Defining a stressed Period

[elhazzag]

Hi,
This is a question regarding how to define a stressed period.
Let's say you have data from year 2000 until now (of daily equity prices).
And you want to find a 2-year stressed period window within that data that you have.
How would you proceed? What if you are interested only in the negative moves (stocks decreasing)
(compute a moving standard deviation of the daily log returns of each equity price for 2 years for example, then averaging it and taking the worst).
Any idea or papers on it that might help would be welcome.

 

[Alexandre Chretien]

Look at VIX

 

[ZHOU QIAO]

As Alexandre mentioned, VIX is very useful. Practitioners also used to look at 200 day moving average as a naive regime filter.

Building on these explanatory variables, we can train Clustering models, Gaussian Mixture Models or more advanced Regime Switching Models. There are extensive literature on these topics though most of them are not practical. However, I've tried and found them useful and seen others applied similar techniques meaningfully.