On the use of mean-variance and quadratic approximations in implementing dynamic investment strategies: A comparision of returns and investment policies

This paper compares two approximation schemes for calculating the optimal portfolios in the discrete-time dynamic investment model, specifically, the mean-variance (MV) and the quadratic approximations, to the exact power function method. Future returns are estimated via the empirical probability assessment approach. The results show that (i) with quarterly revision, the MV model approximates the dynamic model very well; (ii) with annual revision, there are often sharp differences between the power function model and the MV approximation; and (iii) these differences become even larger when the quadratic approximation is used.