Does optimized portfolios outperform the naive diversification: Implications from joint tests Can optimized trading strategies significa ntly outperform the naive diversification when correcting for data-snooping

Abstract

This thesis expands upon the debate surrounding the paper of DeMiguel, Garlappi &Uppal (2009). We investigate the performance of optimized strategies compared to the naive 1/N rule while controlling for data-snooping. Using the Sharpe ratio and theFFC4 alpha as performance measures, we investigate 10 basic portfolio strategies with datasets from the US and Norwegian markets. We attempt to answer two weaknesses of previous studies on the topic by accounting for data-snooping using White’s RealityCheck (WRC) and the Superior Predictive Ability (SPA) test. In addition, we include the alpha measure in order to account for the established factor premiums present in the datasets. When we conduct joint-tests on the US datasets, most of our findings show little empirical evidence towards at least one active strategy significantly outperforming the naive benchmark. In the joint-test results from the Norwegian datasets, we find evidence towards at least one strategy significantly outperforming the benchmark. Our study highlights the difference in capital markets while also calling into question the value of asset allocation from optimized strategies.