Calculate bipower variation of daily stock returns

[Michael Craig]

Hey there,

I have calculated daily excess stock returns over 16 years and I want to calculate the realized bipolar variation of these excess returns like in the paper of Chung&Seo.

The formula is found on page 8 of the paper (equal to the paper from 2017 in the International Review of Economics and Finance but working paper is freely available):

𝜎(𝑡𝑖)^2 ≡ 1/ /(21−2) ∑𝑗=3 |𝑟𝑡 ||𝑟𝑡−1|.

I am not sure how to do that. Probably with the rolling command or so?

Do you have an idea on the code I can use (the rt and rt-1 are simply the daily excess returns in my Stata database).

Regards,
Michael

[Michael Craig]

I may need to provide more information in order to get any help on this issue I guess?

[David Benson]

Michael,

I'm not sure what the bipolar variation is, but you will probably use Stata's built-in rolling command, or something like asreg or asrol (both from SSC) written by Attaullah Shah

You might also take a look at this post here

It would also be helpful if you posted a small sample of your data using Stata's dataex command.

[Michael Craig]

Hi David, sorry for answering so late. I needed to check some issues with the requirements but I applied the asrol command using

bysort ID_S: asrol excessreturn, window (Date 21) stat (sd) gen(std_bipower)

which I guess calculates a daily rolling 21day window standard deviation per stock across the dates (not sure about the correct formula I posted on January 14th). I am not sure if it is working. I have 1.600 stocks and daily data for about 15 years and it is now 2 hours that Stata is calculating on my MacBook from 2013.

I am not sure if something is wrong or if it is normal that it takes so long. I am using Stata SE 15.1.