Analytical Solutions

[LuisReyes]

Hi again,

I am working with a research team interested in Stock-Flow Consistent (SFC) modeling, which consists of simulation exercises based on systems of equations which respect an accounting framework. Most of the members in the team make their models using EViews, but not everybody knows how to program.

Some doubts have come up concerning the capabilities of EViews concerning analytical problem solving and how this could be overcome. Going through the Command and Programming Reference, I did not come across any analytical solving command (perhaps I just didn’t look well enough). Could you please let me know if this is the case?

In any case, people in the team have different objectives and different ways to tackle problems (some use other software), but my feeling is that, in order to make models more coherent to any national case, we must estimate parameters, then use these in a system of equations we build based on economic theory and see what it gives. Up to now (and to my knowledge), SFC models have not used estimated parameters.

My goal is to build a model which estimates parameters (via cointegrated VAR or even SVAR models), insert them automatically in a SFC model, solves and shows different scenarios given predetermined shocks. Could you give me some advice as to what we can do using EViews? I saw a paper where they propose a calibrating method, but this implies a priori solving by hand, running a first model to calibrate, then used calibrated parameters in the model… Is there a command (or way) that keeps us from solving manually?

Thanks in advance.

[trubador]

I am fairly sure that EViews' "Model" object can handle all the stuff you have outlined about your study and you may not even need to use programming features. Moreover, Jean-Louis Brillet's excellent book Structural Econometric Modeling can guide you through both the modeling and EViews implementation.

[LuisReyes]

Thanks a lot for your feedback and for the link.

The link is useful in that it answers my question. Page 45 of Brillet's book reads: "EViews does not solve analytically any equation for the variable". Below, page 117 it reads: "coefficient expressions cannot be obtained analytically".

The command "model" that you mention solves numerically. I already know how to do that, but once you set starting values "far from a solution" the model stops (neither by Gauss-Seidel, Newton nor Broyden).

This is an important problem in larger-than-usual scale models...

Any feedback is still welcome (specially from developers).

[jlbrillet]

Hello from Paris,

I am the author of the "excellent" book.

If you give me more specifics I can give you some general advice.

Best regards,

Jean Louis B.

[LuisReyes]

Hello (also from Paris),

I just realized your message was here. Thanks a lot for this reply. I was wondering if there would be a way to get analytical (instead of numercial) solutions using EViews. Not that this is vital, I already finished the model and this step was not necessary. However, it might be useful to get to know the theoretical steady-state values of a system of equations (as in, for instance Matlab or Mathematica) so that we try, for instance, to calibrate a model using an 'adjusting' variable.