Disentangling reactive versus insensitive components of order flow in a full limit order book setting

Ascertain whether, in a modeling framework that includes limit orders and cancellations (full limit order book data), one can decompose global order flow into a component that reacts to child orders (e.g., stimulated refill) and a component that is insensitive because it reflects preplanned execution of other metaorders, and develop identification methods to quantify these components from publicly available market data.

Background

The authors note that their setting focuses on child orders executed as market orders, and highlight the need to consider limit orders and cancellations to better capture microstructural reactions to metaorders. They pose whether reactive order flow (e.g., liquidity replenishment or stimulated refill) can be disentangled from non-reactive flow driven by other metaorders' schedules.

Solving this identification problem is especially relevant for limit order book models and could clarify why purely statistical models, even when accurately capturing autocorrelations, fail to reproduce empirical metaorder price trajectories. A methodology for separating these components would enable more accurate impact forecasting and cost estimation.