A0215
Title: On the optimal combination of naive and mean-variance portfolio strategies
Authors: Rodolphe Vanderveken - UCLouvain (Belgium) [presenting]
Nathan Lassance - UCLouvain (Belgium)
Frederic Vrins - UCLouvain (Belgium)
Abstract: A disheartening fact in portfolio choice is that the naive equally weighted portfolio often outperforms the estimated optimal mean-variance portfolio out of sample. The value of portfolio optimization is reaffirmed by combining the two portfolios to optimize out-of-sample performance. We show that the seemingly natural constraint that the two combination weights sum to one is unnecessary and has several undesirable consequences. In particular, the resulting portfolio combination overinvests in the mean-variance portfolio and underperforms the risk-free asset for sufficiently risk-averse investors. We derive the combination of the equally weighted and mean-variance portfolios that relaxes the constraint and prove that it avoids the undesirable properties of the constrained combination. Moreover, we demonstrate that even though the optimal combination coefficients suffer from more estimation error than the constrained ones, the optimal portfolio combination delivers better out-of-sample performance for most risk-aversion levels. The empirical analysis confirms the superiority of our approach relative to the previous rule and other benchmarks. In general, our novel portfolio rules deliver out-of-sample gains relative to the equally weighted portfolio and the risk-free asset for any degree of risk aversion, and hence render portfolio theory beneficial to all investors with mean-variance preferences.
