1. Are historical covariances or means more trustworthy as estimators of the future?
2. Why do some statistical packages estimate covariances differently (and different from those we computed in this chapter)? Does the same problem also apply to expected rates of return (means) and betas?