The Granger Centre for Time Series Econometrics

GC 19/05: Dynamic discrete mixtures for high frequency prices



The tick structure of the financial markets entails that price changes observed at very high frequency are discrete. Departing from this empirical evidence we develop a new model to describe the dynamic properties of multivariate time-series of high frequency price changes, including the high probability of observing no variations (price staleness). We assume the existence of two independent latent/hidden Markov processes determining the dynamic properties of the price changes and the excess probability of the occurrence of zeros.  We study the probabilistic properties of the model that generates a zero-in ated mixture of Skellam distributions and we develop an EM estimation procedure with closed-form M step. In the empirical application, we study the joint distribution of the price changes of four assets traded on NYSE.  Particular focus is dedicated to the precision of the univariate and multivariate density forecasts, to the quality of the predictions of quantities like the volatility and correlations across assets, and to the possibility of disentangling the different sources of zero price variation as generated by absence of news, microstructural frictions or by the offsetting positions taken by the traders.

Download the paper in PDF format


Leopoldo Catania, Roberto Di Mari and Paolo Santucci de Magistris


View all Granger Centre discussion papers | View all School of Economics featured discussion papers


Posted on Wednesday 13th March 2019

The Granger Centre for Time Series Econometrics

School of Economics
University of Nottingham
University Park
Nottingham, NG7 2RD