This dissertation attempts to identify and predict earthquakes in the financial market (financial crashes) using Hawkes processes. Models based on Hawkes processes were first used in the earthquake literature. The dissertation shows Hawkes processes also match investors' self-exciting herding behavior around crashes, which has similar characteristics to the self-exciting behavior of seismic activity around earthquakes. In Chapter 2 an Early Warning System based on Hawkes models is developed that indicates the arrival of a crash within a trading week. EWS based Hawkes model outperform EWS based on well-known and commonly used volatility models, proving that these models do not capture all information that can be used to predict crashes. Specification tests for Hawkes models with a specific focus on testing for cross-excitation, are designed in Chapter 3. Chapter 3 as well as Chapter 4, indicate that shocks in one financial market affect the occurrence (and magnitude) of shocks in other financial markets. Moreover, comparing predictions of models with and without cross-excitation, more accurate predictions are obtained including cross-excitation. The last Chapter (Chapter 5) develops methods to estimate non-affine Hawkes models using the information in option prices with the aid of Machine Learning techniques.

Self-excitation, Hawkes processes, Earthquakes, Jumps, Early Warning System, Specification Testing, Lagrange Multiplier tests, Cross-excitation, non-affine models, option prices, Bayesian estimation, Machine Learning, Particle Filtering, Particle Gibbs, jump risk premium
Ph.H.B.F. Franses (Philip Hans) , H.J.W.G. Kole (Erik)
Erasmus University Rotterdam
Department of Econometrics

Gresnigt, F. (2020, February 20). Identifying and Predicting Financial Earthquakes Using Hawkes Processes. Erasmus University Rotterdam. Retrieved from