Markowitz versus 1/N: Evaluating Portfolio Performance Sensitivity to Estimation Errors with An Unbiased Monte Carlo Methodology

Publisher: TBC

Abstract:

Traditional strategy-based evaluations of portfolio theory, often involving joint tests of alpha, risk, and portfolio construction (PC) models, yield inconclusive and contradictory results regarding the performance of the mean-variance (MV) portfolio relative to 1/N. This paper resolves the debate by assessing MV portfolio performance sensitivity to estimation errors, utilising a novel Monte Carlo methodology independent of alpha and risk model choices. The results reveal that, despite MV’s theoretical potential for superior performance, heuristics (e.g., 1/N and long/short 1/N) constitute a significant portion of the realistically optimal portfolio opportunity set. This poses a legitimate challenge beyond Markowitz portfolio theory (MPT).

Source: Manual