A Comprehensive and Quantitative Internal Control Index:

Construction, Validation, and Impact

Hanwen Chen
Xiamen University
[email protected]

Wang Dong
Zhejiang University
[email protected]

Hongling Han
Zhejiang University
[email protected]

Nan Zhou
State University of New York at Binghamton
[email protected]

Current Version: December 12, 2013


We thank Stephanie Dehning Grimm (Discussant) and workshop participants at the 2013 American
Accounting Association Annual Meeting for their helpful comments. Hanwen Chen acknowledges the
research grants from the National Natural Science Foundation of China (No. 71332008) and from the
Education Ministry of China (No. 10JJD630003) for funding the construction of the internal control index
for all public firms in China. Wang Dong acknowledges the research grant from the National Natural
Science Foundation of China (No. 71302060). Hongling Han acknowledges the research grants from the
National Natural Science Foundation of China (No. 71272168) and from the Zhejiang Provincial Natural
Science Foundation of China (No. Y6110216). Please address all correspondence to Hongling Han,
Department of Accounting and Finance, School of Management, Zhejiang University, Hangzhou, China,
Zip 310058.

Electronic copy available at: http://ssrn.com/abstract=2368000

 A Comprehensive and Quantitative Internal Control Index:
Construction, Validation, and Impact

Abstract

We examine whether it is beneficial to improve internal control disclosure by

constructing an internal control index for 99% of all public firms in China, because the
lack of internal control regulations during our sample period presents an interesting
setting, in which the diversity of internal control quality is preserved. Two distinctive
features set our index apart from the information currently available under SOX 404 in
the U.S. First, it comprehensively evaluates a firm’s internal control, based on the COSO
framework. Second, it quantitatively measures a firm’s internal control, using the analytic
hierarchy process (AHP) designed for analyzing complex decisions. We proceed to
validate our index by confirming the known relation between internal control quality and
earnings management. Further, we theorize that our internal control index has a positive
impact on the earnings response coefficient (ERC), and find that better internal control
indeed makes financial reporting more credible to investors.

Keywords: Internal Control Index; COSO; AHP; Earnings Quality; Earnings Response
Coefficient

JEL Classification: M41

Electronic copy available at: http://ssrn.com/abstract=2368000

 A Comprehensive and Quantitative Internal Control Index:
Construction, Validation, and Impact

1. Introduction

Sarbanes-Oxley Act Section 404 (SOX 404) requires publicly traded U.S. firms to disclose

information on internal controls. 1 SOX 404 (a) requires management to file a report on the

company’s internal control over financial reporting (ICFR). The management report must

include (1) a statement of management's responsibility for ICFR, (2) management's assessment

of the effectiveness of ICFR, and (3) a statement identifying the framework used by

management. SOX 404 (b) requires the auditor to issue an attestation report on management’s

assessment of the company’s ICFR. Consisting of only required elements, the report for an

effective internal control, either from the management or the auditor, contains only parsimonious

information. For example, Deloitte and Touche certifies in Microsoft’s 10-K, filed on July 30,

2013, that “the Company maintained, in all material respects, effective internal control over

financial reporting as of June 30, 2013, based on the criteria established in Internal Control –

Integrated Framework (1992) issued by the Committee of Sponsoring Organizations of the

Treadway Commission.” The disclosure for an ineffective ICFR contains additional qualitative

information. For example, American Tower reports in its 10-K, filed on March 14, 2008, that

“We identified a material weakness in our internal control over financial reporting because we

failed to maintain effective controls over the accounting for income taxes.”

Recognizing the limitation in such mandatory disclosure under SOX 404, the current

literature finds evidence in favor of disclosing more information on internal control. Specifically,

1
An accelerated filer—a public firm with an equity market capitalization of more than $75 million—must comply
with SOX Section 404 for its first fiscal year ending on or after November 15, 2004. A non-accelerated filer must
comply with SOX 404(a) for its first fiscal year ending on or after December 15, 2007, but is exempt from
complying with SOX 404(b) under Section 989G of the Dodd-Frank Act.

1
a continuous variable is found to be superior to a dichotomous measure, as the severity of

internal control problems is better captured by the number of internal control weaknesses than by

the dummy for such weaknesses (Feng et al., 2009); a detailed breakdown is found to be more

informative than an aggregate summary, as additional insights can be gained by separating

internal control weaknesses into account-level- and company-level weaknesses (Doyle et al.,

2007; Elder et al., 2009); an in-depth analysis is found to be more valuable than a general

statement, as deficiency information at the level of the individual control is more informative to

auditors than the company-level reporting of effective or ineffective controls required under

SOX 404 (Bedard and Graham, 2012).

Building on this line of research, we investigate whether it is beneficial to improve the

quality of internal control disclosure. Specifically, we construct an internal control index that

covers 99.0% of all Chinese public firms from 2007 to 2010, with the remaining 1.0% missing

necessary internal control information.2 These sample firms are not subject to internal control

regulations, because the Basic Standards of Enterprise Internal Controls—dubbed as the

Chinese version of SOX 404—became effective for such firms on January 1, 2012. The

implementation of SOX 404 homogenizes the internal control quality for firms in the U.S.,

resulting in 95.1% of the accelerated filers reporting effective internal controls for fiscal years

between 2007 and 2012. Consequently, the lack of internal control regulations on Chinese firms

during our sample period presents an interesting setting, in which the heterogeneity of internal

control quality is preserved. Our focus on China in the pre-Chinese SOX 404 period generates

2
Following Wang et al. (2008) and Chen et al. (2010), we examine all firms listed on the Shanghai Stock Exchange
and on the Shenzhen Stock Exchange. Note that we exclude firms listed on the Entrepreneurship Section of the
Shenzhen Stock Exchange, because the Entrepreneurship Section was introduced by the Shenzhen Stock Exchange
in 2009.

2
the internal control diversity necessary for us to study the impact of differential internal control

quality on the capital market.

Two distinctive features set our internal control index apart from the information currently

available under SOX 404 in the U.S. First, our index is a comprehensive evaluation of a firm’s

internal control based on the Internal Control—Integrated Framework (1992) created by the

Committee of Sponsoring Organizations of the Treadway Commission (COSO). We use the

COSO framework, because this framework is identified by Securities Exchange Commission

(SEC) Release No. 33-8810 (June 27, 2007) as being suitable for management to assess the

effectiveness of the company’s ICFR.3 The COSO framework defined internal control as a

process, effected by an entity's board of directors, management and other personnel, designed to

provide reasonable assurance regarding the achievement of objectives in the following four

categories:

--effectiveness and efficiency of operations,

--reliability of financial reporting,
--compliance with applicable laws and regulations,
--safeguarding of assets.4

While specifically including “safeguarding of assets” in its ICFR definition, SEC

acknowledges that its ICFR definition only “encompasses the subset of internal controls

addressed in the COSO Report that pertains to financial reporting objectives” and “does not

encompass the elements of the COSO Report definition that relate to effectiveness and efficiency

of a company's operations and a company's compliance with applicable laws and regulations”

3
In the same release, SEC also cites the Guidance on Assessing Control published by the Canadian Institute of
Chartered Accountants (“CoCo”) and the report published by the Institute of Chartered Accountants in England &
Wales Internal Control: Guidance for Directors on the Combined Code (known as the Turnbull Report) as examples
of other suitable frameworks that issuers could choose in evaluating internal control effectiveness. SEC encourages
companies to examine and select a framework that may be useful in their own circumstances; SEC also encourages
the further development of existing and alternative frameworks.
4
While the first three categories are contained in the 1992 original COSO report, the fourth category “safe guarding
of assets” is added in the 1994 COSO addendum to the Reporting to External Parties volume of the COSO report.

3
(SEC Release No. 33-8238, June 5, 2003). To overcome this limitation, we follow the COSO

framework to also consider “effectiveness and efficiency of operations” and “compliance with

applicable laws and regulations." Relative to SOX 404’s focus on financial reporting, our

internal control index encompasses the full set of internal controls pertaining to the

aforementioned four categories in the COSO Report.

Second, our index is a quantitative measurement of a firm’s internal control using the

analytic hierarchy process (AHP) popular for analyzing decision problems.5 AHP is a structured

technique for organizing and analyzing complex decisions, providing a scientific and simple

method to effect a quantitative analysis for a qualitative problem. Consisting of different levels

with both qualitative and quantitative information, the evaluation system of internal control is

appropriate for the systematic application of AHP. According to Saaty (2008a), AHP first

decomposes a complex decision problem into a hierarchy of more easily comprehensible sub-

problems, and then produces the qualitative and quantitative analyses of each sub-problem. AHP

next makes pairwise comparisons of the same-level items in every sub-hierarchy, analyzing their

relative impact on an element in the hierarchy above them. As discussed in Saaty (2008b), a

critical feature of AHP is that human judgment, in addition to the underlying information, can be

used in performing the evaluations.

Our AHP approach to create the internal control index is more advanced than the weighted

average method used to construct several popular disclosure indices. In constructing the AIMR

5
Based on mathematics and psychology, AHP was developed by Thomas L. Saaty in the 1970s and has been
extensively used in a wide range of decision situations in in fields such as government, business, industry,
healthcare, and education. For a comprehensive review, please see “Analytic hierarchy process: An overview of
applications” by Vaidya and Kumar (2006). A number of AHP software packages and Excel templates are
commercially available. For example, Saaty's consulting company gives access to its AHP software at
http://www.superdecisions.com/.

4
Ranking (e.g., Lang and Lundholm, 1993, 1996),6 an industry-specific subcommittee of security

analysts evaluates the level of corporate disclosure for selected firms within the industry along

three dimensions: annual report information, voluntary disclosures in quarterly reports, and

disclosures from investor relation activities. The final AIMR Ranking is a weighted average of

the disclosure scores in these three categories. Similar weighted average methods are used to

create other disclosure indices, including the international CIFAR Index of average accounting

disclosures (e.g., LaPorta et al., 1998; Leuz et al., 2003) and Standard and Poor’s scores of

international firms’ disclosures (e.g., Khanna et al., 2004).7

Leuz and Wysocki (2006) point out that disclosure measures, such as the AIMR Ranking and

the international CIFAR Index, suffer from a number of limitations, in that the selection and

coding of the relevant disclosures are subjective, that they generally capture the existence of

particular disclosures—rather than the quality of those disclosures, and that the construction of a

single index assigns particular weights to the different disclosure items. Since this criticism also

applies to our internal control index, we first set out to establish the validity of our index. Since

Doyle et al. (2007) and Ashbaugh-Skaife et al. (2008) find that firms with effective internal

controls tend to have high accruals quality, we hypothesize that our internal control index will

also be related to earnings quality, if it properly measures internal control quality. Therefore, we

study the relation between our index and earnings management, and find a significant negative

relation between the two. Our finding that better internal controls improve firms’ financial

reporting provides a validation of our internal control index, because we are able to use the index

to confirm the known relation between internal control quality and earnings quality.

6
AIMR stands for the Association for Investment Management and Research. AIMR changed its name to CFA
Institute in 2004.
7
CIFAR stands for the Center for International Financial Analysis and Research.

5
 When we separate our internal control index into five underpinning COSO components—

control environment, risk assessment, control activities, information and communication, and

monitoring—we find that control environment and information and communication are

negatively related to earnings management. Measuring the firm’s strength in corporate

governance, internal auditing, and human resources, the control environment component

captures the fraud-prevention mechanism established in a firm. Therefore, firms with a strong

control environment or an established fraud-prevention mechanism have less earnings

management. Our finding on control environment is consistent with those studies that document

the importance of corporate governance in containing earnings management (Klein, 2002;

Hunton et al., 2010). The information and communication component is coded, based on whether

a firm has proper channels to collect information internally and externally, whether a firm can

communicate with the board of directors and with outside investors, whether a firm has accurate,

complete, and timely information in accounting or corporate governance, and whether a firm has

a whistle-blowing program to monitor management. Consequently, a firm that has a strong

information system can alert the board of directors on potential accounting irregularities and thus

reduce earnings management.

After establishing the validity of our index, we then investigate the impact of our internal

control index on the capital market. Specifically, we focus on the earnings response coefficient

(ERC) that measures the extent of a security’s abnormal market return in response to earnings

surprises, 8 because ERC captures investor responsiveness to earnings and measures investors’

perceived earnings quality (Dechow et al., 2010). We use both the short- and long window to

examine the relation between our index and ERC. We find that our internal control has a positive

impact on ERC, suggesting that better internal controls make financial reporting more credible to
8
Please see Kothari (2001) for a review of the ERC literature.

6
investors. This finding demonstrates that disclosing comprehensive and qualitative information

on internal controls can be beneficial to the capital market. When we separate our index into the

five COSO components, we find that the control environment component is responsible for

increasing ERC in the short-window analyses, whereas the information and communication

component is responsible for increasing ERC in the long-window analyses. This suggests that

strong internal controls and corporate governance enhance the perceived earnings quality in the

short run, whereas effective communications with investors improve the perceived earnings

quality in the long run.

Our paper contributes to the accounting literature and -practice by innovatively constructing

an internal control index based on the COSO framework and the AHP method. The

comprehensive evaluation system and quantitative scoring system developed in this paper could

form the basis for building similar indices in other markets or other settings. Our index is new to

the accounting literature, extending those aforementioned studies that are in favor of disclosing

more information on internal controls (Feng et al., 2009; Doyle et al., 2007; Elder et al., 2009;

Bedard and Graham, 2012). While internal control is found to be related to earnings quality,

proxyed by earnings properties, such as accruals (Doyle et al., 2007; Ashbaugh-Skaife et al.,

2008), or by external indicators, such as financial restatements (Feng and Li, 2011), we add to

the literature by finding that internal control quality is associated with ERC, a market-based

measure of earnings quality.9 Overall, the documented impact of internal control on the capital

market provides evidence that a comprehensive and quantitative assessment of internal control

can be beneficial to investors.

9
Dechow et al. (2010) classify earnings quality proxies into three categories: (1) properties of earnings that include
earnings persistence and accruals; earnings smoothness; asymmetric timeliness and timely loss recognition; (2)
investor responsiveness to earnings that includes the earnings response coefficient (ERC) or the R2 from the
earnings-returns model; and (3) external indicators of earnings misstatements included in Accounting and Auditing
Enforcement Releases (AAERs), restatements, and internal control procedure deficiencies.

7
 Our paper also contributes to the accounting literature by introducing the AHP approach to

transform a qualitative problem into a quantitative score. Per our prior discussion, the AHP

method is more sophisticated than the weighted average method used to construct several

popular disclosure indices, including the AIMR Ranking and the CIFAR Index. A large number

of studies explore the impact of AIMR Ranking of corporate disclosure on analyst forecasts

(Lang and Lundholm, 1993, 1996), corporate liquidity (Welker, 1995; Healy et al., 1999), cost of

equity capital (Botosan, 1996; Sengupta, 1998), and agency frictions (Nagar et al., 2003; Huang

and Zhang, 2012). Other studies use the international CIFAR index of average accounting

disclosures in annual reports (e.g., LaPorta et al., 1998; Leuz et al., 2003) and Standard and

Poor’s scores of international firms’ disclosures (e.g., Khanna et al., 2004). Different from these

studies, our study is the first to use AHP to create an index of internal control and to apply such

an index to investigate the impact of internal control on the capital market.

In addition, our internal control index has a significant influence on accounting practices.

First, the 2009 and 2010 indices were published in major financial newspapers in China. The

2009 index was simultaneously published by three authoritative financial newspapers in China:

China Securities Journal, Shanghai Securities Journal, and Securities Times on June 11, 2010,

whereas the 2010 index was simultaneously published by Shanghai Securities Journal and

Securities Times on September 6, 2011. Second, our index has garnered attention and recognition

from major firms in a variety of industries. For example, Deloitte highlighted this index on the

website of its Corporate Governance Center. Lutai Textile Inc. announced it on its corporate

website, when the firm was ranked among the top 100 firms with the best internal controls,

according to the ranking of our index. Third, our index has received recognition from regulators

8
in China. In an acknowledgement letter sent to our research team on May 18, 2013, 10 the Public

Company Monitoring Division of the Shanghai Stock Exchange writes:

“We have paid continuous attention to ‘the Internal Control Index for Chinese Public
Companies’ developed by your research team, since the index was published in Shanghai
Securities Journal and other financial newspaper in 2010. Your internal control index created
for the Chinese capital market has significant reference values for our efforts of monitoring
internal control and corporate governance, and will help listed companies improve their
internal control quality. We hope that you will continue to study problems related to internal
control for Chinese public companies and contribute to the orderly development of the
capital market in China.”

The rest of the paper is organized as follows. Section 2 provides the background information

on internal control regulations in China. Section 3 describes the construction of our internal

control index. Section 4 outlines the sample selection. Section 5 validates our internal control

index by examining its relation to earnings management. Section 6 applies our index to study the

impact of internal control on the capital market. Section 7 concludes the paper.

2. Background Information

After the enactment of SOX in 2002, following the Enron and MCI WorldCom accounting

scandals, internal controls have received more and more attention from countries around the

world. The Chinese government also considers strengthening internal controls a priority. Premier

Wen Jiabao emphasizes that “we need to enhance internal control systems, improve corporate

governance, and introduce innovative mechanisms by borrowing from the managerial experience

available abroad” in his State of the Government Report presented in the Fourth Plenary Meeting

of the Tenth People’s Congress on March 5, 2006.11 Following Premier Wen’s address, the State

Council (Cabinet) requested the Ministry of Finance to coordinate with related agencies and to

establish a set of rules and regulations to govern a firm’s internal control system. In June 2006,
10
The original letter with the seal and on letterhead is available upon request.
11
Reported by the official Xinhua News Agency on March 15, 2006.

9
the State-Owned Assets Supervision and Administration Commission (SOASAC), issued A

Guideline of Comprehensive Risk Management for Central Government Owned Enterprises,

requiring that central government-owned enterprises and state owned public companies develop

internal control systems and strengthen risk management. In July 2006, the Shanghai Stock

Exchange (SSE) issued A Guideline of Internal Controls for Listed Firms on Shanghai Stock

Exchange. Further, in September 2006, the Shenzhen Stock Exchange (SZSE) issued A

Guideline of Internal Controls for Listed Firms on Shenzhen Stock Exchange. These two

regulations at the stock exchange level establish rules of internal control implementation and -

reporting for publicly traded Chinese firms.

On July 15, 2006, five Chinese government authorities and regulatory bodies, including the

Ministry of Finance (MOF), the China Securities Regulatory Commission (CSRC), the National

Audit Office (NAO), the China Banking Regulatory Commission (CBRC), and the China

Insurance Regulatory Commission (CIRC), jointly established a Committee on Internal Control

Standards, aiming at research, and stipulated a set of universal, recognizable, and scientific rules

governing firms’ internal controls. After two years of conducting research and seeking feedback,

the Committee on Internal Control Standards issued The Basic Standards of Enterprise Internal

Controls. Dubbed as the Chinese version of SOX 404, the Basic Standards require that a listed

firm issue a self-assessment on its internal controls and that a Certified Public Accountant issue a

report on the firms’ internal controls. In order to prepare firms to comply with the Basic

Standards of Enterprise Internal Controls, the Committee on Internal Control Standards further

released Supplemental Guidelines of Firms’ Internal Controls on April 26, 2010. These

supplemental guidelines include 18 categories in Implementation Guidelines of Firms’ Internal

Controls, one category in Evaluation Guidelines for Evaluation and Assessment of Effectiveness

10
of Firms’ Internal Controls, and one category in Guidelines for Performing Assurance

Engagement in Relation to Assessing Effectiveness of Firms’ Internal Controls. The Basic

Standards of Enterprise Internal Controls became effective for firms dual-listed in stock

exchanges, both in China and abroad on January 1, 2011, and for firms listed on the Shanghai

Stock Exchange, as well as for firms listed on the main section of the Shenzhen Stock Exchange

on January 1, 2012. The enactment of these regulations marks the completion of developing an

internal control system for Chinese firms. Emphasizing risk management and fraud containment,

and building around control standards and evaluation criteria, the internal control system in

China is aimed at creating a properly structured and scientifically designed comprehensive

framework with distinctive levels and coherent sequences.

3. Construction of Internal Control Index

We construct an internal control index for listed firms in China by tracking their internal

control information from financial statements, CSRC filings, government documents, and press

releases. Following Wang et al. (2008) and Chen et al. (2010), we examine all firms listed on the

Shanghai Stock Exchange and on the Main Section of the Shenzhen Stock Exchange.12 Our final

index covers 99.0% of all Chinese public firms from 2007 to 2010 with the remaining 1.0%

missing necessary internal control information. The actual breakdown is as follows.

2007 2008 2009 2010

Companies with
Internal Control Indices 1,507 1,594 1,671 1,891
Total Public Companies 1,539 1,612 1,684 1,896
Percentage 0.979 0.989 0.992 0.997

12
We do not include firms in the Entrepreneurship Section of the Shenzhen Stock Exchange, because the
Entrepreneurship Section was introduced by the Shenzhen Stock Exchange in 2009.

11
 Based on COSO’s (1992) Internal Control−Integrated Framework and innovated to

accommodate unique Chinese conditions, our index considers controls not only related to the

reliability of financial reporting and safeguarding of assets, but also pertaining to the

effectiveness and efficiency of operations and compliance with laws and regulations.

Specifically, our index uses the COSO components—control environment, risk assessment,

control activities, information and communication, and monitoring—as the five first-level

criteria in internal control. These five elements advocated in the COSO framework optimize the

internal control structure and -system, integrate different views on internal controls, and create a

consensus platform and conceptual framework to evaluate internal control quality. Under each

level, there is a sequence of sub-level criteria. For example, the first-level criterion, Control

Environment, can be further divided into six second-level criteria: corporate governance, internal

auditing, human resources, employee quality, social responsibility, and corporate culture. Our

final index has four levels of evaluation criteria, consisting of five first-level criteria, 24 second-

level criteria, 43 third-level criteria, and 144 fourth-level criteria. In addition, the scores for first-

level criteria of control environment, control activities, and information and communication will

be deducted by a pre-specified percentage, if a firms receives regulatory penalties or media

exposure for violating the accounting rules or securities laws, or if a firm experiences major

accidents of operational safety, environmental pollution, or product recalls.

We then use AHP to transform the qualitative information obtained in the four levels of our

evaluation system into a quantitative measurement of a firm’s internal control. The first step in

applying AHP is to model the internal control evaluation problem as a hierarchy. First, we need

to analyze the decision problem in-depth, extracting relevant factors and determining the

relations among different factors. Second, we arrange different factors into an analytic hierarchy,

12
in which each factor belongs to a certain hierarchy and is assigned to a factor in the upper

hierarchy. The internal control evaluation system contains five hierarchies in the following top-

to-bottom order: overall objectives, sub-objectives, standards, sub-standards, and plan execution.

The overall objective level represents the internal control index; the sub-objective level consists

of the five components from the COSO internal control framework (the five first-level items);

and the standard level, the sub-standard level, and the plan execution level are all expansions of

the sub-objective level. Once the hierarchy is built for evaluating internal control, pairwise

comparisons of the same-level items are performed for the sub-objective level, analyzing their

relative importance to the same assigned element in the hierarchy above them. After a

comparison, each of the two items receives a score weighted according to its relative importance.

Appendix 1 presents the detailed description and construction of our internal control index.

4. Sample Selection

We obtain all our financial variables and stock price information from CSMAR and

RESSET. The CSMAR (China Stock Market & Accounting Research) Database is designed and

developed by GTA Information Technology, one of major providers of China data. CSMAR is

the comprehensive database for Chinese business research. CSMAR covers data about the

Chinese stock market, financial statements, and China Corporate Governance of Chinese Listed

Firms. RESSET is a vendor that specializes in providing financial information and service

products in China, and supplies data needed in empirical research and investment analysis in

economics, finance, and accounting areas. Because of the differential requirement in control

variables, our sample size for analyzing earnings management is different from that for analyzing

ERC.

13
 Our sample period is from 2007 to 2010, since this is the period for which we construct our

internal control index. The first column presents the subsample for earnings management

analyses. We intersect the CSMAR and RESSET databases and obtain 6,731 firm-year

observations for public firms in China. First, we exclude 118 observations in the financial

industry, 67 observations without internal control indices, 137 observations with missing

information needed to calculate measures of earnings management, and 246 observations with

missing control variables. This leaves us with a final sample of 6,163 observations for analyzing

earnings management. The second (third) column presents the subsample for short-window

(long-window) ERC analyses. We again start from the 6,731 firm-year observations from the

intersection of CSMAR and RESSET and exclude 118 observations in the financial industry and

67 observations without Internal Control Indices. We further exclude 242 (122) observations

with missing information needed to calculate cumulative abnormal returns (CARs) and 687 (845)

observations with missing control variables. This leaves us with a final sample of 5,617 (5,579)

observations for analyzing short-window (long-window) ERC.

5. Validation of Internal Control Index

5.1 Hypothesis Development

We first set out to establish the validity of our internal control index by confirming a known

relation between internal control quality and earnings quality. Since Doyle et al. (2007) and

Ashbaugh-Skaife et al. (2008) find that firms with effective internal controls tend to have high

accruals quality, we conjecture that our index will be related to earnings quality, if it properly

measures internal control quality. Specifically, we conjecture that better internal controls will

constrain firms’ opportunistic earnings management behavior and have the following hypothesis.

14
 Hypothesis 1. Our internal control index will have a negative impact on earnings
management.

5.2 Research Design

Using an Ordinary Least Square (OLS) regression to examine Hypothesis 1, we model

discretionary accruals as a function of our internal control index and a set of control variables in

Equation (1). For firm i in year t,

DiscretionaryAccrualsit   0  1IC _ INDEX it   2 LNMVit  3 SEGMENTit   4 FOREIGNit  5 STDSALESit

  6 STDCFOit   7 STit  8 LEVit  9 BIG 4it  10 M & A  11INVENTORY   12 OPERCYCLE
  13 MB   14 LNAGE   it (1)

where DiscretionaryAccruals are used to measure firms’ earnings management activities;

IC_INDEX is the Internal Control Index; LNMV is the natural logarithm of a firm’s market

value of equity; SEGMENT is the natural logarithm of a firm’s business segments; FOREIGN is

an indicator variable if a firm has foreign sales; STDSALES is the standard deviation of the ratio

of last three year revenues to total assets; STDCFOS is the standard deviations of the ratio of last

three operating cash flows to total assets; ST is a “special treatment” designation;13 LEV is the

ratio of total liabilities to total assets; BIG4 is an indicator variable for Big 4 clients; M&A is an

indicator variable for mergers and acquisitions in the last three years; INVENTORY is inventory

valued at cost divided by total assets; OPERCYCLE is the operating cycle; MB is the ratio of

market equity value to book equity value; and AGE is the natural logarithm of the number of

years since a firm’s founding.

13
ST stands for “special treatment.” When a public firm has abnormal financial or other conditions that make it
hard for investors to assess its business prospects, the security exchanges will apply special treatments to this firm’s
stock transactions, including (1) adding “ST” to the ticker symbol, and (2) limiting the daily stock price appreciation
or depreciation to no more than five percent.

15
 For the dependent variable, we use three measures of discretionary accruals to capture a

firm’s earning management activities. We first use MJDA to denote the discretionary accruals

from the modified Jones Model (Dechow et al., 2005). Since discretionary accrual estimates are

found to be correlated with firm performance (Kasznik, 1999; Kothari et al., 2002), we also

construct PMJDA as the absolute value of performance-controlled modified Jones Model

discretionary accruals. Following Kothari et al. (2005), PMJDA controls for performance by

adding the prior year’s ROA into the Modified Jones model, when estimating expected accruals.

Following Ashbaugh et al. (2003), we further construct PDCA as the absolute value of

performance-controlled discretionary current accruals. PDCA controls for performance by

including the prior year’s ROA in the estimation of expected discretionary current accruals.

Details are provided in Appendix 2. We estimate the discretionary accruals models in industry

cross-sections year by year.14

The internal control index (IC_INDEX), our main variable of interest, is described in Section

3 and Appendix 1. We follow Doyle et al. (2007) and Ashbaugh et al. (2008) for our set of

control variables and augment them with variables identified in other studies. Firm size

influences discretionary accruals, so we control for size by using the natural logarithm of market

value (LNMV) (Becker et al., 1998).15 Firms with more complex operations are more likely to

have noisier accruals, so we control for the natural logarithm of the number of business segments

(SEGMENT) and for whether the firm has foreign sales (FOREIGN). The volatility of firms’

operations is systematically related to the likelihood of estimation errors in accruals, so we

control for the standard deviations of sales (STDSALES) and operating cash flows (STDCFOS),

14
We obtain very similar results, when we measure discretionary accruals with either the Jones Model (Jones, 1991)
or the performance-matched Jones Model (Kothari et al., 2005).
15
We use the logarithmic transformation of market value (MV) and later firm age (AGE), because the distributions
for these two variables are skewed. Please see Table 2 for the univariate statistics on these two variables.

16
respectively (Francis et al., 2005). Firms under financial distress are more likely to engage in

earnings management, so we control for the “special treatment” designation (ST) and financial

leverage (LEV) (Reynolds and Francis, 2000). Big 4 firms are associated with high accounting

quality, so we control for the Big 4 clients (BIG4). Firms involved in mergers and acquisitions

are likely to have larger abnormal accruals, so we control for such mergers and acquisition

activities (M&A). Fast growing firms are likely to have imprecise accruals due to inventory

build-ups, so we control for inventory investments as a percentage of total assets (INVENTORY)

and use the market-to-book ratio to proxy for growth (MB). The firm’s operating cycle and age

is found to be related to accruals, so we control for the firm’s operating cycle (OPERCYCLE)

and age (LNAGE) (Francis et al., 2005; Stuben, 2010). We include dummy variables in our

regression models to control for industry- and year fixed effects.

5.3 Empirical Findings

Table 2 presents the descriptive statistics for the discretionary accruals sample. The means

(medians) are 0.084 (0.055) for MJDA, the modified Jones model discretionary accruals

(Dechow et al., 1995); 0.079 (0.052) for PMJDA, the performance-controlled modified Jones

model discretionary accruals (Kothari et al., 2005); and 0.079 (0.050) for PDCA, the

performance-controlled discretionary current accruals (Ashbaugh et al., 2003); suggesting that

these three accruals measures are equally effective in capturing the firms’ earnings management

activities. These mean absolute values of discretionary accruals, according to three different

measures, are about 8 percent of beginning total assets, suggesting that the amounts are both

economically and statistically significant.

17
 The mean value of our internal control index is 0.360, suggesting that our sample firms on

average receive only 36.0% of the maximum possible points. The distribution of our index is not

skewed, since the median value is 0.358. Further, our index means are 0.286 for 2007, 0.357 for

2008, 0.386 for 2009, and 0.402 for 2010, with the year-to-year increases in average index scores

significant at the one-percent level. This implies that Chinese firms have started to strengthen

internal controls, and yet still need to improve in many areas.

In addition, our sample firms on average have RMB 7,093 million in market equity value

(MV), two business segments (SEGMENT), 0.54 debt-to-asset ratio (LEV), 4.34 market-to-book

ratio of (MB), 0.175 inventory-to-asset ratio (INVENTORY), and 9.6 years of age (AGE).

43.5% of our sample firms have foreign sales (FOREIGN), 1.8% experience mergers and

acquisitions in the last three years (M&A), 7.3% received a “special treatment” designation for

abnormal financial conditions (ST), and 6.1% use Big 4 international auditors (BIG4). The

standard deviation of revenues (STDSALES) and that of operating cash flows (STDCFOS) have

a mean of 0.138 and 0.052, respectively.

Table 3 uses OLS regressions to examine the effect of our internal control index on measures

of earnings management. In Model 1, we measure the dependent variable earnings quality with

MJDA, the absolute value of Jones model discretionary accruals proposed in Dechow et al.

(1995). Confirming Hypothesis 1, we find that the coefficient on our index is significantly

negative at the one-percent level, suggesting that high internal control quality can be effective in

reducing earnings management activities. The coefficients on LNMV, STDSALES, STDCFOS,

ST, LEV, INVENTORY, and OPERCYCLE are significantly positive at the five-percent level or

better, suggesting that earnings management activities are likely to be greater for firms with a

larger market equity value, greater volatility in sales or operating cash flows, with a ST

18
designation, higher leverage, more inventory build-ups, and longer operating cycle. The

coefficient on LNAGE is significantly negative at the one-percent level, suggesting that firms

with fewer years in existence are more likely to engage in earnings management activities. The

results on these control variables are similar to those in Doyle et al. (2007) and Ashbaugh et al.

(2008), lending further credence to our findings. In Model 2, we measure earnings quality with

PMJDA, the performance-controlled modified Jones model discretionary accruals (Kothari et al.,

2005). In Model 3, we measure earnings quality with PDCA, the performance-controlled

discretionary current accruals (Ashbaugh et al., 2003). Our findings in Models 2 and 3 are very

similar to those reported in Model 1, suggesting that our results are robust to alternative

measures of discretionary accruals. Since our overall results in Models 1-3 are consistent with

the findings in Doyle et al. (2007) and Ashbaugh et al. (2008), we establish the validity of our

index by confirming the known relation between internal control quality and earnings quality.

In Table 4, we further separate our index into its five components: control environment

(CtrEnv), risk assessment (Risk), control activities (CtrAct), information and communication

(InfoCom), and monitoring (Monitor). In Model 1 with MJDA as the dependent variable, the

coefficients on CtrEnv and InfoCom are both significantly negative at the five-percent level.

Measuring the firm’s strength in corporate governance, internal auditing, and human resources,

CtrEnv, the control environment component, captures the fraud-prevention mechanism

established in a firm. Therefore, firms with a strong control environment or with an established

fraud-prevention mechanism have less earnings management. COSO (1992) articulates that “the

control environment is the foundation for all other components of internal control.” Our results

lend support to the importance of the control environment. Our findings are also consistent with

those in Doyle et al. (2007), who document that company-level weaknesses, related to control

19
environment, have a greater impact on a firm’s accrual quality. Company-level weaknesses are

harder to detect by auditing, whereas account-level weaknesses, related to specific accounts and

transactions, are easier to detect by auditing.

The information and communication component, InfoCom, is coded based on whether a firm

has proper channels to collect information internally and externally; whether a firm can

communicate with the board of directors and with outside investors; whether a firm has accurate,

complete, and timely information in accounting or corporate governance; and whether a firm has

a whistle-blowing program to monitor the management. Because enhanced corporate

transparency and reduced information asymmetry can improve the monitoring of management

and thus restrict executives’ opportunistic behaviors (Bushman and Smith, 2001; Jo and Kim,

2007), a firm with a strong information system can alert the board of directors about potential

accounting irregularities and thus reduce earnings management. Our results are very similar

when we replace the dependent variable MJDA with PMJDA in Model 2 or with PDCA in

Model 3. In addition, the signs and significance levels on control variables are similar to those

reported in Table 3.

6. Impact of Internal Control Index

6.1 A Simple Model of Internal Control and ERC

Dechow et al. (2010) classify the earnings quality proxies into three categories: (1) properties

of earnings that include earnings persistence and accruals, earnings smoothness, asymmetric

timeliness and timely loss recognition, target beating, and earnings management; (2) investor

responsiveness to earnings that includes ERC or the R2 from the earnings-returns model; and (3)

external indicators of earnings misstatements that include Accounting and Auditing Enforcement

20
Releases (AAERs), restatements, and internal control procedure deficiencies. Using earning

quality proxies in Category (1), Doyle et al. (2007) and Ashbaugh-Skaife et al. (2008) find that

firms with internal control problems tend to have lower accruals quality. Using earnings quality

proxies in Category (3), Feng and Li (2011) find that internal control quality is negatively

associated with the likelihood of material misstatements. Since the proxies in all three categories

are intended to capture earnings quality, we conjecture that internal control quality will also have

an impact on ERC, the investor responsiveness to earnings in Category (2).

We build a simple one-period model to formalize this relation between internal control and

ERC. Following the information signal model developed by Holthausen and Verrechia (1988),

we make an adaptation, similar to adaptations proposed by Teoh and Wong (1993) and Francis

and Ke (2006), to allow consideration of internal control quality. Specifically, we let ̃ be the

firm’s disclosed value in the financial report. The variable ̃ communicates the firm’s true value

̃ to the capital market with an accounting error ̃ . Mathematically,

̃ ̃ ̃

where the firm’s true value ̃ ( ) and the accounting error ̃ ( ). Similar to Teoh

and Wong (1993), we assume that the precision level for the firm’s true value is a positive

constant.16 Different from Teoh and Wong (1993), we model the precision level for the

accounting error as a strictly increasing function of internal control quality I.

( )

where ( ) . The modeling of is supported by the research evidence in Doyle at al. (2007)

16
Since the firm’s true value ̃ can be viewed as the net present value of all future cash flows (Teo and Wong,
1993), is related to the uncertainty of the firm’s future cash flows (Francis and Ke, 2006).

21
and Ashbaugh et al. (2008) that internal control weaknesses can lead to low accruals quality,17

and also by the research evidence in Chan et al. (2008) that improved internal controls under

SOX 404 can reduce intentional or unintentional accounting errors.

Let and denote a firm’s stock price at time T-1 and T, respectively. The firm

releases its reported value y in time T. Under the efficient market hypothesis, we have the

following.

( ̃)

and

(̃ ̃ ) ( ) ( )

where is obtained by the properties of the multivariate normal distribution. Therefore, the

change in stock price for this firm due to the release of its financial report y is as follows.

( )

where is reported earnings and is expected earnings. Since ERC is intended to capture the

relation between equity returns and unexpected earnings, we have the following.

where is the stock return in response to unexpected earnings . Because the financial

report precision is a strictly increasing function of internal control quality or ( ) with

( ) we have the following.

17
Similar results have been found for financial institutions (Altamuro and Beatty, 2010) and for Canadian firms (Lu
et al., 2011).

22
 ()
()

Taking a first-order derivative with respect to internal control quality I, we have the

following.

()
[ ( )]

because and ( ) . Since implies that ERC is an increasing function of

internal control quality, we have the following hypothesis.

Hypothesis 2. Our internal control index will have a positive impact on ERC.

6.2 Research Design

We use an Ordinary Least Square (OLS) model to test Hypothesis 2. For firm i in year t,

CARit  0  UE * (1   2 IC _ INDEX it  3 NEGit   4 LNMV it  5 LEVit  6 MBit  7 BETAit )   it (2)

where CAR is the cumulative abnormal return, UE is unexpected earnings, NEG is an indicator

variable that takes the value of one, if unexpected earnings are negative, LNMV is the natural

logarithm of the firm’s market value, LEV is the ratio of total liabilities to total assets, MB is the

ratio of market equity value to book equity value, and BETA is the firm’s beta. Detailed

definitions are provided in Appendix 2.

Following Balsam et al. (2003) and Baber et al. (2013), we study both the short-window ERC

and the long-window ERC. As discussed in Baber et al. (2013), if the value-relevant information

contained in earnings is disclosed at the earnings announcement date, then the short-window

ERC is superior to the long-window ERC; if the value-relevant information is revealed

throughout the year, then the long-window ERC is better than the short-term ERC. To estimate

23
the short-window ERC, we adopt an event-study approach and use the dependent variable

CAR_S, defined as the three-day market-adjusted abnormal returns covering the day before, the

day of, and the day after a firm’s earnings announcement. 18 UE is the unexpected earnings per

share, which is equal to the difference between the EPS before extraordinary items for this year

and that for last year, divided by the closing price of the last trading day in April. The short-

window ERC is estimated from Equation (2) by using CAR_S. To estimate the long-window

ERC, we adopt an association approach and use the dependent variable CAR_L, defined as the

twelve-month market-adjusted abnormal monthly return from May in the current year to April in

the following year.

Following Balsam et al. (2003), the internal control index, our main variable of interest, and

control variables are entered into Equation (2) as interactive terms with UE—the earnings

surprise variable—because these variables are expected to influence ERC, a slope coefficient.

After controlling for the other factors, β2 captures the impact of our index on ERC. According to

Hypothesis 2, our index will have a positive impact on ERC, and thus we expect β2 to be

positive. If the earnings of firms with strong internal controls (large index score) are perceived as

more credible, then ERC will be larger for those firms. We control for NEG, an indicator

variable for firms with negative unexpected earnings, because the market views negative

unexpected earnings and positive unexpected earnings differently (Hayn 1995; Basu, 1997;

Balsam, 2003). We also control for firm size—measured as the natural logarithm of market value

(LNMV), following (Bowen et al., 1992)—and growth, measured as the ratio of market equity

18
We use the three-day market-adjusted cumulative abnormal returns to be consistent with our measurement of
long-term cumulative abnormal returns. Our results remain unchanged, when we use the three-day market-model
adjusted cumulative abnormal returns, covering the day before, the day of, and the day after a firm’s earnings
announcement. We estimate market model parameters with daily stock returns over the 250-trading-day estimation
window ending twenty days before the earnings announcement. We require firms to have positive trading volumes
for at least 100 days during this estimation window and eliminate those firms that fail to meet this requirement.

24
value to book equity value (MB), following Collins and Kothari (1989) and Hackenbrack and

Hogan (2002). We control for leverage (LEV), because ERC for a firm with high leverage is

found to be lower than that of a firm with low leverage (Dhaliwal et al., 1991). We further

control for BETA, a measure of systematic risk, because it affects the expected rate of return and

thus can have a negative effect on ERC (Collins and Kothari, 1989; Easton and Zmijewski,

1989). We include dummy variables in our regression models to control for industry- and year

fixed effects.

6.3 Empirical Findings

Table 5 presents the descriptive statistics for the ERC sample. For the short-window

analyses, the sample size is 5,617 firm-year observations. The mean (median) CAR_S—the

three- day market-adjusted cumulative abnormal returns around earnings announcement dates—

is 0.001 (-0.002); whereas the mean (median) RET_S—the three-day unadjusted cumulative raw

returns around earnings announcement dates—is -0.001 (0.000). For the long-window analyses,

the sample size is 5,579 firm-year observations. The mean (median) CAR_L—the one-year

market-adjusted abnormal returns from May in the current year to April in the following year—is

16.2% (13.2%), indicating that our sample firms with the index scores significantly outperform

the capital market over our sample period from 2007 to 2010. Similarly, the mean (median) mean

RET_L—the one year raw return from May in the current year to April in the following year—is

22.8% (19.1%) over our sample period. The mean (median) UE—the unexpected earnings per

share—is -0.002 (-0.003). The means and medians for control variables are similar to those

reported for the discretionary accruals sample in Table 2.

Table 6 uses OLS regressions to examine the effect of our internal control index on ERC.

25
Models 1 and 2 present the short-window ERC results. Model 1 uses the dependent variable

CAR_S, the three-day market-adjusted cumulative abnormal returns. Confirming Hypothesis 2,

we find that the coefficient on UE*IC_INDEX, β2 in Equation (2), is significantly positive at the

one-percent level, suggesting a positive impact of our index on the short-window ERC. The

short-window ERC is larger for firms with a higher index score, implying that the earnings of

firms with strong internal controls are perceived as more credible over the three-day earnings

announcement windows. Model 2 uses the dependent variable RET_S, the three-day cumulative

raw returns, and shows similar results to those reported for Model 1.

Models 3 and 4 present the long-window ERC results. Model 3 uses the dependent variable

CAR_L, the twelve-month market-adjusted abnormal monthly return from May in the current

year to April in the following year. Again confirming Hypothesis 2, we find that the coefficient

on UE*IC_INDEX, β2 in Equation (2), is significantly positive at the five-percent level,

suggesting a positive association between our index and the long-window ERC. The long-

window ERC is larger for firms with a higher index score, implying that the earnings of firms

with strong internal controls are perceived as more credible over the one-year period following

their earnings announcements. Model 4 uses the dependent variable RET_L—the twelve-month

raw return from May in the current year to April in the following year—showing similar results

to those reported in Model 3. The coefficients on control variables in Models 1-4 are largely

consistent with those reported in Chan et al. (2012), in terms of signs and significance levels.

Table 7 studies our index’s five components: control environment (CtrEnv), risk assessment

(Risk), control activities (CtrAct), information and communication (InfoCom), and monitoring

(Monitor). Models 1 and 2 present the short-window ERC results. In Model 1 with the dependent

variable CAR_S, the coefficient on UE*CtrEnv is significantly positive at the one-percent level,

26
suggesting that internal control quality affects the short-window ERC through the control

environment component (CtrEnv). Measuring the firm’s strength in corporate governance,

internal auditing, and human resources, the control environment component captures the fraud-

prevention mechanism established in a firm. Because the control environment component is

found in Table 4 to reduce earnings management significantly, we find here that a strong control

environment will make a firm’s financial statements more credible, enhancing a firm’s perceived

earnings quality around its earnings announcement date.

Models 3 and 4 present the long-window ERC results. In Model 3 with the dependent

variable CAR_L, the coefficient on UE*InfoCom is significantly positive at the five-percent

level, suggesting that internal control quality affects the long-window ERC through the

information and communication component (InfoCom). InfoCom is coded based on whether a

firm has proper channels to collect information internally and externally; whether a firm can

communicate with the board of directors and with outside investors; whether a firm has accurate,

complete, and timely information in accounting or corporate governance; and whether a firm has

a whistle-blowing program to monitor management. Because a firm that has a strong information

system can communicate effectively to both board and investors, we find that the information

and communication component is responsible for enhancing the perceived earnings quality in the

long-term. This component is related to the long-term cumulative abnormal returns, because

firms with strong communications can interact with the capital market effectively, and

information about such firms is likely to be reflected in these firms’ stock prices in a timely

manner. Model 4 uses RET_L and obtains results similar to those reported in Model 3. Overall,

our Table 7 results suggest that strong internal control and corporate governance enhance the

27
perceived earnings quality in the short run, whereas effective communications with investors

improve the perceived earnings quality in the long run.

7. Conclusion

We examine whether it is beneficial to improve internal control disclosure by constructing an

internal control index for 99% of all public firms in China, because the lack of internal control

regulations during our sample period presents an interesting setting in which the diversity of

internal control quality is preserved. Two distinctive features set our internal control index apart

from the information currently available under SOX 404 in the U.S. First, our index

comprehensively evaluates a firm’s internal control based on the COSO framework, expanding

the scope from controls over financial reporting to those over business operations and regulatory

compliance. Specifically, our index builds upon the five COSO elements integral to internal

control: control environment, risk assessment, control activities, information and

communication, and monitoring. Second, our index quantitatively measures a firm’s internal

control using AHP, which is designed for analyzing complex decisions. We find that our internal

control index is negatively associated with earnings management, suggesting that better internal

controls improve firms’ financial reporting. This finding validates our index by confirming a

known relation between internal control quality and earnings quality. Moreover, we find that our

index has a positive impact on ERC, suggesting that better internal controls make financial

reporting more credible to investors. Overall, our findings suggest that a comprehensive and

quantitative evaluation of internal control can be beneficial to the capital market. The

comprehensive evaluation system and quantitative scoring system developed in this paper could

form the basis for building similar indices in other markets or other settings.

28
Appendix 1. Creation of Internal Control Index
I. COSO Based Evaluation System
Released by COSO in 1992, the Internal Control—Integrated Framework is by far the most
authoritative literature on internal control. In this framework, internal control contains five
components: control environment, risk assessment, control activities, information and
communication, and monitoring. Because the five-component structure optimizes the internal
control system, integrates different understandings, and constructs a comprehensive conceptual
framework; it enables the COSO framework to gain acceptance by regulators, such as SEC and
PCAOB, and to form the foundation of SOX 404 on internal controls.

The five government agencies, including the Ministry of Finance, the China Securities
Regulatory Commission (CSRC), the National Audit Office, the Banking Regulatory
Commission, and the Insurance Regulatory Commission, jointly issued The Basic Standards for
Enterprise Internal Control (Basic Standards) on May 22, 2008 and Implementation Guidelines
for Enterprise Internal Control (Guidelines) on April 26, 2010. The Guidelines contain three
parts: Guidelines for Application of Enterprise Internal Control, Guidelines for Evaluation of
Enterprise Internal Control, and Guidelines for Audit of Enterprise Internal Control.
Often referred to as “the SOX in China,” the Basic Standards are the most authoritative
regulations on internal control in China. The Basic Standards and the Guidelines use the COSO
framework as a reference, and take into consideration the internal control implementation
situation in China. In the Basic Standards and the Guidelines, internal control contains the same
five components proposed in the COSO framework. Mainly based on the Basic Standard and the
Guidelines, the design of our index also takes into consideration the requirements in other
regulations, including Internal Control Guidelines for Public Companies Listed in Shenzhen
Stock Exchange, Internal Control Guidelines for Public Companies Listed in Shanghai Stock
Exchange, Code of Corporate Governance for Listed Companies in China, Company Law of the
People's Republic of China, and Guidelines for the Articles of Association of Listed Companies.
The design of our index further incorporates the related internal control research on international
firms and Chinese firms,
Our internal control starts with five first-level items, the same as the five components in the
COSO framework. Each of the first-level items will then contain more evaluation items. As
shown in Appendix Table C, the final evaluation system contains four levels. A code is given to
each item at each level. For example, the codes for the first item in the first three levels are IC1,
IC11 and IC111, respectively. For the fourth level, the code is defined as, for example, IC11101.
The first three numbers (111) represent the first three levels to which this item is affiliated. If
there is no third level for a particular item, then the first two numbers are presented. Overall,
there are five first-level items, 24 second-level items, 43 third-level items, and 144 fourth-level
items. In addition, there is a special sub-category—Punishment and Other Negative Event—for
three first-level items: Control Environment, Control Activities, and Information and
Communication). If a punishment or other negative event exists, then the overall score for this
first-level item will be reduced at a pre-specified percentage.
II. AHP Based Scoring System

29
Since each item in the evaluation system has differential influence on a firm’s internal control,
this means that we need to weight each item differently to improve credibility of the
comprehensive internal control index. Consisting of different levels with both qualitative and
quantitative information, the internal control evaluation system is appropriate for the systematic
application of the analytic hierarchy process (AHP). AHP is a structured technique for
organizing and analyzing complex decisions, providing a scientific and simple method to
perform a quantitative analysis for a qualitative problem. Based on mathematics and psychology,
it was developed by Thomas L. Saaty in the 1970s and has been extensively used in a wide range
of decision situations. According to Saaty (2008a), AHP first decomposes a complex decision
problem into a hierarchy of more easily comprehended sub-problems and then carries out the
qualitative and quantitative analyses of each sub-problem. AHP next makes pairwise
comparisons of the same-level items in every sub-hierarchy, analyzing their relative impact on an
element in the hierarchy above them. As discussed in Saaty (2008b), a critical feature of AHP is
that human judgments, in addition to the underlying information, can be used in performing the
evaluations. The AHP analysis process of our system is as follows.
(i) Hierarchy Construction

The first step in applying AHP is to model the internal control evaluation problem as a hierarchy.
First, we need to analyze the decision problem in-depth, extracting relevant factors and
determining the relations among different factors. Second, we arrange different factors into an
analytic hierarchy, in which each factor belongs to a certain hierarchy and is assigned to a factor
in the upper hierarchy. The internal control evaluation system contains five hierarchies in the
following top-to-bottom order: overall objectives, sub-objectives, standards, sub-standards, and
plan executions. The overall objective level represents the internal control index; the sub-
objective level contains the five components from the COSO internal control framework (the
five first-level items); and the standard level, the sub-standard level, and the plan execution level
are all expansions of the sub-objective level. As shown in Appendix Figure 1, each item in the
2nd, 3rd, 4th level belongs to one and only one item from the upper level.

Appendix Figure 1. Hierarchy for the Internal Control Evaluation System

Overall objective Internal control index

Sub-objective IC1 control environment IC2 risk assessment IC3 control activities IC4 IC5

Standard IC11:corporation governance …… …… …… …… ……

Sub-standard IC111:nstitution establishment

……

Plan execution IC11101 IC11102 IC11103

30
(ii) Judgment Matrix
Once the hierarchy is built for evaluating internal control, pairwise comparisons of the same-
level items are performed for the sub-objective level, analyzing their relative importance to the
same assigned elements in the hierarchy above them. After the comparison, each of the two
items receives a score according to its relative importance. To perform the pairwise comparisons,
we need to establish a judgment matrix according to Saaty’s AHP 1-9 Scale presented in
Appendix Table A. The AHP 1-9 Scale is aimed at improving judgment accuracy and, therefore,
weight credibility.
i

Appendix Table A. AHP 1-9 Scale

Serial Number Importance Level Value

(#)

1 Two factors have equal importance 1

2 The former one is slightly more important 3

3 The former one is strongly more important 5

4 The former one is very strongly more important 7

5 The former one is extremely more important 9

6 The former one is slightly unimportant 1/3

7 The former one is strongly unimportant 1/5

8 The former one is very strongly unimportant 1/7

9 The former one is extremely unimportant 1/9

10 Between 2,4,6,8,1/2,1/4,1/6,1/8

The above technique uses the Delphi method to create the judgment matrix. Under this approach,
experts compare the relative importance between two factors, and assign a value to each factor
based on the AHP 1-9 Scale. Appendix Table B presents the judgment matrix for items at the
sub-objective level. For example, Control Environment is judged to “be slightly more important”
than Risk Assessment, so the value for Control Environment is 3 according to #2 on the AHP 1-9
Scale, and the value for Risk Assessment is 1/3 according to #6 on the AHP 1-9 Scale. The rest of
the values are given analogously. Note that a value of 2 means that it is between a value of 1 and
a value of 3, according to #10 on the AHP 1-9 Scale.

31
 Appendix Table B. Judgment Matrix for Items at the Sub-Objective Level

Control Risk Control Information & Monitoring

Environment Assessment Activities Communication

Control Environment 1 3 2 3 3

Risk Assessment 1/3 1 1/2 1 1

Control Activities 1/2 2 1 2 2

Information & 1/3 1 1/2 1 1

Communication

Monitoring 1/3 1 1/2 1 1

(iii) Weight Calculation and Consistency Check

There are several different methods to calculate the item weight under AHP, including the
eigenvalue method, the logarithmic least square method, and the upper triangular matrix method.
The internal control evaluation system uses the eigenvalue method to calculate the weight. If the
judgment matrix A is a consistent n-dimensional matrix, then the normalized eigenvector from
the solution of is the weight for the related factor, where ( ) . In
addition, we need to check the consistency of the judgment matrix by examining the consistency
ratio. If the ratio is greater than 0.1, the matrix needs to be adjusted. The consistency ratio and
the weight will be calculated automatically when the judgment matrix, such as the one presented
in Appendix Table B, is entered into the YAAHP software.19 Analogously, we can obtain the
weight for each item at the standard level and at the sub-standard level.

Because there are too many items at the plan execution level, it is difficult to perform the
pairwise comparisons. As a result, we use the objective evaluation method to calculate the
weight based on observed values. The objective evaluation method uses the variation coefficient.
Under the variation coefficient method, the weight is based on the variation coefficient that
reflects the differences in information between items. According to Li (2005), “the variation
coefficient method reflects the dynamics of the weight. Since there is not a long history of the
listed company development and the whole environment is not steady, the weight from this
method can better reflect the differences between items.”20 For example, there are three fourth-
level items (IC11101, IC11102, and IC11103) under the item Institution Establishment (IC111).
If the variation coefficient for IC11101 accounts for 1/4 of the sum of the variation coefficients
for these three fourth-level items, then the weight for IC11101 is 0.25.

19
YAAHP is software for AHP, available for download from the website of Beijing Xinshengyun Software Co.
(http://www.foreology.com/product/yaahp.html). Many other AHP software packages and Excel templates are
commercially available. For example, Saaty's consulting company gives access to its AHP software at
http://www.superdecisions.com/.
20
Li, Weian. 2005. Corporation Governance Evaluation and Index Research. Higher Education Press: Beijing.

32
Using the aforementioned methods, we can obtain the weight for each item in the hierarchy.
Without loss of generality, the impact factor for each item at the fourth level is the product of all
the weights for items on its hierarchical path, including the weight of the fourth level item and
excluding the weight of the first-level item. For example, let us focus on the leftmost path in the
internal control evaluation hierarchy shown in Appendix Figure 1. If the weights are 0.25 for
IC11101, 0.2 for IC111, and 0.4 for IC11, respectively, then the impact factor on our index for
IC11101 is calculated as the product of these three values or 0.25*0.2*0.4=0.02.

(iv) Calculation of Internal Control Index

Based on the observed values and calculated weights, our internal control index is the weighted
average of each item. For those items whose scores are listed as “To be standardized” at the
fourth level, the standardized score is calculated as the actual score on this item for the evaluated
firm divided by the maximum score on the same item from all the listed companies. For
example, if the actual number of words in the internal control evaluation report is 5,911 for a
particular firm and the maximum number of words from all internal control evaluation reports is
10,000, then the standardized score for this item is 0.5911 (5,911/10,000). For other items at the
fourth level, the score is calculated according to the evaluation standard, such as 1 for yes and 0
for no. The weighted average calculation method is as follows.

where is our overall internal control index, is the control environment index, is
the risk assessment index, is the control activities index, is the information and
communication index, is the monitoring Index, and is the weight of the ith item at the first
level (i=1,2,3,4,5). Further,

(∑ ) ( )

where is the value of the jth item at the fourth level associated with the ith item at the first
level, is the impact factor of the jth item at the fourth level on the associated ith item at the first
level, Pi is the deduction ratio for the special category—Punishment and Other Negative Event—
at the first level. For example, Pi is equal to 50% if a firm’s executives, directors, or supervisors
receive any punishment from the Ministry of Justice, Ministry of Finance, CSRC or
Shanghai/Shenzhen Stock Exchange.

Our final internal control index is a percentage score with the highest being 100% and the lowest
being 0.

33
 Appendix Table C. Internal Control Evaluation System

First Level Second Third Level Fourth Level Score Example

Level
Sub- Standard Sub-Standard Level Plan Execution Level Xiamen
Objective Level Guomao
Level (Code
600755)

IC11101：Any manual or 1 for yes, 0 1.00

guideline on internal control in for no
the company?
IC111： IC11102：Any outside 1 for yes, 0 1.00
organization to help the for no
Institutional
Arrangement company improve its internal
control quality?

IC11103：A cross-listed 1 for yes, 0 0.00

company? for no
IC11：Corporation Governance

IC11201：Chairman of the 1 for no, 0 for 0.00

IC1：Control Environment

board also the controlling yes

shareholder?

IC11202：Ownership To be 0.02
IC112： percentage of institutional standardized
shareholders
Shareholders
Composition IC11203：Number of To be 4.00
institutional shareholders in the standardized
top 10 shareholder list

IC11204：Average percentage To be 0.30

of shareholders taking part in standardized
the shareholder meeting

IC11301：Board Size 1 for 5-19 1.00

person, 0
IC113： otherwise
Board of Directors IC11302：Number of board To be 6.00
committees standardized

34
 IC11303：Any committee in 1 for yes, 0 1.00
charge of internal control for no
(audit committee or risk
management committee at the
board level)?

IC11304：Number of audit To be 3.00

committee members standardized

IC11305：Percentage of To be 0.33
independent directors on the standardized
board

IC11306：Average attendance To be 0.94

ratio of board meetings for standardized
independent directors

IC11307：Number of To be 0.00
dissenting proposals by standardized
independent directors

IC11401：Number of To be 3.00
members in the board of standardized
supervisors

IC11402：How is the 0 if by 1.00

employee supervisor controlling
nominated? shareholder,1
otherwise
IC114：
IC11403：Percentage of To be 0.33
Board of
employee supervisors standardized
Supervisors
IC11404：Any 1 for both, 0 0.00
legal/accounting experience for otherwise
supervisors?

IC11405：Number of To be 0.00
dissenting proposals by standardized
supervisors

IC115： IC11501：Is the board 1 for no, 0 for 1.00

chairman and the CEO one yes
Management person?

35
 IC11502：Percentage of To be 0.33
executives on the board standardized

IC11503：Any complex salary 1 for yes, 0 0.00

structure for the management? for no

IC11504：Percentage of To be 0.00
shares owned by the standardized
management

IC11505：Are the shares 1 for yes, 0 1.00

owned by the management for no
available for circulation?

IC12101：Any internal control 1 for yes, 0 1.00

department in the company (at for no
IC12：Internal

IC121：
the execution level)?
Audit

Internal Control
IC12102：Does the internal 1 for yes, 0 1.00
Implementation
audit department report to the for no
board?

IC13101: Any human resource 1 for yes, 0 1.00

policy? for no

IC13102: Does the human 1 for yes, 0 1.00

resource policy contain for no
provisions on recruiting?

IC13103: Does the human 1 for yes, 0 1.00

IC13：Human Resource

resource policy contain for no

provisions on training?
IC131：
IC13104: Does the human 1 for yes, 0 0.00
Recruiting and resource policy contain for no
Training provisions on job transferring?

IC13105: Is the company 1 for yes, 0 1.00

considered as a Best for no
Employer?

36
 IC13201: Does the human 1 for yes, 0 1.00
resource policy contain for no
provisions on salaries?

IC13202: Does the human 1 for yes, 0 1.00

resource policy contain for no
provisions on evaluations?

IC13203: Does the human 1 for yes, 0 1.00

resource policy contain for no
IC132：
provisions on promotion?
Incentives
IC13204: Does the human 1 for yes, 0 1.00
resource policy have for no
previsions on rewards or
punishments?

IC13205: Is the salary linked 1 for yes, 0 1.00

to performance? for no

IC13206:Percentage of shares To be 0.00

owned by employees standardized

IC13301: Does the human 1 for yes, 0 0.00

resource policy contain for no
provisions on discharging?

IC133： IC13302: Does the human 1 for yes, 0 0.00

resource policy contain for no
Severance provisions on resigning?

IC13303:Any audit upon the 1 for yes, 0 0.00

departure of the chairman, for no
CEO, and other executives?

IC14101:Highest reward/honor 1 for national 0.00

bestowed on the chairman and level, 0.5 for
IC14：Morality and

the CEO province

Competence

level, 0
IC141：
otherwise
Morality

37
 IC14201: Percentage of the To be 0.66
staff with at least a junior standardized
IC142： college education
IC14202: Any employee 1 for yes, 0 1.00
Competence
training carried out by the for no
company?

IC15101: Any social 1 for yes, 0 1.00

responsibility report released? for no
Responsibility
IC15：Social

IC15102: Any charitable 1 for yes, 0 1.00

donation? for no

IC15103: Does the company 1 for yes, 0 1.00

pay attention to environment, for no
safety, and product quality?

IC16101: Any corporate 1 for yes, 0 1.00

IC161： culture? for no

Corporate Culture IC16102: Any merger or 1 for no, 0 for 0.00

IC16：Culture

restructuring occurred? yes

IC16201: Any legal 1 for yes, 0 1.00

IC162： department? for no

Legal Awareness IC16202: Any counselor? 1 for yes, 0 1.00

for no

IC1P1: Any punishment for the 50% 0.00

firm, executives, directors, and deduction for
supervisors from the Ministry the total
of Justice, Ministry of Finance, scores of the
CSRC, or Shanghai/Shenzhen control
Stock Exchange? environment
Punishment and Other level, if yes
Negative Event IC1P2: Any irregularity 30% 0.00
disclosed by the media and deduction for
other channels? the total
scores of the
control
environment
level, if yes

38
 IC21101: Any objective for 1 for yes, 0 0.00

Objective
IC21：
risk management? for no

Setting
IC22101: Any committee or 1 for yes, 0 1.00
department for risk for no
IC221:
management?
Risk Evaluation
IC22102: Any risk evaluation 1 for yes, 0 1.00
disclosed in the annual report? for no

IC22201: Any internal risk 1 for yes, 0 1.00

disclosed in the annual report? for no

IC22202: Any human resource 1 for yes, 0 0.00

risk disclosed? for no

IC22203: Any management 1 for yes, 0 1.00

risk disclosed? for no
IC2：Risk Assessment

IC22204: Any innovation risk 1 for yes, 0 0.00

IC22：risk recognition

IC222: disclosed? for no

Internal Risk IC22205: Any financial risk 1 for yes, 0 0.00

disclosed? for no

IC22206: Any safety and 1 for yes, 0 0.00

environment risk disclosed? for no

IC22207: Any other internal 1 for yes, 0 0.00

risk disclosed? for no

IC22208:Any significant loss 1 for no, 0 for 1.00

over the past two years? yes

IC22301: Any external risk 1 for yes, 0 1.00

disclosed in the annual report? for no

IC22302: Any economic risk 1 for yes, 0 1.00

IC223: disclosed? for no

External Risk IC22303: Any legal risk 1 for yes, 0 1.00

disclosed? for no

IC22304: Any social risk 1 for yes, 0 0.00

disclosed? for no

39
 IC22305: Any technology risk 1 for yes, 0 0.00
disclosed? for no

IC22306: Any environment 1 for yes, 0 0.00

risk disclosed? for no

IC22307: Any other external 1 for yes, 0 0.00

risk disclosed? for no

IC23101: Any quantitative risk 1 for yes, 0 0.00

IC23：Risk Analysis

analysis in the annual report or for no

IC231: in the internal control
evaluation report?
Risk Evaluation
Method IC23102: Are risks sorted by 1 for yes, 0 0.00
importance or are main risks for no
specified?

IC24101: Any risk response 1 for yes, 0 1.00

IC241:
strategy disclosed? for no
Response
IC24102: Any analysis of risk 1 for yes, 0 1.00
Strategies
IC24：Risk Response

tolerance? for no

IC24201: Any risk 1 for yes, 0 0.00

management measure taken to for no
IC242: control risk?

Risk Management IC24202: Any change inrisk 1 for yes, 0 0.00

Measure evaluation and response from for no
the annual report for last year
to that for this year?

40
 IC31101: Any control for 1 for yes, 0 1.00
incompatible separation of for no
duties?

IC31102: Any control for 1 for yes, 0 1.00

authorizing and approving? for no
IC31: Separation of Duties

IC31103: Is the approving of 1 for yes, 0 1.00

material matters in accordance for no
with existing procedures?

IC32101: Is accounting 1 for yes, 0 1.00

IC32:Accounting

information institutionalized in for no

Control

the company?
IC3：Control Activities

IC33101: Any loss of 1 for no, 0 for 1.00

IC33:Property Control

inventories? yes

IC33102: More asset 1 for no, 0 for 1.00

impairment this year than last yes
year?

IC34101: Any committee or 1 for yes, 0 1.00

department for budgeting? for no

IC34102: Any budget 1 for yes, 0 1.00

IC34:Budget Control

implemented in the company? for no

IC34103: Is annual budget 1 for yes, 0 1.00

discussed in the shareholder for no
meeting or other similar
setting?

41
 IC35101：Any operational 1 for yes, 0 1.00

IC35:Operating
analysis? for no

Control

IC36101：Any department in 1 for yes, 0 1.00

charge of performance for no
IC36:Performance

analysis?
Control

IC36102：Any report resulting 1 for yes, 0 1.00

from the performance for no
analysis?

IC37101：Any early warning 1 for yes, 0 0.00

sign for material risks? for no
IC37:Emergency Control

IC37102：Any emergency 1 for yes, 0 0.00

response system? for no

IC3P01：Number of 20% 0.00

interruptions in operations, deduction for
such as contractual disputes less than 5,
and inventory losses 30% for 5-10,
40% for 10
and above

IC3P02：Any significant 30% 0.00

accident? deduction if
yes

IC3P03：Any significant 30% 0.00

Punishment and Other
pollution accident? deduction if
Negative Event
yes

IC3P04：Any significant 30% 0.00

product safety accident? deduction if
yes

42
 IC41101：Any channel for 1 for yes, 0 1.00

IC41:Information
internal communication of for no

Collection
information?

IC41102：Any channel for 1 for yes, 0 1.00

external collection of for no
information?

IC42101：Number of board To be 12.00

IC421: meetings standardized

Internal IC42102：Number of To be 7.00

Communication meetings for the board of standardized
supervisors
IC4：Information and Communication

IC42201：Any mechanism for 1 for yes, 0 1.00

information disclosure? for no

IC42202：Is the board 1 for yes, 0 1.00

IC42:Information Communication

secretary a “Gold Secretary”? for no

IC42203：Any mechanism for 1 for yes, 0 1.00

IC422: for no
managing relations with
External investors?
Communication
IC42204：Any link for 1 for yes, 0 1.00
investor relations on the for no
corporate website?

IC42205：Any other platform 1 for yes, 0 1.00

for communicating with for no
investors?

IC42301：Are all resolutions 1 for yes, 0 1.00

from shareholder meetings for no
IC423:
disclosed?
Information
IC42302：Are all resolutions 1 for yes, 0 1.00
Transparency
from board meetings for no
disclosed?

43
 IC42303：Are all resolutions 1 for yes, 0 1.00
from meetings of the board of for no
supervisors disclosed?

IC42401：Is the annual audit 1 for no, 0 for 1.00

opinion a qualified one? yes

IC42402：Any accounting 1 for no, 0 for 1.00

error or correction? yes

IC42403：Any change in 1 for no, 0 for 1.00

accounting policies and yes
estimates?
IC424:
IC42404：Any auditor 1 for no, 0 for 1.00
Information switching? yes
Accuracy
IC42405：Any dissenting 1 for no, 0 for 1.00
statements from the formal yes
auditor because of switching?

IC42406：Any problem found 1 for no, 0 for 1.00

and corrected by the yes
independent directors or
supervisors in the financial
reports?

IC42501：Are the periodic 1 for yes, 0 1.00

reports released on the for no
IC425:
scheduled date?
Information
IC42502：Are the periodic 1 for yes, 0 1.00
Timeliness
reports released on time as for no
required?

44
 IC43101：Any information 1 for yes, 0 1.00
department or information for no
security department?

IC43:Information System

IC44101：Any anti-fraud 1 for yes, 0 0.00

IC441: mechanism in the company? for no
IC44:Anti-Fraud

Anti-Fraud IC44102：Any channel for 1 for yes, 0 0.00

Mechanism whistle-blowing? for no

IC442: IC44201：Does the company 1 for yes, 0 0.00

specify the priorities for for no
Anti-Fraud
anti-fraud?
Priority

IC4P1：Any restatement, 20% 0.00

correction, or supplement of deduction for
the financial statements? yes
Punishment and other Negative
Event
IC4P2：Any information 20% 0.00
leakage? deduction for
yes

45
 IC511: IC51101：Any inspection of 1 for yes, 0 1.00
internal control from the for no
Monitoring from
internal audit department?
Internal Control
Department

IC512: IC51201：Did the board of 1 for yes, 0 1.00

IC51:Internal Monitoring Function

supervisors monitor for no

Monitoring from operational activities, such as
the Board of financing activities, asset sales,
Supervisors and acquisitions?

IC51301：Did the audit 1 for yes, 0 0.00

committee discuss the internal for no
IC513: control inspection in its
responsibility report?
Monitoring from
the Board of IC51302：Did the independent 1 for yes, 0 0.00
Directors directors discuss the internal for no
IC5：Monitoring

control inspection in their

responsibility report?

IC514: IC51401：Any monitoring for 1 for yes, 0 1.00

suspicious special events? for no
Special Monitoring

IC52101：Any standard for 1 for yes, 0 0.00

deficiency recognition? for no

IC52102：Any analysis of the 1 for yes, 0 0.00

IC52:Internal Control Deficiencies

reasons for deficiencies? for no

IC52103：Any plan for 1 for yes, 0 0.00

correcting internal control for no
deficiencies?

IC52104：Did the company 1 for yes, 0 0.00

follow the remediating for no
process?

46
 IC53101：Any internal control 1 for yes, 0 1.00
evaluation report? for no

What does the evaluation

report contain?

IC53102：Was the internal 1 for yes, 0 1.00

control well designed? for no

IC53103：Did the internal 1 for yes, 0 1.00

control function well? for no

IC53104：Was it a right 1 for yes, 0 1.00

situation for internal control for no
inspection?

IC53105：Any significant 1 for yes, 0 0.00

risks in executing the internal for no
IC53:Internal Control Disclosure

control?

IC53106：Any evaluation of 1 for yes, 0 0.00

the execution of the internal for no
control inspection?

IC53107：Any plan to 1 for yes, 0 1.00

improve internal control? for no

IC53108：Any plan for 1 for yes, 0 0.00

internal control for the next for no
year?

IC53109：Any other 1 for yes, 0 1.00

information on internal for no
control?

IC53110：Any opinion from 1 for yes, 0 1.00

the board of supervisors on for no
internal control?

IC53111：Any opinion from 1 for yes, 0 0.00

the independent directors on for no
internal control?

IC53112：Any opinion from 1 for yes, 0 0.00

the auditor on internal control? for no

47
 IC53113：Number of pages in To be 167.00
the financial report standardized

IC53114：Number of pages in To be 9.00

the internal control evaluation standardized
report

IC53115：Number of words in To be 5,911.00

the internal control evaluation standardized
report

Note: For item whose score is “To be standardized,” the standardized score is calculated as the actual
value for this item by the maximum value of the same item from all the public companies. For example, if
the actual number of words in the internal control evaluation report for Xiamen Guomao is 5,911 and the
maximum number of words is 10,000 from the internal control evaluation reports for all public firms,
then the standardized score on this item for Xiamen Guomao is 0.5911(5911/10000).

48
Appendix 2. Variable Definition

Panel A. Discretionary Accruals Sample

Variables Definitions
Dependent Variables
MJDA The absolute value of modified Jones model discretionary accruals proposed in Dechow et al.
(1995). The specific model is as follows.
TAit  0  1(1 / LagAsset )   2 ((REV  AR) / LagAsset )it  3 ( PPE / LagAsset )it   it
where TA, the total accruals, is net income before extraordinary items minus operating cash flows
scaled by LagAsset; LagAsset is total assets at the beginning of the fiscal year; ΔREV is equal to
net sales in year t less net sales in year t- 1; ΔAR is equal to accounts receivables in year t less
accounts receivables in year t- 1; PPE is net property, plant and equipment. MJDA takes the
absolute value of modified Jones model discretionary accruals, obtained as the residuals from an
industry-by-industry cross-sectional regression of the above equation. A constant is included in
the estimation model to control for heteroskedasticity, to mitigate problems stemming from an
omitted size variable, and to generate more symmetric accrual measures (Kothari et al. ,2005).

PMJDA The absolute value of performance-controlled modified Jones model discretionary accruals
proposed in Kothari et al. (2005). It is a measure of discretionary accruals based on the modified
Jones model controlling for performance by including the prior year’s ROA in the estimation of
expected accruals. The specific model is as follows.
TAit  0  1 (1 / LagAsset )   2 ((REV  AR) / LagAsset )it  3 ( PPE / LagAsset )it   4 LagROAit   it
where LagROA is the prior year’s return on assets. All other variables have the same definitions
as those for MJDA. A constant is included in the estimation model to control for
heteroskedasticity, to mitigate problems stemming from an omitted size variable, and to increase
the power of the test by generating more symmetric accrual measures.
PDCA The absolute value of performance-controlled discretionary current accruals proposed in
Ashbaugh et al. (2003). The performance-controlled discretionary current accruals is the
discretionary current accruals measure, controlling for performance by including the prior year’s
ROA in the estimation of expected accruals. The specific model is as follows.
CAit  1 (1 / LagAsset )   2 (REV / LagAsset )it  3 LagROAit   it
where CA, current accruals, is net income before extraordinary items plus depreciation and
amortization minus operating cash flows, scaled by beginning of year total assets. All other
variables have the same definitions as those for MJDA. The parameters from an industry-by-
industry cross-sectional regression of the above equation are used to calculate the expected
current accruals estimate with a performance control:
ECAit  1 (1 / LagAsset )   2 ((REV  AR) / LagAsset )it  3 LagROAit   it
The performance-controlled discretionary current accruals is equal to CA minus ECA, estimating
discretionary current accruals by controlling for performance on a firm-specific basis. PDCA
takes the absolute value of performance-controlled discretionary current accruals.
Independent Variables
IC_INDEX Our self-constructed internal control index based on the COSO framework and the AHP method
LNMV Natural logarithm of a firm’s market value of equity at the beginning of the year
SEGMENT Natural logarithm of a firm’s business segments
FOREIGN 1 if a firm has foreign revenue; 0 otherwise
STDSALES The standard deviation of sales, which is scaled by total assets, calculated over the past three
years
STDCFOS The standard deviation of operating cash flow, which is scaled by average assets, calculated over
the past three years

49
ST 1 if a firm has a “special treatment” designation; 0 otherwise. When a public firm has abnormal
financial or other conditions that make it hard for investors to assess its business prospects, the
security exchanges will apply special treatments to this firm’s stock transactions, including (1)
adding “ST” to the ticker symbol, and (2) limiting to the daily stock price appreciation or
depreciation to no more than five percent.
LEV Total liabilities divided by total assets at the beginning of the year
BIG4 1 if a Big 4 client; 0 otherwise
M&A 1 if a firm has conducted merger & acquisitions in the past three years; 0 otherwise
INVENTORY Inventory valued at cost scaled by total assets
OPERCYCLE Log[360/(COGS/Average Inventories) + 360/(revenue/average accounts receivable)]
MB Market value of equity divided by book value of equity
LNAGE Natural logarithm of a firm’s age

Panel B. ERC Sample

Variables Definitions
Dependent Variables
CAR_S The three-day market-adjusted cumulative abnormal returns covering the day before, the
day of, and the day after a firm’s earnings announcement. CAR_S=∑(Rit－RMt), where Rit
is the daily return for firm i in day t and RMt is daily value-weighted return for the market
in day t.
RET_S The three-day cumulative raw returns covering the day before, the day of, and the day after
a firm’s earnings announcement.
CAR_L The twelve-month market-adjusted cumulative abnormal return from May in the current
year to April in the following year. CAR_L=∑(Rit－RMt), where Rit is the monthly return
for firm i in month t and RMt is the monthly return for the market in month t.
RET_L The twelve-month cumulative raw return for a firm from May in the current year to April
in the following year. RET_L=∑Rit, where Rit is the monthly return for firm i in month t.
Independent Variables
IC_INDEX The Internal Control Index developed by the Xiamen University research group
UE Unexpected earnings, which is equal to the difference between the earnings per share
before extraordinary items for this period and those for the last period, divided by the
closing price of the last trading day in April.
NEG Equals 1 if UE<0; 0 otherwise
LNMV Natural logarithm of a firm’s market value at the beginning of the year
MB Market value of equity to book value of equity at the beginning of the year
LEV Ratio of total liabilities over total assets at the beginning of the year
BETA A firm’s beta, calculated with the market model. We require at least 100 trading days
during the 250-day estimation window ending 20 days prior to the release of annual
financial statements

50
References

Altamuro, J. and A. Beatty. 2010. How does internal control regulation affect financial
reporting? Journal of Accounting and Economics 49 (1-2): 58-74.

Ashbaugh-Skaife, D. Collins, W. Kinney, and R. LaFond. 2008. The effect of SOX internal
control deficiencies and their remediation on accrual quality. The Accounting Review 83 (1):
217-250.

Ashbaugh, H., R. LaFond, and B. Mayhew. 2003. Do nonaudit services compromise auditor
independence? Further evidence. The Accounting Review 78: 611–639.

Balsam, S., J. Krishnan, and J. Yang. 2003. Auditor industry specialization and earnings quality.
Auditing: A Journal of Practice and Theory 22 (2): 71-97.

Baber, W., J. Krishnan, and Y. Zhang. 2013. Investor perceptions of the earnings quality
consequences of hiring an affiliated auditor. Review of Accounting Studies, forthcoming.

Basu, S. 1997. The conservatism principle and the asymmetric timeliness of earnings. Journal of
Accounting and Economics 24 (December): 3-37.

Becker, C., M. DeFond, J. Jiambalvo, and K. Subramanyam. 1998. The effect of audit quality on
earnings management. Contemporary Accounting Research 15: 1–24.

Bedard, J., and L. Graham. 2011. Detection and severity classifications of Sarbanes-Oxley
Section 404 internal control deficiencies. Accounting Review 86 (3): 825-855.

Bowen, R., M. Johnson, T. Shevlin, and D. Shores. 1992. Determinants of the timing of quarterly
earnings announcements. Journal of Accounting, Auditing, and Finance 7 (Fall): 395-422.

Botosan, C. 1997. Disclosure level and the cost of equity capital. Accounting Review 72: 323-
349.

Bushman, R., and A. Smith. 2001. Financial accounting information and corporate governance.
Journal of Accounting and Economics 32 (1-3): 237-333.

Chan, K., B. Farrell, and P. Lee. 2008. Earnings management of firms reporting material internal
control weaknesses under Section 404 of the Sarbanes-Oxley Act. Auditing: A Journal of
Practice and Theory 27: 161-179.

Chan, L., K. Chen, T. Chen, and Y. Yu. 2012. The effects of firm-initiated clawback provisions
on earnings quality and auditor behavior. Journal of Accounting and Economics 54 (2-3): 180-
196.

Chen, H., J. Chen, G. Lobo, and Y. Wang. 2010. Association between borrower and lender state
ownership and accounting conservatism. Journal of Accounting Research 48 (5): 973-1014.

51
Collins, D., and S. Kothari. 1989. An analysis of intertemporal and cross-sectional determinants
of earnings response coefficient. Journal of Accounting and Economics 11 (2-3): 143-189.

Committee of Sponsoring Organizations of the Treadway Commission (COSO). 1992. Internal

Control-Integrated Framework. www.coso.org.

Dechow, P., W. Ge, and C. Schrand. 2010. Understanding earnings quality: A review of the
proxies, their determinants and consequences. Journal of Accounting and Economics 50: 344-
401.

Dechow, P., R. Sloan, and A. Sweeney. 1995. Detecting earnings management. Accounting
Review 70: 193-225.

Dhaliwal, D., K. Lee, and N. Fargher. 1991. The association between unexpected earnings and
abnormal security returns in the presence of financial leverage. Contemporary Accounting
Review 8 (1): 20-41.

Doyle, J., W. Ge, and S. McVay. 2007. Accruals quality and internal control over financial
reporting. Accounting Review 82 (5): 1141-1170.

Easton, P., and M. Zmijewski. 1989. Cross-sectional variation in the stock-market response to
accounting earnings announcements. Journal of Accounting and Economics 11(2-3): 117-141.

Francis, J., and B. Ke. 2006. Disclosure of fees paid to auditors and the market valuation of
earnings surprises. Review of Accounting Studies 11: 495-523.

Feng, M., and C. Li. 2011. Does SOX Section 404 curb material misstatements? Working paper,
University of Pittsburgh.

Feng, M., C. Li, and S. McVay. 2009. Internal control and management guidance. Journal of
Accounting and Economics 48: 190-209.

Francis, J., R. LaFond, P. Olsson, and K. Schipper. 2005. The market pricing quality. Journal of
Accounting and Economics 39 (2): 295–327.

Ge, W., and S. McVay. 2005. The disclosure of material weaknesses in internal control after the
Sarbanes-Oxley Act. Accounting Horizon 19 (3): 137-158.

Hackenbrack, K., and C. Hogan. 2002. Market response to earnings surprises conditional on
reasons for an auditor change. Contemporary Accounting Research 19: 195-223.

Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 20 (2):
125-153.

52
Healy, P, A. Hutton, and K. Palepu. 1999. Stock performance and intermediation changes
surrounding sustained increases in disclosure. Contemporary Accounting Research 16: 485-
520.

Holthausen, R., and R. Verrecchia. 1988. The effect of sequential information releases on the
variance of price changes in an intertemporal multi-asset market. Journal of Accounting
Research 26 (1): 82-106.

Huang, P., and Y. Zhang. 2012. Does enhanced disclosure really reduce agency costs? Evidence
from the diversion of corporate resources. Accounting Review 87 (1):199-229.

Hunton, J., R. Hoitash, and J. Thibodeau. 2011. The relationship between perceived tone at the
top and earnings quality. Contemporary Accounting Research 28 (4): 1190-1224.

Jo, H., and Y. Kim. 2007. Disclosure frequency and earnings management. Journal of Financial
Economics 84 (2): 561-590.

Kasznik, R. 1999. On the association between voluntary disclosure and earnings management.
Journal of Accounting Research 37: 57-81.

Khanna, T, K. Palepu, and S. Srinivasan. 2004. Disclosure practices of foreign companies

interacting with U.S. markets. Journal of Accounting Research 42, 475-508.

Klein, A. 2002. Audit committee, board of director characteristics, and earnings management.
Journal of Accounting and Economics 33 (3), 375-400.

Kothari, S. 2001. Capital markets research in accounting. Journal of Accounting and Economics
31 (1-3): 105-231.

Kothari, S., A. Leone, and C. Wasley. 2005. Performance matched discretionary accrual
measures. Journal of Accounting and Economics 39 (1): 163-197.

Lang, M., and R. Lundholm. 1993. Cross-sectional determinants of analyst ratings of corporate
disclosures. Journal of Accounting Research 31: 246-271.

Lang, M., and R. Lundholm. 1996. Corporate disclosure policy and analyst behaviour.
Accounting Review 71: 467-492.

La Porta, R., A. Shleifer, R. Vishny, R., and F. Lopez de Silanes. 1998. Law and finance.
Journal of Political Economy 106: 1113-1155.

Leuz, C., D. Nanda, and P. Wysocki. 2003. Earnings management and investor protection: An
international comparison. Journal of Financial Economics 69: 505-527.

Leuz, C., and P. Wysocki. 2006. Capital-Market Effects of Corporate Disclosures and Disclosure
Regulation. In: Canada Steps Up: Evolving Investor Protection, Research Studies for the Final

53
 Report of the Task Force to Modernize Securities Legislation in Canada, T. Allen et al. (eds),
Volume 2, 181–239.

Lu, H., G. Richardson, and S. Salterio. 2011. Direct and indirect effects of internal control
weaknesses on accrual quality: Evidence from a unique Canadian regulatory setting.
Contemporary Accounting Research 28 (2): 675-707

Nagar, V., D. Nanda, and P. Wysocki. 2003. Discretionary disclosure and stock-based
incentives. Journal of Accounting and Economics 34, 283-309.

Reynolds, J., and J. Francis. 2000. Does size matter? The influence of large clients on office-
level auditor reporting decisions. Journal of Accounting and Economics 30: 375-400.

Saaty, T. 2008a. Decision Making for Leaders: The Analytic Hierarchy Process for Decisions in
a Complex World. Pittsburgh, Pennsylvania: RWS Publications.

Saaty, T. 2008b. Relative Measurement and its Generalization in Decision Making: Why
Pairwise Comparisons are Central in Mathematics for the Measurement of Intangible Factors -
The Analytic Hierarchy/Network Process. RACSAM (Review of the Royal Spanish Academy of
Sciences, Series A, Mathematics) 102 (2): 251–318.

Securities Exchange Commission (SEC). June 5, 2003. Management's Report on Internal Control
over Financial Reporting and Certification of Disclosure in Exchange Act Periodic Reports.
Release No. 33-8238. Washington, D.C.: SEC.

________. June 27, 2007. Commission Guidance Regarding Management’s Report on Internal
Control Over Financial Reporting Under Section 13(a) or 15(d) of the Securities Exchange Act
of 1934. Release No. 33-8810. Washington, D.C.: SEC.

Sengupta, P. 1998. Corporate disclosure quality and the cost of debt. Accounting Review 73: 459-
474.

Teoh, S., and T. Wong. 1993. Perceived auditor quality and the earnings response coefficient.
Accounting Review 68 (2): 346-366.

Vaidya, O., and S. Kumar. 2006. Analytic hierarchy process: An overview of applications.
European Journal of Operational Research 169: 1-29.

Wang, Q., T. Wong, and L. Xia. 2008. State ownership, the institutional environment, and
auditor choice: Evidence from China. Journal of Accounting and Economics 46 (1): 112-134.

Welker, M. 1995. Disclosure policy, information asymmetry, and liquidity in equity markets.
Contemporary Accounting Research 11: 801-827.

54
Table 1. Sample Selection

Discretionary ERC Sample ERC Sample

Accruals Sample Short window Long window

All firms from the Intersection of CSMAR

and RESSET from 2007 to 2010 6,731 6,731 6,731

Excluding Firms：
In Financial Industry 118 118 118
Without Internal Control Indices 67 67 67
With Missing Information to Calculate
Measures of Earnings Management 137
With Missing Information to calculate
CAR 242 122
With Missing Control Variables 246 687 845

Final Sample 6,163 5,617 5,579

CSMAR (China Stock Market & Accounting Research) Database is designed and developed by GTA
Information Technology, one of the major providers of China data. CSMAR is the comprehensive
database for Chinese business research. CSMAR covers data on the Chinese stock market, financial
statements, and China Corporate Governance of Chinese Listed Firms.

RESSET is a vendor that specializes in providing financial information and service product in China and
supplies data needed in empirical research and investment analysis in economics, finance, and accounting
areas. The RESSET Financial Research Database (RESSET/DB) includes 10 series, such as stock, fixed-
income, fund, macroeconomics, and Hong Kong stock. It has over 60 databases, 500 Chinese datasets and
related English datasets, and about 15,000 variables covering a wide range and complete historical
information.

55
Table 2. Descriptive Statistics for the Discretionary Accruals Sample

Variable Mean Q1 Median Q3 Std. Dev.

MJDA 0.084 0.024 0.055 0.109 0.093
PMJDA 0.079 0.023 0.052 0.104 0.085
PDCA 0.079 0.023 0.050 0.102 0.086
IC_INDEX 0.360 0.287 0.358 0.427 0.099
CtrEnv 0.339 0.244 0.324 0.444 0.153
RiskAss 0.165 0.089 0.156 0.204 0.120
CtrAct 0.512 0.389 0.503 0.628 0.160
InfoCom 0.482 0.406 0.473 0.522 0.113
Monitor 0.284 0.162 0.277 0.388 0.145
SEGMENTS 2.051 1.609 2.079 2.485 0.579
FOREIGN 0.435 0.000 0.000 1.000 0.496
OperCycle 4.809 4.219 4.797 5.356 0.969
M&A 0.018 0.000 0.000 0.000 0.134
MV (RMB Millions) 7,093 1,596 3,008 6,299 13,756
LNMV 8.131 7.375 8.009 8.748 1.086
ST 0.073 0.000 0.000 0.000 0.259
LEV 0.543 0.374 0.518 0.647 0.577
BIG4 0.061 0.000 0.000 0.000 0.238
MB 4.340 2.143 3.551 5.497 3.999
STDSALES 0.138 0.052 0.097 0.173 0.139
STDCFOS 0.052 0.021 0.039 0.066 0.047
INVENTORY 0.175 0.070 0.138 0.227 0.154
AGE 9.560 5.000 10.000 13.000 4.950
LNAGE 2.028 1.609 2.303 2.565 0.801
N 6,163

This table provides summary statistics of the sample for analyzing earnings management. All variables are defined
in Appendix 2.

56
Table 3. Effect of Internal Control Index on Discretionary Accruals

Model 1 Model 2 Model 3

Dep. Var. MJDA PMJDA PDCA
IC_INDEX -0.071*** -0.058*** -0.061***
-5.54 -4.86 -5.15
LNMV 0.006*** 0.005*** 0.004***
4.59 3.99 3.83
SEGMENT -0.002 -0.001 -0.002
-0.9 -0.65 -1.14
FOREIGN -0.005** -0.003 -0.004*
-2.29 -1.51 -1.7
STDSALES 0.043*** 0.033*** 0.036***
5.08 4.17 4.59
STDCFOS 0.739*** 0.735*** 0.764***
30.57 32.74 34.21
ST 0.022*** 0.022*** 0.024***
4.99 5.35 5.91
LEV 0.015*** 0.006*** 0.006***
8.14 3.41 3.68
BIG4 -0.008 -0.007 -0.009**
-1.62 -1.48 -2.03
M&A 0.011 0.010 0.009
1.45 1.45 1.25
INVENTORY 0.036*** 0.029*** 0.036***
3.41 3.01 3.74
OPERCYCLE 0.005*** 0.002 0.002
2.81 1.31 1.38
MB -0.000 0.000 0.000
-0.19 0.56 0.39
LNAGE -0.019*** -0.015*** -0.016***
-14.03 -11.94 -12.63
Intercept 0.033** 0.047*** 0.049***
2.15 3.24 3.45
Year dummies included included included
Industry dummies included included included
Adj R Square 0.2524 0.2415 0.2661
N 6,163 6,163 6,163

This table uses OLS regressions to examine the effect of our internal control index on measures of earnings
management. MJDA is the absolute value of modified Jones model discretionary accruals proposed in Dechow et al.
(1995). PMJDA is the absolute value of performance-controlled modified Jones model discretionary accruals
proposed in Kothari et al. (2005). PDCA is the absolute value of performance-controlled discretionary current
accruals proposed in Ashbaugh et al. (2003). All variables are defined in Appendix 2. The regression coefficients are
reported in the upper rows, whereas the t-statistics are reported in the lower rows. *, **, and *** denote two-tailed
significance at the ten-, five-, and one-percent levels, respectively.

57
Table 4. Effect of Components of Internal Control Index on Discretionary Accruals

Model 1 Model 2 Model 3

Dep. Var. MJDA PMJDA PDCA
CtrEnv -0.042*** -0.033*** -0.032***
-4.92 -4.14 -4.13
RiskAss 0.005 0.002 -0.001
0.51 0.2 -0.08
CtrAct -0.016** -0.011 -0.013*
-2.08 -1.53 -1.88
InfoCom -0.026** -0.026*** -0.023**
-2.54 -2.73 -2.38
Monitor 0.010 0.008 0.008
1.20 1.06 1.00
LNMV 0.006*** 0.005*** 0.005***
5.07 4.41 4.23
SEGMENT -0.002 -0.001 -0.002
-0.83 -0.57 -1.08
FOREIGN -0.005** -0.003 -0.003
-2.17 -1.4 -1.61
STDSALES 0.042*** 0.032*** 0.035***
4.98 4.08 4.50
STDCFOS 0.736*** 0.733*** 0.762***
30.44 32.63 34.09
ST 0.022*** 0.022*** 0.024***
4.86 5.25 5.79
LEV 0.015*** 0.006*** 0.006***
8.16 3.43 3.7
BIG4 -0.008* -0.007 -0.009**
-1.71 -1.52 -2.04
M&A 0.012 0.011 0.009
1.52 1.50 1.30
INVENTORY 0.038*** 0.031*** 0.038***
3.64 3.21 3.92
OperCycle 0.004*** 0.002 0.002
2.65 1.19 1.26
MB -0.000 0.000 0.000
-0.24 0.53 0.34
LNAGE -0.019*** -0.015*** -0.016***
-13.89 -11.78 -12.45
Intercept 0.034** 0.048*** 0.050***
2.14 3.23 3.41
Year dummies included included included
Industry dummies included included included
Adj R Square 0.2538 0.2426 0.2669
N 6,163 6,163 6,163

58
This table uses OLS regressions to examine the effect of components of our internal control index on various
measures of earnings management. MJDA is the absolute value of modified Jones model discretionary accruals
proposed in Dechow et al. (1995). PMJDA is the absolute value of performance-controlled modified Jones model
discretionary accruals proposed in Kothari et al. (2005). PDCA is the absolute value of performance-controlled
discretionary current accruals proposed in Ashbaugh et al. (2003). All variables are defined in Appendix 2. The
regression coefficients are reported in the upper rows, whereas the t-statistics are reported in the lower rows. *, **,
and *** denote two-tailed significance at the ten-, five-, and one-percent levels, respectively.

59
Table 5. Descriptive Statistics for the ERC Sample

Panel A: Short-window ERC Sample

Variable Mean Q1 Median Q3 Std. Dev.

CAR_S 0.001 -0.030 -0.002 0.028 0.057
RET_S -0.001 -0.037 0.000 0.038 0.071
IC_INDEX 0.356 0.281 0.352 0.424 0.101
CtrEnv 0.334 0.230 0.319 0.442 0.156
RiskAss 0.166 0.089 0.166 0.205 0.122
CtrAct 0.507 0.382 0.496 0.617 0.161
InfoCom 0.479 0.403 0.469 0.521 0.114
Monitor 0.277 0.153 0.268 0.381 0.144
UE -0.0019 -0.024 -0.003 0.015 0.108
NEG 0.553 0.000 1.000 1.000 0.497
MV (RMB Millions) 8,611 1,523 2,863 6,223 49,737
LNMV 8.099 7.329 7.959 8.736 1.116
LEV 0.633 0.377 0.524 0.653 3.141
BETA 1.023 0.916 1.049 1.156 0.217
MB 3.794 1.735 2.936 4.832 3.711
N 5,617
Panel B: Long-window ERC Sample

Variable Mean Q1 Median Q3 Std. Dev.

CAR_L 0.162 -0.072 0.132 0.357 0.338
RET_L 0.228 -0.025 0.191 0.450 0.357
IC_INDEX 0.356 0.281 0.353 0.425 0.101
CtrEnv 0.335 0.231 0.320 0.443 0.156
RiskAss 0.167 0.089 0.166 0.205 0.122
CtrAct 0.508 0.382 0.496 0.620 0.160
InfoCom 0.480 0.403 0.469 0.521 0.114
Monitor 0.277 0.153 0.269 0.381 0.144
UE -0.002 -0.024 -0.003 0.015 0.106
NEG 0.554 0.000 1.000 1.000 0.497
MV (RMB Millions) 8,634 1,516 2,873 6,258 49,879
LNMV 8.099 7.324 7.963 8.742 1.119
LEV 0.525 0.377 0.523 0.652 0.230
BETA 1.004 0.905 1.035 1.132 0.193
MB 3.767 1.735 2.930 4.815 3.606
N 5,579
This table provides summary statistics for the sample for analyzing earnings response coefficients (ERC). All
variables are defined in Appendix 2.

60
TABLE 6. Effect of Internal Control Index on Earnings Response Coefficients

Short Window Long Window

Model 1 Model 2 Model 3 Model 4
Dep. Var. CAR_S RET_S CAR_L RET_L
UE 0.067* 0.234*** 1.201** 1.246**
1.95 2.81 2.40 2.53
UE* IC_INDEX 0.300*** 0.374*** 1.402** 1.383**
2.91 2.93 2.19 2.20
UE*NEG -0.011 -0.012 -0.325** -0.309**
-0.51 -0.46 -2.57 -2.48
UE*LNMV -0.024** -0.039*** -0.029 -0.036
-2.23 -2.95 -0.54 -0.67
UE*LEV -0.003** -0.004** -0.607*** -0.580***
-2.05 -2.41 -4.07 -3.94
UE*MB 0.001 0.001 -0.003 -0.003
0.54 0.45 -0.24 -0.31
UE*Beta -0.007 -0.012 -0.282 -0.303
-0.32 -0.46 -1.29 -1.40
Intercept 0.002 -0.011 0.078** 0.192***
0.32 -1.48 2.20 5.52
Year dummies included included included included
Industry dummies included included included included
Adj R-Sq 0.0046 0.0224 0.0972 0.2144
N 5,617 5,617 5,579 5,579

This table uses OLS regressions to examine the effect of our internal control index on earnings response coefficients
(ERC). The sample size is 5,617 for the short-window analyses in Models 1 and 2 and 5,579 for the long-window
analyses in Models 3 and 4. All variables are defined in Appendix 2. The regression coefficients are reported in the
upper rows, whereas the t-statistics are reported in the lower rows. *, **, and *** denote two-tailed significance at
the ten-, five-, and one-percent levels, respectively.

61
Table 7. Effect of Components of Internal Control Index on Earnings Response
Coefficients

Short Window Long Window

Model 1 Model 2 Model 3 Model 4
Dep. Var. CAR_S RET_S CAR_L RET_L
UE 0.184** 0.281*** 1.09** 1.105**
2.49 3.09 2.07 2.13
UE*CtrEnv 0.254*** 0.296*** 0.065 -0.030
3.63 3.42 0.14 -0.07
UE*RiskAssess 0.004 0.026 0.414 0.420
0.04 0.21 0.74 0.76
UE*CtrAct 0.035 -0.004 0.075 0.251
0.49 -0.05 0.19 0.64
UE*InfoCom -0.101 -0.015 1.332** 1.225**
-1.00 -0.12 2.52 2.35
UE*Monitor -0.035 -0.030 0.411 0.328
-0.51 -0.36 1.03 0.84
UE*NEG -0.007 -0.011 -0.342*** -0.315**
-0.35 -0.41 -2.62 -2.45
UE*LNMV -0.023** -0.039*** -0.069 -0.072
-2.08 -2.88 -1.19 -1.26
UE*LEV -0.003** -0.004** -0.570*** -0.548***
-2.13 -2.07 -3.63 -3.55
UE*MB 0.001 0.001 -0.003 -0.003
0.64 0.50 -0.31 -0.32
UE*Beta -0.012 -0.016 -0.230 -0.246
-0.57 -0.61 -1.02 -1.11
Intercept 0.002 -0.012 0.078** 0.193***
0.25 -1.53 2.20 5.53
Year dummies included included included included
Industry dummies included included included included
Adj R Square 0.0053 0.0227 0.0974 0.2145
N 5,617 5,617 5,579 5,579

This table uses OLS regressions to examine the effect of components of our internal control index on earnings
response coefficients (ERC). The sample size is 5,617 for the short-window analyses in Models 1 and 2 and 5,579
for the long-window analyses in Models 3 and 4. All variables are defined in Appendix 2. The regression
coefficients are reported in the upper rows, whereas the t-statistics are reported in the lower rows. *, **, and ***
denote two-tailed significance at the ten-, five-, and one-percent levels, respectively.

62