We run two sets of experiments:
1. The experiments from the main text. We run these all together using the batchfile "run_main_text_experiments.bat", but the tab-files and cmfs 
can also be run separately from gemsim (using the fortran option). The batchfile first runs the tab-file and compiles, then runs the cmfs, and finally writes the relevant variables for the tables in the main text to csv-files. To generate the tables we import the results into stata and generate
tables. Somehow the batchfile does not continue after compiling (even when adding the statement "CALL" before it), so the batchfile has to be run 
twice, first compiling and then running the experiments. 
2. The experiments to compare the way variaty-scaling is modelled. This can only be run from the batch-file "experiment_shocks_compare_etk_code", as it
passes a range of parameters from the batch-file to the command files. The results are written to csv and analyzed with a do-file.

The folder with GEMPACK-based code contains the following subfolders:
1. batchfiles
2. cmf. The different cmf have self-explanatory names. We need different command files to call different parameter files and to implement different
sizes of shocks. Switching between the models occurs through changes in the parameter files as explained in the main text. We run the sims with Gragg, 
but also checked Euler. The sims with a granularity parameter of 0.67 require more steps to get accurate results. The sims comparing the three ways 
of modelling variety scaling are based on a single-step solution with a large number of steps. This is done multiple times with different single-steps 
(101,131,161 steps) to compare. The main text sims are based on multi-step simulations.
3. code. There are three tab- and sti-files in this folder:
a. aekm_final. The main tab-file used for the sims in the main text.
b. aekm_final_alt_sshift. The same code as in the main tab-file but using a different, more complicated (and precise) way to do variety scaling
c. gtap_mc_francois98. The code used in Francois (1998) to do variety scaling as described in the text.
4. In data there are two folders, based on respectively a gtap10 and gtap92 aggregation to 11 sectors, 11 regions and 6 factors. gtap10 is used for
the main text and gtap92 for the comparison. There is no particular reason to do so, and the comparison-sims on variety scaling can probably easily 
be redone with the gtap10-aggregated data. As explained in the main text we switch between the different models by switching between prm-files.
5. do_files. Used to make tables. Paths are adjusted to the fact that at the WTO STATA is run through a server.
6. simulations. Results from the sims and the csv files will be written to subfolders of this folder.
7. updated. Will contain the updated data-files.  
