What the Compass Would Have Said

The most damaging thing in macro investing is not being wrong about a recession. It's being wrong about a recession that never happened.

From 2000 to 2026, the US economy had three real recessions: 2001, 2008, and 2020. In that same span, there were at least five serious false alarms — periods where a significant portion of market commentary, professional economists, and even parts of the institutional investment community called for a recession that did not materialize. In two of those episodes, 2015–2016 and 2022–2023, the false calls were nearly universal. The inverted yield curve screamed it. The LEI confirmed it. Credit spreads widened. And the recession never came.

The question worth asking is whether a systematic framework, applied consistently, would have done better. Not just at catching the real recessions, but at correctly filtering the false ones.

The answer, run through the Cycle Compass framework retrospectively, is yes. And the reason it works is structural, not lucky.

The Back-Test

The table below runs the Compass framework across every significant inflection point since 2000, applying the indicator logic as if it had been running in real time. The regime classification it would have produced is compared to what actually happened and to what the consensus called at the time.

Why the Labor Gate Is the Star

If you look at the table carefully, a pattern emerges. Every false alarm — 2015–2016, 2019, 2022–2023 — is filtered by the same mechanism: initial jobless claims never confirmed. The yield curve inverted. The LEI fell. ISM contracted. All the leading indicators waved red flags. And every single time, the Compass held the classification at Late Cycle rather than moving to Contraction, because initial claims, a coincident indicator, didn't confirm.

This is not an accident. It's a structural insight about how recessions actually happen versus how they're predicted.

Leading indicators — yield curves, PMIs, credit spreads — are genuinely forward-looking, but they're also noisy. They capture fear, uncertainty, and financial tightening, which can resolve without a recession if the underlying labor market holds. A company facing tighter credit conditions and a weaker order book does not automatically lay off workers. It first cuts hours, freezes hiring, and runs lean. Layoffs come later, when the situation hasn't resolved.

Initial jobless claims are the first place layoffs show up. When they start rising with consistency — crossing above 240–245k on a 4-week average — the labor market is actually deteriorating, not just threatened with deterioration. That's the difference between a warning and a confirmation.

"The yield curve is a useful but noisy predictor. Every recession has been preceded by an inversion. Not every inversion has been followed by a recession."

— Arturo Estrella & Frederic Mishkin, Federal Reserve Bank of New York Research, 1996 →

The 2022–2023 episode is the clearest illustration. By mid-2022, the Bloomberg Economics recession model was publishing a 100% probability of recession within 12 months. ^[1] The Conference Board's LEI fell for 18 consecutive months, the longest streak since 2008. ^[2] The 2–10 year yield curve inverted to −1.07%, the deepest inversion since 1981. ^[3] Every single leading indicator was screaming recession.

The S&P 500 fell 25% in 2022, partly because of this consensus. Institutional allocations shifted defensively. Retail investors panicked. And then the recession never came. The S&P rallied 24% in 2023, entirely recovering the loss and then some. The investors who held through the false alarm made money. The ones who repositioned defensively based on the leading indicator consensus missed the recovery entirely.

The Compass framework, with its labor gate, would have kept you at Late Cycle throughout. Initial claims peaked at around 260k in March 2023 and then fell steadily back below 230k. ^[4] The gate never fully opened. In June and August 2023 — the two months where claims briefly crossed 245k — the model did register Contraction, but those readings reversed within weeks as claims pulled back. The classification never sustained a move to Contraction, and it never called a recession that didn't exist.

The structural insight in one sentence: leading indicators warn; coincident indicators confirm. A framework that can tell the difference between a warning and a confirmation will, over time, dramatically outperform one that treats all warnings as confirmations.

It Was Early and Right at Real Tops

The other thing worth noting is the 2006–2007 classification. The Compass would have flagged Late Cycle while the consensus was still calling a soft landing. Bernanke's "subprime is contained" comment came in March 2007, nearly a year after yield curve inversion began. ^[5]

Being early to caution is the correct error to make for a risk management tool. An investor who reads Late Cycle in 2006 doesn't lose money. They reduce leverage, stop adding risk, build some cash. If the soft landing had materialized, they'd have missed some upside but preserved capital. When the crisis materialized instead, they were positioned for it.

The asymmetry matters. The cost of false caution is foregone upside. The cost of false confidence is capital destruction. For a fund built around capital preservation as the precondition of compounding, the Compass's tendency to be early-and-cautious rather than late-and-complacent is a feature, not a limitation.

The Honest Limitation: 2020

No macro framework predicted COVID-19. To claim otherwise would be lying. The 2020 recession was exogenous: a global pandemic that shut down economic activity by government decree in a matter of weeks. No yield curve, no PMI, no credit spread was going to flag that in advance.

The data makes this precise. In February 2020 — the NBER-designated recession start — the model was at Expansion, with claims at 211k, a historically tight labor market. In March, as claims spiked to 2.3 million and then 4.6 million in April, the model moved to Late Cycle. It did not reach Contraction during the official three-month recession window. The composite score peaked at 0.58 in April, just below the 0.60 threshold, and by May the recovery was already registering in the data. A monthly model simply cannot fully classify a recession that begins and ends in a quarter.

There is one mild mitigation. The 2018–2019 yield curve inversion had already moved the Compass to Late Cycle going into 2020. An investor following the framework was already positioned cautiously — not fully hedged, but not at maximum risk either. That pre-positioning was the model's real contribution to 2020: not the in-recession read, which was mechanically impossible to complete in three months, but the caution it had established in the 12 months prior.

This is also a good example of why framework-based investing is preferable to prediction-based investing. A prediction model for 2020 that correctly called the pandemic would require a model that predicted the pandemic, which is a different problem entirely. A regime classification model that moves in the right direction when the world changes — even if it can't complete the classification before the shock resolves — is still enormously valuable. It tells you what posture to hold, even when you couldn't have known the world was about to change.

What This Means for How We Invest

The back-test is encouraging. But the more important point is what it implies about process.

The single most common forecasting error of the last 25 years is not failing to call a real recession. It's calling a recession that never happened, repositioning defensively based on that call, and then watching the market rally while sitting in cash or short. That error has cost more than any missed downturn.

The Compass framework doesn't prevent all errors. Nothing does. But it has a specific structural advantage: it requires labor market confirmation before moving to Contraction. That single rule would have saved the 2015–2016, 2019, and 2022–2023 false alarms — three of the most expensive forecasting mistakes of the era.

Right now the Compass reads Expansion at 0.65 confidence. The yield curve has normalized to +0.49% — no longer inverted, but flattening. ISM Manufacturing is at 52.7, still in expansion territory. Initial jobless claims sit at 209k, a historically tight labor market. CPI at 3.9% remains the one persistent warning — inflation above target constrains the Fed's ability to ease if conditions deteriorate.

That's the whole job.