Completion time structures of stock price movements

This paper proposes to model movements in more than a century of daily US stock prices as the outcome of a multi-state marked point process and studies the time it takes for stock prices to complete an up or a down move of a certain size. We present a new econometric specification for a class of dynamic models that account for autoregressive conditional duration effects. We also present a method to account for the effect of time-varying state variables that may change within a duration. We find strong evidence of dynamic dependencies in the direction and speed of stock price movements. Past interest rates are also found to affect the speed and direction of completion times. Out-of-sample prediction results show that forecasts of the direction of moves in stock prices can be greatly improved by including covariates such as interest rates and allowing for dynamics in the econometric specification. © Springer-Verlag Berlin Heidelberg 2005.