Modelling Structural Breaks In Stock Price Time Series Using Stochastic Differential Equations

TL;DR

This paper investigates how quarterly earnings reports cause structural breaks in stock prices by modeling with stochastic differential equations and analyzing parameter changes before and after reports.

Contribution

It introduces a method to detect structural breaks in stock prices using stochastic differential equations fitted with maximum likelihood estimation.

Findings

Earnings reports significantly alter model parameters

Structural breaks indicate models need refitting post-report

Models before and after reports differ notably in parameters

Abstract

This paper studies the effect of quarterly earnings reports on the stock price. The profitability of the stock is modelled by geometric Brownian diffusion and the Constant Elasticity of Variance model. We fit several variations of stochastic differential equations to the pre-and after-report period using the Maximum Likelihood Estimation and Grid Search of parameters method. By examining the change in the model parameters after reports' publication, the study reveals that the reports have enough evidence to be a structural breakpoint, meaning that all the forecast models exploited are not applicable for forecasting and should be refitted shortly.