Optimal Versus Naive Diversification: How Inefficient is the 1/N Portfolio Strategy?

What it asks

DeMiguel, Garlappi and Uppal compare fourteen different optimisation methods against the equally weighted portfolio. Out of sample, 1/N beats most of them. The finding does not mean the theory is wrong: estimation error is large enough to eat up the advantage the theory provides. The amount of data a method would need to beat 1/N is, in practice, usually not available.