New Features in EViews 12 : Models
  
Models
 
Solve Control for Target
Endogenous Variable Specification
EViews 12 offers powerful new features for working with models. New methods allow you to solve your control to match an endogenous target, or to change the specification of the endogenous variable in an equation.
Solve Control for Target
Normally, when solving a model, we start with a set of known values for our exogenous variables, then solve for the unknown values of the endogenous variables of the model. If we would like an endogenous variable in our model to follow a particular path, we can solve the model repeatedly for different values of the exogenous variables, changing the values until the path we want for the endogenous variable is produced. For example, in a macroeconomic model, we may be interested in examining what value of the personal tax rate would be needed in each period to produce a balanced budget over the forecast horizon.
The problem with carrying out this procedure by hand is that the interactions between variables in the model make it difficult to guess the correct values for the exogenous variables. It will often require many attempts to find the values that solve the model to give the desired results. EViews can address this challenge in two ways. The first is to perform a numerical optimization of the exogenous variable values, automating the search for exogenous values that lead to the desired endogenous values. The second approach re-specifies the model in such a way that we can directly solve for the exogenous values.
After creating series holding the trajectory values for our endogenous target variables, we can conduct a search for the precipitating values of our exogenous control variables via the Solve Controls for Targets… item under the model’s Proc menu.
The required parameters are a list of control variables (in the upper text area) and a list of target variables and trajectory series (in the lower text area). These lists may include group objects. The sample over which new control values will be calculated defaults to the current workfile sample. The solved control values will be stored in the original control series by default. The Store solution values… checkbox lets you indicate that the results should be stored in an alternative set of series with the adjoining suffix. This suffix defaults to the active scenario alias suffix, which allows the results to be easily applied to the model via a scenario override. As a convenience, if the trajectory series names are just the target variable names with a suffix applied, you may use the Omit trajectory series… checkbox and adjoining field to shorten the required parameters.
After clicking the OK button, EViews will determine an appropriate lag for each control variable, conduct the search over control values, and then display a summary of the results in the open model window. The image below shows an example results summary for a single control named DGF.
This summary is a transient rather than a persistent model view and will be lost on the next view change if not frozen.
There are three notable limitations of this numeric approach.
The model may not contain leads (positive lags).
There must be an equal number of control and target variables. The overdetermined case with more target variables than control variables is not supported currently.
The procedure may take some time to complete. Depending on the dependencies among the variables and the control lags selected, the optimization may require multiple bidirectional passes over the sample observations, causing the solution time to be far greater than a standard model solve.
It is also possible for the search to fail, with EViews unable to discovery a set of satisfactory control variable values. You can always alter the control series values (the search’s starting values) and try again.
Endogenous Variable Specification
The alternative to a numerical search for exogenous values is to respecify the model such that the exogenous control variables are made endogenous and the endogenous target series are made exogenous. The values of the control series can then be computed using the standard model solve procedure (with the trajectory series overriding the target series). You can automatically rewrite a model’s equations to select a new set of endogenous variables via the Flip Endog/Exog Variables… item under the model’s Proc menu.
The required parameter is a list of endogenous and exogenous variables to flip (in the upper text area). There must be an equal number of endogenous and exogenous variables (but not add factors) that may be specified in any order. The list may include group objects.
EViews will automatically determine a minimal set of equations to rewrite symbolically. The number of equations rewritten will be at least the number of specified exogenous variables, but could be much greater. If a rewritten equation was included in the model via a linked object, e.g., a VAR, system, or another model, then the link to that object will be broken.
If there are equations that you wish to exclude from being rewritten, you may list the endogenous variable for these “locked” equations in the lower text area. Additionally, you may lock all equations that have associated add factors or innovation variances with the appropriate check box. Naturally, endogenous variables being flipped should not be locked.
Once you click the OK button, EViews will determine which equations need to be rewritten and then modify the model object accordingly. A summary of the changes made to the model is displayed in the model window.
This approach has several limitations to be aware of:
Equations containing ARMA specifications will not be altered by this procedure and are treated as locked.
Currently, all rewritten equations lose their add factors. This deficiency will hopefully be addressed in a future update, but for the moment, locking all equations with add factors is the only way to prevent such information loss.
Rewriting equations is a process of algebraic manipulation, since we are essentially solving equations for a new set of left-hand side variables. Consequently, we are limited by the set of invertible mathematical operations.
It is possible that EViews will be unable to rewrite the model’s equations to produce the desired set of endogenous variables. If you have explicitly locked any equations, you may retry with fewer locked equations, hopefully providing EViews with enough options to be successful.
See Model::control and Model::fliptype for command documentation.