1
$\begingroup$

I have constructed an adjusted Mean-Variance portfolio optimization method that optimizes the exposure in a set of X assets.

The portfolio works perfectly fine during normal periods (even when there are negative returns).

The problem is that I am using 2 years of data to construct the portfolio. If a crisis happens (e.g. during 2007-2009), then this affects my weights 2 years later.

enter image description here

Are there any known methods or techniques that I can think of to protect against such movements?

$\endgroup$
  • $\begingroup$ Just so I understand, you are using the most recent two years of data to estimate both the covariance matrix and the expected returns? $\endgroup$ – noob2 Jan 26 '17 at 18:17
  • $\begingroup$ Yes indeed. Simply a rolling window of 2 years. Thats why the weights are not proper when estimating them using a period of stress for a non-stress period. $\endgroup$ – JohnAndrews Jan 26 '17 at 18:21
  • $\begingroup$ For covariance two years is a reasonable period. But for estimating returns 2 years is far too short, the resulting estimated are too variable. You should use much more data, say 2 decades, for returns, or other methods of estimation. $\endgroup$ – noob2 Jan 26 '17 at 20:18
  • 1
    $\begingroup$ Did you consider using a framework such as Black-Littermann which allows you to enforce a certain relaxation of the estimates given your data? I am not sure if you get your question correctly, so I hesitate to write a full answer: Is your concern that a crisis periods leaves you with unreasonable estimates (for example variances way to big), and you want to ensure that your weights reflect that these estimates are probably not valid anymore? $\endgroup$ – muffin1974 Jan 31 '17 at 14:17

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.