Five gauges, one pendulum
Each gauge is scored only against its own ten-year distribution — no composite, no weighting to argue about. Tight credit spreads and a sleepy VIX read as complacency; an inverted yield curve reads as late-cycle regardless of percentile.
| Gauge | Today | 10-yr pctile | Where in its range (shaded = complacency zone) | 10 years, monthly | Read |
|---|---|---|---|---|---|
| Credit HY credit spread |
2.84% | 26th | near average | ||
| Credit IG credit spread |
0.80% | 25th | tight | ||
| Curve 10Y minus 2Y |
+0.47% | 54th | positive (+0.47) | ||
| Curve 10Y minus 3M |
+0.92% | 65th | positive (+0.92) | ||
| Sentiment VIX |
17.09 | 51st | near average |
Late 1999, side by side
6 of 11 comparable indicators are MORE extreme today than at the late-1999 peak (0 equal, 5 less extreme). The count matters less than its direction over time.
| Indicator | Today | Late-1999 peak | Verdict |
|---|---|---|---|
| Valuation | |||
| Shiller Excess CAPE Yield | 1.39 | -1.52 | less extreme |
| CAPE (Shiller PE10) | 39.6 | 44.2 | less extreme |
| Buffett Indicator (Wilshire 5000 / GDP) | 219 | 150 | MORE extreme |
| Psychology | |||
| US household equities as % of financial assets | 41.50 | 38.00 | MORE extreme |
| Leverage | |||
| Margin debt / nominal GDP | 3.20 | 2.70 | MORE extreme |
| Breadth & concentration | |||
| RSP/SPY 60d relative return (%) | 2.63 | n/a | no 1999 peer |
| S&P 500 top-5 weight (concentration) | 28.50 | 18.00 | MORE extreme |
| S&P 500 top-10 weight | 38.50 | 25.00 | MORE extreme |
| Leadership | |||
| QQQ/SPY 126d relative return (%) | 1.25 | 74.96 | less extreme |
| Mania symptoms | |||
| IPOs raised, trailing 12m ($B) | 38.0 | 97.0 | less extreme |
| Profitless IPO share (% of count) | 65.00 | 81.00 | less extreme |
| Crypto total market cap ($T) | 3.60 | 0.00 | no 1999 peer |
| Credit & volatility | |||
| HY credit spread (OAS) | 2.84 | n/a | no 1999 peer |
| IG credit spread (OAS) | 0.80 | n/a | no 1999 peer |
| 10Y minus 2Y | 0.47 | -0.44 | no 1999 peer |
| 10Y minus 3M | 0.92 | 0.29 | no 1999 peer |
| VIX | 17.1 | 23.3 | MORE extreme |
The price of admission
The Shiller Excess CAPE Yield — the cyclically-adjusted earnings yield minus the real 10-year Treasury yield — is the cushion equities offer over bonds. At 1.39% (15th percentile since 1970), the cushion is thin: expensive markets don’t forecast a crash, they forecast low returns and demand fatter discounts on every purchase.
The full 60-year story, with the headlines that defined each regime: The Living Intelligent Investor. What a starting ECY this low has meant for the next decade: The Patience Curve.
The crowd
University of Michigan consumer sentiment is at 49.5 — the 1st percentile of all readings since 1952. Readings this depressed have historically been both a contrarian floor for stocks and a recession lead. We treat it as both: a reason not to flee, and a reason to demand a discount.
What this implies
- New positions: demand a margin of safety at the upper end of the historical range. If the valuation is tight, pass.
- Cash posture: hold or build. Expensive markets reward patience; the pendulum always swings back.
- What this is not: a timing signal. None of these gauges predicts the date of a turn — they price the risk of being early versus the cost of being greedy.
The pendulum is tilting toward greed. Not the kind of euphoria that demands an exit, but the kind of complacency that demands a tighter definition of margin of safety. Be selective. The price of admission to a fair market is patience.
Sources & method. Five-gauge pendulum: daily FRED series HY OAS, IG OAS, 10Y−2Y, 10Y−3M, VIX, each ranked in its own 10-year window. Late-1999 panel as of 2026-08-03: FRED, Yahoo Finance, Shiller’s public dataset, Z.1 household equity share, FINRA margin statistics, and IPO tallies; “MORE extreme” means beyond the late-1999 peak in the late-cycle direction (±3% band counts as equal). Excess CAPE Yield: (1/CAPE) − real 10-year Treasury yield, monthly, latest reading Jul 2026. Consumer sentiment: FRED UMCSENT. Rebuilt each market day from the same pipeline that writes the research logs.