Survival Models for the Duration of Bid-Ask Spread Deviations

TL;DR

This paper introduces a survival regression framework to model the duration of bid-ask spread deviations, capturing intra-day liquidity shocks and their recovery in limit order books.

Contribution

It presents a novel application of survival models to measure the duration of liquidity deviations, improving understanding of intra-day liquidity dynamics.

Findings

Survival models effectively capture spread deviation durations.

Model selection via branch-and-bound improves explanatory performance.

Dynamic liquidity shocks are quantifiable using the proposed framework.

Abstract

Many commonly used liquidity measures are based on snapshots of the state of the limit order book (LOB) and can thus only provide information about instantaneous liquidity, and not regarding the local liquidity regime. However, trading in the LOB is characterised by many intra-day liquidity shocks, where the LOB generally recovers after a short period of time. In this paper, we capture this dynamic aspect of liquidity using a survival regression framework, where the variable of interest is the duration of the deviations of the spread from a pre-specified level. We explore a large number of model structures using a branch-and-bound subset selection algorithm and illustrate the explanatory performance of our model.