Accounting undergraduate Honors theses: The determinants and consequences of disclosure committee adoption

In this paper, I investigate the determinants and consequences of disclosure committee adoption. I find that companies with material weaknesses in internal controls over financial reporting and less readable 10-K filings are more likely to adopt disclosure committees.

University of Arkansas, Fayetteville

ScholarWorks@UARK

Theses and Dissertations

7-2015

The Determinants and Consequences of

Disclosure Committee Adoption

Lyle Roy Schmardebeck

University of Arkansas, Fayetteville

Commons, and theCorporate Finance Commons

This Dissertation is brought to you for free and open access by ScholarWorks@UARK It has been accepted for inclusion in Theses and Dissertations by

an authorized administrator of ScholarWorks@UARK For more information, please contact scholar@uark.edu, ccmiddle@uark.edu

Recommended Citation

Schmardebeck, Lyle Roy, "The Determinants and Consequences of Disclosure Committee Adoption" (2015) Theses and Dissertations.

1188.

http://scholarworks.uark.edu/etd/1188

The Determinants and Consequences of Disclosure Committee Adoption

The Determinants and Consequences of Disclosure Committee Adoption

A dissertation submitted in partial fulfillment

of the requirements for the degree of Doctor of Philosophy in Business Administration

by

Lyle Roy Schmardebeck Brigham Young University Bachelor of Science in Accounting, 2009 Brigham Young University Master of Accountancy, 2009

July 2015 University of Arkansas

This dissertation is approved for recommendation to the Graduate Council

_

Dr Kangzhen Xie

Committee Member

Abstract

After the passage of the Sarbanes-Oxley Act of 2002, the Securities and Exchange

Commission recommended that companies voluntarily adopt disclosure committees to aid in preparing company disclosures In this paper, I investigate the determinants and consequences

of disclosure committee adoption I find that companies with material weaknesses in internal controls over financial reporting and less readable 10-K filings are more likely to adopt

disclosure committees In consequences analyses, using a propensity score matched control sample and a difference-in-differences research design, I find that 10-K filings are longer and less readable after disclosure committee adoption However, consistent with institutional theory,

I do not find evidence of a reduction in information asymmetry or an increase in the

informativeness of earnings following disclosure committee adoption

Acknowledgments

I am grateful for the guidance and instruction of my dissertation chair, Linda Myers, and the members of my dissertation committee, Cory Cassell, James Myers, and Kenneth Xie I also thank Ben Anderson, T J Atwood, Cari Burke, Lauren Cunningham, Andrew Doucet, Taylor Joo, Stacey Kaden, Gary Peters, Michael Stuart, and seminar participants at the University of Arkansas, the University of Missouri, and the Miami Rookie Camp for providing helpful

comments and suggestions I express gratitude to my family for their support and

encouragement Finally, I express gratitude to my wife, Brenna, because I wouldn’t have been able to accomplish my goals without her love, friendship, and dedication

Table of Contents

I Introduction 1

II Background, Prior Literature, and Hypothesis Development 6

A Disclosure Committee 6

B Annual Report Readability 7

C Information Asymmetry 9

D Earnings Informativeness 10

III Research Design 12

A Determinants of Disclosure Committee Adoption 12

B Propensity Score Matching and Difference in Differences 15

C Readability Tests 16

D Information Asymmetry Tests 18

E Earnings Informativeness Tests 20

IV Sample Selection and Data 22

V Results 24

A Determinants Test and Propensity Score Matching 24

B Readability Tests 27

C Information Asymmetry Tests 29

D Earnings Informativeness Tests 29

VI Additional Analyses 30

A Specific Word Types 30

B Disclosure Committee Composition Tests 32

C Internal Control over Financial Reporting 35

VII Robustness Tests 36

A Timing of Disclosure Committee Adoption 36

B Measurement Window of Bid-Ask Spreads and Market Illiquidity 37

C Alternative Measure of Information Asymmetry 37

VIII Conclusion 38

IX References 41

X Appendix A: Sample Disclosure Committee Charter 46

XI Appendix B: Variable Definitions 48

List of Tables

1 Disclosure Committee Sample Summary 52

Panel A: Disclosure Committee and Control Observations by Year 52

Panel B: Disclosure Committee and Control Observations by Fama and French (1997) Industry 53

2 Descriptive Statistics for Determinants Model 55

3 Determinants Model with Fama and French (1997) Industry Classification 56

4 Determinants Model with One Digit SIC Codes 57

5 Propensity Score Matching 58

Panel A: Sample Selection 58

Panel B: Covariate Balance 58

6 Difference in Governance Controls 59

Panel A: Before Adoption 59

Panel B: After Adoption 59

7 Difference in Differences Sample Selection 60

Panel A: Readability Tests 60

Panel B: Information Asymmetry Tests 60

Panel C: Earnings Informativeness Tests 60

8 Number of Words Test 61

9 Gross File Size Test 62

10 Net File Size Test .63

11 Bid Ask Spread Test 64

12 Market Illiquidity Test 65

13 ERC Test 66

14 FERC Test .67

15 Number of Positive Words Test 68

16 Number of Negative Words Test 69

17 Number of Neutral Words Test 70

18 Percentage of Positive Words Test 71

19 Percentage of Negative Words Test 72

20 Percentage of Neutral Words Test 73

21 Disclosure Committee Composition Descriptive Statistics 74

22 Disclosure Committee Composition Number of Words Test 75

23 Disclosure Committee Composition Gross File Size Test 76

24 Disclosure Committee Composition Net File Size Test 77

25 Disclosure Committee Composition Percentage of Positive Words Test 78

26 Disclosure Committee Composition Percentage of Negative Words test 79

27 Disclosure Committee Composition Bid-Ask Spread Test 80

28 Disclosure Committee Composition Illiquidity Test 81

29 Disclosure Committee Composition ERC Test 82

30 Disclosure Committee Composition FERC Test 83

31 Disclosure Committee Adoption and Current Material Weaknesses 84

32 Disclosure Committee Adoption and Future Material Weaknesses 85

33 Number of Words Test Excluding Unclear Adoption Timing Observations 86

34 Gross File Size Test Excluding Unclear Adoption Timing Observations 87

35 Net File Size Test Excluding Unclear Adoption Timing Observations 88

36 Bid-Ask Spread Test Excluding Unclear Adoption Timing Observations 89

37 Market Illiquidity Test Excluding Unclear Adoption Timing Observations 90

38 ERC Test Excluding Unclear Adoption Timing Observations 91

39 FERC Test Excluding Unclear Adoption Timing Observations 92

40 Bid-Ask Spread Test Measuring BA_SPREAD over the fiscal year 93

41 Market Illiquidity Test Measuring ILLIQUIDITY over the fiscal year 94

42 Residual Volatility Test 95

I Introduction

In an effort to assist companies in dealing with additional disclosure requirements and increased scrutiny over financial reporting practices following the adoption of the Sarbanes-Oxley Act of 2002 (SOX), the Securities and Exchange Commission (SEC) recommended that companies adopt disclosure committees “with the responsibility of determining the materiality of information and determining disclosure obligations on a timely basis” (SEC 2002; Bevilacqua 2004) Although disclosure committee adoption is voluntary, regulators and practitioners believe that a disclosure committee is an essential part of a company’s communication process (SEC 2002; Deloitte and Touche 2003; Deloitte 2013)

Disclosure committee responsibilities can include reviewing the design and operation of disclosure controls, determining whether periodic filings (e.g., 8-K or Form 4 filings) are

necessary, and drafting and reviewing quarterly and annual filings (i.e., 10-Qs and 10-Ks) Although the composition of disclosure committees is not regulated, most are comprised of management (e.g., the Chief Executive Officer (CEO), Chief Financial Officer (CFO), Chief Accounting Officer, General Counsel, and Chief Investor Relations Officer) and board members (Bocchio and Daly 2007; McCarthy 2008) Most disclosure committees meet at least quarterly (KPMG 2011a) and report directly to the CEO or CFO (Deloitte 2013; WSJ 2013) Therefore, committee meetings can require a significant investment of management time and company resources Although regulators, practitioners, and the business press highlight the potential benefits of disclosure committee adoption, to my knowledge, no prior studies examine why companies form disclosure committees and whether disclosure committee adoption influences disclosure quality and/or market outcomes

In this study, I examine the determinants and consequences of disclosure committee adoption To perform my analyses, I hand collect a sample of disclosure committee adopters by performing keyword searches of SEC filings on EDGAR Using a sample of 764 disclosure committee adoptions, I find that companies are more likely to adopt disclosure committees when there are prior year material weaknesses in internal controls over financial reporting or when prior year 10-K filings are less readable I also find that disclosure committee adoption is more likely among larger companies and companies engaging a Big N auditor

Next, I investigate how disclosure committee adoption influences corporate disclosures Companies might adopt disclosure committees to facilitate the transfer of information across the

institutional theory, corporate governance mechanisms may be “primarily ceremonial and serve

as symbols of effective oversight” (Beasley et al 2009, p 69) Thus, disclosure committee adoption may not have an effect on disclosure quality

I first examine the impact of adoption on the length and readability of 10-K filings The SEC’s Plain English Rule mandates clear language that avoids “boilerplate” disclosure (SEC 1998) More recently, the SEC has undertaken projects to reduce disclosure overload and

10-K filings are more informative to investors and analysts (Li 2008; Miller 2010; Lehavy et al

formalizing the disclosure process and aiding in the accumulation of financial information, determining the materiality of transactions, and communicating this information to management

in a timely manner (Bocchio and Daly 2007; McCarthy 2008)

suggesting that they remove duplicate, redundant, or unnecessary disclosure from 10-K filings

and by identifying areas where additional disclosure may be useful or necessary (see SEC

Changes Tack on Disclosure Overload Project, available at:

http://blogs.wsj.com/cfo/2014/05/01/sec-changes-tack-on-disclosure-overload-project/)

2011; Loughran and McDonald 2014) If disclosure committee adoption represents a

commitment to improved disclosure quality, then I expect 10-K filings to be shorter and more readable post-adoption However, if disclosure committee adoption is primarily ceremonial, then the length and readability of 10-K filings should be unaffected or 10-K filings should become longer and less readable post-adoption

I also examine whether the adoption of a disclosure committee influences the level of information asymmetry and the informativeness of earnings Prior research suggests that

increased disclosure quality reduces the likelihood that investors discover and trade on private information and therefore decreases information asymmetry (Brown and Hillegeist 2007) Thus,

if disclosure committee adoption improves disclosure quality, then I expect that the level of information asymmetry will decrease post-adoption However, if disclosure committee adoption

is primarily ceremonial in nature or decreases disclosure quality, then I expect no change or an increase in information asymmetry after disclosure committee adoption Prior research also suggests that more informative disclosure can improve the informativeness of earnings (Healy et

al 1999) and can increase the amount of future earnings news reflected in current returns

(Lundholm and Myers 2002) If disclosure committee adoption improves disclosure quality, then I expect the informativeness of earnings to increase post-adoption However, if disclosure committees do not change or reduce disclosure quality, then I expect no change or a decrease in the informativeness of earnings post-adoption

In my consequences tests, I use a difference-in-differences research design and select a control sample of companies that did not adopt a disclosure committee using a propensity score matching method that controls for observable company characteristics (Rosenbaum and Rubin 1983) Specifically, I match each company in the treatment sample (i.e., each of the disclosure

committee adopters) with a company that does not adopt a disclosure committee but is as likely

to adopt given its observable characteristics

I find that companies adopting disclosure committees file longer and less readable 10-Ks post-adoption Specifically, I find that disclosure committee adoption is associated with an 8 percent increase in the number of words in the 10-K filing and an 8 percent increase in gross 10-

K file size, representing an increase in length of 6 pages of single spaced text, on average Thus, disclosure committee adoption does not improve the readability of the annual 10-K filing In addition, I find no relation between disclosure committee adoption and level of information asymmetry or between disclosure committee adoption and the informativeness of earnings Taken together, these results suggest that disclosure committee adoption is associated with longer and less readable 10-K filings, but these longer and less readable 10-K filings do not influence market outcomes

In additional analyses, I examine the relation between disclosure committee adoption and specific word types used in the 10-K filing After disclosure committee adoption, I find an increase in positive, negative, and neutral words However, I do not find a change in the

proportions of positive, negative, or neutral words to total words in the 10-K after adoption These results suggest disclosure committee adoption is associated with an increase in words in general and do not change the sentiment or tone of the 10-K filing Next, I use a subsample of companies that disclose the composition of their disclosure committee and investigate whether the composition of the disclosure committee impacts the length and complexity of the 10-K filing, information asymmetry, and earnings informativeness In these tests, I find some

evidence that the inclusion of senior management on the disclosure committee is associated with longer 10-K filings I do not find robust evidence suggesting that the inclusion of senior

management on the disclosure committee influences information asymmetry or the

informativeness of earnings Last, I examine the relation between disclosure committee adoption and the quality of internal controls over financial reporting I find an increase in likelihood of an internal control weaknesses reported in the year of disclosure committee adoption However, I also find a decrease in the likelihood of internal control weaknesses reported in the year after adoption Taken together, these results suggest that disclosure committee adoptions are a signal

of an improvement in current and future internal controls over financial reporting

My findings should be of interest to academic researchers, regulators, investors, and other stakeholders interested in the determinants and consequences of disclosure committee adoption Consistent with institutional theory, my findings suggest that the disclosure committee adoption

is primarily ceremonial in that although it affects the amount of quantity of disclosure, it does not improve disclosure quality This suggests that the investment of management time and company resources may not benefit shareholders Finally, my finding that disclosure committee adoption increases the length and decreases the readability of 10-K filings could be problematic given recent efforts by the SEC to reduce disclosure overload

The remainder of this paper is organized as follows In Section 2, I provide background information about disclosure committees, review prior literature, and develop my hypotheses In Section 3, I describe my research design Section 4 describes my sample selection and data Section 5 presents the results of my empirical tests Section 6 presents additional analyses Section 7 describes my robustness tests and Section 8 offers concluding remarks

II Background, Prior Literature, and Development of Hypotheses

Disclosure committee adoption represents a significant investment of management time and company resources In general, disclosure committees provide guidelines for disclosure and determine the appropriateness of public disclosures Specifically, disclosure committees should ensure that company filings are fair, accurate, timely, and complete by creating, reviewing, and

Typically, one of the primary responsibilities of the disclosure committee is to draft and review the annual 10-K filing The 10-K is one of the most important documents filed by a public company because it contains both summary information related to prior year performance

suggest that disclosure committees can help managers determine what information to disclose in the 10-K as well as how and when to disclose this information For example, Tysiac (2012) suggests that the disclosure committee can help improve reporting in the Management

Discussion & Analysis section of the 10-K by monitoring such items as financial covenants, regulatory and legislative changes, problems with customers and suppliers, and litigation

developments Disclosure committees can take the lead in benchmarking key disclosures against industry peers and advising management on which disclosures to include or exclude from

company filings (McCarthy and Iannaconi 2010) Moreover, disclosure committees can reduce

Appendix A Among the responsibilities listed, Cardinal Health’s Disclosure Committee has oversight of the preparation of the Company’s 10-K filing

rather than 10-K and 10-Q filings because prior research suggests that 10-K filings are more informative than 10-Q filings to investors

the complexity of financial disclosures made in 10-K filings (KPMG 2011b) Alternatively, according to an audit committee member of a company listed on the New York Stock Exchange, disclosure committees place a burden on management and on the audit committee by adding another layer of bureaucracy to the organization (Tremblay and Gendron 2011)

Companies might adopt disclosure committees to reduce agency costs by providing a signal of improved disclosure quality However, the adoption of a disclosure committee might also be a ceremonial governance choice that doesn’t improve disclosure quality or provide perceived benefits to shareholders As such, disclosure committee adoption could be consistent with institutional theory, which states that some corporate governance structures are merely symbolic and do not actually improve corporate governance (Scott 1987) In this paper, I

examine whether disclosure committee adoption provides real or perceived benefits to

shareholders by investigating whether disclosure committee adoption influences the readability

of the annual report, information asymmetry, and informativeness of earnings

Li (2008) was the first to document the effects of complex language in the 10-K filing; he finds that less readable 10-K filings (i.e., 10-Ks that are less readable) are associated with poor future performance and lower earnings persistence You and Zhang (2009) find that investor underreaction to information in 10-K filings is more pronounced for companies with lengthy 10-

Ks, and Miller (2010) finds lower overall trading volume for companies with longer and less

and analyst forecast uncertainty are greater and analyst forecasts are less accurate (Lehavy et al 2011) Moreover, experimental research also finds evidence that the readability of disclosure

impacts investors For example, Rennekamp (2012) finds that investors experience processing disfluency and rely less on disclosures when readability is poor Taken together, prior research suggests that investors and analysts find longer and more complex 10-K disclosures less useful (Libby and Emett 2014)

Because disclosure committees are often responsible for drafting and reviewing 10-K filings, I examine the impact of disclosure committee adoption on 10-K disclosure quality On the one hand, disclosure committees could improve disclosure quality by helping top

management obtain and summarize information, and draft and review disclosures Having a formalized disclosure process and including management from multiple areas within a company (e.g., legal, investor relations, accounting, and operations) could improve the readability of 10-K filings In addition, disclosure committees could improve the readability of 10-K filings by drafting disclosures that are shorter and more informative On the other hand, disclosure

committee adoption could lead to longer and less informative 10-K filings because committee members may increase the length or decrease the readability of 10-Ks because of regulatory or legal concerns or because committee members may feel obligated to increase disclosure

(regardless of content) given that such a committee has been formed Because the adoption of a disclosure committee could increase or decrease the length and/or readability of 10-K filings, my first null hypothesis is as follows:

H0a: Disclosure committee adoption does not influence the length or readability of a company’s 10-K filing

To test this hypothesis, I examine the change in 10-K length and readability following disclosure committee adoption

Many prior studies use the Fog Index, which measures readability as a function of

Loughran and McDonald (2014) find that the Fog Index is misspecified when used to measure the readability of 10-K filings because many of the “complex words” in the 10-K filing are simple business terms, rather than words that are difficult to read or understand Loughran and McDonald (2014) suggest that the file size of the 10-K filing provides a better measure of

readability because it correlates more strongly than the Fog Index with returns volatility, analyst forecast error, and analyst forecast dispersion Thus, I use the 10-K gross file size and net file size, following Loughran and McDonald (2014), to measure the length and readability of the 10-

K filing The 10-K gross file size variable is the size of the 10-K filing as posted on EDGAR and the net file size variable is the size of the 10-K filing when only textual content is included Loughran and McDonald (2014) use both file size measures to proxy for readability I also follow You and Zhang (2009) and Miller (2010) and use the number of words in the 10-K filing

as a simple proxy for 10-K length and complexity

disclosure quality reduces investor incentives for private information search, thereby decreasing information asymmetry

Because disclosure committees are charged with increasing disclosure quality in general and ensuring that disclosures are complete, accurate, and relevant, changes to disclosures made

by the disclosure committee could reduce information asymmetry by increasing the amount of information available to investors However, because disclosure committees could reduce disclosure or decrease disclosure quality, information asymmetry could increase or be unaffected

by disclosure committee adoption Stated in the null, my second hypothesis is as follows:

H0b: Disclosure committee adoption does not influence information asymmetry

To test this hypothesis, I use two common proxies for information asymmetry – bid-ask spreads and stock illiquidity – and examine the change in information asymmetry after disclosure

committee adoption

Bid-ask spreads are a common proxy for the level of information asymmetry because they capture the adverse selection problem that arises from trading in the presence of

asymmetrically informed investors (Healy et al 1999; Leuz and Verrecchia 2000) Prior

research also uses the level of stock illiquidity to proxy for information asymmetry This

research suggests that a reduction in information asymmetry resulting from an increase in

disclosure increases the liquidity of a company’s stock (Amihud and Mendelson 1986; Diamond and Verrecchia 1991) Thus, higher levels of stock illiquidity represents greater information asymmetry

Decades of accounting research investigate the returns-earnings relation (Kothari et al 2001) This research suggests that current stock returns reflect investor beliefs about current and future earnings When investors perceive current earnings to be more informative about

company value, the relation between current stock returns and current earnings surprise (i.e., the earnings response coefficient (ERC)) will be greater In addition, stock returns reflect investor

beliefs about future performance (Beaver et al 1980) such that future earnings explain a large amount of the variation in current stock returns (Collins et al 1994) Recent literature examines the impact of disclosure on the relation between current returns and current and future earnings This research suggests that companies with large increases in disclosure activity experience an increase in ERCs (Healy et al 1999), and that companies with more informative disclosures experience higher future earnings response coefficients (FERCs) (Gelb and Zarowin 2002; Lundholm and Myers 2002; Choi et al 2011)

The SEC and practitioners suggest that disclosure committees should improve the

relevance, timeliness, and credibility of reported performance Audit firms also suggest that disclosure committees should help to increase the credibility of reported earnings (PwC 2006)

If disclosure committee adoption improves disclosure quality and the relevance of earnings, then ERCs will increase following disclosure committee adoption In addition, if disclosure

committees increase the timeliness and credibility of reported earnings, then FERCs should increase post-disclosure committee adoption However, because disclosure committee adoption may be ceremonial, my third and fourth null hypotheses are as follows:

H0c: Disclosure committee adoption does not affect the market’s response to earnings news

H0d: Disclosure committee adoption does not affect the amount of future earnings news reflected in current returns

To test this hypothesis, I examine the effect of disclosure committee adoption on both ERCs and

FERCs

III Research Design

Examining the consequences of disclosure committee adoption requires controlling for factors occurring concurrently with the formation of disclosure committees One way to do this

is to establish a prediction model for disclosure committee adoption and use this to identify a propensity score matched control sample of companies that do not adopt a disclosure committee

In developing my prediction model, I acknowledge that some companies may use

disclosure committee adoption to signal improvements in disclosure quality as a means to reduce

agency costs Therefore, I expect companies with poor quality disclosures in year t-1 to be more

likely to adopt disclosure committees because this can signal to shareholders that they are

making efforts to improve current and future disclosure quality I also expect larger companies, higher-growth companies, and companies with more complex operations to be more likely to adopt disclosure committees because they can assist management in gathering financial

information and drafting and reviewing disclosures Finally, I include company profitability, leverage, age, and level of institutional ownership, as well as auditor characteristics in the year

+ α3 GROSSFILESIZEit-1 + α4 SIZEit-1 + α5 BTM it-1 + α6 M&Ait-1 + α7 FINANCINGit-1 + α8 SEGMENTSit-1 + α9 FOREIGNit-1 + α10 RET_VOLit-1 + α11 STD_SALEit-1 + α12 STD_CFOit-1 + α13 AR_INit-1 + α14 SPECIAL_ITEMSit-1 + α15 OP_CYCLEit-1 + α16 ROAit-1

+ α17 LEVERAGEit-1 + α18 AGEit-1 + α19 %INST_HOLDit-1 + α20 BIGNit-1

+ α21 AUDITOR_TENUREit-1 + αj INDUSTRYFEit-1 + αk YEARFEit-1 (1)

include governance variables in my models would result in significant sample attrition

However, in untabulated analyses, I find no difference in governance characteristics (i.e., audit committee size, percentage of outsiders on the audit committee, and percentage of outsiders on the board of directors) between treatment and control companies either before or after disclosure committee adoption

where:

disclosure committee during year t, zero otherwise;

a restatement in year t-1 due to misstatements in prior year

financial statements, zero otherwise (Audit Analytics);

a material weakness in internal control over financial

reporting under SOX Section 302 in year t-1, zero

otherwise (Audit Analytics);

10-K filing in year t-1 (available from Bill McDonald’s

website)8;

of common equity (Compustat: CEQ) divided by the market value of equity (Compustat: CHSO x PRCC_F);

a merger or acquisition in year t-1, and zero otherwise (Compustat: SALE_FN);

one and the number of shares outstanding (Compustat:

CHSO) increased by at least 10 percent during the year, or

if M&A is not equal to one and long-term debt increased by

at least 20 percent during the year, zero otherwise;

year t-1 (Compustat Segment File);

t-1 (Compustat Segment File);

http://www3.nd.edu/~mcdonald/Word_Lists.html

STD_SALEt-1 = the standard deviation of sales (Compustat: SALE) for the

prior three years ending in year t-1;

(Compustat: OANCF) for the prior three years ending in year t-1;

and inventory (Compustat: INVT) in total assets (Compustat: AT) in year t-1;

(Compustat: AT) in year t-1;

the number of days sales in accounts receivable plus the number of days sales in inventory;

extraordinary items (Compustat: IB) scaled by total assets (Compustat: AT);

AT);

years during which the company reports total assets on

Compustat (Compustat: AT) greater than zero;

(Thomson Reuters Institutional Holdings);

by a Big N auditor in year t-1, zero otherwise;

French (1997) industry classifications; and

B Propensity Score Matching and Difference-in-Differences

Most sample companies disclose the presence of a disclosure committee in their 10-Q, 10-K, or DEF 14A filings In each instance, I ensure that the disclosure committee was formed and operating before the fiscal year-end date used to measure my consequences (outcome)

variables I estimate Equation (1) using the disclosure committee adoption year for my treatment companies (i.e., I remove the pre-and post-adoption years for companies that adopt a disclosure committee) and all of the available company-years in the sample period for companies that did

company because non-disclosure committee adopters can appear in the sample multiple times I use the results from this estimation to propensity score match each treatment company to a control company in the same year using the caliper method (without replacement) This method allows me to control for observable company characteristics in my consequences analyses

(Rosenbaum and Rubin 1983) The sets of treatment companies and matched control companies comprise my sample companies for all remaining analyses

Because I am interested in the effects of disclosure committee adoption, I use a

difference-in-differences research design Specifically, my sample includes the year before and the year after disclosure committee adoption for both treatment companies (i.e., those adopting disclosure committees) and the propensity score matched control companies (i.e., those not

time-variant characteristics that could influence the outcome variable

adopted the committee in any of the years of my sample

parallel trends assumption that is important for difference-in-differences research designs

(Roberts and Whited 2011)

C Readability Tests

To examine the relation between disclosure committee adoption and readability of the

10-K filing, I estimate the following cross-sectional regression using ordinary least squares (OLS):

READ_VARit = β0 + β1 DCit + β2 AFTERit + β3 DCit x AFTERit + β4 SPECIAL_ITEMSit + β5 M&Ait

+ β6 FINANCINGit + β7 RET_VOL3it + β8 LOSSDit + β9 STD_INCOMEit + β10 DELAWAREit + β11 %INST_HOLDit + β12 SIZEit + β13 BTMit

where:

disclosure committee, and zero otherwise;

SPECIAL_ITEMSt = special items (Compustat: SPI) scaled by total assets

(Compustat: AT);

merger or acquisition during the year, and zero otherwise

(Compustat: SALE_FN);

FINANCINGt = an indicator variable set to one if M&A is not equal to one and

the number of shares outstanding (Compustat: CHSO) increased by

at least 10 percent during the year, or if M&A is not equal to one

and long-term debt increased by at least 20 percent during the year, zero otherwise;

years (CRSP: RET);

the matched treatment observation adopted a disclosure committee

LOSSDt = an indicator variable set to one if the company reports a loss

during the year (Compustat: NI), zero otherwise;

three years;

the state of Delaware, zero otherwise;

Institutional Holdings database);

(Compustat: CEQ) divided by the market value of equity (Compustat: CHSO x PRCC_F);

Segment File);

Segment File);

all other variables as previously defined

adopter and control companies, from before to after disclosure committee adoption

Equation (2) includes controls from prior research that are associated with the length and readability of 10-K filings Specifically, I control for the magnitude of special items

(BTMt), the number of foreign operating segments (FOREIGNt), and the number of business

in 10-K length and readability over time and across industries

Next, I examine the relation between disclosure committee adoption and information asymmetry I use the following general model for these tests:

IA_VARit = γ0 + γ1 DCit + γ2 AFTERit + γ3 DCit x AFTERit + γ4 TURNOVERit-1 + γ5 MVEit-1

where:

BA_SPREAD or ILLIQUIDITY;

as the average high-low estimator over the period from 9 months before the fiscal year-end through 3 months after the fiscal year-end;

ILLIQUIDITYt = the daily ratio of absolute stock returns scaled by the dollar

volume of trading (following Amihud (2002)), averaged over the period starting 9 months before the fiscal year-end through 3 months after the fiscal year-end;

year (CRSP: RET);

GROSSFILESIZEt = the gross file size of the 10-K filing (from McDonald’s website);

all other variables as previously defined

γ3 is the coefficient of interest It represents the difference-in-differences in information

asymmetry, between adopter and control companies, from before to after disclosure committee adoption

My tests utilize two common measures of information asymmetry The first measure is the bid-ask spread I use the high-low spread estimator from Corwin and Schultz (2012) because

it is easy to calculate and generally outperforms other bid-ask spread estimators based on the

daily bid-ask spread measure from three months before the fiscal year-end through nine months after the fiscal year-end to ensure that information in the 10-K filings is priced

The second measure of information asymmetry is a measure of illiquidity

ratio of absolute stock return to the dollar value of trading volume from three months before the fiscal year-end through nine months after the fiscal year-end to capture information in the 10-K

depth.14

stock’s fundamental volatility The high-low spread estimator is a reasonable proxy for

information asymmetry because daily high prices are usually buyer-initiated and daily low prices are almost always seller initiated Thus, the ratio of high-to-low prices on a given day represents both the stock’s fundamental volatility and its bid-ask spread Another benefit of using the Corwin and Schultz (2012) estimator is that it avoids need to match individual buys and sells with TAQ data, which computationally intensive and time consuming

Paul Schultz’s website at http://www3.nd.edu/~scorwin/HILOW_Estimator_Sample_002.sas

unchanged

In my information asymmetry tests, I include controls for lagged company size

prior literature (Cheng et al 2013; Daske et al 2013) Because I am interested in how disclosure committee adoption influences information asymmetry overall, I also control for the readability

indicator variables

Next, I examine the impact of disclosure committee adoption on the informativeness of earnings using the following model:

RETURNSit = δ0 + δ1 EARNINGSit-1 + δ2 EARNINGSit + δ3 DCit + δ4 AFTERit + δ5 DCit x AFTERit

+ δ6 DCit x EARNINGSit-1 + δ7 DCit x EARNINGSit + δ8 AFTERit x EARNINGSit-1 + δ9 AFTERit x EARNINGSit + δ10 DCit x AFTERit x EARNINGSit-1 + δ11 DCit x AFTERit x EARNINGSit

+ δl (CONTROLS x EARNINGSit) + δm (CONTROLS x DCit)

where:

of year t;

items (Compustat: IB), scaled by the market value of common equity (Compustat: PRCC_F x CSHO) at the beginning of year t;

all other variables as previously defined

I estimate Equation (4) using ordinary least squares (OLS) regression with robust standard errors clustered by company (Petersen 2009)

relation between current earnings and current returns A positive coefficient estimate would indicate that disclosure committee adoption increases the informativeness of current earnings

Finally, I examine the impact of disclosure committee adoption on the relevance and timeliness of reported earnings Following prior literature (e.g., Lundholm and Myers (2002), Ettredge et al (2005), Tucker and Zarowin (2006), and Drake et al (2014)), I use the following model for these tests:

RETURNSit = λ0 + λ1 EARNINGSit-1 + λ2 EARNINGSit + λ3 EARNINGS3it+1 to t+3

+ λ4 RETURNS3it+1 to t+3 + λ5 DCit + λ6 AFTERit + λ7 DCit x AFTERit + λ8 DCit x EARNINGSit-1 + λ9 DCit x EARNINGSit

+ λ10 DCit x EARNINGS3it+1 to t+3 + λ11 DCit x RETURNS3it+1 to t+3 + λ12 AFTERit x EARNINGSit-1 + λ12 AFTERit x EARNINGSit + λ14 AFTERit x EARNINGS3it+1 to t+3 + λ15 AFTERit x RETURNS3it+1 to t+3

+ λ16 DCit x AFTERit x EARNINGSit-1 + λ17 DCit x AFTERit x EARNINGSit + λ18 DCit x AFTERit x EARNINGS3it+1 to t+3

+ λ19 DCit x AFTERit x RETURNS3it+1 to t+3 + λjCONTROLS

+ λm (CONTROLS x EARNINGS3it+1 to it+3)

where:

EARNINGS3t+1 to t+3 = the sum of income available to common shareholders before

extraordinary items (Compustat: IB) for the years t+1 through t+3, scaled by the market value of equity (Compustat: PRCC_F x CSHO) at the beginning of year t;

RETURNS3t+1 to t+3 = the buy-and-hold return for the fiscal years t+1 through t+3,

measured from the beginning of fiscal year t+1;

analyst following (NUMBER_ANALYSTSt), asset growth

all other variables as previously defined

I estimate Equation (5) using OLS regression with robust standard errors clustered by company (Petersen 2009)

the FERC A positive coefficient estimate would indicate that disclosure committee adoption increases the amount of future earnings news reflected in current returns The control variables follow Lundholm and Myers (2002), Ettredge et al (2005), and Tucker and Zarowin (2006) and have been shown to influence the associations between current returns and current and future earnings

My sample consists of a treatment group and a control group The treatment group is comprised of all companies that disclose the presence of a disclosure committee in EDGAR filings and the control group is a propensity score matched sample of companies that do not adopt a disclosure committee (as described in Section 3) I collect internal control weakness data, auditor data, and financial statement misstatement data from Audit Analytics, financial data from the Compustat Fundamental Annual database, returns data from CRSP, analyst data from I/B/E/S, and institutional holdings data from Thomson Reuters To minimize the impact of extreme observations, I winsorize all continuous variables at 1 and 99 percent Appendix B provides details about the variable construction

I hand collect disclosure committee adoptions from EDGAR filings (i.e., 10-K, 10-Q, DEF 14A, DEFA 14A, PRE 14A, PREA 14A, 8-K, S-1, 40-F, 20-F, and 424B filings) from 2002

their adoption For my determinants, readability, and information asymmetry tests, my sample period ends in 2012, but for my earnings informativeness tests, my sample period ends in 2009 because I require three future years of earnings and returns I perform keyword searches of each filing for evidence of the presence of a disclosure committee I then read these filings and

This search yields a sample of 764 unique companies disclosing the presence of a disclosure

Table 1 contains the details about my treatment and control samples Panel A reveals that many sample companies adopted a disclosure committee immediately after the passage of SOX Panel B presents the number of treatment and control observations for by Fama and French

DEF 14A filings I also perform searches of EDGAR filings for 1999 through 2001 to

investigate whether companies adopted disclosure committees before the passage of SOX but do not find evidence of disclosure committee adoptions

Disclosure and Controls Committee, Financial Reporting Committee, and Financial Reporting and Disclosure Committee I perform a keyword search using variations of these examples, including identifying all observations with the words disclosure and committee within three words of each other

disclosure committees on company websites and I find that 390 of the S&P 1500 companies disclose the presence of a disclosure committee I find that 333 of these companies were

identified by my EDGAR searches so my EDGAR searches capture about 85 percent of

disclosure committee companies In all analyses, I exclude the 57 companies identified in

Google searches but not in EDGAR searches as having a disclosure committee because I cannot reliably determine the disclosure committee adoption date for these companies

(1997) industry classification My sample companies represent a number of industries, with companies from the Business Services industry comprising the greatest proportion (at 15.7 percent of the sample) and companies from the Agriculture, Tobacco Products, Shipbuilding and Railroad Equipment, Defense, Precious Metals, Coal, and Shipping Containers industries

representing the smallest proportion (at 0.10 percent)

[Insert Table 1 Here]

In Table 2, I present descriptive statistics for my treatment and control samples I find that, on average, treatment companies have less of their assets in accounts receivable and

adoption I also find that treatment companies are more likely than control companies to report a

[Insert Table 2 Here]

Table 3 presents the results of analyses investigating the determinants of disclosure committee adoption The area under the receiver-operating characteristic (ROC) curve is greater than 0.70, suggesting that my model has acceptable discriminatory power (Hosmer and

adoption and control group companies in the pre- and post-adoption periods I do not find significant differences in losses between these groups in the pre- or post-adoption years

Lemeshow 2002) The results suggest that companies with material weaknesses in internal

disclosure committees Additionally, companies with less readable 10-K filings

[Insert Table 3 Here]

Prior research suggests that the inclusion of fixed effects in nonlinear models creates an incidental parameters problem and that the unconditional maximum likelihood estimator is biased (Neyman and Scott 1948; Lancaster 2000), but studies using simulations find that this bias becomes smaller as the group size increases and Katz (2001) suggests that this bias becomes negligible when the group size exceeds 15 To investigate the robustness of my results given the potential for an incidental parameters problem, in Table 4, I re-estimate Equation (1) using industry-fixed effects based on one digit Standard Industry Classification (SIC) codes and find that the coefficient signs and significance are unchanged

[Insert Table 4 Here]

Next, I use the propensity scores from the determinants model to match each treatment company (i.e., disclosure committee adopter) to a control company that did not adopt a

A of Table 5, I am able to match 732 of the 764 disclosure committee companies In Panel B, I

committee during the sample period

of the standard deviation of the propensity scores The standard deviation of the propensity score from my determinants model is 0.02 so I use a caliper of 0.005

present univariate statistics for the sample of treatment and control companies used in my

determinants test Columns (1) and (2) contain the means and standard deviations for my control group and columns (3) and (4) contain the means and standard deviations for my treatment group In column (5), I present p-values for tests of differences in means I do not find

statistically different mean values for any of the variables used in my determinants model,

suggesting that my match procedure successfully matches treatment companies with similar control companies

[Insert Table 5 Here]

In column (6), I report normalized differences (i.e., the differences in means scaled by the average of the two within-group standard deviations) Normalized differences are a more

reliable way to assess covariate balance because they are invariant to sample size (Imbens and Wooldridge 2009; Abadie and Imbens 2011) Imbens and Wooldridge (2009) suggest that normalized differences of greater than 0.250 in absolute value represent less than acceptable covariate balance I find that all normalized differences are below 0.250 in absolute value, suggesting that the match between treatment and control companies provides acceptable

covariate balance

Governance data in machine-readable form is unavailable for half of my sample

companies so include governance variables in my models would result in significant sample attrition As such, I do not include board and audit committee related variables in my models I investigate differences in governance characteristics for those observations that have available governance data In Panels A and B of Table 6, I find no difference in governance

characteristics (i.e., audit committee size, percentage of outsiders on the audit committee, and

percentage of outsiders on the board of directors) between treatment and control companies either before or after disclosure committee adoption

[Insert Table 6 Here]

Table 7 describes the sample selection for my multivariate tests For each test, I require that the treatment and control companies have data available for the pre- and post-adoption years I lose 72 treatment companies (and their matching control companies) for my readability tests, 117 treatment companies for my information asymmetry tests, and 363 treatment

companies for my earnings informativeness tests because of missing data for the treatment or

[Insert Table 7 Here]

The results of my readability tests appear in Tables 8, 9, and 10 In Table 8, I investigate

between treatment control companies in the pre-disclosure committee adoption period I find

treatment companies experiencing an increase in 10-K length relative to control companies following the adoption of the disclosure committee In addition, I find that the joint test of the

treatment companies having longer 10-K filings than control companies in the post-adoption period

data to calculate analyst following and I require three years of future earnings and returns data

[Insert Table 8 Here]

Next, I examine the relation between disclosure committee adoption and the readability

AFTERt is positive and significant (p < 0.05), consistent with disclosure committee adoption decreasing the readability of the 10-K filing I also find that the joint test of the coefficients on

companies having less readable 10-K filings than control companies in the post-adoption period

[Insert Table 9 Here]

DC is insignificant, consistent with no difference in 10-K readability between treatment and

control companies in the pre-adoption period I find a positive and significant relation (p < 0.05)

decreasing after disclosure committee adoption I also find that the joint test of the coefficients

having less readable 10-K filings than control companies in the post-adoption period (p < 0.05)

[Insert Table 10 Here]

Taken together, my tests reveal that disclosure committee adoption is associated with an

effect would need to be to influence the statistical significance of my results For my #WORDS

be needed to change my statistical inferences For the GROSSFILESIZE and NETFILESIZE

C Information Asymmetry Tests

Tables 11 and 12 present the results of my information asymmetry tests If disclosure committees decrease the level of information asymmetry, I expect that bid-ask spreads

committee adoption Alternatively, if the disclosure committee does not increase disclosure quality and the amount of information available to investors, bid-ask spreads and market

illiquidity could be unchanged or could increase

that disclosure committee adoption does not change the level of information asymmetry as captured by bid-ask spreads

[Insert Table 11 Here]

Similarly, in Table 12, I find no relation between disclosure committee adoption and the

[Insert Table 12 Here]

Next, I examine whether disclosure committee adoption impacts the informativeness of earnings If a disclosure committee improves disclosure quality related to current or future earnings news, then I expect that the ERC and FERC will be larger post-adoption The results of

tests, the effect would need to be 1.20 and 1.28 times the effects of DC x AFTER to change my

statistical inferences

information asymmetry and readability In these tests, I include an interaction between DC, AFTER, and GROSSFILESIZE I find that the coefficient on DC x AFTER x GROSSFILESIZE

is insignificant, suggesting that disclosure committee adoption does not influence the relation between information asymmetry and readability

these tests are tabulated in Tables 13 and 14 In Table 13, consistent with prior research, I find a negative relation between past earnings and current returns and a positive relation between

is insignificant This result suggests that disclosure committee adoption does not influence the relation between current returns and current earnings Next, I examine whether disclosure

committee adoption influences the extent to which future earnings are reflected in current

returns

[Insert Table 13 Here]

In Table 14, consistent with prior literature, I find negative relations between past

AFTERt x EARNINGSt and DCt x AFTERt x EARNINGS3t+1 to t+3 are insignificant These results suggest that disclosure committee adoption does not change the informativeness of earnings or

[Insert Table 14 Here]

I also examine the impact of disclosure committee adoption on the specific words in 10-K filings Specifically, I look at the number of positive, negative, and neutral words reported in a

et al (2005)