Real-Time Regime Identification Lag

The dominant failure mode of regime-switching trading models. A model’s smoothed regime inference — which uses the full sample and is available only with hindsight — identifies regimes sharply and matches events like NBER recessions well. But a trader only ever has filtered inference, using data up to the present, which is noisy and lags the true turning point: Shu, Yu & Mulvey (2024) report a median detection latency around 25 calendar days. Dacco & Satchell (1999) prove that only a small real-time misclassification rate is enough to make even the true Markov Regime-Switching Model forecast worse than a random walk. The lag means regime models are useful for de-risking but routinely miss the start of recoveries, capping their tradeable upside.

Connections

• Markov Regime-Switching Model — contradicts, source: https://arxiv.org/html/2402.05272v2
• Dacco and Satchell 1999 — supports, source: https://iaorifors.com/paper/30956