Our U.S. research underscores the difficulty of forecasting returns, risks, correlations and cross-correlations for asset classes, sub-asset classes and investment strategies, especially in the short term.

When using a mean-variance optimization model, looking at different periods can lead to different results (see “History can be misleading”). For each set of asset classes, we examined returns, volatility and correlations from 1988 up to the construction date of the portfolio, to determine the most efficient combination of asset and sub-asset classes. We then created optimized portfolios based on the best performance each asset class achieved in the given period, and attempted to replicate performance during a new time period.

We evaluated these portfolios over the next three-year period, considering how different lengths of time and actual periods affected results. We then compared returns of the optimized portfolios with a benchmark of 60% stocks and 40% bonds. When we tried to replicate the positive results of the original portfolios, in most cases the optimized portfolios did worse than the originals. We also found the volatility of the benchmark and model portfolios differed.

No asset class provides high (or low) relative returns forever. Showing clients potential volatility can help you arrive at the allocation suited to each client’s comfort level.

Fast Facts on Gen X
Read more: