Survey_Graham_1996

Abstract

The data are gathered from all three annual Compustat tapes: Full-Coverage, Primary, Secondary, Tertiary, and Research. These files contain information from 1973 through 1994, although this paper focuses on tax rates calculated for the years 1977-1991. The analysis is performed on data samples with observations deleted if any of the variables has a missing value, with two exceptions. In raw form, the Manzon and tax-exhaustion rates have approximately 15% and 11 % of their values missing in the main sample, respectively. because NOL and ITC information is frequently missing from the Compustat tapes. Consequently, if NOL or ITC data is missing, I reassign the missing values to zero. After this modification, none of the proxies has values missing more frequently than 0.1 % of all observations with nonmissing taxable income. The data is divided into two primary groups. First, a rnrr)forward sample is formed with tax rates from 1977-1979 that are compared to perfect foresight MTRs (which use actual taxable income realized over the full carryforward period). For example, the foresight MTR for 1979 uses actual taxable income data through 1994 to capture the entire 15-year carryforward period. Firms are deleted if taxable income is missing for any year between 1980 and 1994, leaving a sample of 15,246 firm-year observations representing 5,700 firms. The requirement that taxable income be available through 1994 implies that the carryforward sample is subject to survivorship bias. Consequently, a carryback sample is also analyzed. Instead of requiring that taxable income data be available through the entire 15-year carryforward period, the carryback restriction requires that taxable income be available for three subsequent years. Thus, the carryback sample contains all observations that can directly affect year t income via the carry back provision of the tax code, and permits the use of tax rates from 1980 through 1991, representing 12,576 firms and 71,311 firmyear observations.