Stacey H. Chen and Shakeeb Khan, "Semi-Parametric Estimation of
Program Impacts on Dispersion of Potential Wages," Journal of Applied
Econometrics, Vol. 29, No. 6, 2014, pp. 901-919.

All data and programs for empirical applications are zipped in the
file ck-empirical.zip. The data set (data66.dta) used in the article
is public available, downloadable from Kling's (2001) Data Archive on
the website of the Journal of Business and Economic Statistics. A
plain ASCII version of the data (data66.txt) is also provided. 

There are 5225 observations. All of the variables used in our study
are listed in the file data66.dct.  Sample selection rules are
described in the files data66.dct and main20121122.do. The program
file main20121122.do calls three macros. (1) sum1.do calculates and
reports the descriptive statistics of key variables; (2) qr3.do
implements quantile regression models; (3) bsnlsym.do implements
bootstrap iterations.  


Programs for replicating the simulation results in Online Tables A1
and A2 are zipped in the file ck-simulation.zip.


1. The "IV" sub-folder contains programs  for IV estimators: 

(1) IV5000.gau generates the IV results of the first row in Table A1,
for 5000 observations.

(2) IV.gau generates the IV results of Case 1 in Table A1, for sample
sizes 190, 760, and 3040.  

(3) IV1x.gau generates the IV results of Case 2 in Table A2, for
sample sizes 190, 760, and 3040.  We report the scale rates as x_2=0. 

(4) IV2x.gau generates the IV results of Case 3 in Table A2, for
sample sizes 190, 760, and 3040.  We report the scale rates as x_2=0. 


2. The "var" sub-folder contains programs for generating the results
for variance-based pairwise matching:

(1) var5000.gau generates the quantile estimates of the first row in
Table A1, for 5000 observations.

(2) var.gau generates the quantile regression results of Case 1 in
Table A1, for sample sizes 190, 760, and 3040.  

(3) var1x.gau generates the quantile regression results of Case 2 in
Table A2, for sample sizes 190, 760, and 3040.  We report the scale
rates as x_2=0. 

(4) var2x.gau generates the quantile regression results of Case 3 in 
Table A2, for sample sizes 190, 760, and 3040.  We report the scale
rates as x_2=0. 


3. The "QR" sub-folder contains programs for for generating the
results for quantile-based pairwise matching:

(1) qrsimu5000.gau generates the quantile estimates of the first row
in Table A1, for 5000 observations.

(2) qrsimu.gau generates the quantile regression results of Case 1 in
Table A1, for sample sizes 190, 760, and 3040.  

(3) qrsimu1x.gau generates the quantile regression results of Case 2
in Table A2, for sample sizes 190, 760, and 3040.  We report the scale
rates as x_2=0. 

(4) qrsimu2x.gau generates the quantile regression results of Case 3
in Table A2, for sample sizes 190, 760, and 3040.  We report the scale
rates as x_2=0. 

(5) boladth.* are the compiled codes required to run quantile regressions.
These are binary files which require a 32-bit version of Gauss.


Please address any questions to:

Stacey Chen
128 Academia Road Section 2
Institute of Economics
Academia Sinica 
Taipei
Taiwan R.O.C.