Print
E-MailDhammika Dharmapala is the Julius Kreeger Professor of Law at the University of Chicago Law School. This post is based on a recent paper by Professor Dharmapala.
An extensive literature across law, accounting, economics and finance has analyzed the compliance costs of securities regulation. In the US context, the Sarbanes-Oxley (hereafter SOX) legislation enacted in 2002 has been a particular focus of attention. However, its social welfare consequences remain controversial. My new working paper on Estimating the Compliance Costs of Securities Regulation: A Bunching Analysis of Sarbanes-Oxley Section 404(b) brings both a novel dataset and a new empirical approach to bear on this important question. Many of the most significant provisions of pre-SOX—in particular, auditor attestation of internal controls under Section 404(b)—have been applied only to firms at or above a threshold of $75 million of “public float” (i.e. the market value of shares held by non-insiders); firms satisfying this threshold are referred to as “accelerated filers.”
Despite its importance for determining regulatory obligations, public float is not reported in standard financial databases. While previous studies have hand-collected public float for various subsamples of firms, we construct a much larger dataset by using Python code to “scrape” information on public float from firms’ 10-K filings. Using this method, we collect public float information for the universe of reporting entities for fiscal years 1993-2015 (yielding nearly 161,000 observations of public float). We divide this data into pre-SOX (1993-2002) and SOX (2003-2015) periods of approximately equal duration. As suggested in the prior literature, firms can potentially “manage” their public float through insider purchases of stock and other mechanisms. Such strategies entail various costs, which firms would weigh against the net costs of SOX 404(b) in determining what public float to report.
We apply to this new dataset an empirical technique that has not previously been used to study securities regulation. It draws on a growing literature in economics analyzing “bunching” around bright-line tax and regulatory thresholds. In a setting with a size-based threshold, a bunching analysis examines the divergence between the number of firms just below the threshold and the counterfactual density (i.e. the number of firms that would be expected to be found around the threshold, absent the regulation). The counterfactual density is calculated by fitting a flexible polynomial function to the observed density of firms, excluding an interval close to the threshold (and assuming that the density would be smooth absent the threshold). Our application of this approach involves estimating the counterfactual density by fitting a fifth-order polynomial function to the observed density of firm-years in the range of $50-$150 million of public float, excluding the $66-$83 million interval (which is chosen using an iterative process).
In the pre-SOX period, there is no detectable divergence between the actual and counterfactual density around $75 million (nor around an inflation-adjusted equivalent of this nominal amount). In contrast, in the period following the enactment of SOX, there is a substantial “excess” number of firms below the $75 million threshold, relative to the smooth counterfactual density. This bunching is bimodal, being evident both immediately below $75 million and some distance below it (around $70 million). Bunching some distance from the threshold may be reasonable, in that firms may wish to leave a buffer in the event of an unexpected increase in stock price on the last day of the second quarter (on which public float is calculated). The magnitude of the excess mass (which consists of 257 excess observations, relative to a counterfactual number of 1334 observations in the range $66-$75 million) implies that firms around the threshold reduce their public float by about $1.7 million on average (that is, averaged over all firm-years in the region, including those firms that are for various reasons indifferent to the threshold). This estimate is statistically significant, using bootstrapped standard errors. Using a set of additional assumptions, this baseline estimate is consistent with a net compliance cost of approximately $6 million for firms in the region of the threshold, in present value terms (over the time horizon over which a firm may expect to remain below the threshold in the future).
We address a number of potential alternative explanations for bunching. It is possible that negative market reactions upon crossing the threshold may increase the number of public float observations just below the threshold. However, the result is robust to modifying the threshold to account for this possibility. It is possible that some firms crossed the threshold in the past and became accelerated filers, even though they have public float below the threshold (noting that a firm that crosses the threshold typically remains an accelerated filer even if its public float subsequently falls below the threshold). However, the result is robust to excluding firms that reported being accelerated filers in the previous year while having public float below $75 million. Finally, it is possible that firms’ insiders may avoid crossing the threshold to retain private benefits of control, rather than to avoid compliance costs (although it appears unlikely on a priori grounds that private benefits would be substantially affected by the relevant regulations). However, the basic result is robust to omitting firms at which private benefits may be particularly high—those that have dual-class stock or a high value of the entrenchment index constructed by Bebchuk, Cohen and Ferrell.
Nothing in the paper establishes directly that the observed bunching behavior is motivated by the regulatory threshold. However, there is no evidence of bunching at this threshold in the pre-SOX period, and the result is robust to considering various alternative explanations for bunching. Thus, the paper provides a novel form of evidence on one of the central issues in the study of securities regulation. In particular, it provides a direct measure of firms’ perceptions of the net value of the regulation that is difficult to obtain using other empirical approaches. More generally, its empirical approach may potentially be of wide applicability in the analysis of company law and securities law in various countries where bright-line thresholds are used to determine which firms are subject to particular legal rules.
The full paper is available for download here.
