model objects, policy simulations and forecasting: a step ... · out of sample forecasting eviews...
TRANSCRIPT
Model objects, Policy simulations and Forecasting in
E-views: a step by step approach
By
Tinashe Bvirindi
Layout
• Model object creation
• Solving a model (in sample)
• Forecasting out of sample
• Conducting policy simulations and ploting response functions
Modelling in E-views
•A model consists of a set of equations that jointly describe the relationship between a set of variables.
•The equations can be:• Simple Identities,•Results of single equations, or•Results of multiple equation estimators
Modelling in E-views
• The equations are combined in a single object to derivedeterministic or stochastic joint forecasts or simulationsfor all the variables in the model.• Deterministic setting: model inputs are fixed at known values
and a single path is calculated for the output variable• Stochastic setting: uncertainty is incorporated into the model
by adding a random element to the coefficients, equationresiduals or exogenous variables
•Models allow us to conduct policy simulations
Modelling in Eviews
• In Eviews for the model to have a unique solution, there should typically be as many equations as there are endogenous variables
• Each equatin in the model must have a unique endogenous variable assigned to it.
• Any variable that is not assigned as an endogenous variable is considered exogenous to the model.
Creating a model object
To create a model object, click on Object, in the main window and select New Object
Building a model
Now on the type of object, select Model and click OK
Give the model a name of your choice
Model object
Model object
• Equations in Eviews can either be inline or linked • Inline - the equation is specified as text within the model
• linked – the equation brings its specification into the model from an external eviews object e.g. a single equation object
• The advantage of linking is that it allows coupling of the model with the estimation procedure underlying the equations.
• Equations can either be stochastic equations or identities
Creating a linked equation
To create a linked equation, right click on the equation of choice then copy it
Creating a linked equation
Now take the copied equation and then paste it in this window
Creating a linked equation
Select the yes to all option
Linked equations
In this equation window all the pasted equations will appear, with a list of their explanatory variables
The scenario in question appears here
List of variables
Linked equations
To views the variable dependencies and their classifications click on the Variables button
Exogenous variables are labelled and have an X
Endogenous variables are equation variables and have and En
Adds stands for add factors
Advantage of linking the equations• Once we added our equations as linked equations we can go back and
re-estimate our equations and automatically update the model to the new estimates as follows:
Click on Proc, then select the Links button and click update all links and recompile
Adding identities
• In order to add identities, right click the mouse while in the VIEW equation window.
Right click anywhere in this window and click on insert
Adding identities
Once you select insert this dialogue box will appear and then you enter the identityinto the model source edit window and click OK
Creating inline equations• To create an inline equation first copy the equation representations
First open the equation of choice and click on view, then select representations
Creating inline equations
Once this output comes out copy the substituted coefficients
Creating an inline equation
Click on the text toggle/ button and paste the copied equation into this window
Inline equations
In order to view the dependence structure of the variables, click on variables
Then click Yes to save modifications and compile
Inline equations
Click on view to see the equations or the block structure of the model
Block structure
Inline text equation
Solving equations
• Once you have inputed all the equations into the model the next stepis to solve the model
• There are many options available for solving the model in Eviews• For now we concentrate on the basic techniques
• To solve the model simpy click on solve
• However, before we solve the data we may want to input theexogenous variables we wish to use in policy simulations as in linetext
Solving the model
Click on text and type the variable you wish to employ in policy simulationsNB: this technique
is a shortcut and is sometimes not advisable
Solving the model
Click on solve and then click on YesTo save and compile modifications
Solving the model Once you click save this dialogue box will appear
Choose baseline scenarioClick on deterministic
Select static solution for model evaluation
Adjust the sample size over which to solve the model to avoid initialising the model on missing values
Solving the model
• Once you have set all the conditions click OK to solve the model
• You will then receive the following solution message
Workfile appearance after initial solve
The solution of the baseline scenario is saved with and underscore of zero in the workfile
Forecasting- static solution
Plot the baseline lm3 from the static solution against the actual lm3
Forecast- dynamic solution (recursive)
Re-solve the model but this time selecting the dynamic solution and plot the baseline from the dynamic solution against the actual
This result shows how the model Would have performed if we had used it back in 2000
If satisfied with performance of the model against historical date we can use the model to forecast future values of endogenous variables
Forecasting out of sample• First step is to decide on the value of exogenous variables.
• If they are not available the re is need to provide these.
•
Out of sample forecasting Eviews uses a Monte Carlo simulation technique to generate the uncertainty surrounding our forecasts
To generate this graph, click the solve button and in the model solution dialogue select stochastic, and tick the standard deviation box in Active and click OK
Then go to Proc, Make graph, and in the solution series box select mean+2s.d and reset sample period to 2003Q1 to 2011Q4, and click OK
Policy simulations • Having satisfied ourselves of our model’s capabilities in and outside
the sample we may conduct policy simulations
• In the policy simulations, we are mainly interested in the impact that the exogenous variable will have on the endogenous variables and ultimately on our model• Economists- interest is to assess the effect of policy variable on
macroeconomic aggregates
• Risk manager/ supervisor- interest is to determine the level of stress that an external event may induce on the endogenous variables
Policy simulations
• Step 1: create a scenario
In the model object window click on View, then select Scenarios
Policy simulations
Click on create new and then OK
Policy simulations
Click on the Variables button
Policy simulations
Right click on the exogenous variable of interest and the select properties on the drop down menu
Policy simulations
Tick in the use override series in scenario and click OK
Policy simulation
Once the variable is set the text changes to red
Policy simulation- temporary shock
• Lets assume a temporary shock of a sudden increase in nominal income of 10% per quarter for the period 2001Q1 to 2003Q3
• To do policy simulation we make use of a very simple command
Set the period over which the shock takes placeAnd calibrate the shock
Policy simulation
• Once the sample size is set and the shock is specified then we proceed to solve the model
Click on solve in the model object window
Policy simulation
Reset simulation type to deterministic
Set the interval over which you want to estimate
Make sure the set scenario is active and click ok
Policy simulations
Solution message for the scenario
Plotting the results
Whilst still in the model object click on the Proc Button and select the make graph option
Policy simulation: shock vs baseline
Make sure that the actuals and scenario are selected
Reset horizon and click OK
Select list of variables and list endogenous variables
Policy shock simulation
IMPULSE RESPONSE OF LM3 TO LNGDP
Money supply has a lagged response to income, relationship is error correcting, shocks die down after about 5 years
Class exercise1. Estimate the following error correction models:
• D(LSTCKPR) LSTCKPR(-1) LTDC(-1) LCPI(-1) LP_R(-1) C D(LTDC(-1)) D(LCPI(-4)) D(LP_R(-1)) D(LSTCKPR(-1))
• D(LM3) LM3(-1) LNGDP(-1) LCPI(-1) LP_R(-1) C D(LM3(-1)) D(LCPI(-3)) D(LNGDP(-4)) D(LP_R(-4))
• D(LTDC) LTDC(-1) LR(-1) LM3(-1) LNGDP(-1) LNEER(-1) C D(LR(-1)) D(LTDC(-1)) D(LM3(-1)) D(LNGDP(-3)) D(LNEER(-2))
2. Create a model object with the three equations and solve it.
3. Perform an in sample forecast and produce an out of sample forecast showing the level of uncertainty associated with the forecast.
4. Trace the impact of an increase in GDP on money demand, the stock price and domestic credit and comment
5. Plot the impulse response function
References