THE DENOMINATOR
See exactly what is being normalized.
The median uses ten complete calendar-year margins, so a single crisis year cannot dominate a decade.
How the calculation worksTHE VALUATION LAB
A historical view of valuing the S&P 500 using its current sales base and median profit margins.
NORMALIZED-MARGIN CAPELoading historical data
—Price ÷ normalized earnings
Loading historical data
Loading historical data
Loading historical data
Loading historical data
Sales growth in the data: last year +7.4% · 5-yr CAGR +7.7% · 10-yr CAGR +6.8%
HISTORICAL SERIES
Accounting-period history using revised data. These are not values known in real time and are not an investment backtest.
VALUATION VS WHAT FOLLOWED
Each point is one starting quarter. The vertical axis is the S&P 500 return over the ten years that followed, annualized with dividends reinvested, built from Shiller's monthly price and dividend data. The revised CAPE is plotted at its base settings — margin dial at zero, trailing sales, and the margin basis selected above — so the dial and forward-growth controls do not change this chart, but the margin basis does. Shiller CAPE points cover exactly the same starting quarters: no Shiller data appears before the revised CAPE begins. Early decades rest on estimated sales; see Method & sources.
THE DENOMINATOR
The median uses ten complete calendar-year margins, so a single crisis year cannot dominate a decade.
How the calculation worksHOW TO READ IT
Loading calculation.
The gap to Shiller is a difference in measurement. It does not establish fair value or predict the next market move.
THE NUMBERS
| Quarter | Index price | TTM sales / share | Margin used | Normal EPS | Normalized CAPE | Shiller CAPE |
|---|
Values are nominal dollars per index share, margins are percentages, and CAPEs are multiples. Rows marked (est.) are normalized readings that use estimated sales — in the current period, or in at least one year of the margin window. Full inputs are in the workbook through Q3 2025; the Q4 2025 and 2026 inputs are described under Method & sources.
METHOD & SOURCES
The window selector changes the number of annual margins. Five- and fifteen-year settings are sensitivity tests. The downloaded workbook contains the ten-year baseline with editable source inputs and calculation formulas.
The margin dial and the forward sales growth rate adjust the denominator on this page only. The dial blends each period's window-median margin toward that period's current reported margin: at 0% the margin used is the window median; at 100% it is the current reported margin, and the multiple then equals the trailing P/E. At Q3 2026 no current reported margin exists yet, so the margin used there is the window median at any dial setting. The growth control states sales one year forward — forward sales = trailing sales × (1 + g) — at the same rate for every historical quarter. Its presets are a consensus nominal GDP assumption of 4.5% (a fixed assumption for this static page, not a live forecast), the trailing 5-year and 10-year sales growth rates computed from the data on this page, or a rate entered directly. Defaults are a 0% dial and the consensus GDP preset. Neither control changes the downloaded workbook, which holds the ten-year baseline on trailing sales. The margin basis control runs the same analysis on the window's median margin (the default), its top-quartile margin, or its bottom-quartile margin, selected above; the headline, charts, table, download and scatter all follow the selected basis.
The headline also shows a P/E range from the same window's margin quartiles: the 75th-percentile and 25th-percentile margins of the same complete calendar years, with percentiles taken by linear interpolation between closest ranks. Each quartile margin goes through the same dial blend and the same forward-sales growth as the headline, so the range always brackets the headline multiple — the top-quartile margin gives the low end and the bottom-quartile margin the high end — and at a 100% dial setting both ends collapse onto the headline multiple.
Current sales recognize growth already achieved. Margin normalization reduces the effect of unusually high or low current profitability. Using current sales per share reduces the mismatch caused by averaging old EPS at different business and share-count scales.
This version uses aggregate index margins. It retains historical changes in sector and company mix, historical financing costs and taxes, and possible effects from changes in accounting. It is not the company-by-company CAPE proposed in the original discussion, is not fully payout-neutral, and does not support sector-neutral country comparisons.
Peak sales can still overstate sustainable business size. Structural changes can make historical margins less relevant. A lower ratio is not automatically a better valuation signal.
Sales per index share run from 1946, from sources that predate the S&P workbook. Calendar-year anchors: 1946–1991 are a reconstructed proxy history published by DQYDJ (before 1957 it describes predecessor and proxy indices, not the modern 500-stock index) and are flagged as estimates; 1992–1999 are calendar-year sales from the S&P Analyst's Handbook, transcribed in the PCRA dataset, and are reported values; 2000 Q1–Q3 are estimated trailing-twelve-month values from DQYDJ bridging to the reported series; from 2000 Q4 onward, sales are exact sums of four reported quarters from the archived S&P workbook, and all 100 of those trailing observations match the workbook sums exactly. Four further quarters, Q4 2025 through Q3 2026, are appended from continuation sources, described below. Within 1946–1999, the March, June and September observations are DQYDJ's monthly trailing sales proportionally benchmarked to the December anchors: they keep DQYDJ's within-year shape, pass exactly through each December anchor, remove the level seam where the handbook values replace the proxy in 1992, and are flagged as estimates. A normalized reading is flagged (est.) when its own sales figure is estimated or any annual sales figure inside its margin window is estimated. The normalized series starts when the selected window first fills with complete calendar years: Q4 1950 at five years (304 observations), Q4 1955 at ten years (284), Q4 1960 at fifteen years (264). The reported series ends in September 2025; the flagged continuation observations end in September 2026. Dates identify accounting periods, not filing or publication dates. Historical values reflect the retrieved data vintage, so this series must not be treated as a point-in-time strategy backtest.
The archived S&P workbook is dated January 31, 2026 and says the public files were discontinued. Its complete quarterly sales and reported earnings stop in Q3 2025. Forecasts and incomplete quarters are excluded. Four further quarters are appended from continuation sources. Q4 2025 sales per index share of $2,070.73 are DQYDJ's December 2025 reading — the last quarter-end their series reports directly; at the splice, their September 2025 reading matches the workbook sum exactly ($2,042.16), and their quarter-end values match the workbook sums at every quarter-end checked. Q4 2025 earnings of $63.07 are differenced from Shiller's trailing earnings — December $240.63 less September $235.25, plus the workbook's Q4 2024 quarter of $57.69 — giving calendar-2025 earnings of $239.44 and a 2025 margin of 11.56%; on Shiller's own trailing earnings, the same sales give 11.62%. The 2026 quarters are included, with their source flags, for review. Since 2026, S&P publishes a price-to-sales ratio rather than sales per share, and DQYDJ's 2026 quarter-end sales are derived from that ratio and approximate by their own methodology note. Their March 2026 reading of $1,582.31 (a derived ratio of 4.21, against 3.31 in December 2025) implies a one-quarter trailing-sales drop of about 24% — a decline that did not occur: GuruFocus's aggregate index revenue for March 2026 is about 6% above March 2025, and no reported series shows a fall. That reading is therefore rejected here as a methodology artifact of the derived series, and the model's March 2026 sales are instead a straight-line interpolation between the December 2025 and June 2026 quarter-ends — $2,131.77, the midpoint of $2,070.73 and $2,192.80 — flagged as an estimate like the other 2026 observations. June 2026 reads $2,192.80 as published, and that June value is repeated unchanged from July through October, so September 2026 ($2,192.80) is an apparent carry-forward rather than a new report. All three 2026 observations, and every normalized reading computed from them, are flagged as estimates pending reported confirmation; the early-2026 discontinuity in DQYDJ's derived series is a source methodology break, not a measured collapse and rebound in index sales. Reported trailing earnings for Q1 and Q2 2026 are differenced from Shiller's trailing earnings on the same basis as Q4 2025 — calendar-basis trailing earnings of $260.53 at March and $294.20 at June. No reported trailing earnings for Q3 2026 exist in the sources used: Shiller's current vintage and Multpl's earnings table both stop at June 2026 (Multpl: $298.03, against Shiller's $295.39). The Q3 2026 observation therefore carries normalized values only; its current margin and trailing P/E are shown as n.a., and the margin dial uses the window median at that point.
Shiller's monthly history in the workbook runs from January 1871 through October 2026. October 2026 uses the October 2 close, and September/October CPI are estimated in that source. On this page the Shiller series are sampled at quarter-ending months and end in September 2026 (October 2026 is not sampled: it is a partial month priced at the October 2 close): price from 1871, CAPE from 1881, the first quarter Shiller publishes it; the sampled trailing earnings end in June 2026, the last month the current vintage publishes. The Shiller lines span that full history in the charts; the normalized lines begin where the selected margin window begins. The earnings chart's reported line is Shiller's trailing earnings series; the stat cards use the S&P workbook's reported earnings — a different vintage (Q3 2025: $234.06 per share, versus Shiller's $235.25). In the data table, readings before 1988 use Shiller's December trailing earnings, the source of the pre-1988 annual margins. Shiller and the archived S&P file have different earnings vintages.
This is a downloaded data snapshot, not a live feed. Updated source data are required to refresh the series.
Sources retrieved November 15–16, 2026 (Pacific time). The downloadable Excel contains source URLs and the formulas. All 100 trailing-sales observations overlap the published cross-check within $0.006 per index share, and the 100 reported trailing-sales observations from 2000 Q4 in the imported sales file agree with the workbook's own quarterly sums to within $0.000000000005 per index share.