Abstract
The question of how best to account for income taxes in financial reports has plagued the accounting profession virtually since the enactment of the tax. The reason for this question is differences between the amount of taxable income in the tax return and income before taxes in the financial statements. Especially significant are items included in calculating both income amounts but not in the same year, known as timing differences. Three bases for allocating timing difference have been proposed: flow through (FlTh), partial allocation (PITA), and comprehensive allocation (CITA). This study compares these three allocation bases. The criterion used to compare these allocation bases is predictive ability. One of the primary objectives of financial accounting reports is to provide information useful in predicting future operating results of and cash flow from an enterprise to its investors and creditors. Based on this objective, an accounting treatment which allows statement users to make more accurate predictions of future results and cash flows is superior to its less accurate counterparts. A random sample of two hundred firms was selected from the Compustat Industrial tape. Firms with fewer than twelve years data were deleted from the final sample. The amount of tax expense using each allocation basis FlTh, PITA and CITA, and the cash flow for income taxes was observed or computed for each company for the fifteen year period from 1967 through 1982. The relationship between each tax basis and cash flow for taxes was modeled using ordinary least squares regression for periods of ten, eleven and twelve years. The regression models were used to predict cash flow one, two, and three years beyond the prediction period applying each of five predication models. The predicted values were compared with actual amounts to determine comparative predictive accuracy of the allocation bases. Comparison of allocation bases indicates that CITA has the largest prediction error for all models. For most models, PITA outperforms FlTh slightly although this apparent superiority may result, at least in part, from the technique the study used to compute PITA. Tests using analysis of variance show significant differences for only two of ten combinations of error calculation and model. For these two cases, CITA is significantly different from PITA and FlTh. In no case is there a significant difference between PITA and FlTh...
Richardson, Clara Lucile (1985). The predictive ability of alternative bases for interperiod tax allocation : an empirical study. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -430100.