Signal Integrity · Market Stress & Crisis
Is this a crisis — or just stress?
In a selloff it’s the hardest question to answer. Panic on a scare and you sell the bottom; miss a real crisis and you ride it down — and the official recession call arrives months too late. EconCERF gives you one auditable answer, in real time, and shows its work.
How bad is market pressure right now?
Based on
Output — a level
Can be extreme on its own.
Has that stress become systemic?
Requires all three, at once
The dollar
The subset that matters.
The three conditions are the difference — a stress reading becomes a crisis only when all three hold at once. That gate is how we tell the two apart.
March 2020. Was it 2008 all over again, or another quick dip like December 2018? On 11 March, EconCERF classified it a systemic crisis — 57 days before the economy confirmed. And when the noise was loudest — 2018, 2022, April 2025 — it held those at stress, and was right.
9
major events, all sorted correctly
4
became crises
5
held at stress
17 yrs
on our own point-in-time signals
The evidence
Every major event of the last 17 years — and why
Pick any event. It leads with the three actors — volatility, credit, and the dollar — then shows the three conditions with the real numbers, so you see exactly why it was promoted to crisis or held at stress. Nothing asserted; everything shown.
COVID-19 crash
Q1 2020
The three actors
Equity volatility
VIX 82.7
Extreme
Credit
HY 10.9%
Extreme
Dollar · confirmer
+7.4%
Confirms
The three conditions
Systemic — two independent channels
Equity-vol and credit both stressed
Severe — extreme magnitude
VIX 82.7 and HY 10.9% — both extreme
Sustained — 5+ trading days
25 trading days active; went active 11 Mar
Promoted to crisis. Classified as systemic 57 days before the economy confirmed; the dollar’s 7.4% dash-for-cash confirmed.
Led the economy by 57 days.
What you get
One answer you can act on — and defend
The moment it matters, one auditable classification — stress or crisis — with the evidence behind it. Not a prediction, not a vibe: a call you can put in front of your investment committee or your clients, and stand behind.
On the dollar
A confirmer, not a trigger
The dollar is a confirmer, not a trigger. It corroborates a real crisis but never casts a vote — because it also rises in rate-driven selloffs that are not crises (in 2022 it climbed 4.4%, yet the event was correctly held at stress). Our signal is the Federal Reserve’s trade-weighted broad dollar index (FRED DTWEXBGS), not the six-currency ICE “DXY” shown on most trading platforms — they move similarly but are not the same instrument.
Trade-weighted dollar · FRED DTWEXBGS
How it works
The rule, if you want to see it
See the exact rule and its thresholds›
Systemic. Two independent channels — equity volatility and credit — stressed at once. The dollar corroborates but casts no vote.
Severe. VIX ≥ 40 alone, or VIX ≥ 30 together with HY OAS ≥ 500 bps. Credit never counts on its own.
Sustained. Holds for 5+ consecutive trading days (5 on to confirm, 3 off to stand down, 5-day recovery window).
These numbers are read live from the engine (rule v1.0.0) — this page cannot drift from the rule it documents.
Why trust it
No hindsight, and nothing hidden
No hindsight. When we say we would have classified COVID as a crisis on 11 March 2020, we mean with the information available that day — never with the benefit of knowing how it ended.
How we guarantee it ›
- ✓Every input is non-revising market data or a signal recomputed on a trailing-only window — no revised numbers, no look-ahead.
- ✓The credit signal reads the HYG/LQD ETF ratio, with clean history back to April 2007 — not the government spread series that was later truncated.
- ✓The historical signals are produced by the same live scoring function as production, so there is no historical-versus-live split-brain.
Robustness — we moved every threshold
90 outcome cells, exactly 1 changes. The four crises are invariant across every perturbation — not one drops. Of 90 outcome cells, exactly one changes: the Dec-2018 selloff would promote to crisis only if the volatility-elevated bar were loosened to 28 (it is 30 at baseline). Dec 2018 was the most crisis-like of the stress events, so the single boundary sits exactly where it should — between the closest call and the real crises. We report it and leave it; we do not move the line to erase it.
See the full grid ›
What we don’t claim
Built by studying history›
So the 17-year record proves consistency, not fortune-telling. The only true test is forward — and since we froze the rule, every new event (2025, 2026) has sorted correctly.
The dollar confirms, it doesn’t decide›
It agrees with every real crisis but also rises in rate-driven selloffs like 2022, so we show it as corroborating evidence — it never casts a vote toward a crisis.
The 2010 label is a judgment call›
It meets the rule and we count it as a crisis; whether you call it a crisis or a severe scare is an editorial choice, and we say so rather than bury it.
We can see back to 2007›
Our credit signal begins when its instrument began trading. Further back than that, we don’t pretend to have it.
What we can say
- ✓Classifies every major event of the last 17 years correctly — 4 crises and 5 held at stress — on our own point-in-time signals.
- ✓Robust: the conclusion is unchanged when every threshold is varied across an economically sensible range; the single sensitivity (the 2018 boundary) is reported, not tuned away.
- ✓Glass box: every level is documented and traceable, credit reads from HYG/LQD with clean history to 2007, and the historical signals use trailing-only normalization.
- ✓The dollar independently confirms every real crisis on our own data, and is shown as a confirmer — not a trigger — because it also rises in the rate-driven 2022 dollar rally.
What we cannot say
- ✗That we predicted these crises out-of-sample. (The data is point-in-time, but the design saw history.)
- ✗A 17-year live-engine track record. (The engine has not run live for 17 years; this is a historical replay of its signals.)
- ✗“Zero false alarms” without stating the window and definition.
- ✗Any real-time crisis claim attached to the Macro Regime — it lags and confirms, it does not detect.