Cynics often refer to mean-variance optimizers as error maximizers because they believe that small input errors lead to large output errors. In many cases, however, this view arises from a misunderstanding of sensitivity to inputs. Consider optimization among assets that have similar expected returns and risk. Errors in the estimates of these values may substantially misstate optimal allocations, but the return distributions of the correct and incorrect portfolios will nevertheless likely be quite similar. Now consider optimization among assets that have dissimilar expected returns and risk. Errors in these estimates will have little impact on optimal allocations; hence again the return distributions of the correct and incorrect portfolios will not differ much. Some examples can illustrate these points. The first example concerns allocation of an equity portfolio across developed market countries. Exhibit 1 shows estimates of expected return, standard deviation, and correlations. The expected returns are those that would obtain if these assets were fairly valued, given their historical covariances with the world equity market measured from January 1980 through December 2005—which is to say that these returns are proportional to their historical betas. The standard deviations and correlations are calculated from the same historical sample. For our purposes it really does not matter how we estimate these values. What is important is the assumption (and it’s only an assumption) that these means, standard deviations, and correlations will prevail in the future.