Page 1 of 1

Terminal conditions for future values

Posted: Mon Feb 20, 2012 4:42 pm
by ashtongate
Hi all,

Heavy eviews user, but new to models. Running a small macro model with some future values. I am using the terminal conditions option, but am a little confused. Where can I specify the terminal conditions? If I cannot, what is eviews assuming? (e.g., I would like all of the endogenous variables to equal zero at some future date).

Another option is to specify a smooth transition from last observable "actual" to an assumed steady state at the end of the forecast period. Am I committing an egregious error doing this? Is the baseline solution then really the "model forecast"?

Thanks!

Ashtongate

Re: Terminal conditions for future values

Posted: Tue Feb 21, 2012 2:47 pm
by EViews Chris
You can set the terminal values in a couple of different ways. To see the choices, go to the 'Solver' page of the dialog and look for the list box 'Terminal conditions'.

By default, the terminal conditions involve taking the values in the endogenous variables at the end of the sample as fixed.

There are a number of other options you can set which tie down the values at the end of the sample in some other way.

Re: Terminal conditions for future values

Posted: Tue Feb 21, 2012 3:21 pm
by ashtongate
Thanks Chris. I see those general options for the method on the solver tab, but I do not see where I can fix the specific terminal values (or the growth rates). Also, is there a way to code it up in a program? I know you can set the general option in solveopt(t=??), but again, can you fix the specific conditions? I suppose one approach is to use "actuals" as starting values.

Re: Terminal conditions for future values

Posted: Tue Feb 21, 2012 4:31 pm
by EViews Chris
As I said before, using the default settings, the terminal values are the values of the endogenous variables at the end of the sample.

So let's say you are solving the model:

y = .3*y(-1) + .2*y(+1) + x

where the workfile runs from 1950 to 2030 and the model solve sample is from 1980Q1 to 2010Q4.

The solution for y is 'tied down' by the value of y in the last period before the solve sample (1979Q4) and the first value *after* the solve sample (2011Q1) (as well as the values within the solve sample of the exogenous variable x). EViews will be unable to solve the model unless a value for y in 2011Q1 has been specified by the user.

With the other methods (constant levels, constant difference, constant growth rate) solutions for y in the final periods are calculated usisng the chosen rule. In this case, values for the endogenous variables after the end of the solution sample do not need to be specified by the user.

Note that you can't specify what you want the level/difference/growth rate to be in this case - it will be determined as part of the model solution.

In most cases, what you'd really like to see is that what is happening at the end of the forecast sample doesn't really make much difference to the values you are seeing at the beginning of your forecast. If it does, then you probably need to lengthen the solution period until the terminal conditions no longer make any signficiant difference to values within the range of dates you are actually interested in forecasting.

Chris

Re: Terminal conditions for future values

Posted: Thu Feb 23, 2012 8:04 am
by ashtongate
Got it, thanks for you help.