A Comparison of Alternative Dividend Forecasting Techniques

Abstract

Dividend forecasting is a primary element in event-type studies which examine the market's reaction to information conveyed by dividend announcements. This study compares four dividend forecasting models, the naive model, the univariate time series model, the vector autoregressive model, and an analyst forecast, for a random sample of 50 firms for each of three years 1989-1991 and for forecast horizons of one to four quarters. The empirical results indicate that Valueline forecast makes significantly better dividend forecasts than the rest of three models. These results are consistent with rationality in the market for analyst forecast.