Forecasting price directions : a transaction level approach

The main focus of this work is theanalysis of price direction processes at transaction level ofdifferent stocks traded at the NYSE. Three different models havebeen applied to the data, namely the autoregressive conditionalmultinomial model, a probit model with latent ARMA-process anddifferent specifications of Markov chain models. The models havebeen assessed by a description of their strengths and weaknessesin modelling probabilities and by the application of non-nestedtests. The implications of the decimalisation at the NYSE on pricedirection processes were examined in a further application.