Multilayer Modeling of a Market Covariance Matrix

摘要

Asset allocation relies on a sound understanding of the relations between and among all markets under consideration, but estimating the required forward-looking market covariance is far from straightforward. The multilayer approach presented here separates the generation of the matrix into manageable steps. Its key advantages are two. First, the approach deals with relations between markets and factors, rather than relations between markets directly—that is, it deals with a lower order of magnitude. Second, the resulting matrix is consistent, which is indispensable if one is to construct portfolios appropriately.