I am new to mean reversion, and I'd like to run an analysis on a fund (ts with monthly returns only) to see if mean reversion applies and if so, when it will happen.

Most of the examples I found focus on cointegrated securities. Ornstein-Uhlenbeck seems to be a popular model for mean reversion estimate. If someone could please provide me an example of applying OU model on a return-only time series or point me to the right direction, I will really appreciate it!


T, while not necessarily a direct answer to your question, I just wanted to offer a word of caution in applying a mean reversion model to security or fund returns.

In attempting to fit a mean reversion model to returns, you are implicitly stating that you believe that historical returns are good predictors of future returns. Although I don't know what your dataset is, this is generally not the case.

More likely to be mean reverting are the fundamentals underlying these returns. Using a financial markets example this might be something like multiples (P/E or EV/EBITDA).

But if you still wanted to run such a mean reversion analysis, I would recommend starting with a regression with a format similar to the following: use as your independent (x) variable the returns from period t and as your dependent (y) variable the returns from period t+1. If there is mean reversion you should see a negative resulting coefficient, although you should also check the fit of your result as well -- in keeping with my comments above, I would expect the fit to be poor, even if your coefficient is negative.

Hope this helps!

| improve this answer | |
  • $\begingroup$ You're absolutely right and I'm aware of the issue. Indeed it's a common problem when it comes to forecasting. I was just trying to see if there's mean-reverting pattern in the fund return time series. Thanks! $\endgroup$ – T-T Nov 20 '17 at 16:05

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.