Yield curve inversions are commonly treated as a homogeneous signal of credit stress. Using monthly U.S. data from 1986 to 2024 and six inversion episodes defined by the 10-year minus 2-year Treasury spread, I show that the relationship between yield curve inversion and investment-grade credit spreads is strongly heterogeneous across episodes. Pooled regressions yield no significant relationship, but this null reflects cancellation across episodes with opposing signs rather than a stable zero effect: tests of coefficient homogeneity are decisively rejected for the Moody’s BAA–AAA spread and equally for the Gilchrist–Zakrajšek credit spread and its excess bond premium component, and the rejections are robust to an extended set of macro-financial controls and to differenced specifications. Which episode carries predictive content for future spreads is itself measure-dependent: in within-episode predictive regressions, the 2022–2024 episode is the only one in which the yield curve slope predicts subsequent changes in the BAA–AAA spread, whereas for the Gilchrist–Zakrajšek spread and the excess bond premium the only significant episode is 2000–2001. A decomposition shows the 2022–2024 spread widening is concentrated in the excess bond premium, with default-risk compensation declining. The results caution against pooled inference on inversion effects and against generalizing from any single episode.
Boon Chuan Lim (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: