The RATS Software Forum

In my current research I am analysis money demand (in)stability in the US for various monetary aggregates.
I always obtain a - at least from my point of view - very strange result, which is robust across a large number of different specifications.

No matter which monetary aggregate (M1, M2, M2M, MZM), scale variable (GDP, GNP, PDI) or interest rates are included in the model (and no matter what trend-assumptions and dummies I include), I always find inflation to enter an otherwise classic money demand function with a positive sign.

Mr - Yr + 10*(shortrate-ownrate) - 10*Inflation = I(0)

According to model diagnostics and formal stability tests the model seems well specified. The only thing which concerns me are some signs of I(2)-ness in the data (considerable differences in uncorrected and Bartlett-corrected trace statistics).

Does anybody have an idea, what kind of misspecification could cause my strange result or how my result could be explained?

Many thanks in advance!

Poppelwops

I assume you're estimating this with CATS. Are you, in fact, getting the result that you have one and only one cointegrating vector and that the coefficient on pi has a wrong sign which is significant? Since what you're doing has been done before, how do your results compare with the published ones if you restrict yourself to the same period? (The data won't quite match, but the results should be similar).