Multilevel stochastic dynamic process models and possible applications in global financial market analysis and surveillance

Abstract

This paper advances Multilevel Stochastic Dynamic Process (MSDP) models as a new framework for modeling nonstationary and nonlinear time series and complex systems, and discuss possible applications of MSDP models in global financial market analysis and surveillance. Under the heterogeneous market hypothesis, different types of market participants react to the same information differently, characterized by their time horizons or dealing frequencies. Consequently financial prices exhibit multilevel trends, cycles, and seasonality, which provide the very basis for MSDP models. Both the discrete- and continuous-time MSDP models are constructed: Multilevel Structural Time Series (MSTS) with Unobserved Components (UC) and Multilevel Stochastic Differential Equations (MSDE).