This study proposes a nonparametric regime-detection framework based on differential entropy estimation with heavy-tailed kernels.

By combining entropy and tail-index analysis within a rolling-window approach, the method captures shifts in market behavior without relying on restrictive distributional assumptions.

Unlike variance-based measures, entropy remains informative even in environments with extreme tails and unstable higher moments.

The findings also show that volatility and entropy can diverge materially, implying that standard variance measures alone may fail to capture broader uncertainty and systemic instability.

Want more research breakdowns like this? Subscribe for weekly insights