This master's thesis examines the implications of applyingtime-varying covariance structures betweenfour majorasset classes in the US economy within the framework of the conditional Capital Asset Pricing Model (CAPM). The DCC-GARCH-in-mean modelis employed to estimate the time-varying covariance structures. To construct the market portfolio, weights based onmarket valuesare utilized and updated for each time period. The findings reveal significant evidence of time-varying risk premia and risk exposures (beta), thereby supporting the notion of a time-varying covariance structure in the CAPM. However, we also find that intercepts in CAPM equations are significant, in contradiction to the CAPM. Consequently, it is suggested that while there is a need to account for time-varying estimation of market risk, solely relying on this factor may not adequately explain the variations in expected returns. Hence, the results imply the inclusion of additionalfactors to enhance the model's explanatory power.