Vector Errror Correction Model

[brennan6738]

Does anyone know of a resource that gives the modeler an idea of the structure of a certain financial time series before it is modeled? For example, if I am trying to model an interest rate I would think that it should be modeled with a trend, but no seasonality. Or would I want to model an interest rate with an intercept? Or does it sometimes depend on the rate? Rates have a zero lower bound and typically only rise to 10% (or maybe 15% or more for risky assets). Anyone have any ideas?

[hum2004]

hi,
one way to get an idea how to model the VECM are the time series you plan to include. if none of them apears to have a drift term you may try to include the constant in the ce only. if the series appear to be a random walk with drift, you may either have the constant in the var or in the ce and the var. Finally, after estimating the VECM you can test for significance of the constant. A good read is Walter Enders, Applied Econometric Time series, 2nd Ed, see page 350. The same applies to seasonal dummies. If you include them, you can test for their significance. In my work i found that interest rates do have a drift. A couple of Journal Papers i read do aggree.

[tcfoon]

Dear friend,

An alternative approach that you could consider is the Pantula principle and/or the Modified Pantula Principle. The statistical rule will give some insight of which model is the best..option.. This method play the role because sometime not all series under investigation follow the same pattern.

Hope the above helped you...