Does XBRL reduce the level of earnings
management?
Student name: Wang Kaidi Student number: 10824340 Date: 22 June, 2015
Word count: 12024
Supervisor: Dr. Ir. SP (Sander) van Triest
MSc Accountancy & Control, specialization Contol
Abstract
Prior research in XBRL area has found out XBRL’s advantages on financial reporting and capital market. These advantages include enhanced relevance, understandability, transparency and comparability of financial statements, reduced information asymmetry between information users and information providers, improved information quality and information efficiency. Findings suggest that adopting XBRL has the potential to decrease discretionary accruals through these advantages. And SEC’s phase-in XBRL mandate offers me a great opportunity to study XBRL’s impacts on earning management. In this study, earnings management is measured by discretionary accruals, and I use cross-sectional modified Jones model to calculate discretionary accruals. I take the advantage of SEC’s phase in schedule and created two groups: one experiment group that consists of firms that were required to file in XBRL format mandatorily, and the other control group consists of firms that didn’t adopt XBRL, I matched each firm in experiment group with one firm that didn't adopt XBRL based on industry and firm size, and I study the differences in the level of discretionary accruals before and after the adoption of XBRL. First I examine XBRL’s impacts on earnings management using sample that only consists of experiment group, then I add matched group into the final sample, both test results suggest that discretionary accruals were reduced after the adoption of XBRL.
Statement of Originality
This document is written by Kaidi Wang 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.
Acknowledgements
Firstly, I would like to thank my thesis supervisor Mr. Sander van Triest for always responding my questions swiftly, guiding me through this whole process, thanks for his patience and professionalism, I feel I have gained a lot knowledge completing this thesis.
Secondly, I would like to thank my colleagues and friends who have been struggling to complete this task. Thanks for their suggestions and encouragements.
Table of Content
1. Introduction ... 4
1.1 Background ... 4
1.2 Research Question ... 8
1.3 Motivation ... 8
2. Literature review and hypothesis ... 9
2.1 About XBRL ... 9
2.1.1 XBRL for data providers ... 9
2.1.2 XBRL for data users ... 12
2.2 How does XBRL impact earnings management? ... 15
2.3 XBRL’s effects on internal control & capital market ... 17
2.4 Earnings management ... 20
2.5 Hypothesis ... 21
3. Research method ... 22
3.1 Sample & Data ... 22
3.2 Research design ... 25
3.3 Research model ... 26
3.4 Variables ... 27
4. Results ... 28
4.1 Descriptive statistics and univariate data analysis ... 28
4.2 Regression results and Pearson correlations ... 29
5. Matched sample analysis ... 31
5.1 Revised research design ... 31
5.2 Results ... 34 6. Conclusions ... 37 7. Limitations ... 39 References: ... 40 Appendix A ... 44 Appendix B ... 56
1. Introduction
1.1 Background
XBRL is short for eXtensible Business Reporting Language. The notion that came up with XBRL is like bar coding for business information, by applying XBRL technology, identified and standardized tags which symbolized as barcodes are tagged to every individual unit of data, therefore each individual item of data is assigned unique and specific information. And computers, software applications can access and analyze the information incorporated in the data by “scanning” the tagged items. Thus, XBRL provides data in an interactive, context-rich format that information users can easily access the data and download the data directly into analytical software (Debreceny et al.2009).
Specifically, the tags incorporated in each individual unit of data define the content and structure of information. And the tags are standardized, makes it easier for information users to compare the information provided by various information providers across various platforms and over a number of years. The way the companies apply XBRL makes it easier for users to retrieve the information by searching the data tags, for example, users can search similar tags and compare different companies’ choices to record similar transactions/items.1 And now, a growing number of countries have begun to include the use of XBRL into their disclosure systems. According to SEC, these countries including Australia, Belgium, China, Ireland, Japan, Singapore, Spain, Sweden, United Kingdom and United States.2 It offers me a great opportunity to study the impacts of XBRL on capital market, I’ll elaborate upon XBRL’s impacts on the degree of earnings management in this case particular, and I choose to use modified Jones model to measure discretionary accruals as the proxy for earnings management. And in this paper, I use the terms “interactive data” and “XBRL” mutually, the definition of interactive data
1https://www.calcbench.com/Content/resources/Peer%20benchmarking%20made%20easy%20with%20XBR
L.pdf
offered by SEC says that data is deemed as interactive data when the data is tagged by computers or other software application markup language, and the data format is machine-readable and can be directly interpreted and analyzed by computers or software applications. Additionally, in the area of business reporting, XBRL plays a role as such a language.1So, in this paper, I refer to XBRL as a technology back-up for achieving interactive data. Besides, I also use the term “earnings management” and “discretional accruals” interchangeably, because in this paper earnings management is measured by discretionary accruals, the level of discretionary accruals represents the level of earnings management.
Two of the most fundamental parts of internal control and capital market are financial statements and business reporting. And XBRL International, an not-for-profit worldwide consortium which develops and maintains the XBRL standard and related specifications and main proponent of XBRL claims that “one of the core capabilities of XBRL is to create digital, unambiguous accurate and reusable versions of financial statement”2. Jensen and Meckling once point out that according to agency theory,
management has the incentive to use fraudulent accounting to boost earnings to a certain extent in financial statements in order to achieve budgets or extract rents from organizations (Jensen and Meckling, 1976). And current accounting standards allow various earnings-manipulating methods for companies to cook the book, firstly, companies are able to choose different accounting methods for different kinds of transactions under a variety of commercial circumstances. Secondly, management can cook the book by using complicated disclosure and obfuscated presentation. For example, managers may choose to accelerate revenues by recording lump-sum revenues as current sales instead amortizing revenues over the years of actual contract time. And management may use different costing method to calculate the cost of inventory. And current accounting standards also allows companies to change accounting standards from period to period, it gives companies more flexibility to manipulate earnings.
1 https://www.sec.gov/spotlight/xbrl/what-‐is-‐idata.shtml
Adopting XBRL, in some ways may reduce earnings management. I’ll explain how XBRL help enhance the quality of financial reporting in two aspects that are relevant to earnings management. Firstly, XBRL tagged data makes financial statements more relevant. The enormous amount of information incorporated in financial statements always makes it difficult for users to retrieve and analyze the data. For example, if an investor wants to have a clear view of a company’s inventory balance and related footnotes disclosure, he has to look up in the balance sheet for inventory balance and he has to look up for related disclosure in another document or from another source. It becomes more complicated if he wants to compare over a particular period or with other companies. Meanwhile, XBRL makes it possible for machines to extract and analyze relevant information from financial statements a direct and automate process. XBRL offers uninformed investors a chance to directly compare information across various platforms and across different firms much more efficiently through taxonomies defined within XBRL language and XBRL provides a richer data set than those data aggregators (Blankespoor ,Miller and White 2014).
On the other hand, XBRL can help enhance understandability for those who are not professional investors by using taxonomy. The definition of taxonomy offered by XBRL International, an organization which develops and maintains the XBRL Standard and related specification, is “Taxonomies are the reporting-area specific hierarchical dictionaries used by the XBRL community.” And taxonomy incorporate the definitions and concepts of each account into the account names in financial statements and at the same time identify the relationship between each account, furthermore, taxonomy helps clarify the connection between accounts and footnotes disclosures. Moreover, taxonomy links abovementioned definitions, concepts and relationships to FASB Codification to enrich the context. Enriched information enables information users a better understanding of the data and relevant disclosures.1 XBRL, in this way, offers the non-professional investors a chance to be at par with professionals, thereby downgrades informational barriers. XBRL technology helps information users understand the data and make the data more useful to them.
At the very beginning, SEC adopted a voluntary approach when promoting its interactive data voluntary program in April, 2005. Participants including operating companies and mutual fund companies who voluntarily file in XBRL format, thus the number of participants was rather limited. Overall, the participants only consists of less than 2% of the companies that upload their financial statements on SEC website and the majority of participants who joined the voluntary program are firms with large size (Efendi, Smith, & Wong, 2009). Then in year 2009,SEC started to mandate XBRL. The main reason that motivates SEC to mandate XBRL is to improve financial information efficiency, data quality, and ultimately to achieve transparency (Booth, 2007). The SEC interactive data mandate was supposed to be implemented over a three-year period and consists of three stages. SEC requires that in the first year of the phase-in schedule, all large accelerated filers that have a total public float above 5 billion dollars as of the end of the second fiscal quarter of fiscal year 2008 and are under US GAAP framework must file in interactive data on or after June 15, 2009. In the second year of the phase-in period, all other large accelerated filers that are using US GAAP must file their annual and quarterly financial reports in interactive data on or after June 15,2010. In the third year of the phase-in schedule, all remaining filers have the obligation to provide their annual and quarterly financial reports using interactive data as of the end of June 15,20111.
As indicated by the phase-in schedule, an increasing number of participants would engage in this interactive data program, which suggests the development of XBRL will be accelerated, firms and information users would be more familiar with this technology, the transparency brought by XBRL would likely to be more significant. Giving the potential that outsiders may gain a lot knowledge of the companies along with the transparency brought by XBRL, it is reasonable to hypothesize that managers may become less likely to be involved in earnings management since it becomes easier for other parties to detect earnings management and they may take the aftermath of reputational costs and the side effects of caught misreporting into consideration.
1 https://www.xbrl.org
This paper exploits SEC’s phase-in XBRL mandate schedule examines the effectiveness of XBRL in mitigating the degree of earnings management within the regulatory and business environment in United States. To define what is earnings management, I used the definition suggested by Healey and Wahlen (1999): “... managers’ use of judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.” And to measure earnings management, I follow prior studies using discretionary accruals as proxy of earnings management and I use cross-sectional modified Jones model to measure discretionary accruals as Dechow, Sloan and Sweeney conclude that a modified version of the model developed by Jones (1991) exhibits the most power in detecting earnings management (Dechow et al.1995). By investigating the connection between XBRL mandate and earnings management, this study makes a significant contribution towards understanding the interacting role of XBRL and helps the diffusion of XBRL in countries that still haven’t mandate XBRL yet.
1.2 Research Question
Does XBRL reduce the level of earnings management?
The purpose of this paper is to analyze whether the mandatory introduction of XBRL had an impact on earnings management, and more precisely, on the level of discretionary accruals. Earnings management is captured by discretionary accruals that are calculated using modified Jones model.
1.3 Motivation
My examination of changes in firms’ earning management activities is motivated in part by the promotion of XBRL and SEC’s mandate of XBRL, XBRL reporting is still in its start-up phase. And it is expected to improve accessibility and transparency for the adopting organizations. It is reasonable to hypothesize that it would have a significant impact on the level of earnings management. I’m also motivated in part by the literature
documenting increasing earnings management.This paper will contribute to the earnings management literature, although the topic earnings management has been studied multiple times in previous literature, relatively little is known about XBRL’s impacts on earnings management. My main objective is to examine whether the degree of earnings management declined after the adoption of XBRL.
The rest of this paper will be structured as follows: the next section will give a brief introduction of how XBRL works, from data providers and data users’ perspectives respectively. Then literature review part will be structured as how XBRL impacts earnings management, XBRL’s impacts on internal control and capital market, and how XBRL is linked to earnings management. Next, I propose my hypothesis according to prior studies and information I gathered. In section 3, I develop my research method and research design. Section 4 describes the results of my first regression. In Section 5, I create a matched group and expend the sample, and the results of revised research designed are displayed. Section 6 reports the overall conclusion and Section 7 offers the limitations of this study.
2. Literature review and hypothesis
2.1 About XBRL
2.1.1 XBRL for data providers
XBRL US has provided a guide to help starters to be able to know how to prepare financial statements in XBRL format1. I’ll make a short summary of how XBRL works. XBRL’s objective is to enable computers and software applications to directly interpret and process information. Another requirement is that these data must be analyzed accompanied with a rich context. Using the example in Figure 1. Computers couldn't understand the number "our net sales were 131,383" without giving more information about what the actual economic meaning is, XBRL-formatted document will define “net
sales”, and explain of which company the data is belonged to (Company ABC), in which currency the data is reported (dollars), in which fiscal year the data is reported (year 2013). Therefore XBRL provides a context in which the data is applied.
Figure1. Sample XBRL income statement1
Under XBRL, companies use taxonomies to perform tagging, a taxonomy defines relevant terms for a given universe of discourse. Taxonomies are like dictionaries, they contain agreed-upon definitions for all the terminologies used in specific types of business reports. And if the concepts a company needs do not exist in existing taxonomies, the company can extend the taxonomy, making adaptions according to the companies’ specific economic conditions. This adds more flexibility into the application of XBRL. Taxonomy builds an exclusive system that links various ways of information
1 Source: http://xbrl.us/preparersguide/pages/section1.aspx
presentation, definitions of individual items and the relationships between each elements. After the tagging process, the next step is to create a XBRL-encoded file, these files are also known as instance documents. These instance documents must be reviewed in compliance with Commission rules before they are submitted and acknowledged by SEC.
A part of a brief mapping template is shown below:
Figure 2. Sample of mapping process of balance sheet1
On the right column, companies need to add additional descriptions of how they treat this line item. Therefore, through XBRL documents, users can observe the decisions made by the managers.
After tagging, mapping, and firms’ choice of extending the taxonomy, an instance document will be created.
The instance document creation process is shown below:
1 Source: http://xbrl.us/preparersguide/pages/section1.aspx
Figure 3. Instance document creation process1
2.1.2 XBRL for data users
I choose Coca Cola ‘s XBRL-formatted balance sheet for year 2009 to serve as an
1 Source: http://xbrl.us/preparersguide/pages/section1.aspx
example to show what tagged reports look like to data users. This 10-K filing is available on SEC’s website.1 By clicking “Interactive data” you can easily read tagged XBRL
reports online. Figure 4 is a screenshot of Coca Cola Co.’s consolidated balance sheet for year2009.
Figure 4.Coca Cola Co’s consolidated balance sheet for year 2009 (source: SEC website)
And this file can also be opened using excel directly.
By using browser’s extensions, you can have a clearer presentation of the
1http://www.sec.gov/Archives/edgar/data/906107/000119312510040142/0001193125-‐10-‐040142-‐index.ht
company’s reports. For example, I use “Electronic Company Filings”1 on Google
Chrome. Figure 5 is an screen shot of Coca Cola Co’s consolidated balance sheet for year 2009.
Figure 5. Coca Cola Co’s consolidated balance sheet for year 2009 (source: SEC website) And by clicking “Inventories”, you can get an overview of composition of this category (Figure 6).
Figure 6.Overview of composition of inventory of Coca Cola (source: SEC website)
2.2 How does XBRL impact earnings management?
Prior studies conducted in this area mainly focus on:1)the advantages and disadvantages of XBRL and possible impediments of implementing XBRL;2)the association between information efficiency, data quality and XBRL application;3)the role of XBRL for stakeholders.
XBRL is an Internet-based language and it enables organizations’ financial information and data to be disseminated across various platforms. XBRL enables the preparation process of financial reporting become an automatic process, eliminate multiple data entries, however, the output of data processing is of high degree. XBRL analyze data from data pool and in context-rich format, narrow the information barriers between information users and information providers.
SEC has called for the improvements in the preparation of financial statements and disclosure processes. To address the oversight of financial reporting and disclosure process, SEC has focused on a specific issue called “earnings management”, the definition of earnings management offered by SEC is “the practice of distorting the true financial performance of the company”1, And by taking advantage of XBRL’s inherent versatility, the information flow and data analysis during the financial reporting process
can be substantially upgraded. Because current accounting standards allow companies to choose different ways to present the business information and there’re alternative methods to treat one type of transactions, it’s very difficult for information users to process all tremendous amount of information incorporated in financial statements and interpret the meaning of the information correctly. And managers have too much freedom in their accounting choices, they may choose to present business information in complicated ways that make the uninformed investors or non-professionals confused about the financial performance of the company in order to achieve their personal benefits or reach certain objectives. While computers can intelligently interpret the encoded information in XBRL-format files, information users have more time to focus on comparing the information presented in various ways to have a holistic view of the financial statements, gives management little space to exercise their influence over earnings management. Although the companies have the ultimate call to decide what information to display and in which way the information is displayed, with XBRL the information users can have the option to slice and dice the data according to their intentions, and don’t have to accept the data the companies choose to present. A lot of scholars have indicated that the tagging process incorporated in XBRL technology and the fact that reviewing XBRL files is an automatic process make it easier for other parties other than information providers to detect misreporting behavior of the management, thus it’s more difficult for management to exercise manipulation of earnings (Rezaee and Turner 2002). And this helps to reduce the possibility of human fraud, collusion, and mistakes.
“In addition to building a corporate culture of accountability and accuracy, there is a very real need to re-examine the manner in which information is produced, verified and disclosed... We believe that for many enterprises the use of XBRL technology will prove the most suitable, platform-independent way to impose rigor on the framework of reporting. ”(KPMG, 2008) 1 KPMG as a main advocator of XBRL, has pointed out that adopting XBRL can help reveal the way data is produced, and managers’ choice to disclose the information, therefore decision process for financial statements becomes
more transparent.
Prior literature also suggests that external audit plays a vital role in firms’ choices to manipulate earnings. External audit has a function that puts pressure on management to take the aftermath of such manipulation being revealed into consideration. To some extent, external audit prevents management from earnings management (Healy,1985). By being able to use computers or other software applications to intelligently interpret and analyze the financial information, auditor can focus more on investigate the reliability and accuracy of the financial statements, audit quality is therefore taken to a higher level. And scholars have pointed out that by adopting XBRL, firms can enhance audit trail and internal control (Bizarro, Garcia, 2011). Bizarro and Garcia use screenshots of XBRL GL instance documents to show that XBRL-tagged information that is shared between various platforms retains its context and identity, make an automated audit trail possible. They also indicate that by applying XBRL, auditors prey on the fast-generated financial data to pay more attention on internal control procedures and higher-level investigative analyses.
2.3 XBRL’s effects on internal control & capital market
The key objectives for the SEC to mandate XBRL are to improve financial information efficiency, improve data quality, and then to achieve transparency. There are many parties would benefit from the XBRL project, these parties involve regulators, companies, governments, data providers, analysts and investors and accountants. (Booth, 2007).
XBRL makes the information and data more accessible to information users, and XBRL enhance the information searching capabilities of information users, As Diamond and Verrecchia (1991) suggested, through providing information users more access to ways they couldn’t have otherwise can mitigates the information asymmetry between information providers and information users. Empirical research has investigated the link between XBRL and accounting information quality. Yoon et al. examined the effectiveness of XBRL’s impacts on reducing the degree of information asymmetry
specifically in Korean stock market. Their sample include all Korean public companies, the final sample consists of 550 firms, and they employed relative quoted spread as a proxy for information asymmetry. Usually, there are differences between bid price and ask price, and the bid-ask spread is a common proxy for information asymmetry. In this study particular, they compute the relative quoted spread as:𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑞𝑢𝑜𝑡𝑒𝑑 𝑠𝑝𝑟𝑒𝑎𝑑 =
!"# !"#$%! !"# !"#$%
!"# !"#$% !!"# !"#$% !
. To observe the effects of XBRL, they define the pre-XBRL period as from December 2006 to August 2007 and December 2007 to August 2008 as post-XBRL period, they observe a significant and negative association between XBRL adoption and information asymmetry – the regression coefficient of XBRL adoption for the large-sized companies is significant associated with relative quoted spread, and relative quoted spread reduced significantly after the adoption of XBRL, which suggests that the by adopting XBRL may decrease the level of information asymmetry. They also indicate that large firms experienced larger extent of information asymmetry reduction than firms with smaller size (Yoon et.al.2011).
Prior studies have reported XBRL’s ability to improve accounting data, information integrity, and information quality (e.g Vasarhelyi, Chan, and Krahel 2012). Vasahelyi et al. analyze five aspects of XBRL adoption, including current data, disclosure format, historical data, data fidelity & assurance, and third-party data by creating scenarios and summarizing prior studies’ conclusions to examine the potential of the usefulness of XBRL on financial reporting. Regarding current data, they argue that using XBRL can address comparability problems arising from the use of different naming conventions and accounting policies by identifying each line item and tagging the accounting method used. In respect of disclosure format, they argue that the standardization of disclosures will help remove a degree of opacity, improving the comparability and consistency of inter-firm and period analysis and enhancing the usefulness of disclosures. Furthermore, they indicate the use of mapping software can ameliorate the issue of historical data obsolescence. On top of that, they conjecture SEC’s assurance requirements for XBRL filings will enhance the reliability and consistency of financial reporting. Finally, regarding third-party data, they propose that the inclusion of standardized third-party
financial and nonfinancial data will enrich the usefulness of XBRL technology by incorporating new dimensions of automated data analysis. Their main conclusion is that XBRL-tagged statements enhance the usefulness of financial information and aid users in data retrieval and analysis for decision-making.
Efendi et al. (2014) collected 7619 firm-quarterly observations from 474 firms and used post earnings announcement drift as a proxy for information efficiency, PEAD， which is short for post–earnings-announcement drift, is the tendency for a stock’s cumulative abnormal returns to drift in the direction of an earnings surprise for several weeks (even several months) following an earnings announcement (Bernard and Thomas, 1989). And the outcome of their study shows that post earnings announcement drift declines in post-XBRL period for good news, and from that they conclude that XBRL improves information efficiency in the capital market. In their study, good news portfolios refer to portfolios that positive standardized unexpected earnings (SUE), and SUE is measured by: 𝑆𝑈𝐸!,!,! =!"#!,!,!!!"#!,!,!!!
!!,!,! ,where i=firm, q=quarter, t=year, and
EPS= earnings per share, P=stock price.
Of the many advantages numerous scholars have mentioned, transparency is one of the most discussed topics, and it’s one of the key benefits of applying XBRL, it provides a tactic to accelerate the speed of retrieving and integrating all relevant financial information from companies’ financial statements. Hodge et al. (2004) conducted an study where they create various scenarios to investigate the individuals’ ability to collect and integrate information and use that information to make investment decisions with or without the help of search-facilitate technology (XBRL), to be more specific, they recruited 96 second-year MBA students as their participants, and observe their decisions by randomly assigning them to one of their four scenarios in an 2 × 2 between subject design, the randomly assigned scenarios are composed of two elements, one is presentation format, the participants can be assigned to group that uses non-searchable information or group that uses searchable information, the other element is the placement of data, the participants can be assigned to a scenario where the placement of data is
recognition and they can also be assigned to a scenario where the placement of data is disclosure. And their results reveal that XBRL enhances the ability of the uninformed investors to observe management’s behavior on the decisions they make for what information to disclose and what information to not disclose, also XBRL increases the transparency of the information embedded in financial statements, furthermore, they discover that users who use XBRL are superior at acquiring decision-relevant information and integrating that information compared to users who don’t use XBRL. They proved that XBRL enhances the information searching ability of investors. And that the increased transparency brought by XBRL may act as a constraint on management’s incentives to try to maneuver the numbers on financial statements (for example, through accelerating revenues or delaying expenses) or manage to ask for more freedom in accounting choices or the change of accounting methods.
2.4 Earnings management
In this paper, I used the definition of earnings management suggested by Healey and Wahlen(1999): “... managers’ use of judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.” Managers can manipulate numbers in financial statements by exercising judgments. They can, for example, record the lump-sum payments received from customers as current sales where the contract lasts for several years and the payments should be amortized, or they can improve their earnings by exercise reductions in their defined benefit plans. And they are responsible for choosing suitable accounting methods to calculate different costs and expenses, regarding depreciation policies, they can choose straight-line method or accelerated depreciated method. Furthermore, managers have to exercise judgment on working capital management. Managers also have the responsibility to decide to make or defer expenditures. On top of that, managers can create off-balance-sheet items to hide a part of revenues and costs from investors. Besides, they make the ultimate call of how to structure corporate transactions. Therefore, there’re a lot of opportunities that managers can manipulate earnings.
Hunton, Libby and Mazza（2006） designed an experiment to investigate whether greater transparency in comprehensive income report can acts as a constraint on management’s choice on earnings management. In their experiment, participants are composed of 62 financial executives and CEOs, and they make decisions on which available-for-sale security to sell from a portfolio. There are two elements in the 2 x 2 between-subjects design, one is the transparency of comprehensive income reporting, more transparent reporting they are referring to here are financial reports that can be easily retrieved and analyzed to understand the substance of the financial position and performance of the company, and the other one is the results of consensus forecast, which can be above or below projected earnings. And from the outcome of the experiment they find that when projected earnings are below (above) the consensus forecast, participants sell securities that increase (decrease) earnings. Meanwhile, the outcome also indicates that more transparent financial reports diminish the level of both earnings management that increases income and earnings management that decreases income. Thus the results indicate that transparent reporting requirements significantly reduce the level of earnings management.
2.5 Hypothesis
Accordingly, XBRL disclosure enhances the relevance, understandability and comparability of financial statements, and it serves as a effective communication tool for stakeholders, reduces the information asymmetry between information producers and information users. On top of that, XBRL enhances audit quality by enhancing audit trail and internal control. Furthermore, XBRL improves the transparency of financial reporting, and more-transparent disclosure makes it easier to detect earnings management, therefore, I propose the hypothesis: the adoption of XBRL will reduce the level of earnings management by reducing discretionary accruals.
3. Research method
3.1 Sample & Data
The SEC issued a schedule that requires companies to file in XBRL format over a three-year phase-in schedule. SEC requires that in phase 1, Domestic and foreign large accelerated filers that use U.S. GAAP and have a worldwide public common equity float above 5 billion dollars as of the end of the second fiscal quarter of their most recently completed fiscal year, which is fiscal year 2008, will provide their quarterly and annual reports using interactive data for a fiscal period ending on or after June 15, 2009 must file their quarterly and annually reports in XBRL format. In year two of the phase-in schedule, all other large accelerated filers using U.S. GAAP are required to provide their interactive data for a fiscal period ending on or after June 15, 2010. All remaining filers are required to provide their initial interactive data submission as of June 15, 2011. According to SEC’s three-year phase-in schedule, I collect data for post-XBRL filings from SEC’s monthly RSS Feeds (https://www.sec.gov/Archives/edgar/monthly/) from June 15th 2009 to June 14th, 2010 (the first phase). This year being the first year of XBRL mandate provides me a effective setting to investigate the introduction of XBRL’s impacts on earnings management. For the first year of SEC’s three-year phase-in schedule, companies are required to tag the basic financial information displayed on financial statements and also to block tag footnotes. The XBRL-formatted documents are made available on EDGAR, and SEC requires companies to release the XBRL-format financial statements on their corporate website to notify their investors as soon as they submit their financial reports on SEC’s website. And the fact that not all companies are mandatorily required to file in XBRL format provides me an opportunity to set up a control group to examine XBRL’s effects on earnings management from pre-XBRL era to post-XBRL era. The federal securities laws require public companies to submit annual reports on Form 10-K, quarterly reports on Form 10-Q.1
1 http://www.sec.gov/answers/form10q.htm
The data for pre-XBRL filings and for control variables are obtained from COMPUSTAT database. Data required for discretionary accruals estimation will also be collected from COMPUSTAT.
Firstly, I downloaded all the XBRL 10-k filings that filed between June 15th, 2009 and June 14th, 2010, after deducting repeating data, the sample consists of 446 filings in total. (A list of these 446 firms is in Appendix A) And to ensure the sample is consisted of mandatory filers, I removed firms that have reported in XBRL before June 15, 2009, to decide which companies fall into this category, I first manually validate each company by using SEC’s EDGAR RSS Feeds to check their filings prior to June 15, 2009 are tagged with “Interactive Data”. And then I cross reference with Callaghan and Nehmer’s list of voluntary/early XBRL adopters (Callaghan, J & Nehmer, R 2009), removing voluntary XBRL adopters (Appendix B). And in this step I filter out 51 firms in total. Furthermore, I excluded firms in finance industry following prior studies on earnings management. Previous literature always exclude banks and other financial institutions in the sample for earnings management researches and the reason is that their characteristics differ fundamentally from other firms (Peasnell, Pope and Young, 2000). The finance industry is highly regulated and the behavior of accruals may differ vastly from other industries. And here, by deciding which companies falling into this industry, I use two digits SIC codes to filter out those firms (two digits SIC codes ≠ 60 - 67). And to conform to the phase-in rule, I removed firms that had a public float below $5 billion, public float here represents the portion of shares of a firm that are in the hands of public investors as opposed to company officers and directors. SEC requires each company to state the public float on the first page of their 10-K filings. When I collected this data, I found some firms didn’t disclose their public float in their 10-K filings and some firms don’t have 10-K filings for fiscal year 2008, so I also removed these firms from the overall sample. And last, I removed firms with missing values for both pre- and post-XBRL periods. The pre-XBRL period I’m referring to here is from June 15,2008 to June 14, 2009.
There’s an overlap between the pre-XBRL period I defined (15 June, 2008 to 14 June, 2009) and 2007-2008 financial crisis, but most of the filings in my sample are not in this period. To ensure the results are free from bias, I create a control group to take the economic wide factors into consideration, the matched control group design will be described in next section.
Table 1 presents the overview of the final sample composition. Table 1.Overview of sample composition:
XBRL 10-K ‘s filed for fiscal periods between 6/15/2009 and 6/14/2010 446 Less filings for firms that voluntarily filed XBRL in prior period (51) Less firms in finance, insurance and real estate industry (69) Less filings that have market float of less than $5 billion at the end of the second quarter of fiscal year 2008(ie. Voluntary XBRL adopter in phase 1) and firms didn’t disclose their public float in their 10-K filings and firms don’t have 10-K filings for fiscal year 2008
(37)
Less filings that have missing data in Compustat and duplicated filers (14)
Total post-XBRL filings 267
Table 2 summarizes the industry breakdown of the final sample. Table 2.Industry breakdown of the final sample
3.2 Research design
I hypothesize that the level of earnings management will be reduced after the adoption of XBRL, and earnings management is captured by discretionary accruals here, before and after mandatory adoption of XBRL. And I have chosen using the cross-sectional modified Jones model to measure the non-discretionary accruals. The calculation process consists of 3 steps:
Step 1: calculate the estimations (𝛽!, 𝛽!, 𝛽!), the estimations should be estimated by year and industry, industry observations that are less than 20 should be deleted to ensure accuracy.
One-digit
SIC Industry name
Number of XBRL
firms Percentage of total 1 2 3 4 5 7 8
Mining and Construction Light manufacturing and chemicals
Heavy Manufacturing Transportation and public utilities
Wholesale and retail trade Services Health Service Total 39 43 78 52 28 23 4 267 14.6% 16.1% 29.2% 19.5% 10.5% 8.6% 1.5% 100%
The Modified Jones Model: 𝑇𝐴!,! 𝐴𝑇!,!!! = 𝛽!∗ 1 𝐴𝑇!,!!!+ 𝛽!∗ ∆𝑅𝑒𝑣 − ∆𝑅𝑒𝑐 𝐴𝑇!,!!! + 𝛽!∗ 𝑃𝑃𝐸𝑇 𝐴𝑇!,!!!+ 𝜀 Where:
TA=Total accruals =Income Before Extraordinary Items – Operating Activities Net Cash Flow AT= Assets Total
∆Rev=changes in revenues ∆Rec=changes in receivables
PPET=Gross Property Plant & Equipment
Step 2: Use the estimations calculated from the regression above on the regression below to calculate NDA (non-discretionary accruals):
𝑁𝐷𝐴 𝐴𝑇!,!!! = 𝛽! 1 𝐴𝑇!,!!!+ 𝛽!∗ ∆𝑅𝑒𝑣 − ∆𝑅𝑒𝑐 𝐴𝑇!,!!! + 𝛽!∗ 𝑃𝑃𝐸𝑇 𝐴𝑇!,!!!
Step 3: Finally, use the formula below to calculate DA (discretionary accruals): 𝐷𝐴!,! = 𝑇𝐴!,! − 𝑁𝐷𝐴!,𝑡
3.3 Research model
To study the introduction of XBRL’s effects on the level of earnings management, I estimate the following OLS regression model:
𝐷𝐴!,! = 𝛼! + 𝛼!𝑌𝑒𝑎𝑟 2009 + 𝛼!𝑆𝐼𝑍𝐸 + 𝛼!∗ 𝑃𝐸𝑅 + 𝛼!∗ 𝐿𝐸𝑉 + 𝛼!𝑀𝐵 + 𝜀
Earnings management here is captured by discretionary accruals, that’s the dependent variable here. Additionally, because I have two observations for each firm (pre-XBRL observation and post-XBRL observation), to solve for cross-observation correlation for these variables, I cluster standard errors at the firm level.
3.4 Variables
Firstly, I create a dummy variable- Year 2009 to distinguish firms that adopt XBRL in the given time. It equals to 1 if the filing was filed between June 15,2009 to June 14,2010, it equals to 0 otherwise. Therefore, observations in period June 15,2008 to June 14,2009 are in pre-XBRL period and observations in period June 15,2009 to June 14,2010 are in post-XBRL period.
And I mainly rely on previous studies to choose variables and proxies. To filter out other firm-specific characteristics that may have effects on discretionary accruals, I include several control variables in my regression model. I’ll control for the following variables that may affect earnings management, as indicated in prior literature. First, as earnings management represented by discretionary accruals are found to be related to performance, size and leverage (Young,1998), I’ll control for the firm performance of XBRL era. Performance measured using return on asset (ROA), calculating as operating earnings divided by assets, correspondent COMPUSTAT code is ib/at . Furthermore, I’ll control for other determinants f earnings management.: firm size (SIZE, the logarithm of total assets for a firm XBRL), financial leverage (LEVERAGE, total liabilities divided by total assets for a firm XBRL).Furthermore, market-to-book ratio is included as one variable to capture the growth opportunities, since prior studies have pointed out that higher investment levels could lead to more accruals. (Zhang,2007)
Table 3 .Overview of variables:
Variable Name Variable definition
PER
SIZE LEV MB Year 2009
Performance measured using return on asset (ROA), calculating as operating earnings divided by assets
The logarithm of total assets for a firm pre-XBRL
Total liabilities divided by total assets for a firm pre-XBRL Total market value divided by total book value
Dummy variable, it equals to 1 if the filing was filed between June 15,2009 to June 14,2010, it equals to 0 otherwise
4. Results
4.1 Descriptive statistics and univariate data analysis
Table 4 presents descriptive statistics of the variables of my sample used in my main tests. The mean (median) of total accruals, TA is -0.073 (-0.049) in pre-XBRL period, and -0.071 (-0.060) in post-XBRL period, the difference is not significant statistically. And the mean (median) of DA is -0.024(-0.026) in pre-XBRL period and the mean(median) of DA in post-XBRL period is -0.054(-0.043), which has a difference of -0.030 and is statistically significant. Which means there’s a significant amount of decrease in DA after the mandate of XBRL. And from the two-tailed t test, the differences of SIZE, LEV and MB between pre-XBRL period and post-XBRL period are also statistically significant. In summary, the results of descriptive statistics and univariate test result provide some evidence that XBRL reduces the level of earnings management and decreases discretionary accruals.
Table 4.Descriptive statistics for experiment group in pre-XBRL period (2008/06-2009/06) and post-XBRL period (2009/06-2010/06) Pre-XBRL (n=267) Post-XBRL (n=267) Differe nce in Sig.(2-taile d)
Mean Median SD Mean Median SD Means Means
TA DA SIZE LEV PER MB -0.073 -0.024 4.038 0.583 0.060 0.573 -0.049 -0.026 4.018 0.599 0.072 0.510 0.098 0.058 0.455 0.212 0.123 0.420 -0.071 -0.054 4.060 0.554 0.056 0.469 -0.060 -0.043 4.027 0.567 0.054 0.415 0.061 0.095 0.449 0.202 0.070 0.315 0.002 -0.030 0.022 -0.029 -0.004 -0.104 0.7931 0.0000*** 0.0000*** 0.0000*** 0.5810 0.0000*** *,**,*** indicates statistical significance at respectively 0.1, 0.05 and 0.01 per cent level
Where:
TA=Total accruals =Income Before Extraordinary Items – Operating Activities Net Cash Flow, scaled by beginning total assets
DA=Discretionary accruals scaled by beginning total assets SIZE= The logarithm of total assets
LEV= Total liabilities divided by total assets
PER= Performance measured using return on asset (ROA), calculating as operating earnings divided by assets
MB= Total market value divided by total book value
4.2 Regression results and Pearson correlations
Table 5 presents the results of multiple regression of discretionary accruals. Year2009, the variable that measures the XBRL’s impacts on earnings management, is negatively correlated with discretionary accruals, suggesting that DA is significantly lower after XBRL mandate. The adjusted R-square is 0.2379, which offers an explanatory power of 23.79%, it means that 23.79% of the variation between pre and post-XBRL discretionary accruals can be explained by the variables defined in the regression model. The coefficient for Year2009 is -0.0674, specifically. Thus, in the post-XBRL period, discretionary accruals are lower by 6.74 percent of total assets. This result is consistent with my hypothesis. And the results also show that the coefficients for PER (performance) and MB (market-to-book ratio) are also significant. (p-value=0.0450
and 0.0740, respectively), suggesting that post-XBRL discretionary accruals are different from pre-XBRL period when PER and MB are used as dependent variables.
Table 5. Multiple regression of discretionary accruals(using only experiment group) (Std. Err. adjusted for 526 clusters in firms)
Source SS df MS Number of obs 526.0000
F( 5, 520) 4.5700
Model 0.1404 5 .028070855 Prob > F 0.0004
Residual 3.1939 520 .006142057 R-squared 0.2421
Adj R-squared 0.2379
Total 3.3342 525 .006350903 Root MSE 0.1784
DA Coef. Std. Err. t P>t [95% Conf. Interval]
Year2009 -0.0674 .0069865 -3.54 0.0000*** -0.0385 -0.0110 SIZE 0.0078 .0081165 0.52 0.6010 -0.0117 0.0202 PER 0.0729 .0363312 2.01 0.0450** 0.0015 0.1443 LEV 0.0201 .0177019 1.14 0.2560 -0.0147 0.0549 MB 0.0185 .0103385 1.79 0.0740* -0.0018 0.0388 _cons -0.0712 .0320892 -2.22 0.0270 -0.1343 -0.0082 *,**,*** indicates statistical significance at respectively 0.1, 0.05 and 0.01 per cent level
The results of the Pearson correlations of my variables in my study are presented in table 6. Pearson correlations provide an overview of the relationships between each variable. To be specific, Year2009 is statistically significantly negative to DA, which suggests that DA is significantly reduced after the adoption of XBRL in year 2009. The results show that Year2009 is also significantly negatively correlated with MB (market-to-book ratio), which suggest that MB is reduced in the post-XBRL period.
Table 6.Pearson Correlation of variables
(obs=526)
Year2009 SIZE PER LEV MB DA
Year2009 1 SIZE 0.0241 1 PER -0.0194* -0.1625 1 LEV -0.0679* 0.1916 -0.1451 1 MB -0.1404*** 0.2576 -0.2965* -0.1756 1 DA -0.1726*** 0.0379 0.0581*** 0.0389 0.079*** 1
*,**,*** indicates statistical significance at respectively 0.1, 0.05 and 0.01 per cent level Where:
DA=Discretionary accruals scaled by beginning total assets
Year2009=dummy variable, it equals to 1 if the filing was filed during June 15,2009 to June 14,2010, and it equals to 0 otherwise.
SIZE= The logarithm of total assets
LEV= Total liabilities divided by total assets
PER= Performance measured using return on asset (ROA), calculating as operating earnings divided by assets
MB= Total market value divided by total book value
5. Matched sample analysis
5.1 Revised research design
There is one factor that may have a huge impact on my results, which is the financial crisis during 2007 to 2008. From my prior results, DA is reduced probably due to financial crisis. To assess the robustness of my results, and take different economic conditions into consideration, I take advantage of SEC’s phase-in schedule and create a difference-in-difference design. The control group consists of non-XBRL adopters based on industry and size to observe XBRL’s effects on earnings management. This design
take market-wide factors other than the variables I defined in my regression model into consideration, because there’s a passage of time during my investigation, a difference-in-difference design helps to mitigate other factor’s effects to an acceptable level. To highlight the specific method I use, I match each firm in phase 1 period to a non-XBRL filing firm based on industry (the three-digit SIC codes) and size (total asset). For each industry in the XBRL sample, I choose an equivalent number of non-XBRL adopters from non-XBRL sample using the 3-digit SIC code. Because the first group that is required to file in XBRL consists of big companies (with public float over $5 billion) so the matching group’s average size is less than the experimental group’s average size. But the matching group still consists of large companies, from the table shown below, the mean (median) of the control group’s size in year 2009 is 3.393 （3.324）. Because not in every industry can I find exactly the same equivalent number of matching firms, and because of missing data in COMPUSTAT, the control group consists of 201 firms, in contrast, I again choose respondent experiment firms using the 3-digit SIC codes (I delete firms in experiment group that don’t have a matching firm using the matching indicator “industry” and “size”). The relevant descriptives are shown below.
Table 7a.Descriptives across samples
XBRL adopters Non-XBRL adopters Number of Firms
Numbers of observations Size in year 2009(Mean) Size in year 2009 (Median)
201 402 3.989 3.948 201 402 3.408 3.332
Table 7b presents descriptive statistics for matched groups in pre-XBRL period, in other words, in this period, all of these firms did not adopt XBRL. The statistics suggest that, during pre-XBRL period, experiment group firms experienced higher TA and DA than control group firms, probably because as large accelerated filers with large public float, these firms need to manage earnings to attract or maintain investors and satisfy stakeholders. This difference may lead to experiment group’s XBRL adoption exhibiting
more sensitivity to DA.
Table 7b.Descriptives for matched groups in pre-XBRL period XBRL non-adopters in pre XBRL-period (n=201) XBRL adopters in pre XBRL period (n=201) TA Mean -0.092 -0.075 Median -0.054 -0.047 DA Mean -0.045 -0.025 Median -0.019 -0.026 SIZE Mean 3.393 3.968 Median 3.324 3.933 PER Mean 0.007 0.057 Median 0.037 0.071 LEV Mean 0.592 0.565 Median 0.587 0.585 MB Mean 0.946 0.574 Median 0.780 0.498
And to capture XBRL’s impacts on earnings management on both control group and experiment group, I change the research design slightly. Firstly, I create a dummy variable- XBRL, it equals to 1 if the firm is in XBRL sample, it equals to 0 otherwise. I also create a dummy variable Year 2009 to distinguish firms that adopt XBRL in the given time. It equals to 1 if the filing was filed between June 15,2009 to June 14,2010, it equals to 0 otherwise. The interaction item: Year2009*XBRL, therefore captures the impact of XBRL on earnings management.
The abovementioned changes result in the revised regression:
𝐷𝐴 = 𝛼!+ 𝛼!∗ 𝑋𝐵𝑅𝐿 + 𝛼!𝑌𝑒𝑎𝑟 2009 + 𝛼!𝑌𝑒𝑎𝑟 2009 ∗ 𝑋𝐵𝑅𝐿 + 𝛼!𝑆𝐼𝑍𝐸 + 𝛼!∗ 𝑃𝐸𝑅 + 𝛼!∗ 𝐿𝐸𝑉 + 𝛼!𝑀𝐵 + 𝜀
where:
DA=Discretionary accruals XBRL=firms in XBRL sample
Year2009=Dummy variable, it refers to Post-XBRL period, where experiment group adopted XBRL and control group didn’t. It equals to 1 if the filing is filed during June,15,2009 to June 14,2010. It equals to 0 otherwise.
during June 15,2009 to June 14,2010, it equals to 0 otherwise. SIZE= The logarithm of total assets
LEV= Total liabilities divided by total assets
PER= Performance measured using return on asset (ROA), calculating as operating earnings divided by assets
MB= Total market value divided by total book value
Again, to solve for cross-observation correlation, I cluster standard errors by firms.
5.2 Results
Table 8 provides statistics for control group before and after the XBRL split. The table shows that for all control group firms the result of two-tailed t test for the difference in DA before and after the Year2009 is not significant. It provides some evidence that for firms that didn’t adopt XBRL in Year2009, the XBRL mandate didn’t have significant impact on the level of earnings management (discretionary accruals). This result assumes the reduced discretionary accruals for experiment group firms after the XBRL mandate is linked to the adoption of XBRL.
Table 8.Descriptive statistics for control group in pre-XBRL period (2008/06-2009/06) and post-XBRL period (2009/06-2010/06) Pre-XBR L (n=201) Post-XBRL (n=201) Differe nce in Sig.(2-taile d)
Mean Median SD Mean Median SD Means Means
TA DA SIZE PER LEV MB -0.092 -0.045 3.393 0.007 0.592 0.946 -0.054 -0.019 3.324 0.037 0.587 0.780 0.129 0.094 0.360 0.152 0.222 1.067 -0.094 -0.048 3.408 0.016 0.568 0.663 -0.073 -0.020 3.332 0.027 0.555 0.621 0.124 0.113 0.366 0.086 0.226 0.502 0.002 -0.003 0.015 0.009 -0.024 -0.283 0.8081 0.6996 0.0120* 0.3851 0.0000*** 0.0000*** *,**,*** indicates statistical significance at respectively 0.1, 0.05 and 0.01 per cent level
The outcome for the multiple regression of discretionary is presented in table 9, the sample of this regression consists of experiment group and control group, 402 firms and
804 observations taken together. The adjusted R-squared is 0.2379, which indicates that the explanatory power of the regression model is 23.79%. The coefficients for variables: Year2009, XBRL and Year2009*XBRL are all significant and negative (-0.0195, -0.0043 and -0.0754), which suggests that by adopting XBRL in year 2009 reduces the level of earnings management which in turn measured by DA. On the other hand, PER is also positively associated with DA, which indicates that firms with high performance tend to use discretionary accruals to do earnings management. The coefficient of LEV(0.0229) is significant and positive, which suggest that firms with lower leverage tend to engage in earnings management through lowering discretionary accruals.
Table 9. Multiple regression of discretionary accruals (using experiment group and control group) (Std. Err. adjusted for 402 clusters in firms)
Source SS df MS Number of obs 804 F( 7, 796) 22.0700 Model 0.9719 7 . 138841301 Prob > F 0.0000 Residual 5.8000 797 .006290651 R-squared 0.2435 Adj R-squared 0.2379
Total 6.7719 804 .007289418 Root MSE 0.0773
DA Coef. Std. Err. t P>t [95% Conf. Interval] Year2009 -0.0195 .0080 -2.00 0.0460** -0.0318 -0.0003 SIZE -0.0043 .0064 -0.66 0.5090 -0.0169 0.0084 PER 0.2770 .0244 11.33 0.0000*** 0.2290 0.3250 LEV 0.0229 .0131 1.74 0.0820* -0.0029 0.0488 MB 0.0067 .0045 1.46 0.1440 -0.0023 0.0156 XBRL -0.0621 .0088 -3.48 0.0010*** -0.0484 -0.0135 Year2009*XBRL -0.0754 .0105 4.60 0.0000*** 0.0278 0.0691 _cons -0.0293 .0226 -1.73 0.0830 -0.0838 0.0052 *,**,*** indicates statistical significance at respectively 0.1, 0.05 and 0.01 per cent level
Table 10 presents the results of the correlation matrix for all variables defined in chapter 5.1. Discretionary accruals is negatively correlated with indicator variable Year2009 (-0.0417), XBRL (-0.0178) and the interaction item which captures XBRL’s impacts on earnings management, Year2009*XBRL (-0.0143), all results are significant.
Furthermore, Year2009 and XBRL are highly positively correlated with the interaction item: Year2009*XBRL (coefficient 0.5754 and 0.3311, respectively) suggests that using these two variables to explain the decreased discretionary accruals would produce similar outputs. Both Year2009 and XBRL are negatively correlated with DA, provides initial evidence that the level of earnings management is reduced in the post-XBRL periods for the firms that adopted XBRL. And as I stated before, the difference-in-difference design takes the overall economic condition into account, the results are therefore more rigid, and I’m particularly paying attention to the interaction item: Year2009*XBRL, from the table shown below, I thus find some evidence that support my hypothesis: adopting XBRL helps mitigates the level of earnings management measured by discretionary accruals.
Table 10. Pearson Correlations of variables
Year2009 SIZE PER LEV MB XBRL
Year2009 *XBRL D A Year2009 1 SIZE 0.0189 1 PER 0.0148 0.0256 1 LEV -0.0585* 0.1044* -0.2107* 1 MB -0.1462*** -0.0351 -0.1963 -0.2426 1 XBRL -0.5754** 0.3269** 0.1127 -0.0021** -0.0743 1 Year2009*XBRL 0.5754*** 0.3524** 0.1012* -0.0727* -0.1701* 0.3311* 1 DA -0.0417*** -0.0488 0.3263** -0.0253* -0.0178 -0.0178** -0.0143** 1
*,**,*** indicates statistical significance at respectively 0.1, 0.05 and 0.01 per cent level Where:
DA=Discretionary accruals XBRL=firms in XBRL sample
Year2009=Dummy variable, it refers to Post-XBRL period, where experiment group adopted XBRL and control group didn’t. It equals to 1 if the filing is filed during June,15,2009 to June 14,2010. It equals to 0 otherwise.
Year2009*XBRL=An interaction dummy variable, it equals to 1 if the filing is filed by experiment group during June 15,2009 to June 14,2010, it equals to 0 otherwise.
SIZE= The logarithm of total assets
LEV= Total liabilities divided by total assets
PER= Performance measured using return on asset (ROA), calculating as operating earnings divided by assets
MB= Total market value divided by total book value
6. Conclusions
In April, 2009, SEC mandated XBRL, SEC’s initial motivation to mandate XBRL was to follow their voluntary XBRL filing program started in April 15,2005. By collecting comments from accounting firms, associations, corporations and individuals, SEC reached the conclusion that adopting XBRL has the potential to reduce information cost, to increase the transparency of business information and help information users to make better decisions based on enhanced information quality. The list of comments and related information can be found on SEC’s website.1 The primary motivations for the SEC to mandate XBRL are to improve financial information efficiency, improve data quality, and then to achieve transparency. (Booth, 2007).
To answer the research question: does XBRL reduce the level of earnings management, I develop a hypothesis: XBRL reduces discretionary accruals. I take the advantage of SEC’s phase in schedule and created two groups: one experiment group that consists of firms that were required to file in XBRL format mandatorily, which means they never adopted XBRL before the XBRL mandate and it’s an great opportunity to compare the differences adopting XBRL can generate. And the other control group consists of firms that didn’t adopt XBRL,I matched each firm in experiment group with one firm that didn't adopt XBRL based on industry (three-digit SIC codes) and firm size (the logarithm of total asset). This difference-in-difference design helps to mitigate the impacts of 2007-2008 financial crisis to the minimum amount. I first use the experiment group as my sample to test the regression model, the main variable is Year2009 which tests year 2009 (the year those firms adopted XBRL)’s effects on earnings management, the results show that DA is negatively associated with Year2009, which suggests DA is reduced in post-XBRL period than in pre-XBRL period. Then, I added control group into
