The Fair-Parke Program

The Fair-Parke Program

The Fair-Parke (FP) program is a DOS-based, command-line program. It allows one to estimate and analyze dynamic, nonlinear, simultaneous equations models. The models can be rational expectations models, and they can have autoregressive errors of any order. The estimation techniques include OLS, 2SLS, 3SLS, FIML, LAD, 2SLAD, and some versions of Hansen's method of moments estimator. The Parke algorithm is used for 3SLS and FIML estimation. Stochastic simulation and bootstrapping are two of the key options available to analyze models. There are also a number of single equation testing options. For stochastic simulation the draws can be either from estimated distributions or from estimated residuals.

The options for analyzing models include 1) running forecasts, within sample and outside sample, 2) calculating root mean squared errors and mean absolute errors, 3) calculating multipliers, 4) solving optimal control problems, 5) estimating standard errors of forecasts by means of stochastic simulation, 6) estimating standard errors of multipliers by means of stochastic simulation, and 7) estimating the degree of misspecification of a model by means of successive reestimation and stochastic simulation. Also, general nonlinear functions of coefficients can be maximized using the program, which means that maximum likelihood estimates of the coefficients of any model can be obtained after one has written out the likelihood function.

One can move automatically from estimation to solution without any additional work. Rational expectations models are solved using the extended path method of Fair and Taylor. These models can be estimated using Hansen's method or FIML.

For forecasting purposes, the stochastic-simulation option of the FP program can be used to generate interval forecasts, which provide much more information to a business planner or policy maker than simple point forecasts. Also, the interval forecasts are rigorously computed from the model's stochastic properties; they are not just some model builder's opinion.

The FP program can be downloaded in either FORTRAN code to be compiled on the user's machine or in an executable form for PCs. The FORTRAN code is not machine specific, and this allows the program to be compiled on a variety of systems. The FP User's Guide can also be downloaded.

The FP User's Guide is available in pdf format:
fp.pdf.
The executable file, FP.EXE, is in the zipped file:
fpexe.zip.
The FORTRAN code is in the zipped file:
fpfor.zip.
When you unzip this file, you get FP.FOR, which is FORTRAN 77 code that can be compiled. You need not download this file if you are going to use FP.EXE above. The test files are in the zipped file:
fpeg.zip.
Appendix C of the FP User's Guide explains how to run the examples. You should also read the Preface of the User's Guide carefully before starting to use the program.

The FP program requires some work to learn, but once you have made the investment, you can do many advanced things quickly. The best way to learn the program is to work through the examples, including those in the User's Guide. The main references for the program are
Macroeconometric Modeling: 2018
and
Fair (1984).
Each technique in the FP program is discussed in one of these two references.

The US Model in the FP Program: October 31, 2024

The US model is available in Fair-Parke format. You can thus use all the procedures in the program on the model. It is also easy to change the model within the program (including adding more equations or extending the forecast horizon), estimate, and then solve the new version. Both EViews and FP provide these options, although the FP program has many more advanced features and some find is easier to use for large models once you have made the investment in learning it.

The US model files for the FP program are in the zipped file
fmfp.zip.
The US model files are updated quarterly.

To run the US model in the FP program, first unzip fmfp.zip. This yields the files FMINPUT.TXT, FMDATA.TXT, FMAGE.TXT, and FMOUT.TXT. Then type FP > OUT and hit the enter key. Then wait a couple of seconds and type INPUT FILE=FMINPUT.TXT; and hit the enter key. (Don't forget the semicolon.) This will run all the commands in FMINPUT.TXT and store the output in file OUT. When the job is done, compare OUT to FMOUT.txt. The output in these two files should be the same aside from rounding error. The commands in FMINPUT.TXT set up, estimate, and solve the US model. You should study the comments in FMINPUT.TXT to make sure you understand what the program is doing. You are then ready for your own analysis using the model.

The MCJ Model in the FP Program

The reference for the MCJ model is
Macroeconometric Modeling: 2018.
This contains the complete discussion and listing of the model. See also
The MCJ Model Workbook.
You should look over this workbook before working with the MCJ model.

Data from the MCJ model can be downloaded online by going to the output phase when working with the MCJ model and downloading from there.

If you want to download the MCJ model for use on your own machine, this can be done by downloading the FP program and the MCJ model files that go with the FP program. First download the FP program as above. Then download the following zipped file:
MCJ.ZIP
Unzip this file. It contains all the files needed to work with the MCJ model using the FP program.

To work with the MCJ model once you have downloaded the files, run at the DOS prompt:

FP > OUTA (hit return)
INPUT FILE=MC.INP; (typing will not show up on screen, don't forget the ;)

This job loads all the data, estimates the model, and runs a test job. The output from the test job is in the file OUT, and your OUTA file should match OUT subject to rounding error. If so, then everything is probably working properly.

As with the US model, you need to know how to use the FP program in order to work with the MCJ model.

There are some comments in MC.INP. You should read these to make sure you know what is going on. You should note the following:

• A number of "tricks" have to be used in MC.INP to link the quarterly and annual data. The annual data are stored in the first quarter, with the remaining three quarters having "missing" values. (The missing value number is -99.) A lag of 4 for an annual country is a lag of one year, not four years.
• There are two sets of trade share variables (the alphas): A____ and AA____. The AA____'s are the ones predicted by the trade share equations, and the A____'s are the ones that are constrained to sum to 1. The trade flow variables are denoted X____.
• A solution period must begin in the first quarter of a year and end in the fourth quarter of a year. Otherwise the tricks of linking the quarterly and annual data do not work.
• There are gaps in the numbering of the stochastic equations. The PRINTMODEL command allows you to see the numbering of the equations.
• Note the different treatment before and after 1999.1 regarding the EMU countries (dummy variables EU1 and EU2).
• Once you learn how to work with the MCJ model in the FP program, all the FP commands are at your disposal, and so much experimentation can be done.

The MCJ2 Model in the FP Program

Everything is the same for the MCJ2 model except the download is
MCJ2.ZIP