Modeling the information content of quarterly earnings announcements Reducing predictor variables

Investigating corporate characteristics affecting the information content of corporate earnings provides the basis for the optimal presentation of financial information; however, the main problem is the lack of a documented model in the field of earnings information content. Accordingly, the aim of the present study is to model the factors affecting the information content of quarterly earnings announcements using predictor variable reduction methods. This research is exploratory in terms of its purpose. 131 companies on the Tehran Stock Exchange were selected during the years 2007 to 2024. Dynamic averaging, Bayesian, and selection models were used to reduce predictor variables. Based on the results, the BMA model had the highest accuracy among the models studied. Accordingly, 67 identified variables affecting the information content of quarterly earnings in 5 audit categories; financial ratios; macroeconomic variables; corporate and management governance indicators were entered into the BMA model. Based on prior probabilities, 24 variables were identified as variables affecting the information content of earnings. Based on the results, nonlinear models are more efficient in developing the optimal model of earnings information content. Among the nonlinear approaches, Bayesian averaging models are more accurate than other nonlinear approaches.