In this paper we study whether financial frictions should be accounted for in a DSGE mo­del of Russian economy. We compare the baseline two-sector DSGE model of a small open eco­nomy with its version extended by financial accelerator and another version which assumes an agency problem between bankers and depositors. Using calibrated versions of these models, we show how the assumptions regarding the peculiarities of financial market change the transmission mechanisms of macroeconomic shocks. Specifically, the responses of investment and consequently other variables depend on the dynamics of risk premium. In case of financial accelerator model risk premium depends on net worth and leverage ratio of capital owners. In case of agency problem model financial position of bankers drives changes in risk premium. As a result, the risk premium either changes in the same direction in both models or changes in the opposite way. It determines whether the reaction of investment is amplified in case of financial frictions or not. Estimation of all three models using the same data set which does not include data on risk premium allows us to conclude that the baseline model fits the data better than models with financial frictions. However, the difference between the baseline model and the financial accele­rator model is not that substantial. Estimation of two financial frictions models on the full data set which includes data on risk premium shows that the financial accelerator model is strongly preferred to the agency problem model. In addition, impulse response functions from estimated models indicate that accounting for financial frictions can noticeably alter our assessment of transmission of various shocks. For example, if we do not account for financial accelerator, we can underestimate the positive response of output to government consumption shock and underestimate the reaction of output and inflation to monetary shocks. Moreover, financial sector shocks play a non-negligible role in explaining the fluctuations in output and other variables in historical data. We conclude that optimal economic policy decisions require using a combina tion of DSGE models with different financial sector assumptions.