Interindustry-Macro Modeling Software
Inforum models typically come in two forms: macroeconomic models and interindustry-macro models. Interindustry-macro models have been developed by Inforum and it partners for over forty years. As with macroeconomic models, these models can be assembled and estimated in G7, compiled and run from the G7 menus, and the results viewed in G7. Many times, these models are larger and more complex than macro models, and some hand-coding in C++ is necessary. The methods for constructing interindustry-macro models are described in Volume 3 of The Craft of Economic Modeling. While some of the software needed to construct these models are available for download, the full software package is available only to academic partners of Inforum; other requests for non-comercial use will be considered. In addition to G7, the IdBuild program, Fixer and MacFixer programs, and the InterDyme C++ library are needed; all except the InterDyme library are available on the download page. In addition, the Borland C++ 5.5 compiler is needed to compile models; it is available for free download.
The text below describes the InterDyme package. Please see the documentation page and other references noted here for more details. The PDG package on the Downloads page contains all of the freely-available software described here for building interindustry-macro models.
Interdyme is a package of programs for building Interindustry Dynamic Macroeconomic models. At this time, models for the U.S., Germany, France, Spain, Italy, Japan, Korea, China, Russia, Poland and Hungary have been constructed using Interdyme. We speak of these models as "macroeconomic" because they deal with all the concerns of macroeconomics, such as employment and unemployment, inflation, income, interest rates, output, investment, and productivity. But they are genuine interindustry models and build up most of the aggregates from industry-level variables.
Interdyme concentrates on the extensive data management aspects of such models and leaves the user in complete freedom for development of the behavioral equations which determine the character of the model. The development of Interdyme has been strongly influenced by our experience both with the G7/Build system for building aggregate models. The G7/Build system, in fact, is an integral part of Interdyme, and this explanation will assume that the reader is already acquainted with it through the book The Craft of Economic Modeling . Some general acquaintance with input-output concepts is also necessary. Interdyme is written in C++; a minimal acquaintance with this language will also be necessary for effective use of Interdyme.
- Interdyme is visual. Matrices and vectors of both input and output may be viewed on the screen with a scrolling window like a spreadsheet. Graphs of any element of a vector or matrix or macro variable easily can be viewed and printed.
- Interdyme is capable of both historical simulations or forecasts.
- "Macro" equations estimated in G can be introduced as easily as they are in Build. Indeed, Idbuild (Interdyme Build) has been written to work on the macro equations just like Build does.
- "Fixes" to both macro equations and vectors are easily specified in versatile formats.
- Vector and matrix algebra may be used in writing the code.
- The system produces code in C++, which is based around logical modeling objects, such as Time series, Vectors, Matrices, Equations and Databanks. Working with objects yields more concise and understandable code than working with low-level C or Fortran.
- A C++ compiler. We have used Borland 3.x, 4.x, 5.x, Borland C++ Builder versions 3 through 6, and the GNU GCC compiler for Linux, as well as the Delorie DJGPP and Cygwin versions of GCC. Presumably Microsoft, Symantec or Watcom will also work. A free version of the Borland C++ 5.5 compiler is available.
- A Pentium or higher, with at least 32 Mb of memory is recommended.
- The G7 regression program is required to manage time series and perform regression estimation.
- A basic working knowledge of C and C++ is required, but the level of programming expertise is similar to that required to build models in TSP or other regression packages. The most common errors are simple syntax errors, or confusion of data types, such as Matrices, Vectors and Tseries (time series).
- The standard version of Interdyme comes on two disks. One disk contains the programs necessary to run Interdyme, which includes G7, Idbuild , Fixer , Macfixer, and Compare . The other disk contains a sample model, as well as the skeleton source code from which all models are derived. An instruction manual provides examples and pointers for how to input data, develop equations, and build and test a model. Or, if you like,there is a somewhat outdated Overview of Interdyme available.