Judith Scott-Clayton has been generous enough to provide Stata code for 
using the March 2003 and March 2006 Current Population Files with TAXSIM. 
She starts with the Unicon versions of the files rather than the raw files 
sold by Census.

Her letter is below.

Daniel Feenberg



From Judith_Scott-Clayton at ksgphd.harvard.edu Tue Apr 17 16:49:09 2007
Date: Tue, 17 Apr 2007 16:46:53 -0400
From: Judith_Scott-Clayton at ksgphd.harvard.edu
To: feenberg@nber.org
Subject: march cps 2006 taxsim code


Hi Dan,

 

Jean Roth just delivered the most recent March CPS Unicon files, so
I've re-run my program for March 2006.  Unfortunately some of the key
CPS variables were messed up again, but in a different way than in
2003, so I had to edit my code.  Attached is the new program and log
file, along with some output from Unicon that should be helpful for
anyone who wants to extract the data using the Unicon software (they
should be able to do this by uploading the .req file once in the
program, if they get the CD from Jean), or just look at the variable
documentation (the .rpt file).

The main things I had to change:

1. The CPS's tax filing status var (flstat) was messed up for joint
filers - they tried to split joint filers by whether they were both
65+, both <65, or one of each, but they clearly didn't do it right. 
So I used the "age" variable to fix that.

2. The CPS's variable that links dependents with the people who claim
them is only valid for tax non-filers.  Those who filed taxes will
have a number here, but the number will not mean anything.  So I
recoded tax-filers to be non-dependents.  On the bright side, the
2006 variable did not have the problem that 2003 had, with dependents
being linked to line numbers that didn't exist. 

3. There was an error in my previous code: I added cap gains and
losses to get net gains.  But all of the losses were coded as
positive numbers, so I should have been subtracting them.  I fixed
this.

4. For property taxes, I decided that if there were multiple single
filers within a household, and no joint filers or heads of households
in the same household to "claim" the property tax, then these single
filers should split the estimated property tax (previously, I had
them all paying the full estimated amount).

I put a note at the top of my program that if anyone wants to use the
code for another year, they should definitely check it carefully
first, because the CPS variables (particularly the imputed tax filing
status and dependent status vars, which are key to the programming)
apparently can develop new problems from year to year.  If someone
were more ambitious they could create their own tax filing status and
dependent status vars without having to rely at all on the CPS
imputations, but that's more than I have time to do right now.

 

- Judy