Using the methodology of logistic regression and a trichotomous dependent variable specification, this research derives models of future dividend change from financial statement data in two time periods. A general model of future dividend change for both time periods could not be identified. As measured by average absolute prediction errors (AAPE's), the models of future dividend change derived in this research outperform many extant models of dividend change, including the Aharony & Swary (1980) model and the models developed by Kolb (1981).Two parsimonious models of future dividend change had lower AAPE's than the Several non-earnings/non-dividend related variables were found to significantly contribute to the explanation of future dividend changes over earnings/dividend related variables. Although different variables were significant in the two different time periods, they were all related to only two constructs, inventory-related or long-term-debt related variables. |