The SAS programs used in the analysis are named table1.sas - table8.sas 
The corresponding SAS outputs, which contains output not shown in the paper 
but which might be interesting, are presented in tabl1.lst-table8.lst.

The data are in a SAS data set called 

guns.sas7bdat. 

Downloading the data and running the programs will reproduce the results presented 
in the tables in the paper. The only editing required will be that the library 
reference near the top of each program will have to be changed to the directory 
where the guns.7bdat file is stored. These programs were run using SAS 8.02 on a 
Unix server so the library reference is in Unix.

The results are essentially unchanged if the analyses are done with all the variables  
logged. SAS programs that generate Tables 4-8, but with all variables logged are called 
table4.logs.sas, table5.logs.sas, etc. and the corresponding list files are 
table4.logs.lst, table5.logs.lst, etc. These results are mentioned in the paper,
but are not reported to save space.

To verify that the results are also robust with respect to the inclusion or exclusion 
of the arrest rate, delete the relevant arrrest variable (aomaj, aomur, aorap, aorob, 
aoass, aobur) in each equation and run the program.

To verify that our results are robust to the choice of gun measure, edit the programs 
substituting gss for gsshg, pgun for phg. To verify that the results are the same when
using percent gun suicide, substitute pgs for either gsshg or phg. However, in this case,
the coefficient estimates cannot be interpreted as the percent change in crime from a 
given change in guns. See Moody and Marvell 2003 Pitfalls of using a proxy variable in 
studies of guns and crime. Available at http://ssrn.com/abstract=473661.

For analysts who do not use SAS, we have provided the data in Excel format (guns.xls).
The SAS programs can be readily translated into Stata or other statistical languages.