Extracting CAPM Alphas / Betas

[Anela Kien]

Hello Statalist Users,

I would please need an advice on the execution of a CAPM Regression from someone who is familiar with this model.

I know, not all of you are familiar with the CAPM in Finance, but essentially we explain stock returns with a market factor (and sometimes other factor retuns).

How do I code a CAPM Regression in Stata?
I have quarterly returns of stocks as well as the market return (mktrf), risk free return (rf).

I now want to extract the CAPM alphas (the constants) and Betas.

Do I need to use the Fama MacBeth Method for this or not (or when)? Do I use xtfmb or asreg or just the regress command?

Can I use my quarterly returns for this and my calculated quarterly market return as the independent variable?

Thanks a lot in advance!

Best regards,
Anela

[George Ford]

At the most basic level, you regress the % change in the stock price on the % change in the market index.

You can adjust for the risk free rate, dividends, etc.... but this has little effect on the betas.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3312894

Fama-French factors could have a more meaningful effect on the results.

Quarterly data is too aggregated (though you can get a result). Typically you see estimates of a 1,2, 3, or 5 year beta, and with 5 years you'd only have 20 observations, and only 4 observations for a 1-year beta.

You can download stock price data in Stata using getsymbols.

[George Ford]

asreg is a good option if you many stocks

[Anela Kien]

Thanks a lot! That helped me! I calculated the quarterly returns using the Total Return Index, so the dividends are included.
Could you maybe further explain if I should use the Fama MacBeth Method or when to use it?
As a results I would like to get quarterly alphas (or at least yearly) for every stock in my sample as I would like to use them afterwards in order to calculate other things on this quarterly basis.
My database has a time span of about 20 years and I have about 15k stocks in my sample.

Best,
Anela

[George Ford]

As best I can tell, estimating Betas is more art than science. Most practitioners don't use FFM, but academics often do, and it gives them stuff to argue about and publish papers.

[George Ford]

And, in many cases, the constant is zero in the FFM. If you need alpha, then keep that in mind.