Optimized portfolios versus Naive Diversification : could the superior performance of optimized portfolios be attributed to established factor premiums?

Abstract

Since the publication of the study by DeMiguel, Garlappi & Uppal (2009), where theydemonstrate that none of the 14 mean-variance optimization strategies outperform thenaive diversification, several studies claim to defend the superiority of portfolio optimiza-tion strategies relative to the naive diversification (see e.g. Kritzman, Page & Turkington(2010), Tu & Zhou (2011), Kirby & Ostdiek (2012)). However, in a recent study byZakamulin (2017), the author states that the superior performance of these optimizedstrategies appears due to exposures to established factor premiums. Motivated by thestudy of Zakamulin (2017), this thesis evaluates the out-of-sample performance of fourrisk-based strategies relative to the naive diversification across 25 empirical datasets pro-vided by Kenneth French. Additionally, we assess whether the (out)performance could beattributed to established factor premiums. We find that three of four risk-based strategieson average deliver superior performance over the naive diversification in terms of Sharperatio, although the performance on the individual datasets varies significantly. Each risk-based strategy generates statistically significant alphas in the CAPM, both on average,and in nearly each dataset. In addition, we show that the superior performance of theserisk-based strategies compared to the naive diversification, and in terms of CAPM alpha,are mostly generated in bear markets. After controlling for several risk factors throughthe Fama-French five-factor model, the alphas of any risk-based strategy becomes neithereconomically nor statistically significant. The main conclusion that we reach in this thesisis that the superior performance of the risk-based strategies is likely to be attributed toestablished factor premiums.