Optimizing the “Swensen Six” Portfolio

To answer the question of how much the “Swensen Six” portfolio changes when placed under the optimization microscope, here is the following analysis.  The percentage changes from asset class to asset class are not as different as one might expect.

This material is not available for publication elsewhere on the Internet.

[Read more...]

Optimizer Warnings From William Bernstein and Harold Evensky

Lisbon, Portugal

Lisbon, Portugal

A few years ago I worked with a Mean-Variance Optimization (MVO) program designed by David Wilkinson for William Bernstein.  There was a mind numbing amount of manual data entry and there was the perpetual problem of data accuracy.  Was I using the right data and was it current?  The software required entering constraints, otherwise the MVO software would pump out portfolios that made little sense.  Bernstein, in his first book, warns of the inherent dangers in working with MVO software.  Let me quote Bernstein at length.  “Be aware that you are entering a sensitive area for most financial professionals.  Most ‘retail’ investment professionals such as mutual fund salespeople and brokerage ‘account executives’ are at best only dimly aware of portfolio theory and MVO.  Those that are familiar with these areas form the elite of the investment business, and tend to be managers of large investment pools.  These folks treat portfolio theory a little like the trade secrets of a medieval guild,; don’t expect a lot of help from them.”  

Bernstein goes on to explain how optimization might be helpful.  “…suppose you are wondering about the role of, say, precious metals equity (PME) in your allocation.  You would then set up a simple MVO analysis consisting of three assets: the stock and bond portions of your portfolio and PME.  You might then adjust the return of PME up or down in order to determine the returns required for its inclusion in a portfolio.  (Of course, you will need to have a good idea of its SD and correlation with the rest of the portfolio in order to do this.)  If your analysis shows that precious metals equity starts appearing in your portfolio at a return of, say, 5%, then it might be reasonable to use it.  On the other hand, if your analysis shows that a return of 10% is required, you might be wary, as the long-term return of precious metals equity is likely not that high.”

Harold Evensky provides this warning.  “Any wealth manager who unquestioningly accepts the allocations recommended by an optimizer is likely to be a threat to his clients’ financial well-being.”

With these warnings lodged in the frontal cortex, how does one make, what I consider the most difficult decision of investing – what percentage to invest in each asset class that makes up the portfolio?  Back to Evensky’s advice.  An MVO requires the wealth manager to make a decision regarding the investment time horizon and, for each asset class included in the portfolio, and estimate of the:

  • Expected return.
  • Expected standard deviation.
  • Expected correlation with every other asset class.”

Low and behold, the Quantext Portfolio Planner (QPP) software has all this data stored away ready to use.  Fortunately, an Excel guru showed QPP users a way to access the data, and by using Solver in Excel, optimize on any single metric such as projected portfolio return, projected portfolio standard deviation, Return/Risk ratio, portfolio beta, portfolio yield, Diversification Metric, or one designed by the end user.

While there is still a lot of art and “science” in the use of any MVO, by applying reasonable constraints available within Solver, one can remove some of the guesswork from the most difficult decision of portfolio construction.

Optimized Portfolio Using Core ETFs

As mentioned in the last post, I’m working on an optimization worksheet that connects directly with Quantext Portfolio Planner (QPP).  The following analysis is the result of optimizing the Return/Risk to its maximum setting using the array of ETFs shown below.  As one might expect, forcing the Return/Risk ratio to its maximum setting vs. a more modest 0.60 results in a concentration of the investment assets in a few ETFs.

[Read more...]