Capturing common components in high-frequency financial time series: A multivariate stochastic multiplicative error model

We model high-frequency trading processes by a multivariate multiplicative error model that is driven by component-specific observation driven dynamics as well as a common latent autoregressive factor. The model is estimated using efficient importance sampling techniques. Applying the model to 5 min return volatilities, trade sizes and trading intensities from four liquid stocks traded at the NYSE, we show that a subordinated common process drives the individual components and captures a substantial part of the dynamics and cross-dependencies of the variables. Common shocks mainly affect the return volatility and the trade size. Moreover, we identify effects that capture rather genuine relationships between the individual trading variables.