|The US Model in EViews 7: October 30, 2016|
US Model Optimized for EViews: October 30, 2016, version of the US model
If you have EViews on your computer and want to work with the US model in EViews, there are now two ways to do this. First, you can do this by downloading the zipped file usev163.zip. You should download this file by clicking it, and then unzip it. One of the files in the zipped file is readme.txt, which is repeated below. This version is due to Matthew Cushing at the University of Nebraska-Lincoln. It takes advantage of many more of the features of EViews than does the other way (below) of downloading EViews.
This file descibes how to convert the FAIRMODEL into a convenient EVIEWS workfile. The advantages of this workfile over the one traditionally supplied on the FAIRMODEL site are: 1. It uses the native EVIEWS NA rather than the FP 999999 to represent a missing observation. EVIEWS automatically skips missing observations. Future values of the endogenous variables are represented by NAs. This automatically avoids the mistake of estimating equations over forecasted data. 2. It exploits EVIEWS's ability to recognize simple functions in regression commands. The variable LOG(CS/POP) is entered directly into equation 1, so there is no need to define LCSZ and then recover CS from the forecasts of LCSZ. The equations are much easier to interpret, as they look very much like the equations reported in the FAIRMODEL appendix. 3. Coefficient restriction are made explicit in equations 16, 23 and 24, rather than imposing them by redefining variables. 4. Variables have labels. Selecting a variables and pressing F9 gives the description of each variables. 5. OLS equations (eqs 15 and 29) estimated with AR1 errors use Eviews option(arma=cls, optmethod=legacy). These estimates closely match estimates using the FAIRPARK program. First you must edit the SETUP.PRG file by selecting the directory in which your files are located. You can also select the forecast horizon (currently 2022q4.) If you are updating the installation, you will have to change the %DEND variable to reflect the last year of actual data. You can also change %END, if the forecast horizon has changed, or if you simply want to have a longer forecast horizon. You must download FMAGE.DAT and FMDATA.DAT from the FAIRMODEL site. They are in the ZIP file for downloading the model in FP. To convert FMAGE.DAT and FMDATA.DAT to a form that EVIEWS will read you must delete all occurances of: 'END' LOAD ; and SMPL 1952.1 2016.1 END You should then rename them FMAGE_E.TXT and FMDATA_E.TXT before running SETUP.prg
end of readme.txt
US Model in EViews: No Optimization: October 30, 2016, version of the US Model
For this version the zipped file to download is: fmev.zip. You should download this file by clicking it, and then unzip it. The datasets in it are:
The following discussion assumes that you know EViews 7. If not, you should learn EViews first before trying to get the model working in it.
Enter EViews and type LOAD FM. The entire model is now loaded. In the main workfile the model is called FMA. If you double click FMA, you are ready to go. The model has been solved for the 2016:4--2022:4 period in the workfile, and the solution values end with _0. These solution values are the forecast values of the October 30, 2016, forecast within rounding error. You do not need the datasets FMGENR.PRG, FMEQ.PRG, and FMA, which are included in FMEV.ZIP, but they may be useful for reference purposes. FMGENR.PRG generates various variables, and FMEQ.PRG estimates the model.
EViews handles the use of 2SLS with autoregressive errors somewhat differently than does the FP program, and so some of the estimates using EViews differ slightly those using the FP program. EViews sometimes adds instruments from the ones specified, which FP does not. If the ALT2SLS option is used in the FP program (see page 61 of the User's Guide), which is used for the estimation of the US model, then extra instruments do not have to be added to insure consistency.
For the current version (October 30, 2016) there were problems with equations 4 and 26. The coefficient estimates of these two equations were not close to the model's estimates. Again, the autoregressive errors were handled differently. The EViews estimates for equation 4 were retained, but equation 26 was estimated by OLS rather than TSLS. You will see that some of the coefficient estimates for equation 4 are problematic, which causes the EViews forecast values to be somewhat different than the FP forecast values.
Two variables are renamed when using EViews. COS is renamed COSS, and AR is renamed ARR.
This should be enough to get you started. If you want to modify the model in bigger ways than are allowed on this site and/or want to estimate the model, EViews allows you to do this.
NOTE: This site receives no money from EViews; we are simply making the US model datasets available for those who are interested.