Beta calculation

[RupaR]

How would you decide on the time frame to your Beta calculation?
Specifically, if investing around Daily and Hourly commodity prices, what do you consider before estimating the Beta against any Dow Jones Commodity Index as benchmark?
Thank you for your comments.

 

[T.H.]

may i ask why are you using daily and hourly prices ? using weekly prices is more common practice, and of the top of my head consider if there has been increased volatility in the past years. I am referring post 2008 which has been more volatile. If you were to use a 10 year time frame the level of volatility might not be represented accurately and downplayed. So considering volatility I would use data from the past 4-5 years if you want to get an estimate of beta for short term.

 

[RupaR]

Thank you for your comment. I am using daily and hourly prices since trading occurs in these markets. Hence returns are daily and hourly values in each market. My data is High-Frequency, past 4 years.
I agree traditionally 5 years of data is considered for calculating Beta. But this is my concern. Some analysts suggest that the "time frame of beta calculation must align with the investment time horizon as the end-users prefer to use". End-users of my analysis are Energy Day-Traders. They close positions within hours or days. How do I use a 5 year Beta? My analysis requires Alpha from energy commodities trading strategies across months to understand return patterns in summer and winter seasons. But Beta is not seasonal !! Or do I still go ahead with a 5 year Beta? Please share your views here. Thank you.

 

[T.H.]

Well this is one of those cases where you have to use your best judgement. What I would do for the seasonal beta issue is just assign a value of 0 to beta for the summer periods so you can get a result for how the securities behaved only in the winter and vice versa. Kinda of like the sortino ratio where it ignores positive volatility, you would do the same where you would ignore the other season time period in order to get a beta for the other. Also you should look at other trends that would affect the beta during the winter and summer seasons. Im pretty tired and can't really think of an example right now but I hope I managed to explain what im trying to say well. Are you just using the CAPM ? Have you run any regressions ?