Empirical Research Of Earnings Persistence And Earnings Forecasting

Accounting earnings reflect a company's operating results; reflect the essential value of the company's existence– to create wealth for shareholders. Therefore the information users, including managers, shareholders, creditors and government, are very concerned about the earnings information. Earnings persistence is an important feature of accounting earnings; it means that accounting earnings maintain steady in a long period (ZhaoYan 2007). When a firm has high earnings persistence, the firm's earnings change largely during a period; when a firm has low earnings persistence, the firm's earnings change little during a period. Earnings persistence made people can use earnings forecasting models to predict future earnings.This paper concentrates on the earnings persistence of companies of the Chinese security market and analyzes the characteristics of accounting earnings. It select earnings'variation coefficient as the variable of earnings persistence, to analyze the different characteristic of different industries independently. Chinese listed companies are divided into 12 categories. This study selects samples of different industries based on the classification of that. It uses the earnings information of year 1994 to year 2000, calculates the A ( A=σ/EPS) of every sample. The result shows that: the average A of manufacturing industry and wholesale and retail trade industry (1.1298 and 1.6532) are significantly higher than that of IT industry, real estate industry and other combined industries (0.9277,0.3982 and 0.8621) . This means that the earnings of different industries have different characteristics. Earnings of manufacturing enterprises fluctuate a lot during years, its earnings persistence is at a low level of the whole; Earnings of wholesale and retail trade enterprises are of more significant fluctuation, its earnings persistence is also at a relatively low level. The earnings persistence of IT industry and real estate industry are relatively higher.Earnings persistence is a key factor which can decide the impact of post earnings on future earnings. In this study, A is added to the earnings forecasting model (na?ve model), adjusting the post earnings variable, to study if the earnings persistence can affect future earnings. In specific empirical analysis, this paper use EPS as a measure of earnings, use A as a measure of earnings persistence to adjust post EPS. The model is as follow: EPS t +1 =β0 +β1 * 1A*EPSt+ε. The study selects more than 2000 samples of the manufacturing industry, wholesale and retail trade industry, IT industry and real estate industry for regression analysis. Because there are not enough samples in some industries and the samples are used as cross-section data, the regression result was not very significant: only manufacturing industry passed 0.01 level significance test; whole- industries and estate industry passed the 0.1 level significance test.In order to expand the sample number, this paper uses a pool data from 2000 to 2006 of the manufacturing industry. Every year is considered as t period, and use the former seven years EPS to calculate A. Moreover, a mute variable is added to the model to instead if the t period is after2000 (Chinese accounting standards changed in year 2000), EPS t +1 =β0 +β1 * 1A* EPS t+β2*D+ε. Finally, contrast the regression result with that of naive model. Regression results showed that: coefficient of adjusted variable passed 0.01 level significance test. In this paper, the adjusted model's F value is 151.985, while the naive model's F value is 102.734. This means the model which is added earnings persistence variable could better reflect the relationship between post earnings and future earnings. This provide us a new idea: in earnings forecasting field using earnings persistence (A) to adjust post earnings would make better forecasting.After that this paper continues to study the market value of the forecasted surplus. We use our adjusted model calculate the forecast earnings of t+1 period, and study if it have effect of the stock price. Regression analysis result of manufacturing industry's pool data from 2000 to 2006 shows that: the forecasted earnings have significant impact on the related stock price.