Time series analysis of corporate quarterly earnings

This paper has examined the time-series properties of the earnings per share series of twenty companies, observed quarterly during the 1977-93 period. The goodness-of-fit properties of five forecasting models for quarterly accounting data was evaluated. Goodness-of-fit was examined by comparing the standard deviations of each series whenusing the five models. The five models are: 1. Foster’s ARIMA(1,0,0)(0,1,0); 2. Griffin’s ARIMA(0,1,1)(0,1,1); 3. Brown-Rozeff ARIMA(1,0,0)(0,1,1); 4. Winter’s seasonal exponential smoothing; 5. Specific ARIMA model, developed on a each firm’s basis. The main results of this study are: a) individual models are in most of the cases the best or the second preferred models from the five ones analyzed, b) parsimoniously, quarterly earnings per share can be generally described as a seasonal process dependent on the adjacent quarter’s performance and fluctuations ( an ARIMA(1,0,0)(0,1,1) process), and c) quarterly earnings per share models that use a longer past history of the companies (Winter’s models) perform well especially for the banking industry