The effect of time series properties on the predictive value of quarterly earnings for forecasting annual earnings.

Abstract

This study provides further evidence regarding the predictive value of quarterly earnings for improving the forecasts of annual earnings. Using an analytical model, it is shown that for a specific class of time-series models, the predictive values are determined by the time-series properties, as measured by parameter value, of quarterly earnings. In particular, the model demonstrates that the accuracy of annual earnings forecasts increases as additional quarterly reports become available, and that the time-series model parameter value is positively related to both total improvement and the first quarter's relative improvement in annual earnings forecasts. These theoretical predictions are empirically tested using a sample of 235 firms over a five year period from 1980 to 1984. Empirical results are consistent with the theoretical predictions. First, annual earnings forecasts become increasingly accurate as additional quarterly reports are available, suggesting that quarterly earnings are useful for improving the forecasts of annual earnings. Second, there are cross-sectional variations in the degree of the improved accuracy in forecasts. More importantly, time-series properties (parameter value) of quarterly earnings are an important determinant of the variations in both total and relative predictive values. This result is robust with respect to different time-series models, forecast error metrics, and statistical methods.