Hanlon, Mills & Slemrod on An Empirical Examination of Corporate Tax Noncompliance

Michelle Hanlon (University of Michigan, Ross School of Business), Lillian F. Mills (University of Arizona, Eller College of Business and Public Administration) & Joel B. Slemrod (University of Michigan, Ross School of Business) have posted An Empirical Examination of Corporate Tax Noncompliance on SSRN.  Here is the abstract:

This paper offers some exploratory analysis of an extraordinarily rich data set of audit and appeals records, matched with tax returns and financial statements, of several thousand corporations. We find that corporate tax noncompliance, at least as measured by deficiencies proposed upon examination, amounts to approximately 13 percent of “true” tax liability. Second, noncompliance is a progressive phenomenon, meaning that noncompliance as a fraction of a scale measure increases with the size of the company. Other things equal, noncompliance is related to two measures of the presence of intangibles and with being a private company. We find some evidence that incentivized executive compensation schemes are associated with more tax noncompliance, but only with respect to bonuses and not for stock options and other equity-related incentive pay. We uncover no relation between a commonly-studied measure of the quality of corporate governance and the extent of proposed (scaled) tax deficiency. Finally, we find that there is no consistent simple or partial negative association between our measure of tax noncompliance and measures of the effective tax rate calculated from financial statements. These conclusions are preliminary because our central measure of tax noncompliance is the result of an imperfect and perhaps systematically detailed audit of a tax return declaration that may itself be the opening bid in what is expected, often correctly, to be an intense negotiation and formal appeals process. Second, the causal links among tax aggressiveness, executive compensation, and corporate governance are potentially complex, and the analysis presented here at best establishes statistical associations, but certainly does not establish causal relations.