ubmit your analysis on the reasons and what could make the 10-K more useful to the user

ACCOUNTING HORIZONS American Accounting Association Vol. 30, No. 1 DOI: 10.2308/acch-51240 March 2016 pp. 1–21

Why are 10-K Filings So Long?

Richard A. Cazier The University of Texas at El Paso

Ray J. Pfeiffer Texas Christian University

SYNOPSIS: This study provides evidence on relative magnitudes of factors that appear to drive 10-K length. We employ statistical analysis and text analysis software to partition 10-K length into the portions explained by each of

three fundamental determinants: (1) firms’ operating complexity, (2) disclosure redundancy, and (3) residual

disclosure. Our primary analyses shed light on the relative magnitudes of each of these components and on the

extent to which each varies across firms. Disclosure redundancy and operating complexity explain roughly equal

amounts of variation in 10-K length within our sample. However, 10-K length unexplained by redundancy or firms’

operating complexity (residual disclosure) accounts for the largest degree of variation in 10-K length. Our results are

consistent with the notion that a substantial amount of disclosure volume contained in 10-K reports is attributable to

managerial discretion in how firms respond to mandatory disclosure requirements. Our study expands prior literature

that has focused largely on the consequences of 10-K length and provides important insights for policy makers and

regulators seeking to improve disclosure requirements.

Keywords: 10-K length; disclosure volume; redundancy.

Data Availability: Data available upon request.

INTRODUCTION

A n ongoing discussion among regulators and practitioners centers on what can and should be done to address the

increasing volume of firms’ financial disclosures. Little disagreement exists among observers that the volume of

financial disclosure has expanded considerably. In their 2011 analysis of 25 Fortune 500 companies’ annual reports,

the Financial Executives Research Foundation and KPMG document a 16 percent increase in 10-K length and a 28 percent

increase in the length of the notes to financial statements over the period 2004–2010 (KPMG 2011). In this paper, we provide

empirical evidence regarding the determinants of 10-K disclosure volume to help inform the discussion of how regulators

might address this issue.

Various financial reporting stakeholders increasingly decry the current state of mandatory financial disclosure on the part

of business enterprises. 1

In its July 12, 2012 FASB in Focus publication, the FASB summarizes the concerns of stakeholders as

follows:

Many stakeholders have expressed concerns about the relevance and sheer volume of information in notes to financial

statements, and that some information is either missing or difficult to find. (FASB 2012b; emphasis added)

The authors acknowledge helpful comments and suggestions from workshop participants at Lehigh University, two anonymous reviewers, and Paul A. Griffin (editor). We are also grateful to Texas Christian University for providing financial support for this project.

Editor’s note: Accepted by Paul A. Griffin.

Submitted: June 2014 Accepted: July 2015

Published Online: July 2015

1 Examples noted by the FASB in their July 12 Discussion Paper: Disclosure Framework (FASB 2012a, paragraph 1.38) include the 2001 report by the FASB’s Business Reporting Research Project (FASB 2001); the Investors Technical Advisory Committee’s 2007 Agenda Request: available at: http:// www.fasb.org/cs/ContentServer?c¼Document_C&pagename¼FASB%2FDocument_C%2F&cid¼1175801635556 (last accessed July 13, 2015); the SEC’s (2008) Advisory Committee on Improvements to Financial Reporting’s Final Report; the Financial Executives Research Foundation’s 2011 report, Disclosure Overload and Complexity: Hidden in Plain Sight (KPMG 2011), as well as others.

1

In a 2013 speech to the National Association of Corporate Directors (NACD) Board Leadership Conference, SEC

Chairman Mary Jo White, in calling for a ‘‘meaningful review of our disclosure requirements,’’ similarly pointed to disclosure volume and its potentially harmful effects on investors:

I am raising the question here and internally at the SEC as to whether investors need and are optimally served by the

detailed and lengthy disclosures about all of the topics that companies currently provide in the reports they are

required to prepare and file with us. When disclosure gets to be ‘‘too much’’ or strays from its core purpose, it could lead to what some have called ‘‘information overload’’—a phenomenon in which ever-increasing amounts of disclosure make it difficult for an investor to wade through the volume of information she receives to ferret out the

information that is most relevant. (White 2013)

Chairman White goes on to ask:

Are our rules the sole or primary cause of potential disclosure overload or do other sources contribute to it? Or said

another way, are changes to our disclosure requirements the only way to improve the quality of disclosure? We should

consider all sources that may be contributing to the length and complexity of disclosure.

Questions such as those posed by Chairman White and those that underlie the FASB’s Disclosure Framework project are

vitally important to determine what is to be done to address the issue of disclosure volume. Although prior research documents

a detrimental effect of 10-K length on investors’ ability to process relevant information (e.g., Miller 2010; You and Zhang

2009), little systematic evidence exists on the factors that contribute to 10-K length. Our study focuses squarely on providing

such evidence so that policy makers can better assess the likely impact various regulatory initiatives may have on 10-K length.

We begin by framing the factors that contribute to 10-K length as:

(1) The complexity of the firm’s business transactions, which we label ‘‘operating complexity’’; 2

(2) Redundancy in SEC and GAAP disclosure requirements that may result in the same information being repeated in

various sections of the 10-K; and

(3) Firms’ provision of ‘‘residual’’ disclosure, which is unassociated with the first two factors above. 3

We examine a sample of 10-Ks representing 620 firm-years over the ten-year period 2003 through 2012 and find firms’

underlying operating complexity, disclosure redundancy, and residual disclosure each contribute significantly to variation in

10-K length. Our analysis suggests disclosure redundancy and operating complexity explain a roughly equal amount of

variation in 10-K length within our sample. Specifically, they each account for a difference of approximately 15,000 words in

length between the median firm in the lowest quintile of 10-K length and the median firm in the highest quintile of 10-K length

in our sample. However, 10-K length unexplained by redundancy or firms’ operating complexity (i.e., residual disclosure)

appears responsible for a slightly larger degree of variation in 10-K length across the top and bottom quintiles (about 18,000

words) than do either operating complexity or redundancy.

We next provide further evidence on the extent of redundancy in 10-K filings. Given that the SEC permits a significant

amount of disclosure to be incorporated by reference to the notes to the financial statements, providing redundant information is

at least partially a discretionary reporting choice by management. We test which firm characteristics are associated with the

provision of more redundancy in 10-K disclosures. We find redundant disclosure has a strong positive association with firm

size and leverage. Redundancy is decreasing in R&D intensity but increasing in special items. We find some evidence that total

redundant disclosure has increased over the sample period. However, increases in total 10-K disclosure have outpaced

increases in redundant content, such that redundant disclosures have declined as a proportion of total 10-K disclosure.

Finally, we seek to determine which components of 10-K length—operating complexity, redundancy, or residual

disclosure—are responsible for investors’ information-processing problems documented by prior research (e.g., You and Zhang

2009; Miller 2010; Lawrence 2013). Specifically, we look for evidence of return drift following the filing of longer 10-K

reports (You and Zhang 2009). Although we are unable to replicate You and Zhang’s (2009) overall result that longer 10-Ks

are associated with subsequent return drift, we do find some evidence of return drift after partitioning 10-K length into its three

components. We find that 10-K length attributable to residual disclosure is associated with return drift six months following the

10-K filing date. However, we find no evidence that 10-K length attributable to operating complexity or disclosure redundancy

2 In this study, the term ‘‘operating complexity’’ refers to all business transactions including those classified as operating, investing, and financing activities in the financial statements.

3 Increases in disclosure volume over time can also be affected by changes in disclosure regulation. However, the analysis in our study is focused upon comparisons of disclosure volume across firms. In the cross-section, we interpret the relationship between operating complexity, redundancy, and disclosure volume to be conditioned on existing disclosure regulation. One extension of our work would be to consider the effects of changes in disclosure regulation over time.

2 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

is associated with returns subsequent to the 10-K filing. This suggests that the residual disclosure component of 10-K length

can indeed impair market participants’ pricing of securities.

This paper is among the first to provide descriptive evidence on the determinants of 10-K length. This is important because

although prior research documents that 10-K length reflects complexity (e.g., Lehavy, Li, and Merkley 2011; You and Zhang

2009; Miller 2010), no prior evidence exists regarding whether that complexity is primarily a result of the firm’s underlying

operating complexity or simply its disclosure complexity. Our results indicate that cross-sectional variation in 10-K length is

likely more determined by managerial discretion than by firms’ operating characteristics. In addition, we are among the first to

provide meaningful descriptive evidence on the impact of disclosure redundancy on total 10-K length.

We believe these results provide potentially useful insights on an issue of current relevance for financial reporting policy.

Enhanced understanding of the relative importance of various drivers of length will allow the SEC and FASB to take a more

targeted and likely more effective approach to improving existing disclosure requirements. For instance, our evidence on the

extent to which disclosure volume varies in the cross-section after holding firm characteristics constant highlights the concern

that this disclosure may be excessive in many cases. Standard setters may want to investigate whether the significant variation

in disclosure practices in the 10-K reflects a need for greater standardization in disclosure requirements.

In addition, our evidence on the extent to which redundancy impacts overall length provides further impetus for standard

setters to respond to calls for streamlined disclosure. Market participants have proposed several ways in which standard setters

may be able to reduce redundancy, such as providing more concrete guidance on the extent to which cross-referencing between

the notes and other 10-K sections is permissible. In addition, our research supports calls for greater coordination between the

FASB and the SEC to identify and eliminate overlapping disclosure requirements to reduce the volume of redundant

disclosures in the 10-K. However, we believe further research examining the content and determinants of redundant 10-K

disclosures is necessary to shed greater light on the extent to which these redundancies enhance or inhibit the effectiveness of

financial communication in the 10-K.

INSTITUTIONAL BACKGROUND AND RELATED RESEARCH

The 10-K is only one firm-issued disclosure among many, including those seen as often conveying more timely

information such as earnings announcements and press releases. However, a substantial body of research confirms that 10-Ks

contain price-relevant information that investors trade on (e.g., Huddart, Ke, and Shi 2007; Griffin 2003; You and Zhang 2009).

A recent stream of research finds that investors’ ability to process relevant information in the 10-K is impaired by disclosure

complexity (e.g., Lehavy et al. 2011; Rennekamp 2012). Disclosure volume is an important component of that complexity

(e.g., You and Zhang 2009; Miller 2010). We frame 10-K length as a function of a firm’s underlying operational complexity,

voluntary or mandated disclosure redundancy, and residual disclosure. We discuss each of these three components below.

Operational Complexity and 10-K Disclosure Volume

The Securities and Exchange Commission regulates the format and content of firms’ annual reports (10-K) under

Regulation S-K. The SEC’s Regulation S-X prescribes rules related specifically to the financial statements and related notes. In

addition, the FASB requires mandatory footnote disclosure of items relating to a reporting entity’s operations. These reporting

requirements mandate the disclosure of information related to material information in various parts of the 10-K. For instance,

Item 1 of the 10-K requires firms to describe their business and report on material risk factors the business faces. Item 2 requires

firms to report on their properties, Item 3 requires reporting on legal proceedings, etc. Because firms with greater operational

complexity are expected to have more transactions, subsidiaries, and events on which to report, a reasonable inference is that

firm operating complexity is positively correlated to 10-K length. In addition, certain relatively complex operating, financing,

and investing choices by firms involve accounting issues for which there tends to be more extensive required disclosures, such

as derivatives, complex tax arrangements, share-based compensation arrangements, pensions and post-retirement benefits, and

other financial instruments.

Redundant Disclosure Requirements and 10-K Disclosure Volume

Critics of the current reporting regime point out that significant overlap exists between the disclosure requirements

mandated by the SEC and those mandated by the FASB. For instance, significant accounting policies are required to be

reported both in the financial statement notes and in Item 7 of the 10-K (Management’s Discussion and Analysis [MD&A]).

Similarly, redundancies exist in the reporting for income tax disclosures, related party transactions, segment information, and a

variety of other areas (FASB 2001). Information regarding risk factors, business description, and legal proceedings are often

discussed in both the notes and in other sections of the 10-K (KPMG 2011).

Why are 10-K Filings So Long? 3

Accounting Horizons Volume 30, Number 1, 2016

Regulators and practitioners have expressed concern that redundant disclosure requirements add unnecessary length and

complexity to the 10-K (e.g., White 2013). The SEC reports that its staff is identifying ways to eliminate redundant disclosure

requirements (Securities and Exchange Commission [SEC] 2012). 4

However, the SEC also permits significant cross-

referencing between the notes to the financial statements contained in Item 8 and other sections of the 10-K in lieu of presenting

duplicate information (U.S. Code of Federal Regulations 2014, title 17, sec. 229). Our reading of 10-Ks in our sample suggests

that firms vary significantly in the extent to which they present redundant information rather than cross-referencing other

sections of the 10-K. We provide a representative example in Appendix A, which displays MD&A disclosure of recent

accounting pronouncements for two sample firms. The first uses three paragraphs to repeat verbatim information found in the

notes to the financial statements. We emphasize this represents only a small fraction of the total amount of redundancy in this

firm’s 10-K. The second company uses a single sentence to refer the reader to Note 1 of the financial statements to find the

relevant information. Thus, it appears at least some of the redundancy found in 10-K reports is due to managerial discretion.

The 2011 KPMG survey on disclosure overload found that several respondents wanted an interpretive release from the

SEC providing further guidance on the permissibility of cross-referencing redundant disclosures. This suggests that one reason

for variation in firm’s reporting of redundant information is uncertainty regarding exactly how much cross-referencing is

permissible.

Redundancy may be helpful to investors if providing the same information in different sections of the 10-K provides

context for proximate information to be communicated more clearly. Also, to the extent that differentially skilled financial

statement users might look primarily to the notes to the financial statements versus the MD&A, presenting information in both

places (redundantly) potentially benefits users by having information accessible where they expect to find it. Li (2014) reports

that 10-Ks with greater disclosure redundancy are associated with larger stock price reactions around the filing date. She

concludes that disclosure redundancy improves the transparency of relevant financial information. However, casual observation

of redundant content in our sample indicates that potentially less material disclosures are repeated in many cases. Repetition of

less material disclosures may make the most important information more difficult to discern.

Residual Disclosure

Disclosure volume unassociated with firms’ operating structure or redundant disclosures stem primarily from discretionary

managerial choices, which may be associated with management’s perception of the firm’s litigation or other risks, shareholder

demand, or simply management’s idiosyncratic disclosure ‘‘style’’ (e.g., Bertrand and Schoar 2003). In many cases, these

residual disclosures likely include information that critics allege is irrelevant to shareholders (e.g., KPMG 2011; FASB 2012a).

For instance, concerns regarding shareholder litigation or regulatory intervention may lead firms to over-disclose to ensure they

cannot be held liable for not disclosing sufficient information to investors or regulators (Institute of Chartered Accountants in

England and Wales [ICAEW] 2013). In addition, academic literature suggests some disclosure length may be attributable to

managers purposefully seeking to obfuscate information in the 10-K, possibly to mask poor performance (e.g., Li 2008;

Rogers, Van Buskirk, and Zechman 2011). In addition, firms may vary the level of useful voluntary disclosure depending on

factors such as proprietary costs of disclosure.

Examples of generic disclosures are commonly found in risk factor disclosures and in firms’ disclosure of significant

accounting policies. In their 2011 report on disclosure issues, KPMG identifies a representative example in one company’s

disclosure of risk factors: ‘‘Declining economic conditions could adversely affect our results of operations and financial

condition’’ (KPMG 2011, 13). Similarly, companies often report on recent FASB pronouncements or policy changes that apply

to all companies and that are not specific to the company’s own accounting choices.

Finally, because reporting requirements are generally uniform across all industries, and because reporting requirements are

seldom scaled back after they have outlived their usefulness, many corporations report on issues that may no longer be or never

were relevant for their own business. Examples may include mandatory reporting of the historical low and high closing price of

a company’s common stock, the dilution disclosure, and the ratio of earnings to fixed charges (White 2013).

Empirically, we capture this additional disclosure as the part of total 10-K length not represented by our estimates of

complexity or redundancy. We label this component of total length ‘‘residual’’ disclosure. In addition to any irrelevant or

immaterial disclosures, residual disclosure may also capture the presence of any voluntary disclosures firms provide that are

4 Paragraphs BC4–BC6 of the FASB’s March 4, 2014 Exposure Draft on the Conceptual Framework, Chapter 8: Notes to Financial Statements, discuss respondents’ calls for collaboration with the SEC in reducing overlapping disclosure requirements (FASB 2014). However, the exposure draft does not directly reflect such a collaboration. Rather this is largely a FASB-only effort. As explained in these paragraphs, the FASB is concerned about preparers who are not governed by the SEC and about information that might be less complete or less carefully scrutinized in an SEC disclosure than in financial statement notes.

4 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

intended to be useful to investors. Thus, we do not assume that all disclosures unexplained by operating complexity or

disclosure redundancy are necessarily problematic for investors.

SAMPLE CONSTRUCTION

To provide evidence on the determinants of 10-K length, we start with the population of all Compustat firms-years

covering the ten-year period from 2003 to 2012. Because some of our analyses replicate findings of You and Zhang (2009), we

incorporate their sample selection criteria by eliminating 65,736 observations with closing share prices less than $1 dollar or

total market cap less than $200 million. We exclude 4,606 observations missing firm identifiers on Compustat, and another

3,626 observations that are lost in the merge between Compustat and CRSP. We eliminate 3,758 observations with insufficient

return data on CRSP, 2,636 observations with missing earnings announcement dates, 247 observations where the earnings

announcement date is more than 120 days after the fiscal year end, and 241 observations missing SIC data. Finally, we exclude

5,032 foreign issuers because these firms do not file 10-K reports. These sample selection criteria result in 24,462 firm-years

remaining during our sample period.

Because our analysis involves a substantial amount of hand collection of 10-K information, we further narrow our sample

by randomly selecting 650 firm-years from the pool of 24,462 remaining firm-years. We find that 21 of these observations are

either missing 10-Ks on the SEC website or issued a non-standard 10-K that complicates the measurement of 10-K length. An

additional nine of these firm-year observations are missing Compustat data required for constructing the variables in our study.

Thus, our final sample consists of 620 firm-years representing 570 unique firms from the sample period 2003–2012. Table 1

depicts the effect of each of our sample selection criteria on our final sample.

Although this sample is smaller than those examined in studies that parse 10-Ks using computer programs such as PERL

or Python, we note that our hand-collected method helps eliminate measurement errors these programs may make when

working with 10-Ks with non-standard formats. Non-standard formats include 10-Ks incorporating information by reference to

the annual report and 10-Ks with information typically contained in Item 8 included in Item 15 instead.

MODELING THE COMPONENTS OF 10-K LENGTH

Following prior studies (e.g., Lehavy et al. 2011; You and Zhang 2009; Loughran and McDonald 2011; Miller 2010), we

measure 10-K disclosure volume using the number of words in the 10-K filing (TOTAL_LENGTH). We decompose 10-K length into the portion attributable to operating complexity, disclosure redundancy, and residual disclosure as described in the

following subsections.

TABLE 1

Sample Selection

Firm-Years Omitted

Firm-Years Remaining

Start with all Compustat firm-years between 2003 and 2012, inclusive. 110,344

Eliminate:

Observations with closing share price less than $1 or total market cap less than $200 million (65,736 ) 44,608

Observations with missing firm identifiers (CIK, GVKEY, or CUSIP) (4,606 ) 40,002

Observations lost in merge between Compustat and CRSP (3,626 ) 36,376

Observations with insufficient CRSP return data (3,758) 32,618

Observations missing earnings announcement date (2,636 ) 29,982

Observations missing SIC data (241) 29,741

Observations where the earnings announcement date is more than 120 days after fiscal year end (247) 29,494

Foreign issuers (5,032) 24,462

Random selection of 650 observations from the 24,462 observations remaining 650

Missing 10-K or 10-K data are embedded in document with non-10-K data a

(21)

Missing Compustat data required for variable construction (9)

Final Sample 620

Table 1 depicts the impact of each sample selection criterion on our total sample. a

10-K data embedded in a document with non-10-K data generally refer to cases where certain sections of the 10-K are incorporated by reference to Exhibit 13, where Exhibit 13 ( portions of the annual report to shareholders) includes the President’s letter. We discard these observations because measurement of 10-K length becomes ambiguous in these cases.

Why are 10-K Filings So Long? 5

Accounting Horizons Volume 30, Number 1, 2016

Operating Complexity and 10-K Length

To gain an understanding of how operating complexity affects 10-K length, we conducted a preliminary search through a

broad sample of 10-Ks from 2003–2012. Based on our analysis of these 10-Ks, we identified a number of factors that appeared

to be significant drivers of 10-K length. This analysis, coupled with our review of the reporting requirements in Regulations S-

K and S-X, led to our development of the following model of the factors that impact 10-K length:

TOTAL LENGTHi;t ¼ b0 þ b1REDUNDANCYi;t þ b2DELTA ROAi;t þ b3DELTA REVi;t þ b4ACQUISITIONi;t þ b5FY RETi;t þ b6FY VOLATi;t þ b7SPI Di;t þ b8CAP LEASEi;t þ b9OP LEASEi;t þ b10RDi;t þ b11INTANGi;t þ b12SIZEi;t þ b13DEBT RATIOi;t þ b14FCFi;t þ b15DERIVATIVE USEi;t þ b16NUM OPSEGi;t þ b17NUM GEOSEGi;t þ b18REF ARi;t þ b19DEF14i;t þ Industry controls þ Year controls þ ei;t:

ð1Þ

REDUNDANCY is a measure of the degree to which a given firm’s 10-K and financial statements contain identical words, which we compute as described below in the ‘‘Disclosure Redundancy and 10-K Length’’ subsection. DELTA_ROA is the annual change in income before extraordinary items scaled by lagged assets. DELTA_REV is the annual percentage change in sales revenue. ACQUISITION is an indicator variable equal to 1 if the company made an acquisition during the fiscal year, and 0 otherwise.

5 FY_RET is equal to the raw annual return during the reporting period. FY_VOLAT is equal to the standard deviation of monthly returns for the reporting period. SPI_D is an indicator variable equal to 1 if Compustat indicates the company has reported a special item in their income statement, and 0 otherwise. We expect these variables to reflect significant

fiscal-year events that tend to be discussed in the MD&A or in the notes to the financial statements. We predict these variables

to be positively associated with 10-K length.

CAP_LEASE is an indicator variable set equal to 1 if the company reports a capital lease on its balance sheet, and 0 otherwise. OP_LEASE is an indicator variable set equal to 1 if the value of operating lease payments due in one year is greater than 1 percent of assets, and 0 otherwise. We expect the use of leases to be positively associated with 10-K length, given

required footnote disclosures and their possible inclusion in discussion of the firm’s business in Item 1 of the 10-K.

RD is a continuous variable equal to research and development expenditures scaled by lagged assets. Because material research and development activities are typically discussed in the 10-K, particularly in the business description, we expect RD to be positively associated with 10-K length. To account for purchased intangibles, we also include the continuous variable

INTANG, which is equal to the unamortized value of purchased intangible assets scaled by lagged assets. SIZE is equal to the natural logarithm of the market value of equity. We expect size to reflect operating complexity that increases the volume of 10-

K disclosures.

DEBT_RATIO is equal to total liabilities scaled by total assets. Because firms discuss the composition of their debt obligations in the notes to their statements, we expect a higher debt ratio to be associated with greater 10-K length. We also

include a proxy for free cash flows (FCF) measured as the average of operating cash flows scaled by total assets over the prior and current years (Blouin and Krull 2009). Because liquidity concerns are negatively associated with free cash flows and

companies discuss their liquidity position in the MD&A and other sections of the 10-K, we expect FCF to be negatively associated with 10-K length. DERIVATIVE_USE is an indicator variable equal to 1 if the company reports any current or accumulated gains or losses on derivative transactions, and 0 otherwise. We expect derivative use to increase the length of the

10-K.

We capture the number of operating and geographic segments the company reports with the variables NUM_OPSEG and NUM_GEOSEG, respectively. We expect 10-K length to be increasing in both of these variables.

Some firms incorporate certain sections of the 10-K by reference to the annual report, the relevant portion of which is then

included as Exhibit 13 to the 10-K. We include the referenced Exhibit 13 material in our total word count (TOTAL_LENGTH). To control for the possibility that information primarily intended for the annual report may be written differently (e.g., longer or

shorter) than disclosures intended primarily for the 10-K, we create an indicator variable REF_AR equal to 1 when 10-K sections are incorporated by reference and 0 otherwise. Also, a small number of firms include proxy statement information

(items 10–14) directly into the 10-K rather than following the more standard practice of incorporating this information only by

reference. To control for the impact of this information on the 10-K, we also include an indicator variable DEF14, which is equal to 1 for 10-Ks that include proxy statement information directly, and 0 otherwise.

5 The variable ACQUISITION is set equal to 1 for firm-years with non-missing, non-zero values of ACQCSHI and AQC as reported by Compustat (shares issued for acquisition and cash outflows for acquisition, respectively).

6 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

Disclosure Redundancy and 10-K Length

We use two alternative software programs to measure the amount of redundancy contained in a 10-K filing. Both programs

provide output based on a comparison of the text within two documents. Because concerns regarding 10-K redundancy

generally have to do with overlap between the notes to the financial statements and other sections of the 10-K, we split each 10-

K into two documents: one document containing the financial statement notes and a separate one containing all preceding items

of the 10-K (Business and Risk Factors, MD&A, etc.). We retain all tables in the 10-K, so that our assessment of redundancy

across 10-K sections includes disclosures made in tabular form.

The first measure of redundancy we use is the similarity score computed by Ferret Copy Detector, a program designed to

detect plagiarism by comparing word sequences in two documents. Ferret examines the number of distinct three-word

sequences, or trigrams, in two documents and computes the number of shared trigrams occurring in both documents as a

percentage of the total number of distinct trigrams occurring in both documents. The resulting similarity score

(REDUNDANCY) is computed as a decimal between 0 and 1 with higher scores reflecting a higher degree of shared language between the two documents.

Our second measure of redundancy, MATCHED_WORDS, is the output from a plagiarism detection software program called WCopyfind. This program compares pairs of documents and outputs the total number of distinct words that occur in six-

word sequences existing in both documents. 6

Thus, more redundant disclosures result in higher values of MATCHED_WORDS. Strong theoretical and empirical support exists in prior research for using trigrams to identify similar text because

independently written text typically share few matching trigrams. Gibbon, Moore, and Winski (1997) note that 77 percent of all

trigrams contained in a large sample of Wall Street Journal articles occurred only once. Lyon, Malcolm, and Dickerson (2001) find that three-word sequences are optimal for detecting similar text, and trigrams are a common choice in related research (e.g.,

Nelson and Pritchard 2014). Research on the optimal length of word-sequences to use to identify similar text indicates shorter

sequences (two or three words) provide the best results (Barrón-Cedeño and Rosso 2009). Thus, we are confident our three- and

six-word sequences are of sufficient length to capture disclosure redundancies in our context. 7

Both of our measures of redundancy are subject to drawbacks and limitations. As a percentage measure, REDUNDANCY does an arguably better job of capturing the importance of duplicative disclosure in explaining the proportion of a given firm’s

10-K length, relative to other determinants. However, REDUNDANCY likely is a worse measure for our tests modeling total 10-K length measured in words (TOTAL_LENGTH). For instance, if a complex firm and a simple firm both have REDUNDANCY values of 0.15, then the impact of redundancy on total word count is likely to be much greater for the complex firm because it has a longer 10-K to begin with. Thus, we rely primarily on MATCHED_WORDS to make inferences regarding the impact of redundant disclosure on total 10-K length, and we use REDUNDANCY primarily to make inferences in tests examining the determinants of redundant disclosure. We find that our primary results are qualitatively similar using either

measure.

Residual Disclosure and 10-K Length

We proxy for residual disclosure volume as the portion of 10-K length unexplained by operating complexity and disclosure

redundancy, or the residual from Equation (1). By construction, this measurement choice constrains mean residual disclosure to

be equal to 0. Thus, we interpret observations with negative residuals as having more parsimonious disclosures, and firms with

positive residuals as having abnormally long disclosures.

RESULTS

Descriptive Statistics and Correlation Matrix

Panel A of Table 2 presents descriptive statistics for the primary firm-level variables in our study. Panel A displays

distributional statistics for our final sample of 620 firm-years as well as for the broader sample of 24,462 firm-years from which

the final sample was randomly selected (of which 23,407 have sufficient data for Panel A of Table 2). t-tests indicate that the

6 WCopyfind allows for considerable latitude in comparison tests. Following McMullin (2014), we choose to compare six-word sequences with WCopyfind. We also require the sequences to be identically matched.

7 The majority of content identified as redundant by three- and six-word sequences in our sample come from longer sequences of words that often reflect redundant sentences or even paragraphs. We provide the following insight into the distributional features of redundant content in our sample when redundant output is based on six-word sequences: 80 percent of all words contained in redundant output are part of redundant sequences at least 11 words long. Two-thirds of all words in our redundant output are part of redundant sequences at least 21 words long. Redundant output based on trigrams versus six-word sequences is highly similar. In sensitivity analyses, we also find results using redundant ten-word sequences to produce very similar results to those using six-word sequences.

Why are 10-K Filings So Long? 7

Accounting Horizons Volume 30, Number 1, 2016

T A

B L

E 2

D e sc

r ip

ti v

e S

ta ti

st ic

s

P a

n e l

A :

F ir

m -L

e v

e l

V a

r ia

b le

s

S a

m p

le B

ro a

d e r

P o

p u

la ti

o n

V a

r ia

b le

n 2

5 th

P c tl

. 5

0 th

P c tl

. M

e a

n 7

5 th

P c tl

. S

td .

D ev

. V

a r ia

b le

n 2

5 th

P c tl

. 5

0 th

P c tl

. M

e a

n 7

5 th

P c tl

. S

td .

D ev

.

D E

L T

A _

R O

A 6

2 0

� 0

.0 0

8 0

.0 0

7 0

.0 2

0 0

.0 3

3 0

.0 9

3 D

E L

T A

_ R

O A

2 3

,4 0

7 �

0 .0

0 9

0 .0

0 6

0 .0

1 3

0 .0

3 0

0 .0

9 6

D E

L T

A _

R E

V 6

2 0

� 0

.0 0

4 0

.0 8

4 0

.1 5

8 0

.2 0

3 0

.3 8

7 D

E L

T A

_ R

E V

2 3

,4 0

7 0

.0 0

5 0

.0 9

2 0

.1 5

0 0

.2 1

6 0

.3 2

8

A C

Q U

IS IT

IO N

6 2

0 0

.0 0

0 0

.0 0

0 0

.4 3

5 1

.0 0

0 0

.4 9

6 A

C Q

U IS

IT IO

N 2

3 ,4

0 7

0 .0

0 0

0 .0

0 0

0 .4

5 4

1 .0

0 0

0 .4

9 8

C A

P _

L E

A S

E 6

2 0

0 .0

0 0

0 .0

0 0

0 .2

5 5

1 .0

0 0

0 .4

3 6

C A

P _

L E

A S

E 2

3 ,4

0 7

0 .0

0 0

0 .0

0 0

0 .2

7 0

1 .0

0 0

0 .4

4 4

O P

_ L

E A

S E

6 2

0 0

.0 0

0 0

.0 0

0 0

.3 9

1 1

.0 0

0 0

.4 8

8 O

P _

L E

A S

E 2

3 ,4

0 7

0 .0

0 0

0 .0

0 0

0 .4

2 9

1 .0

0 0

0 .4

9 5

IN T

A N

G 6

2 0

0 .0

0 8

0 .0

7 0

0 .1

7 2

0 .2

7 4

0 .2

1 4

IN T

A N

G 2

3 ,4

0 7

0 .0

0 9

0 .0

8 3

0 .1

8 5

0 .2

8 5

0 .2

3 5

R D

6 2

0 0

.0 0

0 0

.0 0

0 0

.0 3

8 0

.0 3

1 0

.0 8

7 R

D 2

3 ,4

0 7

0 .0

0 0

0 .0

0 0

0 .0

3 6

0 .0

3 0

0 .0

8 3

D E

B T

_ R

A T

IO 6

2 0

0 .1

7 6

0 .3

6 6

0 .3

6 8

0 .5

1 5

0 .2

3 6

D E

B T

_ R

A T

IO 2

3 ,4

0 7

0 .1

8 8

0 .3

6 4

0 .3

6 5

0 .5

0 7

0 .2

2 2

F C

F 6

2 0

0 .0

3 5

0 .0

8 0

0 .0

7 6

0 .1

2 9

0 .1

0 5

F C

F 2

3 ,4

0 7

0 .0

3 6

0 .0

8 0

0 .0

8 0

0 .1

3 1

0 .1

0 0

D E

R IV

A T

IV E

_ U

S E

6 2

0 0

.0 0

0 0

.0 0

0 0

.4 5

1 1

.0 0

0 0

.4 9

8 D

E R

IV A

T IV

E _

U S

E 2

3 ,4

0 7

0 .0

0 0

0 .0

0 0

0 .4

2 0

1 .0

0 0

0 .4

9 4

S IZ

E 6

2 0

6 .1

3 9

7 .0

6 6

7 .2

8 1

8 .0

7 8

1 .4

3 0

S IZ

E 2

3 ,4

0 7

6 .1

7 4

7 .0

3 4

7 .2

7 3

8 .0

8 6

1 .3

8 9

F Y

_ R

E T

6 2

0 �

0 .1

1 2

0 .1

1 4

0 .1

8 7

0 .3

7 6

0 .4

9 4

F Y

_ R

E T

2 3

,3 7

9 �

0 .0

8 9

0 .1

3 3

0 .2

1 2

0 .3

8 2

0 .5

2 7

F Y

_ V

O L

A T

6 2

0 0

.2 1

8 0

.3 2

6 0

.3 7

7 0

.4 7

1 0

.2 2

0 F

Y _

V O

L A

T 2

3 ,3

5 4

0 .2

2 3

0 .3

2 3

0 .3

7 0

0 .4

6 6

0 .2

0 4

S P

I_ D

6 2

0 0

.0 0

0 1

.0 0

0 0

.6 4

1 1

.0 0

0 0

.4 8

0 S

P I_

D 2

3 ,4

0 7

0 .0

0 0

1 .0

0 0

0 .6

7 9

1 .0

0 0

0 .4

6 7

N U

M _

O P

S E

G 6

2 0

0 .0

0 0

0 .0

0 0

0 .4

3 6

0 .0

0 0

1 .5

1 6

N U

M _

O P

S E

G 2

3 ,4

0 7

0 .0

0 0

0 .0

0 0

0 .4

6 7

0 .0

0 0

1 .5

5 4

N U

M _

G E

O S

E G

6 2

0 1

.0 0

0 1

.0 0

0 2

.3 3

1 3

.0 0

0 1

.9 7

5 N

U M

_ G

E O

S E

G 2

3 ,4

0 7

1 .0

0 0

1 .0

0 0

2 .4

0 5

3 .0

0 0

2 .0

5 4

P a

n e l

B :

1 0

-K C

h a

r a

c te

r is

ti c s

V a

r ia

b le

n 2

5 th

P c tl

. 5

0 th

P c tl

. M

e a

n 7

5 th

P c tl

. S

td .

D ev

. Q

3 �

Q 1

T O

T A

L _

L E

N G

T H

6 2

0 4

1 ,2

6 9

5 2

,0 0

4 5

5 ,3

3 5

6 5

,5 3

4 1

9 ,9

4 2

2 4

,3 2

2

M A

T C

H E

D _

W O

R D

S 6

2 0

2 ,3

7 5

3 ,6

7 9

4 ,0

1 4

5 ,3

6 7

2 ,4

8 1

3 ,0

4 2

R E

D U

N D

A N

C Y

6 2

0 0

.0 9

0 0

.1 1

7 0

.1 2

2 0

.1 4

8 0

.0 4

5 0

.0 5

8

D E

F 1

4 6

2 0

0 .0

0 0

0 .0

0 0

0 .0

7 6

0 .0

0 0

0 .2

6 5

—

R E

F _

A R

6 2

0 0

.0 0

0 0

.0 0

0 0

.0 8

6 0

.0 0

0 0

.2 8

0 —

IT E

M 1

_ L

E N

G T

H 6

2 0

7 ,6

1 4

1 2

,1 2

2 1

2 ,9

8 0

1 6

,7 7

5 7

,1 2

2 .9

5 9

,1 6

1

IT E

M 7

_ L

E N

G T

H 6

2 0

9 ,5

8 6

1 2

,4 0

3 1

4 ,0

1 9

1 ,6

6 4

0 7

,0 1

1 .1

6 7

,0 6

0

IT E

M 8

_ L

E N

G T

H 6

2 0

1 4

,3 2

3 1

8 ,4

2 8

2 0

,1 1

6 2

,4 4

5 7

8 ,4

4 9

.6 1

0 ,1

6 7

O T

H E

R _

1 0

K _

L E

N G

T H

6 2

0 5

,5 6

0 6

,9 4

8 8

,2 0

9 .4

9 ,4

9 6

5 ,0

0 8

.8 3

,9 3

6

(c o

n ti

n u

ed o

n n

ex t

p a

g e)

8 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

T A

B L

E 2

(c o

n ti

n u

e d

)

P a

n e l

C :

S a

m p

le C

o m

p o

si ti

o n

b y

In d

u st

r y

F a

m a

-F re

n c h

In d

u st

r y

C la

ss ifi

c a

ti o

n n

In d

u st

r y

R e p

re se

n ta

ti o

n In

S a

m p

le In

d u

st r y

R e p

re se

n ta

ti o

n In

P o

p u

la ti

o n

C o

n su

m e r

N o

n d

u ra

b le

s 2

5 4

% 5

%

C o

n su

m e r

D u

ra b

le s

1 7

3 %

2 %

M a n

u fa

c tu

ri n

g 5

5 9

% 1

0 %

E n

e rg

y 2

1 3

% 5

%

C h

e m

ic a ls

1 5

2 %

3 %

B u

si n

e ss

E q

u ip

m e n

t 1

1 5

1 9

% 1

7 %

T e le

c o

m m

u n

ic a ti

o n

s 2

0 3

% 3

%

U ti

li ti

e s

2 3

4 %

4 %

W h

o le

sa le

/R e ta

il S

h o

p s

5 3

9 %

1 0

%

H e a lt

h c a re

5 5

9 %

9 %

F in

a n

c e

1 5

2 2

4 %

2 2

%

O th

e r

6 9

1 1

% 1

2 %

T o

ta l

6 2

0 1

0 0

% 1

0 0

%

P a

n e l

D :

S a

m p

le C

o m

p o

si ti

o n

b y

Y e a

r

Y e a

r n

% o

f T

o ta

l M

e a

n T

O T

A L

_ L

E N

G T

H M

e d

ia n

T O

T A

L _

L E

N G

T H

2 0

0 3

4 9

8 %

4 4

,3 4

5 3

8 ,7

6 4

2 0

0 4

5 9

1 0

% 4

6 ,2

5 9

4 4

,1 3

9

2 0

0 5

7 1

1 1

% 5

0 ,5

7 9

5 0

,5 1

5

2 0

0 6

6 7

1 1

% 5

1 ,5

2 4

4 7

,3 7

0

2 0

0 7

6 4

1 0

% 5

3 ,7

6 8

4 9

,1 3

5

2 0

0 8

5 8

9 %

6 1

,3 9

0 5

5 ,6

8 5

2 0

0 9

6 5

1 1

% 5

4 ,4

5 5

5 2

,6 5

7

2 0

1 0

5 9

1 0

% 6

3 ,2

7 7

6 1

,4 6

7

2 0

1 1

7 1

1 1

% 6

2 ,2

8 5

6 0

,5 7

6

2 0

1 2

5 7

9 %

6 4

,2 3

9 6

1 ,8

8 2

T o

ta l

6 2

0 1

0 0

%

P a n e l

A o f

T a b le

2 p ro

v id

e s

d e sc

ri p ti

v e

st a ti

st ic

s fo

r o u r

ra n d o m

sa m

p le

o f

6 2 0

fi rm

-y e a r

o b se

rv a ti

o n s

a n d

fo r

th e

la rg

e r

p o p u la

ti o n

o f

2 3 ,4

0 7

C o m

p u st

a t

fi rm

-y e a rs

fr o m

w h ic

h th

e sa

m p le

w a s

d ra

w n .

P a n e l

B p ro

v id

e s

d e sc

ri p ti

v e

st a ti

st ic

s re

g a rd

in g

th e

1 0

-K s

o f

fi rm

s in

o u r

sa m

p le

. P

a n e l

C p ro

v id

e s

a b re

a k d o w

n o f

o u r

sa m

p le

b y

in d u st

ry .

In d u st

ry d e fi

n it

io n s

a re

b a se

d o n

th e

F a m

a -F

re n c h

1 2 -i

n d u st

ry c la

ss ifi

c a ti

o n

sc h e m

e a v a il

a b le

o n

K e n

F re

n c h ’s

w e b si

te :

h tt

p :/

/m b a .t

u c k .

d a rt

m o u th

.e d u /p

a g e s/

fa c u lt

y /k

e n .f

re n c h /D

a ta

_ L

ib ra

ry /d

e t_

1 2 _ in

d _ p o rt

.h tm

l. P

a n e l

D il

lu st

ra te

s th

e c o m

p o si

ti o n

o f

o u r

sa m

p le

a c ro

ss e a c h

o f

th e

1 0

y e a rs

in c lu

d e d

in o u r

sa m

p le

, a lo

n g

w it

h th

e m

e a n

a n d

m e d ia

n n u m

b e r

o f

w o rd

s in

th e

1 0

-K (T

O T

A L

_ L

E N

G T

H )

fo r

e a c h

y e a r.

V a ri

a b le

d e fi

n it

io n s

a re

p ro

v id

e d

in A

p p e n d ix

B .

Why are 10-K Filings So Long? 9

Accounting Horizons Volume 30, Number 1, 2016

random sample is not significantly different from the population it was drawn from except for DELTA_ROA, in which the difference is barely significant ( p ¼ 0.046 ). Thus, our sample is representative of the larger Compustat population.

Among the findings of interest in Table 2, Panel A, we note that nearly 44 percent of sample firms made some type of

acquisition during the year, 64 percent reported special items, and the mean firm has 2.3 geographic reporting segments.

Panel B of Table 2 reports descriptive statistics for variables related to sample 10-K filings. Mean 10-K length is

approximately 55,000 words. The median of approximately 52,000 words is close to the mean, suggesting this distribution is

only mildly skewed. The inter-quartile range is 24,322 words, indicating substantial variation in 10-K length. 8

The longest

section of the 10-K tends to be the financial statements and accompanying notes (ITEM8_LENGTH), followed by the MD&A (ITEM7_LENGTH) and the business description and risk factors (ITEM1_LENGTH). These three sections account for over 85 percent of total 10-K length. We also find substantial variation in the amount of redundancy incorporated in firms’ 10-Ks.

REDUNDANCY has an inter-quartile range equal to 50 percent of its median value, and MATCHED_WORDS has an inter- quartile range of nearly 83 percent of its median value.

Panel C of Table 2 provides a breakdown of the sample by industry. Similar to the larger Compustat population, the sample

has a particularly high representation among firms in the Finance and Business Equipment industries. Panel D indicates that the

sample has roughly equal representation across sample years, and that both the mean and median length of 10-Ks has increased

dramatically over our ten-year sample period. Mean TOTAL_LENGTH in 2012 is 45 percent greater than mean TOTAL_ LENGTH in 2003. This result is consistent with evidence elsewhere of significant increases in 10-K length. In its 2011 study of disclosure trends, KPMG reports a 28 percent increase in mean 10-K length (measured in pages) from the slightly shorter time

period covering 2004 to 2010.

Table 3 shows the Pearson correlations among the redundancy measures, the length of the 10-K and key individual items,

and firm size. REDUNDANCY and MATCHED_WORDS are 57 percent correlated, indicating that these two measures are largely capturing the same construct, though slightly differently as the former is a percentage number and the latter an absolute measure

of redundant text. Interestingly, REDUNDANCY is positively correlated with total 10-K length and with the length of the MD&A and notes to the financial statements (ITEM7_LENGTH and ITEM8_LENGTH), but negatively correlated with ITEM1_ LENGTH. This suggests there is greater redundancy between the content of the MD&A and financial statement notes, and less redundancy between the content of the Business and Risk Factors section and the notes. Not surprisingly, larger firms have

longer 10-Ks and higher values of MATCHED_WORDS, but there is no correlation between firm size and REDUNDANCY.

Regression Results

Table 4 displays results from estimating Equation (1). 9

The first column shows a strong positive coefficient estimate on

REDUNDANCY ( p-value ¼ 0.010). To test for a change in disclosure volume over our sample period, we next estimate the

TABLE 3

Correlation Matrix

(1) (2) (3) (4) (5) (6) (7)

REDUNDANCY (1) 1.00 MATCHED_WORDS (2) 0.57 1.00 TOTAL_LENGTH (3) 0.12 0.63 1.00 ITEM1_LENGTH (4) �0.19 0.33 0.58 1.00 ITEM7_LENGTH (5) 0.17 0.51 0.79 0.21 1.00 ITEM8_LENGTH (6 ) 0.18 0.58 0.84 0.25 0.73 1.00 SIZE (7) 0.07 0.17 0.26 0.00 0.29 0.35 1.00

Table 3 presents Pearson correlations for variables related to 10-K length, redundancy, and firm size. Correlations significant at the 5 percent level or lower are in bold. Variable definitions are provided in Appendix B.

8 To provide some context, this inter-quartile range corresponds to approximately 30 pages of single-spaced text on a standard page.

9 Our tables display results from OLS regressions because we are primarily interested in the magnitude rather than the statistical significance of the coefficient estimates. Inferences regarding statistical significance are unchanged when using clustered standard errors, likely because very few firms are repeated in our sample. As an additional robustness check, we re-estimate the model in Table 4 using a Poisson regression to account for the non- negative constraint on 10-K word count. We find that the direction of the relationship between 10-K count and the right-hand side variables are similar under this log-linear specification.

10 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

TABLE 4

Modeling Total 10-K Length

Panel A: Regression Results

Dependent Variable

Independent Variables

TOTAL_LENGTH Ln(TOTAL_LENGTH)

Parameter Est. p-value

Parameter Est. p-value

Parameter Est. p-value

Parameter Est. p-value

Intercept �3,456.38 0.582 �1,520.752 0.791 7,320.77 0.111 9.859 , 0.001*** REDUNDANCY 38,041.000 0.010** 39,085.000 0.008*** 1.003 0.000*** MATCHED_WORDS 3.860 , 0.001*** DELTA_ROA �156.737 0.860 108.912 0.901 16.522 0.982 �0.002 0.889 DELTA_REV �647.428 0.496 �250.002 0.792 �111.075 0.890 �0.008 0.637 ACQUISITION 2,692.081 0.074* 3,059.107 0.042** 2,523.783 0.047** 0.059 0.029** CAP_LEASE 2,987.452 0.064* 2,770.379 0.087* 1,463.953 0.286 0.048 0.098* OP_LEASE �2,359.632 0.142 �2,410.793 0.135 �1,595.162 0.243 �0.047 0.102 INTANG �3,022.347 0.375 �3,635.702 0.286 �2,379.020 0.410 �0.015 0.811 RD �2,555.674 0.806 �1,832.526 0.860 6,437.094 0.462 0.065 0.728 DEF14 11,811.000 , 0.001*** 11,666.000 , 0.001*** 10,968.000 , 0.001*** 0.191 , 0.001*** REF_AR �1,188.779 0.637 �2,191.817 0.376 435.552 0.836 �0.084 0.058* DEBT_RATIO 5,025.656 0.087* 6,008.778 0.040** 1,398.787 0.574 0.145 0.006*** FCF �22,824.000 0.002*** �22,397.00 0.002*** �15,662.00 0.010*** �0.414 0.001*** DERIVATIVE_USE 1,140.141 0.442 1,024.634 0.488 1,840.376 0.142 0.020 0.443 SIZE 3,398.110 , 0.001*** 3,424.719 , 0.001*** 2,158.337 , 0.001*** 0.049 , 0.001*** FY_RET �139.258 0.916 �1,374.626 0.251 �1,697.574 0.095* �0.019 0.369 FY_VOLAT 12,445.000 0.001*** 12,654.000 0.000*** 6,802.187 0.015** 0.181 0.002*** SPI_D 4,177.114 0.006*** 4,086.281 0.007*** 2,503.600 0.048** 0.084 0.002*** NUM_OPSEG 1,017.885 0.020** 1,014.599 0.020** 938.698 0.011** 0.021 0.007*** NUM_GEOSEG 237.775 0.548 190.460 0.630 188.894 0.574 0.008 0.251 TIME_TREND 2,065.816 , 0.001*** 1,062.774 , 0.001*** 0.038 , 0.001***

Industry fixed effects

included? Yes Yes Yes Yes

Year fixed effects

included? Yes No No No

R 2

0.448 0.434 0.595 0.427

n 620 620 620 620

Panel B: Distributional Statistics for the Three Components of 10-K Length

Variable n 25th Pctl. Median Mean 75th Pctl. Std. Dev. Q3 � Q1

OP_COMPLEX_LENGTH 620 33,092 38,694 39,776 46,165 9,660 13,073 REDUNDANT_LENGTH 620 9,168 14,416 15,697 20,824 9,660 11,656 RESIDUAL_LENGTH 620 �8,029 �997 0 5,904 20,622 13,933

*, **, *** Denote statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Panel A of Table 4 presents regression results from estimating Equation (1). Panel B presents distributional statistics for OP_COMPLEX, REDUNDANT_LENGTH, and RESIDUAL_LENGTH.

Variable Definitions: RESIDUAL_LENGTH ¼ residual from estimating Equation (1), represents 10-K disclosure length unexplained by firms’ operating complexity or

disclosure redundancy; REDUNDANT_LENGTH ¼ fitted value from the coefficient estimate on MATCHED_WORDS in Table 4, represents the impact of redundant disclosures

on total 10-K length; and OP_COMPLEX_LENGTH ¼ fitted value from Equation (1) after subtracting the estimated impact of disclosure redundancy (REDUNDANT_LENGTH).

Why are 10-K Filings So Long? 11

Accounting Horizons Volume 30, Number 1, 2016

regression after suppressing the year fixed effects and instead substituting a variable (TIME_TREND) equal to the fiscal year minus 2003. Results for this alternative specification are displayed in the second column of Table 4 and provide robust

evidence that 10-K length is increasing over the sample period, consistent with related research (e.g., Loughran and McDonald

2011).

As discussed in the ‘‘Residual Disclosure and 10-K Length’’ subsection, MATCHED_WORDS likely provides a better measure of the impact of disclosure redundancy on total word length than does our REDUNDANCY measure. Replacing REDUNDANCY with MATCHED_WORDS results in coefficient estimates presented in Column 3 of Table 4. Consistent with our expectation, the coefficient estimate on MATCHED_WORDS is strongly significant ( p-value , 0.001). Finally, to test whether our results are sensitive to the mild skewness in the dependent variable TOTAL_LENGTH, we rerun our regression after replacing the dependent variable with its log transformation and present the results in the last column of Table 4. We find

that inferences are virtually unchanged.

Across the four columns of Table 4, we find that the incidence of acquisitions, capital leases, leverage, free cash flows,

size, return volatility, special items, and operating segments all contribute significantly to 10-K length. There is some evidence

that fiscal year return (FY_RET) is negatively related to 10-K length. This may be attributable to managers using more words to present bad news than good news (e.g., Li 2008). Unsurprisingly, we also find a strong positive coefficient estimate on DEF14, consistent with firms incorporating information in the 10-K that is typically reserved for the proxy statement having longer 10-

Ks. We find no evidence that revenue growth, operating leases, R&D expense, or geographic segments are associated with 10-

K length, possibly because those factors have been subsumed by other regressors in the model.

We use the parameter estimates from Column 3 of Table 4, Panel A to compute the portion of total 10-K length attributable

to operating complexity, disclosure redundancy, and residual disclosure length. We use the fitted values from all parameter

estimates and independent variables except for MATCHED_WORDS to construct OP_COMPLEX_LENGTH, which we use to measure the portion of 10-K length explained by operating complexity. We use the coefficient estimate on MATCHED_ WORDS to compute a fitted value, REDUNDANT_LENGTH, which we use to measure the portion of 10-K length explained by disclosure redundancy. We use the residual as a proxy for residual disclosure, which we name RESIDUAL_LENGTH.

Panel B of Table 4 displays distributional statistics for the three components of TOTAL_LENGTH. We find that OP_ COMPLEX_LENGTH is the largest, with a mean value equal to 39,776 words. This represents approximately 72 percent of the mean value of TOTAL_LENGTH as reported in Table 2. In contrast, the portion attributable to disclosure redundancy is much smaller, as indicated by the mean value of REDUNDANT_LENGTH, which is only 15,697 words. By construction, the mean value of RESIDUAL_LENGTH is close to 0.

Arguably a more important result from Panel B of Table 4 is the inter-quartile range across the three components of total

10-K length because it reveals the degree to which each component of length explains variation in total 10-K length.

REDUNDANT_LENGTH has an inter-quartile range of 11,656 words, which is over 89 percent of the inter-quartile range of OP_COMPLEX_LENGTH (13,073 words). This surprising result indicates that nearly as much cross-sectional variation in 10- K length is explained by variation in disclosure redundancy as by operating complexity. A slightly greater amount of variation

in 10-K length is unexplained by either operating complexity or disclosure redundancy, as indicated by the inter-quartile range

of RESIDUAL_LENGTH (13,933 words). To provide a sense of how operating complexity, disclosure redundancy, and residual disclosure comprise 10-K length

over relatively shorter versus relatively longer 10-Ks, we partition our sample into quintiles based on 10-K length. We then

compute the median values of OP_COMPLEX_LENGTH, REDUNDANT_LENGTH, and RESIDUAL_LENGTH in each quintile and present this information in pictorial form in Figure 1. Figure 1 indicates that variation in 10-K length due to

operating complexity across the five quintiles are relatively modest. The median firm in the top quintile has OP_COMPLEX_ LENGTH only 15,078 words (47,013 � 31,935) longer than the median firm in the bottom quintile, an increase of only 47 percent (15,078 4 31,935). However, the median firm in the top quintile has REDUNDANT_LENGTH that is 16,509 words (23,864� 7,355) longer than the median firm in the bottom quintile, representing an increase of 224 percent (16,509 4 7,355), suggesting disclosure redundancy explains more of the difference between the longest and the shortest 10-Ks than does

operating complexity. Finally, we note that the portion of unexplained (residual) 10-K length accounts for even more variation

from the top to the bottom quintiles (11,179 – [�7,629] ¼ 18,808 words) than either operating complexity or disclosure redundancy.

These findings provide strong support for the assumptions made in prior research that a significant amount of variation in

disclosure volume, even in documents as highly regulated as the 10-K, is attributable to managerial discretion. We note that our

findings regarding the extent of firms’ idiosyncratic levels of disclosure are subject to certain caveats. Specifically, the

theoretical construct we seek to measure as residual disclosure is 10-K disclosure volume unrelated to firm operating

characteristics and disclosure redundancy. However, the residual from our regression model captures both idiosyncratic

disclosure volume as well as the effect of any potential model misspecification (e.g., nonlinearities in the relationship between

12 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

firm characteristics and disclosure volume). Our analysis does not allow us to definitively determine the extent to which our

regression residual reflects model error rather than firms’ idiosyncratic disclosure choices.

To the best of our knowledge, ours is the first systematic evidence of the extent to which disclosure redundancy influences

10-K length. Although we cannot infer from these results the extent to which disclosure redundancy is driven by managerial

opportunism versus other factors, our analysis in the ‘‘What are the Determinants of Disclosure Redundancy?’’ subsection provides some insight to the degree each of these three components is associated with mispricing subsequent to the 10-K filing

date.

SUPPLEMENTAL ANALYSES

Where is the Redundancy in the 10-K?

We next examine each of the three primary sections of the typical 10-K (Business and Risk Factors, MD&A, and financial

statement notes) to pinpoint which sections are affected most by redundant disclosures. We re-estimate the regression in

Equation (1) after substituting alternately ITEM1_LENGTH, ITEM7_LENGTH, and ITEM8_LENGTH as the dependent variable. For brevity of exposition, we report only the results using REDUNDANCY as our redundancy measure in this analysis, though we note that inferences are similar if we use MATCHED_WORDS instead. We tabulate the results in Table 5.

Table 5 indicates that the coefficient estimate on REDUNDANCY is significantly negative in the first column explaining Item 1 length (Business and Risk Factors), but significantly positive in the second column explaining Item 7 length (MD&A).

Since REDUNDANCY is measured as a percentage, this indicates that 10-Ks marked with a higher proportion of redundant disclosures are associated with shorter disclosures in Item 1 and longer disclosures in Item 7. Although anecdotal evidence

indicates that some discussion of business description and risk factors, the primary content of Item 1, can also be found in the

notes to the financial statements, the results in Table 5 suggest that the 10-K redundancy is more associated with the content of

the MD&A. This finding is interesting in light of the SEC’s counsel to not report information in the MD&A section that is

redundant with the financial statement notes (SEC 2003). The third column of Table 5 indicates that a greater level of 10-K

redundancy is associated with longer Item 8 disclosures (financial statements and notes), though this result is expected in this

test because REDUNDANCY is based specifically on a comparison of the financial statement notes to the earlier sections of the 10-K.

One interesting takeaway from an examination of the R 2

in each column of Table 5 is that firm operating characteristics

and disclosure redundancy explain relatively less variation in Item 1 disclosure length (0.342), and relatively more variation in

Item 8 disclosure length (0.472). This suggests that the drafting of the Business and Risk Factors section is more heavily

FIGURE 1 Median OP_COMPLEX_LENGTH, REDUNDANT_LENGTH, and RESIDUAL_LENGTH across Quintiles

of Total 10-K Length

Figure 1 shows the three components of total 10-K length (RESIDUAL_LENGTH, REDUNDANT_LENGTH, OP_COMPLEX_LENGTH) for the median firm in each of the quintiles of total 10-K length measured in words (TOTAL_LENGTH). RESIDUAL_LENGTH is the residual from estimating Equation (1) and represents 10-K disclosure length unexplained by firms’ operating complexity or disclosure redundancy. REDUNDANT_LENGTH is the fitted value from the coefficient estimate on MATCHED_WORDS in Table 4 and represents the impact of redundant disclosures on total 10-K length. OP_ COMPLEX_LENGTH is the fitted value from Equation (1) after subtracting the estimated impact of disclosure redundancy (REDUNDANT_LENGTH).

Why are 10-K Filings So Long? 13

Accounting Horizons Volume 30, Number 1, 2016

influenced by factors that extend beyond the firm’s operating structure, such as management’s assessment of litigation and

other risks, shareholder composition, or simply management’s idiosyncratic ‘‘style.’’ Prior research shows individual managers

impose their style on a host of corporate policies and practices even after controlling for firm-level characteristics (e.g.,

Bertrand and Schoar 2003). In contrast, the length of the financial statement notes (Item 8) appear to be more heavily

determined by firm operating characteristics and disclosure redundancies, consistent with note disclosures being audited and

thus likely more regimented than other sections of the 10-K. The variation in the length of MD&A (Item 7) explained by firm

operating complexity and redundancy appears to lie midway between that of Business and Risk Factors and the Financial

Statement Notes.

What are the Determinants of Disclosure Redundancy?

Although we have documented that disclosure redundancy is an important contributor to overall 10-K length, an

unanswered question is why some 10-Ks incorporate significantly more redundancy than do others. The SEC permits but does

not necessarily require firms to cross-reference between the notes and other sections of the 10-K in lieu of presenting duplicate

information. No established theory in prior accounting literature predicts which accounting topics will be reported more

TABLE 5

Modeling 10-K Section Length Regression Results

Independent Variables

Dependent Variable

ITEM1_LENGTH ITEM7_LENGTH ITEM8_LENGTH

Parameter Est. p-value

Parameter Est. p-value

Parameter Est. p-value

Intercept 4,971.252 0.028** �4,778.595 0.021** �5,162.661 0.034** REDUNDANCY �23,987.00 , 0.001*** 23,175.000 , 0.001*** 24,313.000 0.000*** DELTA_ROA �636.892 0.065* 316.909 0.315 475.557 0.200 DELTA_REV 203.350 0.584 �336.430 0.323 85.881 0.830 ACQUISITION �99.700 0.866 935.715 0.083* 1,426.562 0.025** CAP_LEASE 1,354.369 0.033** 891.390 0.125 1,049.407 0.125 OP_LEASE �1,117.261 0.078* �645.099 0.266 �688.489 0.312 INTANG 992.083 0.458 �1,977.940 0.107 �2,103.812 0.145 RD 14,190.000 0.001*** �7,892.491 0.034** �10,172.000 0.021** DEF14 2,764.903 0.006*** �561.874 0.536 970.233 0.364 REF_AR �1,573.835 0.106 1,186.774 0.183 897.643 0.392 DEBT_RATIO 1,213.918 0.289 �2,696.225 0.010** 3,309.649 0.008*** FCF �1,070.400 0.704 �9,697.554 0.002*** �12,571.000 , 0.001*** DERIVATIVE_USE �511.612 0.378 278.655 0.600 1,118.715 0.074* SIZE 300.841 0.179 1,491.530 , 0.001*** 1,758.957 , 0.001*** FY_RET �295.807 0.530 �453.287 0.293 �729.846 0.150 FY_VOLAT 2,607.830 0.041** 5,458.367 , 0.001*** 4,282.247 0.002*** SPI_D �367.442 0.533 2,053.787 0.002*** 2,448.225 0.000*** NUM_OPSEG 176.307 0.304 32.269 0.837 567.261 0.002*** NUM_GEOSEG �138.015 0.375 55.590 0.696 161.113 0.336 TIME_TREND 967.643 , 0.001*** 313.931 0.000*** 720.213 , 0.001***

Industry fixed

effects included? Yes Yes Yes

Year fixed effects

included? Yes No No

R 2

0.342 0.410 0.472

n 620 620 620

*, **, *** Denote statistical significance at the 0.01, 0.05, and 0.10 levels, respectively. Table 5 re-estimates Equation (1) after replacing the dependent variable TOTAL_LENGTH with the word count of 10-K Items 1, 7, and 8, respectively. Independent variables are identical to those included in Table 4 and defined in Appendix B.

14 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

redundantly in firms’ financial reports. However, we provide some exploratory analysis into which firm characteristics are

associated with greater levels of redundant reporting.

We regress MATCHED_WORDS and REDUNDANCY on the other right-hand side variables in Equation (1) and report the

results in the first and second columns, respectively, of Table 6. We find that firm size and leverage exhibit a consistent,

positive relationship with both measures of disclosure redundancy. Disclosure redundancy is strongly negatively associated

with R&D intensity in both columns. In addition, firms are more likely to provide redundant disclosure when they report special

items, but less redundancy when they have made an acquisition.

Two other interesting findings emerge from Table 6. First is the negative coefficient estimate on REF_AR in both columns,

indicating that firms are less likely to provide redundant information when significant sections of the 10-K are incorporated into

the filing only by reference to the annual report to shareholders. This may be an indication that managers are more likely to cut

back on redundant reporting for disclosures designed specifically for the annual report, because the annual report tends to be a

more ‘‘reader-friendly’’ publication than the 10-K. Second is the positive and significant coefficient estimate on TIME_TREND

in Column 1, which becomes negative and significant in Column 2. This suggests that levels of redundancy as measured by the

number of redundant words is increasing over time, but that increases in redundancy over time fail to keep up with increases in

total 10-K disclosure, resulting in redundant disclosures declining as a proportion of total 10-K length. However, the economic

magnitude of the TIME_TREND coefficient appears to be fairly small, suggesting changes in redundancy in the 10-K over the

sample period are not large.

TABLE 6

Correlating Redundancy Measures with Other Disclosure and Reporting Attributes

Independent Variable

Dependent Variable

MATCHED_WORDS REDUNDANCY

Parameter Est. Pr . jtj Parameter Est. Pr . jtj

Intercept �1,106.249 0.156 0.098 , 0.001*** DELTA_ROA 55.052 0.653 0.003 0.209 DELTA_REV �52.237 0.693 �0.002 0.544 ACQUISITION 32.268 0.877 �0.011 0.012** CAP_LEASE 392.060 0.083* 0.005 0.240 OP_LEASE �203.149 0.367 0.001 0.858 INTANG �301.523 0.527 0.002 0.803 RD �3,079.841 0.033** �0.093 0.001*** DEF14 267.048 0.449 0.009 0.227 REF_AR �861.927 0.013** �0.018 0.010*** DEBT_RATIO 1,336.996 0.001*** 0.014 0.083* FCF �2,041.993 0.042** �0.029 0.143 DERIVATIVE_USE �233.598 0.258 �0.002 0.594 SIZE 360.894 , 0.001*** 0.003 0.041** FY_RET 128.498 0.443 0.004 0.185 FY_VOLAT 1,518.976 0.001*** 0.000 0.973 SPI_D 587.260 0.005*** 0.018 , 0.001*** NUM_OPSEG 7.310 0.905 �0.001 0.317 NUM_GEOSEG 1.974 0.972 0.000 0.889 TIME_TREND 240.535 , 0.001*** �0.002 0.006*** Industry fixed effects included? Yes Yes

Year fixed effects included? No No

n 620 620

R 2

0.25 0.12

*, **, *** Denote statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Table 6 presents from examining two measures of disclosure redundancy (MATCHED_WORDS and REDUNDANCY) as the dependent variable in Equation (1). Variable definitions are provided in Appendix B.

Why are 10-K Filings So Long? 15

Accounting Horizons Volume 30, Number 1, 2016

Components of 10-K Length and Investors’ Ability to Process Relevant Information

Prior research finds that longer 10-Ks are associated with lower trading volume around the 10-K filing date and subsequent

return drift after the filing date (e.g., You and Zhang 2009; Miller 2010). However, previous research does not indicate what

source of 10-K length might be the driver of impaired information processing around 10-K filing dates. Is it operating

complexity, disclosure redundancy, residual disclosure, or some combination of the three?

To address this issue, we build on You and Zhang’s (2009) empirical method to test for a relationship between post-filing-

date return drift and 10-K length in our more recent sample. We re-estimate their model after decomposing total 10-K length

among the components explained by operating complexity, redundancy, and residual disclosure. We tabulate the results of this

analysis in Table 7.

The first column of Table 7 is similar to You and Zhang’s (2009) Equation (4), which regresses post-filing date returns on

the size-adjusted returns during the three days centered on the 10-K filing date (FDR), a measure of 10-K length, and the

interaction between the two. Size, book-to-market, beta, and momentum are included to control for confounding risk factors.

As indicated by the results in Column 1 of Table 7, we are unable to find evidence of significant post-filing-date price drift

using You and Zhang’s (2009) methodology in our sample. 10

Table 7 reports results for six-month subsequent returns as the

TABLE 7

Testing for Stock Price Drift Following 10-K Filings

Independent Variable

Dependent Variable ¼ Six-Month Size-Adjusted Returns

Parameter Est. Pr . jtj

Parameter Est. Pr . jtj

Intercept 0.058 0.419 0.04226 0.574

FDR �1.443 0.126 0.79629 0.395 TOTAL_LENGTH 3.88E-07 0.520 FDR � TOTAL_LENGTH 2.22E-05 0.154 OP_COMPLEX_LENGTH 1.11E-06 0.442 REDUNDANT_LENGTH �9.21E-07 0.501 RESIDUAL_LENGTH 9.74E-07 0.289 FDR � REDUNDANT_LENGTH 1.19E-05 0.697 FDR � OP_COMPLEX_LENGTH �2.88E-05 0.205 FDR � RESIDUAL_LENGTH 5.02E-05 0.023** BETA 0.011 0.644 0.017 0.491 SIZE �0.011 0.204 �0.011 0.223 BM �0.015 0.408 �0.014 0.436 MOMENTUM �0.176 , 0.001*** �0.177 , 0.001*** Industry fixed effects included? No No

Year fixed effects included? No No

n 581 576

R 2

0.055 0.066

*, **, *** Denote statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. Table 7 presents results from regressing six-month size-adjusted stock returns on the determinants used in You and Zhang’s (2009) Table 7.

Variable Definitions: FDR ¼ size-adjusted returns for the three trading days starting on the 10-K filing date; BETA ¼ estimate using daily returns from a value-weighted market model over three years; SIZE ¼ natural logarithm of market value of equity; BM ¼ book-to-market ratio as of the end of the fiscal year; and MOMENTUM ¼ six-month stock return ending in the filing month.

10 You and Zhang (2009) use an indicator variable to capture whether a given filing is longer or shorter than the annual sample median. Although our reported regression results are based on a continuous measure of 10-K length, we are unable to replicate You and Zhang’s (2009) main result using either measure.

16 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

dependent variable, though we similarly find no results for the three-month or 12-month return windows. A possible

explanation is that investor sophistication has increased in our sample period relative to the period examined in their study

(1995–2005).

In the second column of Table 7, however, we find that partitioning total 10-K length produces some evidence that post-

filing-date return drift is associated with the portion of 10-K length attributable to residual disclosure, as evidenced by the

significantly positive coefficient estimate on FDR � RESIDUAL_LENGTH ( p-value ¼ 0.023). The insignificant coefficient estimates on FDR � OP_COMPLEX_LENGTH and FDR � REDUNDANT_LENGTH, which are significantly smaller in magnitude and far from statistically significant, provide no support for the notion that 10-K length associated with operating

complexity or redundancy impair price discovery around 10-K filing dates. 11

What Accounts for the Increase in 10-K Disclosure from 2003–2012?

The primary objective of our study is to shed light on the extent to which cross-sectional variation in 10-K disclosure

volume is driven by variation in firms’ operating complexity, reporting redundancy, and residual disclosure volume.

Determining which time-series factors explain the evolution of disclosure volume over time is beyond the scope of this study.

However, we note a marked (45 percent) increase in mean 10-K length even over our relatively short and sample period. We

next provide some insight into the how much of the increase in disclosure volume during our sample period is attributable to

increases in firm operating complexity versus increases in required or voluntary disclosure.

We first estimate the baseline relationship between 10-K length and firm operating complexity variables in the early part of

our sample period (2003–2006 ). We use the resulting coefficient estimates to create fitted values for expected 10-K length for

observations in the last six years of our sample. If required and voluntary disclosure over our sample period has remained

constant after controlling for changes in firm operating complexity, then these fitted values of 10-K length should approximate

actual 10-K length (TOTAL_LENGTH) from 2007–2012. Figure 2 plots mean actual and fitted ( predicted) 10-K length for each sample year. Figure 2 indicates that changes in

disclosure length over the sample period attributable solely to changes in firm complexity are relatively modest. Mean TOTAL_

FIGURE 2 Are Recent Increases in 10-K Length Attributable to Increases in Firm Complexity?

Figure 2 contrasts changes in mean TOTAL_LENGTH (number of words in the 10-K) over the sample period with the changes in 10-K length that would be predicted solely by changes in firm complexity during the sample period. The bold line is the fitted value of 10-K length based on coefficient estimates resulting from regressing TOTAL_LENGTH on firm operating complexity variables for the period 2003–2006. The relatively flat trend in the bold line suggests growth in 10-K length attributable solely to changes in firm characteristics is relatively modest. The thinner line represents actual mean TOTAL_ LENGTH for each year. TOTAL_LENGTH increases by 45 percent from 2003 through 2012, whereas length predicted solely by increases in firms’ operating complexity increases by only five percent over the sample period.

11 Although our evidence suggests that overall prices are not influenced by redundant disclosures or operating complexity around 10-K filing dates, it is possible that individual retail investors are harmed by lengthy disclosures associated with redundancy or operating complexity.

Why are 10-K Filings So Long? 17

Accounting Horizons Volume 30, Number 1, 2016

LENGTH in 2003 is 44,277 words and increases by 45 percent to 64,343 words in 2012. Mean predicted length in 2003 is 47,133 words, and increases by only 5 percent to 49,561 words in 2012. Thus, the increase in disclosure volume during our

sample period appears to be driven primarily by increases in required or voluntary disclosure rather than increases in firm

complexity.

Other Robustness Tests

Our model of the determinants of 10-K length in Equation (1) is necessarily truncated to include only a subset of all

operational factors that could potentially affect 10-K length. The variables included in our model are based primarily on our

own reading of numerous 10-K documents. One potential concern is that we have omitted important components of operating

complexity from Equation (1) that explain significant variation in 10-K length, leading to an overestimate of the amount of

variation in 10-K length that should be considered residual disclosure. To test the sensitivity of our inferences to this concern,

we add three additional variables to our model of 10-K length suggested by Li (2008). We find that including firm age, earnings

volatility, and an indicator variable reflecting whether the firm is incorporated in Delaware changes the regression R 2

by less

than 1 percent and the coefficient estimate on MATCHED_WORDS by less than 2 percent. We conclude that our allocation of 10-K length to the portions explained by operating complexity, redundancy, and residual disclosure are largely robust to

concerns about omitted variable bias.

Finally, we address the concern that macroeconomic events may affect 10-K length in ways not captured by our firm-

specific variables. Although we expect that macroeconomic events will largely by accounted for by our year and industry fixed

effects, as well as through their effects on variables measured at the firm level, it is possible that external shocks induce

industry-year variation in 10-K length in ways that are identified as residual disclosure under our current models. For instance,

if banks increased their 10-K disclosures during the recent financial crisis, then it is possible that this increased 10-K length

would be categorized as RESIDUAL_LENGTH, which is intended to capture only firm-year specific effects. If this concern is problematic for our analysis, then we should find values of RESIDUAL_LENGTH in the financial services industry to be particularly large around the financial crisis.

Although we find that TOTAL_LENGTH is significantly greater in 2008 for financial services firms, we find no evidence that mean or median RESIDUAL_LENGTH is significantly larger during the financial crisis than at other times during our sample period. Mean and median RESIDUAL_LENGTH are modestly positive in 2008, though smaller in magnitude than in four other years during the sample period. The mean and median are both negative in 2009, and relatively close to zero in 2010.

Thus, our model appears to do a fair job of allocating the effects of macroeconomic shocks on 10-K length to our operating

complexity variable and not to RESIDUAL_LENGTH.

CONCLUSION

Although several studies have examined the consequences of 10-K length on various corporate outcomes, we are the

first to provide systematic descriptive evidence on the structural components that contribute to cross-sectional variation in

10-K length. We argue that 10-K length can be thought of as having three drivers: (1) the portion attributable to operating

complexity; (2) the portion attributable to disclosure redundancies; and (3) residual disclosure. We find that disclosure

redundancy accounts for as much variation in 10-K length as does operating complexity. We also find that residual

disclosure accounts for more variation in 10-K length across firms than do either operating complexity or disclosure

redundancy.

Results from this study should be of interest to both regulators and academics. Both the SEC and the FASB have expressed

recent interest in enhancing the quality of disclosure, with an emphasis on reducing disclosure volume. In calling for a

meaningful review of disclosure requirements, SEC Chairman Mary Jo White questioned whether ‘‘[SEC] rules [are] the

primary cause of potential disclosure overload or [whether] other sources contribute to it’’ (White 2013). Our study suggests

there is a significant discretionary component to disclosure volume. We are the first to provide systematic evidence of the

significant role of disclosure redundancy in determining 10-K length. We also highlight the extent to which firms’ idiosyncratic

levels of disclosure contribute to variation in 10-K length after controlling for firms’ operating characteristics.

Our findings offer several interesting and important questions for future research regarding firms’ disclosure practices

within the 10-K. First, we propose that future research investigate the factors that drive idiosyncratic disclosure, or disclosure

volume not explained by the firm characteristics modeled in the current study. For instance, future research could consider

whether 10-K disclosure volume varies with CEO-level characteristics, internal and external governance mechanisms,

compensation structure, shareholder composition, and litigation concerns. In addition, further research on the nature and

content of disclosure redundancy may be able to shed further light on whether those redundancies are explained primarily by

strategic considerations or simply by managers’ efforts to comply with overlapping disclosure requirements.

18 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

REFERENCES

Barrón-Cedeño, A., and P. Rosso. 2009. On automatic plagiarism detection based on n-grams comparison. Computer Science 5478: 696– 700.

Bertrand, M., and A. Schoar. 2003. Managing with style: The effect of managers on firm policies. The Quarterly Journal of Economics 118 (4): 1169–1208.

Blouin, J., and L. Krull. 2009. Bringing it home: A study of the incentives surrounding the Repatriation of Foreign Earnings under the

American Jobs Creation Act of 2004. Journal of Accounting Research 47 (4): 1027–1059. Financial Accounting Standards Board (FASB). 2001. GAAP-SEC Disclosure Requirements. Steering Committee Report Series.

Norwalk, CT: FASB. Available at: http://www.fasb.org/cs/ContentServer?c¼Document_C&pagename¼FASB%2FDocument_ C%2FDocumentPage&cid¼1175801856648

Financial Accounting Standards Board (FASB). 2012a. Discussion Paper: Disclosure Framework. Norwalk, CT: FASB. Financial Accounting Standards Board (FASB). 2012b. FASB’s invitation to comment on Disclosure Framework. FASB in Focus (July

12). Available at: http://www.fasb.org/cs/ContentServer?site¼FASB&c¼Document_C&pagename¼FASB%2FDocument_ C%2FDocumentPage&cid¼1176160161150

Financial Accounting Standards Board (FASB). 2014. Chapter 8: Notes to Financial Statements. Exposure Draft. Norwalk, CT: FASB. Gibbon, D., R. Moore, and R. Winski. 1997. Handbook of Standards and Resources for Spoken Language Systems. Berlin, Germany:

Mouton de Gruyter.

Griffin, P. A. 2003. Got information? Investor response to Form 10-K and Form 10-Q EDGAR filings. Review of Accounting Studies 8: 433–460.

Huddart, S., B. Ke, and C. Shi. 2007. Jeopardy, non-public information, and insider trading around SEC 10-K and 10-Q filings. Journal of Accounting and Economics 43 (1): 3–36.

Institute of Chartered Accountants in England and Wales (ICAEW). 2013. Financial Reporting Disclosures: Market and Regulatory Failures: Information for Better Markets Initiative. (December). Available at: http://www.ion.icaew.com/ClientFiles/a42b9c80- 6acd-4dca-980a-bac45d9a324d/FRD%20final.pdf

KPMG. 2011. Disclosure Overload and Complexity: Hidden in Plain Sight. Available at: http://www.kpmg.com/US/en/ IssuesAndInsights/ArticlesPublications/Documents/disclosure-overload-complexity.pdf

Lawrence, A. 2013. Individual investors and financial disclosure. Journal of Accounting and Economics 56 (1): 130–147. Lehavy, R., F. Li, and K. Merkley. 2011. The effect of annual report readability on analyst following and the properties of their earnings

forecasts. The Accounting Review 86 (3): 1087–1115. Li, F. 2008. Annual report readability, current earnings, and earnings persistence. Journal of Accounting and Economics 45 (2): 221–247. Loughran, T., and B. McDonald. 2011. When is a liability not a liability? Textual analysis, dictionaries, and 10-Ks. The Journal of

Finance 66 (1): 35–65. Lyon, C., J. Malcolm, and B. Dickerson. 2001. Detecting short passages of similar text in large document collections. In Proceedings of

Conference on Empirical Methods in Natural Language Processing, Pittsburgh, PA, June 3–4. McMullin, J. 2014. Can I Borrow Your Footnotes? Learning and Network Benefits of Footnote Similarity. Working paper, Indiana

University.

Miller, B. 2010. The effects of reporting complexity on small and large investor trading. The Accounting Review 85 (6): 2107–2143. Nelson, K., and A. Pritchard. 2014. Carrot or Stick? The shift from Voluntary to Mandatory Disclosure of Risk Factors. Working paper,

University of Michigan.

Rennekamp, K. 2012. Processing fluency and investors’ reactions to disclosure readability. Journal of Accounting Research 50: 1319– 1354.

Rogers, J., A. Van Buskirk, and S. Zechman. 2011. Disclosure tone and shareholder litigation. The Accounting Review 86 (6 ): 2155– 2183.

Securities and Exchange Commission (SEC). 2003. Commission Guidance Regarding Management’s Discussion and Analysis of Financial Condition and Results of Operations. Release No. 33-8350 (December 29). Washington, DC: SEC.

Securities and Exchange Commission (SEC). 2008. Final Report of the Advisory Committee on Improvements to Financial Reporting to the United States Securities and Exchange Commission. (August 1). Washington, DC: SEC.

Securities and Exchange Commission (SEC). 2012. Center for Audit Quality SEC Regulations Committee Joint Meeting. (March 27). Washington, DC: SEC.

U.S. Code of Federal Regulations. 2014. Standard Instructions for Filing Forms under Securities Act of 1933, Securities Exchange Act of

1934 and Energy Policy and Conservation Act of 1975—Regulation S-K. Title 17, Pt 229, 2014 ed.

White, M. 2013. The Path Forward on Disclosure. Speech presented at the National Association of Corporate Directors, National Harbor, MD. Available at: http://www.sec.gov/News/Speech/Detail/Speech/1370539878806#.Uzx2VPldV8E

You, H., and X. Zhang. 2009. Financial reporting complexity and investor underreaction to 10-K information. Review of Accounting Studies 14 (4): 559–586.

Why are 10-K Filings So Long? 19

Accounting Horizons Volume 30, Number 1, 2016

APPENDIX A

Examples of Redundant and Non-Redundant 10-K Disclosures

The following two sample disclosures are pulled from the MD&A section of the 2012 10-K filings for sample companies.

Both disclosures relate to recent accounting pronouncements. The first disclosure, from AG Mortgage Investment Trust, is

repeated verbatim in the financial statement notes section of the company’s 10-K. The second, from Ford Motor Company,

simply incorporates this information by reference to the financial statement notes.

AG Mortgage Investment Trust

Recent Accounting Pronouncements

In April 2011, the Financial Accounting Standards Board (FASB) issued Accounting Standards Update 2011-03,

Reconsideration of Effective Control for Repurchase Agreements (ASU 2011-03). ASU 2011-03 simplifies the accounting for financial assets transferred under repurchase agreements and similar arrangements, by eliminating the transferor’s ability

criterion from the assessment of effective control over those assets. The guidance is effective for fiscal years and interim periods

beginning after December 15, 2011. We do not believe that the adoption of the amended guidance will have a significant effect

on our consolidated financial statements now or in the future.

In May 2011, the FASB issued Accounting Standards Update 2011-04, Amendments to Achieve Common Fair Value Measurement and Disclosure Requirements in U.S. GAAP and IFRS (ASU 2011-04). ASU 2011-04 amends Topic 820 and does not modify the requirements for when fair value measurements apply; rather, it clarifies the Board’s intent about the

application of existing fair value measurement requirements, and changes to a particular principle or requirement for measuring

fair value or for disclosing information about fair value measurements. For public entities, this guidance is effective for fiscal

years and interim periods beginning after December 15, 2011. Early application by public entities is not permitted. We do not

anticipate any material impact from this guidance now or in the future.

In December 2011, the FASB issued Accounting Standards Updated 2011-11, Disclosures about Offsetting Assets and Liabilities (ASU 2011-11). ASU 2011-11 amends Topic 210 to require additional disclosure information about offsetting and related arrangements. Entities will be required to disclose both gross information and net information about both instruments

and transactions eligible for offset in the statement of financial position and instruments and transactions subject to an

agreement similar to a master netting arrangement. This scope would include derivatives, sale and repurchase agreements, and

reverse sale and repurchase agreements. The objective of this disclosure is to facilitate comparison between those entities that

prepare their financial statements on the basis of U.S. GAAP and those entities that prepare their financial statements on the

basis of International Financial Reporting Standards (IFRS). The guidance is effective for periods beginning on or after January

1, 2013, and interim periods within those annual periods. The Company does not anticipate any material impact from this

guidance.

Ford Motor Company

Accounting Standards Issued But Not Yet Adopted

For information on accounting standards issued but not yet adopted, see Note 1 of our Notes to the Financial Statements.

20 Cazier and Pfeiffer

Accounting Horizons Volume 30, Number 1, 2016

APPENDIX B

Primary Variable Definitions

Variable Name Definition

ACQUISITION Indicator variable set equal to 1 if the company made an acquisition during the fiscal year, and 0 otherwise.

CAP_LEASE Indicator variable equal to 1 if the company reports a capital lease on its balance sheet, and 0 otherwise.

DEBT_RATIO Total liabilities scaled by total assets.

DEF14 Indicator variable equal to 1 for 10-Ks that include proxy statement information directly, and 0 for firms incorporating that information only by reference to the proxy statement.

DELTA_REV The annual percentage change in sales.

DELTA_ROA The annual change in income before extraordinary items, scaled by lagged assets.

DERIVATIVE_USE Indicator variable equal to 1 if the company reports any current or accumulated gains or losses on derivative transactions, and 0 otherwise.

FCF The average operating cash flows scaled by total assets over the current and prior years.

FY_RET Raw annual return over the 12-month fiscal period.

FY_VOLAT The standard deviation of monthly returns for the reporting period.

INTANG The unamortized value of purchased intangible assets, scaled by lagged assets.

ITEM1_LENGTH The total number of words contained in Item 1 of the 10-K (Business and Risk Factors)

ITEM7_LENGTH The total number of words contained in Item 7 of the 10-K (MD&A).

ITEM8_LENGTH The total number of words contained in Item 8 of the 10-K (Financial Statements and Accompanying Notes).

MATCHED_WORDS The number of distinct words that occur in six-word sequences in both the financial statement footnotes and earlier sections of the 10-K.

NUM_GEOSEG The number of reported geographic segments.

NUM_OPSEG The number of reported operating segments.

OP_LEASE Indicator variable equal to 1 if the value of operating lease payments due in one year is greater than 1 percent of assets, and 0 otherwise.

OTHER_10K_LENGTH The total number of words contained in the 10-K outside of Items 1, 7, and 8.

RD Research and development expenditures scaled by lagged assets.

REDUNDANCY The percentage of total three-word sequences (trigrams) contained in the 10-K that occur in both the financial statement footnotes and earlier sections of the 10-K.

REF_AR Indicator variable equal to 1 if certain 10-K items are incorporated by reference, and 0 otherwise.

SIZE The natural logarithm of market value of equity.

SPI_D Indicator variable equal to 1 if the company reports any special items, and 0 otherwise.

TOTAL_LENGTH The total number of words in the 10-K.

Why are 10-K Filings So Long? 21

Accounting Horizons Volume 30, Number 1, 2016

Copyright of Accounting Horizons is the property of American Accounting Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.