Relative Performance of Earnings Prediction Models for Listed Companies in Taiwan

• Main Authors: Lan, Shun-Der, 藍順得
• Other Authors: Yeh Jin-Cherng
• Format: Others
• Language: zh-TW
• Published: 1997
• Online Access: http://ndltd.ncl.edu.tw/handle/93798214555569945337

碩士 === 淡江大學 === 會計學系 === 85 === Recent years, the importance of capital market in Taiwan is getting more and more. The basic analysis of stock is considered important. Since investorspay more attention to earnings forecasts and as their important basis while making investment decisions, it is necessary to compare the preciseness of several earnings prediction models to the purpose of improving the quality of prediction. Besides, in studies of testing the relationship between earningsinformation and stock returns, a precise earnings prediction is one of the important premises for those studies. This study chooses univariate time-series model ( random walk model with a drift ), price-based model and financial analysts as representatives for comparing earnings prediction models, and is based on listed companies. The first is based on 1992 to 1996''s information to compare the three earnings prediction models on two criteria: (1) earnings prediction accuracy, and (2) ability to yield earnings information variables most correlated with stock returns as a comparison of performance. The relative prediction errors are defined as absolute forecast errors. Firm size is then added to study its effect to the preciseness test of different earnings prediction models, and the association between earnings and stock returns. The result are as follows:(1) Fianacial analysts'' prediction accuracy is better than random walk model with a drift and price-based model at a significant level. Later two models'' prediction accuracy has no significant variance.(2) Under the three earnings prediction models, there is no significant contemporaneous association between unexpected earnings and stock cumulative abnormal returns. None of the three is an appropriate proxy for market expectation of earnings.(3) The three earnings prediction models all have better prediction accuracy for big companies at a significant level. Firm size is not systematically ralated to the ralative prediction performance of different earnings prediction models.(4) Fianacial analysts'' forecast is a better proxy for earnings market expectation for big companies. None of the three earnings prediction models is an appropriate proxy for earnings market expectation for small companies.