The Problem With Optimizers
Backward-looking data can lead to distorted results.
Portfolio optimizers--which have been widely available to individual investors and financial advisors for about 30 years--have an understandable appeal. By simply plugging in a few numbers for risk, returns, and correlation coefficients, users can quickly crunch the numbers and get a recommended asset allocation. The portfolio mix should end up on the efficient frontier, which is the mix of assets that either maximizes expected returns for a given level of risk, or minimizes risk for a given level of returns.
But the problem is a simple one: The results depend heavily on which assumptions are entered (a/k/a garbage in, garbage out). Because past is rarely prologue, the results you get vary widely depending on the periods used to calculate longer-term averages. In this article, I’ll go through three examples to illustrate why the “optimal” portfolio can vary dramatically depending on the assumptions. I’ll also give some tips for avoiding common optimizer pitfalls.