A Study On Predicting Future Earnings Of Listed Companies In China Using Financial Basic Information

With the continuous development of market economy,China's capital market has become more prosperous,more and more investors put their investment funds into the capital market for accessing to investment income.To get more investment income,investors need to predict future earnings of listed companies,then adjust investment decisions according to predictions to maximize revenue,so predicting future earnings of listed companies is very important for investors.Financial basic information included in the financial report that listed companies publish regularly can response to the company's past operations,it has been attached to great importance by investors,investors always predict future earnings according to the financial basic information included in the financial report,thus a study on predicting future earnings using financial basic information has strong theoretical and practical significance.The main purpose of this article is to research financial indicators that can affect prediction of future earnings and to build models of future earnings forecast.This paper selected randomly 225 listed companies in China's manufacturing in 2001 as training samples,184 ones in 2002 as test samples.Dependent variable y of logistics regression model and BP neural network model are a standard variable 0-1,when predicting future earnings increase or decrease,it means decrease or increase of earnings per share in year t+1 compared to that in year t.When the company earnings per share in year t+1 compared to that in year t increase,y get 1;instead,y get 0.When predicting future earnings positive or negative,it means positive or negative of earnings per share in year t+1.When the company earnings per share in year t+1 is positive y get 1;instead,y get 0.Dependent variable y of multiple linear regression model is the difference between earnings per share in year t+1 and t.This paper selected 6 financial indicators based on analysis of earnings per share indicator system that was analized in Xiaowei Wang(2001),then selected 11 financial indicators from financial report that can reflect solvency,operational capacity, profitability and development capacity of company a total of 17 financial indicators as independent variables.Firstly,the samples were taken mean difference test,by analyzing the probability of Sig significant,we find there was significant difference between 13 financial indicators,in these indicators,sales margin(χ₄ ),the total profit of main business profit ratio(χ₆ ) such as the 10 indicators passed significant test of the level of 1%,asset-liability ratio(χ₂ ),main business profit margins(χ₅ ) and the ratio of long-term liabilities(χ₁₀) passed significant test of the level of 5%.The result of mean difference test show that indicators selected can be as independent variables of future earnings prediction.When predicting future earnings increase or decrease,we took use of Logistic model to analize independent variables and earnings prediction,then got the test result with method of forward step by step selection.We built prediction model with main business profit margins (χ₅ ),quick ratio(χ₈ ),the ratio of long-term liabilities(χ₁₀),accounts receivable turnover ratio(χ₁₂),total assets growth rate of net profit(χ₁₇).In training samples,prediction accuracy of companies whose future earnings increase is 60.8%,it of companies whose future earnings decrease is 76.6%,total accuracy is 69.8%;in test samples,prediction accuracy of companies whose future earnings increase is 64.8%,it of companies whose future earnings decrease is 53.8%,total accuracy is 59.2%.Prediction result is acceptable but not very good.In order to compare with Logistic model,this paper built BP neural network model with the 5 variables that were selected from the logistic regression model.For training samples, prediction accuracy of companies whose future earnings increase is 74.227%,it of whose future earnings decrease is 75.00%,total accuracy is 74.667%.Compared with Logistic regression model,BP neural network model improves the accuracy of the prediction,but it's not very good,too.Then,we built multiple linear regression model with accounts receivable turnover ratio (χ₁₂),return on equity(χ₁₄),total assets growthrate ofnet profit(χ₁₇).The prediction result show that prediction accuracy of companies whose future earnings increase is 94.51%, prediction accuracy of whose future earnings decrease is 92.47%,total accuracy is 93.48% within 85%confidence interval;prediction accuracy of companies whose future earnings increase is 93.41%,prediction accuracy of whose future earnings decrease is 95.70%,total accuracy is 94.57%within 95%confidence interval.The prediction effection of this model is significantly better than the first two.When predicting future earnings positive or negative,We built Logistic prediction model with 3 independent variables.In training samples,prediction accuracy of companies whose future earnings is positive is 80.6%,it of companies whose future earnings is negative is 73.5%,total accuracy is 79.6%;in test samples,prediction accuracy of companies whose future earnings is positive is 77.7%,it of companies whose future earnings is negative is 77.8%,total accuracy is 77.7%.Prediction result is good.Then,this paper built BP neural network model with the 3 variables that were selected from the logistic regression model.For training samples,prediction accuracy of companies whose future earnings is positive is 83.246%,it of whose future earnings is negative is 79.412%,total accuracy is 82.667%.In test samples,prediction accuracy of companies whose future earnings is positive is 80.12%,it of companies whose future earnings is negative is 61.111%,total accuracy is 78.261%.Compared with Logistic regression model,BP neural network model improves the accuracy of the prediction of companies whose future earnings is positive and total accuracy.This paper built future earnings increased or decreased prediction model and future earnings positive or negative prediction model for the same samples, it is also the innovation.The results of this paper proved that financial basic information in the financial reports of listed companies can be used for predicting future earnings.Finally,this paper points the inadequacies of the research.