Signal Integrity
Most models work in backtests. The question is whether they work before the outcome is known.
EconCERF is built to answer that — and we publish the results, including where it fails.
Economic Regime
Where are we now?
29 economic indicators across 6 pillars
Updated on data releases
Anchored in real economic outcomes
Tells you where you ARE
Market Themes
What is the market pricing next?
74 cross-asset signals across 10 themes
Updated daily
Forward-looking by design
Tells you where you're GOING
These aren't two versions of the same thing. The regime model grounds you in economic reality. The theme model tells you what's coming. When they agree, conviction is high. When they disagree — that's the most valuable signal of all.
Proven under real conditions
Does the model reflect what's actually happening in the economy?
Validation
Tested across every major cycle since 2007 — expansions, recessions, and inflection points — and verified against NBER recession dates and known macro events.
Result
Correctly identifies all major economic regimes with consistent alignment to real-world conditions.
Does the system detect periods of market stress and crisis?
Validation
Evaluated across major stress events including Lehman, the Global Financial Crisis, COVID, SVB, and other volatility spikes over the past 18 years.
Result
All major crises detected, with peak signal conviction occurring during the most severe periods of stress.
When the system signals, does the economy move in that direction?
Validation
Every sustained market theme (10+ day persistence) was tracked against economic outcomes over the following 3 months — including growth, inflation, labor, and financial conditions — across an 18-year history.
Result
Signals confirm direction in 62–100% of cases, demonstrating consistent forward-looking predictive power.
Does the system distinguish meaningful signals from short-term noise?
Validation
Observed that most market signals last only 2–5 days. A persistence filter was applied, requiring signals to hold for at least 10 consecutive days before being considered valid.
Result
Short-term noise is filtered out, leaving only durable signals with materially higher reliability.
Not every theme performs equally. Two themes — Policy Mistake and Reflation Risk — confirm at roughly 55%. Better than a coin flip, but not at the level where we'd call them proven. Two themes — Stagflation Lite and Productivity Boom — are rare by nature. They've only occurred 3-4 times in 18 years, which isn't enough data for statistical confidence.
We show you the tier for every theme: Proven, Moderate, or Limited Data. We don't hide the ones that aren't perfect — we label them so you can calibrate your confidence accordingly.
Each signal is tested the way a skeptical allocator would evaluate it — not just for accuracy, but for reliability under real conditions.
We re-run the full validation every 3-6 months as new economic data accumulates, and after any significant macro event. Last validated: March 2026.