Macro Drawdown Radar · U.S. S&P 500
10% drawdown · Correction overlay
Correction conditions are on Watch. Check the four pillars below to see which inputs contribute most. A correction is possible, not imminent by definition.
Monthly historical benchmark
32 of 79 historical Amber months. Same monthly sample for every horizon: Jan 2000–Sep 2022. This is not a calibrated probability for today’s daily inputs.
A forward peak-to-trough price decline—not necessarily a 10% fall from today’s price. No crash-date prediction.Transparent calculation
Click a pillar to see its current inputs and sources. Score = the weighted average below, with no extra high-pillar boost.
15%: expensive markets are vulnerable, but valuation alone is a weak correction clock.
20%: rates, liquidity, housing and labor provide the economic backdrop.
25%: spread widening and tighter financing confirm deteriorating risk appetite.
40%: price trends and volatility receive more weight for smaller, faster corrections.
Pillar colors use this layer’s 50 / 65 cutoffs; individual inputs are risk percentiles. Real yields, claims, permits and other slow releases retain their latest available observations. Missing historical pillars are reweighted using available weights.
Dates identify observations or period starts—not necessarily release dates. Both layers use one shared automatic refresh.
Backtest · Strict advance-warning rule
31 Jan 2000–30 Sept 2023 · Monthly-average prices; brief intramonth corrections may be absent.
An advance warning must occur within 12 months strictly before the first 10% breach. It need not precede the market peak.
| Peak → breach | Worst decline | Advance warning | First prior Red | Lead |
|---|---|---|---|---|
| 31 Aug 2000 31 Dec 2000 | 43.7%Monthly averages | Red beforehand | 31 Mar 2000 | 9 months |
| 31 Oct 2007 31 Jan 2008 | 50.8%Monthly averages | Red beforehand | 31 Aug 2007 | 5 months |
| 30 Sept 2018 31 Dec 2018 | 11.5%Monthly averages | Amber only | No advance warning | — |
| 31 Jan 2020 31 Mar 2020 | 19.1%Monthly averages | Amber onlyRed arrived on the breach date—not advance | No advance warning | — |
| 31 Dec 2021 31 May 2022 | 20.3%Monthly averages | Red beforehand | 31 Mar 2022 | 2 months |
In the daily cross-check, the new score was Red 26.8% of the time versus 57.3% for the original cycle score—a reduction of 30.5 percentage points. Both warned before 5 of 5 observed events. Five events are too few to establish reliable future accuracy.
In the long monthly test, the new model catches 1953 but loses the old model’s advance-Red warning for 1987. Its Red misses are 1987, late 2018 and 2020; Amber still warned before all three. Keep the independent 20% layer in view.
One drawdown episode runs until the old peak is recovered. Local corrections inside an unrecovered bear market are not separate episodes. This is not an exhaustive 70-year daily correction inventory.
10% layer · False-alarm audit
36 were already in an unrecovered 10% drawdown. A forward decline can represent further losses, not a new independent correction.
Raw monthly misses: not a count of independent bad calls.
Unlinked Red months less than six months apart are grouped.
| Unlinked warning regime | Red months |
|---|---|
| 30 Sept 2015–29 Feb 2016 | 3 |
Removing post-breach or boundary observations changes the question. The raw forward-frequency result remains visible above.
Same monthly sample · Different risk states
Adjacent months overlap. These are historical month frequencies, not independent event probabilities.
| 10% model state | Months | Within 3m | Within 6m | Within 9m | Within 12m |
|---|---|---|---|---|---|
| Green | 145 | 4.8% | 12.4% | 19.3% | 26.2% |
| Amber | 79 | 12.7% | 25.3% | 31.6% | 40.5% |
| Red | 49 | 26.5% | 57.1% | 71.4% | 71.4% |
Green is not “safe”: even low-stress environments can experience a 10% correction. Scores are not percentages, and crossing a cutoff does not create an instant jump in real-world probability.
Plain-English guide
10% means a correction; 20% means a larger bear-market drawdown. A 10% event can eventually deepen into a 20% event.
Expand the Market and Credit pillars. Rising risk in both is more informative than expensive valuations alone.
Read missed events and time on alert before trusting the catch rate. Amber is a research prompt, not an automatic hedge order.
Three plausible weight sets and four Red cutoffs were compared using 1951–1998. Candidates had to warn before at least 80% of the 11 development breaches; the one with the fewest Red months was selected. Watch is fixed 15 points below Red. The selected parameters were then held unchanged for the later comparisons.
The 10% score is a weighted mean: 0.15 × Valuation + 0.20 × Macro + 0.25 × Credit + 0.40 × Market. Unlike the 20% formula, it has no top-two-pillar blend or confirmation bonus.
| Candidate | V / Macro / Credit / Market | Red at | Development caught | Months Red |
|---|---|---|---|---|
| balanced | 15 / 20 / 25 / 40 | 55 | 11 / 11 | 38.2% |
| balanced | 15 / 20 / 25 / 40 | 60 | 10 / 11 | 25.5% |
| balanced · selected | 15 / 20 / 25 / 40 | 65 | 10 / 11 | 15.1% |
| balanced | 15 / 20 / 25 / 40 | 70 | 3 / 11 | 6.1% |
| market led | 10 / 15 / 25 / 50 | 55 | 11 / 11 | 37.0% |
| market led | 10 / 15 / 25 / 50 | 60 | 10 / 11 | 24.8% |
| market led | 10 / 15 / 25 / 50 | 65 | 10 / 11 | 17.0% |
| market led | 10 / 15 / 25 / 50 | 70 | 3 / 11 | 7.5% |
| fast market | 5 / 10 / 25 / 60 | 55 | 11 / 11 | 35.9% |
| fast market | 5 / 10 / 25 / 60 | 60 | 10 / 11 | 24.7% |
| fast market | 5 / 10 / 25 / 60 | 65 | 9 / 11 | 17.0% |
| fast market | 5 / 10 / 25 / 60 | 70 | 5 / 11 | 10.4% |
Monthly price source: Robert Shiller. Daily price source: FRED / S&P Dow Jones Indices. Expand the pillars for individual macro sources.