Reply to 'The Reg SHO Reanalysis Project: Reconsidering Fang, Huang and Karpoff (2016) on Reg SHO and Earnings Management' by Black et al. (2019)

Abstract

In a 2016 paper (Fang, Huang, and Karpoff, 2016), we report that firms exposed to an increase in the prospect of short selling during the Reg SHO pilot program have lower discretionary accruals during the pilot period. Black, Desai, Litvak, Yoo, and Yu (2019, hereafter, BDLYY) argue that this result is not replicable. We show that BDLYY’s claim is incorrect. The accruals result previously was replicated in papers by Massa, Zhang, and Zhang (2015) and Heath, Ringgenberg, Samadi, and Werner (2019), and is easily replicable using data and code that we have shared widely since 2014 – including with the BDLYY team in 2015 – and that we recently posted publicly. The accruals result also is robust to a wide range of specification changes, including those implied by the BDLYY paper, which include: various measures of performance-matched discretionary accruals and total accruals; using our original 2012 Compustat data or currently available 2019 Compustat data; including both firm and year fixed effects; including or excluding other covariates in the difference-in-differences (DiD) tests; and using unbalanced rather than balanced panels. We conjecture that BDLYY’s results are inconsistent with prior results because they rely partly on non-standard accruals measures and/or use samples that differ from those used by Fang et al. (2016), Massa et al. (2015), and Heath et al. (2019). We conclude by discussing two theoretical concerns. First, we reiterate that an observed increase in short selling during the Reg SHO period is neither necessary nor sufficient to establish that the prospect of short selling has a disciplinary effect on earnings management, as managers’ endogenous adjustments affect short sellers’ opportunities and observed short selling. Second, we discuss a concern that the Reg SHO change appears to be too small to explain a wide range of firm outcomes, as recent empirical findings suggest.