1
$\begingroup$

Regular returns (log-differenced prices) have statistical distributions that are bell-shaped and unimodal (one mode/peak) despite being non-normal and fat-tailed.

Cumulative returns, on the other hand, computed from regular returns as $[\prod (1+r)] -1$, are bi-modal (with two modes/peaks). Is there a reason for cumulative return distributions having this shape?

And in spite of cumulative returns being non-stationary unlike regular returns, are they used in any well-known financial models at all?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

depending on the time period and asset considered, it comes from different market regimes.

e.g. if you consider a long enough period, you will clearly have distinct bull and a bear markets with crisis and changes in investors behaviors, with very different returns, there is an interesting article there.

all in all if you consider just 1 regime, would expect the returns to be monomodal.

Improve this answer
$\endgroup$
3
  • $\begingroup$ if this is the reason, then wouldn't more than two market regimes be possible (tri-, quad- modal cumulative return distributions)? if I currently only see two modes, would elongating the historical dataset another two decades surely cause more modes to appear? $\endgroup$
    – develarist
    Commented Jul 13, 2020 at 2:45
  • $\begingroup$ yes the longer the time period the more modes, in the article linked, the time series spans over a hundred years $\endgroup$
    – John
    Commented Jul 13, 2020 at 3:11
  • $\begingroup$ in spite of cumulative returns being non-stationary unlike regular returns, are they used in any well-known financial models at all? If cumulative returns appear in market regime models, is that all they're good for? $\endgroup$
    – develarist
    Commented Jul 13, 2020 at 3:12

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.