panel data and rolling regression

[jonasgidlof]

Dear all Stata users,

I am doing a rolling regression to obtain beta values for a panel of firms. I use a 60-months window for each beta estimated. I have previously taken an econometric course, where I learned about FD, FE and RE estimation. In my rolling code, however, I do not use xtreg, but regular reg (and I have tsset my data, not xtset). Is this wrong? Can I not worry about RD, RE and RE because the rolling regression estimates the beta for one firm at the time, thus treating the data as timeseries for each firm?

I would appreciate any help regarding this.

Jonas

[Victoria Rogers]

I believe that you can just use the regular reg, at least when you would want to save the intercept of each regression instead of the beta

[jonasgidlof]

Hi Victoria,

Thank you very much for your quick reply. I want to save the betas, but I have the dependent variable as excess return. Like this:
[gen excessReturn= trt1m-RF][/reg excessReturn MktRF, r ]

Jonas.

[jonasgidlof]

Im so sorry, the rolling-part did not show in my last post. My full code is:
qui rolling _b _se if excessReturn!=., window(60) saving(betas, replace): reg excessReturn MarketPremium, r

[Victoria Rogers]

You're welcome.

Please do not depend solely on my advice because I'm also rather inexperienced with Stata.
What does the last r stand for in your code? robust?

I ran a similar code once (CAPM, 3 factor etc.) but I didn't use ,r in my rolling code.

By the way, you should double check before you post a topic/thread because your first post contains many mistakes. I also recommend you to click on the A above the place where you enter your message and then select the code you want to post, then click on # and the code becomes a lot more readible for others.

So, click on the A (next to the smiley), then select your code and press on #

qui rolling _b _se if excessReturn!=., window(60) saving(betas, replace): reg excessReturn MarketPremium, r