This paper contains a test of a new aspect of the informational content of dividend; namely, is there information in the timing of the announcements? The empirical evidence indicates that the market expects ‘bad news’ to be delivered late and that these expectations are confirmed. Mean excess returns of stock prices around late announcements are, depending on the assumed returns generating process, either significantly negative or insignificant while significantly positive around the entire population of announcements. Moreover, the proportion and magnitude of dividend reductions associated with late announcements are significantly larger than in the complete universe of announcements.