A new online method for event detection and tracking: empirical evidence from the French stock market

Abstract

In this article we propose a new approach in event studies based on a hidden Markov chain combined with a classical event study model. The number of states informs us about the number of significant events affecting the related market, and the identification of the hidden states determines exactly the delimiters of the event period. Studying each state parameters allows us to examine the events' effect on the related market and to compare results to traditional event analysis. Extensive Monte Carlo simulations and preliminary examination of real data in the French stock market show promising results.