At the NBER we have calculated income tax liabilities using the data in the Tax Model Files for each individual taxpayer, for each year for which data is available. By comparing the calculated amount with the reported amount, certain discrepencies in the files are revealed. Here are the major one we notice. In addition to these, there are many individual taxpayers with anomolous tax liabilities. These are not listed.
For years prior to 1967, the PDF and TXT files disagree about nearly all variable locations. In every case the TXT file is correct.
In 1962 Schedule D the PDF shows a blotch at line 9, which should read 'E22' and shows 'E12' at line 12 where it should show an 'E22'. The 'E12' at line 17 should be a blank.
In 1966 the number for Dividends Received is always zero. The numbers for total ordinary gains and other gains plus and minus don't seem to be added to AGI by the taxpayer, and may be erroneous. There are many taxpayers with a conflict between their filing status and there number of exemptions, it seems that the number of exemptions is the better guide. There are many taxpayers with form of deduction set to standard, but total deductions of far above that amount. Therefore we suggest ignoring the form of deductions code. Total income tax liability calculated is about 6% short of the published total, suggesting missing observations or incorrect weights.
In 1968 FIELD 58 is marked 'RESERVED' in the documentation, but is really business income.
We find that the variables for 'Fully taxable pensions' and 'Partially taxable pensions' have a correlation of .98+ for the three years 1972,4,6 and less than .15 for all other years. We guess that each is the total because that makes the AGI match best, but the result is not exact.
In 1974 Field 59 is shown on the example form as Medical Deduction but it is really Medical Expenses.
In 1977 TXST never equals 2 (Income Averaging). We can't think of any other way to detect averaging, either.
For 1984 on there are many returns with implausible exemption totals. That is, XTOT or XFPT are zero even though the return is not a a dependent return, and one or more exemptions seem to be subtracted from AGI to get reported taxable income.
In 1987 the field for miscellaneous deductions (Field 92) seems to be wrong when present.
Daniel Feenberg
feenberg@nber.org
Inna Shapiro
ishapiro@nber.org