• No results found

The effect of U.S. mandatory XBRL-adoption on earnings management

N/A
N/A
Protected

Academic year: 2021

Share "The effect of U.S. mandatory XBRL-adoption on earnings management"

Copied!
68
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

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

Master Thesis:

The Effect of U.S. Mandatory XBRL-Adoption on Earnings

Management

Name: Robin Oegema

Student number: 5825172

Thesis supervisor: dr. Réka Felleg

Date: 19 June 2016

(2)

Statement of Originality

This document is written by student Robin Oegema, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is 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.

(3)

The Effect of U.S. Mandatory XBRL-Adoption on Earnings Management

Abstract

In this study, I examine whether the mandated XBRL-adoption by publicly traded firms in the United States is related to the extent to which firms manage earnings in the pre-XBRL versus the post-XBRL period. Using absolute discretionary accruals as a proxy for accrual-based earnings management, I find a decrease in accrual-based earnings management in the post-XBRL period, which is consistent with the view that post-XBRL improves transparency and financial information accessibility and therefore constrains managerial opportunism through accruals. I further find that real earning management, measured as the abnormal levels of overproduction and discretionary expenditures, increases in the post-XBRL period. These findings supports the notion that firms use accrual-based and real earnings management as substitutes. In particular, my findings show that firms shift to real earnings management when their ability to manage earnings through accruals is constrained. This implies that XBRL does not decrease earnings management activities altogether, but changes managers’ earnings management strategy.

Keywords: XBRL (eXtensible Business Reporting Language), accrual-based earnings management, real earnings management, transparency of financial reports

Data availability: Data are available from public sources indicated in the text.

Acknowledgement: I am grateful for the suggestions and guidance I have received from my master thesis supervisor, Réka Felleg at the University of Amsterdam.

(4)

4

I. Introduction

In 2008, the Securities and Exchange Commission (SEC) mandated that public companies adopted eXtensible Business Reporting Language (XBRL) for their corporate filings. In line with its ultimate goal to replace the existing filing system with an interactive data disclosure system, the SEC mandated XBRL because it believes XBRL has the potential to increase the accuracy, usability, and the timeliness of financial disclosure and eventually reduce costs for investors (SEC, 2008). XBRL is an international standard used to digitally codify business reporting information with the objective to improve the financial information interchange by representing the information in a standardized way (XBRL International, n.d.).

In this study, I examine whether there is a relation between mandatory XBRL adoption and earnings management in the U.S. In particular, I compare the pre- and post-XBRL period regarding the level of both discretionary accruals and real activities manipulation. My study is motivated by the increased interest for XBRL among regulators and auditors. Although many countries recently adopted or still are moving on to XBRL reporting, empirical evidence of its potential benefits and costs are scarce. In addition, my research is motivated by the findings of previous literature on both XBRL and earnings management. Only a few studies examined the consequences of the change of financial reporting in the U.S. since the mandatory adoption of XBRL in 2008. For example, Kim, Lim, and No (2012) report that the mandated XBRL adoption resulted in a reduction of information asymmetry. Li, Lin, and Ni (2012) in addition, support this by finding an improved information environment and decreased cost of equity capital after the SEC mandate. Before the mandate, Hodge, Kennedy, and Maines (2004) already found evidence that suggested that XBRL facilitates managers in acquiring and integrating financial information and that it improves the transparency of managers’ financial reporting choices. Previous research shows that when financial information is more transparent, investors are better able to scrutinize managerial opportunism (e.g., Hirst & Hopkins, 1998).

(5)

5

Based on the discussion above, Peng, Shon, and Tan (2011) tested whether mandated XBRL adoption in China led to managers engaging in earnings management to a lower degree. They find a decrease in accruals in the post-XBRL period relative to the pre-XBRL period. However, by looking solely at accrual-based earnings management, they neglect that managers may engage in real earnings management when their ability to manipulate accruals is constrained (Zang, 2012). The possibility that mandatory XBRL adoption causes firms to use the two alternatives of earnings management as substitutes is a motivation to conduct this study. Additionally, the results of Peng et al. (2011) are a motivation to investigate the effect of XBRL on earnings management in the U.S., since they mention that their results have limited generalizability because of the specific characteristics of the Chinese corporate disclosure environment.

Based on the evidence by Kim et al. (2012) and Li et al. (2012) of decreased information asymmetry in the U.S. since the SEC mandate, I expect, similarly to Peng et al. (2011), a decrease of accrual-based earnings management after the mandated XBRL adoption. In line with Cohen et al. (2008), Cohen and Zarowin (2010) and Zang (2012) however, I predict that real earnings management has increased since the mandated XBRL adoption.

To test my predictions, I examine a sample of XBRL-filers between 15 June 2009 and 14 June 2014. I use the cross-sectional model developed by Dechow et al. (1995) to estimate the level of discretionary accruals, which proxy for accrual-based earnings management. To examine real earnings management, I follow prior studies (Cohen et al., 2008; Cohen & Zarowin, 2010; Roychowdhury, 2006) by using abnormal overproduction and abnormal cutting of discretionary expenditures as proxies.

The results of my analyses show that firms report a lower level of discretionary accruals in the post-XBRL period than in the pre-XBRL period, which is consistent with the view that XBRL improves transparency and information accessibility and therefore constrain managerial

(6)

6

opportunism through accruals. My additional analyses across different subsamples show that this effect is only applicable to large and midsized firms. I further find that firms report higher levels of real earnings management in the post-XBRL period than in the pre-XBRL period, which is consistent with the notion from prior literature (e.g., Cohen et al., 2008; Cohen & Zarowin, 2010; Zang, 2012) that firms use accrual-based earnings management and real earnings management as substitutes. Additional analysis indicates that this substitutive relation predominantly means that firms shift to real earnings management when their ability to manage earnings through accruals is constrained. Additional test also show that this result is only applicable to large XBRL-filers.

The findings of my study contribute to the existing literature by, to the best of my knowledge, providing the first evidence of the effect of the mandated adoption of XBRL on both accrual-based and real earnings management in the U.S. With these findings I further provide a unique contribution to the extensive literature that is written on earnings management. The findings of this study might be relevant for regulators, investors and managers of (future) adopters of XBRL reporting. Since there is little empirical evidence on the potential benefits and costs of XBRL-adoption, my results may help inform regulators that recently introduced or will introduce XBRL. My findings identify a potential benefit of XBRL, namely that XBRL constrains and decreases accrual-based earnings management. However, because of the shift towards real earnings management, my findings suggest that XBRL-adoption merely changes the strategy of managers to manage earnings. Therefore, it is questionable whether a decrease of total level of earnings management is one of the benefits gained from XBRL-adoption. For investors this may imply that although XBRL could improve information accessibility, they will bear more information risk as firms tend to shift to an earnings management strategy that is more difficult to detect and can be more costly in the long-term.

(7)

7

The remainder of this paper is structured as follows. Section II discusses the theoretical background of this study and develops my hypotheses. Section III outlines the research methodology and Section IV presents the data and the empirical results. Finally, Section V concludes.

II. Theory and hypothesis development Background

On December 17, 2008, the U.S. Securities and Exchange Commission (SEC) voted to adopt new rules that require public companies and mutual funds to provide their financial statements to the SEC and on their corporate website in interactive data format using the eXtensible Business Reporting Language (XBRL) (SEC, 2008). The SEC (2008) believes that XBRL has the potential to increase the accuracy, usability and the speed of financial disclosure. The decision to mandate XBRL adoption is part of the long term goal of the SEC to replace the current Electronic Data Gathering Analysis and Retrieval (EDGAR) system with a new interactive data disclosure system. With the new rules, the SEC intended not only to make financial information easier for investors to analyze, but also to assist in business information processing and automating regulatory filings (SEC, 2009). As a result, the mandated XBRL adoption may improve the quality of information while reducing its cost.

XBRL was developed to help overcome the difficulties that arose since the development of electronic filing of financial statements (Hoffman, 2006). Examples of these difficulties were the multiple file formats of electronic financial statements and the inconsistency of content which hindered the information exchange between organizations and across different countries. Inspired by the need to address these exchangeability issues, Hoffman (2006) proposed to use an open standard digital language for financial reports. This idea ultimately led to the development of XBRL. XBRL is an international standard used to digitally codify information generally required for business reporting (XBRL International, n.d.). The computer language

(8)

8

technique is proclaimed by XBRL International, a global consortium of more than 600 organizational members from more than 35 countries, that engages in the development and maintenance of the XBRL specifications that make up the standard. They define the objective of XBRL as to improve the financial information interchange by representing the information in a standardized way, independently on the adopted system what is already available in the information system of a company. The standardization is reached by using specific taxonomies composed by predefined “tags” to map the internal data of the organization. Tags are used in electronic formats such as HTML and XML, and are used to define or “tag” data using standard definitions (SEC, 2009). In the case of XBRL, which is derived from XML, financial reporting systems and other software applications recognize these tags and process the tagged financial information. XBRL U.S., the U.S. jurisdiction representative of XBRL International, was contracted by the SEC in 2006 to develop the list of tags necessary for financial reporting in an format that is consistent with U.S. GAAP and the regulations of the SEC (SEC, 2009). In addition, the taxonomy was reviewed by the Financial Accounting Standards Board (FASB), which may also be involved in publishing future draft tags together with their accounting standard. Hence, XBRL provides an official, standardized format to prepare, publish and exchange (financial) business information.

In the last years, several countries adopted programs of voluntary or compulsory adoption of XBRL for the generation of financial reports. For instance, USA, Canada, China, Japan, and Israel have already mandated that publicly traded companies submit their filings in XBRL. Some other countries are still moving on to mandate electronic filing of financial accounts. For example, the Dutch parliament recently (14 April 2016) introduced a new mandate for electronically filing financial reports. The legislation phases in XBRL reporting as part of the Dutch Standard Business Reporting (SBR), starting in January 2017.

(9)

9 Prior research on XBRL

Since the development of XBRL in the early 2000s, a substantial body of XBRL research, both empirical investigations and conceptual papers, has appeared in academic journals (Perdana, Robb, & Rohde, 2015). Perdana et al. (2015) conducted an integrative review on XBRL research and identified four key research themes: (a) XBRL’s impact on business, (b) XBRL’s adoption, (c) XBRL’s technical development, and (d) XBRL education. The first theme, XBRL’s impact on business, is a broad topic that covers how XBRL helps organizations to prepare, distribute, and use financial information effectively and efficiently; what the problems and benefits are that are associated with XBRL implementation; and how XBRL can facilitate financial information users in making better-informed decisions. For example, several studies have reported that XBRL has the ability to improve the usefulness of accounting information (e.g., Cohen, 2009; Vasarhelyi, Chan, & Krahel, 2012), the information integrity (e.g., Madden, 2011), and financial information quality (e.g., Valentinetti & Rea, 2012). Other studies highlight XBRL’s potential to increase transparency and reduced information asymmetry because of improved information accessibility (e.g., Kim et al., 2012; Li, Roge, Rydl, & Hughes, 2006; Pinsker & Li, 2008; Yoon, Zo, & Ciganek, 2011). Blankespoor, Miller, & White (2014) on the other hand, do not find evidence of a reduction in information asymmetry during the first years following the SEC mandate. However, their results are also consistent with the notion that some investor learning may be taking place, suggesting that, together with the technology becoming more accessible, reduction of information asymmetry might indeed be reduced in the future. Kim et al. (2012) state that this improved transparency and reduced information asymmetry are the key benefits of XBRL that help achieve good corporate governance. Apostolou and Nanopoulos (2009) underline this by finding that stakeholders can perceive XBRL implementation as signal that companies are willing to apply good corporate governance principles to their business processes. In addition, Premuroso and Bhattacharya

(10)

10

(2008) found that early and voluntary decisions to adopt XBRL reports is associated with superior corporate governance structures and with increased transparency of reported information. Rao, Guo, and Hou (2013) found that extended XBRL taxonomies that lead to improved financial information disclosure, is associated with five factors related to good corporate governance. Although there are various studies that find a positive impact of XBRL on good corporate governance, Alles and Piechocki (2012) nuance this by stating that for XBRL to fundamentally change governance it as to add more value than simply facilitating interactive data exchange. They claim that companies that implement XBRL also have to take into account factors which are beyond XBRL’s scope, such as understanding the value chain and understand the need to align business strategies and processes.

Other studies found an impact of XBRL on business with regard to the audit process. Rezaee, Elam, and Sharbatoghlie (2001) for example, state that XBRL enables auditors to conduct continuous auditing. With continuous auditing auditors can access the information from companies’ databases without relying on conventional financial statements. Rezaee et al. (2001) claim that XBRL can help minimize error-prone activities that are related to manual preparation of financial information because auditors can get the data directly from the databases. With XBRL, the accounting data is tagged at the transactional level, so this also helps auditors to trace the audit trail from the financial statements to the journal entries. Du and Roohani (2007) support these claims by stating that continuous auditing permits interaction between the auditors’ systems and the auditee’s information systems. This enables auditors to moves their focus from data extraction and transformation towards value adding, i.e., evaluation of their clients’ financial statements. In addition, XBRL can also help U.S. companies comply with regulatory requirements such as the Sarbanes-Oxley Act (e.g., Baldwin & Trinkle, 2011.)

Despite these research findings that suggest that XBRL merely has positive impacts on businesses, there are also studies that imply some concerns with regard to the information

(11)

11

reported in XBRL-based financial statements. Debreceny, Farewell, Piechocki, Felden, & Gräning (2010) for example, found that a quarter of all the SEC-XBRL filings up to September 1, 2009 contained calculation errors. Bartley, Chen, and Taylor (2011) support these claims of errors in XBRL filings, however, they found that the errors in XBRL filings have decreased between the time of the implementation of the SEC’s Voluntary XBRL Filing Program (VFP) and the implementation of SEC’s XBRL Mandatory Program. Markelevich, Shaw, and Weihs (2015) also show in their analysis of the Israeli mandated XBRL adoption that the adoption was accompanied by deficiencies in the XBRL-tagged filings and inconsistencies between these XBRL-filings and the traditional annual financial reports. They also observed pervasive data entry errors which went undetected for over a year. The studies of Debreceny et al. (2010), Bartley et al. (2011) and Markelevich et al. (2015) suggest that the quality of XBRL data is an essential characteristic that contributes to the successful spreading of the innovation.

Another concern of XBRL with regard to its impact on business is the comparability of financial statements. Although many studies highlight improved comparability as one of the advantages of XBRL (e.g., Roohani, Furusho, & Koizumi, 2009; Vasarhelyi et al., 2012), other studies argue that due to the possibility to extend or add new elements to the existing taxonomy, problems related to comparability arise. Debreceny and Farewell (2010) for example, find that enabling extensions hinder comparability. When companies consider the preexisting taxonomy as inadequate to represent a line item in their financial statements, XBRL provides firms with flexibility by using custom tags, or extensions. However, Debreceny and Farewell (2010) argue that the use of custom tags can also be attributed to the “mismapping” of financial statement line items to the U.S. GAAP taxonomy. In addition, Dohle, Lobo, Mishra, & Pal (2015) find that financial statement comparability declined in the initial years after SEC mandate. They also give the use of company-specific taxonomies as a possible explanation. These conflicting arguments with regard to comparability of XBRL filings had led to a debate on how to manage

(12)

12

the trade-off between comparability and flexibility. Bonsón, Cartijo, and Escobar (2009) for instance, support the advantages of XBRL’s comparability and argue that XBRL taxonomies should be aligned with each other by developing a common set of global accounting standards. They state that whether the full potential of XBRL can be realized depends on the extent of harmonization or even standardization of a global taxonomy. Zhu and Wu (2010) support this by arguing that if companies would use a standard taxonomy provided by the FASB, the financial statements would be more comparable.

Earnings management

Within the field of accounting research, earnings management is a topic that is researched extensively. One definition of earnings management that is often used in previous studies is the definition of Schipper (1989): “Earnings management is really disclosure management in the sense of a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain, as opposed to merely facilitating the neutral operation of the process” (p. 92). Earnings management is thus defined as an intervention of management in the financial reporting process to further a private gain, either for management itself, or for the firm as a whole. Financial accounting data is adjusted in order to improve on the outcome of the reporting process.

In general, prior research suggests that there are two alternative ways to manage earnings: accrual-based earnings management and real earnings management. The existence of accrual-based earnings can be explained by the role that accruals have in accounting. According to Dechow (1994), the role of accruals is “to overcome problems with measuring firm performance when firms are in continuous operation” (p. 4). Accruals can mitigate these problems by altering the timing of cash flows recognition in earnings, in line with the revenue recognition principle and the matching principle, as evolved by generally accepted accounting principles (Dechow, 2004). Dechow (2004) however, notes that the use of accruals initiated a

(13)

13

new set of problems: Management has discretion over the recognition of accruals, and this discretion can be used to signal management’s private information or to opportunistically manipulate earnings. There are different techniques that can be used to manage earnings through accruals, such as ‘Big Bath technique’ and ‘Cookie Jar Reserve technique’ (McKee, 2005). Big bath techniques are based on the belief by managers that if you have bad news to report, you better can report all the bad news at once and estimate the losses on the high side to improve the result in the next accounting period. The cookie jar reserve technique involves managers’ judgement by overestimating future expenses with the hope that actual expenses will be lower than the provisions. Hereby a reserve is created that might boost future performance. Examples of common areas where cookie jar reserves are created, are in estimating sales returns and allowances, estimating bad debt write-offs, and estimating inventory write-downs (McKee, 2005).

Accrual-based earnings management thus, is based on accounting choices that managers make with the intention to deliberately bias accounting information, namely by discretionary accrual manipulation. Dichev, Graham, Harvey, and Rajgopal (2013) provide insights into the incentives of managers to engage in accrual-based earnings management from a survey among 169 CFO’s. They find that CFO’s rate the following motivations for using accrual-based earnings management as most important: (1) to influence stock price, (2) because there is outside pressure to hit earnings benchmarks, (3) because there is inside pressure to hit earnings benchmarks, and (4) to influence executive compensations. To illustrate, Cheng and Warfield (2005) report that managers use discretionary accruals to report earnings that meet or just beat analysts’ forecasts, especially for firms with higher managerial equity incentives. Another example of accrual-based earnings management can be found in the study of Shivakumar (2000). He shows that companies report abnormally high accruals before seasoned equity offerings.

(14)

14

Although real earnings management has been less studied related to accrual-based earnings management, Graham, Harvey, and Rajgopal (2005) find in their survey and interviews among 400 executives, that managers rather manipulate earnings through real activities than by using accruals. They explain this tendency to use real earnings management as a consequence of the Sarbanes-Oxley Act, which constrained accrual-based earnings management. Real earnings management is the manipulation of normal operational practices, driven by managers’ intention to deliberately influence accounting information to mislead some stakeholders into believing that certain financial reporting objectives have been met in the normal course of operations (Roychowdhury, 2006). Roychowdhury (2006) however, states that the manipulation of real activities in the current period can have a negative effect on future cash flows and therefore reduce firm value. Gunny (2005) confirms this by finding that real earnings management activities have a significantly negative impact on future operating performance. However, contrary to Gunny (2005), Taylor and Xu (2010) provide evidence that firms, on average, avoid manipulating their operations to such an extent as to cause significant negative impacts on their subsequent operating performances. Nevertheless, Graham et al. (2005) find that 78% of their sample admits to sacrifice long-term firm value to smooth earnings in order to meet earnings targets such as zero earnings, previous period’s earnings, and analyst forecasts. Roychowdhury (2006) gives two possible explanations for this preference for real earnings management. First, the manipulation of accruals is more likely to be placed under auditors’ or regulators’ scrutiny than real decisions about normal operational practices. Real earnings management on the other hand cannot influence auditors’ opinions or regulators’ actions, as long it is correctly disclosed in the financial statements. Second, relying solely on accrual manipulation is risky. For instance, if the difference between unmanipulated earnings and the desired earnings threshold at year-end is too large to be exclusively manipulated by accruals, there is no opportunity anymore to manipulate earnings with real activities. A

(15)

15

repeatedly documented example of an actual decision used by managers to influence accounting information is the opportunistic reduction or delay of research and development (R&D) expenditures (e.g., Baber, Fairfield, & Haggard, 1991; Bens, Nagar, & Franco Wong, 2002; Bushee, 1998; Dechow & Sloan, 1991). Other examples are: price discounts, acceleration of sales, alterations in shipment schedules, and delaying of maintenance expenditures (e.g., Graham et al., 2005).

Corporate transparency and XBRL

Information asymmetry or the lack of transparency is one of the conditions that gives rise to earnings management (e.g., Schipper, 1989). This, supported by the study of Maines and McDaniels (2000) for example, suggests that the ease and clarity of financial information is of influence when investors or other stakeholders have limited information-processing ability. Hirst and Hopkins (1998) underline this by finding that investors have less difficulty in recognizing earnings management when financial information is summarized and easy to process (i.e., comprehensive income reporting as required by SFAS No. 130). This suggests that more transparent disclosures facilitate market participants in detecting accrual-based earnings management. In line with the aforementioned studies, Huang and Zhang (2012) also find evidence for enhanced transparency achieved through greater disclosure. Their results show that extensive disclosure impairs managers’ abilities to utilize corporate resources in a self-serving manner. They also suggest that this enhanced transparency allows market participants to monitor and scrutinize managerial opportunism more effectively and at a lower cost.

Although financial reports in XBRL format do not provide new information beyond the information that would have been reported in the traditional format, XBRL can improve financial information transparency and therefore can enable investors to scrutinize managerial opportunism (e.g., Li et al, 2006; Yoon et al., 2011). In addition, XBRL has the potential to

(16)

16

“increase the speed, accuracy and usability of financial disclosure and eventually reduce costs for investors” (SEC, 2008). Hodge et al. (2004) provided the first empirical evidence for this expectation with their experiment among 96 nonprofessional users of financial reports. They found that participants who used XBRL were better able to acquire and integrate financial information than participants who did not use an XBRL-enhanced search engine. Their results further imply that XBRL facilitates managers in making comparisons between firms, and therefore improves the transparency of managers’ financial reporting choices and the effects of those choices.

Research findings with regard to the effects of XBRL on information asymmetry in the U.S. are mixed. Blankespoor et al. (2014) for example, report an increase in information asymmetry during the first years following the SEC mandate. They find that XBRL increases bid-ask spread, liquidity, and trading volume. On the other hand, Kim et al. (2012) claim that mandated XBRL adoption would reduce information asymmetry as a result of decreased information risk in the equity market, especially for outside, uninformed investors. They argue that this decreased information risk is achieved through improved accessibility and transparency of financial information. In addition, Li et al. (2012) find evidence for their prediction that XBRL adoption leads to a reduction in information processing costs and this reduces the cost of equity capital significantly. They report an improved information environment through larger analyst coverage and more stock liquidity. They interpret these results as support for XBRL’s contribution in lowering information asymmetry. In the studies of Blankespoor et al. (2014), Kim et al. (2012), and Li et al. (2012) however, the effect of firm size is not emphasized. Geiger, North, and Selby (2014) therefore, underscore the effect of firm size in their research on the impact of XBRL on information asymmetry for early U.S. adopters. They find some support of reduced information asymmetry, but strong evidence that the effect is greater for larger XBRL adopting firms (significant decrease of bid-ask spreads and

(17)

17

significant increase of trading volume). They therefore argue that whether XBRL affects information asymmetry is a matter of firm size.

XBRL and earnings management

Based on the close relationship between information asymmetry, corporate transparency and earnings management, and motivated by the reported effects of XBRL adoption on corporate transparency, Peng et al. (2011) predict that XBRL adoption incentivizes companies to improve the quality of their reported information, because investors are now better able to place financial reports under more scrutiny. As a result, investors should be able to conduct analyses to a greater extent and in a more efficient way. In particular, Peng et al. (2011) predict that XBRL adoption may discourage companies from engaging in earnings management. To test this expectation, they examine if the mandatory XBRL adoption by firms in China is related to the level of total accruals that firms report. They compare the reported accruals in pre- versus post-XBRL periods. They find that the level of total accruals in the post-XBRL period is lower relative to the pre-XBRL period. Although looking solely at accrual-based earnings management, they interpret these findings as the investors’ improved ability to detect earnings management because of the adoption of XBRL.

Peng et al. (2011) mention that the environment in China, which they used in their research, provides a unique opportunity to examine whether the implementation of XBRL changes the level of reported accruals. In their view, Chinese companies lack strong corporate governance structure and have an increased tendency in engaging in dubious accounting practices. They also state that corporate disclosure environment in China is less stringent than other major world economies, such as the United States. Furthermore, they argue that the reporting environment in China is more structured. Therefore, they think that the results of their study are of limited generalizability. Besides, they expect the findings to be difficult to transfer

(18)

18

to for example the United States because political and cultural differences may cause the investment environment and functioning of capital markets to be quite different.

In sum, XBRL adoption has the ability to decrease information asymmetry and to enhance the transparency and accessibility of financial information. Due to this improved information environment relative to non-interactive formats, XBRL formatted financial reports may enable investors to easier detect opportunistic earnings manipulation, and at a lower cost. In accordance with Peng et al. (2011), managers might as a result be discouraged to manipulate earnings through accruals. Although Peng et al. (2011) mention that their results may not be applicable to XBRL filings in the U.S., I nevertheless expect that the mandated XBRL adoption discouraged managers in the U.S. to engage in accrual-based earnings management as well. This expectation is based on the results of Geiger et al. (2014), Kim et al. (2012), and Li et al. (2012). These studies all reported either a decrease in information asymmetry or an increase in transparency of financial information after the SEC mandate. Similarly and complementary to these studies, it is therefore it expected that mandated XBRL adoption in the U.S. has led to a decrease in accrual-based earnings management. To provide empirical evidence on this expectation, I test the following hypothesis:

H1: Discretionary accruals decreases from the pre-XBRL adoption period to the post-XBRL adoption period.

Where Peng et al. (2011) only focused on accrual-based earnings management, I also examine the relationship between mandatory XBRL adoption and real earnings management. As reported by Graham et al. (2005), managers manipulate earnings through real activities rather than by using accruals, when the latter is constrained. Roychowdhury (2006) explains this by stating that financial reporting users are less likely to place decisions about normal activities under scrutiny than accrual manipulation. Based on the aforementioned studies about increased transparency and investors’ ability of recognizing accrual-based earnings

(19)

19

management (e.g., Huang & Zhang, 2012), it is likely that managers would even have a greater tendency to shift to real earnings management in a situation where accrual-based earnings management is constrained due to a more transparent financial reporting environment (e.g., XBRL-formatted financial reports). Zang (2012) strengthens this suggestion by stating that accrual-based earnings management and real earnings management function as substitutes. She analyzes the implications for managers’ trade-off decisions due to the different costs that result from both earnings management strategies. After examining a sample of 6,500 earnings management suspect firm-years over the period 1987-2008, she finds significant positive relations between the level of real activities manipulation and the costs associated with accrual-based earnings management, and also between the level of accrual-accrual-based earnings management and the costs associated with real activities manipulation. This indicates that managers tradeoff the two earnings management strategies based on the relative costs for their firms.

More studies found evidence for this substitutive relationship between accrual-based earnings management and real earnings management. Cohen and Zarowin (2010) for example, show that in the year of a seasoned equity offering (SEO), the tendency to use real earnings management is positively related to various costs of accrual-based earnings management. Cohen, Dey, and Lys (2008) show that after the enactment of SOX, heightened scrutiny of accounting practice led to a decline in accrual-based earnings management, while real earnings management increased. Chi, Lisic, and Pevzner (2011) report a similar effect after examining the association between audit quality and real earnings management. They find that higher levels of real earnings management is associated with higher audit quality, and they argue that firms are more likely to engage in more extensive real earnings management when their ability to manage accruals is constrained. In a more recent study, Enomoto, Kimura, and Yamaguchi (2015) support this by showing that when the ability to manage accruals is constrained by

(20)

20

stronger investor protection, managers tend to engage in real earnings management instead of accrual-based earnings management.

Under the maintained hypothesis that real earnings management functions as a substitute for accrual-based earnings management, I predict that earnings management through real activities manipulation has increased since the mandatory XBRL adoption. Therefore, my second hypothesis is:

H2: Earnings management through real activities manipulation increases from the pre-XBRL adoption period to the post-pre-XBRL adoption period.

III. Research methodology Sample selection

To test my hypotheses, I require a sample of all firms that submitted their financial reports in XBRL-format since the SEC mandate has been in force on 15 June, 2009. As recommended by Kim et al. (2012), I first use the EDGAR Dashboard to obtain all the interactive data 10-K submissions (excluding trusts and funds) that were filed since 15 June, 2009. I use annual 10-K filings because the incentives to manage earnings may be higher for annual earnings since fiscal year-ends are the most important reporting dates (Oyer, 1998). Table 1 summarizes how the SEC (2009) identified the different phase-in groups for the XBRL-adoption. Since the pre- and post XBRL-adoption period varies by each different phase-in group, I require a sample period that is large enough to be able to determine a pre- and post XBRL-adoption period for each different filer category (i.e., phase-in group). Therefore I obtain all annual financial data between 1 January 2006 and 14 June 2014 from the Compustat database. By taking 14 June 2014 as the end of my sample period, I have exactly five years of data since the mandate has been in force (i.e., for Phase I firms). Similarly, the post-XBRL period for Phase II and III firms is exactly four and three years, respectively.

(21)

21

[Insert Table 1 about here]

Information of each filing from the EDGAR Dashboard includes EDGAR URL, filing date, type of filing and company information such as company name, CIK, SIC and the filer category (i.e., large accelerated filer, accelerated filer, or smaller reporting company) based on the different phase-in groups as identified by the SEC (2009). In order to be able to merge all the mandatory XBRL 10-K filings with the correct firm-year from Compustat, the corresponding fiscal year is identified for each 10-K filing. This is determined based on the filing date of each 10-K filing, the fiscal year-end of the filing company, and the deadlines for submitting 10-K filings, as required by the SEC (2009). First, all the filing dates are deducted with the SEC’s deadline requirements (i.e., 60 days, 75 days, and 90 days for large accelerated filers, accelerated filers, and smaller reporting companies, respectively) to estimate the corresponding year-end of each 10-K filing. Then, all the actual fiscal year-ends are obtained from Compustat and compared with the estimates. When the estimate and the actual year-ends are within a range of 30 days (to correct for early and late filers), the estimate is equated to the actual year-end, which enables me to determine the fiscal year of each XBRL filing. Finally, based on fiscal year, I merge the Compustat database with the population of XBRL filings.

Of the initial population of annual XBRL filings, filed by 8,246 firms between June 2009 and June 2014, I drop all the firms for which no fiscal year-end could be obtained or because of missing firm-years in Compustat. After merging the XBRL filers database with Compustat for the sample period January 2006 to June 2014, I drop duplicate fiscal years for the same firms. In addition I eliminate all observations with fiscal year 2014, because theoretically this fiscal year cannot be part of data from a sample period that ends on 14 June 2014. A sample of 49,336 firm-years remains. I exclude observations with insufficient data in

(22)

22

order to compute the earnings management variables (26,775 observations dropped). To prevent regression model estimates to be imprecise, I follow Roychowdhury (2006) and Zang (2012) by excluding all firm-year observations where there are fewer than 15 observations in any two-digit SIC code in a given year (1,686 observations). Consistent with prior literature on earnings management (e.g., Cheng & Warfield, 2005; Roychowdhury, 2006; Zang, 2012), I eliminate firms in regulated industries (SIC codes between 4400 and 5000) and financial institutions and banks (SIC codes between 6000 and 6999) because managers in these industries might have different motivations to manage earnings since their firms are governed by different accounting rules (2,208 observations).

By using firms after the mandated XBRL-adoption as my treatment sample and the same firms in the pre-adoption period as my control sample, I try to limit the possibility that the observed effects arise from other uncontrolled characteristics than the mandated XBRL-adoption. Therefore, I exclude all firms that have either solely pre-XBRL firm-year observations or solely post-XBRL firm-year observations (525 firms). Firms that participated in the SEC’s Voluntary XBRL Filing Program (VFP) are also excluded because those firms already reported in XBRL-format before the mandatory XBRL-requirement (24 firms). For managers of these firms, as identified by using the study of Callaghan and Nehmer (2009), XBRL-adoption may not be an incentive anymore to adjust their behavior regarding earnings management. The final sample, for the period between fiscal years 2005 and 2013, consists of 2,477 firms and 17,645 firm-year observations. The sample development procedures are summarized in Table 2.

(23)

23 Empirical design

Measurement of accrual-based earnings management

Following prior literature, I use discretionary accruals to proxy for the extent of a firm’s engagement in accrual-based earnings management. Discretionary accruals are the total accruals deducted by the normal level of accruals. One of the most commonly used models to capture accrual-based earnings management is the Modified Jones (1991) model, as proposed by Dechow, Sloan, and Sweeney (1995). Although more recent models exist (e.g., Kothari, Leone, & Wasley, 2005; Larcker & Richardson, 2004), these models contain only slight additions to the model of Dechow et al. (1995). Moreover, some of the most recent studies still use this discretionary accrual model (e.g., Enomoto et al., 2015; Francis, Michas, & Seavey, 2013). Therefore, in order to estimate the level of discretionary accruals, I follow prior literature by using the commonly used model by Dechow et al. (1995):

TACjt /TAjt – 1 = α0 + α1[1/ TAjt – 1] + α2[(ΔREVjt – ΔRECjt)/ TAjt – 1]

+ α3[PPEjt / TAjt – 1] + εjt (1)

where, for firm j and year t, TAC represents total accruals (income before extraordinary items minus cash flow from operations), TA represent the total value of assets, and ΔREV and ΔREC denote the absolute increase or decrease in net sales and net receivables, respectively. PPE is the total gross value of property, plant, and equipment. The use of assets as a deflator is intended to mitigate heteroscedasticity in residuals. As recommended by Kothari et al. (2005), I also include the constant in the above model. Kothari et al. (2005) claim that including a constant in the model estimation is an additional control for heteroscedasticity and it enables the researcher to further mitigate model misspecifications and to address the power of the test. They find that discretionary accrual measures based on models without a constant term are less symmetric, making power of the test comparisons less evident. I estimate the regression of Equation (1)

(24)

24

cross-sectional by each two-digit SIC industry and year, such that the estimated coefficients vary over time and reflect the industry-wide economic conditions during a year. Following Zang (2012), Equation (1) is only estimated for two-digit SIC industries and years with at least 15 observations. My proxy for accrual-based earnings management, denoted by ABS_DA, is the absolute value of the estimated residuals ε.

For robustness purposes, I use an alternative accrual-based earnings management measure that controls for performance, as suggested by Kothari et al. (2005). They argue that accruals of firms with unusual performance are expected to be systematically non-zero, and therefore accruals are correlated with firm performance. They examine two ways to control for performance in estimating discretionary accruals. One is to adjust a firm’s estimated discretionary accruals by subtracting the corresponding discretionary accruals of a firm in the same two-digit SIC industry, matched on the basis of the closest ROA in the current year. Alternatively, a performance variable such as ROA could be included as an additional independent variable in a discretionary accruals regression. I use this second approach by including ROA in the Modified Jones model of Dechow et al. (1995) of Equation (1):

TACjt /TAjt – 1 = α0 + α1[1/ TAjt – 1] + α2[(ΔREVjt – ΔRECjt)/ TAjt – 1]

+ α3[PPEjt / TAjt – 1] + α4ROAjt + εjt (2)

where, for firm j and year t, ROA is the return on assets, calculated as net income scaled by lagged total assets. Similarly to Equation (1), I estimate Equation (2) for two-digit SIC industries and years with at least 15 observations. The absolute values of the estimated residuals (ε), capturing discretionary accruals, are my proxy for accrual-based earnings management (denoted as AB_ROA).

(25)

25 Measurement of real earnings management

I rely on prior studies to develop my proxies for real earnings management. Roychowdhury (2006) considers the abnormal levels of cash flow from operations, abnormal level of production costs, and abnormal level of discretionary expenses to study the level of real activities manipulation. In this study, I exclude abnormal cash flow from operations as a proxy because, as discussed by Roychowdhury (2006), abnormal cash flow from operations can be affected by real activities manipulation in different directions, and therefore the net effect can be ambiguous. Subsequent studies (e.g., Cohen et al., 2008; Cohen and Zarowin, 2010; Zang, 2012) provide evidence that the other two measures, abnormal production costs and abnormal discretionary expenditures, capture real earnings management.

The first real activities manipulation method I focus on, is the reporting of lower cost of goods sold (COGS) through increased production. By increasing the level of production, overhead costs can be spread over a larger number of units, which can result in a decrease of fixed costs per unit and a thus a higher operating margin. Hence, managers can increase earnings through increasing the level of production more than necessary. However, in the longer term firms may have lower cash flows from operations because they incur higher annual production costs relative to their level of sales.

To capture the abnormal level of production costs, I first estimate the normal level of production costs following Roychowdhury (2006):

PRODjt / TAjt – 1 = α0 + α1[1/ TAjt – 1] + α2[REVjt / TAjt – 1] + α3[ΔREVjt / TAjt – 1]

+ α4[ΔREVjt – 1/ TAjt – 1] + εjt (3)

where, for firm j and year t, PROD is the sum of the cost of goods sold and the net change in inventory, and REV represents the net sales. I estimate the regression of Equation (3) cross-sectional for each two-digit SIC industry and year with at least 15 observations. The abnormal

(26)

26

level of productions costs, denoted by R_PROD, is measured as the estimated residuals ε. The higher the residual, the larger is the amount of overproduction, and the greater is the increase of earnings through reducing the cost of goods sold.

Real activities manipulation through abnormal levels of discretionary expenses is attained by decreasing discretionary expenditures that include research and development (R&D); advertising; and selling, general, and administrative (SG&A) expenditures. Short-term earnings can be increased by reducing these expenditures. When these expenditures usually are paid for in cash, it also could lead to higher current period cash flows. However, the inherent risk is that future cash flows may be lower.

Following Roychowdhury (2006), I estimate the normal level of discretionary expenditures with the following equation:

DISXjt / TAjt – 1 = α0 + α1[1/ TAjt – 1] + α2[ΔREVjt – 1/ TAjt – 1] + εjt (4)

where, for firm j and year t, DISX is the discretionary expenditures (i.e., the sum of R&D, advertising, and SG&A expenditures). Following Cohen et al. (2008), R&D and advertising expenditures are set to zero if they are missing, as long as SG&A expenditure is available. Equation (4) is also estimated cross-sectional for each two-digit SIC industry and year with at least 15 observations. The abnormal level of discretionary expenditures, denoted by R_DISX, is then measured as the estimated residuals from the regression. The higher the residuals, the less indication there is of amounts of discretionary expenditures cut by firms to increase reported earnings.

In order to capture the effect of real earnings management through both measures in one comprehensive measure, I aggregate the two real activities manipulation measures into one main proxy, denoted by R_PROXY, by taking their sum. However, the two individual variables have different implications for real earnings management; an increase in R_PROD

(27)

27

indicates an increased tendency to real earnings management, whereas an increase in R_DISX indicates a decreased tendency to real earnings management. As recommended by Zang (2012), I therefore multiply R_DISX with -1 such that higher values of R_DISX indicate higher tendency to real earnings management as well. This adjustment prevents dilutive results when using R_PROXY. I report results corresponding to this main real earnings

management proxy, but report the two individual real earnings management proxies as well as a robustness check.

Model specification

Accrual-based earnings management

My first hypothesis (H1) predicts that the mandatory XBRL-adoption has led to a decrease in accrual-based earnings management in the post-XBRL period, relative to the period prior to the implementation. To test this hypothesis, I estimate the following ordinary least squares (OLS) regression model, which derives from the model used by Peng et al. (2011):

AB_EM = β0 + β1XBRL+ β2GDP + β3CPI + β4DEFICIT + β5ΔREV + β6PPE

+ β7 LAG_ROA + β8 CFO + β9SIZE + β10 MB + β11 LEV + ε (5)

In the above, AB_EM refers to AB_MJ or AB_ROA. My main independent variable is the indicator variable XBRL, which is equal to 1 in the post-XBRL period and equal to 0 in the pre-XBRL period. To confirm my empirical prediction of H1, I expect the estimated coefficient for

XBRL (β1) to be negative.

To control for the general trend of the U.S. economy that could contribute to the change in the level of total accruals and discretionary accruals, I include three additional variables that reflect macroeconomic-activity-related forces that occur during my sample period. Following Peng et al. (2011), I include three variables that capture the general economy in the United

(28)

28

States: GDP, which is the natural log of annual Gross Domestic Product (GDP); CPI, which is the annual consumer price index (measured on base year 2005); and DEFICIT, which is the natural log of trade deficit (total imports minus total exports). I obtain these macro-economic variables from the World Bank website.

To further isolate the effect of XBRL-adoption on accrual-based earnings management, I follow prior accrual-based earnings management literature (e.g., Dechow et al., 1995; Dechow & Dichev, 2002) by controlling for firm’s economic fundamentals that capture its nondiscretionary component of accruals: ΔREV, which is the change in net sales, scaled by lagged total assets; PPE, which is the gross property, plant, and equipment, scaled by lagged total assets; LAG_ROA, which is the previous accounting performance, measured as the lagged ROA (net income scaled by total assets); and CFO, which is cash flow from operations, scaled by lagged total assets.

I also control for firm size (SIZE), growth opportunities (MB), and leverage (LEV). Following Zang (2012), I measure SIZE as the industry-adjusted log value of lagged total assets to specifically control for the relative firm size in the industry. Prior research (e.g., Dechow & Dichev, 2002) suggests that firm size and accruals are related because larger firms tend to be more conservative in accounting decisions, and such conservatism can be reflected in accounting accruals. Market-to-book (MB, calculated as market value of equity divided by the book value of equity) is included to control for the effect of a firm’s growth opportunities, because high-growth firms are more likely to meet or beat analysts’ forecasts (Skinner & Sloan, 2002) and therefore have an increased tendency to engage in accrual-based or real earnings management. LEV, which is measured as the ratio of long-term debt to lagged total assets, is included because managers of highly levered firms tend to manipulate accruals because of their concerns of debt covenant violations (e.g., DeFond & Jiambalvo, 1994).

(29)

29 Real earnings management

My second hypothesis (H2) predicts that the mandatory XBRL-adoption has led to an increase in real earnings management in the post-XBRL period, relative to the period prior to the implementation. To test this hypothesis, I estimate the following ordinary least squares (OLS) regression model, which derives from the models of Roychowdhury (2006) and Zang (2012):

R_EM = β0 + β1XBRL+ β2GDP + β3CPI + β4DEFICIT + β5MVE

+ β6MB + β7 ROA + ε (6)

In the above, R_EM refers to either one of the two individual real earnings management measures (R_PROD and R_DISX), or to the aggregated proxy R_PROXY. Similar to Equation (5), the main independent variable is the indicator variable XBRL, which is equal to 1 in the post-XBRL period and equal to 0 in the pre-XBRL period. To confirm my empirical prediction of H2, I expect the estimated coefficient for XBRL (β1) to be positive.

To improve the comparability between real and accrual-based earnings management models, I also add the from Peng et al. (2011) derived variables GDP, CPI, and DEFICIT in Equation (6) to capture the general macroeconomic trend in the U.S. during my sample period.

MB is included to control for firms’ growth opportunities. The variables are as defined above.

Following Roychowdhury (2006) and Zang (2012), I further include MVE (measured as the log of market equity value) and ROA (measured as net income scaled by lagged total assets) to control for systematic variation in abnormal production costs and abnormal discretionary expenditures correlated with size and current-period firm performance, respectively.

Because I take a cross-sectional approach (i.e., the dependent variables are estimated for each industry-year) in Models (1), (2), (3) and (4), no year dummy has to be included to control for fixed year effects in both Model (5) and (6). Following the explanation used in prior

(30)

30

literature (e.g., Cohen et al., 2008; DeFond & Jiambalvo, 1994), the advantage of a cross-sectional approach is to control for industry-wide economic circumstances, while allowing the coefficients to change across years as a result of possible structural changes. Since this approach estimates coefficients in a given year, the assumption that coefficients are stable across different years is avoided.

In estimating both Equation (5) and (6), I winsorize all continuous variables at the 1st and 99th percentiles of their distributions to reduce the possibility that my findings are driven by outliers. Standard errors are robust for heteroscedasticity, and following Petersen (2009), I estimated standard errors clustered on firm level to account for firm effects in pooled regressions.

IV. Results Univariate analyses

Panel A and Panel B of Table 3 provide summaries of the sample by two-digit SIC industry and by filer category, respectively. The highest number of firms that is represented in the sample of 17,645 10-K filings, is from the Metal, Machinery and Equipment, and Instruments industry, followed by the Business Service, Auto Repair, and Recreation industry. Among the XBRL filers, most firms are categorized as a Large Accelerated Filer, followed by firms categorized as Accelerated Filers.

[Insert Table 3 about here]

Table 4 presents descriptive statistics of the key variables in my main tests. Panel A summarizes the full sample period. The mean discretionary accruals are 0.079 and 0.062 as calculated by the Modified Jones of Dechow et al. (1995) (AB_MJ) and by the performance

(31)

31

adjusted model of Kothari et al. (2005) (AB_ROA), respectively. For comparison, recent literature that make use of the same discretionary accrual measure, such as Demirkan, Radhakrishnan, and Urcan (2012), find a mean of 0.074. The mean level of both abnormal production costs (R_PROD) and abnormal discretionary expenditures (R_DISX) is 0.001. As a result, the mean total level of real activities manipulation (R_PROXY) is 0.002. This is relatively low compared to prior literature such as Zang (2012), who finds a mean of 0.090. However, comparison of these means is complicated by the fact that Zang (2012) uses a sample period between 1987 and 2008.

During my total sample period, firms experience an average increase in net sales (ΔREV) of 8.7 percent. For an average firm, 54.2 percent of total assets consists of property, plant, and equipment (PPE), and 26.6 percent of assets is long-term debt (LEV). Panel A of Table 4 further reports that for an average firm, cash flows from operations is 5.1 percent, previous accounting profitability is -5.6 percent, and current profitability is -4.9 percent of total assets. The mean market-to-book ratio (MB) is 2.691.

[Insert Table 4 about here]

Panel B of Table 4 presents the descriptive statistics comparing the pre XBRL-adoption period with the post XBRL- adoption period. AB_MJ and AB_ROA have a lower mean in the post XBRL-adoption period (0.070 and 0.055) than during the pre XBRL-adoption period (0.086 and 0.067); the difference is statistically significant at the 1 percent level, suggesting that the implementation of XBRL constrains managerial opportunism through accruals-based earnings management. In contrast to the measures of accrual-based earnings management, the proxies for real earnings management show a higher mean in the post XBRL-period than during

(32)

32

the pre XBRL-period, except for R_PROD. The mean R_PROD is -0.002 in the pre XBRL period, and -0.000 in the post XBRL period; however, the difference is not statistically significant. The mean of R_DISX on the other hand, is significantly higher in the post XBRL period (0.009) than during the pre XBRL period (-0.006). R_PROXY, as a proxy for the total level of real earnings management and calculated as the sum of R_PROD and R_DISX, is significant higher in the post XBRL period (0.009) than during the pre XBRL period (-0.004). This comparison indicates that the implementation of XBRL has led to an increase in managerial opportunism through real earnings management.

With respect to the control variables, I further find that firms report lower change in net sales (ΔREV) in the post XBRL period relative to the pre XBRL period, while firms have increased their property, plant, and equipment (PPE); have increased profitability in both previous (LAG_ROA) and current accounting period (ROA); have higher cash flows from operations (CFO); have higher growth potential (MB); tend to be larger (SIZE and MVE); and have increased their leverage ratio (LEV). All these differences are statistically significant at the 1 percent level.

Table 5 provides the Pearson correlations among the variables in my main tests. There is a positively high and significant correlation between R_PROD and R_DISX (Pearson correlation of 0.537, p = 0.000), suggesting that firms use both ways of real activities manipulation at the same time. AB_MJ and AB_ROA are also positively correlated with each other (0.715, p = 0.000), indicating that the Modified Jones model by Dechow et al. (1995) and the Kothari et al. (2005) model reflect similar outputs with regard to accrual-based earnings management. Both AB_MJ and AB_ROA are significantly and negatively correlated with the proxy of total real earnings management, R_PROXY (-0.125 and -0.138, respectively). Consistent with prior literature (e.g., Cohen et al., 2008; Zang, 2012), this suggests that accrual-based earnings management and real earnings management function as substitutes. Moreover,

(33)

33

AB_MJ and AB_ROA are significantly and negatively correlated with the indicator variable XBRL (-0.080 and -0.084, respectively), providing initial univariate evidence that discretionary

accruals are lower in the post XBRL-adoption period than in the pre XBRL-adoption period. With respect to real earnings management, only R_DISX has a significant correlation with

XBRL (0.025, p = 0.001), suggesting that firms tend to report lower abnormal discretionary

expenditures (i.e., increased tendency to real earnings management) in the post XBRL-adoption period than during the pre XBRL-adoption period.

[Insert Table 5 about here]

Consistent with prior literature such as Dechow et al. (1995) and Kothari et al. (2005), discretionary accruals (AB_MJ and AB_ROA) are negatively related to a firm’s size (SIZE and

MVE) and operating cash flows (CFO), and positively related to a firm’s change in net sales

(ΔREV). Discretionary accruals are also correlated with a firm’s current (ROA) and previous (LAG_ROA) performance. Although I attempt to control for firm performance in AB_ROA by following Kothari et al. (2005), I nevertheless find that both ROA and LAG_ROA are significant correlated with AB_ROA (-0.283 and -0.267, respectively). Furthermore, LAG_ROA, CFO, and ROA are high and significantly correlated to SIZE (and MVE), as larger firms are likely to enjoy better performance and higher operating cash flows.

Although not tabulated, I find that the variance inflation factor (VIF) for the main variables of interest are low (average VIF = 1.75), suggesting that there is no multicollinearity problem.

(34)

34 Multivariate analyses

Table 6 presents the results of the effect of mandatory XBRL-adoption on accrual-based earnings management and real earnings management in regressions (5) and (6), respectively. Hypothesis 1 predicts that mandatory XBRL-adoption decreases accrual-based earnings management. When examining the effect of XBRL-adoption on accrual-based earnings management (i.e., Equation (5)), I find that the coefficient on XBRL is negative and significant at the 1% level for AB_MJ (β1 = -0.013, p = 0.000), which is consistent with my hypothesis

(H1). In other words, firms report 1.3% less abnormal discretionary accruals (as a percentage of total assets) after the mandatory XBRL-adoption, relative to the period prior to the XBRL mandate. Table 6 further shows that this result is robust to the use of the alternative measure of accrual-based earnings management, AB_ROA (β1 = -0.009, p = 0.000). Although the

coefficient on XBRL seems to be low for both AB_MJ and AB_ROA, the magnitude of the effect of XBRL-adoption on accrual-based earnings management is comparable to what Cohen et al. (2008) find for the effect of SOX on accrual-based earnings management (β= -0.019, p = 0.000, where absolute discretionary accruals are also measured by using the Modified Jones model). This suggests that, with regard to managing earnings through accruals, mandated XBRL-adoption caused a constraining effect on U.S. firms that is comparable to the passage of the SOX regulation in 2001. Thus, after a decrease in accrual-based earnings management due to the passage of SOX, the level of abnormal discretionary accruals (as a percentage of total assets) reported by U.S. firms, decreased even further after the mandated XBRL-adoption. These findings imply that mandatory XBRL-adoption causes a decrease in accrual-based earnings management, which is consistent with the view that XBRL improves transparency and information accessibility and therefore enables external users of financial statements to easier detect opportunistic earnings manipulation through accruals.

(35)

35

The results of regression (5) arise after controlling for macroeconomic factors and firm fundamentals during the sample period. I find a significantly negative coefficient for GDP and a significantly positive coefficient for CPI and DEFICIT, suggesting that firms report higher accruals in predominantly weakened macroeconomic conditions. I further find that the coefficients for ΔREV and LEV are significantly positive, while coefficients for LAG_ROA and

SIZE are significantly negative. Consistent with Dechow and Dichev (2002), who find that cash

flows from operations is negatively associated with current period accruals, the coefficient of

CFO is significantly negative.

Although of negligible importance, the total explanatory power of the Model, indicated by Adj. R2,is 22.6%. This is lower compared to the Chinese study of Peng et al. (2011), from which the Model is derived (Adj. R2 = 61.7%). However, this difference can be explained by the fact that Peng et al. (2011) put a few additional independent variables in to control for earnings management incentives that are unique to Chinese listed companies.

[Insert Table 6 about here]

Hypothesis 2 predicts that mandatory XBRL-adoption increases real earnings management. When examining the effect of XBRL-adoption on real earnings management (i.e., Equation (6)), I find that the coefficient on XBRL is positive and significant at the 1% level for the measure of total real earnings management, R_PROXY (β1 = 0.028, p = 0.004), which is

consistent with my hypothesis (H2). Put differently, in the period after the XBRL-adoption, firms use 2.8% more abnormal production costs and discretionary expenses (as a percentage of total assets), relative to the pre-XBRL period. This indicates that mandatory XBRL-adoption causes an increase in real earnings management. However, this coefficient is low compared to

(36)

36

the study of Zang (2012), who finds a coefficient of β =0.197 on the effect of the SOX passage for R_PROXY (as measured in this study). Nevertheless, these findings are in line with the view that firms shift from using accrual-based earnings management to real earnings management, when manipulation of accruals is constrained. The results of Equation (6) are consistent across the two separate real earnings management measures as well for the combined variable

R_PROXY. However, the positive coefficient on XBRL for R_PROD is not significant (β1 =

0.002, p = 0.664), whereas for R_DISX the coefficient on XBRL is positive and significant (β1

= 0.026, p = 0.000). This indicates that the results of R_PROXY are driven by merely one of the two measures of real earnings manipulation. In other words, the increase of real earnings management as an effect of the mandatory XBRL-adoption seems to be caused by increasing reported earnings through cutting discretionary expenditures. This implies that firms consider cutting discretionary expenditures as a more effective way of managing earnings through real activities than by increasing the level of production costs. In contrast to the results of regression (5), I find a significantly positive coefficient on GDP and significantly negative coefficients on

CPI and DEFICIT for regression (6). This suggests that firms are more likely to use real

earnings management in predominantly improved macroeconomic circumstances. Consistent with Zang (2012), I further find that the coefficient of MB is negative and significant, while the coefficient of ROA is positive and significant. This suggest that firms with better growth opportunities have a decreased tendency to manage earnings through real activities and that firms with better current-period performance use more real earnings management.

Overall, my findings are consistent with my hypotheses and show that firms, as a result of the mandatory XBRL-adoption, have a decreased tendency to use discretionary accruals to manage earnings. My results further imply that firms have an increased tendency to use real earnings management when their ability to manage earnings through discretionary accruals is constrained. This implies that increased transparency or constraints over accounting discretion

Referenties

GERELATEERDE DOCUMENTEN

Two studies focus on symptomatic rectocele and internal rectal prolapse; both found a significant reduction of symptoms of ODS/constipation in small cohorts of pa- tients (75 and

Evaluation of influence of historical changes in land use along the middle Vistula river reach on flood risk.. Emilia Karamuz (1), Renata Romanowicz (1), and Martijn

At all three concentrations, we observed similar dif- ferences from control activity development, supporting the conclusion that chronic acylated ghrelin application has a

worldwide. In this paper we present a white-box data leakage detection system to spot anomalies in database transactions. We argue that our approach represents a major leap

The computer system of claim 1, wherein said tabular display of results that allows said user to sort results accord ing to date, relevance, author, Source title, and number of

software tools, should be used to create a clear overview of the patent landscape and to assist with the second process step of the general patent circumvention process:

Furthermore, we have derived pairwise fluctuation terms for the velocities of the fluid blobs using the Fokker-Planck equation, which have been alternatively derived using the

vier faktore naam= Iik: (i) ryping: die ryping van fisiese strukture open die moontlikhede tot denkhandelinge; (ii) ervaring, fi= sies sowel as logies-matematies~