A0652
Title: Mutually exciting point processes with latency
Authors: Yoann Potiron - Keio University (Japan) [presenting]
Vladimir Volkov - University of Tasmania (Australia)
Abstract: A novel statistical approach to estimating latency, defined as the time it takes to learn about an event and generate a response to this event, is proposed. Our approach only requires a multidimensional point process describing the arrival time of events, which circumvents the use of more detailed datasets which may not even be available. We consider the class of parametric Hawkes models capturing clustering effects in which latency is defined as a known function of kernel parameters, typically the mode of kernel distribution. Relying on a realistic mixture of generalized gamma kernels, the estimation of model parameters is performed via quasi-maximum likelihood and the feasible limit theory with in-fill asymptotics is derived. As a byproduct, asymptotic theory for a latency estimator, defined as the function of parameter estimates and two tests, is deduced. Numerical studies corroborate the theory. Latency estimates for the US and Canadian stock exchanges vary between 2 and 13 milliseconds from 2020 to 2021. The US firms are found to be more involved in relative latency competition, implying different risk appetites for firms with different latencies.
