Im currently trying to solve the following regression problem. Daniel Moskowitz Momentum Crashes Since my results for the first column are correct, im sure im on the right way but: For the second column i tried the realized variance of the 126 days before the start of the month (sum of the squared log returns) and the "normal variance" of the preceding 126 days. While my beta coefficient gets nearly the same magnitude as wanted, my intercept is always around 0.24.

Knowing that:

How does a regression input have to change, to affect the intercept ? Scaling it obviously only changes the beta coefficient but not the intercept. So does my independent variable have to fluctuate more ?

I know its a strange question, but since i'm working on this for days and not getting the right values, i am quite desperate. So i am trying to find my mistake by isolating the possibilities.

these are my starting points for the two variance "possibilities"

StdDeviation<-foreach(i=1:22666, .combine = cbind) %do%{




###########Realized Variance
Realized.Variance<-rollapply((log(Fama.French.daily$Mkt+1)^2) ,126,sum,by=1)

datasets are available on Kent Daniels Website for the Paper Momentum Crashes and on Kenneth Frenchs website

But as i said, since i know its hard to replicate the specific problem, i'm mainly interested in the general question above.


1 Answer 1


You can think of the intercept as a baseline. By definition, it's what the outcome is when the predictors are all zero. If it's a small number, it means the baseline for your outcome, when your variables are zero, is low.

  • $\begingroup$ Thanks for your reply, so its more like the independent variable that's not right in my regression. $\endgroup$
    – KDMS
    Jun 29, 2019 at 22:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.