Dear community,
What is the code to perform a Fama-MacBeth regression in Stata? I understand how this works theoretically, but I do not understand how this is implemented in Stata. My variables are the 5 factors of the Fama French 5 factor model and 25 portfolios double sorted on size and book-to-market value of equity.
Additionally I have another question as well. That is, in order to test the Fama French 5 factor model, you just regress the factors on one of the portfolios right? In other words, is the correct code to test the 5 factor model:
- tsset date (in order to declare dataset to be time-series data with date as the time variable)
- reg me1bm1 markt smb hml rmw cma (where me1bm1 is the portfolio with lowest marketcap and lowest B/M and the other 5 variables are the 5 factors).
When I use this code I get very strange results, namely that almost all intercepts are significant (which is in contradiction with the Fama French papers). Hence, I am wondering whether there is something wrong with this code. I hope you all could help me with these 2 questions!
Yours truly,
Niek
Why do you expect the F-F factors to still be good predictors many years after F-F ran their regressions? Would you expect that to work if you chose to study the situation before and after the French Revolution?. Isn't it possible those factors today are not the same as F-F found in their much earlier paper? Best, D. Booth
Dear Mr. Booth,
Thank you for your answer. The reason I am researching this is for my master thesis. The topic of this thesis is whether or not the FF 5 factor model holds internationally. Fama and French did research something similar in 2014 and their five factor model is from 2015. In other words, not so many years ago they obtained totally different results than me. Additionally, there are several other papers which do something similar like me for certain regions of the world and their results are in line with Fama French. Since my strange results lead me to think that I did something wrong. Therefore, I am wondering how the correct code would look like if you want to run the FF regressions and also the code for the Fama MacBeth regressions. If one would be so kind to give an example how the could could look like I could compare it with my own code and maybe identify some mistakes.
I would be interested in your results. In accounting they would call the FF result the FFmodel and use those variables forever. I can't see how that could always be right. Best, David Booth
- Badges
- Science topic
- Similar topics
- Papers