Market Gauge: How High Is the S&P 500, and How Good Is the Quality?

Four axes — breadth · valuation · positioning · quality

August 4, 2026 — Market-Level Analysis

The user’s ask: a separate report measuring how high the market is and how good the quality is, across three milestones: (1) breadth/width of the rally (is it broadening beyond semis? — backtrack vs history); (2) a valuation test (PE/PS/etc. vs history); (3) institutional/positioning metrics (CTA + others). Plus anything else useful.

TL;DR — expensive and stretched, but genuinely high-quality. “Priced for perfection,” not “junk bubble.”

⚠️ Protocol & data-hygiene notice

Applies the Two-Step Research Protocol. §1 method · §2 Step-1 draft · §3 Step-2 review · §4 breadth · §5 valuation · §6 positioning · §7 quality · §8 composite verdict.

Rule 4 flags: (a) yfinance columns are alphabetical → indexed by name. (b) CAPE percentile, breadth (RSP/SPY), sector participation and VIX are computed here and reproducible; (c) forward/trailing P/E, P/S and the Buffett indicator are web-sourced point-in-time estimates (ranges, not precise) and labeled as such; (d) CTA figures conflict across dated snapshots (net-long-adding vs earlier capitulation) — directionally “stretched,” exact number provisional.

Section 1 — Fact-base & method

Section 2 — Step 1: Concise Research Draft

Core conclusion: The S&P is expensive on every valuation axis (CAPE ~99th pct), with stretched mechanical positioning and still-narrow leadership — but underpinned by record real earnings. It is a high-price, high-quality, thin-margin-of-safety market: a priced-for-perfection late cycle, not a low-quality bubble. Downside is more likely from a positioning/rate/credit trigger than from valuation alone.

Supporting (claim → evidence):

  1. Valuation is ~99th-percentile extreme → CAPE 41.3 = 98.9th pct since 1881; Buffett ~225% GDP. Evidence: §5 — obtained.
  2. Leadership breadth is historically narrow despite better participation → RSP/SPY 3rd pct; 69% >200dma. Evidence: §4 — obtained.
  3. Earnings quality is genuinely high → record margins + record aggregate earnings. Evidence: §7 — obtained (web).

Opposing (claim → evidence):

  1. High valuation may be “justified” by margins/rates → record profitability supports higher multiples. Evidence: need margin-sustainability / rate path — partly unknown.
  2. Breadth is broadening → bullish, not late-cycle → 3mo RSP/SPY +1.8%, 69% >200dma. Evidence: obtained, but last-week reversal + 3/11 sectors argue it’s early/fragile.

Section 3 — Step 2: Strict Peer Review (draft NOT rewritten)

  1. Facts to verify: exact forward/trailing P/E and P/S (ranges, provider-dependent); the 200dma breadth number (sources cite 69% and 95% — definition-sensitive); the live CTA net exposure (snapshots conflict by date); whether “record margins” are peak-cyclical (mean-reversion risk).
  2. Logical leaps / equivocation: do not conflate participation breadth (% >200dma, healthy) with leadership concentration (RSP/SPY, narrow) — they say opposite things and the headline “breadth is fine” hides the concentration; “high margins justify high CAPE” is a regime assumption, not a law (margins mean-revert); CAPE is a long-horizon return signal, not a timing tool.
  3. Missing counterexamples / competing explanations: CAPE has been “expensive” for a decade and the market kept rising (it forecasts 10-yr returns, not the next year); index composition shifted toward higher-margin tech (some CAPE elevation is structural, not pure froth); a debasement regime lifts nominal everything.
  4. Primary sources to add: S&P/Shiller earnings series; Yardeni/FactSet forward-P/E & margin data; Goldman/DB prime-brokerage CTA notes (primary); Fed/GDP for the Buffett denominator; NYSE/Nasdaq A-D and %>200dma primary feeds.
  5. Speculation, not fact: “priced for perfection”; “thin margin of safety”; that the trigger will be positioning/credit rather than valuation; the “1998→late-1999” mapping (a characterization borrowed from the AI-bubble report).

Section 4 — Breadth: the two faces (is the rally broadening?)

Face A — participation is healthy (bullish): 72% of S&P 500 constituents are above their 200-day MA and 70% above their 60-day MA (computed from 503 members, Aug 4, 2026 — chart in §9.1; matches the web’s ~69%). On a 3-month view the equal-weight index is catching up (RSP/SPY +1.8%).

Face B — leadership concentration is the narrowest in 20 years (cautionary):

RSP/SPY (equal- vs cap-weight) Value
Level now (2005 = 1.00) 0.845
Percentile of 2005–2026 range 3rd
Change last 12mo −1.4% (mega-caps still leading)
Change last 1 week −3.5% (re-narrowed on mega-cap earnings)
Change last 3 months +1.8% (broadening)
Sectors beating SPY (last 1mo) 3 / 11
Sectors beating SPY (last 3mo) 4 / 11

Read (answering the user directly): your instinct is partly right — participation is broadening (69% >200dma, 3-month equal-weight catch-up). But the cap-weight concentration is still near a 20-year extreme, and the improvement is early and fragile: the last week actually re-narrowed −3.5% (mega-cap earnings pulled the index back up while the average stock lagged), and only 3 of 11 sectors beat the index over the past month. Verdict: genuine but unconfirmed broadening on top of historically narrow leadership. A durable regime change would show RSP/SPY rising for months and >6/11 sectors leading — not there yet.

(RSP/SPY is a clean, reproducible breadth proxy: when it falls, a handful of mega-caps are carrying the index; when it rises, the average stock is participating.)

Section 5 — Valuation: expensive on every axis

Metric Aug-2026 Hist. avg/median Percentile Verdict
Shiller CAPE 41.3 16.5 (median) 98.9th (computed, 1881–now; all-time max 44.2 = Dec-1999) Very expensive
Forward P/E ~21 ~17–18 >90th Expensive
Trailing P/E ~28 ~19–20 >90th Expensive
Price/Sales ~3.0 ~1.5 ~99th 2× history
Buffett (MktCap/GDP) ~225% ~100% 99–100th Record

Read: there is no valuation metric that says “cheap.” The single most robust anchor — CAPE at the 98.9th percentile of 145 years, essentially matching the dot-com peak — says forward 10-year real returns are likely low. Caveat (Rule 4 / §3): CAPE is a long-horizon signal, not a timing tool (it’s been elevated for a decade); part of the elevation is structural (index tilted to high-margin tech) and regime (debasement lifts nominal valuations). But on margin of safety, the answer is unambiguous: there is very little.

Section 6 — Positioning & volatility: stretched and asymmetric

Section 7 — Quality: the genuine bull anchor

Section 8 — Composite verdict: how high, how good

Percentile scorecard:

Axis Reading Signal
Valuation — CAPE 41.3 (98.9th pct) 🔴 EXPENSIVE
Valuation — Buffett ~225% GDP (~99th) 🔴 EXPENSIVE
Breadth — concentration RSP/SPY 3rd pct 🟠 NARROW
Breadth — recent trend 3mo +1.8% / 1wk −3.5% 🟡 broadening but fragile
Positioning — VIX 48th pct 🟡 mid / no cushion
Positioning — CTA net long, $100B+ downside 🟠 stretched/asymmetric
Quality — earnings record earnings + margins 🟢 HIGH (real profits) 
How HIGH:     Very. Valuation ~99th percentile (CAPE ~ dot-com peak), Buffett a
              record, positioning stretched. Margin of safety is THIN.
How GOOD:     Genuinely good. Record earnings + margins; participation improving.
              This is quality-led, not junk-led.
The synthesis: "PRICED FOR PERFECTION." A high-price, high-quality, thin-cushion,
              still-narrow late-cycle market. Not a profitless 1999/2000 blow-off,
              but little room for error.
Maps to:      The AI-bubble report's "1998 -> late-1999, loaded but unlit"
              (ai_bubble Addendum C): fragility preconditions present, trigger absent.
Vulnerability: A POSITIONING / RATE / CREDIT shock (CTA unwind, a Fed surprise,
              the Oracle/AI-credit canary), NOT a valuation-only collapse.
What would improve quality: RSP/SPY rising for months + >6/11 sectors leading
              (durable breadth) — would make the highs healthier and less fragile.

Bottom line for the user: the market is historically high — CAPE at the 99th percentile (basically the 1999 level), Buffett at a record, positioning stretched, leadership still the narrowest in 20 years. But the quality is genuinely good: record earnings and margins mean this is a priced-for-perfection, high-quality tape, not a profitless bubble. Your read that “width is getting better” is partly right — participation is improving (69% >200dma) — but the cap-weight concentration is still near a two-decade extreme and last week actually re-narrowed, so treat the broadening as real but unconfirmed. Practically: thin margin of safety + high quality = stay invested but hedged/diversified (this is exactly the 30/30/40 barbell and tail-hedge case), and watch the positioning/credit triggers, not the P/E, for the turn.

Section 9 — Deep-Dive Update (Aug 4, 2026): true breadth, charts, forward returns & valuation peaks

Added per the user’s follow-up: (a) true breadth from the 503 constituents (% above 60-day and 200-day MA), (b) a CAPE → forward-return backtest, charts to see the trends, and a peak/bottom history (“a high P/E like 26 — which year, why, and what happened after?”). All reproducible via run_deep_dive.py.

9.0 Two-Step Protocol (for the new claims)

Step 1 — Concise draft. Core conclusion: today’s valuation (CAPE ~99th pct) sits in the zone that, across 145 years, preceded near-zero-to-negative 10-yr real returns, while breadth confirms healthy-but-narrow participation — reinforcing the §8 “priced-for-perfection” verdict.

Step 2 — Peer review. (1) Verify: the 41.3 print (web) vs my Shiller file that ends Sept-2023 — the percentile/backtest use history through the file, and 41.3 is applied as a marker; the constituent breadth uses current membership (survivorship bias — dropped names excluded). (2) Equivocation: “forward 10-yr real −2.4%” is an average of overlapping start-months (autocorrelated; small effective n≈36) — directionally strong, statistically soft. (3) Missing counter: the CAPE ≥34 sample is dominated by 1998–2000 and 2021 — a narrow set of regimes, not 145 independent draws. (4) Primary sources: Shiller original series; Bunn/Shiller CAPE-return studies; a survivorship-free constituent history. (5) Speculation, not fact: that today “must” deliver −2.4% (it’s a base rate, not a forecast); the peak analogies (each cycle differs).

9.1 True breadth from constituents (computed today)

As of Aug 4, 2026: 72% of the 503 members are above their 200-day MA, 70% above their 60-day MA. This confirms (and slightly upgrades) the web’s ~69% and settles §4’s “Face A” — participation is genuinely broad, and the 60-day (faster) line sitting just below the 200-day says the very-recent momentum is neither overheated nor breaking.

S&P 500 breadth — % of constituents above moving averages, last 10 years + last 12 months

Left: the last ~10 years (2015–2026); right: the last 12 months. Read: healthy participation (both lines ~70%), recovered from the early-2026 dip. Across the decade, breadth swings 20–90% — today’s ~70% is middling-to-healthy, not euphoric. But recall §4 Face B — this broad participation coexists with record cap-weight concentration (RSP/SPY 3rd percentile). Two true things at once.

9.2 Valuation through time + the forward-return backtest

Is forward P/E ~21 really above the historical average? Yes — forward P/E ~21 vs a ~10-yr average of ~17–18 is ~15–20% above trend; the longer-history median P/E is ~15–16, so it’s richer still. But the cleanest long-history anchor is CAPE:

Shiller CAPE — full history 1881–2026 + last-10-year zoom

Left: full history 1881–2026; right: the last ~10 years (2016–2026). Today’s CAPE 41.3 (red) is above the 1929 peak (32.6) and the 2021 peak (38.6), and second only to the 2000 peak (44.2), versus a 145-yr median of ~17. And it’s high even against the recent decade: the 2016–2026 median CAPE is ~31 — today’s 41.3 tops the decade too. (Data-hygiene note, Rule 4: Yale’s Shiller mirror ends Sep-2023 at CAPE 30.8; the 2023–2026 tail (red) is reconstructed from real price with the slow 10-yr-earnings denominator calibrated to the reported 41.3 — a transparent two-anchor interpolation, not a new data source.)

The backtest — starting valuation vs subsequent real total return (annualized, 1881→now):

CAPE decile CAPE range Fwd 1-yr Fwd 3-yr Fwd 10-yr
1 (cheapest) 4.8–9.3 +16.7% +12.8% +11.7%
5 (median) 15.0–16.5 +7.2% +5.0% +6.7%
9 22.4–26.9 +6.2% +5.7% +4.5%
10 (most expensive) 26.9–44.2 +2.9% +0.9% +0.6%
Start at CAPE ≥34 (like today) ≥34 — — −2.4%/yr avg (−5.9% to +1.7%)

Starting CAPE vs forward return — full history (10-yr) + last decade (1-yr)

Left: full history (1881–now), CAPE vs forward 10-yr real return; right: the last ~10 years (2013–2022 starts), CAPE vs forward 1-yr real return.

Read: the relationship is monotonic and strong — the price you pay caps the return you get. The top decile has historically delivered ~0%/yr real over a decade; the ≥34 zone (where we are) has averaged negative. Strikingly, the recent-decade panel slopes down too: even at a 1-year horizon, the decade’s highest CAPE readings (the 2021 peak, ~38) were followed by −10% to −20% real (the 2022 bear), while the cheaper 2013–2016 starts (CAPE ~24) delivered positive years. Caveat (§9.0): the recent-decade 1-yr signal leans heavily on the single 2022 episode — suggestive, not proof; and none of this dates the exact top (the 1-yr signal is noisy in the full sample).

9.2a Adjusting for rates — the equity risk premium (the “vs bonds” view)

The one thing raw CAPE misses: “expensive vs history” implicitly assumes bonds yield what they used to. They don’t — the 10-yr Treasury is ~4.6% now vs ~2% through the 2010s. The rate-aware gauge is the Excess CAPE Yield (ECY) = CAPE real earnings yield (1/CAPE) − real 10-yr yield — i.e., the premium stocks offer over bonds.

Equity Risk Premium — Excess CAPE Yield, 1920–2026 + last 10 years

As of Aug 2026, ECY ≈ +1.0% — below the last-decade median (~2.6%) and the long-run median (~3.5%): the thinnest equity cushion over bonds in the entire decade (it was ~4% in the mid-2010s and ~4.9% at the 2020 low). But — the crucial nuance — it is still positive, unlike the 2000 peak (−2.6%). Because 2000 paired a high CAPE with high real rates, whereas today’s real rates are lower, the rate-adjusted picture is expensive but not the no-premium extreme of the dot-com top.

Two-sided read:

(Data hygiene, Rule 4: ECY history is Shiller’s own column through Sep-2023; the 2023–26 extension is anchored to that last value and moved by the change in CAPE-yield and the nominal 10-yr — a constant inflation expectation cancels in the rate difference. Level is approximate; the trend/percentile is robust.)

9.3 Valuation peaks & troughs — which year, why, what happened after

(Real total-return drawdown and forward-10-yr real CAGR computed from Shiller’s real-TR index.)

Event ~CAPE Driver Next 5-yr real drawdown Next 10-yr real CAGR
1929-09 peak 32.6 Roaring-20s leverage/margin mania −77% −1.4%/yr
1966 peak 24.1 Nifty-Fifty start; pre-stagflation −22% −2.5%/yr
2000-03 peak 44.2 Dot-com internet bubble −43% −2.8%/yr
2007-10 peak 27.5 Housing/credit peak −50% +5.7%/yr
2021-12 peak 38.6 Post-COVID stimulus / mega-cap −24% (partial) −5.8%/yr (ongoing)
1982-07 trough 6.6 Volcker recession; 14% inflation broke 0% +14.3%/yr
2009-03 trough 13.3 GFC bottom 0% +14.3%/yr

On the user’s specific example — “a high P/E like 26”: a trailing P/E in the mid-to-high 20s (CAPE mid-20s to ~30) has clustered at 1929, 1966, 2007 — each a major top that preceded −22% to −77% real drawdowns and a lost decade of real returns. The mirror image is decisive: the two cheapest starts (1982 CAPE 6.6, 2009 CAPE 13.3) delivered +14%/yr real for the next decade. Valuation didn’t time the exact top, but it powerfully set the 10-year payoff — and today’s 41.3 is on the wrong end of that.

9.4 What this adds to the verdict

Reproduce it yourself

cd market_gauge
python run_market_gauge.py     # writes data/*.csv (breadth, sector breadth, valuation, VIX, scorecard)
python run_deep_dive.py        # §9: constituent breadth (60/200dma), CAPE forward returns, ERP, peaks -> charts/*.png

Data files (market_gauge/data/): breadth.csv, sector_breadth.csv, valuation.csv, vix.csv, scorecard.csv, plus §9: breadth_constituents.csv, cape_forward_returns.csv, excess_cape_yield.csv, valuation_peaks.csv. Charts (market_gauge/charts/): breadth_constituents.png, cape_history.png, cape_forward_scatter.png, erp_excess_cape_yield.png. CAPE computed from Yale ie_data.xls; constituents from the datasets/s-and-p-500-companies list; 10-yr yield from yfinance ^TNX.

Sources (accessed Aug 4, 2026): yfinance (RSP, SPY, ^VIX, 11 SPDR sectors); Yale/Shiller ie_data.xls (CAPE 1881–now); web valuation (MacroMicro, investsnips, GuruFocus, worldperatio — forward/trailing P/E, CAPE, Buffett); breadth (Stock Alarm Pro, MacroMicro, CondorEdge — %>200dma, A/D); positioning (Goldman/Deutsche Bank via Yahoo/Hedgeweek/Investing.com — CTA/sentiment). PE/PS/Buffett are point-in-time estimates (Rule 4); CTA snapshots conflict by date.

Two-Step Research Protocol applied (§2 draft + §3 review). Computed metrics reproducible; web-sourced levels flagged. CAPE is a long-horizon, not a timing, signal. Education/analysis only — not investment advice.