This describes the computer codes for the simulations and data for 
"Education, Sectoral Composition and Growth" by Joseph Kaboski.

The data is quite large (~650M).  Data can be acquired at www.ipums.umn.edu.  The dataset contains data from the 1% samples of 1940, 1950, 1960,1970, 1980, 1990, and 2000.
The variables obtained are year, gq, slwt, perwt, age, sex, educrec, higradeg, labforce, occ1950, ind1950, wkswork2, hrswork2, and incwage.
In addition, for 1990 and 2000, educ99 (years of education) variables replace higradeg (highest grade completed). 
In additin, for 2000, uhrswork (usual hours work per week) replaces hrswork2 (hours worked last week).  

The files are then named us19XX.dta, where XX is the year.  
The sequence is then to run:
makedatasets.do (changes variable names and creates relevant variables), 
makeallofit.do (creates an ID to help fill in missing wage data as detailed in the appendix)
makeallofitschool.do (does the same for missing schooling data)
fillinwages.do (fills in missing wages)
fillinschool.do (fills in missing schooling).

The output is then usXX.dta and usXXs.dta for wages and schooling respectively.   

Alternatively, these usXX.dta files can be acquired by emailing me at kaboski.1@osu.edu.

The decompositions are performed by nineoccsandindsgrowthdecomp.do and nineoccsandindsschooldecomp.do, which perform the decompositions 
of wage growth and schooling growth, respectively.

Simulation results can be obtained using:

simulatecompstats.m (performs the comparative statics simulations, Section 4.2)
simulate40506070809000.m (performs the calibrated simulations for 1940-2000, Section 4.3)
simulategrowthdecomp.m (performs the decomposition of the growth effects of schooling into direct and indirect effects, Section 4.4)

The file simulate40506070809000.m has the most documentation within the code, but the others follow a very similar structure.
These files call the following subroutines:

solveendogpm.m
endogpm.m
solveexogpm.m
exogpm.m
ssolve.m