2
$\begingroup$

I have a set of annualized returns over 4 time periods: 10yr, 5yr, 3yr and 1yr.

Is there a way to weight each return to have a "more representative" return?

$\endgroup$
4
  • $\begingroup$ I know this question sounds odd, but I overheard that there is a typical approach to dealing with different time frames in returns. Maybe a exponential-based set of weights? $\endgroup$
    – Victor
    Jun 20, 2014 at 16:44
  • $\begingroup$ What's the goal of your analysis? $\endgroup$
    – John
    Jun 20, 2014 at 18:23
  • $\begingroup$ Just presentation, I won't use it for any calculation $\endgroup$
    – Victor
    Jun 20, 2014 at 19:19
  • $\begingroup$ Ignoring any GIPS-related issues, I think the weighted returns are too complicated for most presentations. That depends on your audience, of course, but people understand a table showing a few different periods of returns so I wouldn't try to fight that too much. If all else fails, do what your boss says. $\endgroup$
    – John
    Jun 21, 2014 at 2:17

1 Answer 1

1
$\begingroup$

You can weight the returns and use them in calculations as shown below.

From this site:-

http://disc.sci.gsfc.nasa.gov/giovanni/additional/users-manual/G3_operation_time_series_stats.shtml

The weighted mean is

enter image description here

and the weighted standard deviation is

enter image description here

So, making up some annualised returns for time spans, 1 yr, 3 yrs, 5 yrs & 10 yrs:

r = {0.01, 0.02, 0.03, 0.04}

and some weights based on decaying relevance with, say τ = 2

a = {e^(-1/τ), e^(-3/τ), e^(-5/τ), e^(-10/τ)}

and fixing their total to be 1

w = a/Σa

{0.660361, 0.242933, 0.0893701, 0.00733595}

the weighted mean return is 0.0144 and weighted s.d. is 0.00972

Note, by using weights that sum to 1 the formulae are simplified, Σw and (Σw)^2 = 1.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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