The goal of asset allocation is to choose a portfolio to hold for some future period. In Chapter 2 we assumed, implicitly, that we know the true expected returns and covariances of the asset classes. In practice, we must estimate these values based on imperfect information. The inputs to optimization are therefore subject to error, and therefore so are the optimal portfolio weights. Some investors believe that estimation error is so severe as to render optimization a hopeless exercise. We disagree with this extreme view, as we discussed in Chapters 5 and 7. Here are some reasons to remain calm:
- Grouping securities into asset classes that are internally homogeneous and externally heterogeneous reduces noise. We need only estimate the properties of a handful of asset classes, as opposed to hundreds or thousands of underlying securities for which we may lack both information and intuition. (See Chapters 1 and 6.)
- If two asset classes are close substitutes for each other, errors in their expected returns lead to large errors in the optimal weights, but these misallocations across highly similar asset classes have little impact on the portfolio's return and risk, if the portfolio weights are reasonably constrained. If, on the other hand, asset ...