Empirical Models of the Intraday Process of Price Changes and Liquidity : a Transaction Level Approach

The development of a feasible and flexible model for the intraday transaction process is an ongoing topic in the empirical analysis of market microstructure. This work attempts to modify and extend the models of the transaction process suggested by Rydberg and Shephard (1998) and Russell and Engle (1998). The main contribution is twofold:First of all, a new and flexible econometric model of the intraday process of price changes is developed. In this work, a classical quantal response model, e.g. a Probit, is augmented by a latent ARMA type dynamic relying on the concept of generalized residuals. This model employs an observation driven dynamic in the sense of Cox (1981). In contrast to competing approaches the model proposed here allows a parsimonious modelling strategy and stationarity conditions of the latent process take a simple form. In addition, this approach allows a straightforward extension to multivariate models. This is of particular importance if one attempts to assess exogeneity or simultaneity relationships among economic variables.The use of this dynamic quantal response model is not only limited to the analysis of price changes. Lately, quantal response models have become increasingly popular in the analysis of time series, e.g. business cycles, interest rate changes, and credit scores.The second major contribution of this work is an attempt to disentangle the relationship between the process of price changes and the process of trade intensity. This exploits the fact that the new dynamic quantal response model proofs to be a valuable building block for multivariate models. To allow for enough flexibility, the analysis of the simultaneity between the transaction intensity and direction of price changes and the size of price changes respectively is carried out in separate models. Empirical evidence on this relationship is provided based on a sample of BUND futures trading at the DTB in Frankfurt.