Risk modeling in the stock exchange with the approach of nonlinear Bayesian models-time-varying parameters

Traditional models do not have sufficient ability to predict the return on investor portfolio due to changes in the external environment (systematic risk) and the internal environment (non-systematic) and this is mainly due to the identification of the explanatory variables and the experimental design of the model.Therefore, the present research, while explaining this issue and in order to adjust the uncertainty problem of the model, by averaging all the models (Bayesian averaging), has determined the effective risks on stock returns in Iran.The present study expresses this failure in identifying explanatory variables and empirical model design. The statistical sample of the research includes 138 listed companies in the period 1390 to 1399.In this study, 62 risks affecting stock returns in the form of 31 indicators in the field of systematic risk and 31 non-systematic indicators entered into nonlinear Bayesian models with time-varying parameters.The results show that among BMA, TVP-DMA, TVP-DMS, BVAR and OLS models, the TVP-DMA model is the most efficient model. According to the TVP-DMA model, 10 non-fragile risks include systematic risks (real GDP growth rate, unofficial market currency, inflation rate, interest rate) non-systematic risks (instantaneous ratio, liabilities, cash flows from operations, return on equity, debt ratio, and price-to-earnings ratio) as the most important risks affecting stock returns. All the mentioned risks, except interest rate and debt ratio, have a positive effect on stock returns.