Mitigating Estimation Risk in Asset Allocation: Diagonal Models Versus 1/N Diversification

Abstract

DeMiguel, Garlappi, and Uppal (DGU, 2009) evaluate 14 models of optimal asset allocation and find that none can consistently outperform the 1/N naive diversification strategy, which highlights estimation-risk concerns. Building from Stevens (1998), we provide a useful dichotomous classification of asset-allocation models based on which elements of the inverse covariance matrix that a model uses: diagonal-only vs. full-matrix. In DGU (2009), the competing active strategies all belong to the full-matrix class. We argue that parsimonious diagonal-only strategies, that use limited information such as volatility or idiosyncratic volatility, are likely to offer a good tradeoff between incorporating limited data information while mitigating estimation risk. Using five different data sets of disaggregate portfolio returns over the 1926-2012 period, we find that 1/N is generally not optimal when compared with these simple diagonal strategies.