Fixups - Example 5

TAXCALC includes a variable _FIXUP which can be set to "fix up" various intermediate quantities to the taxpayer supplied value. This is chiefly intended for use with the PUF data.

If you were experimenting with a change to itemized deductions with the PUF, you might wish to accept the taxpayer calculation of AGI, so that censoring of other income didn't affect your results. For example, by setting _FIXUP to 1, then the AGI calculation will include an "addfactor" such that AGI always matches the taxpayer calculation, and the regression of calculated AGI on taxpayer supplied AGI will be perfect (standard errors are zero, R**2 is one):

_FIXEIC which (when set to 1) causes the EIC calculation to be made with the taxpayer's value for EIC eligible earnings, rather than the sum of wages plus positive schedule C plus combat pay. We do that in order that the marginal tax rate on earned income will include the EIC subsidy.

_NOMTRFIX will stop the smoothing of various step functions in the tax law, including the tax tables for taxpayers with taxable income under $100,000 and the child care credit.

/* Example 5 - FIXUP */ options nocenter; libname t "/home/data/soi/e"; %INCLUDE "taxcalc.sas'; data; set t.e2005; %INIT; _fixup = 1; %COMP; keep e00100 c00100 weight; run; proc reg; model c00100 = e00100;

Notes

The available values for _FIXUP are:
  • 0 or missing: No fixup
  • 1 fix AGI
  • 2 fix taxable income
  • 3 fix tax before credits

Fixes are cumulative. That is, if taxable income is fixed, then AGI is also fixed. Obviously you can't usefully calculate a marginal tax rate or change the data for any E value that _fixup would compensate for.

Note that due to the nature of SOI censoring, it is possible for a fixup to produce worse results downstream of the fix. For example, for a taxpayer with $3 million dollars in gambling winnings, offset by 3 million dollars in gambling losses taxcalc will generate the correct taxable income with no fixup, since both items are omitted from PUF. But if AGI is corrected (by adding $3 million), taxable income will be just that much too high. This is actually a real taxpayer in the sample, and one with a profound effect on the R**2s of test3.