Advances in the development of Dynamic Stochastic General Equilibrium (DSGE) models towards medium-scale structural frameworks with satisfying data coherence have considerably enhanced the range of analytical tools well-suited for monetary policy evaluation. The present paper intends to make a step forward in this direction: using US data over the Volker-Greenspan sample, we perform a DGSE-VAR estimation of a medium-scale DSGE model very close to Smets and Wouters  specification, where monetary policy is set according to a Ramsey-planner decision problem. Those results are then contrasted with the DSGE-VAR estimation of the same model featuring a Taylortype interest rate rule. Our results show in particular that the restrictions imposed by the welfare-maximizing Ramsey policy deteriorates the empirical performance with respect to a Taylor rule specification. However, it turns out that, along selected conditional dimensions, and notably for productivity shocks, the Ramsey policy and the estimated Taylor rule deliver similar economic propagation.