We propose a simple procedure to model multivariate durations. This is done by specifying two conditional models: anautoregressive conditional durationmodel for on a pooled series of durationsand a logit model for the type marks. We estimate the model by maximising a pseudo-likelihood which is equivalent to estimating the autoregressive conditional duration model and the logit model separately. We illustrate this methodology by modelling the joint dynamics of the trade shares of Tabcorp Holdings Limited and Tatts group Limited in the Australian financial market between January 15 and January 31 2009.