Estimating the Likelihood of Future Tax Settlements Using Firm Fundamentals and Text Disclosures

This study develops a measure of the likelihood of future corporate income tax settlements using quantitative and qualitative information found in the Form 10-K. We use machine-learning to identify language in the tax footnote and tax related sentences of the MD&A that is useful in predicting future tax settlements. The incremental explanatory power of our qualitative prediction is economically significant, which suggests that the language in the Form 10-K can be particularly useful in predicting future tax settlements. We also identify quantitative characteristics found in the financial statements that are useful in predicting future tax settlements. We find that the market more negatively values current tax expense for firms that have a higher ex ante likelihood of settling their uncertain tax positions with a tax authority. In addition to providing evidence that qualitative information predicts future outcomes, researchers can use our measure to approximate the likelihood that a firm will have a future tax settlement with a tax authority.