov 202676th2007this page
Nominal broad dollar (H.10, goods and services)Board of Governors, H.10 (level cross-checked with the Performance page's)Jan 2006 = 100; percentile of daily readings122.6513 Nov 202693rd2006this page

Gold's rolling 156-week beta has run from −14.8 to −1.1% per percentage point since 2007, negative in 100% of windows; today it is −4.3 (correlation −0.16), against −6.2 over all 1147 weeks since 2004. The real yield and the spread carry their own pages' dates, which can trail this page's. Not in the table: real broad dollar (the Board's H.10 files the site reads carry the nominal broad index only; no real series is published there, and the site reads no other permitted source for it); sector forward P/E against history (no permitted source for sector forward earnings: FactSet Earnings Insight is transcribed for the index only).

About this exhibit
Table of readings (10-year real yield, corporate spread, gold, gold's beta to the real yield, broad dollar): latest, date and percentile in its own history.
Across assets: where each sits in its own history
Today's reading and its percentile since the date shown. Gold is also read against the real yield: the least-squares beta of GLD's weekly log return on the weekly change in the 10-year real yield over the last 156 weeks, and that beta's percentile in its own rolling history. Descriptions of what has happened, not signals.
Source: U.S. Treasury (par, real par and HQM corporate yields); Yahoo Finance (GLD); Federal Reserve Board (H.10); Macro Review calculations · data to 16 Nov 2026cross.valuation-21

Appendix

The model, the , identification, the decomposition, bases, carry-forward and attribution278 checks, all pass
About this page
The page reads the S&P 500 the way a research report's valuation section does. It leads with the : the return over the 10-year Treasury that makes a two-stage fair-value model return today's price. It then sets out the model and every input, the at the and where the gap to the price comes from, how sensitive that value is, the value on each premium basis beside a calculator for your own inputs, what the price implies, what moved it, and where the market trades against its own history. The index's earnings are Robert J. Shiller's trailing reported EPS for the latest quarter, the same E₀ the Earnings page prints; FactSet's consensus is the author's 11 September 2026 capture, moved with the index since, shown with its attribution. Every value is model-conditional — what these inputs imply, not a forecast or a target — and whether the gap has predicted returns is tested on the valuation history page.

The two-stage fair-value model

P∗=k E0 [(1+gL)+H (gS−gL)]r−gL,H=T2,r=rf+ERPP^{*} = \frac{k\,E_0\,\left[(1+g_L) + H\,(g_S - g_L)\right]}{r - g_L}, \qquad H = \tfrac{T}{2}, \qquad r = r_f + \mathrm{ERP}
gap=P∗P−1and, solved at P∗=P:ERPimplied=gL+k E0 [(1+gL)+H(gS−gL)]P−rf\text{gap} = \frac{P^{*}}{P} - 1 \quad\text{and, solved at } P^{*} = P:\quad \mathrm{ERP}_{\text{implied}} = g_L + \frac{k\,E_0\,[(1+g_L) + H(g_S - g_L)]}{P} - r_f
SymbolMeaningFromAssumed or estimated
E0/PE_0/Ptrailing 12-month reported earnings ÷ price, 3.98%Robert J. Shiller's workbook (the Earnings page's E₀) and the S&P 500's close–
kkpayout: dividends + buybacks over as-reported earnings, 0.84 measured over the four quarters to 2025 Q3 (on operating earnings 0.78; Damodaran measures 0.67–0.73; the former convention 0.85 is a sensitivity row)S&P Dow Jones Indices' buyback release (18 Dec 2025), the author's decision of 3 Oct 2026 (D-2026-10-03-12)–
gSg_Snear-term EPS growth, 5.77% (consensus 15.1% for the implied premium), from a fan with no edge over a fixed meanEarnings page (fan, bridged): its 20 one-year forecasts 2005–2024: error 49.8 against 49.4 for a fixed mean, log score −12.37 against −7.45; FactSetestimated
gLg_Lterminal growth, 4.0%Fed longer-run projections (SEP)assumed
TTtransition period, 6 years (H = 3)a convention of this siteassumed
rfr_f10-year Treasury yield, 5.12%U.S. Treasury–
ERP\mathrm{ERP}equity risk premium: 3.50pp as priced, 5.17pp the yield-gap premium (the model's premium)this pageestimated
rrcost of equity, 8.62% as priced, 10.29% at the model's premiumthis page–

Fuller and Hsia's H-model: growth starts at g_S and falls in a straight line to g_L over T years, then holds. With g_S = g_L it is the one-stage payout model. Model-conditional: the value is what these inputs imply, not a forecast or a target.

Exhibit 17 · The identity, in words

A share price is a claim on earnings, discounted: price = earnings ÷ (discount rate − growth), where the discount rate is the risk-free yield plus the extra return investors ask for holding equity rather than a Treasury — the premium.

The market publishes one price, so it fixes one number and no more: the premium less growth. A larger premium and faster growth leave the price exactly where it was, so that single combination — what the price fixes — is all a price can tell you.

This page holds its own view of each of the four: earnings, growth, the discount rate and the premium. Set the market's four against them, and the move in the index that would take it to the site's fair value is the gap to fair value: negative means the index sits above fair value. The table below reads that move one factor at a time, and then all four together.

                    earnings
price  =  ────────────────────────────
          risk-free + premium − growth
                      └── one gap ───┘

  the market's price      ──▶ ┐
                              ├──▶  the gap to fair value: the move that closes it
  the site's own view     ──▶ ┘

The tables write the same identity as a multiple: PE = k ÷ (risk-free + premium − growth), with k the share of earnings paid out. Turned round, k ÷ PE is the discount rate less growth — so k ÷ PE minus the 10-year Treasury is what the price fixes, and it is the one quantity the market's price hands over.

Where the earnings view comes from. The expected-earnings row is not computed here: it is the output of Earnings, whose top-down band and process fan are read back through the manifest and checked against that page at every update; the same page's one-step prediction errors set the earnings leg of the joint draws further down.

Exhibit 18 · Why it is set up this way

An identity, not a model. The relation above is an accounting statement: given a price, earnings, a discount rate and a growth rate, it holds by construction, and nothing in it is estimated. A model would add a rule for what the price ought to be and put its own errors between the reader and the market's. Keeping to the identity means every disagreement with the market has a named home — earnings, growth, the discount rate or the premium — and nothing is left over: in logs the four steps add up to the whole move. What this page adds is a view of each factor, stated separately and sourced to the page that produces it, rather than one verdict on whether the index is cheap.

A band, not a point. Because the price identifies only the premium less growth, neither of the two is known on its own. The page never chooses between them: it reads the same gap from both sides, prints both readings, and counts the gap once — which is why the growth row and the premium row carry the same effect. That is also why the answer is a range, and every range says what kind it is. The input range draws the two inputs the site estimates — near-term growth from the Earnings page's fan and the premium from past months like this one — and its 10th–90th percentile is the range the page prints: the spread of those inputs carried through the identity, not a probability that the fair value lies inside it. The assumption range sweeps the conventions as well, the payout ratio, the fade and the long-run rate, and is not a probability. The corner box beside the fair-value table, the lowest and highest gap over the corners of three ranges, is a box of combinations. None of them is a forecast; each is model-conditional.

Exhibit 19 · Definitions

How the numbers are made, step by step

  1. Where the market trades. Source: the September review's 11 Sep captures of the S&P 500 trailing earnings yield (Shiller data via multpl.com) and FactSet's forward P/E; SPY's closes and distributions (Yahoo Finance); Robert J. Shiller's workbook. Transform: carry the captured yield and multiple with the S&P 500's price from the day of the capture's close; each measure against its own monthly history. Check: the September review's premium series, quarter-end table, forward premium and matrix reproduced; Shiller's columns parsed and his CAPE reproduced (23 checks, all passing).
  2. The premium, in its regime. Source: U.S. Treasury par and real par yields; SPY and TLT closes. Transform: the premium on each basis; the sign of the 63-day stock–bond correlation picks the same-sign months; today's implied premium plus the yield gap's distance to its median there is the yield-gap premium. Check: the correlation equals the Correlations page's; the same-sign and other-sign months partition the sample (3 checks, all passing).
  3. The discount rate. Source: the Inside the 10-year, Term premia, Policy rules and What is priced pages (copies); the Treasury curve page's 10-year par yield. Transform: rf = expected policy path + term premium; r = rf + equity premium. Check: every copy against its source and the identities between the parts; one ACM series named on each of its two dates (19 checks, all passing).
  4. Growth, near term and long run. Source: the Earnings page's fan (a copy); the FOMC's September projection table. Transform: the fan's eight-quarter median bridged to EPS and annualised; the SEP's longer-run real GDP and PCE inflation medians added. Check: the bridge reproduces the Earnings page's median; one long-run rate on both pages (12 checks, all passing).
  5. The two-stage model and the one gap. Source: steps 1–4. Transform: core.identity: the premium the price implies at the consensus near-term rate, then the model price at the site's expected inputs; the gap in four logged steps. Check: the priced state returns today's price to 1e-12; the reduction to the one-stage identity; the steps add to the total (19 checks, all passing).
  6. Sensitivity and the two ranges. Source: steps 1–5 and the declared ranges in the page's specification. Transform: seven one-way swings (the tornado), with the premium held and on the page's own steps; 20,000 joint draws for the input band and the input + assumption band; the long-run growth × r grid. Check: every bar through the headline in both readings; every endpoint through core.identity to 1e-12; all-middle draws return the headline; R = I exact; quantiles ordered; every range names its kind (22 checks, all passing).
  7. The one-stage objects and the context. Source: the New York Fed's term-premium file; the SEP funds rate; H.10; GLD; the sentiment pages. Transform: the one-stage distribution (an AR(1) 10-year, a Gaussian copula); what moved the price; across assets; the price-of-risk strip. Check: the AR(1) residuals, the copula matrices and R = I within three standard errors; the what-moved identity; every copied reading (58 checks, all passing).
  8. The report section and the cross-check. Source: steps 1–6; the valuation history's implied premium; Aswath Damodaran, NYU Stern. Transform: fair value by premium basis, the bridge in index points, the premium × 10-year table, terminal growth moved, what the price implies, the assumptions and the calculator's defaults; Damodaran's monthly premium and measured payout beside them. Check: every row through core.identity to 1e-12; the headline row is the headline; his own figures reproduce from his workbooks (59 checks, all passing).
  9. Copies from other pages. Source: the other pages' published data, through the site's manifest. Transform: each copied value read from the published version of its page. Check: each source published with its checks passing and still current (42 checks, all passing).
  • Tornado
    One input at a time.

    The tornado is model-conditional: each bar is the headline gap recomputed on the same two-stage model with one input moved to the ends of its range and every other input at its headline value. The bars are drawn, and ranked, with the site's expected premium held at its level (5.17pp on the two-stage basis), so the cost of equity moves only with the 10-year and the premium: payout ratio −55.0% to −34.2%, long-run growth −43.9% to −37.0%, fade −42.5% to −39.6%. Beside each bar is the page's own-steps reading, in which the premium the price implies is re-read at each payout, fade and long-run rate and the regime gap is carried: there the long-run rate enters the priced and the expected state alike and largely cancels (it moves the gap only −41.7% to −41.4%), the payout moves it −45.8% to −39.7%, and the direction reverses for the long-run rate and the fade. For near-term growth, the 10-year, the term-premium model and the premium the two readings are one number. Each bar is labelled with the kind of range behind it: near-term growth (the fan's 10th–90th) and the premium (the same-sign months' interquartile range) are input ranges; the 10-year's ±25bp is the page's scenario step; the term-premium model is a model range — the 10-year as the New York Fed's and the Board's models fit it on 6 Nov 2026 (ACM 4.864% = 4.22% expected path + 0.64pp premium; Kim–Wright 4.887% = 4.01% + 0.87pp; par 4.86%), carried to today by each fit's distance from par; the models split the 10-year differently but fit nearly the same yield, and the identity discounts at the yield, so this bar is short; the payout ratio (0.65–0.95), the fade (4–10 years) and the long-run rate (3.7–4.5%, the lowest and highest FOMC participant's longer-run real GDP plus PCE inflation in the 16 Sep 2026 projections) are declared assumption ranges, printed as such and never called probabilities. Checks: every bar passes through the headline in both readings; every endpoint is recomputed through core.identity and through the Monte Carlo's arithmetic, to 1e-12; the two readings are one number wherever the input does not enter the implied premium; the fade bar's own-steps ends are the fade table's rows and the premium bar the fair-value table's premium band.

  • Ranges
    The two-stage Monte Carlo: an input range and an assumption range.

    The two-stage Monte Carlo is model-conditional, not a forecast. 20,000 draws (seed 20260925) of five inputs through the same identity and the page's own steps (the tornado's diamonds): near-term growth from the Earnings page's fan — its 399-point percentile grid of 8-quarter profits growth, bridged to EPS by 2 × −0.92 log points and annualised, drawn by inverse transform — and the premium from the 64 same-sign months (the empirical distribution, drawn by rank); the 10-year stays at today's par. That is the input band — the spread of the two estimated inputs carried through the identity, not a probability that the fair value lies inside it — (10th–90th percentile −60.7% to −13.9%, median −40.1%; in index points 2,912 to 6,386). The input + assumption band adds the payout ratio, the fade and the long-run rate, each drawn uniformly across its declared range — uniform across each declared range — a sweep of conventions, not a belief about them — giving −64.5% to −14.1%; it is an assumption range, not a probability. Variance shares: each input drawn alone with the others at their middle values, divided by the band's variance — in the input band near-term growth 62.1%, the premium 31.0%, interaction 6.9%; with the assumptions, the payout 0.7%, the fade 1.0% and the long-run rate 0.0%. Drawn instead through a Gaussian copula at the correlation of earnings innovations and premium changes over all quarters (+0.08, the one-stage copula's own matrix), the input band is −60.0% to −15.2%. With the site's expected premium held at its level, as the tornado's bars hold it, the input + assumption band is −64.9% to −12.0% and the shares become the payout 7.5%, the fade 0.1% and the long-run rate 0.8%. Checks: every draw at its input's middle value returns the headline to 1e-12; with R = I the copula returns the independent draws exactly (the same normal draws); each input drawn alone at the tornado's endpoints returns that bar's own-steps ends; the quantiles are ordered; no more than 2% of draws stop at the floor (r − g_L ≤ 0.5pp). The Monte Carlo standard error of the input band's ends is about 0.16 and 0.28pp.

  • Long-run grid
    Long-run growth × required return.

    The long-run growth × cost of equity grid is model-conditional, on the same two-stage model as the headline with near-term growth at the site's expected 5.77%, k = 0.8447933585995308, a 6-year fade and the trailing earnings yield 3.98% held; each cell is the gap to fair value at that long-run rate and cost of equity. Unlike the tornado, the cost of equity is the grid's own axis, so a cell does not re-read the premium. The expected cell (g_L 4.0%, r 10.29%) returns the headline and each row's break-even long-run rate — the g_L at which the index is at fair value, from core.identity's h_model_g_long — returns today's price (both checks); the SEP participants' range (3.7–4.5%) is marked.

  • Measures
    The table at the top.

    The measures table is a description, not a model. Each row prints today's reading, the mean and standard deviation of the last 20 years of monthly observations (240 months) and the percentile of today's reading in the whole monthly series, and names the series' first month. Sources: the forward multiple is FactSet's 11 Sep capture carried with the S&P 500's price, and it has no monthly history here — its mean, standard deviation and percentile are not computed and the quoted five- and ten-year averages (19.8× and 19.0×, FactSet Earnings Insight, 11 Sep 2026) are printed as quotes in their place. The trailing multiple and the earnings yield are one number read two ways (their product is 100, a check) from the Shiller/multpl trailing earnings yield, monthly since 2005; today's reading is the Earnings page's E₀ (Shiller's trailing reported EPS for its latest quarter) over today's close. The cyclically adjusted multiple and the dividend yield are Shiller's own columns (workbook read on 17 Nov 2026): the first is real price ÷ the ten-year average of real earnings through the prior month, which a check reproduces from the workbook in every month Jan 1881 – Nov 2026; the second is trailing 12-month dividends ÷ price, monthly since 1871, and Shiller's dividend column runs a quarter behind his price column (Sep 2026). The two premium rows are the trailing earnings yield less the 10-year par and the 10-year real par (Treasury), monthly averages since 2005. Workbook note: Nov price is Nov 12th close; Oct/Nov CPI estimated; Nov GS10 is Nov 12th value.

  • Two-stage
    The H-model, and what it reduces to.

    The two-stage gap to fair value is a model output and conditional on this model: it is what the index would do if the model's inputs replaced the market's, not a forecast of what the index will do; whether the gap has said anything about the returns that followed is tested on the valuation history page, and this page makes no claim about it. The identity is the H-model — growth starts at a near-term rate g_S and fades linearly to a long-run rate g_L over 6 years, so with H = 3 (half the fade, Fuller and Hsia) the price of trailing earnings E0 at a payout k and a discount rate r is P = k·E0·[(1 + g_L) + H·(g_S − g_L)] ÷ (r − g_L). E0 is observed (the trailing earnings yield, 3.98%: Shiller's trailing reported EPS, the Earnings page's, over the S&P 500's close), so the model needs no index level, and the gap to fair value is defined on price, ln(P_model ÷ P_actual) — the consensus forward multiple never meets the model's own. Inputs: k = 0.8447933585995308; g_S = the Earnings page's default fan of NIPA profits after tax: median 8-quarter profits growth 13.95% (13.06 log points), bridged to EPS by 2 × −0.92 = −1.84 log points to 11.87%, annualised as (1 + 0.1187)^(1/2) − 1 = 5.77% a year (10th–90th −4.87 to 17.85%), against the consensus 15.1% the price is read at — a one-year figure, and the model treats each as the rate the fade starts from; g_L = 4.0% from the September SEP; the fade length is a convention of this site, not estimated — printed as such; r = the 10-year plus the premium. Read at today's price the identity implies a premium of 3.50pp (or, at the expected premium, a long-run rate of 5.79%: this is the two-stage model; the one-stage identity's implied growth is 5.67% because it capitalises forward earnings at 18.7× at one rate for ever, while this one grows trailing earnings at the consensus 15.1% fading over 6 years, a lift of ×1.337 against the ×1.345 the forward multiple already assumes (forward ÷ trailing earnings), so it needs more long-run growth) — one identified gap read two ways, as on the one-stage table, and the two rows' effects are equal by construction. Reduction: with g_S set to g_L the bracket collapses to (1 + g_L) and P ÷ (E0(1 + g_L)) = k ÷ (r − g_L), the one-stage identity — a check on every update, to 0.0 at the page's three discount rates. Any input combination that brings r within 0.5pp of g_L stops that cell and prints the reason instead of an unbounded multiple. Sensitivities: the fade at 4 and 10 years beside the 6 the table uses, and the one-stage gap to fair value at g = 3.0 (−37.6%) and 5.28 (−8.7%). The one-stage values stay on the table beside the two-stage ones so the change is visible rather than silent.

  • Discount rate
    rf = expected policy path + term premium; r = rf + ERP.

    The discount rate is the identity's least examined input, and this page now assembles it from its parts rather than taking today's 10-year as given. The risk-free rate is the Treasury curve page's rf, the 10-year par yield (5.12% on 16 Nov 2026), copied and held to this page's own Treasury file. Inside the 10-year splits the zero-coupon 10-year (4.88% on 6 Nov 2026) into the expected policy path, 3.12% (the expected real short rate 0.82% plus expected inflation 2.30%), and the term premium on 6 Nov 2026, 1.75pp, anchored on the surveys; the four models it sets beside each other put the path at 3.12–4.22% and the premium at 0.64–1.75pp — a model range, not a probability. The term-premium page splits it into the average short rate the market expects over the ten years and the compensation for holding duration: on the New York Fed's model 4.39% plus 0.72pp, a fitted 10-year of 5.111% on 13 Nov 2026 (the Board's model: 4.01% plus 0.87pp on 6 Nov 2026). The rules page prices the hold path — hold at 3.90% through 2028, then the SEP's longer-run 3.2% from Jan 2029 — at an average short rate of 3.340% and adds the same term premium for a fair 10-year of 4.065%, the value What is priced reads; that is −105bp from today's yield. The equity premium is the page's own: as the two-stage model prices it (3.50pp), on the payout basis at the one long-run rate (3.41pp), on the trailing basis (−1.14pp over Treasuries, +1.06pp over TIPS) and the yield-gap premium (5.17pp: today's implied premium plus the distance from today's yield gap to its median in the 64 months since 2005 whose stock–bond correlation had today's sign). r = rf + ERP: 8.62% as priced, 10.29% at the model's premium, 7.57% on the hold path's fair 10-year at the priced premium; with everything else where the price has it, a 10-year at the hold path's fair value would move the model price +29.6%, and at the expected inputs the gap to fair value would be −29.7% instead of −41.5%. Every copied part is compared with its source page's data on each update, and the identities between the parts (fitted = expected + premium; fair = average + premium; one term premium on all three pages) are checks.

  • Implied growth
    The price read as growth.

    Implied growth is a reading of today's price under one identity and its inputs, not a forecast of growth. It inverts the payout identity the page already uses: g = 10-year + premium − k ÷ forward P/E, so at today's 5.12% 10-year, the yield-gap premium 5.08pp, the 0.8447933585995308 payout and the carried 18.66× multiple the price implies 5.67% nominal growth a year, for ever. It is a reading of the price, not a forecast: nothing estimates it, and it moves one-for-one with whatever premium is assumed. It is printed against the single long-run rate the site uses, 4.0% — the September SEP's longer-run medians for real GDP (2.0) and PCE inflation (2.0) added, published 16 Sep 2026 — which makes the gap +1.67pp. This is the one-stage identity (forward earnings at 18.7×, one rate for ever); the two-stage model's long-run rate is 5.79% because it grows trailing earnings at the consensus 15.1% fading over 6 years, a lift of ×1.337 against the ×1.345 the forward multiple already assumes (forward ÷ trailing earnings), so it needs more long-run growth.

  • Sensitivity grid
    Near-term growth × required return.

    The sensitivity grid is model-conditional, on the same two-stage model as the headline: for each near-term growth rate (across) and cost of equity r (down) it prints the gap to fair value ln(model price ÷ index price) with k = 0.8447933585995308, trailing earnings yield 3.98%, the long-run rate 4.0% and a 6-year fade held fixed. The priced cell (growth 15.1%, r 8.62%) returns zero by construction and the expected cell (growth 5.77%, r 10.29%) returns the headline gap — both checks. A cell whose r comes within 0.5pp of the long-run rate is left blank rather than printed as an unbounded multiple. The ERP × 10-year matrix beside it is the one-stage fair multiple at each premium and yield; the gap as a distribution replaces the corners with draws.

  • One gap
    How the other gap figures reconcile.

    One gap. The page prints the two-stage model's gap to fair value (−41.5%) and no other as its answer. It still computes the figures earlier issues printed side by side — the one-stage identity at the long-run rate (−27.6%) and at its two sensitivities, the premium step alone (−26.5%) and the joint distribution's four-quarter median (−17.0%) — and the Detail fold's reconciliation table says, for each, what differs from the headline: the identity (one growth rate for ever, or a near-term rate fading to it), the growth input (the regression's band, the fan, or a sensitivity), the horizon (today's price, or four quarters of draws) and the premium's basis. A check ties each row to the block it comes from.

  • Gap to the model's value
    Four factors, one identity.

    The gap to fair value (the buffer in earlier issues): four factors, one identity. PE = k ÷ (rf + ERP − g) with k = 0.8448 (the measured payout, S&P DJI's dividends + buybacks over as-reported earnings, the four quarters to 2025 Q3; the September issue used its 0.85 convention) and g = 4.0 (the one long-run rate), PE_now the carried forward multiple and rf the 10-year. The identity fixes one number, the premium less growth: ERP − g = k ÷ PE_now − rf. Each row has a priced value, an expected value with a band and the index change if that factor alone converged to its expected mid (the others at their priced values): earnings (1 + E_exp) ÷ (1 + E_priced) − 1, the others through the fair multiple ÷ PE_now − 1. Sign: negative = the index sits above fair value; positive = below it (a cushion).

  • Expected
    What the site expects.

    Expected values (the one-stage table): earnings = the top-down band of the Earnings page, with its default fan (NIPA profits bridged to EPS, calendar 2027 growth) as the alternative view on the same row (4-quarter EPS growth at Q4 2027 under nominal GDP growth of 4, 5 and 6%; consensus CY2027 from FactSet, quoted on 11 Sep 2026, is the priced value); the discount rate = today's 10-year with ±25bp scenarios — Scenarios by design; the policy-path × term-premium grid on Policy rules is the tool for other paths. Growth = the one long-run rate 4.0 (SEP), with the sensitivities 3.0 and 5.28 as the band; the premium = the yield-gap premium: today's payout-basis premium plus the distance from today's yield gap (the trailing earnings yield less the 10-year) to its median and quartiles over the complete months since 2005 whose month-end 63-day stock–bond correlation had today's sign — an assumption that the yield gap reverts, not a median premium (the payout basis is only observed at the author's capture dates). Same-sign months are listed in the hover note.

  • Identification
    Growth and the premium.

    Growth and the premium are identified only jointly. The premium row holds growth at the long-run rate and reads the gap as a premium (+3.41 priced vs the yield-gap premium); the growth row holds the premium at that level and reads the same gap as growth (5.67% priced vs 4.0). Their effects are equal by construction and the joint waterfall counts the gap once, at the premium step (the growth step is zero because the long-run rate is the priced growth at the payout premium).

  • Joint
    The waterfall and its band.

    The joint gap to fair value is a sequential waterfall in logs, earnings → discount rate → growth → premium, from the priced state (consensus earnings, today's 10-year, g = the long-run rate 4.0, the payout premium) to the expected state; the parts sum to the total exactly (a check). The band is the minimum and maximum of the total over the 2⁴ corners of the four bands (each input at its lo or hi, the others free), printed as such — it is not a confidence interval.

  • Distribution
    The one-stage gap drawn jointly.

    The gap-to-fair-value distribution is a model-conditional object, not a forecast. 20,000 joint draws at 4 and 8 quarters of three things, drawn independently in the headline object and through a Gaussian copula beside it (below): (i) cumulative earnings growth G_h, resampled by inverse transform from the Earnings page's default fan — NIPA profits after tax, bridged to EPS (its percentile curve of cumulative profits log growth at a 0.25% step, 20,000 simulated paths from Apr 2026, shifted by the Earnings page's bridge −0.92 log points a year: one bridge at four quarters, two at eight; forward earnings are assumed to grow at the same rate); (ii) the 10-year yield from an AR(1) on 147 quarter-end par yields Jan 1990 – Jul 2026, y_t − ȳ = a (y_(t−1) − ȳ) + e_t with a = 0.9479 (half-life 13.0 quarters), residual s.d. 0.491pp and ȳ = 3.92% = the SEP longer-run funds rate (3.2) plus today's ACM 10-year term premium (0.72pp, 13 Nov 2026), started at today's 5.12%; (iii) the premium, the yield gap resampled from its 64 same-sign months and carried to the payout basis by today's gap, exactly as the gap-to-fair-value table's premium row does. g is the one long-run rate 4.0 in the fair multiple; gap_h = (1 + G_h) × k ÷ (rf_h + ERP − g) ÷ PE_now − 1. Draws with rf + ERP − g at or below 0.5 are dropped (0 of 20,000 at 4 quarters), the page's existing convention near r = g. Sensitivities: ȳ with the 2015–19 ACM average instead of today's (2.91%), and g moved alone, the premium held, to 3.0 (median −29.2%) and 5.28 (median +6.5%) at 4 quarters.

  • Correlation
    The Gaussian copula and the two matrices.

    Correlated draws: the same three marginals are drawn through a Gaussian copula — Z ~ N(0, R), U = Φ(Z), each factor read off its own marginal at U — so only the dependence changes. R is the correlation of three quarterly innovations since 2006: the one-step prediction errors of the earnings process's default fit on NIPA profits (copied from the Earnings page), the quarterly change in the 10-year par yield, and the quarterly change in the payout-basis premium (the trailing basis carried to the payout basis by a constant gap, so the change is the same number). Two matrices are printed: all 82 quarters Jan 2006 – Apr 2026 (earnings–rate 0.29, earnings–premium 0.08, rate–premium −0.48), and the 8 of them whose quarter end sat in today's correlation sign-rule state (bonds with stocks, oil against; earnings–rate 0.01, earnings–premium 0.18, rate–premium −0.83) — 8 quarters is a thin sample and the matrix is printed as such, not smoothed. Both matrices are positive definite as estimated (no nearest-positive-definite correction was needed); the quadrant rule is the correlations page's own (stocks–bonds positive and stocks–oil negative at the quarter end: bonds with stocks, oil against). At 4 quarters P(gap < 0) is 74.0% independent, 84.8% on all quarters and 92.5% in the quadrant; at 8 quarters 62.1%, 68.7% and 74.3%. Check: with R = I the copula reproduces the independent draws' 5th, 50th and 95th percentiles inside three Monte Carlo standard errors of the difference, each standard error derived as √(p(1−p)/n) ÷ f(q) on the two samples and printed in the check. The premium's marginal is the pool of same-sign months, so the copula ties the rank of the premium draw to the ranks of the other two — the correlation is estimated on changes and imposed on levels, which is the assumption the two matrices make visible.

  • Grid
    The percentile grid the exhibit reads.

    The exhibit reads a 99-point percentile grid (the 1st to the 99th percentile of the gap to fair value, stored to four decimals) for the independent draws and for each correlation matrix, so the page draws a density and a cumulative curve without carrying 20,000 draws. Checks: every grid is non-decreasing and its 50th point equals the stored median.

  • Growth convention
    One long-run rate, and the one-stage g.

    The gap-to-fair-value distribution runs on the one-stage identity at the one long-run rate g = 4.0; on the same draws, with the premium held, the median gap to fair value at 4 quarters is −17.0% there and −29.2% at g = 3.0 and +6.5% at g = 5.28 — the September review's g and the trailing-series fan's anchor held for ever, which the one long-run rate does not allow. Both are sensitivities. g in PE = k ÷ (rf + ERP − g) is growth held for ever; the site's one long-run rate (4.0, nominal GDP growth with no wedge) is the base of both identities on this page. The sensitivities are the September review's g (3.0) and the trailing-series fan's anchor held for ever (nominal GDP plus the 1990–2019 EPS wedge); whether they stay the sensitivity set is the author's decision.

  • Variance
    Where the spread comes from.

    Variance decomposition: each factor is simulated alone with the other two at their medians and its variance divided by the variance of the joint gap to fair value. At 4 quarters earnings growth 14.4%, the premium 33.5%, the discount rate 26.3%, interaction remainder 25.9%; at 8 quarters 14.3%, 16.8%, 28.3% and 40.6%. The shares need not add to one: the gap to fair value is a product, so the factors interact, and the remainder is printed rather than distributed.

  • Moved
    What moved expectations.

    What moved expectations: the trailing-basis discount spread rf + ERP − g = k ÷ PE_trailing is only about 3%, so a yield change of tens of basis points is a large share of it and the rates and premium parts are large and offsetting; they are the identity's arithmetic, not estimates. Over each window Δlog P (SPY, unadjusted close) = Δlog E + Δlog PE with E the trailing earnings index (Shiller/multpl earnings yield × the month's average price; the current month is the author's 11 Sep estimate — earnings enter at monthly frequency), and Δlog PE = −Δlog(rf + ERP − g) on the trailing basis with k the measured payout: the 10-year's change is observed, Δg = 0 by convention and ΔERP is the residual that closes the identity. The premium part is therefore everything the yield does not explain.

  • Sentiment
    Six readings, one rule.

    The sentiment strip copies six readings from their pages through the manifest (every copy is checked against the source on each update): the cross-sectional reward for beta over the last 63 sessions and the low-beta excess over cash of the same window, each against the CAPM comparison and with its own-history percentile (the Risk sentiment page's rolling line, which replaced the year-to-date factor fit on 19 Sep 2026), the site's equal-weight financial-conditions index and its percentile (Financial conditions), the excess bond premium and the HQM 10-year spread with percentiles (Corporate spreads), 20-day realised equity volatility and its percentile (Volatility) and the correlation quadrant (Correlations & regime). Rule: the average percentile of (EBP, HQM, vol) below 30 = rich, above 70 = cheap — unchanged by that swap, which enters no part of the rule.

  • Bases
    Four readings.

    Four readings of the same question. (1) Trailing: the S&P 500's trailing 12-month GAAP earnings yield (Shiller data via multpl.com, the input the September issue used) minus the monthly average 10-year Treasury par yield, and minus the 10-year TIPS real par yield. (2) Payout: r = 0.8448 ÷ forward P/E + g, with k the measured payout and g the one long-run rate 4.0% (the September issue used 0.85 and 3.0%; its numbers are reproduced at its own payout and g in the reconciliation), the discount rate implied by a constant-growth payout model; ERP = r − 10-year. (3) Forward: 100 ÷ forward P/E − 10-year. (4) Dividend: SPY's trailing 12-month cash distributions ÷ price + g − 10-year, made only from the ETF and Treasury.

  • Carry
    Proprietary inputs carried with price.

    Proprietary inputs are not refreshed between the author's captures. The forward P/E (FactSet, 19.1 in the report of 11 Sep, priced at the 10 Sep close) is carried to today with the S&P 500's own price from 10 Sep, holding earnings fixed: P/E_t = 19.1 × SPX_t ÷ SPX_10Sep. The trailing earnings yield is no longer the 11 Sep capture carried (3.82% then): since 27 Sep 2026 it is the Earnings page's E₀ — Shiller's trailing reported EPS for the latest quarter — over today's close, one E₀ on both pages; the capture remains the input of the September review's reproductions.

  • Matrix
    The constant-growth identity.

    Fair forward P/E = k ÷ (10-year + ERP − g), k = 0.8448 (the measured payout; the September report's matrix used its 0.85 convention) and g = 4.0%, the one long-run rate. The report's exhibit, reproduced in the reconciliation at the report's own g of 3.0% and payout of 0.85, expresses the matrix in index points using FactSet's forward EPS (7,591.70 ÷ 19.1); this page shows fair multiples and the percentage change from today's carried multiple, and keeps the index-point matrix only as the reproduced issue exhibit.

  • Caveats
    Assumptions, not estimates.

    A constant-growth model is very sensitive near r = g: at a 7% discount rate each 0.25pp of growth moves the fair multiple by about 6%. Growth and the earnings measure (GAAP trailing vs operating forward) are assumptions, not estimates, and the payout is measured on one basis (as-reported earnings) and applied to a forward operating multiple in the one-stage objects; the bases disagree by construction. The trailing earnings yield for the current month is the author's estimate.

  • Licence
    Treasury yields; vendor earnings and prices.

    Treasury yields are free to use. S&P 500 earnings (Shiller) and FactSet consensus are vendor data, shown with their attribution; the FactSet figures are the author's dated captures rather than a live feed. SPY prices and distributions are Yahoo Finance data. The identities and premia are computed here.

Source: The September report's calculations, reproduced; U.S. Treasury; quoted proprietary inputs; reference copies through the manifest.cross.valuation-18

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.