On the accuracy and efficiency of polynomial approximations in optimal macroeconomic policy determination