I'm producing a macroeconomic forecast using the model solver in Eviews.

I'm forecasting several endogenous variables on a quarterly basis, the problem being that the time series are unevenly extended.

For instance, I have the actual value for the CPI up until 2012Q4, since the data is published more regularly than national accounts, whereas I only have the actual value for consumption until 2012Q3.

I've specified the solution sample as being 2012q4-2015q4 and the model solves fine. The problem is that instead of using the most recent actual value for 2012q4, time series such as the CPI use that period's forecasted value. That means the model is not using the most recent data for some of the variables, causing my forecast to be suboptimal.

I have chosen the preferred solution starting values as "Actuals" instead of "Previous period solution" within the solver options. I've also tried the command "model.exclude(actexist=t) y1" which caused my model not to solve due to missing values.

What am I doing wrong here?

## Actuals instead of solved values

**Moderators:** EViews Gareth, EViews Moderator

### Re: Actuals instead of solved values

Hello from Paris,

The Exclude command does it. However there is a big problem: the command actually drops the equation from the model (for the associated periods).

If the equation is estimated, OK : the behavior is replaced by an exogenous decision. The model has one less variable and one less equation.

But if it is an identity, it no longer holds true. For instance if you exclude GDP in :

GDP + Imports = Final Demand + Exports

supply has no chance to be equal to demand.

Of course if you exclude imports which is estimated, there is no problem. It is just as if the country could decide on its import level, instead of letting the economy decide.

To summarize :

You can replace the agents in their decisions.

But you cannot replace an element enforcing a logical constraint.

There are two possibilities :

Drop one estimated variable and reformulate the model

For instance :

Q + M = FD + X

M=f(--)

becomes :

M = FD - Q + X

Q exogenous.

But this is rarely feasible.

You can also use my algorithm (see "switching endogenous and exogenous variables") which is simple of use and works most of the time.

Hope this helps,

Jean Louis B.

The Exclude command does it. However there is a big problem: the command actually drops the equation from the model (for the associated periods).

If the equation is estimated, OK : the behavior is replaced by an exogenous decision. The model has one less variable and one less equation.

But if it is an identity, it no longer holds true. For instance if you exclude GDP in :

GDP + Imports = Final Demand + Exports

supply has no chance to be equal to demand.

Of course if you exclude imports which is estimated, there is no problem. It is just as if the country could decide on its import level, instead of letting the economy decide.

To summarize :

You can replace the agents in their decisions.

But you cannot replace an element enforcing a logical constraint.

There are two possibilities :

Drop one estimated variable and reformulate the model

For instance :

Q + M = FD + X

M=f(--)

becomes :

M = FD - Q + X

Q exogenous.

But this is rarely feasible.

You can also use my algorithm (see "switching endogenous and exogenous variables") which is simple of use and works most of the time.

Hope this helps,

Jean Louis B.

### Who is online

Users browsing this forum: No registered users and 1 guest