No announcement yet.
  • Filter
  • Time
  • Show
Clear All
new posts

  • Carhart 4-Factor Model Regression

    Hi all,

    Hope you're all fine.

    I'm currently trying to regress monthly excess returns on Carhart's four factors. I have x monthly excess returns of i firms over several years and am regressing these on Carhart's four factors RM, SMB, HML and MOM. The regression results in all factors being highly statistically significant and the F-statistic is very high as well. However, the r-squared is very low which seems odd. the model should fit well. Does anyone have an idea on what I missed?

    Many thanks in advance for your help, much appreciated!


    regress ExcReturn MktRF SMB HML Momentum
    Click image for larger version

Name:	Screen Shot 2018-07-11 at 5.49.25 PM.png
Views:	1
Size:	52.6 KB
ID:	1452900

  • #2
    Dear all,

    I understand that regress might be the wrong type of regression. I converted the data into panel data using xtset and regressed again using xtreg. R-squared is now around 15% but still, all variables are highly significant. I still believe that I have to approach this differently. Do I have to aggregate first on monthly level or even on firm level?

    Many thanks for your help.



    • #3
      You'll improve your chances of a useful answer by remembering that we are not necessarily from your area and explain everything.

      When you say you have a low r-square, that really depends on the literature and how predictable something is. As I remember, the r-squares in CAPM model estimations are often quite low at particularly in shorter observation periods (like daily or monthly).

      I thought you said you were using monthly data so I don't understand aggregating to the month level. Your questions are really more appropriate for folks in finance than a general audience like Statalist - these kind of questions are often answered by disciplinary norms along with the substance of the domain.