1. Measure the stock-over-bond gap on a long trailing window and rank it historically
The repeatable method
- Compute stocks' annualized total return minus long-bond total return over a multi-year trailing window (Paulsen uses 76 months), back as far as data allow.
- Mark the prior peaks and note what followed each (bear markets, crashes, long flat periods).
- Rank the current reading. A reading above every prior peak is an extreme by definition — not a timing signal, but a reason to reduce risk.
- Pair it with the driver you think is unsustainable (here: rising rates and oil) so the extreme has a mechanism, not just a percentile.
Here: the Paulsen/Felder chart — prior peaks 1939, 1956 (~25%), 1969, 1981, 1999 (~16%, followed by "the punishing bear market of 2000 to 2003"); today ~28%: "there's never been such a pronounced performance gap between stocks and bonds."
Watch for
- The gap rolling over from a record — historically the damage came as it mean-reverted, via stocks falling rather than bonds rallying.
2. Use two precedents: one matching the kind of divergence, one matching its degree
The repeatable method
- Find the episode where the same mechanism ran: long yields rising sharply while equities rallied anyway.
- Record the size of each leg and exactly how it resolved (magnitude and speed of the drawdown).
- Separately find the episode with the largest measured gap — it may be a different era.
- If both precedents ended badly by different routes (sudden crash vs multi-year bear), the warning is robust to which path this time takes.
Here: kind — 1986–87, long yields "7.2% to 10%" while the S&P rose "nearly 40%," then SPX −28.4% Sep 30–Nov 30, 1987. Degree — the late 1990s, "far more dramatic" on the relative-return chart, ending in 2000–03.
Watch for
- A 10-year yield holding at or above 5% while equities stay near highs — the 1987 configuration; a yield spike coinciding with an oil spike raises the odds of the fast version.
3. Check any "long-run average return" for starting-point bias
The repeatable method
- When a return is quoted as "the long-run average," recompute it from the earliest available date.
- If the numbers differ materially, identify the crash or boom near the start that explains it.
- Use the gap to temper expectations: a recent run far above even the favourable long-run figure is borrowing from future returns.
Here: the S&P's 16%/yr since July 2020 vs the usual 9–10% long-run figure, and only "6.75%" from 1927 — "undoubtedly, this reflects the great stock market crash of 1929" and the bear market to 1932.
Watch for
- Return assumptions (in plans, target-date glide paths, pension models) built on the recent six years rather than on crash-inclusive history.