Faculty of Economic and Business
The relationship between growth and internal
control quality for non-listed firms
A comparison between large firms and SMEs
Master Thesis
Linda Pol 6056903 22 June 2014
First supervisor: Dr. Bo Qin
Abstract
SMEs account for a significant part of the employment rate as well as the gross domestic product, however there is limited research done is this field. The objective of this research is to provide empirical evidence on the relation between growth and the quality of internal control for non-listed firms, in particular small and medium sized enterprises. This is important because there is a gap in literature in this field. It is hypothesized that (1) non-listed firms that grow rapidly are more likely to have lower internal control quality; and (2) the relationship between growth and internal control quality is stronger/weaker for SMEs than for large firms. Two proxies for internal control quality are used, a dummy for whether a firm had a material weakness in internal control and an overall internal control score. Data on the internal control quality is gained through questionnaires conducted among audit managers of Ernst&Young, while the remaining information is hand-collected from annual reports. The results show no significant relationship between growth and internal control quality. However, results show that younger firms, firms that face losses and firms subject to merger, acquisition, reorganization or restructuring are more likely to have lower internal control quality. Results of the second hypotheses show no evidence for a moderating effect of size on the relation between growth and internal control quality. However, robustness tests show that a different measure for size does show a highly significant moderating effect of size on the relation between growth and internal control quality. Furthermore, robustness results show some indication that a larger sample size might lead to more significant results. The main limitation of this research is the sample size.
Table of Contents
1 Introduction ... 4
2 Literature review and hypotheses ... 7
2.1 Definitions ... 7
2.1.1 Internal control ... 7
2.1.2 Material weakness ... 9
2.1.3 Small and medium enterprises ... 9
2.2 Main prior literature ... 10
2.3 Hypotheses ... 12
3 Research design ... 15
3.1 Measurement ... 15
3.1.1 Internal control quality ... 15
3.1.2 Independent variables ... 17
3.1.3 Empirical model ... 20
3.2 Sample ... 21
4 Empirical results and discussions ... 23
4.1 Descriptive statistics and univariate tests ... 23
4.2 Multivariate regression ... 29
5 Robustness tests ... 36
6 Conclusion ... 47
References ... 50
1 Introduction
The objective of this research is to provide empirical evidence on the relation between growth and the quality of internal control for non-listed firms, in particular small and medium sized enterprises. This is important because there is a gap in literature in this field. Prior evidence on this relationship was mainly based on large, listed firms. There is a lack of accordance in literature about whether this relation also applies for SMEs and to what extent. Since SMEs account for a significant part of the employment rate as well as the gross domestic product, insight in this could be valuable.
The importance of internal control over financial reporting is widely recognized (Kerr and Murthy, 2013; Doyle, Ge and McVay, 2007; Ashbaugh-Skaife, Collins, & Kinney Jr., 2007; Kinney, 2000). Companies should have strong internal controls, since weaknesses increase the risk of material misstatements in the annual reports, leading to less reliable financial statements (Hoitash, Hoitash, & Johnstone, 2012, Doyle et al., 2007; Ashbaugh-Skaife et al., 2007). This could eventually lead to bankruptcy. For example, Lehman Brothers, a large US investment bank, went bankrupt in 2008. The bankruptcy report of the examiner extensively describes how the lack of internal control was one of the main reasons for the bankruptcy at Lehman Brothers (Bankruptcy report, 2010).
There are factors that influence the quality of internal control. It is assumed that growth is one of these factors. However, this is only proved for listed US firms
(Chernobai & Yasuda, 2013; Doyle et al., 2007; Ashbaugh-Skaife et al., 2007) and not for non-listed firms. This research examines the following question: “What is the relation between growth and the quality of internal control for non-listed firms, in particular SMEs?” This is examined in two steps. First, the relation is examined for all non-listed firms. Second, a difference is made between large firms and SMEs, this means that there is examined whether ‘size’ is a moderator.
Prior research has shown that growth is an important determinant for material weaknesses in internal control for listed US firms. However, there is a lack in literature whether this also applies to listed firms. This research therefore focuses on the non-listed firms, especially small and medium enterprises (SMEs). Focusing on non-non-listed
firms, especially SMEs, is important for two reasons. First, it is particularly important to gain insight in the determinants for material weaknesses in internal controls for SMEs, since they provide employment to 66.5% of the total employees in the European Union and the SME sector as a whole delivered 57.6% of the gross domestic product (GDP) of the EU (European Commission, 2013). In the US, SMEs account for 64% of the job creation and 46% of the GDP (U.S. Small Business Administration, 2012a). It is therefore public interest that these companies continue to exist, and the quality of internal control is an important factor in the continuity of firms (Hoitash et al, 2012). Second, from an academic point of view, prior research has only focused on large, listed firms. As Tilley (2000, p. 33) stated: “Small firms are not little big firms”. They have a different organizational structure (Buonanno, 2005), different managerial style, ownership and independence (Coviello & McAuley, 1999, as cited in: Brouthers & Nakos, 2004), and different needs (Perrini, Russo, & Tencati, 2007). This suggests that results for SMEs on growth as a determinant for material weakness in internal control might be different from results on large listed firms.
Based upon prior literature it is expected that there is a positive relation between growth and material weakness in internal control for non-listed firms. Furthermore, it is expected that size serves as a moderator on this relation. However, there is no
accordance in literature whether the relation is expected to be stronger of weaker for SMEs comparing to large firms. It is important to study whether size is a moderator, because this could imply different approaches in practice for SMEs and large firms. Size is expected to be a moderator in this relation because large firms and SMEs are very different in their managerial style, organizational structure and independence. Therefore, growth could have a different impact on the internal control.
To provide evidence for my predictions, a questionnaire is developed that scores the internal control of firms. From this questionnaire an average internal control score can be derived as well as whether a company has faced a material weakness in internal control. In this way the research also answers to whether firms that grow fast are having weaker internal controls instead of only focusing at whether a material weakness is more likely for growing firms. Prior research only examines the relation between growth and material weakness in internal control. Therefore, by adding an extra proxy for internal control quality, a broader insight is given on the relation between growth
and the quality of internal control in this research compared to prior research. The questionnaire is based on a model that Ernst & Young (EY) uses to clarify to management of firms what the quality of their internal control is for all different
processes. Furthermore, growth, along with all control variables, is determined through the annual reports of firms as well as additional information requested in the
questionnaire. The sample consists of 82 non-listed EY clients. For these firms, a completely filled out questionnaire was received.
The results have a two-folded societal contribution. First, the results can be used in practice by management of SMEs to create awareness of these determinants and furthermore to alert them to evaluate and reestablish the internal controls in case the organization has encountered one of the determinants. Results can also be used by auditors in their process of testing the internal controls, by alerting them to pay extra attention when firms faced one or more of the determinants in the recent past. Besides, this research contributes to the literature in two ways. First it contributes to the existing literature on internal control as it provides new insights on the applicability of
determinants of internal control to an unknown large part of the economy, SMEs. Besides, it adds to the literature a different way of measuring the quality of internal control. This can be used in future research as well as to confirm results of prior research.
My personal interest for this subject comes from the fact that the field of internal control interests me. Furthermore, I have noticed that smaller firms that grow fast tend to face difficulties in adapting quickly to the changing environment, especially in their financial and accounting departments. This results in a lot of bankruptcies, for example after the dot-com bubble. As a result of the dot-com bubble, a lot of SMEs in the IT environment went bankrupt. It is expected that these bankruptcies were due to the rapid growth that they encountered in the previous years without investing in internal controls. After years of continuous growth, the growth declined and their internal organization and internal controls proved to be not sufficient. I am curious whether I can prove this relationship.
The remainder of this paper is structured as follows. Chapter 2 describes, the most important concepts of the research are defined and prior literature is reviewed upon which the hypotheses are developed. The research method is described in chapter 3.
Chapter 4 discusses the descriptive statistics and results. Chapter 5 shows the robustness checks that are performed. Finally a conclusion is drawn upon the results and a
discussion is provided.
2 Literature review and hypotheses 2.1 Definitions
As can be understood from the research question, the main concepts of this research are ‘internal control’, ‘material weakness’ and ‘small and medium enterprises’. Therefore a definition of these concepts will be discussed below.
2.1.1 Internal control
The Committee of Sponsoring Organizations of the Treadway Commission (COSO) provides a widely-recognized framework that can be used to evaluate internal control (Klamm & Watson, 2009). The COSO definition of internal control is: “Internal control is a process, effected by an entity’s board of directors, management, and other personnel, designed to provide reasonable assurance regarding the achievement of objectives relating to operations, reporting, and compliance” (COSO, 2013). These objectives are: (1) the efficiency and effectiveness of the entity’s operations; (2) the reliability of financial reporting, and; (3) the compliance with laws and regulations. According to COSO (2013), internal control consists of five integrated components. These are the components of internal control that management should design and implement in order to provide reasonable assurance that the goals of the company are met. The components are: control environment, risk assessment, control activities, information & communication and monitoring. The control environment is the set of standards, procedures and structures that provide the basis for carrying out internal control throughout the organization. It contains for example integrity and ethical values and serves as a foundation for the other four components. Risk assessment is about identifying and analyzing all relevant risks related to the company. Control activities are the actions taken by management to mitigate risks to the achievement of the objectives. Information and communication makes sure that employees know their responsibilities and duties in the internal control system. And finally, monitoring is about evaluating to
ascertain that all five components are present and functioning. The COSO framework sets out seventeen principles representing the main objectives associated with the five components (COSO, 2013).
However, the SEC provides another definition of internal control, namely:
“a process designed by, or under the supervision of, the registrant's principal executive and principal financial officers, or persons performing similar functions, and effected by the registrant's board of directors, management and other personnel, to provide reasonable assurance regarding the reliability of financial reporting and the preparation of financial statements for external purposes in accordance with generally accepted accounting principles and includes policies and procedures that
§ pertain to the maintenance of records that in reasonable detail accurately and
fairly reflect the transactions and dispositions of the assets of the registrant;
§ provide reasonable assurance that transactions are recorded as necessary to
permit preparation of financial statements in accordance with generally
accepted accounting principles, and receipts and expenditures of the registrant are being made only in accordance with authorizations of management and directors of the registrant;
§ provide reasonable assurance regarding prevention or timely detection of unauthorized acquisition, use or disposition of the registrant's assets that could have a material effect on the financial statements.” (SEC, 2003).
In conclusion, the concept of internal control is defined in two ways. SOX section 404 uses the definition as composed by COSO. However, the definition that is used by the SEC is derived from the COSO definition, only this one focuses more on the reliability over financial reporting than the COSO definition which is more focused on the total operations and the role of management in the internal control (Doyle et al., 2007). This research focuses on the material weaknesses in internal control of organizations. Since the definition used in this research of material weaknesses is focused on the financial reporting, and the definition of the SEC on internal control is also focused on the financial reporting, the definition of the SEC on internal control is
used in this research. This is also the definition used in prior literature (Doyle et al., 2007; Ashbaugh-Skaife et al., 2007; Hoitash et al., 2008; Hoitash et al., 2012).
2.1.2 Material weakness
In 2005 the Public Company Accounting Oversight Board (PCAOB) introduced the following definition material weakness: “a significant deficiency, or combination of significant deficiencies, that results in more than a remote likelihood that a material misstatement of the annual or interim financial statements will not be prevented or detected” (PCAOB, 2005). The auditor is responsible for determining whether a company has a material weakness or not. In doing so, the auditor should consider the future consequences of the deficiency and the presence of compensating controls into account and whether these compensating controls are effective (PCAOB, 2005).
The U.S. Securities and Exchange Commission changed the definition a little in 2007: “a deficiency, or combination of deficiencies, in internal control over financial reporting such that there is a reasonable possibility that a material misstatement of the company’s annual or interim financial statements will not be prevented or detected on a timely basis by the company’s internal control over financial reporting” (SEC, 2007). The PCAOB then adopted this definition of the SEC in the auditing standards.
In this research the definition of the SEC will be used, because it is the most recent one and it is used by the SEC as well as in the auditing standards set by the PCAOB. This is also the definition used mostly in prior research (Doyle et al., 2007; Ashbaugh-Skaife et al., 2007; Hoitash et al., 2008; Hoitash et al., 2012).
2.1.3 Small and medium enterprises
The focus of this research is on non-listed firms. A further distinction is made between large enterprises and small and medium enterprises (SMEs). Therefore, it is important to determine the definition of an SME. According to the EU law an enterprise can be categorized an SME when the enterprise employs fewer than 250 persons and the annual turnover does not exceed 50 million euros and/or the annual balance sheet total does not exceed 43 million euros (European Commission, 2005). This means that it is compulsory to meet the threshold for the number of employees, while either the
turnover or the balance sheet total needs to be met. Within SMEs a further distinction can be made between medium-sized enterprises, small enterprises and microenterprises. A small enterprise is defined as an enterprise which employs fewer than 50 persons and whose annual turnover and/or annual balance sheet total does not exceed 10 million euros. A microenterprise employs fewer than 10 persons and has an annual turnover and/or annual balance sheet total that does not exceed 2 million euros (European Commission, 2005).
The US definition for SMEs is more complex, since this differs per sector. For some sectors the size standard is the average annual receipts in dollars, while for other sectors the standard is based on the average number of employees and the maximum amounts also differ between sectors (U.S. Small Business Administration, 2012b). For firms with a size standard of annual receipts, the amount of annual receipts should on average not exceed 16.9 million dollars in order to be categorized as an SME. The amounts On average, Average 16.9 . For most firms that have a size standard based on the number of employees, an enterprise can be categorized as an SME when it employs fewer than 500 persons (U.S. Small Business Administration, 2012b).
The definition of an SME differs between the European Union and the United States. The EU applies general requirements on employees and annual turnover and/or annual balance sheet total, while the US has different requirements for different sectors. Since the sample of this research will consist of EU firms and the application of the EU criteria is easier, the EU definition of an SME will be used. All firms that are too large to comply with the requirements for SMEs, are categorized as large firms.
2.2 Main prior literature
Prior research has shown that listed US firms that grow rapidly are more likely to report material weaknesses in internal control. Ashbaugh-Skaife, Collins and Kinney (2007) studied 538 different listed firms that disclosed internal control deficiencies prior to SOX 404 but after SOX 3021 became effective. Growth is defined in their research as
1
SOX 302 requires that “the CEO and CFO of an organization certify and assert to stakeholders that SEC disclosures are truthful and reliable, and that management has taken appropriate steps to satisfy themselves that the disclosure processes and controls in the company they oversee are capable of consistently producing financial information stakeholders can rely on” (SEC, 2003). So under SOX 302 it is the responsibility of management to report material weaknesses in internal control.
“the average percentage change in sales” (Ashbaugh-Skaife et al., 2007, p. 172). They expect that firms that grow rapidly, are more likely to report material weaknesses in internal control because they assume that these firms are more likely to have accounting information systems that are not able to adapt quickly enough to the rapid changes, leading to weaknesses in the system. Besides, they expect problems with the capability of personnel because of the rapid changes in the complexity of the business. They found, in accordance with their explanations, that firms that disclose these weaknesses are faster growing relative to firms that do not disclose internal control deficiencies.
Doyle, Ge and McVay (2007) conducted a similar research, only after the implementation of SOX 4042. Their research was conducted among 779 listed US firms that reported a material weakness between 2002 and 2005 along with a control group. In their research, the variable growth is used as a dummy variable. They expect to find that firms that grow fast are more likely to report a material weakness in their internal control due to the assumption that fast growing firms are not capable to let the internal control grow in the same pace, because the systems are often rigid and hard to transform. They found, similarly to their expectations, that firms that are subject to rapid growth are more likely to report material weaknesses in their internal control.
Chernobai and Yasuda (2013) also examined the relationship between rapid changes in the organizational environment and material weaknesses in the internal control. The difference with the aforementioned studies is that they conducted their research in Japan, where it is also mandatory for listed firms to report material
weaknesses under J-SOX3. They conducted their research among 75 Japanese firms that reported a material weakness during 2008 and 2009. Growth is defined as the “one-year sales growth in percentages” (Chernobai & Yasuda, 2013, p. 1528). They assume that firms that grow fast, adjust their policies completely to this growth and therefore somewhat release the standards for internal control. As a result, these firms are more likely to report material weaknesses in internal control. Furthermore, the assumptions of Doyle et al. (2007) and Ashbaugh-Skaife et al. (2007) are used in their expectations. Chernobai and Yasuda conclude that firms that grow rapidly are more likely to report
2
Under SOX 404 it became mandatory for the company’s external auditor to report on the reliability of management’s assessment of internal control (SEC, 2003).
3
J-SOX is the Japanese equivalent of SOX in the US and requires the same of companies on the internal control as SOX 404 (Chernobai & Yasuda, 2013).
material weaknesses in their internal control, in accordance with the studies of Doyle et al. (2007) and Ashbaugh-Skaife et al. (2007).
In conclusion, prior research shows that listed firms in the US as well as in Japan that grow fast are more likely to report material weaknesses in internal control. This research will examine whether this also applies to non-listed firms, in particular SMEs. In addition, this research adds another proxy for internal control quality, the internal control score. So it is examines whether firms that grow fast are more likely to have a lower internal control score.
2.3 Hypotheses
As reviewed above, there is guidance from prior literature regarding the influence of rapid growth on the quality of internal control for listed firms, however there has not been conducted any research among SMEs. The development of the hypotheses is therefore based on a mix of prior literature for this specific subject on listed firms and prior literature on SMEs. The first hypothesis is about the relationship between growth and the quality of internal control, while the second hypothesis examines whether ‘size’ has a moderating effect on this relationship.
To begin, prior research has shown that firms that grow rapidly are more likely to face material weaknesses in internal control, due to a number of reasons. Frist, firms that grow rapidly are likely to have an accounting information system (AIS) that is not able to adapt quickly enough to the rapid changes, leading to weaknesses in the AIS and therefore in the internal control (Ashbaugh-Skaife et al., 2007). This is because internal control systems are often rigid and hard to transform, which makes it difficult for the systems to ‘grow’ in the same pace as the company (Doyle et al., 2007). Besides that, prior research has shown that companies that grow fast are more likely to encounter personnel problems. This is due to the fact that because of the growth, the complexity of the business increases. Therefore, the capability of personnel is tested, causing the personnel system to be ineffective (Ashbaugh-Skaife et al, 2007). Moreover, prior research has shown that the standards for internal control are more likely to be neglected when a firms is focusing on growth (Chernobai & Yasuda, 2013). But, all
abovementioned is based on research conducted among large listed firms. There is however no indication that these assumptions would not apply for non-listed firms, therefore it is expected that these arguments also hold for non-listed firms.
In conclusion, firms that grow fast are more likely to encounter problems with their AIS and personnel and standards are more likely to be neglected in times of growth. Thus, it is expected that firms that grow rapidly are more likely to face material weaknesses in their internal control systems and more likely to have lower internal control score. The following hypotheses are therefore proposed:
Hypothesis 1a: Non-listed firms that grow rapidly are more likely to face a material weakness in internal control
Hypothesis 1b: Non-listed firms that grow rapidly are more likely to have lower internal control score
Small firms are not ‘little big firms’ because they have different organizational, managerial and ownership structures. Therefore the influence of growth on the internal control can be different for SMEs than for large firms. This research therefore compares the effect of growth on the internal control for SMEs with the effect in large firms. In other words, it is examined whether ‘size’ can be seen as a moderator. Based on prior research it is not clear whether a stronger or a weaker relation is expected for SMEs than for large firms.
As stated in the development of the first hypotheses, growing firms are more likely to face personnel problems. Personnel problems may however in times of growth be more likely for SMEs, than for large firms, since SMEs are less capable of attracting skilled personnel mainly because SMEs are less likely to adopt sophisticated recruitment (Bacon and Hoque, 2005). In times of growth more personnel needs to be attracted, if this personnel is less skilled this leads to weaker internal control (Ashbaugh-Skaife et al., 2007). Therefore, it is expected that the relation between growth and material weakness in internal control is stronger for SMEs than for large firms. Additionally, firms are more likely to neglect internal control standards when growing. This is however more likely for SMEs than for large firms, because within
SMEs there is more opportunity to deviate from the procedures than within large firms and prior research has shown that managers in SMEs are actually more likely to overrule internal controls compared to large firms (Krishnan & Yu, 2012). Therefore it is more likely that internal control standards will be neglected in times of growth within SMEs than in large firms, resulting in the relation between growth and material weakness in internal control being stronger for SMEs than for large firms.
On the other hand, also based on prior research, a weaker relation for growth on the internal control quality is expected for SMEs comparing to large firms. SMEs may be able to enhance internal control quality because of rapid growth. For example, growth creates opportunities for a company to hire new employees. This creates opportunities for improved segregation of duties, which strengthens the quality of internal control (Hoitash et al., 2008). Moreover, as mentioned above, rapidly growing firms are more likely to have an accounting system that cannot adapt quickly enough to the changing environment. However, this is likely to differ between large firms and SMEs, since larger firms tend to have more sophisticated, rigid and complex systems (Ashbaugh-Skaife et al., 2007). More complex, rigid AISs are harder to change quickly to a changing environment, compared to the less complex AISs of SMEs. Therefore, it is expected that the relation between growth and material weakness in internal control is weaker for SMEs than for large firms. Therefore, it can be argued that the internal control in SMEs is already weaker so that growth does not have an extra influence on this.
In conclusion, SMEs are less able to attract capable personnel in times of growth and are more likely to neglect internal control standards when growing material weaknesses in internal control, and that SMEs are more likely to have less skilled personnel. Therefore it is expected that the relation between growth and material weakness in internal control is stronger for SMEs than for large firms and that SMEs are more likely to have a lower internal control score. On the other hand, growth could lead to opportunities for small firms to increase their internal control and AISs of SMEs can adapt to growth more easily than large firms. Therefore, it is expected that the relation between growth and material weakness in internal control is weaker for SMEs than for large firms and that they are less likely to have a lower internal control score. The following hypotheses are therefore proposed:
Hypothesis 2a: The relationship between growth and material weakness in internal control is stronger/weaker for SMEs than for large firms.
Hypothesis 2b: The relationship between growth and internal control score is stronger/weaker for SMEs than for large firms.
Figure 1. Summary of hypotheses
3 Research design 3.1 Measurement
In order to examine these hypotheses, the measures for the dependent variables as well as the independent variables need to be determined. In this chapter, all variables are defined and explained.
3.1.1 Internal control quality
Prior research used the mandatory reporting under SOX section 404 and 302 to determine whether a firm had material weaknesses in internal control (Doyle et al., 2007; Ashbaugh-Skaife et al., 2007). Access to this information for listed firms is easy to gain, since under SOX listed firms are required to establish, maintain and evaluate
Growth Material weakness in internal control Size Hypothesis 2a Internal control score Hypothesis 2b
internal control effectiveness and to report on this in the quarterly and annual statements (Hoitash et al., 2012). Therefore, information about the internal control is publicly known. Because non-listed firms are not subject to SOX and therefore not required to report on the internal control in their annual reports, a different approach is needed in order to examine the quality of the internal control. To do so, a questionnaire is
developed. In the questionnaire the internal control of an organization is scored for each main process, whereas category ‘red’ can be seen as a material weakness (see appendix A). Using this questionnaire, internal control quality can be measured in two ways: (1) it can be determined whether a firm faces a material weakness; (2) an overall internal control score can be determined about the quality of internal control.
No prior research used a questionnaire to assess the internal control quality as they could use the publications under SOX. However, the questionnaire is thoroughly developed along with experts in the field of internal control (as explained later), and is therefore considered a good measure of internal control quality. The questionnaire is based on an EY checklist for interim procedures. The interim procedures are the procedures prior to the year-end auditing of the financial statements during which the quality of internal control is extensively tested. EY uses this checklist for a specific type of non-profit firms, because it provides extra insight on the internal controls. However, it is not widely implemented since every manager is allowed to report to management on their internal control as they prefer. Because these firms are non-listed, the report on the quality of internal control does not have to be published, therefore there is no standard format. However, the checklist used by EY for a type of non-profit firms, can be applied to all types of firms for giving insight on the internal control, when
additional processes are added. As a result, this checklist is chosen as a base for the questionnaire. The advantage of using this design is that most EY managers are familiar with the format and therefore understand it.
The questionnaire uses a likert scale from one to four. The main processes of the most common types of organizations are listed (Romney & Steinbart, 2012). Each process is scored from 1 for category ‘red’ to 4 for category ‘green’. In the end an average score for the internal control of the firm (IC_SCORE) is calculated. This is used as a score for the quality of the internal control of the firm. The category ‘red’ is
one or more processes, this firm is considered to have a material weakness in internal control (ICMW). So, the internal control quality of a firm is expected to be low when: (1) the firm has a material weakness in internal control (ICMW); and (2) the firm has a low internal control score (IC_SCORE).
Before sending the questionnaire to the managers of the audit teams of the selected firms, it is tested extensively. First, among peers and later among people higher in EY, to be sure that the questionnaire provides a good measure to determine the internal control quality. The feedback received contained for example that it was not clear what was meant with ‘business segments’. This is why I included a definition of this concept in the questionnaire. There were no issues regarding the meaning of the internal control table and it was noted that the table provided a clear and complete overview of the internal control of a firm. E).
The questionnaire is conducted among the engagement managers of the audit team of the selected firms. The engagement manager leads the audit for the specific firm, mostly for multiple years and is therefore an expert with regard to firm-specific knowledge. Besides, the engagement manager is always someone in the function of ‘manager’ or ‘senior manager’. This means that they have finished their post-master program and have the title CPA. This implies that they can be seen as experts in the field of audit. E).
The questionnaires are communicated to the engagement managers through email. To higher the response rate a personal email is send to all managers, requesting to fill out the questionnaire as well as to send the annual reports of 2012 and 2011.
Furthermore, strict discretion with regard to the sensitive client information is assured. This entails the assurance that none of the provided numbers could be traced back to the client, nor any client names would be mentioned in the research. The final response rate is 73%.
3.1.2 Independent variables
The most important independent variable used in this research is growth (GROWTH). In accordance with prior research, growth is measured by average growth rate in sales over a period of 3 years (Ashbaugh-Skaife, 2007). As is explained in the hypotheses development, it is expected that there is a positive relationship between
growth and material weakness in internal control, while a negative relationship is expected between growth and internal control score.
Another important independent variable is size. According to the definition in the European law regarding large, medium, small and micro organizations, size should be measured as a combination of number of employees, annual turnover and total assets, as explained earlier. However, to simplify the model, size is measured as one of these conditions, TOTAL_ASSETS, which is in line with prior research (Hoitash et al, 2012). For the first hypotheses, size is used as a control variable as prior literature suggests that the larger the firm, the stronger the internal control due to the availability of enough personnel and resources (Doyle et al, 2007; Ashbaugh-Skaife et al., 2007; Chernobai & Yasuda, 2013). Therefore it is expected that larger firms are less likely to have a material weakness in internal control and more likely to have a high internal control score. For the second hypotheses, it is examined whether size can be seen as a moderator on the relationship between growth and the quality of internal control. As is described in the hypotheses development, it is expected that this is the case, however it is not clear in what way.
Furthermore, various control variables are used. First, the number of years since the foundation of the firm (FIRM_AGE) is included since it is expected that the younger the firm the less time they had to evaluate and improve the internal control, while older firms the more time they had to remove weaknesses from there procedures. Besides, prior research has shown that listed firms that are younger are more likely to have a material weakness in internal control (Doyle et al., 2007; Chernobai & Yasuda, 2013). Therefore, it is expected that there is a negative relation between material weakness in internal control and firm age and a positive relation between internal control score and firm age.
Second, it is assumed that firms with greater complexity are more likely to face problems in their internal control (Ashbaugh-Skaife, 2007). Complexity is measured through number of business segments (#SEGMENTS) as is expected that the more segments a firm has, the more it operates in different industries or markets, the more complex firm operations become. The more complicated the firm operations, the harder it is to implement effective internal controls. This is also in compliance with the findings of prior research for listed firms (Doyle et al., 2007; Ashbaugh-Skaife et al.,
2007). Therefore, a positive relation is expected between material weakness in internal control and the number of segments a firm has and a negative relation between number of segments and internal control score. The definition of business segments used is derived from IAS 14 and is as follows: “A business segment is a distinguishable component of an enterprise that is engaged in providing an individual product or service or a group of related products or services and that is subject to risks and returns that are different from those of other business segments. A business segment is seen as a distinct segment if it generates more than 5% of the total net sales of the organization.” (European Union, 2003).
Third, it is included whether a firm booked a loss (%LOSS) in the years 2010-2012, because it is expected that firms that face losses are more likely to underinvest in their internal control. Besides, it could be that firms with losses have to fire personnel, leading to staffing issues, which influences the internal control in a negative way. Prior literature finds evidence for this with regard to listed firms (Doyle et al., 2007; Ashbaugh-Skaife et al., 2007) Therefore, it is expected that the more losses a firm has, the more likely they face a material weakness in internal control and the lower internal control score they have.
Finally, it is included whether firms faced a reorganization, restructuring, merger or acquisition (M&A_RESTRUCTURING) during the years 2010-2012. It is expected that reorganizations and restructurings often cause personnel problems in terms of segregation of duties, inadequate staff and supervision problems. This has a negative impact on the internal control. Furthermore, firms that participate in a merger or acquisition face challenges to combine both internal control systems into one integrated system. This causes the system to be weaker. Prior literature has shown that this is the case for listed firms (Doyle et al., 2007; Ashbaugh-Skaife et al., 2007). Therefore it is expected that firms that engage in mergers, acquisitions, reorganizations or restructurings are more likely to face a face a material weakness in internal control and are less likely to have a high internal control score.
3.1.3 Empirical model
The following logistic and linear regression models are used in order to assess the extent to which growth and the control variables are associated with internal control quality, using two proxies for internal control quality: (1) material weaknesses in internal control; and (2) the overall internal control score:
ICMW or IC_SCORE =
+ + _ + # + % +
& _ + _ + ,
where ICWM is coded one for firms with a material weakness in internal control, and zero for control firms and IC_SCORE is the overall internal control score for the firm with a value between 1 and 4. Furthermore, hypotheses 2a and 2b examine the moderating effect of size on the relation between growth and internal control quality. The following logistic and linear regression models are used to examine this moderating effect:
ICMW or IC_SCORE =
+ + _ + ∗ _ +
_ + # + % +
& _ + ,
where ICWM is coded one for firms with a material weakness in internal control, and zero for control firms and IC_SCORE is the overall internal control score for the firm with a value between 1 and 4. CGROWTH and CSIZE_TA are the centered variables of GROWTH and TOTAL_ASSETS, where GROWTH and TOTAL_ASSETS are defined in table 1. CGROWTH*CSIZE_TA is the product term of CGROWTH and CSIZE_TA.
Table 1.
Variable definitions
Variables Definitions and data source
IC_SCORE Average score on the internal control of all processes. Source: hand-collected through questionnaires. ICMW Dummy-variable. Equal to one if the firm has a
material weakness (i.e. has at least one process marked ‘red’ for at least one business segment), and zero otherwise. Source: hand-collected through questionnaires.
GROWTH Average growth rate in sales from 2010 to 2012. Source: hand-collected from annual reports TOTAL_ASSETS The log of the number of total assets on the balance
sheet 2012. Source: hand-collected from annual reports
FIRM_AGE The log of the number of years since the foundation of the firm. Source: hand-collected from company.info #SEGMENTS Number of business segments in 2012. Source:
hand-collected through questionnaires.
%LOSS Proportion of the years 2010, 2011 and 2012 that a firm reports negative earnings. Source: hand-collected from annual reports.
M&A_RESTRUCTURING Dummy-variable. Equal to one if a firm is involved in a merger, acquisition, restructuring or reorganization from 2010 to 2012, and zero otherwise. Source: hand-collected through questionnaires.
3.2 Sample
The sample consists of clients of Ernst&Young (EY) as I only have access to the client list and personnel list of EY. To determine the sample, a list of all EY clients is conducted. According to both Dutch and European law, an auditor needs to test the effectiveness of internal control and report to management on this only for all entities that are legally obliged to an audit (NBA, 2014; European Union, 2006). Therefore, all organizations that are not legally obliged to an audit are removed from the sample. Next, the remaining non-profit organizations are removed, since they are outside the scope of the research. Finally, the listed firms are excluded, since this research only focuses on non-listed firms. The sample now contains all non-listed profit firms that are legally
obliged to an audit. From this list, 160 firms are selected randomly. Then, some adjustments are made to the sample. First, when the engagement manager of a firm had already two questionnaires assigned to him, a potential next one is deleted from the sample and replaced by a new one, to higher the response rate. Second, as the percentage of large listed firms is quite low, I included manually large firms, to be sure that at least some large firms were in the sample. This was done by manually going through the next 50 firms of the list and verifying whether they met the abovementioned criteria for a large firm or not. If so, the firm was added to the sample. This led to an additional 5 firms, so in total 165 questionnaires are send to engagement managers of EY.
The final response rate is 73%. Unfortunately not all engagement managers were able to provide the required information of the selected firms. This is due to for example, bankruptcy, change of audit firm or the fact that there was only performed substantive audit procedures, which means that the internal control is not tested. The final sample that is included in this research is therefore 82 firms of which 19 large firms and 63 SMEs.
Table 2. Sample derivation
Description N
All clients EY NL Assurance 2,699
Removal of non-legally obliged audits (1,001)
Removal of remaining non-profit (452)
Removal of listed companies (104)
Sample applicable to research 1,142
Random sample selection 160
Additional large firms 5
Firms selected for research 165
Total respondents 120
Firms not suitable 38
4 Empirical results and discussions
4.1 Descriptive statistics and univariate tests
The number of observation per industry in the sample is summarized in table 34. The wholesale trade industry accounts for the largest part of the sample, followed by 19.5% for the manufacturing industry. Because of the limited observations in this research, no industry-specific control variables are included. However, the distribution is used to determine the growth of the sample firms relative to their industry in the robustness analysis.
Table 3. Industry distribution
Frequency Percent
Agriculture, forestry and fishing 3 3.7
Construction 7 8.5
Finance, insurance and real estate 13 15.9
Manufacturing 16 19.5
Wholesale trade 25 30.5
Retail trade 2 2.4
Services 10 12.2
Transportation, communication, electric, gas and sanitary services
6 7.3
Total 82 100.0
Information concerning the internal control scores is shown in table 4. It appears that none of the processes applies to all processes. This is because different types of firms have different main processes. The Production Process is the less represented process, while the Cash Collections applies to almost all firms. Thirteen firms had a process that was not specified separately in the questionnaire but was important to include in the total internal control score. Furthermore, the table shows that all processes have been scored at least once a ‘one’ and once a ‘four’, while the lowest overall score for a firm is 1.2. Besides, on average, the IT process is scored the lowest followed by the Risk Management process, while the Personnel Process received the highest average score, meaning that the internal control for this process is on average
4
Note: Normally when conducting questionnaires, a number of preliminary tests are required, such as a response analysis, a reliability test and a factor analysis. However these tests do not apply to this setting, as not all questionnaires measure a single construct. All applied questionnaires measure the internal control, however for different firms and therefore a different outcome is expected per questionnaire.
the strongest for the sample firms. The total combined score indicates that on average all firms in the sample have an internal control score of 3.160 which means that the internal control system almost meets the requirements and that there are only some points for improvements. So overall for the sample firms, the internal control system is quite good. Notable also is that of the 82 firms, 12 have been indicated as having a material weakness in internal control. The average internal control score of these firms is 2.475, while the average of the remaining firm is 3.277 as is shown in panel B.
Table 4. Summary internal control scores and material weaknesses
N Mean Std. Deviation
Q1 Median Q3 Min Max
Panel A: internal control score Planning&Control 73 2.990 0.825 2.00 3.00 4.00 1.0 4.0 Risk Management 62 2.850 0.786 3.00 3.00 4.00 1.0 4.0 Purchasing 78 3.370 0.705 3.00 3.00 4.00 1.0 4.0 Sales 78 3.240 0.742 3.00 3.00 4.00 1.0 4.0 Production 34 3.180 0.758 3.00 3.00 4.00 1.0 4.0 Personnel 78 3.450 0.658 3.00 4.00 4.00 1.0 4.0 Cost Accounting 54 3.000 0.801 2.75 3.00 4.00 1.0 4.0 Cash Collections 81 3.470 0.654 3.00 4.00 4.00 1.0 4.0 IT 77 2.690 0.782 2.00 3.00 3.00 1.0 4.0 Other 13 3.000 0.913 2.50 3.00 4.00 1.0 4.0 Total (combined) 82 3.160 0.530 2.88 3.15 3.56 1.2 4.0 N Mean Std. Deviation
Q1 Median Q3 Min Max
Panel B: material weakness (ICMW)
ICMW Sample 12 2.475 0.602 2.00 2.48 2.97 1.2 3.3
Control Sample 70 3.277 0.420 3.00 3.24 3.63 2.3 4.0
In Panel A of Table 5 on page 27 the descriptive statistics of all dependent and independent variables are presented. Besides, it provides the univariate tests that assess the comparison between the firms that report a material weakness in internal control (ICMW sample) and the firms that do not (control sample).
First, the table shows that 15% of the firms have a material weakness in internal control and that the IC_SCORE for the ICMW sample is significantly lower than the score for the control sample. This is in line with the expectations. Furthermore, the table shows a strong significant difference between the two samples for M&A_RESTRUCTURING, indicating that firms that were involved in a merger, acquisition, restructuring or reorganization are more likely to have a material weakness in internal control. This is consistent with the expectations. Moreover, the ICMW sample has a significantly higher mean for the %LOSS than the control sample, which is also consistent with the predictions. The remaining independent variables do not show a significant difference between the ICMW sample and the control sample.
Panel B of the same table provides the correlation coefficients, where the upper right-hand portion of the table presents Pearson correlations and the lower left-right-handed portion of the table provides the Spearman rank-order correlations. The Pearson correlations are discussed here, but it should be noticed that the patterns of the two different correlations are quite similar. The strongest correlation is the correlation between GROWTH and %LOSS and this is a significant negative correlation of 0.338. This makes sense, as it is not very likely that firms that grow also report a loss, however it is not the same as GROWTH is measured through sales, while %LOSS is measured through profit. The second largest correlation is a significant positive correlation between #SEGMENTS and TOTAL_ASSETS, indicating that larger firms are more likely to have multiple segments and are therefore more complex. The correlations between IC_SCORE and the independent variables show that there is a significant positive correlation between IC_SCORE and GROWTH of 0.228, indicating that firms that grow are more likely to have a higher internal control score. This contradicts with the expectations. Furthermore, there is a significant negative correlation between IC_SCORE and %LOSS, indicating that the more loss a company has, the lower the internal control score, which is consistent with the expectations. IC_SCORE does not correlate significantly with the
other variables. However, for FIRM_AGE, TOTAL_ASSETS and M&A_RESTRUCTURING the predicted signs are in line with the signs in the correlation matrix. For #SEGMENTS there is hardly any correlation found with IC_SCORE. In de next section, more formal tests of the hypotheses are conducted.
Table 5
Descriptive Statistics on all variables
Mean Std. Deviation Q1 Median Q2 Min Max
Panel A: Distributional properties Dependent variables ICMW 0.150 0.357 ─ ─ ─ 0 1 IC_SCORE ICMW Sample 2.475*** 0.602 2.00 2.48 2.97 1.200 3.300 Control Sample 3.277 0.420 3.00 3.24 3.63 2.300 4.000 Independent variables GROWTH MW Sample ─0.013* 0.193 ─0.076 0.027 0.096 ─0.568 0.209 Control Sample 0.061 0.379 ─0.046 0.048 0.124 ─2.497 0.809 FIRM_AGE MW Sample 1.310 0.413 0.985 1.308 1.724 0.602 1.944 Control Sample 1.278 0.379 0.988 1.243 1.544 0.477 2.008 #SEGMENTS MW Sample 2.333 1.669 1.000 2.000 3.000 1 6 Control Sample 1.771 1.466 1.000 1.000 2.000 1 7 %LOSS MW Sample 0.472* 0.413 0.000 0.500 0.917 ─ ─ Control Sample 0.257 0.341 0.000 0.000 0.333 ─ ─ TOTAL_ASSETS MW Sample 7.758 0.566 7.396 7.680 7.975 7.053 9.093 Control Sample 7.495 0.669 7.010 7.455 7.896 6.521 10.251 M&A_RESTRUCTURING MW Sample 0.830*** ─ ─ ─ ─ ─ ─ Control Sample 0.390 ─ ─ ─ ─ ─ ─ A B C D E F G Panel B: Correlations ¹ IC_SCORE A ─ 0.228** 0.185 0.002 ─0.231** 0.081 ─0.120 GROWTH B 0.189 ─ ─0.002 ─0.010 ─0.338*** 0.092 ─0.008 FIRM_AGE C 0.149 0.006 ─ 0.202 ─0.079 0.021 0.044 #SEGMENTS D 0.018 ─0.020 0.200 ─ ─0.081 0.297*** 0.171 %LOSS E ─0.236** ─0.389*** ─0.067 ─0.112 ─ ─0.049 0.160 TOTAL_ASSETS F 0.077 0.074 0.019 0.333*** ─0.034 ─ 0.247** M&A_RESTRUCTURING G ─0.104 ─0.080 0.049 0.201 0.154 0.253** ─
***,**,* Indicates significance at the 0.01, 0.05 and 0.10 level or better, based on t-statistic. There are 12 firms in the ICMW sample and 70 firms in the Control Sample. All continuous variables have been winsorized at the 5 and 95 percentile values.
4.2 Multivariate regression
Before conducting the regression analyses some assumption tests were conducted (see appendix B). Based on these tests it can be assumed that there is no multicollinearity or heteroscedasticity in the model.
Table 6 shows the results for the logistic regression to examine the relation between the independent variables and the dummy variable ICMW. The Wald of 31.861 (significant at 0.01) shows that the constant of the model is not equal to 0. Furthermore, the overall model has a χ² of 14.471 and is significant at 0.05, meaning that the probability of finding an effect if there is in fact no effect of the independent variables, taken together, on the material weakness in internal control, is very low. The Nagelkerke R² is 0.286, indicating that 28.6% of the variance in the material weakness in internal control is explained by the independent variables included in the model.
As presented in the table, the coefficients of the variables #SEGMENTS and %LOSS are in the predicted direction, while GROWTH, FIRM_AGE and TOTAL_ASSETS have a coefficient in the opposite direction. However, these coefficients are not significant. The only significant coefficient found in the predicted direction is for M&A_RESTRUCTURING, indicating that firms that were involved in a merger, acquisition, restructuring or reorganization in 2010─2012 are more likely to report a material weakness in internal control. This is in line with the predictions.
Hypothesis 1a proposed that firms that grow are more likely to report a material weakness in internal control. This can however, not be concluded from this model. A possible explanation for this insignificant relation could be that GROWTH is measured through the average growth rate in 2010─2012. This means that also firms that grew a little are included in GROWTH, even though growing a little does not directly lead to a material weakness in internal control. Hoitash et al. (2012) and Doyle et al. (2007) use therefore a different measure of growth in their studies. An additional test on this will be provided in the robustness paragraph. Another explanation could be the limited sample size of this study. A larger sample size provides more predictability, leading to a higher possibility of proving significant relations. However, it could also be that there is indeed no relationship between growth and internal control quality for non-listed firms
Table 6. Logistic regression of the determinants of material weakness in internal control (ICMW)
ICMW= + + _ + # + % +
& _ + _ + ,
Dependent variable = ICMW Independent variables Predicted sign Logit estimate
(χ²) INTERCEPT ─8.825 (2.535) GROWTH + ─3.584 (1.409) FIRM_AGE ─ 0.156 (0.028) #SEGMENTS + 0.177 (0.277) %LOSS + 1.211 (1.424) TOTAL_ASSETS ─ 0.676 (0.903) M&A_RESTRUCTURING + 1.752** (4.260) Wald, χ² 31.861*** Model, χ² 14.471** Nagelkerke R² 0.286 Sample size 82
ICMW is coded one for firms that have a material weakness in internal control (n=12) and zero otherwise (n=70). ***,**,* indicates respectively significance at the 0.01, 0.05 and 0.10 level or better.
Table 7 shows the results for the linear regression model to examine the relation between the independent variables and the IC_SCORE. The model has an R² of 0.109 which means that 10.9% of the variance in the internal control score is explained by the model. However, because of the small sample size, the adjusted R² is much lower with a predictive value of 3.8%. Overall the model has an F-value of 1.536 which is not significant.
The table shows that, except for GROWTH, the coefficients of all independent variables have the predicted sign, however the only significant coefficient found is for FIRM_AGE. The coefficient of 0.278 indicates that if the firm age increases with 1 year, on average the internal control score for that firm increases with 0.278. This is in line with the formulated expectations.
Hypothesis 1b stated that firms that grow rapidly are more likely to have lower internal control quality and therefore a lower internal control score. This hypothesis is not supported by this model. The possible explanations of the lack of a significant relation are the same as for the previous model. It could be that the measure for growth is not accurate or that the sample size is too small. But, it cannot be asserted that there is no relationship between growth and internal control quality for non-listed firms.
Table 7. Linear regression of the determinants of internal control score (IC_SCORE)
IC_SCORE = + + _ + # + % + & _ + _ + ,
Dependent variable = IC_SCORE Independent variables Predicted sign Coefficients
(t-statistic) INTERCEPT 2.030** (2.397) GROWTH ─ 0.508 (1.078) FIRM_AGE + 0.278* (1.722) #SEGMENTS ─ ─0.016 (0.385) %LOSS ─ ─0.214 (1.217) TOTAL_ASSETS + 0.116 (1.042) M&A_RESTRUCTURING ─ ─0.078 (0.641) R² 0.109 Adjusted R² 0.038 F-value 1.536 Sample size 82
***,**,* indicates respectively significance at the 0.01, 0.05 and 0.10 level or better.
Table 8 provides the model for hypotheses 2a, a logistic regression to examine whether size has a moderating effect on the relation between growth and material weakness in internal control. The GROWTH and TOTAL_ASSETS variables are transformed to be the centered variables (CGROWTH and CSIZE_TA) to counteract the problem of multicollinearity between the two variables and the product term
(CGROWTH*CSIZE_TA). There is an ongoing debate about whether it is needed to center variables when examining a moderator effect. In these analyses it is chosen to do so, in accordance with the paper of Dawson (2014). Note however, that for this dataset the results are almost identical to what the results would have been if the variables had not been centered (see robustness tests for results with untransformed variables).
The model in table 8 has a χ² of 19.232 which is significant at 0.01. The χ² has increased comparing to the first model (in table 5), indicating that the product term, CGROWTH*CSIZE_TA adds to the model. Besides, the Nagelkerke R² has increased from 0.286 in the previous model to 0.386 which means that by adding the product term a larger part of the variance in material weakness in internal control is explained by the model.
The directions of the coefficients of the different independent variables are similar to the directions in the first model. The additional independent variable CGROWTH*CSIZE_TA has a negative direction, indication that size is lowering the effect of growth on material weakness in internal control. However, this conclusion cannot be drawn, because the coefficient is not significant. The only significant coefficient found is, again, for M&A_RESTRUCTURING.
Hypothesis 2a proposed that the relation between growth and material weakness in internal control is stronger/weaker for SMEs compared to large firms. Based on this research the 0─hypothesis cannot be rejected, meaning that there is not enough evidence that size is a moderator for the relation between growth and material weakness in internal control. The lack of evidence for this hypothesis could be the result of the small sample size as the improved χ² and Nagelkerke R² show that the moderating factor adds to the model. Another cause could be that TOTAL_ASSETS is not an accurate measure for growth, since European law requires looking at total assets, number of employees and annual turnover in order to determine size (European Commission, 2005). Whether the results change with different measures of size is examined through additional analyses in the robustness tests paragraph.
Table 8. Logistic regression on the moderating effect of size on the relation between growth and the
internal control score.
ICMW= + + _ + ∗ _ +
_ + # + % + & _ + ,
Dependent variable = ICMW Independent variables Predicted sign Logit estimate
(χ²) INTERCEPT ─4.880*** (7.473) CGROWTH + ─3.949 (1.116) CSIZE_TA ─ 0.848 (0.945) CGROWTH*CSIZE_TA +/─ ─2.682 (0.165) FIRM_AGE ─ 0.203 (0.036) #SEGMENTS + 0.137 (0.335) %LOSS + 1.355 (1.409) M&A_RESTRUCTURING + 2.452** (7.473) Wald, χ² 32.556*** Model, χ² 19.232*** Nagelkerke R² 0.386 Sample size 82
***,**,* indicates respectively significance at the 0.01, 0.05 and 0.10 level or better. CGROWTH and CSIZE_TA are the centered variables of GROWTH and TOTAL_ASSETS, where GROWTH and TOTAL_ASSETS are defined in table 1. CGROWTH*CSIZE_TA is the product term of CGROWTH and CSIZE_TA.
The model for examining hypothesis 2b is presented in table 9. As this model also examines the moderating effect of size, only this time on the relation between growth and internal control quality, the centered variables are used.
The model has an R² 15.4% and an adjusted R² of 7.3% which are both higher than the original model (table 7), indicating that the product term contributes to the model. The model has an F-value of 1.895 which is significant at the 0.10 level, while the previous model, without the moderating variable, was not significant. These both imply that the moderating variable improves the overall model.
The coefficients of all independent variables have the same sign as in the first model. The additional independent variable CGROWTH*CSIZE_TA has a positive direction, indication that size is strengthening the effect of growth on internal control quality. However, this conclusion cannot be drawn, because the coefficient is not significant. The only significant coefficient found is, as in the first model, for FIRM_AGE.
The proposed hypothesis 2b was as follows: the relation between growth and internal control quality is stronger/weaker for SMEs compared to large firms. This hypothesis can however not be supported by this model. The lack of a significant moderating effect of size can have multiple causes. First, it could be that the sample size is too small to show significant results. According to the R² and the F-value, the model is improved by adding the product term, so it has some effect. Besides, figure 2 shows that the relation between growth and internal control score is stronger for large firms than for SMEs. This implies that size has indeed a moderating effect on the relation between growth and internal control score, however this is not significant for this sample size. In a larger sample this could lead to a significant effect. Second, as mentioned above, it could be that the measure of size is not accurate. Additional tests will be performed to examine this.
Table 9. Linear regression on the moderating effect of size on the relation between growth and the
internal control score.
IC_SCORE= + + _ + ∗ _ +
_ + # + % + & _ + ,
Dependent variable = IC_SCORE Independent variables Predicted sign Coefficients
(t-statistic) INTERCEPT 2.908*** (13.581) CGROWTH ─ 0.441 (0.995) CSIZE_TA + 0.075 (0.745) CGROWTH*CSIZE_TA +/─ 1.125 (1.495) FIRM_AGE + 0.309** (2.046) #SEGMENTS ─ ─0.017 (0.462) %LOSS ─ ─0.197 (1.245) M&A_RESTRUCTURING ─ ─0.088 (0.785) R² 0.154 Adjusted R² 0.073 F-value 1.895* Sample size 82
***,**,* indicates respectively significance at the 0.01, 0.05 and 0.10 level or better. CGROWTH and CSIZE_TA are the centered variables of GROWTH and TOTAL_ASSETS, where GROWTH and TOTAL_ASSETS are defined in table 1. CGROWTH*CSIZE_TA is the product term of CGROWTH and CSIZE_TA.
Figure 2. Graph of difference between SMEs and large firms
SIZE_DUMMY is a dummy variable equal to zero for firms that are classified as SME according to the European rules on number of employees, balance sheet total and annual turnover and equal to one otherwise.
5 Robustness tests
The validity of the results depends on the quality of the measures used for the most important variables, internal control quality, growth and size. Since the research already includes two ways of measuring internal control quality, this section only focusses on the measures of growth and size. Furthermore, hypotheses 2a and 2b are tested using untransformed variables and an alternative method to test the moderating effect is introduced.
The first robustness test that is conducted relates to hypothesis 1, specifically to the measure for growth. In the main analysis, growth is defined as the average growth rate in sales from 2010 to 2012 in accordance with the paper of Ashbaugh-Skaife et al. (2007). However, using this definition of growth did not lead to any significant results.
As stated earlier, this may be due to the focus on all growth, even though growing a little does not immediately affect the internal control quality. Therefore, the robustness test focuses on more rapid growth in accordance with the papers of Hoitash et al. (2012) and Doyle et al. (2007). Three models are introduced to measure rapid growth: (1) growth is a dummy variable that is equal to 1 if the firms year-over-year sales growth is above the median of sales growth for their industry, and zero otherwise; (2) growth is a dummy variable that is equal to 1 if the firms year-over-year sales growth is in the highest quintile of sales growth for their industry, and zero otherwise; and (3) growth is a dummy variable that is equal to 1 if the firms year-over-year sales growth is in the highest five percent of sales growth for their industry, and zero otherwise. Information on this is retrieved from StatLine, the online database of the CBS (Central Bureau for Statistics). This is a Dutch governmental organization that provides all kinds of information on Dutch organizations.
Table 10 shows the results of the logistic regression models using this new definition of growth. For all three models, GROWTH does not show a significant coefficient, indicating that there is not enough evidence to conclude that there is a relationship between growth and material weakness in internal control. In other words, the results of the main analyses to not change when using a different measure for GROWTH. Model 3 even shows a χ². This is probably due to the fact that there are only four firms in the sample that have sales growth in the highest five percent of their industry, so the sample is too small to show significant results. However, the model shows, in contrast to the main analyses, a significant coefficient for %LOSS in model 1 and 3. This means that firms that book a loss are more likely to face material weaknesses in internal control, which is in line with the initial expectations.
Table 11 shows the results of the linear regression models using the new definition of growth. The table shows also no significant relation for growth, indicating that it cannot be assumed that there is a relation between growth and internal control score. For model 1 and 3 there is however now a small negative significant coefficient for %LOSS, indicating that firms that face losses are more likely to have a lower internal control score, which matches the initial expectations.
To conclude, introducing a different measure of growth does not change the results from the main analyses. Therefore there is still not enough evidence to assume that there
is a relation between growth and material weakness in internal control or growth and internal control quality.
Table 10. Logistic regression of the determinants of material weakness in internal control (ICMW) using
a different measure for GROWTH
ICMW = + _ + _ + # +
% + & _ + _ + + ,
Dependent variable = ICMW
Model 1 Model 2 Model 3 Independent variables Predicted
sign
Logit estimate
(χ²) Logit estimate(χ²) Logit estimate(χ²)
INTERCEPT ̶ 13.994** (4.378) ̶ 9.320* (2.914) ̶ 9.902* (2.813) GROWTH_DUMMY + 0.660 (0.661) ̶ 0.857 (0.479) ̶ 19.032 (0.000) FIRM_AGE ─ 0.349 (0.124) 0.244 (0.068) ̶ 0.164 (0.027) #SEGMENTS + 0.122 (0.298) 0.139 (0.405) 0.145 (0.406) %LOSS + 1.835* (3.085) 1.549 (2.696) 2.045** (4.089) TOTAL_ASSETS ─ 1.110 (1.859) 0.695 (0.972) 0.811 (1.181) M&A_RESTRUCTURING + 2.527** (5.165) 1.848** (4.769) 1.855** (4.547) Wald, χ² 32.556 31.861 29.691 Model, χ² 17.762*** 13.496** 15.464** Nagelkerke R² 0.359 0.269 0.327 Sample size 82 82 82
ICMW is coded one for firms that have a material weakness in internal control (n=12) and zero otherwise (n=70). ***,**,* indicates respectively significance at the 0.01, 0.05 and 0.10 level or better.
Model 1 ̶ Model where GROWTH_DUMMY is coded one for firms that have a year-over-year sales growth within the highest 50% of sales growth for their industry.
Model 2 ̶ Model where GROWTH_DUMMY is coded one for firms that have a year-over-year sales growth within the highest quintile of sales growth for their industry.
Model 3 ̶ Model where GROWTH_DUMMY is coded one for firms that have a year-over-year sales growth within the highest 5% of sales growth for their industry.
