Statistical inference for the doubly stochastic self-exciting process

TL;DR

This paper develops a statistical inference method for a time-varying Hawkes self-exciting process, addressing bias issues in high-frequency data estimation and providing a bias-corrected estimator with proven asymptotic properties.

Contribution

It introduces a novel bias correction technique for estimating time-varying parameters in Hawkes processes and establishes its asymptotic normality.

Findings

Bias correction significantly improves estimator accuracy

Method performs well in finite samples

Empirical analysis demonstrates practical applicability

Abstract

We introduce and show the existence of a Hawkes self-exciting point process with exponentially-decreasing kernel and where parameters are time-varying. The quantity of interest is defined as the integrated parameter $T^{-1}\int_0^T\theta_t^*dt$, where $\theta_t^*$ is the time-varying parameter, and we consider the high-frequency asymptotics. To estimate it na\"ively, we chop the data into several blocks, compute the maximum likelihood estimator (MLE) on each block, and take the average of the local estimates. The asymptotic bias explodes asymptotically, thus we provide a non-na\"ive estimator which is constructed as the na\"ive one when applying a first-order bias reduction to the local MLE. We show the associated central limit theorem. Monte Carlo simulations show the importance of the bias correction and that the method performs well in finite sample, whereas the empirical study discusses the implementation in practice and documents the stochastic behavior of the parameters.