Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Fama MacBeth Regression

From   "Braunfels, Philipp (Stud. SBE / Alumni)" <>
To   "" <>
Subject   st: Fama MacBeth Regression
Date   Thu, 19 Jul 2012 18:55:47 +0200

Dear statlisters,
I want to run a fama macbeth regression and have a question regarding the different options and the two step procedure. 
Upfront, I have an unbalanced panel of 1500 companies over 10 years of monthly data (=100 months). Is it important that the panel is balanced or unbalances? I looked up that the command -xtfmb- performs a fama macbeth regression. However I could not find any information about the different options: "lag(#)", "verbose", and "level" (the help file did not make it clear to me).

So I want to run a regression of three independent variables on excess return: -xtfmb return x1 x2 x3, "option" - 
available options "lag(#)", "verbose", "level(#)"

Question 1) If I run the above model with the "verbose" option, does this mean that step 1 (individual time series regression) and 2 (cross section regression for each month) of the Fama MacBeth procedure are automatically carried out? When I run the model I only get the Cross-section output (step 2) for my 100 months of data. Furthermore, my x1 variable (market return) is zero for all months. Might this be due to some specification error? If I check for the values of my market return with -sum- I get:

Variable |       Obs        Mean    Std. Dev.       Min        Max
     market|    201189    .0028133    .0487276  -.2114486   .1068373

Question 2) is it correct, that if I use the "lag(#)" option e.g. "lag(4)" that I get newey-west adjusted standard errors? How many lags should one use? Furthermore, my -xtfmb. lag(4)- output always omitts my x1 variable (market return). why is that?

Question 3) is it important to define  "level(#)" or is the 95% confidence interval assumed here (which would be suitable for my purposes)?

I am very thankful for any advise since I am very new to the topic of Fama MacBeth regressions.


*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index