Static signal definitions and momentum crash risk: Evidence from state-conditional momentum

This paper examines whether momentum crash risk is partly endogenous to the conventional practice of imposing a fixed signal definition across heterogeneous market states. We test this hypothesis in a transparent walk-forward design that allows the momentum ranking rule to vary over time within a restricted and interpretable class of heuristics. Using a survivorship-bias-mitigated panel of 1508 firms that were S&P 500 constituents during 2006–2025, we find that state-conditional signal adaptation delivers shallower drawdowns and stronger risk-adjusted performance than well-tuned static momentum benchmarks. The key finance result is that the adaptive strategy remains meaningfully exposed to momentum in bull markets but partially decouples from the conventional momentum factor in bear markets and transition months. This pattern suggests that part of momentum crash vulnerability arises at the signal-construction stage rather than only at the exposure-management stage. Multiple-testing controls further indicate that the findings are unlikely to be explained by extensive searching alone.