17 September 2026 in SSRN Electronic Journal
We develop limit order book imbalance volatility (LOBIV), which measures high-frequency instability in bid-ask depth imbalance, and document a depth-specific volatility anomaly. Using 10-level LOB data for CSI 300 constituents from 2014 to 2025, we estimate a multivariate log-ACD model. Shallow-book LOBIV shocks reduce contemporaneous realized volatility but predict higher subsequent volatility, whereas deep-book shocks exhibit the reverse pattern. These effects are widespread, robust to alternative recursive orderings and crisis regimes, and evolve smoothly across book levels. Market-wide tests further show that shallow-book LOBIV positively predicts future weekly returns, while deep-book LOBIV predicts them negatively after standard controls.
27 October 2026 in SSRN Electronic Journal
We develop limit order book imbalance volatility (LOBIV), which measures high-frequency instability in bid-ask depth imbalance, and document a depth-specific volatility anomaly. Using 10-level LOB data for CSI 300 constituents from 2014 to 2025, we estimate a multivariate log-ACD model. Shallow-book LOBIV shocks reduce contemporaneous realized volatility but predict higher subsequent volatility, whereas deep-book shocks exhibit the reverse pattern. These effects are widespread and robust across crisis regimes, remain consistent under an alternative recursive ordering despite weaker shallow-book evidence, and evolve smoothly across book levels. Market-wide tests further show that shallow-book LOBIV positively predicts future weekly returns, while deep-book LOBIV predicts them negatively after standard controls.