EViews.com

Dear Eviews staff,

I wonder why it's so weird reflection of lags in ARDL cointegration specification?
Look at the screenshot. I declare variables without lags in equation specification. Then I estimate cointegration specification and find out that CE equation contains WAGESA variable without lag (however, I would expect to have a lag), and other variables as expected.
Then, the variables at cointegration coefficients table display the first variable without lag, and other variables with a lag. In this table I would expect to have no lags.
I have many more equations where lags appear and disappear in both specifications in some coincident way.
Any suggestions what's going on?
I have Eviews 14, Feb 26 2025 build

Hi!

Thanks for writing in. Why do you expect variables to have a lag? What if a variable has a zero lag?

The behaviour you're observing is not a bug! It occurs when the variable selection process determines that one or more variables have zero lag in the ARDL representation. In your particular instance, log(wagesa) is a variable which was determined to have no (zero) lags. In fact, if you look at the default estimation output where we write something like "Selected model: ARDL(a,b,c,d) using AIC ..." I am confident you'll see that b=0.

Furthermore, why do you expect the cointegrating coefficients table to not show lags? The theoretical representation of the cointegrating relationship is precisely one where we take a look at the relationship between the dependent variable and the regressor variables in their first lag (of course, excluding those which are 0 lag in the default ARDL form, as is your log(wagesa).

I suggest you take a look at our ARDL blog series to learn more how we derive these representations. They might help clarify some of the issues you're bringing up.

https://blog.eviews.com/2017/04/autoreg ... -ardl.html
https://blog.eviews.com/2017/05/autoreg ... rdl_8.html
https://blog.eviews.com/2017/05/autoreg ... -ardl.html

Kind regards!