Investor responsiveness to internal control deficiencies : an earnings quality approach

(1)

MSc Accountancy & Control, variant Accountancy Faculty of Economics and Business, University of Amsterdam

Master thesis:

Investor Responsiveness to Internal Control

Deficiencies:

An Earnings Quality Approach.

Final version

Robert Patrick van Leeuwen (10003990) 23rd of June 2014

First supervisor: Mr. Vincent O’Connell

(2)

Abstract

The goal of this study is to examine how investors respond to internal control deficiencies. For this study an earnings quality approach is chosen instead of an event study. The findings from this study are of particular interest to regulators that are struggling with cost-benefit issues. For this study a total of 3,869 firm year observations for the years 2007 to 2012 is used. Data for these firms is collected from several WRDS databases, including

AuditAnalytics, CRSP and Compustat. This study does not find support for the prediction that investors perceive earnings quality to be lower for firms disclosing material weaknesses under SOX section 302. Instead, this study finds that firms disclosing material weaknesses under SOX section 302 have a higher perceived earnings quality. The prediction that SOX section 404 material weakness disclosures lead to a lower earnings quality is true but insignificant. My results thus do not provide sufficient evidence to say something about the responsiveness of investors to SOX section 404 material weakness disclosures. Also, only weak associations are found between investor response and material weakness disclosures. Real causal relationships are not found so there is insufficient appropriate evidence to answer the research question. Future research might be able to find a stronger relationship as it can examine whether the use of the earnings quality model used in this study is appropriate for investigating the investor responsiveness to internal control deficiencies.

Keywords Internal control; Sarbanes-Oxley Act; SOX section 302; SOX section 404;

Earnings quality; Investor responsiveness; Material weakness

(4)

1. Introduction

At the end of 2006 a collapse on the U.S. housing market triggered a recession in the U.S. financial systems. In the following years the United States suffered a series of financial institutions going bankrupt, with Lehman Brothers being the most commonly known example. Furthermore there were also declines in consumer wealth and economic activity. The 2007 financial crisis is considered to be the worst crisis since the Great Depression (Vyas, 2011). Due to the fact that the U.S. has such a big influence on the world economy this lead to a worldwide economic downfall (Lastra & Wood, 2010).

A lot of literature can be found on the 2007 financial crisis: how it emerged, the lessons that could be learned (Lastra & Wood, 2010) and how investors responded to it (Rotheli, 2010). However, there has not been any research done on how investors respond to internal control disclosures after this time series.

This study thus contributes to the existing literature in that it examines the investor reaction in a new time series. This study will use a dataset that includes data for the years 2007 to 2012. Furthermore, this study also includes earnings quality as a means of finding the investor reaction to internal control deficiencies. Based on Doyle et al. (2007b) I find that using earnings response coefficients (ERCs) is an appropriate way of examining a market reaction. Prior literature has not used this before in this setting for both SOX section 302 material weakness disclosures and SOX section 404 material weakness disclosures. Also, a comparison between the two types of internal control material weakness disclosures has not been made before using earnings quality as a measure of investor responsiveness.

It would be interesting to see whether and how investors value a firms earnings quality when the firm reported material weaknesses in its internal controls. In that way it is possible to say something about how investors value information disclosed by the firm and how investors value information disclosed by the auditor.

From a societal point of view this research would be of particular interest to regulators. Hammersley et al. (2008, p. 163) suggests that this will be of particular interest to regulators who are struggling with cost-benefit issues related to internal control disclosures. After this study regulators know whether and how investors value internal control information. This will help regulators in its cost-benefit analysis (Hammersley et al., 2008, p.163).

Furthermore, this research will also be of interest to auditors and companies who are required to comply with SOX regulations (Hammersley et al., 2008, p. 164).

(5)

There are three categories of internal control deficiencies: control deficiencies, significant deficiencies and material weaknesses. Control deficiencies are the least severe deficiency and material weakness is the most severe deficiency. Differences between these forms occur due to differences in the probability and the magnitude of these financial statement misstatements (PCAOB, 2004).

There are two provisions in the Sarbanes Oxley Act (SOX) of 2002 that require disclosures about the effectiveness of the firms’ internal control systems. These provisions are section 302 and section 404. These are supposed to say something about the quality of the internal controls of the firm. If material weaknesses remain after a firm complies with SOX 404, the auditor is required to issue an adverse opinion on internal controls (PCAOB, 2004).

Section 302 requires that CEO’s and CFO’s evaluate quarterly the design and effectiveness of the internal controls and report an overall conclusion about their effectiveness. Section 404 requires an annual audit of management’s evaluation of the internal controls and of the effectiveness of these internal controls in which they take

responsibility for maintaining adequate internal controls and make assertions concerning their effectiveness. The company’s auditor then must issue a separate opinion on management’s assertions and the adequacy of the internal controls (Palmrose et al., 2004).

Policy makers intended these reports on internal control systems to provide financial statement users with early warnings about potential future financial statement problems that could result from weak internal controls (PCAOB, 2004). The Sarbanes Oxley Act provisions are required for U.S. firms. But, due to the size of the U.S. financial markets the

implementation of SOX by U.S. firms has global implications (Hammersley et al., 2008, p. 143).

Most of the prior literature consists of event-studies. However, this research does not include an event study. Considering the timespan for this thesis this will not be possible. Nevertheless, event studies already done by other researchers are still useful for getting a better understanding of the results of this thesis. The literature in this study is thus subdivided into literature from event studies and literature on the relationship between earnings quality and internal control disclosures.

Overall the findings of prior studies suggest that firms that disclose material

weaknesses under SOX section 302 have lower earnings quality. However, there are no real quantitative results providing an indication of the severity of the impairment of earnings quality. The event studies suggest that investors associate SOX section 302 with more

(6)

negative cumulative abnormal returns. It can thus be said that investors use SOX section 302 disclosures to value a firm.

The findings of prior studies related to material weakness disclosures under SOX section 404 suggest that earnings quality is impaired by material weakness disclosures under SOX section 404. However, there are some circumstances which increase or decrease this effect. Also the event studies suggest that there are size-adjusted returns in the days

surrounding the announcement of SOX section 404 material weaknesses (Ashbaugh-Skaife et al., 2009; De Franco et al., 2005). These size-adjusted returns appear to be more negative for SOX section 404 material weaknesses than for SOX section 302 material weaknesses. Based on prior literature, it can thus be expected that investors more strongly react to SOX section 404 material weakness disclosures than to SOX section 302 material weakness disclosures. To investigate whether the findings from the existing literature hold for the setting of this study, the following research question is developed: “What is the investor’s response, reflected in earnings quality, to internal control deficiencies under SOX section 302 and SOX section 404?”

The sample that is used is collected from the WRDS AuditAnalytics database for SOX section 302 and SOX section 404 data. The WRDS CRSP database is used to collect data about the returns. All other data come from the WRDS Compustat database. The final sample consists of 3,869 firm year observations for the years 2007-2012.

Based on prior literature I expect investors to use the information from SOX section 302 and SOX section 404 disclosures. Furthermore, I also expect the use of this information by these investors to be reflected in the earnings quality of firms. The prediction is that the earnings quality will be lower for firms reporting SOX section 302 or SOX section 404 material weakness disclosures. Also, I expect this relationship to be more negative for SOX section 404 material weakness disclosures than for SOX section 302 material weakness disclosures.

I do not find support for my prediction that investors perceive earnings quality to be lower for firms disclosing material weaknesses under SOX section 302. The results suggest that firms disclosing material weaknesses under SOX section 302 have a higher perceived earnings quality. Furthermore, the results from my prediction that SOX section 404 material weakness disclosures lead to a lower earnings quality are not supported. My results thus do not provide sufficient evidence to say something about the responsiveness of investors to SOX section 404 material weakness disclosures. Also, only weak associations are found

(7)

between investor response and material weakness disclosures. Real causal relationships are not found so there is insufficient appropriate evidence to answer the research question.

The structure of this thesis is the following. Chapter two will consist of background

information on internal control deficiencies, the underlying theories assumed for this study and the discussion of the prior literature. The latter will result in the development of the hypotheses. Chapter three handles the research design. In this chapter the empirical model will be explained in to detail. The fourth chapter will give insights into the data used for the empirical model. This includes the sample selection and descriptive statistics. The

combination of the data and the empirical model will provide results which will be presented and discussed in chapter 5. This chapter also includes a sensitivity analysis. To conclude this study will end with chapter 6, which consists of a conclusion and discussion.

2. Literature review & hypothesis

2.1 Introduction

This chapter focuses on the review of the prior literature and will serve as a good theoretical background for the remainder of this research. In section 3.2 the background literature considering internal control deficiencies is presented. Section 3.3 presents the underlying theories used in this research such as efficient markets theory and investor responsiveness to earnings. Last but not least, section 3.4 handles the discussion of the prior literature and presents the hypotheses.

2.2 Background on internal control deficiencies

This section provides information about internal control disclosures and some background information about what SOX 302 and SOX 404 entails. This section will provide information which helps in enhancing the understanding of this paper.

There are three categories of internal control deficiencies: control deficiencies, significant deficiencies and material weaknesses. Control deficiencies are the least severe deficiency and material weakness is the most severe deficiency. Differences between these forms occur due to differences in probability and the magnitude of these financial statement misstatements. Material weaknesses are associated with possible material misstatements. If material weaknesses remain after a firm complies with SOX 404, the auditor is required to issue an adverse opinion on internal controls (PCAOB, 2004).

(8)

Figure 2.1: Internal Control Deficiency Evaluation (Ramos, 2004, p. 46)

As shown in the diagram in figure 2.1, internal control deficiencies must be evaluated along two dimensions to determine their relative significance (Ramos, 2004, p. 46). Those dimensions are likelihood and significance, with likelihood on the horizontal axe and significance on the vertical axe. If it is probable that a material error could result from an internal control deficiency, then it is considered to be a material weakness which has to be reported (Ramos, 2004, p. 46).

Furthermore, Ramos (2004, p. 47) also describes several indicators of whether a material weakness exists. These indicators are primarily based on the PCAOB Auditing Standard no. 2. This standard provides guidance on how to evaluate internal control deficiencies (PCAOB, 2004). The following situations should be regarded as strong indicators of the existence of a material weakness (Ramos, 2004, p. 47; PCAOB, 2004):

- Reflecting the correction of a misstatement by restating previously issued financial statements

- When the independent auditor identifies a material misstatement that was not identified by the company’s internal control function

- Ineffective oversight of the audit committee on external financial reporting

- Ineffective risk assessment or internal audit function in complex or large companies - Ineffective regulatory compliance function for complex companies in highly regulated

industries

- Identified fraud of any magnitude by senior management

(9)

- Uncorrected significant deficiencies which have been communicated to management and the audit committee

- Ineffective control environment

There are two provisions in the Sarbanes Oxley Act (SOX) of 2002 that require disclosures about the effectiveness of a firms’ internal control system. These provisions are section 302 and section 404. These are supposed to say something about the quality of the internal controls of the firm (Hammersley et al., 2008, p. 145).

Section 302 requires that CEO’s and CFO’s evaluate quarterly the design and effectiveness of the internal controls and report an overall conclusion about their effectiveness. Section 404 requires an annual audit of management’s evaluation of the internal controls and of the effectiveness of these internal controls in which they take

responsibility for maintaining adequate internal controls and make assertions concerning their effectiveness. The company’s auditor then must issue a separate opinion on management’s assertions of the internal company and also give an independent opinion on the adequacy of the internal controls (Palmrose et al.., 2004).

Policy makers intended these reports on internal control systems to provide financial statement users with early warnings about potential future financial statement problems that could result from weak internal controls (PCAOB, 2004). The Sarbanes Oxley Act provisions are required for U.S. firms. Due to the size of the U.S. financial markets the implementation of SOX by U.S. firms has global implications (Hammersley et al., 2008, p. 143).

2.3 Theoretical framework

2.3.1 Efficient Market Hypothesis Theory (EMH Theory)

This section will provide information regarding the underlying theory assumed for the purpose of this study.

Research on the market reaction to certain economic or accounting effects is generally based on one underlying theory: The Efficient Market Hypothesis (EMH) theory (Ittonen, 2011). In efficient market theory accounting is viewed as being in competition with other information sources. Examples of these other information sources are news media, financial analysts and also the market price itself. The EMH theory consists of 3 forms: weak form, semi-strong form and strong form (Scott, 2011, p. 109).

The weak form of efficient market theory implies that the prices of securities reflect all past publicly known information. The definition of the semi-strong form of the efficient

(10)

markets theory is that an efficient securities market is a market where the prices of securities fully reflect all information that is publicly known about those securities. The strong form is instantly reflecting all information, so also ‘insider’ or hidden information (Scott, 2011, pp. 110-111).

For the purpose of this study the semi-strong form of the EMH theory is assumed. The reason for this comes from Ittonen (2011, p. 5) who says that the semi-strong form of EMH theory is a widely assumed theory to investigate the market reaction to accounting effects. To be able to say something about the effects of the reaction of investors to these accounting effects this assumption has to be made. This study will thus also assume the semi-strong form of EMH theory to investigate the investor responsiveness to internal control deficiencies.

2.3.2 Investor responsiveness to earnings

This section will provide information regarding whether the use of earnings response coefficients is a good measure of investor responsiveness.

The use of earnings response coefficients (ERC) as a proxy for investor responsiveness to earnings is widely discussed in prior literature (Dechow et al., 2010). Dechow et al. (2010) outlines and discusses the existing literature using ERCs as a proxy for earnings quality. These different studies are compared and discussed to be able to say something about whether ERCs are a good proxy for earnings quality in different situations (Dechow et al., 2010).

Dechow et al. (2010, pp. 366-371) divide the literature into two sections: (i) the relation between ERCs and earnings measures and (ii) the relation between ERCs and non-earnings measures. The section about earnings measures is subdivided into (i) direct- and (ii) indirect evidence on ERCs as a proxy for earnings quality. The direct evidence uses the study of Liu & Thomas (2000) as a guiding paper. Liu & Thomas (2000) see ERC as a measure of quality. However, Dechow et al. (2010, p. 367) argue that Liu & Thomas (2000) base this notion of ‘quality’ mostly on overall decision usefulness. Liu & Thomas (2000) thus do not comment on any specific subparts of quality. Liu & Thomas (2000) thus give an overall indication that ERCs might be useful proxies for earnings quality but it is not clear whether this is true for all specific subparts of quality (Dechow et al., 2010, p. 367).

The indirect evidence comes from different subcategories made by Dechow et al. (2010, pp. 367-369): (i) accounting methods, (ii) auditor quality & governance, (iii) firm fundamentals and (iv) leverage. The conclusion from these categories can be broadly defined as that there is mixed evidence on ERCs as a proxy for earnings quality. One of the most

(11)

important points made by Dechow et al. (2010, p. 369) is that ERC is not a proxy for

‘unconditional’ earnings quality. This means that the ERC can only say something about the overall earnings quality but cannot distinguish between the contributions of a certain variable (Dechow et al., 2010, p. 369).

The second section is the relation between ERCs and non-earnings information

(Dechow et al., 2010, p. 369). This section looks at whether ERCs are useful for information that is not based on earnings. Dechow et al. (2010, p. 370) point out that understanding this relation is important because the ERC captures the informativeness of earnings with all other available information held constant. The evidence suggests that there is a correlation between ERCs and the availability of other information (Dechow et al. (2010, p. 370). Dechow et al. (2010, p. 370) argue that ERCs can thus only be viewed as a useful proxy if, within the sample, the availability of the other information is homogeneous.

Finally, Dechow et al. (2010, pp. 370-371) give a final notion on the use of internal control deficiencies as a proxy for earnings quality. Dechow et al. (2010, p. 371) argue that the advantage of this is that an independent external source identifies the quality problem. A researcher thus does not need to specify a model to identify misstatements. However, this may also result in potential bias in the selection criteria used by the external source (Dechow et al., 2010, p. 371).

In this study ERCs will be used as a proxy for earnings quality. This follows from the studies of Ghosh & Lubberink (2007) and Schipper & Vincent (2003) that also use ERCs as a proxy for earnings quality in combination with internal control deficiencies.

2.4 Hypothesis development

This section provides an overview and discussion of the existing literature. Based on this discussion the hypotheses are determined and presented. The literature is divided into event studies and other studies. The event studies are primarily focused on the returns surrounding and during the date the disclosures were first made public. The other studies will be more focused on the earnings quality aspect of this study. This is done for both the literature on SOX section 302 and on SOX section 404.

Event studies SOX section 302

In this section the literature concerning event studies on SOX section 302 internal control disclosures will be presented and discussed.

(12)

The research of Hammersley et al. (2008) examined the stock price reaction to management’s disclosure of the internal control weaknesses under section 302 of the

Sarbanes Oxley Act and to the characteristics of these weaknesses. In this way Hammersley et al. (2008) control for other material announcements in the event window. The findings suggest that some characteristics of the weaknesses are informative.

Also, the information content of internal control weakness disclosures depends on the severity of the internal control weakness. Moreover, in a subsample uncontaminated by other announcements, Hammersley et al. (2008) find negative price reactions to the disclosure of internal control weaknesses.

Ittonen (2010) conducted a similar research in which they investigated investor reactions to material internal control weakness disclosures. Similarly to the research of Hammersley et al. (2008) and Beneish et al. (2008), Ittonen (2010) used the abnormal market returns as a measure of the market reaction to news announcements. Ittonen (2010) however, extended his research by including also the change in volatility and the change in systematic risk due to the disclosure of material internal control weaknesses.

The initial results of the research of Ittonen (2010) are different than that found in the research of Hammersley et al. (2008). Ittonen (2010, p. 265) finds that the material weakness disclosure is good news to investors. However, after controlling for the preceding

management’s internal control disclosure (SOX 302) the abnormal return is only positive when the audit report is consistent with the preceding management report.

Beneish et al. (2008) investigate whether the effect of material weaknesses on the cost of capital and on stock prices is associated with audit quality. Similar to Ittonen (2010) and Hammersley et al. (2008) & Beneish et al. (2008) used the cumulative size-adjusted returns. Also both Ittonen (2010) and Hammersley et al. (2008) use the same 3-day window

surrounding the disclosure for firms making material weakness disclosures under SOX 302 Beneish et al. (2008) uses. The findings suggest that there are negative cumulative size-adjusted returns in this 3-day window for firms disclosing on material weaknesses (Beneish et al., 2008, p. 693).

Other studies SOX section 302

There is not much research specifically focused on the effect of SOX section 302 internal control deficiencies on earnings quality. Nevertheless there will be a discussion between the available literature that did. This will serve as a basis for the hypothesis.

(13)

The study of Kalelkar & Nwaeze (2011) examined the impact of SOX on the valuation weights of earnings and earnings components. The goal of this study is to test whether

investor perception of earnings and accruals quality changes due to SOX. Kalelkar & Nwaeze (2011) perceive the valuation weights of earnings the most vulnerable to manipulation before the implementation of SOX. The findings suggest that there are increases in these valuation weights and earnings components after the implementation of SOX (Kalelkar & Nwaeze, 2011, p. 291). Also Brown et al. (2008) find that the implementation of internal control regulation results in the achievement of greater earnings quality through effective internal controls.

The study of Brown et al. (2008) however, does not examine the impact of SOX section 302 disclosures on earnings quality. Brown et al. (2008) examine the German variant of SOX internal control regulation, the KTG. It can thus be asked whether the results of Brown et al. (2008) also hold for SOX. However, the results of Brown et al. (2008) can still provide useful insights into the effect of internal control regulation on earnings quality.

The implementation of SOX, and thus also SOX 302, thus results in greater earnings quality (Kalelkar & Nwaeze, 2011; Brown et al., 2008). However, this result does not hold for firms which have a great percentage of equity shares held by institutional investors (Kalelkar & Nwaeze, 2011, p. 292). It can thus be said that a weakness under SOX section 302 should be associated with impaired earnings quality (He & Thornton, 2013).

The study of Doyle et al. (2007a) examined the determinants of internal control weaknesses. The findings suggest that firms disclosing internal control weaknesses are considered to be smaller, more complex, undergoing restructuring, financially weaker or growing rapidly. In another study done by Doyle et al. (2007b) the relations between accruals quality and internal controls is examined. This study includes firms with at least one

disclosed material weakness. The findings suggest that internal control weaknesses are generally associated with lower accruals quality. Furthermore, Doyle et al. (2007b) find that this relation is driven by company-level control weaknesses. Doyle et al.( 2007b, p. 220) state that this is true as company-level control weaknesses are more difficult to audit.

Overall the findings suggest that firms that disclose material weaknesses under SOX section 302 have lower earnings quality. However, there are no real quantitative results providing an indication of the severity of the impairment of earnings quality. The event studies suggest that investors perceive SOX section 302 with more negative cumulative abnormal returns. It can thus be said that investors use SOX section 302 disclosures to value a firm. This results in the following hypothesis:

(14)

Hypothesis 1: Material weakness disclosures under SOX section 302 cause investor belief

revision that is reflected in positive earnings response coefficients and lower earnings quality.

Event studies SOX section 404

In this section the literature concerning event studies on SOX section 404 internal control disclosures will be presented and discussed.

The empirical evidence suggests that auditors’ Section 404 internal control weakness

disclosures are not associated with abnormal returns around the announcement date (Ogneva et al., 2007; Beneish et al., 2008). Beneish et al. (2008) conclude that the information

environment of ﬁrms that are required to report under Section 404 is richer and this attenuates the surprise or that SOX 404 reports may reﬂect a low materiality threshold for disclosure.

In an additional analysis, Ashbaugh-Skaife et al. (2009) in their working paper find a significant negative market reaction to SOX 404 reports, and also their cross-sectional test indicates that the systematic risks are higher for firms disclosing internal control weaknesses. Similarly, Schneider & Church (2008) find that the bank loan officers assessments are

negatively affected by disclosed internal control weaknesses.

De Franco et al. (2005) use a sample of 102 firms reporting internal control weaknesses without other material news in the event window to investigate whether the market reaction to the internal control weaknesses varies by investor size. The study aims to test whether a large investor reacts differently to internal control weaknesses than a smaller investor. The findings suggest that there are size-adjusted returns in the 3-day window surrounding the disclosure of internal control weaknesses.

Other studies SOX section 404

In this section the prior literature regarding studies that include earnings quality in their internal control deficiencies study will be presented and discussed. Based on this discussion another hypothesis will be presented.

The study of Ghosh & Lubberink (2007) state that SOX section 404 disclosures are, due to more timely identification of internal control deficiencies, leading to an enhancement of investor confidence in financial reporting. Ghosh & Lubberink (2007) build upon this and provide new insights into the timeliness of the identification of internal control deficiencies. The findings suggest that firms with a higher probability of reporting internal control

(15)

weaknesses have lower earnings response coefficients, higher cost of debt, larger analyst forecast errors and less favorable debt ratings and common stock rankings.

The studies of Ashbaugh-Skaife et al. (2008) and of Bedard et al. (2012) are similar to that of Ghosh & Lubberink (2007) and examine the effects of internal control deficiencies (ICD) on a firms’ accrual and earnings quality respectively. Furthermore, both studies also look at the effect of remediation of these ICDs on accrual and earnings quality respectively (Ashbaugh-Skaife et al., 2008; Bedard et al., 2012).

Ashbaugh-Skaife et al. (2008) look whether the firm disclosed an ICD under SOX section 404. In additional analysis Ashbaugh-Skaife et al. (2008) also look at whether this firm also disclosed an ICD under SOX section 302 in the same year. Furthermore, the researchers also look forward to see whether this firm has another SOX section 404 opinion. The results suggest that the accruals for an ICD firm are less reliable and noisier (Ashbaugh-Skaife et al., 2008, p. 237). Also Bedard et al. (2012, p. 70) come to this conclusion. These findings are similar to the findings of Ghosh & Lubberink (2007).

Firms that report a SOX section 404 opinion in successive years, with worse internal controls in the second year, exhibit a significant negative impact on its accrual quality. There is no change in accrual quality for firms with the same SOX section 404 opinion in both years. Firms with improved internal controls exhibit a small positive impact on its accrual quality (Ashbaugh-Skaife et al., 2008).

Bedard et al. (2012) base their findings, in contradiction to Ashbaugh-Skaife et al. (2008) and Ghosh & Lubberink (2007), on specific firm characteristics. First, Bedard et al. (2012) look at firm characteristics of firms with material weaknesses and whether these weaknesses are remediated. After that, the study looks at the differences between these characteristics in its effect on earnings quality.

The findings suggest that there are differences in firm characteristics between firms remediating their weaknesses and firms who are not able to remediate their weaknesses. Firms with for example fewer resources or with weaker governance are less likely to

remediate their weaknesses in a timely manner (Bedard et al., 2012, p. 70). Combining these findings with the effect of different types of material weaknesses on earnings quality suggest that there are types of problems which are less easy to remediate, but which have a great positive impact on the earnings quality if remediated (Bedard et al., 2012, p. 75). Where Ashbaugh-Skaife (2008) find a small positive impact on earnings quality for remediated material weaknesses, Bedard et al. (2012) find a greater impact. However, this only holds for

(16)

some type of weaknesses and is also dependent on firm characteristics (Bedard et al., 2012, p. 75).

Lopez et al. (2009) performed a study based on the underlying assumption of SOX section 404 that the SOX requirements for internal controls should provide value-relevant information for investors. He & Thornton (2013) perform a similar study in which the

investor perceived earnings quality is examined. Arnold et al. (2011) suggest that in order for SOX section 404 information to affect stock prices, users must access that information, assess its implications and incorporate those implications into their decision process. Arnold et al. (2011, p. 252) find that professional investors are more likely to value SOX section 404 disclosures appropriately.

Lopez et al. (2009) find that an adverse opinion on internal control provides the

investor with incremental value-relevant information. For an unqualified opinion this is based on the assumption that investors assess a higher risk of material misstatement, higher cost of capital, lower sustainability of earnings and lower predictability of earnings. Material

weaknesses reported under SOX section 404 are thus associated with lower investor perceived earnings quality (Lopez et al., 2009, p. 243).

In contradiction with Lopez et al. (2009), the initial results of He & Thornton (2013) suggest that investors do not perceive SOX-mandated disclosures informative for investors. However, additional analysis suggests that firms exhibit greater investor perceived earnings quality when firms remediate the material weaknesses in internal controls.

Overall the prior literature suggests that earnings quality is impaired by material

weakness disclosures under SOX section 404. However, there are some circumstances which increase or decrease this effect. Although there has not been a comparison between SOX section 404 disclosures and SOX section 302 disclosures, I use the findings of the prior literature based on event studies as an indication for how strongly investors respond to both types of disclosures. Based on these results I expect that for SOX section 404 material weakness disclosures earnings quality is lower than for SOX section 302 material weakness disclosures. This results in the following hypotheses:

Hypothesis 2a: Material weakness disclosures under SOX section 404 cause investor belief

revision that is reflected in positive earnings response coefficients and lower earnings quality.

(17)

Hypothesis 2b: Material weakness disclosures under SOX section 404 cause investor belief

revision that is reflected in positive earnings response coefficients and lower earnings quality than under SOX section 302 material weakness disclosures.

3. Research design

This chapter focuses on the how the research is designed. This includes measures for the market reaction to internal control deficiencies. Furthermore the important variables and the control variables are explained.

3.2 Empirical Model

This research uses a different model for both hypotheses. Underneath both models are presented and also the different (control) variables are explained. For this model the model used by Ghosh & Lubberink (2007) is used as a basis.

Hypothesis 1

CAR = β

0 + β1(MW302t)+β2E + β2ΔE + β4E·(MW302t) + β5ΔE·(MW302t) + β6E·Leverage

+ β₇ΔE·Leverage + β₈E·Size + β

9ΔE·Size + β10E·Growth + β11ΔE·Growth + β12E·Big4 +

β₁₃ΔE·Big4 + ν

CAR= Cumulative Abnormal Return, calculated as the cumulated raw returns –

value-weighted CRSP market return.

MW302= Indicator variable that equals 1 if a material weakness was reported under SOX

section 302, and 0 otherwise.

E= Earnings before extraordinary items.

ΔE= Change in earnings before extraordinary items between current year and last year.

--- both E & ΔE are deflated by market value of equity (Common Shares Outstanding x Price Fiscal Year Close)

Leverage= Ratio of total debt (long term debt + debt in current liabilities) and total assets. Size= Logarithmic transformation of the fiscal year-end market value of equity.

Growth= Sum of the market value of equity & book value of debt, scaled by the book value

of total assets

(18)

Big4= Indicator variable that equals 1 if the auditor is one of the Big 4 audit firms, and 0

otherwise.

Hypothesis 2a & 2b

CAR = β

0 + β1(MW404t)+ β2E + β2ΔE + β4E·(MW404t) + β5ΔE·(MW404t) + β6E·Leverage +

β₇ΔE·Leverage + β₈E·Size + β

9ΔE·Size + β10E·Growth + β11ΔE·Growth + β12E·Big4 +

β₁₃ΔE·Big4 + ν

CAR= Cumulative Abnormal Return, calculated as the cumulated raw returns –

value-weighted CRSP market return.

MW404= Indicator variable that equals 1 if a material weakness was reported under SOX

section 302, and 0 otherwise.

E= Earnings before extraordinary items.

ΔE= Change in earnings before extraordinary items between current year and last year.

--- Both E & ΔE are deflated by market value of equity (Common Shares Outstanding x Price Fiscal Year Close)

Leverage= Ratio of total debt (long term debt + debt in current liabilities) and total assets. Size= Logarithmic transformation of the fiscal year-end market value of equity.

Growth= Sum of the market value of equity & book value of debt, scaled by the book value

of total assets.

Big4= Indicator variable that equals 1 if the auditor is one of the Big 4 audit firms, and 0

otherwise.

Following Ghosh & Lubberink (2007) the earnings response coefficient (ERC) is defined as the sum of the coefficients β2 and β3. If investors value internal control material weakness

disclosures, β1 is expected to be positive. Also the earnings variables (E and ΔE) are

interacted with MW(302)/ MW(404) to find the interaction coefficient β

4+β5 , based on

which it can be concluded if the earnings quality is different for firms with internal control weaknesses. If investors associate lower earnings quality with firms that disclose internal control problems, β

4 + β5 is expected to be negative (Ghosh & Lubberink, 2007, p. 12).

(19)

Control variables

In this research the control variables Leverage, Size, Growth and Big4 are included to control for other firm characteristics associated with the earnings response coefficient (ERC) which might also be correlated with internal control deficiencies. These follow from previous studies done by Ghosh & Lubberink (2007) and Collins & Kothari (1989).

The first control variable is Leverage. Leverage is included in this model because of contracting considerations. Firms with high leverage are more likely to pursue earnings management with the aim to avoid the possible violation of debt-contracts (Ghosh & Moon, 2005, p. 591). Firms with high leverage are thus associated with lower earnings quality (Ghosh & Lubberink, 2007, p. 12).

Second, Size is used as a control variable. The use of this control variable can be explained using the political cost theory. In large firms, which are politically sensitive, managers are more likely to pursue earnings management to reduce political costs, thereby affecting earnings quality (Ghosh & Moon, 2005, p. 591). Smaller firms are thus associated with greater earnings quality (Ghosh & Lubberink, 2007, p. 18).

The third control variable is Growth. This control variable is primarily used as for valuation purposes (Ghosh & Moon, 2005, p. 591).

Fourth, and last, Big 4 is used as a control variable. This control variable is used as a control variable because generally large audit firms are associated with high audit quality, and therefore also with higher earnings quality (Ghosh & Lubberink, 2007, p. 12).

4. Data

This chapter focusses on the data used in this research. In section 4.2 the sample selection is explained. Section 4.3 describes the descriptive statistics.

4.2 Sample selection

In this section I explain how the sample selection was carried out. Furthermore, I also describe how the data is modified before the regression analysis.

The research question of this study will be answered based on archival research. This means that this study is using existing data from databases to determine an answer to the research question.

The data is collected for U.S. firms for the years 2007 to 2012. The main reason for using U.S. firms is that U.S. firms are mandated to report under SOX requirements. This

(20)

means that these firms are obliged to report on their internal control quality under SOX section 302 and SOX section 404. Furthermore, by using U.S. firm data enough data is

collected to be able to answer the research question. In this way a more credible answer to the research question is expected.

As already stated above, data will be obtained from a time-period from 2007 to 2012. This yields a 5 year time-period which should be sufficient to be able to say something about the market reaction to internal control deficiencies.

Data about Big 4 auditors and about SOX section 302/ 404 internal control material weakness disclosures is collected from the WRDS AuditAnalytics database. The returns information used for the calculation of the cumulative abnormal returns (CAR) is found in the WRDS CRSP database. For all other variables data is found in the WRDS Compustat

database.

After collecting and merging all the necessary datasets from the WRDS databases an OLS regression analysis will be made. Before the regression analysis is actually made, the data is winsorized as recommended by Ghosh & Moon (2005, p. 593). First, I will winsorize the top and bottom 1% of the observations for earnings (variable E) and changes in earnings (variable ΔE). Second, I will remove all observations with the absolute value of the

cumulative abnormal return (variable CAR) greater than 100%. Third, I will winsorize the top and bottom 1 percent of the observations for the control variables Growth, Size & Leverage. This selection method resulted in 3,869 firm year observations for the “full” sample. This sample includes both the data about SOX section 302 material weakness disclosures and the data about the SOX section 404 material weakness disclosures.

4.3 Descriptive statistics

In this section the descriptive statistics will be presented and discussed. Furthermore, this section also includes a comparison between the means for firms with and without material weaknesses. Finally, this section presents and discusses the correlation matrix for dependent and independent variables.

First, in table 4.3.1 the observations per industry of the final sample are presented. The total number of firm year observations is 3,869 observations. As can be seen in table 4.3.1, the manufacturing industry (SIC codes 2000-3990) represents the greatest portion of the total sample with 1,469 firm year observations. This represents approximately 38% of the total sample.

(21)

Table 4.3.1: Observations per industry

Industry name SIC codes Number of observations

Agriculture, Forestry and Fishing 0100-0999 6

Agricultural, Production-Livestock and Animal Specialties 1000-1999 223

Manufacturing 2000-3999 1,469

Transportation, Communication, Electric, Gas and Sanitary 4000-4999 322

Wholesale and Retail Estate 5000-5999 307

Financial 6000-6999 949

Services 7000-8999 576

Public Administration 9000-9999 17

Table 4.3.2 presents the descriptive statistics. In this table the mean, median, standard deviation, minimum and maximum are presented for every dependent and independent variable of the empirical model and also for the control variables Leverage, Size, Growth and Big4.

As can be seen in table 4.3.2 the mean cumulated abnormal return (CAR) small but positive. This is not as expected. In the study of Ghosh & Lubberink (2007) and Ghosh & Moon (2005) this mean is negative. For variable E and for variable ∆E the means are both negative. This is not significantly different from the results of Ghosh & Moon (2005) and Ghosh & Lubberink (2007) as these studies also present negative or close to zero means for these variables. A possible explanation for this is that Ghosh & Moon (2005) and Ghosh & Lubberink (2007) use data from different years and different firms. This makes comparing these studies more difficult.

The descriptive statistics of the variables MW302 and MW404 suggest that the sample firms report few material weaknesses and thus have for a large part effective internal

controls. Approximately 4 percent of the sample report material weaknesses under SOX 302 or SOX 404. The study of Ghosh & Lubberink (2007) uses the probability of a material weakness in their empirical model. It is thus not possible to compare these numbers in detail as Ghosh & Lubberink (2007) use a model to calculate this probability and do not directly use data from a database.

The interaction variables E*MW302 and E*MW404 suggest that firms with material weaknesses have more negative earnings. This effect is a bit larger for the interaction variables ∆E*MW302 and ∆E*MW404.

(22)

Furthermore, 67 percent of the sample is audited by a Big 4 audit firm. The Big 4 audit firms are KPMG, PwC, Deloitte and EY. This statistic means that the largest part of the sample firms should have high quality audits and thus more reliable SOX 302 and SOX 404 disclosures. Firms are financed for approximately 22 percent with debt (Leverage mean 0.2212) . Furthermore I can conclude that this sample consists of high growth firms.

Table 4.3.2: Descriptive statistics

Variable Mean Median Std. Dev. Minimum Maximum

CAR 0.0087584 -0.0020549 0.1322832 -0.5866634 0.99634 E -0.1254702 0.0411079 0. 6767648 -5.275917 0.29101 ∆E -0.1495966 0.0043768 0.9947152 -7.473831 2.1135 MW302 0.0380118 0 0.1982277 0 1 MW404 0.0409724 0 0.1986181 0 1 E*MW302 -0.0149593 0 0.2253728 -5.275917 0.29101 ∆E*MW302 -0.0149755 0 0.3115671 -7.473831 2.1135 E*MW404 -0.0197757 0 0.2693461 -5.275917 0.29101 ∆E*MW404 -0.0222531 0 0.3747278 -7.473831 2.1135 Leverage 0.2212005 0.1556899 0.2212005 0 0.97858 Size 6.215723 6.204414 2.028803 1.624772 11.1654 Growth 1.364639 0.9926282 1.332484 0.0855585 8.07838 Big4 0.6675196 1 0.4711056 0 1

Ghosh & Lubberink (2007, p. 18) state that there are several firm characteristics that are associated with firms with internal control problems . These firm characteristics can be high growth, poor performance, complex business operations and smaller firm size. Capital markets are able to draw conclusions about the internal control problems of the firm based on these firm characteristics (Ghosh & Lubberink, 2007, p. 18). The comparison of the means between firms reporting material weaknesses under SOX section 302 and under SOX section 404 are presented in table 4.3.3.

As can be seen in table 4.3.3 the CAR is higher for firms reporting material weaknesses under both SOX 302 and SOX 404 than for firms reporting no material weaknesses in the internal controls. Furthermore, the earnings are more negative for firms reporting material weaknesses under SOX 302 or SOX 404. What is interesting to see is that firms with a material weakness reported under SOX section 302 in approximately 59 percent of the cases

(23)

also report a material weakness under SOX section 404. Firms disclosing a material weakness under SOX section 404, have in approximately 55 percent of the cases also a material

weakness reported under SOX section 302.

For the variable Leverage it does not make a difference whether a firm discloses material weaknesses or not. The means are approximately the same for every scenario. However, it can be seen that firms reporting material weaknesses tend to have a smaller firm size.

Firms reporting material weaknesses are also less often audited by a Big 4 audit firm. This is even more true for material weaknesses reported under SOX section 404. This means that non-Big 4 audit firms more often come to the opinion that a firm has internal control deficiencies. The growth rate of the companies does not significantly differ between firms reporting material weaknesses and firms not reporting material weaknesses.

As the means in table 4.3.3 suggest, the data largely covers the firm characteristics stated by Ghosh & Lubberink (2007, p. 18). Investors should thus be able to make inferences about the internal control problems of a firm (Ghosh & Lubberink, 2007, p. 19).

Table 4.3.3: Comparison of means between firms with and without material weaknesses

SOX 302 SOX 302 SOX 404 SOX 404

MW No MW MW No MW

Variable Mean Mean Mean Mean

CAR 0.0005562 0.0090825 0.0090825 0.0089455 E -0.3935429 -0.1148777 -0.4826582 -0.1102102 ∆E -0.3939693 -0.1399405 -0.5431238 -0.132784 MW302 1 0 0.5461147 0.0163042 MW404 0.5886497 0.0193316 1 0 E*MW302 -0.3935429 0 -0.2309778 -0.0057303 ∆E*MW302 -0.3939693 0 -0.2718893 -0.0039994 E*MW404 -0.2489679 -0.0107194 -0.4826582 0 ∆E*MW404 -0.2930658 -0.0115522 -0.5431238 0 Leverage 0.2200081 0.2158819 0.2210648 0.215824 Size 5.216286 6.255214 4.93396 6.270483 Growth 1.308813 1.366845 1.337978 1.365778 Big4 0.4821918 0.6748426 0.4092229 0.6785548 23

(24)

Table 4.3.4 presents the correlation matrix for the dependent and independent variables. This matrix tells us whether the dependent and independent variables are correlated with each other and whether this correlation is significant (Ghosh & Moon, 2005, p. 596). As I use different models for MW302 and MW404 I blocked the correlations between these variables as these are not relevant for my model. This is also true for the interaction variables

containing MW302 and MW404.

As can be seen in table 4.3.3 are CAR and E negatively correlated and significant at the 1% level. The correlation between CAR and ∆E is positive and significant at the 1% level. Also MW302 and MW404 are significantly correlated with CAR. MW302 is negatively correlated with CAR at 1% significance, while MW404 is negatively correlated at the 10% level. Also all the interaction variables are positively and significantly correlated at the 1% level with CAR.

Furthermore, it can be seen that MW302 and MW404 are negatively correlated with E and ∆E. This correlation is significant at the 1% level. MW404 is more negatively correlated with E and ∆E than MW302. This is in line with my hypotheses that firms reporting material weaknesses under SOX section 404 experience more negative earnings response coefficient.

Table 4.3.4: Correlation matrix for dependent and independent variables

Variable CAR E ∆E MW302 MW404 E*MW302 ∆E*MW302 E*MW404 ∆E*MW404

CAR 1 E -0.1046 1 0.0000* ∆E 0.0638 0.5036 1 0.0000* 0.0000* MW302 -0.0123 -0.0787 -0.0488 1 0.0014* 0.0000* 0.0000* MW404 -0.0068 -0.1091 -0.0818 0.5492 1 0.0761*** 0.0000* 0.0000* 0.0000* E*MW302 0.0401 0.3222 0.1637 -0.3339 -0.1981 1 0.0000* 0.0000* 0.0000* 0.0000* 0.0000* ∆E*MW302 0.0294 0.1758 0.3067 -0.2418 -0.1704 0.5515 1 0.0135* 0.0000* 0.0000* 0.0000* 0.0000* 0.0000* E*MW404 0.0398 0.3865 0.2258 -0.1691 -0.3552 0.5692 0.336 1 0.0000* 0.0000* 0.0000* 0.0000* 0.0000* 0.0000* 0.0000* ∆E*MW404 0.0243 0.2392 0.3691 -0.1437 -0.2873 0.3334 0.588 0.6242 1 0.0000* 0.0000* 0.0000* 0.0000* 0.0000* 0.0000* 0.0000* 0.0000*

*/**/*** represent significance levels of respectively 1/5/10%

(25)

5. Empirical Results

This chapter focuses on the results of the regressions. The results will be presented in tables and thereafter these tables will be discussed and elaborated upon.

5.2 Investor responsiveness to internal control deficiencies (SOX 302) In this section the regression results are presented and discussed. This will be done for material weaknesses disclosed under SOX section 302. In this section I provide evidence on whether I can accept or reject hypothesis 1.

Table 5.2.1 reports the results of the OLS regression of returns (CAR) on earnings (E and ∆E). Also it reports on the interaction between earnings and reporting material

weaknesses under SOX section 302. Important for my hypothesis is the ERC (β2+β3) and the

interaction coefficients β4+β5. The results are reported in two columns. Column (I) represents

an OLS regression where the control variables are excluded and column (II) represents an OLS regression where the control variables are included.

The adjusted R2 is 0.0111 for column (I) and 0.0124 for column (II) respectively. This indicates that the OLS regression with the control variables included is stronger related to the returns (CAR). However, this relationship remains weak.

Consistent with prior literature I find a positive ERC with high significance of 1% in both columns (coefficient 0.0205, t-statistic 25.30 in column I; coefficient 0.0280, t-statistic 14.42). This is consistent with what was expected for hypothesis 1. More importantly however is that I find positive and significant, at the10% level, results on the interaction coefficient β4+β5 in both column (I) as in column (II) (coefficient 0.0045, t-statistic 1.75 for

column I; coefficient 0.0046, t-statistic 1.77 for column II). This is not consistent with prior literature. My results thus suggest that the earnings of firms reporting material weaknesses under SOX section 302 between 2007-2012 increased with approximately 16 percent (0.0046/0.0280) relative to firms which did not disclose material weaknesses under SOX section 302.

For the control variables only the interaction variables between earnings and size and between earnings and growth are significant. For the interaction between size and earnings this means a negative coefficient with a significance level of 1%. This is not expected by prior literature, which indicates that ERCs should be higher for larger firms (Ghosh & Moon,

(26)

2005). For the interaction between growth and earnings this means a coefficient of 0.0021 and a significance level of 10%. This is consistent with what was expected (Ghosh & Moon, 2005). The interactions between earnings and leverage and between earnings and Big 4 are not significant.

To summarize, the results suggest, consistent with my hypothesis, that investors

anticipated the disclosure of material weaknesses under SOX section 302. However, investors value earnings at a higher price when firms disclose material weaknesses under SOX section 302. It could thus be said that investors perceive the earnings quality of firms to be higher when firms disclose material weaknesses under SOX section 302. This is inconsistent with my hypothesis. Hypothesis 1 can thus only be partly accepted: the ERCs are positive for firms that disclose material weaknesses under SOX section 302 but these material weaknesses do not have a negative impact on the earnings quality is of the firm.

Table 5.2.1: OLS regression results to test for hypothesis 1

I II

Variable Coefficient Coefficient

[t-statistic] [t-statistic] Intercept β0 0.0115 0.0120 [21.80]* [22.64]* MW302 β1 -0.0011 -0.0017 [-0.39] [-0.60] E β2 0.0187 0.0323 [20.47]* [14.46]* ∆E β3 0.0017 -0.0043 [2.78]* [-2.46]** E*MW302 β4 0.0017 0.0014 [0.59] [0.49] ∆E*MW302 β5 0.0278 0.0032 [1.35] [1.54]

Control variables Excluded Included

E*Leverage β6 0.0010 [0.31] ∆E*Leverage β7 _0.0022 [1.02] E*Size β8 _-0.0052 [-8.14]* 26

(27)

∆E*Size β9 0.0019 [4.45]* E*Growth β10 _0.0033 [3.04]* ∆E*Growth β11 -0.0013 [-2.68]* E*Big β12 -0.0003 [-0.15] ∆E*Big4 β13 -0.0003 [-0.22] E+∆E β2+β3 0.0205 0.0280 [25.30]* [14.42]* E*MW302+∆E*MW302 β4+β5 0.0045 0.0046 [1.75]*** [1.77]*** E*Leverage+ ∆E*Leverage β6+β7 0.0031 [1.18] E*Size+∆E*Size β8+β9 -0.0033 [-5.80]* E*Growth+∆E*Growth β10+β11 0.0021 [1.78]*** E*Big4+∆E*Big4 β12+β13 -0.0006 [-0.35] Adjusted R2 0.0111 0.0124

Data are for the years 2007-2012. The total number of firm year observations is 3,869 observations for both SOX section 302 and SOX section 404.

*,**,*** are significance levels representing respectively 1, 5 or 10% based on a two-tailed test. CAR = β₀+ β₁(MW302t)+β

2E + β2ΔE + β4E·(MW302t) + β5ΔE·(MW302t) + β6E·Leverage + β7ΔE·Leverage + β₈E·Size + β

9ΔE·Size + β10E·Growth + β11ΔE·Growth + β12E·Big4 + β13ΔE·Big4 + ν. CAR= Cumulative Abnormal Return, calculated as the cumulated raw returns – value-weighted CRSP market return. MW302= Indicator variable that equals 1 if a material weakness was reported under SOX section 302, and 0 otherwise. E= Earnings before extraordinary items. ΔE= Change in earnings before extraordinary items between current year and last year. Both E & ΔE are deflated by market value of equity (Common Shares Outstanding x Price Fiscal Year Close) Leverage= Ratio of total debt (long term debt + debt in current liabilities) and total assets. Size= Logarithmic transformation of the fiscal year-end market value of equity. Growth= Sum of the market value of equity & book value of debt, scaled by the book value of total assets. Big4= Indicator variable that equals 1 if the auditor is one of the Big 4 audit firms, and 0 otherwise.

(28)

5.3 Investor responsiveness to internal control deficiencies (SOX 404) In this section the regression results are presented and discussed. This will be done for material weaknesses disclosed under SOX section 404. In this section I provide evidence on whether I can accept or reject hypothesis 2.

Table 5.3.1 reports the results of the OLS regression of returns (CAR) on earnings (E and ∆E). Also it reports on the interaction between earnings and reporting material

weaknesses under SOX section 404. Again, the focus is on the ERC (β2+β3) and the

interaction coefficients β4+β5. The results are reported in two columns. Column (I) represents

an OLS regression where the control variables are excluded and column (II) represents an OLS regression where the control variables are included.

The adjusted R2 is 0.0111 for column (I) and 0.0123 for column (II) respectively. This indicates that the OLS regression with the control variables included is stronger related to the returns (CAR). However, this relationship remains weak.

Consistent with prior literature I find a positive ERC with high significance of 1% in both columns (coefficient 0.0189, t-statistic 20.25 in column I; coefficient 0.0323, t-statistic 14.41). This is consistent with what was expected in hypothesis 2a. More importantly however I find negative, but insignificant results on the interaction coefficient β4+β5 in both

column (I) as in column (II) (coefficient 0.0001, tstatistic 0.05 for column I; coefficient -0.0020, t-statistic -0.88 for column II). The sign of the coefficient is consistent with prior literature (Ghosh & Lubberink, 2007, p. 21). If my results were to be significant, my results would thus suggest that the earnings of firms reporting material weaknesses under SOX section 404 between 2007-2012 are discounted with approximately 6 percent (0.0020/0.0323) relative to firms which did not disclose material weaknesses under SOX section 404.

Consistent with the OLS regression results for SOX 302, also for the control variables for the OLS regression for SOX 404 only the interaction variables between earnings and size and between earnings and growth are significant. For the interaction between size and earnings this means a negative coefficient with a significance level of 1%. As already stated in section 5.2 this is not expected by prior literature which indicate that ERCs should be higher for larger firms (Ghosh & Moon, 2005). For the interaction between growth and earnings this means a coefficient of 0.0021 and a significance level of 10%. This is consistent with what was expected (Ghosh & Moon, 2005). Again, the interactions between earnings and leverage and between earnings and Big 4 are not significant.

(29)

To summarize, the results suggest, consistent with my hypothesis, that investors anticipated the disclosure of material weaknesses under SOX section 404. Also, investors value earnings at a lower price when firms disclose material weaknesses under SOX section 404. This would thus result in a lower earnings quality, which is also lower than for material weaknesses reported under SOX section 302. This is also consistent with my hypothesis. However, these last results of how investors value earnings are insignificant for SOX section 404. Hypothesis 2a can thus only be partly accepted: the ERCs are positive for firms that disclose material weaknesses under SOX section 404, but these material weaknesses do not have significant impact on the perceived earnings quality of the firm.

Table 5.3.1: OLS regression results to test for hypothesis 2a

I II

Variable Coefficient Coefficient

[t-statistic] [t-statistic] Intercept β0 0.0113 0.0118 [21.49]* [22.31]* MW404 β1 0.0032 0.0026 [1.16] [0.94] E β2 0.0189 0.0323 [20.25]* [14.41]* ∆E β3 0.0023 -0.0035 [3.60]* [-1.97]** E*MW404 β4 0.0021 -0.0004 [0.79] [-0.17] ∆E*MW404 β5 -0.0022 -0.0015 [-1.20] [-0.83]

Control variables Excluded Included

E*Leverage β6 0.0015 [0.48] ∆E*Leverage β7 _0.0022 [1.01] E*Size β8 -0.0050 [-7.96]* ∆E*Size β9 0.0019 [4.32]* E*Growth β10 0.0034 [3.06]* 29

(30)

∆E*Growth β11 -0.0013 [-2.75]* E*Big β12 _-0.0009 [-0.45] ∆E*Big4 β13 -0.0005 [-0.39] Interaction coefficients E+∆E β2+β3 0.0211 0.0288 [25.36]* [14.63]* E*MW404+∆E*MW404 β4+β5 -0.0001 -0.0020 [-0.05] [-0.88] E*Leverage+ ∆E*Leverage β6+β7 0.0036 [1.36] E*Size+∆E*Size β8+β9 -0.0032 [-5.69]* E*Growth+∆E*Growth β10+β11 0.0021 [1.77]*** E*Big4+∆E*Big4 β12+β13 -0.0141 [-0.80] Adjusted R2 0.0111 0.0123

Data are for the years 2007-2012. The total number of firm year observations is 3,869 observations for both SOX section 302 and SOX section 404.

*,**,*** are significance levels representing respectively 1, 5 or 10% based on a two-tailed test. CAR = β₀+ β₁(MW404t)+β

2E + β2ΔE + β4E·(MW404t) + β5ΔE·(MW404t) + β6E·Leverage + β7ΔE·Leverage + β₈E·Size + β

9ΔE·Size + β10E·Growth + β11ΔE·Growth + β12E·Big4 + β13ΔE·Big4 + ν. CAR= Cumulative Abnormal Return, calculated as the cumulated raw returns – value-weighted CRSP market return. MW404= Indicator variable that equals 1 if a material weakness was reported under SOX section 404, and 0 otherwise. E= Earnings before extraordinary items. ΔE= Change in earnings before extraordinary items between current year and last year. Both E & ΔE are deflated by market value of equity (Common Shares Outstanding x Price Fiscal Year Close) Leverage= Ratio of total debt (long term debt + debt in current liabilities) and total assets. Size= Logarithmic transformation of the fiscal year-end market value of equity. Growth= Sum of the market value of equity & book value of debt, scaled by the book value of total assets. Big4= Indicator variable that equals 1 if the auditor is one of the Big 4 audit firms, and 0 otherwise.

(31)

5.4 Comparison between investor responsiveness to internal control deficiencies under SOX section 302 and SOX section 404

In this section the results presented in section 5.2 and section 5.3 will be compared with each other. This will give a better look at how investors value information on internal control material weakness disclosures. Also it will be possible to say something about whether investors respond stronger to SOX section 404 material weakness disclosures than to SOX section 302 material weakness disclosures. Based on this comparison I will be able to say something about whether hypothesis 2b should be accepted or rejected.

First I will start by giving a small summary of the results presented in section 5.2. This will gradually develop into the comparison. The results for SOX section 302 material

weakness disclosures suggest that investors anticipated the disclosure of material weaknesses under SOX section 302. This was reflected in a positive and significant ERC (β2+β3).

However, investors perceive the earnings quality of firms to be higher when firms disclose material weaknesses under SOX section 302.

The results for SOX section 404 material weakness disclosures have similar results for whether investors anticipated the disclosure of material weaknesses under SOX section 404. Again, this was reflected in a positive ERC. However, the ERC for SOX section 404 material weakness disclosures is a little bit higher than for SOX section 302 material weakness

disclosures (SOX 302 coefficient 0.0280; SOX 404 coefficient 0.0288). Although the difference is small, this is consistent with my expectation reflected in hypothesis 2a.

Also, investors value earnings at a lower price when firms disclose material weaknesses under SOX section 404. This would thus result in a lower investor perceived earnings quality. For SOX section 302 material weakness disclosures the signs were positive and thus higher than for SOX section 404 material weakness disclosures (SOX 302 coefficient 0.0046; SOX 404 coefficient -0.0020). This is also consistent with my expectation reflected in hypothesis 2b. However, the results of SOX section 404 material weakness disclosures related to earnings quality (β4+β5) are not significant. It will thus not be possible to give a definite

accept or reject recommendation for hypothesis 2b.

Hypothesis 2b can thus only be partly accepted: the ERCs are more positive for firms that disclose material weaknesses under SOX section 404, but it will not be possible to say whether the earnings quality of firms is perceived to be lower for SOX section 404 material weakness disclosures than for SOX section 302 material weakness disclosures.

(32)

5.5 Sensitivity Analysis

In the sections presented above, I demonstrated the relationship between investor

responsiveness and the disclosure of internal control deficiencies under SOX section 302 and under SOX section 404. Also a comparison between these results was made. In this section the sensitivity of these results are examined.

First, I include raw returns as the dependent variable instead of cumulative abnormal returns. This is consistent with Ghosh & Moon (2005) and Collins & Kothari (1989). The reason for this is that raw returns are used at a frequent level for the estimation of

contemporaneous returns-earnings models. The results are presented in table 5.5.1 for SOX section 302 material weakness disclosures and in table 5.5.2 for SOX section 404 material weakness disclosures. These tables can be found in Appendix A.

The results do not differ significantly from the results of the initial regression where cumulative abnormal returns are used as the dependent variable. For SOX 302 the ERC (β2+β3) is positive and significant at the 1% level, which is equal as in the initial regression

(initial coefficient 0.0280; raw returns coefficient 0.0343). Also, for the interaction coefficients β4+β5 the sign of the coefficient is equal to that of the initial regression.

However, in the raw returns regression model, this is not significant. This is in contrast with that of the initial regression where β4+β5 is significant at the 10% level.

For material weakness disclosures under SOX section 404 the raw-returns model finds similar results as the results from the initial regression. Again the ERC (β2+β3) is positive and

significant and the interaction coefficient β4+β5 is negative. One difference however, is that in

the raw-returns regression model for the regression including the control variables β4+β5 is

significant at the 10% level, whereas in the initial regression this result was insignificant. However, since the signs of all coefficients are equal to that of the initial regression and there are no significant differences between the models it can be said that the conclusions from the initial regression hold.

Second, consistent with Ghosh & Lubberink (2007), who performed a similar study, I exclude all financial firms from the analysis. This is because financial firms could be subject to special risk factors and accounting standards which may influence the results. (Viale et al., 2009). The results of this sensitivity analysis are presented in table 5.5.3 for SOX section 302 material weakness disclosures and in table 5.5.4for SOX section 404 material weakness disclosures. These tables can be found in Appendix B

Investor responsiveness to internal control deficiencies : an earnings quality approach