Why does it make sense to use single-period Markowitz mean-variance optimization techniques when we're trying to figure out asset allocation across multiple asset classes (bonds, stocks, REITs, etc)?

Minimizing the variance for a target expected return makes sense to me if you're considering a portfolio of equities. However, including bonds for example complicates things - they have systematic differences from equities such as fixed maturities, periodic coupon payments/reinvestment, depend on interest rate expectations etc. Bonds with a maturity shorter than the investment horizon wouldn't exist at the terminal time in the model. These considerations would naturally be amplified if you're including even more heterogenous asset classes into your portfolio and your optimization problem. The return dynamics would obviously be very different from equities and this seems to be glossed over by the model.

However, I frequently see the standard MVO techniques used to determine asset allocations across equities, fixed income and other classes. Is this theoretically valid? Are there superior techniques for optimizing mixed portfolios?

  • $\begingroup$ Issues about how maturity and credit grades are managed, how reinvestment takes place etc. are "lower level" issues to MVO. One can choose one of many bond total return indexes by ML or Barclays etc that incorporate assumptions one likes and then use such an index's returns and covariances in MVO. $\endgroup$ – Alex C Aug 12 '16 at 0:37
  • $\begingroup$ In other words. You can think of MVO based asset allocation as telling you how to combine 2 to 20 ETFs optimally. The details of how each ETF or index manages it's underlying securities/assets is left out of the analysis. $\endgroup$ – Alex C Aug 12 '16 at 8:28
  • $\begingroup$ oh I see - that makes sense. so in practical applications is MVO mostly for optimizing portfolios of ETFs? when you have a portfolio that doesn't consist of ETFs (for instance, some mix of non-index commodities, equity and fixed income), is MVO worthless? what do people use for asset allocation in those cases? $\endgroup$ – beeba Aug 12 '16 at 13:23
  • $\begingroup$ @beeba, did you solve this issue with non-index assets? $\endgroup$ – AK88 Apr 16 '17 at 13:32
  • $\begingroup$ Not really - for non-index assets I used the closest possible publicly traded asset (eg REITs for real estate) to try and proxy returns and correlations, and incorporated these as inputs to a broader multiperiod simulation for optimization (rather than 1 period MVO). Have you come across anything that might be a better approach? $\endgroup$ – beeba Apr 17 '17 at 13:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.