The Aggregated Motor Vehicle Tariff Application (Rational Expectations)

This is a rational expectations version of the Motor Vehicle Tariff application which is the main application of the MONASH Model in the MONASH Model Book.

The results reported in the book were obtained using static expectations and a disaggregated (115-commodity) data base.

Here you can solve this application under rational expectations using a more aggregated (33-commodity) data base.

We assume here that you have already downloaded and solved the static expectations version of this application. In particular, the instructions that we give here assume that you have worked through the static expectations version.

The rational expectations application described here is documented in the CoPS Working Paper IP-81 (May, 2003)
Rational Expectations for Large Models: a Practical Algorithm and a Policy Application
by Peter B. Dixon, K.R. Pearson, Mark R. Picton and Maureen T. Rimmer. (download)

This application is slightly different from the one in the MONASH Model Book. This is because the TABLO Input file used in IP-81 has a few additions to the file documented in the MONASH Model Book. See section 3.1.2 of IP-81 for details. The TABLO Input file for the application in IP-81 is called MONASHRE.TAB. You can see the additions at the end of that file.

No GEMPACK licence is required to solve this aggregated version.

1. Download the example application

Save as MCAMVRE.ZIP (6.0 Mb) in the ARCHIVE subdirectory of the directory in which you installed RunDynam.

2. Load the example application into RunDynam

Start RunDynam running by double clicking on the RunDynam icon on your desktop.

Select, from the File menu, the item
Restore Ingredients from ZIP Archive
and follow the instructions below.

Select the ZIP archive MCAMVRE.ZIP which you downloaded (see above).

You will then be asked to specify the directory to restore the files to. We suggest that you specify the MCAMVRE subdirectory of the directory in which you installed RunDynam. [Probably RunDynam will make this suggestion to you.]

Then agree to all the suggestions that RunDynam makes. At the end you should see a message saying that "Restored application is now the current application.".

3. An overview of the example application

On the Sim Overview page you can see that the Base Case and Policy runs will cover 20 years, starting with data for the year 1997 and producing data for each year from 1998 to 2017.

You can also see that Do Rational Expectations ? is checked and that 30 iterations of the Base, 1 of the Rerun and 20 of the Policy will be carried out.

Otherwise all closure and shocks files are very similar to those used in the static expectations version. The changes are as follows:

For more details, see the file AAHIST.TXT in the EXTRA subdirectory of the Working directory.

The rational expectations methodology is fully documented in the May 2001 version of the MONASH Model book. You should look at sections 44 (Appendix to chapter 5) and 30.2 (chapter 5) for this documentation. You can download PDF versions of the relevant chapters.

The RunDynam tools for solving under rational expectations are fully documented in the special Rational Expectations Help file supplied with RunDynam. You can access this Help file by clicking on the Help button inside the Rational Expectation box on the Sim Overview page, or by selecting Help under the Rational Expectations menu.

4. Running the Base Case

Click on the Run Base Case button near the bottom of the screen.

RunDynam will run the Base 30 times (30 iterations), each time solving for the 20 years of the Base Case. 30 iterations [This will take an hour or two.]

When the whole run has finished, you can look at the results on the Results page in the usual way.

You will also be interested to see if the results have converged. [In the MONASH rational expectations methodology, the first iteration produces static expectations results. Subsequent iterations gradually move closer to the rational expectations results. Convergence is satisfactory if the results of the last few iterations are very close to each other.]

To check convergence, select View Convergence HAR file - Base under the Rational Expectations menu. Click on the Help button on the file open dialog to see what it is that you will be looking at. When this convergence HAR file is loaded into ViewHAR, check that the results from the last few iterations are very close. Look at the results for a few of the macros and, most importantly, the results for investment and capital by industry [the variables are y (investment) and cap_at_t (capital)].

To see how much difference rational expectations makes, compare the results from the first iteration with those from the last. [You can do this from the Convergence HAR file or by loading the spreadsheet files for these two iterations via menu item View convergence CSV file under the Rational Expectations menu item. If you do the latter, click on Help to see what you are looking at.]

5. ReRunning the Base Case with the Policy Closures

Click on the ReRun Base button near the bottom of the screen. For simplicity, say "No" when asked if you want to check that the prerequisites are ok.

RunDynam will solve for the 17 years of the Base Case, this time using the Policy closures rather than the Base closures. In this case, just one iteration is done. [This will take 10-20 minutes.]

6. Running the Policy

Click on the Run Policy button near the bottom of the screen. For simplicity, say "No" when asked if you want to check that the prerequisites are ok.

RunDynam will carry out 20 iterations, each one solving for the 17 years of the Policy. [This may take about an hour.]

When the whole run has finished, you can look at the results on the Results page.

You should compare your results with those in IP-81.

7. Comparing Your Results with those in IP-81

You will want to compare your results with those reported in section 3 of IP-81. Your results should be identical to those reported there.

For example, look at Chart 3.1 in IP-81. This shows the cumulative deviation results for aggregate employment. To get this chart, go to the Results for All/Several Years part of the Results page in RunDynam, and click on Differences - Policy v Base. Make sure that All is selected in the Which Years part, and that Cumulative is the selected Style. Then click on the Spreadsheet button. In the spreadsheet which appears, select the row corresponding to the variable emp_hours (which is aggregate employment). Then use the Chart Wizard to draw a graph for these results. Choose a line graph since that is used in Chart 3.1 in IP-81. What you will see is a graph of the rational expectations results. In Chart 3.1, the value zero is shown for year 2001. Your graph will however start at year 2002. See point 8 below for obtaining the static expectations part of Chart 3.1.

Alternatively, you can obtain the graph in Chart 3.1 via the graphing capabilities of ViewSOL. To do that, click on Graph/ViewSOL (rather than on "Spreadsheet"). When ViewSOL loads the different solutions (this will take a little time), make sure that, under the Time Series.. menu, Show.. Diff and Style.. Cum are selected. Then click on Macros to get the results for all macro variables. Then right click on the emp_hours row. This will produce the rational expectations graph for aggregate employment. [The graph is shown in the Charter program. You may need to switch to that program to see the graph.]

8. The Static Expectations Results in IP-81

As explained in the footnote in section 3.1 of IP-81, the static expectations results shown in the charts in IP-81 are the cumulative differences between the Policy results after just one Policy iteration and the results after the Rerun of the Base. The spreadsheet file containing the Base Rerun year-on-year results is RX2RSSY.CSV. The spreadsheet file containing the year-on-year results after the first Policy iteration is PX2PSSY_PITER1.CSV.

You can use the GEMPACK program DEVIA to obtain the cumulative differences between these two sets of results. To do this, go to a DOS box in the Working directory via menu item File | Shell to DOS in Working Directory. Then run the program DEVIAA.EXE (which you will find in the directory in which RunDynam is installed). When this program begins to run issue the following responses. [Just enter the parts shown in bold. The first response is a carriage return (or Enter).]

Carriage return (to accept the default program options)
px2pssy_piter1.csv (the Policy static expectations results)
rx2rssy.csv (the Base Rerun results)
px2pdevc_piter1.csv (the output file)

This will produce file PX2PDEVC_PITER1.CSV which contains the desired cumulative differences. These are the static expectations results reported in IP-81.

Open this output file PX2PDEVC_PITER1.CSV in your spreadsheet program. You can get the static expectations parts of the charts in IP-81 using the graphing capabilities of your spreadsheet program.

9. Different Numbers of Iterations and Different ADJ_RE Values

This suggests how you can check the results reported in section 2.5 of IP-81 about convergence. As explained there, it is important to be very fussy about the convergence of the Base. Otherwise you may obtain incorrect cumulative deviation results.

We do not suggest that everyone should do this, only those interested in understanding more about the robustness of the results obtained above.

To investigate these issues, you need to recompute the results using different numbers of iterations for the Base (and for the Policy).

Above you were using ADJ_RE values of 0.3. To see this, go to the Model/Data page of RunDynam, click on the ROREXT line, then right click and select menu item View this file from the popup menu which appears. Click on Yes to see the original version of the ROREXT file. When ViewHAR shows you the Contents page, click on header ADRE to see the ADJ_RE values. They should all be 0.3.

Below you will run with smaller ADJ_RE values, namely 0.2 (as was used in Charts 2.2 to 2.5 in IP-81). To change to these values, go back to the Model/Data page on RunDynam and click again on the ROREXT data file. Then right click and select Select or change the filename from the popup menu which appears. In the resulting file dialog, select file MCAROR97V2.HAR. This file has ADJ_RE values set to 0.2 (as you can check). You will use these values in the calculations below. We have found that more iterations are required to get really accurate results when ADJ_RE is 0.2, compared to when ADJ_RE is 0.3 (as they were in the results above).

First we suggest that you compute results with only 20 iterations of the Base (and with ADJ_RE equal to 0.2). To do this, go to the Sim Overview page of RunDynam and change the Sim names to BX1, RX1 and PX1 respectively. This will ensure that the results you are about to obtain do not overwrite the correct results obtained (using names BX2,RX2,PX2). Then change the number of Base iterations from 30 to 20 and recompute all results by selecting menu item Rational Expectations | Do Base, BaseRerun and Policy.

Your cumulative deviation results for real aggregate investment will be similar to the 20 iteration basecase, fully converged policy part of Chart 2.5 in IP-81. As remarked in section 2.5 of IP-81, the cumulative deviation results are not accurate because the Base has not converged sufficiently.

You might like to experiment with different numbers of Policy and Base iterations, and with different ADJ_RE values. [We have supplied ROREXT files with different ADJ_RE values. For example, MCAROR97.HAR has ADJ_RE equal to 0.1, MCAROR97V4.HAR has ADJ_RE equal to 0.4 and MCAROR97V9.HAR has ADJ_RE equal to 0.9.] With the very high ADJ_RE values, you will see considerable oscillation (rather than convergence), as remarked in section 2.5 of IP-81.

Disaggregated Version of this Rational Expectations Application

After you have worked through the application here (with the 33-commodity data base), you may wish to carry out the same application with the disaggregated (115-commodity) data base for MONASH. You can download the relevant files.

Go back to
Motor Vehicle Application Page
MONASH Model Page