You are on page 1of 55

Have the Effects of SOX Section 404(b) Compliance Changed Over Time?

Jennifer McCallen
University of Georgia

Roy Schmardebeck
The University of Tennessee, Knoxville

Jonathan Shipman
University of Arkansas

Robert Whited
North Carolina State University

April 2020

We thank Lauren Cunningham, Tucker Davis, Eddie Hu, Peter Iliev, Commissioner Robert Jackson Jr. (Securities
and Exchange Commission), Clive Lennox, Commissioner Hester Peirce (Securities and Exchange Commission),
Quinn Swanquist, and conference participants at the University of Tennessee Alumni Research Symposium for
helpful comments and suggestions. We also thank Jason Ashby for excellent research assistance.

Electronic copy available at: https://ssrn.com/abstract=3420787


Have the Effects of SOX Section 404(b) Compliance Changed Over Time?

ABSTRACT: We use regression discontinuity design to evaluate the effects of Section 404(b)
of the Sarbanes-Oxley Act (SOX) on audit costs and financial reporting and how these effects
have changed over time. While audit fee premiums associated with 404(b) compliance were
substantial in the initial years of compliance, we find that the premiums declined until 2010
before reversing and returning to early SOX levels by 2015. We also find that the likelihood that
a company selects a Big 4 auditor is greater for 404(b) compliers and that this drives a portion of
the audit fee premium. Although we find some evidence that compliance results in more
effective internal controls, this effect is confined to the initial years following the extension of
management internal control reporting to non-accelerated filers. We find no evidence that the
audit of internal controls over financial reporting yields more informative internal control
reports. Together, while evidence suggests that audit fees associated with internal control audits
have not materially and sustainably declined since inception, we find no evidence that the
increased audit effort is associated with improved internal control quality or reporting. Our
findings directly respond to calls from regulators and academics for research on the updated
Section 404(b) compliance rules recently introduced by the SEC and should be of interest to a
variety of corporate stakeholders.

KEYWORDS: SOX Section 404(b); Internal Control Audits; Compliance Costs; Compliance
Benefits

DATA AVAILABILITY: All data used are publicly available from sources cited in the text.

Electronic copy available at: https://ssrn.com/abstract=3420787


1. Introduction

Although Congress passed the Sarbanes-Oxley Act of 2002 (SOX) nearly two decades

ago, regulators, lawmakers, and academics continue to debate the Act’s costs and benefits. One

of the most prominent provisions, Section 404, requires the evaluation and maintenance of

internal controls over financial reporting (ICFR) (Section 404(a)) as well as auditor testing and

reporting on ICFR effectiveness (Section 404(b)). Accelerated filers must comply with Section

404(b), but non-accelerated filers have been permanently exempted from this requirement. The

SEC recently adopted a rule that will extend the exemption from Section 404(b) to currently

compliant companies with less than $100 million in revenue, effective April 27, 2020. The initial

proposal sparked renewed interest in, and debate of, the relative costs and benefits of mandated

internal control audits, with both regulators (Jackson 2019; SEC 2019) and over 45 accounting

and law academics (Honigsberg and Rajgopal 2019) calling for more recent evidence on the

current costs and benefits of auditor attestation. In this study, we directly address the calls to

extend and expand the findings of early work in this area. In doing so, we inform both current

rule adoptions and future rule proposals by examining the impact of Section 404(b) compliance

on the audit and financial reporting processes.

It is generally accepted that 404(b) compliance imposes some level of costs on issuers in

the form of higher audit fees (Iliev 2010; Alexander, Bauguess, Bernile, Lee, Marrietta-Westberg

2013), though the exact cost is unclear, and is likely to change over time (Alexander et al. 2013;

Coates and Srinivasan 2014). In addition, Section 404(b) compliance also imposes costs on

issuers in the form of management time and effort in preparing for and participating in the

execution of the ICFR audit, though quantifying these costs is considerably more difficult.

Proponents of Section 404(b) argue that compliance prompts organizational changes and fosters

Electronic copy available at: https://ssrn.com/abstract=3420787


improvements in control environments that benefit investors through improved internal controls,

more informative internal control reports, and improved financial reporting quality (Ashbaugh-

Skaife, Collins, Kinney, and LaFond 2009; Feng, Li, and McVay 2009; Coates and Srinivasan

2014; CAQ 2017; Feng, Li, McVay, and Skaife 2015). However, opponents argue that auditor

attestation under Section 404(b) is largely redundant to financial statement audit requirements,

thus minimizing the benefits of any improvements in internal controls and financial reporting.

While the debate over Section 404(b) has continued since the inception of SOX, its

effects likely vary over time. Compliance costs (benefits) may decrease (increase) over time if

increased company and auditor familiarization with 404(b) compliance leads to more efficient

and effective ICFR implementation, testing, and remediation. In addition, changes to auditing

standards introduced by Auditing Standard No. 5 (AS5) and the PCAOB’s efforts aimed at

improving ICFR audits (DeFond and Lennox 2017) also likely affected both the effort exerted

over ICFR audits as well as the effects of ICFR audits on financial reporting. 1 We examine the

current effects of Section 404(b) by using the accelerated filer public float threshold as an

instrument in a “fuzzy” regression discontinuity design (RDD) to obtain causal estimates of the

effects of Section 404(b) compliance. We first estimate pooled effects since the implementation

of Section 404(b) and then examine whether these effects have evolved over time.

Since the passage of SOX, compliance with Section 404 has largely been determined by a

company’s public float in relation to the $75 million threshold. Given this, Coates and Srinivasan

(2014, p. 648) note, “research designed to isolate the effects of SOX is challenging” for a variety

1
DeFond and Lennox (2017) examine the effect of PCAOB inspections on ICFR audits, documenting that audit
firms receiving higher rates of PCAOB identified deficiencies in ICFR audits respond by increasing audit fees,
issuing more adverse internal control opinions, and that these adverse opinions tend to be issued to clients with
concurrent misstatements. As such, their study tests how differences in PCAOB inspections findings between audit
firms affect 404(b) audits. Thus, while informative, it differs from our study in that it examines the effects of
PCAOB inspections while our study focuses on the effects of 404(b) audits.

Electronic copy available at: https://ssrn.com/abstract=3420787


of reasons including the lack of a “control group” of unaffected companies in the United States

(U.S.). Similar to Iliev (2010), we overcome the endogeneity concerns caused by non-random

compliance by using hand-collected public float data to implement a “fuzzy” RDD that uses

regulatory established cutoffs as instruments for compliance. Theoretically speaking, this

approach is effective because “the exact cutoff is not related to firm fundamentals” (Iliev 2010,

p. 1165). We focus on a relatively narrow bandwidth of public float values surrounding the

regulatory cutoff because companies immediately on either side of the cutoff are largely similar

except for their Section 404(b) compliance status.

We evaluate outcomes associated with internal controls and financial reporting because

Section 404(b) focuses on the audit of internal controls. 2 We begin by examining audit fees, the

most direct cost of complying with Section 404(b). In a pooled sample of observations from

2004 to 2015, we estimate a 64 percent audit fee premium associated with 404(b) compliance for

companies near the $75 million regulatory cutoff. When considering the evolution in this effect

over time, we find that the premium declined from a high of more than 92 percent in 2004 to a

low of approximately 18 percent in 2010. After 2011, we observe a resurgence in the premium,

with estimated premiums in 2015 reaching a level similar to early SOX levels (84 percent). The

timing of the shift is consistent with auditors devoting more effort to ICFR audits following the

PCAOB’s renewed focus on ICFR audits beginning in 2010 (DeFond and Lennox 2017).

We also expect Section 404(b) compliance is likely to influence who companies choose

to provide their audit. If, as it is generally suggested, there are significant complexities associated

with an ICFR audit, we expect that compliant companies are more likely to require the expertise

2
We recognize that an ICFR audit could indirectly affect other outcomes (e.g., operational efficiency); however, the
focus of this study is primarily on those effects most directly tied to internal controls, financial reporting, and the
financial statement audit.

Electronic copy available at: https://ssrn.com/abstract=3420787


and resources of a Big 4 auditor. These complexities may also lead to many non-Big 4 auditors

being unwilling or unable to perform ICFR audits. Consistent with this, we find that Section

404(b) compliant issuers are approximately 13 percent more likely to enlist a Big 4 auditor

during the sample period. Over the sample, we observe a general increase in the effects of

Section 404(b) compliance on the likelihood that a company selects a Big 4 auditor with the

effect reaching approximately 32 percent in 2015. Because Big 4 auditors charge higher fees, this

finding suggests that a portion of the 404(b)-audit fee premium is driven by auditor choice,

especially in more recent years. We confirm this is the case using mediation analysis.

Next, we consider the effects of 404(b) compliance on internal control reporting by

leveraging the extension of 404(a) to non-accelerated filers in 2007. Since 2007, both accelerated

and non-accelerated filers must report on ICFR effectiveness; however, only accelerated filers

also require an audit over ICFR. We find that companies complying with 404(b) are slightly less

likely to disclose internal control deficiencies (ICDs) in management ICFR reports than exempt

companies, suggesting that 404(b) may improve the quality of ICFR. 3 However, this effect is

confined to the initial years following the extension of management internal control reporting to

non-accelerated filers. Furthermore, despite ICFR auditor oversight for 404(b) compliers, we do

not find evidence that ICDs for compliers are more predictive of restatements. Specifically, we

find that ICDs predict restatements for 404(b) compliant and exempt companies, but do not find

that the relationship is stronger for compliers. 4 It is possible that this lack of difference is a result

3
For example, if a 404(b) audit identified material deficiencies in ICFR during interim control testing, then
compliant companies could avoid receiving a material weakness by remediating the deficiencies before year end.
4
Following the approach in DeFond and Lennox (2017), we define an ICFR report to be “more informative” when
a restated year received an ICD on the initial financial statements (i.e., ICD positively predicts restatements). We
discuss the potential limitations with this approach in Section 3. We do not use the stock market reaction to ICD
disclosures as our test of informativeness because ICDs are commonly disclosed in companies’ 10-K filings and the
measurement of returns is likely to be confounded by other information disclosed at the same time as the ICD.
Furthermore, ICDs are often preempted by prior 404 reports and/or quarterly SOX 302 disclosures, further
complicating market reaction tests as a measure of informativeness.

Electronic copy available at: https://ssrn.com/abstract=3420787


of improved ICFR for among compliers. If so, we would expect 404(b) compliers to have more

reliable financial statements. However, consistent with the findings of both Fan, Li, and

Raghunandan (2017) and Bhaskar, Schroeder, and Shepardson (2019), our analyses also do not

yield statistically significant evidence indicating that ICFR audits improve financial reporting

quality in the current regulatory environment. Together, while we find some evidence that an

ICFR audit decreased the prevalence of ICDs in the early years of SOX, we find no evidence

suggesting that ICFR audits improve the reliability of ICFR reports or leads to more reliable

financial statements.

As discussed above, the primary focus of this paper is how Section 404(b) impacts the

audit and financial reporting processes. However, one observable internal cost that relates to

these processes is CFO compensation. Therefore, we use hand collected data on CFO

compensation and estimate the effect of 404(b) compliance on CFO compensation. Consistent

with increased responsibilities related to Section 404(b) compliance, we find evidence that the

compensation of CFOs of compliant companies is significantly higher than that of exempt

companies, particularly in recent years.

Our study provides several important contributions to the extant literature. First, we

provide causal estimates of the effects of Section 404(b) on audit costs using RDD beyond the

first year documented in Iliev (2010). Second, we show that the effects on cost are, in part,

driven by the effects of 404(b) on auditor choice. This suggests that Section 404(b) also

influenced the structure of the audit market and that the increased audit costs associated with

compliance are in part driven by differences in auditor choice. Third, we are the first study to

provide evidence on the effects of an ICFR audit on the quality of ICFR as well as the

informativeness of management internal control reports. Finally, although not a primary focus of

Electronic copy available at: https://ssrn.com/abstract=3420787


our paper, we are the first study to provide archival evidence on some of the internal costs

associated with Section 404(b) compliance.

Our study also informs the current regulatory debate over the Section 404(b) compliance

threshold. In particular, we directly address the calls of 48 accounting and law professors for

research updating the findings of Iliev (2010) on the effects of 404(b) in the years following the

inception of SOX (Honigsberg and Rajgopal 2019). Specifically, we provide updated evidence

on the extent of float manipulation among public companies to avoid 404(b) compliance over

time, as well as the effects of Section 404(b) compliance on audit fees. We also provide new

evidence on Section 404(b) compliance by evaluating the effects of 404(b) on the quality of

internal controls and the informativeness of ICDs.

Finally, our findings are particularly relevant to the recent extension of the 404(b)

exemption to companies with less than $100 million in annual revenue. Nearly 60 percent of

compliers in our sample have less than $100 million in revenue and would therefore be affected

by the new exemption. When we limit our sample to the most recent years (2011-2015) and

companies with revenue of less than $100 million, we continue to find evidence of significant

audit fee premiums for compliers but again find no evidence of financial reporting benefits.

Taken together, our findings should be of interest to a variety of stakeholders interested in the

effects of Section 404(b) compliance.

2. Background and Theoretical Development

Perhaps the most controversial and expensive provision of SOX, Section 404, was

introduced with the intention of improving financial reporting quality. It requires management to

implement and maintain an adequate control system, to annually evaluate the effectiveness of

internal controls over financial reporting (section 404(a)), and to obtain an independent audit

Electronic copy available at: https://ssrn.com/abstract=3420787


over ICFR (section 404(b)). The SEC recently proposed an amendment that would extend the

exemption from Section 404(b) compliance to companies with float between $75 and $250

million and less than $100 million in revenues (SEC 2019). This proposal has sparked debate

among regulators, lawmakers, companies, audit firms, and academics regarding the relative costs

and benefits of compliance (Barth, Landsman, Schroeder, and Taylor 2019; Brush 2019; PwC

2019; Honigsberg and Rajgopal 2019; Jackson 2019; SEC 2019). Those involved in this debate

have expressed the need for an updated economic analysis of the costs and benefits associated

with Section 404(b) compliance, prior to moving forward with the proposal.

Some have raised concerns about the costs of Section 404 compliance, especially for

smaller companies as they effectively had to create control systems, which often included

creating and staffing entirely new departments to design, implement, and maintain control

systems (Iliev 2010; Coates and Srinivasan 2014). The SEC estimated that the internal costs of

producing Section 404(a) control report would average $91,000, though critics suggest that this

amount significantly underestimates the total costs of compliance (Coates and Srinivisan 2014). 5

The SEC noted that it “did not have adequate information to publish an informed estimate” of the

costs of auditor attestation under 404(b) (Coates and Srinivasan 2014), though auditing and

producing an additional audit report was likely to cost substantially more than $91,000, even for

smaller issuers. Iliev (2010) estimates that audit costs increased by over 100 percent for

compliers in the initial year of Section 404. Although the costs of compliance likely depend on

the size of the company, the quality of the company’s internal controls, and the company and

auditor’s experience with 404(b) compliance, it is widely accepted that all companies

5
In a study performed for the SEC, CRA (2005) estimates that the Fortune 1000 companies spent an average of
$5.9M in the first year of SOX compliance, and a survey by Financial Executives International (FEI) (2005) reports
average one-year audit fee increases of $1.3 million and internal compliance costs of $3.1 million associated with
SOX Section 404. This suggests that initial SEC estimates may not have captured the full cost.

Electronic copy available at: https://ssrn.com/abstract=3420787


experienced significant costs associated with initial SOX compliance (Clay and Kim 2017). 6, 7 It

is less clear, however, whether these costs remained significant.

Proponents of SOX argued that the costs, while initially substantial, would decrease over

time because familiarity and experience with compliance would decrease auditor and client effort

from an implementation to a maintenance level. These arguments suggested that the benefits of

Section 404(b) would outweigh the costs as the number of years of compliance increased (Coates

2007; Alexander et. al 2013). Early survey evidence supported this view because respondents

suggest that total compliance costs dropped between 15 and 40 percent in the first two years of

compliance (Coates 2007). Nonetheless, costs and the related concerns about inefficiencies

remained high.

Regulators, acknowledging the high costs of Section 404(b), implemented several

regulatory changes aimed at reducing the costs of compliance. Auditing Standard 5 (AS5)

(PCAOB 2007), effective beginning in 2007, reduced the auditor requirements by allowing

auditors (and management) to test controls using a risk-based approach (PCAOB 2007; Doogar,

Sivadasan, and Solomon 2010). This meant that auditors could tailor the ICFR audit to the

complexity of the company, rotate testing of certain controls, and rely on the work of the

company. These revisions sought to reduce the work required by both the auditor and

6
We note that there are also many indirect costs and benefits associated with SOX compliance that are beyond the
scope of this study. These include decreased investment (Albuquerque and Zhu 2019), increases in real earnings
management (Cohen, Dey, and Lys 2008), and increases in de-registrations (Coates and Srinivasan 2014; Leuz,
Triantis, and Wang 2008). See Coates and Srinivasan (2014) for a broader discussion of the indirect costs (and
benefits) associated with SOX compliance.
7
In the initial years of 404(b) compliance, many companies reported internal control deficiencies (ICDs), yet
auditors were still required to opine on ICFR even though the presence of an ICD necessitates that the auditor take a
substantive audit approach instead of a control reliance approach. That is, auditors were still required to perform
testing of internal control effectiveness, despite the knowledge that ICFR testing would not lead to decreases in
substantive audit work. As such, high deficiency levels in initial years of compliance could further contribute to
initially high audit costs that may diminish as companies improve ICFR. For these reasons, the costs of an ICFR
audit could decline over time.

Electronic copy available at: https://ssrn.com/abstract=3420787


management in evaluating the effectiveness of ICFR and were projected to reduce the cost of

compliance by nearly 25 percent (SEC 2009; GAO 2013; PwC 2017). 8 Costs may also decrease

over time if auditors and companies become complacent, or otherwise lax in their focus on

internal controls. For example, concerns associated with notable accounting scandals starting

with Enron (e.g., Worldcom) and regulatory attention to financial reporting quality could have

caused issuers and auditors to make ICFR a priority in the early years following SOX. However,

companies and their auditors may have experienced high costs and/or perceived minimal benefits

after the initial years and become less attentive to controls over time.

Although efficiency gains and regulatory changes may have reduced the cost of Section

404(b), auditors continue to face high levels of scrutiny from both regulators and investors over

the quality of 404(b) attestations. In fact, the PCAOB renewed its focus on evaluating ICFR

attestations in 2010 (DeFond and Lennox 2017) and continues to highlight deficiencies in auditor

testing of ICFR (e.g., PCAOB 2017). This regulatory focus on ICFR audits may increase

pressure on auditors to reduce inspection deficiencies, and as a result, drive increases in audit

effort for ICFR audits, as documented by DeFond and Lennox (2017). While these factors may

influence the magnitude of the 404(b) audit fee premium, we expect some baseline level of

premium throughout the sample period because auditors must perform a substantial amount of

work to support an opinion on the effectiveness of ICFR. 9

Protecting investors by improving financial reporting is an expressed goal of SOX. More

specifically, Section 404 was intended to improve in ICFRs through its requirements on

8
Regulators also implicitly acknowledged the burden of Section 404(b) in the Dodd-Frank Wall Street Reform and
Consumer Protection Act of 2010 by granting permanent exemption from 404(b) for non-accelerated filers as well
as granting a deferral of compliance with 404(b) to qualifying “emerging growth companies” following the IPO (up
to five-years). While these changes do not affect the costs of compliance, they reflect regulators’ concerns about the
burden of 404(b) for smaller issuers, which our paper directly addresses.
9
Since the PCAOB’s focus is on auditors’ attestation of ICFR generally, we expect companies will experience
higher costs regardless of whether ICFR is effective or not.

Electronic copy available at: https://ssrn.com/abstract=3420787


management (404(a)) and the auditor (404(b)). Proponents of Section 404(b) argue that ICFR

audits improve ICFR by helping companies more effectively identify and remediate ICDs. In

turn, this reduction in deficiencies should lead to more reliable financial reporting. While these

reporting benefits are unlikely to manifest immediately as companies and their auditors become

acclimated with the requirements of 404(b), they should accumulate over time if an ICFR audit is

an effective means of improving financial reporting. Given the direct focus on financial reporting

controls, one would expect relatively unambiguous evidence of higher financial reporting quality

for companies complying with 404(b). Although Iliev (2010) documents lower accruals in the

initial year of 404(b) compliance, research using more recent periods finds no difference in

reporting quality between compliers and non-compliers. For example, Fan et al. (2017) find that

while 404(b) compliers had higher quality financial reporting initially, this difference

disappeared after the 2007 extension of 404(a) to non-accelerated filers, suggesting that 404(a) is

a viable alternative to 404(b). Even more surprising, for a sample of companies from 2007-2013,

Bhaskar et al. (2019) document lower financial reporting quality for compliant companies and

conclude that integrated (i.e., ICFR) audits drive this effect.

Although the preceding discussion highlights several reasons why the effects of Section

404(b) compliance may have changed over time, there is limited empirical evidence examining

this question. 10, 11 Before the SEC voted to extend the exemption from 404(b) to companies

10
Ge, Koester, and McVay (2017) evaluate the merits of exemption based on predicted internal control
misreporting. Our study differs from Ge et al. (2017) in many respects. First, we do not estimate misreporting, rather
we use the cutoff of $75 million as an instrument to estimate the differences in outcomes between companies above
and below the cutoff. This research design allows us to make strong causal inferences about the costs and benefits of
Section 404(b) compliance over time. Second, they focus on capital market consequences, while we focus on the
direct costs and financial reporting benefits associated with compliance. Finally, we consider the evolution of these
costs and benefits over time, whereas they primarily focus on pooled costs and benefits.
11
Kinney and Shepardson (2011) provide some evidence that audit fee changes for first time Section 404(b)
compliers between 2004 and 2007 were lower in 2007 than in 2004. However, the purpose of their study was not to
consider how the costs of 404(b) have changed over time, but to draw conclusions about the overall effectiveness of

10

Electronic copy available at: https://ssrn.com/abstract=3420787


under $100 million in revenue, academics called for studies to provide evidence regarding

changes to 404(b) compliance thresholds (Honigsberg and Rajgopal 2019). In the tests that

follow, we answer this call by using RDD to estimate the costs and benefits of Section 404(b)

compliance for companies similar to those affected by the rule. We begin by examining costs in

the form of audit fees because Section 404(b) compliance directly affects the volume of audit

work performed by mandating that auditors test the effectiveness of ICFR. Relatedly, given that

ICFR audits significantly increase the complexity of the audit, we also examine how Section

404(b) compliance affects auditor selection. Finally, we consider whether compliance improves

ICFR quality or results in more informative internal control reports. We estimate each of the

effects of Section 404(b) compliance in a pooled sample and over time.

3. Sample Selection and Research Methodology

3.1 Sample Selection and Descriptive Statistics

Our sample consists of U.S. incorporated companies with fiscal year ends from 2004

through 2015 (i.e., December 2004 through November 2016). 12 We obtain auditor, audit fee,

restatement, and material weakness data from Audit Analytics (AA) and financial data from

Compustat. We require observations to have all data for control variables. In our primary tests,

we adopt a $100 million bandwidth for public float around the $75 million cutoff (i.e., public

float between $25 and $125 million) which results in a sample of 11,013 observations. 13

We present full sample descriptive statistics in Panel A of Table 1. Consistent with our

SOX provisions by comparing companies changing across regimes. Furthermore, their study only informs the costs
of initial compliance, not how the costs of compliance may have changed over time.
12
Because the extension of Section 404(a) to non-accelerated filers occurred on December 15, 2007, we use a
December 1 to November 30 fiscal year for our year fixed effects and yearly regressions to avoid any fiscal year
overlapping two regulatory regimes. Auditing Standard 5 was effective beginning in November of 2007. Because
very few companies have a November fiscal year end, we do not adjust our fiscal years for this event. However, if
we do so, inferences are unchanged.
13
We drop companies within this bandwidth that have less than $1 million in assets or over $500 million in market
capitalization. We also consider additional bandwidths in Section 5.3.2.

11

Electronic copy available at: https://ssrn.com/abstract=3420787


identification strategy that utilizes companies around the compliance threshold, the average

company in the sample is small with negative return on assets (ROA) and approximately 45

percent of sample companies comply with Section 404(b). In Panel B, we bifurcate the sample

on our primary instrument, Float > $75 Million. Consistent with the instrument being a strong

predictor of compliance, compliance rates are much higher above the threshold (91 percent) than

below the threshold (23 percent). 14 Because public float is inherently related to market

capitalization, companies above $75 million in float are larger than companies below $75

million. While we find that companies above $75 million in public float are less likely to

disclose a material weakness (Weak 404(a)), they do not differ in the likelihood that they have a

material restatement (Restate 8K), though the difference is only marginally significant. We also

note higher Audit Fees for companies complying with Section 404(b).

3.2 Identification Strategy and Methodology

The non-random nature of Section 404(b) compliance compromises identification

strategies that rely on observed SOX compliance status. For example, companies may voluntarily

comply, be exempt, or otherwise be non-compliant. For this reason, actual compliance is likely

endogenous to the outcomes we examine. As such, we follow Iliev (2010) and adopt an

instrumental variables approach that takes advantage of the $75 million public float threshold for

compliance set by the SEC. Companies above $75 million in public float at the end of the most

recent second quarter generally must comply with Section 404, while companies immediately

below the threshold are generally exempted. 15 We utilize two-stage least squares regression

14
We do not expect to find 100 percent compliance above the $75 million threshold because some companies over
the threshold are exempt from compliance (e.g., emerging growth companies).
15
In the initial years of SOX, the exemption related to 404(a) (management reports) and 404(b) (auditor’s reports).
Since 2007, even companies below the $75 million cutoff are required to comply with 404(a), though they continue
to remain exempt from 404(b). As such, results from tests before 2007 reflect the effects of 404(a) and 404(b). All
analysis since 2007 reflects just estimated differences associated with 404(b).

12

Electronic copy available at: https://ssrn.com/abstract=3420787


(2SLS) and implement a RDD approach. We assume that companies immediately on either side

of the cutoff are similar and that the company’s float relative to the cutoff affects a company only

through its effect on 404(b) compliance. Therefore, we use the company’s float relative to the

cutoff as an instrument for Section 404 compliance and focus on a narrow bandwidth of float

surrounding this cutoff. 16 We augment the Iliev (2010) approach by adding a second instrument,

$50 million in float, to our 2SLS tests. Companies that were Section 404(b) compliant in the

prior year remain compliant unless the float falls below $50 million in the most recent second

quarter. 17 In effect, the $75 million cutoff applies to companies that were exempt in the previous

year, and the $50 million cutoff applies to companies that were compliant. 18

We begin by using ordinary least squares (OLS) to estimate the first stage regression:

404(b) Compliant = β0 + β1Float75 + β2Float50 + βXControls + ε [1]

where 404(b) Compliant is an indicator for whether the company received an auditor’s 404(b)

internal controls opinion (as identified by Audit Analytics). Float75 (Float50) is an indicator

equal to one if the float at the end of the most recent second quarter exceeds $75 ($50) million.

In general, with RDD and a sufficiently tight bandwidth, controls are unnecessary for unbiased

estimation of the treatment effect. However, because float is largely driven by company size, we

control for size related variables (e.g., Log Assets, Log Market) since company size strongly

relates to some of the dependent variables we explore (e.g., audit fees) and thus may confound

estimates, even in tight bandwidths. Following Iliev (2010), we also include control variables for

16
As noted in Iliev (2010) and Gow, Larker, and Reiss (2016), widespread float manipulation could impair the
effectiveness of this approach. We discuss and assess this possibility in Section 3.3.4.
17
See “Revisions to Accelerated Filer Definition and Accelerated Deadlines for Filing Periodic Reports,” available
at https://sec.gov/rules/final/33-8644.pdf.
18
Because companies with a public float of greater than $75 million are sometimes exempt from compliance and
companies with floats below $75 million may still be required to comply, a straightforward approach using Float75
as an instrument will produce estimates that understate the economic magnitude of the effect, so we use a 2SLS
approach. An added benefit of the 2SLS approach is that we can use multiple instruments. We use the Stata
command “ivreg” to appropriately adjust standard errors for two stage least squares estimates.

13

Electronic copy available at: https://ssrn.com/abstract=3420787


other company characteristics (Sales, Leverage, Receivables, Segments). 19 Finally, we control for

industry fixed effects using Fama-French 12 classifications (and include year fixed effects in all

pooled analyses). All variables are defined in Appendix I.

We then use OLS to estimate the following second stage regression:


DV = β0 + β1404(𝑏𝑏)𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 + βXControls + ε [2]

where the dependent variable, DV, represents either audit fees, auditor choice, or financial

reporting outcomes. We first focus on audit fees, which we operationalize as the natural log of

total audit fees (Audit Fees), because Section 404(b) compliance directly affects the volume of

audit work performed by mandating that auditors test the effectiveness of ICFR. While we


generally expect a positive coefficient on 404(𝑏𝑏)𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 , the magnitude of this effect, and,

importantly, how it has changed across time is unclear ex ante. We also examine how Section

404(b) compliance impacted auditor choice. Since Section 404(b) audits are more complex and

require additional resources and expertise, compliers could be more likely to engage a Big 4

auditor. To examine this, we utilize Big 4, which is an indicator variable set to one when the

company engages a Big 4 auditor, zero otherwise.

For tests of the reporting benefits of 404(b) compliance, we replace the dependent

variable with one of two financial reporting variables representing internal control/reporting

quality: 1) Weak404(a), which is an indicator variable equal to one if the company discloses at

19
We note that Iliev (2010) includes linear, squared, and cubed values of float as additional instruments (i.e.,
included in first stage but not included in second stage – see Table II, Column 4). However, for simplicity, we just
include the two regulatory thresholds as the other functional forms of float should not relate to compliance.
Consistent with this, we note that the inferences throughout this study are unchanged if we follow this approach.
Furthermore, in RDD, researchers often include the “running variable” (i.e., float) as a control in the second stage
(i.e., not an instrument). We do not include the float as a control because float is a size variable adjusted for
ownership structure and it is unclear why this variable would affect the outcomes we evaluate, except through
compliance. Consistent with this expectation, the running variable is nearly always accompanied by an insignificant
coefficient if included in the second stage. Instead, we include size related controls that are not affected by
ownership structure (Log Assets, Log Market, Sales).

14

Electronic copy available at: https://ssrn.com/abstract=3420787


least one 404(a) weakness, zero otherwise, and 2) Restate 8K, which is an indicator variable

equal to one if the client subsequently restated the year’s audited financial statements with Form

8-K Item 4.02 (i.e., non-reliance restatement), zero otherwise. If an ICFR audit improves ICFR


or reporting quality, we expect a negative coefficient estimate on 404(𝑏𝑏)𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 . Controls

in each specification is the same set of controls from the first stage regression.

3.3 Instrument Collection and Validity

3.3.1 Public Float Measurement

Companies calculate float as the aggregate market value of common equity held by non-

affiliates as of the most recently completed second fiscal quarter. Although SEC rules require

companies to disclose float in their 10-K filings, we hand collect the data because float is not

readily available in existing databases. To do so, we accessed companies' 10-K filings on

EDGAR and utilized a Python script to scrape the public float values. We first scraped float

directly from XBRL tags in the financial statements (tagged with dei:EntityPublicFloat). If

public float was not XBRL tagged, we utilized regular expressions and scraped float from the

header of the 10-K. For float values scraped from the 10-K headers, we manually checked the

consistency of the formatting and the rounding of the numbers (e.g., millions, billions).

3.3.2 Validity of the Instruments: Instrument Exclusion

A valid instrument only affects the outcomes through its impact on treatment; an

assumption that can largely be justified only with theory. There exists no theoretical reason to

believe that a discontinuity in the outcomes we evaluate would exist at $75 (or $50) million for

reasons other than 404(b) compliance. Although Section 404 was not in effect until 2004,

companies disclosed public float in 2003 due to accelerated filer rules. This allows us to perform

placebo tests using the cutoff in the year prior to compliance where we examine whether

15

Electronic copy available at: https://ssrn.com/abstract=3420787


differences in outcomes existed prior to treatment. If no relation exists between the cutoffs and

the primary outcomes in our study in 2003, then we can more confidently attribute post-SOX

2SLS estimates to the effects of SOX 404 compliance. Therefore, we regress Audit Fees, Big4,

and Restate 8K on Float75 (Panel A) and Float50 (Panel B) for 2003 – for comparison we also

perform the analysis for 2004 after the inception of Section 404, where we should observe a

discontinuity if there is an effect of Section 404 on the outcomes.

We present the results of these tests in Table 2 Panel A (B) and note that the coefficient

on Float75 (Float50) is insignificant in columns 1, 3, and 5. Conversely, the coefficient on the

instruments after Section 404(b) was implemented (columns 2, 4, and 6) are statistically

significant in four of six specifications, which is generally consistent with Section 404(b) having

an impact in the initial year. In particular, we note an effect in all three specifications with

Float75. These results provide initial evidence that Section 404(b) increased audit fees, increased

the likelihood that a company would select a Big 4 auditor, and decreased material restatement

rates in the initial year of Section 404(b). 20 While we cannot accept the null of valid exclusion,

these tests provide some support in addition to the theoretical exclusion noted above.

3.3.3 Validity of the Instruments: Instrument Strength

Unlike exclusion, we can directly test the strength of the instrument. We test this by

estimating equation [1] and present the results in Table 3. We estimate positive and highly

significant coefficients on both Float50 and Float75. For brevity, we only tabulate the pooled

regression. In untabulated analysis, we estimate equation [1] by year and find positive and highly

significant coefficients on both instruments in all years (t-stat > 6.50 in each year).

20
It is not surprising that the effects do not reveal themselves with Float50 regressions since this instrument should
only play a role for companies that were compliant in the prior year, and 2004 was the first year of compliance.

16

Electronic copy available at: https://ssrn.com/abstract=3420787


3.3.4 Validity of the Instruments: Estimating the Extent of Float Manipulation

As noted by Gow et al. (2016) relating to RDD, “manipulation of running variable may

occur and researchers should carefully examine… this possibility.” 21 Manipulation of public

float to avoid compliance could compromise our identification strategy if two conditions jointly

exist: 1) companies manipulate public float to avoid compliance and 2) companies that

manipulate their float differ systematically from companies that do not in terms of the outcomes

we evaluate. If float manipulation is either not widespread or unrelated to the outcomes we

investigate, our approach will be effective.

Similar to Iliev (2010), we evaluate whether “bunching” in the distribution of public

floats exists just below $75 million (the stronger of the two instruments). We do so by comparing

the distribution of public float to two other benchmark distributions that should be similar except

they should not exhibit bunching: market value of equity from 2004-2015 and public float in

2003. Companies have no incentive to manipulate market capitalization around a $75 million

threshold and we do not expect companies to manipulate their 2003 float because 404(b) did not

go into effect until 2004 (Iliev 2010) also does not find evidence of float manipulation in 2003).

If a greater proportion of companies exist just below $75 million in float (relative to our

benchmarks), it would suggest some manipulation of float to avoid compliance.

We first display histograms of public float in 2003, market value of equity from 2004-

2015 and public float from 2004-2015 in Panel A of Table 4 to graphically illustrate the

distributions. With both market cap and float, we observe a general decline in representation as

21
Manipulating public float to fall below the $75 million threshold does not violate SEC rules. Legal approaches for
reducing public float below the $75 million threshold include repurchasing shares, paying dividends, or increasing
insider or large shareholder holdings. However, many of the small companies in this sample do not pay dividends or
repurchase shares. In addition, corporate insiders (e.g., executives) likely have much of their wealth tied up in the
stock of the company and do not have the liquidity to purchase enough shares to successfully manipulate float.

17

Electronic copy available at: https://ssrn.com/abstract=3420787


the variable increases across the range – though we observe no strong bunching around the

threshold for 2003 float or market value of equity (2003 float is less smooth due to much smaller

sample size). However, we observe a small but noticeable discontinuity at $75 million in 2004-

2015 public float. To estimate the extent of possible float manipulation, we compare the

distribution of company-years in the $20 million range around the threshold to the distribution

for the two benchmark distributions. 22 We tabulate the breakdown in Panel B of Table 4. While

57.9 percent of company-years from 2004-2015 between $65 and $85 million in float fall below

$75 million, our baseline groups suggest that the representation would be between 53.5 percent

(market value of equity) and 55.1 percent (2003 float) absent manipulation incentives. Thus,

compared to the 1,129 observations we observe in this range, we estimate that between 1,043

(0.535 * 1,949) and 1,074 (0.551 * 1,949) of the 1,949 observations would fall below $75

million in public float absent manipulation. Thus, we roughly estimate that between 55 and 86

company-years in our sample successfully manipulated float to avoid compliance from 2004-

2015 (between 4.6 to 7.2 observations per year). This finding does not suggest widespread float

manipulation, so we do not expect the RDD approach to be compromised.

We next consider whether the distribution changes over time. Panel C of Table 4 displays

the percentage of company-years between $65 and $85 million in float that falls below $75

million in float by year. We also display the same distribution for market capitalization for

reference. While the percentages predictably fluctuate given the small sample sizes in this

narrow range, we observe no discernible trend in float distribution, and this is similar to market

capitalization suggesting that manipulation does not appear to change significantly over time.

22
We focus on this narrow range because we expect that companies that fall outside of this range are unlikely to be
potential float manipulators, which would dilute the strength of our tests. However, inferences are similar if we
perform these tests on a broader range of floats (e.g., the $100 million bandwidth used in this study).

18

Electronic copy available at: https://ssrn.com/abstract=3420787


4. Empirical Findings

4.1 The Effect of Section 404(b) on Audit Costs

4.1.1 Audit Fees

In Panel A of Table 5, we estimate equation [2] using Audit Fees as the dependent


variable. The coefficient on 404(𝑏𝑏)𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 in column 1 suggests that Section 404(b)

compliance increases audit fees by 64 percent (e0.496 – 1) for the pooled sample. The remaining

columns provide year-by-year estimates of the effect of Section 404(b) compliance on audit fees.

Consistent with first year “growing pains,” we estimate the largest fee premium for 404(b)

compliance in the first year (92 percent = e0.650 - 1). Other than a small increase in 2007, the

estimated premium consistently declines from 92 percent in 2004 down to a low of 18 percent

(e0.169 – 1) in 2010 (with substantial declines from 2008 to 2010 following the implementation of

AS5). However, the trend in estimated fee premiums reverses in 2011 – reaching a premium of

84 percent (e0.609 – 1) in 2015. The timing of this reversal coincides with increased effort on

404(b) audits following an increased focus on ICFR audits by the PCAOB beginning in 2010. 23

4.1.2 Auditor Choice

Our audit fees analysis differs from Iliev (2010) in that we estimate the combined direct

(i.e., the audit process) and indirect (i.e., auditor choice) effects of Section 404(b) compliance on

audit fees. We do so by intentionally excluding a Big 4 control from our tests because we expect

that the complexities associated with an ICFR audit increase the likelihood that compliant

23
AS5 was implemented in 2007, in part, to reduce costs related to the external audit. While we estimate an increase
in fee premiums in the first year following AS5 (likely as a result of auditors evaluating and implementing audit
approach changes due to AS5), we do find a steady and significant decline over the next three years to a minimum
of 2010. However, the subsequent rebound in the estimated fee premium, despite the continued existence of AS5
suggests that the decline may not be entirely attributable to AS5. Auditor leniency on controls testing could be
responsible for much of the decline since premiums rebounded following increased PCAOB scrutiny in 2010.
Nonetheless, the premium associated with 404(b) could be even greater if AS5 had not been passed.

19

Electronic copy available at: https://ssrn.com/abstract=3420787


companies require the services of a Big 4 auditor. Thus, hiring a Big 4 auditor is, in part, an

outcome of Section 404(b) compliance and a mechanism by which compliance rules influence

audit fees. If the effect of 404(b) on the likelihood that a company selects a Big 4 auditors has

changed over time, this could be the driver some of the observed increase in audit costs.

To isolate the direct effect on the audit process, we re-perform the tests in Panel A after

including a Big 4 control to estimate the effects of 404(b) compliance on audit fees conditional

on the auditor chosen. We display these results in Table 5, Panel B. The estimated premium

follows a pattern similar to Panel A (i.e., declining until 2010, then larger afterward), though the

premiums are lower than Panel A, consistent with 404(b) compliers being more likely to hire a

Big 4 auditor and Big 4 auditors charging higher fees. Furthermore, consistent with auditor

selection driving a larger portion of the premium in more recent years, the difference in estimates

in Panel A and B are greater in recent years. Because the majority of large companies would

enlist Big 4 auditors regardless of 404(b) compliance, the estimates in Panel B (e.g., estimated

fee premium of 47 percent in 2015 in Panel B) may generalize more to those companies (i.e.,

those above the top end of the bandwidth).

To further consider the indirect effect, we examine whether the effect of Section 404(b)

compliance on Big 4 auditor selection has also increased over time. To do so, we replace Audit

Fees with a Big 4 indicator variable (Big 4) and re-perform our tests. We display these results in

Table 5, Panel C. As shown in the first column, Section 404(b) compliant issuers are

approximately 13 percent more likely to enlist a Big 4 auditor during the sample period. We

observe a general increase in the estimated effect of Section 404(b) compliance on the likelihood

of selecting a Big 4 auditor with the effect reaching approximately 32 percent in 2015. This

suggests that over time, effect of 404(b) on the likelihood of selecting a Big 4 auditor has

20

Electronic copy available at: https://ssrn.com/abstract=3420787


increased, supporting prior results suggesting that a larger portion of the 404(b) audit fee

premium is driven by an increased “Big 4” selection effect in recent years. 24

4.2 The Effect of Section 404(b) on Internal Controls and Financial Reporting Quality

4.2.1 Internal Control Effectiveness

An ICFR audit could improve internal control effectiveness if it helps clients identify and

remediate ICDs. We test this possibility by comparing ICD rates in 404(a) management reports

between Section 404(b) compliers and exempt companies. Since non-accelerated filers were not

required to comply with 404(a) until 2007, we limit the sample for this test to 2007 and later

observations. We present univariate statistics in Table 6, Panel A and the accompanying figure in

Panel B. Deficiency rates for exempt companies exceed deficiency rates for compliant

companies in all years (statistically significant in all but two years). However, compliers and

exempt companies both exhibit a fairly consistent decline in ICDs until 2010 followed by a

relatively consistent increase through 2014.

Because company size is a significant driver of ICFR quality, univariate differences

potentially overstate the 404(b) effect, even within a relatively narrow bandwidth. Therefore, we

present RDD estimates in Panel C of Table 6 with Weak 404(a) as the dependent variable. The

pooled regression estimate (column 1) suggests that 404(b) compliers are 2.6 percent less likely

to disclose ICDs. This evidence suggests that 404(b) led to more effective internal controls on

average from 2007 to 2015. The year-by-year tests in columns 2 through 10 do not suggest a

strong trend across time, though if we perform subsample analyses (untabulated) and limit the

pooled sample to observations between 2007-2010, we observe a negative (-0.050) and

24
We note that the percentage of companies in our bandwidth (i.e., smaller companies) selecting Big 4 auditors has
declined over time for both compliant and exempt companies. However, it has declined to a greater extent for
exempt companies than compliers.

21

Electronic copy available at: https://ssrn.com/abstract=3420787



significant (t-stat = -2.47) coefficient on 404(𝑏𝑏)𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 . In a pooled subsample of

observations between 2011-2015, the coefficient is approximately zero (-0.001, t-stat = -0.05).

This finding is consistent with auditors identifying more ICDs during ICFR audits in recent years

(DeFond and Lennox 2017). However, exempt and compliant companies exhibit similar

univariate trends overall (i.e., steady decline in ICD rate from 2007 to 2010, followed by

increase in ICD rate). The similarity in trends irrespective of 404(b) compliance suggests that the

trend in disclosed ICDs for compliers may be partly driven by ICD detections during financial

statement audits or by market-wide trends in control effectiveness rather than just ICD detections

during the ICFR audit. If the trend for compliers were driven solely by changes in the ICFR audit

over time, we should observe a significantly different trend for exempt companies. Thus, while

ICD rates increase beginning in 2011 for compliant companies, a similar trend for companies not

receiving ICFR audits suggests that changes in ICFR audits in response to PCAOB scrutiny are

likely not the only driver of this increase. 25

4.2.2 Internal Control Deficiency Informativeness

If an ICFR audit helps issuers identify a higher proportion of existing ICDs or prevents

minor control deficiencies from being misclassified as material weaknesses, then internal control

reports should be higher quality for compliers (i.e., fewer false positives and false negatives). 26 If

so, then compliers disclosing ICDs should be more likely to be those that “genuinely deserve

(adverse opinions)” (DeFond and Lennox 2017, p. 617) and therefore, ICDs for compliers should

be more predictive of financial reporting failures than for exempt companies. In untabulated

25
Increased auditor attention to internal controls may have increased the likelihood that auditors require
management to disclose identified ICDs even for issuers not complying with Section 404(b). Thus, auditor
procedures during a financial statement audit may have aided in detecting ICDs even when the auditor does not
perform an ICFR audit.
26
If this is the case, the findings in the previous section understate the extent to which 404(b) results in more
effective internal controls.

22

Electronic copy available at: https://ssrn.com/abstract=3420787


analysis, we find that 33.9 percent of exempt company-years and 23.2 percent of compliant

company-years that eventually restate via a non-reliance filing receive ICDs. This higher rate for

exempt companies is inconsistent with ICFRs improving the extent to which ICDs predict

restatements. To assess this in a multiple regression setting, we follow the approach in DeFond

and Lennox (2017) and regress Restate 8k on Weak 404(a); we do so separately for compliers

and exempt companies (Table 7 Panels A and B). Due to the endogenous nature of compliance,

we also present results split based on Float75 (Table 7 Panels C and D). Following the

reasoning of DeFond and Lennox (2017), if compliers have more informative internal control

reports, we expect a more positive coefficient on Weak 404(a) for compliers. 27

As shown in column 1 of each panel in Table 7, adverse 404(a) opinions predict Restate

8K in each sample. Estimates suggest that companies with ICDs are 5.0 (3.9) percent more likely

for exempt (Float75=0) companies and 3.8 (5.7) percent more likely for compliant (Float75=1)

companies to subsequently restate that year’s financial statements than companies without an

ICD. The difference is statistically insignificant (χ2 < 0.50 in both comparisons), providing no

evidence supporting the idea that 404(b) compliance improves internal control report

informativeness. 28 In the yearly regressions (columns 2 through 10) we note no clear trend in the

informativeness of ICDs across time. Similar to the DeFond and Lennox (2017) findings, we find

no evidence that ICDs are less predictive of restatements in recent years for 404(b) compliers

27
DeFond and Lennox (2017, pg. 615) acknowledge that the disclosure of an ICD “merely means that there is a
‘reasonable possibility’ that a material misstatement will not be prevented or detected on a timely basis.” Like them,
we recognize that there are reasons why ICDs wouldn’t result in misstatements (e.g., auditors perform additional
work to offset risk related to poor control quality).
28
PCAOB AS 2110 – Identifying and Assessing Risks of Material Misstatement, requires auditors to “obtain an
understanding of internal control over financial reporting” (AS 2110.05b). AS 2110.18-.40 describe in detail the
requirements for all public company audits, regardless of 404(b) compliance status, which includes an evaluation
whether ICFR have been designed and implemented (PCAOB 2010). The addition of 404(b) compliance requires
those controls to be evaluated for operating effectiveness. Thus, it is possible that the act of evaluating controls for
design and implementation is similarly effective at identifying ICDs as the tests of operating effectiveness.

23

Electronic copy available at: https://ssrn.com/abstract=3420787


despite their increased prevalence. Interestingly, our results suggest that companies exempt from

ICFR audits experienced a similar phenomenon (i.e., increased ICD rates but no decrease in ICD

informativeness), indicating that this may not be a product of the ICFR audit itself. Specifically,

for exempt filers, the coefficient on Weak 404(a) is positive and significant from 2007-2010

(coefficient = 0.058, t-stat=2.75) and 2011-2015 (coefficient = 0.042, t-stat=2.40); these

coefficients are not significantly different from one another (χ2 = 0.33).

4.2.3 Financial Reporting Quality

Results in the previous section provide some evidence that ICFR audits improve internal

control quality, though the relation is insignificant since 2011. However, if 404(b) helps

compliers remediate the most severe ICDs, we may observe similar informativeness of reported

ICDs despite improved ICFRs. If these conditions explain the previous results, then it indicates

that an ICFR audit improves financial reporting. However, prior research suggests that, at least

since 2007, this does not appear to be the case (Fan et al. 2017; Bhaskar et al. 2019). We confirm

the findings of prior research in our setting by replacing the dependent variable in equation [2]

with Restate 8K. We present univariate statistics for Restate 8K in Table 8, Panel A and the

accompanying figure in Panel B. As shown, both compliant and exempt companies exhibit a

similar trend in restatement rates, suggesting the decline in restatements is unlikely to be solely

attributable to ICFR audits. In Table 8 Panel C, we present RDD estimates and find no evidence

that an ICFR audit decreases the likelihood of a material restatement in the pooled sample (t-stat

= -0.10). Although insignificant in most years, we generally find negative coefficients pre-2007

and positive coefficients beginning in 2007. 29 Thus, consistent with prior research, we find no

29
Consistent with Fan et al. (2017), if we limit the sample to observations before 2007, we document that 404(b)
compliers are 2.8 percent less likely to have a Restate 8k (t-stat= -1.83). Consistent with Bhaskar et al. (2019), if we
limit our sample to 2007-2013, we find that compliers are 2.4 percent more likely to have a material restatement (t-
stat = 2.08). We also document similar evidence if we include the entire 2007-2015 period.

24

Electronic copy available at: https://ssrn.com/abstract=3420787


evidence that an ICFR audit improves financial reporting quality after the extension of 404(a) to

non-accelerated filers, suggesting that the previous results do not appear to be a result of

improved ICFR for among compliers.

4.2.4 Restatement Severity

While the results in the prior section suggest that an ICFR audit does not reduce the

likelihood of misstatements, it is possible that 404(b) limits restatement severity. To test this

possibility, we compare the market reactions at the restatement announcement dates between

companies that received an ICFR audit during the misstated years and companies that did not. 30

If 404(b) reduces misstatement severity, we should observe less negative market reactions

surrounding restatement announcements (indicating less severe misstatements). To perform this

test, we collect all available returns data from CRSP for our sample companies and calculate

three-day cumulative abnormal returns, centered on the date of the restatement announcement.

Consistent with expectations, untabulated analyses indicate that the market reaction surrounding

material restatement announcements for sample companies with returns data is significantly

negative, for both compliers (n=127, mean=-2.14%, t-stat = -1.88), and exempt companies

(n=91, mean=-3.15%, t-stat = -2.96). However, the difference between the means is not

statistically significant (t-stat= -0.62).

4.2.5 Detection of Existing Misstatements

One final consideration is that 404(b) audits improve the likelihood that existing

misstatements are eventually corrected. If so, the earlier results could be the result of lower rates

30
If a restatement announcement related to multiple misstated years including both a year that was not subject to
404(b) and a year that was subject to 404(b), then we eliminated the restatement announcement from the sample
because we could not assign the restatement announcement to a compliance bucket. We also acknowledge that these
tests are subject to limitations. To the extent that investors anticipate a restatement to a greater extent for one group
of companies, the market reaction will be mitigated for those companies.

25

Electronic copy available at: https://ssrn.com/abstract=3420787


of undetected misstatements for compliers. To evaluate this, we examine restatement rates

conditional upon future compliance. There is no reason to believe future compliance relates to

the existence of a misstatement in the current year. Therefore, if an ICFR audit increases the

likelihood of detecting an existing misstatement, we would observe a positive coefficient on

404(b) Compliantt+1. We perform this analysis using 7,323 observations from our sample that are

not 404(b) compliant in year t and regress Restate 8k on 404(b) Compliantt+1 as well as controls

and fixed effects. In untabulated analysis, the coefficient on 404(b) Compliantt+1 is positive, but

insignificant (t-stat = 0.77). Thus, we find no evidence that 404(b) compliance increases the

likelihood that existing misstatements are detected. 31

5. Additional Analyses

5.1 Implications of the Recent Regulatory Changes

Consistent with the PCAOB’s renewed emphasis on internal controls, our results suggest

an inflection point in 2011 across several outcomes. Therefore, we re-perform previous test using

a pooled sample 2011-2015 (December 2011- November 2016 fiscal year ends) to help inform

the effects of Section 404(b) in the current regulatory environment. Untabulated analyses suggest

that Section 404(b) compliance is associated with a 60 percent higher audit fees (t-stat = 9.60), a

22 percent higher likelihood of selecting a Big 4 auditor (t-stat=5.54), and 2.0 percent greater

chance of subsequently restating (i.e., lower reporting quality) financial statements (t-stat =

1.93). We find no significant difference in the likelihood that a company has an ICD (t-stat = -

0.05) and continue to find that ICDs predict subsequent material restatements for compliant and

31
These tests and related inferences are similar to tests performed in Bhaskar et al. (2019) which finds no evidence
that compliance with Section 404(b) in 2004 relates to restatements for the sample companies in 2001-2003,
suggesting that 404(b) does not aid in the detection of existing misstatements. Inferences from the tests are
unchanged we use RDD on a sample of companies with Float75=0 and instrument for Compliant 404(b)t+1 with
Float75t+1 and Float50t+1.

26

Electronic copy available at: https://ssrn.com/abstract=3420787


exempt companies, but the extent does not differ significantly based on compliance (χ2 = 0.19).

Because the SEC’s recent rule will extend the 404(b) exemption to include currently

compliant companies with less than $100 million in annual revenue, we also further limit the

2011-2015 sample to company-years with less than $100 million in revenue and re-perform the

pooled tests. 32 In untabulated analyses, we find that 404(b) compliers incur significantly higher

Audit Fees (57 percent) and a higher likelihood of selecting a Big 4 auditor (17 percent). We also

find no significant difference in ICD rates (t-stat= -0.41) and find significantly higher material

restatement rates (2.8 percent) for compliers. For context, we note that the mean 404(b)

compliant issuer with less than $100 million in revenue has annual audit fees of approximately

$370,000. This suggests that the estimated 57 percent premium would result in an audit fee

reduction of $135,000 each year if exempted. Taken with the fact that approximately half of

these companies have negative ROA and cash flow is therefore critical, this evidence suggests

that the exemption from 404(b) could be meaningful to affected companies.

5.2 Section 404(b) Compliance and CFO Compensation

Given the manager time and other internal costs associated with complying with an ICFR

audit, we expect that the above estimates of audit costs represent a lower bound of the true costs

of compliance. While many internal costs are not easily observable, we expect the position and

job responsibilities of the CFO to be more demanding given the complexities associated with a

Section 404(b) audit. Therefore, we hand collect CFO compensation data from DEF 14A filings

and use RDD to estimate whether 404(b) compliance is associated with higher CFO

compensation (CFO Comp). We present the results in Table 9. Our results suggest 404(b)

32
57.9 percent (844 of 1,457) of 404(b) compliant company-years between 2011 and 2015 in our sample would be
exempted under the proposed rule, suggesting that this sample is highly relevant to the regulatory debate. This is
similar to the percentage of compliant companies in the full sample that have less than $100 million in revenue (57.2
percent).

27

Electronic copy available at: https://ssrn.com/abstract=3420787


compliers pay the CFO approximately 9.0 percent higher salaries. However, in the longitudinal

analyses, we note that effect does not manifest until 2007 and after, consistent with it taking time

for the effects of 404(b) on CFO responsibility to affect salary negotiations. In a pooled analysis

of the recent time period (2011-2015), the estimated premium is 20.6 percent (t-stat = 3.57),

suggesting that 404(b) compliers pay significantly higher CFO salaries than exempt companies,

consistent with increased job responsibilities. While the internal costs associated with complying

with an ICFR audit include far more than CFO salary, this analysis illustrates one small area

where ICFR costs extend beyond the direct costs associated with audit fees.

5.3 Robustness Tests

5.3.1 Changing Composition of the Sample over Time

Prior research documents a decline in IPOs (Rose and Solomon 2016) and an increased

number of companies “going private” (Engel, Hayes, and Wang 2007) following SOX, often

attributing these findings to the costs of SOX compliance. In our sample, we observe substantial

and consistent declines in sample size throughout our period with 2015 having approximately

half the number of companies as 2004 in our primary bandwidth ($25 million to $125 million).

This is consistent with companies often perceiving the costs of public company status (including

SOX compliance) to exceed the benefits of public company status (e.g., access to capital). If the

companies with low costs of compliance disproportionately remain public and, in the bandwidth,

this could potentially lead to the observed year over year declines in the estimated cost of

compliance in the first several years of our sample. The fact that we observe substantial

increases in estimated audit fees of compliance following 2010 while the number of companies

in the bandwidth continues to decline suggests that this is not likely a significant concern. If

sample attrition were driving the initial trend in fee declines, we would expect the declines in

28

Electronic copy available at: https://ssrn.com/abstract=3420787


fees to continue as the number of companies continues to decline.

To further alleviate any concern that the decline in public companies affects our

inferences, in untabulated analyses, we examine a subsample of companies that were in the

sample for at least 5 of the first 7 years and find results consistent with previous results. That is,

the estimated fee premium generally declines over time from approximately 78 percent in 2004

to 25 percent in 2010 for this stable sample of observations. Because we required companies to

be in the sample for at least 5 of 7 years, these results provide evidence that the trend observed in

primary analysis does not appear to result from a changing composition of public companies in

our sample. 33

5.3.2 Placebo Tests and Alternative Bandwidths

The inferences in this study rely on attributing the discontinuities at compliance

thresholds to the effects of SOX 404(b) compliance, and not some other factor. Because the $75

million threshold for compliance is fairly arbitrary (as is the $50 million threshold), these

findings are unlikely to be attributable to something else across a narrow range of public floats.

Nonetheless, we perform additional tests to further increase confidence in the documented

results. First, we perform a placebo test that uses an alternative, arbitrary threshold, $150 million

and examine the companies in the $100 million bandwidth of public float around $150 million. If

we draw similar inferences using a Float150 indicator, this suggests that the results may be

spurious. In untabulated analysis, the coefficients on Float150 in the pooled samples with all

dependent variables used throughout the paper are insignificant (t-stat < 1.00).

Second, since RDD estimates could be impacted by the selected bandwidth, we reperform

33
While there is no reason to believe that the increase in estimated compliance premium following 2010 is
attributable to sample composition changes, we re-estimate our results on a subsample of companies that are present
for at least 7 years during our entire sample. We observe an increase in the estimated compliance fee premium
following 2010 on this subsample as well, consistent with prior results.

29

Electronic copy available at: https://ssrn.com/abstract=3420787


our primary tests using alternative bandwidths of $120, $80, and $60 million in public float and

tabulate the results in Appendix II. As shown, inferences are essentially the same regardless of

the bandwidth chosen. In particular, we observe a similar pattern over time to previous audit fee

tests; we continue to find that 404(b) is associated with a lower likelihood of an ICD, but find no

evidence that ICDs are more informative for compliers; and we fail to find any evidence that

404(b) is associated with improved reporting quality. These analyses suggest that our primary

inferences are not the product of the bandwidth used to establish our RDD sample.

6. Conclusion

In this study, we use a “fuzzy” RDD to evaluate various effects of Section 404(b) of the

Sarbanes-Oxley Act since its passage. While evidence suggests that audit fee premiums

associated with internal control audits have not materially and sustainably declined since

inception, we find no compelling evidence that the increased audit effort is associated with

improved financial reporting. Specifically, we continue to document significantly higher audit

costs, but do not find that ICFR audits materially improve ICFRs, the informativeness of ICFR

management reports, or financial reporting quality. Our findings directly respond to calls from

regulators and academics debating the recent Section 404(b) compliance rule changes introduced

by the SEC and should be of interest to a variety of corporate stakeholders.

Our results are subject to several caveats. First, our tests only address potential financial

reporting benefits and do not address potential operational improvements associated with ICFR

audits. Second, the strength of our design in identifying the causal effects limits the

generalizability of our findings. That is, it is possible that costs are more scalable for large

companies and the benefits accrue more quickly, so the effects may differ based on company

size. Nonetheless, the burden of 404(b) requirements on smaller reporting companies is an issue

30

Electronic copy available at: https://ssrn.com/abstract=3420787


monitored by the SEC, as evidenced by the recent rule changes. In addition, even though the

magnitude of estimates may not generalize to larger companies, we expect that the trends we

observe (in particular the increase in compliance costs following 2010) likely do generalize.

31

Electronic copy available at: https://ssrn.com/abstract=3420787


References

Albuquerque, A. M., and J. Zhu. 2019. Has Section 404 of the Sarbanes-Oxley Act Discouraged
Corporate Investment? New Evidence from a Natural Experiment. Management Science
65(7): 2947-3448.
Alexander, C. R., S. W. Bauguess, G. Bernile, Y. H. A. Lee, and J. Marietta-Westberg. 2013.
Economic Effects of SOX Section 404 Compliance: A Corporate Insider Perspective.
Journal of Accounting & Economics 56(2-3): 267-290.
Ashbaugh-Skaife, H., D. Collins, W. Kinney, and R. LaFond. 2009. The Effect of Internal
Control Deficiencies on Firm Risk and Cost of Equity Capital. Journal of Accounting
Research 47(1): 1-43.
Barth, M., W. Landsman, J. H. Schroeder, and D. Taylor. 2019. Re: Amendments to the
Accelerated Filer and Large Accelerated Filer Definitions – File No. S7-06-19. July
2019. Available at: https://www.sec.gov/comments/s7-06-19/s70619-5802113-
187069.pdf
Bhaskar, L.S., J. H. Schroeder, and M. L. Shepardson. 2019. Integration of Internal Controls and
Financial Statement Audits: Are Two Audits Better than One? The Accounting Review
94(2): 53-81.
Brush, J. 2019. Re: File Number S7-06-19 Support Amendments to Accelerated and Large
Accelerated Filer Definitions. May 2019. Available at:
https://www.sec.gov/comments/s7-06-19/s70619-5585914-185429.pdf
Center for Audit Quality. 2017. CAQ Pulse Poll: Certified Financial Advisor Perspectives on the
Sarbanes-Oxley Act. May 2017. https://www.thecaq.org/caq-pulse-poll-certified-
financial-advisor-perspectives-sarbanes-oxley-act
Clay, C., and D. Kim. 2017. Sarbanes-Oxley: 15 Years of Successes and Challenges. Accounting
Today. https://www.accountingtoday.com/opinion/sarbanes-oxley-marks-15-years-of-
successes-and-challenges
Coates, J. 2007. The Goals and Promise of the Sarbanes-Oxley Act. Journal of Economic
Perspectives 21(1): 91-116.
Coates, J., and S. Srinivasan. 2014. SOX After Ten Years: A Multidisciplinary Review.
Accounting Horizons 28(3): 627-671.
Cohen, D., A. Dey, and T. Lys. 2008. Real and Accrual‐Based Earnings Management in the Pre‐
and Post‐Sarbanes‐Oxley Periods. The Accounting Review 83(3): 757-787.
CRA. 2005. Sarbanes-Oxley Section 404. Costs and Remediation of Deficiencies: Estimates
from a Sample of Fortune 1000 Companies.
https://www.sec.gov/spotlight/soxcomp/soxcomp-all-attach.pdf
DeFond, M., and C. Lennox. 2017. Do PCAOB Inspections Improve the Quality of Internal
Control Audits? Journal of Accounting Research 55(3): 591-627.
Doogar, R., P. Sivadasan, and I. Solomon. 2010. The Regulation of Public Company Auditing:
Evidence from the Transition to AS5. Journal of Accounting Research 48(4): 795-814.

32

Electronic copy available at: https://ssrn.com/abstract=3420787


Engel, E., R. Hayes, and X. Wang. 2007. The Sarbanes–Oxley Act and Firms’ Going Private
Decisions. Journal of Accounting and Economics 44(1-2), 116–145.
Fan, Y., C. Li, and K. Raghunandan. 2017. Is SOX 404(a) Management Internal Control
Reporting an Effective Alternative to SOX 404(b) Internal Control Audits? Auditing: A
Journal of Practice and Theory 36(3), 71-89.
Feng, M., C. Li , and S. McVay. 2009. Internal Control and Management Guidance. Journal of
Accounting and Economics 48, 190-209.
Feng, M., C. Li, McVay, S. and H. Skaife. 2015. Does Ineffective Internal Control over
Financial Reporting affect a Firm’s Operations? Evidence from Firms’ Inventory
Management. The Accounting Review 90(2), 529-557.
Financial Executives International (FEI). 2005. FEI Survey on Sarbanes-Oxley Section 404
Implementation.
Ge, W., A. Koester, and S. McVay. 2017. Benefits and costs of Sarbanes-Oxley Section 404(b)
exception: Evidence from small firms’ internal control disclosures. Journal of Accounting
and Economics 63, 358-384.
Government Accountability Office (GAO). 2013. SEC Should Consider Requiring Companies to
Disclose Whether They Obtained an Auditor Attestation. GAO-13-582. (July). U.S.
Government Accountability Office, Report to Congressional Committees, Internal
Controls. Washington, DC.
Gow, I. D., D. F. Larcker, and P. Reiss. 2016. Causal Inference in Accounting Research. Journal
of Accounting Research 54(2): 477-523.
Honigsberg, C., and S. Rajgopal. 2019. Re: Amendments to the Accelerated Filer and Large
Accelerated Filer Definitions; Proposed Rules; Request for Comments [File Number s7-
06-19]. July 2019. Available at: https://www.sec.gov/comments/s7-06-19/s70619-
5849740-188567.pdf
Iliev, P. 2010. The Effect of SOX Section 404: Costs, Earnings Quality, and Stock Prices.
Journal of Finance 65(3): 1163–1196.
Jackson, R. J. 2019. Statement on Proposed Amendments to Sarbances Oxley 404(b) Accelerated
Filer Definiton. Available at: https://www.sec.gov/news/public-statement/jackson-
statement-proposed-amendments-accelerated-filer-definition
Kinney, W., and M. Shepardson. 2011. Do control effectiveness disclosures require SOX 404(b)
internal control audits? A natural experiment with small U.S. public Companies. Journal
of Accounting Research 49(2): 413-448.
Leuz, C., A. Triantis, and T. Y. Wang. 2008. Why Do Firms Go Dark? Causes and Economic
Consequences of Voluntary SEC Deregistrations. Journal of Accounting & Economics
45(2-3): 181–208.
PricewaterhouseCoopers. 2017. SEC Advisory Committee on Small and Emerging Companies.
September 2017. https://www.sec.gov/info/smallbus/acsec/leonard-combs-pwc-acsec-
091317.pdf

33

Electronic copy available at: https://ssrn.com/abstract=3420787


PricewaterhouseCoopers. 2019. Re: File Number S7-06-19. July 2019. Available at:
https://www.sec.gov/comments/s7-06-19/s70619-5875838-188687.pdf
Public Company Accounting Oversight Board (PCAOB). 2007. PCAOB Auditing Standard No.
5. 2007. An Audit of Internal Control Over Financial Reporting That is Integrated with an
Audit of Financial Statements.
https://pcaobus.org/Standards/Auditing/Pages/Auditing_Standard_5.aspx
Public Company Accounting Oversight Board (PCAOB). 2010. PCAOB Auditing Standard
2110. Identifying and Assessing Risks of Material Misstatement.
https://pcaobus.org/Standards/Auditing/Pages/AS2110.aspx
Public Company Accounting Oversight Board (PCAOB). 2017. PCAOB Staff Inspection Brief:
Information about 2017 Inspections. Volume 2017:3.
https://pcaobus.org/Inspections/Documents/inspection-brief-2017-3-issuer-scope.pdf
Rose, P., and S. D. Solomon. 2016. Where Have All the IPOs Gone? The Hard Life of the Small
IPO. Harvard Business Law Review. 83
Securities and Exchange Commission (SEC). 2009. Study and Recommendations on Sections
404(b) of the Sarbanes-Oxley Act of 2002 for Issuers with Public Float Between $75 and
$250 Million. Washington, D.C.
Securities and Exchange Commission (SEC). 2019. Release No. 34-85814; File No. S7-06-19.
Amendments to the Accelerated Filer and Large Accelerated Filer Definitions. Available
at: https://www.sec.gov/rules/proposed/2019/34-85814.pdf

34

Electronic copy available at: https://ssrn.com/abstract=3420787


Appendix I: Variable Definitions

Variables Definition
404(b) Compliant One if the company has a Section 404(b) opinion, zero otherwise
(from Audit Analytics)

404(𝑏𝑏)𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 Predicted value of 404(b) Compliant from equation [1]
Audit Fees Natural log of total audit fees (from Audit Analytics)
Big 4 One if the company has a Big 4 auditor, zero otherwise (from
Audit Analytics)
CFO Comp Natural log of total compensation paid to CFO(s) during the year
(from DEF 14A filings)
Float Public float disclosed in the 10-K (in millions) (from 10-K filings)
Float50 One if the float is greater than $50 million, zero otherwise
Float75 One if the float is greater than $75 million, zero otherwise
Leverage Liabilities (Compustat: LTt) divided by total assets (Compustat:
ATt)
Log Assets Natural log of total assets (in millions) (Compustat: ATt)
Log Market Natural log of market value of equity (in millions) (Compustat:
CSHOt x PRCC_Ft)
Receivables Receivables (Compustat: RECTt) divided by total assetst
(Compustat: ATt)
Restate 8k One if the client subsequently restated the year’s audited financial
statements and the restatement was disclosed in an item 4.02 8-K
filing, zero otherwise (from Audit Analytics)
ROA Net income (Compustat: NI) divided by average total assets
(Compustat: (ATt + ATt-1) / 2)
Sales Natural log of sales (in millions) (Compustat: SALEt)
Segments Natural log of one plus the total number of business and operating
segments reported in the fiscal year (from Compustat)
Weak 404(a) One if the company discloses at least one 404(a) weakness, zero
otherwise (from Audit Analytics)

35

Electronic copy available at: https://ssrn.com/abstract=3420787


Appendix II: Tabulation of Alternative Bandwidth Results

Appendix II presents results from each two stage least squares regression discontinuity test on alternative bandwidths of $60, $80 million, and $120 million around
the primary $75 million threshold used to determine Section 404(b) compliance. All tables in this appendix include the same control variables, fixed effects, and
standard error adjustments as the corresponding table in the manuscript. Sample sizes are not presented for brevity. We estimate the regressions in these panels
using OLS and present two-tailed t-statistics below the coefficients. ***, **, and * indicate p<0.01, p<0.05, and p<0.10, respectively. We use robust standard
errors, clustering by company in the pooled regressions. All variables are defined in Appendix I.

Appendix II - Table 5, Panel A: Effect of SOX 404(b) Compliance on Audit Fees


(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit
VARIABLES Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees
$120 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.526*** 0.661*** 0.631*** 0.513*** 0.590*** 0.505*** 0.349*** 0.225*** 0.453*** 0.529*** 0.435*** 0.472*** 0.668***
(18.30) (12.01) (10.47) (8.22) (10.28) (8.57) (4.04) (2.81) (6.26) (6.93) (5.54) (6.33) (6.47)
$80 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.495*** 0.638*** 0.548*** 0.517*** 0.552*** 0.452*** 0.352*** 0.215** 0.454*** 0.573*** 0.313*** 0.447*** 0.538***
(15.38) (10.00) (7.50) (7.22) (7.97) (5.39) (3.18) (2.15) (4.83) (5.86) (3.24) (4.76) (4.49)
$60 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.522*** 0.642*** 0.590*** 0.531*** 0.622*** 0.420*** 0.315** 0.253** 0.428*** 0.661*** 0.290** 0.487*** 0.535***
(14.69) (8.77) (7.11) (6.57) (7.59) (3.19) (2.00) (2.11) (3.83) (5.51) (2.45) (4.27) (3.49)

Appendix II - Table 5, Panel B: Effect of SOX 404(b) Compliance on Big 4 Auditor Selection
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4
$120 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.151*** 0.080** 0.128*** 0.007 0.139*** 0.133*** 0.198*** 0.173*** 0.206*** 0.298*** 0.159*** 0.176*** 0.314***
(7.25) (2.15) (3.03) (0.16) (3.27) (3.03) (2.91) (2.94) (3.77) (5.13) (2.71) (3.41) (4.40)
$80 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.115*** 0.053 0.075 0.010 0.113** 0.118** 0.149* 0.163** 0.224*** 0.353*** 0.085 0.087 0.218**
(4.87) (1.22) (1.49) (0.21) (2.05) (2.00) (1.66) (2.06) (3.13) (4.60) (1.16) (1.33) (2.52)
$60 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.113*** 0.059 0.083 0.005 0.187*** 0.159* 0.095 0.158 0.200** 0.376*** -0.010 0.099 0.216**
(4.18) (1.18) (1.39) (0.09) (2.88) (1.80) (0.78) (1.63) (2.27) (3.89) (-0.11) (1.23) (1.99)

36

Electronic copy available at: https://ssrn.com/abstract=3420787


Appendix II - Table 5, Panel C: Effect of SOX 404(b) Compliance on Audit Fees controlling for auditor selection (mediation analysis)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit
VARIABLES Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees
$120 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.437*** 0.619*** 0.565*** 0.509*** 0.516*** 0.432*** 0.240*** 0.111 0.315*** 0.335*** 0.328*** 0.342*** 0.409***
(17.39) (12.06) (10.01) (8.77) (9.69) (7.92) (3.01) (1.62) (5.23) (5.04) (4.69) (5.27) (5.02)
$80 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.429*** 0.609*** 0.514*** 0.512*** 0.497*** 0.390*** 0.273*** 0.108 0.307*** 0.347*** 0.254*** 0.381*** 0.369***
(15.04) (10.34) (7.52) (7.70) (7.77) (5.11) (2.67) (1.27) (3.93) (3.92) (2.98) (4.90) (3.79)
$60 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.456*** 0.609*** 0.552*** 0.528*** 0.525*** 0.334*** 0.262* 0.144 0.300*** 0.440*** 0.296*** 0.410*** 0.365***
(14.60) (9.01) (7.05) (7.04) (6.98) (2.83) (1.89) (1.42) (3.14) (3.93) (2.85) (4.42) (2.98)

37

Electronic copy available at: https://ssrn.com/abstract=3420787


Appendix II - Table 6 Panel C: The Association between SOX 404(b) Compliance and Internal Control Weakness Disclosures
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
Weak Weak Weak Weak Weak Weak Weak Weak Weak Weak
VARIABLES 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a)
$120 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 -0.029** -0.056* -0.053* -0.024 -0.024 0.012 -0.047 -0.015 0.008 0.011
(-2.04) (-1.69) (-1.70) (-0.52) (-0.68) (0.33) (-1.32) (-0.31) (0.18) (0.21)
$80 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 -0.022 -0.053 -0.021 -0.069 -0.032 0.009 -0.022 -0.035 0.035 -0.015
(-1.35) (-1.30) (-0.50) (-1.18) (-0.85) (0.18) (-0.51) (-0.58) (0.72) (-0.25)
$60 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 -0.038* -0.062 -0.071 -0.071 -0.053 -0.006 0.001 -0.043 -0.024 -0.006
(-1.87) (-1.32) (-1.01) (-0.89) (-1.21) (-0.09) (0.03) (-0.58) (-0.42) (-0.08)

Appendix II – Table 7, Panel A: Association Between 404(a) Weakness Disclosures and Restatements for non-404(b) Compliant Companies
Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K
$120 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.039*** 0.006 0.053* 0.088** 0.047* -0.016 0.007 0.021 0.100*** 0.058*
(3.59) (0.27) (1.67) (2.12) (1.75) (-0.89) (0.26) (0.83) (3.28) (1.73)
$80 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.064*** 0.008 0.135* 0.194*** 0.096 -0.056* 0.096* 0.089* 0.087 -0.007
(3.59) (0.23) (1.87) (2.68) (1.65) (-1.74) (1.80) (1.74) (1.59) (-0.83)
$60 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.086*** 0.002 0.190* 0.255** 0.106 -0.071 0.158* 0.097 0.146** 0.001
(3.49) (0.04) (1.89) (2.07) (1.31) (-1.31) (1.67) (1.47) (2.02) (0.08)

Appendix II – Table 7, Panel B: Association Between 404(a) Weakness Disclosures and Restatements for 404(b) Compliant Companies
Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K
$120 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.035** 0.013 0.020 -0.002 0.120 -0.008 0.056 0.071 0.099 0.026
(2.08) (0.36) (0.51) (-0.06) (1.36) (-0.23) (0.85) (1.28) (1.54) (0.48)
$80 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.034* -0.033 0.008 -0.019 0.177 -0.018 0.097 0.086 0.123 0.039
(1.83) (-1.09) (0.21) (-0.57) (1.54) (-0.45) (1.23) (1.34) (1.45) (0.57)
$60 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.029 -0.024 0.017 -0.009 0.044 0.008 0.120 0.063 0.180* -0.009
(1.44) (-0.62) (0.41) (-0.21) (0.38) (0.16) (1.28) (1.06) (1.76) (-0.20)

38

Electronic copy available at: https://ssrn.com/abstract=3420787


Appendix II – Table 7, Panel C: Association Between 404(a) Weakness Disclosures and Restatements for Companies with less $75 million in Public Float
Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K
$120 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.033*** 0.013 0.022 0.068** 0.031 -0.018 0.020 0.037 0.087*** 0.042
(3.34) (0.63) (0.91) (1.96) (1.14) (-1.08) (0.76) (1.36) (3.11) (1.30)
$80 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.045*** 0.018 0.026 0.112** 0.062 -0.045* 0.108** 0.103** 0.076 -0.042*
(3.10) (0.63) (0.69) (2.15) (1.12) (-1.74) (2.21) (2.08) (1.60) (-1.91)
$60 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.059*** 0.021 0.043 0.132* 0.051 -0.038 0.163** 0.127** 0.124** -0.060*
(3.10) (0.66) (0.93) (1.81) (0.69) (-0.96) (2.24) (2.09) (2.07) (-1.80)

Appendix II – Table 7, Panel D: Association Between 404(a) Weakness Disclosures and Restatements for Companies with greater $75 million in Public Float
Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K
$120 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.050** 0.005 0.138 -0.041** 0.183 -0.027 0.007 -0.014 0.147* 0.060
(2.37) (0.12) (1.64) (-2.43) (1.63) (-0.62) (0.10) (-0.23) (1.90) (1.03)
$80 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.060** -0.036 0.201* -0.033** 0.264* -0.050 0.048 -0.008 0.161 0.110
(2.24) (-0.97) (1.90) (-2.13) (1.80) (-0.89) (0.59) (-0.11) (1.54) (1.25)
$60 Million Bandwidth
𝐖𝐖𝐖𝐖𝐖𝐖𝐖𝐖 𝟒𝟒𝟒𝟒𝟒𝟒(𝐚𝐚) 0.054* -0.023 0.242** -0.037* 0.104 -0.027 0.065 -0.106** 0.214 0.026
(1.77) (-0.44) (2.12) (-1.75) (0.62) (-0.36) (0.64) (-2.14) (1.64) (0.49)

Appendix II – Table 8, Panel C: Association Between 404(b) Compliance and Financial Reporting Quality – Restate 8-K
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate8K Restate8K Restate8K Restate8K Restate8K Restate8K Restate8K Restate8K Restate8K Restate8K Restate8K Restate8K Restate8K
$120 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 -0.007 -0.046* -0.021 -0.028 -0.008 0.040* -0.008 0.017 0.029 0.055*** 0.009 0.024 -0.039
(-0.91) (-1.81) (-1.01) (-1.27) (-0.34) (1.70) (-0.24) (0.61) (1.51) (2.73) (0.37) (1.11) (-1.33)
$80 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 -0.005 -0.067** -0.013 -0.040 0.009 0.060** -0.015 0.038 0.022 0.039* 0.013 0.032 0.002
(-0.46) (-2.34) (-0.52) (-1.52) (0.31) (2.13) (-0.38) (0.97) (0.82) (1.81) (0.39) (1.33) (0.05)
$60 Million Bandwidth

𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 -0.002 -0.066** 0.014 -0.039 0.023 0.031 -0.004 0.038 -0.005 0.049 0.016 0.041 0.006
(-0.18) (-2.10) (0.48) (-1.40) (0.73) (0.74) (-0.08) (0.74) (-0.15) (1.44) (0.39) (1.27) (0.14)

39

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 1: Descriptive Statistics

Panel A: Full Sample Descriptive Statistics


Standard
Variable N Mean Deviation p(25) Median p(75)
404(b) Compliant 11,013 0.453 0.498 0.000 0.000 1.000
Float (millions) 11,013 63.984 28.178 39.641 58.800 85.644
Float75 11,013 0.328 0.470 0.000 0.000 1.000
Float50 11,013 0.607 0.488 0.000 1.000 1.000
Log Assets 11,013 4.844 1.490 3.821 4.803 6.077
Log Market 11,013 4.347 0.735 3.859 4.356 4.840
Sales 11,013 3.800 1.782 3.166 3.963 4.852
Leverage 11,013 0.601 0.394 0.299 0.551 0.892
Receivables 11,013 0.276 0.263 0.063 0.175 0.478
Segments 11,013 0.744 0.533 0.693 0.693 1.099
Audit Fees 11,013 12.573 0.912 11.912 12.535 13.212
Big 4 11,013 0.359 0.480 0.000 0.000 1.000
Restate 8k 11,013 0.046 0.209 0.000 0.000 0.000
Weak 404(a) 8,755 0.115 0.319 0.000 0.000 0.000
CFO Comp 8,956 12.714 0.633 12.291 12.675 13.117

Panel B: Comparative Descriptive Statistics


Float < $75 Million Float > $75 Million
Variable N Mean N Mean Difference t-stat

404(b) Compliant 7,396 0.228 3,617 0.913 0.685 88.98 ***


Float (millions) 7,396 47.054 3,617 98.601 51.547 >100 ***
Log Assets 7,396 4.633 3,617 5.276 0.643 21.72 ***
Log Market 7,396 4.106 3,617 4.841 0.735 55.86 ***
Sales 7,396 3.571 3,617 4.267 0.696 19.59 ***
Leverage 7,396 0.614 3,617 0.577 -0.037 -4.64 ***
Receivables 7,396 0.283 3,617 0.261 -0.022 -4.05 ***
Segments 7,396 0.724 3,617 0.784 0.060 5.56 ***
Audit Fees 7,396 12.362 3,617 13.003 0.641 36.64 ***
Big 4 7,396 0.297 3,617 0.485 0.188 19.62 ***
Restate 8k 7,396 0.044 3,617 0.050 0.006 1.44
Weak 404(a) 5,309 0.122 3,446 0.104 -0.018 -2.55 **
CFO Comp 5,837 12.636 3,119 12.859 0.223 16.09 ***
Notes: These panels present the descriptive statistics for variables used in our tests. ***, **, and * indicate p<0.01, p<0.05, and p<0.10,
respectively. All variables are defined in Appendix I.

40

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 2: Instrument Exclusion
Panel A: Float75 Regressions
(1) (2) (3) (4) (5) (6)
VARIABLES Audit Fees 2003 Audit Fees 2004 Big4 2003 Big4 2004 Restate 8K 2003 Restate 8K 2004

Float75 -0.006 0.496*** 0.027 0.054* 0.018 -0.043**


(-0.15) (10.43) (1.02) (1.77) (0.84) (-2.11)

Controls Included Included Included Included Included Included


Fixed Effects Industry Industry Industry Industry Industry Industry

Observations 1,278 1,229 1,278 1,229 1,278 1,229


Adjusted R-squared 0.527 0.529 0.273 0.214 0.044 0.029

Panel B: Float50 Regressions


(1) (2) (3) (4) (7) (8)
VARIABLES Audit Fees 2003 Audit Fees 2004 Big4 2003 Big4 2004 Restate 8K 2003 Restate 8K 2004

Float50 0.018 0.312*** 0.007 0.017 0.011 -0.029


(0.48) (6.21) (0.27) (0.54) (0.57) (-1.43)

Controls Included Included Included Included Included Included


Fixed Effects Industry Industry Industry Industry Industry Industry

Observations 1,278 1,229 1,278 1,229 1,278 1,229


Adjusted R-squared 0.527 0.500 0.273 0.212 0.043 0.027
Notes: We estimate these regressions using OLS and present two-tailed t-statistics below the coefficients. ***, **, and * indicate p<0.01, p<0.05, and p<0.10, respectively. We use robust standard errors
and cluster by company. All variables are defined in Appendix I.

41

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 3: Instrument Strength

VARIABLES 404(b) Compliant

Float75 0.504***
(40.24)
Float50 0.248***
(19.60)
Log Assets 0.056***
(7.72)
Log Market -0.010
(-1.41)
Sales 0.015***
(3.47)
Leverage 0.034***
(2.80)
Receivables -0.164***
(-5.01)
Segments -0.021*
(-1.69)

Fixed Effects Industry, Year

Observations 11,013
Adjusted R-squared 0.523
Notes: We estimate these regressions using OLS and present two-tailed t-statistics below the coefficients. ***, **, and * indicate p<0.01, p<0.05,
and p<0.10, respectively. We use robust standard errors and cluster by company. All variables are defined in Appendix I.

42

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 4: Estimating Float Manipulation

Panel A: Histograms of Public Float in 2003, Market Value of Equity from 2004-2015, and Public Float from 2004-2015

Public Float 2003 Market Value Equity 2004-2015 Public Float 2004-2015
.02

.02

.02
.015

.015

.015
Density

Density

Density
.01

.01

.01
.005

.005

.005
0

0
20 40 60 80 100 120 20 40 60 80 100 120 20 40 60 80 100 120
Public Float Market Value of Equity Public Float

Panel B: Comparison of Distributions Above and Below Compliance Threshold

Public Float 2003 Market Cap 2004-2015 Public Float 2004-2015


n Percent n Percent n Percent
< $75 million 134 55.14% 1,093 53.50% 1,129 57.93%
>= $75 million 109 44.86% 950 46.50% 820 42.07%
243 100% 2,043 100% 1,949 100%

43

Electronic copy available at: https://ssrn.com/abstract=3420787


Panel C: Percent of Company-Years Between $65 Million and $85 Million in Float (Market Capitalization) that Fall Below $75 Million

Float Avoidance over Time


100.00%

90.00%

80.00%

70.00%

60.00%

50.00%

40.00%

30.00%

20.00%

10.00%

0.00%
2002 2004 2006 2008 2010 2012 2014 2016
Public Float Market Value of Equity

44

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 5: The Association between SOX 404(b) Compliance and Audit Fees

Panel A: Two Stage Least Squares Estimates of the Effect of SOX 404(b) Compliance on Audit Fees
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit
VARIABLES Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees


𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.496*** 0.650*** 0.562*** 0.532*** 0.594*** 0.447*** 0.314*** 0.169* 0.423*** 0.537*** 0.354*** 0.494*** 0.609***
(16.40) (11.07) (8.61) (7.98) (9.50) (6.56) (3.25) (1.92) (5.28) (6.58) (4.32) (5.78) (5.40)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 11,013 1,229 1,184 1,134 1,009 1,038 940 881 785 765 727 692 629
Adjusted R-squared 0.583 0.577 0.593 0.615 0.577 0.610 0.605 0.543 0.579 0.591 0.577 0.541 0.525

Panel B: Two Stage Least Squares Estimates of the Effect of SOX 404(b) Compliance on Audit Fees, Controlling for the Auditor Selected (Mediation Analysis)
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit Audit
VARIABLES Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees Fees


𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.423*** 0.614*** 0.516*** 0.525*** 0.529*** 0.381*** 0.224** 0.068 0.287*** 0.350*** 0.299*** 0.404*** 0.382***
(16.05) (11.23) (8.54) (8.53) (9.19) (6.13) (2.56) (0.90) (4.28) (4.79) (4.10) (5.54) (4.31)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 11,013 1,229 1,184 1,134 1,009 1,038 940 881 785 765 727 692 629
Adjusted R-squared 0.657 0.634 0.646 0.665 0.640 0.681 0.681 0.644 0.687 0.686 0.676 0.661 0.676

45

Electronic copy available at: https://ssrn.com/abstract=3420787


Panel C: Two Stage Least Squares Estimates of the Effect of SOX 404(b) Compliance on Big 4 Auditor Selection
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4 Big 4


𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.125*** 0.068* 0.090** 0.013 0.123*** 0.119** 0.159** 0.154** 0.212*** 0.290*** 0.083 0.115* 0.275***
(5.68) (1.73) (1.98) (0.29) (2.63) (2.43) (2.07) (2.33) (3.42) (4.58) (1.27) (1.95) (3.52)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 11,013 1,229 1,184 1,134 1,009 1,038 940 881 785 765 727 692 629
Adjusted R-squared 0.223 0.218 0.191 0.206 0.188 0.242 0.233 0.212 0.186 0.237 0.139 0.131 0.093
Notes: We estimate the regressions in these panels using two-stage least squares and present two-tailed t-statistics below the coefficients. ***, **, and * indicate p<0.01, p<0.05, and p<0.10, respectively. We use robust
standard errors, clustering by company in the pooled regressions. All variables are defined in Appendix I.

46

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 6: The Association between SOX 404(b) Compliance and Internal Control Weakness Disclosures

Panel A: 404(a) Weakness Disclosures and Sox 404(b) Compliance


2007 2008 2009 2010 2011 2012 2013 2014 2015
Weak Weak Weak Weak Weak Weak Weak Weak Weak
404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a)

404(b) Compliant = 0 15.29% 11.96% 13.13% 10.94% 11.60% 13.47% 14.35% 17.29% 10.99%

404(b) Compliant = 1 9.07% 6.01% 6.07% 3.80% 7.47% 6.05% 11.03% 10.92% 9.64%

Difference -6.22%*** -5.95%*** -7.06%*** -7.14%*** -4.13%* -7.42%*** -3.32% -6.37%** -1.35%
(-2.98) (-3.37) (-3.71) (-4.03) (-1.93) (-3.39) (-1.27) (-2.21) (-0.54)

Panel B: Figure of 404(a) Weakness Disclosures and Sox 404(b) Compliance


20.00%

18.00%

16.00%

14.00%

12.00%
Exempt 10.00%
Compliant 8.00%

6.00%

4.00%

2.00%

0.00%
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

47

Electronic copy available at: https://ssrn.com/abstract=3420787


Panel C: Two Stage Least Squares Estimates of the Effect of SOX 404(b) Compliance on 404(a) Weakness Disclosure
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
Weak Weak Weak Weak Weak Weak Weak Weak Weak Weak
VARIABLES 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a) 404(a)


𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 -0.026* -0.052 -0.052 -0.037 -0.045 -0.011 -0.032 0.006 0.044 -0.001
(-1.71) (-1.41) (-1.44) (-0.74) (-1.29) (-0.25) (-0.85) (0.11) (1.00) (-0.01)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 7,316 984 1,025 928 869 779 748 704 666 613
Adjusted R-squared 0.053 0.028 0.024 0.031 0.037 0.025 0.077 0.091 0.150 0.048
Notes: Panel A presents the univariate differences in Weak 404(a) by year. We estimate the regressions in Panel B using two-stage least squares and present two-tailed t-statistics below the coefficients.
***, **, and * indicate p<0.01, p<0.05, and p<0.10, respectively. We use robust standard errors, clustering by company in the pooled regression. All variables are defined in Appendix I.

48

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 7: The Association between SOX 404(b) Compliance and Internal Control Weakness Disclosure Informativeness

Panel A: Exempt Companies


Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K

Weak 404(a) 0.050*** 0.003 0.098** 0.115** 0.061 -0.029 0.041 0.061 0.099** 0.035
(3.67) (0.13) (2.15) (2.11) (1.53) (-1.33) (1.24) (1.61) (2.36) (0.94)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 3,742 510 393 335 448 431 401 432 428 364
Adjusted R-squared 0.019 0.001 0.042 0.044 0.019 0.054 0.045 0.002 0.030 0.007

Panel B: Compliant Companies


Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K

Weak 404(a) 0.038** -0.014 0.029 0.004 0.138 -0.008 0.076 0.075 0.112 0.035
(2.04) (-0.40) (0.66) (0.11) (1.42) (-0.21) (1.14) (1.29) (1.44) (0.55)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 3,574 474 632 593 421 348 347 272 238 249
Adjusted R-squared 0.026 0.055 0.018 0.020 0.068 0.037 0.013 0.043 0.040 -0.026

49

Electronic copy available at: https://ssrn.com/abstract=3420787


Panel C: Companies with less than $75 million in public float (Float75 = 0)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K

Weak 404(a) 0.039*** 0.012 0.036 0.080* 0.037 -0.029 0.053 0.078** 0.077** 0.010
(3.33) (0.51) (1.13) (1.87) (0.99) (-1.54) (1.61) (2.05) (2.14) (0.24)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 4,981 627 665 658 607 545 515 513 454 397
Adjusted R-squared 0.018 -0.001 -0.004 0.039 0.042 0.043 0.077 0.018 0.012 0.002

Panel D: Companies with greater than $75 million in public float (Float75 = 1)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Pooled 2007 2008 2009 2010 2011 2012 2013 2014 2015
VARIABLES Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K Restate 8K

Weak 404(a) 0.057** -0.025 0.198* -0.052** 0.215* -0.041 0.024 -0.005 0.182* 0.082
(2.30) (-0.62) (1.84) (-2.34) (1.70) (-0.76) (0.35) (-0.08) (1.89) (1.17)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 2,336 357 361 270 262 234 233 191 212 216
Adjusted R-squared 0.024 0.072 0.045 -0.007 0.060 0.045 -0.005 0.024 0.036 -0.029
Notes: We estimate the regressions in each panel using OLS and present two-tailed t-statistics below the coefficients. ***, **, and * indicate p<0.01, p<0.05, and
p<0.10, respectively. We use robust standard errors, clustering by company in the pooled regressions. All variables are defined in Appendix I.

50

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 8: The Association between SOX 404(b) Compliance and Financial Reporting Quality

Panel A: Univariate Comparison of Restatement Rates Based on 404(b) Compliance


2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Restate Restate Restate Restate Restate Restate Restate Restate Restate Restate Restate Restate
8K 8K 8K 8K 8K 8K 8K 8K 8K 8K 8K 8K

404(b) Compliant = 0 9.49% 5.98% 5.05% 3.74% 4.43% 5.76% 2.39% 2.75% 1.20% 3.08% 2.42% 2.63%

404(b) Compliant = 1 9.36% 5.60% 4.34% 2.95% 5.38% 4.55% 5.23% 3.15% 3.15% 4.04% 2.52% 4.02%

Difference -0.13% -0.38% -0.71% -0.79% 0.95% -1.21% 2.84%** 0.40% 1.95%* 0.96% 0.10% 1.39%
(-0.07) (-0.27) (-0.55) (-0.69) (0.68) (-0.82) (2.22) (0.33) (1.88) (0.69) (0.08) (0.97)

Panel B: Figure of Comparison of Restatement Rates Based on 404(b) Compliance


10.00%

8.00%

6.00%

4.00%

2.00%

0.00%
2002 2004 2006 2008 2010 2012 2014 2016

Exempt Compliant

51

Electronic copy available at: https://ssrn.com/abstract=3420787


Panel C: Two Stage Least Squares Estimates of the Effect of SOX 404(b) Compliance on Material Restatements
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Restate Restate Restate Restate Restate Restate Restate Restate Restate Restate Restate Restate Restate
VARIABLES 8K 8K 8K 8K 8K 8K 8K 8K 8K 8K 8K 8K 8K


𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 -0.001 -0.057** -0.001 -0.024 0.001 0.043 0.013 0.025 0.037 0.041** 0.017 0.032 -0.025
(-0.10) (-2.16) (-0.05) (-0.97) (0.04) (1.60) (0.35) (0.80) (1.62) (2.32) (0.61) (1.59) (-0.74)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 11,013 1,229 1,184 1,134 1,009 1,038 940 881 785 765 727 692 629
Adjusted R-squared 0.022 0.027 0.005 0.020 0.007 0.003 0.016 0.035 0.025 0.023 0.005 -0.017 -0.015
Notes: We estimate the regressions in these panels using two-stage least squares and present two-tailed t-statistics below the coefficients. ***, **, and * indicate p<0.01, p<0.05, and p<0.10, respectively. We use
robust standard errors, clustering by company in the pooled regressions. All variables are defined in Appendix I.

52

Electronic copy available at: https://ssrn.com/abstract=3420787


Table 9: The Association between SOX 404(b) Compliance and CFO Salaries

Two Stage Least Squares Estimates of the Effect of SOX 404(b) Compliance on CFO Salaries
Pooled 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
CFO CFO CFO CFO CFO CFO CFO CFO CFO CFO CFO CFO CFO
VARIABLES Comp Comp Comp Comp Comp Comp Comp Comp Comp Comp Comp Comp Comp


𝟒𝟒𝟒𝟒𝟒𝟒(𝐛𝐛)𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 0.086*** 0.035 -0.005 -0.065 0.126** 0.040 0.142 -0.041 0.272*** 0.207** 0.052 0.238** 0.191
(3.17) (0.75) (-0.10) (-1.00) (2.15) (0.58) (1.36) (-0.47) (3.18) (2.29) (0.49) (2.31) (1.42)

Industry,
Fixed Effects Year Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry Industry
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Observations 8,956 946 924 960 837 902 793 730 636 618 568 549 493
Adjusted R-squared 0.306 0.241 0.224 0.254 0.206 0.307 0.285 0.243 0.292 0.302 0.276 0.177 0.188
Notes: We estimate the regressions in these panels using two-stage least squares and present two-tailed t-statistics below the coefficients. ***, **, and * indicate p<0.01, p<0.05, and p<0.10, respectively. We use robust
standard errors, clustering by company in the pooled regressions. All variables are defined in Appendix I.

53

Electronic copy available at: https://ssrn.com/abstract=3420787

You might also like