Tue Gorgens and Chris Ryan, "A bounds analysis of school completion 
rates in Australia", Journal of Applied Econometrics, Vol. 23, No. 3, 
2008, pp. 287-304.

DATA:

The Australian Youth Survey contains confidential data. Access may be
obtained through the Australian Social Science Data Archive
(http://assda.anu.edu.au/).

FILES:

All files are ASCII files in DOS format. They zipped in the file gr-code.zip.

edusscs.sta (uses doweights.ado, fixsex.sta): Stata program to construct
analysis data and produce figures for tables 1-3.

edubounds.gss (uses Benchmark.g, BoundsB.g, BoundsC.g): Gauss program to
compute bounds, benchmark estimates, and standard errors for tables 4 and 6
and figures 1-4. Note: Censoring figures for table 4 are labeled
"Dp=0/Dc=0", "Dp=0/Dc=1" and "Dp=1/Dc=0" in the Gauss output tables. For
table 6, the upper bound (U-hat) is labeled "UB/Prb", the naive benchmark
estimate (P-hat hash) is labeled "Prb/ran/cen", while the AYS benchmark
estimates are denoted p1011 and P1112 (P-hat) and the ABS-type benchmark
estimates are ar11e and ar12e (P-hat star) for progression from 10 to 11 and
from 11 to 12 respectively. Standard errors for U-hat can be found in the
table headed "One-sided P-values of ABS figures" in the column labeled
"10*SD". Standard errors for the benchmark estimates can be found in the
bootstrap tables.

NOTE:

For simplicity, most of the code used to carry out sensitivity and to
compute joint confidence levels are omitted. Interested users may obtain
this code from Tue Gorgens.