Abstract:
This study aimed to make a comparison of two earnings forecasting models, the mean reversion model and the naive time-series model. The first one is the model testing for the mean reversion behavior in profitability and earnings and whether that behavior leads to a predictable variation in earnings. The second uses only the lagged variables of earnings to forecast the future earnings. The study covers firms listed in the Stock Exchange of Thailand (SET) through the 1992-2003 periods. The study finds that, from the mean reversion model, there is a mean reversion in profitability in Thailand at a rate of 30% a year. This mean reversion in profitability also leads to a predictable variation in earnings, although small. The naive time-series model results show signs of mean reversion as well. When exporting firms are excluded, the results from both models are not much different from the results of all firms. When the forecast values from both models are put to the test of forecasting accuracy, the finding confirms that the mean reversion model gives a superior forecast. However, the forecast values are still far from the actual values.
