I am setting up a model to predict something using Error Correction Model (see pic. to get a general view about the model itself).

enter image description here

Some of my $X$ variables are the interest spread between the A and B govt bill (Sprd) (e.g. 90 day). At this point I have a practical question:

Regarding $\triangle Sprd$, what is the standard procedure to estimate it?:

  1. take the difference as a simple $Sprd_{t}-Sprd_{t-1}$
  2. or apply $Sprd_{t}/Sprd_{t-1}-1$

As far as I know, the invariant of this kind of variable are option 1 (Meucci´s book), but I am not really sure if this is valid is in this case.

I would appreciate any advice about this

  • 1
    $\begingroup$ Hi: you need to be clear and precise about what the response is and the variables that you think ( or want to test if ) the response is effected by. Then you can define the variables accordingly. saying interest rates and spread is not helpful because variables can't be stock variables or flow variables. Maybe you mean the return of the 10 year note minus the return of the 1 year ? or the daily closing log price of stock A minus the daily closing log price of stock B etc. $\endgroup$ – mlofton Mar 13 at 15:19
  • $\begingroup$ @mlofton edited to clarify X variable. Thx $\endgroup$ – Newbie Mar 13 at 15:46
  • $\begingroup$ If $\Delta X_t = Sprd_t/Sprd_{t-1} - 1$, then what is $X_t$? Presumably you want its lagged value to appear in the error-correction term. $\endgroup$ – The Laconic Mar 13 at 15:59

Your Answer

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

Browse other questions tagged or ask your own question.