A Comparison of Quantitative and Qualitative Disclosures of Tax Settlements to Assess Their Favorability
比较了从税务脚注中提取的三种代理变量（两种定量、一种定性）来衡量税务和解有利性的可靠性，发现定性文本分析方法在两项验证中均表现最佳。
ABSTRACT Tax footnotes, particularly disclosures of unrecognized tax benefit (UTB) accruals, contain significant information. To aid large-sample empirical research, and assess reliable measures of the favorability of tax settlements, we compare three potential proxies of footnote information: two derived from quantitative information and one derived from qualitative information. To validate the proxies, we use hand-coded disclosures about the nature of UTB settlements and an empirical model of the revaluation of prior, open tax positions at the time of settlement. We find that the textual analysis method to infer settlement favorability from qualitative text in tax footnotes performs well in both validations. Conversely, the tax rate reconciliation proxy is discriminant only in the first validation, whereas the accrual of interest and penalties proxy relates to settlement favorability only in the second validation. Overall, the qualitative proxy is most reliable and represents a new and useful measure for the empirical tax literature. Data Availability: Data are available from the sources cited in the text. JEL Classifications: C89; D81; H25; M48.