Suppose a one-year Treasury yields 5% and investors expect the one-year rate to be 3% a year from now. A two-year note has to offer about the same return as buying two one-year notes in a row, so it yields roughly (5 + 3) / 2 = 4% a year. Check: $100 grows to 100 × 1.05 × 1.03 = $108.15 the first way and 100 × 1.04 × 1.04 = $108.16 the second. The two-year yield is below the one-year yield, so this stretch of the curve is inverted, and the reason is that the market expects rate cuts. Observed yields also contain a term premium, which can be positive or negative, so a long yield is never a pure forecast.
That is why the shape of the curve gets so much attention. Every U.S. recession since 1969 was preceded by an inverted curve, with the ten-year yield below the three-month rate, and a New York Fed study found that this spread beat other indicators at predicting recessions two to six quarters ahead. But the signal has misfired. The curve was inverted from September 1966 to January 1967 and no recession followed. And from July 6, 2022 to August 26, 2024 the ten-year yield sat below the two-year yield on every trading day, the longest inversion since that series began in 1976. As of November 2026 the NBER has not dated a recession since the one that ended in April 2020.
An inversion, then, is not a recession forecast with a date attached. It says investors expect lower short rates ahead, and that can happen because inflation is expected to fall, not only because the economy is expected to shrink. The level of rates matters as much as the slope, and a quoted rate says little until inflation is taken out (see Real vs. Nominal). This page stays with how rates across maturities are determined. How the Fed moves the short end is in Monetary Policy, and what sets the real rate in the long run is in Loanable Funds.