Instructions for running the polarization program

There are two zip files. replicates the analysis in the AJS paper. provides an example of analysis using individual level data from the GSS.


You need to be able to decipher Stata programming files (ado files) in order to make use of the code here.  If you can, you will find that the program is actually quite simple.  I did this several years ago, so the ado files are written in versions 5 and 6 of Stata.  Updating the ML commands to the recent versions should not be difficult.  They will run fine as they are, however.


Instructions for runs the example.  At the stata prompt, type "do polar". 


Make sure your working directory is the same directory that the files are in.  I.e., when you type “pwd” at the Stata prompt, this should be the directory the unzipped files are in.  To check, type dir *.*


demo.dta is nes data from 1972-1992

small94.dta is nes data from 1994 merges the two data sets and and reshapes them for analysis

(aggregating by year and response to the abortion question) and run the analysis, calling the program mouw.ado, which is

the code for the heteroskedastic probit.


***note: the models in match the results in Table 5 of the paper

(Mouw and Sobel, AJS 2001) p. 929 1980-1994 data.  The models in

match the results in Table 5 1972-1980 data.


Looking at model m2  (i.e., the model following *m2, the lines

eq v: y year*

eq tau: a

eq m: year*

define the parameters to be estimated.

the variance is estimated with dummy variables for each year,

the cutpoint (tau) has a dummy variable a (year>1988) and the

mean has dummy variables for each year.



Obviously, other variables can be added to estimate v, tau, or m.


Instructions for

Type “do indiv” at the Stata prompt.