According to most nancial experts, it is not possible to study the predictability of stock prices without considering the risks affecting stock returns. On the other hand, identifying risks requires determining the share of risk in the total risk and the probability of risk occurrence in different regimes. Accordingly, different DMA models with full dynamics compared to TVP-BMA, BMA and TVP models have been used in the present study to provide this predictability. Findings showed that the DMA model is more efficient than other research models based on MAFE and MSFE indices. The present research was conducted in the period of 1-2003 to 12-2013 (including 144 periods) and was implemented in MATLAB 2014 software space. As the research results show, the bank interest rate coefficient in 45 periods, the rst lag rate of the bank in terest rate in 37 periods, the in ation rate coefficient in 17 periods, rst lag coefficient of in ation rate in 26 periods, oil price coefficient in 78 periods, rst lag rate of oil price in 85 periods, exchange rate coefficient in 64 periods and rst lag rate of the exchange rate in 35 periods have a signi cant effect on stock returns. The nal conclusion shows that the stock variables of oil price and the exchange rate had the highest impact on stock returns during the studied period.