1. Software and operating system used: 

Empirical part of the paper: Stata 

Computational part of the paper: Matlab R2012a on Mac OSX Yosemite Version 10.10.5

--------------------------------------------------------------------------------------------
2. The order in which the programs need to be run:

Empirical part of the paper:
1. GEM_Regressions.do produces Figure 1(GEM), and half of Table 1 and Tables A3-A12 (in Appendix).
2. ES_Regressions.do produces Figure 1(ES), and half of Table 1 and Tables A3-A12 (in Appendix).

Computational part of the paper:
1.MAIN FILE: running main.m produces Tables 2,3 4,5, and A17 and produces mat-files  
for Figure 4,5, A1, and A2; while the figures in the paper  are generated in STATA
using these matfiles, matlab also generates unformated versions of them. 
main.m calls all necesssary m.files and matfiles which are in the subfolder "main". 

2. CROSS-COUNTRY: running main_countries.m  produces mat-file used for Figure 3
main_countries.m calls all necesssary m.files and matfiles which are in the 
 subfolder "countries"; while the figure in the paper is generated in STATA using this matfile, 
matlab also generates  an unformated version of the figure.

3. UNIQUE DISTRIBUTION: running main_uniquedistr.m produces 
Tables A13, A14, A15, A16 for results using a unique talent
distribution for all three education groups.  main_uniquedistr.m 
calls all necesssary m.files and matfiles which are in the subfolder "unique_distr".

4. HIGH SUBSTITUTABILTY: running main_highsub.m produces 
Tables A18, A19, A20 for a parametrization of the model 
with very  high substitutabiliyy between education groups. 
main_highsub.m  calls all necesssary m.files and 
matfiles which  are in the subfolder "high_sub".

--------------------------------------------------------------------------------------------
3. Expected computation time:

Computational part of the paper: 
1. Running main.m to obtain all main results takes 39 minutes

2. Running main_countries takes 46 minutes

3.Running main_uniquedistr.m to obtain results for a unique
talent distribution takes 8.8 minutes

4. Running main_highsub.m to obtain  results  for the model 
with very  high substitutability between education groups 
takes 3.6 minutes

--------------------------------------------------------------------------------------------
4. Seed and random number generator:

Empirical part of the paper:  not used here

Computational part of the paper: not used here

--------------------------------------------------------------------------------------------