The two-ten yield curve has been inverted, on and off, for nearly three years. By every historical measure, the United States should already be in recession. It is not. The conventional explanation — that the cycle is "stretched" — has reached the point where the explanation needs explaining. Something else is happening.
For five decades, the inverted yield curve has been one of the most reliable recession indicators in macroeconomics. Four out of four times, when the two-year Treasury yield exceeded the ten-year, recession followed within twelve to twenty-four months. The mechanism was straightforward: the Federal Reserve tightens short rates faster than the long end can absorb, credit conditions deteriorate, the real economy follows. The signal had a remarkable consistency from the late 1970s through to the 2007 inversion that preceded the financial crisis.
The curve inverted again in July 2022. By every historical playbook, that meant recession was probable by mid-2023, near-certain by mid-2024. Neither happened. The economy grew. Employment held. Earnings expanded. The curve — by April 2026 — has been inverted, partially or fully, for thirty-three months. The longest inversion without recession in the post-war record was eighteen.
The regime explanation
Most attempts to reconcile the indicator with the data have settled on what we will call the "long and variable lags" explanation: the curve still works, the recession is still coming, the lag is just longer this cycle. This explanation is not testable in any falsifiable way. It also requires accepting that the lag has roughly doubled, which is a substantial shift in a relationship that was previously stable for half a century.
We propose a different reading. The curve is not signalling recession because the curve is not the same instrument it was. The Fed's balance sheet management — quantitative easing, reverse repo, the bank term funding programme — has fundamentally altered how the long end transmits information about expected real growth. The ten-year yield in 2026 is shaped by central bank positioning to a degree that was not true in any previous cycle. The curve has been turned into a policy instrument, not just a market reading.
If the curve is the indicator, and the indicator has been re-engineered, the signal it produces is no longer the same signal.
This matters not as a technical observation but because the entire risk-asset playbook of the last thirty years has been calibrated to the curve as a recession indicator. Position sizing rules, allocation models, derivatives pricing, even the framing of macro funds — all assume the curve means what it has always meant. If it does not, then the consensus is mispriced.
What the data actually shows
The curve in 2024-2026 has decoupled from labour market data, from credit spreads, and from leading indicators in a way that the curve in 2007 or 2000 did not. Specifically: the unemployment rate has remained below historical recession-onset thresholds; high-yield spreads have not widened; the Conference Board LEI has stabilised after eighteen months of decline. None of these are conclusive. Together, they describe a configuration that does not match prior inversions.
2y–10y spread vs unemployment, 2007–2026
SOURCEFRED · Meridian Intelligence Desk. The 2008 inversion produced an unemployment spike within fourteen months; the 2022 inversion has produced no comparable response by the same lag.
The implication for portfolios
If the curve no longer reliably signals recession, the cyclical playbook that worked for the last four cycles is exactly wrong for this one. Defensive rotations into long-duration sovereigns; underweighting credit; raising cash on the inversion signal — each of these has been a losing trade since mid-2023. Funds that ran the playbook are underperforming their benchmarks by hundreds of basis points. Funds that ignored the curve and stayed long the cycle have done well.
We are not arguing that recessions cannot happen. They can, and one will, eventually. We are arguing that the indicator that has framed the question for fifty years has been re-engineered, and that any portfolio framework that treats it as the same indicator is producing the wrong answer.
Where this leaves us
Our positioning reflects the regime read. We hold gold long against persistent dollar weakness; we hold equities long with selective sector calls; we are not short duration but we are not long it either. We treat the curve as one input among many, with conviction sharply lower than its historical weighting in our framework would suggest.
For the consensus reading to be right — for recession to materialise on the original lag-extended schedule — we would need to see credit spreads widen, employment soften, and the LEI break decisively below its current level. None of these are in motion. We will continue to flag any of them when they begin.
Until then: the curve is being misread. The playbook needs rewriting.