Evaluating the accuracy and stability of management forecasts based on statistical analysis of companies listed on the Stock Exchange Iran

Shareholders and economists are much concerned about predicting market value of companies which are active in stock exchange markets. The predictions are made through two prevalent methods which are based on statistical analysis or based on experiences of company CEOs. Predictions made by CEOs are usually biased and non-scientific; but experiences in other countries have shown that they can identify the growth direction very well. However, managers of Iranian companies refrain from expressing those issues explicitly. A second method based on regression modeling exists which its levels of accuracy and reliability are examined throughout the present study. 102 active companies in bourse were monitored during an 8 year period of activity up to 2012. Statistical results have shown that although highly accurate regression models could be fitted in order to perform predictions, but extrapolations for future years lacked appropriate accuracy and were unstable. Models used for predicting were: decision tree, CART, neural network (NN) and multi-phase multivariate linear regression (MLR). Hypotheses were tested using t-test, Pearson correlation test, P binomial ratio test and Weiner’s T accumulation meta-analysis test.

Evaluating the accuracy and stability of management forecasts based on statistical analysis of companies listed on the Stock Exchange Iran Keywords:

Iranian bourse , stocks value , analytic prediction , estimation accuracy , prediction reliability