Effect of Reporting Frequency on Reporting Quality

Effect of Reporting Frequency on Reporting Quality

Name: Ari Tuzi

Student number: 12966452 Date: June 7, 2021

Word count: 11,593

MSc Accountancy & Control Specialization: Accountancy

Faculty of Economics and Business, University of Amsterdam Name of supervisor: Mr M. (Máté) Széles MSc

Statement of Originality

This document is written by student Ari Tuzi who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

This research investigates the effect of financial reporting frequency on financial reporting quality. Several studies give different indications of how the financial reporting frequency affects the financial reporting quality. Also, public figures such as Donald Trump pledge for certain financial reporting frequencies, which makes this study relevant. The last change in financial reporting frequency was done in 1970 by the SEC. Therefore, I use Benford’s Law to measure the number of misstatements within financial statements and compare the number of misstatements before and after the change. The main conclusion of this study is that reporting more frequently improves the financial reporting quality. For this reason, the current financial reporting frequency is better than the previous frequency. This result gives reason for investors to demand more financial reports within the same timeframe as

transparency increases. Also, the result contributes to the literature by showing that short- termism is not caused by reporting quarterly. While it also contributes to the practice by incentivizing firms to design their management accounting systems to be able to report more frequently, maybe even heading towards continuous reporting for internal reporting.

Table of contents

1 Introduction ... 6

2 Literature review ... 8

2.1 Agency theory ... 8

2.2 Information asymmetry ... 8

2.3 Short-term focus ... 9

2.4 Earnings management ... 10

2.5 Workload auditor ... 11

2.6 Audit quality ... 12

2.7 Investors’ perspective ... 13

2.8 Financial reporting frequency ... 14

2.9 Financial reporting quality ... 15

2.10 Continuous reporting ... 15

2.11 Hypotheses development ... 16

3 Research method ... 19

3.1 Sample selection ... 19

3.2 Dependent variable – Financial reporting quality ... 20

3.2.1 Benford’s Law ... 20

3.3 Independent variable – Financial reporting frequency ... 22

3.3.1 Event study ... 22

3.4 Control variables ... 22

3.5 Regression model ... 25

3.6 Reliability and validity of proxies ... 25

3.7 Statistical method ... 26

4 Results ... 27

4.1 Descriptive statistics ... 27

4.2 Correlation matrix ... 30

4.3 Ordinary Least Squares regression ... 33

4.3.1 OLS regression MAD statistic ... 33

4.3.2 OLS regression KS statistic ... 34

5 Conclusion ... 35 References ... 37

1 Introduction

Since 1970, the Securities and Exchange Commission (SEC) mandates all firms in the United States (U.S.) to report quarterly instead of semi-annually (Zhou, 2018). It is, however, still unknown whether this change influences the reporting quality of the financial statements.

Trump pledged in 2018 for a return of semi-annual reporting, as he believes managers fixate too much on short-term targets to improve earnings in the quarter reports (Zhou, 2018).

Extending it to semi-annual reporting incentivizes managers to fixate on longer-term goals (Zhou, 2018). However, the assumption is that this delays the fixation by three months, which is still short-term (Kim, 2020).

The reporting process is essential for giving stakeholders information about the firm’s performance and situation (Nichols & Wahlen, 2004). It is, therefore, more informative if firms report their information frequently at high-quality standards. However, for some firms, it is costly and difficult to report frequently (Fiechter et al., 2018). For example, audits on these reports improve the quality but also result in more costs. Therefore, I believe this is a trade-off between quality and costs, which is why it is interesting to investigate whether the frequency of reporting matters.

With recent technological developments, the magnitude of this trade-off decreases as automation becomes more accessible for firms (Cardoni et al., 2020). These developments come through into the accounting segment, where artificial intelligence, process mining, data mining, and others help make reporting processes shorter (Wang & Kogan, 2020). One of those reporting processes is reporting financial information, or even non-financial information depending on the needs of the firms’ stakeholders. If this process runs faster and more reliable, it gives possibilities to report more frequently (Wang & Kogan, 2020).

Filzen (2015) investigates the information content of risk factor disclosures in quarterly reports, where it is evident that quarterly reports about risk factors negatively influence the share value on the stock market. Also, Filzen (2015) finds that these reported risk factors indeed lead to future negative outcomes. For this reason, I assume that the quarter reports give valid information regarding these risks, which is an indication of quality.

On the contrary, Lee (2012) states that longer and more textually complex quarterly reports are difficult for markets to follow and therefore they respond inefficiently to this information. Based on this, I assume that more information in shorter periods leads to less informativeness and thus lower reporting quality.

I formulate the following research question: What is the effect of quarterly vs. semi- annual reporting on reporting quality? By finding the difference in reporting quality between quarterly and semi-annual reporting, stakeholders know whether it is effective for firms to report more frequently. This leads to more demand for transparency from stakeholders towards firms, or more understanding of stakeholders that reporting quality still needs improvements to be able to be frequent.

From the opposing perspective, executives might not want to report frequently, because shareholders evaluate executives based on financial performance measures (Murphy & Jensen, 2011). Subsequently, evaluations are done quarterly, which incentivize executives to use earnings management for example. For this reason, I regard the agency theory as of importance.

A quantitative research method for this research finds solutions for the research question.

Furthermore, this method compares proxies between different periods, which makes it an event study. Additionally, moderating variables filter out other factors that influence the supposedly main effect.

2 Literature review

2.1 Agency theory

As the financial reporting frequency increases to improve the financial reporting quality, the costs of reporting and auditing financial statements for firms increase as well (Fiechter et al., 2018). Subsequently, executives try to lower the audit fees if they believe they need to cut down on expenses, which eventually leads to less audit quality due to less audit effort and resources (Ettredge et al., 2014). The findings of Ettredge et al. (2014) explain the positive relation between audit fee pressure and misstatements during the financial crisis of 2008.

However, if the benefits of the increased financial reporting frequency outweigh the costs, then there is a trade-off. Marshall et al. (2019) find evidence of this trade-off as announcing audited earnings results in higher abnormal returns relative to unaudited earnings. Firms prefer to have higher abnormal returns compared to audit fees. From these insights, I believe that there is a difference in interests between the executives and the capital market. As executives aim to lower the expenses to achieve targets, while shareholders prefer to have reliable financial information.

This trade-off relates to the agency theory. According to Eisenhardt (1989), the agency theory originates from the relationship between the principal and agent, where the principal usually refers to the shareholders of a firm and the agent refers to the executives of the same firm. Eisenhardt (1989) mentions that an agency problem arises under two circumstances: the desires and goals of the principal and agent conflict with each other, and it is too expensive and difficult for the principal to verify the actions of the agent. This is in line with the difference in the desires and goals of the executives and shareholders. Executives try to reach their targets to acquire more compensation, while investors prefer to have reliable information to make investing decisions.

2.2 Information asymmetry

Information asymmetry is the degree of information between managers of firms and the market (Dierkens, 1991). As Dierkens (1991) states that managers of firms have more information about the firm, which they partially provide to the market. My study relates to this because the SEC uses the increased reporting frequency to increase the amount of information given by managers, which lowers the information asymmetry. Dierkens (1991) concludes that information asymmetry exists and influences market reactions. As investors believe a high

degree of information asymmetry makes the financial reports less reliable, which indicates low financial reporting quality (Dierkens, 1991). Regarding my study, it is interesting to investigate whether less information asymmetry through increased financial reporting frequency increases financial reporting quality.

Fu et al. (2012) find evidence that when the SEC mandates firms to report quarterly, the information asymmetry and cost of equity decrease. As the authors analyze the unobservable firm characteristics before and after quarterly reporting. This indicates that the informativeness increases after quarterly reporting. Subsequently, leading to more confidence from the equity market, as the expected returns decreases. This paper captures relevant information for my study, which provides more insights into the predictions of this study.

2.3 Short-term focus

Trump reignited the discussion about the reporting frequency in 2018 (Zhou, 2018). He pledges for a return of semi-annual reporting, as he believes executives focus too heavily on short-term goals at the expense of long-term goals (Zhou, 2018). Black et al. (2021) investigate the short-term focus of executives by testing the association between CEO pay components and non-GAAP earnings disclosure. As CEO pay components are a proxy for short- versus long-term focus, while an aggressive non-GAAP disclosure indicates that the CEO targets more compensation for the period (Black et al., 2021). They find no evidence that short-term bonuses relate to aggressive non-GAAP disclosures. However, they do find that long-term compensation plans cause less aggressive non-GAAP disclosures. This shows that a long-term focus leads to less aggressive behavior from executives to depict a better picture of the firm’s financial position towards investors, which is indicative of the financial reporting quality of semi-annual reports in my study.

Another study regarding the short-term focus of executives is the paper by Benton and Cobb (2019). They examine whether U.S. corporations focus too heavily on short-term performance, undermining the long-term performance. Based on the results, they state that analyst’s earnings projections are key to short-term focusing as executives cut R&D and manipulate the pension assumption. However, when elite cohesion is present within a firm, the likelihood of R&D cuts and pension-assumption manipulation decreases (Benton & Cobb, 2019). Indicating that the board network supports executives by neglecting the short-term focus while maintaining a long-term focus.

This contradicts the findings of Black et al. (2021) as Benton and Cobb (2019) do find evidence that executives show aggressive behavior to adjust the earnings upwards. However, this difference may come from the elite cohesion at firms, which means that not every firm with short-term goals has this problem.

The paper by Fu et al. (2020) confirms this with a study that focuses on the association between financial reporting frequency and corporate innovation. According to Fu et al. (2020), corporate innovation is a proxy for short-termism, since short-term-focused executives do not invest in innovative projects where profits arrive over longer periods. They find that frequent reporting increases short-termism by executives, as they invest less in innovation. This is by the theory of Benton and Cobb (2019), as it shows that managers are short-term focused in certain situations. Therefore, this research assumes that managers are short-term focused, which may be a deciding factor for the tested relation of this research.

2.4 Earnings management

When executives are short-term focused, they tend to lower R&D expenses or manipulate the pension assumption (Benton & Cobb, 2019). The goal is to increase the earnings before the end of the period, which relates to earning management. I believe earnings management relates to this study, as the financial reporting quality may differ between reporting frequencies due to earnings management.

Schipper (1989) states that earnings management is influencing the summary of firm performance through accruals, resulting in less informativeness. This definition points towards accruals to manage earnings, while Cohen et al. (2008) describe another type of earnings management. They state that firms also use real earnings management, which are activities such as giving discounts to customers if they buy products before the end of a period or producing more to lower the fixed costs before the end of the period. The definition of Cohen et al. (2008) depicts more possibilities of how short-term-focused executives may influence the financial reporting quality.

The paper by Tang et al. (2016) focuses on real earnings management in the event of a changing reporting frequency. Tang et al. (2016) investigate the number of detected real earnings management between different financial reporting frequencies. Subsequently, they find that reporting financial information more frequently leads to a higher probability that auditors detect sales-related real earnings management. This indicates that a higher frequency of reporting results eventually leads to less real earnings management.

A contradiction compared to the paper by Benton and Cobb (2019), since Benton & Cobb (2019) state that executives use more real earnings management due to short-termism.

Although, Tang et al. (2016) focus more on the detection of real earnings management, rather than minimalizing it. While Benton and Cobb (2019) specifically explain that the incentives of executives to increase earnings before period-end.

2.5 Workload of auditors

Due to the increase in financial reporting frequency, the workload of auditors increases as well. When the workload increases, it may be difficult for auditors to allocate their time and resources which may cause a decrease in audit quality. Therefore, the difference in financial reporting quality may depend on the workload of auditors.

Heo et al. (2021) investigate whether the workload of auditors affects audit quality. They find that audits during the busy season have greater abnormal accruals and misstatements, which indicates that audit quality is worse when the workload increases. However, the authors also find that the hours in the labor mix differ during the busy season. As senior auditors spend less time on audits compared to junior auditors (Heo et al., 2021). The reason behind this may be that senior auditors spread their time over other audits, while junior auditors take over tasks from the senior auditors. This indicates that the increase in financial reporting frequency decreases the audit quality, which subsequently leads to less financial reporting quality.

López and Peter (2012) also investigate the effect of workload on audit quality but in a different setting. They use the relative concentration of companies with the same fiscal year per auditor’s client portfolio, which is a different proxy compared to the paper of Heo et al.

(2021). Nonetheless, López and Peter (2012) also find that during the busy season the audit quality decreases, and thus lowers the financial reporting quality. The paper by López and Peter (2012) is in line with the paper by Heo et al. (2012) regarding that the workload decreases the audit quality. Even though, the two papers have different proxies and settings, which confirms the presumed effect.

Another study by Persellin et al. (2019) surveys auditors on workloads to obtain their perception. As they aim to find explanations regarding why workload decreases the audit quality (Persellin et al., 2019). They find that auditors’ job satisfaction decreases when the workload is high, as this puts pressure on the auditors. Subsequently, this dissatisfaction leads to less audit quality, due to auditors being demotivated (Persellin et al., 2019). This paper is also in line with the studies by López and Peter (2012) and Heo et al. (2012).

From the papers above, I find no contradictions as the effect is evident. However, the theory itself is contradicting for my overall study. The reason is that the theory of reducing the information asymmetry, by increasing the number of financial reports, increases the financial reporting quality. However, if the number of financial reports increases, then the number of audits increases as well. Subsequently, the increase in audits causes less audit quality, which in turn negatively affects the financial reporting quality.

2.6 Audit quality

Understanding how the audit quality relates to the financial reporting quality helps to explain one of the possible causes of why there could be a difference in financial reporting quality between the reporting frequencies. Gaynor et al. (2016) test this relationship between financial reporting quality and audit quality.

They define a higher quality audit as “one that provides a higher level of assurance that the auditor obtained sufficient appropriate evidence that the financial statements faithfully represent the firm’s underlying economics” (Gaynor, 2016, p. 5). While they define financial reporting quality as higher quality financial reports that “are more complete, neutral, and free from error and provide more useful predictive or confirmatory information about the company’s underlying economic position and performance” (Gaynor, 2016, p. 14).

Gaynor et al. (2016) conclude that audit quality positively influences pre-audit financial reporting quality and financial reporting quality. An interesting insight is that it also positively affects the pre-audit financial reporting quality, which means that the audit quality indirectly influences firms to make higher quality financial reports. Therefore, a possible assumption is that decreasing the workload of auditors, by auditing semi-annually, improves the quality of quarterly reports. As the decrease in workload results in more audit quality, which in turn positively affects the financial reporting quality.

The indicators for audit quality are the audit fees and non-audit fees, according to Markelevich and Rosner (2013). Furthermore, the authors investigate whether the audit fees and non-audit fees have an association with fraudulent reporting. They find that higher audit fees lead to more fraudulent reporting, due to economic bonding and risk premiums. While non-audit fees also lead to more fraudulent reporting due to economic bonding, according to Markelevich and Rosner (2013). Therefore, it is evident that certain audit quality indicators affect the financial reporting quality, as fraudulent reporting equals low-quality reporting.

From the papers above, I believe the audit quality explains possible differences in financial reporting quality between reporting frequencies. As both papers state that the audit quality affects the financial reporting quality, even before auditors audit the financial statements. However, when the number of audits increases due to a higher reporting frequency, the audit fees annually increase as well. Subsequently, the question arises whether this negatively affects the audit quality, even though the audit fees for each audit stay equal.

2.7 Investors’ perspective

One of the main users of financial reports is investors. As executives provide financial reports to investors because investors base their investment decisions on the financial reports.

Therefore, investors must be well-informed through financial reports. Arif and De George (2020) investigate investors’ reactions towards a change in financial reporting frequency. They find that returns are twice more sensitive when firms announce their semi-annual financial reports, relative to the returns of firms that report quarterly. This indicates that investors cannot rely on alternative sources of earnings information, which leads them to overreact to peer-firm earnings news (Arif & De George, 2020). Therefore, the assumption is that investors unsuccessfully manage the information loss from semi-annual reporting, which indicates the results of this study.

On contrary, Brannon and Jennings (2020) state that there are three drawbacks to quarterly reporting, which are short-termism, excessive information, and randomness in the data. As they find that other information crowds out useful financial information, which makes the quarter reports less informative for investors. While randomness in the data refers to random events that affect a quarterly report, which leads to investors over-reacting in the next quarter’s earnings announcement (Brannon & Jennings, 2020). The authors state that investors over-react since managers usually smooth out the earnings in the next quarter. This theory contradicts the paper by Arif and De George (2020), as they state that investors over-react when firms report semi-annually.

Pitre (2012) investigates the effect of increased reporting frequency on nonprofessional investors’ earnings predictions. This study focuses on the perspective of nonprofessional investors when firms report more frequently. Based on the paper by Arif and De George (2020), the expectation is that more frequent reporting leads to better informativeness towards investors. However, Pitre (2012) finds that an increased reporting frequency causes less accurate and more dispersed predictions from nonprofessional investors, indicating that

nonprofessional investors do not understand more frequent reports. Subsequently, this lowers the market efficiency, as the financial reporting quality decreases.

2.8 Financial reporting frequency

Research closely related to investigating the difference in reporting quality between quarterly- and semi-annual reports is the research by Filzen (2015). Filzen (2015) studies the setting of 2005, where the SEC requires firms to disclose risk factors in their quarterly reports, to accommodate stakeholders in having another monitoring option over executives. The author finds that this change of the SEC affects the reporting quality of firms, as abnormal returns are lower when firms disclose risk factors in quarterly reports and these disclosed risk factors decrease future performance. The idea to analyze abnormal returns quarterly comes from the theory that investors are well-informed about risks that possibly affect performance and therefore respond through lower abnormal returns (Filzen, 2015). This responsiveness by investors shows the perception of investors regarding the informativeness of quarter reports.

When future performance drops due to these risks, it indicates that these risks are relevant for investors to be disclosed (Filzen, 2015). As the informativeness increases, the financial reporting quality increases as well.

The contradicting paper by Lee (2012) finds a different view of the informativeness of quarter reports. Lee (2012) investigates whether the information in quarterly reports is efficient enough for investors to understand. He states the unsophisticated investors are less capable of processing extensive quarter reports, which causes a delay in stock market reactions. However, sophisticated investors process the information promptly, which means delayed reactions are only present with unsophisticated investors (Lee, 2012).

Furthermore, the paper by Lee (2012) depicts that unsophisticated investors do not understand quarterly reports, due to that the information given is too excessive. This is relevant for my research, as it indicates the usefulness of quarter reports towards stakeholders.

Therefore, the contribution of my research to the literature is to know whether the processed information by these investors is of high quality in terms of reliability. Meanwhile, Filzen (2015) solely focuses on risk factors. While my research focuses on financial information.

Therefore, the study of Filzen (2015) is partially indicative of the results of my research. The difference between the informativeness of risks and financial information is that risks have a predictive nature, while financial information is more backward-looking.

2.9 Financial reporting quality

Several factors influence financial reporting quality, according to Herath and Albarqi (2017). As their study focuses on identifying factors that affect financial reporting quality in different ways. They state that the following elements are indicators of financial reporting quality: Relevance, comparability, understandability, timeliness, and faithful representation.

The relevant factors that influence the elements of financial reporting quality are the internal reporting system and accounting standards. A change in reporting frequency requires changes in the internal reporting system. Furthermore, the change in reporting frequency is a change in accounting standards, which influences the financial reporting quality. Regarding my study, I believe the elements of financial reporting quality are ambiguous, which makes it difficult to analyze financial statements.

Amiram et al. (2015) have a solution as they analyze the financial reporting quality through one element, namely Benford’s Law. They describe Benford’s Law as a way of analyzing certain documents by focusing on the leading digits of every item, such as financial statements. Benford’s Law captures elements of financial reporting quality, which makes it a comprehensive variable, which is why it is appropriate for my research.

A relevant paper regarding the financial reporting quality is the paper by Beretta and Bozzolan (2008). They investigate the association between quantity and quality of financial statement disclosures, as the general assumption is that a higher quantity equals higher quality.

As I study the quantity of reporting compared to the quality of reports, the paper by Beretta and Bozzolan (2008) relates closely to this. The authors find no evidence that the quantity of disclosures relates to the quality of it, which indicates the assumption that quarterly reporting does not improve financial reporting quality. However, the difference is that they focus on the financial statement disclosures, while I aim for the financial statement items. Also, the meaning of quantity of disclosures according to Beretta and Bozzolan (2008) is the amount of information given in the disclosure. This contradicts my study where the quantity is the number of reports in a fiscal year.

2.10 Continuous reporting

Several studies focus on a new way of reporting, called: continuous reporting. Roohani (2003) states that continuous reporting is a concept where firms report financial information continuously. This continuous reporting provides investors financial reports in real-time, which keeps investors updated at all times, according to Roohani (2003). Subsequently, the benefits

of continuous reporting are that investors pressure insiders of the firm less and the time interval of reporting is minimalized (Roohani, 2003). This shortened time interval prevents events from happening between the period-end and reporting time (Roohani, 2003). On contrary, the downside of continuous reporting is that it is not applicable for every firm’s infrastructure (Roohani, 2003). Continuous reporting is the highest frequency of reporting at low costs, which means that the information asymmetry and agency costs decrease.

Hunton et al. (2004) also focus on continuous reporting, where they examine the stock price volatility. As the stock price volatility indicates how investors respond to continuous financial reporting (Hunton et al., 2004). They state that the stock price volatility decreases when financial reporting frequency increases, which is in line with the paper by Arif and De George (2020). Therefore, they assume that continuous reporting increases investors’

understanding.

However, Gal (2008) contradicts this as he investigates the issues in continuous reporting systems. He states that investors, auditors, and managers cannot handle a continuous reporting system well. As investors have to make their reports instead of getting traditional reports, due to the system providing them financial data only (Gal, 2008). While auditors have to review the measures used by firms to make sure the information is always available (Gal, 2008). Managers require to continually provide disclosures regarding certain financial data (Gal, 2008). Therefore, Gal (2008) concludes that continuous reporting has several issues, which raises doubts regarding the effectiveness of this system.

Continuous reporting is a relevant subject for this study since the reporting frequency increases just like the switch from semi-annual reporting to quarterly reporting. Therefore, the benefits and downsides of continuous reporting are probably the same. Given the information from the papers above, I believe continuous reporting has issues regarding the human aspect of reporting. As computer systems process information rapidly, it is hard for humans to disclose such information at the same speed. However, the speed of reporting for continuous reporting is not present for quarter reporting, which means that humans can keep up with disclosures.

For this reason, I believe the benefits of continuous reporting are relevant for this study.

2.11 Hypotheses development

I seek to answer the following research question: ‘What is the effect of quarterly vs. semi- annual reporting on reporting quality?’. From the perspective of firms, evidence of the agency theory shows that executives have different interests than investors, which leads executives to

decrease the costs associated with financial reporting quality. While executives also suffer from short-termism, which gives them incentives to eventually mislead investors as Benton and Cobb (2019) show in their paper regarding earnings management. Fu et al. (2012) find that when information asymmetry decreases, financial reporting quality increases. Therefore, the assumption is that providing more information leads to fewer earnings management. Based on this information, I predict that quarterly reporting increases the financial reporting quality, as the information asymmetry decreases which makes investors more aware of the financial position of firms through more frequent reporting.

However, increasing the amount of information leads to more audits and thus more work for auditors. Several papers prove that increasing the workload of auditors decreases audit quality. Since an increased workload decreases the auditor’s job happiness, which eventually leads auditors to do worse during audits. The paper by Gaynor et al. (2016) confirms that decreasing audit quality is equal to decreasing financial reporting quality. Therefore, the argument that decreasing the information asymmetry through more frequent reporting is invalid, as the reliability of reporting needs to be audited by auditors who may suffer from the increased workload. In the early 2000s, studies regarding continuous reporting came through, where the idea is to have advanced technology do the reporting continuously. Nonetheless, this still requires human efforts as Gal (2008) mentions. The possibility of auditors providing continuous assurance is slim to none, while it is also impossible for managers to continuously provide disclosures regarding certain financial data. Therefore, my expectation overturns the thought that more frequent reporting leads to less financial reporting quality. Since the workload and possibilities cannot guarantee financial reporting quality.

The studies by Brannon and Jennings (2020) and Pitre (2012) show that investors also suffer from more frequent reporting, as it shows that they are not capable of understanding excessive information. This confirms my last prediction, as it shows that also investors have difficulties with the increased number of reports.

This prediction aligns with the paper by Lee (2012), where he concludes that there are delays in market reactions when firms publish quarter reports. Indicating that investors do not fully understand the information from the quarterly reports immediately, as they are too excessive in volume. Reports require understandability for them to be of high quality. Still, it is unclear what part of the quarterly report is too excessive or hard to understand (Herath and Albarqi, 2017). Filzen (2015) states that reporting risk factor disclosures per quarter increase the financial reporting quality, which gives the impression that non-financial information is

more reasonable to report quarterly. Therefore, the idea alters that reporting financial data quarterly reduces the financial reporting quality, while financial data disclosures may improve the financial reporting quality.

Beretta and Bozzolan (2008) investigate the relationship between quantity and quality of reporting, where they test financial disclosures. However, they contradict the paper by Filzen (2015) as they find no evidence that there is an association between quantity and quality of reporting financial disclosures. This nullifies the idea that reporting financial data disclosures quarterly increases the financial reporting quality. Regarding this subject, I cannot predict whether the financial data disclosures improve the financial reporting quality or not, as I cannot make a clear distinction between the papers by Filzen (2015) and Beretta and Bozzolan (2008).

Based on this information, I construct the following hypothesis:

H1: The change from semi-annual reporting to quarterly reporting significantly influences the financial reporting quality.

As many studies direct towards a decrease or increase in financial reporting quality when the reporting frequency increases, I cannot hypothesize in one direction with my prediction.

3 Research method

3.1 Sample selection

Based on the regulatory change of financial reporting frequency, I begin my sample in 1969 as that is the last year that U.S. firms report semi-annually (Kraft et al., 2018). After 1970, U.S. firms report quarterly, which extends the sample from 1969 to 1970. The initial sample from Wharton Research Data Services (WRDS), gathered from Compustat in North America, consists of 8,378 firm-year observations for all U.S. firms from 1969 through 1970. I exclude firms whose financial year does not end on the 31th of December, due to that their financial results contain a portion of 1969 and 1970 which makes comparing difficult. This decreases the sample size to 4,922 firm-year observations. Afterward, I select the observations with more than 100,000 U.S. dollars in assets during the period. This indicates the size of a firm, where at larger firms the distance between principal and agent is larger. Therefore, I believe it is more relevant to select these firms for the sample. Subsequently, resulting in 1,648 firm-year observations, where both years contain 824 firms from the sample.

As I analyze the difference between these quarterly- and semi-annual reporting, I need data from samples of quarter- and semi-annual reports. However, Compustat contains limited data for quarter- and semi-annual reports in the research period. Therefore, I choose to use the annual financial statements before and after 1970, as these follow after the quarter- and semi- annual reports by respectively three and six months. Below, the sample selection table:

Table 3.1 Sample Selection

Panel A: Summary of the Sample Selection Process

Activity No. financial statements

Initial sample from database 8,378

Less: Observations with other fiscal-year-end dates than the 31^st

of December 3,456

Less: Observations with value total assets lower than 100,000 US

dollars 3,274

Final sample 1,648

Panel B: Distribution of Financial Statements by Year

Year No. financial statements

1969 824

1970 824

3.2 Dependent variable – Financial reporting quality

To compare the financial reporting quality of firms in the period 1969-1970, I measure financial reporting quality to test a difference between reports in 1969 and 1970. This captures the difference in financial reporting quality between quarterly- and semi-annual reporting. In my literature review, I describe the financial reporting quality as the relevancy, comparability, understandability, timeliness, and faithful representation of a financial report towards stakeholders. There are several ways of measuring the relevancy, comparability, understandability, timeliness, and faithful representation of a financial report. Benford’s Law is a comprehensive measure to capture multiple elements of financial reporting quality.

Therefore, I choose Benford’s Law by Amiram et al. (2015) as the measure of financial reporting quality for this study.

3.2.1 Benford’s Law

Measuring the whole financial statement through the theory of Benford’s Law (Amiram et al., 2015) is a way of analyzing certain documents by focusing on the leading digits of every item. Furthermore, Amiram et al. (2015) test Benford’s Law with financial statements, where the financial statement divergence score (FSD score) signifies fewer errors when it is high.

The dependent variable is the FSD score measured through Benford’s Law. This FSD score is based on the KS stat. Benford’s Law analyzes the first digit of every number in the financial statement to find any errors. This way, I repeat it for all the financial statements from the sample before and after 1970. In the end, comparing the FSD score of the firms before and after 1970 concludes which reporting frequency is better in terms of financial statement quality.

3.2.1.1 Benford’s Law – MAD statistic

Further, I conduct Benford’s Law by counting the number of leading digits for each number from 1 to 9 (Amiram et al., 2015). After that, I divide the frequency of each number by the total amount of leading digits (Amiram et al., 2015). Following, this results in an empirical distribution per number (Amiram et al., 2015). Afterward, compare this number with Benford’s Law’s theoretical distribution per number (Amiram et al., 2015). Then, I make the difference between these distributions per number absolute to avoid any negative outcomes (Amiram et al., 2015). Subsequently, I calculate the mean of these absolute differences, which results in the mean absolute deviation statistic (MAD statistic) (Amiram et al., 2015).

Furthermore, this statistic indicates the deviations in the financial statements and calculates the average of it (Amiram et al., 2015).

Calculating the MAD statistic is as follows:

!"# =(|'#_!− )#_!| + |'#_"− )#_"|+|'#_#− )#_#| + |'#_$− )#_$+ ⋯ + |'#_%− )#_%|) .

Where:

MAD = The mean absolute deviation;

ED = Empirical distribution;

TD = Theoretical distribution;

k = Total number of digits.

3.2.1.2 Benford’s Law – KS statistic

Amiram et al. (2015) go further beyond the MAD statistic and indicate that the Kolmogorov–Smirnov statistic (KS statistic) is more precise in certain situations. The MAD statistic does not have a critical value, which makes a judgment based on comparison with critical value difficult (Amiram et al., 2015). Thus, it is only possible to analyze the MAD statistic by comparing it to the MAD statistic of many other firms (Amiram et al., 2015). The KS statistic is the maximum value of the absolute difference between the distributions (Amiram et al., 2015). This value has a critical value, which is dividing 1.36 with the square root of the number of digits used (Amiram et al., 2015). Whenever the critical value is higher than the KS statistic, the deviation is not significant (Amiram et al, 2015). Amiram et al. (2015) state that this is because when the number of digits used increases, the KS. statistic becomes less useful.

Calculating the KS statistic goes as follows:

/0 = !12(|"#_!− '#_!|, |("#_!+ "#_") − ('#_!+ '#_")|, … , |("#_!+ "#_" + ⋯ + "#_%)

− ('#_!+ '#_" + ⋯ + '#_%)|

5678791: ;1:<= =1.36

√C Where:

KS = The Kolmogorov-Smirnov statistic;

ED = Empirical distribution;

TD = Theoretical distribution;

Max = Maximum value of absolute cumulative deviation between distributions;

P = Number of digits used.

3.3 Independent variable – Financial reporting frequency

The independent variable of this study is whether a financial statement is provided in the period of semi-annual reporting or the period of quarterly reporting as I investigate the difference in financial reporting quality between the reporting frequencies. The SEC mandates U.S. firms to report quarterly since 1970, which is the reason why this study focuses on that period. All U.S. firms that report under the ruling of the SEC publish financial information quarterly. Therefore, I cannot create two different samples of U.S. firms, where one sample reports quarterly and the other semi-annually. The comparison of reporting frequency is done in time, where the samples consist of the same firms but differ in periods.

3.3.1 Event study

Several studies use samples that only differ in time as they investigate the effect of a certain event. For example, He et al. (2020) use this method to test the effect of a unique event on the market. My study is similar in this regard, where the event is the change in the regulation of the SEC. The event study takes place before and after 1970, as most big U.S. GAAP reporting firms require to provide financial statements since 1970. Therefore, the sample consists of two parts where; the first part contains 824 firms reporting in 1969, while the second part contains the same 824 firms reporting in 1970.

In a regression, this event is depicted as a dummy variable; QREP. Where 0 and 1 signal firms reporting in 1969 and 1970 respectively. The Beta (b) of this dummy variable represents the factor in which the difference is in financial reporting quality. A positive b indicates that quarterly reporting provides better financial reporting quality relative to semi-annual reporting, while a negative b indicates the opposite. There is no difference if the b is insignificant.

However, there is a downside to using an event study, which is the possibility of factors influencing the dependent variable other than the financial reporting frequency. To make sure the effect is not caused by other factors, it is appropriate to add several factors to this study.

3.4 Control variables

During the period 1969-1970, the possible factors influencing the financial reporting quality could vary. However, I intend to capture the most probable and influential factors in this study, as it is impossible to take into account all smaller factors.

The first possible influencing factor is technological developments; TECH, which decreases the costs and errors in financial statements (Wang & Kogan, 2020). Subsequently, technological developments make more resources available for firms to report more frequently on high standards (Wang & Kogan, 2020). However, I believe technological developments were not as present in 1970 relative to now, but could still affect the financial reporting quality nonetheless. So, the assumption is that technological developments are present in 1970 and possibly affected the financial reporting quality, which makes it relevant to include in the event study.

Karabag (2019) investigates technological developments by focusing on the research &

development (R&D) of firms. However, my sample of firms does not report R&D costs, as in that period, firms did not report their R&D costs. Therefore, I assume capital expenditures (CapEx) in property, plant, and equipment (PP&E) capture the investments in technological developments. So, the proxy for this control variable is the CapEx per firm scaled by total assets.

The second possible influencing factor comes from the economic growth; EG, as I assume that a bad economic growth leads to more pressure on the management, which might eventually lead to earnings management. Therefore, it is interesting to include this factor in the event study as it might explain a possible difference in financial reporting quality. Schröder &

Storm (2020) measure economic growth through the Gross Domestic Product (GDP). As the GDP is a comprehensive proxy for economic activity, by measuring the value of the final goods and services (BEA, 2020). However, GDP cannot be measured across firms as it captures the economic growth of a country, which is not firm-specific. The GDP consists of consumer spending, operating expenses, government spending, and net exports of goods and services. I believe the GDP components are closely related to revenue, cost of goods sold, and profit/loss before interest and tax. For this reason, profit/loss before interest and tax is a suitable measure for economic growth as a control variable across firms scaled by total assets, which is the return on assets (ROA).

The third possible influencing factor is earnings management. Earnings management could be more present when firms’ management have to report more frequently. Sloan (1996) mentions that signaling earnings management is done through the two components of earnings;

cash flow and accruals. The ratio between cash flows and accruals of revenues and costs

typically remains largely consistent throughout time (Sloan, 1996). Whenever this ratio changes drastically where accruals increase relative to cash flows, it may indicate that the management increases the earnings through increased accruals, which may not be justified (Sloan, 1996). Therefore, I use the proportion of accruals of revenues to analyze whether this changes when the financial reporting frequency increases. The financial statement data from Compustat do not include cash flows, which is why I use the proportion accruals within revenues.

The fourth possible influencing factor is firm size. Fan and Wong (2002) state that bigger firms have an ownership structure where control and ownership are distant. Subsequently, the information asymmetry is larger at bigger firms due to this distance between ownership and control (Fan & Wong, 2002). More information asymmetry leads to less financial reporting quality, which may be present during the event study. Therefore, it is important to control this possible effect for the event study. Wakil (2019) uses the total book assets to measure the firm size. The total book assets depict the value invested in the firm, which shows the size of all assets of the firm. For this reason, I use the total book assets as a proxy of the firm size.

The fifth possible influencing factor is the financial position of the firm measured through the Zmijewski bankruptcy model; FIN_POSI. Silvan (2020) uses the Zmijewski bankruptcy model to predict corporate bankruptcies, where he uses the following formula to calculate the prediction:

Zmijewski score = −4.3 − 4.4X_!+ 5.7X_"− 0.004X_# Where:

X1 = (Profit After Tax / Total assets) * 100%

X2 = (Total Debt / Total assets) * 100%

X3 = (Current assets / Debt Current) * 100%

The cut-off of the Zmijewski score is zero, where a score above zero means the company will go bankrupt. While a score lower than zero indicates a healthy financial position. I use this control variable since the assumption is that managers under the pressure of risking bankruptcy are more susceptible to earnings management because managers try to depict a healthy financial situation. When firms have to report more frequently, lower financial reporting

quality is more present due to an increase in earnings management. Therefore, I predict that firms with a score closer to zero have low financial reporting quality.

3.5 Regression model

To link the independent variable to the dependent variable, I provide the following regression model based:

FSDit = a0 + b1QREPit + b2TECHit + b3EGit + b4R_ACC_REVit + b5SIZEit + b6FIN_POSIit + eit

Where:

FSDit = Financial statement divergence for firm i at year t;

QREPit = Quarter report for firm i at year t;

TECHit = Technological development for firm i at year t;

EGit = Economic growth for firm i at year t;

R_ACC_REVit = Ratio accruals/revenues for firm i at year t;

SIZEit = Firm size for firm i at year t;

FIN_POSIit = Financial position for firm i at year t.

3.6 Reliability and validity of proxies

To test the reliability and validity of the proxies used in the regression model, I use Cronbach alpha. The Cronbach alpha is a statistic that shows a value between 0 and 1 (Ursachi et al., 2013). Ursachi et al. (2013) explain that the generally accepted rule is that a Cronbach alpha of 0.6-0.7 is an acceptable level of reliability, whereas 0.8 or higher is very good.

However, the authors mention that an excessively high level of reliability an indication is of redundance. This reliability refers to the consistency relating to the group of variables of the regression model.

The Cronbach alpha for this study is 0.6102 which is acceptable. However, this value lacks one variable, which is SIZE. After testing the Cronbach alpha with all variables, I find that SIZE lowers the Cronbach alpha under 0.6. Subsequently, I regard SIZE as less reliable or inconsistent with the other variables throughout this study. Nonetheless, removing SIZE may give fewer insights into the relationships between the variables from the regression model.

Therefore, I maintain this regression model.

3.7 Statistical method

As I investigate a possible effect of financial reporting frequency on the financial reporting quality through a sample of firms’ financial statements in two years, I use an event study. The difference between the two years is caused by a regulatory change. Subsequently, this regulation affects the way firms report their financial information. With limited time in mind, it is not possible to collect interviews and survey responses to make this study representative. For this reason, I believe an archival research method is appropriate here.

Through Compustat, I collect the financial statements by looking at the simplified financial statements from U.S. firms between 1969 and 1970. Also, I collect the GDP of the involved years to investigate whether the economy grew significantly enough to influence the dependent variable. Furthermore, I use the ordinary least squares (OLS) method to produce an analysis. To visualize the research method, I present Libby boxes below:

Independent variable (X) Dependent variable (Y)

Conceptual

framework Frequency of reporting Reporting quality

Operational framework

Difference annual financial statements from

U.S. firms before and after 1970

Financial Statement Divergence (Benford’s

Law)

Control variables: Economic growth, technological developments, earnings management, financial position, and firm size.

Figure 1: Libby boxes.

4 Results

4.1 Descriptive statistics

The regression model contains multiple variables that may correlate with each other. In table 4.1, I summarize the variables from the regression model into descriptive statistics.

Further, I split the FSD variable into two separate variables, where FSD1 and FSD2 are the MAD statistic and KS statistic respectively. It is interesting to analyze whether the MAD statistic and KS statistic differ in results as this may indicate which statistic is more sensitive.

Also, I winsorize all variables except QREP. With this, the observations in the first percentile change into the values of the second percentile, and the values in the largest percentile change into the values of the 99^th percentile. Though winsorizing, I maintain the same number of observations while adjusting extreme outliers.

Table 4.1 Descriptive Statistics

Variable N Mean Median Std.

Dev. Min Max

QREP 1648 0.500 0.500 0.500 0.000 1.000

TECH 1648 0.064 0.056 0.058 0.000 0.257

EG 1648 0.064 0.052 0.069 -0.089 0.297

R_ACC_REV 1648 0.127 0.139 0.097 0.000 0.477

FIN_POSI 1648 -1.643 -2.947 0.880 -3.403 -0.312

SIZE 1648 825.866 360.867 1266.478 105.087 8104.820

FSD1 1648 0.046 0.045 0.013 0.020 0.086

FSD2 1648 0.115 0.106 0.045 0.042 0.263

The variables in the descriptive statistics are used in the regression model. This table provides the number of observations, the mean, the median, the standard deviation and the minimum and maximum value of the variables.

The number of observations is equal for every variable as I use one sample of 1,648 firm-year observations for all variables. Further, the mean and median show how the data per variable is distributed, which may indicate that certain excessive numbers drive the mean upwards, for example. While the standard deviation indicates the deviation of the mean, which shows how much the variables differ from the mean. The minimum and maximum values of the variables present the range of values within the variable.

Table 4.1 shows that the mean of QREP is 0.500, which indicates that half of the sample report quarterly. This is in line with the sample, where 50% report semi-annually and 50%

quarterly. Due to the sample being halved by semi-annually- and quarterly reporting, the median has the same value which makes the sample normally distributed. Since this is a dummy variable, the standard deviation is 0.500 as the variable can only deviate by this value, where the minimum and maximum values are 0 and 1 respectively.

The control variable TECH is the capital expenditures of firms in the sample. Table 4.1 presents a mean of 0.064, which is higher than the median. This indicates that the distribution of the variables is right-skewed as several firms spend more excessively than the overall sample. The standard deviation is almost equal to the mean, indicating that the capital expenditures vary between almost zero and double the value of the mean. This means several firms have not invested in PP&E, which explains why the variable can deviate towards little more than zero. This is evident from the minimum.

Compared to TECH, the EG is very similar regarding the statistics. This means that firms mostly invest similar to what they earn unless they record a loss that requires investments to recover. Therefore, the difference between TECH and EG is that EG has a negative minimum value, indicating a loss. Subsequently, this slightly increases the standard deviation compared to TECH. Furthermore, the EG is generally between 0 and 0.10 which is not an unusual ROA for many firms.

Table 4.1 indicates that the proportion of accruals within revenues across firms is around 3 to 20%. While the maximum value is 47.7%, indicating that several firms have excessive accruals. This may be caused by earnings management or optimistic behavior.

The FIN_POSI of the sample ranges Zmijewski scores from -3.403 to -0.312, which indicates that all firms were not certain to go bankrupt. However, several firms did score near zero which means that earnings management due to financial pressure could be present in this sample. This in turn lowers the financial reporting quality. Also, the mean is lower than the median, indicating that a minority of firms with a low score decrease the overall mean of this score across the firms within the sample.

The standard deviation of FSD1 is relatively small compared to the mean, indicating that the MAD statistic is mostly similar for many firms. While FSD2 does have a higher standard deviation as the KS statistic captures the highest divergence, instead of the average captured by the MAD statistic. It also shows that the maximum of FSD2 is above the critical value, meaning that the financial statement contains misstatements.

To further analyze the mean of all variables, I perform a two-tailed t-test of the means, comparing the means between semi-annual and quarterly reporting.

Table 4.2 Descriptive Statistics: Two-Tailed T-test of the Means Semi-

annual (N=824)

Quarter (N=824)

Variable Mean Mean Difference in means

P-value (One- sided <)

P-value (two- sided)

P-value (One- sided >) TECH 0.0657 0.0615 0.0636 0.9269 0.1461 0.0731*

EG 0.0735 0.0552 0.0183 1.0000 0.0000*** 0.0000***

R_ACC_REV 0.1270 0.1270 0.0000 0.4938 0.9875 0.5062 SIZE 794.3464 857.3854 -63.0291 0.1562 0.3125 0.8438 FIN_POSI -1.6866 -1.5997 -0.0868 0.0233** 0.0466* 0.9767

FSD1 0.0464 0.0452 0.0012 0.9682 0.0636* 0.0318**

FSD2 0.1165 0.1128 0.0037 0.9534 0.0932* 0.0466**

Table 4.2 provides a two-tailed t-test to compare the means of the variables between two categories, namely: semi- annual reporting and quarterly reporting. Three P-values are given to analyze the significancy of the differences in means. The difference is calculated by subtracting the means of semi-annual by the means of quarter. Thus, ‘One- sided <’ equals difference lower than zero and ‘One-sided >’equals difference higher than zero.

*, **, *** significant at 10%, 5%, and 1% levels, respectively.

During the switch from semi-annual to quarterly reporting, firms increased capital expenditures as shown in table 4.2. It seems that the means of TECH differ at a 10%

significance level when I use the one-sided p-value.

Firms declined in economic performance during the event as the mean of EG is significantly lower when firms report quarterly. While the level of accruals stayed equal, suggesting that firms did not use earnings management even though the performance was declining. The means of SIZE do not differ as no particularities took place where firms grew or declined during that period.

Furthermore, the FIN_POSI of firms is significantly worse when firms report quarterly, which contradicts the EG. I assume that the minority of firms with very low Zmijewski scores causes the difference compared to the group that reports semi-annually.

The most indicative results for this study are the differences in means for FSD1 and FSD2. It shows that the divergence in financial statements is significantly lower when firms report quarterly. However, the cause of this can be due to economic growth or technological developments. Therefore, further testing is needed to identify the cause of this decrease.

4.2 Correlation matrix

To acquire an understanding of the relationships among all variables, I present a Pearson correlation matrix in table 4.3. This matrix shows the magnitude of the correlation between the variables, which indicates how the variables are related to each other. The magnitude is given by a number between -1 and +1, while 0 indicates an insignificant relationship. A number close to -1 or +1 shows that there is more correlation.

From table 4.3, it is evident that TECH correlates significantly with all variables, except QREP. This means that the capital expenditures did not significantly change between 1969 and 1970, which is in line with the two-sided P-value test in table 4.2. Therefore, I assume that TECH does not explain QREP’s effect on FSD1 and FSD2 as TECH is independent of QREP.

However, EG correlates with every variable significantly. The return on assets decreases when firms report quarterly, which means that EG may explain the effect of QREP on FSD1 and FSD2. It also shows that EG correlates negatively with FSD1 and FSD2, thus increasing the quality of financial reporting. This is in line with the assumption that a bad financial position pressures executives into earnings management.

R_ACC_REV correlates with every variable, except QREP and SIZE. This suggests that firms do not increase accounts receivables relative to revenues when they report quarterly.

Therefore, I exclude the possibility that earnings management may explain the effect of QREP on FSD1 and FSD2. Further, R_ACC_REV correlates negatively with FSD1 and FSD2, meaning that the ratio between accounts receivables and revenues increases the financial reporting quality. This does not relate to earnings management as the ratio does not change over time to unrightfully increase revenues through accounts receivables. It simply means that firms with higher levels of accounts receivables relative to revenues have more financial reporting quality.

Since SIZE does not correlate with QREP, R_ACC_REV, and FSD1, it cannot explain the effect of QREP on FSD1. This means that the value of total assets does not improve or undermine the financial reporting quality for this measure of financial reporting quality.

Table 4.3 Correlation Matrix

Variables QREP TECH EG R_ACC_R

EV

FIN_POSI SIZE FSD1 FSD2

QREP 1.000

TECH -0.035 1.000

EG -0.136*** 0.193*** 1.000

R_ACC_REV 0.002 0.475*** 0.114*** 1.000

FIN_POSI 0.049* -0.222*** -0.704*** -0.029*** 1.000

SIZE 0.031 0.077*** -0.078*** -0.029 0.054** 1.000

FSD1 -0.045* -0.187*** -0.098*** -0.244*** 0.049** -0.036 1.000

FSD2 -0.041* -0.147*** -0.075*** -0.173*** 0.001 -0.048** 0.757*** 1.000

Table 4.3 provides the Pearson correlation between variables.

*, **, *** significant at 10%, 5%, and 1% levels, respectively.

Finally, QREP correlates significantly negatively with FSD1 and FSD2, which indicates the outcome of this study. Therefore, I assume that reporting quarterly improves the financial reporting. However, since EG shares the same correlation, EG may explain the correlation between QREP and FSD1 and FSD2. This can have multiple interpretations. One interpretation could be that during the event economic growth was evident, which led to firms improving their reporting. Subsequently, this interpretation means that QREP has no contribution to the increase in financial reporting quality. Another interpretation could be that quarterly reporting pressures firms to improve their performance, while also improving the financial reporting quality.

Following the correlation matrix, I present a variance inflation factor (VIF) test in table 4.4. The goal of this test is to investigate whether the independent and control variables do not lead to multicollinearity. This means that no independent or control variable has the same linear function as one or more other independent and control variables. If multicollinearity is present, then one b in the regression model captures multiple proxies. Subsequently, the regression model contains the same b more than once, which is not required.

Table 4.4 Variance Inflation Factor (VIF) Test VIF 1/VIF

TECH 1.34 0.744

R_ACC_REV 1.38 0.724

EG 2.09 0.479

QREP 1.03 0.975

SIZE 1.02 0.977

FIN_POSI 2.16 0.463

Mean VIF 1.50

Table 4.4 provides the Variance Inflation Factor test for the independent variables of the regression model.

Table 4.4 shows no evidence of multicollinearity since independent and control variables have low VIFs. The highest VIF is 2.16, which is significantly lower than 10.

Therefore, I conclude that the independent and control variables in the regression model do not capture the same influence.

4.3 Ordinary Least Squares regression

To test the hypothesis, I use OLS regression to identify how the independent and control variables individually affect the dependent variable. Also, I evaluate the regression model through the use of the OLS regression. Furthermore, I use the OLS regression twice for the MAD statistic and KS statistic as the dependent variable, which results in two separate tables for each dependent variable.

4.3.1 OLS regression MAD statistic

Below, I present the OLS regression results for the MAD statistic as the dependent variable.

Table 4.5 Ordinary Least Squares Regression MAD statistic

MAD Coef. St.Err. t-value p-value [95% Conf. Interval]

Constant 0.050 0.001 65.05 0.000*** 0.049 0.052

QREP -0.001 0.001 -2.43 0.015** -0.002 0.000

TECH -0.015 0.006 -2.44 0.015*** -0.028 -0.003

EG -0.029 0.007 -3.95 0.000*** -0.043 -0.014

R_ACC_REV -0.029 0.004 -7.51 0.000*** -0.036 -0.021 FIN_POSI -0.001 0.000 -3.22 0.001*** -0.002 -0.001

SIZE -0.000 0.000 -2.29 0.022** -0.000 -0.000

Mean dependent var 0.045 SD dependent var 0.012

R-squared 0.070 Number of obs. 1648

F-test 20.395 Prob > F 0.000***

Akaike crit. (AIC) -9984.013 Bayesian crit. (BIC) -9946.247

*** p<.01, ** p<.05, * p<.1

The overall regression model is significant at a 1% level, which indicates that this model is suitable for the explanation of variation from the mean. However, the model only has an R- squared of 0.070, which means that the model only explains 7% of the variation of the mean.

This means other factors out of the scope of this study explain the residual variation between the variables. Further, the other factors could be many random factors that are difficult to capture in this study. Therefore, I keep the focus on the known variables from the OLS regression, since this model is significant at the 1% level.

Table 4.5 shows that every dollar in TECH results in a decrease of the MAD statistic by 0.015, which means that capital expenditures increase the financial reporting quality slightly. The same principle applies to EG, where every dollar leads to a decrease of 0.029 in the MAD statistic. Therefore, economic growth relieves executives of economic pressure, which lowers earnings management and eventually increases the financial reporting quality.