Epstein-Zin preferences have attracted significant attention within the macro-finance literature based on DSGE models as they allow to substantially increase risk aversion, and consequently generate non-trivial risk premia, without compromising the ability of standard models to achieve satisfactory macroeconomic data coherence. Such appealing features certainly hold for structural modelling frameworks where monetary policy is set according to Taylor-type rules or seeks to minimize an ad hoc loss function under commitment. However, Epstein-Zin preferences may have significant quantitative implications for both asset pricing and macroeconomic allocation under a welfare-based monetary policy conduct. Against this background, the paper focuses on the impact of such preferences on the Ramsey approach to monetary policy within a medium-scale model based on Smets and Wouters (2007) including a wide range of nominal and real frictions that have proven to be relevant for quantitative business cycle analysis. After setting an empirical benchmark that generates a mean value of 100 bp for the ten-year term premium, we show that Epstein-Zin preferences significantly affect the macroeconomic outcome when optimal policy is considered. The level and the dynamic pattern of risk premia are also markedly altered. We show that the effect of Epstein-Zin preferences is extremely sensitive to the presence of real rigidities in the form of quasi-kinked demands. We also analyse how this effect can be linked to a combined e¤ect of capital accumulation and wage rigidities.