A Relaxation On Volatility-Managed Portfolios

We consider the performance results of the volatility-managed portfolio and reaffirm past conclusions on the poor out-of-sample performance of any real-time combination strategy of volatility management. Of the possible reasons to this poor performance, we analyze the effect of estimation risk on the optimal portfolio weighting scheme. From this estimation risk, we propose a relaxation of the term 'volatility-managed' such that we can make use of a multinomial logistic regression (Softmax model) that attempts to incorporate the estimation risk in its formulation as well as a heuristic algorithm that incorporates a model selection process. When compared to three other portfolio strategies including the 1=N, volatility-managed, and mean-variance combination, the results show that the resulting Softmax model is consistently among the top performing strategies with regards to different risk-associated performance measures such as the Sharpe ratio and certainty equivalent returns.